Developing Models for Complicated Outcomes in Acutely Intoxicated ICU Patients With and Without Toxicity Information

To assess the prognostic value of the mean Lethality Risk Score (LRS), reflecting the toxicity of substances involved in an exposure, and to develop prediction models for complicated outcomes in ICU patients with acute intoxication.

Multicenter prospective cohort study.

Seventy-eight ICUs in Europe, Australia, and the Eastern Mediterranean (2020-2023).

A total of 2265 adult patients were admitted to the ICU with acute intoxication as the main reason for admission for at least 4 hours.

None.

The primary outcome was defined as ICU complications or death (22.5%). Clinical data were collected in an electronic capture system. We developed two prediction models: one incorporating laboratory variables ("Laboratory Model"), and one including the mean LRS ("Expo Model"). The mean LRS was calculated as the average of the individual LRS values assigned to the toxicants involved in the exposure. Key predictors were age, severity at presentation, systolic blood pressure, Glasgow Coma Scale score, and hospital type. The mean LRS remained significant only when laboratory variables were not considered. Model performance was assessed using the area under the receiver operating characteristic (AUROC), calibration plots, likelihood ratio tests, and decision curve analysis. Internal validation demonstrated good performance for both models (AUROCs of 0.79 and 0.77; calibration slopes of 0.94 and 0.95; intercepts of -0.06 and -0.06), with minimal overfitting. Decision curve analysis revealed clinically relevant threshold probability ranges of 0-50% for the Laboratory Model and 0-40% for the Expo Model.

The mean LRS had prognostic value only in the absence of laboratory test results. Both models may aid decision-making regarding the management of acutely intoxicated patients in the ICU.