Predictors of In-Hospital Mortality in Patients With Type 2 Diabetes Complicated by Isolated Hyperosmolar Hyperglycemic Syndrome

Introduction

Hyperosmolar hyperglycemic syndrome (HHS) is a severe acute complication of type 2 diabetes mellitus (T2DM), marked by severe hyperglycemia, elevated plasma osmolarity, and dehydration, often accompanied by impaired consciousness [1]. Although the incidence of HHS is lower than that of diabetic ketoacidosis (DKA), its mortality rate is as high as 20–40%, particularly among older adults and patients with multiple systemic comorbidities [2,3]. In recent years, despite improvements in medical care, mortality rates for HHS still vary significantly across different countries and regions [4,5]. This suggests that its prognosis is influenced by multiple factors, highlighting the limitations and uncertainties of current management strategies. Current management of HHS emphasizes rapid fluid resuscitation, insulin therapy, and the treatment of precipitating factors; however, consensus remains lacking regarding the prevention of severe complications, particularly strategies for mitigating thromboembolic events. Accumulating evidence indicates that inflammatory activation and a hypercoagulable state play critical roles in the pathophysiology of HHS and are closely associated with adverse clinical outcomes. However, high-quality evidence regarding the efficacy and safety of prophylactic anticoagulation – such as the routine use of low molecular weight heparin (LMWH) – remains limited, and existing studies report inconsistent findings [6–8]. These uncertainties suggest that the key determinants of early mortality in HHS remain unclear, and the clinical value of targeted preventive strategies requires further validation. Accordingly, the identification of independent mortality risk factors in HHS is crucial for enhancing early recognition and informing the optimization of clinical management strategy. This retrospective cohort study aimed to identify the major risk factors for in-hospital mortality among patients with isolated HHS and to evaluate the potential impact of prophylactic LMWH on mortality risk. This study is expected to provide direct evidence to refine risk stratification for HHS and to inform more evidence-based intervention strategies, thereby addressing the current evidence gap in clinical practice and offering potential avenues for improving outcomes in this high-risk population.

Material and Methods

STUDY POPULATION:

A retrospective cohort analysis was conducted on clinical data from patients with T2DM complicated by isolated HHS and admitted to the Endocrinology and Critical Care Medicine Departments of the First Hospital of Anhui University of Science and Technology between January 2013 and April 2025. Based on the inclusion criteria, a total of 102 patients were ultimately enrolled in the study (Figure 1). Patients were categorized as non-survivors (30 cases) or survivors (72 cases) based on their clinical outcomes during hospitalization. This study was approved by the Ethics Committee of the First Hospital Affiliated to Anhui Institute of Science and Technology (Approval No.2025-KY-Y027-001) and is in accordance with the Declaration of Helsinki. All data used in this retrospective study were derived from past clinical diagnoses and treatments without any intervention, ensuring no unnecessary risks to patients. Therefore, the Ethics Committee has waived the requirement for informed consent.

To protect patient confidentiality, all identifiable information was removed during data extraction, and each case was assigned an anonymized study code. Data access was restricted to authorized researchers, and all analyses were performed using de-identified datasets stored on password-protected institutional computers.

SAMPLE SIZE JUSTIFICATION:

For the planned multivariable logistic regression analysis, a minimum of 10 events per variable (EPV ≥10) is recommended. In this study, there were 30 in-hospital deaths, which allows for the inclusion of up to 3 independent variables in the regression model, ensuring sufficient statistical power. As a retrospective cohort study, the sample size was determined by all eligible cases during the study period, thereby reflecting the real-world clinical population.

INCLUSION CRITERIA: All participants fulfilled the T2DM diagnosis requirements. Inclusion criteria for HHS were based on the 2024 American Diabetic Association guideline consensus: (1) plasma glucose concentration ≥33.3 mmol/L (600 mg/dL); (2) effective plasma osmolality >300 mOsm/(kg-H2O) or total plasma osmolality >320 mOsm/(kg-H2O); (3) serum HCO3– ≥15 mmol/L or arterial blood pH ≥7.30; (4) anion gap <12 mmol/L; (5) Blood β-hydroxybutyrate <3.0 mmol/L or urinary ketones <++, complications related to impaired consciousness, delirium, and/or coma [6].

EXCLUSION CRITERIA:

(1) patients diagnosed with DKA or those with both HHS and DKA; (2) patients with incomplete medical records or missing more than 20% of records, and those discharged against medical advice within 72 hours; (3) key clinical data (eg, baseline laboratory tests) were missing.

DATA COLLECTION:

Clinical data were extracted from the hospital’s electronic medical record (EMR) system using a standardized data abstraction form developed before data collection. Two trained clinicians from the Department of Endocrinology independently retrieved baseline demographic characteristics, clinical presentations, comorbidities, medication information, and laboratory parameters. Extracted variables included age, sex, precipitating factors, Charlson Comorbidity Index (CCI), and whether prophylactic LMWH was administered after admission. Comorbidities are defined according to the CCI [9]. The recorded comorbidities were as follows: coronary heart disease 43.13%, cerebrovascular disease 33.3%, heart failure 38.2%, gastrointestinal bleeding 11.76%, malignancy 4.9%, diabetic complications 44.11%, moderate-to-severe renal disease 14.7%, and others 3.9%. CCI scoring criteria included: myocardial infarction, congestive heart failure, peripheral vascular disease, dementia, cerebrovascular disease, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, mild liver disease, and diabetes without complications (1 point each); diabetes with end-organ damage, hemiplegia or paraplegia, moderate-to-severe renal disease, leukemia, and lymphoma (2 points each); moderate-to-severe liver disease (3 points); and metastatic solid tumor and AIDS (6 points each).

Laboratory test indicators included: blood glucose (BG), serum potassium (K⁺), serum sodium (Na⁺), serum chloride (Cl^–), uric acid (UA), effective plasma osmolality (EPO), C-reactive protein (CRP), D-dimer, creatinine (Cr), blood urea nitrogen (BUN), cystatin C (Cys-C), fibrinogen, fibrin degradation products (FDP), triglycerides, and low-density lipoprotein (LDL). Corrected sodium was calculated using the Al-Kudsi formula:

To ensure data accuracy and consistency, all extracted information underwent cross-checking by a third senior investigator. Discrepancies were resolved through consensus. Missing or inconsistent records were verified directly within the EMR system, and cases with >20% missing data were excluded according to predefined criteria.

STATISTICAL ANALYSIS:

Statistical analyses were performed using SPSS 26.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 10.0 (GraphPad Software, San Diego, CA, USA). Data completeness was assessed before analysis. Cases with more than 20% missing data were excluded; variables with 10–20% missing data were handled using multiple imputation by chained equations; and variables with less than 10% missing data were analyzed using complete-case analysis. Sensitivity analyses were conducted to confirm the robustness of the results. Continuous variables were examined using the Shapiro–Wilk test and presented as mean±standard deviation or median (interquartile range), depending on their distribution. Group comparisons were performed using the independent-samples t-test or Mann-Whitney U test. Categorical variables were expressed as frequencies and percentages and compared using the chi-square test or Fisher’s exact test. Variables identified as potential determinants in the univariable analysis were subsequently included in the multivariable logistic regression model to estimate odds ratios and confidence intervals. To control for potential confounding factors, the multivariable model incorporated variables with a univariable P value <0.05 as well as clinically relevant factors (eg, age, sex, presence of infection, BG, Na+, and EPO). This approach enabled the identification of independent predictors of in-hospital mortality among patients with HHS. Receiver operating characteristic (ROC) curve analysis was used to evaluate the predictive performance of key risk factors. A P value <0.05 was considered statistically significant.

Results

BASELINE CLINICAL CHARACTERISTICS OF SURVIVORS AND NON-SURVIVORS:

A total of 102 patients with HHS were included in this study, with an in-hospital mortality rate of 29.4%. Infection was the primary precipitating factor, with pulmonary infections accounting for 50% of cases, followed by urinary tract infections, pressure ulcer–related infections, and multiple infections (Table 1). Baseline analysis showed that non-survivors had significantly higher levels of Cr, BUN, Cys-C, CRP, D-dimer, and FDP, as well as higher CCI scores, compared with survivors (all P<0.05). By contrast, among patients without contraindications, the proportion receiving prophylactic LMWH after admission was significantly higher in survivors than in non-survivors (P<0.05) (Table 2).

ANALYSIS OF RISK FACTORS FOR IN-HOSPITAL MORTALITY IN PATIENTS WITH T2DM COMPLICATED BY HHS:

To identify independent predictors of in-hospital mortality, logistic regression analyses were performed. In the univariable models, Cr, BUN, Cys-C, CRP, D-dimer, and CCI score were all significantly associated with increased in-hospital mortality risk, whereas prophylactic administration of LMWH was associated with a lower risk of death (OR=0.193, 95% CI: 0.075–0.497, P=0.001). After adjusting for age, sex, and other potential confounders, the multivariable logistic regression analysis confirmed that BUN, Cys-C, D-dimer, and CCI score were independent predictors of in-hospital mortality. Prophylactic LMWH use remained an independent protective factor (Table 3). These findings support the study hypothesis, indicating that renal dysfunction, systemic inflammation, coagulation activation, and comorbidity burden play critical roles in determining outcomes among patients with isolated HHS.

ROC CURVE ANALYSIS FOR PREDICTING IN-HOSPITAL MORTALITY AMONG HHS PATIENTS:

Area Under the Curve (AUC) values from ROC curve analysis demonstrated that Cys-C (AUC=0.899) and BUN (AUC=0.851) exhibited the highest discriminative performance, followed by the CCI score (AUC=0.787) and D-dimer (AUC=0.720) (all P<0.05). Prophylactic LMWH use showed a lower but still statistically significant discriminative ability (AUC=0.693). These findings indicate that Cys-C and BUN, in particular, possess strong predictive value and may serve as potential biomarkers for identifying patients with isolated HHS at increased risk of in-hospital mortality (all P<0.05) (Table 4; Figure 2).

Discussion

LIMITATIONS:

First, this study was a single-center retrospective analysis with a limited sample size, which may introduce selection bias and restrict the external validity of the findings; therefore, large-scale, multicenter randomized controlled trials are needed to further validate these results. Second, several clinically relevant variables (such as HbA1c, body mass index, lifestyle factors, and medical history) were incomplete and thus not included in the analysis. Future prospective studies employing standardized data collection protocols are warranted to incorporate these variables and assess their impact on mortality risk in patients with HHS. Third, this study examined only in-hospital mortality among patients with HHS; whether BUN, Cys-C, D-dimer, and CCI scores are associated with post-discharge mortality remains unclear and requires clarification through future follow-up data. Fourth, as the study recorded only whether patients received LMWH – without detailed documentation of dosage, administration frequency, or indications – future research should focus on determining the optimal dosing strategies, administration regimens, and clinical criteria for prophylactic use across different risk strata to better define its role in reducing thrombosis and improving clinical outcomes. Finally, the multivariable analysis in this study included variables based on a screening threshold of P<0.05, which may introduce model bias. Therefore, these results should be interpreted in conjunction with the clinical context and biological plausibility.

Conclusions

This study demonstrates that elevated BUN, Cys-C, and D-dimer levels, together with higher CCI scores, constitute the principal determinants of in-hospital mortality in isolated HHS. These findings indicate that outcomes are driven less by the degree of hyperglycemia itself and more by the extent of dehydration-induced end-organ injury and hyperosmolarity-related vascular compromise. The integrative pathophysiological profile also supports the protective role of prophylactic LMWH and reinforces aggressive fluid resuscitation and early anticoagulation as essential components of optimal HHS management.