Sex differences in crude mortality rates and predictive value of intensive care unit-based scores when applied to the cardiac intensive care unit

Abstract

Background:

Limited data exists regarding sex differences in outcome and predictive accuracy of intensive care unit-based scoring systems when applied to cardiac intensive care unit patients.

Methods:

We reviewed medical records of patients admitted to cardiac intensive care unit from 1 January 2011–31 December 2016. Sex differences in mortality rates and the performance of intensive care unit-based scoring systems in predicting in-hospital mortality were analyzed. Calibration was assessed by the Hosmer-Lemeshow test and locally weighted scatterplot smoothing curves. Discrimination was assessed using the c statistic and receiver-operating characteristic curve.

Results:

Among 6963 patients, 2713 (39%) were women. Overall in-hospital and cardiac intensive care unit mortality rates were similar in women and men (9.1% vs 9.4%, p=0.67 and 5.9% vs 6%, p=0.88, respectively) and in age and major diagnosis subgroups. Of the scoring systems, Acute Physiology and Chronic Health Evaluation III and Sequential Organ Failure Assessment had poor calibration (Hosmer-Lemeshow p value <0.001), while Simplified Acute Physiology Score II performed better (Hosmer-Lemeshow p value 0.09), in both women and men. All scores had good discrimination (C statistics >0.8). In the subgroups of acute myocardial infarction and heart failure patients, all scores had good calibration (Hosmer-Lemeshow p>0.001) and discrimination (C statistic >0.8) while in diagnosis subgroups with highest mortality, the calibration varied among scores and by sex, and discrimination was poor.

Conclusions:

No sex differences in mortality were seen in cardiac intensive care unit patients. The mortality predictive value of intensive care unit-based scores is limited in both sexes and variable among different subgroups of diagnoses.

Background

Initially developed for the management of life-threatening complications after acute myocardial infarction (AMI),¹ cardiac intensive care units (CICUs) have changed at a rapid pace over the last decades.^2–4 Challenges seen today in CICUs include increasingly older patients with a high burden of comorbidities; complications such as renal, respiratory, or infectious illnesses that necessitate a multidisciplinary approach in the acute phase;^2–4 use of advanced therapeutic options in phases of disease previously considered incurable and differences in acute cardiovascular disease (CVD) between women and men.^5,6 The complexity of patients seen in intensive care settings has led to the use of step-wise, multivariable models for illness severity and risk assessment. Models developed in the general intensive care unit (ICU) population have been used to predict outcomes for CICU patients. These include ICU scoring systems such as the Acute Physiology and Chronic Health Evaluation (APACHE) III,⁷ Simplified Acute Physiology Score (SAPS) II⁸ and Sequential Organ Failure Assessment (SOFA).⁹ These scoring systems use various combinations of demographic, clinical, laboratory, physiological, and comorbidity data to calculate a severity score. The score obtained is then used to predict in-hospital mortality and, in some cases, length of stay (LOS).^7,10,11 No single score has been proven superior to another with regards to the ability to predict mortality, ease of use, and effect on costs.^12,13 APACHE-based systems, however, tend to be more accurate in predicting mortality. They are cumbersome as they require extensive data collection and require proprietary software (http://www.apache.org/foundation/license). SAPS II and SOFA are easier to use but do not predict LOS and are more susceptible to case-mix effects.⁸

Furthermore, these scores were developed in cohorts of ICU patients composed largely of men (63% men in the SOFA original cohort).⁹ None of the scores includes sex as a variable. The influence of sex on each score’s ability to predict outcome has not been well-defined. One study showed that SOFA was more predictive of mortality in women then in men.¹⁴

In the ICU setting, sex differences, with higher mortality in women as compared to men, were shown in patients with AMI,^15,16 cardiac arrest (CA)¹⁷ and after coronary artery bypass grafting (CABG) surgery.¹⁸ Although women hospitalized for a first AMI were less likely to be admitted to a critical care unit, they had a higher age-adjusted mortality than men when admitted to an ICU or CICU.¹⁹

Due to prior evidence of sex differences in outcomes of patients with acute CVD in ICU settings,^15–18 the higher CVD mortality in women compared to men,^5,20–26 and limited knowledge regarding sex influence on predictive value of ICU-based scores, our aim was to evaluate sex-related differences in mortality and evaluate the performance of three commonly used ICU-based scoring systems in predicting in-hospital mortality in a modern CICU population.

Methods

We reviewed electronic medical records (EMRs) of patients admitted to the CICU at Cedars-Sinai Medical Center, an academic tertiary care center in Los Angeles, California, USA, from 1 January 2011–31 December 2016. We extracted data from the EMR that included: sex, age, race, diagnosis per hospitalization, comorbid illness, in-hospital complications, LOS, physiological and laboratory results, Glasgow coma scale, critical care therapies and mortality.. Diagnoses per CICU hospitalization and for comorbidities were extracted by International Classification of Disease-9 codes.

The APACHE III and SAPS II score assess disease severity within the first 24 h of admission and include patient age and chronic health conditions.^7,8 The SOFA score assesses the severity of acute organ failure and can be measured at different time points during admission.⁹ For each score calculation the worst value during the first 24 h following admission was used.

We analyzed the mortality rates expressed in crude percentages of mortality (CICU and in-hospital mortality) in women as compared to men in the overall CICU population and in subgroups of major diagnoses. Sex differences in mortality rates for patients presenting with AMI were analyzed by age.

We assessed the performance of three of the most commonly used ICU-based scoring systems, APACHE III, SAPS II, and SOFA in predicting crude in-hospital mortality in women and men, when applied to the overall CICU cohort and subgroups of major diagnoses by assessing each score’s calibration and discrimination. Calibration assesses the degree of correspondence between the estimated probability of outcome (in-hospital mortality) and the rates of outcomes (crude percentages of mortality) observed. Discrimination assesses the ability of a model to distinguish patients who died from those who survived. The Hosmer-Lemeshow (HL) goodness of fit test was used for assessment of calibration with a p value <0.05 indicating that the model has poor calibration. Locally weighted scatterplot smoothing (LOESS) calibration was built for each score. The discriminative capability of each score was assessed using the C statistic and receiver-operating characteristic (ROC) curve, method of DeLong et al.²⁷ The statistical software used was SAS (version 9.4, SAS Institute, Cary, North Carolina, USA).

Results

Among 6965 patients hospitalized in the CICU, 2713 (39%) were women. Women were older, less likely to have a history of ischemic heart disease (IHD), revascularization, or dyslipidemia (Tables 1–3). Women more frequently had a history of hypertension, anemia, or associated respiratory or malignant diseases. Higher APACHE III scores were seen in women while men had higher SOFA scores.

Table 1.

Baseline characteristics in cardiac intensive care unit (CICU) patients.

Variable	Women n, (%) n=2713 (39)	Men n, (%) n= 4252 (61)	p-Value
Age at admission±SD, years	72.9±16.3	68.9±16.1	<0.0001
Race			<0.0001
White	1917 (70.7)	3183 (74.9)
Black	456 (16.8)	488 (11.5)
Asian	156 (5.7)	252 (5.9)
Native American	5 (0.2)	9 (0.2)
Other/unknown/refused	179 (6.6)	318 (7.5)
CICU LOS, mean±SD	2.98±4.43	3.03±3.96	0.62
Hospital LOS, mean±SD	9.54±15.11	10.42±16.65	0.023
Hospital LOS, median (IQR)	5.0 (2.9–10.1)	4.9 (2.5–11.3)	0.49
Discharge to home	1922 (70.8)	3181 (74.8)	0.0002
IABP	58 (2.1)	138 (3.2)	0.006
Vasopressors use	572 (21.1)	1134 (26.7)	<0.0001
Transfusion	427 (15.7)	657 (15.5)	0.76
Renal replacement therapy	163 (6.0)	306 (7.2)	0.056
Calculated APACHE III, mean±SD	51.3±19.8	49.5±20.8	0.0002
Calculated SAPS II, mean±SD	42.6±18.1	41.8±19.1	0.072
Calculated SOFA, mean±SD	3.8±3.3	4.1±3.4	0.0007

APACHE: Acute Physiology and Chronic Health Evaluation; IABP: intra-aortic balloon pump; IQR: interquartile range; LOS: length of stay; SAPS: Simplified Acute Physiology Score; SD: standard deviation; SOFA: Sequential Organ Failure Assessment.

Table 3.

Comorbidities by International Classification of Disease-9 (ICD-9) categories.^a

Variable	Women n (%) 2397 (39%)	Men n (%) 3788 (61%)	p-Value
History of IHD	493 (20.6)	1096 (28.9)	<0.0001
HTN	393 (16.4)	518 (13.7)	0.004
Dyslipidemia	184 (7.7)	382 (10.1)	0.001
Family history of IHD	2 (0.1)	3 (0.1)	>0.99
Chronic renal failure	28 (1.2)	56 (1.5)	0.37
History of stroke	17 (0.7)	26 (0.7)	>0.99
PAD	12 (0.5)	20 (0.5)	>0.99
History of PCI	23 (1.0)	105 (2.8)	<0.0001
History of CABG	27 (1.1)	80 (2.1)	0.004
History of HF	590 (24.6)	1001 (26.4)	0.114
Diabetes	221 (9.2)	359 (9.5)	0.75
Respiratory	364 (15.2)	494 (13.0)	0.019
Anemia	192 (8.0)	244 (6.4)	0.022
Malignancy	77 (3.2)	85 (2.2)	0.022

CABG: coronary artery bypass grafting; CICU: cardiac intensive care unit; HF: heart failure; HTN: hypertension; IABP: intra-aortic balloon pump; IHD: ischemic heart disease; LOS: length of stay.

^aComorbid diagnoses were available for 6185 patients (89% of cohort).

More men than women were hospitalized for a diagnosis of ST-elevation myocardial infarction (STEMI), non-ST elevation myocardial infarction, or shock. Women were more likely to have a primary diagnosis of respiratory or hematological disease. No sex differences were seen in the frequency of other major CVD diagnosis groups, such as valvular diseases, conduction disorders, arrhythmias, or heart failure (HF).

The mean CICU LOS was 3.01±4.15 days, with no significant differences seen between women and men. The use of critical care specific therapies (such as intra-aortic balloon pump or vasopressors) was lower in women when compared to men. No sex differences were seen in the use of renal replacement therapy or blood transfusions. Fewer women were discharged to home and the 30-day readmission rate was similar among sexes.

Sex differences in crude mortality rates.

There were 647 (9.3%) in-hospital deaths and 418 (6%) CICU deaths. In-hospital and CICU mortality rates were similar in women when compared to men overall (Table 4) and across all age groups (≤50 years; 50–65 years; ≥65 years) (Supplementary Material Table S1). There was a non-significant trend (p=0.06) to a higher CICU mortality rate in women younger than 50 years (7.2%) compared to aged matched men (3.9%). The in-hospital mortality rate was similar between these two groups (10.3% in women vs 8.8% in men, p=0.52). The sex difference in CICU mortality in this age group was not seen in patients hospitalized with a diagnosis of AMI, HF, or cardiac arrest, but was seen in the group of patients hospitalized for other CVD diagnoses.

Table 4.

In-hospital and cardiac intensive care unit (CICU) crude mortality rates.

	Group of disease	Women n (%)	Men n (%)	p-Value
In-hospital mortality	Overall	247 (9.1)	400 (9.4)	0.67
	AMI	48/478 (10)	105/943(11.1)	0.60
	NSTEMI	26/267 (9.7)	57/488 (11.7)	0.47
	STEMI	22/211 (10.4)	48/455 (10.5)	>0.99
	Heart failure	115/1217(9.5)	222/1942(11.4)	0.09
	Cardiogenic shock	65/183 (35.5)	159/455 (34.9)	0.93
	Shock, other	54/108 (50.0)	105/220 (47.7)	0.73
	Cardiac arrest	64/135 (47.1)	108/214 (50.5)	0.58
CICU mortality	Overall	161 (5.9)	257 (6.0)	0.88
	AMI	35/478 (7.3)	83/943 (8.8)	0.36
	NSTEMI	15/267 (5.6)	43/488 (8.8)	0.15
	STEMI	20/211 (9.5)	40/455 (8.8)	0.77
	Heart failure	79/1217 (6.5)	140/1942 (7.2)	0.47
	Cardiogenic shock	47/183 (25.7)	112/455 (24.6)	0.84
	Shock, other	43/108 (39.8)	62/220 (28.2)	0.044
	Cardiac arrest	49/135 (36.3)	81/214 (37.9)	0.82

AMI: acute myocardial infarction; NSTEMI: non-ST elevation myocardial infarction; STEMI: ST elevation myocardial infarction.

Among patients hospitalized for AMI, HF, or cardiac arrest, women and men had similar in-hospital and CICU mortality rates across all age groups (Supplementary Material Table S1). Women hospitalized for shock, other than cardiogenic shock, had higher CICU mortality as compared to men but had similar in-hospital mortality.

A separate analysis showed no sex differences in mortality by body mass index (BMI) categories, except a higher in-hospital mortality in underweight men but not CICU mortality (Supplementary Material Tables S3 and S4 ). In a logistic regression model, BMI was not a predictor of in-hospital or CICU death (p values 0.77 and 0.42, respectively).

Among both women and men hospitalized in CICU, the lowest in-hospital and CICU mortality were seen in patients with AMI or HF. High mortality was seen, in ascending order, among women and men with a diagnosis of cardiogenic shock, shock other than cardiogenic, and cardiac arrest.

Performance of ICU-based scoring systems in predicting outcome

Calibration.

When applied to overall patients hospitalized in CICU, in both women and men, APACHE III and SOFA scores had poor calibration (HL p value <0.001), either under or over-estimating risk as per LOESS calibration curves (Figure 1(a) and (b) ). In both women and men, only SAPS II was well-calibrated (Table 5 and Figure 1(a) and (b)).

Figure 1.

Locally weighted scatterplot smoothing (LOESS) calibration curves for overall CICU patients: (a) women and (b) men. Blue line: Acute Physiology and Chronic Health Evaluation (APACHE); red line: Simplified Acute Physiology Score (SAPS); green line: Sequential Organ Failure Assessment (SOFA). Model calibration that has complete concordance between estimated and observed risk (dashed line).

Table 5.

Predictive significance of intensive care unit (ICU)-based scoring systems in cardiac intensive care unit (CICU) patients, by sex and subgroups of disease.

Group of disease	Severity of illness score	Women		Men
		Hosmer-Lemeshow p-value	C statistic	Hosmer-Lemeshow p-value	C statistic
Overall CICU	APACHE III	<0.001	0.815	0.02	0.826
	SAPS	0.17	0.845	0.17	0.853
	SOFA	<0.001	0.842	<0.001	0.848
AMI	APACHE III	0.010	0.857	0.021	0.881
	SAPS	0.21	0.883	0.34	0.914
	SOFA	0.56	0.814	0.20	0.887
Heart failure	APACHE III	0.47	0.835	0.82	0.787
	SAPS	0.50	0.864	0.34	0.847
	SOFA	0.78	0.779	0.78	0.830
Cardiogenic shock	APACHE III	0.02	0.736	0.54	0.706
	SAPS	0.005	0.759	0.30	0.756
	SOFA	0.18	0.657	0.06	0.700
Shock, other	APACHE III	0.13	0.750	0.38	0.527
	SAPS	0.11	0.700	0.44	0.637
	SOFA	0.31	0.644	0.07	0.518
Cardiac arrest	APACHE III	0.43	0.617	0.50	0.706
	SAPS	0.49	0.661	0.29	0.753
	SOFA	0.07	0.577	0.004	0.720

AMI: acute myocardial infarction; APACHE: Acute Physiology and Chronic Health Evaluation; SAPS: Simplified Acute Physiology Score; SOFA: Sequential Organ Failure Assessment.

In contrast, in women and men hospitalized for AMI or HF, all three ICU-based scores had good calibration. In patients hospitalized for cardiogenic shock, APACHE III and SAPS II had poor calibration in women, but good calibration in men (Table 5 and Figure 2(a) and (b)). SOFA had good calibration in women hospitalized for cardiac arrest but not in men with CA. SAPS II was the only score with consistently good calibration in overall CICU patients and in the majority of diagnosis subgroups, except in women with cardiogenic shock.

Figure 2.

Locally weighted scatterplot smoothing (LOESS) calibration curves for patients with cardiogenic shock: (a) women and (b) men. Model calibration that has complete concordance between estimated and observed risk (dashed line). Blue line: Acute Physiology and Chronic Health Evaluation (APACHE); red line: Simplified Acute Physiology Score (SAPS); green line: Sequential Organ Failure Assessment (SOFA).

Discrimination.

All three scores had good discrimination in the overall CICU population, for both women and men (Table 5). ROCs for in-hospital mortality in overall CICU patients for APACHE, SAPS II, and SOFA were 0.822, 0.850, and 0.845, respectively. SAPS II and SOFA had a significantly higher ROC as compared to APACHE (difference in ROC 0.0281, p<0.0001 and 0.0236, p=0.002, respectively). Similar results were seen in women and men. Good discrimination was also seen in women and men hospitalized for AMI or HF (Table 5).

In contrast, among patients with the highest mortality such as those hospitalized for cardiogenic shock, shock from other etiology than cardiogenic and cardiac arrest, the discriminative value of all three scores was poor and variable among women and men (highest C statistic of 0.759 for SAPS II in women with cardiogenic shock and lowest C statistic of 0.51 for SOFA in men with shock other than cardiogenic) (Table 5). In patients with cardiogenic shock ROCs for APACHE, SAPS II, and SOFA were 0.752, 0.775, and 0.747 respectively, with similar results seen in women and men.

Discussion

Our study has two major findings. First, in a large, contemporary CICU patient population at a tertiary care academic center, no significant sex differences in mortality rates (CICU and in-hospital mortality, expressed as crude percentages of mortality) were observed, across a variety of age groups and major diagnosis subgroups, except in those with a primary diagnosis of shock other than cardiogenic.

Second, the performance of three routinely used ICU-based scoring systems in predicting in-hospital mortality in CICU patients was limited in both women and men, especially in those subgroups of disease with the highest mortality.

Prior work highlighted significant sex and gender differences in clinical presentation, treatment, complications and in both short and long-term outcome in patients with AMI, with women receiving less guideline medical therapy and having higher mortality rates as compared to men.^5,20–26 A complex interplay between sex (biologically determined) and gender (feminine roles and personality traits) also has been shown to be linked to prognosis after AMI,²⁸ predominance of feminine characteristics being related to poorer outcome. In the ICU setting, sex differences in outcome were reported in patients with a primary diagnosis of CVD, with women having a higher mortality after CABG¹⁸ or AMI.^15,16 In one recent study, higher age-adjusted mortality in women hospitalized in CCU/ICU or medical wards was no longer seen after further adjustment for presenting symptoms, delayed arrival and other risk factors.¹⁹ Similarly, sex disparities in management and outcome after STEMI were diminished when standardized protocols and systems of care were used.^29–31 Nevertheless, mortality rates remain higher in young women with STEMI as compared to men.^30,32 A Swedish study looking at mortality post-ICU care did not find a survival advantage in premenopausal women, and male sex was associated with a lower mortality rate in patients with cardiac arrest (female: male odds ratio (OR), 1.34; 99% confidence interval (CI) 1.11 to 1.63; p<0.001).¹⁷ In contrast, a systematic review and meta-analysis showed that women had better survival after CA despite being more likely to present with CA at home and having more often an unshockable initial rhythm.³³ In our analysis women younger than 50 years had a nearly significant higher CICU mortality, but not in-hospital mortality. While prior data highlighted higher mortality in women with CVD, our study found that this higher CICU mortality did not seem to be driven by those women hospitalized for a primary CVD diagnosis, since there were similar mortality rates between women and men across diagnosis subgroups such as AMI, HF, cardiogenic shock, or CA. Only in the subgroup of patients with a diagnosis of shock other than cardiogenic, did women have significantly higher mortality. Also, when classified by BMI categories, the mortality rates were similar among women and men, except for being higher in hospital mortality in underweight men. Although prior work showed an “obesity paradox” in patients with coronary heart disease^34,35 and, furthermore, higher BMI was associated with better prognosis in critically ill patients, BMI was not predictive of mortality in overall CICU patients with a diversity of cardiovascular diagnoses. This might be explained by the limitation of BMI to reflect the exact metabolic status of patients. BMI assessment includes only height and weight and not parameters such as central obesity, waist-to-hip ratio, or others known to be more reflective of metabolic status and reserve and linked with worse cardiovascular outcomes.³⁵

The modern CICU comprises a large spectrum of complex clinical scenarios. Our study is unique in exclusively evaluating CICU hospitalized patients and exploring sex differences in outcomes. Possible explanations for the lack of sex disparities across a spectrum of CVD diagnoses are: highly standardized treatment protocols used in CICU providing horizontal and vertical equity in women and men,³⁶ and our center’s experience in women’s heart disease and the increased awareness of CVD among women.³⁷ Contemporary with our study, in 2016, for the first time since 1983, a change in CVD mortality trajectories was seen (more males (402,851) died of CVD than females (398,086)).³⁸

The limited performance of the three ICU based scores in predicting outcome in women and men can be related to: (a) the scores originally being developed to compare quality of patients care across ICUs and not mortality prediction; (b) time elapsed since development of scores and (c) differences in patient population between ICUs and CICUs. In subgroups of diseases with the three scores’ performance was different in women as compared to men, and sex may influence variables included in the scores’ calculation. Among the three scores, SAPS II score had better calibration in overall CICU patients with good discrimination. The SAPS II score was developed in a relatively more contemporary set of patients. Nevertheless, when applied to subgroups of major diagnoses, especially those with highest mortality (e.g. cardiogenic shock), its calibration and discrimination are low, especially in women, making it difficult for it to be considered a good predictive score in individual patients. A recent study found that SOFA had good discrimination for short-term mortality in CICU patients when done sequentially to include day-to-day changes in these dynamic patients.³⁹ Also in our study, one-step early risk assessment done in the first 24 h showed good discrimination for all scores, including SOFA, in overall CICU patients. Nevertheless, the calibration was poor for APACHE and SOFA. A sequential risk assessment might improve prognostication in both women and men hospitalized in CICU as trends in their status or sudden changes might be better captured.

In contrast to our findings, a prior study showed that APACHE II and SOFA could predict in-hospital mortality with good calibration (HL p values 0.408 and 0.282, respectively) and strong discriminative value (AUC 0.90 and 0.91, respectively) in overall CICU patients and in those hospitalized for acute coronary syndromes (ACS).⁴⁰ Nevertheless, in that study, patients with ACS represented a majority (69%) of all CICU patients and had a mortality of 20% while the overall in-hospital mortality in CICU patients was 18%. In our cohort, patients with a diagnosis of AMI comprised only 20% of total CICU population and both AMI-related and overall in-hospital mortality were lower (11% and 9.3%, respectively), reflecting important changes in CICU patients’ profiles over the last decade. Other customized risk scores such as the Global Registry of Acute Coronary Events (GRACE)²⁶ and Thrombolysis in Myocardial Infarction (TIMI) risk score⁴¹ were previously developed to predicted in-hospital and 30 days mortality, respectively, in patients with AMI. The caveat of these scores is that cannot be applied to patients with other diagnoses admitted to CICU and even in the AMI patients they cannot be applied to those with cardiogenic shock or AMI precipitated or accompanied by a significant comorbidity, trauma, or surgery, as these were excluded from the original cohorts in which the scores were developed. A new recent score, called Observatoire Regional Breton sur l’Infarctus (ORBI), was developed in patients with STEMI for prediction of related cardiogenic shock⁴² but not for in-hospital mortality. While other studies that applied ICU-based scores to CICU patients also reported poor calibration but good discrimination^12,43,44 they included less contemporary CICU patients and reported a much higher overall and AMI-related in-hospital mortality rate as compared to our study. The performance of scores related to sex or subgroups of disease was not explored.

Due to the limited value in predicting outcome, our findings support the development of tailored CICU risk score models, inclusive of all representative clinical scenarios for appropriate outcome prediction and support for therapeutic decision-making. Whether sex changes how the other variables account for total risk estimation, remains to be explored.

Limitations

This was a retrospective, single-center study in a tertiary academic center. The results might not be generalizable to other settings.

We cannot exclude sex differences in other outcomes besides in-hospital and CICU mortality, as these were not captured in our analysis. We also were limited in assessing the cause of death (cardiac or non-cardiac). The precision of statistical methods which evaluate the scores’ performance is related to sample size. And, although the precision might be lower in subgroups of diagnoses (smaller sample size), we were able to see a signal for sex differences in the scores’ performance, that will need to be verified in larger sample size studies.

Conclusions

In the contemporary standardized CICU setting, no sex differences in crude in-hospital and CICU mortality rates were seen across a wide range of CVD diagnoses. The outcome-predictive value of an ICU-based scoring system was limited in both women and men and variable among different subgroups of diagnoses, emphasizing the need to address the influence of sex and to develop specific CICU risk score models.

Table 2.

Cardiac intensive care unit (CICU) discharge diagnosis, by International Classification of Disease-9 (ICD-9) categories.^a

Variable	Women n (%) 2133 (39.3)	Men n (%) 3290 (60.7)	p-Value
Unstable angina	65 (3.0)	110 (3.3)	0.58
AMI	267 (12.5)	488 (14.3)	0.016
NSTEMI STEMI	211 (9.9)	455 (13.8)	<0.0001
Pulmonary embolism	35 (1.6)	57 (1.7)	0.83
Myocarditis	5 (0.2)	16 (0.5)	0.18
Endocarditis	11 (0.5)	31 (0.9)	0.083
Valvular disease	756 (35.4)	1102 (33.5)	0.143
Conduction disorder	392 (18.4)	626 (19.0)	0.57
Cardiac dysrhythmias	364 (17.1)	512 (15.6)	0.15
Cardiac arrest	135 (6.3)	214 (6.5)	0.82
Heart failure	1217 (57.1)	1942 (59.0)	0.15
Cardiogenic shock	183 (8.6)	455 (13.8)	<0.0001
Shock	108 (5.1)	220 (6.7)	0.014
Cardiomyopathy	290 (13.6)	572 (17.4)	0.0002
Respiratory	1068 (50.1)	1525 (46.4)	0.008
Infection	575 (27.0)	844 (25.7)	0.30
Endocrine	1902 (89.2)	2904 (88.3)	0.31
Hematological	1196 (56.1)	1680 (51.1)	0.0003
Neurological	628 (29.4)	973 (29.6)	0.93
Gastroenterology	936 (43.9)	1366 (41.5)	0.086
Genitourinary	1085 (50.9)	2210 (67.2)	<0.0001

AMI: acute myocardial infarction; NSTEMI: non-ST elevation myocardial infarction; STEMI: ST elevation myocardial infarction.

^aCICU discharge diagnoses were available for 5423 patients (78% of cohort). Percentages add up to more than 100% as patients could have more than one CICU diagnosis.