Research Methodology
Domain 3 - Research Methodology
This domain assesses whether you can design a defensible study and judge how well one was conducted - a compulsory component of PGIM training.
Core rules
- Frame the question with PICO (population, intervention or index test, comparator, outcome) and test it against FINER (feasible, interesting, novel, ethical, relevant).
- Design follows the question: RCT for efficacy, cohort for incidence and prognosis, case–control for rare outcomes, cross-sectional for prevalence, diagnostic-accuracy studies for test performance, qualitative designs for meaning and experience.
- Probability sampling (simple random, systematic, stratified, cluster) permits inference; convenience sampling does not. Cluster designs need a design effect and an inflated sample size.
- Bias is designed out, not analysed out: selection, information/recall, interviewer and confounding. Randomisation with allocation concealment handles confounding in trials; matching and restriction handle it in observational work.
- All human research requires ethics review committee approval and informed consent before data collection; PGIM expects this before the study is registered.
Worked example. You want risk factors for leptospirosis in a district reporting 60 cases a year. A cohort would need tens of thousands of person-years to accrue enough cases. A case–control study with 60 notified cases and 120 controls from the same MOH area is feasible. Matching on age, sex and residence controls confounding; the dominant threat is recall bias about water contact, reduced by a structured exposure list and by interviewers blinded to case status.
Traps
- Labelling any questionnaire survey "cross-sectional" and then drawing causal conclusions.
- No sample-size justification, or calculating it for a mean when the outcome is a proportion.
- Consent obtained after the data have been collected.
- Ignoring STROBE or CONSORT when writing up.