FROM MY PT SCHOOL NOTES · EVIDENCE-BASED PHYSICAL THERAPY · CHAPTER 12
Can Real-World Outcomes Research Prove a Treatment Caused Improvement?
Clinical datasets can be useful, but association and causation require different levels of support.
What outcomes were recorded?
The lecture defines outcomes research as examining the impact of clinical care as it occurs in real-world practice. It often uses existing clinical records or outcomes databases, but may also collect data prospectively. Designs can be retrospective or prospective and cross-sectional or longitudinal.
Which differences could distort the result?
Credibility checks from class: are groups reasonably comparable, are patient characteristics and interventions described, are outcomes clearly operationalized, were standardized measures used, and were data collection and missing records handled consistently? The slides discuss risk adjustment for important baseline and prognostic differences.
Does it fit clinical practice?
The lecture uses an adhesive-capsulitis outcomes study by Jewell and colleagues as a class example. The emphasis is on scrutinizing diagnosis codes, intervention categories, differences between included and excluded cases, and whether reported improvement is attributable to the treatment rather than another explanation.
Outcomes may include functioning, participation, symptoms, self-report results or other health events. The chapter references effect sizes, minimal clinically important differences, odds or risk ratios, confidence intervals, and whether the findings apply to similar patients in actual PT settings.
Check your reasoning: A retrospective record review finds improvement after a treatment. Does this alone prove that treatment caused it?
No. Selection, co-interventions, natural history and missing or miscoded data may contribute.
Updated-standard context
Large retrospective clinical datasets can be useful but are not automatically causal evidence. Data coding, confounding, differences in care and missing outcomes can distort interpretation. Do not read the source’s ‘ratios equal one = a coin flip’ phrase literally; a ratio of 1 indicates no association on that ratio scale, not necessarily a 50% probability.
Put it into practice
Scenario: A clinic’s database shows that average mobility scores improved during a new program, but everyone received other care at the same time.
What should you conclude?
Reasoning: The improvement is a real observation if measured reliably, but this before-and-after dataset alone cannot separate treatment effects from natural history, selection or concurrent care.
Common trap: Turning ‘outcomes improved’ into a causal claim. Remember: Ask what comparison would identify an effect.
Based on my Evidence-Based PT class-notes companion, source lecture Chapter 15. Examples and knowledge checks are educational, not official NPTE material. Compare with your course and current primary sources.
