Cohort study Study Guide

📖 Core Concepts Cohort study – longitudinal design that follows a group sharing a defining characteristic (e.g., birth year, exposure) over time, measuring outcomes at repeated intervals. Prospective vs. Retrospective – prospective collects exposure data before outcomes occur; retrospective uses existing records and looks backward. Exposure status – observed naturally; no intervention is assigned. Incidence rate – number of new cases of the outcome per person‑time of follow‑up; primary measure derived from a cohort. Confounders – variables that influence both exposure and outcome; must be measured and adjusted (usually via regression). Nested case‑control – case‑control analysis performed inside an already‑established cohort, preserving temporal ordering while reducing follow‑up burden. --- 📌 Must Remember Cohort = panel study of the same individuals measured repeatedly. No randomization → higher risk of bias; reliance on careful confounder control. Prospective = higher quality (less recall bias, precise exposure measurement). Retrospective = cheaper/faster but limited control over confounding. Key advantage: can estimate risk (incidence) and assess temporal sequence. Key disadvantages: costly, long follow‑up, attrition (loss‑to‑follow‑up). --- 🔄 Key Processes Define cohort – select participants sharing a common characteristic before any have the outcome. Measure baseline exposure & confounders – record status of risk factor and potential confounding variables. Follow participants – collect outcome data at multiple cross‑sectional intervals (or continuously). Calculate incidence – \[ \text{Incidence Rate} = \frac{\text{Number of new cases}}{\text{Total person‑time at risk}} \] Statistical analysis – use regression (e.g., Cox proportional hazards) to estimate the effect of exposure while adjusting for confounders. Interpret – compare incidence between exposed and unexposed; assess strength and plausibility of association. Nested case‑control variant: when a case appears, select matched controls from the same cohort, then perform conditional logistic regression. --- 🔍 Key Comparisons Cohort vs. Randomized Controlled Trial (RCT) Cohort: observational, no randomization, higher susceptibility to confounding. RCT: experimental, random assignment, gold‑standard for causal inference. Prospective vs. Retrospective Cohort Prospective: data collected forward, less recall bias, more costly/time‑intensive. Retrospective: uses existing records, faster/cheaper, limited control over measurement quality. Cohort vs. Nested Case‑Control Cohort: follows entire sample; large resource demand. Nested case‑control: samples only cases + matched controls; saves resources while keeping temporal clarity. --- ⚠️ Common Misunderstandings “Cohort = case‑control” – false; cohort follows participants before outcome, case‑control selects after outcome. “Retrospective cohort is the same as a case‑control” – false; retrospective still follows a defined cohort over time, not just cases vs. controls. “If an exposure is measured later, the study is no longer a cohort” – false; as long as the group is defined before outcome onset, it remains a cohort (even if exposure data are collected later). --- 🧠 Mental Models / Intuition Time‑arrow model: picture a straight line from exposure → latency → outcome. Cohort studies let you watch the arrow move forward; you never have to guess which way it points. “Movie vs. snapshot”: Cohort = full movie (multiple frames showing progression); cross‑sectional = single snapshot; case‑control = picking out a few frames after the climax. --- 🚩 Exceptions & Edge Cases Loss‑to‑follow‑up > 20% may bias incidence estimates; apply sensitivity analyses or weighting. Rare outcomes – prospective cohorts become inefficient; nested case‑control or case‑cohort designs are preferable. Exposure changes over time – treat exposure as time‑varying covariate in regression models. --- 📍 When to Use Which Prospective cohort → when you can afford time/cost and need precise exposure measurement (e.g., new drug safety). Retrospective cohort → when existing records are rich and you need quicker, cheaper evidence (e.g., insurance claims). Nested case‑control → when outcome is rare or follow‑up of the whole cohort is impractical. RCT → when ethical and logistical conditions allow randomization; otherwise, cohort is the next best design. --- 👀 Patterns to Recognize Incidence > prevalence → indicates a cohort (new cases) rather than cross‑sectional (existing cases). Baseline exposure measured before any outcome → hallmark of cohort design. Adjustment for confounders via regression → typical analytic step in cohort studies. Matching within the same cohort → signals a nested case‑control. --- 🗂️ Exam Traps Distractor: “Cohort studies can establish causality without adjustment.” – Wrong; they require careful confounder control. Distractor: “Retrospective cohort studies are the same as case‑control studies.” – Wrong; cohort still follows a defined group forward in time. Distractor: “Loss‑to‑follow‑up is irrelevant if the sample size is large.” – Wrong; systematic loss can bias results regardless of size. Distractor: “If a study collects exposure after outcome, it’s still a cohort.” – Wrong; exposure must be measured before outcome for cohort validity. ---