This systematic review evaluates 26 validated models for predicting hospital readmission risk, focusing on their performance and suitability for clinical or administrative use. The studies were identified through searches of MEDLINE, CINAHL, Cochrane Library, EMBASE, and reference lists. The models were tested in various settings and populations, with a focus on medical patients. The most common outcome used was 30-day readmission, and only one model specifically addressed preventable readmissions. Fourteen models based on retrospective administrative data could potentially be used for standardizing readmission rates and hospital comparisons. However, nine of these models had poor discriminative ability (c-statistics 0.55–0.65). Seven models could identify high-risk patients for early intervention (c-statistics 0.56–0.72), and five could be used at hospital discharge (c-statistics 0.68–0.83). Six studies compared different models within the same population, and two found that including functional and social variables improved model discrimination. The review highlights that most current readmission risk prediction models, whether designed for comparative or clinical purposes, perform poorly. While some models may be useful in certain settings, efforts to improve their performance are needed as their use becomes more widespread. The review also emphasizes the need for models that incorporate broader social, environmental, and medical factors, as well as hospital-level factors, to better understand and predict readmission risk.This systematic review evaluates 26 validated models for predicting hospital readmission risk, focusing on their performance and suitability for clinical or administrative use. The studies were identified through searches of MEDLINE, CINAHL, Cochrane Library, EMBASE, and reference lists. The models were tested in various settings and populations, with a focus on medical patients. The most common outcome used was 30-day readmission, and only one model specifically addressed preventable readmissions. Fourteen models based on retrospective administrative data could potentially be used for standardizing readmission rates and hospital comparisons. However, nine of these models had poor discriminative ability (c-statistics 0.55–0.65). Seven models could identify high-risk patients for early intervention (c-statistics 0.56–0.72), and five could be used at hospital discharge (c-statistics 0.68–0.83). Six studies compared different models within the same population, and two found that including functional and social variables improved model discrimination. The review highlights that most current readmission risk prediction models, whether designed for comparative or clinical purposes, perform poorly. While some models may be useful in certain settings, efforts to improve their performance are needed as their use becomes more widespread. The review also emphasizes the need for models that incorporate broader social, environmental, and medical factors, as well as hospital-level factors, to better understand and predict readmission risk.