Increased Mortality of Older Patients With Acute Respiratory Distress Syndrome

Increased Mortality of Older Patients With Acute Respiratory Distress Syndrome* Mary R. Suchyta, DO; Terry P. Clemmer, MD, FCCP; C. Gregory Elliott, MD, FCCP; James F. Orme, Jr, MD; Alan H. Morris, MD; Jay Jacobson, MD; and Ron Menlove, PhD 1

Objective: To examine the relationship between age and mortality in ARDS patients and evaluate the importance of factors that increase the mortality of older ARDS patients. Design: Prospective inception cohort study. Setting: Community-based referral hospital. Patients: Two hundred fifty-six ARDS patients identified from May 1987 to December 1990. ARDS was defined by the following: (1) Pa0 2/PA02 ~0.2; (2) pulmonary capillary wedge pressure ~15 mm Hg; (3) total static thoracic compliance ~50 mUcm H 2 0; (4) bilateral infiltrates on chest radiograph; and (5) an appropriate clinical setting for ARDS. Main outcome measures: Comparison of organ failure, incidence of sepsis, patient demographics, arterial oxygenation, and level of support in those 55 years and younger and those older than 55 years of age. Withdrawal of support in patients who died. Results: Seventy-two of 112 patients older than 55 years (64%) died vs 65 of 144 patients 55 years and younger (45%) (p=0.002). Examination of patient groups using age identified older than 55 years as a "cutpoint" above which mortality was greater (p=0.002). Older nonsurvivors did not differ from nonsurvivors 55 years or younger with respect to gender, smoking history, ARDS risk factors, ARDS identifying characteristics, APACHE II (acute physiology and chronic health evaluation), number of organ failures, or the incidence of sepsis. In the 48 h prior to death, nonsurvivors 55 years and younger had more organ failure (3.4±0.2 vs 2.8±0.2; p=0.03), higher fraction of inspired oxygen (0.82±0.03 vs 0.68±0.03; p=0.008), and higher positive endexpiratory pressure levels (13± 1 vs 8± 1; p=0.001) than older nonsurvivors. Despite more severe expression of disease, only 32 (50%) nonsurvivors 55 years and younger had support withdrawn. Significantly more nonsurvivors older than 55 years (73%) had support withdrawn (p=0.009). Even in the absence of chronic disease states, withdrawal was more likely for patients older than 55 years (21151) than in those 55 years and younger (3/32; p<0.001). Conclusions: Mortality is significantly higher for patients with ARDS older than 55 years. Decisions to withdraw support are made more often in ARDS patients older than 55 years. These data suggest that age bias may influence decisions to withdraw support. (CHEST 1997; 111:1334-39) Abbreviations: APAC HE = acute phys iology and chronic health evaluation; Cth = thoracic compliance; Fio 2 = fraction of inspired m;ygen; NCR = no cardiac resuscitation; PE EP = positive end-expiratory press ure

p revious studies have demonstrated that older

patients are less likely to survive ARDS. 1 -5 Maunder et al 1 reported a 55% mortality in ARDS patients *From the Shockfrrauma/lntermou ntain Respiratmy Intensive Care Unit, Pulmonary and Medical Ethics Division, Department of Internal Medicine, LDS Hospital and the Division of Respiratory, Critical Care, and Pulmonary Medicine, University of Utah, Salt Lake City. Supported in part by grant HL36787 from the National Insti tutes of H ealth, the Deseret F oun dation, and the Respiratory Distress Syndrome Foundation. 1 D eceased. Manuscript received August 1 , 1996; revision accepted Octobe r 18. 1334

younger than 65 years old and 97% in those older than 65 years old. Napolitano et al2 found that 82% of ARDS patients older than 70 years old died. Similarly, Sloane et aP observed a mortality of 72% in ARDS patients 60 years and older compared to 37% in those younger than 60 years. Gee et al4 reported that 69% of nonhypothermic ARDS patients 60 years and older died, compared with only 12% of patients younger than 60 years old. Finally, we also observed that ARDS patients older than 65 years old were more likely to die (66% vs 53%). 5 Mechanisms that underlie the increased mortality of older ARDS patients remain uncertain. Both Gee Clinical Investigations in Critical Care

et al 4 and Napolitano et al2 found that older ARDS nonsurvivors had less severe arterial hypoxemia than younger nonsurvivors, suggesting that factors other than irreversible hypoxemia conbibute to the increased mortality of older ARDS patients. Gee et al4 observed significantly lower heart rates and cardiac outputs, and higher serum urea nitrogen and creatinine concentrations; and they hypothesized that age-related declines in cardiovascular function contributed to the poor outcome of older ARDS patients. Other factors may increase the mortality of elderly ARDS patients, including diffe rences in the incidence of sepsis, 6•7 organ failure ,7 or the presence of chronic underlying health problems. Age bias, through its influence on clinical decision-making, has been observed in a variety of medical situations, including renal transplantation and elderly ICU admissions.8·9 If age bias exists in ARDS patients, it could contribute to outcome differences in older ARDS patients. To further address mortality issues in ARDS, we examined the clinical features associated with increased mortality in older ARDS patients using a cohort of previously identified patients.

MATERIALS AND METHODS We prospectively identified ARDS patients from May 1987 to December 1990 at the LDS Hospital, a 520-bed tertiary referral center. 10 This report includes all patients from th e previous reports with the addition of 41 patients from April 1990 to December 1990. 5 .1 1 ARDS was defined by the simultaneous presence of the following: (1) Pa0 2/PA0 2 :s0.2; (2) pulmonary capillary wedge pressure :s15 mm Hg; (3) total static thoracic compliance (Cth ) :550 mUcm H 2 0 ; (4) bilateral infiltrates on chest radiograph; and (5) an appropriate clinical setting for the development of ARDS. Data compiled for each patient included the following: (l) demographic information; (2) presumed 1isk factor for ARDS; (3) routin e daily laboratory measurements; (4) cardiovascular monitoring (pulmonaty artery pressures, thermodilution cardiac outputs, vital signs); and (5) pulmonary physiologic measurements (arterial blood gases, Cth , positive end-expiratory pressure [PEEP]). An APACHE II (acute physiology and chronic health eval uation ) score 12 was calculated for the first 24 h in the ICU. A list of risk factors for ARDS was developed from previous descriptions of ARDS populations.13·14 Organ failure was determined daily using the severe dysfunction criteria of Montgo mery et al. 13 A score of 1 was assigned to each organ failure identified at any time during the course of ARDS. Cumulative scores were determined by adding up the total number of organs that failed at least once before ARDS onset and those that failed at least once after ARDS onset. Sepsis was evaluated daily using the severe dysfunction criteria of Montgomery et al.'3 Sepsis evaluations prior to th e development of ARDS were carried out by review of the chart soon aft er the patient was identified b y the prospective ARDS screening. Data obtained for all nonsmvivors dming the 48 h prior to death included number of organs failing (by the crite ria of Montgome1y et alL>), use of dialysis for re nal failure, use of vasopressor agents, evidence of sepsis, 13 and level of ventilatory

support (fraction of inspired oxygen [Flo 2 ], PEEP, and Pa0 2 ). Data were collected at 48 h prior to death. Care was taken to assure that these data points were not merely preterminal (cardiopulmonary arrest) data. Patients vvi.th do-not-resuscitate orders prior to th e 48 h were excluded from nonsurvivor analyses. The number of days of ventilatory support after the development of ARDS also was recorded. Chronic disease was defined as irreversible anoxic brain injury, severe heart disease (New York Heart Association functional class IV), chronic renal failure, hepatic failure, cancer (actively treated disease), severe COPD, or AIDSJ 2 Patient medical records were evaluated regarding \vithdrawal of support. Withdrawal or potential withdrawal of su ppmt was defined as (1) removal of ventilator or (2) removal of pressor support or (3) a no-cardiac-resuscitation (NC R) order or (4) an NCR \vith no escalation of support despite clinical deterioration that would normally mandate escalation of therapy. For purposes of analyses, all of the categories were com bined and referred to as withdrawal. The end point was death or survival at the time of hospital discharge. Data were analyzed b yindependent t test and Pearson x2 analysis and expressed as means±SEM. Examination of patient groups (using 5-year age groupings and x2 analysis) determined the cutpoint for the age at which mortality was significantly different. Contingency table analyses (comparing survivors older th an 55 years or 55 years and younger with nonsurvivors older than 55 years or 55 years and younger) were used to identify differences between survivors and nonsurvivors in upper and lower age groupings. Differences were considered statistically significant at p<0.05. Bonferroni's correction was used to adjust p value significance level because we realized there number of comparisons 15 would be a arge l To assess the interrelationship of the multiple variables, we performed multivariate logistic regression analysis using survival as the dependent variable and age, sepsis, APACHE II, organ fllilure before and after ARDS onset, and gas exchange va1iables at the onset of ARDS (Table 1) as independent variables .

RESULTS

Two hundred fifty-six ARDS patients were identified in the 43-month study period. One hundred thilty-seven (54%) patients died. Mortality was significantly increased after age 55 years (72 of 112 patients older than 55 years died [64%] vs 65 of 144 patients aged 55 years and younger [45%]; p=0.002) (Fig 1). Population Characteristics

The population characteristics of the study patients are shown in Table l. Survivors and nonsurvivors did not differ with respect to multiple epidemiologic variables. There was a trend toward increased mortality in older patients with respect to female gender, a history of tobacco use, and a higher Pa0 2 and lower Fio2 at ARDS onset, but these did not reach statistical significance. In multivariate analysis, we found that age older than 55 years was the only statistically significant factor associated with survival. CHEST I 111 I 5 I MAY, 1997

1335

0.8 0.7 0.6 0.5

0.4 0.3 0.2

0.1

0

AGE (YEARS) FIGURE

l. Mortality in 5-year age groupings of '256 ARDS patients.

Table !-Population Characteristics* 55 yr and Younger SUR (n = 79) Fe male, No. (%) Smoking history positive, No. (%) Lethal disease present, No. (%) Risk factor for ARDS , No. (%) Pneumonia Sepsis Shock Trauma Other APACHE II Episodes of sepsis during ARDS Organs failing prior to ARDS Organs failing during ARDS Sepsis after ARDS onset, No. (%) At ARDS onset Pa0/ PA0 2 PaO, mm Hg Flo 2 PEEP, em H 2 0 Cth, mUcm H 2 0

NS (n = 6.5)

Older Than 55 yr SUR (n=40)

NS (n = 72)

p Value

47 (59) 36 (46) 1 (1)

35 (54) 16 (25 ) 14 (22 )

19 (48 ) 13 (33) 0 (0)

34 (47 ) 32 (44) 15 (21 )

0.07 0.08 0.32

32 (41 ) 27 (34) 1 (1) 10 (13) 9 (11 ) 16.5 1.0 0.7 2.0 14 (18)

20 (31) 16 (25) 9 (14) 9 (14) 11 (17) 20.8 0.7 0.9 2.7 21 (32)

22 (55) 11 (28) 1 (3) 2 (5) 4 (10 ) 18.3 1.0 0.9 1.8 4 (10 )

34 (47) '23 (32) 6 (8) 2 (3) 7 (10) 22.3 0.6 1.2 2.6 13 (18)

0.44 0.64 0.66 0.92 0.98 0.85 0.58 0.78 0.64 0.52

0.15 66 0.82 7 24

0.15 59 0.88 9 21

0.16 63 0.78 5 31

0.15 70 0.83 6 2

0.13 0.09 0.09 0.19 0.89

*All nonpercentile data are means. SUR=survivor; NS = nonsurvivor. 1336

Clinical Investigations in Critical Care

port was withdrawn more frequ ently in older nonsurvivors (p=0.008). At the time that ARDS was identified, no patient had NCR orders. No survivors had an NCR order or any order to withdraw supportive care. Three patients with NCR orders more than 48 h prior to death were dropped from the analyses, as it was postulated that their data might be affected by NCR orders . (Analyses including these patients did not yield statistically different results.) The remaining patients had (1) no escalation of therapy despite clinical deterioration (n=20), (2) withdrawal of ventilatory support (n=15), or (3) withdrawal of ventilatory support (n=48), with a fatal outcome in less than 48 h. Even in the absence of chronic health states, older patients had support withdrawn more frequently (p=0.001). More days of ventilatory support (p =0.005) and higher levels of PEEP and Flo 2 were observed in younger nonsurvivors (p < 0.001). We did not find evidence of an association between gender, organ failure, or pressor administration, and withdrawal decisions. Withdrawal in older nonsurvivors was unaffected by absolute age. (Of 70 nonsurvivors, 17 [24%] were withdrawn in the age group older than 55 to 66 years vs 13 [19%] older than 75 years.)

Table 2-Nonsurvivor Patient Characteristics* 55 yr and Younger (n = 64) Individual organ failure 48 h before death (% with fai lure present ) Live r Hematologic Cardiovascular Neurologic Re nal GI Hemorrhage Se psis Organ failures Vasopressor support, No. (%) Dialysis, No. (%) Days ventilatory support

Flo 2 PEEP, em H 2 0 Pa0 2 , mm Hg

Older Than 55 yr (n=70)

26 1 48 18 46 36 35 17 23 32 35 8 8 6 2 23 34 2.8::'::0.2 1 3.4::'::0.2 39 (56) 45 (70) 6 (8) 10 (15) 11::'::1 14 ::'::2 0.82::'::0.03 0.68::'::0.03 1 8::'::1 1 13::'::1 62::'::2 63::'::10

*All nonpe rcentile data are mean::':SEM. 'Statistically significant (ps 0.009 ).

Nonsurvivor Analyses Analysis of clinical variables for the 48-h period prior to death is shown in Table 2. Older nonsurvivors had significantly fewer organ failures (p=0.003). More young nonsurvivors had liver failure (p=0.009). Fewer older nonsurvivors had hematologic failure, but this did not meet statistical significance. Higher levels of support (PEEP and Flo2 ) with similar oxygenation values were observed in younger nonsurvivors (p<0.001 ).

DISCUSSION

The present study identified a significant increase in mortality for ARDS patients older than 55 years of age. Differences of gender, smoking history, incidence of sepsis, or organ failure (either number or specific organ) were not associated with the increased mortality of older ARDS patients. In fact , an argument could be made that ARDS was less severe in older ARDS patients as they were supported with significantly lower levels of PEEP and Flo 2 (while maintaining similar Pa0 2 ) in the 48 h preceding

Withdrawal of Support The relationship of withdrawal to the nonsurvivor population characteristics is shown in Table 3. Sup-

Table 3-Withdrawal Analysis (All Nonsurvivors)* 55 yr and Younge r

Chronic disease absent, No. (%) 48 h b efore death Organ failures Days ventilatory support All nonsurvivors All nonsurvivors without chronic disease present Vasopressor support, No. (%) F e male, No. (%)

Flo2

PEEP, em H 2 0 Pa0 2 , mm Hg

Olde r Than 55 yr

Withdrawal (n= 32)

No Withdrawal (n =32)

Withdrawal (n=51)

No Withdrawal (n=19)

3 (9)

27 (84)

21 (41 )

9 (47) 1

3.1

3.5

2.8

2.8

16::'::10 26::'::14 20 (63) 16 (50) 0.76::'::0.04 9::'::1 71::'::19

12 ::'::8 13::'::7 25 (78) 19 (59) 0.87::'::0.04 17::'::2 54::'::5

11 ::':: 7 10 ::'::7 23 (45) 22 (43) 0.62 ::':: 0.03 7::':: 1 65::'::3

11::'::8 11 ::'::9 1 16 (84) 11 (58) 0.85::'::0.06 1 9::'::1 1 59::'::6

*All nonpercentile data are mean::':SE M . 1 Statistically significant (p s 0.007). CHEST/111 /5/M AY, 1997

1337

death. Withdrawal was more likely in older ARDS patients in the absence of terminal illnesses. Although our study may not have provided absolute evidence for age bias, our obse1vations are compatible with the concept that age influences decisions to withdraw supportive care, and thereby contributes to the increased mortality of older ARDS patients. Withdrawal was unrelated to absolute age. If age was the only factor in withdrawal, one would expect that withdrawal would increase progressively with age. Our study suggests that withdrawal decisions are complex, and while age is important, it appears that once a patient receives a label of "old," absolute age may not be relevant. Previous investigators have described an increased mortality for older ARDS patients. However, the issue of age bias has not been addressed, nor have these investigators provided data to confirm or refute the existence of age bias. 1-4 Such data would be important, since the present study provides observations from a single institution and it is important to know if our findings are generalizable. Importantly, Gee et al4 observed that the ratio of arterial Po 2 to Fio 2 was greater and the level of PEEP was significantly less for their older ARDS patients who did not survive. In light of these observations, age bias may have contributed to the observed increased mortality of older ARDS patients in their population. We recognize that our center may differ from other hospitals. LDS Hospital is a referral center for the intermountain west. In addition, we were an ARDS referral center related to a trial of extracorporeal carbon dioxide removal during part of the time the patient cohort was identified. During this period, all ARDS patients referred for transport were transported, regardless of severity of illness or age. Only 41 patients were transferred during the period of this study, 10 of whom were older than 55 years. Mortality was 50% in both age groups (15 of 31 who were 55 years and younger died vs 5 of 10 who were older than 55 years died). This fact reflects that patients transported for the trial were not all young survivors, and it is therefore unlikely that the trial referrals accounted for the age bias that we observed. The biological characteristics of aging may conhibute to the higher case fatality rates of older ARDS patients. There is a reduced ability to respond adaptively to environmental change, 16 which may be related to structural and physiologic changes that accompany aging, eg, loss of cells, loss of function. 17 Furthermore, aging may slow or impair the process of tissue repair follovving an inflammatory injmy. Realizing that older patients might have chronic underlying health problems that influence withdrawal decisions (eg, chronic cardiac disease) or 1338

might be more susceptible to the development of terminal disease states during ARDS (eg, anoxic brain injury), we categorized nonsurvivors by the presence of chronic disease. This categmization did not explain the higher frequency of withdrawal of treatment in patients older than 55 years of age. Recognition of age bias in the management of ARDS is important for several reasons. First, ARDS often is not superimposed on chronic underlying disease, for which the long-term prognosis is poor. In the present study, 62 (55%) older patients had underlying chronic illnesses, and 20 (33%) survived. In contrast, 29 (20%) younger patients had chronic illnesses and none smvived. Second, although limited data are available for older ARDS survivors, most ARDS survivors are not severely impaired following recovery.18 Thus, a reasonable quality of life might be anticipated for survivors. Third, the increasing age of large numbers of Americans means that decisions regarding life-sustaining therapy for older ARDS patients will become more common, \vith attendant effects on health-care resources. The present study underscores the distinction between science and values in the management of ARDS . Decisions to withdraw life-sustaining therapy involve both quantifiable information related to disease and prognosis and value judgments of the health-care providers and the patient and/or his or her surrogates.19 In the case of ARDS, scientific evidence suggests that older ARDS patients are more likely to die.l- 5 Although the data are inconclusive, it raises the question of whether the withdrawal of supportive measures may contribute to the increased mortality of older ARDS patients. In light of the above question, we suggest that health-care providers should not conclude that the increased mortality of older ARDS patients demonstrates the medical futility of life-sustaining therapies for such patients. Rather health-care providers should recognize the interplay of science and attitudes when making decisions regarding life-sustaining therapies, and should seek additional scientific data to guide such decisions. Although medical hiage based on limited resources is ethically justifiable, such triage based solely on advanced age is not generally accepted. REFERENCES 1 Maunder RJ, Kubillis PS, Anardi DM, et al. Determinants of survival in the adult respiratory distress syndrome [abstract]. ·Am Rev Respir Dis 1989; 139:A220 2 Napolitano LM, Rutledge R, Meyer AA, et al. Adult respiratory distress syndrome in elderly ICU patients [abstract]. Crit Care Med 1991; l9:S90 3 Sloane PJ, Gee GH, Gottlieb JE, et al. A multicenter registry of patients with acute respiratory distress syndrome. Am Rev Respir Dis 1992; 146:419-26 Clinical Investigations in Critical Care

4 Gee MH , Gottlieb GE, Albertine KH , e t a!. Physiology of aging related to outcome in the adult respiratory distress syndrome. J Appl Physiol 1990; 69:822-29 .5 Suchyta MR, Cle mmer TP, Elliott CG, et a!. The adult respiratory distress syndrome: a report of survival and modifying factors. Chest 1992; 101:1074-79 6 Doyle RL, Szaflardki N, Modin W, et a!. Identification of patients with acute lung injury. Am J Respir Crit Care Med 1995; 152:1818-24 7 Bell RC, Coalson JJ, Smith JD, et a!. Multiple organ system failure and infection in adult respiratory distress syndrome. Ann Intern Med 1983; 99:293-98 8 Kjellstrand CM. Age, sex, and race inequality in renal transplantation. Arch Inte m M ed 1988; 148:1305-09 9 Chellur:i L, Grenv:ik A, Silverman M. Intensive care for critically ill elderly: mortality, costs and quality of life. Arch Intern Med 1995; 155:1013-22 10 Sittig DF, Elliott CG, Wallace CJ, et al. Screening for adult respiratory distress syndrome: use of the HELP hospital information system. J Clin Eng 1989; 4:237-43 11 Suchyta MR, Clemmer TP, Orme JF, e t a!. Increased survival of ARDS patients vv:ith severe hypoxe mia (ECMO ctiteria). Chest 1991; 99:951-55 12 Knaus WA, Draper EA, 'vVagner DP, et a!. APACHE II: a

13

14

15 16

17 18

19

severity of disease classification system. Cr:it Care Med 1985; 13:818-29 Montgomery AB, Stager MA, Carrico CJ, e t al. Causes of mortality in patients \v:ith adult respirat01y distress syndrome. Am Rev Respir Dis 1985; 132:485-89 Zapol WM , Fr:ikker MJ, Pontoppidan H, et al. The adult respiratory distress syndrome at the Massachusetts General Hospital: etiology, progression and smvival rates, 1978-1988. In: Zapol W, Lemaire F, eds. Adult respiratory distress syndrome. New York: Marcel Dekke r, 1991; 367-80 Armitage P. Statistical methods in medical research. Oxford: Blackwell Scientific, 1971; 150-59 Adelman RC. Hormone interaction during aging. In: Schimke RT, e d. Biological mechanism in aging. Washington, DC: US Dept of Health & Human Services, 1980; 686-90 Kmmpe PE, Knudson RJ, Parsons G, et a!. The aging respiratory system. Clin Geriatr Med 1985; 1:143-75 Ghio AJ, Elliott CG, Crapo RO, et al. Impairment in pulmonaty function after adult respiratory distress syndrome. Am Rev Respir Dis 1989; 139:1158-62 Lanken PN, Ahlheit BD, Crawford S, eta!. Withholding and withdrawing life-sustaining therapy. Am Rev Respir Dis 1991; 144:726-31

CHEST I 111 I 5 I MAY, 1997

1339