Over the past two decades, patient safety has undergone an evolution from an emerging domain to a measured indicator and, more recently, to an established cornerstone of healthcare quality, bolstered by robust patient, staff and organisational data.[1] However, the momentum can stall at data capture, leaving meaningful use of data limited.[2] Certainly, patient safety data from incident reporting, staff and patient surveys, organisational audits and many more have been used to drive improvement.

In the UK, the National Reporting and Learning System (NRLS), the national repository of patient safety incidents has been used to issue 150 patient safety alerts from February 2004 to June 2021,[3] thereby mitigating the threat of future incidents. Also in the UK, the “Learning from Litigation Claims” guidance was published in 2021 to support organisations in learning from NHS negligence claims and drive better patient safety.[4] In the US, the analysis of patient complaints has been demonstrated to be an effective way of monitoring diagnostic safety concerns.[5] These examples demonstrate how one data source has yielded singular types of outputs. What is missing is the triangulated use of patient safety data to render more complete and nuanced results. Given that many developed healthcare systems hold millions of data points from diverse sources and given the readily available power of advanced analytic techniques to clean, combine and interrogate these datasets, there is a notable absence of patient safety intelligence driven from multiple data sources. There are many reasons for the deficit in triangulation of patient safety datasets. They are sensitive, not easily shared, inconsistent over time and notoriously incomplete in comparison to other routine healthcare datasets.[6] However, these challenges do not present insurmountable barriers to all prospective data mining, and it is imperative that patient safety datasets are explored alongside each other to gain a holistic picture of safety in any given area. Patient safety improvement is currently compromised due to incomplete and fragmented insights; we could know more, learn more, do more to improve services if our data – and the findings from it - did not remain siloed.

Despite the volume of data and the availability of redacted, publicly available results on patient safety, there is virtually no evidence of a consolidated analysis of them. The NHS Patient Safety Strategy addresses this point and encourages the use of data linkage and advanced analytics techniques to identify emerging issues more quickly.[7] Given the size and scale of the NHS, attempting this required concentrated local efforts before expanding outward to the national level. In January 2020, the NHS London Patient Safety Leadership Forum adopted the challenge set out by the national strategy and led an effort to develop data-driven insights from a range of sources, using advanced analytic techniques. Their ambition centred on the use of publicly available data from acute care providers within London. The integration of patient safety insights can often be compromised because it requires dual expertise: knowledge of the complex safety data landscape as well as the analytics ability to manipulate and draw conclusions from confusing data. The Forum in London addressed this concern by convening a dynamic group of regional NHS policy makers, familiar with patient safety data and clinical priorities as well as a specialist patient safety research centre with dedicated analytic capabilities.

In the UK, the Care Quality Commission (CQC) conducts inspections of health and social care organisations and produces inspection reports that, amongst other considerations, provide a definitive ranking of whether a provider is safe. The CQC safety rating is widely considered the most consistent, reliable, and comprehensive indicator of patient safety in England and uses four ratings: outstanding, good, requires improvement, or inadequate. However, in addition to CQC inspection reports, the English NHS is awash with dozens of other safety data sources, namely NRLS incident reporting, which currently holds more than 20 million records, and patient and staff surveys.[8,9] To date, no study comprehensively triangulated various data sources in patient safety to provide novel insights on the characteristics associated with a positive or negative safety rating. Such analyses could help to understand the practical nuances of what makes a safe organisation, and therefore learn and scale up best practice on patient safety.