CTSI Academy is hosting a Clinical and Translational Research Seminar, four-part series. Lectures will cover a range of topics aimed at facilitating the translation of basic, clinical and population health research.
Laurens Holmes, Jr., MD, DrPH, joined CTSI as Professor and Director of CTSI Education and Training. Dr. Holmes is Board Certified in Public Health, a Fellow of the American College of Epidemiology, and a member of the American Public Health Association. He has over twenty-five years of teaching and research experience in universities both within and outside of the United States. Dr. Holmes has published a number of books in biostatistics and epidemiology, and has developed coursework for both undergraduate and graduate education.
Second Wednesday of the month from 12:00 p.m. – 1:00 p.m.
You are welcome to bring a brown bag lunch!
Health disparities reflects sub-populations morbidity (incidence, prevalence, screening, treatment, severity, prognosis, safety) and mortality outcomes that differ by socio-demographic indicators namely race, ethnicity, race/ethnicity, sex, age, geography, language, disability, income, education, and more recently drug/alcohol, overweight/obesity, acculturation and country of origin. These variances have been conceptualized by the WHO and CDC as absolute and relative. Whereas, the absolute measure assesses the difference in outcomes comparing subpopulations such as the mean or proportion (prevalence difference), the relative measure examines the imbalance in outcomes using one sub-population as the reference group, thus determining the rate, risk or hazard ratio for instance. The underlying issues are: 1) The selection of reference group, 2) When and when not to use relative or absolute measure, and 3) appropriateness of these measures in mapping intervention in narrowing and ultimately eliminating the consistent and perpetual unequal outcomes of health and healthcare in the United States population.
This presentation will highlight:
Clinical and public health decision-making requires an understanding of the pathway to data modeling and analysis in order to draw a reliable and valid inference. Reality in statistical modeling of clinical and translational research data requires sampling as the basis of a design as well as for modeling. This lecture de-emphasizes the notion of P-value as the measure of sample statistic as well as sample precision in the interpretation of data and the application in addressing therapeutics and prevention. Significantly, P-value is not a measure of evidence but reflects the sample size which is of no practical importance to clinical medicine and public health.
Application of statistics is required in quantifying the variation in a sample which may come from biologic variability as well as measurement errors. Because nothing explains everything, there is a need to assess for confounding, effect measure modifier, and quantify the random error. Tabulation analysis is essential in assessing the data as well as facilitating the adjustment of a single confounding variable using stratification model. In contrast, when confronted with many confounding variables, regression model becomes appropriate in performing multi-variable analysis that allows for the adjustment of many confounding variables simultaneously. Since regression hides many attributes of the sample data, tabulation analysis is required prior to any regression model.
Clinical and Translational Research are challenged with the ability to identify and rectify confounding, effect measure modifier (EMM), random error quantification, and bias minimization in the conduct, analysis, and interpretation of findings in improving patient care and public health. Whereas confounding is not a bias per se in our sample assessment, it may lead to a bias estimate of the exposure effect on our clinical and translational outcomes of interest. Similarly, while EMM is not a confounding, the inability to assess and address EMM results in misleading interpretation of our findings, thus masking exposure effect heterogeneity. This lecture stresses the application of reliable tools in assessing and addressing confounding, EMM, systematic and random error prior to model building in the process of scientific evidence discovery.