Number (percentage) of women and men by age category who had bone mineral density measurements or received pharmacological treatment for hip and vertebral fractures combined. Rx indicates pharmacological treatment; DXA, dual X-ray absorptiometry.
Percentages of men (A) and women (B) in the study group who received study-defined guideline-recommended evaluation and treatment and those who did not.
Feldstein A, Elmer PJ, Orwoll E, Herson M, Hillier T. Bone Mineral Density Measurement and Treatment for Osteoporosis in Older Individuals With FracturesA Gap in Evidence-Based Practice Guideline Implementation. Arch Intern Med. 2003;163(18):2165-2172. doi:10.1001/archinte.163.18.2165
Osteoporosis evaluation and treatment guidelines state that, because of the high risk for future fractures, a fracture in an older individual warrants initiation of pharmacological treatment or bone mineral density (BMD) measurement followed by treatment according to BMD. We compared current practice with these guidelines.
We used the electronic data systems of a health maintenance organization to collect fracture, BMD measurement, and pharmacy data for women aged 50 to 89 years and men aged 65 to 89 years who sustained a study-defined fracture during 1998 or 1999. We determined those who had BMD measurement or pharmacological treatment for osteoporosis (bisphosphonate or estrogen) during the 2 years. We compared the evaluation and treatment data with evidence-based clinical guidelines (for women) or expert consensus (for men).
Of 70 513 members in the eligible age groups, 2804 persons sustained study-defined fractures. Overall, only 4.6% of those with fractures had treatment initiated after the fracture. Women sustained 80.7% of the study-defined fractures; 8.4% had BMD measurement and 42.4% received any treatment during the 2 years. Bone mineral density measurement and treatment frequency decrease significantly with age in women. In men, 1.5% had BMD measurement and 2.8% received any treatment. Approximately 51% (51.2%) of women and 95.5% of men in our study population were not evaluated or treated in accord with guideline or expert recommendations.
Evaluation and treatment rates for osteoporosis in older individuals with fractures fall far below national recommendations, especially for men. Intervention strategies should be developed and evaluated to prevent refracture in older individuals with fractures.
THE HUMAN and economic toll of osteoporotic fractures is staggering. One in 2 white women will suffer an osteoporotic fracture in her lifetime.1 The mortality rate from osteoporotic fractures in women is greater than the combined mortality rates from breast and ovarian cancers.2 In 1995, osteoporotic fractures led to 432 000 hospitalizations, 2.5 million physician visits, and 180 000 nursing home admissions, with direct expenditures estimated to be $13.8 billion.3 Osteoporotic fractures result in significant chronic pain and disability. About 20% of hip fracture patients die, and one third receive nursing home care in the first year after fracture.4 The death and disability rate is similar in men after hip fracture.5
Patients who have had a prior osteoporotic fracture are at high risk for future fractures.6- 9 A recent summary of the literature revealed that women with preexisting vertebral fractures had a 4 times greater risk for subsequent vertebral fracture compared with those without a prior fracture. Perimenopausal and postmenopausal women with a prior fracture had twice the risk of subsequent fracture compared with women without a prior fracture.10 Findings were similar in men.5,11
There are now several national and international evidence-based clinical practice guidelines for preventing osteoporotic fractures in women.4,12- 16 Recommendations for men are based on the consensus of experts.17 The guidelines present similar recommendations for preventing osteoporosis and for treatment strategies once osteoporosis is diagnosed. They vary in their recommendations regarding the timing and target population for screening.18 The guidelines agree that a prior fracture in an older individual warrants immediate initiation of treatment (if osteoporosis is clinically apparent or if bone mineral density [BMD] will not change treatment) or BMD measurement followed by treatment according to BMD.
Unfortunately, despite the availability of effective treatment and well-publicized treatment guidelines, there are significant gaps in implementing the guidelines. Osteoporosis has appropriately been referred to as a silent epidemic. Even cases with a clinically apparent fracture are overlooked. Fewer than 5% of patients with osteoporotic fractures are referred for medical evaluation and treatment for osteoporosis.4 It is likely that targeting evaluation and treatment of those at highest risk could be one of the more cost-effective population-based strategies for preventing osteoporotic fractures.19 Cummings19 estimated that, if those with prior fractures were targeted for treatment, there would be a 9% reduction in the incidence of hip fractures during 5 years (eg, from about 10 to 9 fractures per 1000 women aged 65 years).
In 2000, our health maintenance organization (HMO) embarked on an organizational quality initiative to improve the identification, evaluation, and treatment of members with prior fractures and low bone mass. As part of this effort, this retrospective cohort study collected the 2 years of data reported herein to compare BMD measurement and treatment frequencies of HMO members with guidelines and expert consensus. Unlike other recent evaluations of postfracture management of osteoporosis,20- 24 we were able to evaluate a contemporary cohort of patients (including men) with any fracture that has been associated with an increased risk of osteoporosis.
The protocol for this study was approved by the HMO's internal institutional review board.
The study was conducted in a nonprofit, group-model HMO in the Pacific Northwest with about 448 000 members. Thirty-five percent of the membership is older than 50 years, and more than 90% of this group have a prepaid drug benefit. Members in this age group tend to remain in the plan longer than younger groups and to disenroll at a rate of approximately 8% per year. About 54 000 members are Medicare beneficiaries. The plan insures 17% of the population of the greater Portland and Vancouver, Wash, metropolitan areas. The membership is representative of the sex and racial characteristics of the region, with 47% male, 92% white, 2.6% African American, 2.5% Hispanic, 2.5% Asian, and smaller percentages of other ethnic groups. The system maintains 1 hospital and is affiliated with 6 other geographically dispersed hospitals, and it operates 20 ambulatory care medical offices. Care is provided through a professional corporation with 669 physicians and surgeons working with 417 allied health practitioners.
Several electronic systems that store data on HMO members were used to collect data for this study. EpicCare, an electronic medical record, contains information for all patient contacts since 1996, such as patient demographics, medical history, procedures, and visit summaries. International Classification of Diseases, Ninth Revision (ICD-9) codes are embedded in the diagnosis selection screen for immediate coding of medical conditions, and all patient orders are entered electronically by clinicians. Four other electronic systems provide general member information, claims processing information for covered services provided by outside providers, data on inpatient procedures, and pharmacy data on prescriptions dispensed by outpatient pharmacies. Members provide consent to use these data for research purposes as part of their enrollment agreement.
We identified women aged 50 to 89 and men aged 65 to 89 who had continuous HMO membership during calendar years 1998 and 1999. These age groups were selected because during these years women's rate of bone loss accelerates, a significant percentage of men and women develop osteoporosis, and the risk of osteoporotic fracture increases rapidly.15,25,26 An upper age limit of 89 was selected to focus future prevention efforts on those patients who have significant expected longevity and who would be most able to respond to planned interventions.
Using the electronic databases for 1998 and 1999, we identified persons within the study population who had sustained a study-defined fracture—any clinical fracture except skull, facial, finger, toe, ankle, or any open fracture (suggestive of high force). Clinical fractures are those diagnosed by a physician, physician assistant, or nurse practitioner during a clinical encounter. Our study-defined fractures have been associated with decreased bone mass.27 They include ICD-9 codes 805 (vertebra), 807 (rib), 808 (pelvis), 810 (clavicle), 812 (humerus), 813 (radius and ulna), 814 (carpal bones), 815 (metacarpal bones), 820 (neck of the femur), 821 (femur, other, or unspecified), 822 (patella), 823 (distal tibia), and 825 (≥1 tarsal or metatarsal). We excluded open fractures in those categories.
Study-defined fractures were identified from the databases by searching for the first notation of inpatient or outpatient study fracture ICD-9 codes. This method of fracture ascertainment was validated through medical record review for a previous study.28 The percentage of reviewed fracture-related claims confirmed as actual fractures was greater than 93% for all fracture categories reviewed, except for rib (67%) and vertebra (80%).
Bone mineral density measurement procedures were identified through appropriate procedure codes using the HMO's external referral database. All BMD measures were performed using dual X-ray absorptiometry (DXA) by a single osteoporosis diagnostic center under contract with the HMO. This procedure is not available internally. Dual X-ray absorptiometry was performed on a Lunar DPX-L (GE Medical Systems, Milwaukee, Wis) until April 1999 and on a Lunar Prodigy DXA (GE Medical Systems) after that time. Measurements included BMD of the hip (femoral neck) and lumbar spine (L1-L4).
Information on use of pharmaceuticals for the prevention and treatment of osteoporosis during 1998 and 1999 was obtained from the HMO's outpatient pharmacy system. About 95% of all prescriptions written by the HMO's physicians are filled at the HMO's pharmacies, including those for members without a prepaid drug benefit, and are recorded in the pharmacy database. Pharmaceuticals examined for this study included bisphosphonates, estrogens, selective estrogen receptor modulators, and calcitonin. We included all products containing the bisphosphonates alendronate sodium, etidronate disodium, and risedronate sodium. For estrogens, we included the oral and transdermal preparations of esterified estrogens, conjugated estrogens, estradiol cypionate and estradiol valerate, estropipate, ethinyl estradiol, and norethindrone acetate, or their combinations, and excluded vaginal estrogen preparations. Because selective estrogen receptor modulators were used in only 1.7% of those treated and calcitonin in only 3.1%, these are not included in the tables or in analyses.
Pharmacological treatment of osteoporosis was defined for this study in 2 ways: (1) any use (dispensing) during 1998 and 1999 and (2) "new" use (dispensing) within the 6 months after the index fracture. New use for estrogen compounds after a fracture is defined as no use for 6 months before the fracture, followed by any use in the 6 months after the fracture. New use for a bisphosphonate is defined as no use in the 2 years before the fracture, followed by any use in the 6 months after the fracture. We selected a longer prefracture window for bisphosphonates because these agents have a longer protective effect on BMD and for fractures. The 6-month postfracture window was selected because most treatment for the acute fracture should be complete, allowing consideration of secondary prevention; past a 6-month window, it would be more difficult to associate treatment initiation with the fracture. New use was of particular interest to this study because the medication may have been initiated to reduce perceived increased risk related to the study fracture.
The HMO clinicians have access to the HMO's intranet-based clinical guideline for the evaluation and treatment of osteoporosis. The internal guideline available during the study pertained to women only, recommended DXA only for women with 2 or more risk factors, and provided treatment recommendations consistent with prevailing national guidelines.12- 16 Clinicians also had access from all clinical workstations to the Internet and therefore to other prevailing guidelines. For the purposes of this study, we defined guideline-recommended evaluation and treatment as BMD measurement alone, treatment alone, or combination BMD measurement and treatment. We then estimated the percentage of the study population who received guideline-based management.
Two-year clinical fracture frequencies by age and sex and significance testing by χ2 and Cochran-Armitage for trend were performed using SAS version 8.2 (SAS Institute, Inc, Cary, NC).
Of the 70 513 persons in the study population, 2264 women aged 50 to 89 (4.2% of 54 163 women) and 540 men aged 65 to 89 (3.3% of 16 350 men) had study fractures that occurred in 1998 and 1999. Table 1 presents the HMO population by age category and the number and percentage in each age category who sustained new clinical fractures during the 2 years. Two-year fracture frequencies increased with age for women and men, with the frequency for women aged 80 to 89 being 5 times that of women aged 50 to 64 (11.9% vs 2.4%). Men aged 80 to 89 had twice the frequency of fracture as men aged 65 to 79 (5.5% vs 2.8%).
Table 2 displays the rates of BMD measurements and any dispensed treatment in 1998 and 1999 by fracture category. We categorized the fractures as hip, vertebral, and "other" clinical fractures to be able to compare evaluation and treatment with other studies and prevailing guidelines. For all fractures, 8.4% of women and 1.5% of men received BMD measurement, and 42.4% of women and 2.8% of men received any pharmacological treatment for osteoporosis during the study.
Of 490 hip fractures, 74.1% were sustained by women and 25.9% by men. Nearly 5% (4.7%) of the women and 1.6% of the men with hip fractures had a BMD measurement in the 2-year period. Almost 43% of the women and 2.4% of the men with hip fractures received pharmacological treatment.
Of 269 vertebral fractures, 74.7% were sustained by women and 25.3% by men. Eight percent of the women and 7.4% of the men with vertebral fractures had a BMD measurement in the 2-year period. Almost 71% of the women and 13.2% of the men with vertebral fractures received pharmacological treatment.
Of 2044 other clinical fractures, 83.2% were sustained by women and 16.8% by men. Just over 9% of the women and 0.3% of the men with other fractures had a BMD measurement. Almost 39% of the women and 0.9% of the men with other fractures received pharmacological treatment.
Of the other clinical fractures, 255 (15.0%) in women and 34 (9.9%) in men were wrist fractures (ICD-9 code 813.4). Among those with wrist fractures, 25 (9.8%) of the women and 1 (2.9%) of the men received BMD screening by DXA, and 89 (34.9%) of the women and 1 (2.9%) of the men received pharmacological treatment.
Table 3 shows the percentages of subjects who received pharmacological treatment for osteoporosis during the study by the class of agent received. Of the 42.4% of women with fractures who were treated, 77.9% of those treated received estrogen (hip fractures, 68.4%; vertebral fractures, 64.8%; and other clinical fractures, 82.9%). Almost 15% (14.6%) of the women who were treated received bisphosphonate (hip fractures, 21.2%; vertebral fractures, 19.7%; and other clinical fractures, 11.9%), and 7.5% received estrogen and bisphosphonate (hip fractures, 10.3%; vertebral fractures, 15.5%; and other clinical fractures, 5.1%). The 2.8% of men with fractures who were treated pharmacologically all received bisphosphonate.
The final column of Table 3 displays the percentage of the study population who received any new treatment in the 6 months after a fracture. Only 4.6% of the study group received a new dispensing in the 6 months after a study-defined fracture. Thirty-five percent of those who received bisphosphonate received it as a new postfracture dispensing, whereas only 10.1% of those using estrogen received it as a new postfracture dispensing.
Table 4 displays the number and percentage of women and men by age category who had BMD measurements or received pharmacological treatment for all study fractures. Figure 1 graphically demonstrates this for hip and vertebral fractures (fractures with the highest prevalence of osteoporosis) combined. Bone mineral density measurement and treatment frequencies fell as women's ages increased (P<.001). There was a similar but statistically nonsignificant decrease for men (P<.40).
In women, 40.0% of all fractures and 71.4% of hip and vertebral fractures in the group aged 50 to 64 were treated, compared with 37.2% and 43.3% in women aged 80 to 89, respectively. In men, treatment fell from 3.4% of all fractures and 8.1% of hip and vertebral fractures in the group aged 65 to 79, to 1.2% and 2.8% in men aged 80 to 89, respectively. In women, DXA screening fell from 11.6% of all fractures and 11.6% of hip and vertebral fractures in the group aged 50 to 64, to 2.2% for all fractures and 2.4% of hip and vertebral fractures in women aged 80 to 89. In men, DXA screening fell from 1.9% of all fractures and 4.9% of hip and vertebral fractures in the group aged 65 to 79, to 0.6% for all fractures and 1.4% of hip and vertebral fractures in men aged 80 to 89.
Figure 2 displays the percentages of women and men in the study group who received study-defined guideline-recommended evaluation and treatment (ie, those who were not evaluated but were treated, were evaluated but not treated, or were evaluated and treated) and those who did not (ie, those who were neither evaluated nor treated). Approximately 51% (51.2%) of women were not managed according to clinical guidelines. Approximately 96% (95.5%) of men were not managed according to expert consensus.
Our data support the extensive prior work by others that concludes that the incidence of fractures associated with osteoporosis is higher in women, is substantive in women and men, and increases with age in both sexes.29- 31 Clinical guidelines recommend that older individuals with fractures should be directly treated for osteoporosis (if clinically apparent or if BMD measurement would not change treatment) or tested and then treated according to BMD measurement.4,13- 17 Approximately 51% (51.2%) of women and 95.5% of men in our study population were not screened or treated in accord with these recommendations. Only 4.6% of patients in our study population had pharmacological treatment initiated after a study-defined fracture. This small percentage of new postfracture treatment, when combined with the fact that most women (77.9%) who had any treatment during the study received estrogen, suggests that most of the "treatment" identified is incidental and may have been initiated for reasons other than osteoporosis.
Older individuals who have had a prior fracture comprise a sizable group and are easily identifiable during treatment and from medical records. Our data support the conclusion that this high-risk group is not sufficiently recognized and targeted for evaluation and treatment. Evaluation and treatment rates decrease with age, when risk increases, representing a missed opportunity to prevent refracture. This finding is similar to the significant decrease in treatment with age found by Freedman et al23 in women with wrist fractures.
This study has several important strengths. It includes a representative sample of all fractures treated in inpatient and outpatient settings in older individuals in a large HMO. The study includes men, unlike other reports of postfracture screening and treatment for osteoporosis.20- 24 The evaluation and treatment patterns described are based on clinical (not claims) data and include a broad range of physicians and allied health practitioners in various specialties.
We found that BMD measurements were infrequent (eg, only 8.4% for women). This frequency is lower than that reported within 1 to 2 years after hip fracture at 4 other managed care sites (range, 12%-24%),20 but higher than that reported in women who had wrist fractures (3%).23
In our study population, the frequency of pharmacological treatment varied by the fracture type. In persons with fractures, 42.4% of women and 2.8% of men received at least 1 dispensing during 1998 and 1999. We found that 42.7% of hip fractures in women and 2.4% in men, 70.7% of vertebral fractures in women and 13.2% in men, 38.9% of other clinical fractures in women and 0.9% in men, and 34.9% of wrist fractures in women and 2.9% in men received any pharmacological treatment. The treatment frequency we found in women with wrist fractures is higher than that found in a retrospective study23 of a claims database from 30 states (23%), but similar to that found in wrist fracture patients in Canada (38%).32
The overall postfracture treatment frequency found in our study population (4.6%) is lower than that found in a Rochester, Minn, community after a wrist fracture (17%).21 It is similar to a recent study20 of hip fracture patients in 4 other US health care systems, in which postfracture pharmacological treatment for osteoporosis varied from 5% to 44%, and similar to the 6% treatment rate found in Alberta, Canada.33 It is also similar to a case-control study24 in the United Kingdom that found a significant increase in the use of bone drugs only after vertebral fracture. Our low postfracture new treatment rate could be misleading because of the subject HMO's high prevalence of estrogen treatment that predated the fractures. In the case of some patients who were already taking estrogen at the time of a fracture, the clinicians may have concluded that the patients were already treated.
Estrogen was the predominant treatment seen in women in this study. Nearly 78% (77.9%) of the pharmacological treatment was estrogen, similar to the 85% use of estrogen preparations found after wrist fracture.21 It is unclear how much of the treatment was prescribed to preserve bone vs to treat perimenopausal symptoms or to provide what was a presumed cardioprotective effect. A study34 in another large integrated HMO found similar rates of estrogen treatment in women who had a diagnosis of osteoporosis or a prior fracture, compared with women who did not. That study concluded that the high rate of estrogen treatment was likely due to women receiving hormone therapy for reasons other than treatment of low bone mass. Estrogen use in our study population appears to be higher than that reported by others in patients with fracture. Pal22 reported a 19% use rate of estrogen in a survey of women in Great Britain with vertebral or hip fracture. Broy et al20 reported 0% to 14% estrogen use after hip fracture in 4 managed care organizations. The higher estrogen use found in our study may be due to this HMO's active campaigns in the early 1990s to increase use in postmenopausal women. Another study35 of our HMO reported that 50% of women aged 50 to 64 had at least 1 dispensing of hormone therapy in 1995.
Although national guidelines include estrogen as a treatment for osteoporosis, its adequacy for preventing and treating osteoporotic fractures is controversial.36 The Federal Drug Administration has approved hormone therapy for prevention of osteoporosis but not treatment. Randomized controlled trials of hormone therapy are in progress but have not yet assessed its preventive benefits for nonvertebral fractures.37
Bisphosphonate is used less than estrogen for osteoporosis treatment in women. Our findings are similar to the 15% of hip and vertebral fracture patients (women only) reporting bisphosphonate use in Pal's survey,22 but are slightly higher than reported by Broy et al,20 who found that 2% to 10% of hip fracture patients (men and women combined) at 4 managed care organizations received a bisphosphonate.
The reasons for the gap in evaluation and treatment for osteoporosis in older individuals after fracture are unknown. It is likely the result of a combination of patient, clinician, and health care system barriers. For patients, several factors may hinder initiation and adherence to osteoporosis treatment. Lack of patient knowledge about osteoporosis and hormone therapy has contributed to a lack of adherence to treatment.38,39 Low adherence to estrogen therapy in postmenopausal women has been well documented40; one retrospective study41 revealed long-term adherence of 30%.
Adherence has also been a concern with the next most commonly used pharmacological agent, bisphosphonates. Randomized controlled trials of a bisphosphonate found no significant differences in reported adverse effects among subjects taking an active drug compared with placebo.19 However, in an observational study,42 new upper gastrointestinal symptoms were reported in 33% of users of a bisphosphonate 3 months after initiation, and 35% discontinued the medication. The health care benefit, cost, and patient convenience likely also play a role. For example, an older patient with high out-of-pocket prescription drug costs may not consider treatment for osteoporosis a high priority. Transportation and other issues regarding access to BMD measurement and treatment may also come into play.
A large percentage of clinicians remain uncomfortable with use of BMD measurements and osteoporosis treatment. In 1994, 72% of primary care physicians in Rochester, NY, had never used BMD screening.43 Reported barriers to the use of BMD screening were cost, unfamiliarity with guidelines, uncertainty about clinical applicability, questions about the effect on treatment decisions, and availability of DXA devices. Although a recent literature review of prospective studies reveals that the association between low bone mass and most fractures is well established in women, this connection is not yet appreciated by many health care providers.44 Bone mineral density measurement is not available internally in our study HMO and requires an out-of-plan referral. This may partially explain the infrequency of BMD measurement herein. Saadi et al45 documented that, once osteoporosis was identified, treatment rates varied significantly by physician specialty, from a 96% treatment rate for metabolic bone disease specialists to 53% for general internists.
The epidemiology of osteoporotic fractures and the prevalence of osteoporosis in various fracture categories in men are less clear than for women, and randomized controlled trial data are limited. There are no published evidence-based guidelines for evaluation and treatment of osteoporosis in men, to our knowledge. It is therefore premature to conclude based on our data that clinicians are not following evidence-based guidelines for evaluating and treating osteoporosis in men. However, our data provide important baseline information to track over time as this science evolves. Our data suggest that, although the fracture burden in men is lower than in women, it is still significant. Therefore, as the population ages and the understanding of fracture prevention in men evolves, men will become more important to population management of this disabling condition than previously appreciated. A recent literature review revealed that most types of prior fractures increase the risk of future fractures in men and women.10 Therefore, although most clinicians would rightly not order a BMD for an older man with a minor fracture, such as a rib fracture from an automobile crash, serious consideration should be given to an evaluation for modifiable risk factors for fractures from falls and osteoporosis. Further research is needed to guide osteoporosis management in older men who have had fractures, as well as to identify patient, clinician, and system barriers to evaluation and treatment that might be amenable to intervention.
This study has several potential weaknesses. Although the population base in this study is similar to the community in which it resides, it may not be fully representative of that community or the US population at large. Therefore, conclusions from our study may be limited to the HMO population evaluated. The results are based on comprehensive electronic data, and although the incidence and type of fractures were not validated by medical records review, prior work has shown high validity of the data.28 We elected not to try to eliminate all high-force fractures (eg, a closed fracture from a motor vehicle crash). Our rationale is that there is good evidence that there is an increased relative risk for osteoporosis in high-force fractures and lower-force ones.46 Similar relationships between BMD and risk of nontraumatic and traumatic fractures were also found in the Study of Osteoporotic Fractures (S. R. Cummings, MD, written communication, February 2002).
We did not evaluate potential predictors of screening and treatment other than patient age and sex, nor could we evaluate differences in patient management by patient refracture risk level. Future studies should evaluate the effect of the full complement of patient risk factors for fracture, such as concomitant disease and activity status, on BMD measurement and treatment rates. Additional factors that might affect treatment should also be included in future studies, such as life expectancy, quality of life, and contraindications to treatment. Ideally, we would have evaluated whether or not treatment was appropriate after BMD measurement, but we did not have electronic access to BMD results. However, because BMD measurement was so infrequent, assessing patient management for appropriateness of treatment after BMD would not have substantially changed the percentage of individuals receiving guideline-based treatment. Further research is indicated in these areas.
Based on our data, we conclude that fracture in an older individual rarely prompts evaluation and treatment for osteoporosis, and intervention strategies should be developed, implemented, and evaluated to close the osteoporosis evaluation and treatment gap in older individuals with fractures. Intervention strategies should address barriers to patient participation and adherence, as well as clinician barriers to patient risk identification and management. Promising strategies may include combining patient and clinician education with methods to facilitate care, such as care pathways, alerts and reminders, or case management systems.
Corresponding author and reprints: Adrianne Feldstein, MD, MS, Center for Health Research, Kaiser Permanente, 3800 N Interstate Ave, Portland, OR 97227-1110 (e-mail: email@example.com).
Accepted for publication December 10, 2002.
This study was supported by a research consulting agreement through Merck & Co, Inc.
This study was presented in part as a poster at the 8th Annual HMO Research Network Conference; April 9, 2002; Long Beach, Calif.
We gratefully acknowledge the contributions of Thomas Weiss, PhD, and Ya-ting Chen, PhD, for their expert input on the manuscript; Jeff Showell, BA, and Chuhe Chen, PhD, for analysis and statistical support; Martha Swain, BA, for editing; Lin Neighbors, BA, for graphics; and Mary Harper for secretarial support.