Predictive Validity of an Adult Physical Activity "Vital Sign" Recorded in Electronic Health Records

Background: Few have examined predictive relationships between physical activity (PA) and health using electronic health records (EHRs) of patient-reported PA. Objective: Assess initial predictive validity of the Physical Activity "Vital Sign" (PAVS) recorded in EHRs with BMI and disease burden. Methods: EHRs were from November 2011 to November 2013 (n = 34,712). Differences in not meeting Physical Activity Guidelines (PAG) were tested using chi-square analysis between being normal weight versus overweight/obese, and scoring below versus above the 50th percentile of the Charlson Comorbidity Index (CCI). Repeated measures logistic regression was used to determine odds of BMI and CCI classifications according to responses to the PAVS as not meeting PAG. Results: Patients who did not meet PAG according to the PAVS were more likely than normal weight patients to have a higher BMI (BMI 25.0-29.9, OR = 1.19, P =.001; BMI 30.0-34.9, OR = 1.39, P <.0001; BMI 35.0-39.9, OR = 2.42, P <.0001; BMI >= 40, OR = 3.7, P <.0001) and also higher disease burden (above 50th percentile for CCI, OR = 1.8, P <.0001). Conclusions: The strong association of the PAYS found with patient BMI and moderately-strong association with disease burden supports initial predictive validity of the PAYS recorded in EHRs. PA recorded in EHRs may be vastly useful for assessing patient disease and cost burdens attributed independently to PA behavior.