OBJECTIVES

The granularity of ICD-10-CM diagnosis codes affords researchers the opportunity to perform more in-depth analyses, but their clinical validity requires further exploration. The objective of this study was to validate ICD-10-CM diagnostic coding practices for obesity relative to a reference standard of BMI measurements within a broad US, integrated claims/EMR dataset.

METHODS

Patients with a BMI result from 01/01/2016 to 12/31/2017 were identified using the HealthCore Integrated Research Database (HIRD) with EMR data. The reference standard BMI selected was the latest EMR measurement within the intake period that also had at least 6 months of pre- and post-eligibility. The ICD-10-CM diagnosis code for obesity/BMI that was closest in time was extracted and compared to the reference standard BMI. Validation measures such as sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated overall and for disease-specific subgroups.

RESULTS

A total of 591,029 patients with BMI results were identified (mean age = 50±16 years; 65% female). ICD-10-CM diagnosis codes for obesity/BMI were found in 28.0% of patients overall, increasing to 38.4% of patients with cardiovascular disease, 45.3% for type 2 diabetes, and 54.8% for nonalcoholic steatohepatitis. Sensitivity/PPV/specificity/NPV for the overall population were as follows for the various categories: underweight (14.0%/37.8%/99.7%/98.8%), normal weight (8.2%/90.6%/99.7%/75.2%), overweight (15.6%/76.7%/97.8%/71.4%), and obese (42.5%/92.0%/97.5%/71.4%). Sensitivity increased to 52.2%, 54.0% and 62.0% in obese patients with cardiovascular disease, type 2 diabetes, nonalcoholic steatohepatitis, respectively, with little change in PPV.

CONCLUSIONS

Based on these results observed in the HIRD integrated claims/EMR data (high PPV and low sensitivity), use of ICD-10-CM diagnosis codes in the overall population to identify overweight and obese cohorts of patients for research purposes appears reasonable, but not for situations in which identifying presence or absence of these conditions is critical. Variations by specific disease-state populations warrant further consideration for research uses.