Design of Experiments > Retrospective Study
- Retrospective Study
- Terminology Note
- Advantages and Disadvantages
- Study Steps
- Retrospective Cohort Study Definition
- Types of Retrospective Cohort Study
- Retrospective longitudinal study
A retrospective study is an observational study that enrolls participants who already have a disease or condition. In other words, all cases have already happened before the study begins. Researchers then look back in time, using questionnaires, medical records and other methods; Basically, you just dig into the data and see what you find. The goal is to find out what potential risk factors or other associations and relationships the group has in common. The opposite of a retrospective study is a prospective study where participants are enrolled before any of them have the disease or outcome being investigated. When both retrospective and prospective methods are used at the same time, the study is said to be ambi-directional.
Unlike most other studies, a retrospective study collects data that have been previously collected for some other reason than research (Hess, 2004).
In epidemiology (i.e. in clinical studies), “case-control” and “retrospective study” are used synonymously. That’s mostly because when dealing with diseases and conditions, you always want to have a control. A historical epidemiological study without a control would be unthinkable, and perhaps even useless. Therefore, if you look at clinical studies, medical sites, or anything to do with medicine, you’ll find the two terms are interchangeable.
However, in other areas (e.g. education, the social sciences), there are different types of possibility for studies such as a retrospective case series, which do not use controls at all.
- Useful for rare diseases or unusual exposures.
- Smaller sample sizes.
- Studies take less time, because the data is readily available (it just has to be collected and analyzed).
- Costs are generally lower.
- Missing data: Exposure status may not be clear, because important data may not have been collected in the first place. For example, if the study is inverstigating occupational lung cancer rates, information about worker’s smoking habits may not be available.
- Recall bias: Participants may not be able to remember if they were exposed or not.
- Confounding variables are difficult or impossible to measure.
- Retrospective studies are considered to be inferior to prospective studies, so prospective studies should always be used if there is a choice.
- As this is a relatively weak type of study, you cannot make causal statements, although correlations are okay (see: causation vs. correlation). Therefore, getting the study read and/or published may be difficult.
(Adpated from Kalogeropoulos, 2014):
- Clearly define:
- The study population.
- The time period (how far back in time you’ll get data from).
- Outcomes (are you studying a specific disease outcome? An event occurrence? Something else?
- Collect as much data as possible — preferably quantitative (numerical) data.
- Decide how you’ll defend your study before you implement it.
- Carefully design the database so that you’ll be able to easily analyze your results.
- Enlist other people to help, if possible. For example, you may benefit from getting a database expert to help you design your database.
In a retrospective cohort study (also called a historic cohort study), researchers look back at a certain period of time to find a group’s exposure to the same risk factor. A cohort is a defined group, like “nurses,” “10-19 year-olds,” or “college students.” Participants are chosen for a reason, rather than randomly.
For example, researchers may want to investigate whether exposure to glues commonly used in carpentry increases the risk of developing COPD. A cohort consisting of retired carpenters might be selected. A control group is also chosen. This might be made up of delivery drivers and clerical workers who would not have been exposed to the glues. Health records and employment records are used for data sources. Due to the fact that data is collected retrospectively in a non-controlled environment, it’s not possible to make statements about causation.
Although you can’t make statements about causation, you can find associations and possible relationships, potentially paving the way for the more expensive, longer-term prospective study.
Case-control studies involve two groups of people: people who have the disease (cases) and those who do not (controls). A retrospective case-control uses these two groups and looks back to the past for data and possible risk factors. A matched case-control study chooses controls based on some matching factor, like age, weight or severity of disease.
A case-series is just a series of cases. For example, a physician might encounter a series of patients who all have the same disease. They then look back retrospectively to try and find associations between the patients. The difference between a retrospective case series and a retrospective case-control is that the case series lacks a control group. As such, it is a much weaker design than case-control.
A retrospective longitudinal study involves repeated observations of the variables over a long period of time.
Hess, D. (2004). Retrospective studies and chart reviews. Respiratory Care. 2004 Oct;49(10):1171-4.
Kalogeropoulos, A. (2014). Understanding Retrospective vs. Prospective Study designs. Retrieved October 26, 2017 from: http://medicine.emory.edu/documents/research/kalogeropoulos-study-design-talk.pdf
Sahai, H. & Khurshid, A. (1995). Statistics in Epidemiology: Methods, Techniques and Applications. CRC Press.
If you prefer an online interactive environment to learn R and statistics, this free R Tutorial by Datacamp is a great way to get started. If you're are somewhat comfortable with R and are interested in going deeper into Statistics, try this Statistics with R track.Comments? Need to post a correction? Please post on our Facebook page.