Frequency of Diabetic Ketoacidosis and Its Determinants Among Pediatric Diabetes Mellitus Patients in Northwest Ethiopia

Introduction

Diabetes mellitus (DM) describes a group of metabolic disorders characterized by increased blood glucose concentration; and the main threat to human health in the 21st century.¹

In 2017, there were 425 million DM sufferers globally, and 14.2 million of them were in Africa, 77% were in developing countries, and 5.2% of them were in Ethiopia.²

Diabetic ketoacidosis (DKA) is an acute life-threatening complication of diabetes mellitus characterized by the triad of hyperglycemia, acidosis, and ketosis that occurs in the presence of very low levels of effective insulin action.^1,3–7 It is one of the common emergencies among patients with diabetes mellitus and is associated with considerable morbidity and mortality.⁸

It is still the leading cause of death; and mortality risk is substantially increased in patients with chronically poor glycemic control and frequent DKA.³ Frequent DKA is a serious and common health problem.⁹

Globally, around 65,000 children aged under 15 years develop type 1 diabetes each year and up to 80% present with DKA.¹⁰ In Africa, the mortality rate due to DKA is unacceptably higher.⁸ In developing countries, it ranges from 6% to 24%. In Ethiopia, the risk of dying from DKA is higher due to poor medical services in the country.^4,6,11–13

In addition to the risk of mortality, frequent DKA has a major impact on the quality of life of the individuals and their families, accelerated micro-vascular complications, and costs for families and health care.^4,9,14

According to studies conducted in different parts of the world, being female, those diagnosed at the younger age, family history of DM, missing an insulin dose, infection, and psychiatric comorbidities were the common determinants of frequent DKA.^{9,10,15–20,}

The Federal Ministry of Health (FMOH) of Ethiopia developed a National Strategic Action plan (NSAP) for four priority Non-Communicable Diseases (NCD), including DM, by focusing on integration to a three-tier health care system. However, due to limited resources, the country mainly focuses on infectious diseases and pays little attention to non-communicable diseases.² Besides this little attention, the control of diabetes and the prevention of ketoacidosis were hampered by socioeconomic factors; particularly the cost and unreliability of insulin supplies.²¹ As a result, DKA remains a common public health issue, and a major child killer in sub-Saharan African countries, including Ethiopia.¹⁹ Moreover, there is limited data on the incidence of DKA and its determinants in Amhara Regional State, Ethiopia. Therefore, this study assessed the frequency of DKA and its determinants among pediatric diabetes mellitus patients in public hospitals of Amhara Regional State, northwest Ethiopia.

Materials and Methods

Study Design and Setting

An institutional-based retrospective follow-up study was conducted in northwest Ethiopia from September 2015 to February 2018.

The study was conducted at the University of Gondar Specialized and Referral Hospital (UoGSRH) and Felege Hiwot Specialized and Referral Hospital (FHSRH) of Amhara Regional State, northwest Ethiopia. Both hospitals are public hospitals and are expected to serve five million people. Pediatric emergency, inpatient wards, outpatient follow-up, and an Out Patient Department (OPD) for chronic illness are served at both hospitals. A total of 322 and 301 pediatric DM patients were followed at UoGSRH and FHSRH from 2015 to 2018, respectively.

Study Participants

Pediatric DM patients aged <18 years and who started a follow-up from 2015 to 2018 and attend the follow-up clinic for at least two years were included in the study. The option of Signori normalized Poisson proposed methods were used to determine the sample size.^21,22 Based on the signori normalized Poisson proposed approach, using a 95% confidence level and 90% power, the final estimated sample size used for this study was 389. Pediatric DM patients with unknown dates of diagnosis and follow-up were excluded from the study.

Data Collection Procedures

Secondary data were retrieved from patient’s charts using a data extraction form. The extraction form was developed after reviewing the patients’ chart and based on the record of variables in the chart. The data were collected by six well-trained nurses with BScs. Two supervisors who have master’s degrees in public health supervised the data collection processes. The principal investigator also monitored the data collection process.

Statistical Analysis

Epi-data version 4.6 was used for data entry. For nutritional measurements, WHO Anthro version 3.2.2 for children under 5 years old and Anthro plus version 1.0.4 for adolescents were used to categorize nutritional variables. Outliers were identified for continuous variables using a box, scatter, and normal Q-Q plots. For those variables which had outliers, a model was fitted with and without outliers independently, estimates and standard error were estimated and compared, and noted that they did not vary significantly. Therefore, the analysis was made using the original data. Multi-collinearity was checked using variance inflation factor (VIF), that had a global VIF of 1.22 and a maximum of 4.7 which were <10, indicating that multi-collinearity was not a problem.

To determine the association of covariates with the frequency of DKA, the Poisson regression model was fitted. The Poisson regression model assumes that the mean and variance of count variable are equal, known as equidispersion. The other assumption is the skewed distribution of the outcome variable. The distribution of frequency of DKA in this study was right-skewed and the variance (1.96) was greater than the mean (1.01), indicating that the assumption of the Poisson regression model, is violated. The violation of such assumptions helps us to consider methods of dispersion tests. Overdispersion of the data was confirmed by estimating the ratio of residual deviance to the degree of freedom and dispersion test under the “AER” package. The ratio was greater than one (1.58) and the dispersion test was also significant (P-value = 0.00015) which indicated there is overdispersion. Therefore, to account for this overdispersed data, the negative binomial model was considered.

A zero-inflated negative binomial model was also fitted to check whether the data had inflated zeros or not, and a Vuong test was used to determine whether the zero-inflated negative binomial regression model is statistically preferred over the negative binomial regression model. Based on the test, the negative binomial model was the preferred model for the data.

Akaike’s Information Criteria (AIC) was also used to select the best fitted model and a model with smaller AIC was preferred. Variables with a p-value of less than 0.2 in the bi-variable analysis were selected to fit the model in the multivariable analysis. P-values less than 0.05 and (adjusted incidence rate ratio) AIRR with 95% CI was used to declare statistically significant predictors in multivariable analysis.

Results

Socio-Demographic Characteristics of Study Participants

From the total, 389 patients, more than half, 215 (55.3%), of them were males. The average age of the patients was 11 (±4 SD) years. Regarding residence, half, 198 (50.9%), of the patients were from rural areas.

Characteristics of Patients Based on Clinical Variables

Almost all, 382 (98.2%), DM patients had poly symptoms (polyuria, polydipsia, polyphagia) and 109 (28%) had weight loss. About 246 (63%) of the patients had comorbidities with infection (upper respiratory tract infection, pneumonia, malaria, TB, urinary tract infection) and the rest, 18% of them had other comorbidities (trauma, heart diseases, psychiatric illness, asthma, anemia, and skin infections).

One-seventh, 58 (14.9%), of the patients had upper respiratory tract infections, 61 (15.7%) pneumonia, 36 (9.3%) malaria, and the rest of them had other types of infections such as TB, intestinal parasitosis, tinea capitis, scabies, and urinary tract infections. About 26 (6.7%) of the patients had foot ulcers and 41 (10.5%) of them had blurring vision. During the follow up, 79 (20.3%) of the patients had a history of treatment discontinuation. About 111 (28.5%) of the patients had a history of hypoglycemia; and almost all, 368 (94.6%) of them had DKA at diagnosis. Regarding to electrolytes, 124 (31.9%) the patients had normal, while 46 (11.8%) of them had low and 22 (5.7%) had a high level of potassium. Forty-five (11.6%) of the patients had normal, while 125 (32.1%) and 24 (6.2%) of them had low and high levels of sodium, respectively (Table 1).

Table 1 Characteristics of Pediatric DM Patients Based on the Baseline, Follow Up and Nutritional Variables at Public Hospitals of Amhara Regional State, Northwest Ethiopia, from 2015 to 2018 (N = 389)

Pediatric DM Patient Characteristics Based on Measurement Variables

Regarding the baseline measurements, the mean random blood sugar of the patients was 500.8 (±100.9 SD) mg/dl and fasting blood sugar was 272.6 (±80.8 SD) mg/dl. The mean weight was 22.8 (±9.8SD) kg and the creatinine was 0.8 (±0.3 SD) mg/dl (Table 2).

Table 2 Characteristics of Pediatric DM Patients Based on Baseline Measurement Variables at Public Hospitals of Amhara Regional State, Northwest Ethiopia, from 2015 to 2018

The Severity of Diabetic Ketoacidosis at DM Diagnosis

Regarding the severity of DKA, 157 (42.7%) of the patients had mild while 101 (27.4%) and 110 (29.9%) of them had moderate and severe DKA at diagnosis, respectively (Figure 1).

Figure 1 Characteristics of pediatric diabetes mellitus patients based on the severity of diabetic ketoacidosis at Public Hospitals of Amhara Regional State, northwest Ethiopia, from 2015 to 2018.

The glycemic control level meant 132 (33.9%) were well, while 104 (26.7%) and 153 (39.3%) had fairly and poorly controlled glycemic levels.

Precipitating Factors of DKA During Follow-Up

During the follow up, 109 (28%) DKAs were precipitated by infection (5.4% CAP, 10.3% URTI and 2.6% malaria) and 79 (20.3%) of the patients had a history of treatment discontinuation and the rest had other precipitants such as trauma, congenital heart diseases, Down’s syndrome, eating sweet foods/candy and poor injection techniques.

Frequency of DKA

Nearly half, 188 (48.3%) of the DM patients had at least one bout of DKA during the follow up. Among those patients who had DKA, 102 (54.3%) of them were from Felege Hiwot Specialized and Referral Hospital. The minimum and maximum frequencies of DKA were 0 and 6 per individual within the two years of follow-up, respectively. The mean frequency of DKA was 1.01 (Figure 2).

Figure 2 Frequency of diabetic ketoacidosis among pediatric diabetes mellitus patients at Public Hospitals of Amhara Regional State, northwest Ethiopia, from 2015 to 2018.

Determinants of Frequency of DKA

In the multi-variable analysis, infection, treatment discontinuation, sodium level, heart disease and the dose of treatment were statistically significant variables. The expected mean frequency of DKA was 0.06 when the effect of other covariates were constant. Holding other variables constant, the incidence rate of diabetic acidosis was increased by 41% among DM patients with infection comorbidities (upper respiratory tract infection, pneumonia, malaria, TB, urinary tract infection) compared with those who do not have an infection (AIRR = 1.41, 95% CI = 1.05–2.14). Patients with heart disease as their comorbidity were found to be 4.1 times more likely to have a greater frequency of DKA (AIRR = 4.1, 95% CI = 1.17–14.68) compared to those without heart disease and holding other variables constant.

The incidence rate of DKA was 2.91 times (AIRR = 2.91, 95% CI = 2.02–4.22) greater among DM patients who discontinued their treatment compared to patients who had taken their treatment properly, keeping other variables constant.

Regarding electrolytes, patients with low levels of sodium increased the incidence rate of DKA by 88% (AIRR = 1.88, 95% CI = 1.22–2.89). Increasing the dose of treatment by one international unit (IU) decreased the incidence rate of DKA by 4% (AIRR = 0.96, 95% CI = 0.94–0.97) (Table 3).

Table 3 Bi-Variable and Multi-Variable Negative Binomial Model on Determinants of Frequency of DKA Among Pediatric DM Patients at Public Hospitals of Amhara Regional State, North West Ethiopia, from 2015 to 2018 (N = 389)

Discussion

Acute complications of DM are significant contributors to mortality, and frequent DKA has a major impact on the quality of life of the individuals and their families, accelerating micro-vascular complications and costs for families and health care. This study tried to assess the frequency of DKA and identify factors associated with the frequency of DKA among pediatric diabetes mellitus patients.

The present study showed that the average frequency of DKA is 1.01. The minimum and maximum frequencies of DKA were 0 and 6, respectively. This finding is lower than the multicenter study conducted at University Hospital Birmingham, Birmingham Children’s Hospital, City Hospital, Birmingham Heartlands Hospital, and New Cross Hospital, where the average frequency of DKA was 1.1 and 3.2, respectively.²³ Another study in Birmingham, Alabama community teaching hospitals showed that the average frequency of DKA was 2.5.²² This variation may be due to the short duration of follow-up used in this study (2 years) compared to the long-duration (15 years) in Birmingham.

This finding revealed a higher average frequency of DKA than the findings in Iraq, where the average was 0.9.²⁴ A possible reason may be that they used a shorter duration of study, which was one year, than this study, which was two years. A relatively long follow up will identify the extent of the frequency of DKA more accurately than a shorter-duration. Another possible reason may be that the country, Ethiopia, is less developed than Iraq, and the patients may be faced with frequent DKA for various socio-economic reasons.

In terms of factors associated with the frequency of DKA, comorbid infection, heart disease, treatment discontinuation, low levels of sodium and dose of treatment were statistically significant factors. In this study, the proportion of patients with infection was 63.2%, and having comorbidity with infection (upper respiratory tract infection, pneumonia, malaria, TB, urinary tract infection) increases the incidence of DKA, which is supported by the studies conducted in Birmingham, Egypt, Tigray, Tikur Anbessa Specialized Hospital, Jimma University Specialized Hospital, and Kovai Medical Centre.^{4,11,23,25–27} This may be because of DM results in immune dysfunction and makes the patients susceptible to infection, and increasing its virulence in the hyperglycemic environment and finally precipitating complications like DKA.^28,29

Based on this study, having a comorbidity with heart disease significantly increased the frequency of DKA. However, there were no comparable results with this finding. This may be to the fact that other studies may not include heart diseases as associated factor of DKA or might not be significantly associated with the frequency of DKA. The possible reason may be having heart diseases result in multiple organ dysfunctions.

This study found that treatment discontinuation results in an increasing for the incidence of DKA by three fold compared to their counterpart. This finding was comparable with the studies conducted in a tertiary care hospital of Dhaka, Bangladesh, and the Soroka University Medical Center of Israel.^23,30,31 This might be because when the patients discontinue their treatment, they may be faced insulin deficiency, increased insulin counter-regulatory hormones, and peripheral insulin resistance lead to hyperglycemia, dehydration, ketosis, and electrolyte imbalance, which underlie the pathophysiology of DKA.³² In this study, patients discontinue their treatment for various reasons, including the cost of treatment or discomfort with the daily injection. Patients with a low level of sodium at baseline had a positive significant effect the incidence of DKA. This may be due to the nutritional status of the patient and an insulin-induced low level of sodium.³³

The finding of this study showed that a decreasing dose of treatment increased the incidence rate of DKA. In contrast, a study conducted in Denver, Colorado metropolitan area showed that a high dose of treatment increased the rate of DKA.³⁴ The possible reason for the increased rate of DKA as a result of a low dose of treatment may be that lowering the dose of treatment; results in a low production of insulin in the body, and the insulin antagonist hormones will also increase. Finally, the level of glucose will be increased and that leads to an increased hyperglycemic environment and then DKA.

Limitation of the Study

Since the study was based on secondary data, some important risk factors were not available and unable to analyze them.

Conclusion

The findings of this study indicated the average frequency of DKA during follow-up is higher than the expected frequency, which is preferably zero. Comorbidity with an infection, having heart diseases, lowering the dose of treatment, having a low sodium level, and having a history of treatment discontinuation were the factors that increase the frequency of DKA among pediatric DM patients. Parents and health care providers shall emphasize to pediatric diabetic patients to reduce the frequency of DKA, give special attention to infection prevention and the use of treatments for pediatric DM patients. The Federal Ministry of Health has to set priorities and intervention strategies like providing training for health extension workers and awakening them to work with the patients to prevent DKA.