Analytical Studies and Key Statistical Concepts - PDF

Analytical Studies and Key Statistical Concepts Part I Dr Daniel Todkill, Consultant in Public Health Medicine, Field Epidemiology, UK-Health Security Agency / Warwick Evidence, Warwick Medical School Outbreak & Infection Module [email protected] By the end of t...

Analytical Studies and Key Statistical Concepts Part I Dr Daniel Todkill, Consultant in Public Health Medicine, Field Epidemiology, UK-Health Security Agency / Warwick Evidence, Warwick Medical School Outbreak & Infection Module [email protected] By the end of this session you will be able to: Understand why are epidemiological studies important in the context of outbreaks? Be familiar with the main types of epidemiological study designs Cohort studies (use RR) Case Control studies (use odds) Understand the measures of effect of cohort and case control studies How to interpret the outputs of these studies  What is an analytical study in the context of an outbreak investigation? When there is a clear hypothesis (see earlier outbreak lectures!) regarding the source of the outbreak An analytical study allows us to test that hypothesis and estimate the risk of illness associated with a specific exposure They are always observational (i.e. no intervention like a randomised control trial) For example; did the eggs cause the Salmonella outbreak? Study designs and interpreting outputs 4 This Photo by Unknown Author is licensed under CC BY-ND Why do an epidemiological study? understand what happened estimate the extent of the outbreak identify the potential source so it can be controlled and prevented in the future create evidence and knowledge Study designs and interpreting outputs 5 Why do an Descriptive epidemiology can identify a epidemiolog potential source BUT…what if there are cases that do not ical study? have exposure to this source – does that mean it is not responsible? An epidemiological study can give an idea of how likely it is that an exposure is associated with illness Test the hypothesis generated by the descriptive epidemiology Study designs and interpreting outputs 6  From the descriptive epidemiology, you believe eggs to be responsible for the Salmonella outbreak. However; of the 30 people who were ill… Seven of these didn’t report eating eggs. Were the eggs still to blame? Study designs and interpreting outputs 7  Risk = the probability an event will Risk occur …or, better… Risk = the proportion of initially disease-free individuals who develop disease over a defined period of observation Definition does not include perceptions of risk: A dangerous or unknown situation A particularly scary-sounding disease … these are subjective judgements Analytical Study There are two main study designs; Designs 1. Cohort study 2. Case control study Study designs and interpreting outputs 9  Cohort study A cohort study is the study of choice for an outbreak investigation in a well-defined population (i.e. you can identify the full population potentially exposed) In epidemiological terms the cohort is a group of people with something in common, usually an exposure It is an analytical study which aims to obtain additional evidence or refute or support the existence of an association between suspected cause and outcome Study designs and interpreting outputs 10 Cohort study 1/10 divisions in a Roman Legion A group of individuals who share the same experience Often followed up for specific period of time Study designs and interpreting outputs 11 This Photo by Unknown Auhor is licensed under CC BY- Cohort study output Attack rate (risk): is the proportion of people who become ill with a disease in a population initially free of the disease. (Number of cases / Total number of participants) x 100 For each exposure of interest, we calculate attack rates in exposed and in unexposed participants Study designs and interpreting outputs 15 Cohort study outputs Risk ratio/Relative risk (RR) is a ratio of the exposed and unexposed attack rates Relative Risk = (Risk of Disease in exposed) / (Risk of disease in non-exposed) Determines if the attack rate in the exposed is higher than the attack rate in the unexposed attack rate and by how many times Study designs and interpreting outputs 16  Relative Risk RR greater than 1 Risk factor RR = 1 No association RR less than 1 Protective factor Study designs and interpreting outputs 17  Think! Please stop and think of some examples of an outbreak where a Why? cohort study might be appropriate? Study designs and interpreting outputs 18 Examples Weddings are the classic example BBQ’s are another Can usually identify the whole population (guest lists) Can usually contact most guests Can identify who was ill / well Usually there is a limited menu, so can determine who ate what Study designs and interpreting outputs 19  Relative Risk = (Risk of Disease in exposed) / (Risk of disease in non- exposed) Cohort 53/81= 0.6543209 study 65 output 0.1142857 28 Study designs and interpreting outputs 20  Interpretation In this study, people who ate beef had 5.7 times the risk of developing gastroenteritis compared to patients who did not consume beef Study designs and interpreting outputs 21 Study designs and interpreting outputs 22  Case control study A case control study is used when there is not a defined population e.g. increase in cases in a particular geographical area All cases are included and then controls are selected Controls must be from the same population as the cases i.e. they must have had a chance of becoming ill Cases and controls will be asked the same questions Then compare those that were exposed against those that were not Produces an odds ratio Study designs and interpreting outputs 23 Case control study output (1) Odds ratio (OR) is similar to the risk ratio but as you don’t know the total population being studied you have to look at it in a different way A case-control cannot calculate rates / risks (hence, you cannot calculate a Relative Risk from a case-control) The odds ratio estimates the difference of the frequency of exposure between cases and controls Study designs and interpreting outputs 27 2 x 2 tables – An essential tool for epidemiologists Study Results Disease + Disease - Total Exposure a b a+b + Exposure c d c+d – Total a+c b+d a + b+ c + d Using 2 x 2 tables Odds disease in exposed = a/b Study Results Disease + Disease - Total Exposure a b a+b + Exposure c d c+d – Total a+c b+d a + b+ c + d Using 2 x 2 tables Odds disease in exposed = a/b Odds disease in unexposed Study Results =c/d Disease + Disease - Total Exposure a b a+b + Exposure c d c+d – Total a+c b+d a + b+ c + d Using 2 x 2 tables Odds disease in exposed = a/b Odds disease in unexposed Study Results =c/d Disease + Disease - Total Odds ratio = = Exposure a b a+b + Exposure c d c+d – Total a+c b+d a + b+ c + d Case control study output (2) OR greater than 1 Risk factor OR = 1 No association OR less than 1 Protective factor Study designs and interpreting outputs 33  Interpretation A case-control was conducted during an outbreak of Shigella boydii, with 10 cases in total, six having been exposed to chicken. Of the 43 controls, 1 had been exposed to chicken. This gives an OR of 63. What does this mean? Study designs and interpreting outputs 34 A case-control was conducted during an outbreak of Shigella boydii, with 10 cases in total, six having been exposed to chicken. Of the 43 controls, 1 had been exposed to chicken. This gives an OR of 63. Study Results Disease + Disease - Total Exposure a b a+b + Exposure c d c+d – Total a+c b+d a + b+ c + d 6 1 4 42 Study designs and interpreting outputs 35  Interpretation In this study, an OR of 63 means the odds of having eaten chicken were 63 times higher among case-patients than controls. Because the OR is greater than 1, chicken may be a risk factor for illness. The magnitude of the OR suggests a strong association. Study designs and interpreting outputs 36 Study designs and interpreting outputs 37 Think! Please stop and think of some examples of an outbreak where a case control study might be appropriate? Why? Study designs and interpreting outputs 38 Examples Nationally dispersed outbreaks associated with food Take-aways In fact… most outbreaks! Even though we can only estimate the odds of exposure… we can’t usually identify the whole population exposed, hence we need to use case-control. Study designs and interpreting outputs 39 Summary – you should now be able to; Understand why are epidemiological studies important in the context of outbreaks? Be familiar with the main types of epidemiological study designs Cohort studies Case Control studies Understand the measures of effect of cohort and case control studies How to interpret the outputs of these studies (You will not be asked to calculate an OR or RR in your exams… but you might be asked to interpret them!) Acknowledgements EPIET (European Programme of Interventional Epidemiology Training) Dr Valerie DeCrane, Senior Scientist,