View on GitHub

cvd-and-teeth

Revisiting the Cross-Sectional Association Between Missing Teeth and Cardiovascular Disease: BRFSS 2018

Revisiting the Cross-Sectional Association Between Missing Teeth and Cardiovascular Disease: BRFSS 2018 ================ First created on Nov 7, 2019. Updated on Nov 10, 2019

The relationship between oral health and cardiovascular disease has been an important area of research in the last decade. The objective of the current study is to perform an analysis of BRFSS 2018 data based on the previously published work of R. Constance Wiener and Usha Sambamoorthi from 2010. This analysis is meant to be an example of the data analysis and R programming techniques required to perform both a descriptive and regression analysis of publically available data.

In their orignal 2010 analysis, Weiner and Sambamoorth evaluated the association of cardiovascular disease and the number of missing teeth as a risk indicator. In this analysis, the 2018 BRFSS data and initial variables from that study will be used as the basis for demonstrating the products and coding required for (1) an descriptive analysis and (2) a regression analysis using R.

For more information about the original study check out the published article at: https://www.ncbi.nlm.nih.gov/pubmed/24624297

Stage 1: Data Preparation, Cleaning, and Reduction

The initial data set for the project was the 2018 BRFSS Survey Data, and it was quite large. Before conducting any other analyses to be demonstrated here (such as a desciptive analysis or regression analysis), the data needed some additonal cleaning and reduction. The final result was a data set of 52 variables (compared to the 275 variables in the total set). In addition, the number of records was reduced from over 440,000 records to approximately 266,000 records. The 2018 BRFSS Codebook CDC was a key tool in this process.

The three main products of this process included:

For the detailed coding and results, explore the Stage 1 Preparing, Cleaning, and Reducing the Data.

Stage 2: Example Descriptive Analysis

The key product of this analysis is a complete Table 1 documenting the frequency counts for each variable value (1) overall for all participants, and broken down for those (2) with cardiovascular disease and (3) with no cardiovascular disease. Histogram plots of each variable are also provided. And of course this section contains the detailed r-code that produces all of this. Chi-square analysis of all variables were significant at the p<0.001 level.

To see the detailed coding and results, explore the Stage 2 Descriptive Analysis

Stage 3: Linear Regression

A linear regression analysis is conducted here ONLY AS A DEMONSTRATION OF METHODOLOGY AND CODING. Linear regression is typically most useful in the case of continuous outcome data. However, in this study the outcome variable is binomial: cardiovascular disease or no cardiovascular disease. As a result the appropriate modeling technique for analyzing the hypothesis and data in this study will be Logistic Regression. The logistic regression will appear in Stage 4.

A linear regression analyis is provided for 3 models. In addition a step-wise multiple regression was applied which retained everyone of the variables. However, none of these models were very good. The full model was best but only accounted for about 6% of the variance in the data. As a result and log regression will follow.

For the detailed coding and results, explore the Stage 3 Linear Regression.

Stage 4: Logistic Regression (the Real Analysis)

As stated above, the binomial nature of the outcome variable would suggest that logistic regression is the most appropriate method to analyze these data. In fact, every one of the variables in this study is two-level (binomial) or a categorical variable with three or more levels (multinomial). In this analysis, two logistic regression models were constructed with the same parameters as Model1 and the Full Model used in the logistical model. This model resulted in a series of Odds ratios (with upper and lower limits).

For the detailed coding and results from this logistic regression, explore the Stage 4 Logistic Regression.