Abstract P71: The Accuracy of Coded Obesity in an Administrative Database

Abstract

Objective: Obesity is becoming an increasingly prevalent problem. Gathering information on the adiposity of a population is difficult, so being able to take advantage of existing data, such as that in administrative databases, is appealing. The objective of our study was to assess the validity of obesity coding in administrative databases. Methods: This study was conducted using the Alberta Provincial Project for Outcomes Assessment in Coronary Heart Disease (APPROACH) database and the Discharge Abstract Database (DAD) for Calgary. BMI was calculated within APPROACH; BMI ≥30kg/m2 defined obesity. In the DAD obesity was defined by diagnosis codes 278 (ICD-9-CM) and E65-E68 (ICD-10). Databases were linked using provincial health numbers. The sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) of a diagnosis of obesity in the DAD was determined using the obesity diagnosis in APPROACH as the referent. The accuracy of coding obesity was compared across demographic categories and diagnoses. Results: A total of 17,380 subjects included in the analysis. The study population was largely male (68.8%) and had a mean BMI of 26.96 kg/m2. The overall sensitivity of a diagnosis of obesity in the administrative data was 7.75%. However, it was highly specific at 98.98%, with a NPV of 80.84% and a PPV of 65.94%. When considered by year, there were minor variations in the sensitivity of obesity coding in the administrative data, but it remained poor at under 10% throughout. The prevalence of obesity and the PPV was higher amongst those subjects with conditions associated with obesity, including diabetes and hypertension. Of those coded obese in DAD, the majority (72.89%) were Class I obese; of those not coded obese, 84.31% were Class I obese. Conclusions: Obesity coding in the DAD is poor, as reflected in the low sensitivity of the diagnostic code. However, once obesity is coded in this database, it is coded highly accurately. At present, using administrative databases to define cohorts of obese subjects for surveillance is not a viable option.