BMC Geriatr. 2024 Oct 2;24(1):806. doi: 10.1186/s12877-024-05409-z.
ABSTRACT
BACKGROUND: The amount of prediction models for disability in older adults is increasing but the prediction performance of different models varies greatly, and the quality of prediction models is still unclear.
OBJECTIVES: To systematically review and critically appraise the studies on risk prediction models for disability in older adults.
METHODS: A systematic literature search was conducted on PubMed, Embase, Web of Science, Cochrane Library, Cumulative Index to Nursing and Allied Health Literature (CINAHL), China National Knowledge Infrastructure (CNKI), China Science and Technology Journal Database (VIP), and Wanfang Database, published up until June 30, 2023. Data were extracted according to the Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS). The Prediction Model Risk of Bias Assessment Tool (PROBAST) was used to assess the risk of bias and applicability of the included studies. In addition, all included studies were evaluated for clinical value.
RESULTS: A total of 5722 articles were initially retrieved from databases, 16 studies and 17 prediction models were finally included after screening. The sample sizes of studies ranged from 420 to 90,889. Model development methods mainly included logistic regression analysis, Cox proportional hazards regression, and machine learning methods. The C statistic or area under the curve (AUC) of models ranged from 0.650 to 0.853, and nine models had C statistic/AUC higher than 0.75. Age, chronic disease, gender, self-rated health, body mass index (BMI), drinking, smoking and education level were the most common predictors. According to the PROBAST, all included studies were at high risk of bias, and 10 studies were at high concerns for applicability. Only two studies reported following the Transparent Reporting of a Multivariate Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement. After evaluation, only two models reached the standard of clinical value.
CONCLUSION: Although most of the included prediction models had acceptable discrimination, the overall quality and clinical value of the current studies were poor. In the future, researchers should follow the TRIPOD statement and PROBAST checklist to develop prediction models with larger sample sizes, more reasonable study designs, and more scientific analysis methods, to improve the predictive performance and application value.
TRIAL REGISTRATION: The review protocol was registered in PROSPERO (registration ID: CRD42023446657).
PMID:39358747 | DOI:10.1186/s12877-024-05409-z