Graduation Year

2022

Document Type

Thesis

Degree

M.S.P.H.

Degree Name

MS in Public Health (M.S.P.H.)

Degree Granting Department

Public Health

Major Professor

Thomas Bernard, Ph.D.

Committee Member

Rachel Williams, Ph.D., M.S.P.H

Committee Member

Jared Jeffries, Ph.D., M.S.P.H

Keywords

Bayesian network, diagnostic support tool, elbow injuries, Musculoskeletal disorders, musculoskeletal ultrasound, POCUS

Abstract

The upper extremity is commonly injured at the workplace, frequently involving the elbow. Currently, there are not many diagnostic support tools for elbow injuries. Developing a clinical decision support tool would allow for narrowing differential diagnoses and guide management steps in a timely and cost-effective manner. In a descriptive retrospective cohort study, 85 non-contrast elbow MRIs were obtained from a large Workers Compensation insurer database. MRIs were either (a) greater than 2 weeks after first clinic visit, or (b) more than 6 weeks after injury, but (c) not more than 3 months after injury. A Bayesian network-based diagnostic support tool was developed from the elbow MRI results after removing variables that were not sufficiently representative. The common extensor tendon (CET), the most injured structure in this data set in 36 of 85 cases, served as the parent node. The second most injured structure was the distal biceps tendon with injuries present in 19 of 85 cases. By evaluating the most commonly injured structures, most of other injuries were able to be ruled out with limited point-of-care ultrasound examination of the elbow and the prediction model was used to guide clinicians into one of the following management steps: no follow up, conservative management, or surgical referral with advanced imaging (MRI). Finally, a targeted ultrasound algorithm was developed to reduce the point-of-care ultrasound (POCUS) learning curve for less experienced examiners.

Download

Share

COinS