Background
Although only 10 % of Americans ≥ 50 years old are currently diagnosed with osteoporosis (OP) based on dual energy x-ray absorptiometry (DXA) testing [
1], the lifetime risk of a low-impact fracture is 40 % for older women and 13 % for older men [
2]. Hip fractures, related to OP, are associated with increased mortality and serious adverse effects on quality of life (QOL) and health care costs [
3]. Many older adults know very little about their OP risks or the rationale for screening to identify this silent disease [
4,
5]. As suggested by health behavior theories, like the Health Belief Model [
6], knowledge of OP, its risks and how to prevent or improve OP, is a first step for patients in making bone health-related behavior changes (i.e. getting adequate amounts of calcium, vitamin D and weight-bearing exercise, taking appropriate pharmacotherapy, preventing falls).
Some randomized controlled trials (RCTs) have demonstrated that educational interventions improve OP-related knowledge [
7‐
16]. Less is known about whether such interventions affect patient satisfaction or QOL [
17,
18]. Most of these RCTs examining OP patient-education interventions have been multifaceted or lengthy [
7,
12,
13,
15,
16], which are not scalable. To date, only one RCT has examined the effect of an OP education intervention on QOL, reporting a beneficial effect after both three months and one year [
16]. To our knowledge only three studies have shown that patients receiving educational or quality improvement interventions were more satisfied with their bone-health care than usual care groups [
16,
19,
20]. However, two of these studies were conducted only with older women with an OP diagnosis [
16,
19]. The other study only saw improved satisfaction for timeliness in test result notification but not in other measures of satisfaction (e.g. understanding DXA results or treatment options) [
20]. Because tailored interventions individualized to the patient’s characteristics are more effective and preferred by patients than standardized interventions [
21,
22], we developed a tailored, pragmatic patient-activation intervention.
We reported that 90 % of older adults wanted to receive their DXA results by mail [
23]. Therefore, we developed a DXA result letter and a bone-health educational brochure. We then designed and conducted a pragmatic RCT to assess the impact of this intervention on the pathways leading to appropriate pharmacotherapy, health behavior change, and satisfaction with bone-health care, QOL, OP knowledge, and cost-effectiveness (
www.ClinicalTrials.gov Identifier NCT01507662). This article describes the effects of the intervention on three patient-reported outcomes: satisfaction with bone health care, QOL, and OP knowledge at 12 and 52 weeks post-baseline. We hypothesized that the intervention would improve bone-health care satisfaction and OP knowledge at both time points, but would not change QOL.
Methods
Participants
Patients ≥ 50 years old presenting for DXA between February 2012 and August 2014 at the University of Iowa (UI), University of Alabama at Birmingham (UAB), and Kaiser Permanente of Georgia (KPGA) were invited to participate. We excluded non–English speakers and prisoners. Twenty dollar gift cards were provided after the baseline interview.
Design and randomization
We used a double-blind, parallel, pragmatic RCT [
24]. After patients completed their DXA and baseline interviews they were randomized based on their providers using a computer program written in R [
25]. Providers were first ranked within sites based on the number of DXAs they ordered in the previous two years, and then they were randomized within sites using blocks of three into three groups (1:1:1 allocation ratio). Patients of providers in the first group were assigned to the intervention arm, patients of providers in the second group were assigned to the usual care arm, and patients of providers in the third group were randomized to either the usual care or intervention arms (1:1 allocation ratio). We selected three provider randomization groups to assess potential spill-over effects on the main outcomes for usual care patients of providers in the third group.
Procedures
At baseline, research assistants (RAs) at each site used REDCap™ [
26] computer assisted interviewing (CAI) software to interview patients up to four weeks before or three days after their DXA. All KPGA patients and half of the UI patients completed these interviews in person. All UAB patients and the remaining UI patients completed their baseline interviews over the telephone. Three RAs at UI mailed study materials to intervention patients.
Intervention
Intervention materials included a letter describing results of their DXA (lowest T-score and interpretation [OP, low BMD or normal]), a graphic portrayal of their 10-year probability for a major osteoporotic fracture (using FRAX®;
https://www.shef.ac.uk/FRAX/), and a bone-health educational brochure. These materials have been described elsewhere [
27,
28].
Outcomes
Twelve weeks and 52 weeks after their DXA, Iowa Social Science Research Center interviewers telephoned patients at all three sites to conduct follow-up interviews using WinCATI 4 · 2 and 5 · 0 CAI software (Sawtooth Technologies, Northbrook, IL). Interview questions have been described elsewhere [
24].
OP Care Satisfaction. This five-item scale assesses patient satisfaction with notification and understanding DXA results, understanding OP treatments, receiving adequate information to make an informed decision, and overall satisfaction with bone-health care. Response options ranged from strongly agree to strongly disagree, with summary scores ranging from 5 (least satisfied) to 25 (most satisfied). Four of these items were used in a previous study [
20], with a fifth item related to overall satisfaction added for the current study. Because this was the first use of the satisfaction with OP care scale after its development and initial publication, we used exploratory factor and reliability analyses to explore its psychometric properties. Those results revealed a simple factor structure that was unidimensional with principal factor loadings for each item ranging from 0.53 to 0.77, and an internal consistency reliability (alpha) coefficient of 0.77. At baseline, these items were only asked of patients with prior DXAs because they were irrelevant for DXA naïve patients. All patients were asked these questions at the 12- and 52-week interviews.
Quality of Life. We used three QOL measures. The first was the SF-1 (“In general, would you say your health is excellent, very good, good, fair, or poor”), which was scored as 95, 90, 80, 30, and 15 to reflect the underlying health utilities [
29]. The second QOL measure was the EQ-5D-3 L which has five items assessing difficulties with mobility, self-care, activities of daily living, pain, and mood (α = 0 · 70) [
30,
31]. Responses (no difficulties, some, or were completely impaired) were converted to health utilities ranging from 0 (death) to 1 (best health state) [
30,
31]. The third QOL measure was the EuroQol Visual Analog Scale (VAS) which asks patients to rate their health from 1 (worst health state they can imagine) to 100 (best health state they can imagine) [
30,
31].
OP Knowledge. We used the 10-item “Osteoporosis and You” scale [
5,
32] to measure. The five responses ranged from strongly agree (SA) to strongly disagree (SD), which were collapsed into “correct” or “incorrect” responses. Correct responses (SA or A for true or SD or D for false statements) were coded “1” with incorrect responses coded “0”. We summed the responses into a total score (α = 0 · 68). We also examined the subscales (biological, lifestyle, consequences, and prevention and treatment).
Statistical considerations and analysis
PAADRN was powered for guideline concordant treatment as the clinical endpoint. Therefore, for the current analysis, we calculated the statistical power that we would have to detect a standardized effect size of 0.10 with p < 0.05 and attrition from baseline as high as 20 %. Those calculations indicated that we would have 91.3 % power to detect such differences. We used multiple imputation techniques to account for missingness (lost to follow-up, patient refused a specific question or responded “don’t know”). We imputed each item separately and constructed the outcomes based on the imputed values. Our primary analysis was based on intention-to-treat (ITT).
We first compared the outcome measures between the intervention and control groups at baseline and at the two follow-ups using t-tests. We then used linear random effects regression methods to adjust for patient clustering within provider and for pre-specified covariates. For patient satisfaction, we first examined differences between intervention and control groups at 12-weeks and 52-weeks (baseline patient satisfaction was only asked for those with prior DXAs). Among those with prior DXAs, we adjusted for baseline satisfaction in separate models. For QOL and OP knowledge we examined differences between baseline and 12-weeks, and baseline and 52-weeks. We used Bonferroni methods to adjust for testing at two time points (12- and 52-weeks) and for using three QOL measures. We examined minimally important differences (MIDs) defined distributionally as improvements ≥ 0.5 standard deviations (SD) [
33]. For the EQ-5D utility score we also used an anchor-based approach to predict utility scores for the pairwise SF-1 comparisons (adjusting for age, gender, and race).
We also investigated pre-specified heterogeneity of treatment (HTE) effects. These included median splits on preferred approaches to health care decision-making and treatments [
34], those with prior DXAs vs. those without, those on OP-medications at baseline vs. those who were not, those with a history of OP or osteopenia at baseline vs. those who did not, site (UAB vs. KPGA vs. UI), age (<65 vs. 65-75 vs. > 75), men vs. women, Whites vs. non-Whites, education (high school or less vs. some college vs. graduate school), self-rated health (poor vs. fair vs. good vs. very good vs. excellent), having COPD, depression, or prior fracture at baseline (vs. not), FRAX risk (low vs. moderate vs. high), current smoker vs. former smoker vs. never smoked, heavy vs. moderate alcohol consumption, and median splits on weight-bearing exercise.
In sensitivity analyses we used case-wise deletion instead of multiple imputation. Because those results were entirely consistent with the results presented below that used multiple imputation, we only report the latter here. With the Bonferroni adjustments, all p-values were 2-tailed with ≤ 0.025 deemed statistically significant for patient satisfaction and OP-related knowledge, and < 0.0083 deemed statistically significant for the three measures of QOL. Analyses were performed using SAS 9 · 4 (SAS Institute Inc., Cary, NC).
Discussion
There is growing interest in engaging patients in their own healthcare [
35,
36]. Tailoring health communication to be more patient-centered is becoming more common. Yet, it is unclear whether greater access to tailored, DXA testing communication results in any measurable improvements in patient reported outcomes. We designed a, pragmatic, multi-site, RCT to evaluate the effects of a tailored DXA result letter accompanied by a bone-health educational brochure on patient satisfaction with OP care, QOL, and OP knowledge. Our results revealed significantly improved patient satisfaction with OP care in the intervention group compared to the usual care group. Intervention patients had 58 % greater odds of improving by at least an MID (0.5 SD) at 12-weeks (
p < 0.0005) and 34 % greater odds off improving by at least an MID at 52-weeks (
p < 0.005). However, we found no differences in terms of QOL or OP knowledge between the intervention and usual care groups.
These findings are important because, to our knowledge, this is the first positive study to include a comparison group, and to include patients with and without OP. Patient satisfaction is recognized as an important dimension of healthcare quality. Medicare now evaluates patient satisfaction [
37] which will soon be used in determining preventive service reimbursements for doctors and hospitals. The intervention materials were pilot tested to ensure comprehension as well as patient preferences for information and design [
27,
28]. Tailoring test result communications to patients may improve their satisfaction with other types of testing as well.
As important as the effect on satisfaction with OP care is, failure to improve QOL or OP knowledge in our study also must be considered. We did hypothesize that the patient-activation intervention would not affect overall QOL because OP care is a small component of the healthcare received by older adults who may have several comorbidities. Indeed, OP would likely have minimal effects on QOL in the short-term until a fracture occurs, at which point profound effects on QOL. Thus, our results are consistent with prior studies of OP patient education interventions on QOL [
11,
38], in which only one reported a significant improvement among Malaysian women taking bisphosphonates [
16]. Moreover, trials of OP therapies, which demonstrate reduced fracture rates, seldom are powered to detect an effect on QOL for some of the reasons noted.
The absence of an effect of the patient-activation intervention on OP knowledge is surprising and contrary to our expectations. Prior OP-education interventions have reported significant improvements in knowledge [
7‐
16]. In contrast, we found that OP knowledge significantly increased in both intervention and usual care groups, but that the magnitude of these improvements was the same. This may be due to the measure of OP knowledge. First, the reliability of the OP knowledge measure was only marginally acceptable (α = 0.68) [
39]. Second, this was the first RCT to use the “Osteoporosis and You” measure, and the two prior observational studies assessing its psychometric properties included younger women and not men [
5,
32]. Third, significant practice effects (about half of an additional correct answer) were observed and these may have created a ceiling effect that constrained our ability to detect short-term differences. Finally, the null effect may be due to the fact that neither the patient-activation letter nor the educational brochure were targeted to the “Osteoporosis and You” measure, although an ad hoc analysis of the four items most closely reflecting the intervention did not reveal an effect either (data not shown). Although several other measures of OP knowledge were available when our study began, we eliminated them because they were too long and cumbersome [
40,
41] or were designed for younger women who did not have OP [
42]. Improved measures of OP knowledge are needed, particularly among those known to have OP.
Despite its strengths, our RCT had limitations. First, the patient satisfaction with OP health care scale had not been used in RCTs designed to improve bone health. Second, we did not use an OP-specific QOL measure, which might have been more responsive to our patient-activation intervention. Lastly, given the clinical centers used, our study population may not have been representative of all osteoporosis patients.
Acknowledgements
We thank Sylvie Hall, BA (UI), Rebecca Burmeister, MPH (UI), Mollie Giller, MPH (UI), April Miller RT (UI), CBDT, Amna Rizvi-Toner, BA, BS (UI), Brandon Van Cleave (UI), Kara Wessels, BA (UI), Roslin Nelson (KPGA), Kathy Pines (KPGA), Aimee Khamar (KPGA), and Brandi Robinson, MPH (KPGA), and all of the staff at the Iowa Social Science Research Center for recruiting and interviewing all study participants. All except Ms. Miller were compensated from grant funds for their time. We also thank Sylvie Hall, BA (MPH), Mollie Giller, MPH (UI), Ryan Outman, MS (UAB), Marina Reynolds, BSN (UI), Brandi Robinson, MPH (KPGA), and Jessica Williams, PhD (UAB) for coordinating and facilitating recruitment of study participants. We also thank Xin Lu, MS, and Thuy Nguyen, MS (UI) for managing trial data. Finally, we thank the 7,749 patients who participated in PAADRN.
Jeffrey R. Curtis, MD, MPH, Sarah L. Morgan, RD, MD, CCD, FACP, Janet A. Schlechte, MD, PhD, David J. Zelman, MD