How to identify the study design from a question

Disclaimer: This article is written primarily for students appearing for the Kerala University of Health Sciences Third Professional MBBS Examinations in Community Medicine. However, the tips are of a general nature, and may be helpful for others as well.

With many students appearing for examinations, a common concern relates to the correct identification of study design. In this article I will present some common epidemiological study designs, and provide tips on how to identify the appropriate study design.

Background Information:

All epidemiological studies involve exposure to something (risk factor/ intervention), which leads to a condition- referred to as the ‘outcome’. The study design is influenced by the manner in which exposure happens, and whether the outcome has already occurred or not.

Broadly, epidemiological study designs are either observational or experimental.

Tip#1: When exposure is the result of action(s) by the study subjects, and the investigator does not influence such exposure, one is dealing with an observational study.

Example: Someone steps on a banana peel and falls; the investigator merely watches and records the event.

Explanation: The subject stepped and fell on his/her own- the investigator did not cause the banana peel to be placed so that the subject would step on it–> the investigator did not manipulate the situation.

Corollary to Tip#1: When the investigator manipulates circumstances to ensure that some study subjects are exposed while others are not, one is dealing with an experimental study.

Example: The investigator places a banana peel such that some step on it, while others do not; records observations- how many stepped and fell, etc.

Explanation: The exposure was not determined/ controlled by the subjects. Instead, exposure depended on the positioning of the banana peel by the investigator. This is an example of manipulation. Here, the subjects had no say/choice in exposure, while in the previous situation they could choose whether to step on the peel or not.

I. Observational Studies

There are three main observational study designs: Cohort, case-control, and cross-sectional study.

Cohort study: 
All members of the cohort share at least one common characteristic- residents of a place/ born in a certain year (birth cohort), etc.

At the start of the study, all subjects are free from the condition (outcome) under investigation

Subjects are observed over a length of time (determined by the time required for the condition to manifest/ develop), and each subject contributes multiple observations

The investigator does not influence exposure to the risk factor(s) thought to be associated with the outcome.

New cases are identified at each subsequent visit by the study team

At the end of the study, the number of new (incident) cases are compared among those with and without exposure. This yields the Relative Risk or Risk Ratio. Additionally, Attributable Risk and Population Attributable Risk may be calculated.

Case-Control study:

At the start of the study, some subjects already have the outcome (Cases), while others don’t (Controls).

Exposure has happened in the past, and must be elicited by the investigator.

To adjust for differences between cases and controls, a procedure called matching is employed. This makes it less likely that the difference in status (case/ control) is due to factors other than exposure.

In any condition, those with outcome (cases) are likely to respond differently to questions about exposure, as compared to those without it (controls). When such differences occur systematically, one encounters ‘Bias’.

Several kinds of biases may be encountered in a case-control study.

Cross-sectional study:

Also called a prevalence study, this provides a snapshot of the situation at a point in time.

Crucially, each subject provides only one set of observations- subjects are interviewed/ examined only once by the investigator.

As in other observational studies, the investigator collects information about outcome and exposure- at that point in time- and compares the prevalence of outcome among the exposed and unexposed.

II. Experimental Studies

Although there are several kinds of experimental study designs, I will restrict this discussion to Randomized Controlled Trials (RCTs).

Randomized Controlled Trials:

Subjects agree to receive one or other intervention (exposure) before participating in the study.

Randomization is performed. (Randomization is a statistical procedure that makes the arms of an RCT comparable with respect to known and unknown confounders.)

Subjects receive one or other intervention according to the randomization sequence.

Outcome is measured in each arm and compared with the other.

Key Messages:

Tip#2: Both cohort studies and RCTs involve follow-up. The key is identifying if

  • Randomization is done- if Yes, it is an RCT.
  • Exposure is determined by the investigator or not (observational vs experimental)- if observational, it is a Cohort study; if experimental, it is an RCT.

Tip#3: Cohort studies involve relatively common conditions, and are very expensive to conduct. Therefore, such studies can only be justified when the condition under investigation is a major public health problem.

Tip#4: The outcome has not occurred in a cohort study (or RCT). Determine if there is any follow-up of subjects. Then ascertain how exposure occurred (as mentioned above).

Tip#5: Typically, cohort studies emphasize exposure and non-exposure more than outcome. On the contrary, case-control studies emphasize outcome and absence of outcome more than exposure.

Tip#6: Case-control studies involve bias. Therefore, often, sub-questions ask about bias.

Tip#7: Case-control studies can be completed in a relatively short time. Therefore, they are preferred when limited time is available to complete the study- “during community medicine posting”, for instance.

Tip#8: Case-control studies are employed for rare conditions.

Summary:

Step 1: Determine what the exposure and outcome are in the given question.

Step 2: Determine if it is an observational or experimental study by reading the question carefully

Step 3: Ascertain if key words give away the design (read the sub-questions carefully):

  • ‘Randomized/ Randomization’- RCT
  • ‘Bias/ Matching’- Case-control study

Step 4: Look for confirmation within the main question:

  • Emphasis on exposure/ non-exposure: Cohort study (if observational, and follow-up indicated)
  • Emphasis on outcome/ no outcome: Case-control study
  • Short time for conduct of study: Case-control study (if observational, and emphasis on outcome/ no outcome)
  • Investigation of a new drug/ (therapeutic) intervention: RCT (should be accompanied by mention of randomization/ randomized in the question).