The National Diabetes Register in Sweden: An implementation of the St. Vincent Declaration for Quality Improvement in Diabetes Care

Our data represent diabetic patients from ∼75% of 80 HCMs and ∼15% of nearly 900 PHCs, collected from all parts of Sweden. A total of 46 HCMs in 1996, 60 in 1997, 55 in 1998, and 57 in 1999 participated. A total of 162, 174, 139, and 141 PHCs participated in the respective years. Only adult patients (>18 years of age) were registered. All patients gave written informed consent to participate. However, the participation in NDR was nonmandatory for patients and treatment units. The mean number of subjects with diabetes was 314–439 patients per HCM unit and 78–127 patients per PHC unit, implying that the main part of all diabetic patients per unit of HCM and PHC were included in NDR by most of the participating centers.

Registrations at local centers were generally performed by experienced physicians and nurses. All participating diabetic patients were registered on a paper form or via a specially designed database software supplied on diskette or CD-ROM, developed by Lund’s Data Central (LDC), Lund, Sweden. All registrations were sent to a central database and analyzed by use of SAS (SAS, Cary, NJ). The registrations have so far been performed during the year after the clinical data were collected, i.e., data from 1999 were registered in the year of 2000. Yearly reports of all registered diabetic patients separated into HCMs and PHCs have been presented by NDR to Swedish authorities (also available online at www.ldc.lu.SE/ndr).

This health care system is generally organized by 25 regional county councils, and HCMs and PHCs are managed separately. Diabetes care at HCMs is managed by specialists in internal medicine or endocrinology/diabetology and by specialized nurses. Diabetic patients generally visit their physicians and nurses several times per year. The health care at PHCs is normally delivered by two or more specialists in general medicine with individual responsibility for each patient and by specialized nurses. The patients in PHCs visit their general practitioners at least once per year, but more often when complications occur, and they also visit their nurses several times per year.

In this report, we aimed to describe the treatment situation for patients with type 1 and type 2 diabetes in Sweden, irrespective of health care level. Almost all of the type 1 diabetic patients in Sweden are treated in an HCM. Most type 2 diabetic patients, however, are treated in a PHC, whereas those with diabetes complications and multiple drug treatment are often treated in an HCM after being referred from a PHC.

The definition of type 1 diabetes was based on onset of diabetes before the age of 30 years in combination with treatment with insulin only. The corresponding definition of type 2 diabetes was onset of diabetes at the age of 40 years or above, irrespective of type of treatment. This means that all patients with diabetes onset between 30 and 39 years, as well as all patients with diabetes onset before the age of 30 years on oral treatment or on combined oral and insulin treatment were excluded from further analyses. Furthermore, some patients also had to be excluded due to missing data regarding various measured variables, e.g., diabetes duration.

The registered variables presented here were the following: type of diabetes treatment, age at diabetes onset, weight, height, HbA_1c, and BP. Questions (yes/no) regarding smoking and lipid-lowering treatment (from 1997) were also included. The patients were screened by use of local methods and devices. An HbA_1c level <6.5% was regarded as excellent glycemic control. BMI was calculated as weight divided by height squared. A standard recommendation for BP recordings has been published in Sweden (7,8), stating that a mean of two readings (Korotkoff 1–5) in the supine position shall be recorded, using a cuff of appropriate size. These recommendations were further endorsed by instructions from NDR

Mean levels with SDs and proportions were calculated for the variables separately for patients with type 1 and type 2 diabetes. Comparisons were performed between the first year (1996) and the last year (1999) of registrations, regarding mean levels by use of the Student’s t test for unpaired samples and regarding proportions by use of the χ² test. In a subset of data from repeated measurements in the same patients (cohort subanalysis), mean levels were compared between years by use of a paired Student’s t test. Multivariate analyses (standard least squares) were performed with HbA_1c and BP for 1999 as dependent variables and with age, sex, diabetes duration, BMI, the BMI change from 1996 to 1999, and smoking as independent variables. A P value <0.05 was considered to be significant.

All diabetic patients included in the NDR from HCMs and PHCs during the years 1996–1999 are shown in Table 1. The larger proportion of patients that were excluded from this study in 1999 was due to missing data regarding onset of diabetes, mainly in PHCs.

Patient characteristics concerning age, sex, duration of diabetes, BMI, and smoking habits are shown in Table 2. Mean age was lower in type 1 (39 years) than in type 2 (67 years) diabetic patients, but mean duration of diabetes was longer in type 1 (23 years) than in type 2 (9 years) diabetic patients. The proportions of men/women were similar in the type 1 and type 2 diabetic patients, ∼55/45%. Mean BMI levels increased significantly from 1996 to 1999 in both type 1 diabetic men (24.8 to 25.1 kg/m²) and women (24.8 to 25.1 kg/m²). This was also seen in type 2 diabetic men (28.0 to 28.2 kg/m²) and women (28.6 to 28.9 kg/m²). Smoking frequency was unchanged in type 1 diabetic patients from 1996 to 1999 (14.5 to 13.5%) but was higher in younger type 2 diabetic patients <60 years of age (21.5 to 20.4%) than in older type 2 diabetic patients ≥60 years of age (7.8 to 7.6%). The proportion of patients prescribed lipid-lowering drugs increased significantly from 1997 to 1999, both in type 1 (4 to 11%) and type 2 diabetic patients (10 to 22%).

Mean HbA_1c levels decreased significantly from 1996 to 1999, both in type 1 (7.5 to 7.3%) and type 2 (7.0 to 6.7%) diabetic patients (Table 3). The proportion of patients with diabetes that reached the national treatment target level of HbA_1c <7.5% increased significantly from 1996 to 1999 in type 1 (50 to 58%) as well as in type 2 (66 to 73%) diabetic patients. Similar tendencies were seen regarding limits of excellent treatment levels (HbA_1c <6.5%) and poor treatment levels (HbA_1c ≥9.0%).

Mean BP levels also decreased significantly from 1996 to 1999, concerning diastolic BP (130/75–130/74 mmHg) in type 1 diabetic patients and combined systolic and diastolic BP (151/82–147/80 mmHg) in type 2 diabetic patients (Table 3). The proportion of diabetic patients who reached BP target levels ≤140/85 mmHg increased significantly from 1996 to 1999 in type 1 (77 to 79%) as well as type 2 (32 to 42%) diabetic patients.

In a substudy of patients (cohort substudy) with repeated individual data for HbA_1c, a significant trend for better glycemic control was observed from 1996 to 1999 for both type 1 (7.4 to 7.3%) and type 2 diabetic (7.0 to 6.8%) patients (Table 4).

For type 2 diabetic patients in 1999, we found that those treated in PHCs constituted 61% (not shown). Compared with those in HCMs, patients treated in PHCs were older (mean age 69 vs. 64 years; P < 0.001), had a shorter mean diabetes duration (7 vs. 11 years; P < 0.001), had a higher mean BMI (men 28.5 vs. 27.8 kg/m²; women 29.2 vs. 28.2 kg/m²; both P < 0.001), and were less often smokers (11 vs. 13%; P < 0.001). They also had a lower mean HbA_1c (6.5 vs. 7.1%), a higher rate of HbA_1c <7.5% (80 vs. 63%; P < 0.001), a higher mean BP (149/80 vs. 144/79 mmHg; P < 0.001), a lower proportion of BP ≤140/85 mmHg (37 vs. 50%), and less use of lipid-lowering drugs (18 vs. 28%).

Concerning changes of treatment from 1996 to 1999 in type 2 diabetic patients (Table 5) it was found that the main types of treatments at HCMs were insulin treatment, with or without combination with oral drugs (76 to 81%), while the main treatment categories in PHCs were diet alone or oral drugs (76 to 72%). However, monotreatment with oral drugs decreased significantly from 1996 to 1999 in both HCM (18 to 11%) and PHC patients (49 to 44%). On the contrary, combination treatment with oral drugs and insulin increased significantly from 1996 to 1999 in HCM (15 to 24%) and PHC (7 to 14%) patients.

A prospective multivariate analysis was performed in type 2 diabetic patients (n = 5,426), with HbA_1c (1999) and BP (1999) as dependent variables, and with baseline (1996) age, sex, diabetes duration, smoking, and BMI, plus increase of BMI from 1996 to 1999, as independent predictors (Table 6). It was found that both a high HbA_1c value and high systolic and diastolic BP values at follow-up (1999) were independently and significantly related to a high BMI value in 1996 and also to a high increase of BMI from 1996 to 1999. Diabetes duration but not age was positively related to HbA_1c at follow-up (1999). Age, but not diabetes duration, was positively related to systolic BP (1999), while age and diabetes duration were both negatively related to diastolic BP at follow-up.

The main findings of this study, based on a repeated cross-sectional, large-scale screening of various risk factors in patients with diabetes at four independent surveys during 1996–1999 in Swedish healthcare, are the tendencies of decreasing mean HbA_1c and BP values in both type 1 and type 2 diabetic patients. Furthermore, for both type 1 and type 2 diabetic patients a tendency was also demonstrated for improving achievement of treatment target levels of HbA_1c (<6.5% or <7.5%) and BP (≤140/85 mmHg or <130/85 mmHg). The mean decrease of HbA_1c (0.3%) from 1996 to 1999 in type 2 diabetic patients of this study (Table 3) can be approximately calculated to correspond to a reduction of 4% in any diabetes complications and of 8% in microvascular complications, as a conservative estimation based on UKPDS predictions in all diabetic patients (5). Similarly, the mean decrease of systolic BP 3.5 mmHg in type 2 diabetic patients of this study should approximately correspond to a reduction of 3.5% in any diabetes complications, 5% in diabetes-related deaths, 3.5% in myocardial infarctions, and 4% in microvascular complications (9). The trend was further strengthened by the prospective analysis of those patients (cohort substudy) with available repeated measurements from 1996 to 1999 (Table 4), showing a significant decrease in mean HbA_1c levels. The benefit of lowering HbA_1c levels in order to prevent diabetes complications has been clearly demonstrated in recent intervention studies, e.g., DCCT (4), the Stockholm study (10), the Kumamoto study (11), and UKPDS (5).

A troubling finding was the relatively high frequency of smokers in middle-aged patients with type 2 diabetes (<60 years of age), similar to figures for the nondiabetic Swedish population in the same age-group. These findings underline the need for further tobacco preventive measures in middle-aged subjects (12,13), being a highly cost-effective intervention regarding prevention of cardiovascular complications (14,15).

We also documented an increasing use of lipid-lowering drugs from 1996 to 1999 in type 1 as well as in type 2 diabetic patients. Several recent large-scale studies, e.g., 4S (16) and the Heart Protection Study (17), have shown the importance of treating diabetic patients with lipid-lowering drugs in order to decrease the risk for macrovascular complications and cardiovascular mortality. These important findings seem to have influenced clinical practice.

Our definitions of type 1 and type 2 diabetes chosen for this report were based on accessible data in NDR, according to age of diabetes onset and treatment. The epidemiological definition of type 1 diabetes can alternatively be based on treatment with insulin within 1 year after diabetes diagnosis, but such data were not available in NDR. Those subjects with age of diabetes onset <30 years and treatment with insulin only should reasonably well exclude type 2 diabetic patients. Using age of onset ≥40 years for type 2 diabetes assured that all patients with onset age 30–39 years, and therefore difficult to classify with certainty for diabetes category, were eliminated in this study. Our definitions could still be open for some degree of misclassification, e.g., for type 1 diabetes with late adult-onset and for type 2 diabetes in the young, a phenomenon probably more prevalent in the U.S. than in Sweden.

The increasing participation by HCMs all over Sweden from 1996 to 1999, currently 71%, should make the sample of 6,000–10,000 type 1 diabetic patients in this study sufficiently representative for the whole population of ∼30,000–40,000 such patients (in all age-groups) in Sweden. Similarly, the sample of type 2 diabetic patients from HCMs should be a representative sample of this population. One problem thus far has been the lower participation rate (15%) in NDR of the PHC centers in Sweden, making the samples of 17,000–22,000 type 2 diabetic patients in this study less representative for the total Swedish type 2 diabetic population of ∼260,000–270,000 patients. However, the participating centers were distributed in all parts of the country, with a mean number of as high as 78–127 registered patients per PHC unit, implying that the sample should be fairly representative of the typical Swedish PHC center with two or more general practitioners and other staff. It should be noted that a higher participation rate by PHC centers would have yielded a larger proportion of type 2 diabetic patients from PHCs than the rate of ∼60% in this study.

Our findings were somewhat limited, as they were based on a comparison of data from four cross-sectional surveys. We have, however, shown that the favorable trend in glycemic control was still present in a substudy of subjects (cohort study) with repeated measurements (Table 4). In absolute terms, the reduction of HbA_1c was small, but this is in our view a marker for more general trends in the diabetes population and therefore of interest.

We found that type 2 diabetic patients had increasing mean BMI values from 1996 to 1999. Overweight/obese (BMI ≥27 kg/m²) type 2 diabetic patients in 1999 had higher mean HbA_1c and BP values than nonobese (data not shown). This finding was underlined by the prospective analysis demonstrating that both high BMI values in 1996 and a high increase of BMI from 1996 to 1999 were associated with high HbA_1c and BP values in 1999 (Table 6). This implies that further lifestyle measures should be of value for the treatment of obese type 2 diabetes. It should, however, also be kept in mind that treatment with sulfonylureas or insulin could further increase weight while improving glycemic control and preventing microvascular complications (5). Therefore, the two goals of weight control and improved glycemic control while using sulfonylureas or insulin therapy is not easy to accomplish for many patients.

Type 2 diabetic patients treated with oral agents constituted 44% in PHCs but only 11% in HCMs (1999), with a tendency toward a decreasing use of oral agents in monotherapy from 1996 to 1999 in both PHCs and HCMs. Concomitantly, we found an increasing use of the combination therapy of oral agents and insulin from 1996 to 1999, with a higher rate in HCMs (24%) than in PHCs (14%) in 1999.

The tendency seen in NDRs from 1996 to 1999 of a decreasing use of oral agents and of a concomitant increasing use of insulin combined with oral agents in both PHCs and HCMs should reflect ongoing discussions in Swedish health care regarding the best treatment choice for diabetes with progressive secondary failure.

Health care for diabetic patients in PHCs and HCMs should be improved according to empirical data and previous studies. It has been previously shown that a large proportion of elderly patients with diabetes is not receiving a standard of care in accordance with the published guidelines of the American Diabetes Association (ADA) (18,19). Additionally, a Dutch study observed that management of type 2 diabetic patients in primary care only partly adhered to the published Dutch guidelines (20). Management of type 1 diabetes has also been found to be suboptimal (21).

NDR has therefore been initiated as a national tool for a continuous quality assessment of diabetes care in Sweden. One aim is to provide population data regarding clinical practice, complementary to data from previous and ongoing clinical trials. Another aim is to provide local centers with data regarding the quality indicators of diabetes care, also making a comparison possible with the national or regional NDR data. Further quality validation of sampled data should be of value.

Local reports are used by the physicians and nurses for a continuous evaluation and quality assessment of diabetes care, with the aim to increase the degree of fulfillment of treatment goals regarding glucose, BP, and other variables and with the assistance by regional NDR coordinators. NDR also intends to introduce the use of an Internet-based data registration, which will further simplify the process and allow individual patients to follow their own diabetes data by use of the Internet. Validation has been carried out continuously by communication between participating centers and the central database at LDC. Regarding HbA_1c, Swedish HCM and PHC centers nowadays take part in a quality validation assessed by a collaboration between all laboratories measuring HbA_1c (a project called Equalis that started in 1995). This means that HbA_1c measurement has been quality-assured since the beginning of NDR by calibration to the standard Swedish high-performance liquid chromatography Mono-S method (22,23).

NDR is probably among the largest diabetes registers currently known. Comparatively, the National Health and Nutrition Examination Study (NHANES) III is a nationally representative sample of the U.S. noninstitutionalized civilian population, although with somewhat differing HbA_1c reference levels compared with the Swedish levels. In a sample of 1,482 self-reported type 2 diabetic patients from 1988 to 1994 (24,25), a mean HbA_1c of 7.7% was reported in individuals aged 65–74 years compared with a mean HbA_1c of 6.7% in this study with mean age of 67 years. Among NHANES III participants aged 65 years or older, former ADA guidelines goal for glycemic control (HbA_1c <7%) in individuals using diet alone, oral agents, or insulin were achieved by 71, 44, and 27%, respectively, compared with 79–95, 75–79, and 58–68%, respectively, in this study. Another sample of 733 adult type 2 diabetic patients from NHANES III from 1991 to 1994 (26) showed that 76% were treated with oral agents or insulin compared with 72% in PHCs and 93% in HCMs of this study.

In conclusion, the NDR is a tool for continuous quality assessment of Swedish diabetes care. Improving trends based on results from 1996 to 1999 were seen in both type 1 and type 2 diabetes according to decreasing mean HbA_1c and BP values and increasing use of lipid-lowering drugs. These trends could translate into clinical benefits regarding micro- and macrovascular complications and diabetes mortality. Obesity and smoking are still major health problems that need to be addressed in the diabetic population.

NDR is financially supported by the Swedish Board of Health and Welfare, Stockholm.

The NDR Working Group consists of Soffia Gudbjörnsdottir (National Coordinator), Björn Eliasson, Jan Cederholm, Peter Nilsson, Thomas Fritz (representative of primary health care centers), and Christian Berne (President of the Swedish Society for Diabetology). Anders Nilsson and Göran Blohmé were in charge of the registry from 1996 to 2000. Carl-David Agardh implemented the idea of a national diabetes registry in 1995.

We thank all of the participating physicians, nurses, and other staff members who contributed data to NDR. Most of all we thank the diabetic patients who both individually and collectively, via the organization of the Swedish Diabetes Federation, have supported NDR.

Data management and its compilation into the computerized register has been handled on a daily basis by staff at LDC, Lund, Sweden.