Introduction
In rehabilitation care, every patient is unique. Therefore, each patient should be treated differently. Treatments that are tailored to a patient’s individual needs are called ‘personalised care’. Most clinicians are not skilled enough to adequately personalise their treatment. That’s why the PREPARE program aims to deliver a clinical decision support tool (CDST) that helps clinicians to provide such personalized care. This PREPARE tool will use data from previous patients who have received various rehabilitation treatments. By comparing the characteristics, contextual factors and treatment outcomes of previous patients with those of a new patient, the tool can predict the most likely course of symptoms and outcomes of possible treatments for the new patient. In theory, the information from such a tool could enable clinicians to discuss the expected prognosis with the patient and to decide together with the patient which treatment has a predicted outcome that outweighs any adverse events and should therefore be preferred. However, clinical practice can be stubborn and there are many barriers to the use of CDSTs, such as the PREPARE tool.
Crucial aspects for the implementation of CDSTs in practice
A systematic review of reviews conducted by the PREPARE group revealed that numerous studies (i.e., 23 reviews with a total of 615 individual studies) have already focused on perceived barriers and facilitators to the implementation of CDSTs in clinical practice. When taking a helicopter view of all these studies, three crucial aspects for the implementation of CDSTs in clinical practice can be distinguished. First, the design and applicability of CDSTs should be such that they can be easily integrated within the regular workflow without substantial effort, investment and costs. Second, the underlying algorithms that lead to the prediction should be accurate, credible and trustworthy. To achieve this, the algorithms should contain the relevant variables, which should be administered to the right patients in the right way and at the right time. Third, clinicians should have or develop the necessary knowledge and skills to apply CDSTs for their patients and should be open to integrating this tool into their daily routine. For each of these three crucial aspects, there are serious challenges that need to be overcome before CDSTs can really make a difference. In this blog, we want to highlight the third crucial aspect: how to sufficiently equip and motivate clinicians to apply the PREPARE tool with their patients?
How to equip and motivate clinicians to apply CDSTs?
Clinicians may not be eager to adopt a tool such as the PREPARE tool as it may limit them in their own clinical decision-making. A study from our group at HAN University of Applied Sciences showed that physiotherapists tend to overestimate their ability to predict the recovery of their patients. In a large sample of patients with acute low back pain, physiotherapists predicted recovery within three months (yes/no) no better than flipping a coin: approximately 50% correct. Along with their own overestimation, clinicians may also underestimate the potential added value of CDSTs. This suggests that clinicians may first need to realize that their clinical reasoning is suboptimal at best, and that a data-driven CDST may improve this. This could possibly be achieved by confronting them with their suboptimal predictions, especially when a CDST outperforms them. In addition, clinicians may need reassurance that such a tool will not replace their own clinical reasoning, as it is merely a support tool designed to enrich and further improve their clinical reasoning.
As clinicians start to use a CDST, many of them will face issues when explaining the result of a CDST to their patients. Explaining terms like ‘probability’, ‘expected outcome’ and ‘uncertainty’ in a way that patients (especially those with limited health literacy) can understand is a very complex task. In a recent study from Radboudumc in patients with intermittent claudication (which is also one of the nine pilot case groups of PREPARE), it was found that physiotherapists are particularly reluctant to communicate the outcome of a CDST (i.e., the Kompas Tool) to their patients in case of an unfavorable prediction. They do not know how to convey such a ‘negative’ message to their patients without demotivating them or introducing ‘nocebo effects’. This means that appropriate educational activities are needed to train these clinicians in understanding prognostics themselves and explaining it adequately to their patients (especially in case of an unfavorable outcome). This kind of training should help clinicians to integrate the use of a CDST into their routine clinical decision-making process in such way that a CDST assists them in making the right clinical decisions (instead of blindly following the CDST).
Visualizing the prediction in a figure may help a clinician to explain the expected course or treatment outcome to their patient. However, what such a visualization should look like appears to be quite difficult. In a previous project, we have developed a prediction tool to predict whether patients with acute low back pain will recover within 3 months and integrated this in a CDST (‘Practical Prediction Tool’). A visualization (see Figure A) was designed to illustrate the predicted recovery of a current patient as favorable (expected recovery) vs. unfavorable (expected non-recovery) and to compare this with the predicted recovery of 99 comparable previous patients.
Figure A. Prediction visualization from the ‘Practical Prediction Tool’ CDST
Many of the participating patients and physiotherapists appreciated this visualization, but many others did not because they found it too complex. This shows the variation in preferences among end-users for visualizations within CDSTs. An even more complex visualization has been developed for the Kompas Tool, as mentioned above. In this visualization (see Figure B), the expected course of a new patient is displayed, together with an area of uncertainty around this expected course and the actual course of about 200 comparable previous patients (based on which the expected course was estimated). An advantage of this visualization is that clinicians can opt not to show parts of this visualization (e.g., the area of uncertainty or the course of previous patients) to make it easier to understand. This allows the visualization to be personalized to the needs and preferences of each individual patient. However, the visualization is still relatively complex, which may hinder especially patients with limited health literacy to understand their prognosis and treatment possibilities. To prevent visualizations from being too complex, patients and clinicians should be involved in the development of such visualizations.
Figure B. Prediction visualization from the ‘Kompas Tool’ CDST
Conclusion
To summarize, data-driven CDSTs such as the PREPARE tool, in addition to the clinicians’ own clinical expertise (‘gut feeling’) could help these clinicians and their patients to make better treatment choices with more realistic treatment expectations. We emphasize the importance of clinicians being adequately equipped and motivated to apply CDSTs to their patients, and the need for easy-to-understand visualizations of predictions from a CDST, in order to achieve successful implementation of CDSTs in clinical practice. In the PREPARE program, we will address these crucial aspects, by involving end-users in the design of the PREPARE tools and by developing educational activities for clinicians on how to apply these tools for optimization of their clinical decision making process.
Authors: Dr. Jesper Knoop, Dr. Bart Staal, Dr. Esther Janssen, HAN University of Applied Sciences & Radboud University Medical Centre