Rationale: Opioid abuse is increasing, but its impact on critical care resources in the United States is unknown. Objectives: We hypothesized that there would be a rising need for critical care among opioid-associated overdoses in the United States. Methods: We analyzed all adult admissions, using a retrospective cohort study from 162 hospitals in 44 states, discharged between January 1, 2009, and September 31, 2015 to describe the incidence of intensive care unit (ICU) admissions for opioid overdose during this time. Admissions were identified using the Clinical Database/Resource Manager of Vizient, the successor to the University Health System Consortium. Results: Our primary outcome was opioid-associated overdose admissions to the ICU. The outcome was defined on the basis of previously validated ICD-9 codes. Our secondary outcomes were in-hospital death and markers of ICU resources. The final cohort included 22,783,628 admissions; 4,145,068 required ICU care. There were 52.4 ICU admissions for overdose per 10,000 ICU admissions over the entire study (95% confidence interval [CI], 51.8–53.0 per 10,000 ICU admissions). During this time period, opioid overdose admissions requiring intensive care increased 34%, from 44 per 10,000 (95% CI, 43–46 per 10,000) to 59 per 10,000 ICU admissions (95% CI, 57–61 per 10,000; P < 0.0001). The mortality rate of patients with ICU admissions with overdoses averaged 7% (95% CI, 7.0–7.6%) but increased to 10% in 2015 (95% CI, 8.8–10.8%). Conclusions: The number of deaths of ICU patients with opioid overdoses increased substantially in the 7 years of our study, reflecting increases in both the incidence and mortality of this condition. Our findings raise the need for a national approach to developing safe strategies to care for patients with overdose in the ICU, to providing coordinated resources in the hospital for patients and families, and to helping survivors maintain sobriety on discharge.

Problem definition: Translate data from electronic health records (EHR) into accurate predictions on patient flows and inform daily decision making at a major hospital. Academic/practical relevance: In a constrained hospital environment, forecasts on patient demand patterns could help match capacity and demand and improve hospital operations. Methodology: We use data from 63,432 admissions at a large academic hospital (50% female, median age 64 years old, median length of stay 3.12 days). We construct an expertise-driven patient representation on top of their EHR data and apply a broad class of machine learning methods to predict several aspects of patient flows. Results: With a unique patient representation, we estimate short-term discharges, identify long-stay patients, predict discharge destination, and anticipate flows in and out of intensive care units with accuracy in the 80%+ range. More importantly, we implement this machine learning pipeline into the EHR system of the hospital and construct prediction-informed dashboards to support daily bed placement decisions. Managerial implications: Our study demonstrates that interpretable machine learning techniques combined with EHR data can be used to provide visibility on patient flows. Our approach provides an alternative to deep learning techniques that is equally accurate, interpretable, frugal in data and computational power, and production ready. History: This paper has been accepted for the Manufacturing & Service Operations Management Special Section on Responsible Research in Operations Management. Funding: The research was funded by Beth Israel Deaconess Medical Center. Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2021.0971 .

Three dominant models of generalist care exist for adults hospitalized for medical diagnoses in the United States. For some patients, their primary care physician (PCP) serves as the attending physician during a hospitalization. For others, a hospitalist (a generalist physician who focuses primarily on hospitalized patients) serves as the attending physician. Finally, some patients are cared for by a cross-covering generalist who is neither the patient’s PCP nor a hospitalist (eg, physicians in group practices may cover inpatient admissions for the practice for a few weeks per year). These practice patterns reflect trade-offs between familiarity with the patient and expertise in hospital care. The association between these trade-offs and patient outcomes has not been extensively investigated.

As many as one in 10 patients experience dyspnoea at hospital admission but the relationship between dyspnoea and patient outcomes is unknown. We sought to determine whether dyspnoea on admission predicts outcomes.We conducted a retrospective cohort study in a single, academic medical centre. We analysed 67 362 consecutive hospital admissions with available data on dyspnoea, pain and outcomes. As part of the Initial Patient Assessment by nurses, patients rated “breathing discomfort” using a 0 to 10 scale (10=“unbearable”). Patients reported dyspnoea at the time of admission and recalled dyspnoea experienced in the 24 h prior to admission. Outcomes included in-hospital mortality, 2-year mortality, length of stay, need for rapid response system activation, transfer to the intensive care unit, discharge to extended care, and 7- and 30-day all-cause readmission to the same institution.Patients who reported any dyspnoea were at an increased risk of death during that hospital stay; the greater the dyspnoea, the greater the risk of death (dyspnoea 0: 0.8% in-hospital mortality; dyspnoea 1–3: 2.5% in-hospital mortality; dyspnoea ≥4: 3.7% in-hospital mortality; p<0.001). After adjustment for patient comorbidities, demographics and severity of illness, increasing dyspnoea remained associated with inpatient mortality (dyspnoea 1–3: adjusted OR 2.1, 95% CI 1.7–2.6; dyspnoea ≥4: adjusted OR 3.1, 95% CI 2.4–3.9). Pain did not predict increased mortality. Patients reporting dyspnoea also used more hospital resources, were more likely to be readmitted and were at increased risk of death within 2 years (dyspnoea 1–3: adjusted hazard ratio 1.5, 95% CI 1.3–1.6; dyspnoea ≥4: adjusted hazard ratio 1.7, 95% CI 1.5–1.8).We found that dyspnoea of any rating was associated with an increased risk of death. Dyspnoea ratings can be rapidly collected by nursing staff, which may allow for better monitoring or interventions that could reduce mortality and morbidity.