Study design and setting

We conducted a multicenter, patient level, randomized trial in the midst of the usual flow of routine primary care practices. The study was originally designed to recruit patients and randomize them to one of two arms: (i) use of the Osteoporosis Choice decision aid during the clinical encounter (Decision Aid), or (ii) provision of the patient’s FRAX estimate only to the clinician (FRAX). After 5 months of enrollment, we added a usual care arm—in which the FRAX score was not made available during encounters—in order to offer a direct comparison against all major practice patterns at that time.

The Mayo Clinic Institutional Review Board approved the study protocol and all study procedures on March 4 2009, when the trial was submitted for registration at clinicaltrials.gov. Unforeseen delays in the release of the registration record due to errors and corrections created an unexpected lapse between approval and registration. The first patient was enrolled in May 2009, and clinical trials.gov released the registration record (Identifier: NCT00949611) on July 28, 2009. Patients and clinicians gave written informed consent.

Study population

Participating practices (Family Medicine, Preventive Medicine, Primary Care Internal Medicine and General Internal Medicine) were all affiliated to the Mayo Clinic (Rochester, Minnesota, USA). Clinicians (i.e., physicians, nurse practitioners) from participating primary care practices were eligible to take part in the study if they provided care for patients with osteopenia or osteoporosis. English-speaking women from participating practices were eligible if they were over 50, with a diagnosis of osteopenia or osteoporosis, were not taking bisphosphonates or other prescription medications to treat their condition, were identified by their clinician as potentially eligible for bisphosphonates, were available for a six month follow up after randomization, and had no major learning barriers.

Eligible patients were identified through lists of upcoming bone mineral density evaluations and from participating clinicians’ appointment calendars, and approached by study coordinators at the time of their appointment.

Intervention and usual care

The intervention in the first arm (Decision Aid) consisted of the use of the Osteoporosis Choice decision aid by the clinician and patient during the clinical encounter (Fig 1). The decision aid included (a) the individualized 10-year risk of having a bone fracture (estimated using the FRAX calculator) with and without use of bisphosphonates (i.e., showing the absolute reduction with bisphosphonates) represented using an evidence-based pictograph and assuming a treatment-related reduction in overall fractures of 40%[10]; and (b) potential harms and other downsides of using bisphosphonates. The study coordinator prepared each decision aid according to each patient’s characteristics prior to the clinical encounter (See S1 Fig for an example of a filled decision aid). Patients and clinicians were to review the decision aid, deliberate about whether to start bisphosphonates, and make a decision together at that time or at a later time. The intervention in the second arm (FRAX) consisted of giving clinicians a copy of the patient’s individualized 10-year risk of having a bone fracture estimated using the FRAX calculator before the visit for use during the clinical encounter. In the third arm (Usual Care), clinicians discussed risk of fractures and treatment as usual without any research-related intervention. No specific guidance was provided to support decisions about non-pharmacological interventions to reduce falls and fractures in any of the three arms.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The Osteoporosis Choice Decision Aid (unfilled).

https://doi.org/10.1371/journal.pone.0128063.g001

Randomization, allocation concealment, and blinding

Patients were the unit of randomization and were allocated using a computer-generated sequence that randomized them 1:1:1 in a concealed fashion. Stratification factors were baseline fracture risk categorization (<20% 10-year risk not on any osteoporosis medication other than bisphosphonates, <20% with a concurrent osteoporosis medication other than bisphosphonates, >20%), practice type (local vs. referral population) and clinician’s likelihood of prescribing medication (low vs. high) assessed from a scenario-based question at the start of the study. The balance of stratification factors was actively monitored and no imbalance was noted across the three arms.

After randomization, only data analysts remained blind to allocation. Patients and clinicians were aware of the overall objective, presented as improvement in communication between patients and clinicians during the clinical encounter, but remained blinded to the specific aims. Research staff at Mayo Clinic monitored patients’ medical records, administered follow-up surveys, and requested pharmacy records.

Possible contamination at the clinician level (i.e. clinician who, having used the decision aid with a prior patient, recreates elements of the decision aid with a subsequent patient allocated to receive FRAX alone or usual care) was monitored by a detailed review of the available video recorded encounters (see Fidelity assessment below).

Data collection and outcomes

Patients and clinicians completed surveys immediately after their clinical encounters, in addition to providing demographic information upon enrollment into the study.

Clinical encounters were video recorded when both the patient and the clinician consented. Medical records were reviewed at patient enrollment and at six months post encounter. Patients completed a telephone follow-up six months later. We also obtained pharmacy records for the period 3 months prior to enrollment to 9 months post enrollment.

We measured patients’ knowledge, decisional conflict, decision to start medication, adherence to medication, involvement in decision making by the clinician, fidelity to intended intervention, acceptability, satisfaction, and quality of life. Decision quality was designated as a secondary outcome.

We assessed patients’ knowledge post-encounter using a 13-item questionnaire that was developed for and used in our previous study, which included questions pertaining to information found within the decision aid (n = 9) and questions regarding overall osteoporosis knowledge (n = 4)[5,9]. We also asked patients to estimate their 10-year risk of breaking a bone at baseline and with the use of bisphosphonates. Patient decisional conflict was assessed post encounter using the Decisional Conflict Scale, the most commonly used outcome measure in decision aid trials[11]. The decision to start bisphosphonates or not was recorded by patient survey immediately after the encounter. We collected patients’ prescription drug and billing data through pharmacy records and assessed adherence to their prescribed medication through 6-month self-reported data and pharmacy records. Medication adherence was evaluated in two ways[12,13]. First, by calculating the percentage of patients who filled their initial prescription (primary adherence). Then, by calculating the percentage of days covered (PDC), defined as the number of days a patient had a supply of each medication divided by the number of days of eligibility of that medication, and the proportion of PDC >80%, for both, those who filled the prescription and all who received a prescription (secondary adherence). We considered medications within the drug class of bisphosphonates interchangeable.

Patients’ involvement in the decision making process by the clinician was assessed by reviewing video-recording of encounters using the OPTION scale, a third-person observer scale designed for use in reviewing audio recordings of primary care visits[14]. We also assessed, by reviewing the video-recorded encounters, the fidelity with which the decision aid was delivered and used as intended during these encounters using the osteoporosis fidelity checklist[5]. This scale is comprised of 10 items (present/absent scale), and results are presented as the percentage of items present. We assessed patients’ and clinicians’ acceptability of and satisfaction with the decision aid using questions that require them to assess the extent to which they would want for themselves and recommend to others similar decision support as they received/provided during the visit. Patients’ quality of life was measured using the EURO QOL5d health thermometer, which assesses patients current health state on a scale of 0–100[15]. Clinicians’ perspective of patient’s perception of the effectiveness of the decision making process was assessed post-encounter using a modified version of the decisional conflict scale effectiveness subscale[11]. Duration of encounters was estimated based on the recorded time of the clinical encounters[5].

Sample size

We powered our trial to have adequate precision in the estimation of the comparative effectiveness of the interventions on knowledge and acceptability, recognizing that rates of start, stop and adherence to bisphosphonates may require joint consideration of this and our prior trial. While we estimated enrolling 150 participants (enough to have 80% power to detect a 30% difference in adherence between arms at 6 months, but underpowered to detect smaller yet important differences), funding run out before we could reach that target. On the basis of the findings from our initial Osteoporosis Choice trial[5], in which we reported a mean patient knowledge score of 3.9 in the usual care arm compared to a mean score of 5.7 for the decision aid arm (standard deviation of 2.5) out of 13 questions, we estimate the accrued sample would provide about 80% power to detect a difference of this magnitude in this trial, and a 0.75 standard deviation difference on any continuous outcome (e.g., decisional conflict scale) across two arms, with a two-sided test and an alpha of 0.05.

Data analysis

We analyzed all patients as randomized, adhering to the intention to treat principle[16]. We present patient and clinician characteristics as counts and frequencies for categorical data and means and 95% CI or medians and interquartile range, as appropriate for data, for continuous outcomes within each of the three arms. The comparison of the arms was conducted using the Chi-square or Fisher’s exact test for categorical outcomes and for continuous outcomes are compared using the t-test or Wilcoxon rank sum test.

Main findings

We found that using the Osteoporosis Choice decision aid was better than usual care with or without FRAX calculation in improving patient knowledge and patient engagement in deciding whether to start bisphosphonates in at-risk women. Using the decision aid changed the length of the discussion, on average, by less than a minute. Clinicians found the decision aid extremely helpful and would recommend it to others. More patients started a bisphosphonate and filled their prescriptions in the Decision Aid arm compared to the FRAX/Usual Care arm; however this was not statistically significant due in part to the relative small number of actual bisphosphonate starts.

Another important finding was that using the FRAX calculator alone as a clinical decision support tool during the encounter was no different than usual care across all measured parameters. The FRAX calculator provides a specific 10-year risk estimate for fracture, a presumed enhancement over risk unawareness or risk communication inferred from the results of bone mineral density tests. Perhaps the addition of an exact number versus a gestalt range from a clinician is not important or not perceived as different to a patient, or perhaps clinicians did not use the FRAX results to communicate quantitatively with patients. It is also possible that the layout and design of the decision aid better fosters communication of the patient’s individual risk; provision of the FRAX score to the clinician does not necessarily mean that a clinician knows how to use that score in a clinical encounter with a patient or that they will actually do so. The decision aid communicates not just the risk of fracture but also quantifies the potential risk reduction with bisphosphonate therapy. The decision aid also brings various patient important issues (i.e., side effects, cost) to the foreground and serves as an invitation for the patient and clinician to address these. These unique features of the decision aid help to create a conversation centered on what is important to the patient and likely facilitated the increase in patient engagement in the Decision Aid arm compared to the FRAX/Usual Care arm.

Our results are consistent with decision aid trials that show improved knowledge and patient engagement and decreased decision conflict with decision aid use. Cranney et. al studied an osteoporosis decision aid in a convenience sample of 20 women and found that knowledge improved from 47% to 83% and decisional conflict was reduced from 50 to 37[17]. In our study, decisional conflict was lower in the decision aid arm than the FRAX/usual care arm, but did not attain statistical significance. However decisional conflict was relatively low in both arms. Our results are also consistent with previous decision aid trials from our group in which we also detected improvements in knowledge, increased patient engagement, and reduced decisional conflict[5,18].

Our study has several strengths. The practices used in our study were usual primary care practices subject to real world time pressures, with usual patients and providers, strengthening the applicability of our results to real-world clinical settings. This was also a patient-level randomized controlled trial that employed adequate randomization, allocation concealment, blinding of data and outcome collectors and data analysts. We used video recording whenever possible to document fidelity and identify and control for contamination, as well as to document the extent to which the intervention, the decision aid, indeed led to shared decision making.

The main limitations of this study, for those who require impact on clinical outcomes to justify investments in shared decision making, are first our inability to determine the impact of this intervention on patient adherence to bisphosphonates, a central issue in osteoporosis care. In our prior trial using a similar tool, we found a significant improvement in the proportion of patients taking bisphosphonates for more than 80% of days prescribed at 6 months [23 (100%) for the Decision Aid arm vs. 14 (74%) for Usual Care arm, p = .009]. In this study, the results were too imprecise to make a judgment. When the two trials are combined in an exploratory meta-analysis, no significant impact is apparent [24 (60%) for the Decision Aid arm vs. 17 (53%) for FRAX/Usual Care arm, p = .96]. Together, these results show no significant effect on adherence with use of the decision aid compared to clinical decision support or usual care across trials, but the sparse data weakens this inference. In addition, clinicians in these two studies had very limited experience in using the decision aid with their patients. We recently showed that clinicians use decision aids as intended only partially and inconsistently, and that the appropriateness of this use was associated with patient knowledge and involvement in the decision making process, thus suggesting an underestimation of the potential efficacy of decision aids when used as intended[19].

The Osteoporosis Choice decision aid improved knowledge and patient engagement and had a high level of patient and clinician satisfaction. Whether this increased patient engagement and involvement translates to higher uptake of bisphosphonates or better adherence remains unclear. Our results suggest this may still be the case, however larger studies are needed to reliably answer this question. To facilitate the exploration of this tool in practice, our group has made freely available an electronic version of the tool for stand-alone use or integrated into the electronic medical record. The tool can be found here: http://osteoporosisdecisionaid.mayoclinic.org