Biostatistics and Clinical epidemiology

A.Y. 2019/2020
Overall hours
Learning objectives
The aim of the course is to develop knowledge and skills related to the concept of probability, to deal with situations of decision-making uncertainty, to descriptive and inferential statistical methods, necessary to address the problem of biological variability, to aspects relating to critical judgment on the quality of studies published in scientific literature, from which to extract evidence for clinical decisions in the practice of Evidence Based Medicine (EBM).
The student will learn to deal with outcomes of stochastic experiments in terms of probability, having clarified the concepts of the event universe, elementary event and compound event, conditional event, applying the sum and product rules and the Bayes theorem to derive the probability a rear. He will also learn to use the concept of random variable, in the various probability distribution models useful for statistical inference (Gaussian, binomial, poissonian, hypergeometric, Student's t, chi-square model) and to interpret the values of the estimators, calculated on data samples, to make a correct inference, as regards the corresponding parameters of the population.
The student will also learn to recognize the different types of study used in epidemiological-clinical research (observational and experimental studies, cross-sectional and longitudinal studies, cohort and case-control studies), to assess their quality (confounding bias, in observational studies, information bias in experimental studies) and to extract information useful for clinical decision (accuracy measurements from diagnostic studies, estimates of association between exposure and disease from etiological and prognostic studies, measures of efficacy and therapeutic safety from comparative therapy studies).
The student will acquire the skills necessary to set up a database on a spreadsheet and proceed to the calculation of the necessary descriptive and inferential statistics (comparison of averages, simple linear regression analysis, comparison of proportions, analysis of simple and multiple contingency tables, adjustment techniques for confounding, calculation of confidence intervals).
Expected learning outcomes
The student will be able to read a scientific publication at the end, correctly interpreting the various sections of which it is composed: introduction, for the completeness and quality of the premise of the study, for its rationale and objectives; the materials and methods, for the eligibility and exclusion criteria that serve to define the target population, for the definition of the variables detected and considered in various ways in the study, for the appropriateness of the statistical analysis methods; the results, for the critical reading of tables and indexes to be interpreted in their practical meaning and for their correct interpretation for the purposes of application to the clinic; the discussion, due to the different possible interpretations of the results, with particular attention to the weak aspects of the study.
Course syllabus and organization

Single session

Course syllabus
Study designs (examples of observational and experimental studies; secondary studies or literature reviews)
Statistical measures derived from different studies (diagnostic accuracy, disease occurrence, relative risk, efficacy)
Variables: types (explanatory, control, outcome) and measurement scales (nominal, ordinal, interval, rational).
Data quality: accuracy, precision and data management: data matrix, observation unit, use of electronic spreadsheets

Descriptive statistics:
Data tabulation and graph construction: pie charts, bar charts, histograms, cumulative curve, scatter plots, dot plot, box and whisker plot, survival curve
Measures of central tendency and dispersion: centiles, mean, range, standard deviation, interquartile range.

Statistical inference:
Probability models, random variables, expected values (Bernoulli, Binomial, Poisson, hypergeometric, Gaussian)
Random sampling and sampling distributions of statistics (mean, difference of means, proportion, difference of proportions, ratio of proportions, odds ratio, rates, rate difference, rate ratio)
Null hypothesis and alternative hypothesis, rejection area, type I and type II errors, sample size and power of the studies
Confidence intervals (means, proportions, rates; difference of means, difference of proportions, difference of rates; ratio of proportions, ratio of rates, odds ratios)

Data analysis exercises:
Independent and paired samples: comparison of means (Student's t test and z test, confidence interval of the difference between two means); contingency tables analysis: 2x2; Rx2; RxC; RxCxK (Chi square test; confidence interval of association measures for etiological, diagnostic accuracy and intervention studies)
Linear regression analysis and deviance decomposition: the least squares method for estimating parameters.
Confounding factors. Controlling for confounding variables: stratification (Mantel-Haenszel test) and regression models (regression analysis, logistic regression analysis)
Introduction to survival analysis: Kaplan-Meier method, Cox regression and Hazard Ratio (HR) estimation
Prerequisites for admission
Basic mathematics and algebra knowledge are required
Teaching methods
The course is delivered through lectures focused on the concepts and the practical use of elementary methods of statistical analysis, with particular attention to the interpretation of study results and applications in the context of etiological, diagnostic, therapeutic clinical decision. Particular attention will be paid to the critical evaluation of the quality of the different types of study (etiological, diagnostic, efficacy, systematic literature review).
Teaching Resources
Martin Bland (2019) "Statistica medica" Maggioli Editore
Supporting files are available on Ariel website: handouts (pdf), slides (pdf), spreadsheets (xlsx), commented exercises (pptx), scientific articles (pdf)
Assessment methods and Criteria
The examination consists of a 2-hour written test including practical exercises, formulated as open questions or multiple choice questions. Use of pocket calculator and of statistical tables is allowed. A two-step assessment method will be followed. First, some specific exercises (clearly specified) must be solved to reach the sufficiency threshold, without their solution students will not pass the exam. Second, a further series of exercises that must be solved to obtain the final mark. after the correction of the test, the solutions and the mark of each assignment will be communicated through the university website.
MED/01 - MEDICAL STATISTICS - University credits: 3
Lessons: 36 hours
Professor: Duca Piergiorgio