Review of Application of Machine Learning as a Screening Tool for Diagnosis of Obstructive Sleep Apnea.
Level 5 - mechanism / opinion, no new human data
Narrative review without systematic search or original human data
PubMed 36363530 · doi:10.3390/medicina58111574
What was done
Narrative review examining the application of machine learning models that use demographic data, standard clinical variables, and physical findings to screen and triage patients at high risk for obstructive sleep apnea syndrome (OSAS).
What was found
The abstract reports no quantitative performance metrics, numbers, or specific accuracy figures for the machine learning models. It notes that 80% to 90% of middle-aged adults with moderate-to-severe OSAS remain undiagnosed under current screening paradigms, and frames clinical ML models as an emerging triage strategy.
Why it matters
Polysomnography is resource-intensive and impractical for mass screening. Evaluating machine learning approaches that rely on routine clinical features could help identify scalable, low-cost methods for triaging patients who require formal sleep diagnostic testing.
Limits
As a narrative review, the abstract lacks a formal systematic search methodology, risk of bias assessment, study count, and quantitative outcome measures (such as sensitivity, specificity, or AUC) needed to evaluate model performance.
Cited by
- supports Between 80% and 90% of people with obstructive sleep apnea remain undiagnosed.