Patient world model forecasts health campaign outcomes with low error
A Patient World Model for Early Forecasting of Digital Health Campaign Outcomes: Capabilities and Limits
Machine LearningArtificial Intelligence
Summary
Measuring the impact of health campaigns after they finish is common, but predicting their success while they are running is harder. The authors built a model that tracks each patient’s state over time and uses it to forecast future prescription volumes based on observed exposures. This model is much more accurate than existing methods in predicting campaign results weeks or months in advance. They also showed how turning off exposures in simulations highlights the challenges of interpreting cause and effect in real-world data.
What this means in practice
- •For digital marketing teams: Predict future prescription volumes early during health campaigns to better allocate resources and adjust messaging strategies.
- •For healthcare analytics teams: Model patient response dynamics to simulate potential outcomes of altering exposure strategies in promotional campaigns.
Tested on one dataset.
Authors
Yunlong Wang
Abstract
Digital direct-to-consumer (DTC) health campaigns are usually measured after the fact. In-flight forecasting commonly relies on a separate classifier for every cutoff and horizon. We treat this task as a dynamic-system problem and build a compact patient world model. The architecture maintains a latent state per patient, learns exposure-conditioned state dynamics jointly with a weekly conversion hazard, and rolls forward into future conversion curves. We evaluate it on a US campaign dataset with 147{,}173 patients and 5.2 million at-risk person-weeks. In a retrospective evaluation conditioned on recorded future exposures, the model forecasts the remaining new-to-brand prescription volume through week 52 with a relative error of 2.9\% from a week-4 cutoff and 0.8--2.6\% from cutoffs at weeks 8--26. The strongest non-recurrent baseline, a pooled-hazard gradient boosting model given the same survival rollout and information, has relative errors of 13.6--33.1\%. Per-horizon classifiers perform substantially worse. A Fisher-information analysis motivates dense next-exposure supervision when conversions are rare. Removing this auxiliary objective increases prescription-volume error by approximately $2$--$14\times$, while providing no consistent disadvantage on the more common specialist-visit outcome. We also evaluate scenario simulation. Switching all future exposure off raises predicted conversion from 0.31 to 0.89, a pattern consistent with selection effects in observational exposure data. This result highlights the limits of interpreting exposure-conditioned rollouts causally.