Document Type

Capstone Experience

Graduation Date

5-2024

Degree Name

Master of Public Health

Department

Biostatistics

First Committee Member

Lynette M. Smith, PhD

Second Committee Member

Jim Anderson, PhD

Abstract

The original study aimed to show that thyroidectomy does not result in surgical site infection (SSI) in most cases, and thus routine prescription of antibiotics is not necessary. The study looked to see what risk factors could predict the incidence of SSI. This would highlight those individuals who were at most risk of developing SSI, and then antibiotics would only be prescribed to these individuals instead of all or most individuals who undergo thyroidectomy.

This study used NSQIP data to look at incidence of SSI and look for risk factors that may be predictive of SSI. Only surgeries that were considered clean were included. To determine factors that could be significantly associated with SSI the study used Chi-square tests and logistic regression.

It was determined that being 80 or older, male, and a current smoker were all significant predictors of SSI after thyroidectomy as well as being ventilator dependent. If an individual did not have these risk factors, then incidence of SSI was relatively low and routine prescription of antibiotics was not recommended.

The replication analysis included pure replication, Measurement and Estimation Analysis (MEA) and Theory of Change Analysis (TCA). The goal of the pure replication was to reproduce results presented in the tables from the original paper using the NSQIP data and reported statistical methods. For the MEA, the goal was to check the model assumptions used in the univariate Chi-square analysis and logistic regression. The assumptions for the multivariate logistical regression were also checked using residual analysis. The MEA proved that model assumptions were met for both Chi-square and logistic regression. For the TCA, the objective was to determine if there was any interaction between gender and being a current smoker, as both show strong statistical significance based on the multivariate analysis. It was determined that no interaction was present.

Download

Share

COinS