krishnareddy/icddxdescmap

ICD10 Diagnosis Description Mapping Dataset

Overview

This dataset is designed to assist in mapping ICD10 Diagnosis descriptions documented in clinical documents to the standard ICD10 Diagnosis descriptions by CMS (Centers for Medicare & Medicaid Services). The primary objective is to train a model that can map free-form disease text to ICD Codes.

Dataset Details

The dataset consists of the following columns:

• AnnotationString: This column contains the disease text as described in patient clinical documents.
• DXCode: This column contains the standard DX code corresponding to the description in the AnnotationString column.
• ShortDesc: This column contains the standard ICD Description in a short form corresponding to the DXCode.
• LongDesc: This column contains the standard ICD Description in a long form corresponding to the DXCode.

Model Training Objective

The goal of this dataset is to train the LLAMA2 7B Model to generate one of the following outputs based on the provided AnnotationString as input:

Case 1: Expecting DX Code as Output

• Instruction: As an ICD Code helper, the model needs to generate an ICD 10 DX Code for the corresponding disease description.
• Input: <Annotation String>
• Output: <DXCode>

Case 2: Expecting ShortDesc as Output

• Instruction: As an ICD Code helper, the model needs to generate the standard ICD 10 DX Code Short Description for the corresponding clinical text.
• Input: <Annotation String>
• Output: <ShortDesc>

Case 3: Expecting LongDesc as Output

• Instruction: As an ICD Code helper, the model needs to generate the standard ICD 10 DX Code Long Description for the corresponding clinical text.
• Input: <Annotation String>
• Output: <LongDesc>

Usage

To utilize this dataset effectively for training the LLAMA2 7B Model, you can follow these steps:

1. Preprocess your input data to match the format specified in the "Instruction" for the desired case (1, 2, or 3).
2. Use the preprocessed data as input to your model.
3. The model will generate the corresponding output (DXCode, ShortDesc, or LongDesc) based on the case you specified.

Example

Here's an example of how to use this dataset and the model:

Input

# Instruction: As a ICD Code helper, you need to generate ICD 10 Dx Code for corresponding following Disease description
# Input: "Patient presents with persistent cough and fever."

Output (Case 1)

# Output: "J44.9"

Output (Case 2)

# Output: "Chronic obstructive pulmonary disease, unspecified"

Output (Case 3)

# Output: "Chronic obstructive pulmonary disease, unspecified"

Warning: Potential Mapping Errors

This dataset has been harnessed from another rule-based NLP application, and it's important to be aware of the possibility of errors in the mapping of clinical text to standard codes or standard descriptions. While every effort has been made to ensure accuracy, there are inherent challenges in automated mapping processes, especially when dealing with complex medical terminology and variations in clinical documentation.

Users should exercise caution when relying on the generated results and consider the following:

1. Clinical Variation: Clinical documentation can vary significantly in style and terminology, which may lead to inaccuracies in mapping.

2. Complex Cases: Some medical conditions or descriptions may not have a one-to-one mapping to standard codes or descriptions, resulting in potential discrepancies.

3. Data Quality: The accuracy of the mapping is also dependent on the quality and consistency of the input clinical text.

4. Continuous Improvement: Efforts are ongoing to improve the accuracy of the mapping, and users are encouraged to provide feedback and report any discrepancies.

It is advisable to cross-verify the results generated by this dataset, especially in critical healthcare applications where accuracy is paramount.

Please use this dataset responsibly and in conjunction with clinical expertise to make informed decisions.

If you encounter any issues or have suggestions for improvement, please reach out to the dataset maintainers.

Thank you for your understanding and cooperation.

Citation

If you use this dataset or the pre-trained model in your research or applications, please consider citing the source or authors for proper attribution.

---

Feel free to modify and expand this readme file to include any additional information or instructions specific to your use case.