EVIDENCE-BASED PT · CHAPTER 3 · SAMPLING VS. ALLOCATION
Random sampling and random assignment are not the same.
A paper says its participants were randomized. Does that mean the participants represent everyone with the condition? Not necessarily. First ask how they got into the study, then how they got into groups.
1. Who entered the study?
The course separates a target population from the people researchers can reach and those ultimately enrolled. Eligibility rules, recruitment methods and refusals determine who is represented.
Random sampling involves selection by chance from a defined sampling frame. Convenience recruitment from one clinic is different even if participants are later divided into groups by chance.
Quick check: Were 40 volunteers from one clinic necessarily randomly sampled just because treatment assignment was randomized?
Reasoning: No. Recruitment and treatment allocation are different steps.
2. Who received which comparison?
Random assignment uses chance to place enrolled participants in comparison groups. It can help distribute confounding influences on average, but it does not ensure identical baselines, particularly in small trials.
Your lecture also covers blocking, masking, treatment contamination and allocation methods. Assignment concealment before enrollment and blinding after assignment address different sources of bias.
Quick check: Two treatment groups have equal sizes. Does that prove correct random allocation?
Reasoning: No. Equal counts do not establish sequence generation, concealment, balance or complete follow-up.
3. Who might benefit from the findings?
An internally credible trial may enroll people who differ substantially from the patient in front of you. Look at study setting, inclusion and exclusion criteria, baseline health and follow-up.
Ask two separate questions: Can I trust the comparison? Does this sample resemble my patient enough to apply it? Neither question is settled by the word randomized.
Quick check: Can you assume a study in otherwise healthy young adults generalizes unchanged to a medically complex older adult?
Reasoning: No. Applicability needs evaluation even when the study design is strong.
Sources & learning scope
Put it into practice
Scenario: Investigators recruit volunteers from one outpatient clinic and randomly assign them to two treatment programs.
What should you conclude?
Reasoning: Assignment may reduce confounding between treatment groups, but the volunteer sample is not automatically representative of everyone with the condition. Ask separately about recruitment, allocation and generalizability.
Common trap: Treating ‘randomized’ as proof that the population was randomly sampled. Remember: Internal comparison and external applicability require different checks.
Based on my original Evidence-Based PT Chapter 6 Subjects lecture and accompanying research-design notes. Scenarios are newly written illustrations, not copied tests or a validation of a particular article.
