Personalized Allocation of Patients to Physiotherapists with Machine Learning Techniques
(2022) EEML05 20221Department of Biomedical Engineering
- Abstract
- Osteoarthritis (OA) is a heterogeneous inflammatory joint disease that affects around 240-250 million people worldwide. The disease is a prominent cause of disability and a leading source of societal expenses in older adults. As of today, personalized OA treatment is considered essential and is currently being addressed in several osteoarthritis guidelines. To explore the possibilities of precision medicine, machine learning (ML) has been implemented to allocate patients to therapists in the psychotherapy field. It is therefore argued that allocation using ML can be of interest in OA physiotherapy as well. The aim of this study is to implement a supervised classification ML model to predict the best mapping between onboarding patient and... (More)
- Osteoarthritis (OA) is a heterogeneous inflammatory joint disease that affects around 240-250 million people worldwide. The disease is a prominent cause of disability and a leading source of societal expenses in older adults. As of today, personalized OA treatment is considered essential and is currently being addressed in several osteoarthritis guidelines. To explore the possibilities of precision medicine, machine learning (ML) has been implemented to allocate patients to therapists in the psychotherapy field. It is therefore argued that allocation using ML can be of interest in OA physiotherapy as well. The aim of this study is to implement a supervised classification ML model to predict the best mapping between onboarding patient and physiotherapist. The OA patient and physiotherapist data collected between the years 2019-2022 was provided by the Swedish telehealth company Joint Academy (JA). The data, used for this study, consisted of 8 separate subsets. All subsets of data were processed according to general methods. Python was the programming language of choice, where algorithms, e.g. Random Forest Classifier (RF) and gradient booster XGBoost were explored and implemented with the software library sci-kit learn. Four different models were benchmarked against a baseline model. The baseline model received an F1-score of 71.31% on the test set, which corresponds to 20% of the final data set. The final XGBoosted model received an F1-score of 68.24% on the test set. The final model is not appropriate to implement in a care system in its current state. It can be further improved with better feature engineering, improved imputation techniques, and explore different target variables. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9087971
- author
- Engdal Höie, Line Marie LU and Henningsson, Cecilia LU
- supervisor
- organization
- alternative title
- Personlig allokering mellan patient och fysioterapeut med hjälp av maskininlärning
- course
- EEML05 20221
- year
- 2022
- type
- M2 - Bachelor Degree
- subject
- language
- English
- id
- 9087971
- date added to LUP
- 2022-06-14 11:02:41
- date last changed
- 2022-06-14 11:02:41
@misc{9087971, abstract = {{Osteoarthritis (OA) is a heterogeneous inflammatory joint disease that affects around 240-250 million people worldwide. The disease is a prominent cause of disability and a leading source of societal expenses in older adults. As of today, personalized OA treatment is considered essential and is currently being addressed in several osteoarthritis guidelines. To explore the possibilities of precision medicine, machine learning (ML) has been implemented to allocate patients to therapists in the psychotherapy field. It is therefore argued that allocation using ML can be of interest in OA physiotherapy as well. The aim of this study is to implement a supervised classification ML model to predict the best mapping between onboarding patient and physiotherapist. The OA patient and physiotherapist data collected between the years 2019-2022 was provided by the Swedish telehealth company Joint Academy (JA). The data, used for this study, consisted of 8 separate subsets. All subsets of data were processed according to general methods. Python was the programming language of choice, where algorithms, e.g. Random Forest Classifier (RF) and gradient booster XGBoost were explored and implemented with the software library sci-kit learn. Four different models were benchmarked against a baseline model. The baseline model received an F1-score of 71.31% on the test set, which corresponds to 20% of the final data set. The final XGBoosted model received an F1-score of 68.24% on the test set. The final model is not appropriate to implement in a care system in its current state. It can be further improved with better feature engineering, improved imputation techniques, and explore different target variables.}}, author = {{Engdal Höie, Line Marie and Henningsson, Cecilia}}, language = {{eng}}, note = {{Student Paper}}, title = {{Personalized Allocation of Patients to Physiotherapists with Machine Learning Techniques}}, year = {{2022}}, }