Construct validity of electronic health record surveillance indicator definitions of smoking prevalence and hypertension prevalence, treatment and control

Purpose: Electronic Health Record data offers great potential for population health surveillance, but little is known about how EHR-based surveillance indicators should be defined or interpreted outside of the syndromic surveillance context. Here we describe the construct validity of NYC Macroscope electronic health record surveillance indicator definitions of smoking prevalence and hypertension prevalence, treatment and control.

Methods: NYC Macroscope prevalence estimates were compared to estimates from the 2013 NYC Community Health Survey (CHS) and the 2013-14 NYC Health and Nutrition Examination Survey (NYC HANES). Evaluation metrics included the two one-sided test of equivalence, t-test, prevalence difference and prevalence ratio. Internal consistency was assessed across strata defined by sex and age group.

Results: NYC Macroscope (15.3%) estimates of smoking prevalence were between CHS (14.9%) and NYC HANES (17.7%) estimates and fit better with each than the two reference surveys fit with each other. With respect to measures of hypertension prevalence, estimates of diagnosed hypertension (32.3%) compared very well to estimates of self-reported hypertension in CHS (31.6%) and NYC HANES (32.4%), and the fit of a clinically-informed prevalence definition between NYC Macroscope and NYC HANES estimates is excellent (33.7% vs 35.4%). NYC Macroscope estimates of hypertension treatment fit poorly with NYC HANES estimates overall (79.4% vs 63.9%) but fit well for adults age 60 and older (91.5% vs 89.2%). Estimates of hypertension control were higher than NYC HANES estimates both among adults with history of hypertension (65.7% vs 58.6%) and hypertensive adults with treatment (63.8% vs 51.7%).

Conclusion: The construct validity of NYC Macroscope estimates of smoking and hypertension prevalence is excellent based on comparison with CHS and NYC HANES. We still have more to learn about how to define and interpret EHR-based measures of hypertension treatment and control. Population-based estimate comparisons are only the first step. Diagnostic validity must still be established through chart review studies.