Diabetes mellitus in Irish general practice: Level of care as reflected by HbA_1c values

Abstract

Objective: To assess the level of care given to people with diabetes by general practitioners and factors affecting it. Methods: A cross-sectional study of Irish general practitioners, looking at practice characteristics and patient care over the previous 2 years; a nationally representative sample of 27 general practitioners. A total of 1030 people with diabetes were studied, of whom 201 were type 1 and 829 were type 2. Results: The response rate was 27 out of 52 (52%). HbA_1c values were not related to the patient's socioeconomic status. The average HbA_1c for type 1 people with diabetes was 7.81%, and for type 2 it was 7.1%. HbA_1c values were measured 3.02 times for type 1 and 3.16 times for people with type 2 diabetes. This is a good standard of care, especially for type 2 disease. Computerized practices and those patients whose care was shared with the hospital achieved better control, even though HbA_1c levels were checked less frequently with computerization. The use of a protocol in the practices also improved care. Those practices employing a nurse had increased frequency of measurement of HbA_1c and better control on univariate but not on multivariate analysis. Possible reasons for this are discussed.

Conclusion: Diabetes Mellitus is treated to a good standard in Irish general practice, especially type 2 disease. This standard appears to be independent of the patient's socio-economic status, is improved by GPs being computerised, in group practices and by providing care according to a protocol. Shared care also improves control. Employing a practice nurse may also improve care.

• Diabetes mellitus
• general practice
• standard of care

Introduction

Diabetes mellitus is an increasingly common and important chronic disease. The importance of strict control in preventing long-term complications has been established [1–6]. Glycosylated haemoglobin (HbA_1c) provides an indication of the “average” blood glucose over the preceding 3 mo [7]. It is commonly used to measure glycaemic control [1–6], [8]. The American Diabetes Association (ADA) has recommended that HbA_1c levels be kept below 7% [9].

People with diabetes should be reviewed annually, when HbA_1c levels would be checked [10–12]. One would anticipate that with good standards of care, HbA_1c values would be close to the ideal level recommended by the ADA.

Because of the large numbers of patients involved, it is becoming increasingly difficult for people with diabetes to be cared for exclusively by hospital clinics, and general practitioners are taking over the care of their own diabetic patients [13–15]. This paper examined HbA_1c average values and their frequency of measurement retrospectively over a 2-y period in Irish general practice.

Methods

This was a cross-sectional study. General practitioners (GPs) working in the Mid-Western Health Board region were identified from the Primary Care Unit listing. Three of the four Health Board areas were studied, including Limerick city (urban), Limerick county and Clare county (rural). The GPs were categorized as to whether or not they had a secretary, a practice nurse, whether or not their practices were computerized, urban or rural, and whether the practice was group or single-handed.

Doctors in these 10 different categories were then selected from the Primary Care Unit listing using random numbers. Where the doctor selected was in group practice, the entire group practice was studied.

The selection process aimed to match the sample to a nationally representative pattern using data from the ICGP national general practice study [16]. Selected practices were initially contacted by telephone by the principal investigator (RO'C). Following this, a questionnaire with an explanatory letter was posted to the practice (Appendix 1). After 2 wk, a telephone reminder was given.

The questionnaire was divided into two parts. Part 1 was filled in by the participating GP, and part 2 by either the GP or practice nurse. The first part sought practice characteristics. The second part looked at all known diabetic patients in the practice (type 1 and type 2), including their address. The deprivation measure used, based on each patient's address, was the Trinity College Dublin Small Area Health Research Unit's deprivation index, using the 2002 Census data (SAHRU 2002). This divides electoral divisions (EDs) throughout the country into deciles based on their deprivation score, with 1 being the least deprived and 10 the most deprived. Area-based deprivation measures are routinely used in health analysis in many other countries [17]. The SAHRU index was also used to categorize non-participating GPs, using the address of their main centre of practice. The deprivation scores of participating and non-participating GPs were compared. Practice characteristics of non-participating GPs were determined by telephone survey.

Glycosylated haemoglobin (HbA_1c) measurements were all estimated at the same laboratory (at the Mid-Western Regional Hospital, Limerick) during the study period. At the time of the study, HbA_1c values were not DCCT aligned. The day-to-day imprecision or variability of the HbA_1c assay in this laboratory is less than 2.2% as compared to two referenced standards (one normal and one high). The number of HbA_1c measurements taken over the study period was counted and the value averaged for each patient. The median number of measurements was compared between practices.

Diabetic patients were identified from doctor's memory, electronic searching of computerized records, disease registers, and repeat prescriptions for diabetic drugs. For the purposes of the study, doctors were asked to include only patients whose diabetes was diagnosed according to American Diabetes Association (ADA) or World Health Organisation (WHO) criteria [10]. Due to the absence of universal patient registration, the overall population for each practice was estimated using the methodology recommended by the Irish College of General Practitioners (ICGP). All patients in the General Medical Services scheme were automatically included, as well as those private patients seen at any time in the previous 5 years (Dermot Folan, Asst CEO, ICGP; pers. comm.).

Results

Fifty-two questionnaires were distributed. Twenty-seven were returned, giving a response rate of 52%. Data from 1030 people with diabetes were obtained. Of these, 201 had type 1 and 829 type 2 diabetes. Care was shared with a specialist in 84.5% of type 1 patients and 47% of type 2.

The numbers of people with diabetes from different socio-economic groups studied is shown in Table I.

Table I.  The numbers of persons with diabetes from the different socio-economic groups studied.

2002 SAHRU deprivation index		Patients’ type of diabetes
1 = least deprived; 10 = most deprived		Type 1	Type 2	Total
1	Count	37	96	133
	% within patients’ type of diabetes	19.7%	12.1%	13.6%
2	Count	12	35	47
	% within patients’ type of diabetes	6.4%	4.4%	4.8%
3	Count	4	31	35
	% within patients’ type of diabetes	2.1%	3.9%	3.6%
4	Count	7	30	37
	% within patients’ type of diabetes	3.7%	3.8%	3.8%
5	Count	13	40	53
	% within patients’ type of diabetes	6.9%	5.0%	5.4%
6	Count	12	60	72
	% within patients’ type of diabetes	6.4%	7.6%	7.3%
7	Count	23	73	96
	% within patients’ type of diabetes	12.2%	9.2%	9.8%
8	Count	14	134	148
	% within patients’ type of diabetes	7.4%	16.9%	15.1%
9	Count	38	152	190
	% within patients’ type of diabetes	20.2%	19.2%	19.4%
10	Count	28	142	170
	% within patients’ type of diabetes	14.9%	17.9%	17.3%
Total	Count	188	793	981
	% within patients’ type of diabetes	100.0%	100.0%	100.0%

Non-responders

There were 25 non-responders out of 52 (48%). Their characteristics are shown in Table II, which also compares their characteristics to national figures and those of responders. There were no significant differences in the area-based deprivation scores of participating and non-participating doctors.

Table II.  Comparison of practice characteristics of responders, non-responders and national figures (12,14).

Practice factor	National figure (%)	Responders (%)	Non-responders (%)
Secretary	100	96	100
Nurse	90	81	56
Non-computerized	28	19	8
Partially computerized	17	59	24
Fully computerized	55	22	68
Group	40	30	64
Single-handed	60	74	36
Urban	33	48	76
Rural	66	52	24

HbA_1c frequency

To consider this parameter, the socio-economic (SAHRU) deprivation index was condensed into the two categories “higher or less-deprived socio-economic group [1–5]” and “lower or more-deprived socio-economic group [6–10]”. The median number of times HbA_1c was measured was tested between levels of various factors using the Mann-Whitney U-test due to its non-normal distribution. The difference in medians between the two deprivation groups was not significant (p=0.480), with a median number for the higher and the lower groups both equal to 3.0. That is an average of every 8 mo. This is shown graphically in Figure 1.

Figure 1.  Number of times HbA_1c values were measured for higher (less deprived, SAHRU index 1–5) and lower (more deprived, SAHRU index 6–10) socio-economic groups.

When the median frequency values were tested against whether or not the practice had a nurse, they were found to be significantly different (p<0.0005). The median frequency for practices with a nurse was 3 (range 0–19), and for those without a nurse it was 2 (range 0–9). When tested against whether a practice was computerized, the median frequency value was again significantly different (p=0.04). For a computerized practice it was 2 (range 0–12), and for a non-computerized practice it was 3 (range 0–19). When it was tested against whether a practice provided care according to a protocol or just gave care according to patients’ demands, the medians were found to be significantly different (p < 0.0005). The median frequency with protocol use was 3 (range 0–19), and without it was 2 (range 0–12).

When tested against whether or not it was a group practice or single-handed, frequency values were not found to be significantly different (p=0.429). Likewise, shared care had no significant effect. The median frequency value with shared care was 3, and without it was 2 (not significant, p=0.699).

When multiple regression analyses were carried out to find out which factors were still significant after adjustment for the effect of the others, then whether a practice had a nurse was still significant (p <  0.0005), whether the practice was computerized was still significant (p=0.003), whether care was provided according to a protocol was significant (p< 0.0005), and whether a patient had shared care now became significant (p=0.002). Whether or not the practice was a group practice and the patient's area deprivation status were still not significant. The only other variables that showed significant differences were the practices themselves and the sex and age of the patients. There were obviously other factors (not measured in this study) that explained the significant differences between practices; the median frequency was 2 for the younger age group (under 50 y) and 3 for the older age group (50 and over), and 3 for females and 2 for males.

The regression results showing significant variables are given in Table III.

Table III.  Multiple regression results showing significant variables.

	Non-standardized coefficients		Standardized coefficients
Model 1	B	SE	Beta	t	Sig.
(Constant)	3.476	0.572		6.079	0.000
Practice	3.241×10^–2	0.009	0.112	3.588	0.000
Patient's sex	−0.432	0.146	−0.091	−2.958	0.003
Computer	−0.641	0.215	−0.096	−2.986	0.003
Shared care	0.479	0.155	0.101	3.101	0.002
Diabetes care	−0.929	0.158	−0.198	−5.865	0.000
Practice nurse	1.112	0.227	0.159	4.888	0.000
Age	9.294×10^−3	0.004	0.066	2.114	0.035

Dependent variable: HbA_1c, number of times measured.

HbA_1c average values

The average HbA_1c values for the 2-y study period were not significantly different for the recoded deprivation categories. Again, Mann-Whitney U-tests were used due to the non-normality of the distribution of average HbA_1c values. This is shown in Figure 2.

Figure 2.  Average HbA_1c values by socio-economic group. Less deprived: SAHRU index 1–5; more deprived: SAHRU index 6–10.

When the medians of the average HbA_1c values were tested between different practices, the practices themselves showed significant differences (p<0.0005), but most of this variation could be explained by age and gender differences of the practices’ diabetic patients.

When the median average HbA_1c values were tested against whether or not the practice had a nurse, they were found to be significantly different (p=0.001). The value for practices with a nurse was 6.9 (range 5–11.9), and for those without a nurse it was 7.6 (range 5–11.6). When tested against whether a practice was computerized, the median average HbA_1c value was again significantly different (p=0.001). The value for a computerized practice was 6.9 (range 5–11.9), and for a non-computerized practice it was 7.6 (range 5–11.5). When it was tested against whether a practice provided care according to a protocol or just gave care according to patients’ demands, the medians were found to be significantly different (p<0.0005). The median value with protocol use was 6.7 (range 5–11.7), and without it was 7.25 (range 5–11.9).

When tested against whether or not it was a group practice or single-handed, they were not found to be significantly different.

There were no significant differences in median average HbA_1c values with shared care. The value with shared care was 7.4, and without it, it was 6.5 (not significant, p=0.922).

When multiple regression analyses were carried out to find out which factors were still significant after adjustment for the effect of the others, then whether the practice was computerized was still significant (p=0.001), whether care was provided according to a protocol (p<0.0005), and whether a patient had shared care now became significant (p<0.0005). Whether or not the practice was a group practice now also became significant (p<0.0005); the mean was slightly higher in the single-handed practices. The patient's area deprivation status was still not significant. There were no significant differences between practices, whether a practice had a nurse and the gender of the patients. There were significant differences also between the age groups, where the median value was 7.8 for the younger age group (under 50 y) and 6.9 for the older age group (50 and over).

The regression results showing significant variables are given in Table IV.

Table IV.  Multiple regression results showing significant variables.

	Non-standardized coefficients	Standardized coefficients
Model 1	B	SE	Beta	t	Sig.
(Constant)	6.941	0.350		19.844	0.000
Computer	−0.567	0.167	−0.135	−3.392	0.001
Shared care	0.395	0.107	0.132	3.676	0.000
Diabetes care	0.399	0.106	0.134	3.752	0.000
Age	−1.51×10^−2	0.003	−0.163	−4.769	0.000
Group/single practice	0.520	0.133	0.153	3.902	0.000

Dependent variable: HbA_1c average.

Discussion

This study shows that general practitioners’ care of diabetes mellitus, as evidenced by the frequency of measurement and level of HbA_1c, is of a good standard, especially for type 2 disease. It is not related to the patient's socio-economic status. The American Diabetes Association (ADA) has recommended that HbA_1c levels be kept below 7% [9], and in our study this standard was all but reached for those with type 2 disease. However, it has been suggested that this target is unrealistically low since, after 5 y estimated duration of diabetes, average HbA_1c levels in clinics in Britain are 7.4%, and at least 8% thereafter [19]. Moreover, in the UK, the clinical indicators for glycaemic control in both type 1 and 2 diabetes give 3 points for having a HbA_1c checked in the previous 15 mo, 16 points where its last value is 7.4 or less, and 11 points where its last value is 10 or less [11]. Thus, the value of 7.81% in those with type 1 disease could also be considered good control by international standards.

The finding that HbA_1c levels were not affected by socio-economic deprivation is different to experience reported in the UK [20], and requires further study to confirm its accuracy. It seems to indicate active care of diabetes mellitus by Irish GPs across the socio-economic spectrum.

HbA_1c values were significantly lower in practices with a nurse and those which were computerized, although the difference between practices with nurses disappeared after adjustment for the other factors in a multiple regression model. This is not difficult to explain as most practice nurses would work to a protocol. Thus, in multiple regression analysis, the practice nurse factor could be masking (or acting as a proxy for) the protocol factor.

Computerized practices measured HbA_1c values less often than non-computerized practices. However, the HbA_1c values were significantly lower in computerized practices. This suggests that computerization probably improved the process of care, allowing better overall control to be achieved with less testing.

Care was shared with a specialist in 84.5% of type 1 patients and 47% of type 2. Shared care had a significant effect on HbA_1c values and frequency of measurement after adjustment for other variables. The level of shared care was high by comparison to the UK [21]. However, in Ireland, diabetes care has traditionally been delivered in the hospital sector. In the DiSC project [13] for example, 65% of patients with diabetes in the initial survey indicated that they were receiving their diabetes care exclusively from a specialist. This has been changing recently for a number of reasons, including the commitment of GP teams to become more involved in diabetes care and the increasing recognition that the diabetes epidemic makes specialist care for all impossible.

There are some limitations to this study. The response rate was just over 50%. There was a tendency for more single-handed doctors as well as those with a practice nurse, who were non-computerized or partially computerized, to reply. However, the geographical socio-economic coding of the practice area of non-respondents was indistinguishable from that of respondents. It is logical to conclude from this that the socio-economic spectrum of patients attending these practices would be similar, thus giving similar results. Figures for non-computerization, practice secretary, practice nurse, group and single-handed, urban and rural practices were broadly similar to national figures in the study group. There are clear limitations to the use of an area-based deprivation measure such as the SAHRU index as a measure of deprivation. However, in the absence of more specific individual or localized data, there are few alternatives. Although it was requested that only patients whose diabetes was diagnosed according to WHO or ADA criteria be included in the study, this could not be comprehensively confirmed given the retrospective nature of the study.

In conclusion, diabetes mellitus is treated to a good standard in Irish general practice, especially type 2 disease. This standard appears to be independent of the patient's socio-economic status. The standard is improved by being computerized, in group practices and by providing care according to a protocol. Shared care also improves control. Employing a practice nurse may also improve care.

Dr Tessa Greally, Department of Public Health, Mid-Western Health Services Executive, Limerick, Ireland; Dr David Whitford, Dept of General Practice, Royal College of Surgeons of Ireland; Dr James O'Hare, Mid-Western Regional Hospital, Limerick.

Source of funding: the Research and Education Committee of the Department of Health and Children, Dublin, Ireland; the Mid-Western Health Services Executive, Limerick, Ireland.

Appendix 1. Data collection questionnaire and accompanying letter for the study.

Mid-Western Diabetes Study Dear Colleague, This anonymous questionnaire is part of a study undertaken by the University of Limerick Diabetes Group. We are very anxious to get the highest response rate, so we will be most grateful if you could complete the questionnaire and the accompanying data sheet and return both promptly in the enclosed SAE. The aims of the survey are to establish the extent of both formal and informal diabetes care taking place in general practice, to identify the main factors affecting such care. We also aim to see if socio-economic factors, such as where a person lives, influence their diabetes. Please complete all questions as fully and accurately as you can. This should take no more than 15 minutes. The data sheet will take longer to complete but much of the data can be filled in by your secretary or practice nurse. Diabetes is defined as having ever fulfilled the following diagnostic criteria: non-fasting blood glucose of >11.2 mmol/l or a fasting glucose of >7.06 mmol/l. Completed questionnaires qualify for 2 days’ study leave and €100 payment towards the cost of collecting and filling in the data. If you have any questions, please contact me by e-mail or telephone (tel: [phone]; e-mail:[email protected]). The deadline for the return of completed questionnaires is Friday 6th June. Thank you for your time. Dr Ray O'Connor GP/practice characteristicsGuide to data collection sheetfor the previous 2 years For each diabetic patient, please check the following over the previous 2 years: 1. Pt id: Patient identification (year of birth and sex only) 2. Type 1 or 2: The type of diabetes the patient has Type 1 also known as insulin-dependent diabetes (IDDM) Type 2 also know as non-insulin-dependent diabetes (NIDDM) 3. HbA_1cno: The number of times HbA_1c was measured either by the practice or in the clinic. 4. HbA_1caverage: The average of the HbA_1c values for each patient over the previous 2 years. 5. Share care: Whether or not care for that patient is shared with a hospital specialist with an interest in diabetes. Mark yes only if the patient has been seen by the specialist service in the 2-year study period. 6. Has the patient a medical card (y = yes; n = no; o = non-means tested over-70s card)? 7. St benefit: In receipt of a state benefit (e.g. single parent, dole, invalidity pension)? 8. Address: The patient's current address. For reasons of confidentiality, please do not include the house name or number. 9. Smoker: Current smoking status (yes or no)? 10. Where dx: In diabetic patients diagnosed in the last 2 years, where was the diagnosis made? Was it in the practice (pract) or elsewhere such as in hospital (other)? 11. Stage: Current stage of diabetes treatment (applies to type 2 diabetics only; ignore for type 1) 0: controlled on diet only 1: controlled on diet and drugs

Table

Pt id	Type	HbA1c	HbA1c	Share care	M card	St benefit	Address	Smoker	Where dx	Stage	Smoker
Yob/sex	1 or 2	no.	average	y/n	y/n/o	y/n		y/n	Pract/other	0 or 1	y/n