In-Hospital Mortality and Predictors Among People with Diabetes in Lubumbashi, Democratic Republic of the Congo: A Prospective Observational Study

Introduction

Diabetes mellitus (DM) is a chronic metabolic disorder characterised by persistent hyperglycemia due to defects in insulin secretion, insulin action, or both. It is one of the fastest-growing non-communicable diseases worldwide, with major implications for morbidity, mortality, and healthcare systems. According to the International Diabetes Federation,¹ 589 million adults aged 20–79 were living with DM in 2024, including 9.5 million with type 1 DM, of whom 1.9 million were children and adolescents. This number is projected to reach 853 million by 2050. Globally, 43% of people with DM remain undiagnosed, and over 3.4 million deaths were attributed to DM in 2024.¹ In addition, hyperglycemia in pregnancy affects approximately one in five pregnancies. The burden of DM is particularly high in low- and middle-income countries, where over 80% of adults living with the disease reside, and limited access to diagnostic and treatment services often results in late presentation and increased complications.^2,3

In sub-Saharan Africa, the prevalence of DM has increased dramatically over the past two decades. In 2021, the region was home to 24 million adults with diabetes, a figure expected to more than double by 2045.⁴ Despite this growing epidemic, many cases remain undiagnosed, and access to diagnostic and therapeutic services is often limited—consequently, a large proportion of patients present with advanced disease or acute metabolic complications upon hospital admission. Diabetes-related deaths in Africa were estimated at over 216,000 in 2024, and most occurred among individuals under 60 years of age.¹ Several hospital-based studies in sub-Saharan Africa have identified key predictors of in-hospital mortality among people with diabetes, including severe infections such as sepsis, acute metabolic complications (diabetic ketoacidosis and hyperosmolar hyperglycemic state), cardiovascular and neurological events such as stroke and heart failure, as well as chronic comorbidities, particularly chronic kidney disease and poor long-term glycemic control.^5–7

In the Democratic Republic of the Congo (DRC), the prevalence of DM varies widely—from 5.3% to 15.6% depending on the population and study setting.^8–10 The main risk factors for DM included male sex, age ≥40 years, both general and abdominal obesity, family history of diabetes, and comorbid hypertension.⁵ Despite the increasing burden of DM in the DRC, hospital-based data on diabetes-related mortality and its determinants remain limited. Previous studies conducted in DRC have documented a substantial burden of DM complications, including cardiovascular events, chronic kidney disease, and diabetic foot ulcers. For instance, hospital surveys in Kinshasa reported high rates of hyperglycemia, hypertension, and diabetic foot, with mortality rates ranging from 0% to 17.3% depending on the hospital setting.^11,12 More recent analyses using routine health data have confirmed that DM contributes significantly to adult mortality, accounting for approximately 5.4% of deaths in urban areas, with higher proportional mortality observed among women and older adults.^13,14

Community mortality is often influenced by delayed diagnosis, limited access to healthcare, and socioeconomic barriers, whereas in-hospital mortality is more directly related to the severity of acute complications and the quality of clinical management during hospitalisation. Identifying predictors of in-hospital mortality is therefore essential to improve early risk stratification and optimise hospital care in resource-limited settings. These findings highlight the urgent need for robust evidence on predictors of in-hospital mortality to guide clinical management and public health interventions in the DRC. Understanding the magnitude, associated factors, and determinants of in-hospital mortality among people with diabetes is essential for improving prevention strategies, optimising case management, and guiding policy interventions aimed at reducing premature deaths due to diabetes. Therefore, this study aimed to evaluate in-hospital mortality and identify its associated risk factors among patients with DM admitted to Lubumbashi, DRC.

Materials and Methods

Study Area and Period

The study was conducted in Lubumbashi, the second-largest city of the DRC, located in the southeastern part of the country in the Haut-Katanga Province. Lubumbashi is a major economic and industrial hub, with a population of approximately 2.5 million inhabitants, and serves as a regional centre for health, education, and commerce. The study was carried out across three health facilities: Cliniques Universitaires de Lubumbashi, Centre Médical du Centre Ville (CMDC), and MedPark Clinic, all of which provide specialised care for diabetes, including management of medical and surgical complications. Data were collected over 20 months, from January 1, 2022, to August 31, 2023.

Study Design and Population

A hospital-based prospective cohort study was conducted. The source population included all adult patients (≥18 years) with a confirmed diagnosis of diabetes mellitus (type 1 or type 2) admitted to the internal medicine wards of the three selected hospitals during the study period. The study population included all patients from the source population who met the eligibility criteria and for whom complete clinical and laboratory data were available or who provided informed consent for prospective data collection. Patients were followed from hospital admission until discharge, referral, or death, allowing assessment of in-hospital outcomes, including mortality and associated risk factors. This design enabled the collection of comprehensive clinical, sociodemographic, and laboratory data to identify predictors of in-hospital mortality among hospitalised people with diabetes in Lubumbashi.

Eligibility Criteria

Adult patients aged 18 years and above who were hospitalised in the internal medicine wards with a confirmed diagnosis of DM and who stayed at least 24 hours in the hospital were eligible for inclusion. Both patients with complete medical records and those who provided informed consent for prospective data collection were considered.

Patients were excluded if they were not hospitalised, had incomplete medical records, refused to participate, or were unable to provide the necessary information (or their caregivers were unable to do so). Additionally, individuals with secondary diabetes, such as corticosteroid-induced diabetes, DM related to pancreatitis, or gestational DM, were excluded to ensure a homogeneous study population focused on primary DM.

Sample Size Determination and Sampling Technique

The minimum sample size was calculated using Schwartz’s formula:

where P represents the expected prevalence of DM (15.6%),¹⁰ Z is the standard normal value at a 95% confidence interval (1.96), and d is the margin of error (5%). Based on these assumptions, the minimum required sample size was 225 patients.

A probabilistic exhaustive sampling method was employed, whereby all eligible patients admitted to the internal medicine wards of the selected hospitals during the study period were included in the study. This approach ensured comprehensive coverage of the target population and enhanced the reliability of the findings.

Study Variables

The primary outcome variable of this study was in-hospital mortality among patients with DM. Several predictor variables were considered to identify factors associated with mortality. Sociodemographic factors included age, sex, marital status, occupation, and educational level. Clinical factors encompassed the type and duration of DM, comorbidities, diabetes-related complications, admission complaints, anthropometric measurements (weight, height, and body mass index [BMI]), and blood pressure (systolic and diastolic).

Laboratory investigations included fasting plasma glucose, glycated haemoglobin (HbA1c), C-peptide (when available), and other laboratory tests relevant to DM complications. Medication-related variables comprised the type and duration of both antidiabetic and non-antidiabetic treatments. Clinical diagnoses were established by the treating physicians based on clinical presentation and available laboratory findings according to standard diagnostic criteria. Sepsis was defined as a life-threatening organ dysfunction caused by infection, consistent with the Sepsis-3 definition.¹⁵ Hyperosmolar hyperglycemic state was defined by severe hyperglycemia (plasma glucose >600 mg/dL [33.3 mmol/L]), hyperosmolality (effective osmolality >300 mOsm/kg or total serum osmolality >320 mOsm/kg), and absence of significant ketoacidosis, defined as β-hydroxybutyrate <3.0 mmol/L (or urine ketones <2+), with pH ≥7.3 and bicarbonate ≥15 mmol/L.¹⁶ Meningoencephalitis was diagnosed based on compatible neurological symptoms (fever, altered mental status, or seizures) in combination with cerebrospinal fluid findings or neuroimaging when available.¹⁷ Chronic kidney disease was defined as kidney damage or an estimated glomerular filtration rate (eGFR) below 60 mL/min/1.73 m² for at least three months, regardless of the underlying cause.¹⁸ Diabetic nephropathy was classified based on albuminuria levels: microalbuminuria (urinary albumin excretion 20–199 μg/min) and macroalbuminuria (urinary albumin excretion ≥200 μg/min).¹⁹ Diabetic neuropathy was identified by the presence of clinical signs or symptoms of peripheral or autonomic nerve dysfunction, after excluding alternative etiologies.²⁰ Medication adherence was evaluated using patient self-report at admission and review of hospital records, including prior prescription refills and any documented interruptions; self-report is widely used in chronic disease studies and can be supplemented by objective refill data when available.²¹ Collectively, these variables enabled a comprehensive assessment of patient characteristics, disease profile, and management factors to determine predictors of in-hospital mortality among hospitalised people with diabetes.

Data Collection Procedures and Quality Control

As a prospective cohort study, data were collected using a structured data collection form designed to capture sociodemographic characteristics, clinical information, laboratory results, and treatment details throughout patients’ hospital stay. The tool was initially developed in English and subsequently reviewed and validated by a panel of endocrinologists to ensure its relevance, clarity, and completeness. To account for the multilingual context of Lubumbashi, the form was translated into the local language and back-translated into English to verify consistency.

Data were obtained prospectively through daily follow-up of hospitalised patients, supplemented by review of medical records and, when necessary, direct interviews with patients or caregivers to collect missing or clarifying information. Two trained pharmacists and one nurse were responsible for data collection under the supervision of a medical doctor to ensure adherence to study protocols and accuracy of recorded information.

To ensure data quality, all data collectors and the supervisor received one day of intensive training on the data collection procedures, ethical considerations, and proper extraction and recording of clinical information. The principal investigator performed daily checks of all completed forms, and a pretest was conducted on 5% of the study population to assess the clarity, consistency, and validity of the tool. All procedures adhered strictly to standard ethical guidelines.

Data Processing and Analysis

Data were entered into a database and analysed using Stata version 16. Descriptive statistics were reported as frequencies and percentages for categorical variables, and as mean ± standard deviation (SD) for continuous variables, depending on distribution (Shapiro–Wilk test). The in-hospital mortality rate was calculated.

Bivariate analyses were performed using the Chi-square test for categorical variables and appropriate tests for continuous variables. Variables with a p-value < 0.2 in the bivariate analysis or considered clinically relevant were included in a multivariable Cox proportional hazards regression model to identify independent predictors of in-hospital mortality. To reduce the risk of model overfitting, given the limited number of events, the number of predictors retained in the multivariable model was restricted in accordance with commonly recommended events-per-variable considerations. Adjusted hazard ratios (aHRs) with 95% confidence intervals (95% CIs) were reported, and statistical significance was set at p < 0.05. The proportional hazards assumption was assessed using Schoenfeld residuals. Survival probabilities over 60 days were estimated using Kaplan–Meier curves, and differences between subgroups were assessed using the Log rank test.

Ethical Considerations

The study was conducted in strict accordance with the ethical research regulations of the Democratic Republic of the Congo and adhered fully to the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants or their legal guardians before data collection. The study protocol was reviewed and approved by the Medical Ethics Committee of the University of Lubumbashi (Approval No. UNILU/CEM/032/2022). The authors declare no conflicts of interest and no financial relationships that could have influenced the study.

Results

Overview of Study Participants

During the 20-month study period, a total of 341 consecutive people with diabetes were admitted. Seventeen patients were excluded due to not meeting the eligibility criteria, leaving 324 patients for the final analysis. Among these, 37 patients (11.42%) had type 1 DM (T1DM), while 287 patients (88.58%) had type 2 DM (T2DM) (Figure 1).

Figure 1 Flow diagram illustrating patient inclusion, exclusions, and follow-up in admitted people with diabetes mellitus.

Sociodemographic and Clinical Characteristics of Study Participants

A total of 324 people with diabetes were included in the study (Table 1). The mean age was 56.3 ± 14.4 years, with a balanced sex distribution (164 males [50.6%] and 160 females [49.4%]). Over half of the patients (51.9%) had no formal education, while 22.5% had attained higher or university-level education. Most patients were married (61.1%), and the majority were retired, unemployed, or inactive (56.8%). Smoking and alcohol consumption were reported in 3.1% and 13.3% of patients, respectively. Familial history of diabetes was present in 19.4% of patients.

Table 1 Sociodemographic and Clinical Characteristics of Hospitalised People with Diabetes (n = 324), Lubumbashi, DRC

At admission, 32.1% of patients were newly diagnosed. Among previously diagnosed patients, 29.6% reported good adherence to treatment, while 38.3% were non-adherent. The duration since diagnosis varied, with 12.7% diagnosed less than one year ago and 14.5% living with diabetes for 10 years or more.

Clinically, the mean BMI was 26.1 ± 5.6 kg/m². Patients presented with a mean random blood glucose of 316.0 ± 152.4 mg/dL and a mean HbA1c of 9.5 ± 2.8%. The average length of hospital stay was 12.5 ± 5.4 days. Regarding in-hospital outcomes, 41 patients (12.7%) died, while 283 patients (87.3%) survived.

Comorbidities and Metabolic Complications Among Hospitalised People with Diabetes

Among the 324 hospitalised people with diabetes, 55.9% presented at least one comorbidity. Hypertension was the most frequent condition (41.4%), followed by infectious diseases such as pulmonary infections (excluding tuberculosis) (20.4%), urinary tract infection (19.8%), and malaria (14.8%). Metabolic complications, including hyperosmolar state (8.3%) and diabetic ketoacidosis (10.5%), were also recorded. Less common comorbidities included stroke (3.4%), deep vein thrombosis (1.2%), and various malignancies (1.8%). Regarding nutritional status, one-third of patients were overweight (33.0%), and 21.3% were obese (Table 2).

Table 2 Comorbidities Among Hospitalised People with Diabetes (n = 324), Lubumbashi, DRC

Prevalence of Long-Term Diabetic Complications Among Hospitalised Patients

Overall, 23.1% of hospitalised people with diabetes presented at least one long-term complication, with nephropathy being the most frequent (34.7%), followed by retinopathy (31.6%), diabetic foot ulcer (29.5%), and neuropathy (25.3%) (Table 3).

Table 3 Prevalence of Long-Term Diabetic Complications Among Hospitalised People with Diabetes Mellitus (n = 324), Lubumbashi, DRC

In-Hospital Mortality

In this study, 41 of the 324 hospitalised people with diabetes died, resulting in an in-hospital mortality rate of 12.7%. Among the 41 patients who died during hospitalisation, several comorbidities were frequently observed. Infectious conditions were the most common, including sepsis of unspecified origin (13/41, 31.7%), pulmonary infections (excluding tuberculosis) (11/41, 26.8%), gastrointestinal infections (3/41, 7.3%), and meningoencephalitis (3/41, 7.3%). Chronic kidney disease was also present in 3 patients (7.3%). Among the 41 patients who died during hospitalisation, the mean time to death was 4.9 ± 4.1 days (range: 1–15 days). When considering the entire cohort (N = 324) and accounting for censored observations, the restricted mean survival time was 53.0 ± 1.0 days (95% CI: 51.0–55.0).

Predictors of in-Hospital Mortality

Table 4 summarises the results of the bivariable and multivariable Cox proportional hazards regression model used to identify predictors of in-hospital mortality among people with diabetes. After adjustment for potential confounders, increasing age remained a significant predictor of mortality (aHR = 1.05, 95% CI 1.03–1.09, p < 0.001), indicating that each additional year of age increased the risk of in-hospital death by 5%. Several clinical and infectious comorbidities were independently linked to in-hospital death. Patients with stroke had a markedly higher risk of mortality (aHR = 4.53, 95% CI 1.64–12.55, p = 0.004), as did those with meningoencephalitis (aHR = 5.84, 95% CI 1.60–21.24, p = 0.007), chronic kidney disease (aHR = 5.20, 95% CI 1.43–18.89, p = 0.012), and sepsis (aHR = 5.85, 95% CI 2.57–13.34, p < 0.001). Moreover, the presence of a hyperosmolar hyperglycemic state was also significantly associated with increased mortality (aHR = 3.15, 95% CI 1.16–8.60, p = 0.025). Metabolic control also played an important role. Higher glycated haemoglobin (HbA1c) levels were significantly linked to mortality (aHR = 1.16, 95% CI 1.04–1.31, p = 0.011), suggesting that poor long-term glycemic control adversely affected hospital outcomes. In contrast, admission glycemia was not associated with mortality (p = 0.818). In contrast, other variables such as educational level and peripheral artery disease were not significantly associated with mortality after adjustment (p > 0.05). The overall model was highly significant (likelihood ratio chi² = 70.72, p < 0.001), indicating a good fit for identifying key predictors of mortality among hospitalised people with diabetes in Lubumbashi.

Table 4 Bivariable and Multivariable Cox Proportional Hazards Regression to Identify Predictors of in-Hospital Mortality Among People with Diabetes (n = 324)

Kaplan–Meier Survival Curves by Comorbidity Status

The probability of in-hospital death was markedly higher among people with diabetes with a history of stroke compared to those without. This difference resulted in a substantial reduction in overall survival among patients with stroke, with a statistically significant difference between the two groups (Log rank test, p = 0.0007) (Figure 2).

Figure 2 Kaplan–Meier curves showing 60-day overall survival among hospitalised people with diabetes according to stroke status.

The probability of in-hospital death was substantially higher among people with diabetes presenting with a hyperosmolar state compared to those without this complication. This difference translated into a significantly lower overall survival in patients with hyperosmolar state (Log rank test, p = 0.0006) (Figure 3).

Figure 3 Kaplan–Meier curves showing 60-day overall survival among hospitalised people with diabetes according to hyperosmolar state status.

Hospitalised people with diabetes who developed meningoencephalitis exhibited a markedly higher probability of in-hospital death compared with those without the infection. This difference resulted in a significantly lower overall survival among affected patients (Log rank test, p < 0.0001) (Figure 4).

Figure 4 Kaplan–Meier curves showing 60-day overall survival among hospitalised people with diabetes according to meningoencephalitis status.

People with diabetes who developed sepsis had a substantially higher risk of in-hospital death compared with those without sepsis. This difference led to a significantly lower overall survival in patients with sepsis (Log rank test, p < 0.001) (Figure 5).

Figure 5 Kaplan–Meier curves showing 60-day overall survival among hospitalised people with diabetes according to sepsis status.

People with diabetes with pre-existing chronic kidney disease tended to have a higher risk of in-hospital death compared with those without chronic kidney disease. However, this difference in overall survival did not reach statistical significance (Log rank test, p = 0.0791) (Figure 6).

Figure 6 Kaplan–Meier curves showing 60-day overall survival among hospitalised people with diabetes according to chronic kidney disease status.

Discussion

This hospital-based prospective study assessed in-hospital mortality and its predictors among 324 people with diabetes admitted to three hospitals in Lubumbashi, DRC. The overall in-hospital mortality rate was 12.7%, which aligns with findings from similar sub-Saharan African settings. In Ethiopia, three studies reported comparable in-hospital mortality levels among people with diabetes. A prospective observational study in Jimma found a mortality rate of 13.34%,⁶ while a study conducted in Addis Ababa reported 10.6% mortality among diabetic admissions.²² More recently, a prospective study from southern Ethiopia documented an in-hospital mortality of 12.24%, with infections being the leading cause of death.²³ Likewise, a study from Uganda reported an in-hospital mortality rate of 10.8%,²⁴ and a 10-year Nigerian hospital survey found 8.3% mortality among hospitalised people with diabetes.²⁵ Taken together, these findings indicate that the in-hospital mortality observed in our cohort is consistent with the burden reported across sub-Saharan Africa, where mortality among hospitalised people with diabetes typically ranges between 7.2% and over 12%.^6,26,27 The similarities across diverse settings—Ethiopia, Uganda, Nigeria, and now the DRC—suggest a common pattern characterised by late presentation, high prevalence of acute metabolic complications, and limited availability of specialised diabetes care. These concordant mortality levels also highlight systemic challenges such as inadequate outpatient follow-up, delayed recognition of complications, and the heavy contribution of infections and cardiovascular events to early hospital deaths. Understanding this regional context underscores the importance of identifying locally relevant predictors of mortality to inform targeted interventions and strengthen inpatient diabetes management in resource-limited settings.

In our cohort, there was no significant difference in in-hospital mortality between newly diagnosed patients and those with known diabetes, a finding that contrasts with earlier assumptions that newly diagnosed individuals or clinic defaulters would be at the highest risk of death. International evidence presents a more complex picture. For example, a large multicenter study from South Korea involving more than 33,000 hospitalised adults reported that patients with newly diagnosed diabetes had substantially higher in-hospital mortality compared with both non-diabetic and previously diagnosed people with diabetes (HR 1.89; 95% CI 1.58–2.26), largely attributed to unrecognised hyperglycemia and severe metabolic derangements before admission.²⁸ Our findings from Lubumbashi differ from this pattern, and several factors may explain this discrepancy. Newly diagnosed patients in our setting often presented with fewer advanced chronic complications, whereas many known diabetics had longstanding disease with accumulated comorbidities, such as chronic kidney disease, stroke, sepsis, or hyperosmolar states, that were independently associated with mortality in our Cox model. Early initiation of inpatient treatment at the time of diagnosis may also have reduced the risk of acute deterioration among newly diagnosed individuals. Additionally, differences in health-system context, including patterns of care-seeking, availability of screening, and sociocultural perceptions of symptoms, may shape the clinical profile at admission. Taken together, although large international studies suggest that newly diagnosed diabetes may predict worse in-hospital outcomes,²⁸ our results indicate that this association is not universal and may vary across healthcare systems. This reinforces the need for context-specific evidence to better understand mortality determinants among hospitalised people with diabetes in low-resource settings.

Our findings underscore the prognostic importance of long-term glycemic control in determining in-hospital outcomes. In our cohort, higher HbA1c levels were independently associated with increased mortality (aHR = 1.16; 95% CI, 1.04–1.31), indicating that patients with chronically poor metabolic control experienced worse survival during hospitalisation. This aligns with robust evidence from high-income settings. In a national US cohort, Raghavan et al demonstrated a clear dose–response relationship between rising HbA1c and both cardiovascular and all-cause mortality, with hazard ratios increasing steadily above 7%.²⁹ Similar associations have been confirmed in large primary care and hospital-based cohorts. In a study of over 85,000 patients with diabetes in the UK, long-term elevated HbA1c was strongly associated with higher rates of severe infections, hospitalisation, and infection-related mortality, with risks rising progressively across HbA1c categories.³⁰ Comparable findings were observed in cardiovascular populations: in the PROMISE cohort, patients with coronary heart disease and type 2 diabetes with HbA1c ≥7% had significantly higher all-cause and cardiac mortality, as well as increased major adverse cardiac and cerebrovascular events.³¹ Likewise, among hemodialysis patients, HbA1c levels above 8.5% were associated with a marked increase in all-cause mortality, with cardiovascular deaths rising more sharply than other causes.³² Several biological pathways may explain these relationships. Chronic hyperglycemia impairs neutrophil chemotaxis, phagocytosis, and cellular immunity, thereby increasing susceptibility to infections. It also disrupts endothelial integrity and microvascular perfusion, contributing to vascular dysfunction, while accelerating atherosclerosis and myocardial injury.^33,34 These mechanisms delay tissue repair and increase susceptibility to severe infections and cardiovascular events, major contributors to mortality among hospitalised or high-risk people with diabetes. In sub-Saharan Africa, poor glycemic control is often compounded by systemic barriers, including irregular follow-up, delayed diagnosis, limited access to medications, gaps in diabetes self-management education, and frequent use of traditional or herbal remedies alongside biomedical treatment. Patients also rarely self-monitor glucose, have low physical activity, partially adhere to diet and medication recommendations, and have limited knowledge of diabetes complications.^35,36 Additional associated factors include age extremes, gender, low income, lack of insurance, low education, rural residence, family history, longer diabetes duration, treatment complexity, side effects, alcohol use, smoking, comorbidities, and suboptimal management. Conversely, strong family support, effective coping strategies, high diabetes literacy, adherence to lifestyle and medication regimens, and regular follow-up are linked to better glycemic control.^35,36

In addition to elevated HbA1c, our study identified several independent predictors of in-hospital mortality among people with diabetes, including advanced age, stroke, meningoencephalitis, sepsis, chronic kidney disease, and hyperosmolar hyperglycemic state. Each additional year of age at diabetes diagnosis was associated with a progressively higher risk of adverse outcomes, including heart disease, stroke, disability, cognitive impairment, and mortality, reflecting the cumulative burden of comorbidities, frailty, and reduced physiological reserve in older adults.³⁷ In hospitalised people with diabetes, acute complications such as diabetic ketoacidosis, as well as the presence of multiple comorbidities, substantially increase in-hospital mortality, with rural residence and advanced age further exacerbating risk.⁶ These findings highlight the dual vulnerability of older people with diabetes, who are at heightened risk both from acute metabolic crises and from the long-term accumulation of chronic conditions, underscoring the need for proactive management strategies targeting both preventive and acute care.

Infectious complications, particularly sepsis and meningoencephalitis, represent some of the most potent predictors of in-hospital mortality among people with diabetes. Evidence from a cohort study in Western Saudi Arabia demonstrated that diabetic individuals with bacteremia or septicemia exhibited a markedly high mortality rate, especially when advanced age, comorbidities, septic shock, or multi-organ dysfunction were present.³⁸ These findings highlight the pronounced susceptibility of people with diabetes to severe infections and multi-organ failure, underscoring the need for early recognition and prompt empiric antimicrobial therapy in high-risk populations. Concurrently, acute cerebrovascular events such as stroke significantly worsen prognosis in people with diabetes, emphasising the critical influence of neurological complications on survival. These observations are corroborated by large-scale evidence demonstrating that diabetes substantially exacerbates both short- and long-term outcomes following stroke. In a nationwide cohort of 608,890 hospitalised stroke patients, diabetes was independently associated with higher in-hospital mortality, increased odds of sepsis and other complications, as well as elevated long-term mortality and stroke recurrence, irrespective of stroke subtype.³⁹ These findings highlight the profound impact of diabetes on stroke-related outcomes and the importance of vigilant monitoring and management in this high-risk population.

Hyperosmolar hyperglycaemic states are associated with markedly increased mortality (aHR = 3.15; 95% CI 1.16–8.60). Reported inpatient mortality ranges from 10 to 15%, considerably higher than that observed in diabetic ketoacidosis, and HHS is frequently precipitated by infections, comorbidities, or other acute stressors.^40–42 Current clinical guidelines stress the importance of early recognition, prompt fluid resuscitation, meticulous electrolyte management, and rapid treatment of underlying triggers to reduce mortality and mitigate neurological complications.⁴⁰

Chronic kidney disease emerged as a strong predictor of in-hospital mortality among people with diabetes, increasing the risk approximately fivefold (aHR = 5.20; 95% CI 1.43–18.89). This is consistent with evidence from both sub-Saharan Africa and high-income settings, where chronic kidney disease substantially amplifies the risk of cardiovascular events, infections, and poor metabolic tolerance during acute illness.^43–45 In individuals with type 2 diabetes, the coexistence of chronic kidney disease dramatically increases all-cause mortality, with one cohort study reporting a 10-year cumulative mortality of 31% in patients with both conditions, compared to 11.5% in diabetics without kidney disease.⁴⁴ Mechanistic and clinical reviews highlight the bidirectional interplay between diabetes and kidney disease, underscoring the need for meticulous glycaemic control, blood pressure management, and the use of reno-protective pharmacotherapies, including sodium–glucose co-transporter 2 (SGLT2) inhibitors and nonsteroidal mineralocorticoid receptor antagonists, to improve outcomes.^43,46,47 Collectively, these data reinforce that chronic complications, such as chronic kidney disease, alongside acute metabolic decompensations (eg., hyperosmolar hyperglycaemic states, severe infections), are major drivers of hospital mortality among diabetics.

Notably, some predictors identified in our cohort represent baseline vulnerability, including advanced age, chronic kidney disease, and elevated HbA1c, whereas others, such as sepsis, hyperosmolar hyperglycemic state, stroke, and meningoencephalitis, reflect acute severity at admission. Distinguishing these categories clarifies that certain associations indicate chronic susceptibility, while others capture immediate life-threatening complications. These observations are consistent with broader evidence showing that both chronic comorbidities and acute clinical conditions independently contribute to in-hospital mortality among diabetes patients and other high-risk hospitalised populations.⁴⁸

Early recognition and prompt management of acute metabolic derangements, alongside systematic screening and management of chronic complications, are therefore essential strategies to reduce in-hospital mortality. Special attention should be given to older adults and patients with pre-existing organ damage. Strengthening hospital-based care, including rapid access to dialysis for chronic kidney disease and prompt management of hyperosmolar hyperglycemic states, has the potential to substantially improve outcomes for people with diabetes facing both chronic and acute challenges, particularly in resource-limited settings such as Lubumbashi.

The study’s prospective design and thorough follow-up from admission to discharge or death are key strengths, providing reliable insight into in-hospital mortality patterns. However, the analysis is limited to hospitalised patients and may not reflect outcomes in the outpatient setting. Additionally, some comorbidities may have been underreported due to resource limitations in diagnostic testing, which could lead to an underestimation of their true impact on mortality. Another limitation is the relatively limited number of deaths observed during follow-up (41 events), which may affect the statistical stability of the multivariable Cox regression model. Nevertheless, the prospective design and the consistency of the identified predictors with previous studies support the credibility of the findings.

Conclusion

This study demonstrates a high in-hospital mortality rate of 12.7% among people with diabetes, with deaths occurring early during hospitalisation. Independent predictors of mortality included advanced age, chronic kidney disease, poor long-term glycemic control (baseline vulnerability), as well as acute cerebrovascular events, sepsis, meningoencephalitis, and hyperosmolar hyperglycaemic states (acute severity at admission). These findings highlight the importance of early recognition and management of both chronic comorbidities and acute complications. Strengthening hospital-based interventions and prioritising high-risk populations may substantially improve survival outcomes in this vulnerable patient group. Given the relatively small number of deaths (n = 41), caution is warranted regarding the stability of the multivariable model; however, the prospective design and consistency with previous studies support the credibility of the findings.