10

BIOM 643: Applied Biostatistics in Clinical Research

3 semester credits, REQUIRED

This course will combine a didactic introduction to multivariable regression with the “hands-on” analysis of a clinical dataset. Lectures will introduce regression methods for normally distributed, binary, and count data, as well as for time-to-event data. Lectures will also cover both model building and model evaluation. To complement the lectures, each trainee will perform the secondary analysis of a clinical dataset. The datasets will be created from administrative (claims) data. Trainees will be given the opportunity to select (or suggest) a topic that interests them. Trainees can elect to analyze clinical outcomes or health service research data. Trainees will be asked to further review the literature, then to design analyses that answer a clinical hypothesis. Analyses will include descriptive statistics and multivariable regression models. The trainees will be taught basic data management skills as needed to perform their analyses. The trainees will be required to include at least one analysis using each of the regression methods they are taught. Exams will emphasize questions based on the analysis of their own and other trainee projects. The final exam will be a seminar presenting their research findings. Course readings will include selected chapters from Statistical Analysis of Epidemiologic Data, Selvin S., Oxford University Press, 1998, and Multivariable Analysis: A Practical Guide for Clinicians by Mitchell H. Katz, Cambridge University Press, 1999, as well as handouts that detail regression methods.


Professor: Dr. A. Katz and Dr. James Davis

Course Number: 643

Course Title: Applied Biostatistics in Clinical Research

Course Credit: 3 credits

Prerequisite: Applied Clinical Epidemiology and Biostatistics

Placement in Curriculum:
Applied Biostatistics in Clinical Research will follow, the course in appliedepidemiologic and biostatistical methods. It may be taken before or along with Clinical Research Methods

Course Description:
Applied Biostatistics in Clinical Research will combine a didactic introduction to multivariable regression with the hands on analysis of a clinical dataset. The lectures will introduce the regression methods for normally distributed, binary, and count data, as well as for time-to-event data. The lectures will cover both model building and model evaluation. To complement the lectures, each student will perform the secondary analysis of a clinical dataset. The datasets will be created from administrative (claims) data.

Students will be given the opportunity to select (or suggest) a topic that interests them. Students can elect to analyze clinical outcome or health service research data. The students will be asked to further review the literature, then to design analyses that answer a clinical hypothesis. Analyses will include descriptive statistics and multivariable regression models. The students will be taught basic data management skills as needed to perform their analyses. The students will be required to include at least one analysis using each of the regression methods they are taught. Instruction in multiple predictor analyses as a tool for control of confounding and for constructing predictive models.

Topics will include linear regression and logistic regression. Statistical packages will be used throughout. multivariable statistical analysis that includes instruction in survival analysis and analysis of repeated measures and clustered data. The course culminates with student presentations of statistical analyses of their own research projects. Instruction in advanced topics in biostatistics including individualized instruction in biostatistical methods pertaining to the scholars’ research projects will be included. Exams will emphasize questions based on the analysis of their own and other student projects. The final will be a seminar presenting their research findings.
Learning Outcomes: At the conclusion of the course student will be able to:

  • Understand how analyzing longitudinal data differs from cross-sectional data analysis
  • Define the concepts of hazard and cumulative hazard
  • Use the Kaplan-Meier method of estimating survivor functions
  • Understand the proportional hazard assumption
  • Include time-dependent covariates in regression models
  • Identify nested (multilevel) relations and model them appropriately
  • Interpret regression coefficients from survival and multilevel analyses
  • Analyze data with repeated measurements on the study participants
  • Understand random effects and their implications for longitudinal analyses
  • Design and perform the analysis of a secondary dataset
  • Present analysis results at a seminar
  • Write-up a research analysis in a manuscript form

 

Topical Outline:
Exploratory analysis of longitudinal data
Multilevel models of change
Handling repeated measurements
Incorporating time-dependent covariates
Discrete time hazard models
Descriptive analysis of continuous time-to-event data
Fitting proportional hazard regression models
Extensions of regression models to non-proportional hazards
Practicum

Required Text:
Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence by JD Singer and JB Willet, Oxford University Press, 2003

Learning Experiences:
Lecture, discussion, practicum

Evaluation:
Examination 30%
Analysis project 70%