Short-Term Effects of an Educational Program on Health-Seeking Behavior for Infections in Patients With Type 2 Diabetes: A randomized controlled intervention trial in primary care

We recruited 101 general practitioners (GPs) in four regional Municipal Health Services: one in the north, one in the east, and two in the south of the Netherlands, covering 2.7 million people in total. All patients with type 2 diabetes between the ages of 44 and 85 were eligible for inclusion in this study; participating GPs sent a recruitment letter to 30 randomly selected eligible patients from their practices. Patients not able to attend a meeting because of immobility, and individuals with insufficient knowledge of the Dutch language were excluded.

In a previous collaboration with the Dutch Diabetes Patient Association and the Municipal Health Services, we developed an educational program targeted to type 2 diabetic patients. The unpublished methods and data guiding the development of this intervention are included in online appendix 1 (available at http://www.juliuscenter.nl/appendix_venmans_et_al.pdf). The final program consisted of an invitational letter for a meeting (which contained questions as a first “cue” for patients to think about infections), an interactive meeting, a patient information leaflet, a consultation with the diabetes care provider, a Web site, and a newsletter. For the design of the meeting, the domains of the Health Belief Model (15) were used and the domains of knowledge, need for information, and intention were three supplementary domains studied. We addressed a total of nine determinants of health-seeking behavior. Patients who were not able to attend a meeting received written information at their home address with the same content as that discussed during the meetings. Intervention tools and recruitment letters are given in online appendix 2.

Practices were randomly assigned to an intervention or a control group. Both care providers and their enlisted patients with type 2 diabetes were randomly assigned to the same group. We used block randomization, taking into account the region where the practice was located and the size of the practice (single-handed versus group practice). The control group did not receive any intervention.

Before the implementation of the intervention, all patient educators received standardized training, and the patient information leaflet and material content of the educational meeting were carefully pilot tested in small samples of patients from the target group before the main intervention. The intervention was implemented in October 2005 with the invitation to participate in the 2-h educational meeting. Partners were also invited to attend. A total of 22 local meetings were held throughout the health regions (six in three regions and four in the remaining region), and all meetings were held within 3 weeks of each other. The average participation rate was 68%. In the Netherlands, type 2 diabetic patients visit their diabetes care provider four times a year for a regular check-up. As part of the intervention, employees of the Municipal Health Services visited the diabetes care providers and asked them to discuss symptoms and risks of infections during these check-ups. The Web site (http://www.juliuscenter.nl/diabetes/infecties [in Dutch]) was online from October 2005 to March 2006. The educational newsletter was sent 3 months after the start of the intervention.

Evaluation questionnaires were developed on the basis of results of focus groups conducted during the development of the intervention (online appendix 1) and with guidance from the field of vaccination behavior (16,17). Baseline measurements (T0) were assessed in September 2005 and effects (T5) in March 2006 by means of written questionnaires. The requested time for responding was 1 week, and both mailed (up to two) and telephone reminders were used. The first part of the questionnaire consisted of patient characteristics such as age, sex, educational level, type of diabetes treatment, and comorbidity. The remaining components of the questionnaire assessed determinants of health-seeking behavior (online appendix 3). We pretested the questionnaire in a few patients and assessed readability and comprehension. Answering categories were either dichotomous or according to 5-point Likert scales.

Participants were asked to keep a diary for 20 weeks (from October 2005 to March 2006). The diaries were sent to the homes of the participants after the meeting with instructions on how to complete them. We did not send the diaries before the intervention because we did not want to raise awareness among the control subjects. The requested time to respond was 1 week after the conclusion of the 20-week study period. Two written reminders were used. In the diary, symptoms of LRTI and UTI and visits to GPs were reported. The severity of the symptoms was assessed on a 4-point scale from “no discomfort” to “great discomfort.”

The primary outcome measure was an indicator of health-seeking behavior for UTIs and LRTIs, defined as the proportion of participants with a positive score on statements pertaining to the nine above-mentioned determinants of health-seeking behavior. A positive score per person was assigned if results for at least seven of the nine determinants of health-seeking behavior were positive. Per determinant, one or more statements were posed. At least half of the statements per determinant had to be positive. The secondary outcome measures were changes in the nine separate behavioral determinants of health-seeking behavior. With regard to our second hypothesis, outcome measures were the percentages of cases in which there was self-reported contact with the primary care practice. For LRTI, these cases were coughing and shortness of breath and coughing and fever for longer than 2 days. For UTI, these cases were complaints of the urinary tract for longer than 2 days or complaints of the urinary tract and fever for longer than 1 day (females) and just complaints of the urinary tract (males). . These criteria were derived from the advice given during the educational meetings. They were measured by means of the diaries.

Data were analyzed using SPSS software for Windows (version 12.0; SPSS, Chicago, IL). Statements that were measured on 5-point Likert scales were dichotomized in the two most positive answers versus the other three responses. Proportions and means were used to describe baseline characteristics.

Changes in determinants of behavior between T0 and T5 were measured with ANCOVA. This method was used because it adjusts each patient's follow-up score for his or her baseline score and has the advantage of being unaffected by baseline differences (18). ANCOVA was applied using the generalized equation estimation model to adjust for differences in baseline and for clustering at the level of practice (19). This step was performed using SAS Proc Genmod (version 8.02; SAS, Cary, NC). Odds ratios (ORs) and 95% CIs are presented for questionnaires only for those determinants with P < 0.05. Analyses were done according to the intention-to-treat (ITT) principle. In addition, we planned an on-treatment analysis comparing the scores in patients who actually visited the meetings with those of patients who did not and control patients. Subgroup analysis was done for patients with a lower educational level and for patients at high risk for a complicated course of infections. χ² analyses were used to analyze the diaries. Subgroup analysis was done for instances in which the patient reported severe complaints only (great discomfort).

Of the 1,124 patients who were randomly assigned and responded to the baseline questionnaire, 572 were assigned to the intervention group and 552 to the control group (Fig. 1). Baseline characteristics are given in Table 1. The mean ± SD age was 64 ± 9 years, 53% of the participants were male, comorbidity was present in 44% of the patients, and insulin was used by 13% of the patients. Mean duration of diabetes was 6 ± 7 years. Complete outcome data were available for 468 and 472 patients in the intervention and control groups, respectively. Compared with control subjects, participants in the intervention group who were lost to follow-up were more often men than women (55 vs. 42%). Of the patients who were lost to follow-up, the duration of diabetes, the use of insulin, age, and the intention to consult the GP when symptoms of infections were present did not differ between groups.

The meeting was attended by 68% of the intervention group patients for whom complete data were available (compared with 46% of all patients for whom data were incomplete). In total, 25% of the patients indicated that their risk of infection had been discussed with their diabetes care provider, 44% had read the newsletter, and 9% of the patients who had Internet access had visited the Web site.

Hypothesis 1. Compared with control subjects, knowledge of the risks of UTI and LRTI in intervention group subjects increased (31 vs. 4% for UTI and 37 vs. 0% for LRTI). The same was true for knowledge of the symptoms of UTI and LRTI (11 vs. 0% and 31 vs. 4%, respectively). Moreover, bronchitis and cystitis were perceived by more patients as dangerous (21 vs. 5% and 22 vs. 4%, respectively). Confidence to contact the GP if the practice was difficult to reach increased among participants in the intervention group, whereas among the control group this percentage remained at the baseline level (6 vs. 1% in control subjects). Furthermore, the intention of intervention group patients to consult the GP increased when they had symptoms of a UTI (3 vs. −3% in control subjects) and LRTI (4 vs. −5%) (Table 2). Although the baseline figures were already very positive, the intention to seek medical attention because of symptoms that could indicate an infection increased. Compared with the control subjects, the intervention group patients did not change their belief about a decreased risk of serious consequences if they contacted the GP (0 vs. −4%).

Hypothesis 2. At baseline, 28% of the intervention group participants had a positive score on seven of the nine determinants compared with 27% of the control subjects. After the educational program, these percentages were 53 and 32%, respectively, indicating an improvement of 25% in the intervention group compared with 5% in the control subjects (P < 0.001). Analyses done separately for patients with lower educational levels (from 24 to 50% in intervention patients and from 25 to 27% in control subjects) as well as for patients at high risk for a complicated course of infections (from 40 to 57% and from 31 to 43%, respectively) showed no substantial differences with the overall results.

A sensitivity analysis showed that the proportion of participants with a positive score on at least six of nine determinants increased from 50 to 72% (46 to 53% in control subjects; P < 0.001). For at least eight of nine determinants, the proportion of participants increased from 11 to 36% (11 to 13% in control subjects; P < 0.001).

Hypothesis 1. Subgroup analyses for participants of the intervention group who attended the meeting showed a 45% difference in knowledge about risks of LRTI (0% in control subjects). Knowledge about risks of UTI increased in 33% (5% in control subjects); knowledge of symptoms of UTI and LRTI increased in 15% (0% of control subjects) and 35% (4% of control subjects), respectively; and confidence to contact the GP if the practice is difficult to reach increased in 9% (1% in control subjects).

Hypothesis 2. After the educational program, 58% of the intervention group participants scored positively on seven of the nine determinants (ITT 53%). Of the intervention group participants who did not attend the meeting, 42% had a positive score.

Hypothesis 2. In total, 69% of the patients returned the diary, equally divided over the two groups. Of the patients in the intervention group, 43% (23 of 54 patients with symptoms) sought care for symptoms of UTIs, compared with 32% (21 of 65) of the control subjects. For LRTIs, the figures were 49% (51 of 105) and 41% (43 of 106), respectively. There was a trend toward visiting the GP more often. The differences were most obvious for men with complaints of the urinary tract (57 vs. 35%; OR 2.2 [95% CI 0.8–6.4]). Subgroup analyses for serious complaints (great discomfort) revealed an even larger difference for LRTIs (49 vs. 32%; 2.2 [0.8–6.6]); however, the numbers were small.

This is, to our knowledge, the first study that has assessed the short-term effects of an educational program on health-seeking behavior for infections in type 2 diabetic patients. Patient characteristics were comparable with those of typical Dutch type 2 diabetic patients (20,21). The program positively influenced determinants of health-seeking behavior. The proportion of participants with a positive score on at least seven of nine determinants of health-seeking behavior almost doubled from 28 to 53% (compared with an increase from 27 to 32% in control subjects).

The program especially increased the knowledge for the type 2 diabetic patients about the symptoms and risk factors of infections. It enhanced a realistic risk perception. In addition, self-efficacy and the intention to seek medical attention were positively influenced. On some points the effects of the education were limited; e.g., patients did not believe they had a decreased risk of serious consequences when seeking medical attention because of symptoms indicating an infection. An explanation might be the so-called “ceiling effects,” as the baseline data were already positive. The differences found between the on-treatment and ITT analysis supported the fact that the meetings were an essential part of the educational program in light of the fact that the proportion of patients who had read the newsletter (44%) or visited the Web site (9%) was relatively low. Unfortunately, we cannot compare these findings with those of other studies because the designs of most studies do not allow disentangling of the effects of the different components (22).

It is known that many men with complaints of the urinary tract do not visit their GP (23). We found a trend of intervention group patients visiting the GP more often, especially men with complaints of UTIs. However, the differences between the intervention and the control groups were not statistically significant. This result may be due to low numbers, because subgroup analysis for serious complaints of LRTIs revealed larger differences and may indicate that patients with type 2 diabetes perceive the necessity of seeking medical attention even more when complaints of LRTIs are serious.

The results of our study should be considered cautiously. Effects regarding intention to seek medical care were positive but moderate. However, the fact that all determinants changed positively may indicate that health-seeking behavior for infections by this high-risk group of patients will improve. We believe that this program is a first step in changing the perception of type 2 diabetic patients. Small changes may have a considerable effect on a larger scale.

Our study has several limitations. The first is the short duration of the follow-up. Positive effects from educational programs are mostly shown for interventions of a shorter duration. Indeed, one study showed that 1 year after the last session of an educational program, most of the clinical effects are lost (24). Also, our study showed a suboptimal use of the tools developed (newsletter and Web site). Barriers as to why patients did not fully use the tools provided with the program were not identified systematically. Further research should be conducted to assess ways of improving the use of the tools provided and facilitate their implementation in daily practice. A third limitation is the insufficient number of observations of serious complications to permit subgroup analysis. However, we believe that such a relation can be assumed for infections in diabetes care. Delays in care seeking lead to delays in treatment and may in turn lead to serious complications. A fourth limitation is that all measures were self-reported. We could have used medical records to validate health-seeking behavior. Whether the reported behavior is more likely to be overreported because of social desirability or underreported because of recall bias remains unclear. A fifth limitation is the kind of “dummy” intervention that might address concerns about the Hawthorne effect. Indeed, among all control subjects we raised awareness by asking about their behavior. For that reason differences between groups could have been diminished. Therefore, the Hawthorne effect, if any, would have biased our results to accepting the null hypothesis (finding no difference between groups).

This study clearly demonstrated a benefit in health-seeking behavior for type 2 diabetic patients. Results from this study suggest that patients and practitioners should discuss infections during regular check-ups. Materials (such as leaflets) could be used to increase the implementation of the program in daily practice. Diabetes care providers could instruct people with type 2 diabetes to contact the medical practice when symptoms occur. Just this little effort on the part of diabetes care providers could result in large effects on the burden of common infections in people with type 2 diabetes.

This work was supported by the Public Health Fund (U03/175-P230) and the Dutch Diabetes Research Foundation, the Netherlands (2003.00.031).

We thank the general practitioners and diabetic patients who participated in this study and acknowledge M. Smit for her administrative assistance and J. Box for language review.