Improving Public Health Decision-Making Through Alternative Data Sources

Subject: Alternative Data for Public Health Decision-Making

Researcher(s): Khahlil Louisy, Alexander Radunsky, Betsy Gardner, Elizabeth Leigh Zillioux, Tom Joe Dominic, Jynessa June King-Garcia, Iman Abdi, Ahmed Adam El-Halabi, and Mary Margaret Gilbert.

Description:

This initiative formalizes the use of “alternative data” as decision infrastructure for public health when official reporting is delayed, fragmented, or constrained. Building on the Data-Smart series’ analysis of the barriers to accessing, sharing, and evaluating health data, we position nontraditional signals as complements to clinical sources and specify the conditions under which they add decision value for city practitioners.

Our research program advances four strands. First, we develop a typology that links data classes to concrete use cases: environmental sensor networks for exposure assessment; mobility traces for contact patterns and service reach; pharmacy and retail transactions for demand-side syndromic signals; 311 and digital service logs for place-based anomalies. Second, we specify validity, bias, and drift tests for each class and design privacy-preserving fusion pipelines that reconcile heterogeneous sampling frames while protecting individuals and neighborhoods. Third, we create decision models that resolve risk at fine spatial and temporal scales, with an explicit goal of neighborhood and, where feasible, street-level situational awareness suitable for targeted intervention. Fourth, we codify governance patterns: data contracts, purpose limitation, auditable access, and fit-for-purpose consent, that make cross-sector use of these data operational for resource-constrained agencies. These strands translate the broad recommendations in the series into deployable practice.

Evaluation is integral. We measure contribution to decision quality rather than model accuracy alone, using counterfactual and quasi-experimental designs that trace how alerts, forecasts, or risk maps change actions and outcomes. Metrics include lead time gained over official surveillance, precision of hotspot localization, robustness across neighborhood demographics, and downstream effects on service deployment and health burden. The output is a tested playbook and open methods that cities can reproduce within the Community Data Health Initiative ecosystem, creating a durable pathway from heterogeneous signals to decision-grade public health intelligence.