Do Non-Clinical Factors Improve Prediction of Readmission Risk?: Results From the Tele-HF Study

Area of Science:

• Cardiology
• Health Services Research
• Medical Informatics

Background:

• Existing heart failure readmission risk models demonstrate limited predictive accuracy.
• The impact of incorporating patient-reported data on readmission risk prediction is not well-established.

Purpose of the Study:

• To evaluate if patient-reported socioeconomic, health status, and psychosocial factors enhance 30-day heart failure readmission risk prediction.
• To compare a comprehensive model with one using only clinical and demographic data.

Main Methods:

• A cohort of 1,004 patients hospitalized for heart failure were interviewed within two weeks of discharge.
• Data collected included clinical, demographic, socioeconomic, health status, adherence, and psychosocial variables.
• Two risk prediction models were developed: one with clinical/demographic data, and another including patient-reported information.

Main Results:

• The 30-day readmission rate was 17.1%.
• The model incorporating patient-reported data showed a modest improvement in discrimination (C-statistic 0.65) compared to the clinical/demographic model (C-statistic 0.62).
• While extending predicted readmission ranges, overall model performance remained suboptimal.

Conclusions:

• Patient-reported socioeconomic, health status, adherence, and psychosocial variables do not significantly dominate heart failure readmission prediction.
• Inclusion of patient-reported data offers marginal improvement in model discrimination but does not resolve poor predictive performance.
• Further research is needed to identify more effective predictors for heart failure readmissions.