Case volume and specialization in critically ill emergency patients: a nationwide cohort study in Japanese ICUs

Table 3 Odds ratios for hospital mortality

	Model 1	Model 2	Model 3
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Random effects
Median odds ratio	1.40 (1.32–1.49)	1.07 (1.02–1.12)	1.07 (1.04–1.12)
ICU level variance, (SD)	0.12 (0.02)	0.0056 (0.0038)	0.0054 (0.0031)
Fixed effects
ICU-level variables
Quartile of ICU admissions
First quartile		Ref.
Second quartile		0.96 (0.92–1.01)
Third quartile		0.92 (0.88–0.95)
Fourth quartile		0.93 (0.88–0.99)
Emergency patients as % of total ICU admissions
< 25		Ref.
≥ 25 to < 50		0.97 (0.93–1.02)
≥ 50 to < 75		0.99 (0.95–1.02)
≥ 75		0.83 (0.80–0.86)
Admission quartiles and percentage of emergency patients
First quartile, < 75%			Ref.
Second quartile, < 75%			0.95 (0.92–0.98)
Third quartile, < 75%			0.93 (0.89–0.98)
Fourth quartile, < 75%			0.92 (0.89–0.96)
First quartile, ≥ 75%			NA^a
Second quartile, ≥ 75%			0.92 (0.88–0.96)
Third quartile, ≥ 75%			0.70 (0.67–0.73)
Fourth quartile, ≥ 75%			0.78 (0.73–0.83)
Type of hospital
University hospital		Ref.	Ref.
Non-university hospital		1.00 (0.97–1.04)	0.98 (0.94–1.01)
Number of intensivists per ICU bed		1.06 (1.03–1.10)	1.05 (1.01–1.10)
Number of nurses per ICU bed		0.98 (0.96–1.00)	0.98 (0.96–1.00)
Quartile of hospital beds
First quartile		Ref.	Ref.
Second quartile		1.02 (1.00–1.04)	1.01 (0.96–1.07)
Third quartile		1.02 (0.98–1.07)	1.05 (1.00–1.10)
Fourth quartile		0.99 (0.96–1.01)	0.99 (0.95–1.03)

Odds ratios were calculated using a multilevel logistic regression model, allowing for a random effect (a random intercept) model for each ICU. We adjusted ICU-level and patient-level variables as follows: age, sex, BMI (< 18.5, 18.5 to 23, 23 to 27.5, ≥ 27.5), the Japan Risk of Death score, diagnoses at admission and after cardiac resuscitation, emergency surgery, hospitalization period (from FY 2015 through FY 2019, from FY 2020 through FY 2021), number of nurses per ICU beds, number of intensivists per ICU beds, quartile of hospital beds, and type of hospital (university hospital or non-university hospital). The odds ratios for in-hospital mortality associated with patient-level variables are detailed in the Supplementary Table 2
SD standard deviation, OR odds ratio, CI credible interval, Ref. reference, NA not applicable
^aAll first quartile participants were enrolled in facilities with less than 75% emergency patients

ISSN: 2052-0492