A common data model for the standardization of intensive care unit medication features.

Objective: Common data models provide a standard means of describing data for artificial intelligence (AI) applications, but this process has never been undertaken for medications used in the intensive care unit (ICU). We sought to develop a common data model (CDM) for ICU medications to standardize the medication features needed to support future ICU AI efforts. Materials and Methods: A 9-member, multi-professional team of ICU clinicians and AI experts conducted a 5-round modified Delphi process employing conference calls, web-based communication, and electronic surveys to define the most important medication features for AI efforts. Candidate ICU medication features were generated through group discussion and then independently scored by each team member based on relevance to ICU clinical decision-making and feasibility for collection and coding. A key consideration was to ensure the final ontology both distinguished unique medications and met Findable, Accessible, Interoperable, and Reusable (FAIR) guiding principles. Results: Using a list of 889 ICU medications, the team initially generated 106 different medication features, and 71 were ranked as being core features for the CDM. Through this process, 106 medication features were assigned to 2 key feature domains: drug product-related (n = 43) and clinical practice-related (n = 63). Each feature included a standardized definition and suggested response values housed in the electronic data library. This CDM for ICU medications is available online. Conclusion: The CDM for ICU medications represents an important first step for the research community focused on exploring how AI can improve patient outcomes and will require ongoing engagement and refinement.

Getting back on the curve: Reimagining individualized, sustainable residency training for the critical care pharmacist.

In an effort to expedite the publication of articles, AJHP is posting manuscripts online as soon as possible after acceptance. Accepted manuscripts have been peer-reviewed and copyedited, but are posted online before technical formatting and author proofing. These manuscripts are not the final version of record and will be replaced with the final article (formatted per AJHP style and proofed by the authors) at a later time.

Unsupervised machine learning analysis to identify patterns of ICU medication use for fluid overload prediction.

INTRODUCTION: Intravenous (IV) medications are a fundamental cause of fluid overload (FO) in the intensive care unit (ICU); however, the association between IV medication use (including volume), administration timing, and FO occurrence remains unclear. METHODS: This retrospective cohort study included consecutive adults admitted to an ICU ≥72 hours with available fluid balance data. FO was defined as a positive fluid balance ≥7% of admission body weight within 72 hours of ICU admission. After reviewing medication administration record (MAR) data in three-hour periods, IV medication exposure was categorized into clusters using principal component analysis (PCA) and Restricted Boltzmann Machine (RBM). Medication regimens of patients with and without FO were compared within clusters to assess for temporal clusters associated with FO using the Wilcoxon rank sum test. Exploratory analyses of the medication cluster most associated with FO for medications frequently appearing and used in the first 24 hours was conducted. RESULTS: FO occurred in 127/927 (13.7%) of the patients enrolled. Patients received a median (IQR) of 31 (13-65) discrete IV medication administrations over the 72-hour period. Across all 47,803 IV medication administrations, ten unique IV medication clusters were identified with 121-130 medications in each cluster. Among the ten clusters, cluster 7 had the greatest association with FO; the mean number of cluster 7 medications received was significantly greater in patients in the FO cohort compared to patients who did not experience FO (25.6 vs.10.9. p<0.0001). 51 of the 127 medications in cluster 7 (40.2%) appeared in > 5 separate 3-hour periods during the 72-hour study window. The most common cluster 7 medications included continuous infusions, antibiotics, and sedatives/analgesics. Addition of cluster 7 medications to a prediction model with APACHE II score and receipt of diuretics improved the ability for the model to predict fluid overload (AUROC 5.65, p =0.0004). CONCLUSIONS: Using ML approaches, a unique IV medication cluster was strongly associated with FO. Incorporation of this cluster improved the ability to predict development of fluid overload in ICU patients compared with traditional prediction models. This method may be further developed into real-time clinical applications to improve early detection of adverse outcomes.

Teaching and learning in biophotonics: Crossing the bridge between educators and students.

As a rapidly growing field, biophotonics demonstrates an increasingly higher demand for interdisciplinary professionals and requires the implementation of a structured approach to educational and outreach activities focused on appropriate curriculum, and teaching and learning for audiences with diverse technical backgrounds and learning styles. Our study shows the main findings upon applying this approach to biophotonics workshops delivered 2 consecutive years while updating and improving learning outcomes, teaching strategies, workshop content based on student and teacher feedback. We provided resources for a variety of lecture-based, experimental, computer simulation activities. Quality of subject matter, teaching, and overall learning was rated as "Very good" or "Good" by 88%, 76%, and 82% of students in average, respectively. Application of our teaching strategies and materials during short- and long-term workshops/courses could potentially increase the interest in pursuing careers in the biophotonics field and related areas, leading to standardized approaches in designing education and outreach events across centers.

Addressing Differences in Knowledge and Experience in Trauma Care Capabilities Among an International Team of Military Medical Care Providers in a Deployed Setting.

Military medical personnel are crucial in providing life-saving care at the point of injury (POI) in challenging environments such as combat zones and disaster areas. This article examines the specialized training US Military medical personnel undergo before deployment and the increasing trend of deploying as part of multinational forces in operations like those in Afghanistan with NATO and non-NATO countries. Integrating medical teams from diverse backgrounds poses significant challenges in maintaining a cohesive and efficient team due to varying trauma management training standards and medical practices among the allied forces. Tactical Combat Casualty Care (TCCC) training and the development of clinical practice guidelines (CPG) by the US Joint Trauma Service have been key strategies by the US Military to standardize care. However, the variation in trauma management training among NATO allies and the differences in medical subspecialties and approaches can lead to inefficiencies and reduced effectiveness in a multinational trauma center setting. For instance, the approach to trauma care can significantly differ between the US and European countries, impacting the interoperability and teamwork in multinational medical teams. To address these challenges, the article highlights the importance of standardized medical training programs that include cultural awareness to enhance the effectiveness of multidisciplinary, multinational medical teams. It also underscores the necessity for standardized international trauma training in the face of increasing global conflicts and the potential for large-scale combat operations. The article discusses the "Trauma Tuesday" program implemented in Kabul, Afghanistan, as an example of how intensive training and simulation exercises can improve team dynamics, knowledge, and skills in trauma management among a diverse team from various nations. The need for ongoing education and developing a standard for managing trauma patients in international teams is emphasized to ensure effective communication and coordination. The article suggests that multinational trauma training can significantly improve team cohesion and critical life-saving skills, essential for future battlefields where access to definitive care may be delayed. Further research is recommended to explore the best methods for achieving effective multinational medical team integration and training standardization.

The relationship between the morningness-eveningness questionnaire and incident cancer: A historical clinical cohort study.

OBJECTIVE: We conducted a retrospective cohort study to explore the relationship between chronotype measured by the total Morningness-Eveningness Questionnaire (MEQ) score and incident cancer. METHODS: We used clinical and provincial health administrative data on consecutive adults who underwent a Level 1 Polysomnography (PSG) and completed the MEQ between 2010 and 2015 in an academic hospital (Ontario, Canada) and were cancer-free at baseline. Cancer status was derived from the Ontario Cancer Registry. Individuals were followed until death or March 31, 2020. We used multivariable Cox cause-specific regressions to address the research objective. RESULTS: Of 3,004 individuals, 1,781 were analyzed: a median age of 54 years (IQR: 40-64) and 838 (47.1%) men. The median total MEQ score was 63 (IQR: 55-69); 61 (3.4%) were classified as evening (≤41), 536 (30.1%) as intermediate (42-58), and 1,184 (66.5%) as morning chronotypes (≥59). Over a median of 7 years (IQR: 5-8), 120 (6.7%) developed cancer. A U-shape relationship was found between the total MEQ score and an increased hazard of incident cancer, controlling for PSG measures of sleep apnea severity and sleep architecture, demographics, and comorbidities. Compared to the median of 63.0, a total MEQ score greater or less than the median was associated with an increased hazard of incident cancer, with the largest effect for those with a total score ≥76 (e.g., HR of a MEQ total score of 78 vs. 63: 2.01, 95% CI: 1.09-3.71). CONCLUSION: The U-shaped curve may reflect deviations from a standard circadian tendency, which may stress biological systems and influence malignancy risk.

Obstructive sleep apnea in those with idiopathic intracranial hypertension undergoing diagnostic in-laboratory polysomnography.

RATIONALE: The association of obstructive sleep apnea (OSA) with idiopathic intracranial hypertension (IIH) remains unclear, and few studies have used objective in-laboratory polysomnography (PSG) data. Thus, we used PSG data to examine the: 1) association between OSA, and its severity, with IIH and 2) sex differences in OSA severity in those with and without IIH. METHODS: We retrospectively analyzed diagnostic PSG data from January 2015 to August 2023 for patients who were diagnosed with IIH by a neuro-ophthalmologist using the modified Dandy criteria. We selected three age, sex, and body mass index (BMI) matched controls for each IIH patient. We examined potential associations of IIH with OSA using regression. Sex differences were analyzed using ANOVA. RESULTS: Of 3482 patients who underwent PSG, we analyzed 78 IIH patients (16 males) and 234 matched controls (48 males). Five (6.4 %) IIH and 39 (16.7 %) control patients had OSA, defined as AHI≥15. After adjusting for age, sex, BMI, and comorbidities, IIH was negatively associated with the presence of OSA (OR 0.29, 95%CI 0.10-0.87, p = 0.03). However, models that adjusted for acetazolamide use, with or without comorbidities, showed no significant relationship with OSA (OR 0.31, p = 0.20). Males with IIH had a significantly higher age (p = 0.020), OSA severity (p = 0.032), and arousal index (p = 0.046) compared to females with IIH. CONCLUSIONS: IIH treated with acetazolamide was not an independent risk factor for OSA presence or severity. The presence of IIH treated with acetazolamide likely does not warrant routine screening for OSA, but related risk factors may identify appropriate patients.

Access to, Experience with, and Attitudes towards Take Home Naloxone: An Online Survey.

The opioid public health crisis continues to burden individuals, communities, and economies. Public health opinion has emphasized the need for increased access to harm reduction services, but there is a dearth of information on the views and experiences of people who use opioids. Our study aimed to investigate the prevalence of naloxone use, attitudes, and experiences with naloxone among an online community of people who use drugs. We performed a cross-sectional survey looking at experiences with and attitudes towards take-home naloxone. Data is presented descriptively, with analysis of the differences between people who do and do not use opioids using the χ2 and Fisher's exact tests. There were 1,143 respondents, of whom 70% were from the United States. Only 38% of participants who use opioids had received naloxone training, but 56% of these individuals said that they felt comfortable using a naloxone kit. Nearly all respondents (95%) said they would be willing to use naloxone on someone who had overdosed and approximately 90% would want naloxone used on them in case of an overdose. Regarding harm reduction, 24% of respondents said they had access to safe use programs, and 33% said they had access to clean needle exchange programs. A majority of the participants who use opioids were in favor of having naloxone with them when using drugs and believed naloxone should be freely available. This study demonstrates the receptiveness of take-home naloxone and highlights the need for better implementation of naloxone within communities that use opioids.

Machine-learning assisted swallowing assessment: a deep learning-based quality improvement tool to screen for post-stroke dysphagia.

Introduction: Post-stroke dysphagia is common and associated with significant morbidity and mortality, rendering bedside screening of significant clinical importance. Using voice as a biomarker coupled with deep learning has the potential to improve patient access to screening and mitigate the subjectivity associated with detecting voice change, a component of several validated screening protocols. Methods: In this single-center study, we developed a proof-of-concept model for automated dysphagia screening and evaluated the performance of this model on training and testing cohorts. Patients were admitted to a comprehensive stroke center, where primary English speakers could follow commands without significant aphasia and participated on a rolling basis. The primary outcome was classification either as a pass or fail equivalent using a dysphagia screening test as a label. Voice data was recorded from patients who spoke a standardized set of vowels, words, and sentences from the National Institute of Health Stroke Scale. Seventy patients were recruited and 68 were included in the analysis, with 40 in training and 28 in testing cohorts, respectively. Speech from patients was segmented into 1,579 audio clips, from which 6,655 Mel-spectrogram images were computed and used as inputs for deep-learning models (DenseNet and ConvNext, separately and together). Clip-level and participant-level swallowing status predictions were obtained through a voting method. Results: The models demonstrated clip-level dysphagia screening sensitivity of 71% and specificity of 77% (F1 = 0.73, AUC = 0.80 [95% CI: 0.78-0.82]). At the participant level, the sensitivity and specificity were 89 and 79%, respectively (F1 = 0.81, AUC = 0.91 [95% CI: 0.77-1.05]). Discussion: This study is the first to demonstrate the feasibility of applying deep learning to classify vocalizations to detect post-stroke dysphagia. Our findings suggest potential for enhancing dysphagia screening in clinical settings. https://github.com/UofTNeurology/masa-open-source.

Machine learning vs. traditional regression analysis for fluid overload prediction in the ICU.

Fluid overload, while common in the ICU and associated with serious sequelae, is hard to predict and may be influenced by ICU medication use. Machine learning (ML) approaches may offer advantages over traditional regression techniques to predict it. We compared the ability of traditional regression techniques and different ML-based modeling approaches to identify clinically meaningful fluid overload predictors. This was a retrospective, observational cohort study of adult patients admitted to an ICU ≥ 72 h between 10/1/2015 and 10/31/2020 with available fluid balance data. Models to predict fluid overload (a positive fluid balance ≥ 10% of the admission body weight) in the 48-72 h after ICU admission were created. Potential patient and medication fluid overload predictor variables (n = 28) were collected at either baseline or 24 h after ICU admission. The optimal traditional logistic regression model was created using backward selection. Supervised, classification-based ML models were trained and optimized, including a meta-modeling approach. Area under the receiver operating characteristic (AUROC), positive predictive value (PPV), and negative predictive value (NPV) were compared between the traditional and ML fluid prediction models. A total of 49 of the 391 (12.5%) patients developed fluid overload. Among the ML models, the XGBoost model had the highest performance (AUROC 0.78, PPV 0.27, NPV 0.94) for fluid overload prediction. The XGBoost model performed similarly to the final traditional logistic regression model (AUROC 0.70; PPV 0.20, NPV 0.94). Feature importance analysis revealed severity of illness scores and medication-related data were the most important predictors of fluid overload. In the context of our study, ML and traditional models appear to perform similarly to predict fluid overload in the ICU. Baseline severity of illness and ICU medication regimen complexity are important predictors of fluid overload.

Fully automated minicolumn chromatography.

Miniaturized chromatography columns (minicolumns) operated by automated liquid handlers are an integral part of bioprocess purification development. However, these systems can be limited in both their efficiency and accessibility. Because the minicolumn chromatography operation itself is higher throughput, the lower throughput pre- and post-operation activities become the bottleneck of the workflow. Additionally, method writing and operation of the systems while varying multiple parameters, using a design of experiments approach for example, can be error-prone and resource intensive. Here, we have developed a fully automated minicolumn chromatography system to both address these bottlenecks and improve the accessibility of these systems by allowing users to enter chromatography-relevant information through a simplified user interface. Methods have been developed to automate buffer preparation and protein solution titration leveraging modeling and integrated pH probes with feedback control. Chromatogram generation and fraction pooling has additionally been automated to improve the efficiency of post-chromatography operations. We have also demonstrated the flexibility of the system through an example run where both bind-and-elute chromatography and flowthrough chromatography experiments were performed in parallel. Additionally, all methodology and parameters to operate the system have been shared. We hope this will help interested parties improve the efficiency and accessibility of their minicolumn chromatography systems.

Strengthening oropharyngeal muscles as an approach to treat post-stroke obstructive sleep apnea: A feasibility randomised controlled trial.

This study aimed to determine the feasibility of a randomised controlled trial (RCT) evaluating oropharyngeal exercise (OPE) intervention as an alternative therapy for obstructive sleep apnea (OSA) in patients with stroke or transient ischaemic attack (TIA). Despite the high prevalence of OSA in this population, the standard therapy, continuous positive airway pressure (CPAP), is often poorly tolerated. Thirty stroke/TIA patients with OSA unable to tolerate CPAP were randomly assigned to an oropharyngeal exercise or sham exercise protocol. They performed exercises for 6 weeks, 5 days per week, 30 minutes twice per day. Feasibility was ascertained by the proportion of enrolled patients who completed more than 80% of the OPE regimen. Isometric tongue pressures, apnea-hypopnea index (AHI), oxygen desaturation index (ODI), daytime sleepiness, and quality of life (QOL) outcomes were collected at baseline, post-training (6-week follow-up), and retention (10-week follow-up) to document preliminary efficacy. Adherence to study exercises was excellent, with 83% of participants completing more than 80% of the exercises. The isometric tongue pressures were observed to improve in the oropharyngeal exercise group (compared with the sham group), along with a decrease in OSA severity (measured by the AHI and ODI), reduced daytime sleepiness, and enhanced quality of life outcomes following the exercise programme. Only the effects on posterior isometric tongue pressure and daytime sleepiness remained significantly different between groups at the retention session. In conclusion, an RCT evaluating the efficacy of oropharyngeal exercises on post-stroke/TIA OSA is feasible and our preliminary results suggest a clinically meaningful effect.

Sex Differences in Obstructive Sleep Apnea after Stroke.

BACKGROUND AND OBJECTIVES: Obstructive sleep apnea (OSA) is prevalent after stroke and associated with recurrent stroke, prolonged hospitalization, and decreased functional recovery. Sex differences in post-stroke OSA remain underexplored. The objective of this study was to evaluate sex differences in functional outcomes, stroke and OSA severity, and clinical manifestations of OSA in stroke patients with OSA. METHODS: We retrospectively evaluated data from three previously conducted studies. Study patients had an imaging-confirmed stroke and had been found to have OSA (apnea-hypopnea index [AHI] ≥ 5) on either in-laboratory polysomnography or home sleep apnea testing performed within 1 year of their stroke. Linear regression models were used to evaluate study outcomes. RESULTS: In total, 171 participants with post-stroke OSA (117 males [68.4%] and 54 females [31.6%]) were included. Female sex was an independent predictor for greater functional impairment (ß = 0.37, 95% CI 0.029-0.71, p = 0.03), increased stroke severity (ß = 1.009, 95% CI 0.032-1.99, p = 0.04), and greater post-stroke depressive symptoms (ß = 3.73, 95% CI 0.16-7.29, p = 0.04). Female sex was associated with lower OSA severity, as measured by the AHI (ß = -5.93, 95% CI -11.21- -0.66). Sex was not an independent predictor of specific symptoms of OSA such as daytime sleepiness, snoring, tiredness, and observed apneas. CONCLUSION: Females with post-stroke OSA had poorer functional outcomes and more severe strokes compared to males, despite having lower OSA severity. Females with post-stroke OSA also exhibited more depressive symptoms. Understanding sex differences in patients with post-stroke OSA will likely facilitate better recognition of OSA and potentially improve clinical outcomes.

Machine learning based prediction of prolonged duration of mechanical ventilation incorporating medication data.

Rationale: Duration of mechanical ventilation is associated with adverse outcomes in critically ill patients and increased use of resources. The increasing complexity of medication regimens has been associated with increased mortality, length of stay, and fluid overload but has never been studied specifically in the setting of mechanical ventilation. Objective: The purpose of this analysis was to develop prediction models for mechanical ventilation duration to test the hypothesis that incorporating medication data may improve model performance. Methods: This was a retrospective cohort study of adults admitted to the ICU and undergoing mechanical ventilation for longer than 24 hours from October 2015 to October 2020. Patients were excluded if it was not their index ICU admission or if the patient was placed on comfort care in the first 24 hours of admission. Relevant patient characteristics including age, sex, body mass index, admission diagnosis, morbidities, vital signs measurements, severity of illness, medication regimen complexity as measured by the MRC-ICU, and medical treatments before intubation were collected. The primary outcome was area under the receiver operating characteristic (AUROC) of prediction models for prolonged mechanical ventilation (defined as greater than 5 days). Both logistic regression and supervised learning techniques including XGBoost, Random Forest, and Support Vector Machine were used to develop prediction models. Results: The 318 patients [age 59.9 (SD 16.9), female 39.3%, medical 28.6%] had mean 24-hour MRC-ICU score of 21.3 (10.5), mean APACHE II score of 21.0 (5.4), mean SOFA score of 9.9 (3.3), and ICU mortality rate of 22.6% (n=72). The strongest performing logistic model was the base model with MRC-ICU added, with AUROC of 0.72, positive predictive value (PPV) of 0.83, and negative prediction value (NPV) of 0.92. The strongest overall model was Random Forest with an AUROC of 0.78, a PPV of 0.53, and NPV of 0.90. Feature importance analysis using support vector machine and Random Forest revealed severity of illness scores and medication related data were the most important predictors. Conclusions: Medication regimen complexity is significantly associated with prolonged duration of mechanical ventilation in critically ill patients, and prediction models incorporating medication information showed modest improvement in this prediction.

Hydrofluoric acid ingestions: Retrospective evaluations from cases reported to the National Poison Data System 2007-2017.

Background: Hydrofluoric acid (HF) is associated with systemic toxicity, particularly with high-concentration formulations. However, most existing data describe dermal exposures; there is a paucity of data related to outcomes after ingestions. Objective: To determine the morbidity and mortality associated with HF ingestions as reported to the National Poison Data System (NPDS). A secondary objective is to assess for clinical criteria that are associated with serious outcomes after HF ingestion. Methods: We performed a retrospective review of HF ingestions reported to the NPDS from 2007 to 2017. Data including patient demographics, exposure and caller sites, electrolyte abnormalities, treatments, and serious (moderate or major effect or death as documented in NPDS) and non-serious outcomes were abstracted from case narratives. Cases meeting the criteria for a qualifiable HF ingestion were included in the study. Results: During the study period, there were 653 HF ingestions reported to NPDS, of which 142 were included in the final data analysis. Most HF exposures occurred in men (68.3%), and the most common exposure site was at the exposed individual's own residence (78.2%). Nearly half of all exposures (46.5%) were due to transfer into a non-labeled secondary storage container. Total of 45.8% of the cases resulted in a serious outcome. Electrolyte disturbances were associated with an increased risk of a serious outcome. Hypocalcemia was the most frequently reported electrolyte abnormality, occurring in 24.6% of cases. Nine (6.3%) individuals died. Conclusions: Mortality after HF ingestion is low. However, a large cohort of exposures occurred after the transfer of HF to secondary containers. Targeted interventions to reduce this practice are necessary to decrease hazardous chemical exposures.

Associations between short-term ambient temperature exposure and emergency department visits for amphetamine, cocaine, and opioid use in California from 2005 to 2019.

Substance use disorder is a growing public health challenge in the United States. People who use drugs may be more vulnerable to ambient heat due to the effects of drugs on thermoregulation and their risk environment. There have been limited population-based studies of ambient temperature and drug-related morbidity. We examined short-term associations between daily ambient temperature and emergency department (ED) visits for use or overdose of amphetamine, cocaine and opioids in California during the period 2005 to 2019. Daily ZIP code-level maximum, mean, and minimum temperature exposures were derived from 1-km data Daymet products. A time-stratified case-crossover design was used to estimate cumulative non-linear associations of daily temperature for lag days 0 to 3. Stratified analyses by patient sex, race, and ethnicity were also conducted. The study included over 3.4 million drug-related ED visits. We found positive associations between daily temperature and ED visits for all outcomes examined. An increase in daily mean temperature from the 50th to the 95th percentile was associated with ED visits for amphetamine use (OR = 1.072, 95% CI: 1.058, 1.086), cocaine use (OR = 1.044, 95% CI: 1.021, 1.068 and opioid use (OR = 1.041, 95% CI: 1.025, 1.057). Stronger positive associations were also observed for overdose: amphetamine overdose (OR = 1.150, 95% CI: 1.085, 1.218), cocaine overdose (OR = 1.159, 95% CI: 1.053, 1.276), and opioid overdose (OR = 1.079, 95% CI: 1.054, 1.106). In summary, people who use stimulants and opioids may be a subpopulation sensitive to short-term higher ambient temperature. Mitigating heat exposure can be considered in harm reduction strategies in response to the substance use epidemic and global climate change.

Cluster analysis driven by unsupervised latent feature learning of medications to identify novel pharmacophenotypes of critically ill patients.

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 to 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.