Mechanical ventilation in idiopathic pulmonary fibrosis: a nationwide analysis of ventilator use, outcomes, and resource burden

Our study of IPF patients admitted to a nationwide sample of acute care hospitals found 11-12% of IPF patients admitted with a respiratory condition used MV, with no significant change from 2009–2011. Younger, male patients with fewer comorbidities and/or with a non-IPF principal diagnosis (e.g., pneumonia) were more likely to use MV. MV was associated with nearly 10-day longer hospital stays, $37,000 higher cost, and a more than 7-fold increase in mortality (56% versus 7.5%). Less than 10% of patients who used MV were discharged home routinely, compared to more than half of nonusers. Non-invasive ventilation was associated with increased LOS and cost, although to a lesser extent than MV.

The unchanging nationwide use of MV over time, despite IPF treatment guidelines conditionally recommending against MV use, reflects the limited options available to clinicians treating acute worsening of IPF and the difficulty of advance care planning in IPF. As acute worsening leading to respiratory failure can occur quickly and unexpectedly, MV can provide time to evaluate for possible treatable conditions, to assess patient preferences and/or to support gas-exchange while awaiting lung transplant. Lung transplantation remains the only curative and life-prolonging option for select patients with advanced IPF and respiratory failure. Notably, IPF patients who received MV were younger with fewer chronic medical conditions, more often admitted at a teaching hospital, and more frequently coded with a non-IPF principal respiratory diagnosis (e.g., pneumonia). This suggests a nationwide preference for MV use in younger, somewhat healthier, IPF patients or in those with a clinical suspicion of a reversible condition. Possible explanations for this finding are that younger patients with less chronic comorbidity may be potential lung transplant candidates or clinicians may feel compelled to offer them a trial of ventilator support. We cannot ascertain from the data if patients were awaiting transplant or later transferred for transplant evaluation.

The overall economic and health care burden of IPF is well-recognized [23‐27]. This study uniquely highlights the burden associated with MV use in IPF, while reinforcing with nationwide data the poor outcomes reported in prior smaller studies. Hospital cost was more than 4-fold greater and mortality 7-fold greater in IPF patients hospitalized with a respiratory problem requiring MV. While in-hospital mortality (55.3%) was lower than previously reported, this underestimates mortality as a significant number of patients were transferred to short-term hospitals (6.9%) or other facilities (20.8%) where their final vital status is unknown. Only 16.4% of MV users were discharged home. The high mortality and economic burden associated with MV in IPF stresses the need to improve the quality of medical care for IPF patients, including advances in prevention, treatment, and patient-clinician shared decision-making. While recently approved pharmacologic therapies slow disease progression and may reduce acute exacerbations [7, 8, 28], the course of IPF remains unpredictable. Therefore, early patient-centered discussions on treatment expectations, appropriate referrals for transplant and/or palliative care, and coordination of care across providers, remain integral to honoring patients’ values while ensuring high value care. IPF-specific decision aids are needed to help guide patients.

Some conditions that lead to MV may themselves be associated with greater mortality, confounding the interpretation of our findings. We used a method similar to that of Baron and Kenny [22] to test whether MV simply mediated the mortality effects of other variables. Our results suggest that this was not true of a principal IPF diagnosis, as it remains associated with mortality in the models both with and without MV use. Both cardiovascular conditions and bacterial pneumonia had statistically significant effects in the model that did not include MV, and smaller, not statistically significant, effects in the model that included it. This suggests the association between these characteristics and in-hospital death results, at least in part, from their association with MV, which independently increases the risk of death. However, some residual confounding by indication likely still exists.

This study has limitations. First, there is debate on how to identify IPF patients using claims data. The ICD-9-CM code we used has been used before in several publications [1, 6, 23, 24], however a recent validation study (not in the NIS) found it had a positive predictive value of 30-60% [29]. While less than desirable, this positive predictive value is within the range reported in study of ICD-9-CM codes for 32 conditions (median 80.7%, mean 77%, range 23-100%) [30]. Similarly, none of the conditions we identified (e.g., COPD) were confirmed clinically. Identification relied on ICD-9-CM codes, which are designed and used primarily for billing. Our study was further limited in that the NIS does not allow patients to be followed through subsequent outpatient care, repeat hospitalizations, or transfer to other facilities. We could not determine whether a subject died, received a transplant, or was discharged home after being transferred. The study relied on secondary data collected at discharge for administrative purposes, so no clinical information, including disease severity, was available. We could not determine whether medical conditions were present on admission or developed during hospitalization, nor could we determine the order in which diagnoses were made or treatments were given. Further, less severe comorbid conditions common to patients with IPF (e.g., obesity and gastroesophageal reflux) may be undercoded. Finally, as in prior studies [23, 24] we excluded transplant-related expenditures. This exclusion allows for a close look at direct costs of IPF-related care, but underestimates the complete cost of IPF.

A strength of the study is the use of the Nationwide Inpatient Sample, which was designed to inform policy decisions regarding health and health care at national and regional levels. Previous evaluations of IPF MV use and cost have been limited to specific centers or populations (e.g., Medicare and select private insurers) and their findings may be less generalizable. The NIS includes patients with Medicare, Medicaid, and private insurance, as well as the uninsured, making this dataset the best way to produce estimates valid for the overall US population.

In a nationwide sample of IPF patients, MV was used in 11-12% of those hospitalized due to a respiratory diagnosis with no significant change in its use over time. Mechanical ventilation was more frequent in younger male IPF patients, those admitted at teaching hospitals, and those with fewer chronic medical conditions or a non-IPF respiratory diagnosis. Its use was associated with a 4-fold increase in admission cost ($49,924 compared to $11,742) and a 7-fold increase in admission mortality (56% compared to 7.5%). Further advances in IPF treatment and development of IPF-specific decision aids are needed to improve the resource burden, outcomes, and use of MV in IPF.