A Target for Increased Mortality Risk in Critically Ill Patients: The Concept of Perpetuity

Abstract

Background: Emergency medicine is acuity-based and focuses on time-sensitive treatments for life-threatening diseases. Prolonged time in the emergency department, however, is associated with higher mortality in critically ill patients. Thus, we explored management after an acuity-based intervention, which we call perpetuity, as a potential mechanism for increased risk. To explore this concept, we evaluated the impact of each hour above a lung-protective tidal volume on risk of mortality.

Methods: This cohort analysis includes all critically ill, non-trauma, adult patients admitted to two academic EDs between 1 November 2013 and 30 April 2017. Cox models with time-varying covariates were developed with time in perpetuity as a time-varying covariate, defined as hours above 8 mL/kg ideal body weight, adjusted for covariates. The primary outcome was the time to in-hospital death.

Results: Our analysis included 2025 patients, 321 (16%) of whom had at least 1 h of perpetuity time. A partial likelihood-ratio test comparing models with and without hours in perpetuity was statistically significant (χ²(3) = 13.83, p = 0.0031). There was an interaction between age and perpetuity (Relative risk (RR) 0.9995; 95% Confidence interval (CI₉₅): 0.9991-0.9998). For example, for each hour above 8 mL/kg ideal body weight, a 20-year-old with 90% oxygen saturation has a relative risk of death of 1.02, but a 40-year-old with 90% oxygen saturation has a relative risk of 1.01.

Conclusions: Perpetuity, illustrated through the lens of mechanical ventilation, may represent a target for improving outcomes in critically ill patients, starting in the emergency department. Research is needed to evaluate the types of patients and interventions in which perpetuity plays a role.

Keywords: acuity; critical care; critically ill; emergency department; intubation; mechanical ventilation; perpetuity.

Figure A1

The effect of hours in acuity and heart rate are complex and depend upon the starting number of acuity hours or starting heart rate. The association is illustrated below for the 8 h model.

Figure A2

For the 12 h model.

Figure A3

Sensitivity of inferential results to pre-processing choices. ßEstimated coefficients on the log(hazard rate) scale with 95% confidence intervals and partial likelihood ratio test χ² test statistics for each model are shown. Plotted values are colored based on their associated p-value.

Figure A4

Sensitivity of inferential results to pre-processing choices. Estimated coefficients on the log(hazard rate) scale with 95% confidence intervals for each model are shown. Plotted values are colored based on their associated p-value.

Figure A5

Sensitivity of inferential results to pre-processing choices. Estimated coefficients on the log(hazard rate) scale with 95% confidence intervals for each model are shown. Plotted values are colored based on their associated p-value.

Figure 1

Time course for a hypothetical patient.

Figure 2

Patient flow chart. Other includes: missing or clearly incorrect height, weight, systolic blood pressure, mean arterial blood pressure, questionable death status, having a documented hospital discharge before ICU discharge, and charted tidal volumes that occurred after ICU discharge.

Figure 3

8 h model: Estimated adjusted hazard ratios with 95% pointwise confidence intervals associated with a one-hour increase in perpetuity for various ages and values of O₂. Only values of O₂ present for each age are included in the plots. O₂ is triage oxygen saturation.

Figure 4

12 h model: estimated adjusted hazard ratios with 95% pointwise confidence intervals associated with a one-hour increase in perpetuity for various ages and values of O₂. Only values of O₂ present for each age are included in the plots. O₂ is triage oxygen saturation.