Andrade SE, Majumdar SR, Chan KA, Buist DSM, Go AS, Goodman M, Smith DH, Platt R, Gurwitz JH. Low Frequency of Treatment of Osteoporosis Among Postmenopausal Women Following a Fracture. Arch Intern Med. 2003;163(17):2052–2057. doi:10.1001/archinte.163.17.2052
Osteoporosis is a major cause of morbidity. Treatment of osteoporosis reduces the risk of fracture, particularly for postmenopausal women with a history of fracture.
A retrospective study was conducted using the automated databases of 7 health maintenance organizations to evaluate the use of drugs recommended for secondary prevention of osteoporotic fracture. Women 60 years and older with an inpatient or outpatient diagnostic code for a fracture of the hip, vertebra, or wrist between October 1, 1994, and September 30, 1996, and at least 1 year of continuous enrollment with a drug benefit plan following the date of fracture, were identified. The frequency of use of medications for the treatment of osteoporosis (estrogen replacement therapy, bisphosphonates, and calcitonin) during the 1-year period following the date of the initial fracture was estimated overall and according to patient age, fracture site, and year of fracture.
During the study period, 3492 women 60 years and older were diagnosed with a fracture of the hip, vertebra, or wrist, and met the inclusion criteria. Of these patients, 822 (24%) received a drug for osteoporosis treatment during the year following the fracture. The proportion of women receiving treatment for osteoporosis was approximately 2-fold higher among those with a fracture of the vertebra (44%) than among those with a fracture of the hip (21%) or wrist (23%) (P<.001). Of the 2605 women who had not been treated for osteoporosis in the 90 days before a fracture, 14% received treatment for osteoporosis in the year following a fracture. Increasing age was associated with a reduced likelihood of receiving osteoporosis treatment (P<.001).
Most of the older women who had experienced a fracture of the hip, vertebra, or wrist did not receive drug treatment for osteoporosis within 1 year following the fracture. Interventions to improve the detection and treatment of osteoporosis in high-risk patients need to be developed.
OSTEOPOROSIS IS a chronic and progressive condition that leads to decreased bone mass and skeletal fragility; in turn, these conditions can lead to fractures, disability, pain, deformity, and even death.1- 5 It is estimated that at least 10 million Americans, including 25% of all women older than 50 years, have osteoporosis.1,3 The lifetime risk of a typical osteoporotic fracture (ie, of the wrist, hip, or vertebra) is about 40%.1- 4 The annual cost of treating osteoporosis and its sequelae in the United States has been estimated at $13.8 billion,1,3- 5 compared with $7.5 billion for congestive heart failure or $6.2 billion for asthma.6
Ideally, osteoporosis should be prevented before bone mass is lost or fractures occur. Nevertheless, an important complementary strategy is to identify patients who have already had a typical osteoporotic fracture and institute treatments aimed at secondary prevention.4- 7 In postmenopausal women, at least 80% to 90% of fractures of the wrist, hip, or vertebra are associated with osteoporosis,8- 10 and a patient with both osteoporosis and a fracture has approximately a 20-fold risk of future fracture compared with a patient who has neither osteoporosis nor a history of fracture.2,11 Furthermore, the risk of recurrent fracture begins to rise within a year of the index fracture event.12 Because patients with osteoporosis are at such high risk of recurrent fracture, they may also derive the greatest absolute benefits from treatment. Since the early 1990s, physicians have had several well-tolerated and effective therapeutic options, including the bisphosphonates, calcitonin, and estrogen.1,2,6,7,13 These agents have been shown to increase bone mineral density, and the relative reduction in risk for fracture with each treatment is about 40% to 60%.1,2,6,7,13 Moreover, the benefits of treatment extend to all available subgroups of patients, including the elderly, those with multiple previous fractures, and those with the lowest bone density.14
Despite the relative ease with which high-risk patients with a symptomatic fracture can be identified, and the availability of effective treatments, some recent studies have suggested that physicians may be missing important opportunities for secondary prevention, particularly for patients with nonvertebral fractures.4,11,15- 20 For example, the rate of initiation of new osteoporosis treatment has been reported to be between 5% and 16% following a wrist fracture4,11,15- 17; 1% to 9% following a hip fractures15,18,19; and 18% to 39% following symptomatic vertebral fractures.15,20 These previous studies were not population based11,16,18- 20; were not restricted to postmenopausal women11,15,16,18,19; did not look at US data11,15,16,19; had a short duration of follow-up4,18,20; and/or examined only 1 fracture site4,11,16- 20 and were therefore somewhat limited.
We undertook the present study of a large cohort of postmenopausal women enrolled in 7 different managed care organizations across the United States to document the rate of osteoporosis treatment in the year following a fracture and to examine correlates of receiving treatment after a fracture. We hypothesized that our findings would suggest important missed opportunities for secondary prevention, and that older patients would be at particular risk of low dispensing of treatment.
A retrospective study was conducted among patients enrolled in the following health maintenance organizations (HMOs): Fallon Community Health Plan (Worcester, Mass), Group Health Cooperative (Seattle, Wash), Harvard Pilgrim Health Care (Boston, Mass), HealthPartners (Minneapolis, Minn), Henry Ford Health Systems (Detroit, Mich), Kaiser Permanente of Northern California (Oakland, Calif), and Kaiser Permanente Northwest (Portland, Ore). Women 60 years and older who received an inpatient or outpatient diagnostic code for a fracture of the hip, vertebra, or wrist between October 1, 1994, and September 30, 1996, and who were continuously enrolled in a prescription drug benefit plan for least 1 year following the date of fracture, were identified from automated databases at each HMO. We excluded women who had at least 1 diagnosis code that represented pathologic fracture, bone cancer, breast cancer, colon cancer, lung cancer, cancer metastasis, multiple myeloma, or concurrent major trauma during the study period. The date of initial fracture during the study period was identified; in cases of multiple fractures on that date, if one of them was a hip fracture, the women were classified as having a hip fracture.
We reviewed medical records of a random sample of women who fulfilled the selection criteria for nonhip fractures to evaluate the accuracy of the fracture diagnosis in the automated databases. We calculated the proportion of true positives for each nonhip anatomic site at each HMO and we excluded cases of fracture sites other than the hip from HMOs that had a true-positive rate of less than 60%. The confirmation rate was not assessed for hip fractures because diagnosis codes for hip fracture have been reported to have a high positive predictive value.21
Information on patient age, health plan enrollment status, and prescription drug dispensings was obtained from automated databases at each HMO. The outcome of interest, the dispensing of a drug used for the treatment of osteoporosis (estrogen replacement therapy, bisphosphonates, and calcitonin) within the 12 months following the date of fracture, was evaluated using the automated pharmacy dispensing files. Comorbidity was assessed using the Chronic Disease Score (CDS) developed by Clark et al to predict health care utilization, costs, hospitalization, and mortality,22 which is based on age, sex, and dispensings of prescription drugs from October 1994 through April 1995. The CDS is a claims-based risk-adjustment metric that uses drug dispensing information as a marker for chronic illness. The scores have been found predictive of utilization of health care resources, with higher scores reflecting higher health care costs.22 The scores have similarly been found predictive of hospitalization.23
The overall frequency of use of drugs prescribed for the treatment of osteoporosis (estrogen replacement therapy, bisphosphonates, and calcitonin) during the year following the date of the initial fracture was estimated overall, and according to patient age (categorized by 5-year age groups), fracture site (hip, vertebra, or wrist), year of fracture (1996 or before), and CDS (categories based upon quartiles of the overall population). Statistical significance of differences was tested using the Pearson χ2 statistic and the Mantel-Haenszel test for linear association. Logistic regression was used to estimate the strength of the association between patient characteristics and the use of drugs for treatment of osteoporosis following a fracture. The models constructed included variables for patient age, fracture site, year of fracture, CDS, and HMO site. Analyses were also performed, restricting the population to women who had a diagnostic code for a fracture of the hip, vertebra, or wrist between December 1, 1994, and September 30, 1996, and did not receive a dispensing for a drug used in the treatment of osteoporosis during the 90 days before the fracture.
We identified 3492 women 60 years and older who were diagnosed with a fracture of the hip, vertebra, or wrist during the study period and met the inclusion criteria. Of these women, 1572 (45%) were diagnosed with a fracture of the hip, 300 (9%) were diagnosed with a fracture of the vertebra, and 1620 (46%) were diagnosed with a fracture of the wrist (Table 1). Of the 2995 women who were continuously enrolled in the health plan during the 90 days before initial fracture diagnosis from December 1994 through September 1996, 390 (13%) were dispensed a drug used in the treatment of osteoporosis during the 90 days before the fracture. Of these women, 353 (91%) received an estrogen only, 29 (7%) received a bisphosphonate only, 2 (0.5%) received calcitonin only, and 6 (2%) received an estrogen and a bisphosphonate or calcitonin in the 90 days before the fracture.
Of the 3492 women who incurred a fracture, 822 (24%) received a drug for osteoporosis treatment during the year following the fracture; of these, 700 (20%) received an estrogen, 160 (5%) received a bisphosphonate, and 31 (1%) received calcitonin. Table 2 presents the frequency of women receiving a drug for osteoporosis treatment during the 12 months following a fracture, according to the women's characteristics. The percentage of women receiving osteoporosis treatment was significantly higher among those with a vertebral fracture (44%) than among those with a fracture of the hip (21%) or wrist (23%) (P<.001). Increasing age was associated with a significant decrease in dispensation of osteoporosis treatment (P<.001, test for trend). Of the 1355 women diagnosed during 1996, 351 (26%) received osteoporosis treatment in the year following fracture, compared with 471 of the 2137 women (22%) with a fracture before 1996 (P<.01). The level of comorbidity of the women, as indicated by the CDS, was significantly associated with their receiving a drug for osteoporosis treatment (P<.001, test for trend), as 19% of women in the lowest quartile received treatment compared with 28% of women in the highest quartile of the CDS. Results stratified by fracture site indicated similar associations between patient characteristics and treatment in the year following the fracture.
Among the 2605 women who did not receive a drug for osteoporosis treatment before fracture, 353 (14%) received a drug in the year following the fracture (Table 2), as 263 (10%) received an estrogen, 99 (4%) received a bisphosphonate, and 20 (1%) received calcitonin. As in the overall population, fracture site, age, and the year of fracture diagnosis were associated with use of a drug for osteoporosis treatment during the 12 months following a fracture. Among women with a vertebral fracture, 30% received treatment compared with 12% of those with a fracture of the hip or wrist (P<.001). Among women diagnosed with a fracture in 1996, 15% received treatment in the following year compared with 12% of those diagnosed with a fracture prior to 1996 (P = .03). The CDS was not associated with receiving a drug for osteoporosis treatment (P = .24, test for trend) in this subgroup.
Among the 390 patients who received a drug used in the treatment of osteoporosis before fracture, 365 (94%) continued receiving treatment for osteoporosis in the year following fracture. Of the 353 women who received estrogen alone before fracture, only 25 (7%) had their osteoporosis treatment regimen modified in the year following fracture: 20 (6%) were dispensed a bisphosphonate and 5 (1%) were dispensed calcitonin in addition to the continuation of estrogen therapy.
Table 3 presents the odds ratio (OR) estimates and 95% confidence intervals (CIs) for the association of fracture site, patient age, year of fracture diagnosis, and CDS with use of a drug for osteoporosis treatment in the year following fracture (estimates were adjusted for all variables shown in the table). Compared with women with a hip fracture, women with a vertebral fracture were more likely to receive treatment (adjusted OR, 2.61; 95% CI, 2.00-3.41) and women with a fracture of the wrist were less likely to receive treatment (adjusted RR, 0.77; 95% CI, 0.64-0.94). Women with a fracture diagnosis during 1996 were more likely to receive treatment for osteoporosis than women with a diagnosis before 1996. Increasing age was associated with a significant decrease in the dispensing of drugs used for the treatment of osteoporosis (adjusted OR, 0.28; 95% CI, 0.21-0.38 for women 80 years or older compared with women younger than 65 years). Adjusted OR estimates for the associations between fracture sites, patient age, and year of fracture diagnosis among women not receiving a drug for osteoporosis treatment before fracture were similar to those for the overall population.
We found that, in a large population-based cohort of postmenopausal women, only 24% of older women with a symptomatic fracture were dispensed potentially effective treatment for osteoporosis in the year following the fracture. When we restricted our analyses to women who had not received osteoporosis medications before the fracture, presumably because many of them did not have a preexisting diagnosis of osteoporosis, the overall 1-year treatment rate, at 14%, was even lower. In addition, we noted at least 2-fold differences in treatment rates based on the site of the fracture (44% for vertebral fractures vs 22% for nonvertebral fractures) and the age of the patient (31% for those <65 years vs 15% for those ≥80 years).
The association between increasing age and decreasing likelihood of receiving effective osteoporosis treatment was independent of the site of fracture and the burden of existing comorbidity. Age as a risk factor for inadequate treatment has been previously documented for many conditions,24- 26 including osteoporosis.4,27 Nonetheless, given the fact that osteoporosis is, itself, an age-related condition and that increasing age is a powerful independent risk factor for future fracture, we might expect that elderly patients would be more likely—not less likely—to receive treatment than younger patients. A 75-year-old woman has, on average, a life expectancy of 12 years24 and the benefits of osteoporosis treatment, in terms of increased bone mineral density and decreased fracture risk, are seen within a year or two.12,27 This has prompted Ensrud et al14 to suggest that it may never be too late, in life or in the disease process, to prevent fractures with appropriate treatment.
Overall, our results are concordant with those of previous studies that used data from 1993 through 1999.4,11,15- 20 The problem of undertreatment of osteoporosis in patients with a symptomatic fracture has been documented using different methods, across different populations, and in different health care delivery systems. In our well-insured population, with reasonable access to health care and prescription drug benefits, why should undertreatment be so prevalent? Although many barriers to optimal treatment may exist, one of the major contributors to this problem may be at the level of the health care delivery system. There appears to be a clinical "disconnection" between the physicians responsible for treating the symptomatic fracture, and the primary care physicians who are responsible for the detection and treatment of osteoporosis.4,27 We believe that any health care system that does not explicitly provide the means to link acute care providers and primary care providers will be at risk for delivering suboptimal osteoporosis care.
In addition, an element of "clinical inertia" may be present. Clinical inertia, the failure of health care providers to initiate or change treatment when the health status of the patient indicates that such action is necessary, has been described for several other chronic medical conditions including diabetes, hypertension, and hyperlipidemia.28,29 Recommended strategies to avoid clinical inertia include the provision of various forms of education; systematic, targeted reminders and feedback of practice performance; and the development of interventions to address important quality of care problems.29
Another contributor to the low use of osteoporosis treatment may be the view that insufficient evidence exists about the optimal evaluation processes and interventions to prevent osteoporotic fracture. The National Institutes of Health Consensus Statement on Osteoporosis Prevention, Diagnosis, and Therapy (2000) suggests a need for further study of the long-term effectiveness and safety of current drug interventions.30 In addition, the statement suggests a need for determination of the cost-effectiveness of programs. In regard to safety considerations, the recently reported findings from the Women's Health Initiative emphasize the relevance of the recommendations summarized in the consensus statement.31 Such factors may lead some clinicians and investigators to believe that large-scale interventions to increase use of osteoporosis treatments among postmenopausal women following fracture are premature.
The major strength of this observational study is the large, geographically diverse, population-based cohort that was analyzed. In addition, we had high-quality data for the typical osteoporotic fractures and a relatively long period of follow-up to analyze. However, some limitations should be noted. First, the data are from 1994 to 1996. It might be argued that our results simply reflect the fact that the most rigorous trials of osteoporosis treatment were only starting to be published by 1990, and that alendronate (the most potent—but not the only—bisphosphonate available during our study) was not approved until 1995. In fact, we noticed a small (but statistically significant) 4% increase in overall treatment rates from 1994 to 1996. Nonetheless, even studies using data from 1997 through 1999 demonstrate a magnitude of undertreatment similar to our estimates.4,11,16,18,19
Second, we did not have access to any information regarding the results of, or even the performance of, bone mineral density testing. Because our population had health insurance and access to medical care, it is unlikely that a lack of diagnostic testing accounts for our results. In fact, it is more than likely that our patients had greater access to testing than much of the general US population. Furthermore, although bone mineral density testing is useful for diagnosis, it should be reiterated that almost all postmenopausal fractures of the wrist, hip, or vertebra are related to low bone mass.9,10
Third, we made the conservative assumption that estrogen therapy was prescribed for osteoporosis treatment. This assumption is plausible for the women who first started estrogen therapy after their fracture. However, it may be that the women (13%) who were taking estrogen therapy before their fracture were taking it for other indications, such as menopausal symptoms. For example, based on a medical chart review, Cuddihy et al17 reported that 58% of patients taking estrogen therapy before a fracture of the wrist were not actually taking it for the treatment of osteoporosis. Thus, we may have somewhat overestimated rates of osteoporosis treatment in the overall cohort.
Fourth, we did not have access to information regarding over-the-counter medications. This might be of concern, because calcium and vitamin D are often used to treat osteoporosis. In the population that we studied, however, calcium and vitamin D by themselves would be considered inadequate management. This is true to the extent that even the major trials of osteoporosis therapy have used calcium and vitamin D supplements in their placebo control groups. We also did not have data regarding prior fracture history or contraindications to medications, including whether the patient was unable to stand or sit upright after drug administration, which is relevant to the use of bisphosphonates. Lastly, we did not have any information regarding men or premenopausal women. Given this final limitation, and the fact we drew our sample from patients enrolled in managed care, our results may not necessarily be generalized to other patients or settings.
In conclusion, we found that the vast majority of postmenopausal women did not receive treatment for osteoporosis in the year following a typical osteoporotic fracture. Older women were at particularly high risk of undertreatment. This study demonstrates a significant "care gap" between evidence-based best practice and everyday clinical practice. Our findings suggest an opportunity to improve the quality of osteoporosis care. It is unlikely that publication of randomized trials or the creation and dissemination of clinical practice guidelines will be sufficient to improve osteoporosis management in patients at very high risk of recurrent fractures. To close this care gap and improve the quality of care for these patients, we believe that innovative, multifaceted interventions need to be developed to address the barriers related to system, physician, and patient that stand in the way of best practice.
Corresponding author: Susan E. Andrade, ScD, Meyers Primary Care Institute, 630 Plantation St, Worcester, MA 01605.
Accepted for publication November 21, 2002.
This study was supported in part by grant HS10391 from the Agency for Healthcare Research and Quality (Washington, DC) to The HMO Research Network Center for Education and Research in Therapeutics (CERT).
We thank Jackie Cernieux, Parker Pettus, and Rachel Dokholyan for their technical support. We also acknowledge the health care systems, including Henry Ford Health Systems, for their contribution of data.