Introduction

Alcohol use and dependence disorders are associated with chronic health problems and cost more than $184 billion dollars per year in the United States [1]. These disorders affect more than 15% of hospitalized patients [2, 3]. The rate of prior alcohol dependence in critically ill patients may be as high as 20–39% [4–7], and 5% of intensive care unit (ICU) admissions are related to illicit drug use [8]. The contribution of substance dependence and psychiatric disorders on ICU patient mortality and morbidity is unknown. The purpose of this study was to compare mortality and discharge disposition in ICU patients with and without substance dependence or psychiatric disorders. Additionally, we compared the prevalence of pre-existing alcohol or illicit drug dependence and psychiatric disorders with reported population data. Preliminary data of this study were previously published in abstract form [9].

Patients and methods

We retrospectively reviewed all medical records of patients admitted to the LDS Hospital shock trauma and respiratory intensive care unit (STRICU, a mixed medical and surgical unit) between 1 July 2003 and 30 June 2004. The patient electronic medical records reviewed included all longitudinal inpatient, psychiatric, rehabilitation, and outpatient records. Data obtained were patient age, sex, admission diagnosis, mechanical ventilation, length of ICU and hospital stay, mortality, sequential organ failure assessment (SOFA) score [10], acute physiology and chronic health evaluation II (APACHE II) score [11], comorbid illnesses, blood or urine ethanol levels, presence of acute respiratory distress syndrome (ARDS), and discharge disposition. This study had IRB approval and informed consent was waived.

Alcohol dependence and psychiatric disorders were defined using Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) criteria [12]. Substance dependence refers to the combined drug and alcohol dependence and is defined by the DSM-IV as a pattern of substance use leading to clinically significant impairment, with symptoms of tolerance and withdrawal. A pre-existing diagnosis of alcohol dependence, drug dependence or psychiatric disorder was confirmed if documented: (1) on admitting note, (2) by treating physician in any record, or (3) in psychiatric notes.

Statistics

Descriptive statistics were calculated and comparisons between groups were carried out using ANOVA and Z-tests for continuous data and Fischer’s exact tests and chi squares for categorical data. Separate analyses compared: (1) patients with and without substance dependence, (2) patients with and without psychiatric disorders, and (3) patients with both psychiatric and substance dependence to patients with psychiatric disorders.

Multivariable logistic and linear regressions assessed whether psychiatric disorders or substance dependence predicted: (1) hospital mortality, (2) ICU length of stay (LOS), (3) hospital LOS, (4) SOFA scores, and (5) APACHE II scores, using sex, age, trauma, and mechanical ventilation as covariates.

Z-tests for two proportions compared the prevalence of alcohol dependence, substance dependence, and psychiatric disorders in our patients with Utah or US population data, where available. Population data were obtained from the substance abuse and mental health services administration [13]. Psychiatric disorder prevalence was compared to the reported lifetime prevalence data [14]. Standard error for proportions were calculated using the formula \( {\text{SE}}\, = \,\sqrt {\frac{p(1 - p)}{n}} . \) Confidence intervals were calculated using the formula CI = p ± (SE for proportions) (1.96).

Results

There were 742 critically ill patients with a mean age of 49.6 ± 19.2 years, 54% were male, 5.5% developed ARDS, and hospital mortality of 21%. The mean APACHE II scores were 16.5 ± 7.9, SOFA scores were 6.7 ± 4.2, duration of mechanical ventilation was 5.0 ± 6.2 days, ICU LOS was 7.3 ± 10.1 days, and hospital LOS was 12.3 ± 12.9 days.

Substance dependence

Substance dependence occurred in 137 (19%) patients. Ethanol was detected in 40/137 patients (mean 82.4 ± 113 mg/dL). Patients with substance dependence were younger (45 vs. 51 years, p = 0.001), male (67 vs. 51%, p = 0.002), had a lower prevalence of ARDS (2 vs. 6%, p = 0.01), and shorter hospital stay (10 vs. 13 days, p = 0.004) compared to patients without substance dependence. Admission diagnosis was associated with substance dependence (χ 2 = 89.36, p < 0.001) with a lower rate of sepsis, stroke, and cancer and a higher rate of drug overdose and gastrointestinal disease compared to patients without substance dependence. There was no difference in mortality or discharge disposition between the groups.

Psychiatric disorders

The prevalence of psychiatric disorders was 19% (137 patients; Fig. 1). A higher percentage of patients with psychiatric disorders were female (56 vs. 43%, p = 0.002). Psychiatric disorders were associated with admission diagnosis (χ 2 = 77.81, p < 0.001), a lower rate of trauma and liver failure, and a higher rate of drug overdose and pneumonia compared to patients without psychiatric disorders. Patients with psychiatric disorders had a higher rate of congestive heart failure but a lower rate of liver failure compared to patients without psychiatric disorders. There was no difference in mortality for patients with psychiatric disorders compared to those without disorders. Discharge disposition was associated with psychiatric disorder (χ 2 = 36.07, p < 0.001), but when discharge to psychiatric unit was excluded, the association was not significant (χ 2 = 6.58, p = 0.25).

Fig. 1
figure 1

The prevalence of psychiatric disorders by disorder in critically ill patients (black bars) compared to United States population data (gray bars). Significant differences are denoted with * for p values of <0.01 and ** for p values of <0.001. Error bars denote the standard error of the percentages (proportions)

Full size image

Prediction of outcomes

The regression analyses found psychiatric disorders predicted hospital mortality (Wald statistic = 3.05, odds ratio = 1.50, p = 0.08) and substance dependence predicted hospital LOS (R 2 = 0.08, p = 0.01), after controlling for covariates.

Substance dependence and psychiatric disorder

A higher percentage of patients with substance dependence were male (72 vs. 53%, p = 0.03), discharged to a rehabilitation unit (22 vs. 5%, p = 0.02), had a higher mortality rate (61 vs. 24%, p = 0.001), and higher APACHE II scores (18.2 vs. 14.8, p = 0.05) compared to patients with combined substance dependence and psychiatric disorders.

Substance dependence and trauma

Admission diagnosis of trauma was associated with alcohol dependence (p = 0.002), but not drug dependence (p = 0.23) or any substance (drug or alcohol) dependence (p = 0.12). Trauma patients had higher ethanol levels (121.3 ± 124.8 mg/dL) compared to all other patients (36.9 ± 76.5 mg/dL, t = 3.95, p < 0.001).

Population data comparisons

The prevalence of drug or alcohol dependence was higher in our patients compared to Utah (Fig. 2) and US population data (all p values <0.001). The prevalence of psychiatric disorders was lower in our patients compared to US population data (18.5 vs. 22.4%, p < 0.001; Fig. 1). The prevalence of anxiety (4.2 vs. 18.1%, p < 0.001) and posttraumatic stress disorder (0.1 vs. 3.5%, p < 0.01) was lower in our patients compared to US population data, whereas the prevalence of depression (14.3 vs. 6.7%, p < 0.01) was higher.

Fig. 2
figure 2

The prevalence of substance dependence and psychiatric disorders in critically ill patients (black bars) compared to Utah population data (gray bars). Significant differences are denoted with * for p values of <0.01 and ** for p values of <0.001. Error bars denote the 95% confidence intervals for the percentages (proportions)

Full size image

Discussion

Patients with substance dependence differed from our general medical surgical ICU population as they were younger, more often male, had an admission diagnosis of trauma or drug overdose, had a shorter hospital LOS, and lower incidence of ARDS and comorbid illnesses compared to patients without substance dependence. Patients with psychiatric disorders were more often female, had comorbid illnesses, and had an admission diagnosis of drug overdose. None of these differences were associated with mortality or discharge disposition, except in patients with combined psychiatric disorder plus substance dependence, who had a lower mortality and were less likely to be discharged to a rehabilitation unit. When controlling for covariates, mortality was predicted by the presence of psychiatric disorders but not by drug dependence. Hospital LOS was predicted by substance dependence but not psychiatric disorders when controlling for covariates.

The rate of alcohol and substance dependence in our patients fell between the incidence reported by O’Brien (12%) [1] and Uusaro (24%) [7]. However, the prevalence of alcohol and substance dependence, but not psychiatric disorders was two times higher in our patients than occurs in US population data [13, 14]. The rate of alcohol and substance dependence in our patients was twice the rate observed in the US population [13].

Alcohol dependence was associated with an ICU admitting diagnosis of trauma, similar to that reported by Corrigan [15] and Bombardier [16]. Drug dependence was not associated with a diagnosis of trauma, which is different from previous data [15, 16]. While there was a correlation between trauma and alcohol dependence, 57% of our patients with alcohol dependence history were medical ICU patients. Early recognition of substance dependence may allow intervention prior to ICU or hospital discharge, and presumably, aid in recovery.

One limitation is that we likely underestimated the prevalence of alcohol dependence, drug dependence, and psychiatric disorders. The completeness of our electronic medical record reduced potential for under reporting of these disorders, but our study was retrospective. While prospective studies that screen all ICU admits for such disorders are needed, this limitation may be difficult to remedy, as it is often problematic to get patients to admit to such pre-existing disorders [2, 17, 18]. While it is a standard practice to screen for alcohol and drug use in trauma patients at Level 1 trauma centers, trauma patients represent a fraction of patients admitted to ICUs, so to obtain accurate estimates, all ICU patients would need to be screened.

A second limitation is that our data may not be representative of specialty ICUs (e.g., cardiac ICUs) but rather more representative of patients admitted to a mixed ICU of medical and surgical patients. Our STRICU is comparable to a general medical ICU in many centers, with one-third of the patients admitted for trauma and two-thirds are medical patients.

Our data suggest that substance dependence increases hospital LOS and that patients with drug or alcohol dependence are at higher risk for ICU admission compared to the general population.