Sleep-study AI adds to risk ranking in two hospital datasets

162 words · 1 min

A sleep study already records the heart’s electrical activity.

Atrial-fibrillation risk ranking

Emory

C-index

Clinical information0.70

95% CI 0.69–0.72

Clinical + AI score0.75

95% CI 0.73–0.77

Beth Israel Deaconess

C-index

Clinical information0.77

95% CI 0.75–0.79

Clinical + AI score0.81

95% CI 0.79–0.84

One retrospective study. Confidence intervals are 95%; C-index is risk/event ordering, not accuracy, risk reduction or patient benefit. Outcome-specific exclusions apply.

An October 9 paper asks whether that signal, combined with expert-labeled sleep stages, can also help rank later cardiovascular risk. The model was fine-tuned on 15,809 Massachusetts General patients and evaluated in separate hospital cohorts beginning with 9,810 Emory and 12,576 Beth Israel Deaconess patients. Outcome-specific exclusions apply. [1][2]

For atrial fibrillation, Table 5 reports a concordance index rising from 0.70 to 0.75 at Emory and from 0.77 to 0.81 at Beth Israel Deaconess when the AI score is added. This measures the ordering of risk and event timing, not percentage accuracy or an absolute reduction in illness. [1]

These were retrospective sleep-clinic records, with diagnoses drawn from health records. Validation at another hospital is useful, but the same research team performed it. The study does not show that using the score improves treatment or health outcomes. For research teams, the next useful test is prospective: does the extra information change decisions and help patients? [1][2]

Sources & details 2 sources

Discuss in your voice app

Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.

Open ChatGPT

Paste the brief into your voice app to discuss this reporting.

Reporting sources

Koscova et al. · SLEEP, October 9
Original peer-reviewed paper. Methods, outcome exclusions and Table 5 checked. Reported statistical results have not been independently reproduced.
NIH · October 9 research summary
Funder’s account confirms starting cohort sizes and distinguishes stronger outcomes from stroke/MI. It is not an independent replication or an endorsement for individual screening.

Images & figures

Atrial fibrillation only. Points show concordance; intervals show 95% confidence bounds.
Euangelion · original explanatory figure · Source · Usage terms · Download figure