I am a
Home I AM A Search Login

Editor’s picks: Clinical prediction rules: Use the babies and throw the bathwater?



The 2024 Global Year will examine what is known about sex and gender differences in pain perception and modulation and address sex-and gender-related disparities in both the research and treatment of pain.

Learn More >

Over this holiday season we are publishing our Editor’s picks of 2017 for you to read and enjoy again. 

There are easily a thousand clinical prediction rules (CPRs) related to managing musculoskeletal pain. Okay, maybe a thousand is an exaggeration. My point is there are many. All designed with the aim of helping clinicians to make more certain decisions about diagnosis (diagnostic CPRs), prognosis (prognostic CPRs) or likely response to an intervention (prescriptive CPRs). As a clinician faced with a generous serving of uncertainty on a daily basis, this seems like a reasonable cause to me.

It’s undeniable that many of the available CPRs are at an early stage of development or have been developed without using adequate methods[1,2]. This means that many tools are not (at least yet) ready for clinical use. It’s also apparent that many display inconsistent, or at least population-, or outcome-specific accuracy. Understandably, this presents problems when using them. BUT some CPRs are well developed, display adequate accuracy and have been demonstrated to improve patient care. Check out The Ottawa Ankle Rules[3], The Canadian C-Spine Rule[3], or The Keele STarT Back Screening tool[4] as examples. So why don’t most clinicians use them[5]? (Do a quick survey of your colleagues, and I suspect you’ll find some more anecdotal evidence to support this). We asked a bunch (n=28) of Aussie physios, chiros and osteopaths what they thought about CPRs to try shed some light on the situation[6].

We found that misunderstanding the purpose of CPRs was a major barrier to use. These tools were commonly confused with outcome measures, which could foreseeably result in clinicians using CPRs repeatedly (and interpreting a change in CPR score as a marker of clinical change), or perhaps more importantly, failing to recognise the real potential utility of the original score – its predictive value. Which brings me to the next finding. Clinicians indicated that they’d consider using a CPR only if it added value to patient management (above and beyond the inconvenience of changing practice and any perceived negative consequences of use). Sounds very reasonable. But herein lies the problem: If you don’t understand the purpose of a CPR, you’re unlikely to recognise any potential value!

Another finding was that clinicians preferred diagnostic and prognostic CPRs over those that are prescriptive. Prescriptive CPRs have received much bad press due to present-day inadequacies in development [check out the blog post R.I.P. Prescriptive Clinical Prediction Rules]. However, participants in our study didn’t focus on these shortcomings. Instead, their distaste for prescriptive CPRs appeared driven by a desire to maintain autonomy in decision-making. That is, participants wanted a CPR to guide their decision-making, not dictate a course of action. You might call this the “I’m not a robot, don’t tell me what to do” effect, which I think raises a nice question: How then do these tools ideally fit within our practice? Despite the HIGHLY unfortunate nomenclature, clinical prediction RULES probably shouldn’t be considered rules at all (at least not in the dictatorial sense – perhaps in the statistical sense). They are designed to inform decision-making, and should not be a replacement for clinical reasoning.

Finally, clinicians expected a degree of congruence between a CPR, its included components, their personal experience and current clinical reasoning frameworks. This informed their judgment of a CPR’s accuracy. Undeniably, CPRs are not perfect (using this metric or the recommended interpretation of accuracy statistics)… Neither are orthopaedic tests or other clinical signs, but the usefulness of such tests is rarely discounted.

Perhaps it is time to focus finite research resources towards further developing CPRs that best align with clinicians’ preferences (prognostic or diagnostic tools); provide clinicians with adequate education on CPR concept, purpose and ideal fit with practice (so that they are better equipped to judge the merits of using these tools); AND as clinicians, consider expanding our use of the better-developed CPRs to incorporate their results (within their limitations) with our other clinical information?

Disclaimer: This blog focuses on the misunderstandings expressed by some participants in our study for the purpose of highlighting possible ways forward. It does not represent the opinions or beliefs of the entire sample, in which some people displayed a thorough understanding and very balanced view (in my opinion) of the pros and cons of using CPRs.  

About Joan Kelly

Joan is a physio and PhD candidate based at Recover Injury Research Centre, Griffith University, and Allsports Physiotherapy, in the Gold Coast, Australia. She’s interested in using qualitative methods to research how healthcare practitioners make clinical decisions as well as investigate what drives them to change practice. When she isn’t pondering these questions, Joan can be found sipping coffee, riding pushbikes or watching her vegie garden grow.


[1] Kelly J, Ritchie C, Sterling M. Clinical prediction rules for prognosis and treatment prescription in neck pain: A systematic review. Musculoskeletal Science and Practice. 2017;27:155-164.

[2] van Oort L, van den Berg T, Koes BW, et al. Preliminary state of development of prediction models for primary care physical therapy: a systematic review. Journal of Clinical Epidemiology. 2012;65(12):1257-1266.

[3] Stiell IG, Bennett C. Implementation of Clinical Decision Rules in the Emergency Department. Academic Emergency Medicine. 2007;14(11):955-959.

[4] Hill JC, Whitehurst DG, Lewis M, et al. Comparison of stratified primary care management for low back pain with current best practice (STarT Back): a randomised controlled trial. The Lancet. 2011;378(9802):1560-1571.

[5] Knox GM, Snodgrass SJ, Rivett DA. Physiotherapy clinical educators’ perceptions and experiences of clinical prediction rules. Physiotherapy. 2015;101(4):364-372.

[6] Kelly J, Sterling M, Rebbeck T, et al. What are health practitioners’ perceptions of adopting clinical prediction rules in the management of musculoskeletal pain? A qualitative study. BMJ Open (under review).

Share this