Masterclass
Combining orthopedic special tests to improve diagnosis of shoulder pathology

https://doi.org/10.1016/j.ptsp.2014.08.001Get rights and content

Abstract

The use of orthopedic special tests (OSTs) to diagnose shoulder pathology via the clinical examination is standard in clinical practice. There is a great deal of research on special tests but much of the research is of a lower quality implying that the metrics from that research, sensitivity, specificity, and likelihood ratios, is likely to vary greatly in the hands of different clinicians and in varying practice environments. A way to improve the clinical diagnostic process is to cluster OSTs and to use these clusters to either rule in or out different pathologies. The aim of the article is to review the best OST clusters, examine the methodology by which they were derived, and illustrate, with a case study, the use of these OST clusters to arrive at a pathology-based diagnosis.

Introduction

Physical examination of the shoulder involves a series of steps typically beginning with history, progressing with motion and muscle testing, and culminating in the use of orthopedic special tests (OSTs) with the aim of diagnosing shoulder pathology. While the process itself is systematic and straightforward, for evidence-based practitioners, there are numerous problems encountered when trying to arrive at a diagnosis. First, there is little evidence reporting the diagnostic accuracy of critical pieces of the clinical examination such as history, motion testing, and muscle testing causing a greater reliance on OSTs. Second, although there is a great deal of research on OSTs of the shoulder, much of that research is of moderate to low quality (Hegedus et al., 2008, Hegedus et al., 2012). Third, even in those OSTs that come from high quality literature, there are very few that display solid diagnostic metrics, high sensitivity and specificity (Hegedus et al., 2008, Hegedus et al., 2012). Fourth, although sensitivity and specificity are helpful internal test metrics, there are issues in the application of these metrics to clinical practice. Finally, clinicians and researchers improve diagnostic accuracy by clustering OSTs together; however, in some cases, the clusters are used incorrectly or provide metrics that lead to post-test probabilities that are no different than use of a single stand alone test.

Our aims in this paper are to discuss the importance of likelihood ratios and modified probability in the diagnostic process, to explain multivariate modeling and outline the most effective methods to combine tests for either screening or confirmation of diagnosis. For context, we'll briefly review the best test clusters that have been published, and finally, we'll use a case study to illustrate how the best available test clusters should be used to aid in diagnosis.

Section snippets

Likelihood ratios and modified probability

Diagnostic accuracy studies have design consistencies, standardized metrics, and assumptions. First and foremost, all diagnostic accuracy studies enroll populations of individuals with and without the condition of interest; the condition of interest being the diagnosis studied. Those without the condition of interest should be individuals with some other competing health malady that would normally be distinguished in a traditional clinical environment. For example, a typical shoulder study

Multivariate modeling

The goal in any data analysis is to extract from raw information the accurate estimation (Alexopoulos, 2010). The goal when clustering tests is to determine the best combination estimates that produce the strongest likelihood ratios and to do so, multivariate modeling is required. Thus, clustering is simply the act of evaluating a set of tests and measures, in combination, when making a clinical decision or a mathematical assessment. For example, when attempting to detect acromiocalvicular (AC)

Best test clusters

Before presenting the best published test clusters, “best” needs to be put in context. “Best” as used in this manuscript, is defined as those combinations of tests with the strongest likelihood ratios from research with the highest quality. The quality of the tests clusters is judged by using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) document and using a 0–14 (0 = lowest quality) scale (Whiting, Rutjes, Dinnes, Reitsma, Bossuyt, & Kleijnen, 2004). From our past experience(

Case study

Our fictitious patient is 67 years old and has complained of shoulder pain of 4 months in duration. He reported that his pain initiated while walking his dogs (when they jerked the leash he held) but notes that the pain has progressed markedly over the last two months. He is able to raise his arm above his head (with pain) but has noted that his arm now aches consistently, with a more noticeable ache at night. Frequent use of ibuprofen helps modulate his pain but the effects are only temporary

Conclusion

The clinical diagnostic process should be viewed through the lens of odds and probabilities. In order to do so, test clusters from high quality studies should be utilized. In our case study, the patient likely has a rotator cuff tear but we were unable to rule out or in a labral tear and impingement. High quality clinical test clusters with powerful diagnostic characteristics for labral tears do not presently exist and impingement is an all-encompassing term for tendon pathology at the shoulder

Conflict of interest

None.

Funding

None.

Ethical approval

None.

References (36)

  • E.C. Alexopoulos

    Introduction to multivariate regression analysis

    Hippokratia

    (2010)
  • J.M. Beneciuk et al.

    Clinical prediction rules for physical therapy interventions: a systematic review

    Physical Therapy

    (2009)
  • E. Chronopoulos et al.

    Diagnostic value of physical tests for isolated chronic acromioclavicular lesions

    The American Journal of Sports Medicine

    (2004)
  • C. Cook et al.

    Clustered clinical findings for diagnosis of cervical spine myelopathy

    The Journal of Manual & Manipulative Therapy

    (2010)
  • C. Cook et al.

    Orthopedic physical examination tests: An evidence-based approach

    (2013)
  • C. Cook et al.

    Lumbopelvic manipulation for treatment of patients with patellofemoral pain syndrome: development of a clinical predication rule

    The Journal of Orthopaedic and Sports Physical Therapy

    (2008)
  • A.J. Farber et al.

    Clinical assessment of three common tests for traumatic anterior shoulder instability

    The Journal of Bone and Joint Surgery

    (2006)
  • D. Firth

    Bias reduction of maximum likelihood estimates

    Biometrika

    (1993)
  • Cited by (0)

    View full text