Epidemiology - Public Health

Epidemiology

·        Includes 2^nd and 3^rd year PDS notes

·        Study of distribution and determinants of health and disease in populations

·        For Definitions of Risk and Odds Ratios

·        Bradford Hill Criteria for causation:

o  Time sequence: did the factor come before the disease – only necessary factor

o  Strength: large relative risks are seldom artefacts

o  Consistency: has an association been found elsewhere, using different methods?

o  Specificity 

o  Dose-response relationship: does the risk ­ with the level of exposure to the factor

o  Biological plausibility: is a causal risk consistent with current knowledge

o  Experiment: can an intervention study validate the result

·        Summary of study types: (increasing power as you go down)

o  Descriptive (ecological) studies

o  Analytical (observational) studies:

§  Cross-sectional studies

§  Case-control studies

§  Cohort studies

o  Intervention (experimental) studies

Descriptive studies

·        Personal characteristics (age, sex, occupation, social class, etc), place and time 

·        Measures of occurrence (Þ descriptive):

o  Prevalence: Number with disease/Relevant population at a designated time 

o  Incidence: Rate of development of a disease in a group over a period of time = number of new cases during a specific period/total at risk population

o  Need to compare for specific groups, or standardise for compositional differences in populations 

o  Long duration illness ® ­prevalence, short duration ® ¯prevalence 

o  Incidence and prevalence are similar for short illnesses (eg diarrhoea) but not for long illnesses (eg TB)

·        Useful for:

o  Understanding occurrence

o  Suggesting hypothesis – but correlation doesn‟t prove cause

o  Often called ecological studies

·        Don‟t have unaffected people in the series so no information about relative risks

Cross-sectional studies

·        Examine relationship between disease and other variables (eg risk factors) in a defined population at one time

·        If prevalence not high enough, use sentinel populations (ie those with greater risk) 

·        Normally considering prevalence, not incidence Þ also called Prevalence Studies

·        Limitation: time sequence of cause and effect cannot be determined

Case-control studies

·        Factors amongst patients with a disease compared with the frequency among a control group without the disease 

·        Looking backwards to the exposure (ie the outcome has already happened) Þ also called retrospective studies 

·        Advantages:

o  Can be done quickly (compared with cohort study)

o  Are therefore cost effective

o  Can do it with diseases with low prevalence (cohort study better for more prevalent diseases)

o  Defined endpoint

·        Disadvantages:

o  Retrospective, so can introduce bias

o  Selection criteria create difficulty

o  Biases: non-volunteer bias, recall bias, etc

o  Causation not proven – just an association

o   Can only get information on one outcome

·        Selection of cases:

o   Representative of those presenting over a defined period 

o   Select according to diagnostic criteria. If these cast the net widely, then will include people with unrelated disease ® odds ratio moves towards 1 (if random effect)

·        Selection of controls:

o   Representative of the population from which cases drawn but without the outcome

o   Aim is to establish expected exposure in case group 

o   Want to compare exposures between the cases and the population at large. If, for example, you match on age, you won‟t be able to conclude anything about age as a risk factor, so choose controls so you get an unbiased example of risk factors in the population. Matching does reduce the effect of confounding – but there are other ways to deal with that when doing the number crunching

·        Identical methods must be used to collect information from both

·        Can‟t estimate relative risk in a case-control study:

·        Odds Ratio = (A/C)/(B/D) 

·        Relative Risk

·        = [ A/(A+B) ] / [ C/(C+D)]

·        = Risk of disease among exposed/risk of disease if not exposed) 

·        In a case-control study, the ratio between the cases and the controls is fixed by the study design, not the prevalence in the population. The relative risk is therefore meaningless. That is, you can‟t estimate the risk of the disease in the population in those not exposed to the risk factor 

·        But if the condition is rare and there is no bias in controls, then the odds ratio approximates the relative risk ratio well

·        See also Topic: Risks and Odds

Cohort Studies

·        Groups observed over time: forward looking from exposure, waiting for outcome Þ also called Prospective or Follow-up Studies 

·        Either compare within the cohort, or select a comparison cohort

·        Advantages: 

o   ¯Bias over a case-control study

o   Can examine more than one outcome

o   People who don‟t take part don‟t bias the outcome, just its generalisability

·        Disadvantages:

o   Large numbers are required Þ information collected has to be simpler Þ ­ risk of confounding

o   Need to follow people over time Þ expensive and takes a long time

o   Not good for rare diseases 

·        Strengths: People who don‟t take part don‟t bias the result – just affect its representativeness (ie external validity) Þ less susceptible to bias

Intervention or Experimental Studies

·        Test whether removing suspected aetiological factors reduces frequency of disease 

·        Evaluate preventative measures and treatments (® clinical and community trials)

·        Best test of cause and effect

·        Often not ethical nor logistical to do a RCT 

·        Blind assessment ® ¯ information bias