Using a monthly report sent from the clinical biochemistry laboratory to participating general practices, we sought to provide rapid access to a list of patients with the most sub-optimal control who were overdue for their routine HbA1c test. This aimed to assist practices in making more effective use of resources to tackle the backlog of overdue tests by prioritising those who were most at risk.

Anonymised patient-level HbA1c test data, date of test, result, and requesting location were extracted from the Laboratory Information and Management System at Cambridge University Hospitals NHS Foundation Trust. These data were collected initially for a complete 3-year period from 01/11/2017 to 31/10/2020, and then updated monthly on a 36-month rolling basis until the effectiveness of the programme was analysed in October 2023.

Data comparisons were made between two groups of general practices: an ‘Intervention’ group comprising 60 practices (Cambridgeshire and Peterborough Clinical Commissioning Group) who received monthly reports as described below, and a ‘Non-intervention’ group comprising 51 practices who did not receive the reports (East and North Herts Clinical Commissioning Group).

From these data, two data sets were created: the ‘2020’ data set, reflecting the 3-year period prior to the intervention (01/11/2017 to 31/10/2020), and the ‘2023’ data set, reflecting the 3-year period of the intervention (01/11/2020 to 31/10/2023). The 2020 set comprised 911,117 tests in 452,314 patients (mean 2.01 tests/patient) and the 2023 set comprised 1,154,229 tests in 537,132 patients (mean 2.15 tests/patient). From these, a subset of tests in regularly managed patients were identified as described previously [7]. Briefly, these tests were defined as those in people who had more than one HbA1c test during the 3-year period where at least one was ≥ 48 mmol/mol. Hence, these are likely to be those with pre-existing diabetes. Only those people with an identifiable General Practice (GP) were included. For the 2020 period, this included 275,843 tests on 59,206 patients (mean 4.66 tests per patient), and for 2023, 307,525 tests in 65,449 patients (mean 4.70 tests per patient).

Step 1: Data extraction and anonymization

Data were extracted at patient level, using NHS number, laboratory request number, name, sex, and date of birth as patient identifiers. These were anonymised within the laboratory using an in-house coding system. Identifiable data were removed prior to sharing with the wider project team, leaving only an unique identifier, sex, and age within the extract, alongside date of test, requestor code (GP national code, or blank for non-primary care requests), and test result.

Step 2: Data processing

For each patient, the time intervals between tests and change in HbA1c levels between the most recent tests were calculated. Tests were defined as overdue based on previous test result and intervals between tests. Recommended testing intervals were derived from a combination of UK National Institute for Health and Care Excellence (NICE) guidelines and previous studies examining testing interval versus change in HbA1c [3, 4, 6,7,8,9,10,11].

‘Amber ‘sub-group (previous test result of 58–75 mmol/mol): NICE guidelines suggest that an interval of 3–6 monthly is recommended [3, 4]. We allowed a 7-day ‘leeway’ after the 6 months before a test became classified as overdue. Tests in this group therefore became labelled as overdue at 183 + 7 (190) days.

‘Red sub-group (previous test result of > 75 mmol/mol): 1–3 monthly based on previous studies showing that more frequent testing in the highest risk patients is linked to improved diabetes control [3, 4]. Due to the potentially urgent nature of these tests, the system did not allow a 7-day leeway. Tests within this category therefore became overdue at 92 days.

Step 3: Report generation and distribution

The Benchmarking Partnership team then built a reporting system in Microsoft Excel that enabled the generation of an individual PDF-format report for each general practice through a de-anonymisation process within the clinical laboratory.

The reports comprised a list of the highest-risk patients, defined as patients whose most recent HbA1c result was ≥ 58 mmol/mol. This was to ensure that the monthly reports were concise enough to be manageable for GP practices. Selected patients within the reports were highlighted separately within the ‘Amber’ and ‘Red’ sub-groups. This allowed practices to further focus their attention on the most urgent cases, in the event that time and resources were limited.

Reports were distributed each month by the laboratory to each general practice and included a list of all patients currently overdue their HbA1c test at that time. It also included the date and result of the most recent test, change in HbA1c for the most recent test relative to the previous test, and how many days overdue the next test was. It also included the patient’s name, NHS number, and date of birth, which was re-inserted by the laboratory team prior to distributing the reports via e-mail.

An example of the report is shown in Supplemental Figure S1 (with patient identifiers redacted).

Steps 1–2 were also completed for the ‘Non-intervention’ group, but reports were not sent to the practices.

Step 4: Data analysis

For each practice in both ‘Intervention’ and ‘Non-intervention’ groups, the proportion of people overdue a test and the median number of days the latest HbA1c test was overdue (overall and stratified by ‘Amber’ and ‘Red’ sub-groups) was calculated for those with a HbA1c ≥ 58 mmol/mol for the two time periods (2020 and 2023). For the whole patient group, the proportion of patients with a HbA1c of ≥ 58 mmol/mol, the median HbA1c level and the number of tests per patient per year were also calculated for each time period.

Univariate cross-sectional analysis compared the practices in the ‘Intervention’ group with those in the ‘Non-intervention’ group using Mann–Whitney U test. For 2023, multivariable linear regression analysis was also used. Models included adjustment for equivalent baseline (2020) value, GP list size, deprivation score, proportion of male patients, proportion aged > 65 years and diabetes prevalence.

Longitudinal analysis examined the percentage change in parameters between 2020 and 2023 using both univariate and multivariable analysis as for the cross-sectional analysis.

Stata version 18 (Stata Corporation, TX) was used for all statistical analyses. A significance value of p < 0.05 was considered statistically significant.

Data and Resource Availability Statement: The datasets generated and analysed during the current study are not publicly available due to potential commercial sensitivities but are available from the corresponding author upon reasonable request.

This study is part of a service development programme to increase the quality of laboratory test requesting. Hence, it includes a service evaluation and audit of local practice over time with a view to implementing a service development intervention to enhance the clinical laboratory service and improve conformity to recommendations on monitoring intervals. Accordingly, this study was not considered to be research using the decision tool provided by the UK Health Research Authority [17] and did not require NHS Research Ethics Committee review. All data extracted from Laboratory Information and Management Systems were anonymised.