Project undertaken In collaboration with Drs. David Crawford and Keith Berend, Joint Implant Surgeons, Ohio.

​

 

Background

Arthroplasty is the costliest component for Knee Osteoarthritis (OA) care with major added costs when surgery is delayed or undertaken inappropriately.  Decisions made for referrals for surgery as well as those made by surgeons on what to select are largely subjective which contributes to considerable variance in use of Total and Unicompartment Knee Arthroplasty (KA).  The variation potentially contributes to poor outcomes of up to 20 percent. The study objective was to evaluate the prediction for a surgeon’s indication for and selection of KA based on an eTriage decision tool developed by OAISYS Inc. The eTriage quantified disease severity from validated gradings of X-rays and Disability Scores with key demographics which were then used to predict surgical or non-surgical care.

​

Methods:

 A query of Joint Implant Surgeons records was performed to identity 100 patient each that underwent total knee arthroplasty (TKA), medial unicondylar knee arthroplasty (UKA) or non-arthroplasty. Nine patients in the “non-arthroplasty” group had gone on to TKA and 7 more did not have radiographs available. One patient in the “UKA” group had missing radiographs. When these cases were removed a total of 292 patients remained for the study (109 TKA, 99 UKA, 84 non-arthroplasty).  The radiographs collected included a weightbearing frontal, posteroanterior flex, lateral and patellar axial views. KOOS Jr scores were also collected along with patient demographics. These de-identified data were sent to OAISYS for their analysis and to make the care recommendation for TKA, UKA or no surgery. OAISYS was blinded to the actual care performed. OAISYS undertook a compartmental radiographic grading and these, along with the Disability Scores and demographics, were used to predict the care. The eTriage predictions were then cross referenced against the actual care performed. Analysis included accuracy of prediction, sensitivity, and positive predictive value (PPV).

​

Results:

The accuracy of OAISYS predicting no surgery versus any surgery was 89.04%. The sensitivity for a non-surgical recommendation was 92.9%. The sensitivity for a surgical recommendation was 88.9% with a PPV of 96.9%. Forty-nine patients had an eTriage prediction for UKA, when in fact they had a TKA.  Three of the patients that underwent a TKA had a predication for UKA, and 6 patients predicted for surgery were treated nonoperatively.  The eTriage predictions were was in agreement in 71%.

​

Conclusion:

The OAISYS eTriage decision aid demonstrated considerable accuracy for predicting whether patients were likely to undergo a knee arthroplasty procedure. With high sensitivity of a non-surgical treatment prediction and high PPV for surgery, this model would result in very few patients being referred to a surgeon who were not candidates for surgery, with errors being on the non-surgery side of care. With increasing demands for surgery, continuing variance in its use and poor outcomes, the use of a reliable eTriage tool, such as developed by OAISYS,  might be of considerable value to all stakeholders in the predicative care of knee OA and significantly lessen escalating costs.