The Association of ICU Acuity With Outcomes of Patients at Low Risk of Dying.

OBJECTIVE: Many ICU patients do not require critical care interventions. Whether aggressive care environments increase risks to low-acuity patients is unknown. We evaluated whether ICU acuity was associated with outcomes of low mortality-risk patients. We hypothesized that admission to high-acuity ICUs would be associated with worse outcomes. This hypothesis was based on two possibilities: 1) high-acuity ICUs may have a culture of aggressive therapy that could lead to potentially avoidable complications and 2) high-acuity ICUs may focus attention toward the many sicker patients and away from the fewer low-risk patients. DESIGN: Retrospective cohort study. SETTING: Three hundred twenty-two ICUs in 199 hospitals in the Philips eICU database between 2010 and 2015. PATIENTS: Adult ICU patients at low risk of dying, defined as an Acute Physiology and Chronic Health Evaluation-IVa-predicted mortality of 3% or less. EXPOSURE: ICU acuity, defined as the mean Acute Physiology and Chronic Health Evaluation IVa score of all admitted patients in a calendar year, stratified into quartiles. MEASUREMENTS AND MAIN RESULTS: We used generalized estimating equations to test whether ICU acuity is independently associated with a primary outcome of ICU length of stay and secondary outcomes of hospital length of stay, hospital mortality, and discharge destination. The study included 381,997 low-risk patients. Mean ICU and hospital length of stay were 1.8 ± 2.1 and 5.2 ± 5.0 days, respectively. Mean Acute Physiology and Chronic Health Evaluation IVa-predicted hospital mortality was 1.6% ± 0.8%; actual hospital mortality was 0.7%. In adjusted analyses, admission to low-acuity ICUs was associated with worse outcomes compared with higher-acuity ICUs. Specifically, compared with the highest-acuity quartile, ICU length of stay in low-acuity ICUs was increased by 0.24 days; in medium-acuity ICUs by 0.16 days; and in high-acuity ICUs by 0.09 days (all p < 0.001). Similar patterns existed for hospital length of stay. Patients in lower-acuity ICUs had significantly higher hospital mortality (odds ratio, 1.28 [95% CI, 1.10-1.49] for low-; 1.24 [95% CI, 1.07-1.42] for medium-, and 1.14 [95% CI, 0.99-1.31] for high-acuity ICUs) and lower likelihood of discharge home (odds ratio, 0.86 [95% CI, 0.82-0.90] for low-, 0.88 [95% CI, 0.85-0.92] for medium-, and 0.95 [95% CI, 0.92-0.99] for high-acuity ICUs). CONCLUSIONS: Admission to high-acuity ICUs is associated with better outcomes among low mortality-risk patients. Future research should aim to understand factors that confer benefit to patients with different risk profiles.

Identifying Distinct Subgroups of ICU Patients: A Machine Learning Approach.

OBJECTIVES: Identifying subgroups of ICU patients with similar clinical needs and trajectories may provide a framework for more efficient ICU care through the design of care platforms tailored around patients' shared needs. However, objective methods for identifying these ICU patient subgroups are lacking. We used a machine learning approach to empirically identify ICU patient subgroups through clustering analysis and evaluate whether these groups might represent appropriate targets for care redesign efforts. DESIGN: We performed clustering analysis using data from patients' hospital stays to retrospectively identify patient subgroups from a large, heterogeneous ICU population. SETTING: Kaiser Permanente Northern California, a healthcare delivery system serving 3.9 million members. PATIENTS: ICU patients 18 years old or older with an ICU admission between January 1, 2012, and December 31, 2012, at one of 21 Kaiser Permanente Northern California hospitals. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: We used clustering analysis to identify putative clusters among 5,000 patients randomly selected from 24,884 ICU patients. To assess cluster validity, we evaluated the distribution and frequency of patient characteristics and the need for invasive therapies. We then applied a classifier built from the sample cohort to the remaining 19,884 patients to compare the derivation and validation clusters. Clustering analysis successfully identified six clinically recognizable subgroups that differed significantly in all baseline characteristics and clinical trajectories, despite sharing common diagnoses. In the validation cohort, the proportion of patients assigned to each cluster was similar and demonstrated significant differences across clusters for all variables. CONCLUSIONS: A machine learning approach revealed important differences between empirically derived subgroups of ICU patients that are not typically revealed by admitting diagnosis or severity of illness alone. Similar data-driven approaches may provide a framework for future organizational innovations in ICU care tailored around patients' shared needs.

Differential gene expression patterns and interaction networks in BCR-ABL-positive and -negative adult acute lymphoblastic leukemias.

PURPOSE: To identify gene expression patterns and interaction networks related to BCR-ABL status and clinical outcome in adults with acute lymphoblastic leukemia (ALL). PATIENTS AND METHODS: DNA microarrays were used to profile a set of 54 adult ALL specimens from the Medical Research Council UKALL XII/Eastern Cooperative Oncology Group E2993 trial (21 p185BCR-ABL-positive, 16 p210BCR-ABL-positive and 17 BCR-ABL-negative specimens). RESULTS: Using supervised and unsupervised analysis tools, we detected significant transcriptomic changes in BCR-ABL-positive versus -negative specimens, and assessed their validity in an independent cohort of 128 adult ALL specimens. This set of 271 differentially expressed genes (including GAB1, CIITA, XBP1, CD83, SERPINB9, PTP4A3, NOV, LOX, CTNND1, BAALC, and RAB21) is enriched for genes involved in cell death, cellular growth and proliferation, and hematologic system development and function. Network analysis demonstrated complex interaction patterns of these genes, and identified FYN and IL15 as the hubs of the top-scoring network. Within the BCR-ABL-positive subgroups, we identified genes overexpressed (PILRB, STS-1, SPRY1) or underexpressed (TSPAN16, ADAMTSL4) in p185BCR-ABL-positive ALL relative to p210BCR-ABL-positive ALL. Finally, we constructed a gene expression- and interaction-based outcome predictor consisting of 27 genes (including GRB2, GAB1, GLI1, IRS1, RUNX2, and SPP1), which correlated with overall survival in BCR-ABL-positive adult ALL (P = .0001), independent of age (P = .25) and WBC count at presentation (P = .003). CONCLUSION: We identified prominent molecular features of BCR-ABL-positive adult ALL, which may be useful for developing novel therapeutic targets and prognostic markers in this disease.