Roughly one in five hospital patients returns within thirty days, and the industry has largely treated that as a prediction problem. It is better understood as a scheduling problem. The signals that separate a stable discharge from a fragile one are usually present in the chart, the claims history, and the social context well before the patient leaves the building.
Knowable is not the same as known in time
A risk stratification report that lands in a monthly population health review is an audit, not an intervention. By the time it is read, the discharge decisions it describes are weeks old and the readmissions it predicted have already happened. The analysis was correct and useless in the same breath.
The window that matters is narrow: the hours between the decision to discharge and the patient walking out. That is when a follow-up appointment can still be booked, a home health referral can still be arranged, and a medication reconciliation can still change what the patient does at home on day three.
What arriving on time actually requires
It requires the score to be computed continuously rather than in batch, surfaced where the discharge planner already works rather than in a separate portal, and attached to a recommended next action rather than left as a number for someone else to interpret.
It also requires the list to be short. A worklist that flags a third of the census is not prioritization. Care teams have a fixed number of interventions available per shift, and intelligence that ignores that constraint gets ignored back.
The economics follow the timing
For ACOs and value-based care networks, avoidable readmissions are a direct charge against shared savings. The financial return does not come from identifying risk. It comes from the intervention that risk triggered, which means every hour between signal and action is measurable margin.
Framed that way, the question to ask any predictive vendor is not how accurate the model is. It is when the output arrives, who sees it, and what they are expected to do in the next sixty seconds.
About the author
Preventra is Editorial of Preventra, a predictive care intelligence platform for hospitals, value-based care organizations, and health plans.