Measurements
At diagnosis the GPs recorded height and weight, examined both legs for amputations, presence of patellar reflexes, sense of touch of cotton wool and pin prick, presence of dorsalis pedis or posterior tibial pulse, and recorded history of myocardial infarction and/or stroke causing hospitalization [
42]. A fasting blood sample was drawn and a freshly voided morning urine sample was collected. Retinopathy was assessed by practicing ophthalmologists. The centralized methods used for measurement of blood samples and urinary albumin concentration have previously been reported [
42]. Fraction of HbA1c was analyzed by ion exchange high-performance liquid chromatography (reference interval: 5.4-7.4%; the interval may cautiously be translated into 4.8-6.7% using a DCCT-aligned method). Only samples measured within 90 days after diabetes diagnosis were accepted.
The GPs, together with the patients, completed a questionnaire regarding the presence of 16 typical symptoms of diabetes (abnormal thirst, frequent urination, unintended weight loss, fatigue, confusion, visual disturbances, cramp in calves (pain or paresthesia in lower extremities), genital itching, balanitis, recurrent urinary tract infections, stomatitis, recurrent skin infections, foot ulcer, gangrene, angina pectoris, intermittent claudication) and one open category. The questionnaire was based on a literature search and interviews with experienced diabetologists. In questionnaires, patients gave information about angina pectoris, intermittent claudication, cancer disease, lifestyle, socio-demographic factors and SRH [
42]. SRH was measured with a single validated question [
5,
43] (see Additional file
1) which was completed a median (interquartile range) of 12 (5–25) days after diabetes diagnosis.
At the 1-year follow-up GPs handed out a patient questionnaire called “A life with Diabetes” (no reminder was issued). The questionnaire was constructed by the study coordinator, Niels Olivarius, and an experienced sociologist, Dorte Gannik. It was based on: a) a literature review of previous qualitative and quantitative research about patients’ views on living with chronic illness; b) theories about living with a chronic illness drawing their main inspiration from the symbolic interactionist theory; c) a review of existing instruments; d) in-depth qualitative interviews with three newly diagnosed patients with T2DM (performed by Dorte Gannik), and e) discussions with health professionals. A group of people familiar with the construction of questionnaires (GPs and sociologists) reviewed the questionnaire several times. Questions were rephrased and/or removed in cases of unclear wording or nonexclusive response categories. After the questionnaire review, 15 T2DM patients tested the questionnaire, and this led to a further revision. Before the questionnaire was sent to the GPs participating in our study, it was shown again to GPs and sociologists familiar with the construction of questionnaires. They had no further comments on the wording and response categories. The questionnaire contained questions on SRH and health behaviour (the patients’ dietary habits and their indication of change in their way of living after diagnosis). In multiple response questions the patients reported whether they received the necessary support and understanding from family and significant others (social support) and how they were coping (two questions exploring whether the patients felt diabetes was a mental/practical and/or an illness burden in daily life, and their emotional attitudes towards diabetes). The responses concerning social support and coping were summarised into new categories (see Additional file
1). The GPs reported on the patients’ antidiabetic treatment and the number of diabetes-related consultations within the last year.
Statistical analysis
First, the relation between SRH (we combined the two lowest categories, poor and very poor, into one category, “poor”, due to few respondents reporting very poor) and socio-demographic status, diagnostic HbA1c, and other factors indicating the illness severity at diagnosis (Table
1) were analysed bivariately to establish the baseline distribution of SRH. One year later we made a similar comparison between SRH and the patients’ indication of coping, health behaviour, social support, and treatment since these variables are suggested to vary with SRH (Table
2). Illness severity at diagnosis may possibly have influenced the variation of coping, health behaviour, social support, and treatment. Those patients with diabetic complications would hence be more likely to have another health behaviour than those without complications, and therefore our analysis took account of the distribution of these factors (see below). Factors relevant to glycaemic control, like diabetic complications and cardiovascular status, were not measured one year after diagnosis as there would be little meaning in obtaining such information after only one year. Moreover, most patients with T2DM are still in the early stages of disease progression “honeymoon” period one year after diagnosis, and the blood glucose level is typically very low [
44].
Table 1
Distribution of self-rated general health by patient characteristics at diabetes diagnosis in patients with type 2 diabetes
N (%) | 71 (11.6) | 210 (35.1) | 267 (44.6) | 51 (8.5) | |
Sex (%) | | | | | 0.18 |
Women | 28 (39.4) | 93 (44.3) | 141 (48.8) | 27 (52.9) | |
Men | 43 (60.6) | 117 (55.7) | 126 (47.2) | 24 (47.1) | |
Age (Years) | 65.5 | 65.9 | 65.9 | 63.5 | 0.59 |
(56.0-73.7) | (54.1-73.2) | (57.4-74.4) | (57.1-73.1) |
Cohabiting (%) | | | | | 0.64 |
Yes | 48 (68.6) | 151 (71.9) | 185 (69.3) | 32 (62.8) |
Smoking habits (%) | | | | | 0.72 |
Never | 20 (28.2) | 70 (33.7) | 82 (30.7) | 18 (35.3) |
Former | 27 (38.0) | 73 (35.1) | 83 (31.1) | 17 (33.3) |
Current | 24 (33.8) | 65 (31.3) | 102 (38.2) | 16 (31.4) |
Level of physical activity (%) | | | | | <0.0001 |
Much | 8 (11.4) | 24 (11.3) | 10 (3.8) | 1 (2.0) |
Moderate | 51 (72.9) | 145 (69.1) | 170 (63.9) | 27 (52.9) |
Low | 11 (15.7) | 41 (19.5) | 86 (32.3) | 23 (45.1) |
Number of diabetes-related symptomsa | 3 | 3 | 4 | 5 | <0.0001 |
(1–4) | (2–5) | (2–5) | (3–7) |
Haemoglobin A1cb (fract., %) | 9.6 | 9.9 | 9.6 | 9.1 | 0.53 |
(8.0-11.3) | (8.5-11.5) | (8.1-11.5) | (8.2-11.2) |
Body mass index (kg/m2) | 28.9 | 29.4 | 29.4 | 29.2 | 0.49 |
(26.0-32.2) | (26.2-32.6) | (26.3-32.9) | (25.8-33.6) |
Cancer (%) | | | | | 0.73 |
Yes | 4 (5.7) | 8 (3.8) | 20 (7.5) | 5 (9.8) | |
Complications (%) | | | | | |
Retinopathy | | | | | 0.72 |
Yes | 3 (4.4) | 11 (5.7) | 9 (3.7) | 3 (6.4) | |
Neuropathy | | | | | 0.64 |
Yes | 12 (17.1) | 39 (18.8) | 48 (18.2) | 13 (29.5) | |
CVD | | | | | <0.0001 |
Yes | 9 (12.7) | 46 (21.9) | 90 (33.7) | 25 (49.0) | |
Urinary albumin | | | | | 0.40 |
Normal | 37 (55.2) | 117 (60.6) | 132 (55.7) | 23 (52.3) |
Microalbuminuria | 27 (40.3) | 72 (37.3) | 89 (37.6) | 19 (43.2) |
Proteinuria | 3 (4.5) | 4 (2.1) | 16 (6.8) | 2 (4.6) |
Table 2
Distribution of self-rated general health by patient characteristics one year after diagnosis of type 2 diabetes
N (%) | 131 (21.9) | 263 (43.9) | 186 (31.1) | 19 (3.2) | |
Sex (%) | | | | | 0.52 |
Women | 59 (45.0) | 123 (46.8) | 98 (52.7) | 9 (47.4) | |
Men | 72 (55.0) | 140 (53.2) | 88 (47.3) | 10 (52.6) | |
Health behaviour:
| | | | | |
Has changed way of living after diagnosis (%) | | | | | 0.22 |
Yes | 99 (76.2) | 211(82.4) | 153 (84.5) | 16 (88.9) | |
Food habits (%) | | | | | 0.055 |
Diabetes dieta | 57 (44.2) | 124 (49.0) | 82 (46.3) | 8 (42.1) | |
Full diet without sugar | 50 (38.8) | 110 (43.5) | 79 (44.6) | 19 (57.9) |
Diet as non-diabetics | 22 (17.1) | 19 (7.5) | 16 (9.0) | 0 |
General practice visits per year | 3.7 | 3.5 | 3.7 | 3.6 | 0.51 |
(2.6-4.8) | (2.6-4.7) | (2.8-5.0) | (2.7-4.3) |
Treatment: Diet as only treatment to reduce blood glucose (%) | 93 (71.0) | 143 (56.8) | 91 (51.3) | 6 (33.3) | 0.0005 |
Social support (%) | | | | | 0.0002 |
Full support | 92 (73.0) | 165 (65.5) | 93 (52.8) | 9 (50.0) | |
Handles it myself | 26 (20.6) | 71 (28.2) | 54 (30.7) | 4 (22.2) |
Feel alone, misunderstood | 8 (6.4) | 16 (6.4) | 29 (16.5) | 5 (27.8) |
Coping:
| | | | | |
Illness burden (%) | | | | | <0.0001 |
No burden | 85 (68.0) | 139 (59.1) | 64 (36.0) | 1 (5.3) | |
Minor burden | 30 (24.0) | 68 (26.5) | 56 (31.5) | 5 (26.3) |
Some burden | 9 (7.2) | 45 (17.5) | 40 (22.5) | 5 (26.3) |
Major burden | 1 (0.8) | 5 (2.0) | 18 (10.1) | 8 (42.1) |
Attitudes towards diabetes (%) | | | | | <0.0001 |
The illness is unproblematicb | 49 (38.0) | 113 (43.6) | 81 (44.3) | 4 (23.5) | |
Work/have worked with the illnessc | 78 (60.5) | 134 (51.7) | 71 (38.8) | 5 (29.4) |
It is a strain | 2 (1.6) | 12 (4.6) | 31 (16.9) | 8 (47.1) |
Variables indicating the patients’ coping strategies and social support are regarded as latent variables. To ensure that a co-variation between SRH and these variables does not reflect a common underlying latent variable, we performed a graphical Rasch analysis [
45] conditional on the exogenous variables age, diagnostic HbA1c, smoking habits, and physical activity.
With a multivariate analysis we estimated how the covariates influenced change in SRH from diagnosis until one year later. Illness severity at diagnosis may influence the variation of the covariates: patients with diabetic complications may have health behaviours which are different from those without complications, and an analysis accounted for the distribution of these factors.
The expected level of the two SRH measurements was modelled using a cumulative probit model for ordinal data. One interpretation of this model is that the ordinal measure of SRH is the realization of a continuous, latent SRH, assumed to be normally distributed, with a mean that may be allowed to depend on the covariates, and a variance of one. The larger this latent SRH, the better the SRH. The expected differences in the person’s level of SRH at diagnosis were estimated using baseline information. To avoid a scenario where an effect of a covariate on the level of SRH at diagnosis was misinterpreted as an effect on the estimated change between the two time points, the covariates were also used for modelling the mean SRH at diagnosis. The expected change in SRH, defined as latent SRH one year after diagnosis minus latent SRH at diagnosis, was allowed to depend on the patients’ indication of the impact of diabetes on daily life and their evaluation of social support, indication of change in their way of living, and other variables shown to be significant for the physical condition of diabetes and life in general. (These are age, sex, cohabiting status, smoking habits, physical activity, number of diabetes-related symptoms, HbA1c, BMI, cancer, peripheral neuropathy, diabetic retinopathy, CVD, urinary albumin, change in way of living after diagnosis, food habits, clinic visits per year, antidiabetic treatment, social support, illness burden, and attitudes towards diabetes). The test of whether a variable relates to change in SRH corresponds to a test of interaction with time in the cumulative probit model.
A patient’s SRH at two different time points may be correlated. The analysis therefore took account of a possible correlation of measurements within one person to avoid possible incorrect conventional confidence intervals for the different parameters by using generalized estimating equation methods (PROC GENMOD, SAS version 8.2). The hypotheses were tested using a generalized Wald test. Hypotheses regarding the effect of ordinal, categorical variables on change in latent SRH or at the level of latent SRH at diagnosis were tested as a trend test using a model where the ordinal variable was included as a continuous variable. In the multivariate analyses we performed backwards elimination of variables based on p ≤ 0.25. The nominal statistical significance level was <0.05.