Driving data quality improvements to support elective recovery
December 6, 2022 • Reading time 5 minutes
As the NHS continues its progression towards the era of data-driven healthcare provision, the importance of high-quality data continues to grow. Commissioners and providers, now more than ever, rely on data to signal under- and overprovision of services, so that limited resources can be reallocated where they are most needed, and more patients can receive the care they need, quicker.
With data lying at the heart of this process, it is crucial to maintain and improve its quality. This can present a challenge to operational and business intelligence teams that are already strained under the ever-growing administrative burden of numerous internal and national data collections and reporting.
Edge Health worked with a large NHS trust, supporting them with improving their waiting list data quality. The details of this work are presented below. Off the back of this work, all records on the inpatient and outpatient waiting lists were validated by the operational teams, with more than 370 erroneous records corrected within the first 2 months of use. Internal processes were also put in place to minimise the chances of further errors occurring and to correct those that do fall through the cracks.
The importance of high-quality accurate data in the NHS is growing
Data forms the backbone of healthcare decision making infrastructure. “How many patients are waiting?”, and “What are they waiting for?” are the key questions that determine the focus of short- and long-term service planning efforts. In the light of exploding waiting lists, answering these questions is of paramount importance – knowing the answers allows Trusts to put their efforts into the aspects of delivery that would make the biggest difference to patient outcomes.
Waiting list data is prone to error, resulting in delayed treatment
In order to answer these questions, one would need to dive into what is known as the PTL. PTL stands for Patient Tracking List which is exactly what it says on the tin – a list of patients waiting for elective care. In the past to find an answer to this question, one would physically go through stacks of paper files stored in a hospital. Unsurprisingly, this was a very error-prone process, with patient files falling through the cracks (potentially quite literally). Luckily, now that the NHS is moving digital, the days of sifting through the physical paper files are over (fingers crossed…). However, digital data can still be (and often is) full of errors.
We have worked with many datasets across all care settings and have seldom found a dataset without errors. It is easy to get desensitised to the concept of data quality issues, but we should always remember that in the context of healthcare, there will be a personal story behind every number. Data quality issues can lead to patients going missing and not being called in for their procedure, valuable resources mis-allocated or misinformed staffing decisions made. Knowing the cost of inaccurate data, a large Trust decided to partner with Edge Health to tackle the data quality issues in their PTL.
Edge partnered with a large Trust to support them with improving their waiting list data
We worked closely with the Trust’s internal team to identify and fix their data quality issues. We did this as follows:
1. Identify criteria for erroneous records
Firstly, we held a series of workshops with the operational leads to establish the criteria for erroneous records and to come up with a way of identifying those through manual inspection of the records held in the database. For example, we discovered that lots of patients with pre-op activities were disappearing from the waiting list before completing their treatment – pre-op activities often closed the referral to treatment pathway. As a result, patient records were falling through the digital cracks of the PTL.
2. Fix existing errors
Once the approach was signed off and we knew what to look out for, we extracted erroneous records from the internal databases and provided these to the operational leads, who were able to take action to correct them. For example, when a patient was lost and not invited for treatment on time, the team were able to reach out to them and call them in. Many errors (such as typos) were fixed simply by correcting the data stored in the database.
3. Establish source of errors and review policies
We then set off to identify the source of the errors so that a process could be developed to prevent these from occurring in the future. We discovered that data quality errors tend to fall into two categories:
- Data entry (e.g., typos or mistakes made by the user when inputting the data)
- Data processing (e.g., loss or duplication of data, misattribution, and other issues occurring when the data is being processed and manipulated)
Standard measures to counteract data entry errors such as logic checks at the point of entry, user training and BI/engineering team training were then put in place to minimise the chances of these errors occurring.
4. Automate process for error discovery
Once the approach was signed off, we worked with the data engineering team to automate the process of error discovery so that the quality of data can be monitored and the trends can be reviewed over time. To facilitate that, we developed an automated data pipeline that would run every day, detect erroneous records and write them into a table in the Trust’s database.
5. Build a dashboard
We then built an interactive PowerBI dashboard that allows operational leads to access this information, review existing erroneous records directly, and correct them. In addition to that, it provides high level oversight of the state of data quality in the Trust, as well as data quality trends, which is helping the executive team steer the Trust in the direction of complete, high quality PTL data.
6. Work with the team to develop internal processes
Upon the completion of the project, internal processes were put in place for dashboard maintenance and operational use. Every day, the data in the dashboard would get refreshed so that the operational teams can have the most up-to-date information and use it to correct the erroneous records. The teams would also meet every week to review trends and identify priority areas for data quality improvement