593

Abstract

Introduction: ICU discharge has remained a subjective decision and providers continue to have difficulty developing an objective mechanism predicting post discharge safety or adverse outcomes. The purpose of this study is to prospectively determine the performance of a DRS designed to predict post-ICU discharge death and readmission in a 450 bed tele-ICU program. Methods: All pts in 25 ICUs from 13 hospitals and 4 states, from 1/1/2013 to 1/28/2013 had their risk of post-ICU death and readmission determined (without clinical intervention) using a previously described DRS as part of 24/7 tele-ICU surveillance system (Philips). Based on the DRS, patients were classified as either: high, medium, or low risk of death and readmission within 48 hours, per predefined thresholds. Adverse outcomes (post-discharge deaths and/or readmissions within 48 hrs) were compared to the DRS. Statistical analysis of DRS performance was based on ROC curves, positive (PPV) and negative (NPV) predictive values, sensitivity, and specificity. Results: 1,866 pts with a mean (SD) age of 63.5 (16.5) yrs; 51% male; 79% medical, surgical or mixed ICU patients. None of the 455 pts classified as low risk had an adverse outcome. 1,021 pts were moderate risk with 20 (2%) having an adverse outcome and 123 pts were high risk with 25 (20%) having an adverse outcome. There were 4 adverse outcomes among 267 pts who were not scored at DC due to missing data. Within 48 hrs of ICU DC, there were 13 readmissions (0.7%) and 37 deaths (2.0%). 32 of the 37 deaths had an active 'comfort measures only' (CMO) order at time of DC. Of the 13 readmissions, none were classified as low risk and 4 were high risk. Of the 37 deaths, none were classified as low risk while 21 were high risk. The ROC predicting the combined endpoint was 0.82 (0.75-0.90) and were similar when excluding the 32 CMO-related deaths (0.81; 0.69-0.93). The sensitivity, specificity, PPV and NPV for the high risk threshold were 51%, 95%, 20% and 99% respectively; low risk threshold were 100%, 25%, 3.5%, and 99% respectively. Conclusions: This DRS allows facilities to have evidence based confidence that low risk patients can be safely transferred out of the ICU. None of the patients discharged having adverse outcomes were low risk. Less than 7% of discharges were identified as high risk with 1 in 5 of these dying or being readmitted within the subsequent 48 hrs. Larger numbers of patients are needed in order to validate that the above statistical outcomes are maintained.