Ruud Selles

Personal

Research area

My ambition is to use innovative research approaches to find better ways to treat disorders of the upper extremities and to extend these innovative approaches to other research areas in rehabilitation medicine.

The stakes are high: most stroke survivors suffer from hand and wrist impairments. Musculoskeletal hand and wrist disorders, too, are very prevalent. And yet, we lack both evidence-based treatments for many of these disorders and the knowledge of how individual patient characteristics determine treatment outcomes. This presently results in overtreatment, under treatment and non-optimal treatment. There is a real need for a breakthrough. Over the last five years, I have developed a fundamentally different research approach by collecting clinical data and analyzing it using innovative statistical and data science techniques. To do so, I have:

Developed smart collection approaches for real-world clinical data to measure treatment effects and study prognostic markers in daily clinical practice
Performed comparative effectiveness studies with real-world clinical data to analyze variability in treatment effect and compare treatment options
Developed clinical prediction models to evaluate and predict differences in individual responses to treatment

This is what I am most proud of

By means of leading several randomized controlled trials and comparative effectiveness studies, I contributed to the current evidence on the effects of treatments for upper extremity stroke rehabilitation (studying effects of mirror therapy on motor learning in stroke), and treatment of thumb base osteoarthritis (studying effect of exercise and splinting on preventing surgery with hand therapy and comparing different surgical approaches).

To obtain evidence for the effect of a large number of treatments and develop clinical prediction models for different hand and wrist disorders, I initiated and co-developed a smart outcome measurement system for hand surgery and rehabilitation that has grown into a dataset that is unique in the world (including 60.000 patients and >100 different types of treatment) with highly detailed treatment and outcome information. An increasing number of national and international partners have started to work with this unique data set, providing evidence for the effectiveness of many different upper extremity disorder treatments. We are now extending this approach to innovate clinical data collection at Rijndam Rehabilitation Center.

To better understand and predict upper extremity motor recovery after stroke, I co-developed a dataset on stroke recovery that we used, combined with innovative modeling approaches, to predict treatment response in individual patients. One of the results of that is that we can now screen patients early after stroke to determine who are eligible for future upper extremity motor rehabilitation programs based on predicted poor, moderate or good motor recovery. Last year, we published two prediction models of stroke recovery in leading international journals.

Milestones for the coming 2 years

I identify four interconnected themes for the coming years:

Theme I: Smart collection of observational clinical data
A personalized treatment approach requires high-volume and high-quality datasets and a high level of detail about the patient characteristics (phenotype/biomarkers) and treatment outcomes. Such datasets were traditionally not available for studying the treatment of upper extremity disorders. With both the Hand Wrist Study Cohort and the Profits data, we now have internationally unique datasets available to take steps that other research groups presently cannot. To go beyond the current state of the art, even more high-quality and high-volume data sets need to be developed or exploited. Together with Rijndam, we have defined implementing routine collection of observational clinical data as a key target in the following years to and I plan to contribute to this ambition. This will lead to innovative data sets currently not available in this field.

Innovation in data collection will be in many forms in the following years. Among others, there is a need to combine data from multiple clinical data sources to increase the number of patients and the amount of outcome and biomarker information about each patient. For example, there is enormous potential in electronic data such as electronic patient records, pharmacy registers, appointment information, and billing systems. In addition, increasing, sensor data will become available

Theme II: Information about the effectiveness of treatments
Knowledge about doing comparative effectiveness research on observational clinical data is rapidly evolving and more and more datasets are becoming available. In terms of clinical application, the data sets mentioned earlier have tremendous potential for further comparative effectiveness studies on many upper extremity disorder treatments. The consortium may already hold detailed outcome information for almost all elected treatments for upper extremity motor disorders. By analyzing these data in collaboration with national and international partners, I want to increase the evidence base for many treatments.
Theme III: Developing clinical prediction models
To allow the selection of the right treatment for the right patient at the right time, I want to develop and extend clinical prediction models of upper extremity disorders. This will sometimes require collecting more data or data of higher quality but also innovation in data science to develop such models. .
Theme IV: Implementing and evaluating our findings in clinical practice
Improving clinical care on an individual patient level should be the primary goal of routine outcome measurement. This needs to be obtained by reducing the costs of measurements on the one hand and using the data for clinical care through, for example, individual patient data dashboards and extreme value detection. To do so, we need to further develop tools to process the collected data and use it to inform patients and caregivers so that clinical decision-making and shared decision-making are enabled. For example, we recently developed a tool for real-time dynamic prediction modeling within the Gemstracker data collection environment. With this tool, clinicians can ask for a prediction based on an electronic patient record at any time. How this impacts decision making will be evaluated, and, where possible, we will evaluate if implementing a data-driven treatment approach leads to better treatment outcomes.