• Testing a 30-day readmission risk calculator in a veteran population with heart failure: A pilot study

      Gannuscio, Jacqueline R. (2012)
      index HF hospitalization between September 2007 and October 2010 were reviewed for the presence of 15demographic and clinical risk predictors, the majority of which were from a VA Readmission Risk Calculator. Additional variables specific to HF population were added to the risk calculator and included ejection fraction, substance abuse, and Black race. Binary and multiple logistic regression models were used to predict 30-day ACR. C-statistic was calculated to assess how good the model is in predicting who will be readmitted. Results. The patients studied were mostly male (98%), black (73.1%), and averaged 68 years old (SD 13.2). Of the 271 patients, 79 (29%) had at least 1 readmission; 8.1% had >1 readmission within 30 days of discharge. In bivariate logistic regression, patients with Creatinine > 2 were more than two times more likely to be readmitted (OR=2.35: 95% CI 1.32, 4.19). Patients with COPD had a similar likelihood of readmission (OR=2.36; 95% CI 1.25, 4.47), as did patients with renal failure (OR=2.41; 95% CI 1.25, 4.62). Black race, an added HF specific variable, had a significant influence on the likelihood of readmission (2.60; 95% CI 1.31, 5.16). In multivariate logistic regression with all of the predictors, only COPD (OR=2.70; 95% CI 1.32-5.52) and Black race (OR=2.07; 95% CI .97, 4.37) significantly predicted readmission. The C-statistic for the original model was .52, and improved only to .61 with the additional variables. Conclusion. The VA IPEC Readmission Risk Calculator derived in a medical-surgical population does not predict all-cause 30-day ACR after an index heart failure hospitalization. The addition of HF specific variables also did not improve the model. The study was limited by small sample size and use of a non-heart failure specific model. A future implication is that a heart-failure specific model with better C statistics could be tested and potentially be integrated into an electronic medical record so that an alert with an automated risk score could be developed and implemented. The impact of interventions based on risk assessment an open field of investigation.