Respiratory support in acute heart failure with preserved vs reduced ejection fraction

Abstract Background There is little evidence addressing the use and differential impact of respiratory support in acute heart failure (AHF) patients with preserved (HFPEF) vs reduced (HFREF) ejection fraction. Therefore, our objective was to determine the usage and clinical outcomes of critical care respiratory support in AHF across the two populations. Hypothesis Respiratory support would be associated with adverse outcome in both HFPEF and HFREF. Methods We identified HFPEF, HFREF, invasive mechanical ventilation (IMV), and noninvasive ventilation (NIV) using International Classification of Disease‐Ninth Edition codes in the National Inpatient Sample between January 1, 2008 and December 31, 2014. We determined rates of IMV and NIV use. We identified predictors of need for IMV and NIV and the association between ventilation strategies and in‐hospital mortality in HFPEF vs HFREF. Results 1.3 million AHF‐HFPEF and 1.7 million AHF‐HFREF hospitalizations were included; 5.98% of AHF HFPEF hospitalizations included NIV and 0.57% included IMV. Among HFREF hospitalizations, fewer (4.1%) included NIV and more (0.93%) included IMV. In HFPEF hospitalization, NIV use was associated with 2.24‐fold increased risk for death compared to no respiratory support in an adjusted model (HR 2.24 95% CI 2.05‐2.44) and IMV use was associated with 2.85‐fold increased risk for death (HR 2.85 95% CI 2.30‐3.53). This increased risk of in‐hospital mortality was similar among HFREF patients. Conclusions Use of respiratory support is increasing among patients with both HFPEF and HFREF and associated with substantially increased mortality in both heart failure subtypes.


| INTRODUCTION
Acute heart failure (AHF) hospitalization is responsible for millions of hospital visits per year 1 and substantial morbidity, mortality, and reduced quality of life. 2 The heart failure epidemic involves both heart failure with preserved and reduced ejection fraction, and approximately half of patients with heart failure have preserved ejection fraction. 3 Heart failure with preserved ejection fraction (HFPEF) and heart failure with reduced ejection fraction (HFREF) have important differences in pathophysiology, 4 cause of death, 5 and comorbidities. 6 Specifically, patients with HFPEF frequently suffer comorbid pulmonary disease, 7 skeletal muscle abnormalities, 8 obstructive sleep apnea 9 and obesity 10,11 and right ventricular involvement and pulmonary vascular disease [12][13][14] which contribute to a syndrome distinct from that of HFREF.
AHF patients often require respiratory support secondary to pulmonary edema, reduced cardiac output, respiratory muscle fatigue, and pleural effusion and the burden of respiratory failure in AHF is increasing. 15 Respiratory support can be provided with either noninvasive ventilation (NIV) with facemask or invasive mechanical ventilation (IMV) through an endotracheal tube. [16][17][18] Need for respiratory support in AHF is increasing and associated with substantially worsened outcomes including mortality. 16,[19][20][21] Yet, there is little evidence addressing the differential impact of respiratory failure or mode of respiratory support in AHF patients with HFPEF vs HFREF. Given the pathophysiologic differences between the two syndromes, the incidence, causes, and prognostic significance of respiratory failure in the two populations may differ. Moreover, there is little evidence addressing critical care needs in the HFEPF population, which is a requisite for planning randomized trials of support in this important population. Understanding differences in respiratory support in the two populations would inform intervention, risk stratification, and planning of future trials.
To address these gaps in knowledge, we conducted a nationwide retrospective cohort study to characterize the use of IMV and NIV in AHF subpopulations of HFPEF vs HFREF as well as to identify predictors of requirement for respiratory support and associated hazard for mortality in the two populations. We hypothesized that the epidemiology of usage of IMV and NIV would be different between the two diseases and the predictors of need for IMV and NIV would differ between HFPEF and HFREF patients. Finally, we hypothesized that need for respiratory support would be associated with adverse outcome in both HFPEF and HFREF.   Table S1.
We adhered to the NIS methods advocated by Khera et al. 23

| Exposures and outcomes
Our exposure of interest was receipt of IMV or NIV compared to no support in patients with HFPEF and HFREF. IMV use was identified using ICD-9-CM code 96.7x for mechanical ventilation or 96.0 for endotracheal intubation which are specific for mechanical ventilation. 21,24 NIV was identified with code 93.90. 25 We considered only IMV and NIV within the first 24 hours because our clinical interest was respiratory failure related to initial AHF presentation rather than other factors such as surgery, pneumonia, or other hospital acquired condition. We used standard ICD-9 codes and NIS-provided Elixhauser comorbidities to define other clinical covariates of a priori clinical interest. A list of all ICD-9 codes is displayed in Supplemental Table S1.
We first considered the outcome of use of IMV or NIV vs neither modality in both the HFPEF and HFREF populations. Thus, our primary outcome for an initial set of analyses was IMV or NIV use stratified by HFPEF vs HFREF. Then, to determine the association of respiratory support with outcome in each of HFPEF and HFREF, we divided each of the HFPEF and HFREF populations into three groups: AHF using IMV, AHF using NIV, and AHF using no ventilation. To assess the association of IMV and NIV use with outcome in the HFPEF and HFREF populations, we used a primary outcome of inhospital mortality within 30 days.

| Statistical analysis
Clinical characteristics and outcomes were compared between groups with the Pearson χ 2 test for categorical variables and one-way analysis of variance for continuous variables. Annual rates of IMV and NIV use within HFPEF and HFREF were calculated per 1000 AHF patients during the study period. Multivariable regression models adjusting for age and sex were constructed according to mode of ventilation in the HFPEF and HFREF populations to identify changes in mortality over time, adjusting for clustering of patients within hospitals. To determine the predictors of need for IMV and NIV in each heart failure population, we performed survey-weighted logistic regression models with IMV and NIV as dependent variables and adjusted for factors of a priori interest based on our conceptual model. To determine the association between ventilation strategies and in-hospital mortality in HFPEF and HFREF, we performed univariable and adjusted Cox proportional hazards models censoring at hospital discharge or 30 days of hospital stay, whichever came first. Survival curves were generated with the Kaplan-Meier method and compared using the log-rank test.
Analysis was performed using Stata/MP version 13.0 (StataCorp Inc., College Station, Texas). A two-tailed P value less than .05 was considered statistically significant.

| Clinical characteristics and outcomes of respiratory support in HFPEF and HFREF
Baseline characteristics of the study cohort are displayed in Table 1. HFPEF patients were older than HFREF patients, a higher percentage of HFPEF patients were female, and a higher percentage of HFPEF patients were of white race.
Within the HFPEF group, patients treated with any respiratory support were younger, more likely to have chronic lung disease, chronic kidney disease, and to be obese. Within the HFREF group, those treated with invasive ventilation were younger. Those treated with NIV had higher prevalence of chronic lung disease and chronic kidney disease, diabetes, obesity, and smoking status. The NIV-treated AHF patients were more likely to have chronic lung disease, obesity, and smoking status in both HFREF and HFPEF groups.
Crude outcomes are shown in Table 2. Baseline rates of cardiogenic shock and in-hospital arrest were lower in the HFPEF group compared to the HFREF group; however, requirement for respiratory support in both groups was associated with higher rates of cardiogenic shock and in-hospital arrest. Both HFREF and HFPEF patients manifested high mortality if respiratory support was required. Inhospital mortality for HFPEF patients treated with NIV was 4.3% and HFPEF treated with IMV 19.1%. In-hospital mortality for HFREF patients treated with NIV was 4.7% and HFPEF treated with IMV 22.7%. Mortality has not changed over time for IMV or NIV treated patients in either group ( Figure 2). For HFREF patients, increasing age was associated with lower odds of treatment with IMV but higher odds of treatment with NIV, female sex was associated with higher odds of support with both modalities. For HFREF patients, Hispanic and Asian-Pacific Islander race were associated with higher odds of respiratory support as was chronic lung disease. Obesity was associated with higher odds of NIV but not IMV. Similar to HFPEF, cardiogenic shock was associated with substantially higher odds of IMV use and moderately increased odds of NIV use while in-hospital arrest was associated with very high odds of IMV and not associated with NIV use.

| Factors associated with respiratory support in HFPEF and HFREF
F I G U R E 1 Usage of noninvasive and invasive ventilation in acute heart failure hospitalizations with reduced and preserved ejection fraction. P < .05 for all trends increasing over time

| Epidemiology of respiratory support in HFPEF vs HFREF
Our data support that respiratory support in AHF admissions is common irrespective of LVEF, and HFPEF patients are more likely to require NIV while HFREF patients slightly more likely to require IMV.
Given the fact that over one million patients are hospitalized with heart failure each year nationally 26 and that as many as half of these heart failure patients have preserved LVEF, [27][28][29] our data confirm that many thousands of HFREF and HFPEF patients alike are exposed to

| Limitations
The main limitations of our study are related to the NIS as an administrative database. Thus, our study is unable to assess patient-level factors such as specific LVEF or other echo parameters, labs assessed and drugs delivered during the hospital stay and disease-specific cause of death. The NIS also does not provide an opportunity to evaluate postdischarge events or the specific mode of in-hospital death.
Classification of heart failure relies on accurate coding; however, the ICD-9 diagnosis and procedure codes we use are consistent with prior literature and are well validated. Finally, as an observational risk factor analysis, our study can suggest associations but not demonstrate causality and as such, future studies in these patient populations are needed. Specifically, while our findings describe the association of respiratory support with outcome in HFPEF and HFREF, that is not to imply that patients who require respiratory support should not receive it. Rather, our findings should serve as an impetus to study the optimal mode of respiratory support and also provide general prognostic information for clinicians.

| CONCLUSION
We demonstrate that the use of respiratory support with IMV and NIV is increasing in AHF hospitalizations with both HFPEF and HFREF. Requirement for respiratory support is associated with substantially higher risk for mortality irrespective of underlying HFPEF or HFREF, and mortality has not declined over time in either population.
Studies assessing the mechanism of respiratory failure in HFPEF and HFREF and novel treatment strategies are needed to address these enlarging and high-risk patient population.