Association of Antibiotic Treatment With Outcomes in Patients Hospitalized for an Asthma Exacerbation Treated With Systemic Corticosteroids

Key Points

Question  Among patients hospitalized for an asthma exacerbation and treated with corticosteroids, is the addition of antibiotic therapy associated with better outcomes?

Findings  In this cohort study of 21 628 patients hospitalized for an asthma exacerbation and treated with corticosteroids, 8927 (41.3%) received antibiotics on day 1 of hospitalization. Compared with patients who were not treated with antibiotics, treated patients had a statistically significantly longer but not clinically important increase in hospital stay, higher hospital cost, lower rate of treatment failure, and no significant difference in risk of antibiotic-related diarrhea.

Meaning  Antibiotic treatment may be associated with marginally longer length of stay but lower risk of treatment failure in adult patients hospitalized for asthma treated with corticosteroids.

Importance  Although professional society guidelines discourage use of empirical antibiotics in the treatment of asthma exacerbation, high antibiotic prescribing rates have been recorded in the United States and elsewhere.

Objective  To determine the association of antibiotic treatment with outcomes among patients hospitalized for asthma and treated with corticosteroids.

Design, Setting, and Participants  Retrospective cohort study of data of 21 628 adults hospitalized for asthma exacerbation and treated with systemic corticosteroids in 540 US acute care hospitals from January 1, 2015, through December 31, 2016.

Exposures  Early antibiotic treatment, defined as treatment with an antibiotic initiated during the first 2 days of hospitalization. Patients not treated with antibiotics or treated on day 2 or later were included in the nontreated/late-treated group. Patients with documented infection diagnosed at admission were excluded.

Main Outcomes and Measures  The primary outcome measure was hospital length of stay. Other outcomes included treatment failure (initiation of mechanical ventilation, transfer to the intensive care unit after hospital day 1, in-hospital mortality, or readmission for asthma within 30 days of discharge), hospital costs, and antibiotic-related diarrhea. Multivariable adjustment, propensity score matching, propensity weighting, and instrumental variable analysis were used to assess the association of antibiotic treatment with outcomes.

Results  Of the 21 628 patients, the median (interquartile range [IQR]) age was 46 (33-59) years, 15 662 (72.4%) were women and 9616 (44.5%) were White, and Medicare was the primary form of health insurance for 5499 (25.4%). Antibiotics were prescribed for 8927 patients (41.3%) on day 1; 3022 patients (14.0%) were started after day 1. Compared with patients not treated with antibiotics or treated after day 1, patients treated with antibiotics on day 1 were older (median [IQR] age, 48 [35-60] vs 45 [32-57] years), more likely to be White (48.7% vs 41.5%) and smokers (4.5% vs 3.2%), and had a higher number of comorbidities (eg, congestive heart failure, 5.7% vs 5.4%). Those treated with antibiotics on day 1 had a statistically significant longer hospital stay compared with those not treated on day 1 (mean [SD], 2.81 [2.27] vs 2.57 [2.45] days; difference, 0.11 days, 95% CI, 0.03 to 0.19; median [IQR] 2 [1-4] vs 2 [1-3] days). In propensity score–matched analysis, receipt of antibiotics on day 1 was associated with a marginally longer but not clinically meaningful increase in hospital stay (length of stay ratio, 1.06; 95% CI, 1.04 to 1.09), higher cost of hospitalization (median [IQR] cost, $4320 [$2754-$6716] vs $3861 [$2479-$6236]) but lower risk of treatment failure (7.1% vs 8.2%; difference, −1.08%, 95% CI, −1.93% to −0.24%; adjusted OR, 0.86; 95% CI, 0.77 to 0.97). Multivariable adjustment, propensity score weighting, and instrumental variable analysis yielded similar results.

Conclusions and Relevance  Among adult patients hospitalized for asthma exacerbation and treated with corticosteroids, antibiotic therapy initiated on day 1 of hospitalization was associated with a slightly longer but not clinically important increase in hospital length of stay, higher hospitalization cost, and lower risk of treatment failure. These findings highlight the need to perform randomized clinical trials to determine the role of antibiotic prescribing among patients hospitalized for asthma exacerbation.

Quiz Ref IDAsthma is the most common chronic lung condition in the United States, where it affects 24.6 million individuals.¹ Asthma exacerbations are responsible for 1.7 million emergency department visits, 440 000 hospitalizations, and more than $50 billion in health care expenditures each year in the United States.¹ Current guidelines for the treatment of patients hospitalized for an asthma exacerbation call for objective assessment of lung function, controlled oxygen administration, inhaled short-acting β₂-agonist bronchodilators, and systemic corticosteroids, but several studies have documented limited adherence to these guidelines and variation in acute and chronic asthma care.^2,3 Although recent reviews published in the Cochrane database have not found sufficient evidence in favor of antimicrobial treatment in asthma exacerbations and current guidelines recommend against routine use of antibiotic therapy,⁴ inappropriate use of antibiotics has been documented in several countries including the United States. In a recent study of a large national sample, we found that nearly 49.1% of patients hospitalized for asthma received treatment with antibiotics in the absence of documentation of an indication for antibiotic therapy.⁵

The evidence surrounding use of antibiotics in patients with asthma is limited.^6,7 The most recent trial, Azithromycin Against Placebo for Acute Exacerbations of Asthma, showed no benefit of short-term treatment with azithromycin when added to a regimen of oral or intravenous corticosteroids.⁷ Critics question the external validity of the trial and whether the lack of benefit was a result of the fact that patients who could have benefited from an antibiotic had already received one and were therefore not enrolled. In the absence of more definitive information from randomized clinical trials, we sought to evaluate the association between use of antibiotics when prescribed in addition to corticosteroids and outcomes among a large, representative sample of patients hospitalized for asthma exacerbation. We hypothesized that antibiotic therapy would not be associated with additional benefit.

Quiz Ref IDWe conducted a retrospective cohort study by using data collected from 540 acute care hospitals in the United States that participate in Premier Inpatient Database, an inpatient, enhanced administrative database developed for measuring health care quality and use. Participating hospitals are primarily small to medium-sized nonteaching hospitals located mostly in urban areas in all regions of the United States. In addition to the information contained in the standard hospital discharge file, such as the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Clinical Modification (ICD-10-CM) codes, hospital, and physician information, Premier includes a date-stamped log of all billed items, including medications dispensed and diagnostic tests performed. Approximately 75% of participating hospitals submit cost data; the rest submit calculated costs based on each hospital’s specific cost to charge ratio.⁸ The institutional review board at Baystate Medical Center, Springfield, Massachusetts, approved the study, which was not considered human subjects research because the data set does not contain any identifiable patient information.

We included patients 18 years or older who were admitted to the hospital from January 1, 2015, through December 31, 2016, for a principal diagnosis of asthma or a principal diagnosis of acute respiratory failure combined with a secondary diagnosis of asthma (we used ICD-9-CM and ICD-10-CM diagnostic codes in accordance with the Centers for Medicare & Medicaid Services definition).⁹ We restricted the analysis to patients treated with systemic corticosteroids (oral or intravenous corticosteroids at a dosage equivalent to 20 mg/d of prednisone on day 1), because systemic corticosteroids are recommended for patients with moderate to severe asthma exacerbation. We excluded patients with a potential indication for antibiotic therapy, including those with a secondary diagnosis of acute or chronic bronchitis, chronic obstructive pulmonary disease, emphysema, or bronchiectasis; those with a diagnosis of sinusitis, sepsis, pneumonia, urinary tract infection, or skin or soft-tissue infection present on admission; and patients with a blood or sputum culture ordered on day 1 of hospitalization. In addition, we excluded patients transferred from another acute care hospital because we did not know whether they were treated with antibiotics before arrival as well as patients transferred to another hospital because we could not ascertain their outcomes. If a patient had more than 1 eligible admission during the period of the study, we randomly selected 1 admission to avoid survival bias.

All patients were classified according to the treatment they received on the first calendar day of hospitalization (day 1). The early antibiotic treated group included patients started on antibiotic on day 1 of hospitalization or in the emergency department. Patients with antibiotic treatment started after day 1 of hospitalization were grouped with those not treated. Antibiotics were grouped based on previously published results⁴ into the following categories: macrolides, quinolones, cephalosporins, and tetracyclines.

Our primary outcome was hospital length of stay measured in days. Our main secondary outcome was a composite measure of treatment failure defined as initiation of invasive or noninvasive mechanical ventilation, transfer to the intensive care unit or in-hospital mortality after hospital day 1, or readmission for asthma exacerbation within 30 days of discharge.¹⁰ We also examined hospital cost and potential adverse effects of antibiotic use, including allergic reactions and antibiotic-associated diarrhea defined as treatment with metronidazole or oral vancomycin hydrochloride initiated after hospital day 2 or readmission within 30 days for diarrhea or Clostridium difficile infection.

In addition to patient demographic characteristics, primary payer, and principal diagnosis, we classified comorbidities by using the comorbidity score as described by Gagne et al.¹¹ The score was devised to predict mortality and 30-day readmissions and has better predictive ability than Elixhauser or Charlson scores. We also examined medications typically used in the treatment of asthma exacerbation, including short- and long-acting bronchodilators and methylxanthines, and medications used for smoking cessation therapy. We counted the number of admissions for asthma in the previous year as a proxy for asthma severity and degree of control.

Using absolute standardized differences, we compared the characteristics of patients with asthma who received early antibiotic therapy with those patients who did not receive antibiotics and those in whom antibiotics were prescribed later in the hospital stay; a difference greater than 10% was considered meaningful.¹² To estimate the association of antibiotic therapy with length of hospital stay, treatment failure, total cost, and other outcomes, we developed a multivariable predictive model as a function of patient, treatment, and hospital characteristics. Identity link function was used for length-of-stay and cost outcomes, and logit link functions for the treatment failure outcomes. To further reduce the threat of confounding by indication (in which patients with the most clinically severe asthma presentations would be the most likely to receive antibiotics), we developed a propensity score for early antibiotic treatment by using all patient characteristics, early treatments, comorbidities, and selected interaction terms. We matched patients who did not receive early antibiotic treatment with those with similar propensity who received antibiotics on day 1, and we carried out a conditional logistic regression analysis after accounting for the match.^13,14 Unadjusted (accounting for clustering), covariate-adjusted, and propensity score–adjusted models for each outcome were evaluated. In addition, we used standardized mortality ratio weighting to obtain estimates of average treatment effect among treated patients.¹⁵ To further address the threat of residual selection bias due to unmeasured confounders not addressed through propensity adjustment, we performed an instrumental variable analysis. We used a grouped-treatment approach, a form of instrumental variable analysis in which all patients treated at the same hospital are assigned a probability of treatment with an antibiotic equal to the overall treatment rate at that hospital. This grouped rate was substituted for individual patient exposure to treatment in the logistic regression model. Grouping treatment at the hospital level overcomes the issue of confounding by indication at the patient level, while still accounting for patient-level covariates and outcomes.¹⁶

We performed several sensitivity analyses. First, we repeated the analyses redefining early antibiotic treatment as started on day 1 or day 2 because in this administrative data set, only calendar days are counted. Thus, if a patient is admitted late in the evening, their first hospital day could be as short as 1 hour. These shortened first hospital days may not adequately represent clinician decisions about initial treatments. Second, we excluded patients in whom antibiotic therapy was initiated after day 1 of hospitalization from the nontreated group. Third, we limited the sample to nonsmokers younger than 70 years who were not treated with mechanical ventilation or admitted to the intensive care unit on day 1 of admission to further exclude patients for whom administration of antibiotics may be considered more acceptable in light of the severity of presentation and the difficulty differentiating chronic obstructive pulmonary disease from asthma. Furthermore, because macrolides were the most frequently used antibiotics, we compared the outcomes within the following 2 groups: patients treated with a macrolide alone vs those not treated with antibiotics and patients treated with a macrolide alone vs those treated with any other antibiotic.

For each model, adjusted odds ratios (ORs) for treatment failure or ratios of hospital length of stay and cost with associated 95% CIs for antibiotic treatment were calculated. All tests of significance were 2-sided. All statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc).

A total of 21 628 patients met all the inclusion criteria (eFigure in Supplement 1). The median (interquartile range [IQR]) age of these patients was 46 (33-59) years; 15 662 (72.4%) were women and 9616 (44.5%) were White; 6750 (31.2%) had private insurance, 6037 (27.9%) had Medicaid, and 5499 (25.4%) had Medicare. Among the entire cohort, the most frequent comorbidities were hypertension (9594 [44.4%]), obesity (6124 [28.3%]), diabetes (4787 [22.1%]), and depression (2762 [12.8%]); 3216 (14.9%) were admitted at least once in the previous year (Table 1). Median (IQR) length of stay was 2 (1-3) days; in-hospital mortality occurred in 29 patients (0.1%), 691 (3.6%) patients were intubated after day 1, and 1689 (7.8%) experienced the combined outcome of treatment failure (Table 2).

Quiz Ref IDOverall, 8927 patients (41.3%) received antibiotics on the first hospital day. The most frequently prescribed antibiotics were macrolides (4490 patients [50.3%]), quinolones (3111 [34.8%]), and third-generation cephalosporins (1772 [19.8%]). An additional 3022 patients (14.0%) received antibiotics at some point after hospital day 1, and of these, 2339 (10.8% ) were started on day 2 of hospitalization. Compared with patients who did not receive or received late antibiotic therapy, treated patients were older (median [IQR] age, 48 [35-60] vs 45 [32-57] years), more likely to be White (48.7% vs 41.5%), and more likely to have Medicare insurance (28.5% vs 23.3%). They were also more likely to have a diagnosis of acute respiratory failure (10.9% vs 9.6%), to receive smoking cessation therapy (4.5% vs 3.2%), and to have comorbid diagnoses of congestive heart failure (5.7% vs 5.4%), chronic pulmonary disease (11.4% vs 10.3%), diabetes (23.6% vs 21.1%), renal failure (4.2% vs 3.9%), obesity (29.8% vs 27.2%), and depression (13.1% vs 12.5%). Of all the patients, 631 (7.1%) in the early antibiotic therapy group experienced treatment failure compared with 1058 (8.3%) who did not receive early antibiotic therapy or received late treatment (Table 1).

A total of 7577 patients treated with antibiotics on day 1 of hospitalization were successfully matched to patients with a similar propensity score who did not receive early antibiotics, and most patient characteristics were balanced. Five hundred thirty-nine patients (7.1%) in the early antibiotic therapy group experienced treatment failure compared with 621 patients (8.2%) who did not receive early antibiotic therapy (P = .01). The mean (SD) length of hospital stay was 2.81 (2.27) days in the early antibiotic therapy group vs 2.57 (2.44) days in the untreated group (difference, 0.11; 95% CI, 0.03 to 0.19), and the median (IQR) length of stay in the early antibiotic therapy group was 2 (1-4) days vs 2 (1-3) days in the untreated group (P < .001). In the early antibiotic–treated group, 147 patients (1.6%) had a diagnosis of antibiotic-associated diarrhea compared with 149 (1.2%) in the untreated group; difference, 0.26%; 95% CI, −0.11 to 0.64 (Table 2).

Treatment with antibiotics was associated with a small but not clinically important increase in hospital length of stay and higher hospitalization costs in all multivariable analyses, including the propensity score–matched cohort (length-of-stay ratio, 1.06; 95% CI, 1.04 to 1.09 [Figure, A]; median [IQR] cost [in US $], $4320 [$2754-$6716] vs $3861 [$2479-$6236]; mean [SD], $5662 [$5855] vs $5302 [$6959]; difference, $360; 95% CI, $155 to $566; OR, 1.10; 95% CI, 1.08 to 1.12). In the analysis accounting for clustering of patients within hospital and adjusted for other variables, the propensity score–matched cohort analysis, and the standardized mortality ratio weighting analysis, there was a lower odds of treatment failure in the early treatment group (whole cohort: 7.1% vs 8.3%; difference, −1.08%; 95% CI, −1.93% to −0.24%; adjusted OR, 0.86; 95% CI, 0.77 to 0.96; propensity score–matched OR, 0.86; 95% CI, 0.77 to 0.97). In the instrumental variable analysis, which used the hospital antibiotic prescribing rate (percentage of patients hospitalized with asthma who received antibiotics), the OR for 100% hospital rate vs 0% was 0.67 (95% CI, 0.45 to 0.99). All-cause 30-day readmissions among survivors were not different between the 2 groups (propensity score–matched analysis OR, 1.01; 95% CI, 0.90 to 1.13), and the association remained nonsignificant in the group treatment analysis (OR, 1.19; 95% CI, 0.83 to 1.72). There was no significant difference in 30-day readmissions for asthma between the groups (propensity score–matched analysis OR, 0.91; 95% CI, 0.77 to 1.08; instrumental variable analysis OR, 0.98; 95% CI, 0.61 to 1.57). The risk for antibiotic-related diarrhea was higher in the antibiotic-treated patients (adjusted OR, 1.34; 95% CI, 1.05 to 2.17), but the association became nonsignificant in the propensity score–matched cohort (adjusted OR, 1.19; 95% CI, 0.90 to 1.57). For all the above analyses, the standardized mortality ratio weighting produced similar results.

Several sensitivity analyses were conducted. In the analyses that included patients treated with antibiotics on day 1 or 2 in the early treated group, the treatment failure results were attenuated and became nonsignificant (propensity score–matched analysis: OR, 0.96; 95% CI, 0.83 to 1.10; instrumental variable analysis: OR, 1.08; 95% CI, 0.71 to 1.62) and the difference in length of stay increased (propensity score–matched analysis: OR, 1.18; 95% CI, 1.15 to 1.20) (Figure, B). Unadjusted and propensity matched outcomes of patients started on antibiotics on day 1 or 2 of hospitalization compared with those not treated or started after day 2 are presented in the eTable in Supplement 1. In the analysis that excluded patients treated with antibiotics after day 1, the results were similar to those that included patients treated with antibiotics on day 1 or 2 (propensity score–matched analysis: OR, 0.99; 95% CI, 0.87 to 1.14) except that patients treated with antibiotics had 3 times greater odds of diarrhea than those not treated (OR, 3.18; 95% CI, 2.14 to 4.74). The main results remained consistent in the analysis restricted to nonsmokers younger than 70 years who did not have acute respiratory failure as the principal diagnosis and did not receive mechanical ventilation on day 1 of admission. Analyses that compared patients treated with macrolides only with those not treated with antibiotics and compared those treated with macrolides with those treated with other antibiotics produced similar results in regard to treatment failure and cost. Patients treated with macrolides only were significantly less likely to have antibiotic-related diarrhea than those treated with other antibiotics (1.2% vs 2.1%; P = .002).

Quiz Ref IDIn this observational study of more than 20 000 patients hospitalized for asthma at more than 500 US hospitals, we found that antibiotic treatment initiated on day 1 of hospitalization was associated with a slightly longer but not clinically important increase in length of stay and higher hospital costs. However, early antibiotic therapy was associated with lower risk of a composite measure of treatment failure, which included noninvasive or invasive ventilation, transfer to the intensive care unit, in-hospital death, or 30-day readmission for asthma. These findings challenge current guidelines that recommend against the use of antibiotics in the absence of concomitant infection. In addition, the findings highlight the need for rigorous randomized clinical trials to determine the role of antibiotics in patients hospitalized with an asthma exacerbation.

Evidence for the role of antibiotic treatment in patients with asthma exacerbation comes from 6 trials that enrolled a total of 681 adults and children.¹⁷ Most trials analyzed resolution of symptoms or measurements of lung function and did not investigate outcomes, such as the need for mechanical ventilation, readmission, or death. One trial of 278 adults with asthma exacerbation treated for 10 days with telithromycin (Telithromycin in Acute Exacerbations of Asthma [TELICAST]) found that antibiotic-treated patients had greater reduction in symptoms but no change in peak expiratory flow or other outcomes; owing to adverse effects, use of this antibiotic has been discontinued in the United States.⁶ The most recent trial that compared the addition of azithromycin to standard treatment that included systemic corticosteroids (Azithromycin Against Placebo in Exacerbations of Asthma [AZALEA]) in 199 adults hospitalized for asthma did not show any benefit for symptoms. The difference in the results between the 2 trials could be related to the fact that only 34% of patients in the TELICAST study were treated with systemic corticosteroids. Our results add to the uncertainty of the effect of antibiotics in addition to systemic corticosteroids in patients hospitalized for asthma exacerbation.

In the absence of obvious bacterial infection, why should the early antibiotic treatment examined in this study be associated with lower risk of treatment failure? One possibility is that although we used diagnostic codes to exclude patients with a diagnosis of infection, some patients included in the analysis may have had infections that went unrecognized or not coded by the primary team. Early antibiotic therapy would be expected to improve the outcomes of these patients. Although the results suggest that empiric antibiotic use may be beneficial, in addition to confirming results in clinical trials, further research is needed to identify which patients are most likely to benefit.

Validating known biomarkers, such as the procalcitonin level, for guiding targeted antibiotic therapy is one strategy that could encourage appropriate decisions for antibiotic prescribing.¹⁸ Two recent studies from China^19,20 have shown that a procalcitonin-guided strategy for patients hospitalized for exacerbation of asthma reduced antibiotic prescription (48.9% for patients who received procalcitonin testing vs 87.8% of those who did not; P < .001) and antibiotic exposure (relative risk, 0.56; 95% CI, 0.44-0.70), with no differences in clinical recovery, hospital length of stay, number of asthma exacerbations, or number of hospital readmissions during the 12-month follow-up period.

A second possible explanation for the apparent benefits we observed in this analysis is that macrolide antibiotics have anti-inflammatory effects independent of their role as antibacterial agents.²¹ Systematic reviews of randomized clinical trials report benefits of macrolides on asthma symptoms and exacerbation in patients with uncontrolled persistent asthma.²² However, this evidence does not exist yet for hospitalized patients. In addition, we did not find any difference between macrolides and other antibiotics.

Another explanation is that antibiotics may be administered reflexively to patients who are experiencing clinical deterioration, regardless of whether antibiotic therapy might be beneficial. If that is the case, then late treatment may simply be a marker for clinical deterioration, and inclusion of patients treated on hospital day 2 could bias the results of this sensitivity analysis to make routine antibiotics appear less beneficial.

Our aim was to evaluate the outcomes associated with antibiotics when given routinely early after hospitalization for asthma. In attempting to analyze the data using intention-to-treat principles, we limited the early treatment group to those patients in whom antibiotics were started on day 1. Patients started on antibiotics on hospital day 2 were included in the not-treated/late-treated group and represented 18.4% of this group. As described in the Methods, the second hospital day can begin as early as 1 hour into the hospitalization and can last as long as 47 hours. Thus, many of the patients included in the late-treated group may have been treated with antibiotics close to the time of admission. In the sensitivity analysis in which we included patients treated on hospital day 2 in the early treatment group, the association between early antibiotic therapy and treatment failure still favored antibiotic therapy but was no longer statistically significant.

Quiz Ref IDAlthough we controlled for potential confounders and used several analytic strategies, we cannot exclude the possibility of residual confounding by indication in which antibiotics would be preferentially given to patients with higher acuity of illness. To reduce this threat, we used robust statistical analyses, including propensity score matching, standardized mortality ratio weighting, and instrumental variable analysis, the last being a strategy to specifically reduce the risk of confounding from unmeasured factors. Because this was an observational study, we recognize that our results show an association between lower risk of treatment failure and antibiotic treatment and do not demonstrate causality. A second limitation is that, given the nature of our data set, we did not have access to physiological measures of disease severity, such as the results of oxygen saturation testing, or information on patient symptoms. Nevertheless, our analyses accounted for previous hospital admission and other treatments as a measure of severity. Third, we did not have access to previous pulmonary function test results to confirm the diagnosis of asthma or its severity. We used a set of ICD-9-CM and ICD-10-CM diagnostic codes, which are recommended by the Centers for Medicare & Medicaid Services, and we included only patients treated with oral or intravenous corticosteroids. Excluding patients with a secondary diagnosis of chronic obstructive pulmonary disease further strengthened our study. Of note, an earlier study that assessed the association between antibiotic treatment and outcomes in patients hospitalized for exacerbation of chronic obstructive pulmonary disease also found that antibiotic treatment was associated with better outcomes.²³ Finally, our study was restricted to inpatient events and 30-day readmission to the index hospital.

Among patients hospitalized for asthma exacerbation and treated with corticosteroids, antibiotic therapy initiated on day 1 of hospitalization was associated with a slightly longer but not clinically important increase in hospital length of stay, higher hospitalization costs, and lower risk of treatment failure. These findings highlight the need to perform randomized clinical trials to determine the role of antibiotic prescribing among patients hospitalized for asthma exacerbation.

Accepted for Publication: August 20, 2018.

Published Online: January 28, 2019. doi:10.1001/jamainternmed.2018.5394

Retraction and Replacement: This article was retracted and replaced on January 19, 2021, to fix errors in the abstract, key points, text, tables, figure, and Supplement 1 (see Supplement 2 for the retracted article with errors highlighted and Supplement 3 for the replacement article with corrections highlighted).

Corresponding Author: Mihaela S. Stefan, MD, PhD, Institute for Healthcare Delivery and Population Science, Department of Medicine, University of Massachusetts Medical School, Baystate, 3601 Main St, Third Floor, Springfield, MA 01199 (mihaela.stefan@baystatehealth.org).

Author Contributions: Dr Stefan had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Stefan, Au, Lindenauer.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Stefan, Au, Lindenauer.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Stefan, Shieh, Pekow.

Obtained funding: Stefan, Lindenauer.

Administrative, technical, or material support: Stefan, Spitzer, Lindenauer.

Supervision: Stefan, Pekow, Lindenauer.

Conflict of Interest Disclosures: Dr Krishnan reported serving on a data monitoring committee for Sanofi and receiving grants from the National Institutes of Health and the Patient-Centered Outcomes Research Institute. Dr Au reported serving on a data monitoring committee for Novartis, serving as a consultant to Gilead Sciences, serving on the pulmonary examination writing committee for the American Board of Internal Medicine, and serving as Deputy Editor for the Annals of the American Thoracic Society. Dr Lindenauer reported receiving grant support from the National Heart, Lung, and Blood Institute. No other disclosures were reported.