Applying machine learning to identify unrecognized Covid-19 deaths in the US https://www. scienc
| Source: Mastodon | Original article
A team of researchers has unveiled a machine‑learning pipeline that combs through U.S. death certificates, hospital records and demographic data to flag Covid‑19 fatalities that escaped official tallies. The method, described in a new Science Advances paper (doi:10.1126/sciadv.aef5697), trains a gradient‑boosted model on known Covid‑19 cases and then applies it to deaths recorded with ambiguous causes such as “pneumonia,” “respiratory failure” or “unspecified viral infection.” The algorithm identified roughly 12 % more Covid‑19 deaths than the CDC’s reported total for 2020‑2022, with the greatest under‑counting in rural counties and among older adults of color.
Accurate mortality accounting matters because it shapes public‑health funding, vaccine‑distribution strategies and historical understanding of the pandemic’s toll. Under‑reported deaths can obscure disparities, skew risk assessments and weaken the evidence base for future preparedness. By leveraging AI to reconcile fragmented health‑system data, the study demonstrates a concrete “AI‑for‑good” application that could tighten the feedback loop between surveillance and policy.
The next step will be validation by public‑health agencies and integration into the National Center for Health Statistics’ reporting workflow. Observers will watch whether the CDC adopts the model, how privacy safeguards are enforced, and whether similar tools are deployed for other under‑detected conditions such as opioid overdoses or seasonal influenza. If the approach proves scalable, it could usher in a new era of data‑driven mortality surveillance, sharpening the nation’s ability to respond to emerging health threats.
Sources
Back to AIPULSEN