Heart Snapshot: a broadly validated smartphone measure of VO₂max for collection of real world data

Abstract

<jats:title>Abstract</jats:title><jats:p>Expanding access to precision medicine will increasingly require that patient biometrics can be measured in remote care settings. VO<jats:sub>2</jats:sub>max, the maximum volume of oxygen usable during intense exercise, is one of the most predictive biometric risk factors for cardiovascular disease, frailty, and overall mortality.<jats:sup>1,2</jats:sup> However, VO<jats:sub>2</jats:sub>max measurements are rarely performed in clinical care or large-scale epidemiologic studies due to the high cost, participant burden, and need for specialized laboratory equipment and staff.<jats:sup>3,4</jats:sup> To overcome these barriers, we developed two smartphone sensor-based protocols for estimating VO<jats:sub>2</jats:sub>max: a generalization of a 12-minute run test (12-MRT) and a submaximal 3-minute step test (3-MST). In laboratory settings, Lins concordance for these two tests relative to gold standard VO<jats:sub>2</jats:sub>max testing was <jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.66 for 12-MRT and <jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.61 for 3-MST. Relative to “silver standards”<jats:sup>5</jats:sup> (Cooper/Tecumseh protocols), concordance was <jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.96 and <jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.94, respectively. However, in remote settings, 12-MRT was significantly less concordant with gold standard (<jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.25) compared to 3-MST (<jats:italic>p</jats:italic><jats:sub>c</jats:sub>=0.61), though both had high test-retest reliability (ICC=0.88 and 0.86, respectively). These results demonstrate the importance of real-world evidence for validation of digital health measurements. In order to validate 3-MST in a broadly representative population in accordance with the <jats:italic>All of Us</jats:italic> Research Program<jats:sup>6</jats:sup> for which this measurement was developed, the camera-based heart rate measurement was investigated for potential bias. No systematic measurement error was observed that corresponded to skin pigmentation level, operating system, or cost of the phone used. The smartphone-based 3-MST protocol, here termed <jats:italic>Heart Snapshot</jats:italic>, maintained fidelity across demographic variation in age and sex, across diverse skin pigmentation, and between iOS and Android implementations of various smartphone models. The source code for these smartphone measurements, along with the data used to validate them,<jats:sup>6</jats:sup> are openly available to the research community.</jats:p>

Prospective validation of smartphone-based heart rate and respiratory rate measurement algorithms

“…51 Monitoring one’s HR is also an easy and effective way to assess and adjust exercise intensity, or enable smartphone-based measurement of cardiorespiratory fitness. 52–54…”

Prospective Validation of Smartphone-based Heart Rate and Respiratory Rate Measurement Algorithms

“…35 Monitoring one's HR is also an easy and effective way to assess and adjust exercise intensity, or enable smartphone-based measurement of cardiorespiratory tness. [36][37][38] With further clinical validation across broad populations, such smartphone-based measurement could also be useful in various settings, most notably telehealth where vital sign measurement is challenging due to the remote nature of the patient encounter. 39,40 Though patients can in principle count their own HR or RR, this can be error prone due to factors such as biases that acute awareness of the selfexamination can cause.…”