Long‐Term Risk of Cardiovascular Events in Patients With Chronic Kidney Disease Who Have Survived Sepsis: A Nationwide Cohort Study
Background Long‐term cardiovascular outcomes after sepsis in patients with chronic kidney disease are not well known. We aimed to examine the risk of subsequent cardiovascular events in patients with chronic kidney disease discharged after hospitalization for sepsis in Taiwan.
Methods and Results Using complete claims data for patients with chronic kidney disease from Taiwan's National Health Insurance Research Database, we identified patients with sepsis who survived hospitalization between 2000 and 2010. Each sepsis survivor was propensity score–matched to one nonsepsis hospitalized control patient. Cox regression models were used to estimate the hazard ratios (HRs) of clinical outcomes, including major adverse cardiovascular events (myocardial infarction and ischemic stroke), hospitalization for heart failure, and all‐cause death. Among 66 961 sepsis survivors, the incidence rates of all‐cause mortality and major adverse cardiovascular events during the study period were 288.51 and 47.05 per 1000 person‐years, respectively. In comparison with matched hospitalized nonsepsis control patients, sepsis survivors had greater risks of major adverse cardiovascular events (HR, 1.42; 95% CI, 1.37–1.47), myocardial infarction (HR, 1.39; 95% CI, 1.32–1.47), ischemic stroke (HR, 1.46; 95% CI, 1.40–1.52), hospitalization for heart failure (HR, 1.55; 95% CI, 1.51–1.59), and all‐cause mortality (HR, 1.56; 95% CI, 1.54–1.58). The results remained unchanged in analyses of several subgroups of patients, and were similar in analyses accounting for the competing risk of death.
Conclusions Our findings highlight the association of sepsis with a significantly increased long‐term risk of cardiovascular events among survivors in the chronic kidney disease population.
Chronic kidney disease (CKD) affects about 10% of the general adult population worldwide, and most cases are complicated by sepsis and cardiovascular disease (CVD).1 In preclinical studies, potential biologic plausibility has shown that a sepsis‐induced inflammatory cascade was responsible for adverse cardiovascular effects through endothelial dysfunction, platelet activation, or atherosclerosis progression.2 In addition, a growing body of evidence supports the positive association between sepsis and future cardiovascular events (CVEs), which has been observed consistently in patients with numerous types of infection, ranging from respiratory or urinary tract infection3 to pneumonia requiring hospitalization4 and severe sepsis requiring intensive care unit (ICU) admission.5, 6, 7 However, most studies have involved small selected populations (eg, older or community populations), the inclusion of controls with imbalanced baseline conditions, and/or lack of consideration of the competing risk of death.5, 8, 9 Few studies have specifically addressed the role of sepsis in subsequent cardiovascular risk in populations of individuals with CKD, in whom sepsis remains the leading cause of hospitalization.10
Although studies conducted during the past decade have found a significant increase in the risk of CVEs following sepsis episodes in the dialysis population,11, 12, 13 a knowledge gap remains for nondialysis patients with CKD. Since the long‐term mortality rate of sepsis survivors remains high,14 it cannot be fully ascribed from preexisting chronic illness before onset of infection.5 To systematically address the impact of sepsis on further cardiovascular consequences, we designed a contemporary nationwide population‐based cohort study to evaluate the long‐term risk of CVD among dialysis and nondialysis patients with CKD who survived sepsis in comparison with propensity score–matched nonsepsis hospitalized controls.
The institutional review board of Taipei City Hospital approved this study (TCHIRB‐1030603‐W), and the need for a full ethical review was waived because we utilized deidentified claims data exclusively from Taiwan's National Health Insurance Research Database (NHIRD), which collects information for more than 99% of Taiwan's 23 million inhabitants. This information includes patient demographics, diagnoses, procedures, and prescriptions administered at outpatient, inpatient, and emergency services. Diagnostic information is based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM). We have described this database in detail in previous studies.15, 16, 17, 18
We identified all patients with CKD aged ≥20 years in Taiwan between January 1, 2000, and December 31, 2011. CKD cases were defined according to the registry of ICD‐9‐CM code 585 at 1 or more inpatient or 2 or more outpatient visits.19, 20 The accuracy of CKD diagnoses recorded in the database has been validated with a positive predictive value of 90.4%21 and a negative predicted values of over 90%.22 Most patients in Taiwan with coded diagnoses of CKD are categorized as having stage 3 to 5 according to the estimated glomerular filtration rate–based definition (ie, <60 mL/min per 1.73 m2).21 Two cohorts were established based on the presence or absence of sepsis in CKD claims. The sepsis cohort comprised all patients with CKD who had first‐time discharge diagnoses of sepsis (ICD‐9‐CM code 038.x) and received antibiotic treatment during hospitalization. Our previous study validated the specificity of coded diagnoses of sepsis in Taiwan's NHIRD.16 For the control cohort, we identified CKD patients who were hospitalized with nonsepsis diagnoses. We used index discharge data to examine subsequent cardiovascular outcomes, and excluded patients in both cohorts who died during hospitalizations. The index date was defined as the first day of discharge from hospitalization.
We collected data on the following baseline covariates: (1) demographic covariates (age, sex, year of index date, month of index date, monthly income, urbanization level, hospital level, and Charlson Comorbidity Index); (2) concomitant use of medications associated with CVD (antiplatelet agents, insulin, oral antihyperglycemic drugs, diuretics, β‐blockers, calcium channel blockers, angiotensin‐converting enzyme inhibitors/angiotensin receptor blockers, statins, and steroids); and (3) relevant comorbidities, defined by ICD‐9‐CM codes, which are not included in the Charlson Comorbidity Index calculation (Table 1). Because of potential confounding between the sepsis and nonsepsis cohorts, we calculated a propensity score (probability of hospitalization for sepsis) for each patient in both cohorts by accounting for baseline covariates (Table S1). We matched each patient in the sepsis cohort to a control patient based on dialysis status and similarity of propensity score, which was generated by nearest‐neighbor matching without replacement, using a caliper width equal to 0.1 of the SD of the logit of the propensity score.
The clinical outcomes of primary interest were hospitalization with the principal diagnosis of myocardial infarction (ICD‐9‐CM code 410.x), ischemic stroke (ICD‐9‐CM code 433.x, 434.x, or 436) or heart failure (ICD‐9‐CM code 428.x), and all‐cause mortality. We also considered a composite outcome—major adverse cardiovascular events (MACEs)—that included myocardial infarction and ischemic stroke. Previous studies have shown good diagnostic accuracy for the detection of comorbidities such as ischemic stroke22, 23 and myocardial infarction24 using ICD‐9‐CM codes. All patients were followed until death or December 31, 2012.
Descriptive statistics were used to characterize baseline demographic and clinical variables of the study cohort. We used a standardized difference to check for balance between the sepsis and control cohorts after matching. We calculated the incidence rates of MACEs in the two cohorts using Poisson distribution. The log‐rank test was used to assess differences in the incidence rate of CVEs following sepsis between cohorts.
We used matched Cox regression models with a conditional approach and stratification, with results reported as hazard ratios (HRs) and 95% CIs for outcomes of the interest by dialysis status. Interaction test between dialysis and outcomes of the interest was also performed. In addition, we conducted subgroup analyses to examine differences in the presence or absence of covariates between sepsis survivors and matched controls. We used the SQL Server 2012 (Microsoft Corporation, Redmond, WA) for data linkage, processing, and sampling, and SAS version 9.3 (SAS Institute, Cary, NC) for propensity score calculation. All other statistical analyses were performed with STATA statistical software (version 12.0; StataCorp, College Station, TX). Statistical significance was defined as 2‐sided P<0.05.
Characteristics of the Study Population
A total of 554 863 patients with CKD between January 2000 and December 2011 were identified. Among those patients, we identified 123 796 episodes of hospitalization for sepsis. During hospitalization for sepsis, 62 776 (50.7%) patients required ICU admission, 48 365 (39.1%) received mechanical ventilation, and 57 569 (46.5%) received vasoactive agents.
A total of 66 961 patients with CKD, including 44 352 with nondialysis CKD and 22 609 with dialysis, survived through discharge from hospitalization for sepsis and were included in the sepsis cohort. Sepsis survivors with nondialysis CKD were older (mean age 73.6 years) than those with dialysis (mean age 66.3 years). Baseline clinical characteristics and comorbidities between sepsis survivors with nondialysis CKD and dialysis are shown in Table S2. Overall, the mean age of the sepsis cohort was 71.0±13.2 years, and 52.6% of these patients were men. The median Charlson Comorbidity Index score was 8 (interquartile range, 6–10). The prevalence of comorbid conditions was as follows: diabetes mellitus, 67.9%; hypertension, 91.5%; end‐stage renal disease, 33.8%; cerebrovascular disease, 57.5%; coronary artery disease, 64.8%; heart failure, 48.0%; dyslipidemia, 53.9%; and cancer, 27.2%. A total of 65 265 patients with CKD and sepsis were matched according to propensity scores with 65 265 nonsepsis hospitalized control patients with similar baseline clinical characteristics (Table 1).
Long‐Term Risks of All‐Cause Mortality and MACEs
During the mean 2.5‐year follow‐up period, the incidence rates of all‐cause mortality and MACEs in the sepsis cohort were higher than in the control cohort (288.51 versus 177.71 and 47.05 versus 32.1 per 1000 person‐years, respectively). The Cox regression model showed that the sepsis cohort had significantly higher risks of subsequent MACEs (HR, 1.42; 95% CI, 1.37–1.47), myocardial infarction (HR, 1.39; 95% CI, 1.32–1.47), ischemic stroke (HR, 1.46; 95% CI, 1.40–1.52), heart failure (HR, 1.55; 95% CI, 1.51–1.59), and all‐cause mortality (HR, 1.56; 95% CI, 1.54–1.58), compared with the matched control cohort. When death was considered as a competing risk, the risks of MACEs, ischemic stroke, myocardial infarction, and heart failure remained significantly increased, but were attenuated, in the sepsis cohort. In analyses stratified according to dialysis status, the results remained unchanged in dialysis and nondialysis patients with CKD (Table 2).
Subgroup Analyses of the Risks of Mortality and MACEs
In subgroup analyses (Figure Panel A and Panel B; Tables S3 and S4), the increased risks of all‐cause mortality and MACEs remained consistent in the sepsis groups. Compared with matched controls, the effect of sepsis on the risk of all‐cause mortality was significantly greater in younger patients, those with lower Charlson Comorbidity Index scores, those without relevant comorbidities (eg, hypertension, diabetes mellitus, dialysis, or heart failure), those with greater numbers of organ failures, those admitted to the ICU, those with shock status, and those who received mechanical ventilator support during hospitalization. However, the effect of sepsis on the higher risk of major CVEs was not noted in some subgroups of patients, such as those with higher numbers of organ failures or shock status.
In this large national CKD cohort with long‐term and complete follow‐up, sepsis survivors had a 1.4‐fold higher rate of MACEs and a 1.6‐fold higher rate of all‐cause mortality than did matched nonsepsis hospitalized controls. The associations were also consistent across dialysis and nondialysis CKD subpopulations. The associations remained significant, but less marked, in analyses that accounted for the competing risk of death. In addition, similar associations were observed in several subgroup analyses, including those conducted according to age, sex, baseline comorbidities, and sepsis severity.
The reported risks of hospitalization for septicemia in US Medicare patients with CKD who were and were not receiving dialysis were 8‐fold and 3‐ to 4‐fold greater, respectively, than in patients without CKD,1 and short‐term outcomes after sepsis were substantially worse in nondialysis and dialysis patients with CKD than in the general population.25, 26 However, little is known about the long‐term risk of CVD after discharge from hospitalization for sepsis in the CKD population. Secondary analysis of data from a prospective study12 that enrolled 2358 dialysis patients in the United States in 1996 and 1997 showed that sepsis or bacteremia, as a time‐dependent covariate, was associated with increased future CVEs (including myocardial infarction, congestive heart failure, stroke, and peripheral vascular disease) during a median follow‐up period of 3.2 years, but this analysis was limited by the inclusion of a selected population and insufficient power and size. Another study of a US cohort of older patients (aged 65–100 years) on dialysis in 2001 and 2002, which used a self‐controlled case series method, showed a significant increase in cardiovascular risk (myocardial infarction and stroke) by 25% in the first 30 days and 18% in the first 90 days after infection‐related hospitalization.11 However, that study did not examine long‐term clinical outcomes or include nondialysis patients with CKD or those younger than 65 years. Thus, the results of our study, which included a broad range of patients with CKD, not only provide strong support for existing evidence from patients on dialysis and/or older patients, but also further extend findings to nondialysis patients with CKD and the young and middle‐aged population, which has received less attention.
Results of our subgroup analyses suggest that the greater risks of all‐cause death and CVEs after sepsis are due to the greater severity of acute sepsis, reflected by factors such as ICU admission, greater number of organ failures, shock status, and receipt of medical ventilation. Yende et al5 also found that ICU survivors of sepsis from a US Medicare cohort had significantly increased risks of death and CVEs in comparison with hospitalized control patients with infection who were not admitted to the ICU; organ dysfunction and shock status during severe sepsis appeared to slightly, but not significantly, increase cardiovascular risk in that study. The impact of sepsis per se on subsequent cardiovascular risk in the CKD population may be explained in several ways. First, sepsis may provoke acute kidney injury (AKI), especially in patients with impaired renal function. Despite recovery from acute kidney injury, long‐term coronary or stroke risk has been found to persist following acute kidney injury episodes.27, 28 Second, chronic inflammation has been shown to contribute to the link between CKD and CVD.29 Discharge from hospitalization for sepsis does not necessarily represent complete clinical remission of sepsis; ongoing subclinical inflammation30 may aggravate the progress of atherosclerosis or increase vulnerability to atherosclerotic plaques, especially in at‐risk populations with CKD.31, 32 These adverse influences of sepsis‐related inflammation may provide important molecular clues (including data on cytokines, radicals, or adhesion molecules) regarding the pathogenesis of CVEs.33
Our study has several limitations. First, definition of the primary exposure and outcomes was based on ICD‐9‐CM codes, rather than clinical diagnostic criteria, although the accuracy of these codes has been validated. Thus, potential misclassification or the presence of subclinical disease is of concern; however, we believed that any misclassification is nondifferential for the sepsis and control cohorts, and that the undetected presence of subclinical disease would more likely lead to underestimation of the associations examined. Second, the retrospective observational study design prevented examination of the underlying causality of associations between sepsis and CVD. Third, detailed information on in‐hospital parameters, such as APACHE II scores, biochemical data, and inflammatory markers, was not available in our health insurance claims database. We performed subgroup analyses based on the number of organ failures, site of infection, ICU admission, shock status, and receipt of ventilation as proxies for estimated sepsis severity. Finally, data on sepsis survivors' functional or cognitive status at hospital discharge were not included in our dataset. Nonetheless, we collected comprehensive registry data on demographic characteristics, comorbidities, and concomitant medication use after discharge from hospitalization for sepsis to serve as an overview of individual complexity.
Our findings underline the significant long‐term cardiovascular consequences in patients with CKD who have survived sepsis. We urge physicians to increase vigilance related to modifiable cardiovascular risk factors in these patients, which may help to improve overall post‐sepsis clinical outcomes. Further research is required to elucidate the nature of the interaction between sepsis and subsequent CVEs and to facilitate the identification of novel targets for intervention that can mitigate these adverse consequences in patients with CKD.
Sources of Funding
This work was supported in part by the Novel Bioengineering and Technological Approaches to Solve Two Major Health Problems in Taiwan sponsored by the Taiwan Ministry of Science and Technology Academic Excellence Program under grant number: MOST 105‐2633‐B‐009‐003; Wan Fang Hospital under grant number: 105swf01; Taipei City Government under grant number: 104XDAA00124; Taipei Veterans General Hospital (V104A‐003; V104E4‐003; V105A‐003); Taipei Veterans General Hospital‐National Yang‐Ming University Excellent Physician Scientists Cultivation Program (No. 104‐V‐B‐044).
Table S1. Propensity Score Model Results of Probability of Diagnosis of Sepsis
Table S2. Demographic and Clinical Characteristics of the CKD Patients With and Without Dialysis Who Were Hospitalized With a Diagnosis of Sepsis
Table S3. Subgroup Analysis of Risk of All‐Cause Mortality Among Patients With Sepsis and the Matched Hospitalized Control Cohort
Table S4. Subgroup Analysis of Risk of Major Cardiovascular Events Among Patients With Sepsis and the Matched Hospitalized Control Cohort
This study was based in part on data from the NHIRD provided by the Bureau of National Health Insurance (BNHI) of the Department of Health and managed by the NHRI. The conclusions presented in this study are those of the authors and do not necessarily reflect the views of the BNHI, the Department of Health, or the National Health Research Institute.
- ↵United States Renal Data System . 2015 USRDS Annual Data Report: Epidemiology of Kidney Disease in the United States. Bethesda, MD: National Istitutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases; 2015.
- ↵Merx MW, Weber C. Sepsis and the heart. Circulation. 2007;116:793–802.
- ↵Yende S, Linde‐Zwirble W, Mayr F, Weissfeld LA, Reis S, Angus DC. Risk of cardiovascular events in survivors of severe sepsis. Am J Respir Crit Care Med. 2014;189:1065–1074.
- ↵Lone NI, Gillies MA, Haddow C, Dobbie R, Rowan KM, Wild SH, Murray GD, Walsh TS. Five‐year mortality and hospital costs associated with surviving intensive care. Am J Respir Crit Care Med. 2016;194:198–208.
- ↵Ou SM, Chu H, Chao PW, Lee YJ, Kuo SC, Chen TJ, Tseng CM, Shih CJ, Chen YT. Long‐term mortality and major adverse cardiovascular events in sepsis survivors. A nationwide population‐based study. Am J Respir Crit Care Med. 2016;194:209–217.
- ↵Dalager‐Pedersen M, Sogaard M, Schonheyder HC, Nielsen H, Thomsen RW. Risk for myocardial infarction and stroke after community‐acquired bacteremia: a 20‐year population‐based cohort study. Circulation. 2014;129:1387–1396.
- ↵Dalrymple LS, Mohammed SM, Mu Y, Johansen KL, Chertow GM, Grimes B, Kaysen GA, Nguyen DV. Risk of cardiovascular events after infection‐related hospitalizations in older patients on dialysis. Clin J Am Soc Nephrol. 2011;6:1708–1713.
- ↵Foley RN. Septicemia in the United States dialysis population, 1991 to 1999. J Am Soc Nephrol. 2004;15:1038–1045.
- ↵Shih CJ, Chu H, Chao PW, Lee YJ, Kuo SC, Li SY, Tarng DC, Yang CY, Yang WC, Ou SM, Chen YT. Long‐term clinical outcome of major adverse cardiac events in survivors of infective endocarditis: a nationwide population‐based study. Circulation. 2014;130:1684–1691.
- ↵Shih CJ, Ou SM, Chao PW, Kuo SC, Lee YJ, Yang CY, Tarng DC, Lin CC, Huang PH, Li SY, Chen YT. Risks of death and stroke in patients undergoing hemodialysis with new‐onset atrial fibrillation: a competing‐risk analysis of a nationwide cohort. Circulation. 2016;133:265–272.
- ↵Ou SM, Shih CJ, Chao PW, Chu H, Kuo SC, Lee YJ, Wang SJ, Yang CY, Lin CC, Chen TJ, Tarng DC, Li SY, Chen YT. Effects on clinical outcomes of adding dipeptidyl peptidase‐4 inhibitors versus sulfonylureas to metformin therapy in patients with type 2 diabetes mellitus. Ann Intern Med. 2015;163:663–672.
- ↵Lin MY, Chiu YW, Chang JS, Lin HL, Lee CT, Chiu GF, Kuo MC, Wu MT, Chen HC, Hwang SJ. Association of prescribed Chinese herbal medicine use with risk of end‐stage renal disease in patients with chronic kidney disease. Kidney Int. 2015;88:1365–1373.
- ↵Wu VC, Wu CH, Huang TM, Wang CY, Lai CF, Shiao CC, Chang CH, Lin SL, Chen YY, Chen YM, Chu TS, Chiang WC, Wu KD, Tsai PR, Chen L, Ko WJ; Group N . Long‐term risk of coronary events after AKI. J Am Soc Nephrol. 2014;25:595–605.
- ↵Wu VC, Wu PC, Wu CH, Huang TM, Chang CH, Tsai PR, Ko WJ, Chen L, Wang CY, Chu TS, Wu KD; National Taiwan University Study Group on Acute Renal Failure (NSARF) Group . The impact of acute kidney injury on the long‐term risk of stroke. J Am Heart Assoc. 2014;3:e000933. DOI: 10.1161/JAHA.114.000933.
- ↵Epstein SE, Zhu J, Najafi AH, Burnett MS. Insights into the role of infection in atherogenesis and in plaque rupture. Circulation. 2009;119:3133–3141.