Predictors of In-Hospital Mortality in Helicobacter pylori–Negative, Non-Variceal Upper Gastrointestinal Bleeding: A Retrospective Cohort Study

Introduction

Upper gastrointestinal bleeding (UGIB) is a common medical emergency with significant clinical and epidemiological implications. It remains a major cause of hospitalization, blood transfusion, and mortality worldwide.¹ Historically, peptic ulcer disease secondary to Helicobacter pylori infection has been the most prevalent etiology of UGIB, particularly in developing regions.² However, the prevalence of H. pylori has progressively declined in many populations due to eradication campaigns, improved sanitary conditions, and the widespread use of proton pump inhibitors, which has altered the etiological profile of UGIB.³

UGIB is broadly classified into variceal and non-variceal etiologies.⁴ The most common causes of non-variceal UGIB include peptic ulcer disease, gastritis/duodenitis (erosive or hemorrhagic), vascular lesions (eg., angiodysplasia, Dieulafoy’s lesion), malignancies, and bleeding related to antithrombotic or anti-inflammatory therapy.⁵ Although H. pylori infection is not a direct cause of UGIB, it is a well-established etiological factor for bleeding peptic ulcers and chronic active gastritis, which may predispose to hemorrhage.⁶ Importantly, H. pylori-associated gastritis is typically non-erosive and distinct from the erosive-hemorrhagic gastritis frequently observed in Nonsteroidal Anti-Inflammatory Drug (NSAID) users or critically ill patients.⁶ Notably, persistent H. pylori infection in the context of NSAID use has been linked to a higher risk of recurrent bleeding, underscoring the complex interplay between infectious, pharmacological, and mucosal factors in UGIB pathogenesis.^4,6

Accurate diagnosis of H. pylori infection during acute UGIB is challenging. The rapid urease test (RUT), commonly used at endoscopy, has reduced sensitivity due to blood, recent proton pump inhibitor (PPI) use, and mucosal disruption, often yielding false-negative results.^7,8 Test accuracy improves when biopsies are obtained ≥48 hours after hemostasis.^7,9 Accordingly, guidelines recommend deferring H. pylori testing or using alternative methods (histology, stool antigen, or urea breath test) after the acute phase.^6,9 In clinical practice, however, many patients are classified as H. pylori-negative based solely on a negative RUT during index endoscopy, potentially leading to misclassification.⁶

This has contributed to a rise in H. pylori-negative UGIB, increasingly driven by antithrombotic therapy, comorbidities (cardiovascular, metabolic, hepatic), and advanced age rather than NSAIDs alone.¹⁰ Bleeding peptic ulcers still occur in this group, often idiopathically or due to antiplatelet/anticoagulant use.^4,6 Endoscopically, lesions range from ulcers (classified by Forrest criteria⁷) to non-ulcerative mucosal injury, including erosive gastropathy and vascular lesions linked to endothelial dysfunction.^7,11 This heterogeneity underscores the need for a focused characterization of H. pylori-negative UGIB to guide personalized management.⁴

Clinically, advanced age and male sex have been linked to worse outcomes in UGIB, though gender differences may reflect comorbidity burden rather than biological factors.¹² Hypertension, type 2 diabetes, and chronic kidney disease are the most prevalent comorbidities, all associated with higher bleeding risk and impaired hemodynamic response.^13,14 In H. pylori-negative UGIB, these conditions contribute not only to mucosal injury but also to more complex clinical courses and greater care needs.⁶ Presentation with hematemesis typically indicates acute, high-volume bleeding, whereas melena often reflects slower, distal bleeding, commonly seen in older patients with altered gastrointestinal transit.^15,16 Although initial symptoms may influence diagnostic and therapeutic timeliness, mortality in UGIB is now driven less by the bleeding source and more by systemic factors.^15,17

Despite advances in endoscopic and multidisciplinary care, in-hospital mortality remains significant, particularly in patients with comorbidities or organ failure.^7,16 Age >60 years, cardiovascular events, and encephalopathy are well-established mortality predictors and are incorporated into validated scores such as Rockall,¹⁸ Glasgow-Blatchford,¹⁹ and AIMS65.²⁰ Timely recognition of these factors is critical for risk stratification and resource optimization. This study aims to characterize the etiology, clinical profile, and independent predictors of in-hospital mortality in patients with H. pylori-negative, non-variceal UGIB at a quaternary hospital in Colombia, with a focus on comorbidity burden, antithrombotic exposure, and systemic risk thereby addressing a key evidence gap in Latin American gastroenterology.

Materials and Methods

Study Population and Sample

A retrospective, observational, and analytical study was conducted at a fourth-level referral hospital located in Barranquilla, Atlántico (Colombia), between January 1, 2021, and December 31, 2023. Inclusion criteria were: (i) age ≥18 years; (ii) clinical suspicion of UGIB, defined by the presence of hematemesis, coffee-ground emesis or gastric aspirate, melena, or maroon stool (in cases suggestive of rapid or massive upper GI bleeding); and (iii) confirmation of UGIB by esophagogastroduodenoscopy performed during the index hospitalization, showing a visible source of bleeding (eg., ulcer, erosive gastritis, vascular lesion).

Cases were initially identified through the hospital’s electronic medical records using ICD-10-CM diagnosis codes for UGIB: K92.0 (hematemesis), K92.1 (melena), and K92.2 (gastrointestinal hemorrhage, unspecified). This automated query retrieved all patients potentially diagnosed with UGIB during the study period. All retrieved records were then manually reviewed by two independent investigators to verify eligibility.

All included patients were in-patients in whom active H. pylori infection was ruled out during the index endoscopy using direct diagnostic methods on gastric biopsies: RUT in 248 patients (87.0%), histopathological examination (hematoxylin-eosin and/or Giemsa staining) in 212 patients (74.4%), or both in 175 patients (61.4%). A patient was classified as H. pylori-negative if all performed tests were negative. Indirect tests (serum IgG, stool antigen, urea breath test) were not used during acute hospitalization, per current guidelines.⁹

Patients were excluded if they had variceal bleeding (esophageal or gastric, of portal or hepatocellular origin), a confirmed H. pylori infection during hospitalization or follow-up, incomplete records, significant missing data, or readmission for recurrent UGIB within the same episode. A census sampling strategy was used, identifying all eligible cases from electronic medical records and departmental databases (Gastroenterology and Internal Medicine).

Variables

The variables were defined and classified into sociodemographic, clinical, paraclinical, outcome, and diagnostic categorization domains. Sociodemographic variables included age (in years, analyzed as both a continuous and dichotomous variable: ≤60 and >60 years), with the cut-off selected based on its inclusion in validated UGIB prognostic scores (eg., Rockall and AIMS65) and its established association with increased mortality risk in UGIB;^18,20,21 sex (male or female), and health insurance affiliation scheme.

Clinical variables encompassed the initial presentation of UGIB, categorized as: (i) hematemesis (with or without melena) or (ii) melena without hematemesis (ie., isolated melena). Additional clinical variables included vital signs at admission (blood pressure, heart rate) and comorbidities including hypertension, type 2 diabetes mellitus, chronic kidney disease, liver cirrhosis, ischemic heart disease, decompensated heart failure, prior ischemic stroke, active malignancy, as well as NSAID or antithrombotic use, all identified through documented medical history and medication reconciliation at admission. Validated prognostic scores including the admission Rockall score¹⁸ were calculated for all patients using available admission data.

Paraclinical variables included endoscopic and histopathological findings, categorized as gastric or duodenal ulcer, chronic gastritis, vascular lesions, or other diagnoses based on available biopsy reports. Medical complications during hospitalization were also recorded and defined as follows: (1) cardiovascular events: acute coronary syndrome, acute heart failure, or clinically significant arrhythmias confirmed by clinical assessment and supporting diagnostics (ECG, troponin, or echocardiography); (2) pneumonia: new pulmonary infiltrate on chest imaging plus at least one of: fever (>38°C), leukocytosis (>11,000/μL), or purulent respiratory secretions;²² (3) Complicated urinary tract infection (UTI): positive urine culture with systemic signs such as fever, leukocytosis, or sepsis;²³ (4) Acute kidney injury (AKI): defined according to KDIGO criteria as an increase in serum creatinine by ≥0.3 mg/dL within 48 hours or a 1.5-fold rise from baseline known or presumed to have occurred within the prior 7 days;²⁴ and (5) encephalopathy: acute altered mental status not attributable to primary neurological disease, classified as hepatic (in cirrhotic patients with elevated ammonia), septic/metabolic (in the context of infection or organ failure), or unspecified.²³

The primary outcome was in-hospital mortality (discharge status: survivor vs. deceased). Additional Outcome variables included length of hospital stay, categorized based on institutional clinical care pathways and literature-derived benchmarks for UGIB: expected stay (≤3 days) reflects uncomplicated cases eligible for early discharge; adjusted stay (4–7 days) represents standard recovery; and prolonged stay (>7 days) indicates complications, clinical instability, or need for extended monitoring.¹⁷

Ethical Considerations

The Research Ethics Committee of Clínica de la Costa, through Minute No. 527 of 2025, and the Institutional Review Board of the Faculty of Health Sciences at Universidad Simón Bolívar, through Minute No. 640–017-2025, approved the conduct of this study. In accordance with Resolution 8430 of 1993 issued by the Colombian Ministry of Health, the study was classified as risk-free, as it was a documentary, observational, and retrospective investigation based solely on the review of medical records. Therefore, both committees authorized access to and analysis of the data without requiring informed consent from patients, in compliance with the ethical regulations applicable to this type of research.

Statistical Analysis

The normality of continuous variables was assessed using the Kolmogorov–Smirnov test. As all continuous variables showed non-normal distributions, they are reported as median (range). Categorical variables are presented as absolute and relative frequencies. To compare continuous variables (eg., age, length of stay) between two independent groups (eg., female vs. male; hematemesis vs. melena; survivors vs. non-survivors), the Wilcoxon rank-sum test was used. For comparisons across three or more groups (eg., hospital stay categories: expected, adjusted, prolonged), the Kruskal–Wallis test was applied. Associations between categorical variables were evaluated using Pearson’s chi-squared test or Fisher’s exact test when expected cell counts were <5. Given the exploratory nature of the study and the low number of events (n=13 deaths), a sample size calculation was not performed a priori; however, post hoc analysis indicated that our cohort provided sufficient power (≥80%) to detect large effect sizes (OR ≥3) for major predictors.

Multivariable logistic regression was used to identify factors independently associated with in-hospital mortality. Candidate variables for the initial model were selected based on clinical relevance and univariable analysis (p < 0.20). A backward stepwise elimination procedure was applied, removing the least significant variable at each step (p > 0.05) until only statistically significant predictors (p < 0.05) remained in the final model. This approach balances model parsimony and predictive accuracy in settings with limited events. All analyses were conducted using R (version 4.3.0),²⁵ and a two-sided p-value < 0.05 was considered statistically significant.

Results

Baseline Characteristics

Of 342 patients initially identified through ICD-10 codes, 285 met all inclusion criteria and were included in the final analysis. The main reasons for exclusion were variceal bleeding (n=18), confirmed H. pylori infection (n=12), incomplete medical records (n=15), misclassified lower gastrointestinal bleeding (n=7), and recurrent bleeding within the same hospitalization (n=5) (Figure 1).

Figure 1 Study Flow Diagram of Patient Selection.

The cohort consisted of 285 adults (median age 60 years, IQR 39–73; 54% male) hospitalized for H. pylori-negative, non-variceal UGIB (Table 1). Hematemesis (± melena) was the most common presentation (53%), followed by isolated melena (47%). The median length of stay was 6 days (IQR 3–15), and the admission Rockall score was 2 (IQR 1–3). Comorbidities were prevalent: hypertension (42%), type 2 diabetes (17%), and chronic kidney disease (8.4%). Notably, NSAID use was rare (2.8%), whereas antithrombotic exposure was common (28.8%), including low-dose aspirin (18.2%), dual antiplatelet therapy (DAPT) (4.9%), vitamin K antagonists (3.2%), and direct oral anticoagulants (DOACs) (2.5%). Endoscopically, only 12% of cases showed overt ulcers; the majority presented with erosive or hemorrhagic gastropathy (45%), vascular lesions (6%), or non-specific mucosal irritation (35%).

Table 1 Baseline Characteristics of Hospitalized Patients with H. pylori-Negative Non-Variceal UGIB

Endoscopically, the most common findings were erosive or hemorrhagic gastropathy (45%) and non-specific mucosal irritation (35%), with gastrointestinal ulcers present in only 12% of cases and vascular lesions (eg., angiodysplasia, Dieulafoy’s) in 6%. Among the 35 patients with ulcers, the Forrest classification revealed that the majority had low-risk stigmata (Forrest III, 51%).

Histopathological findings showed that gastrointestinal ulcers were significantly more common in patients with isolated melena than in those with hematemesis (± melena) (17% vs. 8%; p = 0.023), whereas chronic gastritis was the most frequent finding overall and did not differ significantly between groups (33% vs. 26%; p = 0.23) (Figure 2). Additionally, patients with isolated melena were more likely to have an adjusted-length hospital stay (58% vs. 42%), while those with hematemesis (±melena) more often experienced prolonged stays (54% vs. 46%) (p = 0.03 for both comparisons).

Figure 2 Distribution of histopathological findings in non-variceal UGIB without Helicobacter pylori infection, by admission diagnosis (isolated melena vs. hematemesis ± melena). Abbreviations: UGIB, Upper Gastrointestinal Bleeding; AG, Acute Gastritis; CG, Chronic Gastritis; GIU, Gastrointestinal Ulcer.

Factors Associated with In-Hospital Mortality

The overall in-hospital mortality rate was 5% (n=13). Deceased patients were significantly older (median 76 vs. 60 years; p=0.003) (Figure 3) and had a higher admission Rockall score (median 4 vs. 2; p<0.001) compared to survivors (Table 2). Antithrombotic use was markedly higher in non-survivors: low-dose aspirin (62% vs. 16%, p<0.001), DAPT (31% vs. 3.7%, p<0.001), and DOACs (38% vs. 4.0%, p<0.001). In contrast, NSAID use did not differ significantly (7% vs. 2.6%, p=0.34). Key complications associated with death included cardiovascular events (38% vs. 6.6%, p=0.002), acute kidney injury (23% vs. 4.8%, p=0.029), and encephalopathy (15% vs. 0.4%, p=0.006). Chronic kidney disease was also more frequent among deceased patients (31% vs. 7.4%, p=0.017). There were no significant differences in sex, admission diagnosis (hematemesis vs. isolated melena), hospital stay duration, or histopathological findings.

Table 2 Clinical Factors Associated with In-Hospital Mortality in H. pylori-Negative Non-Variceal UGIB

Figure 3 Comparison of age (left) and hospital stay duration (right) between deceased and surviving patients with non-variceal UGIB without H. pylori infection. Abbreviation: UGIB, Upper Gastrointestinal Bleeding.

Independent Predictors of Mortality

In multivariable analysis, four factors were independently associated with in-hospital mortality: age >60 years (adjusted OR: 2.9; 95% CI: 1.7–10.2; p = 0.05), cardiovascular complications (adjusted OR: 7.8; 95% CI: 1.9–32.0; p = 0.004), encephalopathy (adjusted OR: 8.5; 95% CI: 2.2–33.0; p = 0.009), and concomitant antiplatelet plus anticoagulant therapy (adjusted OR: 5.3; 95% CI: 1.6–17.5; p = 0.006). All other variables including comorbidities, endoscopic findings, and hospital stay duration were not statistically significant in the final adjusted model (Table 3).

Table 3 In-Hospital Mortality Predictors in Non-Variceal UGIB Without H. pylori Infection. Multivariable Logistic Regression

Discussion

In this cohort of 285 patients hospitalized for H. pylori-negative UGIB, the in-hospital mortality rate was 5% consistent with contemporary reports from high-income settings.¹⁷ The median age of 60 years reflects the epidemiological shift toward older, multimorbid patients in the context of declining global H. pylori prevalence.²¹ Critically, mortality was not driven by the bleeding source or clinical presentation, but by baseline physiological reserve and systemic comorbidities, reinforcing a patient-centered model of risk stratification.

Notably, traditional NSAID use was rare (2.8%), whereas exposure to antiplatelet or anticoagulant therapy was substantially more common. Based on admission medication records, low-dose aspirin, dual antiplatelet therapy, and oral anticoagulants were documented in nearly 29% and were associated to in-hospital mortality aligning with regional trends in aging populations with high cardiovascular risk.^26,27 This suggests that in our Latin American cohort, UGIB arises less from classic ulcerogenic drugs and more from antithrombotic-related mucosal injury, stress gastropathy, or vascular fragility.^15,28 Endoscopically, only 12% of cases showed overt ulcers; the majority presented with erosive or hemorrhagic gastropathy (45%) or non-specific mucosal irritation (35%), further supporting a non-ulcerative, comorbidity-driven pathogenesis.⁴

Advanced age (>60 years) emerged as the strongest demographic predictor of death (adjusted OR: 3.14, p=0.04), corroborating decades of evidence that aging impairs hemostatic response, delays healing, and increases vulnerability to hemodynamic stress.^17,20 This finding validates the inclusion of age as a cornerstone of prognostic scores such as Rockall and AIMS65,^18,20 which performed well in our population despite its unique etiological profile.

Cardiovascular complications during hospitalization such as acute heart failure or ischemic events were the most potent clinical predictor of mortality (adjusted OR: 7.8, p<0.001). Gastrointestinal bleeding can precipitate myocardial ischemia through hypovolemia-induced hypotension or exacerbate preexisting cardiac dysfunction, creating a vicious cycle of multiorgan decompensation.^29,30 This underscores the importance of cardiovascular monitoring and early intervention in high-risk UGIB patients, particularly those with known heart disease.⁴

Encephalopathy was also associated with a markedly elevated risk of death (adjusted OR: 8.5, p=0.008); however, this finding must be interpreted with caution, as it was observed in only three patients. While encephalopathy is a recognized marker of poor prognosis in cirrhotic patients,^31,32 its presence in non-cirrhotic or critically ill individuals likely reflects global organ dysfunction, sepsis, or metabolic derangement rather than a direct causal factor.³² Thus, we consider this association hypothesis-generating rather than definitive.

The overall comorbidity profile hypertension (42%), type 2 diabetes (17%), and chronic kidney disease (8.4%) further supports a vascular-endothelial pathogenesis rather than classic ulcerogenic mechanisms.²¹ These conditions promote microangiopathy, impair mucosal perfusion, and reduce healing capacity, creating a “fragile gut” susceptible to bleeding even in the absence of overt ulcers or erosions.^16,33 This pathophysiological shift may explain why traditional ulcer-focused interventions show limited efficacy in this subgroup.³³ Our findings affirm that validated prognostic scores remain applicable even in H. pylori-negative UGIB, as age, cardiovascular disease, and markers of systemic instability (eg., encephalopathy) are integral to tools like Rockall, AIMS65 and Glasgow-Blatchford score.^18–20 This supports their routine use in risk stratification, regardless of H. pylori status.

This study has limitations. It is retrospective and single center, limiting generalizability. The small number of encephalopathy cases precludes robust inference. By design, we did not include a H. pylori–positive comparator group, as our aim was to characterize this specific subgroup rather than contrast etiologies. Likewise, the study was not intended to assess predisposing factors for the development of UGIB, but rather to identify in-hospital mortality predictors among patients already diagnosed with non-variceal, H. pylori–negative bleeding. Nevertheless, this work strengthens the evidence base for H. pylori–negative UGIB in Latin America, a region underrepresented in global literature. It demonstrates that in this population, mortality is less about the bleeding lesion and more about the patient’s systemic resilience.

In conclusion, among hospitalized patients with H. pylori-negative non-variceal UGIB, in-hospital death is primarily predicted by advanced age, cardiovascular complications and antithrombotic use. The unexpectedly low NSAID exposure and high antithrombotic use suggest alternative pathophysiological pathways emphasizing the need for region-specific, comorbidity-focused management strategies in an aging, multimorbid era of gastrointestinal bleeding.