Temporal Trends and Variation in Bronchoscopy Use for Acute Respiratory Failure in the United States

Abstract

Background: National data on bronchoscopy for the evaluation of acute respiratory failure are lacking, and the limited available data suggest wide variation in use.

Research question: How commonly is bronchoscopy performed among hospitalizations with acute respiratory failure? How has use changed over time and across hospitals?

Study design and methods: This was an observational cohort study of adult hospitalizations (2012-2018) treated with invasive mechanical ventilation (IMV) using the National Inpatient Sample, which represents 97% of all hospitalizations in the United States. We measured the proportion of hospitalizations treated with IMV who underwent bronchoscopy and assessed trends in bronchoscopy use over time. Multilevel linear regression models were used to quantify hospital-level variation, adjusting for differences in patient and hospital characteristics.

Results: We identified 6,101,070 IMV-treated hospitalizations (2012-2018), of whom 609,405 underwent bronchoscopy; among hospitalizations receiving bronchoscopy, mean age was 61 years, 41.8% were women, and in-hospital mortality was 30.8%. The percentage of IMV-treated hospitalizations receiving bronchoscopy increased from 9.5% (95% CI, 9.1%-9.9%) in 2012 to 10.8% (95% CI, 10.4%-11.2%) in 2018 (P < .001 for difference). In 2018, bronchoscopy use varied from 0% to 57.1% among 1,787 hospitals, and in multilevel models adjusted for patient and hospital characteristics, 16.0% of the variation was explained at the hospital level. The median OR was 2.13 (95% CI, 2.05-2.21), indicating 113% increased odds of receiving bronchoscopy if moving from a lower-use to a higher-use hospital.

Interpretation: Bronchoscopy use among hospitalizations treated with IMV has increased over time. The large variation in use of bronchoscopy across hospitals suggests potentially unwarranted practice variation and need for further studies to clarify which patients benefit from bronchoscopy.

Keywords: acute hypoxic respiratory failure; bronchoscopy; invasive mechanical ventilation.

Figure 1

A, B, Euler figure (proportional Venn diagram) showing Consolidated Standards of Reporting Trials diagram for primary and subgroup analysis (A) and sensitivity analysis (B) of included hospitalizations (2012-2018). A, In the primary analysis, a total of 6,101,070 hospitalizations received IMV, of whom 609,405 (10.0%) received bronchoscopy. In subgroup analyses, among 72,510 hospitalizations with hemoptysis who received IMV, 34,745 (47.9%) underwent bronchoscopy, and among 1,035,855 hospitalizations with an immunocompromising comorbidity who received IMV, 146,010 (14.1%) underwent bronchoscopy. B, In sensitivity analysis, among 9,735,619 hospitalizations with acute hypoxic respiratory failure, 724,815 (7.4%) underwent bronchoscopy. AHRF = acute hypoxic respiratory failure; IMV = invasive mechanical ventilation.

Figure 2

Line graph showing temporal trends in bronchoscopy use for acute respiratory failure (2012-2018). Unadjusted proportion of hospitalizations receiving bronchoscopy in the United States (2012-2018) among IMV-treated hospitalizations (gray), IMV-treated hospitalizations with hemoptysis (red), IMV-treated hospitalizations with an immunocompromising comorbidity (orange), and hospitalizations with acute hypoxic respiratory failure (blue). IMV use, hemoptysis, immunocompromising comorbidity, and acute hypoxic respiratory failure were identified via diagnostic coding, as described in the text and in e-Tables 1 through 4. Dots represent point estimates and whiskers represent 95% CIs. IMV = invasive mechanical ventilation.

Figure 3

A, B, Graphs showing hospital variation in bronchoscopy use among IMV-treated hospitalizations in 2018 (adjusted for patient and hospital characteristics): empty model (reliability adjustment^a only) (A) and fully adjusted model (B) for patient and hospital characteristics (risk and reliability adjusted). Each panel illustrates substantial variation in bronchoscopy in 2018 that persisted across hospitals after adjusting for patient- and hospital-level characteristics. We excluded hospitals with fewer than 25 IMV-treated hospitalizations to measure variation across hospitals reliably. The x-axis represents hospitals ranked by increasing use of bronchoscopy, and the y-axis represents the percentage of IMV-treated hospitalizations at each individual hospital receiving bronchoscopy in the mixed-effect logistic regression model. The blue dots denote the point estimates and the whiskers denote the 95% CIs for each hospital. The number of hospitalizations, bronchoscopy range, median OR, and ICC are presented for each panel. ^aReliability adjustment is a method to reduce apparent variation resulting from small case volume at individual hospitals. ICC = intraclass correlation coefficient; IMV = invasive mechanical ventilation.

Figure 4

Map showing the proportion of IMV-treated hospitalizations receiving bronchoscopy across regions defined by US Census division in 2018. Proportions are adjusted for both patient- and hospital-level characteristics. US Census division tracts: 1 = Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, and Vermont; 2 = New Jersey, New York, and Pennsylvania; 3 = Illinois, Indiana, Michigan, Ohio, and Wisconsin; 4 = Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota; 5 = Delaware, District of Columbia, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, and West Virginia; 6 = Alabama, Kentucky, Mississippi, and Tennessee; 7 = Arkansas, Louisiana, Oklahoma, and Texas; 8 = Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, and Wyoming; 9 = Alaska, California, Hawaii, Oregon, Washington. IMV = invasive mechanical ventilation.