CM
● MODULE 03 / 20
Study Designs & Causal Reasoning
Match a research question to its design and recognize bias, confounding and uncertainty.
Quick revision ↓ Medical Dictionary
FULL CHAPTER EXPLANATION
Deep Learning — Module 03
Connect population needs and measurements to prevention, practical action and evaluation.
01
Descriptive and cross-sectional studies
Descriptive epidemiology characterizes person, place and time and generates hypotheses. A cross-sectional survey measures variables at a defined time, often estimating prevalence. When exposure and outcome are measured together, temporal order may be unclear. Ecological studies compare group-level measurements; a relationship between districts does not prove the same relationship for individuals. Choose the unit of analysis before interpreting the result.
02
Cohort and case-control designs
A cohort groups people by exposure and follows or reconstructs subsequent outcomes, allowing risk or rate comparisons when denominators are available. A case-control study samples by outcome and compares prior exposures. Controls should represent the exposure distribution in the population that generated the cases. Ordinary case-control sampling does not directly estimate disease risk from the proportion of sampled participants who are cases.
03
Trials and allocation
Random allocation aims to balance known and unknown prognostic factors in expectation. Allocation concealment prevents recruiters predicting assignment; blinding reduces behavior or assessment differences after assignment. They solve different problems. Intention-to-treat analysis generally preserves the randomized comparison by analyzing assigned groups. Loss to follow-up, nonadherence and selective reporting can still undermine interpretation. Randomization is not the same as random population sampling.
04
Bias and confounding
Selection bias arises when inclusion or retention distorts the comparison. Information bias arises from measurement or classification error; recall differences can affect reported exposure. Confounding mixes the exposure effect with another cause associated with exposure that is not simply an intermediate step. Restriction, matching, stratification and appropriate adjustment may help. Effect modification is variation in effect across groups and may be a finding worth reporting, not an error to erase.
05
From association to action
A causal claim needs temporal order and a defensible explanation of alternative causes, measurement and design. Strength, consistency, dose-response and biological coherence can support reasoning but are not a mechanical checklist that proves causation. Statistical significance does not remove bias. A systematic review uses explicit methods to identify and assess evidence; meta-analysis pools suitably comparable results but cannot repair weak source studies merely by adding them together.
High-Yield Revision
- Sampling by outcome is the hallmark of case-control design.
- Concealment and blinding address different risks.
- Confounding and effect modification are not synonyms.
Common Exam & Concept Pitfalls
- Do not infer individual relationships from group averages alone.
- Do not treat statistical significance as proof of causation.
Sources & Further Reading
Curriculum and reference review: 24 September 2026. These original study notes summarize established principles. Use current programme schedules, local reporting requirements and clinical protocols for practice. Faculty review is pending.
- PM&DC MBBS Curriculum 2024 — Community Medicine and Public Health, printed pages 61–65
- Measurement and screening: CDC Principles of Epidemiology (archived foundational text) · WHO screening programme guide
- Services and surveillance: WHO primary health care · Pakistan NIH surveillance bulletins
- Immunization: Pakistan FDI official schedule · CDC timing and spacing principles. Apply Pakistan programme guidance to local schedules.
- Family health: WHO maternal health · Contraception · Infant feeding
- Nutrition: WHO malnutrition · Child anthropometric indices
- Environment and work: WHO drinking water · WASH · NIOSH hierarchy of controls
- Vectors and exposure: WHO dengue · Malaria prevention · Rabies · Snakebite
- Long-term health: WHO noncommunicable diseases · Mental health promotion and prevention
- Preparedness: WHO Health Emergency and Disaster Risk Management Framework
RAHBAR study structure: population need → measurement → prevention → action → evaluation. Use institutional curriculum and faculty guidance for local sequencing, practical requirements and assessment specifications.