Diagnostic accuracy of unattended automated office blood pressure measurement in screening for hypertension in Kenya

Abstract

Despite increasing adoption of unattended automated office blood pressure (uAOBP) measurement for determining clinic blood pressure (BP), its diagnostic performance in screening for hypertension in low-income settings has not been determined. We determined the validity of uAOBP in screening for hypertension, using 24-hour ambulatory BP monitoring as the reference standard. We studied a random population sample of 982 Kenyan adults; mean age, 42 years; 60% women; 2% with diabetes mellitus; none taking antihypertensive medications. We calculated sensitivity using 3 different screen positivity cutoffs (≥130/80, ≥135/85, and ≥140/90 mm Hg) and other measures of validity/agreement. Mean 24-hour ambulatory BP monitoring systolic BP was similar to mean uAOBP systolic BP (mean difference, 0.6 mm Hg; 95% CI, −0.6 to 1.9), but the 95% limits of agreement were wide (−39 to 40 mm Hg). Overall discriminatory accuracy of uAOBP was the same (area under receiver operating characteristic curves, 0.66–0.68; 95% CI range, 0.64–0.71) irrespective of uAOBP cutoffs used. Sensitivity of uAOBP displayed an inverse association (P<0.001) with the cutoff selected, progressively decreasing from 67% (95% CI, 62–72) when using a cutoff of ≥130/80 mm Hg to 55% (95% CI, 49–60) at ≥135/85 mm Hg to 44% (95% CI, 39–49) at ≥140/90 mm Hg. Diagnostic performance was significantly better (P<0.001) in overweight and obese individuals (body mass index, >25 kg/m2). No differences in results were present in other subanalyses. uAOBP misclassifies significant proportions of individuals undergoing screening for hypertension in Kenya. Additional studies on how to improve screening strategies in this setting are needed.