Acute Respiratory Failure in the United States

Acute Respiratory Failure in the United States* Incidence and 31-Day Survival Carolyn E. Behrendt, PhD

Study objectives: To estimate the incidence of acute respiratory failure (ARF) in the United States and to analyze 31-day hospital mortality among a cohort of patients with ARF. Design and setting: Retrospective cohort drawn from the Nationwide Inpatient Sample of 6.4 million discharges from 904 representative nonfederal hospitals during 1994. Patients: All 61,223 patients in the sample whose discharge records indicated all of the following: acute respiratory distress or failure, mechanical ventilation, > 24 h of hospitalization, and age > 5 years. Results: An estimated 329,766 patients discharged from nonfederal hospitals nationwide in 1994 met study criteria for ARF. The incidence of ARF was 137.1 hospitalizations per 100,000 US residents age > 5 years. Incidence increased nearly exponentially each decade until age 85 years. Overall, 35.9% of patients with ARF did not survive to hospital discharge. At 31 days, hospital mortality was 31.4%. According to the proportional hazards model, significant mortality hazards included age (> 80 years and > 30 years), multiorgan system failure (MOSF), HIV, chronic liver disease, and cancer. Hospital admission for coronary artery bypass, drug overdose, or trauma other than head injury or burns was associated with a reduced mortality hazard. Interaction was present between age and MOSF, trauma, and cancer. A point system derived from the hazard model classified patients into seven groups with distinct 31-day survival probabilities ranging from 24 to 99%. Conclusions: The incidence of ARF increases markedly with age and is especially high among persons > 65 years of age. Nonpulmonary hazards explain short-term (31-day) survival. (CHEST 2000; 118:1100 –1105) Key words: acute disease; adult; aged; child; hospital mortality; incidence; respiratory insufficiency; survival analysis; United States Abbreviations: APACHE ⫽ acute physiology and chronic health evaluation; ARF ⫽ acute respiratory failure; BMT ⫽ bone marrow transplant; CABG ⫽ coronary artery bypass graft; ICD-9-CM ⫽ International Classification of Diseases, 9th Revision, Clinical Modification; MOSF ⫽ multiorgan system failure

picture of the epidemiology of acute respiratory A failure (ARF) is emerging, but significant gaps remain. Surveys of ICU patients have estimated the incidence of ARF to be 77.6 cases per 100,000 population age ⱖ 15 years in Sweden, Denmark, and Iceland and 88.6 cases per 100,000 residents in Berlin.1,2 The incidence of ARF in the United States, however, is unknown. Rates of hospital mortality among ICU patients with a diagnosis of ARF or requiring mechanical ventilation range from 28 to 58%.1–12 Higher mortality rates have been observed

*From Children’s Hospital, San Diego, CA. Manuscript received November 16, 1999; revision accepted March 28, 2000. Correspondence to: Carolyn E. Behrendt, PhD, 4065 3413 Paseo del Campo, Palos Verdes, CA 90274; e-mail: [email protected] 1100

among ARF patients with AIDS or hematologic malignancy (65% and 83% mortality, respectively).13,14 Independent hazards for ARF mortality include older age,1,2,4 – 6,8,11,12,14 severe chronic comorbidities (HIV, active malignancy, cirrhosis),1,5,6,8 –10,12 certain precipitating events (trauma,3,6,8,11 drug overdose,8,9 bone marrow transplant [BMT]13), and multiple organ system dysfunction or failure (MOSF).3,4,7,9 –13 Mortality has also been associated with acute lung injury 9,12 or bilateral infiltrates on chest radiograph,1 and with an elevated acute physiology score1,6,7 or APACHE (acute physiology and chronic health evaluation) score.9,10 To my knowledge, no study to date has applied survival analysis to a representative cohort of US ARF patients, surgical as well as medical, children as well as adults. Such a cohort can be extracted from Clinical Investigations in Critical Care

the Nationwide Inpatient Sample, a database of all patients discharged from a representative sample of 904 nonfederal hospitals throughout the United States during 1994.15 Using data on these ARF patients, the current study estimates age-specific incidences of ARF and constructs a proportional hazards model to explain the associated hospital mortality. These findings will enhance the epidemiologic picture of ARF, increasing the information available to support clinical decision making, counseling of ARF patients and their families, and investigation of new therapies. Materials and Methods Data Source The 1994 Nationwide Inpatient Sample was designed to include acute care discharges from representative hospitals across the United States during a 1-year period.15 A systematic random sample of 904 hospitals was drawn from 2,135 nonfederal general and specialty hospitals stratified by geographic region, urban or rural location, control (government nonfederal, private not-forprofit, and private investor-owned), teaching status, and bed size. All discharges from the selected hospitals were included, for a total of 6,385,011 inpatients. Available data included patient demographics, sampling weights, International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) discharge codes for diagnoses and procedures, duration of hospitalization, discharge disposition, and hospital characteristics. Definitions Patients in the Nationwide Inpatient Sample were considered to have ARF if they had a diagnostic code for acute respiratory distress or failure (ICD-9-CM 518.5, 518.81, or 518.82) together with a procedure code for continuous mechanical ventilation (ICD-9-CM 96.7). Such ventilation included positive end-expiratory pressure but excluded continuous positive airway pressure, intermittent positive-pressure breathing, and oxygen by face mask or nasal cannula. Total hours of ventilation were not recorded in the database, so no minimum period of mechanical ventilation was specified. Instead, ARF patients were limited to those who remained in the hospital at least 24 h. Infants and children ⬍ 5 years were excluded. Additional clinical conditions were ascertained using ICD9-CM codes. MOSF was defined as one or more of the following in addition to respiratory failure: acute renal failure, cardiac arrest, shock, acute cerebrovascular event, disseminated intravascular coagulation, acute intestinal vascular insufficiency, acute hepatic failure, head injury with loss of consciousness for ⬎ 24 h or until death, and burns involving ⱖ 20% of body surface. Trauma refers to a primary diagnosis of crushing or internal injury, open wounds, superficial injury or contusion, sprains or strains, fractures other than isolated hip fracture, spinal cord injury, other injuries, or ARF with lung contusion. Head injury with loss of consciousness for ⬎ 24 h or until death and burns involving ⱖ 20% of body surface were categorized as MOSF rather than trauma. Statistical Analysis National sampling weights were used in all analyses (SUDAAN 7.5.3 software; Research Triangle Institute; Research Triangle

Park, NC) except the generation of survival plots (SAS 6.12 software; SAS Institute; Cary, NC). For the national estimates of ARF incidence, US Census Bureau estimates of the resident population as of July 1, 1994, served as the population denominators.16 The SEs of proportions and 95% confidence intervals around proportional hazard estimates were calculated using Taylor series linearization.17,18 This method takes into account the intracluster correlation that may result from the survey design, which included all patients at selected hospitals rather than selected patients from all hospitals. The follow-up period began at 24 h after hospital admission and continued through 31 days. After this time, the accelerated nature of several hazards (age ⱖ 80 years, MOSF, coronary artery bypass graft [CABG], and drug overdose) undermined the requisite assumption of proportional hazards.19 Survivors included patients who were discharged alive at any time and also those who died in the hospital after the follow-up period; survival times were censored at 31 days. Terms were retained in the multivariate model if they were significant at p ⬍ 0.0001 and improved the ⫺2 log-likelihood ratio.

Results Patients Among the Nationwide Inpatient Sample, 62,642 patients with acute respiratory distress or failure received mechanical ventilation and were hospitalized ⱖ 24 h. After excluding the 2.3% of these patients who were ⬍ 5 years of age, a total of 61,223 ARF patients remained. Their median age was 69 years (5th to 95th percentile range, 30 to 87 years). Children aged 5 to 17 years comprised 1.4% of patients. Half (51.1%) of all patients were male. The highest frequency of head injury, burns, and other trauma combined (20.8%) was among patients aged ⬍ 30 years, whereas the greatest prevalence of congestive heart failure (47.2%) and COPD (39.0%) was among patients aged ⱖ 50 years. The frequency of asthma decreased steadily with age, from 16.6% among children aged 5 to 17 years to 2.9% among adults aged ⱖ 80 years. Incidence The ARF patients in the sample corresponded to 329,766 discharges nationwide or 137.1 hospitalizations per 100,000 US residents aged ⱖ 5 years. ARF incidence increased markedly with age, resulting in an 88-fold difference in risk between the youngest and oldest age groups (Fig 1). Survival Overall, 35.9 ⫾ 0.3% of ARF patients did not survive to hospital discharge. Most (87.5%) deaths occurred within the 31-day follow-up period. Median length of stay (5th to 95th percentile range) was 13 days (3 to 55 days) among survivors and 10 days (1 to 50 days) among nonsurvivors. CHEST / 118 / 4 / OCTOBER, 2000

1101

Figure 1. Cases and incidence of ARF in the United States, 1994, by age. Bars denote the numbers of ARF cases; diamonds indicate incidence per 100,000 US residents. Age-specific incidence estimates are, from left to right, 9.7, 21.6, 32.3, 52.3, 99.9, 231.3, 493.5, 765.5, and 852.9 cases per 100,000.

A cohort of 61,113 ARF patients was appropriate for survival analysis. Patients admitted to the hospital for BMT (n ⫽ 86 or 0.14%) were excluded. Because of the variable lag time between admission for BMT and onset of ARF, their survival curve could not be aligned with that of the rest of the cohort. Another 24 patients (0.04%) who lacked data on survival status at follow-up were also excluded. At the end of 31 days, hospital mortality among the cohort was 31.4 ⫾ 0.2%. As shown in Table 1, hazards for 31-day mortality included age, MOSF, HIV, chronic liver disease, and cancer. Mortality hazard was significantly reduced among patients admitted for CABG, drug overdose, or trauma other than head injury or burns (hereafter referred to simply as trauma). Significant interaction was present between age and MOSF, trauma, and cancer. Mortality hazard increased at age 30 years (but only among patients without MOSF or trauma) and again at age 80 years. MOSF was a significant hazard among all age groups but had proportionally greater impact among patients aged ⬍ 80 years, especially those ⬍ 30 years 1102

and those admitted to the hospital for trauma. Cancer was a significant hazard until age 80 years; thereafter, mortality among patients with and without cancer did not differ. Trauma patients aged 30 to 79 years who did not develop MOSF had a mortality rate as low as that among uncomplicated patients aged ⬍ 30 years; these groups were combined to serve as the model’s referent category. The fit of the model was further improved by combining cancer and chronic liver disease into a single term (severe chronic comorbidity) and by combining drug overdose and CABG into another term (low-risk precipitating event). The model was not improved by including a term for sepsis. A point system, shown in the far right column of Table 1, converted the hazard profiles of individual patients into scores from ⫺1 to 5 (Table 2). The observed survival curves associated with these scores (Fig 2) were each significantly different from the next (log-rank test, p ⬍ 0.001 for all pairs). The steepest curve was the least smooth, owing to the small number of patients (n ⫽ 101) with the maximum score. Clinical Investigations in Critical Care

Table 1—Thirty-one-Day Hospital Mortality Among a Nationally Representative Sample (n ⴝ 61,113) of Patients With ARF*

Variables MOSF‡ Yes Age ⱖ 80 yr Age 30–79 yr Age 18–29 yr Age 5–17 yr No Age ⱖ 80 yr Age 30–79 yr without trauma§ Age 30–79 yr with trauma Age 18–29 yr Age 5–17 yr HIV infection Yes No Severe chronic comorbidity Cancer age ⬍ 80 yr, no liver disease Liver disease, no cancer age ⬍ 80 yr Both comorbidities Neither comorbidity Low-risk precipitating event Drug overdose or poisoning CABG Neither event

n

Observed 31-d Mortality¶, %

Proportional Hazard (95% CI)

Hazard Points†

3,879 13,773 503 179

59.9 50.5 44.3 32.4

8.55 (7.77, 9.41)

⫹4

5.86 (5.34, 6.43)

⫹3

8,228 31,024 1,267 1,570 690

32.3 21.3 11.0 10.9 6.5

3.69 (3.36, 4.05) 2.07 (1.90, 2.27)

⫹2 ⫹1

1,163 59,884

52.0 31.0

2.00 (1.86, 2.15) 1.00

⫹1

4,990 1,872 152 54,099

45.2 54.2 63.6 29.3

1.83 (1.77, 1.89)

⫹1

1,280 1,140 58,693

17.9 7.8 32.2

1.00

0

1.00 0.39 (0.35, 0.43)

⫺1

1.00

*CI ⫽ confidence interval. †When these points are totaled according to an individual patient’s characteristics, the resulting score corresponds to one of seven 31-day survival curves observed among ARF patients with the same score (see Fig 2). ‡The frequency of MOSF (29.9% overall) was significantly higher among those ARF patients who underwent CABG (51.9%) or who had chronic liver disease (48.2%) and significantly lower among trauma patients (22.5%), children ages 5 to 17 years (20.7%), and patients admitted for drug overdose or poisoning (12.1%). §Trauma does not include head injury or burns; these conditions are categorized instead as MOSF. ¶Mortality was not associated with sex, pneumonia, aspiration pneumonitis, CHF, COPD, hypertension, diabetes mellitus, interstitial lung disease, neuromuscular disease, GI hemorrhage, pancreatitis, embolism, aneurysm, ketoacidosis, inhalational injury, or near-drowning.

Discussion As in previous studies of ARF incidence, the current case definition excluded persons who were not admitted to a hospital as well as inpatients who

Table 2—Survival Among ARF Patients (n ⴝ 61,113) by Hazard Score Score

n

Deaths Through 31 Days, No.

31-d Survival* (SEM), %

⫺1 0 1 2 3 4 5 Total

267 4,231 25,344 13,435 11,274 6,461 101 61,113

2 320 4,794 4,668 5,389 4,010 78 19,261

99.3 (0.4) 92.6 (0.3) 81.2 (0.3) 65.3 (0.4) 52.4 (0.4) 38.0 (0.5) 24.0 (3.1) 68.6 (0.2)

*Percent cumulative survival at 31 days, calculated using national sampling weights.

did not receive mechanical ventilation or who were discharged within 24 h.1,2 Unlike earlier studies, the current study did not restrict ARF cases to ICU patients and did not specify a minimum period of mechanical ventilation. These broader inclusion criteria help to explain why the current study yielded a higher incidence estimate than previous studies. The true incidence of ARF may have been underestimated, however, because the Nationwide Inpatient Sample did not include patients admitted to federal hospitals and because some ARF patients may not have been coded as such at discharge. The gap between the current estimate (137.1/ 100,000 residents aged ⱖ 5 years) and that previously published from Berlin (88.6/100,000 residents) narrows once the Berlin estimate has been adjusted for age.2 Adjustment is necessary because, whereas the entire population of Berlin served as the estimate’s denominator, its numerator was restricted to ARF cases ⱖ 14 years. Assuming that children ⬍ 14 CHEST / 118 / 4 / OCTOBER, 2000

1103

Figure 2. Plot of survival among ARF patients (n ⫽ 61,113) by hazard score.

years comprised 20% of the Berlin population (as was the case in the US population16), the ageadjusted incidence of ARF in Berlin would be approximately 110.8/100,000 residents aged ⱖ 14 years. It is unclear why another European survey of ARF yielded a much lower estimate of 77.6 per 100,000 population aged ⱖ 15 years.1 The incidence of ARF was found to increase nearly exponentially with each decade until age 85 years. Comparable incidence estimates by age have not been published. The current 36% rate of mortality before discharge was similar to the 37% rate previously reported among ARF patients admitted to 40 US hospitals6 and approached the 41% and 43% rates reported among ARF patients in two European surveys.1,2 Whether ARF incidence or mortality differs significantly between the United States and Europe cannot be determined at this time, because of varying case definitions among studies to date. Because discharge data reflect the entire course of hospitalization rather than the initial day of follow-up, the current survival analysis and the point system derived from it are explanatory rather than prognostic. The current survival analysis con1104

firmed reports of associations between ARF mortality and older age, MOSF, HIV, cancer, chronic liver disease, trauma, and drug overdose.1–13 The increase in mortality at age 30 years and again at age 80 years observed in the current study has been noted in previous studies of patients receiving mechanical ventilation and ARF patients.5,8,12 Also consistent with earlier studies was the current lack of association between ARF mortality and sex, pneumonia, COPD, congestive heart failure, and diabetes.1–5,9,10,14 The current data did not confirm an independent association between ARF mortality and sepsis.3,9 The current survival analysis reflects the limitations as well as the strengths inherent in the Nationwide Inpatient Sample database. The lack of data on acute lung injury and acute physiologic or APACHE score and the necessity of excluding BMT patients precluded testing of these potential hazards in the multivariate model. However, the presence of large numbers of children, young adults, and the elderly within the sample permitted the significant interaction between age and other hazards for ARF mortality to be detected for the first time. Clinical Investigations in Critical Care

Conclusion The incidence of ARF increases markedly with age and is especially high among persons ⱖ 65 years of age. Nonpulmonary hazards explain short-term (31day) survival. ACKNOWLEDGMENT: The author thanks Alvaro Mun˜oz, PhD, for valuable advice on the survival analysis.

References 1 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; 159:1849 –1861 2 Lewandowski K, Metz J, Deutschmann H, et al. Incidence, severity, and mortality of acute respiratory failure in Berlin, Germany. Am J Respir Crit Care Med 1995; 151:1121–1125 3 Vasilyev S, Schaap RN, Mortensen JD. Hospital survival rates of patients with acute respiratory failure in modern respiratory intensive care units. Chest 1995; 107:1083–1088 4 Stauffer JL, Fayter NA, Graves B, et al. Survival following mechanical ventilation for acute respiratory failure in adult men. Chest 1993; 104:1222–1229 5 Swinburne AJ, Fedullo AJ, Bixby K, et al. Respiratory failure in the elderly: analysis of outcome after treatment with mechanical ventilation. Arch Intern Med 1993; 153:1657– 1662 6 Knaus WA, Sun X, Hakim RB, et al. Evaluation of definitions for adult respiratory distress syndrome. Am J Respir Crit Care Med 1994; 150:311–317 7 Jimenez P, Torres A, Roca J, et al. Arterial oxygenation does not predict outcome of patients with acute respiratory failure needing mechanical ventilation. Eur Respir J 1994; 7:730 – 735

8 Cohen IL, Lambrinos JL. Investigating the impact of age on outcome of mechanical ventilation using a population of 41,848 patients from a statewide database. Chest 1995; 107:1673–1680 9 Epstein SK, Vuong V. Lack of influence of gender on outcomes of mechanically ventilated medical ICU patients. Chest 1999; 116:732–739 10 Knaus WA. Prognosis with mechanical ventilation: the influence of disease, severity of disease, age, and chronic health status on survival from an acute illness. Am Rev Respir Dis 1989; 140:S8 –S13 11 Gracey DR, Naessens JM, Krishan I, et al. Hospital and posthospital survival in patients mechanically ventilated for more than 29 days. Chest 1992; 101:211–214 12 Pascual FE, Matthay MA, Bacchetti P, et al. Assessment of prognosis in patients with community-acquired pneumonia who require mechanical ventilation. Chest 2000;117:503–512 13 Montaner JSG, Hawley PH, Ronco JJ, et al. Multisystem organ failure predicts mortality of ICU patients with acute respiratory failure secondary to AIDS-related PCP. Chest 1992; 102:1823–1828 14 Epner DE, White P, Krasnoff M, et al. Outcome of mechanical ventilation for adults with hematologic malignancy. J Invest Med 1996; 44:254 –260 15 Healthcare Cost, and Utilization Project Nationwide Inpatient Sample, release 3, 1994, CD-ROM. Rockville, MD: US Department of Health and Human Services, Agency for Healthcare Research and Quality, 1996 16 Statistical abstract of the United States 1995: resident population by age and state 1994. 115th ed. Washington, DC: US Bureau of the Census, 1995; 33 17 Binder D. On the variances of asymptotically normal estimators from complex surveys. Int Stat Rev 1983; 51:279 –292 18 Binder D. Fitting Cox’s proportional hazards models from survey data. Biometrika 1992; 79:139 –147 19 Lawless JF. Statistical models and methods for lifetime data. New York, NY: John Wiley and Sons, 1982; 279

CHEST / 118 / 4 / OCTOBER, 2000

1105