CPR APPLICATIONS FOUNDATION · CHAPTER 20 · ABNORMAL
Abnormal Foundations of Biostatistics & Epidemiology
Organize the chapter before drilling into individual CPR concepts.
Supplemental foundation map: this page orients the detailed CPR Applications for Chapter 20. It is not part of the canonical CPR object inventory.
Move through Chapter 20
Normal Foundations
Abnormal Foundations · You are here
Treatment Foundations
20.18 · SOURCE CPR FOUNDATION
Concept
Biostatistical interpretation fails through bias, random error, confounding, invalid inference, and causal overreach. A result can be precise but biased, significant but unimportant, or strongly associated but not causal. Step 1 frequently tests why a conclusion is not justified.
Pattern
Step 1 gives an abstract where the authors overstate results, use the wrong denominator, ignore loss to follow-up, rely on recall, fail to blind outcome assessment, or confuse association with causation. The task is to identify how the reasoning failed.
Rule
- Systematic bias threatens validity by distorting estimates in a particular direction; bias may increase or decrease the apparent effect.
- Random error threatens precision. It widens confidence intervals and reduces power.
- Association does not prove causation. Temporality, confounding control, biologic plausibility, and study design matter.
Applications
- Randomized trial with appropriate allocation, blinding and follow-up → generally supports internal validity; a narrow CI indicates precision but does not rule out bias
- Unblinded outcome assessment → observer bias
- Wide CI → random error or small sample concern
- Observational association after measured-confounder adjustment → causal inference still depends on design, unmeasured confounding and other biases
- Significant result from many subgroup analyses → false-positive concern
- Tip: A precise wrong answer is still wrong
CPR APPLICATIONS · ACTIVE REASONING
Identify threats to valid inference
Commit to an answer before requesting feedback. Full worked explanations remain available if scripts are unavailable. Answers are not transmitted or stored.
Reveal the worked answer and changed-variable challenge
Best answer: Differential recall bias. Cases may report remote exposure differently because of disease awareness, distorting the observed association.
Transfer: How might prospective exposure records reduce the concern?
Reveal the worked answer and changed-variable challenge
Best answer: Confounding by smoking. Smoking can relate to both measured exposure and outcome, creating a spurious or exaggerated association if inadequately addressed.
Transfer: Would adjustment for measured smoking patterns eliminate every possible unmeasured confounder?
Reveal the worked answer and changed-variable challenge
Best answer: Multiplicity increases the chance of a false-positive finding. When many hypotheses are tested, nominal significance in one analysis may arise by chance without proper adjustment or confirmation.
Transfer: How would prespecification and independent replication affect the confidence in an observed interaction?
Change the mechanism
Read all comparisons without scripts
- Case–control retrospective recall: Differential recall by case status can bias exposure estimates.
- Observational exposure-outcome association: Measured and unmeasured confounding can undermine causal interpretation.
- Many post-hoc subgroup tests: The familywise chance of false-positive findings rises without multiplicity control.
Explain it back: Identify the mechanism, name the most important competing explanation, and predict what changes when the context shifts.
Unscored educational reasoning. Not a validated assessment or individualized clinical recommendation.
How to use this foundation map
- Identify the organizing mechanism.
- Recognize the pattern.
- Restate the rule without looking.
- Test transfer with applications and question practice.
Pair CPR with the 5P Approach™ when solving cases and questions.