Exploration of reasons for primary care testing (the Why Test study): a UK-wide audit using the Primary care Academic CollaboraTive

Background Rates of blood testing have increased over the past two decades. Reasons for testing cannot easily be extracted from electronic health record databases. Aim To explore who requests blood tests and why, and what the outcomes of testing are in UK primary care. Design and setting A retrospective audit of electronic health records in general practices in England, Wales, Scotland, and Northern Ireland was undertaken. Method Fifty-seven clinicians from the Primary care Academic CollaboraTive (PACT) each reviewed the electronic health records of 50 patients who had blood tests in April 2021. Anonymised data were extracted including patient characteristics, who requested the tests, reasons for testing, test results, and outcomes of testing. Results Data were collected from 2572 patients across 57 GP practices. The commonest reasons for testing in primary care were investigation of symptoms (43.2%), monitoring of existing disease (30.1%), monitoring of existing medications (10.1%), and follow up of previous abnormalities (6.8%); patient requested testing was rare in this study (1.5%). Abnormal and borderline results were common, with 26.6% of patients having completely normal test results. Around one-quarter of tests were thought to be partially or fully unnecessary when reviewed retrospectively by a clinical colleague. Overall, 6.2% of tests in primary care led to a new diagnosis or confirmation of a diagnosis. Conclusion The utilisation of a national collaborative model (PACT) has enabled a unique exploration of the rationale and outcomes of blood testing in primary care, highlighting areas for future research and optimisation.


Introduction
Routine data from primary care electronic health records has demonstrated a more than threefold increase in the use of laboratory tests in UK primary care between 2000 and 2016, 1 with significant variation in testing rates between GP practices. 2This rise in testing has taken place in the context of significant uncertainty and lack of evidence to determine which tests are 'necessary', with guidelines for chronic disease monitoring mostly based on expert opinion. 3timates have suggested that 25% of primary care laboratory tests might be 'unnecessary', 4 with research demonstrating unwarranted variation 5 and overuse of specific tests including thyroid function tests, liver function tests, prostate-specific antigen tests, and vitamin D tests. 6This may lead to further blood tests, imaging, appointments, and referrals, a process sometimes referred to as the 'cascade effect'. 7The concept of the cascade effect has been around for over 30 years 8 but is rarely measured, 9 and the overall frequency and implications of cascade testing on primary care workload is unknown.
Reduction in unwarranted variation in testing rates has been frequently cited as an aim, 10 particularly in the current UK context of rising workload, 11 a primary care workforce crisis, and concerns about socioeconomic inequalities in health.A prerequisite to achieving this aim is to first understand the rationale for blood testing in primary care, and the outcomes of testing.This information cannot be obtained easily from current electronic health record data.

Background
Rates of blood testing have increased over the past two decades.Reasons for testing cannot easily be extracted from electronic health record databases.

Aim
To explore who requests blood tests and why, and what the outcomes of testing are in UK primary care.

Design and setting
A retrospective audit of electronic health records in general practices in England, Wales, Scotland, and Northern Ireland was undertaken.

Method
Fifty-seven clinicians from the Primary care Academic CollaboraTive (PACT) each reviewed the electronic health records of 50 patients who had blood tests in April 2021.Anonymised data were extracted including patient characteristics, who requested the tests, reasons for testing, test results, and outcomes of testing.

Results
Data were collected from 2572 patients across 57 GP practices.The commonest reasons for testing in primary care were investigation of symptoms (43.2%), monitoring of existing disease (30.1%), monitoring of existing medications (10.1%), and follow up of previous abnormalities (6.8%); patient requested testing was rare in this study (1.5%

Conclusion
The utilisation of a national collaborative model (PACT) has enabled a unique exploration of the rationale and outcomes of blood testing in primary care, highlighting areas for future research and optimisation.

Keywords
blood tests; collaborative research; clinical decision-making; diagnosis; overtesting; primary health care.

RESEARCH | e134
including GP registrars, GPs, and allied health professionals (hereinafter 'PACT members'), by extracting data on recent blood tests undertaken in the GP practice where they were working.Information about the study was disseminated via the PACT newsletter, social media, Clinical Research Networks, GP registrar newsletters, and the newsletter of a national GP leadership programme ('Next Generation GP').PACT members completed an online expression of interest and consent form.A GP partner or practice manager was then required to complete a practice agreement form.
Purposive sampling was used to recruit the first seven pilot GP practices, include a range of PACT team members and a range of electronic health records systems (EMIS, SystmOne, and Vision), and identify any problems with datacollection tools before the wider rollout.All PACT members who expressed an interest were invited to take part, with an aim of recruiting at least 50 practices.

Training
PACT members were required to watch two short training videos and code three fictitious clinical cases using a computerised database (REDCap) before commencing data collection (see Supplementary Box S1).
A pass mark of >70% for each of the three test cases was set; all participants exceeded this on the first attempt for each of the three training cases, with mean scores of 95% for case 1 (range 73%-100%), 94% for case 2 (range 80%-100%), and 94% for case 3 (range 73%-100%).Supplementary Table S1 shows training case scores for GPs, compared with GP registrars and allied health professionals.

Data collection
The research team provided PACT members with an automated search to identify a random sample of eligible participants from their GP practices' electronic health records system.Eligible patients were anyone aged ≥18 years having a blood test in primary care during April 2021.This period was chosen pragmatically to capture usual practice following the early waves of the COVID-19 pandemic, and to allow sufficient time for follow up before issues with primary care blood bottle shortages in the UK during August 2021.Pregnant women were excluded manually by the PACT members because of biochemical differences in reference ranges for routine bloods.
Each PACT member reviewed the notes of 50 patients and manually extracted anonymised data into a REDCap database on: patient demographics (age and gender), type of clinician who requested testing, primary reasons for testing (plus optional secondary reason for testing), symptoms triggering testing, test results, and the outcomes of testing (new diagnosis, medication change, lifestyle recommendation, referral, hospital admission, further tests, reassurance, or 'none of the above').Symptoms triggering testing were subcategorised using the International Classification for Primary Care (ICPC2). 14parate components of a test (for example, individual analytes of a full blood count) were grouped and counted as one test.For each test the results were categorised by PACT members into normal ('all test results are within the laboratory specified reference ranges'), borderline ('one or more tests are very slightly outside of the laboratory specified reference range'), or abnormal ('one or more tests are definitely outside of the normal range').
Categorical and free-text data were collected on how test results were coded, actioned, and communicated, which will be reported separately. 15The final question PACT members were asked to respond to for each patient was: 'In your clinical opinion, were the tests necessary?'.This could be categorised into a) yes, all tests were necessary; b) some tests were necessary, but not all; or c) no tests were necessary.It was emphasised that this question relies on clinical judgement and these findings should be seen as exploratory in nature.

Analysis
Results were analysed using simple descriptive statistics.Logistic regression analysis was used to estimate the association between the frequency of abnormal results and patient age group, gender, the reason for testing, and the type of clinician who requested the test.Results were presented using odds ratios (ORs) and corresponding 95% confidence intervals (CIs) to quantify the strength and direction of the association between each independent variable and the frequency of abnormal blood test results.All statistical analyses were performed using Stata (version 17).

How this fits in
Previous research has shown a more than threefold increase in the use of laboratory tests in UK primary care between 2000 and 2015, with significant variation in testing rates between GP practices.In this study, around one-quarter of tests were thought to be partially or fully unnecessary when reviewed retrospectively by another clinician.Around half of tests (48.8%) did not lead to any change in management or reassurance; 13

Results
Eligibility and consent forms were received from 149 PACT members, from which a total of 57 PACT members (from 57 GP practices) were recruited; 92 PACT members who expressed an initial interest did not complete the relevant study documentation required for participation or withdrew from the study (Figure 1).
Recruited practices came from England (n = 46), Scotland (n = 4), Wales (n = 5), and Northern Ireland (n = 2); demographics of participating practices are shown in Table 1.The majority of participating practices had list sizes between 5000 and 15 000 (38.6% 5001-10 000; 33.3% 10 001-15 000), this compares with an average practice list size of 9544 in England. 16actices were recruited from all regions of the UK, with slightly higher numbers in the South West Peninsula (12.3%) and East Midlands (10.5%).Practice-level Index of Multiple Deprivation data show a higher number of practices were recruited in more deprived than less deprived areas, including 19.3% in the most deprived areas, in keeping with PACT aims of broadening participating in research.
After exclusions, data on a total of 2572 patients were included in the analysis (Figure 1).Age and gender of included patients are shown in Table 1.The cohort was 58.1% female, with the majority of tested patients between the ages of 50 and 79 years, in keeping with previous research exploring the demographics of primary care testing. 1 Table 2 shows the tests performed and frequency of borderline and abnormal results.The most commonly performed test was urea and electrolytes (U&Es) followed by full blood count (FBC), Calculated using Fingertips data for practices in England 16 and using GP practice postcode IMD for devolved nations: Scotland, 17 Wales, 18 Northern Ireland. 19IMD = Index of Multiple Deprivation.The mean number of tests done simultaneously was 4.5 tests (standard deviation [SD] 2.4) per patient (counting FBC, U&E, and LFTs as a single 'test', rather than counting each analyte separately) (Table 3).If all simultaneous tests performed on an individual patient were considered, ≥1 of these tests were coded as 'abnormal' in 1176 (45.7%) patients; hereinafter this is referred to as 'abnormal' (data not shown).In 712 (27.7%) ≥1 tests were coded as 'borderline' with no 'abnormal' results; hereinafter 'borderline'.In 684 (26.6%) patients, all tests were within the laboratory specified reference range; hereinafter 'normal'.
Table 4 shows which member of the primary care team requested blood tests, the number of tests requested on average by each type of clinician, and proportion of tests that were abnormal.Tests were most commonly requested by GPs (47.0%).Logistic regression (adjusted for age, gender, and reason for testing) showed lower rates of abnormal test results for nurse practitioners (OR 0.54, 95% CI = 0.36 to 0.79, P = 0.002), tests requested according to practice protocols (OR 0.74, 95% CI = 0.57 to 0.98, P = 0.03), and tests requested by secondary care (OR 0.58, 95% CI = 0.40 to 0.85, P = 0.005) compared with tests requested by GPs.
Table 3 shows the primary reasons for testing, mean number of tests requested, and the frequencies of abnormal results.The commonest reason for testing was investigation of symptoms (43.2%), followed by monitoring of existing disease (30.1%), monitoring of existing medication (10.1%), and follow up/repeat of previous abnormal result (6.8%).Testing to investigate symptoms was associated with the largest number of simultaneous blood tests (mean 5.5 tests) followed by monitoring of existing disease (mean 4.3 tests).Starting new medications and follow up/repeat of previous abnormal result were associated with the smallest number of tests (mean 2.4 tests, respectively).
Monitoring of existing disease yielded the highest frequency of abnormal results (56.4%), followed by follow up/repeat of previous abnormal result (49.4%), and investigation of symptoms (42.0%) (Table 3).In patients having testing to investigate symptoms (n = 1111), the most frequently recorded symptoms were 'general and unspecified' (20.1%), followed by digestive symptoms (17.0%) and musculoskeletal symptoms (12.2%).For full details of symptoms triggering testing see Supplementary Table S2.Supplementary Table S3 shows both primary and secondary reasons for testing; secondary reasons were optional  The final question in the study was 'In your clinical opinion, were the tests necessary?'.Overall, in 1927 (74.9%) patients, all tests were felt to be necessary, in 538 (20.9%) patients, some tests were felt to be necessary but not all, while in 107 (4.2%) patients, no test was felt to be necessary.Supplementary Table S5 shows how the frequency of tests that were felt to be necessary varied, depending on the indication for testing.

Summary
The commonest reasons for testing in primary care were investigation of symptoms (43.2%), monitoring of existing disease (30.1%), and monitoring of existing medications (10.1%).Only around half of tests in primary care were requested by GPs, reflecting the multidisciplinary nature of UK primary care.
On average, 4.5 tests were requested simultaneously per patient, and abnormal and borderline results were common, with only 26.6% of patients having completely normal test results.Around one-quarter of tests were thought to be partially or fully unnecessary when reviewed retrospectively by another clinician.Overall, 6.2% of tests in primary care led to a new diagnosis or confirmation of diagnosis.Nearly half of tests (48.8%) did not lead to any change in management or reassurance.Tests actioned to monitor existing disease led to the highest frequency of abnormal results; this is to be expected given that patients with chronic conditions such as type 2 diabetes or chronic kidney disease would be expected to have abnormal tests as a result of their condition.

Figure 1 .
Figure 1.Participant flowchart.a Total figure slightlyhigher than expected because: two practices had problems with the automated searches identifying a large number of ineligible patients and therefore 'topped up' their dataset; five practices completed 51 and one practice completed 53 data-collection forms without explanation; and for three small practices the search returned <50 eligible patients.b More than one reason for exclusion could be selected.

Table 1 . Demographics of practices and patients
a

Table 2 . Tests performed and frequency of borderline and abnormal results a Name of test Normal results, n (%) Borderline results, n (%) a Abnormal results, n (%) a Total tests, n
a'Borderline results' defined as 'very slightly outside of the laboratory specified reference range'; 'abnormal' defined as 'definitely outside of the normal range'.b Tests with n<100 merged into 'other' category.HbA1c = glycated haemoglobin.U&E = urea and electrolytes.

Table 3 . Primary reason for testing and frequency of abnormal results (n = 2572)
a Proportion of abnormal tests, compared with reference group (monitoring tests), adjusted for patient age, gender, and type of clinician requesting the test.N/A = not applicable.OR = odds ratio.SD = standard deviation.

Table 5 . Outcomes of blood testing (n = 2572) Consequences of testing n(%) a
The category 'none of the above' was used to identify tests where no change in outcomes could be identified following testing; the authors avoided using the wording 'no change in outcomes' to reduce potential subjective interpretation of what could be defined as a 'change in outcome'.
a Total >100% as >1 option could be chosen simultaneously.b

Table 4 . Member of the healthcare team made the clinical decision to request the blood test (n = 2572) Healthcare team member Patients tested, n (%) Number of tests per patient, mean (SD) Frequency of abnormal results, % OR of receiving an abnormal result (95% CI) a P-value
Odds of receiving an abnormal test result by clinician group compared with GP testing (adjusted for age, gender, and reason for testing).bLocum GP category wasnot available for the pilot practices (n = 7) so locums were included within the 'GP' category in the pilot practices.c Unavailable due to low number of participating paramedics.OR = odds ratio.SD = standard deviation. a