Health Apps Work. The Evidence Still Wobbles.

A major umbrella review in The Lancet Digital Health synthesised 78 systematic reviews, 496 primary studies and 177,373 participants to assess app-based interventions for non-communicable diseases.

The overall signal was positive. App-based interventions significantly improved outcomes including blood pressure, glucose control, bodyweight, BMI, waist circumference, anxiety, depression, stress, physical activity, sedentary time, medication adherence and quality of life.

But statistical significance is not the same as evidentiary certainty. Of the 78 reviews, 57—73%—were rated critically low quality. Another 17 were low quality, three were moderate quality and only one was high quality.

Across 31 outcomes, five had highly suggestive evidence, eight suggestive evidence, eleven weak evidence and seven non-significant evidence. None reached the highest “convincing” evidence category. Heterogeneity was also very high across many principal outcomes.

My takeaway: we should stop asking whether “health apps work” as though they were one homogeneous intervention. Evidence for a broad category cannot substitute for validation of a specific product, version, population, comparator and outcome.

Integration into healthcare should therefore be evidence-stratified. Implementation and reimbursement require clinically meaningful effects, independent product-specific replication, transparent attrition reporting and evidence of durability, safety, equity and real-world engagement. Continuous evaluation is essential when the software itself changes.