Validation of the Instant Blood Pressure App
Embargo until
2019-05-01
Date
2016-05-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Johns Hopkins University
Abstract
Introduction: The application of portable technology to healthcare is known as mobile health (‘mHealth’). Instant Blood Pressure (IBP) is an mHealth app that measures blood pressure (BP) using the internal sensors in an iPhone, no cuff required. It was a popular app and user reviews document its use in management of hypertension and other BP-related conditions. It has never been independently validated.
Methods: We enrolled adults from 5 ambulatory clinics in 2015, excluding those with an internal device, active arrhythmias, or who were unable to use the app. Participants guessed their BP then had two order-randomized pairs of BP measurements taken from IBP and a sphygmomanometer (‘standard’ measurement). Mean absolute differences for BP were calculated comparing each IBP measurement to an average of the two standard measurements. We also calculated mean relative differences (IBP minus standard), British Hypertension Society (BHS) accuracy grading, and sensitivity and specificity for the detection of hypertensive measurements. Mean absolute differences and mean relative differences were calculated for successive same-device BP measurements. We regressed systolic and diastolic BP on IBP on age, sex, height, weight, and HR.
Results: Of the 85 participants, 52% were women, mean (SD) age was 56.6 (16.3) years, BMI was 27.6 (5.7) kg/m2; 53% self-reported hypertension. Mean absolute difference was 12.4 (10.5) mm Hg for systolic and 10.1 (8.1) mm Hg for diastolic. Mean relative difference was -1.2 (16.2) mm Hg for systolic and 7.1 (10.8) mm Hg for diastolic. IBP achieved the lowest possible BHS accuracy grades. Sensitivity and specificity of IBP for detection of hypertensive-range BP were 0.22 and 0.92. For BP, successive IBP measurements varied less than successive standard measurements. Regression analysis found that 68% and 85% of the variability of IBP systolic and diastolic BP results were attributable to age, sex, height, and weight and HR.
Conclusions: BP measurements from an mHealth app with >148,000 copies sold were inaccurate. The low sensitivity for detection of hypertension means that 78% of hypertensive BPs were misclassified as non-hypertensive. We suspect that the algorithm may derive its results from population curves of BP for entered age, sex, height, and weight.
Description
Keywords
Mobile Health, mHealth, Mobile Apps, Digital Health, Blood Pressure, Blood Pressure Measurement, Hypertension, Validation