Risk of Noncommunicable Diseases After Hospitalization for Community-Acquired Pneumonia

Abstract

Background

Community-acquired pneumonia (CAP) is a leading cause of hospitalization and mortality. Increasing evidence suggests that CAP may contribute to long-term morbidity through the development of noncommunicable diseases (NCDs). We evaluate the risk for new NCDs following hospitalization for CAP.

Methods

This was a retrospective matched cohort study of adults hospitalized between 1 June 2015 and 31 May 2016, using the Epic Cosmos database, with 5 years of follow-up. Adults hospitalized with CAP were matched 1:3 by age, sex, and race with adults hospitalized for non-CAP medical reasons. NCDs were identified using International Classification of Diseases, Tenth Revision (ICD-10) codes. For each NCD category (cardiovascular, pulmonary, central nervous system, metabolic, renal, and neoplastic), analyses were restricted to patients without a prior history of that specific NCD, thereby defining separate at-risk populations for each outcome. Absolute and population-based measures of risk were calculated. Exploratory analysis estimating the national excess of NCDs attributable to CAP was performed.

Results

Among 207 848 CAP patients and 623 544 matched controls, CAP hospitalization was associated with significantly increased risk of new NCDs over 5 years. Pulmonary disease showed the greatest absolute burden with a number needed to harm of 14 (95% confidence interval [CI], 13–15). The estimated national excess of NCDs attributable to CAP in 1 year was 146 542 (95% CI, 132 677–161 716), driven primarily by pulmonary, cardiovascular, and renal diseases.

Conclusions

CAP hospitalization is associated with a substantial and sustained increase in risk for new NCDs, supporting CAP as an acute infection that may trigger long-term multimorbidity.

Community-acquired pneumonia (CAP) remains a major cause of hospitalization in the United States (US), with an estimated 1.5 million admissions annually [1]. While acute management and short-term outcomes have been the focus of most research, increasing evidence demonstrates that patients hospitalized with CAP may develop long-term morbidity, including the development of noncommunicable diseases (NCDs) [2]. Understanding the impact of long-term sequelae is considered an emerging and important area of pneumonia research [3, 4].

Noncommunicable diseases, including cardiovascular, pulmonary, central nervous system (CNS), metabolic, kidney, and neoplastic diseases, are leading contributors to global morbidity and mortality [5]. CAP may precipitate deterioration of preexisting NCDs or trigger new-onset NCDs, with large database studies demonstrating increased risk for cardiovascular disease, CNS disease, and other chronic diseases following CAP hospitalization [6–8]. However, comprehensive analyses across multiple NCD categories using large-scale real-world data remain limited.

This study aimed to evaluate the incidence and burden of new NCDs as long-term sequelae in hospitalized patients with CAP using a large-scale, nationwide electronic health record database. We hypothesized that patients hospitalized with CAP would have an excess risk of developing new NCDs compared to those hospitalized for other medical reasons. By quantifying excess NCD burden and estimating the national impact, our findings aim to reframe the understanding of pneumonia not only as an acute respiratory illness but also a driver of long-term multimorbidity.

METHODS

Study Design

This was a retrospective matched cohort study of patients hospitalized in the US from the Epic Cosmos database. Epic Cosmos is a de-identified, limited real-world evidence dataset aggregating patient-level records from >300 million individuals from more than 1762 hospitals and 40 700 clinics across health systems in the US [9]. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [10] and was reviewed and deemed exempt by the institutional review board WCG (WIRB-Copernicus Group).

Study Population

CAP Cohort Inclusion Criteria

The CAP cohort included adults aged 18 years or older who were hospitalized between 1 June 2015 and 31 May 2016, with a principal diagnosis of pneumonia, identified using International Classification of Diseases, Tenth Revision (ICD-10) codes of J12–J18, which encompass all pneumonia, community- and hospital-acquired.

CAP Cohort Exclusion Criteria

To restrict the cohort to CAP, patients were excluded if there was evidence of hospital-acquired pneumonia or ventilator-associated pneumonia, defined by the presence of a concurrent ICD-10 code for nosocomial infection (Y95) or ventilator-associated pneumonia (J95.851). Patients were also excluded if they died within 30 days of hospitalization to ensure adequate follow-up for the assessment of long-term outcomes.

Non-CAP Cohort Inclusion Criteria

The non-CAP cohort included adults aged 18 years or older who were hospitalized between 1 June 2015 and 31 May 2016, to medical departments without concurrent ICD-10 codes of J12–J18.

Non-CAP Cohort Exclusion Criteria

Patients were excluded if the reason for admission to hospital was related to labor and delivery, or if they died within 30 days of hospitalization.

Eligible Hospitalization

For both cohorts, only the first eligible hospitalization was included. This hospitalization was defined as the index hospitalization. Patients were followed for 5 years after the first eligible hospitalization.

Matching Variables

Patients hospitalized with CAP were matched 1:3 to patients hospitalized without CAP based on sex, age, and race. Exact matching was used for sex, and nearest neighbor propensity score matching was used to match for age (with a caliper of 5 years’ difference) and race. Sex, age, and race were abstracted from the Epic Cosmos database.

Selection and Definition of NCDs

The NCDs included in this study were selected based on 2 criteria: clinical relevance and the ability to be reliably identified using standardized ICD-10 diagnostic codes in administrative data. Conditions such as health-related quality of life, functional status, frailty, and cognitive decline without a formal diagnosis were not included because they cannot be reliably ascertained using ICD-10 coding alone. The medical conditions used to define an NCD were as follows: (1) CNS disease, defined by the presence of dementia or stroke; (2) metabolic disease, defined by the presence of diabetes mellitus type 2; (3) pulmonary disease, defined by the presence of chronic bronchitis, emphysema, chronic obstructive pulmonary disease, asthma, bronchiectasis, or pulmonary embolism; (4) renal disease, defined by the presence of kidney failure; (5) cardiovascular disease, defined by the presence of myocardial infarction, angina, ischemic heart disease, valve disorders, cardiomyopathy, arrhythmia, or heart failure; and (6) neoplastic disease, defined by the presence of lung cancer, breast cancer, prostate cancer, or colon cancer. NCDs were abstracted for cases and controls from the Epic Cosmos database using ICD-10 codes. ICD-10 codes for each condition are listed in Supplementary Table 1, split by NCD-defining pathology.

Definition of New NCDs as Long-term Sequelae of CAP

Patients were followed for 5 years after the index hospitalization. For each NCD category, analyses were restricted to patients who were free of that specific NCD prior to the index hospitalization. Patients with a documented history of a given NCD before hospitalization were excluded from the population at risk for that specific outcome but were retained for analyses of other NCD categories for which they had no prior history. Thus, separate at-risk populations were defined for each NCD. For example, when evaluating the development of new pulmonary disease, all patients in both the CAP and non-CAP cohorts with a history of pulmonary disease prior to hospitalization were excluded from that analysis. A new NCD was considered a long-term sequela of CAP if the diagnosis was recorded >30 days after hospitalization, or if the diagnosis was recorded within 30 days of hospitalization and persisted beyond 30 days, to minimize misclassification of acute complications or transient conditions related to the index hospitalization.

Comparison of New NCDs Among Cohorts

The frequency, percentage, adjusted hazard ratios, and cumulative incidence of new NCDs in the 5 years following hospitalization were compared for the CAP and non-CAP cohorts.

Epidemiological Measures of Risk

Absolute and population-based measures of risk were calculated to quantify the excess burden on new NCDs associated with hospitalization for CAP.

The attributable risk increase (ARI), also referred as the risk difference, was defined as the absolute difference in the 5-year cumulative incidence of each NCD between patients hospitalized with CAP and matched patients hospitalized without CAP, representing the excess risk attributable to CAP.

The number needed to harm (NNH) was calculated as the inverse of the ARI (1/ARI) representing the number of CAP hospitalizations required to produce 1 excess diagnosis of new NCD.

The attributable fraction among the exposed (AFe) was calculated as the proportion of NCD cases among CAP patients that could be attributable to CAP hospitalization, defined as the ARI divided by the cumulative incidence in the CAP cohort.

The population attributable fraction (PAF) was calculated to estimate the proportion of all new NCD cases in the total hospitalized population attributable to CAP, incorporating both the excess risk associated with CAP and the prevalence of CAP hospitalizations.

For all measures, 95% confidence intervals (CIs) were calculated and reported [11].

Time to Event Analysis

To assess time to development of new NCD, for each population at risk, Cox proportional hazards regressions were performed. Age was included as a covariate, and matched subgroups (eg, cases and their matched controls) were included as a stratifying variable. Adjusted hazard ratios and 95% CIs were reported.

Cumulative incidence curves were created to visualize time to development of new NCDs for each disease, with CAP and non-CAP cohorts each plotted separately with 95% CIs.

National Estimate of CAP-Related New Excess of NCDs

For each NCD, we estimated the percentage at risk based on the observed rates in our CAP cohort. To estimate the number at risk nationally for each NCD, we multiplied this percent at risk by the national incidence of CAP from a previous study performed by our research group evaluating CAP hospitalizations between 1 June 2015 to 31 May 2016 [1]. To estimate the number of new excess CAP-related NCD diagnoses, we divided the number at risk by the observed NNH for each NCD. Wald 95% CIs [12] were used to calculate the percentage at risk. The 95% CI for estimate counts was calculated using the χ² approximation of the Poisson distribution [13]. To be most conservative in estimation, the upper and lower boundaries for CIs for each step in the calculation were carried forward (eg, the lower bound for the population at risk was calculated based on an estimate using the lower bound for the percentage at risk, and the upper bound for the population at risk was calculated based an estimate using the upper bound for the percentage at risk).

Statistical Analysis

Patient characteristics were depicted as median and interquartile range for continuous variables or frequency and percentage for categorical variables. Patients with missing demographic variables were removed from analysis. Differences in patient characteristics and rates of sequelae were compared using standardized mean differences between the matched cohorts. P values of <.05 were considered statistically significant. All analysis was performed in R version 4.4.1 software. The R packages tidycmprsk [14], ggplot2 [15], and ggsurvfit [16] were used for analysis.

RESULTS

Cohort Characteristics

The CAP cohort consisted of 207 848 patients. The matched non-CAP cohort consisted of 623 544 patients. Supplementary Figure 1 depicts the study flowchart to reach these cohorts. Median age was 69 years in both cohorts. Demographics and medical history are shown in Table 1.

Table 1.

Patient Characteristics^a

Variable	Hospitalized With CAP (N = 207 848)	Hospitalized Without CAP (N = 623 544)	SMD
Age, y, median (IQR)	69 (57–81)	69 (57–80)	0.010
Male sex	101 650 (48.9)	304 950 (48.9)	<0.001
Race
White	156 004 (75.1)	470 527 (75.5)	0.009
Black or African American	31 090 (15.0)	87 687 (14.1)	0.025
Other	20 754 (10.0)	65 330 (10.5)	0.016
Past medical historyb
Pulmonary disease	105 939 (51.0)	177 725 (28.5)	0.472
Cardiovascular disease	125 236 (60.3)	307 159 (49.3)	0.222
Renal disease	84 453 (40.6)	183 571 (29.4)	0.236
Metabolic disease	65 082 (31.3)	173 874 (27.9)	0.075
CNS disease	36 710 (17.7)	96 886 (15.5)	0.057
Neoplastic disease	19 995 (9.6)	49 600 (8.0)	0.059

Data are presented as No. (%) unless otherwise indicated.

Abbreviations: CAP, community-acquired pneumonia; CNS, central nervous system; IQR, interquartile range; SMD, standardized mean difference.

^aPatient characteristics depicted are for the full matched cohort without exclusions to define the population at risk.

^bPast medical history includes any preexisting disease based on International Classification of Diseases, Tenth Revision codes.

Epidemiological Measures of Risk

The population at risk and development of new NCDs for each NCD-defining pathology are depicted in Supplementary Table 2. Derivation of the population at risk for each NCD and the epidemiological measures of risk are depicted in Supplementary Figures 2–7. Table 2 depicts the ARI, NNH, AFe, and PAF for each NCD. Pulmonary disease had the highest ARI (7.2% [95% CI, 6.9%–7.4%]), which yielded the lowest number of CAP hospitalizations needed to develop an excess NCD (NNH, 14 [95% CI, 13–15]). The AFe and PAF were also greatest among patients at risk of developing new pulmonary disease (AFe, 33.2% [95% CI, 32.2%–34.1%]; PAF, 8.4% [95% CI, 8.1%–8.8%]).

Table 2.

Epidemiological Measures of Risk

Noncommunicable Disease	Pulmonary	Cardiovascular	Renal	Metabolic	CNS	Neoplastic
CAP cohort population at risk	101 909	82 612	123 395	142 766	171 138	187 853
CAP cohort new diagnoses	22 037	25 896	27 828	12 711	20 781	8114
CAP cohort rate per 100a	21.6	31.3	22.6	8.9	12.1	4.3
Non-CAP cohort population at risk	445 819	316 385	439 973	449 670	526 658	573 944
Non-CAP cohort new diagnoses	64 443	81 138	84 704	35 357	60 128	23 037
Non-CAP cohort rate per 100a	14.4	25.6	19.3	7.9	11.4	4
ARI	7.2%	5.7%	3.3%	1.0%	0.7%	0.3%
NNH	14	18	31	97	138	328
AFe	33.2%	18.2%	14.6%	11.7%	6.0%	7.1%
PAF	8.4%	4.4%	3.6%	3.1%	1.5%	1.8%

Abbreviations: AFe, attributable fraction among the exposed; ARI, attributable risk increase; CAP, community-acquired pneumonia; CNS, central nervous system; NNH, number needed to harm; PAF, population attributable fraction.

^aRepresents rate per 100 of 5-year cumulative incidence.

Time to Event Analysis

Patients with CAP had significantly higher risk for new NCDs during the 5 years of follow-up (Figure 1). Adjusting for age, patients hospitalized with CAP had 2.04 (95% CI, 1.98–2.11) times increased risk of developing new pulmonary disease, 1.86 (95% CI, 1.79–1.93) times increased risk of developing new cardiovascular disease, 1.70 (95% CI, 1.63–1.78) times increased risk of developing new metabolic disease, 1.60 (95% CI, 1.56–1.65) times increased risk of developing new renal disease, 1.49 (95% CI, 1.43–1.56) times increased risk of developing new neoplastic disease, and 1.36 (95% CI, 1.33–1.40) times increased risk of developing new CNS disease. Cumulative incidence curves demonstrated early separation between cohorts that persisted over 5 years (Figure 2).

Figure 1.

Adjusted hazard ratios for development of new noncommunicable diseases in hospitalized patients with community-acquired pneumonia. Abbreviations: CNS, central nervous system; NCD, noncommunicable disease.

Figure 2.

Cumulative incidence of new noncommunicable diseases (NCDs) following hospitalization. New NCDs depicted are pulmonary disease (A), cardiovascular disease (B), renal disease (C), metabolic disease (D), central nervous system (E), and neoplastic disease (F). The shaded areas around the community-acquired pneumonia (CAP) and non-CAP curves represent 95% confidence intervals.

National Estimates

Applying national CAP incidence, the estimated excess new NCD diagnoses attributable to CAP in 1 year included a total of 55 749 pulmonary disease, 35 150 cardiovascular disease, 30 485 renal disease, 11 273 metabolic disease, 9498 CNS disease, and 4387 cancer (Table 3). This yields an estimated 146 542 (95% CI, 132 677–161 716) total excess new NCDs attributable to CAP annually.

Table 3.

Estimations of US Burden of Community-Acquired Pneumonia–Associated New Excess Noncommunicable Disease Diagnoses in the 5 Years Following Hospitalization

Noncommunicable Disease	Pulmonary	Cardiovascular	Renal	Metabolic	CNS	Neoplastic
Proportion of patients without NCD prior to hospitalization with CAP	49.0%	39.7%	59.4%	68.7%	82.3%	90.4%
Estimated eligible US population hospitalized with CAP at risk for new NCD	780 481	632 693	945 034	1 093 388	1 310 678	1 438 692
Number needed to produce an excess diagnosis of NCD	14	18	31	97	138	328
Estimated frequency of excess CAP-associated new NCD diagnoses during 5-y follow-up	55 749	35 150	30 485	11 273	9498	4387

Abbreviations: CAP, community-acquired pneumonia; CNS, central nervous system; NCD, noncommunicable disease.

DISCUSSION

This nationwide analysis demonstrates that adults hospitalized with CAP had significantly increased risk of new NCDs over 5 years compared to matched controls. The excess risk spans renal, pulmonary, cardiovascular, CNS, neoplastic, and metabolic diseases, supporting the concept that CAP hospitalization may trigger long-term health decline and emergence of multimorbidity.

These findings are consistent with prior studies showing increased cardiovascular risk within the first year after pneumonia [6, 7]. Similarly, persistent cognitive impairment has been reported following pneumonia, particularly in older adults [8, 17, 18]. This study extends those observations by performing a comprehensive evaluation of NCDs and quantifying the US burden of new NCDs attributable to pneumonia using real-world data across a nationally representative cohort.

The epidemiological risk measures reported in this study highlight the substantial clinical and public health relevance of hospitalization for CAP as a driver of long-term morbidity. The observed attributable risk increases indicate that CAP is associated with a meaningful absolute excess risk for multiple NCDs over 5 years, particularly pulmonary, cardiovascular, and renal diseases. The low number needed to harm for pulmonary disease (NNH = 15) and cardiovascular disease (NNH = 18) indicates that relatively few CAP hospitalizations are required to result in 1 additional excess NCD. The attributable fraction among the exposed demonstrates that a substantial proportion of incidence NCDs among CAP survivors, with nearly one-third for pulmonary disease, can be attributed to the pneumonia hospitalization itself, supporting the concept of CAP as a trigger or accelerator of chronic disease. At a population level, we estimate that approximately 150 000 excess new diagnoses of NCDs may result from a single year of hospitalizations with CAP in the US, a burden that is largely unrecognized. Collectively, these findings indicate that CAP is as an infectious disease with both acute and long-term outcomes. Additionally, these results support the growing movement toward posthospitalization recovery clinics and integrated care models that include chronic disease surveillance for CAP survivors.

We identified an increase in cancer detection following hospitalization for CAP, driven by lung cancer. Because lung cancer screening is recommended for smokers after an episode of CAP, the findings in our study may reflect increased screening, closer clinical follow-up, or differences in health-seeking behavior among patients with CAP. In a recent publication, Chia et al demonstrated in mice that pulmonary infection with influenza or severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can trigger metastatic progression in dormant cancer cells through inflammatory cytokines and immune-cell expansion [19]. This investigation offers a plausible biologic mechanism that may contribute to the association between CAP and the increased cancer detection observed in our study.

Chronic low-level inflammation is increasingly recognized as a central mechanism underlying the long-term sequelae observed in patients hospitalized with CAP [20]. Even after clinical resolution, many patients exhibit persistent elevation of proinflammatory cytokines such as IL-6 at the time of hospital discharge, which has been associated with long-term sequelae [21]. A persistent low-grade inflammatory state associated with endothelial dysfunction, thrombosis, and immune dysregulation may accelerate the progression of a subclinical NCD, leading to a clinical diagnosis following CAP hospitalization. Another mechanistic consideration is the development of widespread dysbiosis following disruption of the microbiomes in the gut–lung axis [20]. These mechanisms mirror observations from respiratory pathogens such as influenza [22, 23] or SARS-CoV-2 [24, 25] acting as triggers for chronic health conditions.

Strengths and Limitations

Strengths of this study include the use of Epic Cosmos, a large, diverse national cohort from an electronic medical record system representing approximately 42% of the acute care hospital market share [26]. Additional strengths include rigorous exclusion of preexisting conditions to define new disease onset; matching of the populations by age, sex, and race; and long-term follow-up using real-world data. Because we previously conducted a population-based study estimating CAP hospitalizations in the US during the same period, we aligned the Cosmos study window accordingly. This allowed us to estimate the national burden of CAP-associated new NCDs.

This study has several limitations. As a retrospective study relying on administrative coding, misclassification is possible for both pneumonia and NCD diagnoses. In this study, CAP was identified using ICD-10 codes, which may be subject to misclassification. Patient characteristics and illness severity can influence diagnostic coding accuracy. For example, sicker patients with alternative diagnosis may present with radiographic abnormalities and be misclassified as having CAP. In both cohorts, hospitalization due to CAP was not assessed during the study follow-up outside the study period. It is also important to consider that some NCDs may have been present but undiagnosed or undocumented prior to the index hospitalization. Increased healthcare utilization with closer clinical follow-up, more laboratories, and more imaging studies after hospitalization due to CAP may have led to the detection of previously unrecognized conditions. This surveillance bias should be considered when interpreting our findings. Since patients with a prior NCD were excluded from the population at risk, we could not evaluate if hospitalization due to CAP was associated with a deterioration of a prior NCD. Additionally, despite matching based on age, sex, and race, significant differences in past medical history were observed in the matched cohort. Despite excluding patients with prior history from the population at risk for each individual NCD, patients hospitalized with CAP may exhibit greater clinical frailty and thus predispose them to new NCDs. An important consideration when evaluating populations of patients with CAP is that frailty may act as a shared risk factor for development of CAP and NCDs. Thus, the associations observed in this study may reflect underlying frailty acting as a predictor for development of CAP and NCDs.

Although our study demonstrates a robust and consistent association between CAP hospitalization and the subsequent development of NCDs, the observational design precludes causal inference. Residual confounding, unmeasured clinical, socioeconomic, and educational factors, and potential differences in healthcare utilization between groups may contribute to the observed associations and were not accounted for in matching. Therefore, our findings should be interpreted as evidence of increased risk rather than proof of a direct causal relationship between CAP and the development of NCDs. To reach the national estimates of approximately 150 000 excess annual NCD diagnoses, we rely on several assumptions. First, that the Epic Cosmos dataset is broadly representative of the US hospitalized population. Second, that published national estimates of CAP hospitalizations derived from our prior population-based study in the city of Louisville, Kentucky [1] provide an appropriate denominator. Third, that the outcome-specific populations at risk observed in Cosmos reasonably reflect those of the broader US population. Given these assumptions, our national estimates should be interpreted as hypothesis generating rather than precise estimates of US disease burden.

Future research in long-term sequelae of CAP may explore the role of specific pathogens in the development of NCDs. Pathogens with different virulence, tissue tropism, and inflammatory host response may favor the development of specific NCDs. When a patient is hospitalized with CAP, therapeutic interventions are focused on the prevention of acute sequelae. Longitudinal biomarkers to define patients at risk and interventions to prevent or mitigate the development of long-term sequelae need investigation. Vaccination against respiratory pathogens is already central to CAP prevention. The possible role of these vaccines in the prevention of long-term sequelae will also need further study.

CONCLUSIONS

This nationwide analysis indicates that hospitalization for CAP is associated with a substantial and sustained increase in risk for new-onset NCDs. These findings support the reframing of pneumonia as not only an acute infection but also a driver of long-term multimorbidity. Future research is necessary to define mechanisms and interventions to mitigate the hidden burden of CAP-related NCDs.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.