Design
We will evaluate the PORTDA-diab in a randomised pre-post intervention trial using a 2×2 factorial intervention design with a control group (see flowchart in Figure ). The factorial design is considered the most efficient method to study factors which may influence the outcomes. The inclusion of a control group receiving usual care but undergoing pre-post intervention measurement will control for Hawthorne effects as well as for independent effects or changes in the study population.
Study setting and participants
We will test the intervention in routine primary care settings in The Netherlands. Primary care practices will be recruited in the Groningen region where a diabetes disease management program has been implemented. Most practices have a practice nurse who conducts quarterly consultations with the diabetes patients. In The Netherlands, it is common that primary care physicians delegate these tasks of chronic care. Practice nurses are trained to conduct practice examinations, risk assessments, patient education, and counselling. They can make medication treatment changes which have to be approved by the physician. To ensure a minimum level of communication skills training in all practices and increase the study participation rate, providers are offered a training session in motivational interviewing before the study starts. In case they already have followed such a course, they are offered €250 as alternative compensation for participation. In addition, all participants will receive a 2-h training in risk communication and an instruction video with simulated consultations showing ‘good and bad examples’ of applying four basic principles of risk communication, that is, to use natural frequencies, use positive and negative phrasing, explicit uncertainty, and be open and refrain from imposing options. All practices use electronic medical record systems supporting structured care protocols. Furthermore, all practices receive yearly performance reports as part of the regional monitoring of diabetes care. We will thus evaluate the additional effect of using the PORTDA-diab over the current disease management program.
Eligible patients include people with type 2 diabetes managed in primary care. Excluded are patients who had a myocardial infarction within the preceding year, experienced a stroke, suffer from heart failure, angina pectoris, or have a terminal illness, and patients who were above 65 years of age at diagnosis, because the calculated risks and treatment goals in these patients are not sufficiently evidence based. After identification of eligible patients in the electronic medical record system, the healthcare providers will confirm that the selected patients satisfy these criteria. In addition, they will exclude patients with dementia or cognitive deficits, who are blind or not able to read Dutch, since such patients are not expected to benefit from this type of intervention. Practices will recruit patients by distributing information packages containing an invitation letter, information about the project, and an informed consent form. Patients can contact the study coordinator for additional information about the study or an independent general practitioner to discuss questions regarding participation. Patients will be offered €10 in compensation for time spent to fill in the study questionnaires. Invitation letters will be followed up by telephone contact by the research team when needed.
Treatment allocation
For pragmatic reasons, participating practices will be randomly allocated to the paper-based or computer version. This randomisation will be stratified by practice size (below or above 2,500 patients) and organisation (single or more primary care physicians), using a 1:1 computer-generated allocation sequence within the strata (Figure ).
Within each practice, consenting patients will be randomised to receiving: (1) the PORTDA-diab presenting only CHD outcomes; (2) the PORTDA-diab presenting additional outcomes; or (3) the control group, using a blockwise scheme to conceal the allocation process for the healthcare provider. Although this design of randomising patients within practices has the disadvantage of potential contamination at provider level, it ensures that variations in communication skills and practice organisation are balanced between the intervention and control group. In addition, low participation and high drop-out rates can be expected when practices are randomised to a non-intervention control arm. It is not possible to use the PORTDA-diab for other patients than the intervention patients. It is, however, possible that providers will use some of the aspects learned from working with the novel approach in their usual care.
Intervention
The intervention consists of: (1) PORTDA-diab software installed on a computer linked to the electronic medical record system enabling up-to-date data retrieval of the patient’s clinical situation and treatment status; (2) a set of treatment cards which can be used during consultation, summarising positive and side effects of available treatment options; and (3) a short instruction protocol for the healthcare provider about the decision aid and treatment cards (Table ). The decision aid offers personalised information on treatment goals and options to patients immediately before they consult their healthcare provider. The same information will be available to the healthcare provider for further use during consultation. In practices randomised to the computer version, PORTDA-diab will be installed on a 17 inch laptop or workstation that can be used by the patients. In practices randomised to the paper-based version, a printed version of the personalised information generated by the PORTDA-diab software will be offered to the patients at the practice. Intervention patients will be asked to come to the practice 15 min in advance to go through this additional information. All intervention patients should receive the printed version at the end of the consultation. The short version of the PORTDA-diab presents treatment effects on reducing the risk of getting a myocardial infarction, and comprises five pages or four screens. The extended version covers seven pages or eight screens, and presents treatment effects also on other complications which affect functional activities that are of concern to patients, that is, the risk of getting a stroke, amputation, blindness, and renal failure [
9,
57,
58].
| Table 1Instruction protocol for using the PORTDA-diab and treatment cards |
The information presented will be generated automatically integrating routinely registered information from the electronic medical records as well as evidence-based information on diabetes treatment and outcomes as summarised in the Dutch Primary Care Guidelines on type 2 Diabetes [
59]. The patient data are extracted real time from the medical records but can be corrected or completed by the provider when needed. The information starts with a summary of the patient’s current situation and treatment (Figure ). The patient’s risk information is presented using bar graphs with simple explanations of critical treatment goals and treatment options [
60]. Using the most recent information on the patient’s risk factor levels and the UKPDS risk engine algorithm [
61], the patient’s overall risk of getting a myocardial infarct within 5 years is calculated. This overall risk will be presented in relation to potential improvements and to a reference group (Figure ). Such a reference is important since people can misperceive their initial risk. The patient’s current risk is thus compared to his/her expected risk when achieving the recommended treatment goals for all risk factors (‘optimal treatment’). At the same time, it is compared to the risk of a patient with similar patient characteristics but without diabetes, as calculated by the Framingham risk score [
62]. Should the Framingham risk score be higher than the ‘optimal treatment’ risk, the former will be equated to the latter. Next, the impact of achieving recommended goals for separate risk factors will be presented together with possible treatment options. Only risk reductions are presented for risk factors that are not yet at target level. These target levels are derived from the Dutch Primacy Care Guidelines, being <53 mmol/mol (<7%) for HbA1c, <140 mmHg for systolic blood pressure, <2.5 mmol/L for LDL-cholesterol, and not smoking [
59]. Patients are explicitly encouraged to think about which of the risk factors they prefer to address first. This will enable prioritisation that may be needed when several changes in treatment are required. In the short version, only the impact on the risk of getting a myocardial infarction is presented (Figure ). In the extended version, Figure is replicated presenting the impact on the other outcomes [
63]. For myocardial infarction and stroke, bars will be visible only if controlling the corresponding risk factor will lead to an absolute risk reduction of 1% or more. For amputation, blindness, and renal failure, bars will be visible at or above an absolute risk reduction of 0.1%. These cut-points are chosen because patients in the pilot phase mentioned that smaller reductions were not meaningful or relevant for them.
Treatment options and treatment cards
On the final page or screen, the treatment options are summarised for those risk factors not yet at target level. Both lifestyle and medication options can be presented. The medication options are tailored to whether or not the patient is already treated and/or is already receiving maximum treatment. Maximum treatment is defined per risk factor as: using insulin with or without additional oral glucose-lowering medication, using three or more antihypertensive drugs, using a statin at maximum dosage. Possible pros and cons of the options are summarised. In addition, the healthcare providers can use a set of treatment cards during consultation to discuss and compare the benefits and risks of specific treatment options (Figure ).
Control group
Patients in the control group will receive care as usual, including any education or information as deemed necessary by the healthcare provider.
Effect modification
Effects of the decision aid may depend on characteristics of the patients that affect their involvement, understanding, and patient-provider communication about disease management. Therefore, the following characteristics will be included as possible modifying factors: age, gender, educational level, socioeconomic status, duration of diabetes, number and type of risk factors present, co-morbidity, and number of drugs used.
Outcome measures
In correspondence with the aim of the decision aid, our primary outcome is empowerment of patients for making shared decisions about setting and achieving treatment goals [
18]. As secondary outcomes, we include a set of additional patient outcome and process of care measures. These will provide insight into possible related positive and negative effects and a better understanding of how the novel approach may affect diabetes care.
Primary patient outcome:
a) diabetes empowerment (Diabetes Empowerment Scale, subscale ‘Setting and Achieving Goals’ [
64])
Secondary patient outcomes:
b) satisfaction with diabetes care (PEQ-D [
65])
c) beliefs about medication (BMQ [
66])
d) negative emotions (PAID [
67])
e) health status (EQ-5D [
68])
f) predicted coronary heart disease risk [
61])
Secondary process of care outcomes:
g) percentage of patients with (intensified) antihypertensive treatment after insufficiently controlled blood pressure levels
h) percentage of patients with (intensified) glucose-lowering treatment after insufficiently controlled HbA1c-levels
i) percentage of patients treated with lipid-lowering drugs
j) percentage of patients with (micro)albuminuria treated with a RAAS inhibitor
The Diabetes Empowerment Scale consists of three subscales measuring different aspects of patient empowerment: managing psychosocial aspects of diabetes (9 items); assessing dissatisfaction and readiness to change (9 items); and setting and achieving goals (10 items) [
64]. It has been translated into various languages, and is considered a suitable tool in evaluating empowerment-based education programmes. The subscale on setting and achieving goals will be included as primary outcome. The Patients’ Evaluation of Quality of Diabetes Care is a 14-item scale assessing patients’ judgements about the quality of their diabetes care, which was found to be suitable for care delivered by physicians as well as nurses [
64]. The Beliefs about Medication Questionnaire comprises two sections: the BMQ-Specific, which assesses representations of medication prescribed for personal use, and the BMQ-General, which assesses beliefs about medicines in general [
66]. It has been translated into Dutch and used in many studies. The BMQ-Specific assesses beliefs about the necessity of and concerns about prescribed medication. The BMQ-General assesses beliefs that medicines are harmful and overused. The Problem Areas in Diabetes (PAID) is a 20-item questionnaire available in Dutch, assessing diabetes-specific emotional functioning [
67]. The EQ-5D measures the general health status on five dimensions: mobility, self-care, daily activities, pain and symptoms, and negative emotions, and by a VAS-scale measuring overall health status on a scale from 0 to 100 [
68].
Process evaluation
In the post-intervention questionnaire, all patients will be asked to complete a checklist, which assesses contamination of control patients and exposure of intervention patients to the intervention. Intervention patients have to rate how easy or difficult they rate the decision aid information on a five-point Likert scale. Patients allocated to the computer version will in addition be asked how easy or difficult they rate its navigation properties. Since they will receive the paper version after the consultation, they will be asked for a comparison of the two versions. Providers’ experiences with the decision aid will be assessed in qualitative interviews using a topic list based on previous assessments regarding the implementation of innovations in practice. To assess adherence to the intended use (treatment fidelity), the actual use of the decision aid before and during consultations will be logged automatically in the computer version. In addition, all providers will be asked to complete a short checklist about the PORTDA-diab and the (shared) decision-making process after each consultation with an intervention patient.
Data collection
Patient data will be collected with structured questionnaires and automated extraction procedures from the medical records. Patients’ empowerment, beliefs, satisfaction, and perceptions will be measured using mailed surveys that all patients will receive in the month before and 3 to 4 months after a scheduled quarterly consultation for diabetes. Data on actual management and clinical outcomes in the 6 months before and after the intervention period will be collected from the Groningen Initiative to Analyse Type 2 diabetes Treatment (GIANTT) database. This GIANTT database includes longitudinal data collected from primary care medical records using validated automated extraction procedures [
69].
Sample size
Sample size was calculated using the primary patient outcome, that is, change in diabetes empowerment scale (DES). Since there is no published minimum level of change considered to be relevant [
70], we used a change of 0.2 in the DES score to estimate the sample size. With an expected range in its baseline value of 3.2 to 4.0, this represents a 5% to 6% change in the DES score. To be able to detect an absolute change of 0.2 in the DES score, a total of 150 patients per study arm is needed to achieve 80% power at 5% significance (standard deviation (sd) =0.62). It is therefore estimated that we will need a sample of 18 to 20 GPs, each with at least 20 to 30 participating diabetes patients. This number is expected to be sufficient to detect differences of 0.36 on the PEQ-D (sd=1.1), 6.5 on the PAID (sd=20.0), 0.26 on the BMQ (sd=0.8), and 5.0 on the UKPDS risk estimated (sd=15.5). These are all conservative sd estimates derived either from the original validation studies or available data on the UKPDS risk scores for the GIANTT population. Primary care practices in the recruitment region have on average around 45 type 2 diabetes patients eligible for this study.
Statistical analysis
Descriptive statistics will be used to summarise demographic, clinical, and other patient measures. The primary analysis will test pre-post intervention changes in patient outcomes in the combined intervention group as compared to the control group using Student t-test and Wilcoxon’s two-sample statistics. Furthermore, we will examine differences in primary and secondary outcomes stratified by presentation format and medium in each intervention arm. For the process of care outcomes, z-approximation statistics will be used. Data will be analysed on an intention-to-treat basis, and results of two-sided tests between study arms will be regarded significant at P<0.05. We will explore the influence of the practice on the outcomes by multilevel modelling. Also, for hypothesis generation, interactions will be tested to identify possible factors at patient level that might modify the effect of the intervention.
Ethical approval and trial registration
The study will be conducted in accordance with the recommendations provided in the Dutch Code of conduct for health research. The medical ethics committee of the University Medical Centre Groningen approved the study (ARB number NL29042.042.09). The GIANTT project and its data collection and database are registered at the CBP (College Bescherming Persoonsgegevens, number 1250778). The trial is registered at the Dutch Trial register NTR1942 with the acronym PORTDA-diab.
1,2 Mathijs Dun,1,2 Jan Schuling,3 Flora M Haaijer-Ruskamp,1,2 and Jaco Voorham1,2
tailors the information on treatment goals and options to the individual patient,
