Comparison of Comorbidity Indices Between Electronic Health Records (EHR) Derived Database and Claims Data Among Patients With Metastatic Breast Cancer


Asfaw AA1, Ascha M2, Yerram P1, Reiss S3, Brake S1, Wadé NB1
1Flatiron Health, Inc, New York City, NY, USA, 2Flatiron Health, Inc, Oakland, CA, USA, 3Flatiron Health, Inc, Merrick, NY, USA

Presentation Documents

OBJECTIVES: Electronic health records (EHR) and healthcare claims are primarily collected for non-research purposes, therefore it’s important to understand comorbidity data completeness and their comparison. This study compared EHR and claims on individual comorbidities and Charlson comorbidity index (CCI).

METHODS: Patients diagnosed with HR+/HER2- metastatic breast cancer (mBC) between 1/2011 and 9/2020 receiving first-line treatment were selected from a linkage of mBC de-identified Flatiron Health Research Database (FHRD) and Komodo Health claims data. Claims-based coding algorithms identified comorbidities using ICD codes and individual comorbidity status was abstracted from patient charts in EHR. Aggregate CCI kappa, sensitivity, specificity, predictive values, and hazard ratio associated with each data source were used to assess alignment.

RESULTS: Study included 654 patients from FHRD who had overlapping claims coverage. Patients with comorbidities according to either data source had similar baseline characteristics. Proportions of patients having a CCI score of 0 was 56% vs 55% for FHRD and claims, for CCI score of 1, 24% vs 23%, for a CCI score of 2+, 20% vs 22%. Kappa of dichotomized CCI (any comorbidity vs none) comparing FHRD to a reference of claims was 0.39 (0.31, 0.46); sensitivity, 0.65 (0.60, 0.71); specificity, 0.73 (0.68, 0.78); PPV, 0.67 (0.61, 0.72); NPV, 0.72 (0.67, 0.77). Kappa of multiple CCI categories (0, 1 and 2+) was 0.32 (0.26, 0.37). Patients without any EHR-derived comorbidities had comparable hazard of mortality as those without any claims-derived comorbidities (HR: 1.07, (0.93, 1.23), p=0.35), and similar hazards of mortality were observed for CCI scores of 1 or 2+.

CONCLUSIONS: Proportions of patients having comorbidity and corresponding survival based on CCI score were similar between FHRD and claims. The data sources were discordant in the specific patients in the comorbidity categories. Using both data sources maximizes information selecting patients having CCI comorbidities.

Conference/Value in Health Info

2022-11, ISPOR Europe 2022, Vienna, Austria

Value in Health, Volume 25, Issue 12S (December 2022)




Epidemiology & Public Health, Organizational Practices, Study Approaches

Topic Subcategory

Best Research Practices, Disease Classification & Coding, Electronic Medical & Health Records


No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Oncology

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on Update my browser now