Development and validation of a risk assessment tool for uncontrolled type 2 diabetes among patients in South Karnataka, India

Introduction

Diabetes mellitus (DM) is one of the serious persistent challenges confronting healthcare systems globally. By 2040, the International Diabetes Federation projects that 642 million people will have Diabetes.1 Among 422 million people worldwide with diabetes, the majority live in low-income and middle-income countries (LMICs), and 1.5 million deaths are directly attributed to diabetes each year. Over the past few decades, there has been a consistent rise in both the incidence and prevalence of diabetes.2 The projections for DM are worrying. T2DM, which accounts for 90% of all cases of diabetes, has spread throughout the world and is now a leading cause of disability and death in even younger age groups. Previously thought to be a disease of the ‘wealthy’ Western countries, T2DM now affects people of all ages in many developing nations, including China and India, which has reached epidemic levels.3

In addition, the proportion of adults aged 20–79 years with undiagnosed diabetes was estimated to be 90% of the diabetic population in LMICs compared with 33% in high-income countries. Diabetes is a global health concern that poses a threat to the economy of developing nations, especially India. India’s large population and socioeconomic transitions have made it an epicentre of diabetes. Diabetes is more prevalent among Indians than other ethnic groups for several reasons, including a lack of awareness of the disease and its complications, inadequate access to healthcare, high costs, non-adherence to treatment and a lack of accountability.4

According to studies, more than 60% of newly diagnosed diabetics are unaware of their condition until a complication manifests, and one-third of all diabetics go undiagnosed. Because of the way that the disease progresses, early detection is becoming more and more important.5 It is well known that diabetes is a major risk factor for premature death. Long-term complications of diabetes are traditionally classified as microvascular or macrovascular. Macrovascular diabetes complications, such as coronary heart disease, stroke and peripheral vascular disease, as well as microvascular complications like end-stage renal disease, retinopathy and neuropathy, as well as lower-extremity amputations, account for a large portion of the diabetes burden.6 Lifestyle changes and pharmacological treatment can prevent or delay complications of T2DM. Therefore, identifying people who are at risk of developing uncontrolled diabetes in the future is critical to reducing the diabetes burden and its complications.7

To lessen this impact, India needs an effective, non-invasive screening tool for DM such as the diabetes risk assessment tools, which are convenient and practical for early screening and detection of diabetes to prevent diabetes-related morbidity, lower the cost of healthcare and enhance the quality of life.8 In recent years, multivariate risk scores have been developed to predict the risk of developing diabetes in healthy individuals. These risk scores are recommended in the most recent practice guidelines for preventing diabetes and are used in prevention programmes in some Western nations.9

Although many diabetic risk prediction models are available for use, a tool to predict risks for patients with uncontrolled T2DM is lacking. This study was proposed to develop and validate a non-invasive and self-administered risk assessment tool for patients with uncontrolled T2DM in southern India.

Objectives

• To develop a non-invasive and self-administered risk assessment tool for patients with uncontrolled T2DM in southern Karnataka, India.

• To validate the risk assessment tool for uncontrolled diabetes among adults.

Material and methods

Sampling

A total of 2335 patients are registered with the diabetic clinic of JSS Hospital, Mysuru, India. This list includes patients residing in other states of Karnataka. All were invited to participate in the study and 604 responded positively. 59 participants did not match the inclusion criteria and were excluded from this study. Finally, a total of 545 participants were included in the analysis (figure 1).

Request permission

Figure 1

Methodology flow chart. T2DM, type 2 diabetes mellitus.

Results

Out of the 545 study participants, 308 (56.5%) were males and 237 (43.5%) were females. The mean age of the study participants was 55.61±12.185 years. The prevalence of uncontrolled diabetes in our study was 59.9%. 91% of the participants were married, and 1% of them were divorced. Most study participants had completed middle school (19.8%) and high school (19.4%). Hindus (492, 90.3%) make up the majority of participants, followed by Muslims (36, 6.6%) and Christians (17, 3.1%).

In the current study, 269 participants (49.4%) were exercising regularly, compared with 279 participants (50.6%) who were not. 88.1% of participants regularly took their medication while 11.9% did not. 73.2% of the participants saw a doctor or underwent lab tests (RBS, FBS, PPBS, BP and HbA1c) at least once every 3 months, compared with 146 (26.8%) who did not. 319 (58.5%) of the study participants had a history of hypertension while 238 (43.7%) had a history of dyslipidaemia. The majority of study participants (56.5%) did not maintain a healthy diet while 83 (15.2%) were smokers and 99 (18.2%) consumed alcohol.

Tables 1–3 show the sociodemographic, personnel, medical and anthropometric factors associated with uncontrolled diabetes.

Education of the participants, physical activity, duration of diabetes, type of medication, diabetic diet, regular health check-ups, history of hypertension and dyslipidaemia, smoking history and alcohol consumption were factors significantly associated with uncontrolled diabetes (p<0.05). These factors were taken up for univariate regression analysis (table 4).

The variables selected for the multivariable logistic regression analysis were physical activity (OR 2.197, 95% CI 1.549 to 3.118), duration of diabetes (5–10 years, OR 0.645, 95% CI 0.449 to 0.925, >10 years, OR 2.024, 95% CI 1.304 to 3.141), diabetic diet (OR 1.655, 95% CI 1.134 to 2.416), regular health check-ups (OR 1.718, 95% CI 1.148 to 2.571), hypertension (OR 2.317, 95% CI 1.603 to 3.294), smoking (OR 3.253, 95% CI 1.830 to 5.785) and alcohol consumption (OR 2.567, 95% CI 1.553 to 4.243). The β-coefficient of these variables were multiplied by 10 and rounded to the nearest integer. Thus, the final risk score for predicting uncontrolled T2DM was developed (table 5).

For an optimum cut-off point of >13.50, the risk assessment model showed a moderate sensitivity of 71.3%, specificity of 61%, PPV of 73.2% and NPV of 58.3%. The ROC curve was plotted for the model with an AUC of 0.726, 95% CI 0.683 to 0.769 (figure 2).

Request permission

Discussion

Diabetes has already become a leading threat to public health globally in LMICs like India, where the burden has risen significantly in recent decades and will continue to rise in the coming decades. This could greatly influence morbidity and mortality associated with diabetes and, thus, the overall healthcare expenditure in India. A multifaceted approach is required to stop the spread of diabetes and its related complications, including early detection of diabetes, screening for its complications, providing optimal care at all levels of care for those who already have the disease and primary prevention of diabetes in those with pre-diabetes.3

This exploratory study was done among patients with T2DM to develop and validate a non-invasive and self-administered risk assessment tool for patients with uncontrolled T2DM. Our study reported the prevalence of uncontrolled diabetes to be 59.9%. A study done by Ganesh et al5 reported the prevalence of uncontrolled diabetes was 65.4% which was higher than our findings. In a study by Mahapatra et al,15 the prevalence of uncontrolled diabetes was 46.43% and in a study by Kanungo et al,16 the prevalence was 47% which was lesser when compared with our study findings. The difference in the prevalence of uncontrolled diabetes can be attributed to different study settings, distinct criteria for uncontrolled diabetes and different geographical locations.

Our study highlighted the factors affecting uncontrolled diabetes. Taking education as a factor, a significant difference was observed as 63 (19.3%) illiterates had uncontrolled diabetes and 23 (10.6%) had diabetes under control and among professionals, 11 (3.4%) had uncontrolled diabetes, contrary to 14 (6.4%) with controlled diabetes. Patients with a low education level have more complications and they are also more unaware of their hypoglycaemic symptoms causing severe complications.17

Regular exercise was thought to be crucial for managing DM as it regulates the blood glucose levels and actions of insulin. Our study revealed that out of 327 uncontrolled diabetics, 136 (41.6%) of them had a regular physical activity while 133 (61%) people who had a history of regular physical activity were found to have their glucose levels controlled. In a study done by Rasheed et al,18 it was found that 27% with uncontrolled diabetes and 40% of people with controlled diabetics had a regular physical activity, which was lesser compared with our findings.

Hypertension is an important risk factor in DM. In our study, 218 (66.4%) individuals with uncontrolled diabetes had associated hypertension and 101 (46.3%) among controlled diabetes had hypertension. In another study by Rasheed et al,18 in a controlled group of DM 30% of individuals had a history of hypertension while in the uncontrolled group hypertension was present in 47% of individuals and was lesser than our study findings.

In our study, we found out that 58.7% and 27.2% of subjects with a disease duration of 5–10 years and more than 10 years belong to the uncontrolled diabetic groups, respectively, and we can see that as the duration of the disease increases the diabetes is becoming uncontrolled. Another study done by Badedi et al found that individuals with lower HbA1c followed the diabetes diet recommendations and took their medications as prescribed by doctors. On the other hand, patients with poor glycaemic control frequently took multiple medications and had diabetes for a longer period. A longer duration of diabetes and polypharmacy are known to be significantly linked to poor glycaemic control.19 Early detection and effective control of diabetes are the main management priorities. According to studies, diabetes control deteriorates as the disease duration exceeds 10 years.5

Among the subjects in our study, 74.3% followed a regular diabetic diet and had their diabetes status under control and 36.4% of uncontrolled diabetics were not following a proper diabetic diet. Likewise, 79.4% of diabetics who had regular health check-ups within 3 months had their glycaemic level under control and 30.9% of uncontrolled diabetics had their check-ups only after 3 months. The American Diabetes Association defines self-dietary management as the essential first step in providing diabetics with the knowledge and skills they need about treatment, nutritional considerations, medications and complications.20

In our study among the diabetic subjects, 15.2% were smokers and 18.2% were alcoholics. We observed diabetics who smoked and consumed alcohol, there was a higher rate of uncontrolled diabetes than their counterparts. A prospective study of cigarette smoking, alcohol use and risk of diabetes in men, done by Rimm et al, observed lower rates of diabetes among men who did not smoke and among men who consumed alcohol. After simultaneously controlling for each other and other known predictors of non-insulin-dependent DM, the risk of drinking alcohol and smoking became more pronounced.21

In our study, education of the participants, physical activity, duration of diabetes, type of medication, diabetic diet, regular health check-ups, history of hypertension and dyslipidaemia, smoking history and alcohol consumption were factors that had a significant association with uncontrolled diabetes (p<0.05). These factors were taken up for univariate regression analysis. After the analysis, variables such as physical activity, duration of diabetes, diabetic diet, regular health check-ups, history of hypertension and dyslipidaemia, smoking history and alcohol consumption were taken up for multivariable regression followed by the development of final risk score for predicting uncontrolled diabetes with a cut-off point of >13.50 with a sensitivity of 71.3% and specificity of 61%, PPV 73.2% and NPV of 58.3%. The ROC curve was plotted for the predictive model with an AUC of 0.726.

In a study by Lindström and Tuomilehto,22 to develop a risk score for diabetes—age, BMI, waist circumference, history of antihypertensive drug treatment and high blood glucose, physical activity, and daily consumption of fruits, berries or vegetables were selected as categorical variables. The Diabetes Risk Score value varied from 0 to 20. To predict drug-treated diabetes, the score value ≥9 had a sensitivity of 0.78 and 0.81, specificity of 0.77 and 0.76, and PPVs of 0.13 and 0.05 in the 1987 and 1992 cohorts, respectively. The Diabetes Risk Score was designed to be a screening tool for identifying high-risk subjects and increasing awareness of the modifiable risk factors and healthy lifestyles.

Later in our study, the validity of the risk score was assessed on data collected from patients with T2DM residing in urban Mysuru. The risk score was applied to this data which showed a slightly decreased AUC value of 0.700, 95% CI 0.653 to 0.746. The cut-off point based on maximum sensitivity and specificity was the same at 13.50. The sensitivity, specificity, PPV and NPV at this cut-off point were 70.2%, 62%, 74.8% and 56.4%, respectively. The study by Lindström and Tuomilehto22 analysed the performance of the Diabetes Risk Score cross-sectionally in identifying subjects who had either fasting or 2-hour glucose levels exceeding the threshold of diabetes. The ROC curves (not shown) indicated the good performance of the Diabetes Risk Score also in the cross-sectional setting (AUC=0.80 for both surveys). For cut point Diabetes Risk Score of ≥9, sensitivity was 0.77 (95% CI 0.66 to 0.85) and 0.76 (95% CI 0.67 to 0.83), specificity was 0.66 (95% CI 0.64 to 0.68) and 0.68 (95% CI 0.66 to 0.70), PPV was 0.07 (95% CI 0.06 to 0.09) and 0.12 (95% CI 0.10 to 0.15), and NPV (the probability of not having diabetic glucose levels if Diabetes Risk Score was <9) was 0.99 (95% CI 0.98 to 0.99) and 0.98 (95% CI 0.97 to 0.99) in the 1987 and 1992 oral glucose tolerance tests, respectively.

Another study was done by Mohan et al23 to develop and validate a simplified Indian Diabetes Risk Score (IDRS) for detecting undiagnosed diabetes in India. IDRS used four risk factors: age, abdominal obesity, family history of diabetes and physical activity. Beta coefficients were derived from a multiple logistic regression analysis using undiagnosed diabetes as the dependent variable. The beta coefficients were modified to obtain a maximum possible score of 100. ROC curves were constructed to identify the optimum value of IDRS for detecting diabetes by WHO consulting group criteria. The AUC for ROC was 0.698 (95% CI 0.663 to 0.733). An IDRS value ≥60 had the optimum sensitivity (72.5%) and specificity (60.1%) for determining undiagnosed diabetes with a PPV of 17.0%, NPV of 95.1% and accuracy of 61.3%.

A study was done by Adhikari et al24 to validate the MDRF-IDRS in a south Indian population in coastal Karnataka, in which IDRS score of ≥60 had the best sensitivity (62.2%) and specificity (73.7%) for detecting undiagnosed diabetes in this community. The MDRF-IDRS was calculated using age, family history of diabetes, physical activity and waist measurement. ROC curves were constructed to identify the optimum value (≥60%) of IDRS for determining diabetes as diagnosed using WHO consulting group criteria.

Our study was subject to various limitations. First, its base population may not be entirely representative of South Indians overall, as it is based on people living in Southern Karnataka. Second, a small sample size was employed in the data collection process. It is necessary to conduct more research with bigger sample sizes. Finally, the results show low sensitivity, specificity, PPV and NPV. More research must be done with bigger sample sizes and improved sensitivity, specificity, PPV and NPV.

Conclusion

This study developed ‘Diabetes Care’ which is a web-based simple, non-invasive and inexpensive tool for identifying individuals at risk of developing uncontrolled T2DM in the future. The website also provides detailed information regarding the disease and lifestyle modification, risk prediction and prognosis of the disease for the affected and precautionary management of its acute and chronic complications. This tool is specifically designed for the population of southern Karnataka.