document
stringlengths 53
1.08M
| summary
stringlengths 13
9.56k
|
---|---|
Introduction and Theoretical Background
---
Depression in Older Adults
The world population is rapidly aging: Between 2015 and 2050, the percentage of people in the world over the age of 60 will nearly double, from 12% to 22% . Depression among older adults is one of the most serious public health problems facing modern societies . The appearance of depression is associated with serious consequences, including disability, functional decline, diminished quality of life, increased mortality, increased service utilization, and high levels of suicide in adults . In older adults, depression is connected with a marked reduction in cognitive abilities which, in turn, is commonly accompanied by a decrease in social and physical activities . Alongside coronary heart disease, cancer, and cardiovascular diseases, depression is a major public health problem that has become a common chronic disease in older adults .
Depression symptoms greatly influence both physical and cognitive functioning of older adults. Longitudinal studies have reported that depressive symptoms are connected to functional decline, as determined by both self-reported and objective measures of physical performance . They also contribute to limitation of basic activities of daily living in high functioning older adults initially free from disability . As depression deepens and more symptoms surface, the likelihood of becoming disabled increases. Moreover, depressive symptoms may accelerate the disablement process in older adults already exhibiting early signs of disability , and individuals with chronic depressive symptoms have greater declines in functioning compared to those who remained nondepressed .
Late-life depression is also associated with an increased risk of decline in cognitive functioning. Older adults with depression are more likely to have concomitant cognitive deficits, especially executive cognitive functioning deficits, or are subsequently more likely to develop dementia . Older adults with depression often develop cognitive impairment following onset of depression. Thus, depression might be a risk factor or an early symptom of dementia .
One of the risk indicators for depression is lack of social support and social networks. Many of depressed older adults are also lonely, and a correlation has been found between depression and loneliness. Depression with feelings of loneliness leads to more pronounced motivational depletion and serious consequences, including social isolation, reduced selfcare, decreased mobility, and poor diet . participants reported loneliness feelings in the week prior to the interview. The loneliest group was that of age 75+; women reported higher levels of loneliness than men, as did Arabs compared with Jews.
Today, loneliness is also perceived as a biological structure, similar to hunger, thirst, or pain, which are internal mechanisms activating behavior that prevents harm to the person. Hunger makes us seek food; loneliness prompts us to seek social relationships . Correlations have been reported between the feeling of loneliness and several physical health problems: cardiovascular diseases, chronic diseases, cancer, stroke, high blood pressure, and mental illness , low levels of emotional wellbeing, and a high level of suicidal thoughts . The problems caused by loneliness lead to decreased quality of life, increased mortality, poor recovery from illness, and high hospitalization rates among older adults .
Depression in older adults may result from a variety of reasons apart from loneliness feelings, including malnutrition. Some studies have reported a strong and independent association between nutritional deficit and depression, demonstrating that depression increased the risk of impaired nutritional status , whereas others have shown a modest association, or no association . Significantly, over 10% of adults with depression residing in the United States also suffer from malnutrition . Malnutrition impacts quality of life by undermining individual autonomy to perform necessary, instrumental, and social activities of daily living .
Malnutrition is defined as a state in which a deficiency, excess or imbalance of energy, protein, and other nutrients causes measurable adverse effects on tissue/body form , function, and clinical outcome . It is more prevalent and increases among older adults . Although malnutrition's etiology is multifactorial, adverse physiological, psychological, and social causes of malnutrition in older adults are consistently reported in the literature . Aging is accompanied by physiologic changes that can negatively impact nutritional status: Sensory impairment may result in reduced appetite and poor oral health and dental problems can lead to difficulty chewing, inflammation, and a monotonous diet that is poor in quality. The progressive loss of vision and hearing may also limit mobility and affect the elderly's ability to shop for food and prepare meals .
Along with physiologic changes, older adults may also experience profound psychosocial and social changes contributing to poor nutritional status. These include cognitive impairment, heavy use of medication, periods of lengthy hospitalization isolation, retirement from paid work, bereavement, increasing frailty, and loneliness and depression . These factors influence the ability of older adults to meet dietary needs or to digest, absorb, utilize or excrete nutrients that are ingested. The outcome is reduced energy intake and lean body mass, resulting in a reduced metabolic rate and a proportional decline in total energy expenditure that may lead to malnutrition .
Apart from the possible direct connection between loneliness feelings and depressive symptoms, loneliness feelings can have widespread implications for the mental and social lives of older adults, and these implications can explain some of the effects of loneliness feelings on depressive symptoms. Specifically, one mediator might be at play -malnutrition.
---
The Current Study
The current study was conducted in Israel in 2020 during a Covid-19 pandemic quarantine, providing a unique opportunity to assess the effect of loneliness feelings on depressive symptoms, mediated by malnutrition, among older adults from different cultures during a particularly stressful period.
We posited three hypotheses: 1. Loneliness feelings due to the Covid-19 pandemic is associated with depressive symptoms. 2. Loneliness feelings due to the Covid-19 pandemic are indirectly associated with depressive symptoms through malnutrition; older adults feeling lonely will report higher levels of malnutrition, which will be associated with increased depressive symptoms. 3. Loneliness feelings, depressive symptoms, and malnutrition levels will differ between older adults from different cultures as a result of Covid-19 imposed quarantines.
---
Method
---
Study Design and Participants
The research employed a cross-sectional study of a convenience sample of 201 Jewish and Arab older adults, aged 65 and over, representing the two main ethnic groups living in Israel. Inclusion criteria were age 65 and over and the ability to speak and understand Hebrew or Arabic.
---
Procedure
The study was approved by the Research Ethics Committee of the college at which the research took place. Recruitment of participants was random and the final sample comprised 100 Jews and 101 Arabs. Researchers explained the study objectives and procedure to the participants, including their right to withdraw freely at any time. Strict confidentiality was maintained. Data collection was performed by professional interviewers through telephone interviews, adhering to Covid-19 quarantine restrictions, using appropriate translated, validated, and structured questionnaires. Data collection took place from April to May 2020.
---
Measures
---
Independent variable
Loneliness. Loneliness was measured by a single direct question: «Do you sometimes feel lonely?» with four options: never, seldom, sometimes, often.
---
Dependent variable
Depressive symptoms. Depressive symptoms were measured by the Geriatrics Depression Scale developed by Yesavage and Brink . The purpose of the questionnaire was to determine participants' depressive symptoms by using a simple and reliable tool that does not require the time and skills of a professional interviewer. The tool is composed of 15 items, in a yes / no response formats .
---
Mediator
Malnutrition. Malnutrition was measured by the Determine Nutrition Screening Initiative developed jointly by the American Diabetes Association, the American Family Doctors Association, and the National Council of Old Age. The purpose of the questionnaire was to detect older adults at risk for malnutrition. The tool is composed of 10 items, in a yes / no response format.
---
Covariates
The study controlled for socioeconomic variables. Background variables included gender, age, marital status, and years of education. Age and years of education were both defined as continuous measures. Gender was coded as dichotomous . Marital status was coded as "with partner" = 1; or "without a partner" = 0.
All instruments were translated into Hebrew and Arabic by bilingual translators. The complete questionnaire underwent a pilot test. The questionnaire took approximately 15 minutes to complete, the verbal instructions were comprehensible, and there was no need for further changes before administering the questionnaire.
---
Data Analyses
Descriptive statistics were employed to calculate the means and standard deviations of the continuous variables and the percentage and frequency of the categorical variables. In the second stage, bivariate analyses were performed to examine the association between depressive symptoms and the independent variable, mediator variable, and socio-economic variable using an independent t-test, one-way ANOVA, Pearson or Spearman correlation tests.
Mediation analyses were then computed in which the selected mediator was entered to test the components of the mediation model using the bootstrapping method to assess the indirect effects of the mediation model . Thus, the meditation model was examined by directly testing the significance of the indirect effect of the independent variable on the dependent variable through the mediator , while controlling for background variables that were identified earlier as significant in the bivariate analyses. This method is based on regression analysis, calculating the direct effect , total effect and indirect effects of an independent variable on a dependent variable. The total and specific indirect effects were calculated through bootstrapping set at 5,000 samples. Confidence intervals were calculated using this method by sorting the lowest to highest of these 5,000 samples of the original dataset, yielding a 95-percentile confidence interval . All analyses were run using SPSS 25.0 with the PROCESS statistical program . All estimated effects reported by PROCESS are unstandardized regression coefficients.
---
Results
Of the participants, 105 were women and 96 were men, ranging in age from 65 to 95 . Years of education ranged from 6 to 21 and 71.1% had a partner. The group had equal numbers of Arab and Jewish participants. There were significant differences based on ethnicity with regard to education, with Jewish participants reporting more years of education than Arab participants; the Cohen's effect size value was high. Arab participants reported a significantly higher level than Jewish participants of loneliness , malnutrition , and depressive symptoms . For most variables, the Cohen's effect size values were relatively high .
Table 2 presents the bivariate tests between the study variables with depressive symptoms as the dependent variable. Results revealed that all demographic variables, except for gender and marital status, were significantly related to depressive symptoms. Advanced age was positively correlated with depressive symptoms. Lower educational level was correlated with higher reported depressive symptoms. An ethnic difference in depressive symptoms was also found with Arab participants reporting higher depressive symptoms than Jewish participants. Both the independent and mediator variables were significantly related to the dependent variable-depressive symptoms. High malnutrition and high loneliness were positively associated with depressive symptoms.
---
The Mediation Analyses
Using PROCESS model 4, we tested hypotheses two and three, whether malnutrition mediated the relationship between loneliness feelings and depressive symptoms controlling for covariates . The results indicated a significant total direct effect of loneliness feelings on depressive symptoms, a significant direct effect, and a significant indirect effect through malnutrition. The results also showed that loneliness feelings were associated with higher malnutrition scores and that malnutrition was positively associated with depressive symptoms. Finally, ethnicity was associated with depressive symptoms. However, no significant associations were found between depressive symptoms and the other covariates: gender, age, education, and marital status.
---
Discussion
The general purpose of this study was to examine the extent to which loneliness feelings are connected with depressive symptoms of older adults from different cultures during isolation by conducting research during a Covid-19 pandemic quarantine in Israel. We measured whether older adults who suffered from higher degrees of loneliness feelings due to the Covid-19 quarantine had higher levels of depressive symptoms and what potentially accounted for this association. The findings confirmed the first hypothesis that older adults suffering from a high degree of loneliness feelings suffer from higher levels of depressive symptoms. These results are consistent with other studies executed before and during the Covid-19 pandemic, reporting that loneliness feelings are associated with mental wellbeing .
The mediation model confirmed the second hypothesis that the connection between loneliness feelings and depressive symptoms is accounted for by level of malnutrition. The association between loneliness feelings, depressive symptoms, and level of malnutrition can be explained by two different aspects: dietary behavior and the influence nutrition has on depressive symptoms. With regard to dietary behavior, loneliness feelings may affect appetite and nutrient intake through a decline in mood, physical functioning, or cognition . These various declines combine with the difficulty people have eating alone. Appetite can be further inhibited by changes in social status, particularly when older adults experience loneliness and/or bereavement due to loss of a spouse or friends of the same age-group . Eating in the company of others can help prevent malnutrition. It increases caloric intake and is related to healthier food habits , and maintains the motivation of older adults to eat and cook, providing them with opportunities for social interaction and connectedness .
The current study took place during a Covid-19 quarantine when older adults, as a high-risk group, were counseled to stay at home with their permanent partners only and to avoid, as much as possible, from going shopping and to depend on home deliveries. Hence, it is likely that older adults were forced not only to eat alone, but to be alone most of the time, resulting in insufficient food intake and, consequently, malnutrition.
The second aspect concerns the connection between depression and the diet of older adults. Recent studies have suggested that depressive symptoms are more prevalent in individuals with impaired nutritional status than in other older patients. It has been observed that individuals with specific nutritional deficiencies such as lack of folic acid and vitamin B12 as well as antioxidant vitamins had more depressive symptoms than those with normal nutritional status . The mechanism that might be at work here is that older adults who suffer from loneliness feelings tend to eat less and lack a healthy appetite and, for that reason, they will not consume all the nutrients they need and will suffer from malnutrition or depressive symptoms. This mechanism accelerates during a quarantine, when even older adults who do not suffer from loneliness ordinarily suffer from isolation. The third study hypothesis was that research variable levels will differ between cultures, particularly due to the extremity of the Covid-19 quarantine. Indeed, Arab older adults reported greater loneliness feelings, higher level depressive symptoms, and greater malnutrition compared to the Jewish older adults. The results with regard to depressive symptoms are consistent with studies reporting that being a member of an ethnic or racial minority is a risk factor for depression. Specifically, Arab respondents have reported higher depressive symptoms than Jewish respondents ; however, there are also contradictory results .
Concerning the higher levels of loneliness and malnutrition, the Covid-19 quarantine may have posed new challenges to the Arab study participants that explain the study findings. Arab society in Israel is known for family consolidation and solidarity. The extended family maintains a multi-generational structure, with grown children obligated to filial piety responsibilities . The Covid-19 quarantine forced new ways of communication upon family members, exposing the absence of digital literacy among Arab older adults: for Arab older adults, Internet use does not protect against loneliness . Moreover, the extended family often lives together, with the daughters and daughters-in-law cooking and serving the older adults . The quarantine isolated the Arab older adults from their extended family, resulting in reduced help with cooking and serving meals. Moreover, the strong connection between feeling loneliness and malnutrition cited above amplified the problem.
---
Conclusion and Implications
The present study was executed during a Covid-19 pandemic quarantine. The results indicate that there is a connection between loneliness feelings, depressive symptoms, and malnutrition. The primary conclusion is that loneliness feelings are a serious problem facing all older adults since they negatively affect both depressive symptoms and malnutrition. In order to overcome these feelings, it is important to connect older adults to social network technologies and teach them how to use these technologies. Families can also receive guidance for meeting while maintaining social distancing. Another option is to encourage neighbors in the same building to talk to each other and find ways for mutual support.
Research indicates that members of the Arab society in Israel are a high-risk group and much more vulnerable to loneliness feelings, depressive symptoms, and malnutrition in times of crisis such as the Covid-19 pandemic. The lives of many Arab older adults changed from being surrounded and supported by their extended family to being alone during the quarantine and to suffering from malnutrition. In order to fight these two related outcomes, one possible solution is to provide psychological-nutrition intervention by telephone. This would provide by missing human contact and nutrition guidelines and encouragement to cook and eat healthier foods. We should point out three main limitations of the current study. One is the cross-sectional study design, which does not allow for prediction of a causal relationship between the variables. A future study should use longitudinal data to examine the relationship between loneliness feelings and depressive symptoms. A further limitation might be the use of only one question concerning loneliness feelings. However, previous studies have also used one question in order to describe loneliness feelings . Third, a generalization of the findings is limited because the sample and the sampling procedure do not guarantee the representativeness of Jewish and Arab older adults. The sample was conducted by telephone and included only older adults who answered the telephone at that moment. Those who did not answer or did not have a telephone are not represented in this study. These various factors may have biased the results.
Despite these limitations, the present study provides initial insights into the mechanisms of the association between loneliness feelings, malnutrition, and depressive symptoms during periods of quarantine imposed isolation.
---
Conflict of Interest Statement: The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
---
Ethical standards:
The number of the approval from Tel Hai College is 3/2020-5. | Research on loneliness conducted in different countries has demonstrated that it is a common, universal phenomenon, although its prevalence varies between societies and cultures (11). In the United States, more than 19% of older adults aged 65 and older reported loneliness feelings (12); in Australia, the comparable figure was 40% (13). In a study of loneliness in Israel ( 14), it was found that nearly half (47. |
Introduction
There are disagreements about, and contradictory indications of, the state of qualitative inquiry as a global endeavour. Denzin , for example, argues that qualitative inquiry faces a global onslaught from conservative and neo-liberal critics, especially in the field of educational research, and especially in the United States following the writing into legislation of the requirement for federal educational research funding to be allocated to so-called 'scientific' research . 'Scientific' research has been extensively defined in US legislation, and includes reference to "measurements or observational methods that provide reliable and valid data across evaluators and observers… evaluated using experimental or quasi experimental designs…with a preference for random assignment experiments…" ). The argument of critics is that qualitative research is not scientific, or not scientific enough, and cannot produce definitive evidence about 'what works' in social policy interventions. Qualitative research produces too many disconnected, non-cumulative studies that do not provide convincing evidence for central education, health, business and so forth . A key point that Atklinson and Delamont make is that qualitative research is still widely funded in countries such as the UK paper for the UK Department of Education, the terms of the debate may be changing in the UK). Delamont and Atkinson also note the development of a number of successful interdisciplinary journals in the field along with increases in the numbers of issues per year of these journals and thus the numbers of papers published. They also note the continuously expanding market for qualitative research methods texts and sourcebooks such as Sage Handbooks and Major Works, not least Denzin and Lincoln's Handbook of Qualitative Research, now in its fourth edition .
There are some disagreements in the field then, in some respects representing different interpretations of the prospects for qualitative inquiry in the USA and the UK. However, what I want to argue in this paper is that these disagreements rather miss the point. The issue is not whether or not qualitative approaches to social research are developing across disciplines and continuing to receive funding and policy attention in diverse countries around the world, but rather to what purposes are qualitative methods being put? What research agendas are being pursued, and who sets them? It is clear that qualitative methods are indeed being taken up widely across regions and disciplines, with the scale and scope of the annual International Congress of Qualitative Inquiry from which this special issue derives being part of the positive evidence. It is also undeniable that, internationally, social research in general, qualitative approaches to social and educational research in particular, have been under very specific attack for fifteen year or more . Criticism is manifested in different ways, and with different levels of severity, in different countries and different disciplines. The pressure derives from concerns about the quality and the utility of social and educational research more generally, not just concerns about qualitative approaches . It also derives from an increased government focus on value-for-money in research, and how social research might better serve social policy .
Qualitative methods of inquiry can be seen to play to this policy and utility agenda as much as to a social justice agenda, or indeed a disinterested scientific agenda. Qualitative methods can be deployed in policy-driven research as well as in more open-ended inquiry, and thus may survive and even prosper, but not necessarily in ways that all qualitative researchers will welcome. Researchers and evaluators in many applied research fields including health care and education have noted the reduced timescales now associated with qualitative work, and the narrowness of many research agendas being pursued . This is also the case in what we might term the more traditional disciplines of anthropology and sociology. Mills and Radcliffe report on the increasingly truncated timescales for anthropological fieldwork and the use of a more limited range of ethnographic methods among 'freelance development consultants ' and in development agencies such as the World Bank. Thus while it can be argued that qualitative research remains widely supported, the form that such research takes might be said to be somewhat limited, and even of poor quality, given arguments for exploratory and extensive field immersion that are apparent in many ethnographic textbooks.
Such an orientation towards the use of qualitative methods is apparent in a range of policy development settings and intervention agencies. Thus for example, Valerie Caracelli , writing from the perspective of the US Government Accountability Office, argues for the inclusion of qualitative methods in evaluation studies "to assure contextual understanding" . She states that:
Recently, there has been an acknowledgement about how ethnographic studies can inform agency actions and how it can be used to study culture in organisations .
However whether truncated timescales and narrow policy agendas can be easily reconciled with the theory and practice of 'ethnographic studies' is a moot point. Moreover studying 'culture in organisations', in order to 'inform agency actions' seems to suggest the utilization of qualitative inquiry as as technology of government, rather than as an approach to understanding emergent issues that policy may not want to address.
A recent UK Cabinet Office report produced in the context of the move toward evidenceinformed policy and practice presented guidelines for judging the quality of qualitative approaches and methods. Interestingly, one of the key quotes in the report used to justify the use of qualitative methods comes not from the epistemological or methodological literature, but from a civil servant, a government department "research manager": I often commission qualitative research when it's about users or stakeholders and . . . It is in this context of the potential co-option of qualitative inquiry to the agendas of policy that I want to focus on a particular sub-issue of this wider debate: that of research selectivity and concentration in the UK and its implications for qualitative research methods.
Significant changes are being enacted in the UK, with regard to university finance, governance and the nature of scholarly activity, that carry potentially very severe implications for the practice of social research in general, and for qualitative approaches to research in particular. There is an argument that social research is increasingly being nationalised and corporatized in the UK -seen by many social researchers as well as policymakers as an arm of government and more important to universities for its incomegenerating potential than for its contribution to knowledge and the public good. Thus arguments about whether or not qualitative approaches to research remain popular are rather irrelevant. What is at stake is the legitimacy and quality of the research which is being undertaken, and the purposes to which it is being put. As noted above, policymakers, and some social science scholars themselves, have argued that social and educational research, particularly qualitative research, is too often conceived and conducted as a small scale 'cottage industry', producing too many small scale, non-cumulative studies which do not provide firm evidence for decision-making. It is axiomatic to such criticisms that providing evidence for policymaking is indeed the proper role for social science.
Additionally, governments around the world are seeking better value for money from their investment in research and university teaching, and this has involved restricting and focusing resource allocation. Selectivity and concentration of research resources are particularly being pursued in the UK. The Conservative dominated coalition government is cutting public spending in the wake of the 2008 banking crisis and global recession; so there are fewer resources available for research than might otherwise have been the case, and selectivity and concentration have become even more severe in the UK over the last couple of years. Concentration of research is effected both through focused core allocations from the higher education funding agencies and through highly competitive bidding to research councils and foundations as described below. Success leads to further success and to relatively few universities securing the overwhelming majority of available funding. This in turn produces the issue of universities seeking research funding first and foremost for their own corporate survival, rather than for the public benefits that may accrue. It also leads to social research becoming part of a nationalized approach to managing national social problems, rather than being part of an international community pursuing better understanding of the nature of social problems and what produces them.
Funding agencies and individual universities are now concentrating resources on fewer research units and programs, and are taking decisions to develop a 'big science' model of social science. This is being pursued by funders supporting fewer, larger projects, with explicit policy encouragement for researchers to develop cross-institutional, mixed method approaches, to address the supposedly 'big issues' of our time: health and well being, an aging population, sustainable growth, and so forth. These issues are indeed important, and research evidence should be produced to interrogate and inform public debate. But such issues are being presented as part of a common-sense, taken-for-granted trade-off of government funding in exchange for social scientists serving policy. Critique, diversity of perspective, and the insight into complexity which detailed qualitative studies can provide are potentially being marginalized. Social science is being reconceptualised as a technical service to government rather than developed as a democratic intellectual resource for the community.
The remainder of the paper will provide some brief illustrative examples of policy initiatives and how this change is being accomplished.
---
Funding universities
University policy and funding is located in the UK government department of Business, Innovation and Skills -evidence in itself of where government priorities lie.
Universities are not housed in the Department for Education . A recent BIS White Paper stated quite clearly that "we intend to maximize the impact of our research base on economic growth" . It went on:
To compete effectively the UK must harness its strengths in…research…and its expertise in areas such as design and behavioural science… .
In essence the White Paper argues that investment in research should be oriented to those areas that promise most economic return, with "behavioural science" being deployed to understand and change people's behaviour in relation to key threats to economic development such as poor health and global security. Moreover the White Paper goes on to assert that major social and economic challenges "can only be resolved through interdisciplinary collaboration" and thus government will "actively support strong collaborations" across disciplines and institutions. Research, including social research, will be marshaled and directed in the national economic interest.
Such policy discourse then sets the tone for the activities of intermediary agencies such as research funding councils and individual universities. While individual research councils can set their own agenda, their budgets derive from government are unlikely to stray too far from policy imperatives. Similarly universities, while independent, must compete for specific forms of funding under common national rules and thus common institutional structures of research prioritization, monitoring and compliance emerge across institutions. The research allocation which universities receive from the Higher Education Funding Council is based on quality ratings derived from an accountability exercise originally called the Research Assessment Exercise , and now known in its current iteration as the Research Excellence Framework . As a result of these exercises funding has become highly concentrated in historically elite institutions, with, for example, only four universities receiving 32% of all HEFC research funding in 2009-10, and 25 universities receiving 75% of funding . Of course it might be argued that not receiving such funding is not necessarily such a bad thing, since it leaves the majority of universities outside the mainstream funding agenda and thus outside this particular mechanism of government control. However the mainstream agenda remains dominant since university managers simply cannot ignore it. Some research success, however small, is regarded as better than none, especially with respect to the reputational gains associated with research activity and associated student fee income. For all practical purposes, when it comes to research funding in the UK, the REF is the only game in town.
Clearly, this level of selectivity begs serious issues for university finance, management and The pursuit and measurement of 'Impact' is likely to drive research activity further towards applied and policy-oriented research. Moreover the pursuit of specific research tenders and contracts is also likely to be increasingly emphasized by individual universities, as noted above, both to fill gaps in core funding left by increasing selectivity and concentration, and to maximize the metrics available for the 'Environment' element of the profile. So this element of core research funding, derived from the Higher Education Funding Council, which is intended to provide the underpinning platform for basic university research, is becoming increasingly selective and concentrated, and is now being significantly oriented towards applied rather than basic research. Thus while, technically, the agenda for this research activity remains under the control of individual scholars and research groups, increasingly they need to think about developing research programs which fit with government, university and departmental priorities, in order to maximize quality gradings and income. In turn universities and departments are developing strategies and programs to manage the REF accountability procedures, and research impact and environment issues more generally, such that research is becoming an increasingly planned, bureaucratized and managed activity. Ironically, given the current emphasis in some quarters on the 'gold standard' of scientific method, university research is ceasing to be scientific in the sense of the disinterested pursuit of knowledge and becoming a corporate, commercial activity undertaken on behalf of individual universities themselves and their financial survival. In turn, the social relations of research within universities and departments are changing.
It might of course be argued that 'twas ever thus' -universities have always had to fund their activities and individual scholars and research groups have always had to manage and mediate the relationship between funding opportunities and the research that they think is important to design and undertake. My argument is that this balance of calculation is under severe pressure in the UK at the present time -both with respect to the overall funding available and with respect to the position of qualitative research activity within this funding environment. It might be further argued that this balance should change, and that the pursuit of more applied policy-oriented research should be undertaken; a second paper could be written about the issues at stake and the arguments deployed. My point for moment however, is simply that such change does indeed seem to be occurring.
---
Funding social science
Still, to reiterate, in principle the research agenda pursued with this funding regime remains under the control of faculty and research groups. However, another irony of the current situation in the UK is that, if anything, the funding strategy of the research council -the Economic and Social Research Council -further compounds the problem rather than providing any sort of counter weight. In principle the ESRC is an 'arms length' body, independent of government, allocating awards for excellent social science research in response to competitive bids refereed by peer review. It might be thought, therefore, that the ESRC's agenda would reflect the research agenda of the social scientific community as a whole. However all research council funding derives from government and as such is clearly influenced by government priorities; the ESRC is no exception. The ESRC shapes the content of the social science research agenda much more directly than government policy, but does so in large part, because its funds derive from government. Again, of course, there are many intermediary processes and activities, with ESRC officials seeking to maximize funding available in difficult times, in return for responding to government priorities. Likewise peers review proposals on merit but these proposals have already been produced in response to priority areas and specific funding calls.
The ESRC works within the context of an overarching cross-research council strategy. There are seven research councils in the UK, distributing funds across the natural and social sciences, humanities, and medicine. Each council has become progressively more managerial over recent years, not simply responding to bids from the scholarly community for funding, but actively shaping the agenda around which bids can be made -establishing priorities and issuing specific calls for proposals, as noted above. The assumptions embedded in this statement of strategy reflect the development of an institution which sees itself governing social science rather than supporting it -defining priorities, selecting "future leaders" and concentrating PhD training in a few selected centres so that those leaders are in any case selected from an increasingly narrow institutional and intellectual base. The use of the term "training" for the development of future scholars also seems indicative of a directive rather than a supportive role. The "Deliver Plan" goes on to address this specifically and states:
We will develop national capability through: …Broadening the skill of all social science PhD students by emphasizing transferable skills training…We will require institutions to provide training on core topics such as impact, public engagement and media training…to ensure the continuing pipeline of excellent researchers for the Nation .
The Strategic Plan and Delivery Plan also identify a need for the longer, larger projects to be interdisciplinary and involve cross-institutional collaborations which are defined as "essential in studying and resolving complex challenges" . In effect social science is being re-conceived and re-structured as the "behavioural science" arm of government, so that social science can 'influence behavior and inform interventions'.
It might be argued that because ESRC's budget and level of activity is very small when compared with overall funding for social science in the UK, its influence will be similarly small. The ESRC budget is c. £166M in 2014/15 . The total 'science' budget for UK higher education is £4.6B . Detailed breakdowns between different natural and social science allocations are difficult to identify but it is likely that ESRC does not fund more than 10% of overall social science teaching and research activity including that supported from student fees. Thus, again, there is a case for suggesting that most UK social science research is beyond the reach of ESRC and, in turn, government policy agendas. But, as with the REF, influence far exceeds scale of activity, as universities and research groups seek to bid for research council funding despite success rates dipping well below 20%, and develop postgraduate training activities to mirror ESRC provision so that they are not excluded from future funding possibilities. Nevertheless further concentration of funding might be construed as an opportunity as well as a challenge for qualitative social research, to reach out to community support and other forms of charitable, foundation and European Union funding. An interestingly critical issue for the current policy of selectivity concentration is at what point might the nationally state-supported funding base shrink so low that the research council's influence over the sector as a whole will disappear? Furthermore, there is a separate paper to be written about whether or not, and if so in what ways, social science should develop so that it can 'influence behavior and inform interventions' on behalf of government. Many would argue that social science should engage more directly with the public and help to inform democratic debate and decision-making, though such activity is not necessarily co-terminous with simply acting in response to and in support of government policy .
Likewise, with respect to the content of research 'training', some would point to statistics indicating that only around 20% of research council supported research students secure academic posts in universities and thus a wider training in 'employability' is important . Yet by the same token the statistics indicate that 80% go on to work in researchrelated and more general administrative, teaching and managerial roles in other public sector, charitable and commercial organisations. If these figures are accurate then not only is such a problem not the responsibility of ESRC to solve, it is not actually a problem at all, since PhD graduates are already securing employment outside academia and the influence of social science training could be said to be extending well beyond university departments. In fact students often value these broader elements of their doctoral programs as academic employment opportunities are so limited. Nevertheless time spent on 'public engagement' and 'media training' is time not spent on fieldwork and data analysis -the core of any qualitative research training. These matters are not necessarily clear-cut, but my point for the moment is simply that these developments are shaping social science in general in the UK at the present time, qualitative educational and social research in particular.
---
Short term consequences and implications
The first and most obvious consequence is that relatively small scale funding to undertake specific qualitative and case study-type work is no longer available from ESRC. Hitherto ESRC ran a specific 'small grant' scheme which was able to accommodate small scale exploratory investigations, pilot projects and post-doc projects from early career researchers. It was particularly suitable for supporting individual scholars to undertake detailed case study work over the period of a year or eighteen months, by 'buying their time' to concentrate on research for short periods; and of course because the grants were small, more could be funded. In the absence of knowing in advance which project would show most theoretical and empirical advance, funding more small scale projects seems sensible. This support is no longer available. The smallest 'open call' grant now available is £200K -still modest by international standards -but substantial enough to indicate that a larger scale mixed methods approach would be more likely to win funding when evaluated against 'value-for-money' criteria. Additionally ESRC used to support a 'first grant' scheme, which was similarly oriented, as the name implies, towards early career researchers who, again, often applied for small scale funding for qualitative work. This Thus qualitative work must now be conceived and proposed either in terms of much larger scale longitudinal ethnographic investigations , or as part of a large scale, cross disciplinary and cross-institutional mixed methods research design. Again, involvement in such work is no bad thing, and no doubt will contribute to the continuing popularity which Delamont andAtkinson claim for qualitative methods in the UK, but it should not be the only qualitative research that secures support. Moreover, when such large scale investigations are only funded because government wants to 'influence behavior and inform interventions', be they mixed method or ethnographic, and perhaps especially when they are ethnographic, then they beg many questions about the legitimate role of social science in a democracy. As noted above, qualitative research 'is a wonderful vehicle if you want to understand the motives of people'.
The key issues here comprise both the scale of endeavour now expected, and the control of the research agenda. Social scientists are being positioned, by government and by some leading members of the social science community as experts in social policy whose function is to respond to social problems. Again, social science should indeed be prepared to respond to social issues, but not to the exclusion of critical social inquiry. Furthermore the 'expertise' of the community depends on what we might term a hinterland of basic social theory and social research. From this hinterland expertise can be drawn and framed in response to particular issues -the move from basic to applied research. But if social science can only now act in relation to prescribed issues, problems and large scale empirical research designs then, over time, it will lose the ability to construct its objects of study independent of the context of inquiry. It will in effect cease to be social science in any meaningful sense.
With respect to the social relations of research, opportunities for early and mid career social scientists to develop their own intellectual trajectories are likely to decrease as areas of strategic investment are defined by funding agencies. Moreover large scale grants are likely only to be awarded to senior researchers who have a track record of managing and 'delivering' on previously funded work. Thus early career researchers must now attend to bidding for research grants, especially in areas of applied research, and look to secure collaborative funding with more senior colleagues. This is now a core feature of any social science career. Developing scholars must also become able to design impact strategies and build networks, including with policymakers as well as other scholars, in order to become included in collaborative research designs. In effect involvement in social research is being reconstructed as a technical and professional career, and indeed a quasi-governmental career, rather than as a contribution to science, or as an independent and critical service to the community and the democratic process. The very nature of the purposes and practices of social research is being changed.
My argument is not that that these matters are entirely novel or uncontentious. Social science has had an uneasy relationship with government funding for many years . Similarly, outcomes are not easily determined.
There are many mediating interpretations, actions and institutional processes between grandiose policy documents, often written first and foremost to secure funding rather than control it, and the activities of individual researchers and research groups. Yet the material and discursive context of interpretation and mediation means the direction of travel becomes ever more difficult to resist. Nor is my argument that none of these things should be happening; to reiterate, universities, research groups and individual scholars have always had to fund their work and certainly should be responsive to the needs of social policy development. However, responsiveness to policy and to wider contexts of action should take their proper place in the overall career trajectories of social researchers and scholars more generally. Intellectual curiosity, about how society operates and with what effects and consequences for individuals and social groups, must remain the driving force of social science and scholarly activity. Otherwise it simply becomes a technology in service of government, rather than an independent and critical intellectual resource for democratic debate.
---
Longer term trends and prospects
A more general issue is whether or not the social science community is right to accept that proximity to government will necessarily ensure its sustainability. Part of the backcloth to the current debate is the uncertain status and legitimacy of both science and government at the present time. Government, and the process of mainstream electoral politics, is itself generally unpopular and under pressure 'to deliver', especially with respect to economic competence and with regard to the provision of public services. Can we always assume the benign intent and impact of government? What reasons are there for state intervention in the lives of ordinary citizens? How appropriate is it for social research to attempt to 'influence behaviour and inform interventions'? It is at least arguable that government intervention can dis-empower communities, and it certainly locates agency in government and professional bureaucracies, including those of social science, rather than local communities. A different approach would involve social research helping to build communities' capacities to develop themselves, rather than simply providing evidence for central policymaking and the development and evaluation of government intervention programs. In this respect it may be the case that deriving legitimacy for social research from proximity to government is self-defeating. Such a strategy links social research to an inherently unpopular institution and at one and the same compromises the basic claim for the legitimacy of science -that of disinterested inquiry. Of course many qualitative researchers also want to pursue a social justice agenda, not just a social scientific agenda. But here too, collaborating with local organisations, institutions and communities, rather than government, would seem to hold much more promise with respect to both the quality of the research and its potential 'impact' on social and economic life.
A second element of an alternative approach to the further development of qualitative research must be to maintain involvement in international networks and debates, and resist attempts to render qualitative research into a parochial set of research techniques to be pursued in the national interest. The use of qualitative research methods has a long and distinguished history in education and the social sciences. Discussions of qualitative methodology have been at the forefront of many decades of debate over whether, and if so in what ways, we can conduct enquiry and build knowledge in the social sciences. It might even be argued that it is qualitative methods, or perhaps, more generally, a qualitative sensibility and approach, that constantly questions the development of social research as a technology.
The strengths of qualitative research are at their most manifest when used to address both the substantive topic under investigation and the way in which that topic is constituted and realised in action by the social processes at play. The phenomenon under study is not simply taken as given.
Thus local activity and development can draw intellectual sustenance from global debates. A key strength of a qualitative approaches to social research is face-to-face engagement with participants. This must be maintained as both an ethical and political strength, as well as an epistemological strength. Many recent international discussions of quality in qualitative research revolve around issues of engagement, deliberation, ethical process and responsiveness to participant agendas, along with the need to maintain a critical perspective on both the topic at hand and the power of particular forms of knowledge . It is these strengths of a qualitative approach that should be privileged in any discussion of their inclusion larger scale social research activities, not in order to 'influence behaviour' but in order to maximise the possibilities for the democratic development of research procedures. If research is to serve the periphery as much as the centre, in political debate and decisionmaking, then engagement of research participants in both setting the research agenda and evaluating the outcomes of the process must be developed as central to the future development of social research methodology. | There are disagreements about, and contradictory indications of, the state of qualitative inquiry as a global endeavour. The paper argues that these disagreements rather miss the point. The issue is not whether or not qualitative approaches to social research are developing across disciplines and continuing to receive funding, but rather to what purposes are qualitative methods being put? What research agendas are being pursued, and who sets them? The paper illustrates these issues with respect to the effects of research selectivity and concentration in the UK and argues that qualitative research should continue to develop as an iuntellectual resource for the community rather than as a technical service to government. policy-making. What is required is the 'gold standard' of randomised controlled trials (RCTs) for establishing cause and effect in social programs. Such criticism is not confined to the USA. In a speech to the UK's Economic and Social Research Council in 2000, titled "Influence or Irrelevance" the then Secretary of State for Education, David Blunkett (2000), asserted that Many feel that too much social science research is inward-looking, too piecemeal, rather than helping to build knowledge in a cumulative way, and fails to focus on the key issues of concern to policy-makers, practitioners and the public (p.1). More recently, similar arguments have arisen once again in a paper from the UK Department for Education. In language paralleling the US debate, the paper asserts that randomized assignment of treatment to samples is the most appropriate way to conduct research in general, educational research in particular, and that more experimental research should be undertaken in education, specifically via RCTs (Goldacre 2013). Such criticisms combine critique of methods employed, with critique of purposes pursued, arguing that social research in general, qualitative approaches to social and educational research in particular, are not responsive enough to the needs of government. So significant challenges to qualitative inquiry are apparent, but others are more cautious in their assessment of these challenges. Fielding (2010) for example shares Denzin's concerns with respect to 'the dirigisme and anti-intellectualism of the gold standards lobby' (p.127) but notes that circumstances vary internationally and that threats to qualitative research should not simply be read from a North American perspective. Others still, argue that qualitative inquiry is healthy, diverse and growing, both geographically in terms of its global reach, and in terms of its diffusion across disciplines and across applied research fields such as |
Introduction
Numerous studies have demonstrated that social networks have a positive effect on finding a job [1][2][3][4]. Along with the gradual introduction of social network analysis into research on China, many studies have directly focused on the Chinese labor market [5][6][7][8][9][10][11][12][13]. In general, job seekers who use social networks are able to attain higher earnings and a better occupational status than job seekers who do not use social networks. Studies on China also tend to place more emphasis on the institutional environment, linking social network analysis to the market reform process [7,11]. As an informal source of resource allocation, social networks play an important role alongside formal sources such as the market and the government. In the early stages of China's market reform, when the rules and norms of the market were not sufficiently clear, the role of social networks as informal channels for resource allocation gradually increased. As marketization deepened over time, the market rules became clearer and stronger, and the space for network-based activity declined. This means that the effect of social networks followed an inverse U-shaped trend as marketization intensified over time.
However, recent research has shown that the percentage of individuals who use social networks when searching for a job does not follow an inverse U-shaped pattern as supposed; rather, it has continued increasing and, according to recent data analysis, has reached 80% . This means that it is now common to obtain help from one's social networks when seeking a job. We have strong reason to believe that those actors who use their social networks when seeking a job are not a small, special group; given that nearly 80% of job seekers utilize their social networks, the heterogeneity within this group is likely very large. Against the background of an increasing percentage of job seekers using social networks and the increase in heterogeneity, a new but very important question arises: Are the returns to social networks among those actors who use social networks when seeking a job consistent? When we focus on job seekers who use social networks, do the returns to social networks vary from person to person or group to group? This question discusses the return of social networks, which is in fact the endogeneity of social networks. Endogeneity has been a hot topic in the fields of social networks and social capital, and the existence of endogeneity has led the scholars to doubt whether there are really returns to social capital accordingly. In this study, we fully examine the complete causality and further point out the existence of group differences in the returns to social capital while discussing the returns to social capital. This will help us to answer the endogenous challenge of social networks and help us to have a more complete and accurate understanding of the returns of social networks in the labor market.
---
Effects of Social Networks: The Origin of Heterogenous Returns
Social network analysis has attracted abundant interest in and research on the effects of social networks since the 1970s, the beginning of the rise of social network research. Regarding whether social networks play a role in the labor market and, if so, how, various theoretical perspectives, such as the network strength approach [1,4], network resource approach [3], network structure approach [14], and network signal approach [15], have been discussed. Whether considering strong or weak ties under the network strength approach, information or favorable treatment under the network resource approach, or structural holes under the network structure approach, the effect of social networks has been comprehensively demonstrated. Among the many fields that have incorporated social network analysis, the study of the labor market has always been at the forefront of social network analysis, and the job-search process is a good context in which to examine the role of social networks. As discussed above, the effects of social networks may be diverse, as most job seekers use social networks [16]. When exploring the role of social networks, a failure to account for differences in the relationship between social networks and status attainment imposes limitations. When we move away from an overall analytical perspective and instead analyze the differences in the returns to social networks, then there are two possible theoretical assumptions: positive returns and negative returns.
The social capital embedded in social networks has both instrumental and expressive effects [17]. Most of the previous studies on the effects of social networks in the labor market have focused on the income effects of social networks. In fact, the process of acquiring status in the labor market and the corresponding outcomes involve both the instrumental effects represented by income and the expressive effects of the job search process and its outcome. One of the most direct expressive effects is job satisfaction after the job search. Job satisfaction refers to the degree of satisfaction with the job found and is a subjective evaluation made by the job seeker. It is a global evaluation and may include dimensions such as satisfaction with the job itself, satisfaction with the salary package, satisfaction with the opportunities for promotion, and satisfaction with colleague relationships. The economic behavior of actors is embedded within a social structure, and job seekers might not simply choose the job with the highest salary and income when seeking an occupation but may be motivated by a global evaluation of various other non-instrumental indicators, such as colleague relationships and career development. Therefore, when we analyze the returns to social networks, we should not limit ourselves to the instrumental indicator of the returns in terms of earnings but should consider both instrumental and expressive returns. Both of these dimensions have two possible theoretical relationships with the use of social networks: positive return or negative return.
---
Positive Returns Hypothesis
In the social capital theory, Nan Lin describes a pyramidal social structure in which each actor is located at a certain point on the pyramid [3]. Social resources also exhibit this pyramidal structure. The higher up the actors are within this structure, the fewer positions and occupiers there are [3]. There are very few positions and occupiers at the top level, where those who control the largest number of valuable resources are located. The further up the structure one goes, the fewer actors and the more resources there are. At the bottom of the structure, the number of actors is larger, and the resources are fewer. In his theory, Lin proposed the "strength-of-position" proposition: the better the position of origin is, the more likely the actor is to access and use social capital, which is likely to be of higher quality [17].
The strength-of-position proposition is consistent with traditional structural theory, which describes the structural advantages of each actor and extends to social capital research. The initial position refers to both the actor's ascribed position and achieved position. The ascribed position is the position inherited by the individual from their parents and relatives; the achieved position refers to the social position acquired and occupied through self-effort. A better initial position results not only in more social resources being available to the actor but also in human capital, political capital, and family background advantages, among others. When social capital is combined with other advantages, it can result in combined advantages such that the whole is greater than the sum of its parts.
However, actors with an average or worse initial position are at a double disadvantage: on the one hand, due to structural constraints, the social resources that they can mobilize through their social networks are inferior in both quantity and quality to those of actors with a better initial position. On the other hand, they also lack other relevant advantages that could be combined with their social capital in order to exploit the combined advantages.
When an actor has a weak endowment of all types of capital, the whole may be less than the sum of its parts. In summary, the better the initial position of the actor, the more likely that actor is to use their social network. Moreover, the more likely the actor is to use their network, the greater the return from the social network is likely to be. In this paper, we take current employment earnings and job satisfaction as the returns of getting a job. So, we state the research hypothesis of positive returns to social networks as follows:
Hypothesis 1.1: The more likely an actor is to use their social networks, the higher the current employment earnings they would obtain.
---
Hypothesis 1.2:
The more likely an actor is to use their social networks, the more satisfied with work they would be.
---
Negative Returns Hypothesis
Bourdieu proposed the concept of social capital in contrast to economic capital and cultural capital at the beginning of his first elaboration of capital in "The Forms of Capital" [18]. Social capital, similar to economic capital, human capital, and cultural capital, is one of "the forms of capital", and the return to such capital in terms of the income of job seekers may exhibit diminishing marginal effects with an increase in capital types. The importance of social capital is relatively minor when the actor has a sufficient number of other types of capital that can generate returns, while the importance of social capital is inevitably higher when the actor has access to fewer types of capital or even to social capital only.
Returning to the pyramidal social structure described by Nan Lin, those actors who are initially in a better position have relatively superior social capital, human capital, and political capital. According to the previous literature, human capital and political capital have a positive effect on actors in the Chinese labor market [19][20][21]. Therefore, in addition to social capital, other types of capital, such as human capital and political capital, provide higher rewards to those actors who are initially better positioned. The return to the social capital embedded in social networks is a return to only one of many types of capital and is simply icing on the cake.
Accordingly, for those actors with a poor initial position, the returns to social capital account for a greater share of their overall earnings due to their relative lack of other types of capital. The use or nonuse of social networks has a relatively greater or even decisive impact on such actors' earnings. For these actors, the returns to social capital obtained through the use of social networks are relatively more important and can often be a much-needed helping hand. Therefore, the better the initial position of the actor is, the more likely that actor is to use his or her social network. However, the less likely the actor is to use his or her network, the greater the returns to the social network are likely to be. Accordingly, we obtain our research hypothesis on the negative returns to social networks.
---
Hypothesis 2.1:
The less likely an actor is to use their social networks, the higher the current employment earnings they would obtain.
---
Hypothesis 2.2:
The less likely an actor is to use their social networks, the more satisfied with work they would be.
---
Two Types of Selection
Bias: Who Prefers to Use Social Networks? Who Benefits More from Social Networks? Pretreatment and Treatment Effect Heterogeneity On the basis of the principles of homogeneity and endogeneity, Ted Mouw raised suspicion of the social network effect in 2003 [22,23]. This resulted in a great deal of social network research in response to Mouw's challenge [24]. A variety of methods have been used to address endogeneity. These include the Heckman two-step selection model, propensity score matching, instrumental variables, simultaneous equations system, and suspicion [25][26][27][28][29][30][31]. Recently, Brand, Yu, and their colleagues pointed out in a series of studies that when we want to accurately estimate a causal relationship, we should consider all of the endogeneity along the causal chain: the pretreatment heterogeneity bias and the treatment-effect heterogeneity bias [32][33][34]. Specifically, first, people are not randomly assigned to an experimental group and a control group ; rather, people with certain characteristics are more likely to become members of the experimental group. On the other hand, people with these characteristics are not only more likely to be in the experimental group, but they are also likely to obtain benefits that differ from those of their peers without these characteristics [33,35].
Endogeneity is a universal and unavoidable problem in all social sciences. However, since Mouw's criticism of how endogeneity has been handled in social network analysis, the approach has been questioned and challenged. This is because the establishment of social networks-interactions among actors-is always purposeful and highly subjective. This makes it difficult for samples to satisfy the requirement that treatment status be randomly distributed and leads to estimation bias in the regression analysis. Given this difficulty, combined with the challenge of the complete endogeneity along the causal chain mentioned above, both of which affect the research on social networks in job search, we need to rethink the returns to social networks in the labor market. If there is endogeneity bias in the form of pretreatment heterogeneity bias, then people with certain characteristics are more likely to use social networks in when job searching. If there is endogeneity bias in the treatment effects, then people with those certain characteristics are not only more likely to use social networks but the returns that they obtain from their social networks are likely to vary, either increasing or decreasing with an increase in the likelihood of use. These two types of endogeneity bias are exactly what we try to analyze in this paper. Recent research from China has pointed out that the use of social networks in job searching is not randomly distributed and that those people who use social networks share common behavioral traits [36]. Actors who are strongly relational are more likely to use social networks when searching for a job. A preliminary study focused on investment in everyday relationships to discuss relational tendencies [36]. It has been demonstrated that the more actors invested in everyday relationships, the higher their tendency to use guanxi when engaging in instrumental actions, such as searching for a job. Apart from attitudes toward relationalism, are there any other external characteristics that identify a behavioral tendency toward relationalism? However, there is still a further problem: since social networks are not randomly selected, are the returns to social networks the same among different groups of people? This paper focuses on those people who use social networks in their job search and explores whether the effects of social networks vary from person to person. Specifically, in light of the two competing theoretical hypotheses proposed above, is it true that the more likely people are to use social networks, the higher the returns to social networks are? Or is it the case that the more likely people are to use social networks, the lower the returns are?
---
Data, Measures, and Methods
---
Data
We used data from the 2014 Job-Search Networks survey to examine the heterogeneity in the effects of social networks on job search. JSNET2014 contains data on individuals residing in 8 large cities throughout China that is representative data with a total sample size of 5480. This survey used probability proportional to size sampling, in which a random sample was taken from three levels for each city. In each randomly selected neighborhood, households were selected from a household list through a systematic sampling technique. To obtain a complete list of both permanent and migrant households, a map of each neighborhood was drawn, and a list of the households that live there was made. From each randomly sampled household, an adult aged 18-69 was selected to be the respondent by randomly selecting birthdays. Above all, JSNET provides a representative sample and reliable results.
The JSNET data are an excellent source of information about social networks and job search, which are the key variables in our research. Unfortunately, there are no more data on job search networks update after 2014. Therefore, so far, these data are the most appropriate and the latest data on the measurement of social networks and occupationally related variables.
---
Measures
---
Dependent Variable
Earnings. "Earnings" here refers to current employment earnings, which are the occupational earnings of the worker at the beginning of their entry into their current job. This study focuses on the relationship between the social networks used by job seekers in their job search and the subsequent occupational earnings obtained, and to strictly preserve the temporal chain of causality, we directly used the occupational earnings at job entry. The rapid economic development over the past 40 years since China's reform and opening up has resulted in substantial price increases. Therefore, we used the Consumer Price Index for each city in each year to eliminate the effects of inflation, allowing occupational earnings to be comparable across each generation of new workers. Moreover, due to the right-skewness of the original income distribution, after eliminating the effects of inflation, we then took the logarithm of the earnings to bring the distribution of earnings, the dependent variable, closer to a normal distribution.
Job satisfaction. Job satisfaction is a global indicator. We focused on job seekers at job entry and on the following five dimensions of satisfaction: the work itself, pay and benefits, opportunities for promotion, colleague relationships, and upper management. Each item was measured with a 5-point Likert scale ranging from "very unsatisfied" to "very satisfied". We summed up the scores across the five items and found the average, which represents the overall satisfaction of job seekers with the results of their job search.
---
Independent Variables
The Use of Social Networks. The use of social networks is both the most dominant independent variable in the model for predicting income and is the dependent variable in the model for predicting the likelihood of using social networks . In coding the variables, "the use of social networks" was defined not only for those respondents who directly answered that they had used social networks when searching for their job but also for those who mentioned using social networks in other follow-up questions. This is because in Chinese social life, the use of social networks, i.e., the use of connections, is often seen as synonymous with "going through the back door", which often has a negative connotation of unfairness. Some respondents were reluctant or hesitant to answer directly that they had used their social networks, but in the follow-up questions, they expressed indirectly that they had done so.
---
Control Variables
We measured respondents' attitudes toward relationalism by their willingness to "handle things through social networks". This is a continuous variable, with a higher score indicating that respondents were more receptive to the idea of handling things through social networks and thus that they are more accepting of relationalism.
The control variables in this paper also include personal attributes, family background, and structural variables such as gender, age, hukou, education, Party membership, father's education, father's Party membership, labor market sector, an indicator for high-income industries, occupational requirements, and year of job entry. All variables used in this study can be found in Table 1.
---
Methods
To test whether the income return to social networks varies according to the propensity of the actors to use social networks when searching for a job, we used the stratificationmultilevel method [33] to develop our model for empirical analysis. Our approach is similar to propensity score subclassification [27] in that the whole sample was divided into a number of balanced subsamples, which were then analyzed in the next step. The specific analysis in this study followed four steps.
Step 1. First, we needed to discuss the probability with which social networks are used during job search; this probability is unknown and needs to be estimated. We developed a logit regression model with a dummy variable as the dependent variable, and the resulting predicted probability is the probability that the actors use their social networks, with higher values indicating a higher probability.
Step 2. We constructed balanced propensity score strata following a principle that we call "homogeneity within groups, heterogeneity between groups". Specifically, there should be no significant differences in the average values of covariates and in propensity scores between the treatment and control groups. In this study, "homogeneity within groups" means that the two types of job seekers, those who use social networks and those who do not, were randomly distributed across the subsamples defined by the propensity scores and that there were no significant differences in the distribution of those independent variables that influence the use of social networks. "Heterogeneity between groups" means that the different subsamples had significantly different probabilities of using social networks: the higher the stratification of a subsample, the more likely it is to represent the individuals in that subsample who use social networks.
Step 3. Next, we estimated the propensity score stratum-specific treatment effects in terms of the instrumental returns and emotional returns to social networks. We can eliminate the heterogeneity bias in the use of social networks because of our within-group homogeneity. We can directly use multiple linear regression models to estimate the earnings and emotional returns to social networks in each subsample, which means that we can obtain the treatment effects for each stratum.
Step 4. Finally, we observed the relationship between the social network effects and the ranking of each subsample in terms of propensity scores. If the social network effects are greater in the higher ranked subsamples, then there is a positive selection effect related to the use of social networks in the labor market: the more likely someone is to use social networks, the higher the return to their social network is in terms of earnings or job satisfaction. In contrast, if the social network effects are lower in the higher ranked subsamples, then there is a negative selection effect, and the more likely people are to use social networks, the less rewarding the use of those social networks will be. Specifically, to conduct this analysis, we fit the data using a variance weighted least squares regression with the ranking of the subsamples as the independent variable and the social network effects for each subsample as the dependent variable. Since the sample for this step is simply the number of subsamples, the sample size was too small for direct estimation via ordinary least squares . VWLS uses standard deviations as weights, which relaxes the requirement for homoskedasticity and overcomes the small-sample problem of OLS.
---
Results
---
Descriptive Statistics
Table 1 presents descriptive statistics for the sample and compares both of the groups we focused on: those who use social networks when searching for a job and those who do not use social networks. The continuous variables are described in terms of their means, the categorical variables are described in terms of percentages for all categories other than the reference category, and the last column shows the results of a significance test. Clearly, there were a number of significant differences between job seekers who use social networks and those who do not. Most notably, job seekers who use social networks had higher entry-level earnings than those who do not, with the former earning an average of 340 per month ) and the latter earning only 226 ). The same trend was observed for job satisfaction: job seekers who use social networks had a mean job satisfaction value of 3.6, while job seekers who did not use social networks when seeking employment had lower job satisfaction, a value of 3.47 . However, it is not yet clear as to whether the differences in earnings and job satisfaction observed at this point are the result of job seekers' use of social networks because the binary analysis indicates that the use of social networks when seeking a job is not a random event but is influenced by numerous demographic and social factors. For example, the average age of job seekers who use social networks was found to be 29 years old, which is significantly higher than that of job seekers who do not use social networks ; those who seek jobs in the private labor market sector were found to also more likely to use social networks than those in the public labor market sector. In addition, factors such as the year of job entry, household registration at the time of the job search, education, and attitudes toward relationalism, as well as occupational requirements, all affect whether social networks are used in the job search process.
---
Selectivity in Using Social Networks: The Logit Model
As the descriptive analysis in Table 1 shows, the distributions of the two types of job seekers, those who use social networks and those who do not, were not consistent across many variables. To obtain an unbiased estimate of the social network effect, we first looked at the selection in social network use: who is more likely to use their social networks in their job search? We constructed a logit model with the use of social networks in job search as the dependent variable in order to predict the likelihood of using a social network. The variables included in the model to control for factors influencing the use of social networks in job search were gender, age, household registration, education, Party membership, parental education, parental Party membership, work-unit sector, an indicator for high-income industries, occupational requirements, and the year of job entry. The results are shown in Table 2. From the results shown in Table 2, it is easy to see that the use of social networks when seeking a job is indeed influenced by many factors, both personal characteristics such as age, education, household status, and attitudes toward relationalism, as well as structural factors such as labor market sector, occupational requirements, and the year of job entry. Again, this finding verifies that the use of social networks in job searching is not random but that those who use social networks share certain common behavioral characteristics [37]. People do not randomly become social network users or nonusers; rather, people with certain common characteristics become social network users. The logit model can provide a predicted probability of using social networks, and this probability value is important, as it is the basis for the stratification conducted in the next step.
---
Propensity Score Strata
On the basis of the predicted results of the logit model presented in Table 2, we divided the ranked predicted probability values to generate six subsamples that were heterogeneous between groups but homogeneous within groups. These subsamples are referred to as strata one through six in Table 3. We divided the sample in descending order so that a higher stratum number indicates a higher likelihood of using social networks in the job search. Specifically, the probability that job seekers in the first stratum used their social networks in the job search process was the smallest-less than 0.1-and the number and proportion of job seekers who actually used social networks in this subsample was also the smallest, with only 84 out of 393 people using social networks. The probability of using a social network was slightly higher in the second stratum than in the first and ranged from 0.1 to 0.19. The number and proportion of job seekers who actually used social networks were also higher than those in the first stratum, with 124 out of 354 respondents using social networks. Finally, respondents in the sixth stratum had the highest probability of using a social network, with each person having a probability of more than 85%. The number and percentage of those who actually used their social networks were also the highest at 437 out of 589. Each column in Table 3 represents a stratum, and each row represents an independent variable that helps predict the likelihood of using a social network. Ideally, within each stratum of the subsample, the effects of each independent variable on the dependent variable would be statistically insignificant, which means that the use of social networks is not confounded by any other factor in the subsample and can be considered a purely exogenous variable. However, in practice, it is very difficult to obtain such a perfect match, and there are many cases where the individual variables are still significant. For example, the variable "paternal Party membership" in the fifth stratum has a p-value of 0.044, indicating that in this stratum, paternal Party membership still influences whether respondents use their social networks. As suggested by Xie and his colleagues [33,34], we need to further control for these still significant variables in the next step of the model when we predict the effect of social networks.
After obtaining the six propensity score strata, we then calculated the effects of the use of social networks in each stratum and subsequently fit the relationship between the number of strata and the effects of social networks using VWLS. The difference between VWLS and OLS is that VWLS relaxed the assumption of homoskedasticity. In fact, the variances did vary across the different subsamples. In addition, we also found the average treatment effect for the overall sample, following the conventional method for propensity score matching. The final results of the analysis are presented in Table 4 and Table 5 . Turning first to Table 4, the table is divided into three panels. The top panel reports the coefficient, standard error, and Z value of the social network effect for each of the six subsamples. As the subsample ranking increased , the effect of social networks on earnings became progressively smaller and eventually become negative . It should be mentioned that because the subsamples in each stratum were created according to the empirical stratification results, the subsamples themselves were not representative, and the results for those subsamples were not generalizable to higher-level groups. In fact, this step is only an intermediate step in the entire analysis process, and the results are only meaningful after weighted summation as in propensity score stratified matching or regression fitting using variance weighting as in the SM model.
The middle and bottom panels of Table 4 present the results of a composite analysis based on the effects of each stratum. The middle panel presents the calculated average earnings effect of social network use for the overall sample according to the propensity score subclassification method.
b = ∑ K k=1 [ * b k ] Var = ∑ K k=1 [ˆ2) * Var] SE = Var Z = b/SE
where b is the average effect of social network use, k = 1, 2, . . . , 6 indexes the number of strata, b k is the coefficient representing the effect of social network use on earnings in each stratum, n k is the number of observations in each stratum, N is the number of observations in the whole sample, and SE is the standard error of b. When the average treatment effect b is divided by its standard error, SE, the Z value used for statistical inference is obtained. In the overall sample, job seekers who use social networks had significantly higher earnings than those who do not , which is consistent with the findings of previous studies.
The bottom panel of Table 4 shows the relationship between the social network earnings effect and the propensity to use social networks estimated via VWLS. The results indicate that there was a significantly negative relationship between the strata ranking and the social network effect.
Table 5 is structured in the same way as Table 4. Looking at the top of the table reveals that within each subsample, the coefficient representing the effect of social network use on job satisfaction showed steadily increases. The results in the middle panel were calculated in the same way as the earnings effect and indicate that in the overall sample, there was no significant difference in job satisfaction between job seekers who use social networks and those who do not. The bottom panel of the table shows the relationship between the effect of social networks on job satisfaction and the propensity to use social networks obtained via VWLS. The results show that there was a significant positive effect between the strata ranking and the social network effect, and this effect was statistically significant. That is, the higher the propensity to use social networks when seeking employment, the higher the job satisfaction of job seekers. This result is the opposite of that for the earnings effect.
---
Conclusions and Discussion
The job search process for laborers is actually the process of allocating labor resources through the labor market. From the planned economy in China's pre-reform period to its dual-track system at the beginning of the reform and then into the era of the market economy as the reform deepened, resources have been allocated through three channels: redistribution, the market, and social networks [4,17,19,38]. Of these, redistribution channel and the market channel are formal channels for resource allocation, while the social network channel is an informal resource allocation channel. Theoretically, as marketization increases over time, market rules become increasingly clear, institutional rigidity becomes stronger, the power of formal channels increases, and the use of informal channels such as social networks should decrease. However, the most recent empirical data show that the proportion of workers using social networks during the job search process has increased to nearly 80% of the samples [39]. These realistic data suggest that the use of social networks in job searching is no longer restricted to a special group, and the internal heterogeneity within the group of social network users is gradually increasing [40]. This leads us to ask: for this internally heterogeneous group, is there also variability in the effect of social networks? That is, are the returns to social networks consistent across those job seekers who use them? Focusing on the use of and returns to social networks during job searching, this paper attempts to address the two-part endogeneity issue: First, the endogeneity in pretreatment effects, which means whether the use of social networks is a random event; second, the endogeneity in treatment effects, in this study meaning whether the return to social networks is consistent among those actors who use them when searching for a job. We employed a large set of survey data on social networks and used a heterogenous treatment effects model in our empirical analysis. We obtained the following main conclusions:
First, the use of social networks in job searching is not random, and the propensity of job seekers with different characteristics to use social networks is not the same. In the process of job searching, which is an instrumental action, the choice of job seekers to use or not to use social networks is not a random event. Rather, there are certain characteristics that lead some people to use their social networks, while others choose not to. These characteristics are both individual and structural.
Most previous research on social network effects has been based on analyses of regression models. Research based on OLS assumes that all the behaviors of the study occur randomly; however, the creation and the use of social networks-interactions among actors-is purposeful and highly subjective. It is difficult to satisfy the requirement of random distribution [22,23]. In our empirical analysis, we found that there do indeed exist some shared characteristics that result in essential differences between the two types of people: those who use social networks and those who do not.
Second, the returns to social networks are different for job seekers with different propensities to use social networks. This means that in the labor market job search process, there are two different selection biases working at the same time: there is not only pretreatment heterogeneity, which means that the use of social networks suffers from selection bias, but also treatment effect heterogeneity, which means that the effects of social networks are also biased due to endogeneity. Even within the group of those who use their social networks when searching for a job, the returns to that use differ from person to person.
Third, there is negative selection in the instrumental effect of social network use in the job search process but positive selection in the expressive effect: the higher the propensity to use one's networks, the lower the income return to social networks and the higher the job satisfaction; the lower the propensity to use one's networks, the higher the income return and the lower the job satisfaction.
Endogeneity has been a major challenge in social network analysis in recent years, especially in research on social network effects in labor markets [22,23,41]. Scholars have tried to respond to this endogeneity challenge, which takes many forms, such as the homogeneity of social networks [24], the selectivity of social networks [12,13,37], and reverse causality [42]. Previous studies on social network effects have not found heterogeneous returns to the use of these networks, which may have been due to the limitations of their research methodology. This study identified endogeneity in both pretreatment effects and treatment effects and estimated the effect of social network use while considering the complete endogeneity problem. People differ not only in their use of social networks but also in how they are affected by that use. This study provides ideas for analyzing the heterogenous returns to social networks. We first estimated the probability of using social networks and then stratified the sample on this basis to calculate the effects of social networks among those with different propensities to use their social networks. Finally, we tested the relationship between the ranking of the strata and the effect of social network use. Nan Lin has reviewed 31 studies on the relationship between social capital and status attainment from the last two decades of the last century, and 30 of those studies demonstrated a link between social capital and status attainment [17]. Are these links the same for different actors? Nan Lin's strength-of-position proposition states that the better the initial position of an actor is, the more likely they are to access and mobilize higher levels of social capital. This study builds on this proposition and further explores the differences in returns related to different propensities. The results show that in a pyramidal social structure, there are not only differences in the distribution of resources but also differences in the returns to resources.
---
Limitation
This paper also has some limitations. In this study, we first explored the heterogeneity in the returns to social networks; however, more advanced research may be conducted in the future. In our exploration of the use of social networks, we used various external individual-level characteristics to comprehensively analyze the propensity to use social networks. We believe that a third dimension could be introduced in future research that enables an in-depth exploration of the relationship between group characteristics and the propensity to use social networks and the differences in the corresponding returns. For example, information about class [43], a concern of sociological research, could be included. Are there differences in the propensity of members of different classes to use their social networks, and are there differential returns? This would further enrich the social implications of social network research.
Second, this study was concerned with the returns to social networks at the beginning of the job seeker's entry into their career. In addition to the role of social networks in the job search process, which is to help job seekers find jobs and earn high salaries, social networks may also have long-term effects and delayed returns. In the real labor market, job seekers' use of social networks to obtain jobs may not necessarily generate benefits at the beginning of their entry into the workforce. Whether those who find jobs through social networks are sheltered by that network after they enter their job, thus obtaining long-term professional development, rapid promotions, and job security, is an interesting question [44]. The collection of panel data will help us answer this question effectively.
In addition, this study was concerned with the use of and returns to social networks in the labor market, and the labor market segmentation model is a classical theoretical model in labor market research. The object of analysis in this paper was the use of social networks in the labor market as a whole. A prominent feature of China's social development in the past 40 years since its reform and opening up is inequality. What kind of differences might exist in the use of and returns to social networks in different periods and regions? Whether institutional differences, industrial differences, and differences in occupational levels have caused labor market segmentation for different reasons and whether this segmentation has led to differences in the use of and returns to social networks are specific topics to be discussed in depth in future research.
---
Data Availability Statement:
The data presented in this study are available on reasonable request from the corresponding author.
---
Author Contributions: Conceptualization, X.G.; data curation, Q.C.; formal analysis, Q.C.; funding acquisition, X.G.; investigation, X.G.; methodology, X.G. and Q.C.; project administration, X.G.; writing-original draft, X.G.; writing-review & editing, X.G. All authors have read and agreed to the published version of the manuscript. | With the advancement of social network research over time, a research consensus has been reached that the use of social networks in job searching can provide positive returns. This study focused on the heterogeneity in the returns to social networks. Using Job Search and Social Networks (JSNET) survey data on urban residents of China, we examined the differences in the propensity to use social networks in job searching and the returns to social networks in terms of job seekers' earnings and job satisfaction using propensity score stratification and the heterogeneous treatment effects model (HTE model). The use of social networks in job searching was found to be nonrandomly distributed, and the propensity to use such networks varied according to job seeker characteristics. For job seekers with different propensities, the returns to social networks were also different. Moreover, there was negative selection in the instrumental effect of social network use in the job search process but positive selection in the expressive effect: the higher the propensity to use one's networks, the lower the income return to social networks and the higher the job satisfaction; the lower the propensity to use one's networks, the higher the income return and the lower the job satisfaction. |
IntrOduCtIOn
Due to the rapidly ageing population and increasing longevity, in addition to increasing improvement in medical care, a growing number of people are living for long periods of time with chronic disease. The WHO reported that chronic diseases kill 40 million people per year, which is equivalent to 70% of all deaths globally. Each year, 15 million people die from a chronic disease, and >80% of these 'premature' deaths occur in low-income and middle-income countries. 1 In recent years, the number of individuals with chronic diseases in China has continued to rise, with many residents facing serious health challenges. There were 260 million people with chronic diseases in China in 2012, and this number has increased by an average of 8.9% each year. Deaths from chronic diseases account for 86.6% of total mortality. As a result of the rapidly ageing population, the number of elderly people in China in 2050 is predicted strengths and limitations of this study ► The China Health and Retirement Longitudinal Study is the first national survey focusing on middle-aged and elderly individuals, and provides detailed information about respondents and their living spouses. ► This was the first analysis of foregone care among Chinese middle-aged and elderly people with chronic diseases. ► This paper evaluated foregone care only among middle-aged and elderly adults who perceived a need for healthcare but did not seek treatment; however, it did not take into account those not perceiving the need for healthcare; therefore, the prevalence of foregone care may have been underestimated to some extent. ► Cross-sectional data cannot be used to draw causal conclusions.
Open Access to total >400 million, accounting for one-third of the total population, and the average number of chronic diseases that individuals >60 years have is estimated at 2.2. China has entered a period of high economic burden caused by chronic diseases, which account for approximately 70% of the total economic disease burden. 2 According to a report by WHO in 2015, the direct medical costs for chronic diseases in China were >US$500 billion, and by 2030, rapid ageing may increase the financial burden of chronic diseases by 40%. Access to equitable healthcare, according to need, and regardless of demographic, ability to pay or social background, is an important goal for worldwide healthcare service systems. 3 China has implemented a series of healthcare reforms during the past two decades, including the expansion of health insurance coverage and the reformation of the health service system, with the ultimate goal of affordable and equitable healthcare for all. The reforms of the healthcare systems have minimal benefits at the beginning and progress with expanded benefits. 4 However, healthcare in China remains highly unequal across different subpopulations, and both the demand for, and inequality in, healthcare utilisation have increased significantly in this country, due to its rapidly ageing population. Several studies have addressed this issue by examining factors associated with healthcare utilisation among the elderly in China. [5][6][7][8] Foregone care is another aspect of healthcare access; an individual with forgone care needs is defined as one who does not use healthcare, despite perceiving a need for it. Forgone care is an important aspect in the assessment of healthcare system performance, as it represents the gap between the perceived need and actual utilisation of healthcare services from the citizens' perspective.
Ageing is accompanied by an increase in the prevalence of non-communicable chronic diseases. 9 In China, the elderly people are more sensitive to the economic burden of diseases overall because of the high prevalence of chronic diseases and low incomes of this population, and the absence of social security mechanisms. 10 Thus, elderly individuals with chronic diseases may have a higher prevalence of foregone care compared with those who do not have such conditions. Several studies have empirically tested the associations between individual and community factors and foregone care. These studies indicate an association between higher rates of foregone care and female sex, younger age, rural living, lack of health insurance, lack of financial support, low levels of education and poor health. [11][12][13][14][15][16] Although foregone care is an area of high interest for researchers and policy-makers, there is a dearth of research on this subject among middle-aged and elderly individuals with chronic disease in China. The overall goal of this study was to identify the prevalence of foregone care and its associated factors among middle-aged and elderly people with chronic conditions. Two specific objectives were defined: comparison of the prevalence of foregone care between middle-aged and elderly individuals with and without chronic diseases and using Anderson's behavioural model, to identify factors associated with foregone care among middle-aged and elderly individuals with chronic diseases.
---
MethOds And dAtA data source/sample
This study was based on data from a nationally representative household survey of middle-aged and elderly people, the China Health and Retirement Longitudinal Study . CHARLS is a biennial survey conducted by the China Centre for Economic Research at Peking University. The sample was drawn in four stages. First, county-level units were sampled directly. These counties cover 28 of 30 provinces in mainland China, other than Tibet. Second, village and community units within county units were chosen with the help of the National Bureau of Statistics, using recently updated village-level population data. The sample used administrative villages in rural areas and neighbourhoods in urban areas as primary sampling units . Three PSUs were selected within each county-level unit, using probability proportional to size sampling, to select a total of 450 PSUs. Third, household units were selected in each PSU. The sampling frame was constructed using Google Earth base maps, and a computer-assisted personal interview programme was then used to sample households and to conduct the interviews using laptops. Finally, all age-eligible sample households who were willing to participate in the survey were interviewed.
CHARLS baseline data include detailed information about respondents and their living spouses. The main questionnaire includes information on basic demographics, family, health status and functioning, healthcare and insurance, work, retirement and pension, income, expenditure and assets. In the 2013 wave, the survey sample consisted of 18 605 individuals ≥45 years old and their spouses. In this paper, we excluded spouses ≤45 years old, and the remaining 18 377 observations were included in the analysis. statistical analysis Binary multivariate logistic regression was applied to estimate ORs of foregone care. The dependent variable 'foregone care' was dichotomised into 0 = 'reported utilisation of healthcare' and 1 = 'reported non-use of healthcare'. All explanatory variables were recorded as binary variables.
---
Variable selection
The dependent variables were three kinds of foregone care: non-use of outpatient services, non-use of inpatient services and non-attendance for physical examination. Participants were asked the following two questions: 'Have you been ill in the last month?' and 'What was the main reason for not seeking outpatient treatment?", and Open Access the responses were used to estimate 'non-use of outpatient services' . Participants were asked, 'In the past year, did a doctor suggest that you needed inpatient care but you did not get hospitalised?' and the response was used to estimate 'non-use of inpatient services' . Participants were asked, 'Did you undergo a physical examination in the last 2 years?' and the response used to estimate 'non-use of physical examination' .
Independent variables were selected based on Andersen's healthcare utilisation model . This model has guided systematic investigations into the factors that can lead to the use of healthcare services, including predisposing, enabling and need factors. 17 Predisposing factors include demographic characteristics and both social structure and health belief-related factors. In our analyses, we included sex, age and marital status as demographic characteristics and education, employment status and family size as social structure factors. Enabling factors are those that make health service resources available to an individual; they can be measured by determination of family attributes, such as income, access to health insurance or other sources of third-party payment. Enabling factors in the community in which the family lives can also affect the use of services. Community attributes include the number of healthcare facilities and personnel in a community and the region of the country, in addition to the rural-urban nature of the community. In our analyses, access to public insurance, income levels and financial transfers received and given were considered as family components, while area of residence and medical facilities were considered indicators of community attributes. Need factors were both perceived need and need diagnosed or evaluated by a healthcare professional; the former was based on self-reported health status, and the latter included chronic diseases, depression, body mass index and limitations on activities of daily living . Variables included in the multivariate model are presented in table 1.
---
results sample characteristics
The total sample consisted of 18 377 respondents ≥45 years old. Missing data were automatically removed during data processing. No chronic disease was reported by 28.88% of respondents, with 71.12% reporting ≥1 chronic disease. In CHARLS, the chronic conditions included hypertension, dyslipidaemia, diabetes, cancer, chronic lung, stomach, memory-related, and/or liver diseases, heart problems, stroke, kidney, emotional and nervous problems, arthritis or rheumatism and/or asthma. Patients with chronic diseases were divided into two categories, including individuals with one chronic and those with multiple chronic diseases in one person.
Descriptive statistics of the characteristics of the individuals with, versus without, chronic conditions are presented in table 2. Chronic disease prevalence was slightly higher among middle-aged and elderly women than men. Middle-aged and elderly individuals aged 45-55 years accounted for 43.28% of those without chronic disease, which was higher than the proportion of those with one or multiple chronic diseases; however, with increasing age, the proportion of individuals with one or multiple chronic diseases exceeded that of people without these conditions, indicating that the risk of chronic disease increases with age. The proportion of people without chronic diseases, and with one or multiple chronic diseases gradually increased among middle-aged and elderly people who were divorced/ single/separated or unemployed/retired, indicating that the risk of chronic disease was relatively high among this group. Middle-aged and elderly people with lower educational levels and small family size also had a higher risk of chronic diseases.
Regarding enabling factors, the ratios of population subgroups with social health insurance did not differ significantly. The ¥1496-8126 income group had the highest probability of having more than two types of chronic disease, and the proportion of people without chronic disease was greater in the high-income, compared with the low-income, group. Compared with middleaged and elderly people who did not receive support from others, people with economic support exhibited an increasing trend towards lack of chronic disease, 73.75% towards one chronic disease and 78.14% towards multiple chronic diseases; people who provided economic support to others exhibited the opposite trend.
Among need factors, middle-aged and elderly individuals whose self-reported health was fair or poor had an increasing trend towards having no chronic disease , one or multiple chronic diseases, and those whose self-reported health was good or excellent exhibited the opposite trend. These results indicate that the risk of chronic disease was negatively associated with health status. Depression and probable depression were also associated with a significant increasing trend towards having one or more chronic diseases. There was also an increasing trend in probability from people without chronic disease, to one chronic, and then to multiple chronic diseases among middle-aged and elderly people with BMI >24 or with any ADL.
---
Prevalence of foregone care
The data presented in table 3 demonstrate the prevalence of foregone care among middle-aged and elderly individuals. Overall, the prevalence of non-use but oughtto-have-used outpatient services within the last 4 weeks was 10.21%, and the prevalence of non-use of inpatient services was 6.84%. Comparisons of foregone care among Open Access different groups demonstrated that the underutilisation rate was higher among people with multiple chronic diseases. Furthermore, the results demonstrate that individuals without chronic disease had the highest probability of not having undergone a physical examination in the last 2 years.
---
Multivariate regression analysis
Binary logistic models were constructed to examine the determinants of foregone care among middle-aged and elderly Chinese people with chronic diseases. Estimated coefficients describing the predicted effects of the variables are presented in table 4. We also calculated overall significance values for a multivariate logistic regression Open Access With the exception of sex, all other predisposing factors were predicted to have significant impacts on foregone healthcare. Specifically, middle-aged and elderly people >55 years old were less likely to report outpatient or physical examination foregone care compared with those <55 years old. Individuals who were married or had a larger family had a higher likelihood of foregone physical examination, while those with high levels of education had a significantly lower likelihood of foregone physical examination. Middle-aged and elderly people with chronic diseases who were retired without pensions exhibited a higher likelihood of outpatient and inpatient foregone care compared with retired people with pensions. Open Access
---
Open Access
Among enabling factors, we found a significant association between social health insurance and non-use of outpatient and inpatient services; hence, enrolment in social health insurance schemes may significantly reduce the likelihood of outpatient and inpatient foregone care among middle-aged and elderly people with chronic diseases. The highest income group demonstrated a significantly reduced likelihood of forgone inpatient and physical examination care compared with the lowest income group. Receipt of financial transfers increased the likelihood of non-use of outpatient services, and providing transfers was associated with reduced probability of foregone physical examination. Compared with middle-aged and elderly people living in urban areas, those in rural areas were significantly more likely to forego outpatient and physical examination care.
Among need factors, significant associations were detected between self-reported health, multiple chronic diseases, depression, and outpatient, inpatient, and physical examination foregone care. Middle-aged and elderly people with chronic diseases with fair, good or very good self-reported health had a lower likelihood of outpatient and inpatient foregone care compared with those with poor health status, and the likelihood decreased with increasing health status; however, individuals who self-reported their health as fair had a higher likelihood of foregone physical examination. Middle-aged and elderly people who had multiple chronic diseases demonstrated a higher probability of foregone outpatient and inpatient care, while exhibiting a lower likelihood of foregone physical examination. Individuals with depression had a higher likelihood of outpatient, inpatient and physical examination foregone care. Middleaged and elderly people with BMI values >24 had a lower probability of outpatient and inpatient foregone care. Any ADL increased the possibility of outpatient service foregone care. Drinking alcohol did not have a significant impact on any kind of foregone care, while smoking significantly increased the likelihood of foregone physical examination.
dIsCussIOn Along with the rapidly ageing population in China, the number of chronic diseases continues to rise and the elderly face serious health challenges. Although many studies have analysed healthcare utilisation in diverse population groups, few have examined foregone care in middle-aged and elderly chronically ill Chinese patients. This study, focusing on middle-aged and elderly people, was performed to analyse the prevalence of foregone healthcare and its associated factors in China.
In this study, we found that 10.21% and 6.84% of respondents had foregone outpatient and inpatient care, respectively; however, the prevalence of physical examination foregone care was relatively high . This may be attributable to the fact that Chinese social healthcare insurance schemes do not cover physical examinations; therefore, many middle-aged and elderly people are apt to forego physical examinations because of the associated economic burden. The prevalence of outpatient and inpatient foregone care rose, alongside increasing number of chronic diseases. Furthermore, the results of regression analysis confirmed that respondents with multiple chronic diseases were more likely to forego outpatient and inpatient care. These data indicate that a greater demand for health services is associated with a higher probability of foregone outpatient and inpatient care. However, the likelihood of physical examination foregone care exhibited the opposite trend; people without chronic disease were most likely to forego physical examination and the prevalence decreased with increased number of chronic diseases. This result suggests that middle-aged and elderly individuals with chronic disease may have stronger awareness of preventative healthcare.
Among predisposing factors, our results indicate that the possibility of foregone outpatient and physical examination care decreases with increasing age. This is consistent with previous results literature reports. 11 18 People in older age groups generally have more need for healthcare, related to more serious conditions, and may be more flexible with their time when seeking healthcare compared Open Access with younger individuals. Middle-aged and elderly people with chronic diseases who have fixed incomes accessed more healthcare and exhibited a relatively low likelihood of foregoing care. This finding indicates that some medical financial assistance programmes may need to be extended, particularly with respect to unemployed and elderly retired people without pensions, to facilitate further reduction of the possibility of foregone care. In addition, our analyses revealed that middle-aged and elderly people with chronic diseases who were married and had large families exhibited a relatively high probability of foregone physical examination, consistent with a number of existing literature reports. 19 We did not identify any association between sex and foregone care, which is also in line with several previous publications. 12 20 Of enabling factors, our results demonstrate a strong association between socioeconomic variation and foregone care. Specifically, we found a significant association between household income level and foregone care; people with chronic diseases in the highest income group reported less foregone inpatient and physical examination care compared with those in the lowest income group; however, we did not find a significant association between income level and outpatient foregone care, similar to previous studies. 11 16 18 These findings indicate that due to higher hospitalisation costs, income levels remain an important impact factor influencing the possibility of inpatient foregone care; however, for outpatient services, there were no significant differences in the likelihood of foregone care among different income groups, indicating that the universal coverage of medical insurance in China may be gradually improving the equity of outpatient care utilisation among middleaged and elderly people with chronic disease from different income levels. We also found that although social healthcare insurance can significantly reduce the level of outpatient and inpatient foregone care among chronically ill elderly patients, these schemes may not significantly impact physical examination foregone care among elderly patients with chronic disease. Hence, policy efforts may need to focus on chronic elderly patients having urban resident medical insurance or new cooperative medical insurance, and strategies may need to be adjusted to promote the effective utilisation of preventive healthcare. Middle-aged and elderly people with chronic diseases who have received financial transfers from others exhibited a slightly higher likelihood of outpatient foregone care, while those who had made transfers to others had a significantly lower level of physical examination foregone care. Residents in rural areas had a significantly higher likelihood of outpatient and physical examination foregone care, as previously described. 21 These findings may reflect the rural-urban income gap and differences in medical security in China.
As Andersen described, need factors best describe the true needs for healthcare. 17 Consistent with previous studies, we found that middle-aged and elderly people with multiple chronic diseases or depression were more likely to forego outpatient and inpatient care. 8 22 23 This may be because of high demands for medical services or lack of patient compliance among this group, likely leading to a relatively high probability of foregone care. Regarding physical examinations, multiple chronic diseases were associated with a reduced possibility of foregone care; however, depression was associated with an increased likelihood, consistent with published data. 8 24-27 Self-reported health was negatively associated with outpatient and inpatient foregone care, except for individuals whose self-reported health was excellent. This indicates that the superior health of respondents was associated with a lower likelihood of foregone care. 28 ADL limitations did not affect inpatient and physical examination foregone care; however, these factors significantly increased the likelihood of outpatient foregone care. This finding maybe related to the fact that people with ADL often have some difficulty seeking outpatient treatment in a timely manner. 29 The findings of this study may provide evidence on healthcare underutilisation and associated factors among middle-aged and elderly people. Some limitations of our study must be acknowledged. We evaluated foregone care only among middle-aged and elderly adults who perceived a need for healthcare but did not seek treatment; however, we did not take into account those who did not perceive the need for healthcare. Some patients objectively require healthcare, but do not perceive that need; therefore, the prevalence of foregone care may have been underestimated to some extent. In addition, cross-sectional data cannot be used to draw any causal conclusions about why these individuals may have foregone care.
---
COnClusIOn
Our findings reveal that the prevalence of foregone physical examination is relatively high among middle-aged and elderly people, and that this prevalence decreases with an increased number of chronic diseases. As a result of their greater demand for health services, respondents with chronic diseases demonstrated relatively higher likelihoods of both outpatient and inpatient foregone care. Hence, the adjustment of health insurance policies and improvements in the hierarchical medical system in China may be helpful to the effective utilisation of preventive care and outpatient and inpatient services. Some predisposing factors, including age, marital status, employment, education and family size; enabling factors, including social health insurance, income level and area of residence; and need factors, such as self-reported health, multiple chronic diseases and ADL, significantly affected foregone care among middle-aged and elderly people. Therefore, we may need to address these socioeconomic and demographic factors for reducing foregone care and enhancing the effective use of health services among Open Access middle-aged and elderly people in China. Finally, further work, including longitudinal studies, is required to gain a better understanding of the causal relationships between foregone care and its associated factors.
---
The study was approved by the ethics committee of Nanjing Medical University; reference number: 2017613.
Provenance and peer review Not commissioned; externally peer reviewed.
data sharing statement Extra data can be accessed via the Dryad data repository at http:// datadryad. org/ with the | Objective In general, published studies analyse healthcare utilisation, rather than foregone care, among different population groups. The assessment of forgone care as an aspect of healthcare system performance is important because it indicates the gap between perceived need and actual utilisation of healthcare services. This study focused on a specific vulnerable group, middle-aged and elderly people with chronic diseases, and evaluated the prevalence of foregone care and associated factors among this population in China. Methods Data were obtained from a nationally representative household survey of middle-aged and elderly individuals (≥45 years), the China Health and Retirement Longitudinal Study, which was conducted by the National School of Development of Peking University in 2013. Descriptive statistics were used to analyse sample characteristics and the prevalence of foregone care. Andersen's healthcare utilisation and binary logistic models were used to evaluate the determinants of foregone care among middle-aged and elderly individuals with chronic diseases. results The prevalence of foregone outpatient and inpatient care among middle-aged and elderly people was 10.21% and 6.84%, respectively, whereas the prevalence of foregone care for physical examinations was relatively high (57.88%). Predisposing factors, including age, marital status, employment, education and family size, significantly affected foregone care in this population. Regarding enabling factors, individuals in the highest income group reported less foregone inpatient care or physical examinations compared with those in the lowest income group. Social healthcare insurance could significantly reduce foregone care in outpatient and inpatient situations; however, these schemes (except for urban employee medical insurance) did not appear to have a significant impact on foregone care involving physical examinations. Conclusion In China, policy-makers may need to further adjust healthcare policies, such as health insurance schemes, and improve the hierarchical medical system, to promote reduction in foregone care and effective utilisation of health services. |
Introduction
Background. Explanations for monogamous marriage have centered around the prevalence of this practice in Eurasia , linking its emergence to the development of idiosyncratic features of societies in the region. These include, for example, the establishment of large nation states and democracy ; the spread of Christianity ); the onset of industrialization and urbanization , and of economic development more generally . Consistently, cross-cultural analyses show that societies scoring high on scales measuring "societal complexity" tend toward monogamous marriage .
By focusing on the cross-cultural distribution of marriage strategies, these studies fail to account for their history. Restrictions on polygynous marriage appear in the earliest historical records, long predating the development of aspects of social organization typically associated with Eurasian societies and with "societal complexity" and "modernization" more generally . In this paper I reconstruct the pattern of change in marriage strategies in the history of societies speaking Indo-European languages, using cross-cultural data in the systematic and explicitly historical framework afforded by the phylogenetic comparative approach . In the companion paper I extend this framework to reconstructing the history of residence strategies in the same set of societies . Finally, this framework is extended in Fortunato to investigate the co-evolution of monogamous marriage and neolocal residence; on this analysis, it cannot be excluded that the observed association between marriage and residence strategies is the artifact of a history of descent of these societies from a common ancestor.
Phylogenetic tree-building analyses of linguistic data support Renfrew's hypothesis for the origin of the IE language family, which envisages an expansion from Anatolia with the spread of agriculture starting between 7000 and 6000 BCE . This scenario involves an early split of IE languages from their sister group, the extinct Anatolian family, which comprises the extinct languages Hittite, Palaic, Lydian, Luwian, and Lycian; together, IE and Anatolian form the Indo-Hittite language family . The other main competing hypothesis, proposed by Gimbutas , places the homeland of the language family in the Pontic steppes north of the Black Sea and posits an expansion driven by nomadic horse-riding pastoralists starting after 4000 BCE . Consistent with the former scenario, I use the terms "Proto-Indo-European" for the hypothetical ancestor of IE languages and "Proto-Indo-Hittite" for the hypothetical ancestor of Indo-Hittite languages, as well as for the hypothetical "proto-societies" that spoke them. Other classifications, based on alternative explanations for the origin of the IE language family, use PIE for the ancestor of IE and Anatolian languages . To avoid confusion, throughout this and the companion paper I have changed instances of the latter usage to the one stated here.
---
Related Research.
Early IE marriage strategies have been reconstructed on the basis of linguistic and ethnographic evidence. The linguistic evidence is, at best, tenuous. The correspondence of the Middle Irish and Avestan terms for "concubine, wanton woman" suggests that PIH society recognized some form of polygynous mating, if not polygynous marriage; this interpretation is supported by the use of specialized terms to designate legitimate children . A second line of evidence relates to the concept of widowhood: while it is possible to reconstruct a PIH term for "widow," there is no corresponding term for "widower." One interpretation of this pattern is that male widowhood was not recognized in PIH society because men married polygynously, such that the death of one wife did not affect their marital status . Implicit in this interpretation are, however, a number of unrealistic assumptions, for example that all PIH men married multiple wives and that they rarely, if ever, outlived their wives; both represent demographic impossibilities.
Based on the ethnographic evidence, Murdock reconstructed "an Eskimo type of social structure in the prehistory of the Indo-European peoples." The defining features of this type of social organization are the presence of the Eskimo system of cousin terminology and the absence of exogamous unilineal kin groups; additional typical characteristics include the presence of monogamy, independent nuclear families, and neolocality, but variant sub-types characterized by non-neolocal marital residence may feature polygyny and extended families . In the Eskimo type of cousin terminology all cousins are equated with each other but differentiated from siblings; unilineality refers to kin groups organized around principles of matri-or patrilineal descent; neolocality indicates residence of married couples apart from the kin of either spouse. The reconstruction was inferred from the social organization of five societies speaking IE languages, representing four of the major sub-groups of the language family. Specifically, the social systems of Yankees and Ruthenians are of the Eskimo type, the social systems of Albanians and Ossets include features suggesting "Eskimo antecedents," while the social system of the Kurds is not incompatible with the Eskimo type. Quite apart from the small number of societies upon which the reconstruction was based, Murdock's approach presents a serious methodological issue. According to Fox , modern IE terminologies for kin and affines tend to show a much stronger bias toward the nuclear family than was the case for early IE terminological systems. This brings into question the validity of inferences about past social organization drawn from kinship and affinal terminologies .
Finally, Fortunato and Mace used the ethnographic evidence in a phylogenetic comparative framework to test the hypothesis of co-evolution of bridewealth with polygyny and of dowry with monogamy in a sample of 51 societies speaking IE languages. This analysis reconstructed monogamy with dowry as the most likely state at the root of a phylogenetic tree representing the historical relationships among the 51 societies; in addition, the tree included Hittite, thus the root of the tree corresponded to PIH. The reconstruction was obtained using the maximumlikelihood phylogenetic comparative method developed by Pagel and implemented in Discrete ; this method estimates the evolutionary scenario that is most likely to have produced the observed distribution of states of the two traits across taxa at the tips of the tree. The tree was generated through phylogenetic tree-building analysis of the corresponding 51 speech varieties in Dyen et al.'s IE basic vocabulary database, using a maximum-parsimony optimality criterion, as implemented in PAUP* 4.0b4a ; this approach finds the tree or set of trees that optimizes the degree of fit to the data. However, the use of a "best" tree for phylogenetic comparative analysis is problematic, because results are affected by the topology and other parameters of the tree used . Thus, while controlling for the effect of descent on the distribution of cultural practices, Fortunato and Mace's reconstruction may be contingent upon the phylogenetic tree model used to represent how the societies are related.
---
Rationale and Objective.
For the most part, previous attempts to infer early IE marriage strategies from linguistic and ethnographic data have failed to use a systematic and explicitly historical approach. Where such an approach has been taken, the available methods did not account for uncertainty in the phylogenetic tree model used to represent population history.
Here I use a phylogenetic comparative method, in a Bayesian reversible jump Markov chain Monte Carlo framework, to reconstruct the pattern of change in marriage strategies in the history of societies speaking IE languages. Using a phylogenetic tree to represent how the taxa are related, phylogenetic comparative methods infer likely evolutionary scenarios that produced the observed distribution of the attribute of interest across the taxa. The Bayesian MCMC framework uses a single tree or a sample of trees to represent the relationships among the taxa; use of a tree sample removes dependence of the inferences upon any single phylogenetic hypothesis. This is particularly important for application of the phylogenetic approach to crosscultural data, because the reticulate nature of the interactions linking human societies cannot be captured by any single tree model . In a similar way, the RJ-MCMC implementation of the approach removes dependence of the inferences upon any single model of trait evolution; the model of trait evolution specifies, for example, whether a trait is likely to be acquired and lost at the same or at different rates. This is also crucial in the analysis of cultural traits, because the mechanisms of change are usually unknown for these traits. More generally, the Bayesian MCMC framework estimates the degree of statistical uncertainty in the parameters of interest to the comparative question ; this provides an indication of the degree of confidence that can be placed in any particular inference about the evolution of the trait. A non-technical introduction to the approach is presented in Fortunato .
Phylogenetic comparative methods assume that traits are transmitted vertically ; consequently, high rates of horizontal transmission of traits across societies may invalidate their application to cross-cultural data . Recent simulation analyses show, however, that phylogenetic comparative methods outperform non-phylogenetic methods under a wide range of simulated scenarios and levels of horizontal transmission . In any case, kinship and marriage systems appear to be "conservative" features of social organization ; in the context of linguistic and genetic variation, the effect of descent is strongest at the supra-regional level, while the effect of contact prevails within regions . This makes traits relating to kinship and marriage systems especially suited for analysis within the phylogenetic comparative framework.
---
Data and Methods
I used data on marriage strategy from the Ethnographic Atlas for a sample of societies speaking IE languages. The cross-cultural data were mapped onto a sample of trees representing how the societies are related, obtained by Pagel et al. through tree-building analysis of Dyen et al.'s IE basic vocabulary database. Finally, on the cross-cultural data and tree sample I used the phylogenetic comparative method developed by Pagel and colleagues to reconstruct ancestral states of marriage strategy.
Cross-Cultural Data. The EA is distributed in electronic form through the World Cultures journal. Variable identifiers in this section follow Gray's EA codebook. The data in binary form are in the supporting information file; a map of their geographical distribution is in Figure 1. The cross-cultural sample was collated by matching societies in the EA with speech varieties in Dyen et al.'s IE basic vocabulary database, as described in the SI file.
I coded societies included in the cross-cultural sample as monogamous or polygynous based on EA variable 9, which scores societies on the prevailing form of family organization . I collapsed the five categories for polygynous marriage, thus ignoring the distinction between limited and general polygyny, between general polygyny with sororal and non-sororal co-wives, and between general polygyny with co-wives occupying the same or distinct dwellings. This produced a sample with 18 of 27 societies coded as monogamous and 9 coded as polygynous .
Tree Sample. I used Pagel et al.'s posterior probability sample of 750 phylogenetic trees to represent how societies in the cross-cultural sample are related by way of descent from a common ancestor. Trees are present in the sample in proportion to their posterior probability, which is the probability of the tree conditional on the data and model of word evolution used in the tree-building analysis, and can be interpreted as the probability that the tree is correct . For example, the Ibero-Romance speech varieties included in the tree-building analysis by Pagel et al. share an ancestor in 84% of the trees in the sample; the probability that they are a "monophyletic" group is thus 0.84, given the data and model of word evolution used in the tree-building analysis. As discussed below, use of a tree sample instead of a single "best" tree amounts to incorporating phylogenetic uncertainty in the comparative analysis; the degree of phylogenetic uncertainty at several of the nodes in the Pagel et al. tree sample emphasizes the importance of doing so in this case . Pagel et al. inferred the posterior probability distribution of trees from Dyen et al.'s IE basic vocabulary database, using the Bayesian MCMC phylogenetic tree-building method developed by Pagel and Meade , as described in the SI file. I obtained the tree sample from Mark Pagel. I pruned the trees to retain only the speech varieties corresponding to the 27 societies in the cross-cultural sample, plus the outgroup Hittite, using Andrew Meade's program BayesTrees . "Outgroup" taxa provide information on the direction of change in the data by virtue of being distantly related to the groups under investigation, the "ingroup" taxa; they are used in tree-building for 1 in the SI file.
Colors express the marriage strategy , as per Figure 4; numbers correspond to taxa in Figure 4 and to entries in the relevant column in Table 1 in the SI file.
determining ancestor-descendant relationships . I preferred this "pruning" strategy, over the alternative of generating a tree sample for the speech varieties included in the comparative analysis , because the accuracy of phylogenetic tree-building grows with increased sampling of taxa . Hittite was retained for consistency with previous work but was assigned no marriage strategy data for the purpose of the comparative analysis.
Comparative Analysis. Reconstruction of ancestral states using the phylogenetic comparative method developed by Pagel and colleagues is performed using BayesMultistate, available as part of the BayesTraits package from http://www.evolution.rdg.ac.uk/BayesTraits.html. Unless otherwise specified, the information in this section is based on Pagel andMeade , Pagel et al. , and on the BayesTraits manual , to which I refer the reader for a more detailed description of the method and its application.
A non-technical discussion of the method is in the SI file.
Given the cross-cultural data and tree sample, BayesMultistate uses parameters q MP and q PM to describe the evolution of the trait "marriage strategy" on a tree. q MP and q PM measure the instantaneous rates of change, respectively, from monogamy to polygyny and from polygyny to monogamy ; they are used to define the probabilities of these changes, the probabilities of the two states at internal nodes on the tree, and the likelihood of the data, which is the probability of the data given the tree and the model of trait evolution specified by the rates ). In the likelihood calculations BayesMultistate treats taxa that are not assigned comparative data, like the outgroup Hittite in this case, as taking any state with equal probability.
In Bayesian RJ-MCMC mode, BayesMultistate uses RJ-MCMC methods to estimate the posterior probability distributions of rate parameters and of ancestral states at internal nodes on a tree, and of the possible models of trait evolution specified by the rate parameters . Four model categories are possible in this case: that q MP and q PM take distinct positive values, that they take the same positive value, or that either one is set to zero while the other takes a positive value. The posterior probability of a parameter value is a quantity proportional to its likelihood of having produced the observed data and represents the probability of the parameter value given the data and model of trait evolution. Schematically, the posterior probability distributions are estimated by running RJ-MCMC chains that sample states in the model of trait evolution in proportion to their posterior probability, across trees in the tree sample; a state in the model consists of model category and values of the rate parameters and ancestral state probabilities. Combining estimates over the sample produced by a chain amounts to "averaging" inferences over uncertainty in the phylogeny, in the parameters of the model of trait evolution, and in the model itself.
I performed five sets of analyses, each comprising five separate chains started from random seeds. One set estimated the posterior probability distributions of states M and P at internal nodes on the consensus tree summarizing the tree sample. The means of the posterior probability distributions of states M and P at a given node, denoted p and p, are multiplied by the posterior probability of the node itself, denoted p, to produce the combined probabilities of the two states at the node, denoted p and p; p ϩ p ϭ 1, thus p ϩ p ϭ p. This means that if reconstruction of the node itself is uncertain [i.e., if p Ͻ 1], the value of p sets an upper limit to the confidence that can be placed in the ancestral state reconstructions for the node. As a rule of thumb, confidence can be placed in reconstructions with combined probabilities Ն0.70.
The other four sets of analyses were used to assess explicitly the relative "fit" of states M and P at nodes PIH and PIE. For each node, one set of analyses was run with the node fixed to state M and one with the node fixed to state P. The posterior probability distributions of log e values sampled by the chains reflect how well a given fossil state fits the node; a measure called the "Bayes factor," which is used to compare posterior probability distributions , provides an indication of the strength of the evidence in favor of one state over the other at the node. The Bayes factor for state M over state P is denoted B MP . 2log e is approximated as twice the difference between log e [H] for a chain fixed on state M and log e [H] for a chain fixed on state P, where log e [H] is the natural logarithm of the harmonic mean of the likelihood values. In theory, values of 2log e Ͼ 0 represent evidence for state M and values of 2log e Ͻ 0 evidence for state P. Specifically, the evidence for a given state is "weak" for 0 Ͻ ͉ 2log e ͉ Ͻ 2, "positive" for 2 Ͻ ͉ 2log e ͉ Ͻ 5, "strong" for 5 Ͻ ͉ 2log e ͉ Ͻ 10, "very strong" for ͉ 2log e ͉ Ͼ 10 . In practice, however, the harmonic means of likelihood values may vary across runs: they are expected to converge to the same value if the chains are run to infinity. Consequently, I take the conservative approach recommended by Pagel and Meade , which disregards any evidence for either state given by ͉ 2log e ͉ Ͻ 2.
I determined the RJ-MCMC chain specifications through preliminary maximum-likelihood and MCMC runs, all with nodes not fossilized. These specifications ensure that the RJ-MCMC chains sample parameter space adequately and ultimately converge to the posterior probability distribution of states in the model of trait evolution. I ran the RJ-MCMC chains for 10 8 iterations, sampling every 10 3 , with an additional burn-in of 10 6 , and rate deviation set to 50. All chains used a uniform prior on the models and an exponential prior on the rate parameters; the mean of the exponential prior was seeded from a uniform hyperprior on the interval 0 -10. The shape of the prior distribution and the interval of the hyperprior only marginally affected the mean of the posterior probability distributions of log e values and of ancestral states, returning qualitatively similar results.
For each set of analyses, convergence to the posterior probability distribution of states in the model of trait evolution was assessed by comparing the samples returned by the separate chains, through visual inspection of time-series plots of log e values, the posterior probability distributions of model categories, and the average deviation of parameter estimates across runs. The near-independence of sampling events was judged from the autocorrelation of the log e values of successive states sampled by the chains. In all cases, these diagnostics indicated that the chains sampled the target distributions adequately.
---
Results
For each set of analyses, I compared the log e [H] values for the states in the model of trait evolution sampled by the separate chains at a. "Z" denotes rates assigned to the zero class. "0" and "1" denote two distinct non-zero rate classes; rates with the same value are assigned to the same non-zero rate class. For example, under model category 1 q MP and q PM are assigned to the same non-zero rate class, that is, they take the same positive value; under model category 2 q MP is assigned to the zero rate class, that is, it is set to zero, while q PM is assigned to a non-zero rate class, that is, it takes a positive value; under model category 3 q MP and q PM are assigned to distinct non-zero rate classes, that is, they take different positive values. b. The relative cumulative frequency of a model category is obtained by summing the absolute frequency of sampled points in the model category to the absolute frequencies of sampled points in all preceding categories, if any, and then dividing by the total number of sampled points in all categories ͓e.g. /10 5 ϭ 0.993 for model category 2͔.
convergence; I present results for the chain that returned the median value of the log e [H]. Below I discuss the ancestral state estimation at nodes on the consensus tree. The fossilization of nodes PIH and PIE largely confirms the ancestral state estimation and is discussed in the SI file.
---
Estimation of Ancestral States on the Consensus Tree.
At convergence, the chain sampled states in the model of trait evolution with mean log e Ϯ SD ϭ Ϫ11.43 Ϯ 0.93 . Below I example, q MP and q PM were assigned to the same rate class in 87.8% of the 10 5 sampled points . Over the four model categories, rates were assigned to 1.01 Ϯ 0.08 non-zero classes ; this indicates that the evolutionary transitions that produced the observed distribution of states of marriage strategy across societies in the sample can be described by a simple model of trait evolution based on only one non-zero rate class .
Nodes PIH and PIE reconstructed as monogamy with high posterior probabilities under model categories 1, 3, and 4, and as polygyny with p ϭ 1 under model category 2 . In the latter category, transitions from monogamy to polygyny are excluded, because q MP is set to zero . This forces nodes PIH and PIE to reconstruct as polygyny: under this category, any variation in states of marriage strategy at the tips of the tree is the result of transitions from polygyny to monogamy. In other words, the presence of polygyny at the tips could not be accounted for if PIH and/or PIE reconstructed as monogamy. However, models in category 2 returned a mean log e value 1.67 units worse than models in the category that returned the best mean log e value , and 1.47 units worse than the mean log e value over the four model categories. This suggests that the evolutionary scenario described by model category 2 is unlikely to have produced the observed distribution of states of marriage strategy across societies in the sample.
Posterior Probability Distributions of Ancestral States. Over the four model categories, nodes PIH and PIE reconstructed as monogamy with high posterior probabilities . The posterior probability distribution is more strongly skewed toward high values for state M at node PIE, as reflected in the higher value of p at this node .
Monogamy reconstructed with high posterior probabilities through to nodes A and B on the consensus tree summarizing the tree sample, but phylogenetic uncertainty limits the confidence that can be placed in these inferences . Node D and node F reconstructed as monogamy with high posterior probabilities. Node E reconstructed as polygyny with high posterior probability. The uncertainty in the reconstructions at the base of the consensus tree means that a host of scenarios can explain the distribution of states of marriage strategy at the tips . Polygyny was acquired at least once on the tree, between nodes PIE and E, and possibly a second time in the branch leading to Albanian G. A minimum of two reversals to monogamy occurred, one in the branch leading to Panjabi ST and one in the branch leading to Singhalese.
---
Discussion
The phylogenetic comparative analysis of marriage strategies across societies speaking IE languages provides evidence in support of PIE monogamy; this pattern likely extended back to PIH, but the evidence is only suggestive. Polygyny evolved at least once, possibly twice, throughout the history of IE-speaking societies, but uncertainty in some of the "deep" reconstructions means that alternative evolutionary scenarios cannot be excluded. This uncertainty emphasizes the need for an explicitly historical approach when deriving inferences about past social organization from cross-cultural data, and to cross-cultural comparison more generally. In turn, the uncertainty in the phylogenetic model used to represent how the societies are related by descent stresses the importance of avoiding "best tree" approaches in phylogenetic comparative analysis.
As discussed above, Fortunato and Mace used a "best" tree in the phylogenetic comparative analysis of marriage strategies and strategies of wealth transfer at marriage in IE-speaking societies. The present findings suggest that their reconstruction of PIH monogamy is not contingent upon the phylogenetic tree used. Further, while the outgroup Hittite was not assigned marriage strategy data in the present analysis, it was coded as monogamous in Fortunato and Mace's ; this was necessary because the phylogenetic comparative method they used did not allow missing data . The present findings indicate that the reconstruction by Fortunato and Mace is not contingent upon the state they assigned to the outgroup. Applying a phylogenetic comparative approach to the marriage transfer strategy data, in a Bayesian MCMC framework, Fortunato et al. showed that the reconstruction of PIH dowry is similarly robust to phylogenetic tree model and coding of the outgroup.
More generally, these reconstructions push the origin of monogamous marriage into prehistory, well beyond the earliest instances documented in the historical record. This implies that the archaeological and genetic evidence for the nuclear family in prehistoric populations may reflect a monogamous marriage strategy; on their own, the archaeological and genetic data can at best provide clues about a monogamous mating pattern. For example, Haak et al. found evidence, through analysis of aDNA samples, of genetic relatedness of one adult male and one adult female with two children recovered in one burial at the site of Eulau, Germany, a late Neolithic community attributed to the Corded Ware culture; they argued that this establishes "the presence of the classic nuclear family in a prehistoric context in Central Europe" . Similarly, Bentley et al. identified a nuclear family, comprising an adult male, an adult female, a mature female, and two children, through analysis of the isotopic signatures of skeletal remains in the communal grave at Talheim, Germany, an early Neolithic community attributed to the Linear Ware culture; previous analysis of the teeth had revealed high similarity between the male and the children, suggestive of genetic relatedness. The Corded Ware and Linear Ware cultures are archaeological horizons of northern and central Europe associated with populations speaking IE languages . The phylogenetic comparative analysis shows that monogamous marriage prevailed among prehistorical IE-speaking societies located in Europe. At least to the extent that evidence from a single grave can substantiate claims such as Haak et al.'s , this suggests that the burial patterns may reflect the monogamous marriage strategy of the Eulau and Talheim communities.
Larger-scale analyses of Y-chromosome data provided evidence for a later increase in effective male population size compared to effective female population size, across European and worldwide samples. One possible interpretation of this pattern is that the difference in effective population size was caused by a regime of polygynous mating, which resulted in greater variance in reproductive success for males than for females . Further, Dupanloup et al. interpreted the delayed increase in the effective male population as evidence for a relatively recent shift from polygynous to monogamous mating; based on Pritchard et al.'s estimates for the timing of expansion of Y-chromosome data, they linked this shift to the emergence of food production, between 10,000 and 5000 years ago in Europe and Asia and more recently elsewhere, when "Nuclear families replaced the polygamous, extendedfamily compounds typical of hunting-gathering populations" . Dupanloup et al. use the concept of "nuclear family" as synonymous with non-extended and monogamous family organization, although in anthropology it is used to designate a family unit comprising parents and their dependent children ; technically, therefore, nuclear families exist in societies practicing monogamous and polygamous marriage . The reconstruction of PIE monogamy presented in this paper, with monogamy likely extending back to PIH, falls within the temporal interval identified by Dupanloup et al. , suggesting that monogamous marriage had emerged in Eurasia by that time. However, Dupanloup et al.'s interpretation holds only to the extent that the shift to the "nuclear family" coincided with a shift from polygynous to monogamous mating, with consequent reduction in the variance in male reproductive success. The historical and ethnographic evidence suggest that variance in male reproductive success is comparable in societies practicing monogamous and polygynous marriage . In any case, this pattern is to be viewed against the background of moderately polygynous mating that is believed to have characterized our species' evolutionary past based on morphological and genetic data.
Finally, and most importantly, the phylogenetic comparative analysis confutes explanations linking the emergence of monogamous marriage to the development of features of social organization typically associated with the "complex," "modern" societies located in Eurasia. Indeed, investigation of the factors resulting in the shifts from monogamy to polygyny identified by the phylogenetic comparative analysis, and in the corresponding reversals, would be of particular interest. Shifts in marriage strategies may reflect, for example, changes in subsistence systems, as well as historical contingencies such as the diffusion of religious beliefs . Recent theoretical work situates variation in marriage strategies in the context of variation in ecological factors, linking the prevalence of monogamy across Eurasian societies to the development of intensive modes of production . Diamond has convincingly argued that the relative greater "complexity" of these societies, and their consequent "modernization," can also be attributed to ecological determinants that facilitated the emergence of food production in the region. This raises the possibility that ecological factors act as a confounding variable in the observed relationship between marriage strategies and indicators of "societal complexity" and "modernization." An appeal to "complexity" and "modernization" as the terminus of explanation amounts to providing a proximate answer to ultimate questions about the evolution of marriage strategies. | Explanations for the emergence of monogamous marriage have focused on the cross-cultural distribution of marriage strategies, thus failing to account for their history. In this paper I reconstruct the pattern of change in marriage strategies in the history of societies speaking Indo-European languages, using cross-cultural data in the systematic and explicitly historical framework afforded by the phylogenetic comparative approach. The analysis provides evidence in support of Proto-Indo-European monogamy, and that this pattern may have extended back to Proto-Indo-Hittite. These reconstructions push the origin of monogamous marriage into prehistory, well beyond the earliest instances documented in the historical record; this, in turn, challenges notions that the cross-cultural distribution of monogamous marriage reflects features of social organization typically associated with Eurasian societies, and with "societal complexity" and "modernization" more generally. I discuss implications of these findings in the context of the archaeological and genetic evidence on prehistoric social organization.To access this article as a PDF pay-per-view download via BioOne, please click here. |
Introduction
Endometriosis is a chronic, multisystemic disease of inflammation affecting approximately 10% of the female population . It is defined as the extra-uterine growth of endometrial glands and stroma and may present with a variety of symptoms such as pelvic pain and infertility, resulting in a significant negative impact on individuals' health and quality of life . The most widely regarded theory on the origin of endometriosis dates back to Dr Sampson, who postulated that endometriotic implants resulted from retrograde menstruation back in the 1920s, although reports of aberrant endometrial tissue growth date back to as early as 1860 .
---
Historical perspective
John A Sampson was interested in exploring the reason for infertility seen in patients with endometriosis, at a time of social concern regarding declining birth rates among upper-class women in the United States . In the context of such societal panic, Dr J Meigs proposed a theory that endometriosis was linked to contraceptive use and delayed childbearing seen most commonly in the 'well-to-do' . This theory gained ground for several decades, substantiated by methodologically flawed research demonstrating increased rates of endometriosis among private White patients compared to the ward Black patient, a dichotomy ridden with confounding and bias . Although some evidence to the contrary started to emerge in the 1950s, it was not until Dr Chatman presented his work in the 1970s that the view of low prevalence of endometriosis in Black patients began to shift . Nevertheless, by this point, a strong bias regarding the impact of race/ethnicity in the epidemiology of endometriosis was perpetuated in the medical community, evidenced by the narrative in medical education literature suggesting a rarity of endometriosis amongst non-white patients, present until the twentieth century . More recent texts continue to suggest lower prevalence of endometriosis diagnosis in Black and potentially higher prevalence among Asians, compared to White women .
---
Summary of evidence regarding race/ethnicity and endometriosis
Several publications dating back to the 1920s have investigated the epidemiological risk factors for developing endometriosis, including race/ethnicity. This literature was synthesized in a systematic review and meta-analysis by Bougie et al. in order to estimate the risk of endometriosis among various racial/ethnic groups. The review included 18 studies and identified that compared to White women, Black and Hispanic woman were less likely to be diagnosed with endometriosis , while Asian women were more likely to have this diagnosis . Significant heterogeneity was present in the analysis for all racial/ethnic groups, which may have stemmed from clinical variation of included study participants and definition of outcomes, as well as methodological differences in study design. Two studies were not included in the meta-analysis as they reported outcomes of interest in a format not compatible with data synthesis. First, Missmer et al. examined the incidence of surgically diagnosed endometriosis in the Nurse's Health Study II and found that Black women had lower rate of endometriosis diagnosis compared to White women , whereas Asian women had similar rates of disease compared to White women. Hispanic women had lower rates of endometriosis diagnosis compared to White women, although this did not reach statistical significance . These findings should be interpreted cautiously as the studied population was predominantly Caucasian. Secondly, Zhao et al. examined the prevalence of endometriosis-related hospitalization based on the Nationwide Inpatient Sample in 1991 and 1992 in the United States. They observed significantly lower rates of endometriosis diagnosis in Black , Hispanic , and Asian Pacific women compared to White women . A more recent retrospective cohort study using claims electronic health records estimates that 70% of patients diagnosed with endometriosis were White, 6% Hispanic, 9% Asian, and 4.7% non-Hispanic Black .
This work highlighted several important themes pertinent to race/ethnicity and endometriosis. First, many of the included studies were of poor methodologic quality and at significant risk of selection bias as well as confounding, particularly from socioeconomic status. Secondly, the majority of patients included in the studies were of White racial origin. Lastly, the majority of studies focused on disease prevalence, without exploring potential variability in presenting symptoms, diagnostic delays, or therapeutic response based on race. Literature looking at uterine fibroids, another gynecologic condition, suggests that there is significant variability amongst different ethnic groups in terms of symptom burden and clinical presentation, not necessarily in alignment with objective measures of disease burden . Flores-Caldera et al. recently presented the results of their international collaborative cross-sectional study 3:2 which established the phenotypic profile of Hispanic/ Latinx patients with endometriosis. Specifically, they identified substantial severity of symptoms, particularly dysmenorrhea and dyspareunia, high pain catastrophizing scores, and overall negative impact on quality of life .
Finally, the primary presenting symptom of endometriosis, pelvic pain, may limit clinical consideration of this diagnosis of this condition amongst non-White patients . Historically, medical education has perpetuated stereotypes surrounding Black patients and their experience of pain . Significant racial and ethnic disparities remain across different areas of pain care , with minorities receiving lesser quality pain care than non-Hispanic white patients . These encounters may be rooted in implicit and explicit biases held by healthcare providers including the notion that non-white patients have a higher pain threshold . Similarly, when stereotypes surrounding the prevalence of diagnoses like endometriosis are perpetuated within the broader community, racialized patients may be less likely to seek medical attention for their symptoms.
---
Race/ethnicity and endometriosis representation in medical education
Endometriosis and race have been discussed in medical education with heavy influence from the gynecology community. The integration of evidence-based medicine has stressed the importance of using high-quality evidence to support clinical practice; however, we must appreciate that there are many long-standing beliefs that are held as 'mantra' in medicine, supported mostly by flawed confirmation bias.
The perception of endometriosis as less prevalent in Black patients is widespread amongst foundational textbooks of gynecology, including but not limited to Williams Gynecology, Blueprints Obstetrics & Gynecology, and Speroff's Clinical Gynecologic Endocrinology and Infertility . Textbooks are important to examine as they are widely distributed as educational tools based on expert opinions, and until the recent shift to online resources, formed the foundation of medical education. For example, an excerpt from the sixth edition of Novak's Gynecology, published in 1961 states, 'There seems no doubt that endometriosis is much more common in the white private patient than in the dispensary clientele'. . By the 16th edition published in 2020, the section states that endometriosis 'is found in women from all ethnic and social groups'. . Whereas Novak's revised its content to remove all references pertaining to race and endometriosis, other textbooks removed blatant commentary while still alluding to an ongoing 'controversy' . Other examples of racial bias are more nuanced. For instance, the 2013 edition of Blueprints of Gynecology features a corresponding multiple-choice clinical vignette in which 'Her ethnicity is Caucasian' is correctly identified to increase suspicion for endometriosis . Similar commentary can be found in other gynecology textbooks dating back to the 1960s with the nature and severity of these assertions changing over time. It is of note that many of these comments are made without any appropriate citations.
---
The uses and misuses of race/ethnicity in medicine
The consideration of race and ethnicity in medicine and biomedical research has been a long-standing, charged, and complex debate. There is an established history of racial injustice in medicine and a hesitance to repeat past mistakes . There are two general positions: either there is strong utility to the inclusion of race in research and medical practice or race has no biological origin and thus should not be included in medicine . To understand each side, it is imperative to define this terminology, which has been used inconsistently. Race and ethnicity are primarily social constructs . They arose through geographical, social, and cultural forces, as opposed to defined biologic constructs. Race is often classified based on continental origin, and historically, genetic variation has been related to geographical mating patterns . Ethnicity is a further construct, related to geography but also considering religion, culture, and language. They are often a result of endogamous mating within continents; thus, they have genetic variation but less than continentally defined groups . Both race and ethnicity can influence socioeconomic status, resulting in unequal access to opportunities and resources and disproportionate morbidity and mortality .
Race/ethnicity is often used as a marker for underlying genetics. Epidemiologic and clinical research often divide endometriosis R37 3:2 participants into such categories to investigate hypotheses between environmental and genetic risk factors . Although social determinants of health and access to care must be considered, racial/ethnic differences are often seen despite controlling for such factors. For instance, in a study by Karter et al. , the rate of diabetic complications was evident despite using the same health maintenance organization and after adjustment for various social determinants. Conversely, while these risks are often reported as 'intrinsic differences' between races, they are likely capturing the risk of inequities from exposure to structural racism .
Single gene disorders are an example of successful discovery based on race/ethnic considerations. Noting that certain groups display certain diseases, geneticists hunt for a cause, leading to discoveries of the genes for Tay-Sachs, cystic fibrosis, and thalassemia . Mendelian disorders are often traced back to particular groups, such as Ashkenazi Jews, French Canadians, the Amish, or certain European backgrounds , but it should be noted that these groups are not defined by race. Genetic variation is still more commonly observed within continental populations, as opposed to among them . Common chronic diseases, such as cardiovascular disease, diabetes mellitus, and kidney disease, have been seen worldwide, confirming that all populations are susceptible, and variation is more likely due to environment. While race can help target screening for disease-associated mutations, the only way to diagnose a DNA-sequence variant is to test for it . It should also be noted that despite all the literature on the topic, race has not been defined in genetic terms .
Further application of race/ethnicity has been in drug responsiveness. It is well known that functional variants of genes encoding drug-metabolizing enzymes exist and understanding a patient's background can allow clinicians to predict drug responsiveness and tailor therapies accordingly . For instance, there has been substantial work in targeting 'race-specific' medical therapy in cardiovascular disease. The Clarification of Oral Anticoagulation through Genetics trial showed a difference between the responsiveness of populations of European and African ancestry with respect to warfarin dosing, necessitating a race-specific approach to treatment . However, researchers must be cautious, as other randomized trials were interpreted to show different responses by race, but further analysis of the results revealed this conclusion was a type I error . These assumptions may lead to clinicians withholding certain medications from certain racial groups, thereby exacerbating the differential care seen between groups.
While the above examples describe positive attempts at using race in medicine, there are far more examples of its misuse, of which one of the most interesting areas is in diagnostic algorithms. Used as a concrete proxy for bias, this can be seen throughout all fields of medicine. Vyas et al. explored some of the most common diagnostic algorithms. By including race in these calculations, they suggest that race-based medicine is being propagated, guiding decisions that may further direct resources away from minorities.
One of the most widely described algorithms incorporating race is that of estimated glomerular filtration rate . The formula predicts higher eGFR, meaning better kidney function, for Black patients. This is supported by evidence that higher serum creatinine levels are seen in Black people, potentially due to increased muscularity . This assumption may result in delayed referrals, and indeed Black people have higher rates of end-stage renal disease. On the other hand, ignoring race in this algorithm may lead to the prediction of worse kidney function in such patients and result in overtreatment and inappropriate drug dosing . Similarly, the Kidney Donor Risk Index finds that black donor kidneys perform worse, and given that Black patients are more likely to receive organs from Black donors, it is not surprising that they have long wait times for renal transplantation .
Algorithmic inclusion of race can be seen across the medical specialties. The vaginal birth after Caesarean risk calculator predicts a lower chance of success if a patient identifies as black or Hispanic, which may deter clinicians from offering a trial of labour . Indeed, non-White Americans have higher rates of Caesarean sections, and it is well known that Black patients have increased rates of maternal mortality . While some algorithm developers do offer sources for these adjustments, these are often found to be outdated and biased. The racial distinctions seen in large datasets are more likely reflecting the toxic effects of racism, such as its physiological consequences, discrimination, and access to care . Borrell et al. cautions that when using standardized algorithms, clinicians should consider whether the inclusion of race would decrease health inequities, leading to better health outcomes.
The consideration of race has also led to knowledge gaps in medical research. Years of insufficient funding for research in minority populations have led to questionable generalizability of medical advances to such groups. For instance, less than 2% of National Cancer Institute-funded clinical trials and less than 4.5% of federally funded pulmonary research have included minority populations . The National Institutes of Health requires reporting of all racial or ethnic groups, and despite this, there is a paucity of information and minimal progress in including minority groups in large trials .
Endometriosis is another condition affected by the misuse of race. As reviewed above, it has historically been viewed as a condition of White women and a systematic review of the literature identified a strong focus of research on White women, with minimal data on minority groups . While some studies postulate that endometriosis is higher in Asian women and lower in Black women compared to their Caucasian counterparts, other studies comparing women of different races with equal indications and socioeconomic status have failed to note a difference . It is commonly thought that those women of African descent rarely have endometriosis, yet it is one of the most common reasons for African American women in the United States to undergo gynecologic surgery . It has also been shown that in private patients admitted for such surgeries, the prevalence of endometriosis was similar between African American and Caucasian patients . Interestingly, Kyama et al. reported a significantly lower rate of endometriosis in African-Indigenous women compared to African American, indicating that race alone is not an explanation, and more likely due to lack of awareness, lack of access to laparoscopy, limited training on diagnosis and treatment, lack of research interests, and lifestyle factors . In their retrospective study, Shade et al. reviewed charts of African American patients that had undergone surgery for endometriosis and noted that 93% of patients demonstrated uterine endometriotic implants. Although this was a retrospective study without a comparator group, the authors suggest that this finding may indicate a variation in disease presentation compared to other racial/ethnic groups.
The debate regarding the use and misuse of race/ ethnicity in medicine is ongoing and fraught with complexities. While race is not a reliable proxy for genetic difference, we must acknowledge that differences do exist between people of different racial categories and this is clinically meaningful . Ultimately, replacing race with genetic ancestry would result in more informative, evidence-based approaches, yet this technology
is not yet readily available outside the research environment . While ignoring race may improve equality, it is only through the equity that racial disparities can be tackled .
---
Phenomics and endometriosis
The advent of genome-wide association studies has gathered a great amount of information suggesting highconfidence genotype-phenotype associations between specific genomic loci and a large number of diseases, including diabetes, obesity, Crohn's disease, and hypertension. Recognizing the importance of considering phenotypic variation among endometriosis patients, Vigano et al. conducted a comprehensive review of the relationship between morphometric traits and endometriosis. They identified some association between BMI and particularly pigmentary trains/presence of nevi and diagnosis of endometriosis. More importantly, they drew attention to the consideration of genomic contribution to the phenotype of endometriosis. Although currently the mechanisms underlying genotype-phenotype relationships remain only partially explained and must be interpreted in the context of multiple limitations, including the inherent variability in quality of published data and the higher order complexity of the genotype-phenotype relationship, it is easily imaginable that a catalogue of nearly all human genomic variations and their relative impact on human diseases will be available within our lifetime . We must appreciate that even with expansion of genomic information that will become available, the majority of phenotypic variation seen amongst endometriosis and other medical conditions will stem from genotype-environment interaction . Genetic variation between populations tends to be geographically structured, as expected from the partial isolation of human populations during much of their history . In this regard, race may be considered first-order approximation to the geographically structured phenotypic variation in the human species . The influence of environmental exposure, resulting from geographic variation, on the development of endometriosis has been suggested .
---
Conclusions and future work
In summary, we see that there is a strong historical bias regarding the epidemiological impact of race/ethnicity on 3:2 the prevalence of endometriosis. Although the summation of current evidence suggests that endometriosis may be less common amongst Black and Hispanic women, compared to White women, it is imperative to recognize the significant methodological flaws and bias driving the studies performed to date. Furthermore, it is important to question the relevance of this information in the provision of outstanding and individualized patient care. Recognizing the narrative on this topic to date, we suggest the following clinical, education, and research priorities moving forward: . Moreover, it is important to note that race should not be used as a proxy for biological markers or genetic predisposition for a disease . Race also should not be used to simplify systemic contributors when teaching around the underlying cause for disparities in diagnoses . Medical trainees must be taught social determinants of health in order to holistically serve diverse patient populations.
3. Further investigation into the epidemiologic risk factors predisposing the development of endometriosis should be encouraged in order to identify at-risk individuals and implement early detection and appropriate treatment of the condition. Research exploring the unique presentation and treatment of endometriosis amongst patients may explore race and ethnicity but needs to include these factors in a sensitive manner, reflecting on underlying social constructs. Any researcher including race/ethnicity when studying endometriosis should reflect a priori on the reason and implications of including this factor in their study.
We advocate for the adaptation of an individualized and patient-centred approach to the management of endometriosis to achieve more equitable and improved care provision for all endometriosis patients.
---
Declaration of interest
O Bougie has participated in a speakers' bureau and received research grants and consulting fees from Bayer Pharma, Allergan, Hologic and AbbVie. The authors confirm these sponsorships had no involvement in the study. I N and C W have no disclosures.
---
Author contribution statement
All authors contributed to the research and manuscript preparation. They all approved the final submitted version. | Endometriosis is a chronic, multisystemic disease often presenting with significant phenotypic variation amongst patients. The impact of race/ethnicity on the prevalence of endometriosis, as well as disease presentation, is a question of interest which has been explored for the last century. This narrative review explores the historical perspective of endometriosis and race/ethnicity as well as the evidence available to date. Furthermore, we discuss the potential implication of the bias perpetuated on this topic, specifically in the areas of medical education, research, and clinical care. In consideration of these intersecting realms, we suggest priorities for future consideration of race/ethnicity as it pertains to the delivery of care for endometriosis patients. |
"Introduction\n\nFamily is a very important social institution of the society. It is one of the basi(...TRUNCATED) | "Given that social institutions are the basis of any society, they must ideally operate in unison wi(...TRUNCATED) |
"HOW ARE OLDER PEOPLE FARING POST-PANDEMIC? EVIDENCE FROM THE ENGLISH LONGITUDINAL STUDY OF AGING\n\(...TRUNCATED) | "questions were included to explore caregivers' feelings of preparedness and family support. Of all (...TRUNCATED) |
"Introduction\n\nPhysical inactivity can be a fundamental cause of various public health problems, e(...TRUNCATED) | "Neighborhood built environment may influence residents' physical activity, but evidence of non-majo(...TRUNCATED) |
"Background\n\nProviding cost-effective care for the increasing numbers of complex elderly and/or co(...TRUNCATED) | "Background: Patients with chronic conditions and multiple comorbidities represent a growing challen(...TRUNCATED) |
End of preview. Expand
in Dataset Viewer.
README.md exists but content is empty.
Use the Edit dataset card button to edit it.
- Downloads last month
- 34