Abstract 18756: Association of Morphologic Variability with 6-Month Mortality in a General Cohort of Intensive Care Unit Patients: Results from the MIMICII Database

Abstract

Background: Morphologic variability (MV) has been previously shown to identify patients at increased risk of mortality following acute coronary syndrome in retrospective analyses of randomized controlled clinical trial data. We assessed the relationship of MV and mortality in a general cohort of intensive care unit (ICU) patients.

Methods: MV was calculated for ICU patients within the MIMICII database. Data from all ICU admissions with at least 10 hours of continuous ECG sampled at 9-bit quantization or greater during the first 24 hours were used (n=359). Patients were dichotomized into risk groups at the highest quartile. The primary endpoint was in-hospital or 6-month post-discharge mortality previously determined in MIMICII from the Social Security Death Index (n=83).

Results: MV was associated with mortality on univariate analysis and in multivariate models adjusted for the Simplified Acute Physiology Score (SAPS) and the Sequential Organ Failure Assessment (SOFA) Score (Table). These results were consistent when the multivariate models were further adjusted for the Elixhauser comorbidity scores of cardiac arrhythmia and congestive heart failure (adjusted hazard ratio 2.46; 95% CI 1.30 - 4.66; p=0.005). The model constructed for the overall cohort using SAPS and SOFA showed improved discrimination and reclassification when MV was added (IDI 0.022, p=0.027; NRI 0.349, p=0.008).

Conclusions: MV has value in risk stratification of the general cohort of ICU patients and provides complementary information to existing ICU scoring systems.