Role of Biostatistics
The biostatisticians and data managers of the Department of Biostatistics play a collaborative role in the research studies in which they participate. As true collaborators their main goal is the successful completion of each project. Ideally, this role begins when the idea for the research project begins to take shape. The biostatisticians can then have immediate input on the optimal study design. An efficient experimental design can save money, time and subjects (animal or people). Sample size/power calculations early in the process can lead to an efficient design and eliminate study hypotheses for which meaningful results cannot be obtained within the constraints of the available sample. The study design also leads directly to a preliminary statistical analysis plan drafted by the biostatistician. The design stage of the study is also the appropriate time for the data managers to begin to collaborate on the creation of the data collection forms and to devise the appropriate data management plan.
When the study begins, finalizing the data collection forms and sorting out the details of the protocol are the first priority. The second priority is creating the study database management system. The database will mirror the collection forms and often include additional tables to be used by the research laboratories. They also form a part of the comprehensive data collection and management system that can include data entry via multiple methods, quality control, report generation and even randomization and subject scheduling. This type of complete data system is crafted to fit the needs of each individual study and to minimize data collection and entry as well as protocol violations. Timely report generation is also a key component for tracking the progress of the study and keeping the principal investigator and other personnel updated on any problems or concerns.
The statistical analysis of research data is the task most closely associated with biostatisticians. In the current environment where statistical analysis software is readily available to all and easily implemented by anyone, the ability to run t-tests and regression analyses is not a major contribution. However, current research in the health sciences often involves the generation of data from study designs that violate many of the assumptions of traditional statistical analysis. Correlated data from longitudinal and clustered study designs are commonplace. Missing data patterns must be carefully scrutinized before an appropriate data analysis method can be chosen. Huge amounts of data are now routinely generated with new tools in areas like bioinformatics and genomics. Observational studies based on huge clinical or claims databases must correct for inherent biases. An understanding of the requirements for the traditional methods and current knowledge of new areas of statistical analysis is required in research to compete for funding and to publish results of completed studies. New computationally intensive methods are often needed when assumptions are violated. Only a professionally trained biostatistician can evaluate these issues, choose a correct technique and even devlop new methodology to handle specific problems in a research project.
Methodological development may be extensive and require the biostatistician to obtain additional funding as PI of his or her own grant or it may require a smaller extension of existing methods that can be developed within the initial analysis. In either case, it will improve the analysis and maximize the information obtained from the study data. A strong methodological publication record is a key component in many new areas of research.
Biostatistics also plays an important role in preparing study results for both presentations and papers. Concise presentations of descriptive and inferential results using graphs and tables are part of our training. After all, a graph is worth a thousand words, and a key statistic is worth at least a thousand data points. The biostatistician also contributes the statistical methods section and often is a key player in the interpretation of the results. Authorship for members of the biostatistics team should be based on the accepted criteria of most journals. These criteria generally cite both study design and statistical analysis as intellectual input sufficient for authorship. Monetary support, either as salary support on grants or reimbursement for time, does not preclude or replace authorship.
For a comprehensive review of HIPAA regulations and how they affect research on this campus, please refer to http://www.iu.edu/~vpgc/compliance/hipaa-privacy-and-security/index.shtml. Although the primary investigator is ultimately responsible for ensuring HIPAA compliance, the Department of Biostatistics plays an integral role in the proper storage and use of PHI. Our data management experts are here to help you with ideas for data collection systems and transfer. In addition, all staff members have received basic HIPAA training and can help you identify problem areas both at the grant/protocol development stage and once the project begins in the research process stage.
Often investigators would like to conduct analyses using data collected from studies that have been closed (i.e. IRB approval has been terminated). These studies most often will need to be re-opened before we can access the data. Please see Conducting Analyses on Retrospective Data to determine whether or not your study will need to be reopened.
It is absolutely essential that investigators include Biostatistics as early as possible in the development process. This should be well before the grant/protocol deadline to take full advantage of the available expertise in designing the study. As R.A. Fisher noted way back in 1938, "To consult a statistician after an experiment is finished is often merely asking him to conduct a post mortem examination. He can perhaps say what the experiment died of."