Musculoskeletal (MSK) disorders are among the most widespread health issues globally and the leading cause of disability. These conditions place a significant burden on both patients and society. Traditional treatments for MSK disorders often lack solid evidence or provide only moderate and short-term effects. Due to the diverse nature of the patient group, it is challenging to apply guidelines and systematic reviews to clinical practice, as they are often based on average group differences from randomized studies. To optimize treatment, it should be tailored individually, focusing on effective communication and shared decision-making. However, this is challenging due to the many factors influencing the patient’s experience and symptoms.
One potential solution is to use artificial intelligence (AI) to personalize treatment at the individual level or through stratified treatment tailored to subgroups of patients with similar characteristics.
Objective
The main goal of the project was to improve the treatment of patients with MSK disorders in primary care, from stratified treatment in general practice to personalized treatment in physiotherapy. To achieve this, the project developed a digital clinical decision support system for physiotherapists based on AI and further developed a decision support system for general practice based on stratified treatment for general practitioners. The project also tested the effectiveness of these systems in clinical practice and developed a prototype for an AI-based decision support system for general practitioners.
Results
The project developed and evaluated digital clinical decision support systems for physiotherapists and general practitioners. The systems were evaluated in two randomized controlled trials with 729 patients among physiotherapists and 298 patients in general practice. The results showed that the decision support system for physiotherapists did not provide better treatment outcomes compared to the control group receiving usual care. Process evaluation and qualitative studies revealed that physiotherapists appreciated parts of the system, but it did not influence treatment choices as expected. The study among general practitioners also did not show improved treatment outcomes with the stratified approach. However, the process evaluation provided valuable insights into factors that promote and hinder the use of the decision support system among general practitioners.
The project also aimed to develop a prototype for an AI-based decision support system for general practitioners. After the study among physiotherapists showed that the AI system did not improve treatment outcomes, we decided to focus on a patient-centered approach. In the final work package, we conducted a user-centered design process with co-design workshops involving the project group, experts, frontend developers and end-users. This resulted in a redesigned prototype of shared digital screens, based on principles from Human-Computer Interaction (HCI) and Computer-Supported Collaborative Work (CSCW).
The prototype was tested in a field experiment during consultations between four general practitioners and twelve patients. We explored how an interactive graphical user interface could facilitate patient-centered care. The main results from the field experiment among general practitioners were summarized in six points: 1) The influence of system design on interaction, 2) Structure of the consultation, 3) Patient involvement, 4) Validation of the patient’s symptoms, 5) Educational tool and “third party” in the consultation, 6) Shared understanding. The project demonstrates how a user-centered approach can improve digital tools for healthcare professionals, aiming to promote patient-centered care.
The project also included four master’s theses that used data from the project to address clinically relevant issues: 1) The relationship between health literacy and treatment outcomes, 2) The relationship between sleep and psychological stress, 3) Expectations as a prognostic factor, 4) Childhood experiences and their association with treatment outcomes.
This project shows that while AI and digital decision support systems have potential, there are still challenges related to development, implementation, and effective use in clinical practice. Further research and development are needed to realize the potential of personalized treatment for MSK disorders.
Prosjektet utviklet og bygget digitale, kliniske beslutningsstøttesystem for fysioterapeuter og fastleger. Systemene ble testet ut i en randomisert kontrollert studie blant fysioterapeuter med 729 pasienter og i en randomisert kontrollert studie i allmennmedisin med 299 pasienter. Studien blant fysioterapeutene viste at beslutningsstøttesystemet, som brukte metoder fra kunstig intelligens, ikke ga en bedre behandlingseffekt, sammenlignet med ei kontrollgruppe som fikk vanlig oppfølging uten bruk av systemet. Prosessevalueringen og kvalitative studier viste at fysioterapeutene likte veldig godt deler av beslutningsstøttesystemet, men systemet påvirket ikke valg av behandling slik som intensjonen med systemet var. Studien blant fastlegen fant heller ikke en forbedret behandlingseffekt basert på den stratifiserte tilnærmingen. Her ga også prosessevalueringen nyttig kunnskap om fremmere og hemmere for bruk av beslutningsstøttesystemet blant fastleger. Prosjektet hadde også som mål å utvikle en prototype for et KI-basert beslutningsstøttesystem for fastleger. Siden studien blant fysioterapeutene viste at det KI-baserte beslutningsstøttesystemet ikke ga bedre behandlingseffekt, valgte vi å ikke basere prototypen på den samme metoden fra kunstig intelligens. Derfor valgte vi å videreutvikle og forbedre beslutningsstøttesystemet for fastlegene med fokus på pasientsentrert tilnærming. Et viktig spørsmål var å undersøke hvordan delte digitale skjermer kan fremme pasientsentrert behandling i allmennmedisin. Hovedresultatene fra denne delen av prosjektet ble oppsummert omkring 6 punkter:1) Innflytelsen av design av systemet på interaksjon mellom fastlege og pasient, 2) Struktur på konsultasjonen, 3) Involvering av pasient, 4) Validering av pasientens symptomer og plager, 5) Pedagogisk verktøy og «tredjepart» i konsultasjonen, 6) Felles forståelse.
Prosjektet har avdekket flere hemmere og fremmere for å utføre klinisk forskning i primærhelsetjenesten i Norge. Prosjektet har også bidratt til økt fokus på bruk av kunstig intelligens i helsetjenesten, ved at prosjektgruppa har deltatt på mange ulike forum nasjonalt og internasjonalt, både for forskere, pasienter, beslutningstakere, og politikere. Prosjektet har bidratt til forsterket tverrfaglig samarbeid mellom medisinske og teknologiske miljø på NTNU. Satsningen på kunstig intelligens innen helse ved NTNU har bidratt til at forskningsfeltet fikk en nyopprettet stilling fra Rektor ved NTNU innen kunstig intelligens og Helse, som betyr mye for den videre satsningen på forskningen innen feltet.
Musculoskeletal disorders (MSDs) is the number one cause of years lived with disability andreduced health worldwide. In Norway, every fourth patient in primary care suffers from MSDs. Treatment effects are however modest and knowledge
of best practice limited. The SUPPORTPRIM project will address these challenges in two main steps:1) To optimize person-centered care, we will employ innovative methods from artificial intelligence in terms of Case-Based Reasoning to build a clinical decision support system (cDSS ) based on patient data already collected in primary care physiotherapy. Case-Based Reasoning aims to solve new problems based on the solutions to similar problems in the past. In other words, previous MSD cases will be used to help similar cases in the future, just as humans learn from their own experience. We will then assess the efficacy of the cDSS in physiotherapy practice in a randomised controlled trial.2) This effort will be expanded to general practice by implementing The STarT MSK screening Tool as basis for stratified care for MSD patients. The efficacy of the stratified care will be assessed in a randomized controlled trial in general practice. Finally, the cDSS from physiotherapy practice will be extended and adapted to fit general practice. State-of-the-art personalized treatment plans are envisioned to benefit a much larger proportion of MSD patients than a "one-size-fits-all" approach. In addition, SupportPRIM facilitates and emphasizes the co-decision process between the patient and physiotherapist. The innovative approach using artificial intelligence to develop personalised decision support systems are far beyond current state of art, is highly relevant for other patient groups in health care, and can be integrated in future medical record systems.