edoc

Clinical prediction models in orthopaedics: from data collection to model development and validation

Stojanov, Thomas. Clinical prediction models in orthopaedics: from data collection to model development and validation. 2024, Doctoral Thesis, University of Basel, Faculty of Medicine.

[img]
Preview
PDF
Available under License CC BY (Attribution).

2805Kb

Official URL: https://edoc.unibas.ch/96732/

Downloads: Statistics Overview

Abstract

Rotator cuff disease affects up to 20% of the general population and is a major cause of shoulder pain and disability. Non-operative treatment is usually firstly counseled, but surgery, with the performance of arthroscopically assisted rotator cuff repair (ARCR) is often required. The lack of studies with proper methodology impairs health policy makers in drawing conclusions regarding the safety and the effectiveness of this procedure. In that context, the ARCR_Pred study, a nationwide multicenter cohort funded by the Swiss National Science Foundation and Swiss Orthopaedics, aimed to standardize outcomes, and develop clinical prediction models. This PhD thesis focuses on the implementation of the study, the description of overall and patient characteristics, and the development of prediction models for post-operative stiffness (POSS) and functional outcomes, notably through the patient-reported Oxford Shoulder Score (OSS). In the first thesis manuscript, the successful implementation of the ARCR_Pred study was highlighted: 973 patients were included from 18 Swiss and 1 German orthopedic centers between June 2020 and November 2021, achieving high follow-up rates. After consideration of case-mix variables, the analysis showed no clinically relevant outcome differences between public and private clinics at last follow-up. The second manuscript is a systematic review summarizing the evidence related to prognostic factors for POSS. However, the low quality of the summarized evidence highlighted the need to complement our findings with an expert opinion. A Delphi process was therefore initiated and based on these results, three models were built and compared in the third manuscript. The ARCR_Pred-POSS model had the best discrimination and calibration performance (AUC=0.719 and calibration slope=1.022). Decision-curve analysis showed superior performance of the ARCR_Pred-POSS model in comparison to the surgeon prediction of POSS over the whole range of threshold probabilities. At the risk probability threshold 15%, the model showed a sensitivity of 57% and a specificity of 73%. A user-friendly responsive web-based application was developed. Regarding the prediction of the OSS, the second systematic review identified 22 potential prognostic factors for the improvement of shoulder function after an ARCR, which correspond to the fourth and last manuscript of the thesis. These findings were also completed by an expert opinion involving 52 healthcare professionals (including surgeons and experienced physiotherapists), which led to the identification of 101 factors. The ARCR_Pred-OSS model remains to be developed and validated on the ARCR_Pred data, which would add a patient-relevant outcome to the ARCR_Pred prediction tool. The evaluation of the impact of the use of such prediction tools on surgical and patient-reported outcomes remains also to be assessed. Model optimization should first be sought, by assessing the predictive ability of new factors and modern modeling techniques. Prediction tools in surgery should also incorporate patient and healthcare professionals’ preferences.
Advisors:Müller, Marc Andreas
Committee Members:Christ-Crain, Mirjam and Audigé, Laurent and Aghlmandi, Soheila and Zwahlen, Marcel
Faculties and Departments:03 Faculty of Medicine > Departement Biomedical Engineering
UniBasel Contributors:Müller, Marc Andreas and Christ-Crain, Mirjam and Audigé, Laurent
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:15611
Thesis status:Complete
Number of Pages:201
Language:English
Identification Number:
  • urn: urn:nbn:ch:bel-bau-diss156119
edoc DOI:
Last Modified:04 Feb 2025 10:09
Deposited On:28 Jan 2025 13:18

Repository Staff Only: item control page