Intermediate Epi Final Paper
Diabetes is a leading cause of death in the United States and affects approximately 11.3% of individuals over the age of 20.(1) Previous cross-sectional studies using the National Health and Nutrition Examination Survey (NHANES) conducted in 2012 show the estimated prevalence of diabetes was between 12% and 14% among US adults, with a higher prevalence found amongst individuals who were Non-Hispanic Black/African American, Non-Hispanic Asian and Hispanic.(2) In addition, when comparing prevalence rates from several years of NHANES data, an increase in prevalence of diabetes in the overall population has occurred from 1988 through 2012. Similar trends have been seen in other countries around the world. Throughout the literature, Diabetes, and specifically Type II diabetes, has been considered a serious public health problem around the world, due to the many chronic complications related to the disease and the high cost associated with treating those who become diabetic.(3-6) The estimated annual cost related to diabetes and complications related to the disease in 2012 in the United States was $245 billion.(1) With an estimated 20.6 million people in the United States who are affected by Type II diabetes, it is important to understand all the potential risk factors associated with the disease to allow for the implementation of appropriate public health interventions.(7) Surveys such as NHANES and the Behavioral Risk Factor Surveillance System (BRFSS) provide a way to measure the prevalence, incidence, morbidity and mortality of diabetes in the population as well as surveillance of the associated risk factors.(8)
If you need assistance with writing your essay, our professional essay writing service is here to help!Find out more
Many of the traditional risk factors associated with the development of Type II diabetes have been well documented and often are associated with the clinical factors related to the onset of the disease. Some of the these traditional risk factors that have been documented include sedentary behavior, obesity, hyperlipidemia, hypertension, and cardiovascular disease.(3, 4, 8) More recently however, there has been an increased focus on the social determinants of health that are potential risk factors for diabetes. Some examples of these SDH risk factors in a urban setting include easy access to healthy foods, less opportunity for regular exercise, and in a rural setting include reduced access to medical care.(1) Another social determinant of health risk factor that has been study as a potential risk factor for Type II diabetes is socioeconomic status. Previous studies have examined the association between socioeconomic status and the prevalence and incidence of diabetes in individuals and have found a relationship between socioeconomic status and diabetes.(1, 3, 5, 7, 9, 10)
A retrospective cross-sectional study was conducted using the a data set from the Behavioral Risk Factor Surveillance System (BRFSS) which has been described in detail previously.(11-14) BRFSS was created by the Centers for Disease Control (CDC) in 1984, with the purpose of collecting information to assess the risk factors associated with yearly morbidity and mortality as well as the prevalence of behaviors related to health promotion. The survey is conducted on a monthly basis across all 50 states, including several US territories. Study participants include adults who were aged 18 years or older, who resided in the United States at the time that the survey was conducted, and were noninstitutionalized. In households in which multiple adults resided, one adult was selected at random for participation in the interview.(11)
The sample of participants used in the cross-sectional study was from the 2014 administration of the BRFSS survey and consisted of 464,664 individuals. The BRFSS telephone survey used random digit dialing to select sample records randomly from the list of all available telephone numbers. The sample designs across all the states that participate are required to meet the BRFSS standard, with the justification that the records selected are a probability sample of all households in the participating states. This criteria was met by all states who participated in 2014. As mentioned in the sample description provided by the CDC, “Fifty-one projects used a disproportionate stratified sample (DSS) design for their landline samples.”(13) In this design, landline telephone numbers are divided into two strata, a high-density and a medium-density. The numbers are divided into the two stratum based on the number of houses that are listed in its hundred block. The strata are then sampled to obtain the probability of all households that contain a telephone.
Cellular telephones were sampled using the Telecordia database telephone exchange, form which one telephone number is selected at random from a set interval.(13) The intervals for the BRFSS were formed by dividing the population count of telephone numbers in the frame by the desired sample size. For the 2014 BRFSS cellular telephone sample, individuals who were targeted were aged 18 or older, had a working cellular phone, and resided in a private residence or housing at a college.
A power analysis was not conducted since this was a hypothesis generating cross-sectional study.
Data was collected for BRFSS through landline telephone surveys, with the addition of cellular telephone based surveys in 2011.(11) Trained telephone interviewers collect data from a randomly selected adult in a household through the use of a semi-standardized survey. For the 2014 BRFSS survey, the Computer-Assisted Telephone Interview (CATI) system was used by interviewers for data collection.(13) The system contained the core questions in addition to state specific questions. The average length of time for completion of the survey was 18 minutes for the core component and 5 to 10 minutes for the additional state component.
Interviewers were trained repeatedly by a state coordinator or interviewer supervisor under guidelines specific to BRFSS.(13) Individuals selected as interviewers typically had experience conducting telephone surveys, but additional training was provided on the BRFSS questionnaire in addition to related procedures prior to being approved to work on BRFSS. Protocols required regular evaluation of interviewer performance which was conducted using a monitoring system. The system allowed evaluators to monitor interviewers by listening to only the interviewer at an on-site location or to both the respondent and interviewer from a remote location. Systematic monitoring of interviewers took place for a set amount of time on a monthly basis. Data was collected to quantify the performance of each interviewer.
Interviewers placed calls to respondents 7 days per week, during both daytime and evening hours.(13) A standard protocol was followed to ensure rotation of calls over days of the week and times of the day.
The 2014 BRFSS Survey included questions related to 22 measures that covered demographic, preventative health practices, behavioral health and chronic conditions. The outcome assessed in the study was diabetes and the exposure was annual household income. Potential confounders that were assessed included sex, age, education level, self-reported race/ethnicity, perceived health status, physical health, mental health, poor physical or mental health interference with usual activities, health care coverage, exercised in past month, currently has asthma, current employment status, self-reported weight, self-reported height, calculated BMI, smoked at least 100 cigarettes in entire life and current smoking status. Exercise in the past month was assessed as a potential effect measure modifier.
Prior to conducting a statistical analysis of the BRFSS data, certain variables were recoded. Age was recoded into five categories consisting of individuals from 18 to 29 years, 30 to 39 years, 40 to 49 years, 50 to 59 years, 60 to 69 years, 70 to 79 years, and 80 years or older, with the age group 60 to 69 years being used as the reference group. Education level was recoded into two categories consisting of individuals who were a high school graduate or less and those that were some college or more, with some college or more being used as the reference group. The variables for race and Hispanic, Latino/a, or Spanish origin were recoded into a single variable labeled as self-reported race/ethnicity with five categories consisting of Non-Hispanic White, Non-Hispanic Black/African American, Hispanic or Latino/a, Non-Hispanic Other and Missing, with Non-Hispanic White being used as the reference group. Perceived health status was recoded into four categories consisting of excellent or very good, good, fair or poor, and missing, with excellent or very good being used as the reference group. Healthcare coverage was recoded into three categories consisting of yes, no and missing, with yes being used as the reference group. Exercised in the past month was recoded into three categories consisting of yes, no and missing, with yes being used as the reference group. Currently has asthma was recoded into three categories consisting of yes, no and missing with no being used as the reference group. Ever had a depressive disorder was recoded into three categories consisting of yes, no and missing. Ever had diabetes was recoded into two categories consisting of yes and no. Current employment status was recoded into four categories consisting of employed, unemployed, retired or unable to work, and missing, with employed being used as the reference group. Annual household income was first recoded into 5 categories and later recoded into three categories consisting of less than $49,999, $50,000 or more, and missing with $50,000 or more being used as the reference group. Has smoked a least 100 cigarettes in entire life was recoded into three categories consisting of yes, no, and missing with no being used as the reference group. Amount of days currently smoking was recoded into four categories consisting of current smoker, former smoker, never smoked and missing with never smoked being used as the reference group. In general, categorical variables that including a missing consisted of individuals who responded with don’t know/not sure, refused to provide a response, or for those that were not asked or were missing a response in the survey data when compiled.
Several continuous variables in the BRFSS data were also recoded into categorical variables prior to conducting logistic regression. The number of days that a participant reported their physical health to be not good in the past month was recoded into four categories consisting of 0 days, 1-10 days, 11-20 days, 21-30 days, with 0 days being used as the reference group. The number of days that a participant reported their mental health to be not good in the past month was recoded into four categories consisting of 0 days, 1-10 days, 11-20 days, 21-30 days, with 0 days being used as the reference group. The number of days that a participant reported poor physical or mental health that interfered with usual activities in the past month was recoded into four categories consisting of 0 days, 1-10 days, 11-20 days, 21-30 days, with 0 days being used as the reference group.
Certain respondents observations in the BRFSS data for continuous variables required a conversion to a standard scale prior to being recoded into categorical variables to be used in the logistic regression. Self-reported weight was provided in kilograms for some respondents which was converted to pounds to create a uniform measurement of weight across all respondents in the data. Self-reported weight was then recoded into five categories consisting of 50-149 lbs, 150-299 lbs, 300-449 lbs, 450-599 lbs, and greater than 600 lbs, with 150-299 lbs being used as the reference group. Self-reported height was provided in meters/centimeters for some respondents which was converted to inches to create a uniform measurement of height across all respondents in the data. Self-reported height was then recoded into five categories consisting of less than 60 inches, 61 to 69 inches, 70 to 79 inches, and greater than 80 inches, with 61-69 inches being used as the reference group. After converting self-reported weight and height, a new variable of Body Mass Index (BMI) was computed using a standard calculation of height and weight. Once calculated for each respondent, BMI was then recoded into four categories based on the clinical categories associated with BMI and consisted of underweight (<18.5), normal (18.5 to 24.9), overweight (25 to 29.9) and obese (>30), with obese being used as the reference group.
A univariable analysis was conducted in order to provide descriptive statistics for respondents in the BRFSS data. For the categorical variables, the frequency and percentage of the total respondents in the data set were calculated for each variables categories. For the continuous variables, the mean, median, standard deviation, range, and interquartile range were calculated.
A bivariable analysis was conducted to determine if there was a difference across all variables for respondents in regards to the outcome and exposure of interest. Chi-Square tests were used to determine if there was a difference in respondents who reported as being diabetic and those who did not across all categorical variables, and t-tests were used for the continuous variables. In order to determine the differences in respondents who reported as having certain levels of annual household income, Chi-Square tests were used for all categorical variables, and ANOVA was used for continuous variables.
Our academic experts are ready and waiting to assist with any writing project you may have. From simple essay plans, through to full dissertations, you can guarantee we have a service perfectly matched to your needs.View our services
A stratified analysis was conducted to assess the presence of potential confounders as well as interactions. Odds ratios for the crude association between diabetes and annual household income were calculated along with the stratum specific odds ratios. The analysis was conducted using logistic regression. The exposure variable of annual household income was originally recoded into four categories, and was recoded prior to conducting logistic regression into two categories, comprising of individuals with an annual household income of less than $49,999 and the other of individuals with an income greater than or equal to $50,000. The low income category was set as the positive outcome and the high income category as the negative outcome.
A multivariable analysis was conducted to assess the association between the exposure of annual household income and the outcome of diabetes. Logistic regression was used to calculate the crude odds ratios and the reference group for each variable was set to the category with the highest frequency of participants. An interaction term of annual household income and exercise in the past month was also assessed.
A fully-adjusted logistic regression model was conducted in which the interaction term of exercise in the past month was assessed and all additional variables where treated as confounders. The confounders that were considered in the fully-adjusted model included sex, age, education level, self-reported Race/Ethnicity, perceived health status, physical health, mental health, poor physical or mental health, asthma, current employment status, self-reported weight, self-reported height, Body Mass Index, and current smoking status. One variable that had been identified as collinear was removed from the regression equation and consisted of smoked at least 100 cigarettes in entire life. The variable history of depressive disorder, which was neither considered a confounder or an effect measure modifier, was not assessed in this analysis. An interaction term of annual household income and exercise in the past month was also assessed. Hosmer-Lemeshow goodness-of-fit test was conducted for the fully-adjusted model. Akaike information criterion and Bayesian information criterion were calculated for the model to allow for a comparison to the final model.
The final model was conducted using logistic regression and several variables were removed in the process of testing to achieve the most parsimonious model. The variables that were considered confounders for the final model included respondents sex, age, education level, self-reported race/ethnicity, health care coverage, current employment status, and Body Mass Index. An interaction term of annual household income and exercise in the past month was also assessed. Model assumptions were tested and influential outliers were removed to improve model fit. Hosmer-Lemeshow goodness of fit test was conducted for the final model. Akaike information criterion and Bayesian information criterion were calculated for the model to indicate whether the model fit improved from the final model after the removal of variables.
1. Clark ML, Utz SW. Social determinants of type 2 diabetes and health in the United States. World journal of diabetes 2014;5(3):296-304.
2. Menke A, Casagrande S, Geiss L, et al. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988-2012Prevalence of and Trends in Diabetes Among US AdultsPrevalence of and Trends in Diabetes Among US Adults. JAMA 2015;314(10):1021-9.
3. Agardh E, Allebeck P, Hallqvist J, et al. Type 2 diabetes incidence and socio-economic position: a systematic review and meta-analysis. International Journal of Epidemiology 2011;40(3):804-18.
4. Elgart JF, Caporale JE, Asteazarán S, et al. Association between socioeconomic status, type 2 diabetes and its chronic complications in Argentina. Diabetes Research and Clinical Practice 2014;104(2):241-7.
5. Evans JMM, Newton RW, Ruta DA, et al. Socio-economic status, obesity and prevalence of Type 1 and Type 2 diabetes mellitus. Diabetic Medicine 2000;17(6):478-80.
6. Grintsova O, Maier W, Mielck A. Inequalities in health care among patients with type 2 diabetes by individual socio-economic status (SES) and regional deprivation: a systematic literature review. International journal for equity in health 2014;13:43-.
7. Krishnan S, Cozier YC, Rosenberg L, et al. Socioeconomic Status and Incidence of Type 2 Diabetes: Results From the Black Women’s Health Study. American Journal of Epidemiology 2010;171(5):564-70.
8. Ali MK, Siegel KR, Laxy M, et al. Advancing Measurement of Diabetes at the Population Level. Current Diabetes Reports 2018;18(11):108.
9. Rabi DM, Edwards AL, Southern DA, et al. Association of socio-economic status with diabetes prevalence and utilization of diabetes care services. BMC health services research 2006;6:124-.
10. Robbins JM, Vaccarino V, Zhang H, et al. Socioeconomic status and diagnosed diabetes incidence. Diabetes Research and Clinical Practice 2005;68(3):230-6.
11. Silva NM. The Behavioral Risk Factor Surveillance System. The International Journal of Aging and Human Development 2014;79(4):336-8.
12. Pierannunzi C, Hu SS, Balluz L. A systematic review of publications assessing reliability and validity of the Behavioral Risk Factor Surveillance System (BRFSS), 2004–2011. BMC Medical Research Methodology 2013;13(1):49.
13. Centers for Disease Control and Prevention (CDC). Behavioral Risk Factor Surveillance System Overview: BRFSS 2014. Atlanta, Gerogia, 2015, (U.S. Department of Health and Human Services, Centers for Disease control and Prevention
14. Nelson DE, Powell-Griner E, Town M, et al. A comparison of national estimates from the National Health Interview Survey and the Behavioral Risk Factor Surveillance System. American journal of public health 2003;93(8):1335-41.
Cite This Work
To export a reference to this article please select a referencing stye below:
Related ServicesView all
DMCA / Removal Request
If you are the original writer of this essay and no longer wish to have your work published on UKEssays.com then please: