Association of medication attitudes with non-persistence and non-compliance with medication to prevent fractures

Abstract

Summary

Our objective was to assess the association of self-reported non-persistence (stopping fracture-prevention medication for more than 1 month) and self-reported non-compliance (missing doses of prescribed medication) with perceived need for fracture-prevention medication, concerns regarding long-term harm from and/or dependence upon medications, and medication-use self-efficacy (confidence in one’s ability to successfully take medication in the context of their daily life).

Introduction

Non-persistence (stopping medication prematurely) and non-compliance (not taking medications at the prescribed times) with oral medications to prevent osteoporotic fractures is widespread and attenuates their fracture reduction benefit.

Methods

Cross-sectional survey and medical record review of 729 patients at a large multispecialty clinic in the United States prescribed an oral bisphosphonate between January 1, 2006 and March 31, 2007.

Results

Low perceived necessity for fracture-prevention medication was strongly associated with non-persistence independent of other predictors, but not with non-compliance. Concerns about medications were associated with non-persistence, but not with non-compliance. Low medication-use self-efficacy was associated with non-persistence and non-compliance.

Conclusions

Non-persistence and non-compliance with oral bisphosphonate medication have different, albeit overlapping, sets of predictors. Low perceived necessity of fracture-prevention medication, high concerns about long-term safety of and dependence upon medication , and low medication-use self-efficacy all predict non-persistence with oral bisphosphonates, whereas low medication-use self-efficacy strongly predicts non-compliance with oral bisphosphonate medication. Assessment of and influence of these medication attitudes among patients at high risk of fracture are likely necessary to achieve better persistence and compliance with fracture-prevention therapies.

Appendix: psychometric (measurement) properties of survey scales

All survey items were pre-tested in detail with volunteers with osteoporosis (who were not participants in the main study) to assess whether or not they were clear and interpreted in the way we intended. We randomly split the study sample into two halves, and examined the psychometric properties of the multi-item scales separately in each half. Principal components analysis was done to establish the factor structure of all items within the multi-item scales in both groups. Internal consistency reliability of the multi-item scales was assessed using Cronbach’s alpha, and item-scale correlations were evaluated. The unidimensionality of the summated rating (Likert) scales were established using principal components of each scale separately, and a ratio of the first to second eigenvalue greater than 3.0 was considered to be strong evidence of unidimensionality.

Perceived need for fracture-prevention medication was assessed by a seven-item scale, adapted for those with osteoporosis and at high risk of fracture, analogous to the disease-specific necessity scales of Horne and colleagues [16]. The raw summed scores were squared to achieve a normal distribution. The internal consistency reliability in each half of the study sample, respectively, was 0.87 and 0.86. There was also strong evidence of this being a unidimensional scale, in that principal components analysis showed the ratio of the first to second eigenvalue to be 4.77 and 4.09, respectively, in each half of the study sample. The item-scale correlation ranges were 0.69 to 0.80 and 0.74 to 0.82, respectively, in each half of the study sample.

Concerns about medications was measured by an 11-item scale that assessed patient perceptions regarding the perceived long-term safety of and dependence upon medications, and whether or not medications in their view are over-prescribed. Importantly, this scale does not assess actually experienced side effects (adverse reactions) to any medications. The internal consistency reliability of this scale was 0.85 in both halves of the study sample. The ratio of the first to second eigenvalue was 3.92 and 3.55, respectively, in each half of the study sample. The item-scale correlation ranges were 0.58 to 0.68 and 0.56 to 0.73, respectively, in each half of the study sample.

Medication-use self-efficacy was measured with a seven-item scale derived from that of Resnick and colleagues [27]. In our pre-test, participants expressed difficulty rating their self-efficacy with the original linear rating scale of Resnick and colleagues, and hence we converted this to a Likert scale with five response categories. We also eliminated items in Resnick’s original that referred to side effects or medication costs. We conceived of self-efficacy as confidence in the ability to execute medication-use behavior in the context of one’s daily life and that side effects or concerns about costs may lead one to choose not to take medication but would not influence the confidence to take it should they so choose. Our scale nonetheless had an internal consistency reliability in the two study sample halves, respectively, of 0.94 and 0.93, and was strongly unidimensional (ratio of first to second eigenvalues of 10.58 and 10.38, respectively, in the two study halves). The item-scale correlation ranges were 0.82 to 0.90 and 0.81 to 0.90, respectively, in the two study halves.

Principal components analysis with orthogonal rotation of all three scales together within both study halves showed the loadings of all items onto their hypothesized factor to be 0.51 or higher, and the loadings onto the other two factors to be 0.20 or lower.