Occurrence and Impact of Gastrointestinal Bleeding and Major Adverse Cardiovascular Events during Sepsis: A 15-Year Observational Study

Objective Sepsis patients are at risk of gastrointestinal bleeding (GIB) and major adverse cardiovascular events (MACEs), but few data are available on the occurrence of GIB and MACEs and their impact on sepsis outcomes. Methods The medical claims records of 220,082 patients admitted for sepsis between 1999 and 2013 were retrieved from the nationwide database. The adjusted odds ratios (aORs) of composite outcomes including the hospital mortality, intensive care unit (ICU) admission, and mechanical ventilation (MV) in patients with a MACE or GIB were estimated by multivariate logistic regression and joint effect analyses. Results The enrollees were 70.15 ± 15.17 years of age with a hospital mortality rate of 38.91%. GIB developed in 3.80% of the patients; MACEs included ischemic stroke in 1.54%, intracranial hemorrhage (ICH) in 0.92%, and acute myocardial infarction (AMI) in 1.59%. Both ICH and AMI significantly increased the risk of (1) ICU admission (aOR = 8.02, 95% confidence interval (CI): 6.84–9.42 for ICH and aOR = 4.78, 95% CI: 4.21–5.42 for AMI, respectively), (2) receiving MV (aOR = 3.92, 95% CI: 3.52–4.40 and aOR = 1.99, 95% CI: 1.84–2.16, respectively), and (3) the hospital mortality (aOR = 1.08, 95% CI: 0.98–1.19 and aOR = 1.11, 95% CI: 1.03–1.19, respectively). However, sepsis with GIB or ischemic stroke increased only the risk of ICU admission and MV but not the hospital mortality (aOR = 0.98, 95% CI: 0.93–1.03 for GIB and aOR = 0.84, 95% CI: 0.78–0.91 for ischemic stroke, respectively). Conclusions GIB and MACEs significantly increased the risk of ICU admission and receiving MV but not the hospital mortality, which was independently associated with both AMI and ICH. Early prevention can at least reduce the complexity of clinical course and even the hospital mortality.


Introduction
Sepsis is a complex syndrome induced by severe infection and involving acute organ failure [1]. Despite advances in drugs and treatment modalities, management of sepsis patients is a critical care challenge [2,3]. Decreasing the occurrence of sepsis-associated complications would be expected to improve hospital mortality and the clinical course by reducing the need for intensive care unit (ICU) admission and mechanical ventilation (MV) support. Few data on the incidence and impact of major adverse cardiovascular events (MACE) and gastrointestinal bleeding (GIB) in sepsis patients are available even though the concurrent development of MACE and GIB in sepsis patients is not unusual [4].
An analysis of over 119,000 patients hospitalized with sepsis between 2003 and 2012 and included in a nationwide database in the USA estimated that the incidence of GIB was 5.4% (6,571/119,684 patients). Concurrent GIB was found to increase sepsis mortality by 9% [5]. Sepsis-associated atrial fibrillation, coagulopathy, hemodynamic instability, and prolonged systemic inflammation act to trigger acute ischemic stroke. Ischemic stroke events are not unusual in patients with sepsis and thrombocytopenia, but the cause appears to be complex [6][7][8][9].
An analysis of data from over 2.6 million sepsis patients included in a national inpatient database in the USA from 2002 to 2011, found that 4.5% (118,183/2,602,854 patients) had a concurrent, nonprimary diagnosis of AMI during hospitalization. Non-ST-elevation AMI was the diagnosis of 71.4% of those cases. Hospital mortality was higher in sepsis patients with AMI (35.8%) than those with sepsis alone (16.8%, P < 0.001; adjusted odds ratio (aOR) � 1.24, 95% confidence interval (95% CI): 1.22-1.26). Invasive management concurrent AMI was associated with reduced mortality compared with conservative management (OR 0.47, 95% CI: 0.44-0.50) [10]. e treatment of sepsis patients with concurrent GIB and a MACE is complicated by difficulties in choosing among antiplatelet, anticoagulation, and hemostasis drugs. e use of predisposing medications such as antiplatelet drugs, anticoagulants, and proton pump inhibitors (PPIs) before or during the course of sepsis also complicates treatment.
is study used the 15-year nationwide database of Taiwan that included data from 1999 to 2013. e data were from 220,082 patients who were first admitted for sepsis to determine the frequency of occurrence of GIB and MACEs in the course of sepsis. e impacts and interactions of MACE and GIB on the composite outcomes of the hospital mortality, ICU admission, and receiving MV were analyzed.

Data Source.
e study database included anonymized patient and claims information retrieved from the National Health Insurance Research Database (NHIRD) of Taiwan. e records of 220,082 inpatients who were first admitted with a diagnosis of sepsis between 1999 and 2013 were included in the analysis. e NHIRD is maintained by the National Health Insurance Program, which was launched by the National Health Insurance Administration (NHIA) in 1995, and currently provides coverage for more than 23.03 million residents (>99% of the entire population). e NHIRD included the data from the clinic, district hospital, regional hospital, and medical center.

Study
Participants. Sepsis patients were identified by ICD-9-CM discharge diagnosis code 038 from the NHIRD. e positive predictive value of the sepsis (92.3%) and septic shock (97.0%) diagnoses have been previously validated [12,15].
All the enrolled sepsis patients should include a main diagnosis coding of sepsis in the first or second diagnostic coding plus a coding representing the infection origin within the first three diagnoses. e infection origin coding was referred to Angus et al. in 2001 [16]. Besides, GIB or MACE could not be the first diagnosis code or have been entered before a diagnosis code for sepsis [17,18].

Potential Confounders.
We systematically identified the potential confounders in the claims data. e identified confounding factors were age, sex, insurance premium (as a proxy of household income), level of urbanization, baseline comorbidities, and medications. e baseline comorbidities were (1) (7) cancer (ICD-9-CM: 140-208). Drug use was identified by claims indicating use for more than 1 week within a one-year period prior to the index date.

Selection Process.
Patients <18 or >100 years of age or infected with human immunodeficiency virus were excluded from the study. In the patients with repeated admissions, only data from the first hospitalization for sepsis between 1999 and 2013 were included in the analysis. e date of admission for the first hospitalization for sepsis was defined as the index date. Comorbidities were identified by ICD-9-CM codes of diagnoses made within a one-year period prior to the index date.
2.6. Ethical Approval. As the database contained deidentified data for research, the study was exempted from obtaining informed consent from the participants. is study was approved by the Institutional Review Board of Taichung Veterans General Hospital (CE18102A) and China Medical University (CMUH104-REC2-115).

Statistical Analysis.
Differences in demographic characteristics, baseline comorbidities, drug use (including aspirin, clopidogrel, warfarin, metformin, nonsteroidal anti-inflammatory drugs, statins, PPIs, steroids, and immunosuppressants), and the composite outcomes (total hospital mortality, ICU admission, and MV) were compared by the chi-squared or two-sample t-test.
Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for each variable in the logistic regression model. Adjusted ORs (aORs) for total hospital mortality, ICU admission, and receiving MV were obtained after adjusting for potential confounders including age, sex, insurance premium (a proxy for household income), urbanization level (a proxy for the accessibility of medical care), and comorbidities [15].
e Kaplan-Meier analysis was conducted to compare the cumulative incidence of hospital mortality between the patients with and without GIB and MACE, respectively.
Joint effect analysis was used to analyze the synergistic impact of sepsis complications including GIB, ischemic stroke, ICH, and AMI, on total hospital mortality, ICU admission, and MV. e 16 possible combinations of the four complications were evaluated using uncomplicated sepsis as the reference. e aORs of each combination of complications were calculated by logistic regression after adjusting for age, sex, insurance premium, urbanization level, and baseline comorbidities. e statistical analysis was performed with SAS 9.4 (SAS Institute, Inc., Cary, NC, USA). P values ≤0.05 were considered significant.

Logistic Regression Analysis of Total Hospital Mortality, ICU Admission, MV, and Complications of Sepsis.
After adjusting for age, sex, insurance premium, urbanization level, and baseline comorbidities, the aOR of GIB for total hospital mortality was 0.98 (95% CI, 0.93-1.03), the aOR for ICU admission was 1.30 (95% CI, 1.23-1.38), and the aOR for MV was 1.32 (95% CI, 1.26-1.40) with uncomplicated sepsis as the reference (

Kaplan-Meier Analysis with the Log-Rank Test.
In the Kaplan-Meier analysis, the patients with sepsis complicated with AMI and ICH had a higher cumulative incidence of hospital mortality than those without AMI or ICH (log-rank test, P < 0.001) (Figures 1 and 2). However, the opposite phenomenon was observed in patients with ischemic stroke and GIB (log-rank test, P < 0.001) (Figures 3 and 4).

Joint Effect Analysis of GIB and MACE on Hospital Mortality, ICU Admission, and MV.
e results of joint effect analysis shown in Table 3 summarize the sepsis outcomes if two or more complications occurred at the same time. e patients may have needed contrasting treatment during the sepsis course. For example, GIB needs hemostasis and cessation of antiplatelet drugs, and ischemic stroke needs antiplatelet drugs. GIB plus any MACE complicated the sepsis course by increasing ICU admissions and receiving MV. e combination did not affect total hospital mortality. Similar results were observed for other combinations such as AMI plus ischemic stroke. No specific combination of thrombotic complications such as ischemic stroke and AMI or hemorrhagic complications, such as GIB and ICH, significantly increased the risk of hospital mortality. However, the occurrence of more than one complication changed the clinical course, for example, by increasing the risk of ICU admission and MV. e combination of three or four complications was omitted because there were very few cases.

Discussion
To the best of our knowledge, this is the first and largest cohort study to comprehensively describe the individual and combined impact of GIB and MACE complications of sepsis in patients with a primary diagnosis of sepsis at the time of admission. Analysis of nationwide claims data in a cohort of sepsis patients found that GIB and MACE were associated with significantly increased risks of ICU admission and receiving MV for critical care and treatment of respiratory failure. Except for AMI and ICH, the complications did not affect mortality. In conclusion, GIB and MACE may not have a serious effect on hospital mortality as serious as was previously thought, and their occurrence will undoubtedly increase the complexity of the sepsis and hospital course. e use of preventive medications such as antiplatelet drugs, anticoagulants, statins, and PPIs should be monitored and balanced throughout the sepsis course.

AMI during
Sepsis. An analysis of a national database in the USA by Smilowitz et al. in 2016 found that 118,183 of 2,602,854 sepsis inpatients (4.5%) had a concurrent diagnosis of AMI during hospitalization. e hospital mortality was higher in sepsis patients with concurrent AMI (aOR � 1.24), which is in line with our finding of an aOR of 1.11 (95% CI, 1.03-1.19) for sepsis complicated by AMI [10]. Both studies found that AMI not only complicated the sepsis course but was also associated with a significant increase in total hospital mortality. Smilowitz et al. reported that patients who were managed more invasively had much lower mortality than those managed conservatively.

Ischemic Stroke and ICH.
Acute ischemic and hemorrhagic strokes are the fifth leading cause of death in the United States, and those who survive with subsequent longterm disabilities cause a heavy national socioeconomic burden [19]. Chronic hypertension, DM, atherosclerotic disease, and exposure to environmental toxins including air pollution have been identified as risk factors. A casecrossover study by Amelia et al. found that recent hospitalization for infection was associated with an increased risk of stroke, and that severe sepsis was associated with newonset atrial fibrillation, which also increased stroke risk [20].
A population-based cohort study of inpatients in Denmark found that about 80% of the cardiovascular events in those admitted with bacteremia occurred within 6 months [21].

GIB during Sepsis.
A database analysis of patients with septic shock in the USA by Siddiqui et al. reported that the incidence of GIB was 5.4% (6,571/119,684) in those hospitalized patients between 2003 and 2012. e occurrence of GIB was associated with a 9% increase in mortality from 45% to 54% [5]. In this study, the GIB occurred in 3.80% of the patients, and the occurrence of GIB increased mortality to 46% compared with 39% in uncomplicated sepsis. However, the adjusted OR did not find a significant association of GIB with an increased risk of mortality (aOR � 0.98, 95% CI, 0.93-1.03)). GIB is not an infrequent complication of sepsis patients, and it increases the complexity of care. Siddiqui   Adjusted OR: adjusted for age, sex, insurance premium, urbanization level, and comorbidities in the logistic regression model. * * P < 0.01; * * * P < 0.001. AMI, acute myocardial infarction; CI, confidence interval; GIB, gastrointestinal bleeding; ICH, intracranial hemorrhage; OR, odds ratio. 8 Emergency Medicine International complexity of sepsis case management but did not influence hospital mortality [18].

Concurrence of GIB and
MACEs. Joint effect analysis found that no combination of any two GIB and MACE complications had a significant effect on mortality as we previously thought, but any combination significantly increased the complexity of the sepsis course and the incidence of respiratory failure. If GIB or MACE complications were caused by sepsis and can be improved or resolved by supplemental therapy or prophylaxis, the complexity of the sepsis course would be reduced. However, hospital mortality would remain unchanged under the current best supportive care.

Limitations.
is study had some limitations, but they did not detract from the correctness of our main study results. First, as a large epidemiologic study using an administrative database, it was inevitably lacking in laboratory data such as inflammatory markers and lactate levels. Composite outcomes, including total hospital mortality, ICU admission (a proxy of critical condition), and receiving MV (a proxy of respiratory failure), were used to estimate the impacts of GIB and MACE during sepsis because they were not affected by the laboratory data alone, but those outcomes reflected poor laboratory data. Second, in the administrative database, we could not distinguish whether PPIs were used for ulcer treatment or prophylaxis. However, prophylactic PPI use did not play a role in the sepsis course [24]. ird, different pathogens, such as bacteria, virus, and fungus, can influence the occurrence of associated complications. For example, systemic salmonella infection may cause abdominal aortic infected aneurysm [25]. However, in the NHIRD, information on the definite pathogen that caused sepsis was unavailable, except for some rare specific codings, for example, ICD-9-CM: 481, pneumococcal pneumonia; ICD-9-CM: 002, typhoid/paratyphoid fever; ICD-9-CM: 01, pulmonary tuberculosis [16].

Conclusion
GIB and MACE were associated with a significantly increased risk of ICU admission and MV but not with total hospital mortality, which was independently associated with AMI or ICH alone. Early prevention of GIB and MACEs can at least reduce the complexity of clinical course and even the hospital mortality.