Presentation (CTI)

OBJECTIVES: Body mass index (BMI) is an important measurement in real-world evidence-based studies. However, variability in how BMI is captured between recorded lab values and ICD diagnosis codes can affect study validity. This abstract evaluates agreement between obesity-related ICD codes and BMI lab values in the context of TriNetX data.
METHODS: Using the TriNetX claims-EMR dataset Linked, patients were identified with a BMI vitals measurement on or after January 1, 2016. An index BMI value was selected from the first eligible date where the patient was 18+, had at least 1 year prior of continuous enrollment and no record of a positive pregnancy result in the 6 months prior. Patients were grouped by whether or not they had an obesity-related ICD code within 6 months before or after the index BMI date. Obesity categories were classified as underweight/normal weight (<25 kg/m²), overweight (25-29.9 kg/m²), obese (30-39.9 kg/ m²) and severe obesity (>=40 kg/ m²). Calculated measures of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were used to assess accuracy for each obesity category in the cohort.
RESULTS: 2,488,363 adult patients had a BMI lab measure on or after January 1, 2016. 793,697 (39%) of which had a proximal obesity-related ICD code in addition to meeting all study criteria. Among all patients, measures of specificity and PPV were consistently high across all BMI categories. The specificity of severe obesity was 99.1%, 97.5% for obese, 98.1% for overweight, and 98.9% for normal/underweight patients. PPV was 88.8% for normal/underweight, 81.5% for overweight, 86.5% for obesity, and 83.5% for severe obesity. Sensitivity measures were low because 61% of patients had missing ICD-10 codes but improved with each increase in obesity category.
CONCLUSIONS: Obesity-related ICD codes are consistent with recorded BMI vital measurements in categorizing patients from the TriNetX EMR-claims databases.