Association between blood alcohol concentration and mortality in critical illness

Journal of Critical Care 30 (2015) 1382–1389

Contents lists available at ScienceDirect

Journal of Critical Care journal homepage: www.jccjournal.org

Association between blood alcohol concentration and mortality in critical illness☆,☆☆ Christine R. Stehman, MD a, Takuhiro Moromizato, MD b, Caitlin K. McKane, BS, RN c, Kris M. Mogensen, MS, RD, LDN, CNSC d, Fiona K. Gibbons, MD e, Kenneth B. Christopher, MD f,⁎ a

Department of Emergency Medicine, Indiana University School of Medicine, Indianapolis, IN Department of Medicine, Okinawa Hokubu Prefectural Hospital, Okinawa, Japan Department of Nursing, Brigham and Women's Hospital, Boston, MA d Department of Nutrition, Brigham and Women's Hospital, Boston, MA e Division of Pulmonary and Critical Care Medicine, Department of Medicine, Massachusetts General Hospital, Boston, MA f The Nathan E. Hellman Memorial Laboratory, Renal Division, Department of Medicine, Brigham and Women's Hospital, Boston, MA b c

a r t i c l e

i n f o

Keywords: Alcohol Ethanol Critical care Intensive care Mortality Sepsis

a b s t r a c t Objective: In animal models of renal, intestinal, liver, cardiac, and cerebral ischemia, alcohol exposure is shown to reduce ischemia-reperfusion injury. Inpatient mortality of trauma patients is shown to be decreased in a dosedependent fashion relative to blood alcohol concentration (BAC) at hospital admission. In this study, we examined the association between BAC at hospital admission and risk of 30-day mortality in critically ill patients. Design: We performed a 2-center observational study of patients treated in medical and surgical intensive care units in Boston, Massachusetts. Setting: Medical and surgical intensive care units in 2 teaching hospitals in Boston, Massachusetts. Patients: We studied 11850 patients, 18 years or older, who received critical care between 1997 and 2007. The exposure of interest was the BAC determined in the first 24 hours of hospital admission and categorized a priori as BAC less than 10 mg/dL (below level of detection), 10 to 80 mg/dL, 80 to 160 mg/dL, and greater than 160 mg/ dL. The primary outcome was all-cause mortality in the 30 days after critical care initiation. Secondary outcomes included 90- and 365-day mortality after critical care initiation. Mortality was determined using the US Social Security Administration Death Master File, and 365-day follow-up was present in all cohort patients. Adjusted odds ratios (ORs) were estimated by multivariable logistic regression models with inclusion of covariate terms thought to plausibly interact with both BAC and mortality. Adjustment included age, sex, race (white or nonwhite), type (surgical vs medical), Deyo-Charlson index, sepsis, acute organ failure, trauma, and chronic liver disease. Results: Thirty-day mortality of the cohort was 13.7%. Compared to patients with BAC levels less than 10 mg/dL, patients with levels greater than or equal to 10 mg/dL had lower odds of 30-day mortality; for BAC levels 10 to 79.9 mg/dL, the OR was 0.53 (95% confidence interval [CI], 0.40-0.70); for BAC levels 80 to 159.9 mg/dL, it was 0.36 (95% CI, 0.26-0.49); and for BAC levels greater than or equal to 160 mg/dL, it was 0.35 (95% CI, 0.27-0.44). After multivariable adjustment, the OR of 30-day mortality was 0.97 (0.72-1.31), 0.79 (0.57-1.10), and 0.69 (0.54-0.90), respectively. When the cohort was analyzed with sepsis as the outcome of interest, the multivariable adjusted odds of sepsis in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.72 (0.50-1.04) or 0.68 (0.51-0.90), respectively, compared to those with BAC less than 10 mg/dL. In a subset of patients with blood cultures drawn (n = 4065), the multivariable adjusted odds of bloodstream infection in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.53 (0.27-1.01) or 0.49 (0.29-0.83), respectively, compared to those with BAC less than 10 mg/dL. Conclusions: Analysis of 11850 adult patients showed that having a detectable BAC at hospitalization was associated with significantly decreased odds of 30-day mortality after critical care. Furthermore, BAC greater than 160 mg/dL is associated with significantly decreased odds of developing sepsis and bloodstream infection. © 2015 Elsevier Inc. All rights reserved.

1. Introduction ☆ Funding: None. ☆☆ Institution where work was performed: The Nathan E. Hellman Memorial Laboratory, Renal Division, Brigham and Women's Hospital. ⁎ Corresponding author at: Renal Division, Department of Medicine, Brigham and Women's Hospital, 75 Francis Street, MRB 418, Boston, MA 02115. Tel.: +1 617 272 0535. E-mail address: [email protected] (K.B. Christopher). http://dx.doi.org/10.1016/j.jcrc.2015.08.023 0883-9441/© 2015 Elsevier Inc. All rights reserved.

Alcohol abuse and dependence is highly prevalent in the population [1]. It is estimated that between 10% and 30% of critically ill patients have an alcohol use disorder (AUD) defined as alcohol dependence or harmful use of alcohol [2-8]. The alcohol-attributable disease burden is generally seen in younger patients and is primarily related to cirrhosis

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

of the liver, neuropsychiatric disorders, and unintentional or intentional injury [9]. Epidemiologic studies support that the risk of coronary heart disease, cardiomyopathy, diabetes, and stroke is reduced with light to moderate alcohol intake [10-13]. The relationship between blood alcohol concentration (BAC) and inhospital mortality has been explored outside the intensive care unit (ICU) with some studies indicating an increase in mortality [14-16], others showing a decrease in mortality [17-23], and others indeterminate [24-29]. A recent large observational study of all level 1 and 2 trauma units in the State of Illinois demonstrated that inpatient mortality was decreased in a BAC dose-dependent fashion and showed slightly lower proportion of infections in patients with blood alcohol present [30]. Biological data show that acute alcohol administration improves experimental ischemia-reperfusion injury and subsequent organ dysfunction [31-43]. Given that alcohol use is likely to be highly prevalent in patients admitted to the ICU, and the possible alteration of inflammation related to acute alcohol use [40,42,43], and the importance of inflammation in critical care outcomes, we sought to elucidate the effect of alcohol on critical illness mortality. We performed a multiyear 2-center observational cohort study of critically ill patients among whom BAC was measured within 24 hours of hospitalization. The objective of this study was to test our hypothesis that alcohol intoxication at hospital presentation is associated with a decreased odds of 30-day all-cause mortality after critical care. 2. Materials and methods 2.1. Source population We extracted administrative and laboratory data from individuals admitted to 2 teaching hospitals in Boston, Massachusetts: Brigham and Women's Hospital (BWH), with 793 beds, and Massachusetts General Hospital (MGH), with 902 beds. The 2 hospitals provide primary as well as tertiary care to an ethnically and socioeconomically diverse population within eastern Massachusetts and the surrounding region. 2.2. Data sources Data on all patients admitted to BWH or MGH between August 3, 1997, and January 5, 2007, were obtained through the Research Patient Data Registry (RPDR), a computerized registry that serves as a central data warehouse for all inpatient and outpatient records at Partners HealthCare sites, which includes BWH and MGH. The RPDR has been used for other clinical research studies [44,45]. Approval for the study was granted by the Partners Human Research Committee Institutional Review Board. 2.3. Study population During the study period, there were 54 392 unique patients, 18 years or older, who received critical care [46]. Critical care was determined by Current Procedural Terminology (CPT) code 99291 (critical care, first 3074 minutes) assignment during hospital admission and is previously validated in the RPDR database [44]. Exclusions included 2372 patients assigned CPT code 99291 who received care only in the emergency department, 205 foreign patients as vital status in this study is determined by the Social Security Administration Death Master File, 166 patients with missing laboratory data, and 39 799 patients without BAC measured. Thus, 11850 patients constituted the total study population. 2.4. Exposure of interest and comorbidities The exposure of interest was the BAC determined in the first 24 hours of hospital admission and categorized a priori as BAC below

1383

level of detection (b 10 mg/dL), 10 to 80 mg/dL, 80 to 160 mg/dL (the legal limit in most countries), and greater than 160 mg/dL [47,48]. Race was either self-determined or designated by a patient representative/health care proxy. Patient admission “type” was defined as “medical” or “surgical” and incorporates the diagnosis-related group methodology [49]. We used the Deyo-Charlson Index to assess the burden of chronic illness [50], which is well studied and validated [51,52]. We used the validated International Classification of Diseases, Ninth Revision, coding algorithms developed by Quan et al [51] to derive a comorbidity score for each patient from individual physician billing codes obtained from all outpatient and inpatient encounters at BWH or MGH before hospital discharge. Sepsis is defined by the presence of any of the following International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), codes: 038.0-038.9, 790.7, 117.9, 112.5, or 112.81, 3 days before critical care initiation to 7 days after critical care initiation [53], an approach that we have validated in our database [54]. Acute kidney injury (AKI) was defined as RIFLE class Injury or Failure occurring between 3 days before critical care initiation and 7 days after critical care initiation [55]. We classified patients according to the maximum RIFLE class (class Risk, class Injury or class Failure) defined as a fold change in serum creatinine from preadmission serum creatinine [54,55]. Acute organ failure was adapted from Martin et al [53] and defined by a combination of ICD-9-CM and CPT codes relating to acute organ dysfunction assigned from 3 days before critical care initiation to 30 days after critical care initiation [44,54]. Acute failure is the summation of the number of acute organ failure categories (respiratory, cardiovascular, renal, hepatic, hematologic, metabolic, and/or neurologic) [53] present by ICD-9-CM code assignment. For severity of illness risk adjustment, we used the acute organ failure score, an ICU risk prediction score created from demographics (age and race), patient admission “type” as well as ICD-9-CM code–based comorbidity, sepsis, and acute organ failure covariates, which has similar discrimination for 30-day mortality as Acute Physiology and Chronic Health Evaluation II [56]. We determined the traditional ICD-derived Injury Severity Score (ICISS) via the product of all survival risk ratios for an individual patient's traumatic ICD-9 codes with more severe injuries having lower ICISS scores [57,58]. Alcohol use disorders were determined by ICD-9-CM codes at any time before discharge (alcohol abuse: 305.0-305.03; alcohol dependence: 303.0-303.93) [59]. Neighborhood poverty rate was defined as the percentage of each neighborhood's residents with incomes below the federal poverty line [44] and determined via submission of patient addresses for geocoding linked to US Census data at the census tract level [60-62]. To determine neighborhood socioeconomic disadvantage, we used geocoded residential address data [63] from electronic health records then linked the zip + 4 data to the Area Deprivation Index developed by Singh et al [64] and linked to the 2000 US census by Kind et al [65]. Chronic liver disease was determined by ICD-9-CM codes 571.x, 70.54, and 703.2 at any time before discharge [66]. MELD Score (Model For End-Stage Liver Disease) was calculated with the United Network for Organ Sharing modifications. Diabetes mellitus is defined by ICD-9-CM code 250.xx in the 2-years before hospital discharge [67,68]. Early ICU admission is defined as ICU admission within 48 hours of hospital admission. Inotropes or vasopressors were considered to be present if prescribed 3 days before critical care initiation to 7 days after critical care initiation [69,70]. Using electronic pharmacy records, exposure to inotropes and vasopressors was determined in the 7 days after the critical care initiation date for dopamine, dobutamine, epinephrine, norepinephrine, phenylephrine, milrinone, and vasopressin. Mechanical ventilation is defined as intubation (CPT 31500) or mechanical ventilation management (CPT 94656 or 94657) or tracheostomy (CPT 31600) performed after critical care initiation [71]. For the presence of malnutrition, data were collected in a subset of cohort patients at the time of initial nutrition consultation by a registered dietitian between 2005 and 2007 performed 10 days before

1384

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

critical care initiation to 10 days after critical care initiation [69]. Malnutrition was considered to be present if the patient had a diagnosis of nonspecific or specific protein-calorie malnutrition. Malnutrition was considered not to be present if the patient was not at risk for developing malnutrition or at risk for developing malnutrition but did not meet criteria for malnutrition [69]. All patients who had blood cultures drawn 48 hours before 48 hours subsequent to critical care initiation were identified [72]. Blood cultures were defined as positive if aerobic, anaerobic, or fungal blood cultures grew identifiable organisms. Patients with positive blood cultures were considered to have bloodstream infections [72-75]. Single coagulase-negative staphylococci isolates were not considered to be positive blood cultures.

analyses [84,85]. Using logistic regression, propensity scores were calculated for each cohort subject to estimate the probability for the presence or absence of a BAC greater than or equal to 10 mg/dL. Covariates selected for the propensity score development included age, sex, race, patient type, Deyo-Charlson Index, sepsis, acute organ failure, trauma, and chronic liver disease. Two smaller cohorts were obtained where a subject (with BAC ≥10 mg/dL) was matched to a control subject (with BAC b10 mg/dL) based on the propensity score. We used Mahalanobis metric matching within calipers defined by the propensity score to match the smaller cohorts [85] by using the matching algorithm “psmatch2” [86]. All P values are 2 tailed, with values less than .05 considered statistically significant. All analyses were performed using Stata 13.1 MP statistical software (College Station, TX).

2.5. Assessment of mortality

3. Results

Information on vital status for the study cohort was obtained from the Social Security Administration Death Master File, which has a reported sensitivity and specificity for mortality of 92.1% and 99.9%, respectively [76-79]. We have previously validated the accuracy of the Social Security Administration Death Master File for inhospital and out-of-hospital mortality in our administrative database [44].

Table 1 shows characteristics of the study population. Most patients were men (67%), white (72%), and had medically related diagnosis-related groups (61%). The mean age at hospital admission was 50.9 (SD, 20.3) years. The mean BAC was 35.7 (SD, 84.1) mg/dL. The study cohort was different than the parent ICU 54392 patient cohort (Supplementary Table 1). Patient characteristics of the study cohort were stratified according to BAC levels (Table 2). Factors that significantly differed between stratified groups included age, sex, race, Deyo-Charlson index, diabetes, AUD, poverty rate, vasopressors and inotropes, mechanical ventilation, sepsis, acute organ failure, AKI, MELD score, ICISS, trauma, chronic liver disease, Medicaid, and socioeconomic disadvantaged status. Primary diagnosis differed between patients with and without detectable BAC (Supplementary Table 2). Table 3 indicates that BAC, age, Deyo-Charlson index, acute organ failure, AKI, patient type, trauma, AUD, and chronic liver disease are significant univariate predictors of 30-day mortality. Inhospital mortality rate was 12.7%. Thirty-day, 90-day, and 365-day mortality rates were 13.7%, 16.4%, and 21.2%, respectively.

2.6. End points The primary end point was all-cause 30-day mortality. Secondary outcomes included 90-day mortality, bloodstream infections, and sepsis. 2.7. Power calculations and statistical analysis We performed an a priori power analysis based on observations from Friedman's regional study [30]. We assume that patients with BAC greater than 160 mg/dL compared to those with BAC less than 10 mg/dL will have an absolute decrease in 30-day mortality of 4.3%. With an α error level of 5% and a power of 80%, the estimated sample size for our 2sample comparison required for our primary end point (30-day mortality) is 1089 in the BAC less than 10 mg/dL group and 1089 in the BAC greater than 160 mg/dL group. 2.8. Statistics Categorical variables were described by frequency distribution and compared across BAC groups using contingency tables and χ 2 testing. Continuous variables were examined graphically (eg, histogram, box plot) and in terms of summary statistics (mean, SD, median, interquartile range) and then compared across exposure groups using 1-way analysis of variance. The primary outcome considered was 30-day mortality. Unadjusted associations between BAC groups and 30-day mortality were estimated by bivariable logistic regression models. Adjusted odds ratios (ORs) were estimated by multivariable logistic regression models with inclusion of covariate terms thought to plausibly associate with both BAC levels and 30-day mortality. For the primary model (30day mortality), specification of each continuous covariate (as a linear vs categorical term) was adjudicated by the empiric association with the primary outcome using Akaike's Information Criterion [80]; model fit was assessed using the Hosmer-Lemeshow χ 2 goodness-of-fit test [81]. Models for secondary analyses were specified identically to the primary model. Unadjusted event rates were calculated with the use of the Kaplan-Meier methods and compared with the use of the log-rank test. Locally weighted scatter plot smoothing [82,83] was used to graphically represent the relationship between BAC level and risk of 30-day mortality. We tested the significance of the interaction using the likelihood ratio test. To reduce potential bias from the nonrandomized use of alcohol in the cohort, we constructed propensity scores for BAC greater than or equal to 10 mg/dL and used these in the primary and secondary

3.1. Primary outcome Blood alcohol concentration at hospital admission was a strong predictor of 30-day mortality (Table 4 and Fig. 1). The odds of 30-day mortality in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 64% and 65% less, respectively, compared to those with BAC less than 10 mg/dL. Blood alcohol concentration remained a significant predictor of odds of 30-day mortality after adjustment for age, sex, race, patient type (medical vs surgical), Deyo-Charlson index, sepsis, acute organ failure, trauma, and chronic liver disease. The adjusted odds of 30-day mortality in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 21% and 31% less, respectively, compared to those with BAC less than 10 mg/dL (Table 4). Locally weighted scatter plot smoothing plot (Fig. 2) shows a near inverse linear association between BAC levels and risk of 30-day mortality up to BAC levels near 200 mg/dL. Table 1 Patient characteristics of the study population n Sex, n (%) Female Male Race, n (%) Nonwhite White Age, mean (SD) Patient type, n (%) Medical Surgical Trauma, n (%) Sepsis, n (%) Vasopressors/inotropes, n (%) Mechanical ventilation, n (%)

11 850 3956 (33.4) 7894 (66.6) 3304 (27.9) 8546 (72.1) 50.9 (20.3) 7198 (60.7) 4652 (39.3) 6677 (56.4) 1130 (9.5) 2995 (27.4) 4586 (42.0)

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

1385

Table 2 Patient characteristics by BAC at hospital admission BAC

n Age, mean (SD) Sex, n (%) Male Female Race, n (%) White Nonwhite Patient type, n (%) Medical Surgical Deyo-Charlson index, n (%) 0 1-2 3 ≥4 Acute organ failure, n (%) 0 1 2 3 ≥4 Chronic liver disease, n (%) MELD score, mean (SD)a AUD, n (%) Medicaid, n (%) Socioeconomic disadvantaged, n (%)b Malnutrition, n (%)c Diabetes mellitus, n (%) Early ICU, n (%) Mechanical ventilation, n (%) Sepsis, n (%) Vasopressors/inotropes, n (%) AKI, n (%)d Trauma, n (%) ICISS, mean (SD) e Acute organ failure score, mean (SD) Bloodstream infection, n (%)

P

b10 mg/dL

10-79.9 mg/dL

80-160 mg/dL

N160 mg/dL

9283 54.0 (20.2)

638 40.5 (17.5)

703 38.1 (16.2)

1226 40.0 (15.0)

5828 (62.8) 3455 (37.2)

498 (78.1) 140 (21.9)

576 (81.9) 127 (18.1)

992 (80.9) 234 (19.1)

6792 (73.2) 2491 (26.8)

424 (66.5) 214 (33.5)

496 (70.6) 207 (29.5)

834 (68.0) 392 (32.0)

5805 (62.5) 3478 (37.5)

333 (52.2) 305 (47.8)

371 (52.8) 332 (47.2)

689 (56.2) 537 (43.8)

2097 (22.6) 3677 (39.6) 1307 (14.1) 2202 (23.7)

289 (45.3) 247 (38.7) 55 (8.6) 47 (7.4)

349 (49.6) 277 (39.4) 35 (5.0) 42 (6.0)

581 (47.4) 507 (41.4) 74 (6.0) 64 (5.2)

3402 (36.7) 2754 (29.7) 1694 (18.3) 874 (9.4) 559 (6.0) 867 (9.3) 11.9 (7.0) 1494 (16.1) 854 (9.2) 1340 (17.5) 260 (60.8) 1901 (20.5) 9003 (97.0) 3860 (41.6) 986 (10.6) 2619 (28.2) 656 (7.9) 4567 (49.2) 0.64 (0.19) 9.6 (5.1) 590 (16.7)

277 (43.4) 189 (29.6) 99 (15.5) 48 (7.5) 25 (3.9) 49 (7.7) 10.1 (5.3) 310 (48.6) 86 (13.5) 126 (23.2) 21 (70.0) 52 (8.2) 629 (98.7) 255 (40.0) 54 (8.5) 177 (27.7) 19 (3.0) 489 (76.7) 0.59 (0.18) 7.1 (4.6) 18 (12.2)

343 (48.8) 208 (29.6) 94 (13.4) 35 (5.0) 23 (3.3) 59 (8.4) 9.1 (3.9) 438 (62.3) 85 (12.1) 119 (20.2) 25 (73.5) 53 (7.5) 687 (97.7) 280 (39.8) 34 (4.8) 163 (23.2) 27 (3.9) 588 (83.6) 0.57 (0.18) 6.4 (4.2) 10 (7.4)

530 (43.2) 349 (28.5) 222 (18.1) 89 (7.3) 36 (2.9) 125 (10.2) 8.7 (3.5) 985 (80.3) 113 (9.2) 218 (22.5) 21 (67.7) 85 (6.9) 1204 (98.2) 579 (47.3) 56 (4.6) 288 (23.5) 26 (2.1) 1033 (84.3) 0.58 (0.18) 6.9 (3.9) 16 (6.6)

b.0001* b.0001

b.0001

b.0001

b.0001

b.0001

.28 b.0001* b.0001 b.0001 b.0001 .33 b.0001 0.013 0.001 b.0001 0.001 b.0001 0.001 b.0001* b.0001* b.0001

Note: P values determined by χ2 test, unless designated by asterisk (*), then P value determined by analysis of variance. a MELD score data available in 8214 cohort patients. b Socioeconomic disadvantaged are patients with area deprivation index greater than 113; data available in 9764 cohort patients. c Nutrition data available in 523 cohort patients. d Acute kidney injury data available in 10865 cohort patients. e ICISS data available in 6677 cohort patients.

3.2. Subanalyses Although statistical power was compromised, when patients younger than 50 years were excluded (n = 5821), the odds of 30-day mortality in patients with BAC greater than 160 mg/dL was 0.33 (0.21-0.50), relative to the BAC less than 10 mg/dL group. The multivariable adjusted odds of 30-day mortality in patients with BAC greater than 160 mg/dL was 0.48 (0.31-0.75), all relative to the BAC less than 10 mg/dL group. When patients with chronic liver disease were excluded (n = 10750), the odds of 30-day mortality in patients with BAC greater than 160 mg/dL adjusted for age, sex, race, patient type, Deyo-Charlson index, sepsis, acute organ failure, and trauma was 0.64 (95% confidence interval [CI], 0.48-0.85), relative to the BAC less than 10 mg/dL group. With limited statistical power present, when patients with BAC less than 10 mg/dL were excluded (n = 2567), the odds of 30-day mortality in patients with BAC greater than 160 mg/dL was 0.65 (0.46-0.94), relative to the BAC 10 to 79.9 mg/dL group. The multivariable adjusted odds of 30-day mortality in patients with BAC greater than 160 mg/dL was 0.70 (0.48-1.02), relative to the BAC 10 to 79.9 mg/dL group. Furthermore, when trauma patients were excluded (n = 5173), the odds of 30-day mortality in patients with BAC greater than 160 mg/dL adjusted for age, sex, race, Deyo-Charlson index, sepsis, and acute organ failure was 0.74 (0.55-0.98), relative to the BAC less than 10 mg/dL group.

When the cohort was restricted to those with AUD (n = 3227), the odds of 30-day mortality in patients with BAC greater than 160 mg/dL adjusted for age, sex, race, Deyo-Charlson index, sepsis, and acute organ failure was 0.71 (0.49-1.01), relative to the BAC less than 10 mg/dL group. We next analyzed a series of secondary outcomes including bloodstream infection, sepsis, and acute organ failure. In a subset of patients with blood cultures drawn (n = 4065), the odds of bloodstream infection in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.53 (0.27-1.01) or 0.49 (0.29-0.83), respectively, compared to those with BAC less than 10 mg/dL after adjustment for age, sex, race, Deyo-Charlson index, and patient type. When the cohort was analyzed with sepsis as the outcome of interest, the odds of sepsis in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.72 (0.501.04) or 0.68 (0.51-0.90), respectively, compared to those with BAC less than 10 mg/dL after adjustment for age, sex, race, Deyo-Charlson index, and patient type. When cohort patients were grouped according to presence or absence of blood alcohol (BAC ≥ 10 mg/dL vs BAC b10 mg/dL; Supplementary Table 3), acute organ failure was statistically lower in the BAC greater than or equal to 10 mg/dL group (55%) compared to 63% of those with BAC less than 10 mg/dL group [χ 2(4, n = 11 850) = 84.74; P b 0001]. When the cohort was analyzed with the presence of acute organ failure as the outcome of interest, the odds of

1386

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

Table 3 Unadjusted associations between covariates and 30-day mortality

Age (per 1 y) Sex Male Female Race Nonwhite White Patient type Surgical Medical Deyo-Charlson index 0 1-2 3 ≥4 Acute organ failure 0 1 2 3 ≥4 Acute organ failure score (per 1 point) AKI Trauma Sepsis Chronic liver disease AUD BAC b10 mg/dL 10-79.9 mg/dL 80-159.9 mg/dL N160 mg/dL

OR

95% CI

P

1.04

1.03-1.04

b.0001

0.73 1.00

0.66-0.81 Referent

b.0001

0.82 1.00

0.73-0.93 Referent

.002

0.82 1.00

0.74-0.92 Referent

0.001

1.00 3.33 4.77 6.69

Referent 2.77-4.00 3.86-5.88 5.54-8.09

b.0001 b.0001 b.0001

Fig. 1. Time-to-event curves for the primary end point. Unadjusted event rates were calculated with the use of the Kaplan-Meier methods and compared with the use of the logrank test. Categorization of blood alcohol level is per the primary analyses. The global comparison log-rank P value is less than .0001.

1.00 1.86 3.05 4.15 8.61

Referent 1.58-2.19 2.57-3.62 3.41-5.04 6.95-10.67

b.0001 b.0001 b.0001 b.0001

adjustment for age, sex, race, Deyo-Charlson index, trauma, sepsis, and patient type.

1.18 4.07 0.55 2.58 1.40 0.49

1.17-1.19 3.44- 4.82 0.50-0.61 2.24-2.98 1.19-1.65 0.43 -0.56

b.0001 b.0001 b.0001 b.0001 b.0001 b.0001

1.00 0.53 0.36 0.35

Referent 0.40-0.70 0.26-0.49 0.27-0.44

b.0001 b.0001 b.0001

acute organ failure in patients with BAC greater than or equal to 10 mg/ dL were 0.87 (0.79-0.95) compared to those with BAC less than 10 mg/ dL after adjustment for age, sex, race, and patient type. When the cohort was analyzed with AKI as the outcome of interest, the odds of AKI in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.92 (0.61-1.40) or 0.47 (0.31-0.72), respectively, compared to those with BAC less than 10 mg/dL after adjustment for age, sex, race, DeyoCharlson index, trauma, sepsis, and patient type. Finally, when the cohort was analyzed with renal replacement as the outcome of interest, the odds of renal replacement in patients with BAC 80 to 160 mg/dL or greater than 160 mg/dL were 0.68 (0.29-1.57) or 0.47 (0.22-0.97), respectively, compared to those with BAC less than 10 mg/dL after

3.3. Effect modification Analyses based on fully adjusted models were performed to evaluate the BAC-mortality association, and P for interaction was determined to explore for any evidence of effect modification. We individually tested for effect modification by trauma, diabetes mellitus, chronic liver disease, mechanical ventilation, sepsis, vasopressor and inotrope use, creatinine, hematocrit, and white blood cell count by adding an interaction term to the multivariate models. Trauma, diabetes mellitus, and chronic liver disease were not effect modifiers of the association between BAC and 30-day mortality (P interaction: 30-day mortality in all cases N .28). The effect modification analysis showed that the association between BAC and 30-day mortality was modified by sepsis (P interaction: 30-day mortality = .047), mechanical ventilation (P interaction: 30-day mortality b .001), and vasopressor and inotrope use (P interaction: 30-day mortality b .001). In all cases where effect modification is present, the effect size differs (hence, the effect modification), but the directionality of the association is maintained. Furthermore, individually running the adjusted model with and without terms for poverty, socioeconomic disadvantage, comorbidities (chronic liver disease and diabetes mellitus), measures of disease severity (mechanical ventilation, sepsis, and vasopressor and inotrope use), principle diagnosis category or laboratory measurements at the time of critical care (creatinine, hematocrit, and white blood cell count), the

Table 4 Unadjusted and adjusted associations between BAC and mortality after critical care b10 mg/dL 30-d mortality Crude Adjustedb Adjustedc Adjustedd PS-matched cohorte 90-d mortality Crude Adjustedb Adjustedc Adjustedd PS-matched cohorte

OR (95% CI)

10-79.9 mg/dL

80-160 mg/dL

N160 mg/dL

HL χ2 P

OR (95% CI), P

OR (95% CI), P

OR (95% CI), P

a

1.00 (1.00-1.00) 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a

0.53 (0.40-0.70), b.0001 0.97 (0.72-1.31), .85 0.79 (0.60-1.06), .12 0.85 (0.58-1.23), .38 1.05 (0.77-1.43), .76

0.36 (0.26-0.49), b.0001 0.79 (0.57-1.10), .16 0.62 (0.45-0.86), .004 0.84 (0.58-1.21), .35 0.69 (0.49-0.97), .031

0.35 (0.27-0.44), b.0001 0.69 (0.54-0.90), .006 0.57 (0.45-0.74), b.0001 0.70 (0.52-0.93), .015 0.68 (0.51-0.89), .005

1.00 .21 .68 .21 1.00

1.00 (1.00-1.00)a 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a 1.00 (1.00-1.00)a

0.46 (0.36-0.61), b.0001 0.90 (0.67-1.20), .48 0.70 (0.53-0.92), .011 0.81 (0.56-1.15), .24 0.98 (0.73-1.31), .88

0.35 (0.26-0.46), b.0001 0.81 (0.60-1.11), .20 0.60 (0.44-0.81), .001 0.88 (0.62-1.24), .47 0.70 (0.51-0.96), 028

0.30 (0.24-0.38), b.0001 0.65 (0.50-0.83), .001 0.50 (0.39-0.63), b.0001 0.64 (0.48-0.85), .002 0.62 (0.48-0.81), b.0001

1.00 .55 .44 .51 1.00

Note: PS indicates propensity score calculated to estimate the probability for the presence or absence of BAC greater than or equal to 10 mg/dL. a Referent in each case is BAC less than 10 mg/dL. b Model 1: Estimates adjusted for age, sex, race, patient type (medical vs surgical), Deyo-Charlson index, sepsis, acute organ failure, trauma, and chronic liver disease. c Model 2: Estimates adjusted for sex, the acute organ failure score, trauma, and chronic liver disease. d Model 3: Estimates adjusted for age, sex, race (medical vs surgical), Deyo-Charlson index, acute organ failure, ICISS, and chronic liver disease (n = 6933). e Propensity score–matched cohort: n = 4922, with 2461 with BAC less than 10 mg/dL and 2461 with a BAC greater than or equal to 10 mg/dL.

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

Fig. 2. Blood alcohol concentration vs risk of 30-day mortality. Locally weighted scatter plot smoothing used to represent the near inverse linear association between BAC level and risk of 30-day mortality.

30-day mortality estimates in each case are similar (data not shown). This indicates that the BAC–30-day mortality relationship is not materially confounded by severity of illness, comorbidities, case-mix, or poverty. We assessed the odds of death in a smaller cohort of propensity score–matched patients (n = 4922). Again, the propensity score matched on the baseline characteristics (BAC b10 mg/dL, n = 2461; BAC ≥10 mg/dL, n = 2461; Supplementary Table 4). In the matched cohort, crude all-cause mortality rates were 8.9% (95% CI, 7.8-10.0; 219 deaths) in patients with BAC less than 10 mg/dL and 7.0% (95% CI, 6.08.1; 170 deaths) in patients with BAC greater than or equal to 10 mg/dL. In the matched cohort, the adjusted odds of 30-day mortality in the group of propensity score–matched patients with BAC greater than or equal to 160 mg/dL was 1.6-fold less than that of patients with BAC less than 10 mg/dL (Table 4).

4. Discussion In this study, we investigated whether BAC at presentation to the hospital was associated with mortality in critically ill patients. Our data demonstrate that the presence of blood alcohol at hospital admission is associated with a significant decrease in the odds of 30-day mortality in critically ill adults, which remained present on multivariable analyses. In addition, we show that the presence of blood alcohol decreases the odds of bloodstream infection, sepsis, and organ dysfunction. However, as our study is observational and not interventional, a causal nature of the relationship between alcohol exposure and outcomes cannot be concluded by these data alone. Our study compliments and extends the observations made by Friedman's regional study in patients admitted with trauma, which demonstrated that inpatient mortality was decreased in a BAC dosedependent fashion [30]. Biologic evidence shows decreased sequela in experimental injury models when alcohol is present. In animal models of renal, intestinal, liver, cardiac, and cerebral ischemia, alcohol exposure is shown to reduce ischemia-reperfusion injury [31-43]. Intriguing ethanol preconditioning experiments in animal models show that ethanol prevents postischemic adhesive interactions between leukocytes and endothelial cells, which can lead to organ dysfunction and death [42,87]. Furthermore, in experimental ethanol exposure, evidence exists for an antiadhesive and anti-inflammatory state in postcapillary venules mediated by ethanol dependent NADPH oxidase–derived oxidants [40,42,43]. After adjustment for potential confounders, we show that patients with detectable BAC have decreased odds of bloodstream infection, sepsis, and organ dysfunction. We hypothesize that the impaired early inflammatory response related to alcohol exposure may be beneficial in terms of the severity of subsequent organ dysfunction.

1387

Prior studies have evaluated surrogates for alcohol consumption and outcomes in critical illness. Alcohol dependence when diagnosed by ICD-9 coding is associated with sepsis, septic shock, and hospital mortality among ICU patients [7]. In a small prospective study, critically ill patients with septic shock who have chronic alcohol abuse as identified by the Short Michigan Alcohol Screening Test have increased risk of acute respiratory distress syndrome [6]. In another small study of critically ill patients, chronic alcohol consumption determined by patient interview and the Short Michigan Alcohol Screening Test showed an increase in a composite infection outcome including central venous catheter infections, ventilator-associated pneumonia, and complicated catheterrelated urinary tract infections [4]. Beyond the harm to society and increased health care costs, alcohol exposure also may have harmful biological effects [1,9,88-91]. In traumatic injury, acute alcohol exposure appears to have deleterious effects with regard to suppression of proinflammatory cytokine release, decreased neutrophil recruitment and phagocytic function, and alterations in the proliferative phase of wound healing, specifically wound angiogenesis [92-96]. We examined several factors to explain our findings. First, outcomes in critical care are known to depend on severity of illness. In our cohort, the patients with a detectable BAC have a significantly decreased age and significantly less sepsis, comorbidity, and vasopressor or inotrope use (but not mechanical ventilation) compared to patients with BAC less than 10 mg/dL (Table 2). Severity of illness is estimated in our study using an acute organ failure score, which strongly correlates with mortality [44,53,97,98] (Table 3). Despite adjustment for severity of illness surrogates and an absence of confounding of the BACmortality relationship related to mechanical ventilation or vasopressor or inotrope use, insufficient adjustment for severity of illness may have contributed to our observed results. The present study has potential limitations. In general, observational studies may be limited by the absence of a randomly distributed exposure, potential confounding, and reverse causation. In our study, ascertainment bias may exist as the patient cohort under study had BAC status measured for an unknown reason that may be absent in other critically ill patients. These differences may decrease the generalizability of our results to all critically ill patients. Sepsis is defined based on administrative ICD-9-CM codes that have high positive and negative predictive values for true cases of sepsis but likely underestimate actual sepsis [53,54]. As determination of the Deyo-Charlson Index relies on inpatient and outpatient daily billing codes from the hospitals under study, we are unable to use data from physicians outside our institutions to determine comorbidity. Despite adjustment for multiple potential confounders, there may still be residual confounding contributing to observed differences in outcomes. Specifically, the presence of blood alcohol may be a marker for the general condition of the patient, for which we are unable to fully adjust. The present study has several strengths. For example, our study has sufficient numbers of patients to ensure the adequate reliability of our estimates (n = 11850; 30-day mortality rate, 13.7%). We have sufficient statistical power to detect a clinically relevant difference in 30-day mortality if one exists. The Deyo-Charlson Index accounts for chronic conditions and is well studied and validated [51,52]. Finally, all cohort patients under study had vital status follow-up for at least 365 days using the Social Security Administration Death Master File, which is validated for inhospital and out-of-hospital mortality in our data set [44]. 5. Conclusions In summary, these data demonstrate that BAC at hospital admission is strongly associated with survival in a large cohort of critically ill patients. Although it has been hypothesized that alcohol dependence or abuse may play a deleterious role in critical care outcomes [4,6,7], our work presents important evidence with regard to outcomes in patients with acute alcohol exposure. In conjunction with the basic science evidence [31-43], we believe that our results provide clinical evidence of

1388

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389

a graded association between BAC at the time of hospital admission and improved outcomes of critically ill patients. Further longitudinal studies are required to confirm our findings and establish mechanisms underlying these observations. Acknowledgment This manuscript is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman, MD, PhD. We thank Shawn Murphy and Henry Chueh and the Partners HealthCare Research Patient Data Registry group for facilitating use of their database. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jcrc.2015.08.023. References [1] Hasin DS, Stinson FS, Ogburn E, Grant BF. Prevalence, correlates, disability, and comorbidity of DSM-IV alcohol abuse and dependence in the United States: results from the National Epidemiologic Survey on Alcohol and Related Conditions. Arch Gen Psychiatry 2007;64:830–42. [2] Baldwin WA, Rosenfeld BA, Breslow MJ, Buchman TG, Deutschman CS, Moore RD. Substance abuse-related admissions to adult intensive care. Chest 1993;103:21–5. [3] de Wit M, Wan SY, Gill S, Jenvey WI, Best AM, Tomlinson J, et al. Prevalence and impact of alcohol and other drug use disorders on sedation and mechanical ventilation: a retrospective study. BMC Anesthesiol 2007;7:3. [4] Gacouin A, Legay F, Camus C, Volatron AC, Barbarot N, Donnio PY, et al. At-risk drinkers are at higher risk to acquire a bacterial infection during an intensive care unit stay than abstinent or moderate drinkers. Crit Care Med 2008;36:1735–41. [5] Maxson PM, Schultz KL, Berge KH, Lange CM, Schroeder DR, Rummans TA. Probable alcohol abuse or dependence: a risk factor for intensive-care readmission in patients undergoing elective vascular and thoracic surgical procedures. Mayo Perioperative Outcomes Group. Mayo Clin Proc 1999;74:448–53. [6] Moss M, Parsons PE, Steinberg KP, Hudson LD, Guidot DM, Burnham EL, et al. Chronic alcohol abuse is associated with an increased incidence of acute respiratory distress syndrome and severity of multiple organ dysfunction in patients with septic shock. Crit Care Med 2003;31:869–77. [7] O'Brien Jr JM, Lu B, Ali NA, Martin GS, Aberegg SK, Marsh CB, et al. Alcohol dependence is independently associated with sepsis, septic shock, and hospital mortality among adult intensive care unit patients. Crit Care Med 2007;35:345–50. [8] Suchyta MR, Beck CJ, Key CW, Jephson A, Hopkins RO. Substance dependence and psychiatric disorders are related to outcomes in a mixed ICU population. Intensive Care Med 2008;34:2264–7. [9] Rehm J, Mathers C, Popova S, Thavorncharoensap M, Teerawattananon Y, Patra J. Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders. Lancet 2009;373:2223–33. [10] Zakhari S. Molecular mechanisms underlying alcohol-induced cardioprotection: contribution of hemostatic components. Introduction to the symposium. Alcohol Clin Exp Res 1999;23:1108–10. [11] Puddey IB, Rakic V, Dimmitt SB, Beilin LJ. Influence of pattern of drinking on cardiovascular disease and cardiovascular risk factors–a review. Addiction 1999;94: 649–63. [12] Rehm J, Room R, Graham K, Monteiro M, Gmel G, Sempos CT. The relationship of average volume of alcohol consumption and patterns of drinking to burden of disease: an overview. Addiction 2003;98:1209–28. [13] Reynolds K, Lewis B, Nolen JD, Kinney GL, Sathya B, He J. Alcohol consumption and risk of stroke: a meta-analysis. JAMA 2003;289:579–88. [14] Fabbri A, Marchesini G, Morselli-Labate AM, Rossi F, Cicognani A, Dente M, et al. Blood alcohol concentration and management of road trauma patients in the emergency department. J Trauma 2001;50:521–8. [15] Pories SE, Gamelli RL, Vacek P, Goodwin G, Shinozaki T, Harris F. Intoxication and injury. J Trauma 1992;32:60–4. [16] Waller PF, Stewart JR, Hansen AR, Stutts JC, Popkin CL, Rodgman EA. The potentiating effects of alcohol on driver injury. JAMA 1986;256:1461–6. [17] O'Phelan K, McArthur DL, Chang CW, Green D, Hovda DA. The impact of substance abuse on mortality in patients with severe traumatic brain injury. J Trauma 2008; 65:674–7. [18] Blondell RD, Looney SW, Krieg CL, Spain DA. A comparison of alcohol-positive and alcohol-negative trauma patients. J Stud Alcohol 2002;63:380–3. [19] Kraus JF, Morgenstern H, Fife D, Conroy C, Nourjah P. Blood alcohol tests, prevalence of involvement, and outcomes following brain injury. Am J Public Health 1989;79:294–9. [20] Salim A, Ley EJ, Cryer HG, Margulies DR, Ramicone E, Tillou A. Positive serum ethanol level and mortality in moderate to severe traumatic brain injury. Arch Surg 2009; 144:865–71. [21] Tien HC, Tremblay LN, Rizoli SB, Gelberg J, Chughtai T, Tikuisis P, et al. Association between alcohol and mortality in patients with severe traumatic head injury. Arch Surg 2006;141:1185–91 [discussion 1192]. [22] Ward RE, Flynn TC, Miller PW, Blaisdell WF. Effects of ethanol ingestion on the severity and outcome of trauma. Am J Surg 1982;144:153–7.

[23] Yaghoubian A, Kaji A, Putnam B, De Virgilio N, De Virgilio C. Elevated blood alcohol level may be protective of trauma patient mortality. Am Surg 2009;75:950–3. [24] Huth JF, Maier RV, Simonowitz DA, Herman CM. Effect of acute ethanolism on the hospital course and outcome of injured automobile drivers. J Trauma 1983;23: 494–8. [25] Jurkovich GJ, Rivara FP, Gurney JG, Fligner C, Ries R, Mueller BA, et al. The effect of acute alcohol intoxication and chronic alcohol abuse on outcome from trauma. JAMA 1993;270:51–6. [26] Madan AK, Yu K, Beech DJ. Alcohol and drug use in victims of life-threatening trauma. J Trauma 1999;47:568–71. [27] Shandro JR, Rivara FP, Wang J, Jurkovich GJ, Nathens AB, MacKenzie EJ. Alcohol and risk of mortality in patients with traumatic brain injury. J Trauma 2009;66:1584–90. [28] Shih HC, Hu SC, Yang CC, Ko TJ, Wu JK, Lee CH. Alcohol intoxication increases morbidity in drivers involved in motor vehicle accidents. Am J Emerg Med 2003;21:91–4. [29] Zeckey C, Dannecker S, Hildebrand F, Mommsen P, Scherer R, Probst C, et al. Alcohol and multiple trauma: is there an influence on the outcome? Alcohol 2011;45: 245–51. [30] Friedman LS. Dose–response relationship between in-hospital mortality and alcohol following acute injury. Alcohol 2012;46:769–75. [31] Miyamae M, Camacho SA, Zhou HZ, Diamond I, Figueredo VM. Alcohol consumption reduces ischemia-reperfusion injury by species-specific signaling in guinea pigs and rats. Am J Physiol 1998;275:H50–6. [32] Miyamae M, Diamond I, Weiner MW, Camacho SA, Figueredo VM. Regular alcohol consumption mimics cardiac preconditioning by protecting against ischemiareperfusion injury. Proc Natl Acad Sci U S A 1997;94:3235–9. [33] Wang Q, Kalogeris TJ, Wang M, Jones AW, Korthuis RJ. Antecedent ethanol attenuates cerebral ischemia/reperfusion-induced leukocyte-endothelial adhesive interactions and delayed neuronal death: role of large conductance, Ca2+−activated K+ channels. Microcirculation 2010;17:427–38. [34] Collins MA, Neafsey EJ, Mukamal KJ, Gray MO, Parks DA, Das DK, et al. Alcohol in moderation, cardioprotection, and neuroprotection: epidemiological considerations and mechanistic studies. Alcohol Clin Exp Res 2009;33:206–19. [35] Pagel PS, Krolikowski JG, Kehl F, Mraovic B, Kersten JR, Warltier DC. The role of mitochondrial and sarcolemmal K(ATP) channels in canine ethanol-induced preconditioning in vivo. Anesth Analg 2002;94:841–8 [table of contents]. [36] Louboutin JP, Marusich E, Gao E, Agrawal L, Koch WJ, Strayer DS. Ethanol protects from injury due to ischemia and reperfusion by increasing vascularity via vascular endothelial growth factor. Alcohol 2012;46:441–54. [37] Chen CH, Budas GR, Churchill EN, Disatnik MH, Hurley TD, Mochly-Rosen D. Activation of aldehyde dehydrogenase-2 reduces ischemic damage to the heart. Science 2008;321:1493–5. [38] Churchill EN, Disatnik MH, Mochly-Rosen D. Time-dependent and ethanol-induced cardiac protection from ischemia mediated by mitochondrial translocation of varepsilonPKC and activation of aldehyde dehydrogenase 2. J Mol Cell Cardiol 2009;46:278–84. [39] Wang Q, Sun AY, Simonyi A, Kalogeris TJ, Miller DK, Sun GY, et al. Ethanol preconditioning protects against ischemia/reperfusion-induced brain damage: role of NADPH oxidase-derived ROS. Free Radic Biol Med 2007;43:1048–60. [40] Dayton C, Yamaguchi T, Kamada K, Carter P, Korthuis RJ. Antecedent ethanol ingestion prevents postischemic P-selectin expression in murine small intestine. Microcirculation 2004;11:709–18. [41] Yuan Q, Hong S, Han S, Zeng L, Liu F, Ding G, et al. Preconditioning with physiological levels of ethanol protect kidney against ischemia/reperfusion injury by modulating oxidative stress. PLoS One 2011;6:e25811. [42] Yamaguchi T, Dayton CB, Ross CR, Yoshikawa T, Gute DC, Korthuis RJ. Late preconditioning by ethanol is initiated via an oxidant-dependent signaling pathway. Free Radic Biol Med 2003;34:365–76. [43] Yamaguchi T, Dayton C, Shigematsu T, Carter P, Yoshikawa T, Gute DC, et al. Preconditioning with ethanol prevents postischemic leukocyte-endothelial cell adhesive interactions. Am J Physiol Heart Circ Physiol 2002;283:H1019–30. [44] Zager S, Mendu ML, Chang D, Bazick HS, Braun AB, Gibbons FK, et al. Neighborhood poverty rate and mortality in patients receiving critical care in the academic medical center setting. Chest 2011;139:1368–79. [45] Braun A, Chang D, Mahadevappa K, Gibbons FK, Liu Y, Giovannucci E, et al. Association of low serum 25-hydroxyvitamin D levels and mortality in the critically ill. Crit Care Med 2011;39:671–7. [46] Bazick HS, Chang D, Mahadevappa K, Gibbons FK, Christopher KB. Red cell distribution width and all-cause mortality in critically ill patients. Crit Care Med 2011;39: 1913–21. [47] Taylor B, Rehm J. The relationship between alcohol consumption and fatal motor vehicle injury: high risk at low alcohol levels. Alcohol Clin Exp Res 2012;36:1827–34. [48] Christophersen AS, Skurtveit S, Grung M, Morland J. Rearrest rates among Norwegian drugged drivers compared with drunken drivers. Drug Alcohol Depend 2002; 66:85–92. [49] Rapoport J, Gehlbach S, Lemeshow S, Teres D. Resource utilization among intensive care patients. Managed care vs traditional insurance. Arch Intern Med 1992;152: 2207–12. [50] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–83. [51] Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130–9. [52] Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011;173:676–82.

C.R. Stehman et al. / Journal of Critical Care 30 (2015) 1382–1389 [53] Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med 2003;348:1546–54. [54] Moromizato T, Litonjua AA, Braun AB, Gibbons FK, Giovannucci E, Christopher KB. Association of low serum 25-hydroxyvitamin D levels and sepsis in the critically ill. Crit Care Med 2014;42:97–107. [55] Braun AB, Litonjua AA, Moromizato T, Gibbons FK, Giovannucci E, Christopher KB. Association of low serum 25-hydroxyvitamin D levels and acute kidney injury in the critically ill. Crit Care Med 2012;40:3170–9. [56] Elias KM, Moromizato T, Gibbons FK, Christopher KB. Derivation and validation of the acute organ failure score to predict outcome in critically ill patients: a cohort study. Crit Care Med 2015;43:856–64. [57] Osler T, Rutledge R, Deis J, Bedrick E. ICISS: an international classification of disease9 based injury severity score. J Trauma 1996;41:380–6 [discussion 386–388]. [58] Stephenson S, Henley G, Harrison JE, Langley JD. Diagnosis based injury severity scaling: investigation of a method using Australian and New Zealand hospitalisations. Inj Prev 2004;10:379–83. [59] Smothers BA, Yahr HT, Ruhl CE. Detection of alcohol use disorders in general hospital admissions in the United States. Arch Intern Med 2004;164:749–56. [60] Krieger N, Waterman PD, Chen JT, Soobader MJ, Subramanian SV. Monitoring socioeconomic inequalities in sexually transmitted infections, tuberculosis, and violence: geocoding and choice of area-based socioeconomic measures–the public health disparities geocoding project (US). Public Health Rep 2003;118:240–60. [61] Subramanian SV, Chen JT, Rehkopf DH, Waterman PD, Krieger N. Racial disparities in context: a multilevel analysis of neighborhood variations in poverty and excess mortality among black populations in Massachusetts. Am J Public Health 2005;95:260–5. [62] Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R. Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). J Epidemiol Community Health 2003;57:186–99. [63] Duncan DT, Sharifi M, Melly SJ, Marshall R, Sequist TD, Rifas-Shiman SL, et al. Characteristics of Walkable Built Environments and BMI z-Scores in Children: Evidence from a Large Electronic Health Record Database. Environ Health Perspect 2014;122:1359–65. [64] Singh GK. Area deprivation and widening inequalities in US mortality, 1969–1998. Am J Public Health 2003;93:1137–43. [65] Kind AJ, Jencks S, Brock J, Yu M, Bartels C, Ehlenbach W, et al. Neighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study. Ann Intern Med 2014;161:765–74. [66] Hug BL, Lipsitz SR, Seger DL, Karson AS, Wright SC, Bates DW. Mortality and drug exposure in a 5-year cohort of patients with chronic liver disease. Swiss Med Wkly 2009;139:737–46. [67] Chen G, Khan N, Walker R, Quan H. Validating ICD coding algorithms for diabetes mellitus from administrative data. Diabetes Res Clin Pract 2010;89:189–95. [68] Zgibor JC, Orchard TJ, Saul M, Piatt G, Ruppert K, Stewart A, et al. Developing and validating a diabetes database in a large health system. Diabetes Res Clin Pract 2007;75:313–9. [69] Robinson MK, Mogensen KM, Casey JD, McKane CK, Moromizato T, Rawn JD, et al. The relationship among obesity, nutritional status, and mortality in the critically ill. Crit Care Med 2015;43:87–100. [70] Purtle SW, Moromizato T, McKane CK, Gibbons FK, Christopher KB. The association of red cell distribution width at hospital discharge and out-of-hospital mortality following critical illness. Crit Care Med 2014;42:918–29. [71] Kind AJ, Smith MA, Frytak JR, Finch MD. Bouncing back: patterns and predictors of complicated transitions 30 days after hospitalization for acute ischemic stroke. J Am Geriatr Soc 2007;55:365–73. [72] Mendu ML, Zager S, Gibbons FK, Christopher KB. Relationship between neighborhood poverty rate and bloodstream infections in the critically ill. Crit Care Med 2012;40:1427–36. [73] Shorr AF, Jackson WL, Kelly KM, Fu M, Kollef MH. Transfusion practice and blood stream infections in critically ill patients. Chest 2005;127:1722–8. [74] Blaine KP, Tuohy MJ, Wilson D, Procop GW, Tisch DJ, Shrestha NK, et al. Progression to bacteremia in critical care patients colonized with methicillin-resistant Staphylo-

[75]

[76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86]

[87] [88] [89]

[90]

[91]

[92] [93]

[94] [95]

[96]

[97]

[98]

1389

coccus aureus expressing Panton-Valentine leukocidin. Diagn Microbiol Infect Dis 2010;68:28–33. Miller RS, Norris PR, Jenkins JM, Talbot III TR, Starmer JM, Hutchison SA, et al. Systems initiatives reduce healthcare-associated infections: a study of 22,928 device days in a single trauma unit. J Trauma 2010;68:23–31. Cowper DC, Kubal JD, Maynard C, Hynes DM. A primer and comparative review of major US mortality databases. Ann Epidemiol 2002;12:462–8. Sohn MW, Arnold N, Maynard C, Hynes DM. Accuracy and completeness of mortality data in the Department of Veterans Affairs. Popul Health Manag 2006;4:2. Schisterman EF, Whitcomb BW. Use of the Social Security Administration Death Master File for ascertainment of mortality status. Popul Health Manag 2004;2:2. Newman TB, Brown AN. Use of commercial record linkage software and vital statistics to identify patient deaths. J Am Med Inform Assoc 1997;4:233–7. Akaike H. A new look at the statistical model identification. IEEE Trans Autom Control 1974;19:716–23. Hosmer DW, Lemeshow S. Goodness of fit tests for the multiple logistic regression model. Communications in Statistics– Theory and. Methods 1980;9:1043–69. Cleveland W. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979;74:829–36. Cleveland W, Devlin S. Locally-Weighted Regression: An Approach to Regression Analysis by Local Fitting. J Am Stat Assoc 1988;83:596–610. Rosenbaum P, Rubin D. Constructing a Control Group Using Multivariate Matched Sampling Methods That Incorporate the Propensity Score. Am Stat 1985;39:33–8. D'Agostino Jr RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998;17:2265–81. Leuven E, Sianesi B. PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing; 2003. Eltzschig HK, Collard CD. Vascular ischaemia and reperfusion injury. Br Med Bull 2004;70:71–86. International classification of diseases and related health problems. 10th revision. Geneva: World Health Organization; 2007. Harwood H. Updating estimates of the economic costs of alcohol abuse in the United States: estimates, update methods, and data. Bethesda, MD: National Institute on Alcohol Abuse and Alcoholism; 2000. Alexandre PK, Roebuck MC, French MT, Chitwood DD, McCoy CB. Problem drinking, health services utilization, and the cost of medical care. Recent Dev Alcohol 2001;15: 285–98. Cryer PC, Jenkins LM, Cook AC, Ditchburn JS, Harris CK, Davis AR, et al. The use of acute and preventative medical services by a general population: relationship to alcohol consumption. Addiction 1999;94:1523–32. Greiffenstein P, Molina PE. Alcohol-induced alterations on host defense after traumatic injury. J Trauma 2008;64:230–40. Radek KA, Matthies AM, Burns AL, Heinrich SA, Kovacs EJ, Dipietro LA. Acute ethanol exposure impairs angiogenesis and the proliferative phase of wound healing. Am J Physiol Heart Circ Physiol 2005;289:H1084–90. Radek KA, Kovacs EJ, DiPietro LA. Matrix proteolytic activity during wound healing: modulation by acute ethanol exposure. Alcohol Clin Exp Res 2007;31:1045–52. Radek KA, Kovacs EJ, Gallo RL, DiPietro LA. Acute ethanol exposure disrupts VEGF receptor cell signaling in endothelial cells. Am J Physiol Heart Circ Physiol 2008;295: H174–84. Fitzgerald DJ, Radek KA, Chaar M, Faunce DE, DiPietro LA, Kovacs EJ. Effects of acute ethanol exposure on the early inflammatory response after excisional injury. Alcohol Clin Exp Res 2007;31:317–23. Beier K, Eppanapally S, Bazick HS, Chang D, Mahadevappa K, Gibbons FK, et al. Elevation of blood urea nitrogen is predictive of long-term mortality in critically ill patients independent of "normal" creatinine. Crit Care Med 2011;39:305–13. McMahon GM, Mendu ML, Gibbons FK, Christopher KB. Association between hyperkalemia at critical care initiation and mortality. Intensive Care Med 2012;38: 1834–42.