Study Types Flashcards
(26 cards)
What are the defining features of cross-sectional survey?
Cross Sectional Survey – Positivist / Subjectivist, observational descriptive
- Give a snapshot of what is going on in a specific time point: prevalence, health, attitudes, behaviours, lifestyles and demographics.
- Primarily descriptive but can be inferential if a random sample reflective of the population is collected.
- Data: Categorial nominal (named) or binary (yes/ no), Ordinal (linkert scale), Counts – discrete (number of eggs in a sample) and continuous numerical (weight).
- Won’t tell whether a disease is increasing or decreasing – no incidence
What are the defining features of Ecological survey?
- Analysis of data collected for other purposes, e.g. data from a disease registry. Can work out prevalence (like a cross sectional survey), but just using data from a secondary party.
What are the defining features of Qualitative survey?
- Try and find a reason why.
What are the defining features of Cohort survey?
Cohort study – Positivist – observational descriptive
- Longitudinal study: examine the flow over time. Measures incidence. Follow group of individuals overtime. Or separate exposed and non-exposed group followed for an outcome.
- Opposite structure to case control.
- Prospective Cohort Studies: Starting point, follow into future to see what happens. Compare exposed and non-exposed.
- Retrospective Cohort Studies: Look through existing medical records, try and find a causality. Spit exposure and non-exposure.
What are the defining features of Case-control survey?
Case control study – Subjectivist? Observational analytical
- Establishes an association between the disease and outcome: causality. Do not rely purely on correlation though.
- Recruit two groups with one control (no disease) and one exposed (diseased), check histories and try and compare histories for exposure
- Checks risk factors. There is no ‘at risk’ population. Display’s risk in the form of an Odds ratio
What are the defining features of Nested Case control?
- Whereby the cohort gathered have already been exposed to a risk, measured using a corhot study – a known baseline survey has already been done. It’s a case control ‘nested’ in a selected corhort population (rather than being drawn randomly from a population)
- Baseline data are available, and both cases and control are drawn from the same cohort, reducing the risk of bias
What are the defining features of Experimental?
- Control for other factors in an attempt to establish causality. Highly structured. Closed environment. Little outside influence. Reduce variability outside.
What are the defining features of RCT?
Randomised control trial – Positivist – experimental, explanatory
- Measures the effect of a therapy or intervention. Draw sample using simple random sampling. Representative of population. Can also use block allocation & blinding.
- Strict procedure – simple random sampling, allocation concealment (allocating participants is hidden from administrators of the trial), blinding (participants received treatment, single, double or triple blinded depending on the concealment).
What type of research questions do Surveys tend to address or what are the designs best suited to measuring?
Cross-sectional surveys: when want a snapshot of a disease = Prevalence.
What type of research questions do Case report tend to address or what are the designs best suited to measuring?
Case report: new observation, unusual observation. Something of note
What type of research questions do Cohort study tend to address or what are the designs best suited to measuring?
Cohort Study: Longitudinal study, follow some people over time, to establish incidence, incidence rate and causality. Use risk.
What type of research questions do Case-control report tend to address or what are the designs best suited to measuring?
Case-control Study: Where have two groups, control and exposed. When want to establish a correlation, hint towards causality (only if the effect size is huge). Uses odds.
What are the strengths of Survey?
- Many different variables / risk factors compared at once
- Relatively quick, easy to gather and amass data
- Inexpensive
- Much of the data is easily available on record
- Good for prevalence measurement
- Good for public health planning
What are the weaknesses of Survey?
- Temporality – different results at different times – only getting a snapshot
- Cannot measure the incidence, only one point in time
- Prone to bias, unless have a very careful sampling method
- Cluster sampling requires large samples
- Difficult to assign causality
What are the weaknesses of case-control?
- Length of time may limit the study – effects reliability
- Loss of cohort, people drop out – effects reliability
- Measurement: may change over time! – effects reliability
- Selecting a control group is very subjective – there is no ideal control group. Hospital = confounding variables: Selection bias.
- Case definition is hard. A lot of illnesses have the same symptoms with a reproducible definition of case – can’t just rely on a clinical diagnosis, need an objective specific diagnostic technique. (look for definitions in literature)
- Interviewer bias – knows the status & captures or interprets information differently
- Recall bias – case has more knowledge of disease or recalls the information more precisely
What are the strengths of case-control?
- Conducted quickly and cheap
- Good for rare conditions – because tracking back from the two groups, can easily exclude many factors until pinpoint
- Good for conditions with a long latency period – because not having to wait around for a particular disease to occur!
- Good where the population is not stable and where follow up may be difficult – because don’t have to!
- Good when there are multiple exposures – can include it into the survey design
What are the strengths of Cohort Study?
- Identify strong casual links – but need a lot of evidence
- Measures incidence
- Understand the development of the disease
- Can compare groups
- Can look at multiple exposures
- Samples not biased by knowledge of outcome status, they’re defined before disease outcome becomes known (contrast to case control study)
- High validity – true incidence is known
- Good cost: benefit ratio
- Easy to assess cause and effect
What are the weaknesses of Cohort Study?
- Complex calculations required to work out sample size – gPower. Need alpha, baseline prevalence, MoE, confidence level (unlike cross sectional survey)
- Need a baseline survey with an accurate description
- Measurement accuracy needs to be determined in advance, as this can change through the duration of the study (introduction of a better diagnostic procedure)
- Clear case definition and diagnostic measure
- Follow up can be difficult over a long period of time
- Selection bias – content bias
- Unsuitable for rare conditions, need a huge amount of time for sufficient statistical power and cases to appear during the study
What different types of bias are surveys susceptible to?
- Bad survey design – introduce systematic bias
- Sampling bias: wrong target population, need to make sure it is representative: therefore, comes down to sampling method, should you use stratified? Large geographical distribution, cluster? Is the simple random sample logistically too challenging? (Simple random? Stratified? Cluster? Systematic?)
- Systematic bias: error in measurement and sampling. people not available, reduces the target population.
- Sampling frame: bias introduced if the sampling frame trying to represent isn’t representative of the population.
- Measurement bias: diagnostic test accurate? Survey / questionnaire accurate? Incomplete surveys wrongly filled out. Open interviews may sway answers. Face-to face may deter sensitive questions from being answered correctly.
- Selection bias: distorted outcomes arising from the sampling strategy - group selected for study may differ to the actual population
- Performance bias: non-normal behaviour of participant arising from performance within the study. Would arise if the participants didn’t want to disclosure sensitive information for example.
- Reporting bias: Only reporting results that are positive or have relevance
What different types of bias are case-controls susceptible to?
- Classification bias: lots of people change, move from one interest group to another. Start showing same behaviours as control group.
- Sampling bias: definition of case is very hard to pin-point
- Control groups are hard to amass without them having a confounding variable associated
- Interviewer bias / research bias – knows the status of condition, interprets or pushes towards a specific opinion
- Recall bias – the case has more knowledge of disease, therefore recalls information precisely
- Reporting bias: Only reporting results that are positive or have relevance
What different types of bias are cohort studies susceptible to?
- People withdraw from study.
- Attrition bias: distortion due to depletion in the sample as the study progresses
- Reporting bias: Only reporting results that are positive or have relevance
Which study designs measure incidence & which prevalence?
Prevalence - Cross sectional Survey: point and period prevalence - Cohort Study Incidence - Cohort study
What is the starting point of a case-control study? (selecting cases v controls / exposure / disease status?)
1st is case definition: how to define when people are ill (diagnostic criteria)
2nd recruit controls of same population & match. Have to be the same demographic - need to avoid confounding variables
Case-control studies - how are cases defined & controls selected?
Need to identify key suspect. Can’t measure relative risk (unlike a cohort study), because only looking at an assembled group of ill and healthy individuals – we don’t know why they are ill! It is not representative of the population on the whole. We can only look at differences in exposures.
Therefore, we test the null hypothesis that there is no different between ill and non-ill group. Then we can measure the magnitude of effect. Work out an odds ratio.
Controls are selected from the same demographic, but need to limit confounding influence. Never perfect.
