OBJECTIVES: Body mass index (BMI) scores are used to define obesity, an important indicator of health status and a risk factor for numerous chronic diseases. This study aimed to assess the value of enhancing BMI records from the Komodo Research Dataset (KRD), a claims database, with EHR-derived BMI data from Komodo Clinical Observations (KCO), a database of structured clinical variables data.

METHODS: A retrospective study was conducted on adult patients (>18 years) from January 1, 2017 to September 30, 2023. The number of patients with BMI records identified from KRD alone using ICD-10 codes (Z68.1, Z68.2, Z68.3, Z68.4) was compared with those identified from KRD or KCO. The frequency of BMI records per patient in both datasets was also analyzed. To assess the quality of BMI obtained from KCO, the average BMI from KCO was compared with those in KRD (median of ICD coded range) for patients present in both databases. Findings were reported for the overall study period and stratified by year and gender.

RESULTS: KRD and KCO together identified 49.1% more patients with at least one BMI record (n=148,178,206) compared with KRD alone (n = 99,400,510). The frequency of BMI observations per person remained similar by supplementing with KCO (5.2 (standard deviation [SD] 8.8) vs KRD alone 5.5 (SD 8.8). Among patients with both KRD and KCO records, their average BMI from KCO (31.5 [SD 7.0]) is within range of that from KRD (32.2 [SD 9.2]) with 78.9% having their KCO BMI within 2-point range of their KRD BMI, demonstrating the reliability of KCO-derived BMI. Similar results were observed by year and gender.

CONCLUSIONS: Incorporating clinical observation values from EHR with claims-based diagnosis increases the identification of patients with BMI records and their frequency of BMI records without compromising the quality of BMI scores.