Real-World Outcome Measurement for Prescription Digital Therapeutics

Prescription digital therapeutics (PDTs) — software applications that deliver evidence-based behavioral interventions like cognitive behavioral therapy — have achieved FDA clearance but cannot generate the real-world outcome data that insurance companies require for reimbursement. Pear Therapeutics, which held the first-ever FDA de novo clearance for a digital therapeutic and had clinical trial data supporting its substance use disorder app, went bankrupt in 2023 because payers consistently denied coverage. The gap between regulatory approval (which evaluates efficacy in trials) and payer reimbursement (which demands evidence of cost-effectiveness in practice) has no established bridge for software-based interventions.

Substance use disorders affect over 46 million Americans, and fewer than 10% receive any treatment. Mental health conditions account for roughly $280 billion in annual US healthcare costs. Digital therapeutics could deliver evidence-based interventions at a fraction of the cost of in-person therapy and without the workforce shortage constraints. Pear's apps were prescribed by doctors and used by patients — the clinical pathway worked. But without reimbursement, the business model collapses. This problem extends beyond Pear: nearly every standalone digital therapeutics company (including Akili Interactive, which also shut down in 2024) has faced the same barrier. The underlying challenge is that payers don't know how to evaluate whether a software intervention is working in the real world.

Pear Therapeutics conducted randomized controlled trials showing its reSET app improved substance use disorder treatment retention. It achieved FDA clearance for three products. But insurance payers require a different evidence standard: proof of cost savings in a real-world population over time, not just efficacy in a controlled trial. Traditional drugs demonstrate real-world effectiveness through pharmacy claims data and medical outcomes tracked over years. PDTs have no equivalent measurement infrastructure — there's no standard way to track whether a patient engaged with the software, whether their condition improved, and whether the intervention reduced downstream healthcare utilization. Pear tried to generate this evidence through hospital partnerships, but the data collection was fragmented across health systems with different EHR platforms, different outcome metrics, and no standardized reporting. The result: plenty of anecdotal evidence but nothing systematic enough to satisfy an insurance actuary.

Two things would help: (1) a standardized, automated outcome measurement framework specifically designed for digital therapeutics — one that can passively collect engagement, symptom, and utilization data across diverse health systems and link it to insurance claims outcomes; (2) a cost-effectiveness modeling approach that payers accept as valid for software interventions, analogous to the QALY frameworks used for drugs. Adjacent fields with relevant approaches include remote patient monitoring (which has solved some data integration challenges), real-world evidence generation for pharmaceuticals (which has developed claims-based outcome tracking), and adaptive clinical trial designs (which allow evidence to accumulate during deployment).

A student team could: (1) design a prototype data pipeline that integrates PDT engagement data (app usage, session completion, in-app assessments) with EHR outcome data (diagnosis codes, emergency visits, hospitalizations) and insurance claims data, demonstrating feasibility with synthetic or de-identified data; (2) develop a standardized outcome metric framework for a specific condition (e.g., substance use disorder) that maps PDT engagement patterns to clinically meaningful outcomes payers would accept; (3) build a cost-effectiveness simulation model that estimates the insurance cost impact of a digital therapeutic compared to standard care, identifying what data inputs are needed to make the model credible. Relevant disciplines include health informatics, data science, health economics, and software engineering.