Observational Research for Therapies Titrated to Effect and Associated With Severity of Illness: Misleading Results From Commonly Used Statistical Methods

Characteristics and Outcomes for Low-Risk Hospital Admissions Admitted to the ICU: A Multisite Cohort Study

Supplemental Digital Content is available in the text.

IMPORTANCE:

Prognostication following ICU admission can often be determined based on known risk factors, including demographics and illness severity; however, little is known about outcomes of patients deemed to be “low-risk” at the time of hospital admission who subsequently are admitted to the ICU.

OBJECTIVES:

The objectives of this study were to determine the characteristics, outcomes, and costs for patients requiring ICU admission despite having lower predicted mortality when they were admitted to the hospital.

DESIGN, SETTING, AND PARTICIPANTS:

In this historical cohort study, we used a prospectively maintained ICU registry that included all ICU admissions to The Ottawa Hospital for patients 18 years or older from January 2011 to December 2016. We classified patients as low-risk using the Hospital-patient 1-year Mortality Risk at admission score, a hospital admission score validated to predict 1-year mortality.

MAIN OUTCOMES AND MEASURES:

The primary outcome was inhospital mortality. Secondary outcomes included adverse events, resource utilization, and costs.

RESULTS:

Of the 17,173 total ICU patients, 3,445 (20.1%) were classified as low-risk at hospital admission. Low-risk patients were younger (48.7 vs 67.5 yr; p < 0.001) and had a lower Multiple Organ Dysfunction Score (2.37 vs 4.14; p < 0.001). Mortality for low-risk patients was significantly lower than for non–low-risk patients (4.1% vs 25.4%; p < 0.001). For low-risk patients, multivariable logistic regression showed mortality was independently associated with older age (odds ratio, 1.02 per 1 yr; 95% CI, 1.00–1.03 per 1 yr), Multiple Organ Dysfunction Score (odds ratio, 1.42 per 1 point; 95% CI, 1.31–1.54 per 1 point), fluid management adverse events (odds ratio, 2.84; 95% CI, 1.29–6.25), hospital-acquired infections (odds ratio, 1.60; 95% CI, 1.02–2.51), and mechanical ventilation (odds ratio, 1.98; 95% CI, 1.20–3.26).

CONCLUSIONS AND RELEVANCE:

Despite their robust premorbid status, low-risk patients admitted to the ICU had significant inhospital mortality. Fluid management adverse events, hospital-associated infections, multiple organ dysfunction, and mechanical ventilation are important prognostic factors for low-risk patients.

Modelling hospital outcome: problems with endogeneity

Background

Mortality modelling in the critical care paradigm traditionally uses logistic regression, despite the availability of estimators commonly used in alternate disciplines. Little attention has been paid to covariate endogeneity and the status of non-randomized treatment assignment. Using a large registry database, various binary outcome modelling strategies and methods to account for covariate endogeneity were explored.

Methods

Patient mortality data was sourced from the Australian & New Zealand Intensive Society Adult Patient Database for 2016. Hospital mortality was modelled using logistic, probit and linear probability (LPM) models with intensive care (ICU) providers as fixed (FE) and random (RE) effects. Model comparison entailed indices of discrimination and calibration, information criteria (AIC and BIC) and binned residual analysis. Suspect covariate and ventilation treatment assignment endogeneity was identified by correlation between predictor variable and hospital mortality error terms, using the Stata™ “eprobit” estimator. Marginal effects were used to demonstrate effect estimate differences between probit and “eprobit” models.

Results

The cohort comprised 92,693 patients from 124 intensive care units (ICU) in calendar year 2016. Patients mean age was 61.8 (SD 17.5) years, 41.6% were female and APACHE III severity of illness score 54.5(25.6); 43.7% were ventilated. Of the models considered in predicting hospital mortality, logistic regression (with or without ICU FE) and RE logistic regression dominated, more so the latter using information criteria indices. The LPM suffered from many predictions outside the unit [0,1] interval and both poor discrimination and calibration. Error terms of hospital length of stay, an independent risk of death score and ventilation status were correlated with the mortality error term. Marked differences in the ventilation mortality marginal effect was demonstrated between the probit and the "eprobit" models which were scenario dependent. Endogeneity was not demonstrated for the APACHE III score.

Conclusions

Logistic regression accounting for provider effects was the preferred estimator for hospital mortality modelling. Endogeneity of covariates and treatment variables may be identified using appropriate modelling, but failure to do so yields problematic effect estimates.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01251-8.

Practice Patterns in the Initiation of Secondary Vasopressors and Adjunctive Corticosteroids during Septic Shock in the US

RATIONALE

A central component of septic shock treatment is the infusion of vasopressors, most commonly starting with norepinephrine. However, the optimal approach and practice patterns to initiating adjunctive vasopressors and corticosteroids are unknown.

OBJECTIVES

To characterize practice pattern variation in the norepinephrine dose at which secondary vasopressors and adjunctive corticosteroids are initiated and to identify factors associated with a treatment strategy favoring secondary vasopressors to a treatment strategy favoring adjunctive corticosteroids among patients with septic shock on norepinephrine.

METHODS

We used a multicenter ICU database to identify patients with septic shock who were started on norepinephrine followed by an additional vasopressor or corticosteroids. We used multi-level models to determine the hospital risk-adjusted norepinephrine dose at which additional vasopressors and corticosteroids were started, the percentage of variation in norepinephrine dose at the time of adjunctive treatment associated with hospital of admission, and the factors associated with choosing an 'additional vasopressor first' strategy vs. a 'corticosteroid first' strategy.

RESULTS

Among 4,401 patients with septic shock on norepinephrine, 1940 (44.0%) were started on adjuncts (1357 received an 'additional vasopressor first' strategy and 583 received a "corticosteroid first strategy). The hospital risk-adjusted norepinephrine dose at which vasopressors were initiated ranged 6.4 mcg/min (95% CI 5.9-7.0 mcg/min) to 92.6 mcg/min (95% CI 72.8-113.0 mcg/min). The hospital risk-adjusted norepinephrine dose at which corticosteroids was initiated ranged 3.0 mcg/min (95% CI 2.4-3.8 mcg/min) to 32.7 mcg/min (95% CI 24.9-43.0 mcg/min). 25.1% (intraclass correlation coefficient 95% CI 24.8%, 25.5%) of the variation in norepinephrine dose at which additional vasopressors were initiated was explained by hospital site after adjusting for all hospital- and patient-level covariates. Hospital of admission was strongly associated with receiving an 'additional vasopressor first' over a 'corticosteroids first' strategy (median odds ratio 3.28 (95% CI 2.81-3.83)).

CONCLUSIONS

Practice patterns for adjunctive therapies to norepinephrine during septic shock are variable and are determined in large part by hospital of admission. These results inform several future studies seeking to improve septic shock management.