Disentangling the effect of illness perceptions on health status in people with type 2 diabetes after an acute coronary event

This study is a secondary analysis of an RCT [19]. Between October 2011 and August 2013 type 2 diabetes patients with a first acute coronary event were recruited via their cardiologists in thirteen hospitals in three regions across the Netherlands. An acute coronary event was defined as a percutaneous transluminal coronary angioplasty (PTCA), a coronary artery bypass graft (CABG) procedure and/or a myocardial infarction (MI). After hospital discharge, individuals were invited to participate if 1) they were known with type 2 diabetes for more than one year; 2) were older than 35 years; 3) had sufficient knowledge of the Dutch language; 4) were able to fill in questionnaires; and 5) had no other serious illnesses or conditions that would prevent full participation. Exclusion criteria were 1) having had a previous acute coronary event and 2) being discharged from the hospital longer than two weeks. After informed consent, participants were randomised in a 1:1 ratio (parallel groups) to the intervention or the attention control group [20]. Randomisation, stratified on region, was generated at the person level by a computerized random-number generator at the research centre. The study protocol was registered at the trial registration (Nederlands trial register; NTR3076) and approved by the Medical Ethical Committee of the University Medical Center Utrecht (Protocol number 10–403).

The full details of the design, rationale and theoretical framework of this RCT have been described previously [18]. The intervention aimed to strengthen beneficial illness perceptions and alter less beneficial perceptions, with the purpose to decrease diabetes-related distress and improve health status. For example, both type 2 diabetes and the acute coronary event were discussed during the intervention and questions about the (consequences of) both diseases were considered in order to increase coherence. Furthermore, feelings of uncontrollability of the conditions were discussed to identify aspects of the disease that could be controlled. Seven trained diabetes nurses visited the participants in the intervention group three times at their homes (Table 1). During the visits, the nurses used a questionnaire on illness perceptions and strategies of motivational interviewing to stimulate beneficial perceptions of both type 2 diabetes and the coronary event, and they tried to challenge less beneficial conceptions.

Participants in the attention control group received one consultation (approximately 15 min) by telephone, within three weeks after hospital discharge, to discuss how they felt and functioned in the period after discharge. Two members of the research team conducted the consultations, in accordance with a semi structured protocol. They were instructed to give the patients personal attention, without giving advice or support regarding treatment and self-management. The focus of the consultation was on the topics which were of high importance to the patient. Open questions were used to stimulate the discussion (e.g. “Can you tell me how you have been doing since you got home?”).

Patients completed a set of questionnaires at their home, after hospital discharge but before the first home visit/phone consultation (T0) and five months later (T1). The validated Dutch version of the Brief Illness Perception Questionnaire (BIPQ) was used to measure illness perceptions [21]. It comprises eight items on a 0–10 scale on the dimensions: 1. Consequences (“How much does your illness affect your life?”); 2. Timeline (“How long do you think your illness will continue?”); 3. Personal control (“How much control do you feel you have over your illness?”); 4. Treatment control (“How much do you think your treatment can help your illness?”); 5. Identity (“How much do you experience symptoms from your illness?”); 6. Illness concern (“How concerned are you about your illness”); 7. Coherence (“How well do you feel you understand your illness?”); and 8. Emotional representation (“How much does your illness affect you emotionally?”). Regarding the beneficial and less beneficial perceptions, it is known that people who perceive little control over and coherence of their disease have a lower quality of life [16]. In addition, an open question assessed the perceived cause(s) of illness, by which respondents list their three most important factors they believe to have caused their illness. The BIPQ shows good test-retest reliability and is designed as a one-item domain instrument [22]. At baseline and follow-up participants received a version of the BIPQ that measured their perceptions of the diabetes (diabetes-BIPQ) and a version that measured their perceptions of the acute coronary event (acute coronary event-BIPQ).

Self-reported health status was measured using the Euroqol Visual Analogue Scale (EQ-VAS) [23], a subjective measurement of health status, measured on a graded vertical line ranging from 0 to 100. Participants were asked to mark a point on the EQ-VAS that best reflected their actual health status; a score of 100 representing the best and a score of 0 the worst imaginable health state.

Data about HbA1c levels, other comorbidities (cancer, chronic respiratory diseases, joint problems, other) and microvascular complications (eye problems, kidney problems, neurological problems and foot problems) were extracted from hospital files.

For the analyses described in this manuscript, we based our post-hoc calculation on health status measured with the EQ-VAS. With a sample size of 161, a mean score of 69.5 on the EQ-VAS, a standard deviation of 16.4, a two-sided significance level of 5% and a power of 80% we were able to detect a difference of 7.0 on the EQ-VAS. The minimal important difference of the EQ-VAS is assumed to be > 3.0 [24]. Therefore, our sample size of 161 was large enough to detect a clinically relevant difference on the EQ-VAS.

Data were analysed according to the intention-to-treat principle. Thus, participants who did not complete the intervention, or did not want the telephone consultation, were asked to complete the follow-up questionnaires as well. Descriptive statistics were used to describe the study sample. Changes in illness perceptions were analysed using ANCOVA, for the items of the diabetes-BIPQ and acute coronary event-BIPQ separately. In the model, treatment allocation (intervention or control group) was included as factor and the baseline score on the BIPQ item as covariate.

For the illness perceptions items that were influenced by the intervention, we assessed their potential mediating role to explain the intervention effect on participants’ health status by estimating a series of linear regression models as proposed by Baron and Kenny [24]. In a first model the effect of the independent variable (here, the intervention) on the mediator (here, change in an illness perception variable from baseline to 5 months later) was estimated. In a second model the effect of the independent variable (here, the intervention) on the dependent variable (here, change in health status from baseline to 5 months later) was estimated. And in the final, third model, the effects of both the independent variable (intervention) and the mediator (change in an illness perception variable) on the dependent variable (change in health status) were estimated.

Following Baron and Kenny, mediation by participants’ illness perception is established, if the intervention has a significant effect on participants’ illness perception(s) in the first model and on their health status in the second model, and participants’ illness perception(s) significantly affects their health status in the third model. This mediation is illustrated in Fig. 1, by the causal pathway Intervention → Illness perception (path a) → Health status (path b). The figure shows that the mediation effect is an indirect effect of the intervention on health status via its impact on participants’ illness perception(s). If the indirect effect is significantly different from zero (using Sobel’s test, as recommended by Baron and Kenny [24]) we conclude that the intervention effect on health status is mediated by participants’ illness perception(s). A p-value of p < 0.05 was considered significant. All analyses were performed using SPSS version 20.0.

All analyses were performed using SPSS version 20.0.

Mean scores on the diabetes-related perceptions at baseline and follow-up are presented in Table 3. A significant intervention effect on coherence (F = 8.19, p = 0.005) was found; only participants in the intervention group perceived a better understanding of their diabetes at follow-up compared to baseline (T0: 6.5 ± 2.9; T1: 7.3 ± 2.5, p = 0.003). A significant intervention was also found on perceived treatment control of the diabetes (F = 14.01, p < 0.001); the participants in the control group showed decreased feelings of treatment control after five months (T0: 7.6 ± 2.0; T1: 6.7 ± 2.3, p = 0.007), while these perceptions did not change in the intervention group (Fig. 3).

Regarding the perceptions of an acute coronary event, an intervention effect was found on perceived treatment control of (F = 7.81, p = 0.006) (Table 4). Again, perceptions of treatment control decreased in the control group (7.5 ± 2.0; T1: 6.9 ± 2.6, p = 0.036), but not in the intervention group. Furthermore, the perceived personal control over the coronary morbidity significantly improved in the intervention group (T0: 4.8 ± 2.7; T1: 5.6 ± 2.9, p = 0.004) and did not change in the control group, but the intervention effect was not significant. Furthermore, the control group had less concerns regarding the acute coronary event at follow-up compared to directly after discharge (T0: 5.6 ± 3.0; T1: 4.9 ± 2.8, p = 0.020), while concerns did not change in the intervention group (Fig. 3).

A significant improvement on health status was found after the intervention (T0: 69.9 ± 17.3; T1:76.8 ± 15.6, p < 0.001) as well as a significant intervention effect (β = 0.20, p = 0.016; Fig. 4a–c).

Regarding the effects of participants’ change in illness perceptions on their health status, few significant effects were found. A better perceived coherence regarding the diabetes was associated with a better health status (β = 0.18, p = 0.035). The effect of the intervention on health status was slightly lower when coherence regarding the diabetes was taken into account compared to the model without coherence (Fig. 4a). Nevertheless, the mediating effect was not significant, as demonstrated by the Sobel z-test (z’ = 1.54, p = 0.12). The total-, direct- and indirect effects are reported in Table 5. Perceived treatment control regarding the diabetes (Fig. 4b) and perceived treatment control regarding the acute coronary event (Fig. 4c) were not associated with health status and hence were not considered to act as mediators.