Smartphone camera apps can detect AF, but false positives a concern
A systematic review and meta-analysis of four smartphone camera applications that detect atrial fibrillation reported high sensitivity and specificity of the apps for diagnosing AF, but a modest predictive value, which researchers concluded could generate a high number of false-positive results.
The analysis focused on 10 primary diagnostic accuracy studies (n = 3,582 participants) of four smartphone camera apps: Preventicus, PULSE-SMART, Cardiio Rhythm (Cardiio Inc.) and FibriCheck Qompium). These applications obtain a photoplethysmogram signal from a fingertip pulse via the user’s smartphone camera to detect AF. The researchers measured sensitivity and specificity with bivariate random-effects meta-analysis and modeled positive and negative predictive values for different population groups.
For all apps combined, sensitivity was 94.2% (95% CI, 92.2%-95.7%) and specificity was 95.8% (95% CI, 92.4%-97.7%).
Additionally, the researchers reported high sensitivity and specificity of each individual smartphone app.
For the detection of AF in an older asymptomatic population, the positive predictive value for all apps combined was between 19.3% (95% CI, 19.2%-19.4%) and 37.5% (95% CI, 37.4%-37.6%), while the negative predictive value was between 99.8% (95% CI, 99.83%-99.84%) and 99.9% (95% CI, 99.94%-99.95%), according to the study abstract.
“Our results have provided both users and health care professionals who care for these patients the diagnostic accuracy of these applications collectively and individually for the commercially available applications,” Jack O'Sullivan, MBBS, DPhil, postdoctoral research fellow in cardiovascular medicine at Stanford University, and colleagues wrote. “We believe it is plausible that patients with an established diagnosis of paroxysmal AF may use these smartphone camera applications to monitor for relapses of AF (ie, monitoring device). Patients with paroxysmal AF alternate between sinus rhythm and AF. When these patients develop symptoms suggestive of AF, many are instructed to take their pulse and try to self-diagnose a relapse of AF. A confirmed relapse of AF may lead some patients to take rate-controlling medication or to seek medical attention.”
For AF detection in individuals aged 65 years and older with hypertension, the positive predictive value was 20.5% (95% CI, 20.4%-20.6%) to 39.2% (95% CI, 39.1%-39.3%) and the negative predictive value was 99.8% (95% CI, 99.8%-99.8%) to 99.9% (95% CI, 99.9%-99.9%), according to the results.
“Regarding AF screening, if any of these applications reach a diagnosis of sinus rhythm in a healthy, asymptomatic person, it is likely this person does not have AF (ie, true-negative),” the researchers wrote. “Conversely, we cannot draw the same conclusion from a positive result. In fact, our model suggests that if these applications detect AF in an asymptomatic person, the result is most likely to be a false-positive. From these results, it would appear premature to use these devices among healthy individuals or to use them to screen an asymptomatic population, including those older than 65 years with hypertension.”
This study included primary diagnostic accuracy trials that assessed the accuracy of smartphone applications that use a smartphone camera for the diagnosis of AF. Researchers defined smartphone applications as capable of AF diagnosis if it had the ability to produce at least a single-lead ECG and analyze a heart rhythm tracing.
“Our results do not directly address another proposed use of these applications, which is chronic disease management. Patients with AF often alternate between sinus rhythm and AF; the ability for patients to self-diagnose when they are in AF may enable them to seek (or not seek) medical care as appropriate,” the researchers wrote. “Although we were not able to formally model the PPV for a population monitoring AF, it is plausible that it would be higher than what we observed for the screening population (given that those with paroxysmal AF have a far greater pretest probability than an asymptomatic population that is being screened for AF). Future research addressing this unknown would be advantageous.” – by Scott Buzby
Disclosures: O’Sullivan reports no relevant financial disclosures. Please see the study for all other author’s relevant financial disclosures.