Derivation and validation of diabetes risk score for urban Asian Indians

doi:10.1016/j.diabres.2005.02.016

Diabetes Research and Clinical Practice

Volume 70, Issue 1, October 2005, Pages 63-70

https://doi.org/10.1016/j.diabres.2005.02.016 Get rights and content

Abstract

Objective:

Simple risk scores for identifying people with undiagnosed diabetes have been developed, mostly in Caucasian groups. This may not be suitable for Asian Indians, therefore this study was undertaken to develop and validate a simple diabetes risk score in an urban Asian Indian population with a high prevalence of diabetes. We also tested whether this score was applicable to South Asian migrants living in a different cultural context.

Research design and methods:

A population based Cohort of 10,003 participants aged ≥20 years was divided into two equal halves (Cohorts 1 and 2), after excluding people with known diabetes. Cohort 1 (n = 4993) was used to derive the risk score. Validation of the score was performed in the other half of the survey population (Cohort 2) (n = 5010). The validation was also done in a separate survey population in Chennai, India (Cohort 3) (n = 2002) (diagnosis of diabetes was based on OGTT) and in the South Asian Cohort of the 1999 Health Survey for England (n = 676) (fasting glucose value ≥ 7 mmol/l and HbA1c ≥ 6.5% were used for diagnosis). A logistic regression model was used to compute the β coefficients for risk factors. The risk score value was derived from a receiver operating characteristic curve.

Results:

The significant risk factors included in the risk score were age, BMI, waist circumference, family history of diabetes and sedentary physical activity. A risk score value of >21 gave a sensitivity, specificity, positive predictive value and negative predictive value of 76.6%, 59.9%, 9.4% and 97.9% in Cohort 1, 72.4%, 59%, 8.3% and 97.6% in Cohort 2 and 73.7%, 61.0%, 12.2% and 96.9% in Cohort 3, respectively. The higher distribution of risk factors in the UK Cohort means that at the same cut point the score was much more sensitive but also less specific. (sensitivity 92.2%, specificity 25.7%, positive predictive value of 21.6% and negative predictive value of 93.7%).

Conclusions:

A diabetes risk score involving simple non-biochemical measurements was developed and validated in a native Asian Indian population. This easily applicable simple score could play an important role as the first step in the process of identifying individuals with an increased likelihood of having prevalent but undiagnosed diabetes. The different distribution of risk factors with the migrant Asian Indians living in England and the different relationship between sensitivity and specificity for the same score demonstrate that risk scores and cut-points developed and tested even within one ethnic group cannot be generalized to individuals of the same ethnic group living in a different cultural setting where the distribution of risk factors for diabetes is different.

Introduction

The prevalence of diabetes is increasing worldwide and is a serious public health problem in many countries. The cost of treating diabetes and, in particular, dealing with its complications, is a major concern especially in developing countries such as the Indian sub-continent where the prevalence of diabetes is high [1]. As a large proportion of the people with prevalent Type 2 diabetes are undetected and many have diabetic complications at the time of clinical diagnosis [2], it has been suggested that screening could be one approach to reduce the burden of morbidity attributable to diabetes. However universal screening of all adults in the population is not practical nor recommended [3].

One critical uncertainty in the debate about whether or not to screen for diabetes, is the most appropriate method for identifying individuals with undiagnosed diseases. A common strategy is to use simple methods to identify a population of individuals at increased risk of having undiagnosed diabetes who can then be invited to attend for more definitive testing.

A number of different groups have developed and evaluated simple risk scores for predicting prevalent but undiagnosed diabetes. Some of these have required people to complete a questionnaire, whereas others have utilized information already available to health care practitioners. The common characteristic of these questionnaires is that they all utilize non-blood based risk factors and are used to identify target groups who could be invited for biochemical measurements [4], [5], [6], [7], [8]. Lindstrom and Tuomilehto had used the score to predict incident cases [5]. These risk factors questionnaires have invariably been developed in Caucasian groups and may not be generalizable to other populations where the distribution of risk factors and their associations with prevalent undiagnosed diabetes differ. Only one previous study has tested the validity of a score that was originally derived and tested in a Caucasian population [4] in a different ethnic minority population in the UK [8]. However, no study has previously developed and tested a risk score specifically in a South Asian population.

Therefore the aims of this study were (1) to develop and test the validity of a simple diabetes risk score in the South Asian Indian population and (2) to assess its performance in a migrant Indian population in the UK.

Section snippets

Materials and methods

Four different Cohorts were used in this study. Cohorts 1 and 2 were two halves of the National Urban Diabetes Survey (NUDS) conducted in the year 2000 in six major cities of India [9]. Adults aged ≥20 years from an urban population were randomly selected for the study. Subjects from all socio-economic strata were included. The total number of subjects screened was 11,186 (men 5258, women 5928). A Cohort of 10,003 participants was divided into two equal halves (Cohorts 1 and 2), after excluding

Results

Table 1 shows the characteristics of all the four study Cohorts. Table 2 shows the β coefficients and the assigned scores for each risk factor included in the model. It also shows the prevalence of the risk factors in Cohort 1. The prevalence of risk factors was similar in Cohorts 1–3. The maximum risk score was 42. Increasing age, BMI ≥25 kg/m², waist circumference (≥85 cm for men, ≥80 cm for women), positive family history of diabetes and low levels of physical activity were significantly

Discussion

This study has shown that a risk score developed specifically for an Asian Indian population was valid for the population living in India. The intensity of risk conferred for diabetes varies between different ethnic groups, therefore it becomes necessary to develop risk scores specific for different groups. A low risk threshold for age, BMI, upper body adiposity and a high propensity of familial aggregation of diabetes are well known in Indians [13]. Moreover the impact of socioeconomic factors

Acknowledgements

We thank Dr. Mrs. K. Satyavani for helping in the data analysis and Ms. K. Subhaa for the secretarial assistance.

References (17)

A. Ramachandran et al.
Low risk threshold for acquired diabetogenic factors in Asian Indians
Diab. Res. Clin. Prac.
(2004)
World Diabetes Foundation, Executive Summary, Diabetes Atlas 2,...
M.I. Harris et al.
Onset of NIDDM occurs at least 4–7 years before clinical diagnosis
Diab. Care
(1992)
N.J. Wareham et al.
Should we screen for type 2 diabetes? Evaluation against National screening committee criteria
BMJ
(2001)
S.J. Griffin et al.
Diabetes risk score: towards earlier detection of Type 2 diabetes in general practice
Diab. Metab. Rev.
(2000)
J. Lindstrom et al.
The diabetes risk score: a practical tool to predict type 2 diabetes risk
Diab. Care
(2003)
C. Glumer et al.
A Danish diabetes risk score for targeted screening: the inter99 study
Diab. Care
(2004)
B.P. Tabaei et al.
A multivariate logistic regression equation to screen for diabetes: development and validation
Diab. Care
(2002)

There are more references available in the full text version of this article.

Cited by (122)

External validation of non-invasive diabetes score in a 15-year prospective study
2022, American Journal of the Medical Sciences
A novel scoring system called Non-invasive Diabetes Score (NDS) was developed. The model showed prominent discrimination and calibration in the original study population. However, before a new model could be adopted in clinical practice and acquire widespread use, it is necessary to confirm that it also performs well in external validations in different settings of people. The aim of this study was to investigate whether the novel user-friendly score predicting diabetes mellitus (DM) could have satisfying performance in predicting DM in Southwest China in a 15-year prospective cohort study.
This prospective cohort study was carried out based on a general Chinese population of 711 individuals from 1992 to 2007. We excluded 24 of them at baseline because they were diabetics. The end point was DM, and the risk was calculated using the model formula.
During a follow-up of 15 years, 74 (10.77%) patients reached the end point. Evaluation of this model in our cohort, with Harrell's C-index of 0.662 (95% CI: 0.600-0.723) for the whole cohort and 0.695 (95% CI: 0.635-0.756) in sensitivity analysis, indicated the possibly helpful discrimination. The calibration capability in our cohort was useful that the observed incidence of diabetes mellitus was near the predicted.
Our external validation suggested NDS had possibly helpful discrimination and satisfying calibration for predicting DM during 15-year follow-up.
Validation of the Finnish Type 2 Diabetes Risk Score (FINDRISC) with the OGTT in Health Care Practices in Europe
2021, Diabetes Research and Clinical Practice
/hypothesis.
To determine the best cut-off threshold value of the Finnish Diabetes Risk Score (FINDRISC) for the detection of diabetes and non-diabetic hyperglycaemia in people 35 years or older at primary health care settings in Europe.
Cross-sectional study in 11,444 adults from primary health care centres using community and opportunistic screening approaches. All participants completed the FINDRISC questionnaire and underwent a 2-hour oral glucose tolerance test (OGTT). The FINDRISC performance was assessed by the area under the curve (AUC) using receiver operating characteristics (ROC) analysis. The sensitivity, specificity, Youdeńs index, positive and negative prediction values for different FINDRISC cut-offs were calculated.
The optimal FINDRISC value for detecting both diabetes or glucose impairment in the community - screened sample was 14 point with the associated AUC 0.75,5 (95 %CI 0.73,7–0.77,3). The optimal score in the opportunistic screening sample was 16 with the associated AUC only 0.60,4 (95% CI 0.56, 4–0.64, 4).
The FINDRISC is a non-invasive tool useful for detecting people with unknown diabetes and glucose impairment in people visiting primary health centres in Europe.
Diagnostic accuracy of the Finnish Diabetes Risk Score (FINDRISC) for undiagnosed T2DM in Peruvian population
2018, Primary Care Diabetes
To assess the diagnostic accuracy of the Finnish Diabetes Risk Score (FINDRISC) for undiagnosed T2DM and to compare its performance with the Latin-American FINDRISC (LA-FINDRISC) and the Peruvian Risk Score.
A population-based study was conducted. T2DM and undiagnosed T2DM were defined using oral glucose tolerance test (OGTT). Risk scores assessed were FINDRISC, LA-FINDRISC and Peruvian Risk Score. Diagnostic accuracy of risk scores was estimated using the c-statistic and the area under the ROC curve (aROC). A simplified version of FINDRISC was also derived.
Data from 1609 individuals, mean age 48.2 (SD: 10.6), 810 (50.3%) women, were collected. A total of 176 (11.0%; 95%CI: 9.4%–12.5%) were classified as having T2DM, and 71 (4.7%; 95%CI: 3.7%–5.8%) were classified as having undiagnosed T2DM. Diagnostic accuracy of the FINDRISC (aROC = 0.69), LA-FINDRISC (aROC = 0.68), and Peruvian Risk Score (aROC = 0.64) was similar (p = 0.15). The simplified FINDRISC, with 4 variables, had a slightly better performance (aROC = 0.71) than the other scores.
The performance of FINDRISC, LA-FINDRISC and Peruvian Risk Score for undiagnosed T2DM was similar. A simplified FINDRISC can perform as well or better for undiagnosed T2DM. The FINDRISC may be useful to detect cases of undiagnosed T2DM in resource-constrained settings.
Screening models for undiagnosed diabetes in Mexican adults using clinical and self-reported information
2018, Endocrinologia, Diabetes y Nutricion
Citation Excerpt :
For this reason, several screening models have been developed to account for the ethnical and cultural differences between countries and/or regions. These include Kuwait,12 Oman,13 The Netherlands,14,15 USA,16,17 India,18–20 China,21–23 Brazil,24 Egypt,25 Thailand,38,42 the Middle East and North Africa.26 Additionally, Vistisen et al.27 developed a global screening model considering the differences between most geographical regions (i.e. Asia, Greenland, Europe, Australia, Middle East, Eastern Mediterranean, Africa and North America).
Prevalence of diabetes in Mexico has constantly increased since 1993. Since type 2 diabetes may remain undiagnosed for many years, identification of subjects at high risk of diabetes is very important to reduce its impact and to prevent its associated complications.
To develop easily implementable screening models to identify subjects with undiagnosed diabetes based on the characteristics of Mexican adults.
Screening models were developed using datasets from the 2006 and 2012 National Health and Nutrition Surveys (NHNS). Variables used to develop the multivariate logistic regression models were selected using a backward stepwise procedure. Final models were validated using data from the 2000 National Health Survey (NHS).
The model based on the 2006 NHNS included age, waist circumference, and systolic blood pressure as explanatory variables, while the model based on the 2012 NHNS included age, waist circumference, height, and family history of diabetes. The sensitivity and specificity values obtained from the external validation procedure were 0.74 and 0.62 (2006 NHNS model) and 0.76 and 0.55 (2012 NHNS model) respectively.
Both models were equally capable of identifying subjects with undiagnosed diabetes (∼75%), and performed satisfactorily when compared to other models developed for other regions or countries.
En México, la prevalencia de diabetes se ha incrementado consistentemente desde 1993. Dado que la diabetes tipo 2 puede mantenerse sin diagnóstico por muchos años, es de suma importancia la identificación temprana de los sujetos con alto riesgo de tener la enfermedad con la finalidad de reducir su impacto y prevenir así las complicaciones asociadas.
Desarrollar mecanismos de fácil implementación para la detección de sujetos con diabetes no diagnosticada con base en las características de la población adulta mexicana.
Los modelos fueron desarrollados usando datos de las Encuestas Nacionales de Salud y Nutrición (NHNS) 2006 y 2012. Las variables utilizadas para desarrollar los modelos de regresión logística multivariada fueron seleccionadas mediante un procedimiento de pasos hacia atrás. Los modelos finales se validaron usando datos de la Encuesta Nacional de Salud (NHS) 2000.
El modelo obtenido de la NHNS 2006 incluyó edad, circunferencia de cintura y presión arterial sistólica como variables explicativas, mientras que el modelo NHNS 2012 incluyó edad, circunferencia de cintura, estatura e historia familiar de diabetes. La sensibilidad y la especificidad obtenidas del proceso de validación externo fueron 0,74, 0,62 (modelo NHNS 2006) y 0,76, 0,55 (modelo NHNS 2012), respectivamente.
Ambos modelos desarrollados fueron igualmente capaces de identificar sujetos con diabetes no diagnosticada (∼75%), y mostraron un desempeño satisfactorio en comparación con otros modelos desarrollados para otras regiones y países.
Gestational diabetes mellitus risk score: A practical tool to predict gestational diabetes mellitus risk in Tanzania
2018, Diabetes Research and Clinical Practice
Citation Excerpt :
To assess the performance of the prediction model, the receiver operating characteristic (ROC) curve for GDM risk score was built by plotting the sensitivity on the y-axis and the false positive rate (1- specificity) on the x-axis. Moreover, the area under the curve (AUC) was used to measure the power to discriminate high-risk from low-risk individuals [14,26]. The more accurate discriminating the test, the steeper the upward portion of the ROC curve, and the higher the area under the curve (AUC).
Universal screening for hyperglycemia during pregnancy may be in-practical in resource constrained countries. Therefore, the aim of this study was to develop a simple, non-invasive practical tool to predict undiagnosed Gestational diabetes mellitus (GDM) in Tanzania.
We used cross-sectional data of 609 pregnant women, without known diabetes, collected in six health facilities from Dar es Salaam city (urban). Women underwent screening for GDM during ante-natal clinics visit. Smoking habit, alcohol consumption, pre-existing hypertension, birth weight of the previous child, high parity, gravida, previous caesarean section, age, MUAC ≥ 28 cm, previous stillbirth, haemoglobin level, gestational age (weeks), family history of type 2 diabetes, intake of sweetened drinks (soda), physical activity, vegetables and fruits consumption were considered as important predictors for GDM. Multivariate logistic regression modelling was used to create the prediction model, using a cut-off value of 2.5 to minimise the number of undiagnosed GDM (false negatives).
Mid-upper arm circumference (MUAC) ≥ 28 cm, previous stillbirth, and family history of type 2 diabetes were identified as significant risk factors of GDM with a sensitivity, specificity, positive predictive value, and negative predictive value of 69%, 53%, 12% and 95%, respectively. Moreover, the inclusion of these three predictors resulted in an area under the curve (AUC) of 0.64 (0.56–0.72), indicating that the current tool correctly classifies 64% of high risk individuals.
The findings of this study indicate that MUAC, previous stillbirth, and family history of type 2 diabetes significantly predict GDM development in this Tanzanian population. However, the developed non-invasive practical tool to predict undiagnosed GDM only identified 6 out of 10 individuals at risk of developing GDM. Thus, further development of the tool is warranted, for instance by testing the impact of other known risk factors such as maternal age, pre-pregnancy BMI, hypertension during or before pregnancy and pregnancy weight gain.
Trend in the clinical profile of type 2 diabetes in India - Study from a diabetes care centre in South India
2020, Diabetes and Metabolic Syndrome: Clinical Research and Reviews
Study the changing clinical and therapeutic profiles of type 2 diabetes (T2D) patients during a 10-year period in a diabetes care centre in Southern India.
Retrospective, cross-sectional data of newly registered and review patients at four periods between 2009 and 2018 were analysed (n = 50,322). Clinical findings, anthropometry, blood pressure (BP), glycaemia, lipids, treatment, and co-morbid conditions were analysed. We studied the trends in age and gender distributions, body mass index (BMI), glycaemia, (Glycosylated haemoglobin A1c (HbA1c) levels), BP, cholesterol, triglycerides and therapeutic regimen during this period. Trend analyses were done.
Approximately 60% of patients were men. Percentages in 30–40 years increased, ∼60% were aged 50–69 years and proportion of older patients decreased (p < 0.0001). In 10 years, 85.9% was overweight, obesity increased from 22.1% to 25.0% (p < 0.0001) and <13% maintained normal BMI. HbA1c <7.0% remained approximately at 22%, percentage with moderate glycaemic control (HbA1c 7.0–7.9%) increased significantly, HbA1c of ≥9.0% decreased from 35.1% to 29.1% (p < 0.0001). Use of monotherapy decreased. Prevalence of hypertension increased from 16.2% to 21.6% (p < 0.0001); use of Angiotensin Receptor Blockers (ARB) and calcium channel blockers increased, Angiotensin Converting Enzyme Inhibitors and thiazides decreased (p < 0.0001). Increased use of statins paralleled with reduction in total cholesterol and LDLc.
Increasing percentages of younger patients and obesity, use of multiple drugs and reduction in HbA1c were the important observations. Rising prevalence of hypertension, increased use of ARB and statins with better control of dyslipidaemia was observed. Achievement of ideal HbA1c and BP were suboptimal.

View all citing articles on Scopus

View full text

Derivation and validation of diabetes risk score for urban Asian Indians

Abstract

Objective:

Research design and methods:

Results:

Conclusions:

Introduction

Section snippets

Materials and methods

Results

Discussion

Acknowledgements

Diab. Res. Clin. Prac.

Onset of NIDDM occurs at least 4–7 years before clinical diagnosis

Diab. Care

Should we screen for type 2 diabetes? Evaluation against National screening committee criteria

BMJ

Diabetes risk score: towards earlier detection of Type 2 diabetes in general practice

Diab. Metab. Rev.

The diabetes risk score: a practical tool to predict type 2 diabetes risk

Diab. Care

A Danish diabetes risk score for targeted screening: the inter99 study

Diab. Care

A multivariate logistic regression equation to screen for diabetes: development and validation

Diab. Care