Course objectives 1. Understand the methodology for developing clinical prediction rules. 2. Assess clinical prediction rule methodology in prescriptive rules used in rehabilitation. 3. Discuss methods to evaluate individual patient change during a PT episode of care. 4. Learn systematic methods for prognostic and evaluative assessments of patient outcomes, using low back pain as a case example. Types of clinical prediction rules Diagnostic: factors relate to a specific diagnosis or condition Prognostic: factors predictive of outcome Prescriptive: Factors target the most effective interventions Identify treatment effect modifiers through history and/or clinical examination Use treatment effect modifiers to select treatment interventions. Advocates argue these are a method for personalized intervention and optimal success Development of clinical prediction rules 1. 2. Validation i. Testing the validity of the rule as it was designed ii. Updating may be required 3. Impact i. Adoption and utilization by providers ii. Improve patient outcome or decrease cost 4. Implementation i. Dissemination for routine use in daily practice ii. Inclusion in clinical practice guidelines Clinical prediction rule validation stages Level Characteristics I II III IV Highest clinical value. Validity established in multiple practice settings, patient populations, with provider diversity. Includes impact analysis for quality of care, cost-effectiveness, outcomes, or efficiency. Confidently applied in patient care. Validity established through broad validation studies. Impact unknown. Validation in sample similar to development study. Not generalizable to broad patient base in routine care. analysis only. Apply only to patients with characteristics consistent to derivation sample. 1
Diagnostic Accuracy Terminology Dethroning the Clinical Prediction Rule Sensitivity (Sn) Probability of a positive test in someone with the pathology. (TP / (TP+FN)) Specificity (Sp) Probability of a negative test in someone without the pathology. (TN / (TN+FP)) Positive Likelihood Ratio (+LR) Ratio of positive test result in someone with condition to positive test result in someone without condition. (Sn / (1-Sp)) Negative Likelihood Ratio (-LR) Ratio of negative test result in someone with condition to negative test result in someone without condition. ((1 Sn)/Sp) Reliability Extent to which the test will yield the same result on successive trials. This course involves a detailed discussion of the studies referenced for each of the following examples. The study design, method, and results (including tables and figures) are shown during the course presentation. They are not included in this handout due to copyright restrictions. Example 1: Clinical prediction rule identifying patients likely to respond to patellofemoral taping Lesher JD, Sutlive, TG, Miller, GA, Chine, NJ, Garber, MB, Wainner, RS. Development of a clinical prediction rule for classifying patients with patellofemoral pain syndrome who respond to patellar taping. J Orthop Sports Phys Ther 2006;36(11):854-866. doi:10.2519/jospt.2006.2208 Key points regarding the development of clinical prediction rules shown in this study Diagnostic quality of the predictive factors is critical to the quality of clinical prediction rules Diagnostic utility can be assessed statistically and clinically Poor reliability of clinical examination factors diminishes the stability of the rule Including poorly performing factors in a clinical prediction rule will lead to future problems The rule should make clinical sense Example 2: Clinical prediction rule identifying patients with LBP who may benefit from lumbar manipulation Flynn et al. A clinical prediction rule for classifying patients with low back pain who demonstrate short-term improvement with spinal manipulation. Spine 2002;27:2835 2843. doi:10.1097/00007632-200212150-00021 Validation Childs et al. A clinical prediction rule to identify patients with low back pain most likely to benefit from spinal manipulation: A validation study. Ann Intern Med 2004;141:920-928. doi:10.7326/0003-4819-141-12-200412210-00008 Different patients from the same sample were used in the derivation and validation The validation was a well-controlled RCT;the derivation failed to include a comparison group 2
Hancock et al. Independent evaluation of a clinical prediction rule for spinal manipulative therapy: a randomised controlled trial. Eur Spine J 2008 17:936 943. doi:10.1007/s00586-008-0679-9 Baseline characteristics were not significant No differences in outcome were attributed to the PT CPR status, treatment, or time was not significant when modelling for pain or disability Validity of the manipulation CPR is not supported Manipulation differed significantly and was predominantly non thrust techniques Cleland et al. Comparison of the effectiveness of three manual physical therapy techniques in a subgroup of patients with low back pain who satisfy a clinical prediction rule a randomized clinical trial. Spine 2009;34:2720 2729. doi:10.1097/brs.0b013e3181b48809 Baseline characteristics are not different Significantly greater improvement in pain and disability for both thrust groups at 1 and 4 weeks Significant difference in disability but not pain at 6 months The results of the study support the generalizability of the CPR to an additional thrust manipulation technique, but not to a nonthrust manipulation technique. CPR + patients respond favorably to thrust manipulation CPR patients were excluded Prohibits comparing response to SMT by CPR status Learman et al. No differences in outcomes in people with low back pain who met the clinical prediction rule for lumbar spine manipulation when a pragmatic non-thrust manipulation was used as the comparator. Physiotherapy Canada 2014; 66(4);359 366. doi:10.3138/ptc.2013-49 Cook CE, Learman KE, O Halloran BJ, et al. Which prognostic factors for low back pain are generic predictors of outcome across a range of recovery domains? Phys Ther. 2013;93(1):32 40. doi:10.2522/ptj.20120216 The results of this study in comparison to prior work Consistent with Hancock et al. 2008 Conflict with Childs et al. 2004 & Cleland et al. 2009 CPR validity upheld only in derivation population CPR fails to adequately predict treatment outcome CPR factors are prognostic; not prescriptive 3
Example 3: Lumbopelvic manipulation for the management of patellofemoral pain syndrome Iverson et al. Lumbopelvic manipulation for the treatment of patients with patellofemoral pain syndrome: Development of a clinical prediction rule. J Orthop Sports Phys Ther 2008;38(6):297-312. doi:10.2519/jospt.2008.2669 Biologic plausibility is unsubstantiated Prior studies did not evaluate symptom response Lumbopelvic manipulation rationale remains theoretical 49 subjects were used to evaluate 41 factors 10-15 observations per final predictor is suggested Observation:predictor ratio dependent on probability Sample size is too small and underpowered 95% CI demonstrates poor estimate precision The regression model is overfit to the data Clinical Prediction Rule Stage Clinical Utility Lumbar manipulation for LBP Flynn et al. Spine 2002 Lumbar manipulation for PFPS Iverson et al. JOSPT 2008 Lumbar stabilization for LBP Hicks et al. Arch Phys Med Rehab 2005 Pilates exercise for LBP Stolze et al. JOSPT 2012 Thoracic manipulation for neck pain Cleland et al. Phys Ther 2006 Traction for neck pain Raney et al. Euro Spine J 2009 Cervical manipulation for neck pain Puentedura et al. JOSPT 2012 Hip mobilization for knee OA pain Currier et al. Phys Ther 2007 -- None Validity not established Rabin et al. JOSPT. 2014 4
Critical clinical prediction rule evaluation: all rules should have these elements 1. Biologic plausibility Clinical prediction rule factors Treatment 2. in studies comparing a control group 3. Adequate sample size 4. Multivariate regression modelling 5. Validation in a unique sample 6. Clear statements of the CPR stage 7. CPR should demonstrate stability 8. Critical consumers of the research product Assessment Battery for Low Back Pain Domain Measure Demographics & History Age, gender, race/ethnicity, education, socioeconomic status, social support, history of LBP, medical history Pain severity NPRS or VAS Functional disability Oswestry Disability Index Roland Morris Disability Questionnaire Quality of Life EuroQol EQ-5D VR-12/SF-12 PROMIS Negative Affect Patient Health Questionnaire-2 (PHQ-2) Beck Depression Index Fear Avoidance Belief/Catastrophizing Fear Avoidance Belief Questionnaire (FABQ) Tampa Scale of Kinesiophobia OSPRO-YF Resilience Factors Pain Self-Efficacy Questionnaire 2-item short form OSPRO-YF Health Behaviors Alcohol: CAGE Questionnaire alcoholism screen Sleep: Pittsburgh Sleep Quality Index Tobacco screening Table adapted from Boissoneeault et al. JOSPT 2017;47(9):588-592. doi:10.2519/jospt/2017.0607 5