Abstract
BACKGROUND: Ventilator-associated pneumonia (VAP) is one of the most common hospital-acquired infections in ICUs and is associated with significant morbidity and mortality. Gram-negative bacteria cause 55–85% of hospital-acquired pneumonia and are associated with increased mortality.
METHODS: This study sought to describe mortality rates and 30-d readmission rates among intubated and mechanically ventilated subjects with Gram-negative pneumonia and to explore associated risk factors for mortality and rehospitalization using data from the 2013 Healthcare Cost and Utilization Project (HCUP) National Readmission Database. The study sample included adults age ≥ 18 y who were hospitalized with invasive, continuous mechanical ventilation; were discharged between February 1, 2013, and November 30, 2013; and had a primary or secondary diagnosis of Gram-negative bacterial pneumonia. Logistic regression was used to identify subject characteristics significantly associated with mortality and readmissions.
RESULTS: Using the HCUP projected sample of 32,683 intubated and mechanically ventilated subjects with Gram-negative pneumonia, the mortality rate during the index hospitalization was 24.3%. More than one fifth of subjects (22.9%) who survived the index hospitalization were readmitted within 30 d of discharge. Among subjects with readmissions, 18% occurred within 3 d of discharge, 39% occurred within 7 d of discharge, and 65% occurred within 14 d of discharge. Subjects with prior hospitalization within 30 d of the index hospitalization had a higher risk of readmission with an odds ratio of 1.70 (95% CI 1.48–1.94).
CONCLUSIONS: Mortality was high and readmissions were substantial among intubated and mechanically ventilated subjects with Gram-negative pneumonia.
Introduction
Ventilator-associated pneumonia (VAP) is one of the most common hospital-acquired infections in ICUs and is associated with significant morbidity and mortality.1,2 Data suggest VAP occurs in 9–27% of all intubated patients.1,3–5 Among intubated patients, the risk of acquiring VAP is estimated to be between 1.2 and 8.5 cases per 1,000 ventilator days.6 Critically ill patients who develop VAP have a considerably higher risk of mortality relative to similar patients without VAP.2
While data are available regarding the morbidity and mortality associated with VAP, scant data are available specific to the mortality and associated risk factors among intubated and mechanically ventilated patients with Gram-negative pneumonia. There are also limited data on the risk for readmission among intubated and mechanically ventilated patients with Gram-negative pneumonia. Thirty-day readmission has become an important metric because hospitals with excess 30-d readmissions following a pneumonia diagnosis are at risk for payment reductions by the Centers for Medicare and Medicaid Services and some private payers.7,8 Pneumonia is among the top 5 medical conditions associated with the highest 30-d rehospitalization rate among Medicare beneficiaries.9 Research has shown that the 30-d readmission rate associated with pneumonia was 22.4% in the Medicare population, and 62.6% of these readmissions occurred within 15 d after discharge from the initial hospitalization.10
The underlying etiology of pneumonia is important for the effective treatment of VAP. Gram-negative bacteria is estimated to cause 55–85% of hospital-acquired pneumonia.11,12 With the potential public health crisis posed by resistant Gram-negative bacteria, understanding the epidemiology of Gram-negative VAP is increasingly important.13,14 To address this gap in the literature, we sought to describe mortality rates and 30-d readmission rates among intubated and mechanically ventilated subjects with Gram-negative pneumonia and to explore associated risk factors for mortality and rehospitalization among these subjects using data from the 2013 Healthcare Cost and Utilization Project (HCUP) National Readmission Database (NRD).
Quick Look
Current Knowledge
Ventilator-associated pneumonia (VAP) due to Gram-negative organisms is one of the most common hospital-acquired infections in ICUs and is associated with significant morbidity and mortality. Critically ill patients who develop VAP are twice as likely to die compared to similar patients without VAP. With the rise in resistance to antibiotics, the difficulty in treating VAP, and the potential public health crisis posed by resistant Gram-negative bacteria, understanding the burden of illness associated with Gram-negative VAP is increasingly important.
What This Paper Contributes to Our Knowledge
In a large cohort of intubated and mechanically ventilated subjects with Gram-negative pneumonia, the mortality rate neared 25%. Additionally, nearly one-fifth of subjects who survived to discharge were readmitted within 30 days. Subjects with prior hospitalization were also more likely to be readmitted.
Methods
Study Design
This was a retrospective, observational cohort study using data from the 2013 HCUP NRD. HCUP is the most comprehensive source of hospital data in the United States, containing information on in-patient care, ambulatory care, and emergency department visits. It is part of a family of databases that were developed with support from the Agency for Healthcare Research and Quality.15 The databases are derived from administrative data and contain encounter-level, clinical, and nonclinical information including all listed diagnoses and procedures, discharge status, patient demographics, and charges for all patients, regardless of payer type or insurance status (eg, Medicare, Medicaid, private insurance, uninsured). Each hospitalization is listed as an individual entry and includes up to 25 diagnoses and 15 procedures as classified by International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes.16 The HCUP databases have been used in multiple publications to study health care costs and policies for VAP, as well as readmissions.16–19
The NRD is constructed from data contributed by 21 geographically dispersed states, accounting for 49.3% of the total U.S. resident population and 49.1% of all U.S. hospitalizations. Weighting criteria are provided along with the nationwide sample data to produce national estimates. The NRD is also designed to support national readmission analyses based on patient linkage numbers that could be used to track the patient across hospitals within a state. It contains > 100 clinical and nonclinical variables for each hospitalization, including timing between admissions for a patient, length of in-patient stay, patient demographics, insurance type, total charges, and hospital costs.
Subject Population
The study sample included adults age ≥ 18 y (as of the index date) who were hospitalized with invasive, continuous mechanical ventilation (ICD-9-CM: 96.70, 96.71, 96.72; Current Procedural Terminology [CPT]: 94002, 94003); were discharged between February 1, 2013, and November 30, 2013; and who had a primary or secondary diagnosis of Gram-negative bacterial pneumonia (ICD-9-CM: 482.0, 482.1, 482.2, 482.82, 482.83) (Fig. 1). The admission date of the earliest hospitalization meeting the above criteria during the subject identification time frame was defined as the index date, and all subsequent hospitalizations were evaluated as potential rehospitalizations. Subjects with index hospitalizations with a primary or secondary discharge diagnosis of viral pneumonia, fungal pneumonia, atypical organisms, Gram-positive bacterial pneumonia, or pneumonia occurring secondary to infectious disease were excluded (ICD-9-CM: 480.x, 481, 482.3x, 482.4x, 482.81, 482.84, 483.0, 483.1, 483.8, 484.1, 484.3, 484.5, 484.6, 484.7, 484.8, 486, 487.x, 488.x). The pre-index baseline period included the 30 d prior to admission for the index event, and the post-index follow-up period included the 30 d after discharge for the index event.
Selection of study population.
The study collected demographic, clinical, and admission characteristics. Subject characteristics included age, gender, and payer source. Clinical characteristics included the severity of illness classification (All Patients Refined Diagnosis-Related Groups [APR-DRG] severity based on loss of function), components of the Charlson comorbidity index, the presence of sepsis, and an indicator of hospitalization within 30 d prior to index due to chronic pulmonary disease, lower respiratory tract infection, or any cause. Admission characteristics included the presence or absence of an emergency department visit and discharge disposition.
Outcomes Variables
This study evaluated the mortality rate, readmission rate, and time to readmission among intubated and mechanically ventilated subjects with Gram-negative pneumonia. Mortality and risk factors associated with mortality were evaluated in the full study population, while the readmission rate, time to readmission, and associated risk factors were evaluated in the subgroup of subjects discharged alive from their index hospitalization. Mortality rate was defined as the number of pneumonia admissions with discharge disposition of death divided by the total number of subjects included in the study population. Readmission rate was defined as the number of subjects rehospitalized within 30 d of the index hospitalization discharge among those discharged alive from the index admission. Time to readmission was defined as the number of days between the index admission and the first readmission.
Statistical Analysis
All analyses incorporated weighting to national estimates of in-patient discharges treated at community hospitals (excluding rehabilitation and long-term acute care facilities) in the United States each year based on the sampling strata and discharge-level weights provided in the NRD. This methodology incorporates data from the American Hospital Association Annual Survey of Hospitals and the State In-patient Database to construct discharge-level weights as described by HCUP.20,21 Weights were constructed by summarizing the number of discharges among community hospitals in the U.S. NRD stratum, defined by hospital characteristics (census region, urban/rural location, teaching status, hospital size, and hospital control) and subject characteristics (gender and age category).20,21
Continuous variables were presented as means ± SD, medians, minimums, maximums, and interquartile ranges (IQRs); categorical variables were presented as frequencies (n) and percentages. Logistic regression models were fit to evaluate subject characteristics significantly associated with mortality and readmissions. The models evaluated clinical and demographic characteristics available in the dataset: age, gender, presence or absence of sepsis, hospitalization within 30 d prior to index due to chronic pulmonary disease, lower respiratory tract infection, hospitalization for any cause, and comorbid conditions used in calculating the Charlson comorbidity index. Each covariate was initially evaluated in bivariate analysis against either mortality or readmission. For multivariate models, variables were considered for inclusion when the bivariate association with mortality or readmission was P < .10 and the phi correlation coefficient was P > .25 for each respective model. The odds ratios (ORs) of subject death during index hospitalization and of readmission by selected variables within the final models are reported. SAS 9.4 (SAS Institute, Cary, North Carolina) was used to manage and analyze the data, including all statistical analyses.
Results
Study Population
After applying appropriate weighting to national estimates of in-patient discharges treated at community hospitals, the projected sample was composed of 32,683 subjects with a diagnosis for Gram-negative pneumonia. More than half of the included subjects (61%) were male, and 54% were ≥ 65 y old. The majority of subjects (63%) were in a Medicare plan, followed by private insurance (15%), Medicaid (14%), and self-pay (3%). Most subjects (78%) utilized the emergency department during their hospitalization (Table 1). Among the 24,756 subjects who were discharged alive, the demographics are similar to the full study population (Table 1).
Subject Characteristics by Mortality Status
Mortality Rate and Time to Readmission
The mortality rate among included subjects during the index hospitalization was 24.3%. More than one fifth of subjects (22.9%) who survived the index hospitalization were readmitted within 30 d of discharge. Among subjects with readmissions, 18% occurred within 3 d of discharge, 39% occurred within 7 d of discharge, and 65% occurred within 14 d of discharge (Fig. 2).
Flow chart.
Risk Factors Associated With Mortality
Among 7,927 (24.3%) subjects who died during hospitalization, 63.9% were ≥ 65 y old, 22.2% had a previous hospitalization, and 60.5% had sepsis (Table 1). The final multivariate model identified several statistically significant risk factors associated with mortality (Fig. 3). In the present analysis, subjects with concomitant sepsis had the highest risk of mortality (adjusted OR [aOR] 2.60, 95% CI 2.35–2.88), followed by subjects age ≥ 65 y (aOR 1.88, 95% CI 1.66–2.14) and those with any prior hospitalization within 30 d (aOR 1.34, 95% CI 1.20–1.49). Comorbidities upon admission with the highest risk of mortality included cancer (aOR 2.45, 95% CI 2.14–2.81), liver disease (aOR 1.91, 95% CI 1.69–2.15), AIDS/HIV (aOR 1.60, 95% CI 1.05–2.41), renal disease (aOR 1.33, 95% CI 1.21–1.47), and congestive heart failure (aOR 1.15, 95% CI 1.15–1.26). Presence of diabetes (with or without complications) was found to be associated with a lower mortality risk (aOR 0.80, 95% CI 0.72–0.89). Of note, diabetes with and without complications was combined into a single variable for the multivariate analysis. In the bivariate analysis, diabetes with complications had an OR of 1.065 (95% CI 0.86–1.31, P = .56), while diabetes without complications had an OR of 0.82 (95% CI 0.73–0.91, P < .001). Because the ORs were numerically similar in the bivariate analyses and the majority of subjects had diabetes without complications, the 2 diabetes classifications in the Charlson comorbidity index were collapsed into a single diabetes variable for inclusion in the multivariate analyses.
Risk factors associated with in-hospital mortality among intubated and mechanically ventilated subjects with Gram-negative pneumonia in final logistic regression model. Bars denote adjusted odds ratio, whiskers show 95% CI. AIDS = acquired immunodeficiency syndrome; HIV = human immunodeficiency virus; OR = odds ratio.
Risk Factors Associated With Readmission
Among subjects who were readmitted, mean age was 63.6 y, with 52.7% being age 65 y or older; almost two thirds (65.3%) were insured by Medicare (Table 2). The final multivariate model identified several statistically significant risk factors associated with readmission (Fig. 4). Presence of a prior hospitalization within 30 d of the initial hospitalization was associated with the highest risk of readmission (aOR 1.67, 95% CI 1.48–1.94). Among comorbidities, renal disease was found to have the highest risk for readmission (aOR 1.31, 95% CI 1.16–1.47), followed by liver disease (aOR 1.23, 95% CI 1.05–1.44), chronic pulmonary disease (aOR 1.16, 95% CI 1.05–1.28), and congestive heart failure (aOR 1.14, 95% CI 1.03–1.27). Subjects who had pneumonia as a secondary diagnosis (vs primary) had a lower risk of readmission (aOR 0.78, 95% CI 0.64–0.95).
Risk factors associated with 30-day readmission among intubated and mechanically ventilated subjects with Gram-negative pneumonia in final logistic regression model. Bars denote adjusted odds ratio, whiskers show 95% CI. OR = odds ratio.
Subject Characteristics by Readmission Status
Discussion
Given the limited number of studies on burden of illness involving subjects with VAP due to Gram-negative bacteria, we sought to describe mortality rates and 30-d readmission rates and associated risk factors for each outcome among intubated and mechanically ventilated subjects with Gram-negative pneumonia using data from the 2013 HCUP NRD. Overall, the incidence of mortality was found to be considerable in this study of adult intubated and mechanically ventilated subjects with Gram-negative pneumonia. One fourth of subjects in this study died during the index hospitalization, and even higher mortality rates were observed among subjects with sepsis, cancer, liver disease, and increased age. Of note, more than half of the subjects in the study were ≥ 65 y old and male, and almost one fifth of subjects had experienced a previous hospitalization. The mortality rates observed in this study among intubated and mechanically ventilated subjects with Gram-negative pneumonia are largely consistent with other studies that examined mortality in subjects with pneumonia of varying etiologies requiring intubation or mechanical ventilation. Studies examining subjects with severe community-acquired pneumonia requiring mechanical ventilation have reported mortality estimates ranging from 46% to 56%.22,23 A retrospective hospital database study noted that the mortality rate among mechanically ventilated subjects with pneumonia was 25.5%.24 Mortality estimates among subjects with VAP who require intubation range from 24% to 50% and up to 76% in some high-risk settings.25 The populations found to be at greatest risk for mortality in this study were also largely similar to previous analyses. Elderly subjects and subjects with septic shock were previously reported to be at an increased risk of mortality.26 While previous studies reported that subjects with diabetes were at an increased risk,26 the presence of diabetes was associated with a lower mortality risk in this analysis. It is important to note that the majority of subjects in our analyses had diabetes without complications, and this probably reflects a patient population with less severe disease as opposed to a true protective effect.
Our results indicate that readmission rates among intubated and mechanically ventilated subjects with Gram-negative pneumonia were high and the majority of readmissions among these subjects occured shortly after discharge. Previous research has reported similar findings among studies of subjects with pneumonia caused by varying etiologies (ie, pathogens). A retrospective database study noted that the 30-d readmission rate among subjects with pneumonia was 22.4%, and 62.6% of those readmissions occurred within 15 d after discharge from the initial hospitalization.10 These findings are similar to our results, which indicate that more than one fifth of subjects were readmitted within 30 d of discharge after initial hospitalization, and that, of all readmissions, more than one third occurred within 7 d of discharge and almost two thirds occurred within 14 d of discharge.
Among subjects who were discharged from the hospital, those who were at greatest risk included those with prior hospitalization, renal disease, and liver disease. Some of these findings are consistent with previous research. A study of subjects with health care-associated pneumonia noted 4 factors associated with readmission even after adjustment: admission from long-term care (OR 2.2), immunosuppression (OR 1.9), prior antibiotics (OR 1.7), and prior hospitalization (OR 1.7).27 Another study of health care-associated pneumonia reported the following factors to be associated with an increased risk of readmission: significantly longer hospital initial stay, higher Charlson comorbidity index, and age >60 y.28 An additional study that examined pneumonia subjects indicated that comorbidities were often responsible for readmissions; heart failure (7.4%), COPD (6.1%), and septicemia (3.6%) were specifically mentioned as leading causes related to readmission after pneumonia hospitalization.29 It is important to note that many of these risk factors for readmission may not be modifiable, but they can help identify patients who may require closer monitoring and more intensive home care upon discharge.
Limitations of this study should be noted. First, there are inherent limitations that exist within claims databases, as these data are not collected for purposes of research. Intubated and mechanically ventilated subjects with Gram-negative pneumonia were identified using ICD-9-CM diagnosis codes, which do not directly distinguish between community- and hospital-acquired infections. While all subjects were hospitalized with both continuous mechanical ventilation and a diagnosis of Gram-negative bacterial pneumonia, sufficient data are not available to confirm hospital-acquired infection, and the possibility of a subject with undiagnosed community-acquired infection being misclassified as hospital-acquired does exist. In addition, because ICD-9-CM diagnosis codes were used to identify Gram-negative bacterial pneumonia, some true cases of Gram-negative pneumonia that were not coded with sufficient sensitivity will be missed in subject selection, while some cases coded as Gram-negative pneumonia may not be confirmed by laboratory test results.
There are also some additional limitations of this dataset. The NRD does not track subjects who were hospitalized in one state and readmitted or transferred to a hospital in another state, as each of the state-specific databases use different coding for their subject linkage numbers; thus, readmission may be underreported. The risk factors observed or identified in this analysis were also limited to those available in the administrative dataset. The NRD annual files contain in-patient records for patients discharged in a calendar year. The files include patients admitted in the prior year and discharged in the current year, while excluding patients admitted to a hospital in the current year but discharged in the next year. Therefore, 30-d readmissions for subjects admitted in the latter part of the year may not be captured if the subsequent admission crossed into the next year. Lastly, HCUP NRD does not contain treatment-specific information, number of days in the ICU, or number of days on ventilation, which may influence the mortality and readmission rates reported for subjects in this analysis.
Strengths of the study should be mentioned as well. The 2013 HCUP is highly representative of the hospitalized population in the United States, making the study results generalizable to this population. In addition, the NRD is a unique database that allows one to examine readmissions in any hospital within the same state as the initial hospitalization. The research also addresses a gap in the existing literature regarding readmissions and mortality among intubated and mechanically ventilated subjects with Gram-negative pneumonia, and most of the literature examines pneumonia irrespective of the etiology.
Conclusions
Our results indicate that mortality was high in mechanically ventilated subjects with Gram-negative pneumonia, as approximately 1 in 4 subjects included in this study died during hospitalization. The findings also indicate that readmissions were substantial among intubated and mechanically ventilated subjects diagnosed with Gram-negative pneumonia. More than 1 in 5 mechanically ventilated subjects with Gram-negative pneumonia were readmitted to the hospital within 30 d after discharge, and the majority of readmissions occur within 14 d of discharge. The main risk factors for mortality identified in this analysis were the presence of sepsis, the presence of cancer, the presence of liver disease, and age ≥ 65 y; the major risk factors for 30-day readmission observed in this study included any prior hospitalization, the presence of renal disease, and the presence of liver disease. Further prospective research is needed to confirm these findings and to explore potential health care intervention strategies that would improve outcomes in this patient population, especially among those at greatest risk.
Acknowledgment
The authors thank Gerald O'Brien MD, employed at Bayer at the time this study was conducted, for his assistance on the review of the data.
Footnotes
- Correspondence: Thomas P Lodise PharmD PhD, Albany College of Pharmacy and Health Sciences, Albany, New York, 12204. E-mail: thomas.lodise{at}acphs.edu
This study was funded by Bayer HealthCare, and Xcenda was contracted by Bayer HealthCare to assist in the study and publication. Dr Lodise has disclosed a relationship with Bayer HealthCare. Dr Spilsbury-Cantalupo is a Bayer HealthCare employee. Dr Law and Ms Liao were employees of Bayer US, LLC at the time the study was conducted. Drs McCart and Eaddy are employees of Xcenda.
- Copyright © 2021 by Daedalus Enterprises
