E-healthcare for diabetes mellitus type 2 patients – a randomised controlled trial in Slovenia

This was an interventional, randomised controlled study of patients with DM type 2, treated only with a diet regime and/or tablets. Slovene diabetics treated with insulin are usually treated by diabetes care units, and were therefore not included in this study.

Patients were selected from 22 solicited family practices from 6 different regions in Slovenia (Posavje, Zasavje, Štajerska, Gorenjska, Primorska, and Ljubljana). To be included, family practices had at least 1,000 patients, a nurse with secondary-school training, and a qualified nurse with higher education (the diabetes care coordinator).This cluster sampling was chosen to give the best estimate for the Slovenian population of patients with DM type 2 not using insulin.

A computerised randomisation programme assigned patients to the interventional or the control group through a balanced randomisation process using the last four patients. Randomisation was carried out for all practices simultaneously, but not at the level of individual practices. Due to its design, the study could not be blinded; the staff at each practice was aware of patient allocation.

The randomisation process is presented in Figure 1.

Figure 2

The inclusion period was 3 months and the next criteria were the same for both groups:

between the ages of 18 and 75;

The time frame for the follow-up of each patient was one year.

Patients randomised in the control group had the conventional care for diabetic patients, according to Slovenian professional guidelines. Patients in the interventional group had, in addition to conventional care, access to the eDiabetes application.

Nurse practitioners could comment about measurements made by patients in the interventional group and give advice about lifestyle. General practitioners gave advice about medications. Some medications, but not a significant number, were altered during the study period for both groups.

The comparison between the diabetes care offered to the control group and the interventional group is shown in Table 1.

Table 1

Comparison between conventional and eDiabetes care.

The eDiabetes application, designed in collaboration with experienced ICT experts, consisted of a web portal for patients and health-care providers (26), with a patient-oriented interface for individualised care, and a web server for a repository and education material (http://eoskrba.si). The educational material included informational handouts, web addresses, articles, and instructions to help diabetic patients better manage their condition.

Every two weeks patients in the experimental group recorded data, including their body weight, blood pressure, diet, and physical activity. The last parameter was evaluated on a 7-point Likert scale. Every 6-8 weeks patients completed the COOP-WONCA charts. These are comprised of seven parameters designed to measure functional health status using words and pictures to represent and assess: physical fi tness, feelings, daily activities, social activities, change in health, overall health, and pain. These charts are rated using a five-point scale, with higher scores indicating a worse functional status (3, 4).

The application automatically sent users reminders by simple e-mail and SMS: “Please enter your blood sugar/or other parameters/ into the eDiabetes application”. These reminders were sent if a user had not entered body weight, blood pressure, physical activity, and diet data within 2 weeks of the deadline or not completed the COOP-WONCA questionnaire within 8 weeks. Reminders were also e-mailed to the diabetes care coordinators.

In case of medical emergencies, a warning was issued to immediately contact the medical emergency services. More than 1000 SMS and e-mail reminders were sent during the study. E-mail communication between the patients and family practices was less common, with fewer than 100 such e-mails registered during the study. Nurse practitioners and physicians mostly commented directly with patients on the eDiabetes inputs during office visits.

The primary outcome measure was the change from baseline of HbA1c at 1 year. This outcome was calculated as the difference between baseline and final HbA1c for both groups.

The secondary outcome measures were the change from baseline of: (a) HbA1c at 6 months, (b) Body Mass Index (BMI) at 1 year, (c) Patients’ Functional Health Status indicated by the COOP-WONCA Questionnaire at 6 months and 1 year, (d) blood lipids at 1 year, and (e) diastolic and systolic blood pressure at 6 month and 1 year.

The minimal clinically important difference (MCID) was set to one standard error of the mean in the outcome score (27, 28).

A sample size of 60 patients per group was necessary to achieve 80% power at a two-sided 5% significance level and a hypothetical dropout rate of up to 15%.

No adjustment of the sample size was made for the potential clustering of scores within a single practice.

The statistical analysis of data collected in the eDiabetes application and from the patient health record was performed using SPSS Statistics 21 and the EXCEL version of Microsoft Office 2010. The unequal variances t test was used to assess changes of the target interval outcomes between the two groups at baseline and within each group at 1 year. The Wilcoxon rank sum test was used to assess differences in the COOP-WONCA results (29). Multivariate analysis was used to find factors that might be related to changes of HbA1c. The degree of statistical significance was defined as p<0.05.

In terms of the distribution by age, gender, medication and diabetes duration in years, the computerised randomisation process resulted in the almost ideal allocation of patients in both study arms (p>0.05). Despite some evidence about lower Internet use among less educated and older diabetic patients (9, 14, 19), these two sub-groups were adequately represented among those meeting the inclusion criteria. Statistically significant differences (p<0.05) between patients in interventional and control group were found in 2 groups with highest education. Statistically important differences were not found when merging participants from the “College or university” and “Master, PhD, or specialisation” subgroups in one group, which means that distribution of participants by level of education could not affect the results.

The main barriers to more extensive enrolment in the study and better adherence of enrolled patients were: low motivation of practitioners and nurses in some family practices, patients’ lack of computer and/or access to the Internet (n=29), and patients’ lack of sufficient e-skills (n=17). The importance of PHC provider adherence to different forms of web-based diabetic care is well documented in the literature (5, 13, 16-18, 20-22).

The results of this research confirmed the hypotheses about improvement in treatment outcomes compared to usual healthcare, and rejected the presumption of eDiabetes application impact on the functional health status of patients with DM type 2.

A significant negative correlation found between the number of inputs to the application and the average values of self-recorded systolic blood pressure (p=0.004), as well as to the EHbA1c-3 values (p=0.006) showed that a higher number of inputs correlated with lower final HbA1c and SBP values. This can be understood as confirmation of the impact of application on the improvement of key parameters and, consequently, as the most important result of the study.

The result of the regression analysis – the list of four significant predictors that help explain 81.1% of the final HbA1c values – is also a valuable part of this study. The connections between the final and initial HbA1c values, as well as between self-measured blood glucose and final cholesterol levels, were expected.

The main limitation of this randomised controlled trial was poor adherence of some primary-care practices and, consequently, a smaller number of enrolled patients in both groups.

However, the strong correlation (p<0.001) between the final HbA1c values and the average self-measured systolic and diastolic blood pressure values was surprising.

There were very few similar reports in the reviewed literature, although significant changes in serum lipids have often been reported (5, 6, 11, 12, 14, 20, 22).

Statistically significant reductions of the HbA1C level in the interventional group, at the first follow-up and the final check-up, were already reported in several previous meta-analyses and systematic reviews (5, 6, 11, 12, 14, 20, 30). However, the patient population of this trial had well-regulated diabetes with slightly elevated baseline HbA1c (7.1% for the interventional and 6.8% for the control group).

The findings of the lower HbA1c and self-measured fasting blood glucose (SELF-BG) for the interventional group, compared to initial values and to the control group, were consistent with previous studies (5, 11, 12, 19-22, 31).

Despite the improvement of laboratory results in the interventional group, there is no clear impact of separate factors, such as nurse practitioners or diabetes care coordinators, educational materials accessible from the application, SMS and e-mail remainders.