Health data can be divide into two types: retrospectively and prospectively collected.
Retrospective/historical studies are sometimes referred to as observational studies. However, some observational studies can be prospective also.
i) Cohort studies
A cohort of patients, selected based on them meeting certain criteria, are examined over some window of time. The window for each patient may be at the same calendar time or not, or over their lives from a certain age until an event (death, for example). By holding certain characteristics equal we can judge what effect environment or genetic factors has on measured characteristics of interest, eg lifespan. Moreover, since the individuals are followed through time the incidence or onset of disease across patients can be examined.
ii) Cross-sectional studies
These take a sample across the population in an instant of time, eg in different age brackets or sexes or socio-economic class and compare measures of interest for example height, weight, smoking status. This can be linked to an outcome of interest such as prevalence of heart disease.
iii) Case-control studies
Individuals are chosen retrospectively who are matched (preferably, to avoid confounding) except in some chosen variable, eg whether they smoke or not. The smokers are the cases and the non-smokers are the controls. The effect of smoking on the prevalence of lung cancer for example can then be estimated.
i) Cohort studies
Individuals chosen because they meet certain criteria are followed forwards in time and observations about them made. Unlike a retrospective study, individuals may drop out of the study unexpectedly (for reasons assumed not to be related to the measurements being taken). For example, they may move away from the catchment area or decide not to participate anymore. This is an observational study because no intervention is made.
ii) Cross-sectional studies or surveys
At a certain time a study is made across the population of interest. A collection of variables and the outcome of interest are measured. Thus inferences can be made about causal links between measured variables and the outcome of interest. Unlike retrospective cross-sectional studies individuals cannot always be relied upon to consent to being measured. They may not answer the phone or wish to fill in a questionnaire.
iii) Intervention cohort studies
These are informal clinical trials, where individuals are given a drug, say, that may improve their health. The patients will simply be tested to see how their health improves. Judgments about which type of patients respond better to treatment might be made based on measurements of certain variables. A formal study like this, where before and after measurements are compared within individuals is called a crossover trial.
iv) Intervention case-control studies or controlled clinical trials
There are usually two 'arms' to the trial - individuals given the intervention and individuals not given the intervention. The 'gold standard' is to randomise individuals to each arm, and further to 'blind' them against knowing whether they have received the treatment or a 'placebo'. Biases are then protected against.
These are the seven types of collected data in health studies. The statistical evidence that the studies provide is used to tailor future care, introducing new drugs, procedures and recommending life-style choices such as quiting smoking to improve health.