Summary

COVID-19 resulted in large backlogs of diabetes patients, with no way to manage risk for those waiting. Working in conjunction with diabetologists, service leads and analytics, Guy’s & St Thomas’ NHS Foundation Trust (GSTT) defined, coded and validated clinical rulesets to objectively identify patients on waiting lists who might be deteriorating between appointments. Over 120 high-risk patients were identified and appointments were expedited based on their pathology tests. These patients were disproportionately represented by ethnic minority groups and areas of high deprivation. Clinical evaluation indicated that harm was prevented in ~40% of cases.

Innovation

Data-Led Prioritisation (DLP) was co-designed with diabetologists, service leads and analytics experts. It brings together the best of what is available but often difficult to aggregate, using technology. The project delivered impact within 18 months. Adopting surgical pathways’ well-recognised terminology of P1-P4 (highest to lowest priority) in Outpatients provided clarity, without creating more risk prioritisation processes. The clinician-facing dashboard was co-designed with colleagues from the service. It presented each patient’s risk category against all relevant events, pathology data and wait time for next scheduled routine appointment, reducing assessment time to two minutes from over five. In addition, a bespoke clinic administrator dashboard showing appointment allocations according to risk and guide booking practice around clinical urgency was developed. The front end was designed in Power BI, which is available to every NHS organisation. The trust is implementing a new Electronic Health Record (EHR) and the technology was developed with sustainability in mind. The outputs are easily integrated with other data portals and new systems can be incorporated. The new EHR will enable further development of DLP so that patients have a summary of their pathology data and risk score. DLP optimises existing knowledge, processes and data to harness the power of available information.

Dissemination and Sustainability

The DLP pilot was specific to the diabetes department at GSTT, but learnings and methodology are being shared with diabetes departments across the country and Europe. An iteration is being created with South East London to facilitate population health management and address diabetics with comorbidities. The clinical diabetes team has drafted papers for publication and presented them at European conferences. It is hoped that the HSJ award for ‘Driving Change through Data and Analytics’ will support the dissemination of the methodology and associated benefits. Five ‘fast follower’ specialities with similar serology to diabetes are working to adopt the DLP methodology for risk score data. A ‘clinical library’ of rules across the specialities will be available to other trusts wanting to develop DLP.

Implementation of New Technologies

The enterprise data warehouse was used to collate and integrate data from trust clinical systems. By leveraging a range of data points across various systems, DLP provides clinicians with the most up-to-date view of a patient’s health in one place using a ‘risk scoring engine’, not automated decision making. The front end can be embedded in the trust’s existing infrastructure and deliver insights through a simple traffic light system. No new technology had to be purchased or licensed, so actionable insights were delivered in months. Clinicians are alerted to new patient risk factors seamlessly and without logging into multiple care systems. COVID-19 necessitated bulk cancellations of outpatient activity, resulting in enormous backlogs. GSTT’s diabetes follow up waiting list was ~8,000 patients, with 10% not booked. Manually assessing the waiting list to identify deteriorating patients was not feasible. The project was clinically led and translated the rules that diabetologists use into simple rulesets that could be applied to data. Overlaying the risk scoring assessment with patients’ next booked appointment slot revealed highrisk patients who were not due to be seen for over three months and stable patients who were being seen regularly. A strategy was developed, centred around these principles: treat those patients in most clinical need and measure this through reduced harm; ensure that the solution actively reduces existing health inequalities; be clinically driven and leverage the knowledge of clinicians when designing, shaping, implementing and evaluating the solution; ease the operational burden associated with booking patients; fully utilise trust data. Customer input helped shape the health analytics model. During the pilot phase, appointments for >120 high-risk diabetes patients were expedited. Clinical evaluation indicated that harm was prevented in ~40% of these cases. A total of 600 low-risk diabetes patients were contacted to put back their appointments. Clinicians reported that the tool assured them that they were making informed decisions about the waiting list. The methodology also helped to standardise decisions across the department and streamlined administrative processes. Leveraging all of the information in the GSTT data warehouse had ancillary benefits. Patient safety was at the heart of the programme. A panel of 10 diabetes specialists identified the most appropriate clinical risk criteria, focused on the modifiability of risk and accessibility of information in the EHR database. This included a robust evaluation of risk criteria proposed by a national organisation for risk stratification in diabetes. The model used seven risk criteria, including new data on HBa1c, EGFR lab results, A&E attendances and diabetic eye treatments. An independent medical statistician developed the validation methodology. Blind tests were carried out with 450 diabetes patients and nine clinicians to ensure the sensitivity of the model to detect high-risk people. When there was discordance, cases were reviewed to obtain consensus on the final risk category status. The tool detected 83% of cases that were flagged as concerning by clinicians, as well as 81% of lower-risk patients. Next came ‘real world testing’. Over 350 checks were run on the stability of the data before it was ingested into the scoring platform. Expediting appointments for high-risk patients has patient and system benefits, with associated cost savings. Having identified 17% of the cohort as low risk and demonstrated that ~42% of those can avoid an appointment saved ~500 follow-up appointments a year, worth approximately £70,000 to the trust if they can be used to see more patients awaiting their first consultation.