Predictors of hyperosmolar hyperglycemic state among adult type 1 and type 2 diabetes mellitus patients at public hospitals in Harari region,Eastern Ethiopia: A survival analysis

Abstract

Background

The hyperosmolar hyperglycemic state (HHS) is a rare but life-threatening complication of diabetes. Evidence on hyperglycemic emergencies, including HHS and their predictors in Ethiopia, is limited, and predictors specific to HHS are not well characterized. Therefore, this study aimed to determine the incidence of hyperosmolar hyperglycemic state and its predictors among adult diabetic patients in eastern Ethiopia.

Method

A retrospective cohort study was conducted among 461 adults with diabetes from January 1, 2019, to December 31, 2024. Data were collected using the Kobo Toolbox and analyzed with STATA. Survival probabilities were estimated using the Kaplan-Meier curve. Predictors of HHS were identified using the Weibull proportional hazards regression model. Hazard ratios (HR) with their respective 95% confidence intervals (CI) and p-values were computed.

Results

From the total of 452 patients included in the study, 45 (9.96%), 95% CI (7.51%-13.08%) developed hyperosmolar hyperglycemic state. The incidence rate of HHS was 5.36 cases per 100 person-years of observation, 95% CI (4.00-7.18), which was 3.89 and 5.76 per 100 person-years among T1DM and T2DM, respectively. Factors significantly associated with HHS included being rural area residents (AHR: 1.99, 95% CI: 1.01–3.92), having comorbidities (AHR: 3.01, 95% CI: 1.28–7.05), medication non-adherence (AHR: 2.14, 95% CI: 1.04–4.40), and lack of community health insurance (AHR: 2.04, 95% CI: 1.03–4.04).

Conclusion

HHS occurred more frequently among patients with T2DM. Strengthening follow-up and targeted interventions, particularly for rural residents, patients with comorbidities, those without community health insurance, and those with medication non-adherence, may reduce HHS occurrence.

Keywords

hyperosmolar hyperglycemic state predictors diabetes mellitus Ethiopia

Introduction

Diabetes mellitus (DM) is a chronic metabolic disorder characterized by persistent hyperglycemia.¹ As the global prevalence of diabetes continues to rise, particularly in low- and middle-income countries, the burden of its acute and chronic complications has become a significant public health concern.² Among the acute complications, the hyperosmolar hyperglycemic state (HHS) stands out as a rare but life-threatening emergency, predominantly affecting individuals with type 2 diabetes mellitus (T2DM).^1,3 HHS is distinguished by extreme hyperglycemia, marked hyperosmolarity, and dehydration, often without significant ketoacidosis.⁴ If not promptly recognized and managed, HHS carries a high risk of morbidity and mortality, with reported case fatality rates ranging from 10% to 20% in various settings.^4,5

The pathophysiology of HHS involves a complex interplay of insulin deficiency and increased counter-regulatory hormones, which lead to profound hyperglycemia and osmotic diuresis.³ Despite advances in diabetes care, HHS remains a diagnostic and therapeutic challenge, especially in resource-limited settings where delayed presentation and limited access to care may worsen outcomes.⁶

Globally, the burden of hyperglycemic emergencies is shifting toward low- and middle-income countries (LMICs). In Sub-Saharan Africa and South Asia, the incidence of hyperglycemic emergencies is estimated to account for up to 30% of diabetes-related hospital admissions.⁷ The hyperosmolar hyperglycemic state (HHS), in particular, has a significantly higher case fatality rate (10–50%) compared to diabetic ketoacidosis (2–5%) due to delayed clinical recognition and limited diagnostic infrastructure.^8,9

A Danish nationwide cohort study found an incidence of HHS of 3.9 per 10,000 person-years in type 2 diabetes mellitus and 16.5 per 10,000 person-years in type 1 diabetes, noting that one-third of HHS cases were newly diagnosed diabetes.¹⁰ Conversely, a previous study in Ethiopia reported an incidence of HHS of 0.9 per 100 person-years in type 1 diabetes and 2.4 per 100 person-years in type 2 diabetes.¹¹ Studies conducted in different parts of the world have identified several predictors of HHS, including infections, poor glycemic control (high HbA1c), use of glucocorticoids and benzodiazepines, dehydration, acute illnesses, medication non-compliance, and comorbidities.^10–13 Moreover, HHS often overlaps with DKA(14), yet it disproportionately affects adults with type 2 diabetes and carries a higher mortality risk.^14,15

Most studies on HHS have been conducted in high-income countries, and their findings may not be directly applicable to the Ethiopian context due to differences in healthcare infrastructure, patient demographics, and disease patterns. Even the few available studies on all hyperglycemic emergencies often lack regional specificity or comprehensive analysis. Understanding the local epidemiology of HHS is crucial for developing targeted prevention and management strategies.³

Given these gaps, this study was designed to determine the incidence of hyperosmolar hyperglycemic state and to identify its predictors among adult diabetic patients in public hospitals in the Harari region, eastern Ethiopia. By elucidating the factors associated with the development of HHS, the findings aim to inform clinical practice and health policy, ultimately contributing to improved outcomes for diabetic patients in the region.

Methods and materials

Study design, setting, and period

A retrospective cohort study was carried out among adult diabetes mellitus patients receiving follow-up care from January 01, 2019 to December 31, 2024 (cohort follow-up period) and data were abstracted retrospectively from medical records during January 01 to January 31, 2025, in public hospitals located in the Harari region, which is situated in the Harari regional state, approximately 526 kilometers from Ethiopia’s capital, Addis Ababa. The Harari region has an estimated population of 283,000, with a male-to-female ratio of 1:1.¹⁶ The region is served by two public, two private, one police, and one non-government hospital. Additionally, there are 9 health centers, 29 private clinics, 26 health posts, and one regional laboratory serving the community. The study was conducted in two public hospitals within the region: Hiwot Fana Comprehensive Specialized University Hospital (HFCSUH) and Jugal General Hospital (JGH). HFCSUH, a large teaching hospital with 210 beds, serves as a major referral center for approximately 5.8 million people in eastern Ethiopia.¹⁷ Jugal General Hospital, on the other hand, has 95 beds and primarily serves the population of Harar town. Currently, around 1,500 DM patients visit these two hospitals for regular follow-up care, with 852 patients at HFCSUH and 648 at JGH.¹⁸

Populations

All adult DM patients on chronic follow-up for treatment at public hospitals in the Harari region were considered as the source population. The study population consisted of all adult DM patients on chronic follow-up between January 01, 2019, and December 31, 2024.

Eligibility criteria

Adult DM patients enrolled in the chronic follow-up care at public hospitals in the Harari region, who were diagnosed between January 1, 2019, and December 31, 2024, were included in the study. However, patients who developed HHS at the time of their initial diabetes diagnosis and those who were transferred in from other facilities were excluded.

Sample size and sampling procedures

The final sample size of 461 was determined using the Schoenfeld formula via Stata software version 17, employing power analysis for the Cox proportional hazard model.¹⁹ This was calculated by considering a 17.7% probability of an event (cumulative incidence of HHS) and an AHR of 0.5 from a previous study.¹⁰ We assumed a 95% confidence interval, 80% power, and 20% of withdrawal probability.

All public hospitals in the Harari region, namely Hiwot Fana Comprehensive Specialized University Hospital and Jugal General Hospital, were selected. Before sample allocation, the number of newly diagnosed DM patients who visited both hospitals for follow-up care within the study period was determined.

The total sample size of 461 was distributed proportionally between the two hospitals based on the number of diabetic patients receiving follow-up care at each facility. A list of medical record numbers for adults diagnosed with diabetes was compiled using diabetic follow-up logbooks from both hospitals. Finally, the required participants were randomly selected from the sampling frame (a list of diabetic patients enrolled between January 1, 2019, and December 31, 2024) using Excel-generated random numbers (Figure 1).

Figure 1.

Schematic presentation of the sampling procedure for DM patients at public hospitals of Harari region, Eastern Ethiopia, 2025.

Study variables

The incidence of HHS was the main outcome variable. Independent variables included socio-demographic factors (such as age, sex, residence, and community health insurance), clinical factors (including BMI, family history of diabetes, recent infections, type of diabetes, glycemic control, duration of diabetes, presence of comorbidities, systolic and diastolic blood pressure, and chronic diabetic complications), and treatment factors (such as frequency of diabetes follow-up, type of medication used, duration of treatment, and medication non-adherence).

Data collection and quality control

Data were collected using a data extraction format adapted from various relevant studies.^{11,12,20–22} The internal consistency of the adapted tool was assessed using Cronbach’s alpha coefficient, yielding a value of 0.815, indicating good reliability. This format consisted of five sections: general information, socio-demographic characteristics, clinical characteristics, treatment-related parameters, and the follow-up form. Initially, all DM patients on chronic follow-up between January 1, 2019, and December 31, 2024, were identified from the DM registration book. Eligible adult patients were then randomly selected using Excel-generated random numbers, after excluding individuals who developed HHS at their initial DM diagnosis and those who were transferred in. Data were collected from patients’ charts using a pretested data extraction format. Medical registration numbers were used to identify reviewed records. The data collection was conducted by two BSc public health professionals experienced in data collection, supervised by a trained BSc nurse working in chronic follow-up and an MPH public health professional.

To ensure data quality, data collectors and supervisors received a one-day training session focused on the data extraction system, data collection tools, and the study’s objectives. Before the main data collection, a pretest was conducted at HFCSUH on 5%²³ of the sample size, and necessary adjustments were made to the data extraction format based on the findings. A week before data collection began, patient charts were cross-checked to confirm the adequacy of the instruments, ensure the checklist of tools was completed on time, and verify the completeness of data in the charts. Close monitoring and supervision were conducted throughout the data collection period by both the principal investigator and the supervisors. Data collectors received daily feedback to address any issues. The collected data were reviewed for completeness before analysis. To minimize misclassification bias, uniform ascertainment criteria were applied for clinical variables with multiple diagnostic criteria. Missing data were addressed using a complete case analysis.

Operational definitions

Event: The occurrence of HHS.

Censored: Adults diagnosed with DM who did not develop HHS during the follow-up study (transfer out, died, lost to follow-up, and still had not developed HHS as the study ended).

Incomplete records: Incomplete records are charts that lack complete information for the variables of the date of DM diagnosis, the date when HHS occurred, and follow-up history after diagnosis.

HHS status: ‘Yes’ for HHS indicates that the patient experienced their first incident of HHS (RBS >600 mg/dl with minimal or absent urine ketone),²³ based on the clinical decisions of the physicians and confirmed from the patient’s medical records. ‘No’ for HHS indicates that the patient did not experience any incidents of HHS during the follow-up period and is considered censored.

Glycemic control: Good glycemic control was defined as HbA1c of < 7% or average fasting blood sugar (FBS) over three months of 70–130 mg/dl. Poor glycemic control was defined as HbA1c of ≥ 7% or average three-month FBS of > 130 or < 70 mg/dl.²⁴

Medication non-adherence: Any documentation in the medical record indicating missed doses, irregular/inconsistent medication use, defaulting, or discontinuation of prescribed anti-diabetic medications during follow-up.²⁵

Adult : Defined as individuals aged ≥ 18 years at the time of enrollment.

Data processing and analysis

After completing data collection, the data were exported from the Kobo Toolbox to Microsoft Excel and then imported into STATA software version 17 for further cleaning and analysis. Basic descriptive analyses were conducted. Each patient was followed until either the occurrence of HHS or one of the censoring events, whichever came first. Survival analysis was employed as the statistical model to assess the associations between HHS incidence and its predictors. The HHS-free survival probability was calculated in months, based on the time interval between diabetes mellitus (DM) diagnosis and either the date of HHS occurrence or censoring. The incidence rate of HHS was also computed. Life tables were used to estimate the survival probabilities at different time intervals following the diagnosis of DM. Kaplan-Meier survival curves and log-rank tests were performed to estimate survival probabilities and to statistically compare survival across different predictor categories, respectively.

To identify predictors of HHS, both the Cox proportional hazard (PH) model and three parametric models (Weibull, exponential, and Gompertz) were fitted. The most parsimonious model was selected based on the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and log-likelihood values. Based on these criteria, the Weibull PH regression model was identified as the most efficient for describing the data.

The Schoenfeld residuals test was employed to evaluate the proportional hazards assumption for individual predictors as well as for the overall model using global tests.¹⁹ Each predictor had a P-value greater than 0.05, and the overall global test had a P-value of 0.47, confirming that the proportional hazards assumption was not violated. A log-log plot of survival was used to evaluate the predictors, and the variables in the final model yielded parallel curves, confirming the proportional hazards assumption. Multicollinearity was assessed using the variance inflation factor (VIF = 1.76), suggesting no significant multicollinearity among the independent variables.

A Weibull PH regression model was used to identify the predictors of HHS. Variables with p-values < 0.25 in the bi-variable analysis were candidates for a multivariable Weibull PH regression model. Hazard ratios (HR) with their corresponding 95% CI were used to show significance and strength of the association with the outcome variable. Variables with a p-value < 0.05 in the multivariable Weibull PH regression model were considered significant and independently associated with the outcome variable. The goodness of fit of the model was assessed by using the Cox–Snell residual technique. The cumulative hazard plot follows a straight line through the origin with slope, indicating that the model fits the data well (Figure 2).

Figure 2.

Cox–Snell residual plots for Weibull PH model of adult patients with diabetes at public hospitals of Harari region, Eastern Ethiopia, 2025. (N=452).

Ethical considerations

Ethical clearance and approval were obtained from the Institutional Health Research Ethics Review Committee (IHRERC) of Haramaya University College of Health and Medical Sciences. A cooperation letter was issued by Haramaya University College of Health and Medical Sciences to the respective hospitals. Informed, voluntary, written, and signed consent was obtained from the hospital head. After obtaining ethical approval, data were collected from patient charts using a checklist, with codes used to maintain confidentiality. The completed checklist was stored securely in a locked location. The entered data was protected with a password on the computer. The data was used solely by the principal investigator.

Result

Socio-demographic characteristics

From a total of 461 adult diabetes patient cards that were reviewed, 452 were included in the final analysis, while 9 (2%) patients with incomplete information were excluded. Out of 452 adults with diabetes, half (50%) of the participants were males, and 245 (54.20%) were rural dwellers. The mean age at the time of diabetes diagnosis was 46 years, with a standard deviation of 16 years. Those who had no health insurance constituted 158(34.96%) of the diabetes patients (Table 1).

Table 1.

Incidence of HHS and socio-demographic characteristics of diabetes patients at public hospitals of Harari region, Eastern Ethiopia. (N=452).

Variables	Category	HHS status		Total (%)
Variables	Category	Developed HHS(n=45)	Censored (n=407)	Total (%)
Health facility	HFCSUH	27	232	259(57.30)
Health facility	JGH	18	175	193(42.70)
Age at diagnosis	18-44	22	174	196(43.36)
	45-64	17	180	197(43.58)
	≥ 65	6	53	59(13.05)
Sex	Male	26	200	226(50.00)
Sex	Female	19	207	226(50.00)
Residence	Urban	17	201	245(54.20)
Residence	Rural	28	135	207(45.80)
CBHI	Yes	19	275	294(65.04)
CBHI	No	26	132	158(34.96)

CBHI, community-based health insurance.

Clinical and treatment characteristics of patients with diabetes mellitus

According to this study, 344 (76.11%) of the DM patients had T2DM, and the majority of the patients, 397(90.6%), had no family history of diabetes. Of the total, 157 (34.73%) had an acute infection, and 202 (44.69%) of the patients had uncontrolled blood sugar. Over half of the participants (55.97%) had comorbidity. Of the total, 34 (7.52%) had chronic complications of DM. More than one-fourth, 126(27.88%), were overweight, and 158(34.96%) patients had irregular follow-up. Most of the patients, 312 (69.03%), took oral hypoglycemic agents at diagnosis, and 94 (20.80%) of adult patients with diabetes had medication non-adherence (Table 2).

Table 2.

Clinical and treatment-related results of adult patients with diabetes at public hospitals of Harari region, Eastern Ethiopia. (N=452).

Variables	Category	Frequency (%)
Type of DM	Type2DM	344(76.11)
Type of DM	Type1DM	108(23.89)
DM duration	<3 years	350(77.43)
DM duration	≥ 3 years	102(22.57)
Family history of DM	No	397(90.64)
Family history of DM	Yes	41(9.36)
Glycemic control	Good glycemic control	250(55.31)
Glycemic control	Poor glycemic control	202(44.69)
Trauma	No	436(96.46)
Trauma	Yes	16(3.54)
Body mass index status	Normal	287(63.50)
	Overweight	126(27.88)
	Obesity	25(5.53)
	Underweight	14(3.10)
Acute infection	Yes	157(34.73)
Acute infection	No	295(65.27)
Comorbidities	Cardiovascular disorder	158(34.95)
	Chronic respiratory disorder	24(5.30)
	Renal disease	12(2.65)
	Liver disease	9(1.99)
	HIV/AIDS	8(1.76)
	Benign prostate hyperplasia	7(1.54)
	Psychiatric disorder	6(1.32)
	Others**	20(4.42)
	More than one comorbidity	57(12.61)
Chronic complication of DM	Yes	34(7.52)
Chronic complication of DM	No	418(92.48)
Type of treatment	Oral hypoglycemic agent	312(69.03)
	Insulin	131(28.98)
	Oral and Insulin	9(1.99)
Medication non-adherence	No	358(79.20)
Medication non-adherence	Yes	94(20.80)
Follow-up frequency	Regular	294(65.04)
Follow-up frequency	Irregular	158(34.96)

*Others include carbuncle, thyrotoxicosis, surgical site infection, necrotizing fasciitis, sexually transmitted infections, and acute otitis media.

**Others include dyslipidemia, hypothyroidism, epilepsy, hyperthyroidism, and goiter.

Incidence of hyperosmolar hyperglycemic state

Out of 452 patients followed over five years, 45 developed HHS (9.96%; 95% CI: 7.51%–13.08%), while the majority (90%) were censored (Figure 3). The overall incidence rate was 5.36 cases per 100 person-years (95% CI: 4.00–7.18). Specifically, the rate was higher among patients with T2DM (5.76) than those with T1DM (3.89) per 100 person-years. Incidence was notably higher in rural residents (8.28; 95% CI: 5.71–11.99) and older individuals aged ≥ 65 years (6.18; 95% CI: 2.78–13.76) per 100 person-years. Throughout the follow-up period, the highest incidence of HHS was recorded between the 12-to-18-month follow-up intervals (Table 3). HHS-free survival probability declined over time, reaching 82.11% by the end of the follow-up period. The median HHS-free survival time for the entire cohort was undefined, as the survival curve did not cross the 50% threshold, indicating that more than half of the cohort remained HHS-free at the last recorded time point. (Figure 4).

Figure 3.

Bar graph showing the status of HHS among adult patients with diabetes at public hospitals of Harari region, Eastern Ethiopia. (N=452).

Table 3.

Life table showing the survival to develop HHS among adult diabetic patients at public hospitals of Harari region, Eastern Ethiopia. (N=452).

Interval in months	Patients at risk	No of HHS cases	Censored	Cumulative survival	SD. error	[95% CI] L U
(0 6]	452	9	57	0.0213	0.0070	0.0111 0.0404
(6 12]	386	9	75	0.0465	0.0108	0.0295 0.0730
(12 18]	302	15	78	0.1009	0.0170	0.0723 0.1399
(18 24]	209	4	29	0.1194	0.0190	0.0871 0.1625
(24 30]	176	4	48	0.1426	0.0217	0.1054 0.1914
(30 36]	124	1	20	0.1501	0.0228	0.1110 0.2013
(36 42]	103	1	12	0.1589	0.0242	0.1174 0.2131
(42 48]	90	2	28	0.1810	0.0282	0.1328 0.2441
(48 54]	60	0	22	0.1810	0.0282	0.1328 0.2441
(54 60]	38	0	38	0.1810	0.0282	0.1328 0.2441

Figure 4.

Kaplan–Meier survival curve of HHS-free survival estimate of adult patients with diabetes at public hospitals of Harari region, Eastern Ethiopia. (N=452).

Predictors of hyperosmolar hyperglycemic state

In the bivariable analysis, variables with a p-value <0.25 were entered into a multivariable Weibull proportional hazards regression model to identify independent predictors of hyperosmolar hyperglycemic state. Ten variables (CBHI (community-based health insurance), Residence, Medication non-adherence, glycemic control, Types of DM, Trauma, acute illness, comorbidities, follow-up frequency, and family history) were identified as the predictors of HHS at p < 0.25. In the multivariable Weibull PH regression analysis, four variables were identified as predictors of HHS. These were rural residence, comorbidities, CBHI, and medication non-adherence (Table 4).

Table 4.

Bivariate and multivariate Weibull proportional hazards regression of predictors of HHS among DM patients at public hospitals in the Harari region, East Ethiopia.

Variable	Category	HHS status		CHR (95% CI)	AHR(95% CI)	P-value
Variable	Category	Developed HHS(n=45)	Censored (n=407)	CHR (95% CI)	AHR(95% CI)	P-value
CBHI	Yes	19	275	1	1
	No	26	132	3.13 (1.73- 5.66)	2.04 (1.03-4.03)	0.040
Residence	Urban	17	201	1	1	1
	Rural	28	135	2.36 (1.29- 4.32)	1.99 (1.01-3.92)	0.046
Glycemic control	Controlled	13	237	1	1	1
Glycemic control	Uncontrolled	32	170	4.02 (2.11-7.69)	1.90 (0.87-4.12)	0.104
Types of DM	Type 1	7	101	1	1
	Type 2	38	306	3.87(2.83- 5.97)	2.15 (0.90-5.14)	0.082
Trauma	No	42	394	1	1
Trauma	Yes	3	13	3.70(1.14- 12.00)	1.92(0.55-6.68)	0.300
Comorbidities	No	8	191	1	1
Comorbidities	Yes	37	216	3.77(1.75-8.11)	3.01(1.28-7.05)	0.011
Follow-up frequency	Regular	21	273	1	1
Follow-up frequency	Irregular	24	134	2.24(1.25-4.04)	1.24(0.63- 2.44)	0.529
Acute recent illness	No	21	274	1	1
Acute recent illness	Yes	24	133	2.68(1.49- 4.83)	1.63(0.83-3.19)	0.149
Medication non-adherence	No	25	333	1	1
Medication non-adherence	Yes	20	74	4.06(2.25- 7.34)	2.14(1.04-4.40)	0.038
Family history of DM	No	36	361	1	1
	Yes	9	46	1.88(0.83-4.23)	1.19(0.50- 2.81)	0.691

CBHI, community-based health insurance.

Accordingly, patients from rural areas had a 99% higher hazard of developing HHS compared to those from urban areas (AHR: 1.99, 95% CI: 1.01, 3.92). Diabetes mellitus patients without CBHI had a two-fold higher probability of HHS than those with CBHI (AHR: 2.04, 95% CI: 1.03, 4.04). Regarding treatment adherence, the risk of HHS was 2.14 times higher among patients with medication non-adherence compared to those who were adherent (AHR: 2.14; 95% CI: 1.04, 4.40). Similarly, the hazard of HHS was three times higher in patients with comorbidities than in those without (AHR: 3.01; 95% CI: 1.28, 7.05) (Table 4).

Discussion

This study investigated the incidence and predictors of HHS among diabetic patients in Eastern Ethiopia. The incidence rate of HHS in this study was much higher than the national reports from the United States (0.9 per 1000 person-years)²⁶ and Denmark (20.4 per 10,000 person-years(16.5 and 3.9 among people with known type 1 and type 2 DM, respectively)),¹⁰ and higher than study conducted Bahar Dar, Ethiopia (2.1 per 100 person-years (0.9 and 2.4 among T1DM and T2DM, respectively)).¹¹ The higher incidence rates of HHS in this study may be attributed to its institution-based design, in contrast to nationwide studies. The incidence rate of HHS in this study was higher among individuals with type 2 diabetes mellitus, in contrast to a study conducted in Denmark, which reported a higher incidence rate among individuals with type 1 diabetes mellitus.¹⁰ This discrepancy highlights the importance of considering specific population characteristics and regional factors when assessing the risk and incidence of HHS. The cumulative incidence in this study was lower than reported rates in China (46.8%),¹¹ Atlanta (35%),²⁷ and Nigeria (58.3%).¹³ These discrepancies may be attributed to differences in study populations, socioeconomic status, culture, lifestyles, and environmental factors, all of which influence glycemic control and diabetes management.²⁸

This study identified a significant link between rural residence and HHS, which is consistent with a study conducted in Iraq.²⁹ The association could be due to limited access to healthcare, poorer glycemic control, and socioeconomic issues, all of which contribute to poor diabetes management and its complications.³⁰ This aligns with findings from other Ethiopian contexts, where the ‘distance decay’ effect is interpreted through Andersen’s Behavioral Model. Service utility decreases as distance increases, which often leads to delayed presentation of hyperglycemic emergencies.³¹

The incidence of HHS was greater among DM patients with comorbidities in comparison to those without comorbidities. This finding is consistent with studies conducted in Bahar Dar, Ethiopia¹¹ and in Sidama, Ethiopia.³² This may be due to comorbidities and their associated medications, which can disrupt the normal functions of insulin and other counter-regulatory hormones, either directly or indirectly. Furthermore, diabetic patients with chronic medical comorbidities often experience poor glycemic control.³³

In this study, patients without community health insurance were positively associated with HHS. This agrees with a study conducted in Bahar Dar, Ethiopia¹¹ and South Africa.³⁴ The reason might be that patients without community health insurance are less likely to afford regular self-blood glucose monitoring, and the costs of follow-up examinations.³⁵ Furthermore, the lack of CBHI acts as a financial barrier to secondary prevention, such as routine blood glucose monitoring. In Ethiopia, CBHI membership has been shown to increase outpatient department visits by up to 43%, suggesting that those without insurance are more likely to delay care until an acute crisis like HHS occurs.³⁶

Medication non-compliance was positively associated with HHS in this study. This finding aligns with a previous study conducted in China,³⁷ in Nigeria,¹³ and in Bahar Dar, Ethiopia.¹¹ This may be due to medication non-compliance, which disrupts the management of diabetes, leading to increased risks for severe complications like HHS.³⁸ This also aligns with the Health Belief Model. Previous evidence from the Harari region, Ethiopia, indicates that diabetic patients often have low perceived severity regarding complications and low self-efficacy for self-care practices. When patients do not perceive the immediate threat of asymptomatic hyperglycemia, adherence to daily regimens drops.³¹

Strengths and limitations of this study

A key strength is the utilization of five years of follow-up data, which allows for the assessment of hyperglycemic hyperosmolar state (HHS) incidence over time and offers a comprehensive understanding of the condition. However, the retrospective design limits the inclusion of all potential factors influencing HHS, and the possibility of misclassification bias may affect the validity of the results. Specifically, because the HHS diagnosis relied on clinical records, some cases may have been miscoded as DKA or mixed states. This likely resulted in an underestimation of the true HHS incidence rate. Additionally, reliance on physicians’ clinical judgments for defining and diagnosing HHS may introduce inconsistency or unreliability. Other limitations include the inability to assess time-varying exposures and limited generalizability beyond public hospital attendees.

Conclusion

In this study, 10% of diabetes patients developed a hyperosmolar hyperglycemic state. The incidence rate of hyperosmolar hyperglycemic state was found to be higher among patients with T2DM compared to T1DM. Diabetes programs should adopt a T2DM-targeted HHS prevention and early-detection policy. Rural residence, comorbidities, medication non-adherence, and lack of community health insurance were found to be significantly associated with the hazard of HHS. To mitigate these risks, the Harari Regional Health Bureau should prioritize the expansion of CBHI enrollment specifically for chronic disease patients to remove financial barriers to monitoring. Furthermore, decentralizing diabetes follow-up care to rural health centers could reduce the distance decay effect identified in this study. Clinical practice should incorporate targeted patient education to improve medication adherence and early symptom recognition. Finally, we recommend that future prospective follow-up studies incorporate time-dependent covariates to account for variables that change over time and to address the missed variables.

Footnotes

Acknowledgments

We would like to thank institutional health research ethics review Committee of Haramaya University College of Health and Medical Sciences for its ethical approval. We would also like to thank the Harari region public hospital administrative bodies and card room workers for their cooperation. Finally, we would like to thank data collectors and supervisors for their commitment during data collection.

ORCID iDs

Birhanu Shegene

Alemayehu Tesfaye

Dawit Firdisa

Ethical considerations

The studies involving humans were approved by the Institutional Health Research Ethics Review Committee of the Haramaya University College of Health and Medical Sciences (IHRERC/149/2024). The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants’ legal guardians/next of kin due to retrospective nature of the study.

Author contributions

BS, AG, and AT were involved in the conception and design, acquisition of data, analysis, and interpretation of data. DF and SD were involved in data curation, software used, and revising contents until the final version was prepared for publication. Finally, all the authors read and approved the final version and agreed on all aspects of this work.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors upon request and without undue reservation.*

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting, or dissemination plans of this research.

References

American Diabetes Association . Diagnosis and classification of diabetes mellitus. Diabetes care 2013; 33(Supplement_1): S62–S69. https://doi.org/10.2337/dc10-s062

Fitrianingsih

Sauriasari

Yunir

, et al.

Socioeconomic and clinical profile of diabetic foot ulcer patients with national health insurance during pandemic COVID-19 in Indonesia

2025; 15: 21–33.

Orłowska

Kapłan

Ostański

, et al. Hyperosmolar hyperglycemic syndrome: A comprehensive review of clinical presentation, diagnosis, and treatment strategies in hyperglycemic crises. Journal of Education, Health and Sport 2024; 55: 217–230. https://doi.org/10.12775/jehs.2024.55.014

Woode

Kitabchi

. Hyperglycemic crises in diabetes mellitus: diabetic ketoacidosis and hyperglycemic hyperosmolar state. Management of Pediatric Obesity and Diabetes 2011; 23: 327–348.

Pasquel

Umpierrez

. Hyperosmolar hyperglycemic state: a historic review of the clinical presentation, diagnosis, and treatment. Diabetes care 2014; 37(11): 3124–3131. https://doi.org/10.2337/dc14-0984

Umpierrez

Korytkowski

. Diabetic emergencies—ketoacidosis, hyperglycaemic hyperosmolar state and hypoglycaemia. Nature Reviews Endocrinology 2016; 12(4): 222–232. https://doi.org/10.1038/nrendo.2016.15

Haile

Fenta

. Magnitude, risk factors and economic impacts of diabetic emergencies in developing countries: A systematic review. PloS one 2025; 20(2): e0317653. https://doi.org/10.1371/journal.pone.0317653

Djakaridja

Famoussa

Tatiana

, et al. ETIOLOGIES OF INFECTIONS IN DIABETIC PATIENTS HOSPITALIZED AT BOUAKE UNIVERTY TEACHING HOSPITAL, CÔTE D’IVOIRE. African Journal of Infectious Diseases 2023; 17(2 Suppl): 1–5. https://doi.org/10.21010/ajidv17i2s.1

Mairghani

James

Mehta

, et al. Global Monitoring of Hyperglycaemic Emergencies in Adults with Diabetes: A Systematic Review of Incidence and Outcomes (2015–2024). PROSPERO: International prospective register of systematic reviews 2024; 27: 17.

10.

Rosager

Heltø

Fox Maule

, et al. Incidence and characteristics of the hyperosmolar hyperglycemic state: a Danish cohort study. Diabetes Care 2024; 47(2): 272–279. https://doi.org/10.2337/dc23-0988

11.

Abate

Semachew

Emishaw

, et al. Incidence and predictors of hyperglycemic emergencies among adult diabetic patients in Bahir Dar city public hospitals, Northwest Ethiopia, 2021: A multicenter retrospective follow-up study. Frontiers in public health 2023; 11: 1116713. https://doi.org/10.3389/fpubh.2023.1116713

12.

Desse

Eshetie

Gudina

. Predictors and treatment outcome of hyperglycemic emergencies at Jimma University Specialized Hospital, southwest Ethiopia. BMC research notes 2015; 8(1): 553. https://doi.org/10.1186/s13104-015-1495-z

13.

Edo

. Clinical profile and outcomes of adult patients with hyperglycemic emergencies managed at a tertiary care hospital in Nigeria. Nigerian Medical Journal 2012; 53(3): 121–125. https://doi.org/10.4103/0300-1652.104378

14.

Hassan

Mushtaq

Mahmoud

, et al. Overlap of diabetic ketoacidosis and hyperosmolar hyperglycemic state. World journal of clinical cases 2022; 10(32): 11702–11711. https://doi.org/10.12998/wjcc.v10.i32.11702

15.

Wang

Hao

, et al. Type 2 diabetes–prevention diet and all-cause and cause-specific mortality: A prospective study. American journal of epidemiology 2022; 191(3): 472–486. https://doi.org/10.1093/aje/kwab265

16.

CSA . Population Size by Sex, Region, Zone and Wereda. [Available from. https://www.statsethiopia.gov.et/ (2023).

17.

Cheru

Edessa

Regassa

, et al. Incidence and predictors of chronic kidney disease among patients with diabetes treated at governmental hospitals of Harari Region, eastern Ethiopia, 2022. Frontiers in Public Health 2024; 11: 1290554. https://doi.org/10.3389/fpubh.2023.1290554

18.

HMIS . Hiwot Fana Comprehensive Specialized University Hospital and Jugal General Hospital HMIS report. Haramaya University; Harari Regional Health Bureau, 2024.

19.

Schoenfeld

. Partial residuals for the proportional hazards regression model. Biometrika 1982; 69(1): 239–241. https://doi.org/10.1093/biomet/69.1.239

20.

Gebre

Assefa

. Magnitude and associated factors of diabetic complication among diabetic patients attending Gurage zone hospitals, South West Ethiopia. BMC research notes 2019; 12(1): 780. https://doi.org/10.1186/s13104-019-4808-9

21.

Negera

Weldegebriel

Fekadu

. Acute Complications of Diabetes and its Predictors among Adult Diabetic Patients at Jimma Medical Center, Southwest Ethiopia. Diabetes, metabolic syndrome and obesity : targets and therapy 2020; 13: 1237–1242. https://doi.org/10.2147/DMSO.S249163

22.

Tittel

Sondern

Weyer

, et al. Multicentre analysis of hyperglycaemic hyperosmolar state and diabetic ketoacidosis in type 1 and type 2 diabetes. Acta diabetologica 2020; 57(10): 1245–1253. https://doi.org/10.1007/s00592-020-01538-0

23.

Adeyinka

Kondamudi

. Hyperosmolar hyperglycemic syndrome. StatPearls Publishing, 2023. InStatPearls [Internet].

24.

Association

. Standards of Medical Care in Diabetes—2020 Abridged for Primary Care Providers. Clinical Diabetes 2020; 38(1): 10–38.

25.

Araya

Gebrezgabiher

Tekulu

, et al. Medication Non-Adherence and Associated Factors Among Diabetic Patients Visiting General Hospitals in the Eastern Zone of Tigrai, Northern Ethiopia. Patient preference and adherence 2020; 14: 2071–2083. https://doi.org/10.2147/PPA.S278148

26.

Centers for Disease Control and Prevention (CDC . National diabetes statistics report. U.S. Department of Health and Human Services, 2017. -estimates of diabetes and its burden in the United States background.

27.

Pasquel

Tsegka

Wang

, et al. Clinical outcomes in patients with isolated or combined diabetic ketoacidosis and hyperosmolar hyperglycemic state: a retrospective, hospital-based cohort study. Diabetes care 2020; 43(2): 349–357. https://doi.org/10.2337/dc19-1168

28.

Alibeji

Stewart

. Cultural considerations in diabetes management. America’s Pharmacist, 2016.

29.

Mansour

Abdu-Alla

. Predictors of diabetic ketoacidosis among patients with type 1 diabetes mellitus seen in the emergency unit. British Journal of Medicine and Medical Research 2016; 11(10): 1–12. https://doi.org/10.9734/bjmmr/2016/21384

30.

Rahman

Gasbarro

Alam

, et al. Rural‒urban disparities in household catastrophic health expenditure in Bangladesh: a multivariate decomposition analysis. International Journal for Equity in Health 2024; 23(1): 43. https://doi.org/10.1186/s12939-024-02125-3

31.

Ayele

Tesfa

Abebe

, et al. Self care behavior among patients with diabetes in Harari, Eastern Ethiopia: the health belief model perspective. PloS one 2012; 7(4): e35515. https://doi.org/10.1371/journal.pone.0035515

32.

Tiruneh

Mekonnen

Getnet Zemeskel

, et al. Incidence and predictors of hyperglycemic emergencies among adult diabetic patients at public hospitals of Sidama Region and Gedeo Zone, Southern Ethiopia: a survival analysis. Therapeutic advances in endocrinology and metabolism 2024; 15: 20420188241303416. https://doi.org/10.1177/20420188241303416

33.

Mamo

Bekele

Nigussie

, et al. Determinants of poor glycemic control among adult patients with type 2 diabetes mellitus in Jimma University Medical Center, Jimma zone, south west Ethiopia: a case control study. BMC endocrine disorders 2019; 19: 1–11. https://doi.org/10.1186/s12902-019-0421-0

34.

Motala

Mbanya

Ramaiya

, et al. Type 2 diabetes mellitus in sub-Saharan Africa: challenges and opportunities. Nature reviews Endocrinology 2022; 18(4): 219–229. https://doi.org/10.1038/s41574-021-00613-y

35.

Tsega

Getaneh

Taddesse

. Are Ethiopian diabetic patients protected from financial hardship? PloS one 2021; 16(1): e0245839. https://doi.org/10.1371/journal.pone.0245839

36.

Yihdego

Tajvar

Sari

. Assessment of the Effect of Community-Based Health Insurance Scheme on Health-Related Outcomes in Ethiopia: A Systematic Review. Iranian Journal of Public Health 2024; 53(10): 2239–2250. https://doi.org/10.18502/ijph.v53i10.16701

37.

She

Wang

, et al. Clinical profiles, outcomes and risk factors among type 2 diabetic inpatients with diabetic ketoacidosis and hyperglycemic hyperosmolar state: a hospital-based analysis over a 6-year period. BMC Endocrine Disorders 2020; 20: 1–9. https://doi.org/10.1186/s12902-020-00659-5

38.

Smollin

. Current Medical Diagnosis & Treatment 2024. McGraw-Hill Education, 2024.