Cluster analysis driven by unsupervised latent feature learning of intensive care unit medications to identify novel pharmaco-phenotypes of critically ill patients

Abstract

Unsupervised clustering of intensive care unit (ICU) medications may identify unique medication clusters (i.e., pharmacophenotypes) in critically ill adults. We performed an unsupervised analysis with Restricted Boltzmann Machine of 991 medications profiles of patients managed in the ICU to explore pharmacophenotypes that correlated with ICU complications (e.g., mechanical ventilation) and patient-centered outcomes (e.g., length of stay, mortality). Six unique pharmacophenotypes were observed, with unique medication profiles and clinically relevant differences in ICU complications and patient-centered outcomes. While pharmacophenotypes 2 and 4 had no statistically significant difference in ICU length of stay, duration of mechanical ventilation, or duration of vasopressor use, their mortality differed significantly (9.0% vs. 21.9%, p < 0.0001). Pharmacophenotype 4 had a mortality rate of 21.9%, compared with the rest of the pharmacophenotypes ranging from 2.5-9%. Phenotyping approaches have shown promise in classifying the heterogenous syndromes of critical illness to predict treatment response and guide clinical decision support systems but have never included comprehensive medication information. This first-ever machine learning approach revealed differences among empirically-derived subgroups of ICU patients that are not typically revealed by traditional classifiers. Identification of pharmacophenotypes may enable enhanced decision making to optimize treatment decisions.

A Common Data Model for the standardization of intensive care unit (ICU) medication features in artificial intelligence (AI) applications

“…The first iteration used just generic drug name in addition to basic patient demographics and ICU outcomes, excluding all other potential drug features, but still showed the presence of pharmacophenotypes. [24] The second employed a significantly expanded list of drug features in line with this proposed CDM, showing again the ability of AI to incorporate vast amounts of data into its predictive algorithms. [40] Understanding how to translate these pharmacophenotypes into clinically meaningful subgroups or interventions for a bedside clinician remains an ongoing area of investigation.…”

“…For the feasibility of this initial process, the team focused on two key areas: first, medication products (e.g., aspirin 325mg tablet) and second, medication products mostly likely to be used in the ICU as defined by a previously validated metric (the medication regimen intensity-intensive care unit (MRC-ICU) Score). [13 14 24-31] Because these medications in the MRC-ICU have been previously identified as common to ICU care and having unique characteristics that make their use associated with increased ICU patient care complexity and a requirement for expert oversight, they were deemed an appropriate initial list of representative medications for CDM development. [13 14 24-31]…”

“…[13 14 24-31] Because these medications in the MRC-ICU have been previously identified as common to ICU care and having unique characteristics that make their use associated with increased ICU patient care complexity and a requirement for expert oversight, they were deemed an appropriate initial list of representative medications for CDM development. [13 14 24-31]…”

“…From here, the derived medication list comprised of 30,550 discrete medication order entries for 991 ICU patients. [24] When a filter for the generic drug name, dose, and administration route was applied, a total of 1,868 unique medication products were identified. When only those medications incorporated in the MRC-ICU Scoring Tool were considered, a total of 889 discrete medication products remained for review and coding by the panel.…”