Open Collections

Yu, Maggie Miao

University of British Columbia

Background: Hospital readmissions significantly impact patient outcomes and healthcare costs globally. While prediction models are widely used to identify patients at high risk of readmissions, their discriminative ability often falls short of achieving adequate accuracy. Integrating patient-reported outcome measures (PROMs) may enhance these models by considering patient perspectives on their own health conditions. Objective: To investigate the contribution of PROMs to the performance of prediction models for hospital readmissions by (1) systematically reviewing the literature on readmission prediction models, (2) assessing the incremental value of incorporating PROMs on the performance of risk prediction models, and (3) exploring extra predictive value of PROMs across various clinical and geographic subgroups. Methods: A systematic review of readmission risk prediction models that included PROMs was conducted through searching electronic databases. A retrospective population-based cohort study was performed using data from the British Columbia Acute Inpatient Survey linked to administrative healthcare databases. The study cohort comprised patients discharged from acute care facilities between September 2016 and March 2017 who had completed PROMs. Subgroups were stratified by the presence of ambulatory care sensitive conditions and distance from major hospitals relative to patients’ residences. Prediction models were constructed using both statistical methods and machine learning techniques. Model performance was compared with and without PROMs data for the overall cohort and subgroups. Results: The systematic review highlighted the potential of PROMs in predicting hospital readmission, although the effectiveness varied depending on how PROMs data were collected and used in the prediction models. The analysis included 9,148 patients discharged from BC’s acute care hospitals, with PROMs completed between 26- and 60-days following discharge. While global summary of PROMs were influential predictors, the inclusion of PROM data led to only modest improvement in model performance. Subgroup analyses indicated the need for adjusting risk thresholds to accommodate specific clinical and demographic characteristics. Conclusion: Incorporating PROMs into prediction models for unplanned hospital readmission or death within one year modestly enhanced model performance. However, the marginal gains in predictive accuracy must be balanced against the benefits to patients and the costs associated with collecting these data.

Text

2025-06-30

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/88551

Population and Public Health

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/