The CQC held vast quantities of notifications, comprising inconsistently categorised, variably complete, and largely free-text statutory notification returns. They were manually interpreted by inspectors who decided on an appropriate risk response. CQC was seeking to develop a decision support tool using artificial intelligence (AI) to help their inspectors sift, present and link information and support decision making. Success would be measured against four key indicators: increased efficiency, assessable data, tacit knowledge leveraged, increased consistency.
The project involved processing a significant amount of unstructured data to apply a range of natural language processing (NLP) and textual analytical techniques. The processed output notification data contained a number of structured and classified risk and temporal aspects that provide meaning and risk or safeguarding impacts.
This is used to provide actionable insight to inspectors, who are responsible for monitoring or undertaking an inspection of a care provider. Importantly, it also helps inspectors manage and prioritise their inspection workload.
Benefits of the project included previously unknown unextracted information being made available to inspectors for review; consistency of processing unstructured data and analysing across boundaries, provider organisations and types; automatic processing of manual and time intensive tasks and finally a methodology allowing CQC to improve organisational memory, and support inspectorate decision making.