Incidence, predictors and prognosis of respiratory support in non-ST segment elevation myocardial infarction

Abstract

Background

The incidences of invasive mechanical ventilation and non-invasive ventilation among patients with non-ST segment elevation myocardial infarction and associated prognosis are not well characterized.

Methods

We conducted a retrospective cohort study of patients with admission diagnosis of non-ST segment elevation myocardial infarction using the US National Inpatient Sample database between 2002–2014. The exposure variable was invasive mechanical ventilation or non-invasive ventilation within 24 h of admission, compared to no respiratory support. The primary outcome was in-hospital mortality. We determined the association between respiratory support and mortality using Cox proportional hazard models.

Results

A total of 4,152,421 non-ST segment elevation myocardial infarction hospitalizations were identified, among whom 1.3% required non-invasive ventilation and 1.9% required invasive mechanical ventilation. Non-invasive ventilation use increased over time (0.4% in 2002 to 2.4% in 2014, p<0.001) while there was no definite trend in invasive mechanical ventilation use. Revascularization was lower for non-ST segment elevation myocardial infarction hospitalizations including invasive mechanical ventilation (23.9%) and non-invasive ventilation (14.5%) compared to 36.5% of those without respiratory support (p<0.001). In-hospital mortality was 3.1% for non-ST segment elevation myocardial infarction without respiratory support compared to 9.2% with non-invasive ventilation (adjusted hazard ratio 1.86, 95% confidence interval 1.74–1.98) and 37.2% with invasive mechanical ventilation (adjusted hazard ratio 3.03, 95% confidence interval 2.88–3.19). Mortality for non-ST segment elevation myocardial infarction-non-invasive ventilation is improving over time while mortality for non-ST segment elevation myocardial infarction-invasive mechanical ventilation is increasing over time.

Conclusion

Mechanical respiratory support in non-ST segment elevation myocardial infarction is used in an important minority of cases, is increasing and is independently associated with mortality. Studies of the optimal management of acute coronary syndrome complicated by respiratory failure are needed to improve outcomes.

Introduction

Non-ST segment elevation myocardial infarction (NSTEMI) is a common reason for hospitalization and is associated with adverse outcomes.^1–3 Over one million patients are hospitalized annually with acute coronary syndromes in the USA,⁴ and NSTEMI now outpaces ST segment elevation myocardial infarction (STEMI) as the underlying cause.¹ Patients with NSTEMI are older with greater comorbidity burden compared with patients with STEMI^1,3 and may manifest a different epidemiology of critical illness complications which could in turn connote worse prognosis. Importantly, the temporal improvements in NSTEMI survival have lagged behind those reported in STEMI. The incidence of respiratory failure in STEMI is increasing over time,⁵ and the mortality in STEMI patients with respiratory failure is not declining.⁵ NSTEMI represents a larger percentage of contemporary cardiac intensive care unit (CICU) admissions compared to STEMI,⁶ yet the incidence, trend, and prognosis of respiratory failure in NSTEMI has not been well characterized.

Understanding the epidemiology of respiratory failure requiring the provision of invasive mechanical ventilation (IMV) and non-invasive ventilation (NIV) and their associated prognosis in NSTEMI is important to inform optimal CICU intensivist staffing,^7–9 for clinician prognosis, for planning cardiac critical care intervention trials,¹⁰ and to benchmark CICU quality improvement efforts.¹¹ To address these knowledge gaps, we performed a nationwide retrospective cohort study to determine the incidence, predictors, and prognosis of requiring IMV and NIV in NSTEMI. We hypothesized that respiratory failure manifested by requirement for IMV and NIV would be increasing in NSTEMI patients, associated with non-cardiac comorbidities and predictive of substantially higher in-hospital mortality.

Methods

Study population

We used the National Inpatient Sample (NIS) from the Agency for Healthcare Research and Quality (AHRQ) between 2002–2014. The NIS is a large inpatient database incorporating hospital admissions from a stratified sample of 20% of hospitalizations nationwide. Use of provided weights enables national estimates.^12,13 The Johns Hopkins Institutional Review Board deemed NIS research “not human subjects research” due to the fact that the database is publicly available and contains no patient identifiers nor protected health information. The Healthcare Cost and Utilization Project (HCUP) approved use of the data set.

Diagnoses and procedures in the NIS are identified with International Classification of Disease, 9th ed., Clinical Modification (ICD-9-CM) codes. The first diagnosis in the NIS is the principal diagnosis which is the primary reason for hospital admission. We included all hospitalizations with a primary diagnosis code for NSTEMI (410.7) which has been used in prior administrative studies to identify NSTEMI in the literature,^14,15 and ICD-9 codes have been shown to be over 90% sensitive and specific for identification of acute myocardial infarction.¹⁶ A list of all ICD-9 codes used to classify the study population is displayed in the Supplemental Material. We were cognizant of the methodological checklist for NIS research published by Khera et al.¹² including identifying data points as hospitalizations rather than patients, avoiding analyses clustered at the state, hospital, and physician level, avoiding use of ambiguous or nonspecific secondary ICD-9 codes to infer in-hospital events, and using survey weighting.

Exposures and covariates

The exposure of interest was use of IMV or NIV within the first 24 h of NSTEMI hospitalization, each compared to no mechanical respiratory support. IMV use was determined by the presence of ICD-9-CM code 96.7x for mechanical ventilation or 96.0 for endotracheal intubation and NIV by the presence of ICD-9 CM code 93.90 for non-invasive ventilation; use of these codes is consistent with previously published methods for identification of IMV and NIV;^5,17–20 ICD-9 codes for IMV have been shown to be over 99% specific for receipt of IMV^18,21 and for NIV 86% sensitive and 92% specific.²² We assessed IMV and NIV use within the first 24 h of hospitalization to increase confidence that the respiratory failure was due to NSTEMI rather than due to a hospital acquired complication or secondary to sedation for a procedure. The ICD-9 codes used to identify other covariates of a priori clinical interest are displayed in Supplemental MaterialTable 1.

Outcomes

The primary study outcome was in-hospital mortality to 30 days. We also assessed IMV and NIV use in a model to identify factors associated with receipt of respiratory support and to present trends in use of IMV and NIV in NSTEMI hospitalization over time.

Statistical analysis

Across the exposure categories, the Pearson test and one-way analysis of variance were used to compare categorical and continuous variables as appropriate. Rates of IMV and NIV use per year were calculated per 1000 NSTEMI admissions. Mortality among NSTEMI patients receiving IMV and NIV over time is presented using multivariable regression models adjusting for age and sex. We used survey-weighted logistic regression models with IMV and NIV as the dependent variable to determine factors associated with IMV and NIV use. We performed survival analysis to determine the association between IMV and NIV use and mortality. In our Cox proportional hazards models, we censored at hospital discharge or admission day 30. In all models, we adjusted for factors identified a priori based on a conceptual model including demographics, comorbidity index, presence of cardiogenic shock and presence of in-hospital arrest. We used the Kaplan-Meier method to generate survival curves for each ventilation strategy and used the log-rank test to test group differences. We performed several sensitivity analyses. First, we excluded patients with in-hospital cardiac arrest who would be expected to have high rates of IMV use. Next, we excluded patients with in-hospital arrest and cardiogenic shock to determine the association with outcome in a non-shock, lower risk population. Analyses were performed with Stata/MP version 13.0 (StataCorp Inc., College Station, Texas, USA). A two-tailed p-value <0.05 was considered statistically significant.

Results

Use of respiratory support in NSTEMI

A total of 4,152,421 hospitalizations for NSTEMI were identified; among whom 1.3% required NIV and 1.9% required IMV within the first 24 h. Of NSTEMI hospitalizations, 2.0% received NIV and 3.7% received IMV at any point in the hospitalization. Early NIV use increased over time (0.4% in 2002 to 2.4% in 2014, p<0.001), while there was not a significant overall trend in IMV use which qualitatively decreased from 2002 to 2006 then increased thereafter (Figure 1). Among hospitalizations for NSTEMI that also included cardiogenic shock, 3.1% included early NIV, 24.2% included early IMV, and 72.7% no respiratory support.

Factors associated with respiratory support in NSTEMI

Demographics and clinical characteristics for NSTEMI requiring IMV and NIV versus no support are shown in Table 1. NSTEMI hospitalizations requiring respiratory support were more frequently older, female, had chronic pulmonary disease, chronic kidney disease, diabetes, and had the highest composite comorbidity index. Univariable and adjusted models to identify clinical factors associated with IMV and NIV use are shown in Table 2. For IMV: non-White race, female sex, higher comorbidity index, chronic pulmonary disease, weekend admission were associated with use of IMV. Cardiogenic shock and in-hospital arrest were associated with IMV use. Obesity, smoking, and older age were associated with slightly lower odds of IMV use. For NIV: older age, female sex, weekend admission, and non-White race were modestly associated with NIV use. The strongest predictors of NIV use included higher comorbidity index, chronic pulmonary disease, obesity and cardiogenic shock. In-hospital arrest was associated with lower odds of NIV use. Findings were similar in sensitivity analyses excluding first hospitalizations with in-hospital cardiac arrest and next with hospitalizations including in-hospital arrest or cardiogenic shock (Supplemental MaterialTable 2). Rates of sepsis and and any pneumonia were all higher among patients requiring respiratory support compared to NSTEMI patients who did not: rate of sepsis was 13.7% for NSTEMI treated with IMV compared to 4.1% for NIV and 2.0% for those requiring no support. Rate of any pneumonia was 22.2% for those requiring IMV, 25.7% for NIV and 7.5% for those requiring no support.

Outcomes for NSTEMI requiring IMV and NIV

In-hospital outcomes for patients with NSTEMI requiring IMV and NIV compared to no respiratory support are shown in Table 3. In-hospital mortality was 9.2% for NSTEMI requiring NIV and 37.2% for NSTEMI requiring IMV compared to 3.1% for NSTEMI requiring no respiratory support (p<0.001 for both comparisons). NSTEMI patients requiring either form of respiratory support were less likely to undergo coronary revascularization with PCI or CABG: for NSTEMI without respiratory support, 36.5% had revascularization compared to only 14.5% of those treated with NIV and 23.9% of those treated with IMV (p<0.001 for both). Among NSTEMI patients who did not undergo revascularization with PCI or CABG, in-hospital mortality was 42.4% for those requiring IMV, 10.3% for those requiring NIV, and 4.1% for those requiring no respiratory support. For those who received revascularization, mortality rates were lower in all cases – 20.1% for those requiring IMV, 2.6% for those treated with NIV, and 1.2% for those treated with no respiratory support. Hospital length of stay, charges, and costs were all higher for IMV and NIV compared to no respiratory support. Mortality for NSTEMI requiring NIV declined over the study period while mortality for NSTEMI requiring IMV increased over the study period (Figure 2).

Survival curves for NSTEMI treated with IMV, NIV, and no respiratory support are displayed in Figure 3. After adjustment for demographics, comorbidity, cardiogenic shock, and in-hospital arrest, NIV (hazard ratio (HR) 1.86, 95% confidence interval (CI) 1.74–1.98) and IMV (HR 3.03, 95% CI 2.88–3.19) remained associated with mortality. When hospitalizations with in-hospital cardiac arrest were excluded, findings were similar – NIV (HR 1.84, 95% CI 1.72–1.97) and IMV (HR 3.64, 95% CI 3.45–3.83) were associated with mortality in adjusted models with greater magnitude of hazard for IMV compared to the full cohort. When hospitalizations with in-hospital cardiac arrest and cardiogenic shock were excluded, findings were similar and magnitude of hazard was higher – NIV (HR 1.93, 95% CI 1.80–2.08) and IMV (HR 5.13, 95% CI 4.86–5.42) were associated with mortality in adjusted models. In the adjusted model including respiratory support and adding revascularization status, undergoing revascularization was independently associated with 63% lower odds of mortality (OR 0.37, 95% CI 0.35–0.39, p<0.001) irrespective of need for respiratory support.

Discussion

In our nationwide study of the use and outcomes of respiratory support in hospitalization for NSTEMI, we report several major findings. First, respiratory support is required in a clinically important minority of NSTEMI hospitalizations and the use of NIV is increasing significantly over time. Second, non-cardiac comorbidities, cardiogenic shock, and cardiac arrest are associated with a need for respiratory support as are lower rates of coronary revascularization. Finally, outcomes are substantially worse in NSTEMI when respiratory support is required; in-hospital mortality in NSTEMI requiring NIV is improving over time while IMV is increasing. Our findings provide estimates of the burden of respiratory failure in NSTEMI which may be useful for planning CICU intensivist staffing and critical care cardiology education. Our findings characterize the magnitude of risk in NSTEMI associated with concomitant respiratory failure and should serve as an impetus to identify strategies to mitigate this risk, perhaps through improved management of cardiac and noncardiac comorbidities, increased revascularization rates, and studies of novel critical care therapeutics in this high-risk patient population.

Use of respiratory support in NSTEMI

We demonstrate that respiratory failure is present in an important minority of NSTEMI patients – approximately one in 33 NSTEMI hospitalizations required some form of respiratory support. In comparison, one in 23 patients with STEMI required IMV or NIV.⁵ Although a minority, considering the national burden of acute coronary syndrome,² this minority represents thousands of patients nationwide with NSTEMI and respiratory failure. These patients likely represent an even larger fraction of CICU admissions – Bohula et al. reported that NSTEMI was the most common primary diagnosis at CICU admission (17% of CICU admissions) and respiratory insufficiency (over 25% of all CICU admissions) the most common indication for critical care.⁶ Jentzer et al. reported an increasing use of respiratory support over time at a single academic medical CICU.²³ Our findings add to these prior results noting that the use of NIV in NSTEMI is increasing and IMV seems to be increasing after a period of initial decline at the national level. Therefore, clinicians and healthcare systems providing care for NSTEMI patients should be prepared to provide concomitant respiratory support; educational efforts in cardiac critical care,²⁴ and CICU staffing patterns^9,25–27 should account for an increasing need for respiratory support in acute coronary syndromes. Moreover, clinicians providing critical care services to NSTEMI patients should also be cognizant of these trends, given that cardiac care is a perceived deficiency among critical care trainees.²⁸

Associated factors

We report factors associated with respiratory support, some of which are intuitive. Respiratory failure in acute coronary syndromes can result from several possible causes including pulmonary edema, pump failure, global shock, pleural effusions, acute valvular disease or mechanical complication, cardiac arrest or ventricular arrhythmia, as well as secondary non-cardiac diseases. In that setting, the association of cardiac arrest and cardiogenic shock with IMV is expected. The magnitude of association does suggest that patients at risk for cardiac decompensation are also at risk for respiratory failure and requiring respiratory support. The non-cardiac comorbidities identified, such as chronic pulmonary disease, suggest that optimal management of these conditions could forestall or treat the respiratory failure given that heart disease and lung disease are synergistic.^29,30 Of note, the counterintuitive negative association of smoking and obesity with IMV likely represents age-mediated “obesity paradox” and “smoking paradox” effects in the NIS, as previously described.⁵ We also describe higher rates of pneumonia and sepsis in NSTEMI patients requiring respiratory support. Although there are limitations given use of secondary ICD-9 codes for these definitions, these findings suggest substantial non-cardiac comorbidity among NSTEMI patients who require respiratory support.

Outcomes of respiratory failure in NSTEMI

Our results demonstrate that NSTEMI patients who require non-invasive or invasive respiratory support are at significantly and independently higher risk of prolonged length of stay, hospital complications, and in-hospital mortality – 87% higher mortality if NIV is required and over three-fold increased mortality if IMV is required. The findings are similar to those reported in other populations^5,31 and are intuitive. The magnitude of hazard that we describe should prompt further studies into means to improve this high rate of adverse outcomes. Patients with NSTEMI and respiratory failure are at increased risk as a function of severity of underlying cardiac condition, non-cardiac comorbidities, and iatrogenic preventable harm from critical care as with ventilator-induced lung injury.³² These factors all represent treatable targets to study in future intervention trials. It is noteworthy that NSTEMI patients requiring respiratory support were less likely to undergo revascularization; reasons for this disparity should be explored in future work. We demonstrate that those patients who underwent revascularization had lower in-hospital mortality rates amongst those who are at highest risk for adverse outcomes. The high rates of sepsis and pneumonia in NSTEMI patients who require respiratory support supports a conceptual model that evidence-based critical care management of non-cardiac comorbidities is needed in this high-risk patient population We also report that mortality in NSTEMI patients treated with NIV is improving, which could represent increasing experience with the technique.¹⁹ However, IMV mortality is increasing over time which lends further urgency to improve these adverse outcomes, particularly given declining mortality in the acute coronary syndrome population at large.^2,15 This increase could be due to the increasing comorbidities seen in the cardiogenic shock population.^33,34

Limitations

Limitations of our study include those inherent to the NIS – as an administrative database, data on ventilator settings, laboratories, and medical therapies are lacking. A second limitation is retrospective design. Prospective studies with granular data are needed and underway.⁶ Inherent in the study design, our inference is limited in that we rely on ICD-9 codes to identify NSTEMI and other exposures. However, our approach is consistent with other reported studies^14,15 for identification of NSTEMI. Although we attempted to mitigate this concern by the use of NSTEMI as the principal discharge diagnosis, we could not differentiate NSTEMI due to acute coronary syndrome (type 1) from NSTEMI due to supply-demand mismatch (type 2). Nor is it possible to determine the underlying etiology of respiratory failure, be it secondary to pulmonary edema, mechanical complication, respiratory muscle fatigue, or other cause. By focusing on IMV and NIV use only during the first 24 h, the respiratory failure is most likely proximal to the acute coronary syndrome in aggregate. Finally, our observational design is such that we report associations rather than causal factors, and future studies with causal inference are needed. Finally, we chose a priori to focus on NSTEMI rather than all myocardial infarctions because of the different clinical implications, important demographic differences, and differences in comorbidity burden and prognosis between STEMI and NSTEMI.

In conclusion, approximately one in 33 NSTEMI hospitalizations will require respiratory support with NIV or IMV, NIV use is increasing significantly over time, and IMV and NIV use are independently associated with poor in-hospital outcomes. Prospective epidemiological studies and intervention trials are needed in this high-risk patient population to expand knowledge and improve outcome.

Supplemental Material

Supplementary material is available at European Heart Journal: Acute Cardiovascular Care online.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Conflict of interest: The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Authors have no conflicts of interest related to this paper. T Metkus performs consulting unrelated to this subject matter for BestDoctors Inc. and Oakstone/EBIX. T Metkus received royalties for a textbook publication for McGraw-Hill publishing, unrelated to this subject matter.

References