CLINICAL INVESTIGATIONS
Quality of Osteoporosis Care of Older Medicare Recipients with Fragility Fractures: 2006 to 2010 Stephen K. Liu, MD, MPH,*†‡§¶ Jeffrey C. Munson, MD, MSCE,*‡#¶ John-Erik Bell, MD, MS,**¶ Rebecca L. Zaha, MPH,* John N. Mecchella, MD, MPH,*‡§¶ Anna N. A. Tosteson, ScD,*†‡¶ and Nancy E. Morden, MD, MPH*†¶
OBJECTIVES: To assess uptake of postfracture care guidelines in community-dwelling Medicare recipients with fractures. DESIGN: Retrospective observational cohort study. SETTING: Claims-based study using U.S. Medicare administrative inpatient, outpatient (2003–2010), and prescription (2006–2010) data. PARTICIPANTS: Individuals aged 68 and older who survived at least 12 months after a fracture of the hip, radius, or humerus. MEASUREMENTS: Poisson regression modeled factors, including participant characteristics, comorbidities and hospital referral region (HRR), associated with bone density testing or osteoporosis pharmacotherapy in the 6 months after fracture. Models were repeated for participants with no osteoporosis care observed before fracture (attention na€ıve). RESULTS: In 61,832 individuals with fractures, mean age was 80.6, 87.0% were female, 88.5% were white, 2.6% were black, and 62.1% were attention na€ıve at the time of fracture; 21.8% received testing, pharmacotherapy, or both in the 6 months after fracture. In adjusted models, factors associated with significantly lower likelihood of receiving this care were black race, male sex, and an upper extremity fracture (vs hip). In models restricted to attention-na€ıve participants, the same factors were associated with lower relative risk of receiving care. Adjusted
From *The Dartmouth Institute for Health Policy and Clinical Practice, Lebanon, New Hampshire; †Department of Community and Family Medicine, Geisel School of Medicine at Dartmouth, ‡Department of Medicine, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire; §Leadership Preventive Medicine Residency Program, Dartmouth-Hitchcock Medical Center, ¶Multidisciplinary Clinical Research Center in Musculoskeletal Diseases, Lebanon, New Hampshire; # Veterans Affairs Medical Center, White River Junction, Vermont; and **Department of Orthopedics, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire. Address correspondence to Stephen K. Liu, Dartmouth-Hitchcock Medical Center, Section of Hospital Medicine, One Medical Center Drive, Lebanon, NH 03756. E-mail: [email protected] DOI: 10.1111/jgs.12507
JAGS 61:1855–1862, 2013 © 2013, Copyright the Authors Journal compilation © 2013, The American Geriatrics Society
HRR-level care rates ranged from 14.7% to 22.9% (10th to 90th percentile). The proportion receiving care increased from 2006 to 2009. CONCLUSION: Postfracture osteoporosis care was uncommon, particularly in black and male participants. Care increased over time, but for most, a fracture was insufficient to trigger effective secondary prevention, especially for participants who were not receiving prefracture osteoporosis attention. Clinicians and policy-makers must consider effective remedies to this persistent care gap. J Am Geriatr Soc 61:1855–1862, 2013.
Key words: Medicare; osteoporosis; fragility fracture
F
ragility fractures associated with osteoporosis are associated with substantial morbidity, mortality, and healthcare costs in older adults.1–3 Individuals with hip, humerus, or radius fractures have a high risk of recurrent fracture.4 Individuals who sustain a hip fracture have an estimated 5 to 8 times greater 3-month risk of death.2 Appropriate attention to osteoporosis can reduce the risk of recurrent fractures and the associated effect on quality of life and longevity.5,6 Evidence-based guidelines recommend bone density testing and osteoporosis pharmacotherapy after a fracture.7,8 In 2004, the National Committee on Quality Assurance (NCQA) introduced an osteoporosis care quality measure endorsed by the National Quality Forum (NQF) that assesses receipt of appropriate bone density testing and pharmacotherapy within 6 months of a fragility fracture for women aged 67 and older.9,10 Despite the development and dissemination of treatment guidelines and the NCQA quality measure aimed at promoting osteoporosis care in individuals with fractures, studies have shown persistent low testing and treatment rates.11–15 Research has also documented treatment disparities according to fracture type (hip vs upper extremity) and in populations typically considered at low risk of osteoporosis, such as blacks and men.11,14–19
0002-8614/13/$15.00
1856
LIU ET AL.
Previous studies of postfracture care have largely been limited to select individuals in a single geographic region or institution, have small study populations, rely on survey methods, or assess inpatient treatment only.13,16,17,20,21 Many predate the emergence and broad dissemination of pharmacotherapy guidelines and the NCQA quality measure.11,14,16,17 Research examining guideline-concordant postfracture osteoporosis management specifically in Medicare beneficiaries, the largest payer for such care, has likewise been limited. Previously published Medicare studies of osteoporosis care have used managed Medicare plan (Medicare Advantage) data or examined care in a broad group of individuals with osteoporosis, not specifically the group at highest risk for a fracture: those who have already incurred one.22–24 Medicare Advantage plans vary and provide care through an integrated delivery model; as a consequence, observations may not be generalizable to other Medicare Advantage plans. They are probably even less generalizable to the larger fee-for-service Medicare population for whom care delivery is not necessarily as highly integrated or coordinated. The current study used fee-for-service Medicare administrative data to advance the understanding of osteoporosis management through examination of care in a cohort of participants who should probably be the highest priority for attention to osteoporosis: fragility fracture survivors (individuals who sustained a fragility fracture and survived at least 12 months). By restricting the cohort to community-dwelling individuals who were continuously insured for inpatient, outpatient, and prescription services, this analysis provides a conservative and valid assessment of recent uptake and dissemination of fragility fracture care guidelines and an indication of quality of postfracture care in older adults in the United States. Understanding factors associated with the likelihood of receiving recommended care will help inform policies and practices aimed at improving care and reducing the burden of osteoporosis in this population of vulnerable older adults.
METHODS Using Medicare claims data, fragility fracture survivors were identified, and the use of bone density testing and osteoporosis pharmacotherapy was assessed in this cohort after a fracture. Factors associated with the likelihood of receiving this care were then evaluated.
Setting and Design From a 40% random sample of Medicare beneficiaries, U.S. residents aged 68 and older who experienced a fragility fracture between May 1, 2006, and December 31, 2009, and were continuously enrolled in fee-for-service Medicare Part A (insurance for inpatient services) and Part B (insurance for outpatient services) for at least 36 months before the index fracture and at least 12 months after the fracture were identified. This 36-month “look back” period was used to distinguish incident from prevalent fractures and initial from subsequent fractures. The time period was also used to ensure Medicare enrollment in the time before fracture so that prefracture bone density testing might be ascertained. This study design resulted in
NOVEMBER 2013–VOL. 61, NO. 11
JAGS
the exclusion of incident fractures occurring between age 65 (Medicare enrollment) and 68, the earliest entry age. To capture use of osteoporosis pharmacotherapy, cohort members were required to have continuous Medicare Part D (prescription drug insurance) enrollment for at least the 4 months before and 12 months after the index fracture and at least one Part D prescription drug fill claim in the 12 months after the fracture. Fragility fractures were defined as at least one Current Procedural Terminology code for hip fracture repair or at least one International Classification of Diseases, Ninth Edition (ICD-9) diagnosis code for distal radius or proximal humerus fracture plus at least one upper extremity radiography claim within 7 days () of the index ICD-9 diagnosis claim date (Table S1). Vertebral fractures were included as a covariate in the analysis (see below), as an indicator of higher fracture risk, but not included as an index fracture for cohort inclusion because the onset of these fractures is difficult to determine using claims data.
Cohort Exclusions To ensure a relatively homogeneous cohort of participants newly experiencing a fragility fracture, individuals with a nonvertebral fragility fracture in the 36 months before the index fracture were excluded. Individuals were also excluded if they were hospitalized in an acute care facility for longer than 90 days of the first 6 months after index fracture, their original Medicare eligibility was due to disability or end-stage renal disease, they had one or more diagnoses for cancer (other than nonmelanoma skin cancer) at any time in claims records analyzed, they were enrolled in hospice at any point in the observation period, or they did not survive 12 months after the index fracture. Individuals were also excluded if any prescriptions were filled at a long-term care pharmacy25 to optimize prescription claim capture and to establish a cohort of community-dwelling participants most appropriate for osteoporosis care.
Outcomes The principal outcome was attention to osteoporosis, defined as bone density testing or receipt of osteoporosis pharmacotherapy within 6 months of an index fracture. This outcome and time frame were based upon the NCQA/NQF quality measure for Osteoporosis Management in Women who had a Fracture.9 Receipt of osteoporosis pharmacotherapy (henceforth referred to as pharmacotherapy) was defined as one or more prescription fills for pharmacotherapy appearing in the Part D Prescription Drug Event File (PDE) or a claim for zoledronic acid infusion in the Medicare Part B files. Medications considered pharmacotherapy included oral estrogens, raloxifene, calcitonin, teriparatide, and intravenous and oral bisphosphonates (Table S2). Bone density testing was identified through Part B claims (Table S3). Secondary outcomes included receipt of testing or pharmacotherapy within 12 (rather than 6) months of index fracture, pharmacotherapy alone, bone density testing alone, and receipt of both.
JAGS
NOVEMBER 2013–VOL. 61, NO. 11
Covariates Information was obtained on the covariates age (68–70, 71–75, 76–80, >80), race and ethnicity (white, black, Hispanic, or other), sex, calendar year of fracture, and Part D low-income subsidy status (a measure of poverty and indicator of very low prescription cost share) using Medicare denominator files and claims data for inpatient and outpatient services.26 Charlson comorbidities present in more than 2% of the cohort were used to capture the burden of comorbid illness and were recorded if diagnosed once on an inpatient claim or twice on an outpatient claim during the 36-month look-back period or in the 12-month period after the index fracture27 (Table S4). In the main models, these were categorized as zero, one, two, or three or more comorbidities. A diagnosis of vertebral fracture and bone density testing occurring in the 36 months before fracture and pharmacotherapy preceding the index fracture (defined as a Part D fill in the 4 months or a Part B record for zoledronic acid infusion in the 12 months before index fracture) were also identified. Participant ZIP codes were used to assign each cohort member to a Dartmouth Atlas of Healthcare hospital referral region (HRR).28
Analysis Poisson regression with robust variance estimation, clustered according to HRR, was used to model bone density testing and pharmacotherapy in the 6 months after fracture. These models included anatomic location of the index fracture (hip, radius, humerus) and the covariates listed above. Poisson models were also used to derive fully adjusted HRR-level percentage of participants receiving attention to osteoporosis in HRRs with at least 50 cohort members.
Secondary Analyses In secondary analyses, the main analyses were repeated examining bone density testing and pharmacotherapy independently as outcomes. Models for the main outcome were also repeated allowing a 12-month postfracture period. To assess the effect of individual comorbidities rather than comorbidity count categories, models were run including individual comorbidities (rather than summary counts). For these models, two broad mental illness states were added to the Charlson comorbidities: depression and serious mental illness (schizophrenia, bipolar and other nonorganic psychoses). Main models were also repeated separately in subcohorts with no evidence of testing or pharmacotherapy before fracture (attention na€ıve subcohort), women only (population for the NCQA/NQF quality measure), and hip and nonhip fracture type.
RESULTS Sixty-one thousand eight hundred thirty-two individuals with fractures meeting the inclusion criteria were identified. Distribution of fracture type was 37.3% hip, 19.8% proximal humerus, and 42.9% distal radius (Table 1). Mean age was 80.6 7.0, 87.0% of the participants were
OSTEOPOROSIS CARE AFTER FRAGILITY FRACTURE
1857
female, 88.5% were white, and 2.6% were black. Overall, 9.6% of the cohort had received pharmacotherapy before the index fracture and 34.1% had received bone density testing in the 36 months before fracture. Before fracture, 62.1% had no observed testing or pharmacotherapy; this subcohort is described as attention na€ıve. The attentionna€ıve participants were more likely to be older, have higher comorbidity counts, be male, and be of nonwhite race and ethnicity than the overall cohort. Characteristics of the female-only subcohort, analyzed separately because of their specific targeting by the NCQA/NQF quality measure, paralleled the overall cohort, which was 87% female. The unadjusted prevalence of postfracture osteoporosis care was examined, and low achievement of care was found. Overall, 21.8% of the cohort received testing or pharmacotherapy in the 6 months after fracture; 7.5% of these received bone density testing, 11.7% received pharmacotherapy and 2.6% received both (Table S5). Greater attention to osteoporosis was seen with the progression of calendar time; 16.8% of participants fracturing in 2006, 22.1% in 2008, and 30.5% in 2009 received attention to osteoporosis within 6 months (Table 2). In the attentionna€ıve subcohort, osteoporosis care after fracture was less common; 11.8% received testing or pharmacotherapy in the 6 months after fracture (7.2% bone density testing, 3.1% pharmacotherapy, 1.5% both). In the subcohort of women, 23.9% received testing or pharmacotherapy in the 6 months after fracture. Models fully adjusted for participant characteristics and comorbidity count categories were run on the main outcome (testing or pharmacotherapy within 6 months) for the overall cohort and separately for attention-na€ıve participants and women (Table 3). Factors most strongly associated with a lower likelihood of attention to osteoporosis in the full cohort were aged 80 and older (vs 68–70; relative risk (RR) = 0.83, (95% confidence interval (CI) = 0.78–0.88), male sex (RR = 0.45, 95% CI = 0.41– 0.49), and black race (vs white; RR = 0.80, 95% CI = 0.70–0.92). Participants with humerus fractures (RR = 0.85, 95% CI = 0.82–0.90) and radius fractures (RR = 0.94, 95% CI = 0.91–0.97) were less likely to receive osteoporosis care than those with a hip fracture. Participants with testing (RR = 1.29, 95% CI = 1.25– 1.34) or pharmacotherapy (RR = 4.84, 95% CI = 4.65– 5.02) before fracture were more likely to receive postfracture attention to osteoporosis. Fully adjusted models including only the attentionna€ıve subcohort paralleled those of the full cohort but with stronger estimates: aged 80 and older (vs 68–70; RR = 0.63, 95% CI = 0.57–0.70), male sex (RR = 0.39, 95% CI = 0.36–0.40), black race (vs white; RR = 0.80, 95% CI = 0.67–0.95), humerus or radius fracture (vs hip; RR = 0.66, 95% CI = 0.61–0.70 and RR = 0.77, 95% CI = 0.72–0.80, respectively). In the models restricted to women, lower likelihood of attention to osteoporosis was associated with aged 80 and older (vs 68–70; RR = 0.83, 95% CI = 0.79–0.86) and black race (vs white; RR = 0.82, 95% CI = 0.74–0.91). Region of residence was associated with substantial differences in osteoporosis care. In the 285 (of 306) HRRs with 50 or more cohort members, the unadjusted median
1858
LIU ET AL.
NOVEMBER 2013–VOL. 61, NO. 11
Table 1. Characteristics of Older Medicare Recipients with Fragility Fractures Occurring May 2006 to December 2009: Overall Cohort, Attention-Naïve Subcohort and Female Subcohort
Characteristic
Overall, N = 61,832
Age, mean SDa 80.6 (7.0) Age, n (%)a 68–70 5,978 (9.7) 71–75 11,719 (19.0) 76–80 11,620 (18.8) >80 32,515 (52.6) Sex, n (%) Male 8,018 (13.0) Female 53,814 (87.0) Race, n (%)b White 54,744 (88.5) Black 1,618 (2.6) Hispanic 3,292 (5.3) Other 2,178 (3.5) Fracture location, n (%) Hip 23,048 (37.3) Proximal humerus 12,275 (19.8) Distal radius 26,509 (42.9) 18,824 (30.4) Part D low-income subsidy, n (%)c Year of fracture, n (%) 2006 9,722 (15.7) 2007 17,981 (29.1) 2008 17,548 (28.4) 2009 16,581 (26.8) Number of comorbidities, n (%)d 0 16,380 (26.5) 1 16,551 (26.8) 2 12,084 (19.5) ≥3 16,817 (27.2) 5,941 (9.6) Osteoporosis pharmacotherapy before fracture, n (%)e Bone density test 21,074 (34.1) during 36-month look-back period, n (%) Vertebral fracture 2,354 (3.8) during 36-month look-back period, n (%)
AttentionNaïve Subcohort, n = 38,376
Female Subcohort, n = 53,814
81.3 (7.2)
80.8 (7.0)
3,481 6,638 6,663 21,594
(9.1) (17.3) (17.4) (56.3)
4,916 10,052 10,074 28,772
(9.1) (18.7) (18.7) (53.5)
7,314 (19.1) 31,062 (80.9) 33,708 1,233 2,112 1,323
(87.8) (3.2) (5.5) (3.5)
47,608 1,455 2,824 1,927
(88.5) (2.7) (5.3) (3.6)
15,500 7,583 15,293 12,994
(40.4) (19.8) (39.9) (33.9)
19,232 10,758 23,824 16,939
(35.7) (20.0) (44.3) (31.5)
6,249 11,443 10,999 9,685
(16.3) (29.8) (28.7) (25.2)
8,499 15,616 15,267 14,432
(15.8) (29.0) (28.4) (26.8)
9,413 10,061 7,638 11,264
(24.5) (26.2) (19.9) (29.4)
14,759 14,653 10,461 13,941 5,853
(27.4) (27.2) (19.4) (25.9) (10.9)
20,410 (37.9)
1,138 (3.0)
2,100 (3.9)
JAGS
Table 2. Unadjusted Prevalence of Bone Density Testing or Osteoporosis Pharmacotherapy in the 6 Months After Fragility Fracture: Overall Cohort, AttentionNaïve Subcohort, and Female-Only Subcohort Overall, N = 13,452 Characteristic
Attention-Naïve Subcohort, n = 4,534 %
Overall 21.8 11.8 Agea 68–70 23.3 14.0 71–75 24.7 14.4 76–80 23.6 13.5 >80 19.8 10.2 Sex Male 7.2 5.7 Female 23.9 13.3 Raceb White 21.9 11.9 Black 13.2 8.6 Hispanic 19.7 11.9 Other 26.9 13.4 Fracture location Hip 21.1 12.8 Proximal humerus 19.6 9.8 Distal radius 23.4 11.9 Part D low-income subsidyc No 23.1 12.5 Yes 18.7 10.4 Year of fracture 2006 16.8 10.3 2007 16.0 10.1 2008 22.1 13.3 2009 30.5 13.1 Number of comorbiditiesd 0 25.2 13.4 1 23.1 12.9 2 21.3 11.8 ≥3 17.5 9.5 Bone density test during 36 month look-back period No 16.2 Yes 32.4 Osteoporosis pharmacotherapy before fracturee No 14.8 Yes 87.3 Vertebral fracture during 36-month look-back period No 21.5 11.8 Yes 28.3 14.1
Female Subcohort, n = 12,879
23.9 26.9 27.7 25.8 21.4
24.1 14.4 21.8 29.1 23.5 21.6 25.3 25.7 20.0 18.4 17.5 24.3 33.8 27.2 25.0 23.4 19.8 18.5 32.9 16.2 87.4 23.7 30.1
a
At time of fragility fracture. From Medicare denominator file. c Indicator of income 2% of the cohort (diabetes mellitus, cerebrovascular disease, lung disease, liver disease, renal disease, peptic ulcer disease, congestive heart failure, dementia) and comorbidities selected for clinical importance in osteoporosis or medication adherence (depression). a At time of fragility fracture. b From Medicare denominator file. c Indicator of income
Quality of Osteoporosis Care of Older Medicare Recipients with Fragility Fractures: 2006 to 2010 Stephen K. Liu, MD, MPH,*†‡§¶ Jeffrey C. Munson, MD, MSCE,*‡#¶ John-Erik Bell, MD, MS,**¶ Rebecca L. Zaha, MPH,* John N. Mecchella, MD, MPH,*‡§¶ Anna N. A. Tosteson, ScD,*†‡¶ and Nancy E. Morden, MD, MPH*†¶
OBJECTIVES: To assess uptake of postfracture care guidelines in community-dwelling Medicare recipients with fractures. DESIGN: Retrospective observational cohort study. SETTING: Claims-based study using U.S. Medicare administrative inpatient, outpatient (2003–2010), and prescription (2006–2010) data. PARTICIPANTS: Individuals aged 68 and older who survived at least 12 months after a fracture of the hip, radius, or humerus. MEASUREMENTS: Poisson regression modeled factors, including participant characteristics, comorbidities and hospital referral region (HRR), associated with bone density testing or osteoporosis pharmacotherapy in the 6 months after fracture. Models were repeated for participants with no osteoporosis care observed before fracture (attention na€ıve). RESULTS: In 61,832 individuals with fractures, mean age was 80.6, 87.0% were female, 88.5% were white, 2.6% were black, and 62.1% were attention na€ıve at the time of fracture; 21.8% received testing, pharmacotherapy, or both in the 6 months after fracture. In adjusted models, factors associated with significantly lower likelihood of receiving this care were black race, male sex, and an upper extremity fracture (vs hip). In models restricted to attention-na€ıve participants, the same factors were associated with lower relative risk of receiving care. Adjusted
From *The Dartmouth Institute for Health Policy and Clinical Practice, Lebanon, New Hampshire; †Department of Community and Family Medicine, Geisel School of Medicine at Dartmouth, ‡Department of Medicine, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire; §Leadership Preventive Medicine Residency Program, Dartmouth-Hitchcock Medical Center, ¶Multidisciplinary Clinical Research Center in Musculoskeletal Diseases, Lebanon, New Hampshire; # Veterans Affairs Medical Center, White River Junction, Vermont; and **Department of Orthopedics, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire. Address correspondence to Stephen K. Liu, Dartmouth-Hitchcock Medical Center, Section of Hospital Medicine, One Medical Center Drive, Lebanon, NH 03756. E-mail: [email protected] DOI: 10.1111/jgs.12507
JAGS 61:1855–1862, 2013 © 2013, Copyright the Authors Journal compilation © 2013, The American Geriatrics Society
HRR-level care rates ranged from 14.7% to 22.9% (10th to 90th percentile). The proportion receiving care increased from 2006 to 2009. CONCLUSION: Postfracture osteoporosis care was uncommon, particularly in black and male participants. Care increased over time, but for most, a fracture was insufficient to trigger effective secondary prevention, especially for participants who were not receiving prefracture osteoporosis attention. Clinicians and policy-makers must consider effective remedies to this persistent care gap. J Am Geriatr Soc 61:1855–1862, 2013.
Key words: Medicare; osteoporosis; fragility fracture
F
ragility fractures associated with osteoporosis are associated with substantial morbidity, mortality, and healthcare costs in older adults.1–3 Individuals with hip, humerus, or radius fractures have a high risk of recurrent fracture.4 Individuals who sustain a hip fracture have an estimated 5 to 8 times greater 3-month risk of death.2 Appropriate attention to osteoporosis can reduce the risk of recurrent fractures and the associated effect on quality of life and longevity.5,6 Evidence-based guidelines recommend bone density testing and osteoporosis pharmacotherapy after a fracture.7,8 In 2004, the National Committee on Quality Assurance (NCQA) introduced an osteoporosis care quality measure endorsed by the National Quality Forum (NQF) that assesses receipt of appropriate bone density testing and pharmacotherapy within 6 months of a fragility fracture for women aged 67 and older.9,10 Despite the development and dissemination of treatment guidelines and the NCQA quality measure aimed at promoting osteoporosis care in individuals with fractures, studies have shown persistent low testing and treatment rates.11–15 Research has also documented treatment disparities according to fracture type (hip vs upper extremity) and in populations typically considered at low risk of osteoporosis, such as blacks and men.11,14–19
0002-8614/13/$15.00
1856
LIU ET AL.
Previous studies of postfracture care have largely been limited to select individuals in a single geographic region or institution, have small study populations, rely on survey methods, or assess inpatient treatment only.13,16,17,20,21 Many predate the emergence and broad dissemination of pharmacotherapy guidelines and the NCQA quality measure.11,14,16,17 Research examining guideline-concordant postfracture osteoporosis management specifically in Medicare beneficiaries, the largest payer for such care, has likewise been limited. Previously published Medicare studies of osteoporosis care have used managed Medicare plan (Medicare Advantage) data or examined care in a broad group of individuals with osteoporosis, not specifically the group at highest risk for a fracture: those who have already incurred one.22–24 Medicare Advantage plans vary and provide care through an integrated delivery model; as a consequence, observations may not be generalizable to other Medicare Advantage plans. They are probably even less generalizable to the larger fee-for-service Medicare population for whom care delivery is not necessarily as highly integrated or coordinated. The current study used fee-for-service Medicare administrative data to advance the understanding of osteoporosis management through examination of care in a cohort of participants who should probably be the highest priority for attention to osteoporosis: fragility fracture survivors (individuals who sustained a fragility fracture and survived at least 12 months). By restricting the cohort to community-dwelling individuals who were continuously insured for inpatient, outpatient, and prescription services, this analysis provides a conservative and valid assessment of recent uptake and dissemination of fragility fracture care guidelines and an indication of quality of postfracture care in older adults in the United States. Understanding factors associated with the likelihood of receiving recommended care will help inform policies and practices aimed at improving care and reducing the burden of osteoporosis in this population of vulnerable older adults.
METHODS Using Medicare claims data, fragility fracture survivors were identified, and the use of bone density testing and osteoporosis pharmacotherapy was assessed in this cohort after a fracture. Factors associated with the likelihood of receiving this care were then evaluated.
Setting and Design From a 40% random sample of Medicare beneficiaries, U.S. residents aged 68 and older who experienced a fragility fracture between May 1, 2006, and December 31, 2009, and were continuously enrolled in fee-for-service Medicare Part A (insurance for inpatient services) and Part B (insurance for outpatient services) for at least 36 months before the index fracture and at least 12 months after the fracture were identified. This 36-month “look back” period was used to distinguish incident from prevalent fractures and initial from subsequent fractures. The time period was also used to ensure Medicare enrollment in the time before fracture so that prefracture bone density testing might be ascertained. This study design resulted in
NOVEMBER 2013–VOL. 61, NO. 11
JAGS
the exclusion of incident fractures occurring between age 65 (Medicare enrollment) and 68, the earliest entry age. To capture use of osteoporosis pharmacotherapy, cohort members were required to have continuous Medicare Part D (prescription drug insurance) enrollment for at least the 4 months before and 12 months after the index fracture and at least one Part D prescription drug fill claim in the 12 months after the fracture. Fragility fractures were defined as at least one Current Procedural Terminology code for hip fracture repair or at least one International Classification of Diseases, Ninth Edition (ICD-9) diagnosis code for distal radius or proximal humerus fracture plus at least one upper extremity radiography claim within 7 days () of the index ICD-9 diagnosis claim date (Table S1). Vertebral fractures were included as a covariate in the analysis (see below), as an indicator of higher fracture risk, but not included as an index fracture for cohort inclusion because the onset of these fractures is difficult to determine using claims data.
Cohort Exclusions To ensure a relatively homogeneous cohort of participants newly experiencing a fragility fracture, individuals with a nonvertebral fragility fracture in the 36 months before the index fracture were excluded. Individuals were also excluded if they were hospitalized in an acute care facility for longer than 90 days of the first 6 months after index fracture, their original Medicare eligibility was due to disability or end-stage renal disease, they had one or more diagnoses for cancer (other than nonmelanoma skin cancer) at any time in claims records analyzed, they were enrolled in hospice at any point in the observation period, or they did not survive 12 months after the index fracture. Individuals were also excluded if any prescriptions were filled at a long-term care pharmacy25 to optimize prescription claim capture and to establish a cohort of community-dwelling participants most appropriate for osteoporosis care.
Outcomes The principal outcome was attention to osteoporosis, defined as bone density testing or receipt of osteoporosis pharmacotherapy within 6 months of an index fracture. This outcome and time frame were based upon the NCQA/NQF quality measure for Osteoporosis Management in Women who had a Fracture.9 Receipt of osteoporosis pharmacotherapy (henceforth referred to as pharmacotherapy) was defined as one or more prescription fills for pharmacotherapy appearing in the Part D Prescription Drug Event File (PDE) or a claim for zoledronic acid infusion in the Medicare Part B files. Medications considered pharmacotherapy included oral estrogens, raloxifene, calcitonin, teriparatide, and intravenous and oral bisphosphonates (Table S2). Bone density testing was identified through Part B claims (Table S3). Secondary outcomes included receipt of testing or pharmacotherapy within 12 (rather than 6) months of index fracture, pharmacotherapy alone, bone density testing alone, and receipt of both.
JAGS
NOVEMBER 2013–VOL. 61, NO. 11
Covariates Information was obtained on the covariates age (68–70, 71–75, 76–80, >80), race and ethnicity (white, black, Hispanic, or other), sex, calendar year of fracture, and Part D low-income subsidy status (a measure of poverty and indicator of very low prescription cost share) using Medicare denominator files and claims data for inpatient and outpatient services.26 Charlson comorbidities present in more than 2% of the cohort were used to capture the burden of comorbid illness and were recorded if diagnosed once on an inpatient claim or twice on an outpatient claim during the 36-month look-back period or in the 12-month period after the index fracture27 (Table S4). In the main models, these were categorized as zero, one, two, or three or more comorbidities. A diagnosis of vertebral fracture and bone density testing occurring in the 36 months before fracture and pharmacotherapy preceding the index fracture (defined as a Part D fill in the 4 months or a Part B record for zoledronic acid infusion in the 12 months before index fracture) were also identified. Participant ZIP codes were used to assign each cohort member to a Dartmouth Atlas of Healthcare hospital referral region (HRR).28
Analysis Poisson regression with robust variance estimation, clustered according to HRR, was used to model bone density testing and pharmacotherapy in the 6 months after fracture. These models included anatomic location of the index fracture (hip, radius, humerus) and the covariates listed above. Poisson models were also used to derive fully adjusted HRR-level percentage of participants receiving attention to osteoporosis in HRRs with at least 50 cohort members.
Secondary Analyses In secondary analyses, the main analyses were repeated examining bone density testing and pharmacotherapy independently as outcomes. Models for the main outcome were also repeated allowing a 12-month postfracture period. To assess the effect of individual comorbidities rather than comorbidity count categories, models were run including individual comorbidities (rather than summary counts). For these models, two broad mental illness states were added to the Charlson comorbidities: depression and serious mental illness (schizophrenia, bipolar and other nonorganic psychoses). Main models were also repeated separately in subcohorts with no evidence of testing or pharmacotherapy before fracture (attention na€ıve subcohort), women only (population for the NCQA/NQF quality measure), and hip and nonhip fracture type.
RESULTS Sixty-one thousand eight hundred thirty-two individuals with fractures meeting the inclusion criteria were identified. Distribution of fracture type was 37.3% hip, 19.8% proximal humerus, and 42.9% distal radius (Table 1). Mean age was 80.6 7.0, 87.0% of the participants were
OSTEOPOROSIS CARE AFTER FRAGILITY FRACTURE
1857
female, 88.5% were white, and 2.6% were black. Overall, 9.6% of the cohort had received pharmacotherapy before the index fracture and 34.1% had received bone density testing in the 36 months before fracture. Before fracture, 62.1% had no observed testing or pharmacotherapy; this subcohort is described as attention na€ıve. The attentionna€ıve participants were more likely to be older, have higher comorbidity counts, be male, and be of nonwhite race and ethnicity than the overall cohort. Characteristics of the female-only subcohort, analyzed separately because of their specific targeting by the NCQA/NQF quality measure, paralleled the overall cohort, which was 87% female. The unadjusted prevalence of postfracture osteoporosis care was examined, and low achievement of care was found. Overall, 21.8% of the cohort received testing or pharmacotherapy in the 6 months after fracture; 7.5% of these received bone density testing, 11.7% received pharmacotherapy and 2.6% received both (Table S5). Greater attention to osteoporosis was seen with the progression of calendar time; 16.8% of participants fracturing in 2006, 22.1% in 2008, and 30.5% in 2009 received attention to osteoporosis within 6 months (Table 2). In the attentionna€ıve subcohort, osteoporosis care after fracture was less common; 11.8% received testing or pharmacotherapy in the 6 months after fracture (7.2% bone density testing, 3.1% pharmacotherapy, 1.5% both). In the subcohort of women, 23.9% received testing or pharmacotherapy in the 6 months after fracture. Models fully adjusted for participant characteristics and comorbidity count categories were run on the main outcome (testing or pharmacotherapy within 6 months) for the overall cohort and separately for attention-na€ıve participants and women (Table 3). Factors most strongly associated with a lower likelihood of attention to osteoporosis in the full cohort were aged 80 and older (vs 68–70; relative risk (RR) = 0.83, (95% confidence interval (CI) = 0.78–0.88), male sex (RR = 0.45, 95% CI = 0.41– 0.49), and black race (vs white; RR = 0.80, 95% CI = 0.70–0.92). Participants with humerus fractures (RR = 0.85, 95% CI = 0.82–0.90) and radius fractures (RR = 0.94, 95% CI = 0.91–0.97) were less likely to receive osteoporosis care than those with a hip fracture. Participants with testing (RR = 1.29, 95% CI = 1.25– 1.34) or pharmacotherapy (RR = 4.84, 95% CI = 4.65– 5.02) before fracture were more likely to receive postfracture attention to osteoporosis. Fully adjusted models including only the attentionna€ıve subcohort paralleled those of the full cohort but with stronger estimates: aged 80 and older (vs 68–70; RR = 0.63, 95% CI = 0.57–0.70), male sex (RR = 0.39, 95% CI = 0.36–0.40), black race (vs white; RR = 0.80, 95% CI = 0.67–0.95), humerus or radius fracture (vs hip; RR = 0.66, 95% CI = 0.61–0.70 and RR = 0.77, 95% CI = 0.72–0.80, respectively). In the models restricted to women, lower likelihood of attention to osteoporosis was associated with aged 80 and older (vs 68–70; RR = 0.83, 95% CI = 0.79–0.86) and black race (vs white; RR = 0.82, 95% CI = 0.74–0.91). Region of residence was associated with substantial differences in osteoporosis care. In the 285 (of 306) HRRs with 50 or more cohort members, the unadjusted median
1858
LIU ET AL.
NOVEMBER 2013–VOL. 61, NO. 11
Table 1. Characteristics of Older Medicare Recipients with Fragility Fractures Occurring May 2006 to December 2009: Overall Cohort, Attention-Naïve Subcohort and Female Subcohort
Characteristic
Overall, N = 61,832
Age, mean SDa 80.6 (7.0) Age, n (%)a 68–70 5,978 (9.7) 71–75 11,719 (19.0) 76–80 11,620 (18.8) >80 32,515 (52.6) Sex, n (%) Male 8,018 (13.0) Female 53,814 (87.0) Race, n (%)b White 54,744 (88.5) Black 1,618 (2.6) Hispanic 3,292 (5.3) Other 2,178 (3.5) Fracture location, n (%) Hip 23,048 (37.3) Proximal humerus 12,275 (19.8) Distal radius 26,509 (42.9) 18,824 (30.4) Part D low-income subsidy, n (%)c Year of fracture, n (%) 2006 9,722 (15.7) 2007 17,981 (29.1) 2008 17,548 (28.4) 2009 16,581 (26.8) Number of comorbidities, n (%)d 0 16,380 (26.5) 1 16,551 (26.8) 2 12,084 (19.5) ≥3 16,817 (27.2) 5,941 (9.6) Osteoporosis pharmacotherapy before fracture, n (%)e Bone density test 21,074 (34.1) during 36-month look-back period, n (%) Vertebral fracture 2,354 (3.8) during 36-month look-back period, n (%)
AttentionNaïve Subcohort, n = 38,376
Female Subcohort, n = 53,814
81.3 (7.2)
80.8 (7.0)
3,481 6,638 6,663 21,594
(9.1) (17.3) (17.4) (56.3)
4,916 10,052 10,074 28,772
(9.1) (18.7) (18.7) (53.5)
7,314 (19.1) 31,062 (80.9) 33,708 1,233 2,112 1,323
(87.8) (3.2) (5.5) (3.5)
47,608 1,455 2,824 1,927
(88.5) (2.7) (5.3) (3.6)
15,500 7,583 15,293 12,994
(40.4) (19.8) (39.9) (33.9)
19,232 10,758 23,824 16,939
(35.7) (20.0) (44.3) (31.5)
6,249 11,443 10,999 9,685
(16.3) (29.8) (28.7) (25.2)
8,499 15,616 15,267 14,432
(15.8) (29.0) (28.4) (26.8)
9,413 10,061 7,638 11,264
(24.5) (26.2) (19.9) (29.4)
14,759 14,653 10,461 13,941 5,853
(27.4) (27.2) (19.4) (25.9) (10.9)
20,410 (37.9)
1,138 (3.0)
2,100 (3.9)
JAGS
Table 2. Unadjusted Prevalence of Bone Density Testing or Osteoporosis Pharmacotherapy in the 6 Months After Fragility Fracture: Overall Cohort, AttentionNaïve Subcohort, and Female-Only Subcohort Overall, N = 13,452 Characteristic
Attention-Naïve Subcohort, n = 4,534 %
Overall 21.8 11.8 Agea 68–70 23.3 14.0 71–75 24.7 14.4 76–80 23.6 13.5 >80 19.8 10.2 Sex Male 7.2 5.7 Female 23.9 13.3 Raceb White 21.9 11.9 Black 13.2 8.6 Hispanic 19.7 11.9 Other 26.9 13.4 Fracture location Hip 21.1 12.8 Proximal humerus 19.6 9.8 Distal radius 23.4 11.9 Part D low-income subsidyc No 23.1 12.5 Yes 18.7 10.4 Year of fracture 2006 16.8 10.3 2007 16.0 10.1 2008 22.1 13.3 2009 30.5 13.1 Number of comorbiditiesd 0 25.2 13.4 1 23.1 12.9 2 21.3 11.8 ≥3 17.5 9.5 Bone density test during 36 month look-back period No 16.2 Yes 32.4 Osteoporosis pharmacotherapy before fracturee No 14.8 Yes 87.3 Vertebral fracture during 36-month look-back period No 21.5 11.8 Yes 28.3 14.1
Female Subcohort, n = 12,879
23.9 26.9 27.7 25.8 21.4
24.1 14.4 21.8 29.1 23.5 21.6 25.3 25.7 20.0 18.4 17.5 24.3 33.8 27.2 25.0 23.4 19.8 18.5 32.9 16.2 87.4 23.7 30.1
a
At time of fragility fracture. From Medicare denominator file. c Indicator of income 2% of the cohort (diabetes mellitus, cerebrovascular disease, lung disease, liver disease, renal disease, peptic ulcer disease, congestive heart failure, dementia) and comorbidities selected for clinical importance in osteoporosis or medication adherence (depression). a At time of fragility fracture. b From Medicare denominator file. c Indicator of income
