Trends in the Use of Noninvasive and Invasive Ventilation for Severe Asthma

We examined inpatient hospitalizations for asthma within the Pediatric Health Information System (PHIS) data set. This is a multi-institutional database that draws from 52 children’s hospitals and is maintained by the Children’s Hospital Association. The PHIS data set contains a wide range of encounter-specific data, including clinical and financial elements. Data quality is ensured via partnerships between the Children’s Hospital Association and participating members.

We queried discharge data from January 2014 to December 2018 for children ages 2 to 17 admitted with a primary diagnosis of asthma. Encounters were identified on the basis of International Classification of Diseases, Ninth Revision (ICD-9) and International Classification of Diseases, 10th Revision (ICD-10) codes (ICD-9 codes: 493.0x, 493.1x, 493.2x, 493.8x, and 493.9x; ICD-10 codes: J45.2x, J45.3x, J45.4x, J45.5x, and J45.9xx). Four hospitals were excluded from our analysis: 2 because they lacked data for some study years and 2 because of reporting errors arising in the transition from ICD-9 to ICD-10.

We examined patient demographic data for each admission, including age, sex, race, and insurer. Collected outcome measures included the use of IMV and NIV, admission to the ICU, hospital length of stay (LOS), and mortality. The use of IMV was identified by using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), International Classification of Diseases, 10th Revision, Procedure Coding System (ICD-10-PCS), and Current Procedural Terminology (CPT) procedure codes (ICD-9-CM: 96.04, 96.05, and 96.7x; ICD-10-PCS: 5A1935Z, 5A1955Z, 5A1955Z, 0BH17EZ, and 0BH18EZ; CPT: 31500). We defined NIV as the use of continuous positive airway pressure or bilevel positive airway pressure, identified by their respective procedure codes (ICD-9-CM: 93.90; ICD-10-PCS: 5A0935x, 5A0945x, 5A0955x, and 5A19054; CPT: 94660 and 94662). We used the PHIS ICU flag to identify hospitalizations that included ICU stays. Rates of NIV and IMV use were calculated as the percentage of admissions wherein those treatment modalities were used. For some analyses, hospitals were divided into quartiles according to the frequency of NIV use in 2018.

We report descriptive and summary statistics for the data measures described above. Associations between continuous variables were evaluated by using Pearson correlation, and changes over the study period were evaluated with weighted least squares. Comparisons between IMV and NIV rates were performed by using the Wilcoxon rank test, and differences in LOS among groups of patients and hospitals were assessed by using the Mann–Whitney U test. All statistical analyses were conducted within the Jupyter Notebook environment by using Python 3.7.²⁰ 

We identified 95 204 discharges with a primary diagnosis of asthma at 48 hospitals from 2014 to 2018. During the study period, annual admissions for asthma per hospital ranged from 55 to 1358 (median of 355; interquartile range [IQR]: 207–527). Demographic data for patients, including age, sex, race, and insurer, are presented in Table 1. Across all years, ICU care was provided within 18 449 admissions (19%). Among all patients, 90 407 received no assisted ventilation, 4104 received NIV, 587 received IMV, and 106 received both. Initiation of IMV or NIV could have occurred in the referring emergency department, during medical transport, or in the ICU after admission.⁶  During the study period, there was a 29% reduction in the number of asthma admissions per hospital, with a median of 434 in 2014 (IQR: 308–640) and 306 in 2018 (IQR: 185–495; P < .01). Concomitantly, there was also a 43% reduction in the number of ICU asthma admissions per hospital (median of 97 in 2014 [IQR: 62–157] vs 55 in 2018 [IQR: 34–83]; P = .02).

From 2014 to 2018, patients admitted for asthma were more likely to receive NIV than IMV, with a median NIV use rate of 1.87% (IQR: 0.44%–4.10%) and an IMV use rate of 0.62% (IQR: 0.48%–1.04%; P < .01). In 2018, the median rate of IMV use for patients admitted to the ICU was 3.6% (IQR: 2.2%–5.5%), whereas the median rate of NIV use was 10.4% (IQR: 1.7%–18.7%). From 2014 to 2018, there was no significant change in the hospital-specific frequency of IMV use (median of 0.6% [IQR: 0.3%–1.1%] in 2014 vs 0.6% [IQR: 0.2%–1.3%] in 2018; P = .43; Fig 1), but NIV use increased in many centers. Across all hospitals, the fraction of patients with asthma receiving NIV increased from a median of 1.5% (IQR: 0.3%–3.2%) in 2014 to 2.1% in 2018 (IQR: 0.3%–5.6%; Fig 1). However, this modest aggregate increase was not statistically significant (P = .15) and belied considerable variability among centers. Stratification of hospitals into NIV use frequency quartiles revealed that some centers seldom employed the modality, whereas others used it often and with increasing frequency (Fig 2A). Across the entire study period, centers in the bottom 3 NIV use quartiles had stable (or even decreasing) rates, whereas those in the top quartile more than doubled their NIV use from 4.8% (IQR: 2.8%–7.5%) in 2014 to 13.2% (IQR: 7.4%–15.2%) in 2018 (P = .02). In Fig 2B, we present these differences more explicitly. Hospitals in this high-use quartile were drawn from all regions of the country and included 4 of the 10 centers with the highest volume of asthma admissions. Overall, but with notable exceptions, hospitals with more asthma admissions tended to use NIV more frequently (P < .01).

Regarding intubation and invasive ventilation, the IMV rate among top-quartile NIV users was slightly higher than that of bottom-quartile users (0.72% [IQR: 0.59%–1.16%] vs 0.55% [IQR: 0.39%–0.67%]), but the difference was not statistically significant (P = .21). Although the NIV rate more than doubled for this top quartile of NIV users over the study period, their IMV rate remained stable (2014 median = 0.76% [IQR: 0.42%–1.14%] vs 2018 median = 0.78% [IQR: 0.55%–1.39%]; P = .36). Across all 5 years and 48 institutions, the percentage of patients who received both NIV and IMV was low, seen within only 106 admissions (0.11% of all admissions). One of the limitations of PHIS is that we are unable to identify the order in which these interventions were administered. These few cases could represent failure of NIV, with progression to IMV, but could also represent postextubation support with NIV. We identified only 71 inpatient deaths due to asthma and no change in mortality across the study period (overall rate of 0.07% in 2014 and 0.05% in 2018; P = .89). The number of patients receiving NIV, IMV, and ultimately dying was small and beneath the reporting threshold permitted by our data use agreements.

Patients receiving NIV exhibited an LOS distribution that was intermediate between those who received no mechanical assistance and those who received IMV (Fig 3). However, because LOS distributions are skewed and short stays dominate average LOS calculations, this had little impact on aggregate LOS measures. Overall average LOS was therefore similar among centers with high and low NIV use (∼2 days; IQR: 1–2 days).

Although overall ICU use declined nationally, ICU admission rates were much greater in institutions with high NIV use. In 2018, the median ICU admission rate for hospitals in the highest quartile of NIV use was 30.0%, compared with 15.9% for all other hospitals (P < .01). Even so, more than one-quarter (26.6%) of patients with asthma receiving NIV were not admitted to the ICU. Among patients who received NIV, those who received it in the ICU had longer admissions than those who received it outside of the ICU (3 days [IQR: 2–5 days] versus 2 days [IQR: 1–3 days]; P < .01).

In this national sample of children’s hospitals, we found that NIV has replaced IMV as the most common mode of mechanical ventilation in children admitted for asthma. However, we observed considerable practice variation, with NIV used primarily in a subset of hospitals that included centers with the highest asthma admission volumes in the country. Mortality, IMV use, and overall average LOS were similar among centers with high NIV use and low NIV use, but ICU admission rates were much higher in centers with high NIV use.

These trends in NIV and IMV use occurred in the context of significant overall declines in both general and ICU admissions for asthma. Advances in outpatient care may be responsible for these declines, and, if so, illness severity among remaining admissions could be increasing. Although this might bring need for more inpatient ventilatory support, and NIV may be filling some of this need, it does not explain the dramatic practice variability we observed, the similar IMV rates across centers, and the absence of an association between NIV and IMV use.

Preferential use of NIV over IMV departs from previous practice descriptions^5,6,8,11  but is consistent with a more recent single-center report.⁹  Given the risks and morbidity attendant to IMV in patients with asthma, NIV has been proposed as a safer, intermediate alternative that potentially reduces the need for endotracheal intubation.^12,13  Indeed, the overall median IMV rate of 0.52% suggests that intubation for asthma has become uncommon. It is now lower than the 1% overall intubation rate that has been previously reported,¹¹  and the 3.7% IMV rate among ICU patients is far below previous reports of 6% to >20%.^5,6,8–10 

We were not able to discern a statistical association between NIV use and IMV use. Even in centers with high NIV use, where the NIV rate more than doubled over the study period, we did not observe an impact on IMV rates. It is possible that the absence of a statistical association simply reflects site practice variation and the relatively low frequency of this target outcome. However, the clinical effectiveness of NIV in asthma remains uncertain,¹⁹  and it is equally possible that the increased uptake of NIV has had no impact on intubation rates.

ICU admission is a primary driver of hospitalization costs,⁹  and ICU crowding is a problem in many institutions.^21,22  In the sample here, ∼75% of patients receiving NIV were admitted to an ICU, and, in many institutions, NIV therapy occurs only in the ICU. Although we did observe that high-NIV users had higher ICU admission rates than low-NIV users, ICU admission is driven by many factors rooted in the need for continuous monitoring, treatment, and reassessment. Because we could not control for severity of illness, it is not possible to directly attribute higher ICU use to higher NIV use. We also observed that patients with asthma receiving NIV had significantly longer LOS than those who did not. Given the limitations of retrospective analyses, we believe that the impact of NIV use on patient outcomes, overall costs, and hospital crowding are best determined prospectively.

Our findings should also be considered in the light of several other limitations. First are all limitations inherent in retrospective studies using administrative data sets.^23,24  These range from inaccuracies within single-patient records to systematic errors in coding or data handling. Here, we sought to reduce such errors by examining large numbers of patients, changes over time, and multiple institutions. Second, we used primary codes to identify mechanical ventilation use and ICU flags to identify ICU admission. When both were present, we assumed that mechanical ventilation was provided within the ICU, but some patients may have begun NIV outside of an ICU and been transferred later. Third, because we relied on primary diagnoses and did not exclude patients with comorbidities, some patients may have received NIV or IMV for treatment of conditions other than asthma (eg, patients with obstructive sleep apnea, bilevel positive airway pressure dependence, or complex neuromuscular or congenital heart disease). To the extent that such patients were present in our sample, numbers here may overestimate NIV and/or IMV use in uncomplicated asthma. Finally, because the PHIS data set only contains information from participating children’s hospitals, our findings may not reflect care trends in non-PHIS institutions.

NIV is now the predominate mode of ventilatory support in the care of children with asthma, with considerable practice variation. High NIV use was not associated with decreased IMV rates or with decreased mortality. Researchers should use future studies to clarify the impact of NIV on the quality, costs, and outcomes of asthma care.

 
• CPT

  Current Procedural Terminology

 
• ICD-9

  International Classification of Diseases, Ninth Revision

 
• ICD-9-CM

  International Classification of Diseases, Ninth Revision, Clinical Modification

 
• ICD-10

  International Classification of Diseases, 10th Revision

 
• ICD-10-PCS

  International Classification of Diseases, 10th Revision, Procedure Coding System

 
• IMV

  invasive mechanical ventilation

 
• IQR

  interquartile range

 
• LOS

  length of stay

 
• NIV

  noninvasive ventilation

 
• PHIS

  Pediatric Health Information System