go back (01aug2003/27jan05)
DrTim homepage link
CH1 Introduction
Intro to Clinical Statistics by Timothy Bilash MD
August 2003
www.DrTimDelivers.com
based on:
Review of Basic & Clinical Biostatistics by
Beth Dawson, Robert Trapp (2001) CH1
 Three types of investigation^{1}*
 Scientific Investigation (Experiment)
 test hypothesis for consistency
 control conditions (expose all) or use control groups (compare to nonexposed)
 randomize to improve accuracy when cannot control exposure
 all exposure risks must be known (controlled and/or randomized)
 all outcomes results must be observed (controlled and/or randomized)
 descriptive statistics summarize results within known group.
 look for changes in outcome for an individual within group
 most accurate means we have to determine causes and effects
 EvidenceBased Medicine (BioStatistics)
 uses sampling statistics to determine average behavior and correlations
 exposure (risk) is not known for sure (controlled), but determined statistically
 outcome (result) is not known for sure (controlled), but determined statistically

uses inferential statistics to determine which variables are exposures and which are outcomes
 not necessarily causal, because unable to control after the fact and nonrandom for many factors. in particular, there is expsoure bias, sampling bias and timedependent effects.
 uses statistics to determine behavior of the group as if they were all the same individual
 Consensus (Popularity)
 determines what is correct by how many believe it, or what is assumed to be how many believe it (Gossip)
 bias very common, creates consistency by arbitrarily excluding data, very unreliable
 those who control the information and promote the information determine its accuracy
 Biostatistics
 mathematics
 using statistics to
 summarize data
 reach certain conclusions or absence of conclusions
 determine the reliablity of results thru statistical techniques
 Epidemiology
 study of health and illness in human populations (groups)
 patterns of health or disease and the factors that influence these patterns
 evidencebased medicine is NOT science
 also called clinical epidemiology
 the application of populationbased information to decision making about individuals
 evidencebased medicine is mathematics
 math functions and summaries
 outcomes not directly linked causally to an individual, but to the population summary of many individuals
 requires proper summary
 requires inferring application to an individual via membership in the group
 uses biostatistics
 Errors in published reports
 1986 Williamson et al
 assessed 30 journals
 only about 20% of 4235 research reports met criteria for validity
 subset of 8 articles
 inadequately designed and analyzed reported 80% positive findings
 properly designed and analyzed had only 25% positive findings
 thus poor studies more likely to find effect
 more recent articles indicate that the problems have not improved
 Avram
 Newer statistical concepts
 number needed to treat (to prevent one undesirable outcome)
 likelihood ratio (analyze risk factors to outcomes)
 odds ratio (predict outcomes from risk factors)
 power (number of subjects needed to detect and effect)
 risk
 multivariate methods
 statistically (mathematically) complex
 logistic regression
 cox proportional hazard model
(*1 Author's Note revised 01.27.2005)
Back
www.DrTimDelivers.com
[an error occurred while processing this directive]page views since Sept2007