Incidence and predictors of diabetic ketoacidosis among children with type 1 diabetes mellitus in Western Amhara Referral Hospitals, Northwestern Ethiopia, 2018: A Retrospective Follow-Up Study CURRENT

Background Diabetic ketoacidosis remains a major cause of morbidity, hospitalizations and mortality in children with established type 1 diabetes mellitus. Therefore, this study aimed to determine the incidence and predictors of diabetic ketoacidosis among children with established type 1 diabetes mellitus at Western Amhara region. Method Institution-based retrospective follow-up study was done on 393 children with established type 1 diabetes mellitus registered between September 2013 and September 2017 in Western Amhara referral hospitals. The collected data was entered into Epidata version 4.2 and further analysis were done using STATA version 14.1. Negative Binomial Poisson Regression analysis model was used. Result The cumulative incidence and incidence density rate of diabetic ketoacidosis among children with established mellitus referral and a higher rate of diabetic ketoacidosis.


Introduction
In children diabetic ketoacidosis is an acute life threatening and recurrent medical emergency that requires frequent hospitalizations, treatment and monitoring for multiple metabolic abnormalities and vigilance for complications. It is the leading cause of morbidity and mortality that continued to be a major public health concern [1]. It has considerable costs to the health care systems and adds burden of costs to the patients and families [2]. Despite the improvement in the treatment and care of type 1 diabetes mellitus and the development of guidelines, diabetic ketoacidosis is still a major cause of hospitalization and the leading cause of death in children with type 1 diabetes mellitus [1,[3][4][5][6][7]. According to international diabetic federation's estimation, the incidence of diabetes mellitus in Ethiopian children (0-14) was 30 per 1000 populations. The mean diabetes related expenditure per a child was 29 United States Dollars [8].
Diabetic ketoacidosis has an overall mortality between 15 and 31 per 10,000 patients, and cerebral edema (CE), which is a devastating complication, accounts for between 57% and 87% of DKA related deaths [12]. Even in developed countries, there is significantly excess mortality from ketoacidosis among children with type 1 diabetes mellitus [25]. The incidence of DKA varies considerably between different countries and studies [26]. The majority of DKA cases occur in patients with previously diagnosed diabetes [24]. It occurs at a rate of 1-10/100 patient-years in children with established diabetes globally. The most recent rate of DKA was 4.81/100 patient-years in children with established type 1 diabetes mellitus in Germany [10]. Up to 90% of children with type 1 DM reported one or more episodes of DKA over the last 6 months in Sub Saharan Africa. Other studies show a variation from 25-90% [11,27].
In developing countries like Ethiopia, the risk of dying from DKA is greater. Even though many patients with diabetes mellitus in Ethiopia keep dying from DKA, there is little documentation about DKA. This creates double burden of communicable and noncommunicable diseases [28]. Therefore, this study aimed to determine the incidence and predictors of diabetic ketoacidosis among children with type 1 diabetes mellitus at Debremarkos and Felegehiwot referral hospitals, 2013-2017.

Methods
Study design, study setting, Study period and populations Hospital based retrospective follow up study was done using routine hospital data at Western Amhara referral hospitals, from September 2013 to September 2017. The study was conducted at two referral hospitals in Amhara Regional State (Northwestern Amhara);

Debre Markos Referral Hospital and Felegehiwot Referral Hospital. Debremarkos Referral
Hospital is found in Debre Markos City administration located at 299 km Northwest of Addis Ababa. The Hospital provides service for more than 3.5 million people [29]. A database from Information Technology prepared excell sheet was seen and about 190 children with type 1 DM were registered during the follow-up period. Felegehiwot Referral Hospital is found in Bahirdar city Administration that was located 563km from Addis Ababa.The hospital served the population in the region and those from Benshangul Gumuz population totally more than 7 million people as a referal center [30] Data collectors, procedure and quality control Two BSc nurses were recruited to participate in data collection. Two health officers were recruited for supervision. Data collectors and supervisors were trained for one day about data collection procedure and the purpose of the research. A week before the actual data collection, checklist was pretested on 22(5%) of the diabetic children records from the same area. Amendments were done on the checklist after the pretest. On site supervision was carried out during the whole period of data collection on daily basis by the supervisor and principal investigator. At the end of each day, checklist was reviewed and crosschecked for completeness, accuracy and consistency by the supervisors and principal investigator and corrective measures was under taken. Exclusion criteria were considered.

Operational Definitions
Established type 1 diabetes mellitus: at least with diabetic duration of one week. Descriptive results were presented using frequency, percentage, mean, standard deviation, median, Interquartile range, cumulative incidence and incidence density were calculated and some of them were presented in graphs and tables. Incidence rate ratio was also calculated and reported. The analysis was started by testing the significance of the association between each predictor and the dependent variable using Bivariable

Sociodemographic characteristics
A total of 393 (90.13%) registered patients' charts that fulfil the inclusion criteria were reviewed. Of these, almost half, 193 (49.1 %) were females. The median age of the patients was 10 (IQR:5-12) years. Slightly more than half, 224 (57%) of them were residing in rural area. Nearly half, 215 (54.71%) of them were given diabetic care by father and mother. The rest were given by father 54 (13.74%), mother 61(15.52%) and siblings only 63 (16.03%). Two hundred twenty two (56.5 %) of the study subjects had community health insurance. Regarding ethnicity almost all, 385 (98%) of patients were Amhara and the rest 8 (2%) were Gumuz (table 1).
Three hundred ninety three patients who were followed for different periods in five years  Diabetic children who were in the age group 10-15years had 2.61 times higher (ARR: 2.61, p value < 0.001) probability of developing diabetic ketoacidosis than those who were in the age group 6-10 years. Children whose primary caregivers were siblings only had 87% increased (ARR:1.87, p value < 0.001) rate of DKA as compared to patients whose primary care givers were father and mother living together. But, those who were given care by their father only had a 50% increased (ARR: 1.51, p value = 0.004) rate of DKA in comparison with those living with their father and mother together. Children without community health insurance had a 51% lower (ARR; 0.49, p value < 0.001) risk of DKA than children who had community health insurance. Diabetic children who did not have any diabetic education were at about 52% more (ARR: 1.52, p value = 0.011) rate of diabetic ketoacidosis than those who were given (table 3)..

Discussion
In children, type 1 diabetes mellitus is the most common form of diabetes [16] where diabetic ketoacidosis occurred most [4]. The current study attempted to determine the incidence and predictors of diabetic ketoacidosis among type 1 diabetic children. The current study found that the cumulative incidence and incidence density rate was 63.9% and 41.5 per 100 person-years respectively, which was consistent with many studies in sub-Saharan Africa [11,27].
In other way, this finding was much higher than the findings in developed countries [10, 15, [31][32][33][34][35][36]. This might be due to difference in study period, sample size and other sociodemographic characteristics observed between the previous and our population. Moreover, currently, the diagnostics and screening ability becomes advanced than the previous that might play significant role detaction rate.
Moreover, this study found that the cumulative incidence and incidence density rate amon chldren with type 1 DM was lower than compared with findings in Tikur Anbessa Specialized hospital (74.4%) [37] and than that of Tanzania (89.9%) [11], Benin teaching hospital (77.1%) [24], Romania (67%), Taiwan (65%) [36]. The lower cumulative incidence might be explained by the improved diabetes education and prevention programs that have been implemented over the last decades. Similarly, the variation in the cut of point for DKA might contribute for the difference [7,12,38].
In other way, the cumulative incidence and incidence density rate was higher than the studies done in Saudi Arabia (33.1%) [15], Swedish cohort study (1.4 episodes per 100 patient years [32], Hvidoere study (4/100 patient years) [33], United states (31.5%) [34], Germany (4.81/100 patient-years) [10], across five nations from England, Wales, the United State, Austria, and Germany (7.1%) [35], Systematic review in Sweden (14%), Canada (18.6%), Finland (22%) and Hungary (23%) [36]. This higher incidence density rate in the current study could be explained by the difference in economic development (well developed diabetic care setup) in those countries; the diagnostic modality was also different because DKA was diagnosed using PH value < 7.3; the variation of the cut off point for DKA, the difference in the age group which was considered 'children'could be much contributary for the lower incidencein those countries [10, 11, 14-16, 19, 26].
The current study also explored the potential predictors for the occurrence of DKA among diabetic children. In the present study, age at diagnosis of DM was identified as a significant predictor for DKA, age group 10-15 years was highly associated with DKA, which was consistent with many studies [10, 11, 14, 16, 17, 19, 20]. Greater personal responsibility, and less parental monitoring and endocrine changes leading to insulin resistance would be responsible for poor DKA control in this age group [10,11,19,39].
In addition, in this study, patients who missed insulin had an increased rate of DKA as compared to those with no omission of insulin. Other studies [9, 11, 16, 23] also confirm that omission of insulin was an important predictor of DKA. Omission of insulin causes hyperglycemia, intracellular starvation leading to stimulation of counter regulatory hormones which accelerates lipolysis and ketoacidosis [14,16,19,28,39,40].
DKA was significantly associated with primary caregiver other than father and mother together. Those children, who lived with their siblings only or fathers only, had increased risk of DKA. This was in line with other studies [14,16,19]. The reason may be parents (father and mother together) are probably more committed to their children with a chronic illness and ensure better compliance with medication. Children under care of their parents (both father and mother) also might enjoy a stable family structure that is supportive [10,11,14,17,19].
Higher insulin dose administration which is greater than1.2 U/kg/day was also increased the rate of DKA. This finding was also in good agreement with previous studies [11,14,19]. This seemed to be as a result of suppression of endogenous insulin which in turn decreases basal insulin [12]. But, those children on insulin dose less than 0.6 U/kg/day was found to be protected from risk of DKA. showed that diabetic children with no community health insurance had decreased rate of DKA. This might be due to many children who did not have community health insurance most probably afforded to cover the expense. In addition, patients with community health insurance might not have close diabetic follow up. And the status of community based health insurance could not be revised the next year and not well documented.
In the current study, diabetic education was significantly associated with the rate of DKA which was consistent with many studies [10, 11, 14-16, 19]. Patients who did not have any diabetic education had increased rate of DKA.

Strength of the Study
As to the researcher's knowledge, it was the first study to establish the incidence and predictors of DKA among established type 1 DM in the study areas and even in the country. It was a cohort study that can strongly suggest cause and effect relationship between variables. The study areas were wider referral hospitals with large catchment area.

Limitation of the Study
Since the study was a retrospective study, it had its own limitation associated with poor documentation. Many other socio-demographic, clinical and economic factors for DKA couldn't be assessed. There was some difficulty comparing findings due to limited data in Ethiopia and inaccessibility to other country study report. The finding was based on patient data on the card according to the physician diagnosis. Sometimes, it might lead to misdiagnosis of DKA or other precipitants or predictors. The variation in the cut off point for the diagnosis of DKA and the difference in the age group which was considered 'children' were the other challenges of the study; it might distort the result. In conclusion, this study revealed that a substantial number of patients had at least one episode of diabetic ketoacidosis and a higher rate of diabetic ketoacidosis. Age at diagnosis, primary caregiver, frequency of clinical visits, omission of insulin, baseline insulin dose, baseline insulin type, diabetic education, and community health insurance membership were identified as significant predictors of diabetic ketoacidosis. The most suffered age group was from 10-15 years.

Conclusions
This study revealed that a substantial number of patients had at least one episode of diabetic ketoacidosis and a higher rate of diabetic ketoacidosis. Age at diagnosis, primary caregiver, frequency of clinical visits, omission of insulin, baseline insulin dose, baseline insulin type, diabetic education, and community health insurance membership were identified as significant predictors of diabetic ketoacidosis. Early identification of patients at risk for this higher incidence of diabetic ketoacidosis and intervention programs targeting these predictors should be implemented.   NB: In the patient characteristics (variables) column, the numbers in brackets were number of episodes (events) in each category.