Risk for diabetes among Texans
MetadataShow full metadata
Diabetes is a disease in which a person's body fails to properly use and store glucose. Glucose is retained in the bloodstream, and consequently causes the person's blood glucose or "sugar" to rise too high. There are two major types of diabetes: Type 1 diabetes and Type 2 diabetes. With Type 1 diabetes, the body completely stops producing insulin. People with Type 1 diabetes must take daily insulin injections. With Type 2 diabetes, the body produces insufficient amount of insulin to convert food into energy, or in the case of insulin-resistance, the body can't properly use the insulin it does produce. The American Diabetes Association reports about 17 million people (6.2% of the population) suffer from diabetes and many of them are not even aware that they have the disease. Diabetes was the sixth-leading cause of death according to analysis of 1999 U.S. death certificates (Diabetes Week, February 3, 2003). Each year, an estimated 12,000-24,000 people become blind because of diabetic eye disease. In addition, more than 38,000 people with diabetes begin treatment for kidney failure each year, and about 86,000 undergo diabetes-related lower extremity amputations. The total of direct and indirect costs of diabetes in the US is nearly $132 billion a year. In Texas, diabetes contributed 13,553 deaths in 1998 and 15,130 deaths 2000. There were 911,039 diagnosed diabetes patients (about 6.2% of the adult population in the age of 18 years or older) and 450,504 people with undiagnosed diabetes (about 3.6% of the adult population in the age of 20 years or older) in Texas in 2001 (TDH Diabetes Council Report, 2001). Because diabetes is a serious, costly, and increasingly common chronic disease that can cause devastating complications and often result in disability and death, appropriate predictors need to be identified. The probability that a person will develop diabetes needs to be estimated, so that people can formulate healthy lifestyles to reduce their risk for developing diabetes and its complications.
Many studies have been done to determine the risk factors for diabetes (Anastasia C. Thanopoulou, et al., 2003; Karin M. Nelson, et al., 2002; Bahman P. Tabaei, William H. Herman, 2002; HU FB Sigal RJ, Rich-Edwards JW, et al., 1999; Hu FB, van Dam RM, Liu S., 2001). Identified non-modifiable risk factors include age, ethnicity and family history of diabetes; identified modifiable risk factors include diet, obesity, physical inactivity, alcohol consumption, tobacco smoking and hypertension. These genetic, environmental, and metabolic risk factors are interrelated and contribute to the development of type 2 diabetes. Multivariate logistic regression has been used in previous diabetes studies (0. Rolandsson, et al. 2001; Philip S. Mehler, et al., 1998; Bahman P. Tabaei, William H. Herman, 2002). Findings from these studies indicated that age, race, body mass index, physical activity, alcohol consumption and family history of diabetes were significant predictors for diabetes.
The purpose of this study was to identify predictors of diabetes and to estimate the probability of diabetes prevalence in Texas. This study used existing data made available by the Texas Department of Health (TDH). Data were collected by Texas Behavioral Risk Factor Surveillance System (BRFSS) in a 2001 survey. Texas BRFSS used the Centers for Disease Control and Prevention "2001 Behavioral Risk Factor Surveillance System Questionnaire" with disproportionate stratified random sampling (DSS) to collect data. There were 5916 participants older than 18 years. Seventeen variables listed in Appendix A were entered into logistic regression model. Of these 5916 participants, 1221 had no missing values in the 17 variables used as analysis sample to generated final logistic model. For the purpose of validation of the final logistic regression equation, the final model was applied to TDH BRFSS survey data of year 1999. Fourteen variables listed in Appendix B were entered into the validation model. Of the 4990 survey participants in year 1999, 1633 had no missing values in the above 14 variables and were used as a validation sample. Stepwise selection identified age, race, blood pressure, blood cholesterol, and body mass index as significant predictors. The final model for prediction of probability of diabetes presence was:
Log [P / (1 -P)] =-3.3389 + 1.1776 (Age_55-64 years)+ 1.1505 (Age_64+ years) + 1.1763 (Non-Hispanic black) + 0.9279 (Hispanic) + 0.5611 (High blood pressure) + 0.5691 (High blood cholesterol)- 0.8746 (Alcohol drink status_yes)- 0.6198 (Leisure time physical activity_yes) + 0.6968 (Obese), Where as P = estimated probability of presence of diabetes
This study showed that diabetes presence was more likely associated with age groups older than 55 years than with 18 - 34 years group (adjusted odds ratio= 3.247, with 95% CI: 1.793 - 5.878 for age 55 - 64 years group; adjusted odds ratio= 3.160, with 95% CI: 1.731 - 5.769 for 65+ age group), with being non-Hispanic black (adjusted odds ratio= 3.242, with 95% CI: 1.637 - 6.422) and Hispanic (adjusted odds ratio= 2.529, with 95% CI: 1.417 -4.513) rather than being non-Hispanic white; with high blood pressure (adjusted odds ratio= 1.753, with 95% CI: 1.005 -2.911) and high blood cholesterol (adjusted odds ratio= 1.767, with 95% CI: 1.079-2.892) rather than normal blood pressure and normal blood cholesterol; with obesity (adjusted odds ratio= 2.007, with 95% CI: 1.237 - 3.256) rather than body mass index less than 25. This study also found that adjusted odds ratios were 0.417 (95% CI: 0.249- 0.700) for alcohol drinkers and 0.538 (95% CI: 0.330 - 0.878) for people who did leisure time physical activity, indicating that appropriate consumption of alcohol and physical activity protect people from diabetes.
By applying the estimated coefficients in the final equation, a person's probability to be a diabetic can be calculated. For example, a physically active non-Hispanic white person who is a moderate level alcohol drinker aged younger than 55 years, with normal blood pressure, normal blood cholesterol, BMI less than 25, only has a 0. 79% probability to be a diabetic. For a physically inactive non-Hispanic black person who is alcohol nondrinker older than 65 years, with high blood pressure, high blood cholesterol, and body mass index greater than 30, the predicted probability to be a diabetic will increase to 69.32%. However, this study suggests that when apply this prediction equation to population, prevalence of diabetes will be underestimated because lack of other important predictors in the final model. Therefore, a threshold of predicted probability of0.10 generated from ROC curve analysis should be considered for the purpose of prevention and early detection. Individuals who had predicted probability greater than this threshold could be identified as a group of people at high risk, and they need further medical diagnostic investigation. When this model is applied to the population, a lower (higher) cutpoint of predicted probability would be considered to meet expectation of a higher (lower) sensitivity and a lower (higher) specificity.
In this sample, more than 70% of participants had a low intake (less than 5 servings per day) of fruit and vegetable, leading to a crude odds ratio of 0. 77 with 95% confidence interval: 0.49- 1.21 for group of people who had a low intake of fruit and vegetable, compared to group of people who had more than 5 servings of fruit and vegetable per day. This result could be generated from survey information bias. Because the survey questionnaire only considered information 'in the past 30 days', some diabetics diagnosed before this period, might have changed their habits and eaten more fruits and vegetables after being diagnosed with diabetes. This fin<;iing implies that the data should include information before and after diabetes had been diagnosed to minimize the information bias.
The final logistic regression model for this study yields a relatively low coefficient of determination (0.22). This result suggests that variables not accounted for in this study, such as family history, vitamin supplements, and other related chronic diseases, might explain a significant proportion of the variance in diabetes presence. A recommendation for future study is that information about family history, vitamin supplements and comorbidities should be collected and analyzed in future studies. A more accurate predicted probability of diabetes presence will be generated by the enrichment of the information.
CitationLi, J. (2003). Risk for diabetes among Texans (Unpublished thesis). Texas State University-San Marcos, San Marcos, Texas.
This item is restricted to the Texas State University community. TXST affiliated users can access the item with their NetID and password authentication. Non-affiliated individuals should request a copy through their local library’s interlibrary loan service.