Study Settings
The study was conducted at six teaching hospitals (two in Tokyo, one in Chiba, one in Gunma, one in Nagasaki, one in Saitama) and two teaching clinics (one in Tokyo and one in Chiba) in Japan.
Study Population and Participants
The target population included patients affected by diabetes mellitus, hypertension, and/or dyslipidemia, who were newly suggested to start drug treatment as a medical option by their physicians. The inclusion criteria were as follows: 1) patients aged 20 years or older, receiving outpatient care at an institution where the study was conducted, who were being newly considered by a physician for initiation of treatment with antidiabetic, antihypertensive, or antidyslipidemic drugs (including patients who were previously under continuous treatment with drugs other than the ones used in the study); and 2) patients with a time interval of one week or more between the physician’s recommendation and their decision.
Study Design
We performed a pre-post quasi-experimental study. The study was not designed as a randomized controlled trial because the intervention directly affected patient-physician communication, and the validity of data collected from the control group would not be ensured without dividing the study into two distinct periods: a pre-intervention period, in which the physician was not affected by the intervention and an intervention period. During the pre-intervention period, patients received a booklet called “Guide to Shared Informed Consent Between Patients and Healthcare Professionals” (hereinafter referred to as guide) that explained the principles of informed consent (IC). After data collection for the six-month pre-intervention period was completed, a 12-month intervention period in which the intervention was administered to the enrolled study participants was launched.
Intervention
Similar to the pre-intervention period, patients were given the guide for study participants during the intervention period. Simultaneously, patients were asked to fill out a “medical decision-making support template” (hereinafter referred to as template) to express their preferences and values to the physician in charge for reaching a decision about starting drug treatment in writing. The patients were asked to express themselves freely on five items. An example of how to fill a support template is shown in Figure 1, alongside the five items.
Patients could either fill the template on the same day they received it, after receiving outpatient care, or fill it at home and mail it to the hospital within two weeks from the date of consent acquisition. The researcher attached the returned completed template to the patient’s electronic medical records. The physician in charge of the patient was then advised to examine the content of the template before beginning the next consultation.
Outcomes and Independent Variables
The primary outcomes for evaluating the intervention’s effect were the decisional conflicts and regrets elicited by the patients’ decisions. The Decision Conflict Scale (DCS), developed by O’Connor and translated into Japanese by Kawaguchi et al., was used to assess the patients’ decisional conflicts [22,23]. The Decision Regret Scale (DRS), developed by Jamie and translated into Japanese by Tanno et al., was used to assess the patients’ decision regrets [24,25].
We evaluated two decision-making statuses three months after study enrollment as secondary outcomes. First, we assessed the percentages of patients who already started or decided to start drug treatment, those who refused to start it, and those who postponed their decision. Based on a chart review of the patients’ medical records, the patients’ decision status was classified as “decided starting drugs” if they had already started or decided to start treatment with the prescribed drug. Conversely, it was classified as “decided NOT starting drugs” if they refused to start treatment with the prescribed drug. In cases where patients could not decide, their status was classified as “still considering.” Second, we observed the concordance between the physicians’ recommendation for starting drug treatment at the time of enrollment and the patients’ actual decision three months after enrollment. During enrollment, physicians were asked to answer how much they recommended starting drug treatment by selecting one of the following five options: 1) “strongly recommend,” 2) “would rather recommend,” 3) “medically neutral,” 4) “would rather not recommend,” and 5) “do not recommend.” Based on the answers provided at enrollment, cases in which options 1 or 2 were chosen were listed as “recommend drug,” while cases in which options 3, 4, or 5 were chosen were listed as “not recommend drug.” Three months after enrollment, the physicians’ recommendation recorded at enrollment and the patients’ actual decision recorded three months later were defined as “concordant” if patients’ medical records for cases listed as “recommend drug” had a status of “decided starting drugs,” or if their medical records listed as “not recommend drug” had a status of “decided NOT starting drugs” or “still considering.” Otherwise, the physicians’ initial recommendation and patients’ actual decisions were defined as “discordant.”
The following information was collected at enrollment: patients’ gender and age, whether they lived alone, whether they visited the outpatient clinic with family members, the name of the disease for which drug treatment was to be started, and whether they were continuously taking other medications alongside the newly recommended drug. We also collected data on the patients’ health locus of control as a confounding variable that may affect the primary outcome, as reported in past literature [26,27].
Data Collection
We used three methods of data collection: a questionnaire survey for patients, a medical record survey, and a survey for the physician in charge. These surveys were administered to both patients and their physicians by the researchers at each institution during enrollment. Patients were asked to complete a questionnaire that included a baseline DCS and the Japanese version of the health Locus of Control scale (LOC) [28,29]. Physicians were asked to rate their degree of recommendation for prescribing the target drug. Two months after enrollment, the researchers distributed questionnaires containing DCS and DRS to patients by mail. Patients mailed their completed questionnaires to the research supervising institution. The baseline LOC scale scores were calculated for the 6 domains included, with a minimum score of 5 points and maximum of 30 points. The 5 subscales score and total score of the DCS measured at baseline and 2 months later were scaled from 0 to 100 points.
Three months after enrollment, researchers at each institution collected data on secondary endpoints based on a chart review of the patients’ medical records. Each institution started enrolling the pre-intervention group after receiving approval from the Ethics Committee. After completing the six-month pre-intervention period, the 12-month intervention group was established. The pre-intervention enrollment period was from September 2016 to August 2017 and the intervention enrollment period was from March 2017 to March 2018.
Statistical Analysis
After data entry and collection were completed, propensity score matching was done after performing logistic regression analysis with patients’ age, sex, whether they lived alone, whether they were accompanied, and whether they were already prescribed medications. LOC/DCS subscale scores at baseline were confounding variables, and control/intervention was the dependent variable. Match tolerance was set to 0.05. A Student T-test was conducted to compare the DRS and DCS mean scores between the two matched groups. The two secondary endpoints were analyzed for frequency comparison with the chi-squared test.
The primary outcome used for the sample size estimation was the comparison of the DRS mean score. Based on the Japanese version of the DRS, the clinically expected effect size was set to 0.5. The required sample size was estimated using a two-tailed test with a power of 0.8 and significance level of 0.05. Consequently, we estimated that 128 patients (64 each in the control and intervention groups) would be needed.