Eleven Health’s Early Warning System for Sickle Cell Crises
Sickle cell disease (SCD) is a genetic disorder of red blood cells that affects millions of people worldwide. Their cells become sickle-shaped due to a mutation in the beta-globin gene; these sickled cells can then become trapped, leading to complications including pain, stroke and organ failure.
Many SCD patients experience daily, chronic pain as well as intermittent and unpredictable acute painful episodes called vaso-occlusive crises (VOCs). VOCs are the leading cause of hospitalisations and emergency department visits for people with SCD, with hospital stays sometimes lasting for weeks. VOCs have a significant impact on patients’ quality of life due to the impact on their mental health, social life, work and school, as well as carrying a costly healthcare burden.
Currently, there is no cure for SCD and treatment is considered palliative. Therefore, the main therapeutic goal is to reduce VOCs and avoid hospitalisation.
Monitoring disease activity levels through patient-reported outcomes is one way of quantifying the degree to which these goals are met. However, this has traditionally been monitored through subjective measures such as patient-reported pain levels. This is commonly recorded through paper diaries and is associated with poor compliance and unreliable data. Subsequently, there is a large unmet need for objective patient-reported outcome instruments that accurately capture the patient experience of SCD.
Objective physiological measures of pain are easier to collect, cheaper, and more consistent. These measures include blood pressure, respiratory rate, oxygen saturation, temperature and pulse; changes to these vital signs have been reported to correlate with acute pain. Additionally, the collection of these objective measures over time would lead to a richer understanding of pain and its causes, ultimately allowing for the development of pain forecasting models.
Wearables are an ideal way to implement non-invasive real-time monitoring of body movement and physiological signals. These data points can be used for automatic pain assessment, improved pain management, and the creation of personalised patient care. This technology also has the potential to prevent costly unplanned hospital admissions.
There is a growing trend of applying machine learning techniques (ML) in various clinical settings. This includes several studies conducted by Duke University that utilised ML to investigate pain management in SCD, wherein they leveraged natural language processing, text mining and ML to predict pain change in 424 clinical notes of 40 SCD patients.
Several studies have employed ML techniques to predict pain from electronic health records (EHR), proving the feasibility of ML techniques in predicting pain scores and incidences of VOCs.
Patient data-driven predictions
Eleven Health is a London-based, patient-owned company that focuses on patients living with SCD and Thalassemia at a global level, and is powered by RwHealth’s Data Science Platform. Eleven combines primary and secondary care records of each patient alongside psychological wellbeing, pain, genomics and live data collected in real-time using wearables to create a digital platform as a solution for improving intervention, and prevention of VOC and end-organ damage.
Eleven provides SCD patients with consumer-grade, clinically validated wearables that facilitate the collection of continuous biometric data. This allows Eleven to build a complete view of each patient’s vitals data, not just at the time of VOC. In addition, patients complete daily pain and psychological scores and traditional surveys to create real-time insight into patient quality of life.
Using live patient data and ML techniques, Eleven has developed an advanced risk stratification model, called the Early Warning System. The Early Warning System combines patient biometrics, psychological and pain scores, weather, and previous patient admission data to predict patient risk scores for VOCs.
The risk stratification is built from each patient’s own 3-month baseline of real-time biometrics, identifying data points with the strongest links to patient-related pain (which is used as a proxy for VOCs). This creates a weighting system of the most critical data points in informing risks of high pain.
Eleven’s algorithm uses each patient’s live biometric data to identify any deviations from the patient’s own baseline to calculate a personalised daily risk score, with a minimum accuracy of 85% for 3-7 days ahead. This enables flagging of early indicators of upcoming health deterioration or complications, allowing for swifter intervention and prevention of further harm.
During the first 8 months of the Eleven programme, the development of predictive insights, including the Early Warning System, have seen improved patient outcomes including:
- 0 progressions to VOC
- 0 hospital admissions
- Fewer GP appointments
- 59% reduction in average pain score
- 90% reduction in days above average pain
- Improved sleep, activity and psychological scores
Initial findings from Eleven has demonstrated the feasibility of leveraging ML algorithms to develop Early Warning Systems that accurately forecast pain in SCD patients using objective physiological data before the point of VOC. Predictive pain insights have the potential to enable clinicians to provide targeted interventions and to empower patients’ self-management of SCD, avoiding costly unplanned interventions and improving individual quality of life.
To learn more about our work on sickle cell disease, email firstname.lastname@example.org
If you are a journalist looking for expert comment for news or features – or to arrange interviews, filming and photography – please email email@example.com