Introduction

Diabetes is a chronic condition associated with reduced physical functioning [1]. There is increasing interest in using brief health status measures to monitor health changes and health service needs in people with diabetes. Those measures provide important information on population health over time and enables informed decisions to be made about the implementation and evaluation of public health action. Furthermore, those measures can be used for the measurement of disparities within countries and comparisons across countries [2].

Self-rated health (SRH) is a single-item measure that is one of the most widely used measures of general health in population health research. The most commonly used form asks people to characterize their health as excellent, very good, good, fair, or poor. SRH has received strong support as an independent predictor of morbidity, disability, health service utilization and mortality [3,4], even when adjusted for objective health indicators. It has been recommended for use in health monitoring by the World Health Organization, the US Centers for Disease Control, the Institute of Medicine in the United States and the European Commission [5-8]. Despite its extensive use as a brief health status measure with consistent predictive power in cohort studies, it is not clear what SRH exactly measures and why it has such strong associations with morbidity and mortality. There is increasing evidence that biological, physiological, and psychological factors are major determinants of SRH [9]. Self-rated health may reflect a subjective summary of all the information people have available about their understanding of health, including information on physical functioning in everyday life, lifestyle conditions that have adverse effects on health, information from health professionals, specific disease characteristics, mood and personality factors [9]. Therefore, SRH might be a dynamic evaluation of health, incorporating past health experience with current health conditions and future health expectations [10,11]. Using data from a large American community sample, Zajacova et al. [12] found that a substantial proportion of individuals (40%) changed their ratings of health across two interviews (one month apart). Han et al. [11] reported that a decline in SRH was a stronger predictor for mortality than a single measurement occasion in a sample of older women. They concluded that the rating of current health status is important but that the way participants arrive at their current health state might be even more important.

Although it is well known that SRH is a dynamic perception of health status that changes over time [13], relatively little is known about changes and the trajectories of SRH in people with chronic medical conditions. Most of the prospective studies have focused on general population samples [14,15]. These studies have identified different patterns of SRH over time; for example, persistently good health, good but declining health, persistently fair health, and fair but declining health [16]. The temporal change of SRH might be different for people with diabetes due to the progress and management of the condition.

The aims of the present paper were to: (1) determine whether a set of SRH trajectories can be identified in a community sample of people with diabetes; (2) describe sociodemographic and clinical characteristics associated with these trajectories; and (3) examine the effect of different trajectories of SRH on functioning outcome.

Identification of trajectory groups of SRH may provide a deeper understanding of SRH as a risk factor for variation in health outcomes in people with diabetes and might help identifying high risk groups in this population.

Materials and Methods

Results

A total of 2003 individuals with type 1 or type 2-diabetes participated in the DHS at baseline (53% female). After excluding those who refused to participate in a follow-up interview (n=246), 1757 individuals formed the baseline sample for the longitudinal cohort. The number of participants with complete information on SRH at baseline and the three follow up interviews was 1746, 1287, 1142 and 1144, respectively. The number of deceased participants at the three follow-up interviews was 32, 17 and 12, respectively. For the trajectory analysis, we included participants who a) had complete information on SRH at baseline and b) participated in at least two follow up interviews or were deceased. A total of 1288 people had complete information on SRH in at least two follow-up assessments and 948 of those participants participated in the four years follow-up assessment.

The participants and those who dropped out did not differ in SRH scores at baseline and follow-ups. Participants who dropped out had poorer functioning status (p<0.001), more diabetes specific complications (p<0.001) and suffered more often from depressive symptoms (p<0.001) at baseline than those who did not drop out.

Mean SRH scores decreased minimally over time: the scores for baseline and the three follow-up assessments were 71.8 (SD=24.2), 71.2 (SD=25.8), 68.6 (SD=27.8) and 68.5 (SD=28.4), respectively.

Four distinct trajectories of SRH were identified: (1) persistently poor self-rated health (P-SRH; 10.1 %); (2) progressively increased self-rated health (I-SRH; 10.4%); (3) progressively decreased self-rated health (D-SRH; 7.3 %); and (4) persistently good self-rated health (G-SRH; 72.2 %). The results of the trajectory analysis are presented in Table 1. The four identified SRH group trajectories are shown in Figure 1. SRH scores for each survey year were calculated from the beta coefficients presented in Table 1. SRH scores of group 1 (P-SRH) and group 2 (G-SRH) did not change much over time: there was a small linear decrease of SRH scores over time. In contrast, members of the D-SRH group had high SRH scores at baseline, similar to the G-SRH group and low SRH after three years, similar to the P-SRH group. Members classified as belonging to the I-SRH group had low SRH scores at baseline and medium to good SRH scores after three years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Self Rated Health (SRH) Trajectories, with 95% Confidence Intervals, for the Four Group Model (with Age, Sex and Education as Covariates), in the Montreal Diabetes Health and Well Being Study, 2008-2011.

Note. Group 1: persistently poor SRH; Group 2: progressively increased SRH; Group 3: progressively decreased SRH; Group 4 persistently good SRH.

https://doi.org/10.1371/journal.pone.0083088.g001

The mean posterior probability value of each group in the four-group model was 0.96 or higher.

Sociodemographic and clinical baseline characteristics are presented in Table 2. There were important differences between the four trajectory groups with respect to sociodemographic and clinical characteristics. Participants in the G-SRH group had a higher educational level, better lifestyle-related behaviors (lower rates of smoking, physical inactivity and obesity) and better functioning (lower rates of disability, fewer chronic conditions and lower rates of minor or major depression) at baseline compared to the other three groups. Those in the P-SRH group had more diabetes-specific complications, higher rates of poor functioning, higher rates of depressive symptoms, were more often smoker and more often physically inactive than participants in the other three groups. The I-SRH had poorer functioning and poorer lifestyle-related behaviors compared to the D-SRH group at baseline.

At three years follow-up assessment, 79.3% of participants in the P-SRH group had a history of depressive symptoms (baseline to three years follow up). Rates of depression history for the I-SRH, D-SRH and G-SRH groups were 57.3%, 67.4 %, and 27.9 %, respectively. Rates of poor functioning history were 86.7%, 68.9%, 82.6% and 24.4% for participants in the P-SRH, I-SRH, D-SRH and G-SRH groups, respectively.

Table 3 presents the results of the logistic regression for the association between poor functioning at 4 years follow-up and SRH trajectories. After adjusting for potentially confounding factors (poor functioning history, depression history, type of diabetes, duration of diabetes, age sex and education), the P-SRH trajectory was associated with a six times increased odds of poor functioning after 4 years (OR=5.9) while both D-SRH and I-SRH trajectories were associated with a more than two times increased odds of poor functioning after 4 years (OR=2.3and OR=2.0, respectively) compared to the G-SRH trajectory. Both D-SRH and I-SRH groups had a lower odds of poor functioning at 4 years follow-up compared to the P-SRH group.

In sensitivity analyses we repeated our analyses for participants with complete SRH information at all three follow-up assessments. In these analyses, the number of groups and group characteristics did not differ systematically.

Discussion

In this prospective community study of people with diabetes, we have identified four distinct trajectories of SRH over a three-year period. The vast majority of the study participants had no major change of SRH status over the course of three years: 72 % were assigned to a persistently good SRH trajectory, while 10% were assigned to a persistently poor SRH trajectory. In contrast, perception of health status changed for the other two groups: one group changed from good to poor SRH, while the other group changed from poor to medium/good SRH.

Our study adds to the current literature in three ways. First, we have evaluated short-term change of SRH in community sample of people with a chronic condition (diabetes), rather than evaluating long-term change of the SRH in a general population sample. Second, using trajectory modeling, we have identified one subgroup of participants with an improved self-rated perception over time. Most of the previous prospective studies on the course of SRH in community based samples have identified stable or decreasing trajectories of SRH [16,31-33]. Our results suggest that there might be additional (short-term) trajectories of SRH in people with chronic conditions like diabetes: for example, a trajectory where perception of health improves over time. The course of the chronic condition might affect health perception. For instance, changes in treatment or changes in lifestyle behaviors might improve diabetes management and reduce complications, which in turn might result in a better perception of health status.

Third, we found that SRH history might be a better predictor for poor functioning than a single assessment of SRH. For example, the progressively decreased SRH group and the persistently poor SRH group had similar poor SRH scores at three years follow-up assessment, but the risk of poor functioning at four years follow-up was different: those with a persistent poor perception of health status were at higher risk for poor functioning than the other group where health perception changed over time, even after controlling for poor functioning history and other potentially confounding variables. Furthermore, the two groups with changing health perception over time - D-SRH and I-SRH - had different SRH scores at three years follow-up, but both groups had a similar risk for poor functioning at the 4-year follow up.

SRH is a widely used measure; the usefulness of this single question is well established [9,34]. Nevertheless, it is not clear what exactly SRH measures. It is likely that it is an umbrella construct under which different dimensions such as physical, mental, social, and general well-being reside. Those dimensions may vary between individuals, where some may consider well-being as more important than symptoms.

The SRH measure is often dichotomized into two categories (fair/poor vs. excellent/very good/good) in epidemiological research. In the present study we used a coding suggested by Diehr et al. [18] which can be interpreted as the estimated probability of future health, or more generally as a global measure of health status. This coding takes into account that the response categories excellent, very good and good are similar but slightly different indicators for good health status, while the response categories fair and poor are similar but slightly different indicators for poor health status. The high stability of SRH observed in our study might be in part explained by this coding scheme: participants who switch from excellent to very good SRH or from very good to good SRH have only a minor change in SRH scores, while those who switched form excellent/very good/good to fair/poor SRH have a major change in SRH scores.

A limitation of this coding, as well as of the SRH measure, is the lack of medium health category.

The role of SRH as a risk factor for poor health outcomes has been shown in the literature [35,36], although most studies have focused on mortality or disability pension in the general population. Idler and Benyamini [37] suggested that SRH might be a dynamic evaluation of health status, judging the trajectory of health and not only current health at a defined point in time. SRH might also reflect resources linked to one’s ability to cope with health threats (both external resources such as education, social support, access to care and internal resources such as optimism, vigour and perceived control). Our results support the idea that SRH is a dynamic evaluation of health status, but it seems to be a somewhat robust dynamic evaluation. SRH is associated with many aspects of health. Hence one might expect that SRH would fluctuate along with these aspects. But the present study suggests a high stability of SRH over time in people with diabetes, at least in a short time interval (3 years). It is likely that SRH is only sensitive to major changes in the different health dimensions and/or other factors not directly related to health.

The study has limitations. SRH is a very brief measure and its longitudinal validity is threatened by ceiling effects (people in the highest category of SRH cannot improve). The PHQ-9 and the WHO-DAS-II are brief questionnaires for the assessment of depression and functioning. Functioning is complex, multi-dimensional phenomenon and a global self-report instrument might not cover all aspects of functioning. Furthermore, we have no information about functioning history before baseline and between follow-up assessments. We also have no information on clinical parameters like blood glucose levels. Another limitation is that our other study variables were self-reported and taken at one point in time, which may not fully capture a person’s life-time experiences. The sampling frame was limited to those with landline telephones, which might result in selection bias. Attrition might be another source of bias, although there was no evidence of different SRH scores in those lost to follow-up. Finally, our findings may not be valid for people with undiagnosed diabetes.

In conclusion, our results confirm that current SRH is an important predictor for disability in diabetes, but the trajectory of SRH seems to be even more important. Health professionals should pay attention not only to SRH per se, but also changes in SRH over time (e.g., decline of SRH from very good to fair or poor) among people with diabetes. A decline in SRH over time indicates that health status is deteriorating and that there is also an increased risk of poor functioning.