To the best of our knowledge, this was the first study to investigate the different effects of specific diabetic complications on HRQoL and health preference in a Chinese population. Patients with history of any of the four major DM-related complications (heart disease, stroke, ESRD and STDR) had lower SF-12v2 and SF-6D health preference scores than the general healthy population sample, as well as uncomplicated DM. After adjusted for socio-demographic and clinical parameters, heart disease was associated with lower GH and SF, but not SF-6D health preference. Stroke, ESRD and STDR were associated with lower PF, RP, BP, SF and PCS and SF-6D health preference. None of the complications were associated with lower MH and MCS.
Compared to the general healthy population sample, uncomplicated DM had lower scores in GH, VT and PCS, but higher scores in SF, RE, MH and MCS. As a result, the SF-6D health preference in uncomplicated DM was higher than that of the general healthy population sample. A few previous studies had compared the HRQoL of diabetic patients with non-diabetic patients. The findings on the impact of DM on HRQoL and health preference were inconclusive. A study in Singapore population showed that the HRQoL (measured by the SF-36 health survey) and SF-6D health preference score in diabetic patients without vascular complications were not significantly different from that of non-diabetic subjects [11]. A study in Spain also showed that health preference measured by the EQ-5D was not significantly lower in diabetic subjects without vascular complications compared to non-diabetic subjects [45]. Studies that did not adjust for diabetic complications found that diabetic patients had worse HRQoL than non-diabetic patients [7, 46,47,48]. There were several explanations for the discrepancy between our findings and previous studies. First, the lower HRQoL in previous studies might be due to the presence of DM-related complications in some study subjects. Second, all the previous studies used self-reported diagnosis of DM, while our study defined DM and complications by documented diagnosis in medical records by chart review. Third, subjects with DM tend to seek different ways of physical and emotional rehabilitation in person or in group. Their active response to their situation probably has a role in their better quality of life. Literature showed that both individual and group-based rehabilitation can improve DM patients HRQoL [49]. Forth, compared to our study subjects, subjects in these studies had lower proportion of being married [7, 46,47,48]. Our study showed that divorce/separated (coefficient − 0.041, P = 0.031, data not shown) and widowed (coefficient − 0.044, P = 0.009, data not shown) was negatively associated with health preference with married as reference group.
Although subjects with heart disease might not feel obvious body pain or role limitation, and there was minimal impairment of their physical functioning and role functioning (Tables 5 and 6), they still reported worse general health (decrement: -5.02, P = 0.062) and social functioning (decrement: -5.95, P < 0.05) probably because they were conscious and being cautious of life-threatening disease. Mild, reversible disease (nephropathy and NPDR/pre-PDR) did not lead to significant decrease in any SF-12v2 domains because they are asymptomatic. However, when the disease progressed to severer stage and became irreversible (ESRD and STDR), almost all the SF-12v2 domains had significant lower scores except MH. Therefore, it is important to screen and treat early these complications in DM subjects to preserve HRQoL.
None of the four major diabetic complications (heart disease, stroke, ESRD and STDR) were associated with lower mental aspect of HRQOL, which was consistent with overseas studies using SF-36 or SF-8 [11, 12, 50]. In our study, heart disease was found to have lower PCS than the general healthy population sample and subjects without complications, but the decrease was no longer significant after adjusting for clinical covariates and co-existing complications. Previous studies found that ischemic heart disease was associated with lower PCS in Singapore population [11, 13], the US [12] and UK [50]. Only one study in Singapore population (including 60% Chinese subjects) reported the impact of diabetic complications on 8 domains of HRQoL [11]. They found severe retinopathy was associated with lower scores in PF and RP, which was consistent to our study. Stroke was only associated with lower PF in Singapore population, while we found that patients with stroke had lower scores in RP, BP, SF, RE and PCS. There were only 10 subjects with stroke in the Singapore study, which could lead to wide variation in the scores.
Most previous studies on the health preference scores of diabetic patients employed EQ-5D [20, 51, 25, 52], one study in Singapore population (60% of the study subjects were Chinese) used SF-6D [11]. The SF-6D scores were 0.79 for non-diabetic subjects and 0.78 for uncomplicated DM in Singapore population [11], and the difference was not significant. The SF-6D score were lower than those found in our Hong Kong Chinese population, which were 0.86 for the general healthy population sample without DM and 0.88 for uncomplicated DM. Compared to uncomplicated DM, decrements of 0.02 to 0.03 in the SF-6D scores were observed for diabetic patients with coronary heart disease, stroke, severe retinopathy and severe nephropathy respectively in the Singapore study. Another study in Australian overweight or obese subjects used the SF-6D and found that DM was not associated lower SF-6D preference score but CHD decreased health preference by 0.054 [53]. In the Chinese population in our study, the decrements in health preference were 0.031, 0.051, 0.047 and 0.037 for subjects with heart disease, stroke, ESRD and STDR. The SF-6D scores in our study were calculated by the Hong Kong Chinese population-specific scoring algorithm [40, 41], while the Singapore study and Australia study adopted the algorithm derived from UK general population [54]. In regarding to sociodemographic parameters, the subjects in the Singapore study was younger than our subjects (48 ± 11 vs 65 ± 10 years-old), lower proportion of subjects were separated/divorced/widowed (5.7% vs 20.8%). Only 60% subjects were Chinese and 40% were Malay and India in the Singapore study. Our study showed that age was not significantly associated with changes in health preference while divorce/separated and widowed was negatively associated with health preference with married as reference group. The lower health preference in the Singapore study was not likely to be explained by the differences in sociodemographic characteristics. The differences in scoring algorithm and population composition might cause the differences in the health preference.
This study found that subjects with two or more complications showed significantly lower scores than those with only one complication in physical component scores and SF-6D score, but unexpectedly MH and MCS scores were not significantly decreased with the increase in the number of complications. A study in Norway population using EQ-5D also showed a marked difference in health preference scores for DM patients with only one complication (0.80) and subjects with two or more complications (0.64) [18].
There were several strengths in this study. First, the numbers of subjects with and without diabetic complications were large enough to detect the relative differences in HRQoL and health preference. Second, purposeful sampling enabled us to assess the impact of different complications on HRQoL and health preference. Third, disease status was defined by documented clinical diagnoses, which are more reliable than self-reported diseases. Fourth, comprehensive demographic and clinical parameters were included in the regressions to determine the independent associations between specific diabetic complications and health preference.
The limitations of our study should be considered when interpreting the study results. First, we did not differentiate T1DM and T2DM in this study because T2DM accounted for over 90% of all diabetic patients [2], although this is not likely to affect our conclusions. Second, it was a cross-sectional study in which only association but not causation could be established. Third, we used telephone survey to collect HRQoL which might select respondents with better HRQoL and biased towards higher HRQoL scores. The same bias should have been applied to all the subjects, so the relative differences among different disease states were less likely to be biased. Fourth, the drop-out rate in the GOPC sample was relatively high (46.0%) due to incomplete information collection during the recruit period for 514 subjects, although all of them completed the telephone interview. The high drop-out rate caused a reduction in the sample size. Since all of these 514 subjects were randomly lost due to accidental cause instead of refusal, this was not likely to cause selection bias. Fifth, education level is an important sociodemographic factor, but we did not include it into analysis due to high missing rate (25%). Living place area was an important sociodemographic factor to reflect the socioeconomic status, of which was not collected in telephone survey.