OBJECTIVES

To utilize a pre-existing multi-center EMR database to identify patients of interest, usually ICD-10-CM or drug prescribed will be used. However this may not be appropriate for studies in China. We aim to summarize challenges in identifying study population using EMR databases in China and propose key steps.

METHODS

In-depth review of three EMR databases in China and feasibility assessment to test different patient identification algorithms.

RESULTS

Key challenges in identifying study population in EMR databases are: 1. diagnoses might be initially recorded as free-text in natural language in HIS and transferred to the database before coded by the hospital IT service; 2. there were at least 6 different versions of ICD codes in China and there was not a wide-accepted mapping between these versions; 3. The most widely used ICD-10-codes BJ cannot be directly mapped to ICD-10-CM (International version); 4. The accuracy of diagnosis coding in HIS depends on experiences and knowledge of hospital IT service; 5. Some multi-center databases coded diagnoses from free-texts to ICD codes and the reliability of the coding process depends on the capacity of database’s medical team and IT team. 6. “primary diagnosis” and “comorbid conditions” might not be stored as structured and stand-alone data fields which can be directly used.To identify patients in EMR databases in China, ICD-10 codes may be useful but cannot be used as the only tool. Chinese keywords can be very powerful to identify patients however the selection and keyword and searching strategy need to be careful designed, tested and validated. Drugs may be helpful depending on the conditions of interest.

CONCLUSIONS

For accurately identifying patients of interest in EMR databases in China, an algorithm using the combination of ICD codes and condition-specific Chinese keywords need to be developed and validated to meet the special needs of a study.