A Data-Driven Medication Regimen Complexity Score for Critically Ill Patients: MRC-ICU 2.0.

Area of Science:

• Pharmacoeconomics and Health Outcomes Research
• Critical Care Medicine
• Health Informatics

Background:

• The original Medication Regimen Complexity-Intensive Care Unit (MRC-ICU) score (1.0) was developed heuristically and validated in a small, single-center cohort.
• The MRC-ICU score is linked to patient outcomes, ICU complications, and critical care pharmacist workload.
• There is a need for improved predictive capabilities of the MRC-ICU score using data-driven methods in diverse patient populations.

Purpose of the Study:

• To reweight the MRC-ICU score using data-driven methodologies in a large, multicenter cohort of adult ICU patients.
• To optimize the updated MRC-ICU score (version 2.0) for predicting hospital mortality, ICU fluid overload (FO), and invasive mechanical ventilation (IMV) use.
• To compare the predictive performance of the updated MRC-ICU scores (2.1 and 2.2) against the original MRC-ICU 1.0 and established severity-of-illness scores (APACHE II, SOFA).

Main Methods:

• Retrospective, observational cohort study involving 19,117 adult ICU patients from two academic health systems (2015-2023).
• Machine learning techniques, including Principal Component Analysis and Random Forest, were employed for score development and optimization.
• Two versions of the updated score were created: MRC-ICU 2.1 (predicting mortality, FO, IMV) and MRC-ICU 2.2 (predicting mortality, FO, adjusted for prolonged IMV).

Main Results:

• The updated MRC-ICU 2.0 scores demonstrated improved discrimination over MRC-ICU 1.0, with Area Under the Receiver Operating Characteristic (AUROC) increases of +0.03 to +0.08.
• MRC-ICU 2.1 and 2.2 did not consistently outperform APACHE II and SOFA in predicting mortality.
• Incorporating MRC-ICU 2.0 scores into models with APACHE II or SOFA resulted in statistically significant improvements in discrimination (AUROC increases of +0.01 to +0.13).

Conclusions:

• The updated MRC-ICU 2.0 score consistently improved discrimination compared to MRC-ICU 1.0 across various outcomes and datasets.
• MRC-ICU 2.0 performance was comparable to SOFA and APACHE II but did not consistently surpass them.
• MRC-ICU 2.0 provides additional predictive value when used with traditional severity-of-illness scores, indicating its utility as a complementary clinical measure.