Family Practice Advance Access published online on September 1, 2008
Family Practice, doi:10.1093/fampra/cmn054
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Evaluating the impact of a national pilot screening programme for type 2 diabetes in deprived areas of England
a School of Health and Related Research, University of Sheffield, Sheffield
b Public Health Sciences, University of Edinburgh, Edinburgh
c Information Services Division, NHS National Services Scotland, Edinburgh, UK
Correspondence to Elizabeth Goyder, School of Health and Related Research, University of Sheffield, 30 Regent Street, Sheffield S1 4DA, UK; Email: e.goyder{at}sheffield.ac.uk
Received 15 February 2008; Revised 2 July 2008; Accepted 31 July 2008.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Background. A pilot programme designed to systematically screen for type 2 diabetes was introduced in 24 general practices in England selected for their high levels of socio-economic deprivation and multi-ethnic populations.
Objective. To evaluate the impact of screening on the prevalence of type 2 diabetes.
Methods. A prospective audit of screening activity in pilot practices and comparison of the change in prevalence of diabetes in pilot and comparison practices were conducted.
Results. Of 41 400 individuals invited for screening from a population of 165 828 in pilot practices, 25 356 (61%) were screened. Three hundred and fifty-eight (0.22%) new cases of diabetes were detected among those screened. Only 69% of those with a positive screening test had diagnostic testing recorded and only 19% had a record of an oral glucose tolerance test. The absolute increase in the prevalence of diagnosed diabetes was 0.53% in pilot practices and 0.42% in comparison practices.
Conclusions. The ‘real world’ nature of the programme and dependence on routine data collection systems makes results more difficult to interpret but also enabled problems with implementation, not evident from previous research, to be identified. It is likely that the low diagnostic yield was largely due to a high level of ad hoc screening activity outside the pilot protocol and inadequate access to diagnostic testing after a positive screening test. In particular, implementation of screening for diabetes in primary care should not be undertaken without robust assessment of the resources required for diagnostic testing and follow-up and adequate clinical audit.
Keywords. Deprivation, primary care, screening, type 2 diabetes.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Screening for type 2 diabetes in high-risk populations is widely recommended in the UK1,2 and US.3 Epidemiological studies in the UK, Australia and the US have previously suggested that between 30% and 50% of all diabetes is undiagnosed.4–6 Despite the absence of conclusive evidence from randomized trials7,8, screening may be cost-effective because earlier diagnosis and appropriate treatment may reduce the risk of complications.9
As part of a national evaluation of a pilot diabetes screening project, we examined the impact of introducing systematic screening on new diagnoses of diabetes in everyday practice rather than in a research setting.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Over 2 years, between 2003 and 2005, the UK National Screening Committee established diabetes screening pilots in 24 general practices in eight areas of England who were responsible for providing primary care for a total population of 165 828 people. The areas selected had socio-economically deprived and ethnically diverse urban populations that were expected to have a high prevalence of both diagnosed and undiagnosed diabetes.4 They were all based in Teaching Primary Care Trusts (PCTs) which agreed to take part in the pilot and which included approximately 20–30 general practices in each PCT. In each PCT, from the practices that volunteered to take part in the study, three general practices were randomly selected as pilots and three as comparison practices.
The pilot practices were asked to use their clinical information systems to identify patients aged over 40 years with a body mass index (BMI) greater than 25 kg/m2 and offer measurement of a random capillary blood glucose (RCBG) using a glucometer. Choice of glucometer and methods of quality assurance varied depending on local practice. A result of 
6 mmol/l was suggested to indicate eligibility for diagnostic testing. The diagnosis of diabetes was recommended to be based on an oral glucose tolerance test (OGTT) [with cut-points of fasting blood glucose (FBG) >7.0 mmol/l or 2 hours >11.1 mmol/l] but a single FBG above 7.0 mmol/l was deemed an acceptable alternative if an OGTT was not performed.10
Each area had a facilitator who offered practical support to the pilot practices and the practices were reimbursed for the costs of additional screening activity. Information was collected from pilot practices on numbers of people who were eligible and who were invited for screening, numbers screened and numbers newly diagnosed within the screening programme.
Information on pilot activity was aggregated across all pilot practices to assess the overall activity. Prevalence data (estimated from total population and total number of cases of diagnosed diabetes) at baseline and 2 years later were collected from both pilot and comparison practices to examine how many additional diagnoses of diabetes were made outside the screening programme. Data on change in the prevalence of diabetes over 2 years from the start of the pilots were aggregated across areas and across all practices to assess the relative impact of the pilot activity on prevalence.
Individual level data on screened patients (included age, sex, ethnicity and clinical information) were collected using either a data collection form developed for the pilot programme (available in paper and electronic Microsoft Word formats) or a template developed locally using the practice clinical information system. We examined the characteristics of those with positive screening and diagnostic tests, and those not receiving diagnostic tests, despite a positive screening test result. Chi-square tests were used to assess the statistical significance of differences between groups (using a test for trend, as appropriate, for ordered categorical variables).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Prevalence data at baseline and 2 years later were available from 24 pilot and 17 comparison practices. Six practices dropped out and one practice was closed before data collection was complete. Where pilot practices dropped out at an early stage (four practices), they were replaced with comparison practices which were re-classified as pilot practices, reducing the number of comparison practices. Individual level data were available from 21 pilot practices. The other three practices were unable to identify and extract pilot screening activity data (as distinct from all glucose measurement activity data) from clinical systems for the evaluation. The overall losses of data are significant and are summarized in Figure 1 which shows the size of the data sets on which the results reported below are based and the reasons for data losses.
|
Reported screening activity
Practices reported inviting 41 418 patients for screening (25% of a total population of 165 828) and screening a total of 25 356 people (61% of those invited). Of those screened, 8367 (33%) had a positive screening test. A total of 358 newly diagnosed cases were reported (4.3% of those with a positive screening test and 1.4% of those screened). This overall activity reported was not consistent with the individual patient data provided, suggesting practices had difficulty identifying the records of screened individuals and identifying the number of new diagnoses that could be ascribed to screening. Some practices had widened the original inclusion criteria to include patients aged over 30 years or to include all patients over 40 years of age regardless of BMI. Most practices had limited the screening criteria by excluding patients with coronary heart disease (CHD) because they had already been screened as part of an annual CHD review. Some practices also excluded any patient who had a normal blood glucose recorded in their general practice records (and in one area, hospital laboratory records) in the previous 2 years.
Characteristics of the screened population
Analyses of the data for the screened population are based on data from 12 145 patients for whom the results of a RCBG screening test was recorded (48% of the total number reported screened). The characteristics of the screened population and those found to be positive on screening or diagnostic testing are shown in Table 1. It cannot be assumed that this subgroup is representative of all those reported by practices as being screened and it is possible that data was better recorded for those with a positive screening test, increasing the prevalence of positive screening and diagnostic testing in this subgroup.
|
Increasing age, BMI and non-white ethnicity were all associated with screening-detected diabetes in this population. Out of 3789 screened patients with a RCBG over 6 mmol/l, 1191 (31%) did not have the results of diagnostic testing recorded on their screening record, even after updating records from practice information systems. These individuals are assumed to represent a combination of patients who did not undergo diagnostic testing (for whatever reason) and those whose result was not recorded on the practice system, or whose data were not coded in a way that meant it could be subsequently identified as the result of diagnostic testing.
Analysis based on the subgroup of 2598 individuals with a positive screening test recorded shows that younger age, higher BMI and higher RCBG were associated with having diagnostic testing (chi-square test results in Table 1). Only 48% of those over 80 years had a record of diagnostic testing, and only 61% of those of normal weight had diagnostic testing, despite a positive screening test. Those with a RCBG result between 6 and 7 mmol/l were also less likely to have diagnostic testing than those with a RCBG result of greater than 7 mmol/l (66% versus 71%, P < 0.001) (Table 1).
Of those with a diagnostic test result recorded, 774 (30%) had undergone an OGTT (as required by World Health Organization diagnostic criteria for diabetes), the remaining 1824 (70%) having had only a FBG measured, of whom only 383 (15%) had a second fasting glucose recorded. Non-white patients were less likely to have an OGTT performed than white patients (P < 0.001, Table 1).
Overall impact on diabetes prevalence
The overall prevalence results are shown in Table 2 for six areas that had the same pilot and comparison practices at the beginning and end of the pilot period. In the other two areas, comparison practices replaced pilot practices that dropped out after selection. Overall, between 2003 (pre-pilot) and 2005 (post-pilot), the prevalence of known diabetes across 17 practices increased from 3.5% to 4.1%, an increase in prevalence of 0.54% (range –0.37% to 2.32%) in the pilot practices. In comparison practices, prevalence increased from 3.3% to 3.7%, an increase of 0.42% (range –0.16% to 1.60%).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Overall, 1 in 70 people (1.4%) screened within the pilot programme was given a new diagnosis of diabetes. Restricting screening to the original project criteria (over 40 years with a BMI over 25 m/kg2) would have increased the yield to 3.75% (based on those with complete data), giving a ‘number needed to screen’ of 27. The number needed to screen to diagnose one additional case in the pilot programme was increased by a number of factors including the widening of inclusion criteria, the lack of follow-up for almost a third of those with a positive screening test and the low level of glucose tolerance testing. The sensitivity of capillary blood glucose as a screening test depends on the choice of cut-off.11 The observed increase in prevalence in comparison practices suggests that many of the additional cases diagnosed by the pilot screening would still have been diagnosed in the absence of the programme. Moreover, the diagnosis was based on one result (either from fasting glucose or OGTT) above the cut-point for diagnosis of diabetes, rather than two results from different days as recommended for a clinical diagnosis of diabetes. Where a second result was available, it was only diagnostic in 71% of cases, suggesting the possibility of over-diagnosis of diabetes from using a single measurement.
Both the low additional yield from systematic screening and the relatively large increase in prevalence in both pilot and comparison practices can be explained by an increase in both ad hoc and systematic screening as part of normal clinical practice. It is likely that the overall increase in recorded prevalence of diabetes was also influenced by a new contract for primary care services introduced in 2004. This new contract linked payment for management of chronic diseases (including diabetes) to recording and reporting of process measures for those on disease registers. There was significant variation in the impact of screening between practices in each area and between areas which was unrelated to the initial prevalence, suggesting that it would be difficult to predict the overall impact of introducing systematic screening at practice level.
Strengths and limitations of this study
This was the first attempt to introduce systematic population screening for diabetes in deprived communities across England and provided an opportunity to explore the potential impact of screening in primary care. Both the strengths and major weaknesses of this analysis are closely related to the ‘real world’ nature of the pilots as discussed below.
Advantages and disadvantages of real world evaluation outside research settings
This project brought into sharp focus some of the major problems with attempting to evaluate a new programme in the real world in which as evaluators we had virtually no control over the design or delivery of the intervention and when data collection relied on routine clinical practice. Undoubtedly, if the pilot had been established as a trial, it would have been possible to reduce the bias caused by the large variations between practices in implementation and high levels of missing data on screening activity. However, some of the useful lessons were learnt precisely because this was a real world pilot programme rather than a research project. Even in a pragmatic trial with a systematic process evaluation,12 some problems with implementation will only become apparent once an intervention is introduced without the level of control by investigators implicit in a randomized trial. Issues that tend not to be problematic in a research setting because of the prescriptive nature of a research protocol may be much more problematic in normal clinical practice. Four major practical difficulties with implementation of screening emerged which had not been previously identified, probably largely because they do not arise to the same extent in a research context: inconsistency in implementing the screening protocol; lack of quality control; lack of adequate diagnostic testing after a positive screening test; and lack of systems for routine data collection on screening.
Issues of bias and external validity
Caution must be applied in generalizing from the findings as the pilot practices were all volunteer practices in relatively deprived urban areas. Comparison and pilot practices were all volunteers and were matched on area. However, they were not formally matched for size or population characteristics (age, ethnicity or deprivation) and so may not have been completely comparable. In addition, the practices which dropped out may have had different attitudes to screening and might therefore have had lower prevalence than those providing data.
The evaluation was only commissioned once the pilot was in progress, so that it was not possible for the evaluation team to directly influence the data collection processes. A major obstacle to evaluating the screening pilots was the lack of a common data collection system and consequently incomplete data collection in some practices. Only half the practices chose to use the pilot data collection form. Locally developed templates appeared initially to be a practical solution but it proved difficult to extract information from them because information systems did not have either consistently used fields for screening test results or the capacity to link the screening record to the results of subsequent diagnostic testing. Problems may have been compounded when staff who were recruited to assist with the screening pilots were not familiar with the information systems. It was not always straightforward to identify screened individuals if there was no specific or unique way of coding, or flagging, these cases or distinguishing from those who had a blood glucose result for a different reason.
Comparison with existing literature
Previous reports of general practice-based systematic testing for diabetes have also reported low yields.13–16 Screening offered to patients aged over 45 years in a single UK practice had an uptake of 35% and a yield of 15 new cases, 1.7% of those screened and representing an absolute increase in prevalence of 0.3%.13 Interventions that have targeted high-risk groups, based on age and BMI or other recorded risk factors,17,18 have achieved higher diagnosis rates. For example, in people aged over 50 years, with BMI over 27 kg/m2, screening increased prevalence to 13.6%, from a baseline of 11%.17 A study of primary care screening in France with a low yield of cases (0.67%) revealed that 75% of patients with risk factors had a FBG measured in the previous year, even in the absence of systematic screening.15 Low yields may also be explained by lower uptake of testing by high-risk individuals.13,16
The relatively small excess in prevalence of newly diagnosed diabetes for the South Asian relative to the white population in this study compared to other studies may be due to the practices having previously actively screened more of their South Asian patients as non-white ethnicity is widely recognized as a significant risk factor for diabetes. In addition, the largely White populations may have had a relatively high prevalence of previously undetected diabetes as they included the most socio-economically deprived areas.
There is very little evidence on which to assess current provision and impact of screening for different populations, as although there is considerable screening activity and systematic screening is often undertaken as part of CHD and hypertension management, there is no routine audit of screening activity or outcomes.
Implications for clinical practice
Despite the limitations, these results strongly suggest that the impact of additional systematic screening in high-risk populations is likely to be modest. This may be largely explained by current high levels of both ad hoc and systematic screening of individuals with cardiovascular disease and hypertension. The number of new diagnoses of diabetes will be further reduced by low uptake, failure to follow-up positive screening tests and lack of access to OGTTs. Despite the low diagnosis rate, practice staff often felt the pilots were very successful in providing opportunities to raise awareness of diabetes and advise patients on weight loss, diet and other lifestyle issues.
|
Conclusions |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Diabetes screening and cardiovascular risk reduction programmes in general have proved feasible and popular, despite limited evidence of effectiveness from randomized trials to date and relatively low rates of new diagnoses in primary care settings. It is likely that primary care will in future have an increasing responsibility for primary prevention and early detection of diabetes and cardiovascular risk and future programmes will be more effective if lessons are learnt from the experience of pilot programmes.
This screening pilot identified a number of practical obstacles to effective systematic screening which contributed to a low yield of new cases of diabetes. Firstly, the efficiency of screening would be increased by more rigorous adherence to eligibility criteria and complete follow-up of those who screen positive. Before programmes are implemented, there should be adequate organizational resources for both screening and for adequate diagnostic testing and follow-up in the screened population. Secondly, any screening or risk reduction programme needs to be introduced with adequate systems to ensure the collection of complete and accurate data and an agreed protocol for systematic audit of both process and outcomes.
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
Funding: The pilots and evaluation were funded by the Department of Health. The authors are entirely independent of the funding body and the funding body was not involved in the decision to submit this paper for publication. The views expressed are those of the authors and not necessarily those of the Department of Health or the UK National Screening Committee.
Ethical approval: Ethical approval was obtained from the Trent Multi-Centre Research Ethics Committee (04/MRE04/52).
Conflicts of interest: None.
|
Acknowledgments |
|---|
The authors would like to thank the pilot practice staff, the PCT facilitators and the UK National Screening Committee team in Oxford and David Graham, who all worked extremely hard to collect the data that made the evaluation possible.
|
Notes |
|---|
Goyder E, Wild S, Fischbacher C, Carlisle J and Peters J. Evaluating the impact of a national pilot screening programme for type 2 diabetes in deprived areas of England. Family Practice 2008; Pages 1–6 of 6.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsDeclarationReferences |
|---|
1 Department of Health. National Service Framework for Diabetes (2001) London: Stationery Office.
2 Diabetes UK. Position Statement. Early Identification of People with Type 2 Diabetes (2006) http://www.diabetes.org.uk/documents/professionals/earlyid_type2_ps.doc (last accessed 18 August 2008).
3 ADA. Screening for type 2 diabetes. Diabetes Care (2004) 27:s11–s14.
4 Riste L, Khan F, Cruickshank K. High prevalence of type 2 diabetes in all ethnic groups, including Europeans, in a British inner city: relative poverty, history, inactivity, or 21st century Europe? Diabetes Care (2001) 24:1377–1383.
5 Colagiuri S, Hussain Z, Zimmet P, Cameron A, Shaw J, AusDiab. Screening for type 2 diabetes and impaired glucose metabolism: the Australian experience. Diabetes Care (2004) 27:367–371.
6 Cowie CC, Rust KF, Byrd-Holt DD, Eberhardt MS, Flegal KM, Engelgau MM. Prevalence of diabetes and impaired fasting glucose in adults in the U.S. population: National Health and Nutrition Examination Survey 1999–2002. Diabetes Care (2006) 29:1263–1268.
7 Wareham NJ, Griffin SJ. Should we screen for type 2 diabetes? Evaluation against National Screening Committee criteria. BMJ (2001) 322:986–988.
8 HSTAT. Screening for Type 2 diabetes. Chapter 19. Guide to Clinical Preventive Services. 3rd edn. Evidence Syntheses, formerly Systematic Evidence Reviews. 2000. http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat3.chapter.3895 (accessed on 25 April 2007).
9 Waugh N, Scotland G, McNamee P, et al. Screening for type 2 diabetes: literature review and economic modelling. Health Technol Assess (2007) 11:1–125.[Web of Science][Medline]
10 Gabir MM, Hanson RL, Dabelea D, et al. The 1997 American Diabetes Association and 1999 World Health Organisation criteria for hyperglycaemia in the diagnosis and prediction of diabetes. Diabetes Care (2000) 23:1108–1112.
11 Rolka DB, Narayan KM, Thompson TJ, et al. Performance of recommended screening tests for undiagnosed diabetes and dysglycaemia. Diabetes Care (2002) 24:1899–1903.[Web of Science]
12 Oakley A, Strange V, Bonell C, Allen E, Stephenson J. Process evaluation in randomised controlled trials of complex interventions. BMJ (2006) 332:413–416.
13 Janssen PGH, Gorter KJ, Stolk RP, Rutten GEHM. Low yield of population-based screening for Type 2 diabetes in the Netherlands: the ADDITION Netherlands study. Fam Pract (2007) 24:555–561.
14 Lawrence JM, Bennett P, Young A, Robinson AM. Screening for diabetes in general practice: cross sectional population study. BMJ (2001) 323:548–551.
15 Cogneau J, Balkau B, Weill A, Liard F, Simon D. Assessment of diabetes screening by general practitioners in France: the EPIDIA Study. Diabet Med (2006) 23:803–807.[CrossRef][Web of Science][Medline]
16 Christenson JO, Sandbaek A, Lauritzen T, Borch-Johnsen K. Population-based stepwise screening for unrecognised Type 2 diabetes is ineffective in general practice despite reliable algorithms. Diabetologia (2000) 47:1566–1573.[CrossRef]
17 Greaves CJ, Stead JW, Hattersley AT, Ewings P, Brown P, Evans PH. A simple pragmatic system for detecting new cases of type 2 diabetes and impaired fasting glycaemia in primary care. Fam Pract (2004) 21:57–62.
18 Klein Woolthuis EP, de Grauw WJC, van Gerwen WHEM, et al. Identifying people at risk for undiagnosed type 2 diabetes using the GP's electronic medical record. Fam Pract (2007) 24:230–236.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||