CMPH Journal

Introduction

Osteoporotic fragility fractures are common, with around 3.5 million new fractures occurring annually in the European Union [1] with an estimated economic burden of € 37 billion in 2010 [1]. These fractures are frequently only managed by surgery or immobilisation with a cast, without any further investigation or prescription of appropriate anti-osteoporosis treatment [2]. This is of particular concern due to the increased probability of refracture following an osteoporotic fracture [3,4] as well to the elevated mortality risk following certain types of fracture [5-8]. The gap between best standards of care for osteoporotic fractures and everyday clinical practice has been demonstrated in many different countries and healthcare systems [7,9-13]. Narrowing this gap and improving standards of care is an important public health challenge, which requires a clear vision of the barriers that prevent patients with osteoporosis from receiving the quality of care that is recommended in practice guidelines [14]. Barriers to a better standard of care may exist at the level of care organisation [2,15,16], at the level of knowledge and attitude of physicians [1719] and at the level of patients’ perceptions [20,21]. Although the first two levels have been widely studied, there is relatively little information available, particularly quantitative information, on the perspective of the patient. In this context, we have implemented a research programme to collect information on the experiences and expectations of patients who have experienced an osteoporotic fracture regarding to secondary prevention for osteoporosis.

Methods

In a first qualitative phase of the study, a literature review was performed and patients who had experienced an osteoporotic fracture were interviewed in extenso by trained investigators in order to identify potential barriers to care and expectations for change. In the second quantitative phase, individuals who had experienced an osteoporotic fracture were sent a questionnaire asking them to rank the importance of each of the items identified in the first phase using the Best and Worst scaling (BWS) method [22]. A scientific committee consisting of rheumatologists, health economists, methodologists and a patients’ association representative oversaw the design and implementation of the study.

Participants

For the qualitative phase of the survey, 24 patients with a recent osteoporotic fracture (including hip or vertebral fractures) were identified by the French patient association (Association Française de Lutte Anti-Rhumatismale; AFLAR). These individuals were volunteers chosen to include a range of patient profiles in terms of age, gender, education, socioeconomic status, education, access to care (rural/urban) and type of fracture.

For the quantitative phase, members of the METASKOPE panel were contacted. METASKOPE is a permanent sample of 20,000 volunteer households constituted by the quota method to be representative of the French population with respect to principal sociodemographic variables. The panel is used to collect information about health or other topics of interest using questionnaires. Individuals in the METASKOPE panel who reported experiencing a non-traumatic fragility fracture within the preceding three years had been identified in a previous study [23] and these were sent the study questionnaire by post. These individuals make up the study panel. In case of failure to return the questionnaire or incoherent or missing data, a follow-up telephone call was made.

Qualitative phase: identification of barriers and expectations

An iterative process was used to identify items potentially related to barriers to care or health expectations [24]. First of all, a literature review was performed in order to understand perceptions of fracture victims regarding expectations from fracture management and to barriers preventing their expectations being met [22,23,25,26]. The recommendations of the 2009 PRISMA statement for performing systematic reviews [27] were followed.

In the next step, semi-structured individual interviews were conducted with the 24 patients who had experienced a fracture. Patients were chosen based on their experience with or knowledge of the condition; they were not representative or randomly selected but were included because they exemplified key cases or possessed in-depth knowledge of the area being explored. The goal was to recruit a broad patient sample in order to capture as wide a range of individual patient experience as possible and to minimise the chances that important themes were missed (purposive sample), rather than to aim for representativeness with respect to the osteoporosis population in general. For this reason, the panel was chosen to include a range of patient profiles typical of patients with osteoporosis in terms of age, gender, socioeconomic status, education, access to care (rural/urban), type of fracture and treatment.

The sample size was determined by the principle of saturation, with participants included as long as they provided new, relevant information. Individual semi-structured interviews were conducted by independent professionals based on an interview grid built from the literature review. The interview method was somewhere between a free (or non-directive) interview, where a general question allows the respondent to answer at length, and a structured interview, where short, closed questions are asked in a fixed order. A content analysis was performed on the collected verbatim derived from transcripts of the interviews. Themes were not identified in advance and were derived from the recorded data. Concepts were grouped by theme by the data analyst. No formal coding system was used.

Finally, two focus groups brought together six and five patients designated by the AFLAR patient association to ensure no major theme was forgotten. The scientific committee then finalised the final item list to a set of 21 barriers and 21 expectations and validated the wording.

Quantitative phase: experimental design and rating of items

During the quantitative phase, panellists were asked to rate the relative importance of the items on the checklist using the BWS method [22]. In order to limit attention difficulties for the panellists, three separate questionnaires were constructed, each presenting seven choice tasks for barriers and seven tasks for expectations. In each choice task, participants had to choose the item of five items presented that they considered the most important (best) and the one they considered the least important (worst). The questionnaires were constructed using a balanced incomplete blocks design [28]. To avoid any selection bias in responses, the same total number of items was evaluated across the three questionnaires. The three questionnaires were randomly attributed to the panellists, each of whom completed a single questionnaire.

Statistical analysis

The evaluation of the study population is purely descriptive. All statistical analyses were conducted using SAS 9.4 and Excel 2016 software, with the exception of the latent class analysis, which was performed using Sawtooth software (Sawtooth Software, Inc., 2013). A threshold of 5% was considered statistically significant.

Priorities

For each barrier or expectation item, an importance score was computed by subtracting the number of times the item was chosen as the least important from the number of times it was deemed most important. A positive importance score indicates that the item was more often chosen as “best” rather than  “worst” and a negative score indicates the reverse. A score of zero indicates that the item was considered best as often as worst, or that none of the panellists ever chose it as best or worst. The mean importance score was then calculated by dividing the individual importance scores by the number of respondents. This metric represents the average importance score for the panellist population and can be ranked on an interval scale to display the hierarchy of importance attached to each item. A high mean importance individual score, associated with a low coefficient of variation, indicates a strong consensus within the group to declare this barrier or expectation as the most important.

Preference heterogeneity

Segmentation of patients was performed using the latent class analysis method in order to identify groupings of participants who provided similar patterns of priorities [29,30]. Only patients for whom both questionnaires were complete for all item-pairs were included in the latent class analysis. The goodness-of-fit of the final model was assessed using the Akaike Information Criterion (AIC). An Analysis of Variance (ANOVA) was performed in order to identify any individual mean importance scores which differed significantly between classes. A post hoc analysis using the Tukey HSD test was then performed to determine whether overall group averages differed significantly between classes.

Ethics

The survey was conducted in accordance with the ESOMAR International Code on Market and Social Practice, the EphMRA Code of Conduct, relevant current French and European legislation, and Good Epidemiological Practice guidelines. Analyses performed using the METASKOPE panel were approved by the French Data Protection Agency (Commission Nationale de l’Informatique et des Libertés; CNIL). In addition, before answering the questionnaire, panellists provided informed consent for the study.

Results

Study population

In the qualitative phase, 24 patients participated in individual interviews and 11  in focus group discussions. The 24 patients were all women aged from 53 to 88 years (7 aged over 75 years) and came from diverse socioeconomic backgrounds.

In the quantitative phase, the BWS questionnaire was sent to the 464 panellists reporting non-traumatic fractures. Twenty-nine panellists declined to participate, thus reducing the sample to 435 panellists. Overall, 357 questionnaires were returned. A follow-up telephone call was made to 108 panellists in order to correct errors or replace missing data in the questionnaires and corrections were made for 75 questionnaires. In total 324 questionnaires could be analysed. For thirteen panellists, certain item pairs were not rated and these were excluded from the latent class analysis, which was therefore performed using the data from 311 questionnaires. The mean age was 68 years and 78.7% of panellists were women. The most frequently reported fracture locations were the wrist and the ankle.

Qualitative phase: identification of items

A preliminary systematic review of the literature identified 193 publications discussing barriers to treatment and expectations in osteoporosis. These included two previous systematic reviews of the same topic [25,26]. 58 potentially relevant items were identified from the literature review and proposed during the individual faceto-face interviews of patients with osteoporotic fractures. From these interviews a list of 42 criteria deemed the most important was derived, including 21 items for barriers and 21 for expectations. Key themes identified in the face-to-face interviews and examples of verbatim are presented in Table 1.

Quantitative phase: hierarchical rating of items

The relative importance of the different items related to barriers to care is presented in Table 1.

Table 1: Best-Worst Case Analysis of Barriers.

Mean importance scores ranged from -0.36 to +0.42, although the 95% confidence intervals were all rather broad, given the limited number of panellists scoring each item. For seven items, the score was significantly higher than zero, indicating that these items were considered to be important by panellists. Three items stood out as having particularly high scores (between 0.35 and 0.42), with Item 6 (My fracture has nothing to do with my osteoporosis) being the most important, followed by Items 19 (You do not hear enough about osteoporosis and its treatment) and Item 1 (I prefer to take natural treatments and pay attention to what I eat). The other items with a score significantly higher than zero were Items 21, 2, 14 and 13, which all relate to dysfunctions in the healthcare system. However, the confidence interval of Item 6 did not overlap those of the items from Item 21 to Item 13, indicating that it is significantly more important than these items. Item 13 was also rated as significantly less important than Items 6, 19 et 1.

For eight items (N°8 18, 7, 10, 4, 3, 11, 15 and 16), the score was significantly lower than zero, indicating that these items were considered to be relatively less important by panellists. The least important obstacle was Item 16 (I do not have psychological support for the management of my osteoporosis) and its confidence interval did not overlap with the seventeen highest-ranked items, indicating that it was significantly less important than the latter.

Considering the inter-items ranking the most important obstacle were Item 6 followed by Item 13 and 11. The confidence intervals of those item overlapped, indicating that respondents did not differentiate between the importance of the 3 obstacles. The confidence interval for Item 6 did not overlap with any of the 18 lowest-ranked obstacles, indicating that respondents differentiated significantly the importance of this obstacle from the latest and specifically from Items 21, 2, 14 et 13 describing the dysfunctions in the healthcare system. The least important obstacle was Item 16 “I do not have a psychological support” and its confidence interval did not overlap with the 16 highest-ranked item indicating that it was significantly less important than those.

With respect to expectations for care, values ranged from -0.71 to 0.34, although there was considerable overlap between coefficients of variation, which precludes unambiguous ranking of the items (Figure 1).

Figure 1: Hierarchical rating of expectations for care.

The number of expectations with a mean importance score significantly greater than zero (eleven) is higher than that for barriers. The absolute score is very similar for the first five items (2, 1, 5, 16 and 10), and the coefficients of variation of all eleven significant items overlap. The seven items with scores significantly lower than zero, considered relatively unimportant by panellists, were on the other hand better differentiated.

Preference heterogeneity

The latent class analysis only included the 311 patients with complete questionnaires. Since the classification of items for the expectations was not particularly discriminating, responder profiles were only evaluated for the barriers to care item set. The variation in the item scores between panellists was best explained by partition into three classes. Significant differences between scores were observed for around half the items (Tukey HSD test; Figure 2).

Figure 2: Latent Class Analysis.

This was notably the case for Item 1 (Class 2 > Class 1 > Class 3), Item 19 (Class 2 ≡ Class 3 > Class 1), Item 21 (Class 1 ≡ Class 3 > Class 2).

Class 1 (the independents)

Class 1 constitutes 18.3% of the panellists and is represented by individuals who probably understand osteoporosis well. They do not consider lack of awareness about osteoporosis to be an important barrier to the implementation of prevention. For example, they rank Item 19 (You do not hear enough about osteoporosis and its treatment), which ranks second in the overall population, in last place (Figure 2). In second last place, they rank Item 8 (I do not know much about osteoporosis treatments or their efficacy). However, they consider the gaps in healthcare provision as important barriers (Items 13, 17 and 21 are all ranked in the top ten). On the other hand, they are happy to manage their osteoporosis outside the health system, with Item 1 (I prefer to take natural treatments and pay attention to what I eat) ranked as most important.

Class 2 (the unaware)

Class 2 accounts for 31.1% of the panellists and corresponds to individuals who are not particularly concerned about the risk of osteoporosis. Item 5 (I am not exposed to a risk of osteoporosis) is ranked fifth in this group (Figure 2). They do not seem to be aware of the system’s failures as they rank Item 21 (After my fracture, the hospital did contact my GP for follow-up; ranked fourth overall) in last place and Item 17 (No-one has explained to me about osteoporosis) in 18th place (ninth overall). On the other hand, as in the overall group, Item 19 (You do not hear enough about osteoporosis and its treatment) is ranked second.

Class 3 (the victims of the system)

Class 3 accounts for 50.5% of the panellists and corresponds to individuals who are eager for medical management of osteoporosis and mostly encounter barriers related to lack of communication and coordination in care (Figure 2). Item 19 (You do not hear enough about osteoporosis and its treatment) was ranked in first place. Item 21 (After my fracture, the hospital did contact my GP for follow-up) was ranked second and Item 2 (I have more important health problems than my osteoporosis) third. On the other hand, they did not attach importance to barriers related to medication (Items 10, 11 and 12) and Item 1 (I prefer to take natural treatments and pay attention to what I eat), which ranked in third place overall, was ranked 17th in this group.