Paper Submission & Registration
8th Dutch Bio-Medical Engineering Conference
Go-previous
13:40   Pain
Chair: Esther Tanck
13:40
15 mins
Towards optimized personalized referral advice and patient satisfaction in chronic musculoskeletal pain
Wendy Oude Nijeweme - d'Hollosy, Annemieke Konijnendijk, Romee Bessembinder, Floor Leeferink, Jose Broeks
Abstract: Of all Dutch adults, 1 in 5 has musculoskeletal pain[1]. More than 50% of these people have limited functioning. Experienced disease burden can be reduced by starting an appropriate treatment as early as possible. However, it appears that more than 30% of the patients are referred back to the general practitioner (GP) with a rereferral to somewhere else. This results in an inefficient care path. Therefore, the PReferral project was started in September 2020 and aims to optimize existing care pathways and referrals in healthcare based on artificial intelligence leading to a proof-of-concept decision support system for the GP. In preparation of the PReferral project, an exploration study of individual healthcare pathways, referrals and patient satisfaction was performed in the Netherlands in May 2020. During this study, 84 Dutch people with chronic musculoskeletal pain filled in an online survey. Both qualitative and quantitative methods were used for data analysis. Respondents had an average of 3 different referrals for treatment (SD = 3.2, range 0-17). The GP referred to 34 different regular and alternative care providers in primary, secondary and tertiary care. Referrals were most often made to physiotherapists (n = 55), orthopaedists (n = 37) and neurologists (n = 34). According to the respondents, the choice for referral was usually based on specific complaints (63%), patient's preference (8%) or professional guidelines (13%). Respondents were satisfied with their GP's referral, but rated satisfaction with their entire healthcare pathway low on a 0-10 scale (average = 5.5, SD = 2.0). Communication among care providers was also perceived as very poor on a 0-10 scale (average = 5.2, SD = 2.5). The performed exploration study also showed that healthcare pathways for people with chronic musculoskeletal pain is very diverse. Although this study does not make any statements about the correctness of the referral, the large number of referrals from the GP indicates a suboptimal process. In addition, the low patient satisfaction is a clear signal that current care pathways and referrals should be optimized. [1] Position paper ‘Revalidatie bij chronische pijn aan het houdings- en bewegingsapparaat,’ Conv. Heal. Deal “Chronische Pijn,” 2016.
13:55
15 mins
Visual support of electrical pain threshold measurements: Construction and preliminary results of the QScale
Niels Jansen, Leonie Hartl, Marjolein Thijssen, Jan R. Buitenweg
Abstract: Introduction. Chronic pain is often linked with a higher pain sensitivity, which manifests itself as a lower pain threshold (PT) [1]. For the assessment of the PT, the subject has to indicate if the sensation elicited by a stimulus is perceived as painful [2]. In our experiments using electrocutaneous stimulation, subjects often express difficulties in determining the transition from non-painful to painful sensations, which might indicate that their criterion formation is poorly experimentally controlled. We wondered if the addition of visual representations of non-painful and painful sensations to the verbal task instruction could bring the criterion formation under improved experimental control. A first prerequisite for this is that (1) an ordered set of such visual representations can be constructed and (2) that the order of these visual representations is associated with the stimulus strength. Methods. In two experiments we evaluated these prerequisites. In the first experiment the QScale was constructed by asking 24 subjects to make drawings of their perceived sensations evoked by stimuli of various strengths. Thereafter, 3 naïve subjects grouped the drawings and via a graphics program, the QScale was developed consisting of in total 4 qualities (A-D), each with 4 intensities (1-4). In the second experiment, 9 participants were subjected to 70 randomized bursts of stimuli around their own PT and asked to indicate if a certain quality with corresponding level of the QScale was perceived. Results. All qualities were chosen more frequently and at higher intensity levels as the stimulus strength increased. The first two qualities (A and B) were found to be frequently (>50%) chosen already from the lowest stimulus intensities below the PT, while the last two qualities (C and D) were only frequently chosen at or above stimulus intensities corresponding to the PT. Discussion. These preliminary results indicate that qualities A and B seem to represent a generic sensation of electrical stimulation, while qualities C and D are more strongly associated with painful sensations, suggesting that the QScale might be used as a visual support for electrocutaneous PT measurement.
14:10
15 mins
Classification of patients with chronic low back pain and high or low central sensitization by gait outcomes using machine learning methods
Xiaoping Zheng, Michiel Reneman, Jone Ansuategui Echeita, Herbert Kruitbosch, Egbert Otten, Claudine Lamoth
Abstract: A major issue for interventions for patients with chronic low back pain (CLBP) is the heterogeneity of the patient population. Many studies have shown in a controlled laboratory setting that gait of patients with CLBP is different in terms of less variability in trunk rotations, lower velocity in preferred walking speed, shorter stride length. One of the explanations for the inconsistent finding could be the presence of central sensitization since movements may be changed during pain. Also, results might be different when assessed in an uncontrolled environment compared to walking during daily life activities. Therefore, this study aimed to examine if patients with CLBP with high or low probability of CS (a CS Inventory score of 40-100 (CLBP+), or lower than 40 (CLBP-)) could be classified based on gait performance outcomes, obtained from gait cycles during daily life activities. Forty-three patients with CLBP were included (24 CLBP- and 19 CLBP+). Patients wore a 3D accelerometer for about one week. From each patient, 4 days of accelerometer-data was selected randomly. For each day data, continuous gait cycles (628 for CLBP-, 571 for CLBP+) were extracted using a zero-cross method. For all gait cycles in one day, 36 gait outcomes were calculated, representing variables related to pace, regularity, synchronization, smoothness, local stability, and predictability of gait. A Random Forest classifier was trained to classify CLBP- and CLBP+ groups based on gait outcomes and SHapley Additive exPlanations (SHAP) method was used to explain the differences between groups in gait outcomes. The low and high CS groups were classified with a Random Forest method with the F1-score of 0.82, an accuracy of 81%. Eight gait outcomes were identified by SHAP as the most important in classifying 2 groups. They were index of harmonicity-V and harmonic ratio-ML (smoothness), and sample entropy-AP (predictability), and maximal-Lyapunov exponent-V/ML (stability), and stride frequency variability-ML/AP (pace), and walking regularity-ML (regularity). The accurate classification results indicate that patients with CLBP and with high or low CS walked differently. The SHAP method shows that patients with high CS level exhibited lower smoothness, lower local stability, and less predictable in gait.
14:25
15 mins
Inappropriate healthcare referrals of patients with chronic musculoskeletal pain? We can do better: a multimodal artificial intelligence approach
Wendy Oude Nijeweme - d'Hollosy, Duncan Jansen, Mannes Poel
Abstract: At least 25% of all healthcare referrals are considered inappropriate, leading to extra burdens on patients by, for example, re-referrals, excessive waiting times, uncoordinated care, and duplicated testing [1]. Physicians can make better informed choices to refer patients by also including relevant information from multimodal data sources, such as electronic health records, clinical images, and sensor data. However, these available data are too voluminous to be interpreted by single physicians themselves. Therefore, Clinical Decision Support Systems (CDSSs) that support physicians during referral decisions are indispensable. CDSSs can be knowledge-driven, i.e. based on guidelines and expert knowledge, or data-driven, i.e. based on Machine Learning (ML) models trained by data. In our research line, we combine both approaches in a multimodal Artificial Intelligence (AI) methodology for building CDSSs on healthcare referral. Several obstacles need to be overcome to develop integrated CDSSs that capitalize on the strengths of both multimodal data-driven and knowledge-driven approaches. The first challenge is to combine multimodal datasets for ML on the same task, because different types of inputs complicate ML on one task. The second challenge is to merge data-driven and knowledge-driven approaches in developing CDSSs with continuous learning loops, e.g. learning from the quality of previous patient referral cases. The methodology will be validated and demonstrated in different projects focused on the referral of chronic musculoskeletal pain. PReferral (Personalised Referral) is the first pioneering project in this research line and has been started in September 2020. When validity of combining both knowledge-driven and data-driven approaches in a multimodal AI methodology is confirmed, this can help to develop multimodal AI referral CDSSs in other healthcare domains as well. If these type of CDSSs can support referrers to reduce inappropriate referrals from 25% to possibly 10%, this will not only significantly diminish the burden of incorrect referrals for patients but also lead to substantial annual healthcare cost reductions (> €150 million in the Netherlands). [1] M. Naseriasl, D. Adham, and A. Janati, “E-referral Solutions: Successful Experiences, Key Features and Challenges- a Systematic Review,” Mater. Socio Medica, vol. 27, no. 3, p. 195, 2015, doi: 10.5455/msm.2015.27.195-199.


end %-->