Observational studies are designed to predict some sort of relationship between variables. They usually start with an observation of some sort of association. Like chocolate consumption being linked with Nobel prizes. An article in the New England Journal of Medicine shows a correlation between the national consumption of chocolate and the national yield of Nobel prizes . A good summary of the strengths and weaknesses of this association study is found at http://www.reuters.com/article/2012/10/10/us-eat-chocolate-win-the-nobel-prize-idUSBRE8991MS20121010article
Association studies piqué curiosity. They report on association, though, not cause. There is no evidence that chocolate consumption leads to higher academic performance, any more than umbrellas cause rain. Studies that examine cause go two ways. Cohort studies follow a large cohort prospectively to determine how exposure to an educational event or process, or a cohort’s specific characteristics relates to outcomes. An example may be, does attending a medical school with PBL predict future competence as a clinician? So you might follow a cohort, some of whom have done PBL and others who have not, and track their performance. A nice Australian example is Sarah Larkins’ study of medical students. Sarah looked at scores on a depression and anxiety instrument in medical school, and tracked whether high scores predicted poor performance as GP registrars (the answer was yes).
Case-control studies start with the outcome, and then do a retrospective analysis to determine how the outcome relates to exposure to an educational event or process, or a cohort’s specific characteristics. An example might be looking at registrars requiring remediation, and determining whether they are more likely to have had prolonged exposure to general practice in medical school. Case control studies are cheaper, because they do not need to track participants over a long period of time like cohort studies, and they are more useful for outcomes that are rare. Seeing as less than 1% of our registrars need formal remediation, you could track a cohort for years, and only have one or two with the outcome you are studying. Case control studies avoid this dilemma, but they do have the disadvantage of recall bias: they are asking retrospective questions.
One issue with predictive studies is deciding what you will do with the data. It is important to ask a question that can not only be answered, but can lead to helpful educational change. For instance, if a hypothetical study found that participants with blue eyes are more likely to choose rural terms, it is unlikely we will change anything. Could you imagine the outcry if eye colour was used as a selection criterion?? Idle curiosity masquerading as predictive observational studies is research waste. Unless of course it is associating chocolate intake with Nobel prizes. That has spawned all sorts of debate, and probably increased the average chocolate consumption of researchers around the world.