Family Practice Vol. 21, No. 1, 57-62
© Oxford University Press 2004, all rights reserved.
Article |
A simple pragmatic system for detecting new cases of type 2 diabetes and impaired fasting glycaemia in primary care
Mid-Devon Primary Care Research Group, Wyndham House Surgery, Wyndham Road, Silverton, Devon EX5 4HZ, a Peninsula Medical School, Barrack Road, Exeter EX2 5DW and b Somerset RDSU, Taunton and Somerset Hospital, Taunton TA1 5DA, UK
E-mail: colin.greaves{at}pms.ac.uk
Received 9 December 2002; Revised 15 July 2003; Accepted 8 September 2003.
Greaves CJ, Stead JW, Hattersley AT, Ewings P, Brown P and Evans PH. A simple pragmatic system for detecting new cases of type 2 diabetes and impaired fasting glycaemia in primary care. Family Practice 2004; 21: 57–62.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background. Although whole population screening for type 2 diabetes is not currently considered to be justified, targeted screening within higher risk groups may be more cost-effective, and more pragmatic.
Objectives. Our aim was to investigate the feasibility and performance of a pragmatic system for identifying patients with type 2 diabetes and impaired fasting glycaemia (IFG).
Methods. A clustered observational survey of the prevalence of diabetes and IFG was carried out in randomly selected patients from four at-risk groups. Patients were identified by computerized searching of practice databases for age and body mass index (BMI) risk criteria. Sixteen practices in South West England screened 1287 Caucasian patients from four groups with progressive levels of theoretical risk (age >70 and BMI 
33, age >65 and BMI 31, age >60 and BMI 29, and age >50 and BMI 27). Fasting plasma glucose was measured and repeated if abnormal to determine the prevalence of new cases in each group. BMI and age data were validated against measures taken at the clinic.
Results. The response rate was 60.6% and the prevalence of new cases of type 2 diabetes in each group was 4.7% [95% confidence interval (CI) 2.8–7.7], 5.7% (95% CI 4.0–8.2), 3.8% (95% CI 2.4–6.0) and 2.6% (95%CI 1.4–4.7), respectively. An additional 5.2–8.4% had IFG.
Conclusions. Targeted screening by searching existing GP records for age and BMI criteria is feasible for use in general practice in the UK. Screening of patients with a BMI of 27 and aged >50 by fasting glucose identified a substantial prevalence of undetected type 2 diabetes and IFG. The relative costs and benefits as well as the pragmatic advantages of different systems need further evaluation.
Keywords. Body mass index, diabetes, feasibility studies, primary care, screening.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The issue of screening for type 2 diabetes and for pre-diabetic hyperglycaemia is receiving considerable attention.1–4 This has been prompted by the recognition that a large proportion of patients with diabetes already have complications at diagnosis5 and that intensive blood glucose control and lifestyle interventions cannot only improve prognosis for those with diabetes,6 but can also prevent or delay the onset of diabetes for those with impaired fasting glycaemia or impaired glucose tolerance.7,8
Although whole population screening is not considered justified, targeted screening within higher risk groups may be more cost-effective, and more pragmatic.1,3,9 Such targeted screening may use a purely opportunistic approach (identification of risk factors during health care visits). However, it is also possible to identify systematically at-risk patients for subsequent screening. At-risk patients can be identified by various methods including patient questionnaires,10,11 identifying groups using criteria based on epidemiological risk [e.g. age >50, body mass index (BMI) >27),2,3 or the use of more complex risk scores.12
Despite the potential advantages to patients and health care systems, there are still a number of practical difficulties with systematically targeting and testing patients in a general practice setting. First, the tests used for screening may be cumbersome. The emerging gold-standard test for diabetes screening is the oral glucose tolerance (OGT) test,13,14 but the procedure requires two appointments with timing constraints (being 2 h apart), which may be inconvenient for both staff and patients. Secondly, the methods proposed for targeted screening are labour intensive. Questionnaire methods involve mailing large numbers of patients and collating the resulting data, and complex risk factor formulae may be inconvenient for GPs and nurses to apply in consultations.
Alternatively, a more pragmatic screening programme could be designed to exploit the current strengths of UK primary care. One of these strengths is the increasing complexity and comprehensiveness of the patient's computerized medical record.
This study was therefore designed to investigate the idea of computerized searching of routinely collected data as the starting point for a targeted screening programme. As certain risk factors such as family history are not readily recorded, this study used age and BMI, which are more widely available. Both these factors are strongly associated with an increased risk of type 2 diabetes.15,16 Age and BMI were therefore selected as the basis of a simple pragmatic system for computerized identification of high risk patients that could be easily generalized to the majority of UK practices.
To establish the potential feasibility of this system for identifying hyperglycaemic illness, the study assessed the prevalence of previously undetected diabetes and impaired fasting glycaemia (IFG) in four groups of patients with selected age and BMI criteria.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Patients were recruited from 16 practices in Somerset, and North and East Devon through the local primary care research network (SaNDNet). Sixteen practices (eight from each health authority) were randomly selected from 42 volunteers, stratified as to whether they were training practices and whether they were rural (receiving rurality payments). Practices with <3500 patients were excluded, as was one practice with low BMI coverage (<60% of patients aged >50 with BMI recorded in the last 10 years). The practices were representative of the two Health Authority areas in terms of rurality, training status and number of patients per GP. The range of computer systems in use included EMIS (nine), Meditel (two), VAMP (two), HMC (one), Torex6000 (one) and the Exeter System (one). The population prevalence of diagnosed type 2 diabetes for the 16 practices (population 149 379) prior to the study was 2.36%.
Each practice was asked to sample 100 patients for testing, 25 from each of four groups. The groups were specified by stepped age and BMI criteria (Table 1), representing increasing levels of theoretical risk of type 2 diabetes. Lists of patients meeting the inclusion criteria for each group were generated by computer searching, and numbered. BMI was defined as having ever been recorded above each threshold (latest BMI recorded cannot be searched for in many practice database systems). Selection of patients for invitation to screening was determined by applying random numbers generated by the lead researcher. Due to the overlapping criteria, a patient could be selected for more than one group. Patients with known diabetes (type 1 or type 2) were excluded, as were non-Caucasians and patients with learning difficulties.
|
Practice nurses were trained to co-ordinate the project and run the screening clinics. Selected patients were written to with a provisional clinic appointment, and followed up by telephone to confirm attendance and the fasting procedure. At the clinic, weight (kg) and height (m) were measured using recently calibrated equipment. Age was self-reported on a brief questionnaire. A fasting venous blood sample was taken and analysed for blood plasma glucose by local NHS laboratories using NHS-approved testing procedures (based on glucose oxidase and hexokinase reactions) and equipment (e.g. Vitros GLU slide, Beckman LX20).
Patients with elevated fasting blood plasma glucose (6.1 mmol/l) were invited for repeat testing. Patients with a fasting blood glucose of 7.0 mmol/l on both tests were diagnosed as having diabetes, and those with a FPG of 6.1–6.9 mmol/l on both tests were identified as having IFG. Patients with one reading 7.0 mmol/l and one of 6.1–6.9 mmol/l were classed as having IFG. All patients were informed of their results, and those with abnormal results were referred to their GP.
Statistical considerations
Based on pilot data on the known prevalence of type 2 diabetes in the same age/BMI groups, we assumed a detection rate of 10% and an intra-practice correlation of 0.01. A sample of 100 patients from each of 16 practices was calculated to give confidence intervals (CIs) around the estimated percentage of ±2%. Prevalence figures are reported with 95% CIs taking into account the clustered sampling design as appropriate.17 The figures presented as ‘numbers needed to screen’ and ‘numbers needed to test’ do not take into account the likely efficacy of treatment, as may be the case in other studies.18
Ethical approval was provided by Exeter & North Devon, Bath, East Somerset and West Somerset Local Research Ethics Committees.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Recruitment
In total, 1287 patients were recruited, of which 508 (39.5%) were male, giving 1644 data points across the four groups (see Fig. 1 and Table 1). As the grouping criteria are nested within each other, a patient could be selected for more than one group. Of the 1287 participants, 251 were selected into two groups, 47 were selected into three groups and four were selected into all four groups.
|
Uptake of screening
The rate of response to the invitation for screening was 60.6% (95% CI 55.7–65.6), based on data returned by 15 practices. To check for response biases, the age and gender of the most inclusive group (age >50, BMI
27) were compared with a random sample from the same practices (n = 1700) of those who were eligible for inclusion in this group, but not invited for screening. No significant sampling biases due to either age or gender were found.
Coverage and accuracy of practice data
BMI data in the over-50 population were available for 76.8% (95% CI 71.7–81.9) of patients. The BMI on record was compared with the current weight and height measured at the clinics. This indicated that 328 (20.0%) of the sample were misclassified due to either out of date or inaccurate BMI data in the practice record. However, when the data were reanalysed excluding those misclassified, this did not substantially affect the results.
Practice data on age were available for 100% of patients. In 27 cases, self-reported age differed from that on the practice computer by >1 year. This could potentially have led to the misclassification of 11 patients (0.7%).
Outcome of screening
The numbers of new cases of type 2 diabetes and IFG detected are shown in Table 2. There were no significant differences between men and women in the detected prevalences of IFG (chi-squared >0.6) or type 2 diabetes (chi-squared >0.3).
|
Of those attending the screening clinic, the ‘number needed to test’ (Table 3) to detect hyperglycaemia (IFG or type 2 diabetes) was only 7.7 for group 1 (>70, BMI
33). However, this group comprised just 0.6% of the practice population. In group 4, although the number needed to test was greater (12.8), this group comprised 11.1% of the practice population. The trade-off between the overall number of cases detected and loss of screening efficiency is therefore relatively modest, and seems to favour the use of wider screening criteria (e.g. >50 and BMI 27).
|
Incorporating subjects excluded due to non-response and unrecorded BMI data, the ‘numbers needed to screen’ to detect a case of hyperglycaemia were 16.5, 15.3, 17.8 and 27.5 in groups 1–4. This is the number of people in the population meeting the inclusion criteria that one would need to apply the system to in order to detect one case.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Feasibility and effectiveness
A simple and pragmatic screening programme using fasting blood glucose estimations and BMI/age data to stratify patients seems feasible in general practice. The screening of patients with a BMI >27 aged >50 by fasting glucose identified a substantial number of new cases of type 2 diabetes and IFG (Table 2). Further stratification by age and BMI did not result in substantial increases in detection rates.
The ‘number needed to screen’ to detect a case of either diabetes or IFG (between 15 and 28) compares well with other screening initiatives. The equivalent numbers are 35 for diagnosing men at high absolute risk of cardiovascular disease,19 and 146 for identifying a case of breast cancer by mammography screening of women aged 50–64.20
In terms of resource use, it was possible to run screening clinics after only brief nurse-training. The computerized identification of patients for screening took <1 h per practice. The clinic time per patient was 
10–15 min. This compares with a minimum of 2 h and two appointments for OGT testing.
The pragmatic advantages of this system therefore seem highly appealing. However, it is worth noting that the different types of hyperglycaemia identified by the fasting plasma and glucose tolerance tests only overlap by 20–25%.21 Hence, the added benefit of adding glucose tolerance testing to the protocol seems worth further investigation.
The uptake of 61% also compares favourably with other screening programmes currently running in primary care, including mammography (63%) and cervical cytology (71%),22 and is considerably higher than the 35% reported by Lawrence et al.23 in a single practice study of screening by fasting plasma glucose.23 The response rate for repeat testing was 100%, which is also higher than that achieved by Lawrence et al. (75%).23 These differences may reflect our use of telephone calls to confirm appointments, and re-contacting any patients not attending for follow-up testing.
Comparison with other potential screening systems
Alternative systems for targeting diabetes screening, including risk questionnaires, and the calculation of risk scores may be considerably more labour intensive, and OGT testing is likely to be less acceptable to patients than the fasting glucose test. However, these options may provide different population coverage, more efficient targeting of at-risk patients or more sensitive identification of cases. Alternatively, the detection rate of the computerized searching system described here could be improved by efforts to improve BMI recording, by the development of software patches to allow searching for ‘latest BMI’ and potentially by the use of glucose tolerance testing.
The trade-offs between pragmatic feasibility (including uptake), sensitivity and cost-effectiveness of the different possible systems clearly require further investigation. Furthermore, as the cost of treating new cases is considerable, it is important to establish clear evidence for the cost-effectiveness of early treatment of diabetes and hyperglycaemia, and of different systems for targeting detection.
Limitations
The 77% BMI recording rate may cause some sampling bias, as patients with weight-related problems, heart disease and hypertension are more likely to have their weight monitored. The population prevalence of diabetes in the age/BMI groups defined may therefore be slightly lower than we found here. Inaccuracies in BMI data and the inability of most GP computer systems to retrieve the patients' most recent BMI (rather than the highest recorded) also detract from the performance of the targeting system.
Despite these limitations, the system as used in practice (where imperfect data are a reality) did detect a substantial number of cases, and this finding is likely to be replicable across a wide range of general practices. The main problem with having missing and inaccurate BMI data is that this leaves a number of people outside the screening system who would otherwise be selected, and this problem would need to be addressed by any practices adopting the system (e.g. by improving BMI recording).
It should also be noted that using fasting plasma glucose testing might underestimate the true prevalence of type 2 diabetes. Based on the known performance limitations of the fasting glucose test, 30–50% of those diagnosed with IFG may have been diagnosed with diabetes if a follow-up OGT test had been employed,13 26 and this practice is recommended for any future implementation of this system.
Finally, the sample is only representative of practices in South-West England. Detection rates will be higher in populations with larger Asian and Afro-Caribbean populations27 (SW England is 98% Caucasian), and may also vary with lifestyle and socio-economic factors between different settings.
Future directions
Further studies are clearly needed to account for ethnic risk factors. The National Screening Committee have recommended9 that researchers need to establish which subgroups of the population it would be cost-effective to offer screening to; the feasibility and practicality of identifying people at risk; response rates to an invitation for screening; and the relative benefits including cost-effectiveness, health benefit and potential negative effects of being tested and diagnosed. They also recommend investigation of the utility of different screening tests, and of optimum screening intervals.
The current work addresses some of these issues for a single practically oriented system, but much more work is needed to compare and contrast different methods. The feasibility, effectiveness and cost-effectiveness of different targeted screening methods for diabetes could be assessed further using randomized trials comparing different systems. The inclusion of other risk criteria such as family history or blood pressure might improve the diagnostic efficiency of screening, but the effects on feasibility would need to be considered.
Ultimately, however, considering diabetes screening strategies in isolation may be unwise. The idea of stratifying patient lists by age and BMI could be integrated with other current initiatives in primary care, particularly the primary and secondary prevention of coronary heart disease and screening for hypertension. Integrating the simple screening system presented here with such initiatives would seem a logical and efficient use of scarce primary care resources.
Conclusions
Existing practice data on age and BMI can be used to stratify the population to identify high-risk groups. Targeted screening within these groups identified a substantial number of new cases of type 2 diabetes and IFG. The relatively flat distribution of cases by age/BMI favoured wider screening criteria (i.e. age >50, BMI 27). The trade-offs between pragmatic feasibility, sensitivity and cost-effectiveness of different systems require further investigation. However, alternative systems need to demonstrate substantial advantages above and beyond the simple pragmatic screening programme described here to justify their preferred use in general practice.
|
Acknowledgments |
|---|
We are grateful to Trish Brown for her help on this project as research nurse, and to the practice nurses, practice managers and partners of the practices who took part in the study (Beckington Family Practice, Cannington Health Centre, Castle Place Surgery, Chiddenbrook Surgery, Fremington Medical Centre, Frome Medical Practice, Saffron Surgery, Glastonbury Health Centre, Honiton Group Practice, Queens Medical Centre, St Leonard's Medical Practice, St Thomas Health Centre, Summervale Medical Centre, Taunton Road Medical Centre, Westlake Surgery and Wyndham House Surgery). We are also grateful to Joy Choules and to Sylvia Smith for their secretarial and administrative support. The authors gratefully acknowledge the funding for this project from the NHS South West R&D Directorate which was awarded to SaNDNet (Somerset and North and East Devon Primary Care Research Network).
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1 Wareham NJ, Griffin SJ. Should we screen for type 2 diabetes? Evaluation against National Screening Committee criteria. Br Med J 2001; 322: 986–988.
2 Diabetes UK. Early Identification of People with Type 2 Diabetes. Position Statement. London: Diabetes UK; 2001.
3 American Diabetes Association and National Institutes of Diabetes Digestive and Kidney Diseases. The prevention or delay of type 2 diabetes. Diabetes Care 2002; 25: 1–8.
4 Narayan KMV. Targeting people with pre-diabetes. Br Med J 2002; 325: 403–404.
5 Harris MI, Klein R, Wellborn TA, Knuiman MW. Onset of NIDDM occurs at least 4–7 years before clinical diagnosis. Diabetes Care 1992; 15: 815–819.[Abstract]
6 UK Prospective Diabetes Study Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet 1998; 352: 837–853.[CrossRef][ISI][Medline]
7 Tuomilehto J, Lindstrom J, Eriksson JG et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 2001; 344: 1343–1350.
8 Knowler WC, Barrett-Connor E, Fowler SE et al. for the Diabetes Prevention Program Research Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002; 346: 393–403.
9 Department of Health. National Service Framework for Diabetes: Standards. London: Department of Health; 2001.
10 Ruige JB, De Neeling JN, Kostense PJ, Bouter LM, Heine RJ. Performance of a NIDDM screening questionnaire based on symptoms and risk factors. Diabetes Care 1997; 20: 491–496.[Abstract]
11 American Diabetes Association. Clinical practice recommendations 2000: screening for type 2 diabetes. Diabetes Care 2000; 23: S20–S23.
12 Griffin SJ, Little PS, Hales CN, Kinmonth AL, Wareham NJ. Diabetes risk score: towards earlier detection of type 2 diabetes in general practice. Diabetes Metab Res Rev 2000; 16: 164–171.[CrossRef][ISI][Medline]
13 Modan M, Harris MI. Fasting plasma glucose in screening for NIDDM in the US and Israel. Diabetes Care 1994; 17: 436–439.[Abstract]
14 Gerstein HC. Fasting versus postload glucose levels: why the controversy? Diabetes Care 2001; 24: 1855–1857.
15 Harris MI. Epidemiologic studies on the pathogenesis of non-insulin-dependent diabetes mellitus (NIDDM). Clin Invest Med 1995; 18: 231–239.[ISI][Medline]
16 Jarrett RJ. Epidemiology and public health aspects of non-insulin-dependent diabetes mellitus. Epidemiol Rev 1989; 11: 151–171.
17 Kish L. Survey Sampling. London: Wiley; 1965.
18 Law MR. The number needed to screen—an adaptation of the number needed to treat. J Med Screen 2001; 8: 114–115.
19 Scottish Intercollegiate Guideline Network. Lipids and the Primary Prevention of Coronary Heart Disease. SIGN Guideline 40. Edinburgh: SIGN; 1999.
20 Mant J. Is this test effective? In Dawes M, Davies P, Gray A, Mant J, Seers K, Snowball R (eds). Evidence-based Practice: A Primer for Health Care Professionals. An Evaluation of the Prevalent Round of the Breast Screening. Edinburgh: Churchill Livingstone; 1999: 133–157.
21 De Vegt F, Dekker JM, Jager A et al. Relation of impaired fasting and postload glucose with incident type 2 diabetes in a Dutch population: the HOORN study. J Am Med Assoc 2001; 285: 2109–2113.
22 Richards SH, Bankhead C, Peters TJ et al. Cluster randomised controlled trial comparing the effectiveness and cost-effectiveness of two primary care interventions aimed at improving attendance for breast screening. J Med Screen 2001; 8: 91–98.
23 Lawrence JM, Bennett P, Young A, Robinson AM. Screening for diabetes in general practice: cross sectional population study. Br Med J 2001; 323: 548–551.
24 Williams DR, Wareham NJ, Brown DC et al. Undiagnosed glucose intolerance in the community: the Isle of Ely Diabetes Project. Diabet Med 1995; 12: 30–35.[ISI][Medline]
25 Harris MI. Undiagnosed NIDDM: clinical and public health issues. Diabetes Care 1993; 16: 642–652.[ISI][Medline]
26 Harris MI, Eastman RC, Cowie CC, Flegal KM, Eberhardt MS. Comparison of diabetes diagnostic categories in the US population according to the 1997 American Diabetes Association and 1980–1985 World Health Organization diagnostic criteria. Diabetes Care 1997; 20: 1859–1862.[Abstract]
27 UK Prospective Diabetes Study Group. UK prospective diabetes study XII: differences between Asian, Afro-Caribbean and White Caucasian type 2 diabetic patients at diagnosis of diabetes. Diabet Med 1994; 11: 670–677.[ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  E Goyder, S Wild, C Fischbacher, J Carlisle, and J Peters Evaluating the impact of a national pilot screening programme for type 2 diabetes in deprived areas of England Fam. Pract., September 1, 2008; (2008) cmn054v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. L. Norris, D. Kansagara, C. Bougatsos, and R. Fu Screening Adults for Type 2 Diabetes: A Review of the Evidence for the U.S. Preventive Services Task Force Ann Intern Med, June 3, 2008; 148(11): 855 - 868. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. P Klein Woolthuis, W. J. de Grauw, W. H. van Gerwen, H. J. van den Hoogen, E. H van de Lisdonk, J. F. Metsemakers, and C. van Weel Identifying people at risk for undiagnosed type 2 diabetes using the GP's electronic medical record Fam. Pract., June 14, 2007; (2007) cmm018v2. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||