Key words: alcoholism; ARDS; cirrhosis; lung injury; mortality; sepsis; smoking

Transcription

1 Risk Factors for ARDS in the United States* Analysis of the 1993 National Mortality Followback Study Terri TenHoor, MD; David M. Mannino, MD, FCCP; and Marc Moss, MD Objective: To identify specific comorbid factors that are present in US decedents with ARDS. Design: We searched the 1993 National Mortality Followback Study for all decedents who had a code for ARDS mentioned on their death certificate. We also searched for comorbid conditions both on the death certificates (sepsis, medical or surgical misadventures, cirrhosis) and in the study database (current or former smoking, use of alcohol at least 3 d/wk, race, gender, and age). We calculated proportional mortality ratios (PMRs) for these risk factors. Results: Of the 19,003 decedents for whom data were available, 252 decedents, representing an estimated 19,460 US decedents, had ARDS listed on their death certificate. PMRs among decedents with ARDS were significantly increased for medical or surgical misadventures (PMR, 11.8; 95% confidence interval [CI], 3.8 to 36.7), sepsis (PMR, 5.6; 95% CI, 2.0 to 16.0), nonwhite race (PMR, 2.6; 95% CI, 1.4 to 5.0), and cirrhosis (PMR, 2.2; 95% CI, 1.1 to 4.6). PMRs were increased but not statistically significant for current smokers (PMR, 1.2; 95% CI, 0.5 to 3.0) or former smokers (PMR, 1.8; 95% CI, 0.7 to 4.3) compared to never smokers, and drinking alcohol on > 3 d/wk in the year prior to death, when compared to drinking alcohol less than < 3 d/wk (PMR, 1.8; 95% CI, 0.6 to 4.9). Conclusions: The results of this study confirm the positive associations between ARDS mortality and the presence of sepsis and cirrhosis, and suggest possible new relationships between ARDS mortality and nonwhite individuals and patients with medical or surgical misadventures. (CHEST 2001; 119: ) Key words: alcoholism; ARDS; cirrhosis; lung injury; mortality; sepsis; smoking Abbreviations: CI confidence interval; ICD-9 International Classification of Diseases, Ninth Revision; NMFS National Mortality Followback Study; PMR proportional mortality ratio ARDS is characterized by the sudden onset of severe hypoxemia and diffuse bilateral pulmonary infiltrates in the absence of left atrial hypertension. 1 ARDS has been reported to occur in critically ill patients with a variety of diagnoses and can present in individuals without any prior history of lung disease. Over the last 30 years, great strides *From the Division of Pulmonary and Critical Care Medicine (Drs. TenHoor and Moss), Departments of Medicine, Emory University School of Medicine and Crawford Long Hospital of Emory University, Atlanta; and Air Pollution and Respiratory Health Branch (Dr. Mannino), Division of Environmental Hazards and Health Effects, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA. Funding was provided by grant R01-AA A2 from the National Institutes of Health. Manuscript received February 22, 2000; revision accepted August 31, Correspondence to: Marc Moss, MD, Crawford Long Hospital of Emory University, Suite 5310, 550 Peachtree St, NE, Atlanta, GA ; marc_moss@emory.org have been accomplished in unraveling the complex pathogenesis of ARDS; however, studies that have advanced our knowledge of the epidemiology of ARDS are relatively few. 2 A better understanding of the incidence and mortality from ARDS is mandatory for clinical trial design and the development of new therapies that will improve clinical outcome for patients with ARDS. Hudson and colleagues 3 studied 695 patients with a variety of diagnoses associated with the development of ARDS. In trauma patients, mortality was significantly higher in those individuals who developed ARDS (58%) when compared to trauma patients who did not develop ARDS (12.9%). In 1998, Monchi and colleagues 4 studied the early predictive factors of survival in 259 patients receiving mechanical ventilation who met ARDS criteria. They determined that the risk of death in ARDS patients was related to several demographic and physiologic fac- CHEST / 119 / 4/ APRIL,

2 tors, including severity of illness on the first day of ARDS, length of mechanical ventilation, cirrhosis, and right ventricular dysfunction. In an earlier study, Doyle and colleagues 5 examined patients with acute lung injury and determined that the presence of nonpulmonary organ dysfunction, chronic liver disease, and sepsis were predictors of a higher mortality. The results of the above studies are based on data from single institutions. These data registries have the advantage of systematically collecting predefined data on a relatively large number of patients over time. 6 However, information derived from data registries may not generalize to other clinical settings or academic centers. 6 In this study, we sought to identify preexisting factors that are associated with ARDS mortality. Using a large US national mortality data set, we investigated the effect of various comorbid conditions and demographic factors on the proportional mortality ratio for ARDS. Materials and Methods The 1993 National Mortality Followback Study (NMFS) is a multicomponent survey conducted by the National Center for Health Statistics designed to supplement information from death certificates in the vital statistics file with information on important characteristics of the decedent. The survey included individuals aged 15 years who died in 1993 in the United States. A total of 22,957 death certificates or approximately 1% of the total cohort was sampled. In order to produce more robust analyses, certain causes of death (ie, homicide, suicide, and accidental injury), black decedents, and certain age groups were oversampled, accounting for 45.5% of the total sample. The decedent s next of kin or other appropriate person familiar with the decedent s life history was contacted and asked to participate in the survey. The overall participation rate was 83%, with data obtained on 19,003 decedents. Data on these decedents were assigned weights to reflect the 2.2 million decedents in the United States in We used the International Classification of Diseases, Ninth Revision (ICD-9) to categorize the decedents by disease. 7 We searched the 1993 NMFS for subjects who had a code for ARDS (ICD-9 code of 518.4, 518.5, or 518.8) listed anywhere on their death certificate, and excluded subjects who had any listing of congestive heart failure (ICD-9 code of ). A total of 252 individuals or 1.3% of the sample of decedents met these criteria, representing a total estimated 19,460 decedents from ARDS or 0.8% of all US decedents in Additional predetermined codes searched for and used in the analysis included sepsis (ICD-9 code of ), cirrhosis (ICD-9 code of ), injury (ICD ), and medical or surgical misadventures (ICD-9 code of E870- E879.9). 7 A medical or surgical misadventure included accidental perforations during medical or surgical care, foreign objects left in body, failure of sterile precautions during a procedure, failure in dosage, contamination of blood or other fluids, and related problems. Most of these diagnoses were selected due to previous studies 3 5 that identified a positive association between the specific condition and ARDS mortality. The database was also searched to identify the race (white vs nonwhite), gender, age ( 64 years or 64 years), and use of alcohol or cigarettes of the decedent. In order to stratify the decedents according to their tobacco and alcohol use, we used information obtained during the interview with family members of the decedent. The interview included questions regarding the decedent s general use of tobacco and consumption of alcohol in the year prior to death. With regard to alcohol use, we stratified decedents into two categories: those who were reported to have used alcohol at least 3 d/wk during the final year of their life and those who used alcohol less frequently. With regard to tobacco use, we stratified decedents into three categories: current, former, and never smokers. Both current and former smokers were reported to have smoked at least 100 cigarettes in his/her entire life, and former smokers had stopped smoking and not started again. Due to the fact that this is a study of decedents, our primary outcome variable was the proportional mortality ratio (PMR) for the risk of death from ARDS. The PMR is defined as the proportion of exposed subjects assigned to a specific cause of death compared with the proportion of unexposed subjects assigned to the same specific cause of death. 8 This measure describes the relative importance of an exposure on a specific cause of death in relation to all deaths in a population group. This concept is best illustrated by the following equation: A/A B PMR C/C D where A exposed individuals who died of a specific disease; B exposed individuals who died of all other diseases; C unexposed individuals who died of a specific disease; and D unexposed individuals who died of all other diseases. Statistics All analyses were performed using SAS software (SAS Institute; Cary, NC) or SUDAAN (Research Triangle Institute; Research Triangle Park, NC) and are reflective of the weighted number of decedents. 9,10 SUDAAN is needed to adjust for the complex sample design in calculating the variances and confidence intervals (CIs). Multivariate logistic regression analysis was determined with all of the independent variables (sepsis, cirrhosis, surgical misadventures, injury, race, sex, age, current smoker, former smoker, and alcohol use at least 3 days weekly during the final year of life) included in the final model. In this model, the dependent variable was the categorical variable: death from ARDS or death from all other causes. Results Data were available for 19,003 decedents in the 1993 NMFS, which was weighted to reflect the 2.2 million annual deaths. The overall weighted study population included 49% female and 51% male decedents. Nearly 86% of the decedents were white, and 14.1% were nonwhite. Over 1.6 million (74.4%) of the decedents were 64 years old at the time of their death. In the entire study cohort, 4.1% (n 91,370) had a diagnosis of sepsis, 1.2% (n 26,470) had a diagnoses of a surgical or medical misadventure, 0.6% (n 12,450) had cirrhosis, and 8.4% (n 185,720) had major injury listed on their death certificate. With regard to smoking, 21.5% (n 476,220) were current smokers and 32.5% 1180 Clinical Investigations in Critical Care

3 (n 719,820) were former smokers. In regard to alcohol use, 10.2% (n 226,520) drank alcohol 3 d/wk during the last year of their life. The estimated number and percentage of each comorbid condition and demographic factor for decedents with ARDS and decedents without ARDS are displayed in Table 1. In a univariate analysis, decedents with ARDS were statistically more likely to have a concurrent diagnosis of sepsis, cirrhosis, or a medical/surgical misadventure (Table 2). In addition, the unadjusted PMR of ARDS was statistically higher for nonwhite decedents when compared to white decedents. Decedents with ARDS were similar to decedents without ARDS with regard to gender and age. Nearly 16% of the ARDS decedents drank alcohol on average three or more times a week during the year prior to death, and 63.5% of ARDS decedents were classified as either current or former smokers (Table 1). The unadjusted PMR for ARDS was higher for both individuals who drank alcohol on average 3 d/wk, in former smokers, and patients with injury. However these differences did not reach statistical significance (Table 2). The results of the multivariate model demonstrated that sepsis, surgical misadventure, cirrhosis, and nonwhite race were significantly overrepresented in decedents with ARDS when compared to all other decedents. There was a trend toward finding effects of former cigarette smoking and alcohol use on ARDS mortality (Table 2), but CIs for both of these findings were wide and included unity. With regard to injury, the adjusted PMR was 0.9, indicating that the initial association noted in the univariate analysis was likely due to the presence of a confounding variable. Discussion In this study, we examined the impact of several comorbid conditions and demographic variables on ARDS mortality using a large national database. The 1993 NMFS is unique in that information is obtained from both death certificate records as well as from extensive survivor interviews. Our results indicate that individuals who died with ARDS were more likely to have a diagnosis of sepsis, medical or surgical misadventure, or cirrhosis. In addition, the adjusted PMR was significantly higher for nonwhites when compared to whites. Finally, both former cigarette smoking and alcohol use showed independent trends toward an effect on ARDS mortality, although these effects were not statistically significant. The association between sepsis and ARDS is well supported in the medical literature. Based on a meta-analysis of 83 articles, a causal association between ARDS and sepsis is supported with the highest level of clinical evidence. 11 This relationship between sepsis and ARDS is also extremely common. Several studies have reported that greater than one third of the ARDS cases are observed in patients with preexisting sepsis. It also appears that patients with sepsis-induced ARDS have a higher mortality compared to ARDS patients associated with other risk factors such as trauma. 3 The results of our study demonstrate a markedly increased adjusted PMR for sepsis of 5, strengthening the association between sepsis and ARDS. Cirrhosis of the liver is a comorbid condition that also appears to influence both the incidence and severity of ARDS. Although patients with cirrhosis of the liver are predisposed to develop sepsis, 15 only a Table US Decedents With and Without ARDS (ICD ) and Comorbid Conditions of Sepsis (ICD ), Cirrhosis (ICD ), Surgical Misadventures (ICD-9 E870 E879.9), and Injury (ICD ); or Comorbid Risk Factors* Variables No. (total 252) Estimated Decedents With ARDS (total 19,460) Percent (SE) No. (total 18,751) Estimated Decedents Without ARDS (total 2,193,220) Percent (SE) Sepsis 37 4, (8.4) , (0.4) Medical or surgical misadventure 12 2, (4.6) , (0.2) Cirrhosis (0.4) , (0.5) Injury 65 2, (8.3) 8, , (0.3) Nonwhite 95 5, (6.8) 6, , (0.3) Male 116 9, (9.1) 11,756 1,119, (0.2) Age 64 yr 72 14, (6.0) 6,665 1,633, (0.8) Current smoker 85 4, (6.6) 6, , (0.8) Former smoker 50 8, (9.4) 3, , (0.9) Alcohol use 3 d/wk 48 3, (6.2) 3, , (0.6) * Unless otherwise indicated, data are presented as No. Numbers are unweighted, whereas the estimated number of decedents and the percentages are weighted. CHEST / 119 / 4/ APRIL,

4 Table 2 PMRs of 1993 US Decedents With ARDS (ICD ) Both Unadjusted and Adjusted for Other Covariates Variables Unadjusted PMR (95% CI) Adjusted PMR (95% CI) Sepsis 6.4 (2.3, 17.5) 5.6 (2.0, 16.0) Cirrhosis 3.7 (2.2, 6.1) 2.2 (1.1, 4.6) Medical or surgical misadventure 10.3 (2.7, 39.0) 11.8 (3.8, 36.7) Injury 1.9 (0.7, 5.0) 0.9 (0.6, 1.4) Nonwhite 2.6 (1.3, 4.9) 2.6 (1.4, 5.0) Male 1.0 (0.5, 2.1) 0.8 (0.4, 1.8) Age 64 yr 0.9 (0.5, 1.7) 1.1 (0.6, 2.0) Current smoking 1.0 (0.5, 2.2) 1.2 (0.5, 3.0) Former smoking 1.5 (0.7, 3.2) 1.8 (0.7, 4.3) Alcohol use 3 d/wk 1.7 (0.7, 4.2) 1.8 (0.6, 4.9) few reports have examined a possible association between cirrhosis and the development of ARDS. Matuschak and colleagues 16 retrospectively examined 29 patients with severe liver disease awaiting transplantation, and reported that their incidence of ARDS was higher than a random control group of ICU patients. Recently, Doyle et al 5 reported that the mortality from acute lung injury was increased in 26 patients with chronic liver disease, when compared to acute lung injury patients without liver disease. Finally cirrhosis was identified as the single most important predictive variable of mortality in a cohort of 259 patients with ARDS. 4 Proposed mechanisms for the association between cirrhosis and ARDS include the following: increased pulmonary leukotriene B 4 and C 4 concentration, nitric oxide expression, impaired systemic clearance of endotoxin, and altered systemic glutathione homeostasis We have previously reported 12 an association between a history of chronic alcohol abuse and the development of ARDS. In a study of 351 critically ill patients with at least one at-risk diagnosis, the incidence and severity of ARDS were significantly increased in those patients with a history of chronic alcohol abuse. The effect was most significant in those patients with sepsis as their at-risk diagnosis. In our present study, we have attempted to quantify the amount of alcohol use associated with an increase in ARDS mortality. There appears to be a trend toward a higher PMR for ARDS in those decedents who drank alcohol on average greater than three times per week in their last year of life. In this study, we also reported a higher adjusted PMR of 1.8 for individuals who were former smokers. Iribarren and colleagues 21 have recently reported the possible association between cigarette smoking and the development of ARDS. They examined 120,000 health-plan subscribers, of whom 56 individuals developed ARDS. ARDS was independently related to current cigarette smoking of 20 cigarettes per day, with an odds ratio of One explanation for the nonstatistically significant adjusted PMR for these variables may be a lack of power due to the sample size in our study. To detect a 50% increase in the effect of alcohol use or tobacco use on ARDS mortality, given that ARDS occurs among 1% of all decedents, we would need a sample size of 34,000 decedents to have a power of 0.80 with a 95% CI. In this study, we also identified two potentially new factors that are associated with ARDS mortality: medical or surgical misadventure, and nonwhite race. The category misadventure is clearly a heterogeneous group of disorders that are linked by the delivery of nonoptimal medical care. It appears that the association of these misadventures is not totally related to the development of sepsis, as the effects of the misadventures on ARDS mortality remained in a multivariate analysis that included sepsis in the model. Some of these deaths may be due to the occurrence of transfusion-related lung injury, which has been reported in the medical literature. 22 In addition, the association between medical or surgical misadventures and ARDS mortality is derived from an actual total of 12 cases and future studies will clearly be necessary to better define and quantify this possible relationship. The effect of race on ARDS mortality has only recently been explored. In this study, we reported an increased adjusted PMR for ARDS in decedents of a nonwhite race. It is possible that this difference may be due to factors such as access to health care, which has been found with other illnesses such as myocardial infarction and stroke. 23,24 It is also possible that nonwhite individuals are more likely to develop specific at-risk diagnoses associated with the development of ARDS that were not included in our multivariate analysis. Iribarren et al 21 reported that white individuals are more likely to develop ARDS when compared to African Americans. Differences in our results may be caused by variability in the outcome measures (incidence of ARDS vs ARDS mortality) or discrepancies 1182 Clinical Investigations in Critical Care

5 in classification between the two studies (African American vs all nonwhite). Clearly, further studies are needed to determine whether the association between nonwhite race and ARDS mortality is truly significant. In contrast to several other reports, we did not identify an increase in the PMR for age. Suchyta and colleagues 25 reported a mortality of 64% in ARDS patients 55 years old, and a mortality rate of only 45% in those ARDS patients 55 years old. This difference was statistically significant, and there were no differences between the two groups in regard to gender, smoking history, at-risk diagnosis, or severity of illness. 25 Maunder et al 26 reported a 55% mortality in ARDS patients 65 years old and 97% in those 65 years old. One possible explanation for the differences in conclusions may be the use of a PMR as opposed to a true mortality rate. The use of a PMR to evaluate risk of death has some epidemiologic limitations. The premise of a proportional mortality study is that if the exposure causes a specific fatal illness, there should be proportionately more deaths from that illness among dead people who had been exposed than among dead people who had not been exposed. However, a PMR unfortunately cannot distinguish between whether the exposure is associated with an excess of deaths in the disease of interest or actually a reduction of deaths in the reference or control population. 8 In addition, proportional mortality studies cannot determine the extent to which the exposure or risk factor increases the incidence of the disease of interest or alters the prognosis for individuals with the disease. The most striking limitation to this study is the use of ICD-9 codes for the identification of patients with ARDS. It is unclear whether ICD-9 coding was used consistently in this study. Two studies have examined the validity of ICD-9 coding for ARDS for the diagnoses of ARDS in hospital discharge records. Using a similar ICD-9 definition of ARDS, Rubenfeld and colleagues 27 reported a sensitivity of 74% and a specificity of 97% when compared to careful chart review utilizing the American European Consensus Conference Criteria for ARDS. Thomsen and Morris 28 demonstrated that the combination of ICD-9 discharge diagnosis codes (518.5, , and ) had a sensitivity of 88% and a specificity of 99% for ARDS diagnosis at a hospital in Salt Lake City. However, the incidence of the diagnosis of ARDS in both of these studies was substantially low, so that the positive predictive value for ICD-9 coding was 11 to 38%. In addition, the use of ICD-9 coding in identifying the other independent variables, such as sepsis, cirrhosis, and surgical or medical misadventure, may be imprecise. Clearly the applicability of ICD-9 codes for ARDS mortality needs to be examined more closely in future studies. Other limitations include that the NMFS was conducted in 1993, and some of the associations that we identified between ARDS mortality and comorbid factors may have changed over time. Finally, even though we performed logistic regression analyses to account for confounding variables, it is possible that we did not adjust for some important confounding variables that may alter our results. The results of this study are important for several reasons. We have confirmed a positive association between ARDS mortality and several important medical disorders, such as sepsis and cirrhosis, using a national database. We have also identified two potentially new factors, medical/surgical misadventure and nonwhite race, that are associated with an increased proportional mortality ratio for ARDS. These findings require confirmation by others. In addition, future studies are necessary to define the utility of definitions of ARDS based on ICD-9 codes. References 1 Bernard GR, Artigas A, Brigham KL, et al. Report of the American-European Consensus Conference on acute respiratory distress syndrome: definitions, mechanisms, relevant outcomes, and clinical trial coordination. Am J Respir Crit Care Med 1994; 151: Luhr OR, Antonsen K, Karlsson M, et al. Incidence and mortality after acute respiratory failure and acute respiratory distress syndrome in Sweden, Denmark, and Iceland. Am J Respir Crit Care Med 1999; 159: Hudson LD, Milberg JA, Anardi D, et al. Clinical risks for development of acute respiratory distress syndrome. Am J Respir Crit Care Med 1995; 151: Monchi M, Bellenfant F, Cariou A, et al. Early predictive factors of survival in the acute respiratory distress syndrome: a multivariate analysis. Am J Respir Crit Care Med 1998; 158: Doyle RL, Szaflarski N, Modin GW, et al. Identification of patients with acute lung injury: predictors of mortality. Am J Respir Crit Care Med 1995; 152: Rubenfeld GD, Angus DC, Pinsky MR, et al. Outcomes research in critical care: results of the American Thoracic Society Critical Care Assembly workshop on outcomes research. Am J Respir Crit Care Med 1999; 160: US Public Health Service. International classification of diseases, ninth revision. Washington, DC: US Government Printing Office, 1980; DHHS Publication No. (PHS) Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven, 1998; SAS language: reference, version 6. Cary, NC: SAS Institute, Shah BV, Barnwell BG, Hunt PN, et al. SUDAAN user s manual: release Research Triangle Park, NC: Research Triangle Institute, Garber BG, Hebert PC, Yelle J-D, et al. Adult respiratory distress syndrome: a systematic overview of incidence and risk factors. Crit Care Med 1996; 24: Moss M, Butcher B, Moore FA, et al. The role of chronic alcohol abuse in the development of acute respiratory distress CHEST / 119 / 4/ APRIL,

6 syndrome in adults. JAMA 1996; 275: Sloane PJ, Gee MH, Gottlieb JE, et al. A multicenter registry of patients with adult respiratory distress syndrome: physiology and outcome. Am Rev Respir Dis 1992; 146: Knaus WA, Sun X, Hakim RB, et al. Evaluation of definitions for adult respiratory distress syndrome. Am J Respir Crit Care Med 1994; 150: Wyke RJ. Problems of bacterial infection in patients with liver disease. Gut 1987; 28: Matuschak GM, Rinaldo J, Pinsky MR, et al. Effect of end-stage liver failure on the incidence and resolution of the adult respiratory distress syndrome. J Crit Care 1987; 2: Matushak GM. Lung-liver interactions is sepsis and multiple organ failure syndrome. Clin Chest Med 1996; 17: Kaplan JE, Saba TM. Platelet removal from the circulation by the liver and spleen. Am J Physiol 1978; 235:H314 H Matuschak GM, Pinsky MR, Klein EC, et al. Effects of D-galactosamine-induced acute liver injury on mortality and pulmonary responses to Escherichia coli lipopolysaccharide: modulation by arachidonic acid metabolites. Am Rev Respir Dis 1990; 141: Altomare E, Vendemiale G, Albano O. Hepatic glutathione content in patients with alcoholic and non alcoholic liver diseases. Life Sci 1988; 43: Iribarren C, Jacobs DR, Sidney S, et al. Cigarette smoking, alcohol consumption, and risk of ARDS. Chest 2000; 117: Popovsky MA. Transfusion-related acute lung injury. Transfusion 1995; 32: Kunst AE, Groenhof F, Mackenbach JP. Occupational class and cause specific mortality in middle aged men in eleven European countries: comparison of population based studies. BMJ 1998; 316: Williams JE, Massing M, Rosamond WD, et al. Racial disparites in coronary heart disease mortality from in the state economic areas surrounding the ARIC study communities. Ann Epidemiol 1999; 9: Suchyta MR, Clemmer TP, Elliot CG, et al. Increased mortality of older patients with acute respiratory distress syndrome. Chest 1997; 111: Maunder RJ, Kubillis PS, Anardi DM, et al. Determinants of survival in the adult respiratory distress syndrome [abstract]. Am Rev Respir Dis 1989; 139:A Rubenfeld GD, Caldwell ES, Steinberg KS, et al. ICD-9 codes do not accurately identify patients with the acute respiratory distress syndrome (ARDS) [abstract]. Am J Respir Crit Care Med 1998; 157:A68 28 Thomsen GE, Morris AH. Incidence of the adult respiratory distress syndrome in the state of Utah. Am J Respir Crit Care Med 1995; 152: Clinical Investigations in Critical Care