Outcomes of nighttime refusal of admission to the intensive care unit: The role of the intensivist in triage

Defining Intensivists: A Retrospective Analysis of the Published Studies in the United States, 2010-2020

OBJECTIVES

The Society of Critical Care Medicine last published an intensivist definition in 1992. Subsequently, there have been many publications relating to intensivists. Our purpose is to assess how contemporary studies define intensivist physicians.

DESIGN

Systematic search of PubMed, Embase, and Web of Science (2010-2020) for publication titles with the terms intensivist, and critical care or intensive care physician, specialist, or consultant. We included studies focusing on adult U.S. intensivists and excluded non-data-driven reports, non-U.S. publications, and pediatric or neonatal ICU reports. We aggregated the study title intensivist nomenclatures and parsed Introduction and Method sections to discern the text used to define intensivists. Fourteen parameters were found and grouped into five definitional categories: A) No definition, B) Background training and certification, C) Works in ICU, D) Staffing, and E) Database related. Each study was re-evaluated against these parameters and grouped into three definitional classes (single, multiple, or no definition). The prevalence of each parameter is compared between groups using Fisher exact test.

SETTING

U.S. adult ICUs and databases.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Of 657 studies, 105 (16%) met inclusion criteria. Within the study titles, 17 phrases were used to describe an intensivist; these were categorized as intensivist in 61 titles (58%), specialty intensivist in 30 titles (29%), and ICU/critical care physician in 14 titles (13%). Thirty-one studies (30%) used a single parameter (B-E) as their definition, 63 studies (60%) used more than one parameter (B-E) as their definition, and 11 studies (10%) had no definition (A). The most common parameter "Works in ICU" (C) in 52 studies (50%) was more likely to be used in conjunction with other parameters rather than as a standalone parameter (multiple parameters vs single-parameter studies; 73% vs 17%; p < 0.0001).

CONCLUSIONS

There was no consistency of intensivist nomenclature or definitions in contemporary adult intensivist studies in the United States.

Triage of Critically Ill Patients: Characteristics and Outcomes of Patients Refused as Too Well for Intensive Care

BACKGROUND

The appropriate selection of patients for the intensive care unit (ICU) is a concern in acute care settings. However, the description of patients deemed too well for the ICU has been rarely reported.

METHODS

We conducted a single-centre retrospective observational study of all patients either deemed "too well" for or admitted to the ICU during one year. Refused patients were screened for unexpected events within 7 days, defined as either ICU admission without another indication, or death without treatment limitations. Patients' characteristics and organisational factors were analysed according to refusal status, outcome and delay in ICU admission.

RESULTS

Among 2219 enrolled patients, the refusal rate was 10.4%. Refusal was associated with diagnostic groups, treatment limitations, patients' location on a ward, night time and ICU occupancy. Unexpected events occurred in 16 (6.9%) refused patients. A worse outcome was associated with time spent in hospital before refusal, patients' location on a ward, SOFA score and physician's expertise. Delayed ICU admissions were associated with ICU and hospital length of stay.

CONCLUSIONS

ICU triage selected safely most patients who would have probably not benefited from the ICU. We identified individual and organisational factors associated with ICU refusal, subsequent ICU admission or death.

Association of Sepsis Diagnosis at Daytime and on Weekdays with Compliance with the 3-Hour Sepsis Treatment Bundles. A Multicenter Cohort Study

Rationale: Compliance with sepsis bundles is associated with better outcomes, but information to support structural actions that might improve compliance is scarce. Few studies have evaluated bundle compliance in different time periods, with conflicting results.Objectives: To evaluate the association of sepsis identification during the daytime versus during the nighttime and on weekdays versus weekends with 3-hour sepsis treatment bundle compliance.Methods: This was an observational, multicenter study including patients with sepsis admitted between 2010 and 2017 to 10 hospitals in Brazil. Our exposures of interest were daytime (7:00 a.m.-6:59 p.m.) versus nighttime (7:00 p.m.-6:59 a.m.) and weekdays (Monday 7:00 a.m.-Friday 6:59 p.m.) versus weekends (Friday 7:00 p.m.-Monday 6:59 a.m.). Our primary outcome was full compliance with the 3-hour sepsis treatment bundles. We adjusted by potential confounding factors with multivariable logistic regression models.Results: Of 11,737 patients (8,733 sepsis and 3,004 septic shock), 3-hour bundle compliance was 79.1% and hospital mortality was 24.7%. The adjusted odds ratio (adjOR) for 3-hour full bundle compliance for patients diagnosed during the daytime versus during the nighttime was 1.35 (95% confidence interval [CI], 1.23-1.49; P < 0.001) and was more pronounced in the emergency department (adjOR, 1.55; 95% CI, 1.35-1.77; P < 0.001) than in nonemergency areas (adjOR, 1.19; 95% CI, 1.04-1.37; P = 0.014). Overall, there was no association between diagnosis on the weekends versus on weekdays and 3-hour full bundle compliance (adjOR, 1.08; 95% CI, 0.98-1.19; P = 0.115), although there was an association among those diagnosed in nonemergency areas (adjOR, 1.15; 95% CI, 1.00-1.32; P = 0.047). The lower compliance observed for sepsis diagnosed during the nighttime was more evident 2 years after implementation of the quality improvement initiative.Conclusions: Compliance with sepsis bundles was associated with the moment of sepsis diagnosis. The place of diagnosis and the time from campaign implementation were factors modifying this association. Our results support areas for better design of quality improvement initiatives to mitigate the influence of the period of sepsis diagnosis on treatment compliance.

Assessment of Time-Series Machine Learning Methods for Forecasting Hospital Discharge Volume

IMPORTANCE

Forecasting the volume of hospital discharges has important implications for resource allocation and represents an opportunity to improve patient safety at periods of elevated risk.

OBJECTIVE

To determine the performance of a new time-series machine learning method for forecasting hospital discharge volume compared with simpler methods.

DESIGN

A retrospective cohort study of daily hospital discharge volumes at 2 large, New England academic medical centers between January 1, 2005, and December 31, 2014 (hospital 1), or January 1, 2005, and December 31, 2010 (hospital 2), comparing time-series forecasting methods for prediction was performed. Data analysis was conducted from February 28, 2017, to August 30, 2018. Group-level data for all discharges from inpatient units were included. In addition to conventional methods, a technique originally developed for allocating data center resources, and comparison strategies for incorporating prior data and frequency of model updates, was conducted to identify the model application that optimized forecast accuracy.

MAIN OUTCOMES AND MEASURES

Model calibration as measured by R2 and, secondarily, number of days with errors greater than 1 SD of daily volume.

RESULTS

During the forecasted year, hospital 1 had 54 411 discharges (daily mean, 149) and hospital 2 had 47 456 discharges (daily mean, 130). The machine learning method was well calibrated at both sites (R2, 0.843 and 0.726, respectively) and made errors greater than 1 SD of daily volume on only 13 and 22 days, respectively, of the forecast year at the 2 sites. Last-value-carried-forward models performed somewhat less well (calibration R2, 0.781 and 0.596, respectively) with 13 and 46 errors of 1 SD or greater, respectively. More frequent retraining and training sets of longer than 1 year had minimal effects on the machine learning method's performance.

CONCLUSIONS AND RELEVANCE

Volume of hospital discharges can perhaps be reliably forecasted using simple carry-forward models as well as methods drawn from machine learning. The benefit of the latter does not appear to be dependent on extensive training data and may enable forecasts up to 1 year in advance with superior absolute accuracy to carry-forward models.