Clinical Prediction Rules

One of the most important parts of a chiropractor’s daily work with patients is the need to establish a proper and correct diagnosis. Diagnosis is key to the requisite decisions for appropriate care than then follow. As teach students, we generally use heuristic models, where the thinking is linear and directly correlated. That is, students are taught that if a given orthopedic test is positive, it has a specific meaning; if McMurray’s test is positive it means that there is a torn meniscus, for example. As we gather information from an examination, we combine it with our clinical experience and knowledge to close in on what feel is the correct diagnosis. It is for this reason that at Palmer we have focused on training our students to understand the importance of sensitivity and specificity of diagnostic tests, and to derive likelihood ratios from that information.

But this is imperfect at best. Clinical prediction rules (CPR) are designed to try to enhance the accuracy of a clinician’s diagnostic decisions- and the treatment decisions that follow. A clinical prediction rule is defined as “a clinical too that quantifies the individuals contributions that various components of the medical history, physical examination, and basic laboratory results make toward the diagnosis, prognosis, or likely response to treatment in an individual patient.” (1) CPRs are useful in situations where the decision making is complicated and complex.

CPRs a redeveloped using a 3-step process. The first step involves the actual creation or derivation of the rule. Step 2 involves the testing or validation of that new rule, and the final step assesses the impact the rule has had on actual clinical behavior; that is, does it end up making a difference in practice? In order to derive a new rule, it is necessary to locate and identify the factors that might have predictive power. This information could be drawn from the history, physical examination or from lab or other testing. We could then look at a series of patients to see if any of our proposed predictors are present in a large percentage of those patients. Consider, for example, a positive straight leg raise test in patients with prolapsed lumbar discs. To validate a CPR, we need to demonstrate that if we repeatedly apply it with our patients, it leads to the same results (either diagnostically or prognostically). So, in effect, we are now testing the rule in a larger and new population of patients. In the testing process and the research that is done, one can generate either likelihood ratios or odds or risk ratios. These are concepts we are now well aware of.

Here are a couple of relevant CPR papers:

• Schenk R, Dionne C, Simon C, Johnson R. Effectiveness of mechanical diagnosis and therapy in patients with back pain who meet a clinical prediction rule for spinal manipulation. J Man Manip Ther 2012;20:43-49
• Stolze LR, Allison SC, Childs JD. Derivation of a preliminary clinical prediction rule for identifying a subgroup of patients with low back pain likely to benefit from Pilates-based exercise. J Orthop Sports Phys Ther 2012;42:425-436

References

1.       McGinn T, Wyer P, Wisnivensky J, et al. Clinical prediction rules. In: Guyatt G, Rennie D, Meade MO, Cook D. Users’ guide to the medical literature, 3^rd edition. New York, NY; MgGraw Hill Medical, 2008:491