In-Hospital Mortality and Its Predictors among Hospitalized Diabetes Patients: A Prospective Observational Study

Background Diabetes mellitus (DM) is one of the leading health emergencies of the 21st century and among the top ten causes of death among adults globally in 2017. Although Ethiopia has been victimized by the growing prevalence of DM, data regarding in-hospital mortality among admitted diabetic patients in Ethiopia, specifically in Jimma Medical Center (JMC), are lacking. Objective The aim of the study is to assess in-hospital mortality and its predictors among DM patients admitted to Jimma Medical Center. Methods A hospital-based prospective observational study was employed involving 120 diabetes patients admitted to JMC from October 01, 2020, to June 30, 2021. Data were collected on variables related to the patient, disease, medication, and clinical outcomes. Data were entered into Epidata version 4.6.0.4 for cleaning and exported to SPSS version 23.0 for analysis. Kaplan–Mayer and cox-regression analyses were used to compare the survival experience and to determine the predictors of clinical outcomes, respectively. Hazard ratio with its two-sided p value <0.05 was considered to declare the statistical significance. Result Of 120 DM patients, 81 (67.5%) of them were males. The in-hospital mortality was 13.34% (16/120). Rural residence (AHR: 3.46; 95% CI (1.12, 9.81)), age (AHR: 1.03; 95% CI: (1.001, 1.059)), admission with diabetic ketoacidosis (AHR: 5.01; 95% CI (1.12, 21.88)), and multiple comorbidities: five comorbidities (AHR: 9.65; 95% CI (1.07, 19.59)) and six comorbidities (AHR: 14.02; 95% CI (1.74, 21.05)) were independently associated with in-hospital mortality. On the other hand, exposure to nonantidiabetic medications decreased the hazard of mortality by 86.5% (AHR: 0.135; 95% CI (0.04, 0.457)). Conclusion This study showed the rate of in-hospital mortality was noticeably high. The study showed that rural residence, age, DKA, and having comorbidities (five and six) were the statistically significant predictors of in-hospital mortality. In contrast, the use of nonantidiabetic medications such as statins, ASA, and other antihypertensive agents before admission remained protective. Thus, proper strategies have to be devised to improve in-hospital mortality among admitted DM patients.


Introduction
Diabetes mellitus is one of the most common medical conditions prevalent all over the globe. In 2021, it is estimated that 537 million people, representing 10.5% of the global adult population, have DM. Tis number is expected to increase to 643 million (11.3%) in 2030 and 783 million in 2045. African region, including Ethiopia, is anticipated to have the greatest increase in the number of people with diabetes. Diabetes prevalence in Ethiopia in the last 17 years (2000-2016) ranges from 0.3% at Debre Berhan Referral Hospital to 7% in Harar town [1,2].
Diabetes mellitus is the leading cause of global morbidity and mortality. It attributed for 12.2% of all global deaths between the age of 20-79 years and 32.6% of all deaths occurring in the productive age group. African region, especially Sub-Saharan Africa (SSA), is largely inficted [3,4]. International Diabetes Federation reported an estimated 416,163 diabetes-related deaths in Africa, and the majority of those deaths occur in people aged ≤60 years [5]. A study from central Ghana showed in-patient diabetes mortality rates increased from 7.6 per 1000 to 30 per 1000 deaths from 1983 to 2012, respectively. Te reported average 28-day mortality rate was 18.5% [6]. In Ethiopia, DM-related mortality rate ranges from 2% to 21% [7,8]. It is estimated that 26,448 diabetes-related deaths occur in adults 20-79 years, in 2021 [5].
Several studies showed that in-hospital mortality among DM patients was signifcantly associated with older age, gender, hypertension, hyperlipidemia, burden of comorbidities, infection, poor glycemic control, lack of foot care, long standing diabetes, and prolonged hospital stay [9][10][11]. Despite diferent initiatives undertaken by the Ethiopian Diabetes Association and the country's National Strategic Action Plan (NSAP) for prevention and control of noncommunicable disease (NCD) including diabetes, currently, the country has been challenged by the growing magnitude of diabetes. Ethiopia is among the top 4 countries with the highest adult diabetic populations in Sub-Saharan Africa (SSA) [12], but important data regarding in-hospital mortality were inadequate [13]. Terefore, this study was aimed to assess in-hospital mortality and its predictors among patients with DM admitted to JMC.

Study Area and Period.
Te study was conducted at Jimma Medical Center (JMC). It is located in Jimma town, 352 km southwest of Addis Ababa, the capital city of Ethiopia. It is the only teaching hospital in southwest Ethiopia, providing services for approximately 15,000 inpatients. Moreover, 160,000 outpatients, 11,000 emergency cases, and 4500 deliveries will attend the center annually, and its catchment population is estimated to be over 20 million. Tis study was conducted from October 01, 2020, to June 30, 2021, at the emergency department, medical, surgical, and gynecology/obstetrics wards of JMC.

Study Design and Population.
A hospital-based prospective observational study was employed. All patients admitted to JMC with the diagnosis of DM were considered as source population and admissions to the emergency department, medical, surgical, and gynecology/obstetrics wards with the diagnosis of DM during the data collection period and fulfll the eligibility criteria were considered as sample population.

Eligibility Criteria.
Patients aged 18 years and those who spent at least ≥24 hours in the hospital were included in the study. Whereas, patients diagnosed with gestational DM, who refused to participate, and patients or caregivers who were unable to provide appropriate information during data collection were not considered eligible.

Sample Size Determination and Sampling Technique.
Te sample size for the study was determined by using the single population proportion formula. Considering the proportion of in-hospital mortality rate (p � 11.2%) by Kefale et al. [14], Z (standardized normal distribution value at 95% CI) of 1.96, margin of error (d) of 5%, and 10% nonresponse rate, the initial sample size calculated was 153 patients. However, the number of diabetes patients admitted to the emergency room, medical, surgical, and gynecology/ obstetrics wards over nine months in 2019 was only 382. Hence, the fnite population correction formula was applied and the corrected sample size became 109 patients. Adding 10% for the nonresponse rate, the fnal sample size became 120 patients. As the number of admissions was limited during the time of data collection, all admissions who met the inclusion criteria were recruited in the study using a consecutive sampling technique, and every patient was followed until discharge, referral to other facilities, or death.

Study Variables.
In-hospital mortality was considered as an outcome variable. Predictor variables include patientrelated factors (sociodemographic variables (age, sex, marital status, educational status, residence, and occupation)), disease-related factors (type of DM, duration of DM, length of previous hospital stay, DM-related complications, comorbidities, previous hospitalization, admission blood glucose level, systolic blood pressure (SBP), and diastolic blood pressure (DBP)), and medication-related factors (types of medications (antidiabetics and nonantidiabetic medications) and duration of treatment).

Data Collection Procedures and Quality Control.
Once the data collection tool was developed, it was reviewed by a group of endocrinologists and validated based on their recommendation. It was initially designed in English, then translated to the local language (Afan Oromo and Amharic), and back-translated into English by language experts to assure its consistency. Te semistructured questionnaire was designed to extract information through face-to-face interviews (sociodemographic data and some parts of the clinical characteristics of the patients), and the patients' medical charts were also reviewed (to extract data on clinical characteristics of the patients uncovered by interviews, clinical outcome, medication prescribed after admission, vital signs, and laboratory data). Data were collected by two trained pharmacists (B. Pharm) and one BSC Nurse; while one medical doctor was assigned to supervise the data collection process. All the protocols of COVID-19 were considered during data collection.
To ensure the data quality, data collectors and a supervisor were trained for one day before starting data collection on: how to collect the data, the contents of the questionnaire, ethical issues, how to obtain additional information from the treating physicians, and patient interviews. Te data collectors were also strictly supervised, and the principal investigator reviewed all flled formats daily. Moreover, a pretest was conducted on 5% of the participants before the actual data collection to check the consistency and validity of the data collection tool.

Data
Processing and Analysis. Data were entered into Epidata version 4.6.0.4 and exported to Statistical Package for Social Sciences (SPSS) version 23.0 for cleaning and analysis, respectively. Mean and standard deviation (SD) were used to summarize continuous variables. Categorical variables were expressed in percentage and frequency. Descriptive analysis was performed, and the results were presented by the texts, tables, and fgures. Kaplan-Meier (log-rank test) was used to compare the survival experience of the patients. Te Cox-proportional hazard model was used to determine predictors of in-hospital mortality. Bivariate Cox regression analysis was conducted, and the variables with p value less than 0.25 were considered for the multivariable regression analysis. Te hazard ratio was used as a measure of the strength of association, and the variables with p value <0.05 on the multivariable Cox regression were used to declare statistical signifcance.

Overview of the Study Participants.
Out of 130 consecutive patients admitted over nine months, 10 patients were excluded and 120 patients were included in the fnal analysis. Among those included in the analysis, 89 (74.17%) of them were admitted to the medical ward ( Figure 1).

Sociodemographic Characteristics.
Eighty one (67.5%) of study participants were males. Te mean (+SD) age of the participants was 50.21 ± 19.35 years. About one-third (37.5%) of them were farmers, and nearly half 58 (48.3%) of them had no formal education (Table 1).

Reasons for Hospitalization.
Diabetic ketoacidosis (DKA) was the most common reason for hospitalization. It accounted for 59 (49.2%) of the admissions. Admissions related to infections were 34 (28.33%) and that of diseases of circulatory system were 16 (13.34%). Te most common infection responsible for admission was pneumonia 15 (12.5%), and the most common cardiovascular disease attributed for admission was heart failure 5 (4.17%) ( Table 2).
Overall, 87 (72.5%) patients had at least one acute or chronic comorbidity. Hypertension was the most common type of comorbidity contributing to 51 (58.62%) of the cases, followed by pneumonia 35 (40.23%) ( Table 3).
Both antidiabetic and nonantidiabetic medications were used in the management of admitted diabetic patients. Among those, cephalosporins were the most common among anti-infectives and prescribed for 63(52.5%) patients. Similarly, antilipidemic agents 47 (39.16%) were commonly prescribed cardiovascular agents (Table 6).

Predictors of in-Hospital
Mortality. Cox proportional hazard regression was conducted to identify predictors of mortality. In bivariate analysis, sex, age, residence, educational status, DM-related admission, newly diagnosed DM, DKA as admission diagnosis, history of antidiabetic medications, history of nondiabetic medications, number of comorbidities, presence of diabetic complications, and status of RBS immediate before discharge or at death were associated with death (p < 0.25).
Tis implies that infectious conditions could be the leading causes of death. However, a study from Nigeria reported the highest mortality among patients presented with hypoglycemia, stroke, and DFU [10]. Te proportion of mortality was also higher than the study from Harari region of Ethiopia (4.4%) [18] and WHO's (5.2%) report [19]. Moreover, among the deceased patients, 81.25% were diagnosed with T2DM and 56.25% of them were above the age of 60. In this study, majority of the patients were known diabetics, and consequently, they were presented with multiple comorbidities and DM-related complications. Tis might contribute to the amplifed mortality.
Te current study also identifed the predictors of inhospital mortality among admitted DM patients. Accordingly, the hazard of mortality was 3.46 times higher for rural  residents (AHR: 3.46; 95% CI (1.12, 9.81)). Tis fnding was consistent with the studies from China [20] and USA [21,22]. In the studies from China and USA, limited access to primary health care, health literacy, lifestyle choice, and economic burden were reasons for in-hospital mortality among rural residents. In the current study, complications at admission, lower educational status, and delayed arrival to the hospital were common among rural residents, and this may attribute to the increased in-hospital mortality. Moreover, a unit increase in age of the patient increased the hazard of the in-hospital mortality by 1.03 times (AHR: 1.03; 95% CI (1.001, 1.059)). Tis fnding was consistent with the studies from USA [15,23] and Jordan [24]. Tis could be due to the fact that old age could be more prone to increased hostile cardiovascular risk factors such as macrovascular complications, other impacts of aging, and patients' tendency to be immunosuppressed. Similarly, the risk of mortality was 5.01 times higher in patients admitted with DKA (AHR: 5.01; 95% CI: (1.12, 21.88)). Te fnding was in line with the studies from    China [20], Nigeria [10,25], and Addis Ababa, Ethiopia [7]. However, the fnding was inconsistent with the study from Portugal where high rates of hospital mortality have been associated with DFU's [26]. Te increased hazard of mortality due to DKA may be related to cerebral edema, burden of comorbidities such as precipitating factors, and some patients were also presented with septic shock. In this study, in-hospital mortality was also signifcantly associated with number of comorbidities. Te hazard of mortality was 9.65 and 14.02 times higher in patients with fve and six comorbidities (AHR: 9.65; 95% CI (1.07, 19.59)) and (AHR: 14.02; 95% CI: (1.74, 21.15)), respectively. Tis fnding was similar to studies from Italy [27,28], Israel [29], to Brazil [30]. Co-morbidities may be associated with increased disease severity, complicates the clinical course of diseases, and attenuates the body's natural defense mechanism against diseases by afecting multiple-body systems.    Contrarily, mortality was 86.5% lower among those exposed to nonantidiabetic medications such as statins, ASA, ACEI, BB, and CCB prior to the current admission (AHR: 0.135; 95% CI (0.04, 0.457)). Tis result supplemented the study from Iceland where statin use was associated with 53% reduction in all-cause mortality and 50% reduction in cardiovascular mortality in DM patients [31]. Similar fnding was reported from America where statin therapy in older people (≥65 years) without CVD decreased the risk of all-cause mortality by 14%, CVD death by 20%, and stroke by 15% [32]. Moreover, the fnding of this study supplemented the study conducted in America, among COVID-19 patients with DM receiving statins in whom 12% reduction in the adjusted risk of inhospital mortality was reported [33]. It also supported the ADA guidelines recommendation which promotes the use of low-dose aspirin for diabetic patients with 10-year CVD risk ≥10% [34]. Similarly, other studies also found that ACEIs reduced all-cause mortality, cardiovascular mortality, and cardiovascular events in patients with DM [35,36].
Tough its prospective nature and longer study period (over nine months) provided better data quality, the study sufers from several limitations. First, it was a single-center study. Second, RBS was used in the study; rather than HgA1C which better describes the status of glucose control in the last three months. Tird, it might be difcult to generalize the fndings of this study to the entire DM population due to small sample size.

Conclusions
In this study, the rate of in-hospital mortality was high. More than one-eighth of admitted DM patients died in hospital. Te study showed that rural residence, age, DKA, and having comorbidities (fve and six) were the statistically signifcant predictors of in-hospital mortality. In contrast, the use of nonantidiabetic medications such as statins, ASA, and other antihypertensive agents before admission remained protective.

Data Availability
All relevant data that support the fndings of this study are within the manuscript.

Ethical Approval
Te study was conducted after securing ethical approval from the Institutional Review Board of Jimma University (IRB No: IRB000236/2012). Ten, ofcial permission was obtained from the JMC clinical director before data collection was commenced. Te copy of the ofcial permission letter was then submitted to each head of the wards for ofcial permission to conduct the study. Te data collection was conducted by coding each data collection formats with secret codes.

Consent
A written informed consent was also obtained from all participants.