Clinical prediction models for the early diagnosis of obstructive sleep apnea in stroke patients : a systematic review

Abstract

Background: Obstructive sleep apnea (OSA) is a common sleep disorder characterized by repetitive cessation or reduction in airflow during sleep. Stroke patients have a higher risk of OSA, which can worsen their cognitive and functional disabilities, prolong their hospitalization, and increase their mortality rates. Methods: We conducted a comprehensive literature search in the databases of PubMed, CINAHL, Embase, PsycINFO, Cochrane Library, and CNKI, using a combination of keywords and MeSH words in both English and Chinese. Studies published up to March 1, 2022, which reported the development and/or validation of clinical prediction models for OSA diagnosis in stroke patients. Results: We identified 11 studies that met our inclusion criteria. Most of the studies used logistic regression models and machine learning approaches to predict the incidence of OSA in stroke patients. The most frequently selected predictors included body mass index, sex, neck circumference, snoring, and blood pressure. However, the predictive performance of these models ranged from poor to moderate, with the area under the receiver operating characteristic curve varying from 0.55 to 0.82. All the studies have a high overall risk of bias, mainly due to the small sample size and lack of external validation. Conclusion: Although clinical prediction models have shown the potential for diagnosing OSA in stroke patients, their limited accuracy and high risk of bias restrict their implications. Future studies should focus on developing advanced algorithms that incorporate more predictors from larger and representative samples and externally validating their performance to enhance their clinical applicability and accuracy.