Prediction equations and point system derived from large-scale health check-up data for estimating diabetic risk in the Chinese population of Taiwan

https://doi.org/10.1016/j.diabres.2010.12.022Get rights and content

Abstract

Aim

To develop tools for predicting diabetes development in middle-aged Chinese adults living in Taiwan.

Methods

This study made use of data from 24,899 non-diabetic adults aged ≥35 years who received health examination service from a private health check-up clinic during the period of 1994–1996 and had one or more examinations before December 31, 2006. The proportional hazard model and the receiver operating characteristic (ROC) curve method were used respectively to construct the prediction equation and assess the model's performance. A point system is developed for the ease to calculate diabetes risk.

Results

Increased risk of diabetes development was associated with older age, lower education level, alcohol abstinence, abdominal obesity, elevated body mass index (BMI), blood pressure (BP), triglycerides, and impaired fasting glucose. Model 1, incorporating personal socio-demographic and lifestyle characteristics, BMI, and waist circumference (WC), had an area-under-curve (AUC) of 0.717. The AUC increased to 0.726 (model 2) when BP was introduced and to 0.823 (model 3) when both BP and clinical chemistry measures were added. The AUCs in the testing set for models 1, 2, and 3 were 0.688, 0.694, and 0.799 respectively.

Conclusions

These predictive equations of diabetic risk were easy to use by clinical professions and general subjects.

Section snippets

Background

Prevalence and incidence rates of diabetes and diabetic complications are rapidly increasing in the world. In Taiwan, diabetes and associated cardiovascular consequences such as heart disease, stroke and end-stage renal disease are among the leading causes of death. Although diabetes has become more prevalent in recent years, studies [1], [2] have shown that diabetes can be prevented and its progression halted by appropriate lifestyle interventions. In order to stop disease progression early in

Study population

The MJ Health Screening Center is a chained private membership cooperation with four health check-up centers located in northern (Taipei), north western (Taoyuan), central (Taichung), and southern (Kaohsiung) Taiwan. These clinics provide routine health examination to members. A total of 24,899 non-diabetic subjects aged more than 35 year-old participated the exam during the period of 1994–1996 and had at least one or more health examinations during the period of 1997–2006. Use was made of the

Results

There was no significant difference between the training set and the testing set in baseline characteristics such as age, gender, BMI, and WC (Table 1). There was also no difference in cumulative diabetes incidence between the two sets (Table 1). However, diabetic patients had significantly higher levels of baseline age, body mass index, waist circumference, triglycerides, systolic blood pressure, diastolic blood pressure, and fasting glucose (Table 2).

The average and the standard deviation of

Main study findings

Six risk prediction equations for diabetes, varying in class and number of variables included in the model, were constructed in the current study. The performance of these equations differed in AUCs which ranged from 0.713 to 0.835. The simplest prediction model (model 1), including only sociodemographic and anthropometric variables, performed reasonably well. The second model (model 2), adding BP status information to model 1, performed better. The third model (model 3), including all the

Conflict of interest

There are no conflicts of interest.

Acknowledgements

This study was supported by a grant from the National Science Council, Taiwan (NSC-95-2314-001-012-MY3). We thanked MJ Health Life Company to provide their assistance in data acquirement.

References (29)

  • K. Chien et al.

    A prediction model for type 2 diabetes risk among Chinese people

    Diabetologia

    (2009)
  • C. Glumer et al.

    Risk scores for type 2 diabetes can be applied in some populations but not all

    Diabetes Care

    (2006)
  • M.H. Zweig et al.

    Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine

    Clin Chem

    (1993)
  • L.M. Sullivan et al.

    Presentation of multivariate data for clinical use: the Framingham Study risk score functions

    Stat Med

    (2004)
  • View full text