Reading path

How to Read an Observational Study

Most medical questions can never be settled by a randomized trial, so much of what we know rests on observational studies that simply watch what happens to people over time. This path walks you through the ways those studies can mislead, from confounding and selection bias to the quiet traps of time and interpretation, and the methods researchers use to get closer to cause. By the end you will be able to read an observational finding and judge for yourself how much of it to believe.

The path, step by step

1. 1 Confounding and Causation: Why a Real Association Can Still Lie to You

   6 min read

   Start here, because every later idea builds on the hardest lesson in observational research: two things can move together without one causing the other.

2. 2 Understanding Selection Bias, the Quiet Distorter

   4 min read

   Before trusting any association you have to ask who ended up in the study, and this shows how the answer can bend a result before a single number is calculated.

3. 3 Confounding by Indication: Why the Reason for Treatment Distorts Drug Studies

   5 min read

   This is the confounder that haunts observational drug studies specifically, since the reason a patient was prescribed a treatment is often the very thing that predicts their outcome.

4. 4 Immortal Time Bias: Why a Treatment Can Look Better Than It Is

   5 min read

   A subtler trap where the way follow-up time is counted can hand a treatment credit it never earned, and learning to spot it changes how you read survival claims.

5. 5 The Table 2 Fallacy: Why Not Every Number in a Regression Is a Causal Effect

   5 min read

   Now that you know bias exists, this explains why statistical adjustment does not rescue you, and why not every adjusted number in a results table deserves to be read as a cause.

6. 6 Negative Controls: Catching Bias by Looking Where There Should Be No Effect

   5 min read

   A practical detective's tool: by looking where there should be no effect at all, researchers can expose hidden bias that the main analysis would otherwise conceal.

7. 7 Instrumental Variables: Estimating a Cause When You Cannot Measure Every Confounder

   5 min read

   When you cannot measure every confounder, this method finds a variable that mimics random assignment, opening a path to causal answers from non-randomized data.

8. 8 Mendelian Randomization: Using Inherited Genes as a Natural Experiment

   5 min read

   A powerful real-world use of that idea, using the genes people inherit at conception as a natural experiment that ordinary confounding cannot reach.

9. 9 Does Treating Sleep Apnea Prevent Heart Attacks?

   5 min read

   A fitting capstone: a strong, repeated observational link that a randomized trial failed to confirm, showing exactly why the whole toolkit in this path matters.

Each step is a full article on the Reading the Evidence blog.

Back to all reading paths