Macrovascular Complications in Patients with Diabetes Mellitus: Incidence and Impact on Survival in Kazakhstan

Background and aim: Diabetic patients are at an increased risk for the development of macrovascular complications such as acute myocardial infarction (AMI), stroke and lower-limb amputations (LLA). This study aimed to explore a. the incidence of hospital admission for macrovascular complications (AMI, stroke, and LLA); b. to assess the impact of hospital admission on survival in a large population with diabetes mellitus living in Kazakhstan. Materials and methods: Retrospective observational study using a nationwide anonymized electronic database of 98.469 hospitalized diabetic patients from Kazakhstan between November 2013 and December 2019. The incidence of hospital admissions for AMI, stroke and LLA were obtained to calculate their all-time cumulative incidence, and survival rate at follow-up. Results: The all-time cumulative incidence of hospital admissions was 1.30% for AMI, 1.94% for stroke and 2.94% for LLA. The incidence of macrovascular complications was statistically signicantly higher in males compared to females (p-value<0.05). 29.03% of diabetic patients with AMI, 25.16% with stroke and 29.80% with LLA died during the follow-up period. Individuals with AMI had 3.58 (95% CI 3.20; 4.01) times, with stroke 3.86 (95% CI 3.52; 4.24) times and with LLA 3.63 (95% CI 3.38; 3.88) times higher hazard of 6-year death compared to diabetic patients free of these complications. The stratied survival analysis by sex indicated the lower survival in women than in men, and the lower survival in older age groups. Conclusion: The results from this study shows that cumulative incidence of AMI and stroke among diabetic patients admitted in the hospitals in Kazakhstan between 2013-2019 years was similar to the estimates from other countries, but the incidence of LLA was signicantly higher in Kazakhstan. Patients with diabetes mellitus (DM) in Kazakhstan are at high risk of excess mortality if they suffer from macrovascular complications. More research is required to explore the reasons for the high incidence of those complications, in order to propose systematic solutions for lowering the incidence and improve survival.


Introduction
The prevalence of DM is increasing from year to year. Thus, if in 2017 DM cases reached staggering 451 million worldwide (among 18-99 years old), by 2045, it is projected to hit 693 million cases [1]. The mortality rate of DM was around 5 million (among 20-99 years old) in 2017 that is almost 10% of all deaths worldwide [1]. According to the International Diabetes Federation, the prevalence of DM in Kazakhstan was 6.2% (1.15 mln individuals) in 2019 [2]. The study NOMAD estimated a period prevalence between 2014-2016 of 8.2% [3]. High prevalence and mortality rates have a dramatic impact on healthcare expenditures. In 2017, the global expenditure on DM patients, aged 18-99 years, reached 850 billion USD, and it is expected to rocket up to 958 billion USD by 2045 [1].
Patients with diabetes mellitus are at high risk of cardiovascular complications with approximately a twofold increased risk compared to non-diabetic subjects. These complications are also associated with an increased risk of death [4]. In particular, diabetic patients with acute cardiovascular diseases (CVD) such as acute myocardial infarction (AMI), stroke and lower limb amputation (LLA) have a 4-fold increase in mortality rate compared to diabetic patients without such complications [5,6].
According to the Global Burden of Diseases, from 1990 through 2019, the disability adjusted life years (DALYs) for diabetic patients increased to 24.4%, ranking the disease at the 6th place among causes of the burden of disease [7]. In order to slow down the increasing public health impact of DM, it is crucial to explore complications of DM patients and address targeted measures against them [8].
Obviously, there are huge gaps in the valid knowledge about the incidence of DM-related complications, including acute cardiovascular and cardio-cerebral events, as well as lower limb amputations. Most research evidence comes from high-income countries that represent about 10% of the global population, while there are scanty data on the situation in Central Asia, particularly in Kazakhstan.
With these concepts in mind, we aimed to a. Determine the incidence of hospital admission for macrovascular complications (AMI, stroke, and LLA); b. Assess the impact of hospital admission on survival in a large population with DM living in Kazakhstan.

Data source and study population
This retrospective observational study used a nationwide anonymized database of hospitalized diabetic patients from Kazakhstan obtained from the Republican Center of Electronic Healthcare. Inclusion criteria was a) age > 18 years, and b) hospital admission with a diagnostic code of DM (all types) between 3 November 2013 and 31 December 2019. The database consisted of 652,048 observations. One single hospitalization may have multiple observations, as these represent claims submitted for speci c medical procedures that may occur during the same hospitalization. Patients were identi ed by RpnID -a unique individual population registry number, which is used across all electronic health systems. 91,108 diabetic patients with unique RpnID numbers were identi ed. Apart from that, 48,559 (7.4%) observations did not include RpnID. A probabilistic approach for imputation of missing data was applied by nding duplicates among existing RpnID numbers based on the following ve variables: birth date, sex, ethnicity, home address and region of residence. Following this, 17,342 observations that did not include RpnID were replaced by existing RpnID numbers. The remaining 31,217 observations were assigned with new 7,361 unique RpnID, which had the identical information for birth date, sex, ethnicity and region. Overall, rstly, after deleting 4,336 duplicates of RpnID by all variables, and then, 549,243 duplicates of RpnID separately for observations with and without complications of interest the database consisted of 98,469 diabetic patients with corresponding RpnID numbers, including an array of demographic and social covariages. The ow-chart of the data cleaning process is illustrated in Fig. 1.

Variables
We identi ed incident cases of the three DM-related complications of interest applying International Classi cation of Diseases codes, Tenth Revision (ICD-10) and International Classi cation of Diseases codes, Ninth Revision (ICD-9): AMI, both haemorrhagic and ischaemic strokes, LLA (including toe, foot, amputation below knee or other unspeci ed LLA). The speci c codes utilized were: for AMI ICD-10 codes of I21, I22, I23, I24, for stroke ICD-10 codes of I61, I63, I62, I64 codes, for LLA ICD-9 codes of 84.10, 84.11, 84.12. For descriptive statistics, categorical age (< 54, 54-65, > 65 years), sex (male, female), ethnicity (kazakh, russian, other), region of residence (north, south, east, west and central) and residency (urban, rural) variables were used. The primary outcomes were incidence of hospital admissions for the above mentioned DM-related complications and following death.

Data analysis
Statistical calculations were performed using STATA 16 statistical software (StataCorp) [9]. The descriptive statistics of study population and all-time cumulative incidence of DM-related complications was calculated and presented as percentages. Additionally, case fatality rates for each complication were estimated and compared at their 1st, 7th and 28th days. Following this, crude and strati ed by age categories and sex, non-parametric Kaplan-Meier survival curves were assessed for AMI, stroke and LLA among hospitalized diabetic patients. Patients were followed up from the day of hospital admission (day 1). The last day of follow-up was the death date or the last day of observation period (December 2019).
The log-rank test was used to compare survival between groups. Statistical signi cance was considered at p-value < 0.05. Cox proportional hazards regression was used to estimate unadjusted effect size estimates.

Ethics
The study was approved by the Institutional Research Ethics Committee (NU IREC #203/29112019).

Baseline characteristics
The population analyzed in this work consists of 98,469 patients with a mean follow-up period of 3.42±1.78 years (range 0 to 6.17 years). 37.28% were younger than 54 years, and 29.15% were older than 65 years. 59.82% were females. Ethnicity information was divided into three groups: Kazakhs (60.14%), Russians (19.29%) and Others (20.57%). In addition, 59.14% of the patients were from urban settings.
Sex-strati ed incidence of three complications was calculated. The incidence of AMI and stroke was higher in males compared to females (1.45% vs 1.19%, p<0.001, and 3.67% vs 2.45%, p<0.001, respectively, while the incidence of stroke was similar (1.95% in males vs 1.93% in females, p-value 0.06).
In addition, age-strati ed incidence of AMI, stroke and LLA variables were calculated. The incidence of AMI was 0.57% in the younger than 54 years group, and it was 2.5 fold lower than in the 54-65 years group (1.43%), 3.6 times lower than in the older than 65 years group (2.08%), (p-value<0.001). The incidence of stroke showed the same trend and it was 0.93%, 2.26% and 2.85%, respectively. Almost the same situation in the LLA variable. Incidence of 1.26% of the younger than 54 years group was the lowest, and it was lower for 2.5 times than in the 54-65 years group (3.16%) and 3.8 times lower than in the older than 65 years group (4.85%), (p-value<0.001) (Figure 2b). year death compared to diabetic patients free of complications.
Even though, the incidence of DM-related complications was higher in men than in women, the sexstrati ed survival analysis showed that men had better survival than women, while age-strati ed analysis demonstrated the patients older than 54 years showed lower survival than the same individuals younger than 54 years (Figures 4a-4c, Figures 5a-5c).

Discussion
This is the rst and largest observational retrospective study with electronic health data from all over Kazakhstan, shedding a light on the incidence of macrovascular complications and their impact on survival among diabetic patients and corroborating that the global un-meet need associated with macrovascular complications also affects the Kazakhstani diabetic patients.
According to the results, cumulative incidence of AMI, stroke, and LLA was 1.30%, 1.94%, and 2.94%, respectively. Those complications have a signi cant impact on survival of diabetic patients.
Ischemic heart disease and cerebrovascular disease represents the main causes of life lost to premature death in Kazakhstan [10]. Cardio-metabolic factors related with DM and their vascular complications, as dietary risks, high systolic blood pressure and high body mass index were the highest ranked risk factors for disease burden in Kazakhstan. Despite of progress due to reform over the past few years resulting in improvements in prevention and management of non-communicable disorders in the country [11], those health problems still have a signi cant public health impact and call for continuing strengthening of the health system to respond to their signi cant burden, including the evaluation of those interventions [12]. Along with increasing incidence and prevalence of DM in Kazakhstan, about 80% of diabetic patients are overweight or obese and uncontrolled elevated blood pressure is also highly prevalent, factors that contribute to the development of DM-related micro-and macrovascular complications [13][14][15].
The incidence of AMI found in this work is lower than has been reported in Africa, Americas, Europe and Eastern Mediterranean, but higher than in South-East Asia and Western Paci c; stroke was higher in Europe and Western Paci c, and lower in Africa, Americas, South-East Asia and Eastern Mediterranean. Studies from Spain and Israel with hospitalized diabetic patients reported a similar 2.0-3.0% incidence of AMI [16,17].
LLA were signi cantly higher in Kazakhstan than in any of those regions [18]. LLA has shown incidence ranging from 0.02-2.48% [19][20][21]. The higher observed incidence of LLA in our study could be associated with high prevalence of risk factors among diabetic patients in Kazakhstan [19], resembling the trends in low-and middle-income countries (LMICs), where the incidence is increasing, possibly due to poor control of vascular risk factors among diabetic patients.
We observed a higher incidence of AMI, stroke and LLA in men [19][20][21][22][23][24][25][26], while mortality was higher in women compared to men [26][27][28]. Although there is still a gap in explanation of this observation, previously it was found that diabetic women had a higher overall CVD risk at baseline [29], including higher body fat percentage and higher abdominal fat, a factor that is associated with insulin-resistance, but also were less likely to reach recommended levels of low-density lipoprotein (LDL) and cholesterol [30,31]: men may be treated more intensively [32]. But, there could be other socio-economic factors associated with sex-differences in mortality after macrovascular complications in diabetic patients [33]. Regarding LLA, men could seek less foot care and have a greater risk for development of fool ulcers, which is associated with higher incidence of LLA [20].
The incidence of all macrovascular complications in our study increased with age: factors like a higher presence of comorbidities, obesity, low level of physical activity, hyperglycemia and a longer duration of the disease are risk factors positively associated with CVDs and LLA [34]. Older age has been consistently associated with higher incidence of stroke and AMI [35]. However, fewer studies reported the incidence strati ed by age categories for AMI, stroke and LLA [17,20,28].
Mortality rates for these complications were almost similar: 29.03% for AMI, 25.16% for stroke and 29.80% for LLA during the follow-up. The 5-year mortality rate due to AMI and stroke in other studies varied between 13.9%-50% [36 -40] having been reported to be 62% for LLA [41].
Mortality re ects an increased risk of post-hospitalization mortality [37,38,[42][43][44]. Overall, diabetic patients suffering from AMI, stroke or LLA have a signi cantly higher risk of death during 6 year of follow up [4,45]. Some differences observed in the estimates from other studies could be partially explained by the diversity of data sources and methodologies applied in each study, by the heterogeneity in the characteristics of study populations, but may also be associated with the variations in care for diabetic patients in different countries.

Strengths and limitations.
Overall, this observational retrospective study has certain strengths. First, we have investigated the incidence of macro-vascular complications using a large database that covered all hospitalization cases with DM from all regions of Kazakhstan. We assume that the ndings re ected the real situation of incidence and mortality after those events. To our knowledge, this is the rst study in Kazakhstan and Central Asia to utilize such a large database, therefore representing an important contribution to understand the epidemiology of DM-related macrovascular complications in Kazakhstan.
Inevitably, this research has several limitations. Even though this study utilized a representative sample of diabetic patients, some major variables were not available for the analysis, including the type of DM, duration of DM, treatments, laboratory measures or comorbidities, including obesity, or health behaviors such as smoking or alcohol consumption, or lifestyle factors, such as level of physical activity and eating behavior. It is possible that the hospitalized population might be at a more progressed stage of the disease and have been prescribed different medications, which might affect consequent complications.
Secondly, we used a probabilistic approach for imputations of missing RpnID, thus, our estimates of incidence could be underestimated. Third, there could be potential misclassi cation bias as diagnosis of DM and DM-related complications were de ned by ICD-9 and ICD-10 codes. Fourth, the mortality rates for AMI and stroke could be underestimated as some cases could be coded with codes for main hospitalization reasons which might not be related to them.

Conclusion
This data demonstrates that patients with DM in Kazakhstan are at high risk of excess mortality if they suffer from macro-vascular complications, AMI, stroke and LLA. More research is required to characterize sub-populations of diabetic patients at higher risk for the incidence of those complications, but it is also important that Kazakhstan continues to improve the quality of care of its healthcare system to provide the integrated care required for managing a complex condition as DM.

Consent for publication
Not applicable.

Availability of data and materials
The data that support the ndings of this study are available from Republican Center for Electronic Health of the Ministry of Health of the Republic of Kazakhstan, but restrictions apply to the availability of these data, which were used under the contract-agreement for the current study, and so are not publicly available. Data is however available from the authors upon reasonable request and with permission of Ministry of Health of the Republic of Kazakhstan.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was supported by grants from the Nazarbayev University Faculty Development Research Grant Program 240919FD3913 and 080420FD1916.
Authors' contributions ASS contributed to conception, study design, data acquisition, analysis, interpretation of the data, literature search, wrote and drafted the rst manuscript and provided nal approval for the submission. BO, TM contributed to data analysis, interpretation of the data, literature search, wrote and drafted the rst manuscript and provided nal approval for the submission. AS, NG, AG: Analyzed and interpreted the data; Reviewed the manuscript. All authors read and approved the nal manuscript.  Figure 1 Flow chart of data cleaning process.  Forest plots of hazard ratios of mortality with corresponding 95% CI among diabetic patients strati ed by age and sex groups (controls are diabetic patients without complications of interest). Figure 6a-