A Prognostic Model for 6-Month Mortality in Elderly Survivors of Critical Illness

Abstract

PURPOSE: Nearly 500,000 elderly ( 65) intensive care unit (ICU) survivors are discharged to skilled-care annually. Almost half are re-hospitalized and up to 65% die within 6 months. We developed and validated a prognostic index for 6-month mortality for elderly ICU survivors after discharge to skilled-care. METHODS: We examined charts and claims data from 871 consecutive patients aged 65 who had their first medical-ICU admission lasting >24 hours between 2006 and 2009, and were discharged to skilled-care (excluding hospice). We used the Single Level Clinical Classification software to group primary ICD-9 discharge diagnoses, and calculated a Charlson Co-morbidity Index (CCI) for each patient based on secondary ICD-9 diagnoses present on admission. Mechanical ventilation, cardiac arrest, and hemodialysis were identified by ICD-9 codes. A logistic regression model was derived from data on patients admitted from 2006- 2007, and then validated on those admitted from 2008-2009. RESULTS: Mean age was 80 (SD 8.4) and 79 (SD 8.5) years, and the 6-month mortality was 38.7% and 39.7% in the development and validation sets, respectively. Independent predictors in the multivariate model included admission from skilled-care, ICU length-of-stay, older age, higher CCI, and sepsis. A CCI 8 was the strongest independent predictor for the entire cohort (OR 10.7, 95%CI 4.3 to 26.9), and when patients with cancer were excluded (OR 11.1, 95%CI 1.9 to 64.0). The area under the receiver operating characteristic curve was 0.70 for both the derivation and validation cohorts, and the Hosmer-Lemeshow Goodness-of-Fit test p = 0.75 and 0.45 for the derivation and validation cohorts. CONCLUSIONS: Clinical variables available at hospital discharge can predict death within 6-months for elderly ICU survivors discharged to skilled-care. CLINICAL IMPLICATIONS: This prognostic model may help physicians communicate prognoses to patients and their surrogates, and help risk-adjust populations of elderly frail patients for studies that test palliative and therapeutic interventions. Future research should focus on external validation of the model and identification of other physical and biological markers of frailty at hospital discharge that may enhance model prediction.