touchpy is a Python implementation of the software tools developed in the H-CUP (Healthcare Cost and Utilization Project) at AHRQ (Agency for Healthcare Research and Quality) and it provides functions to map ICD-9 or ICD-10 code to AHRQ comorbidity measures.

This project is still under development now and needs testing for performance and correctness for ICD-10 codes.

For ICD-9 codes, the performance and correctness has been tested on millions of admission records against the SAS program from HCUP.

• touchpy only requires Python 3 (≥ 3.3) and its standard modules. No extra module is needed.

• The input data has to be a comma-separated values (CSV) file, which contains headers or column names for ICD-9 or ICD-10 diagnostic codes, and diagnosis-related group (DRG) codes, etc. in the first row.

• There has to be at least one column of diagnostic codes and the column names of diagnostic codes have to start with DX or dx (not case-sensitive). In addition, the principal diagnosis (usually DX1) must be given before secondary diagnoses.

• There should be only one column of DRG codes and the column name has to start with DRG or drg (not case-sensitive). (If multiple columns of DRG codes are detected, the last one will be used.)

• Missing diagnostic codes should be indicated by blanks (white spaces), NA, or na (not case-sensitive).

First of all, we need manually download or clone this repository by

git clone [email protected]:wenjie2wang/touchpy.git
cd touchpy

A Makefile is provided to simplify usage. We should be able to generate output CSV file by simply calling make icd9 or make icd10 in the terminal for ICD-9 or ICD-10 codes, respectively. We will then be asked to enter the path of the input CSV file. The output file will be generated under the current directory named after the input file with a trailing _touch tag.

For example, suppose we want to generate comorbidity measures from ICD-10 codes and the input file is data/sample_icd10.csv, we may simply call make icd10 and enter data/sample_icd10.csv in the terminal. The output file will be sample_icd10_touch.csv, which contains the following 29 Elixhauser comorbidity measures: CHF, VALVE, PULMCIRC, PERIVASC, HTN_C (either HTN or HTNCX), PARA, NEURO, CHRNLUNG, DM, DMCX, HYPOTHY, RENLFAIL, LIVER, ULCER, AIDS, LYMPH, METS, TUMOR, ARTH, COAG, OBESE, WGHTLOSS, LYTES, BLDLOSS, ANEMDEF, ALCOHOL, DRUG, PSYCH, DEPRESS.

For the first time usage, we need first generate dictionaries for ICD-9 or ICD-10 codes by

python3 parse_icd9.py   # for ICD-9 codes
## or
python3 parse_icd10.py  # for ICD-10 codes

The corresponding Python dictionary objects will be generated under dict/ with a text file named icd9_dictionaries.txt or icd10_dictionaries.txt under the current directory. We should not modify or remove these files manually.

Then we may call the main script by

python3 icd9.py   # for ICD-9 codes
## or
python3 icd10.py  # for ICD-10 codes

Similarly, we will asked to enter the path of the input CSV file. The output file will be generated under the current directory named after the input file with a trailing _touch tag.

Wenjie Wang, with Python dictionaries generated by parsing the SAS scripts from HCUP.

touchpy is free and open source software, licensed under GPL (≥ 3).