Effect of New York State Regulatory Action on Benzodiazepine Prescribing and Hip Fracture Rates

Context Early studies showed that benzodiazepine use was associated with increased rates of hip fracture. Soon after, New York State adopted legislation to control benzodiazepine prescriptions. New Jersey did not. Contribution The authors observed rates of benzodiazepine prescribing and hip fractures in Medicaid patients for 12 months before and 21 months after New York began to track rates of physicians' benzodiazepine prescribing. Prescription rates decreased abruptly by 60%, but hip fracture rates did not change. In New Jersey, concurrent hip fracture rates did not change. Caution Prescribed dosages of benzodiazepines were unknown. Implication Controlling benzodiazepine prescribing may not reduce hip fractures, possibly because the 2 are not causally related. The Editors Concerns about benzodiazepine abuse, misuse, and adverse events, including hip fractures among elderly persons, have prompted state and national policies intended to regulate access to benzodiazepines. In 1990, after publication of the first landmark studies describing the risk for hip fractures associated with benzodiazepine use (1, 2), the U.S. Congress passed the Omnibus Budget Reconciliation Act, which allowed states to restrict coverage of benzodiazepines in Medicaid programs or exclude them from coverage. Although no state excluded benzodiazepines from coverage after this act, about one third of the states imposed limits on the number of prescriptions covered, required authorization before a patient could fill a prescription, or implemented other statewide policies that restricted access (3). Since January 2006, benzodiazepines have been explicitly excluded from coverage through the Medicare Part D drug benefit (4). On 1 January 1989, the New York State Department of Health implemented a triplicate prescription policy (TPP) for benzodiazepines. Since then, all physicians in New York State are required to obtain; pay for; and use special serially numbered, triplicate forms to prescribe benzodiazepines. Pharmacists forward 1 copy of the prescription form to state health authorities for surveillance. The TPP allows monitoring of each physician's prescribing, each pharmacy's dispensing, and each patient's receipt of benzodiazepines. The policy constitutes a barrier to accessing benzodiazepines and resulted in an immediate and sustained decrease of 55% overall in the monthly number of benzodiazepine recipients in a continuously enrolled Medicaid cohort (5), and benzodiazepine prescribing decreased by 30% in a privately insured sample and by 44% statewide (6). A neighboring, demographically similar state, New Jersey, did not regulate benzodiazepine prescribing, and use of benzodiazepines did not change (5). Perceived benefits and risks of restricted access to benzodiazepines influence the decisions to exclude benzodiazepines from coverage. Use of benzodiazepines has been associated with cognitive dysfunction and postural imbalance among elderly persons, and hip fractures are the most serious individual and public health risk because they often lead to disability and death in this group (7, 8). One expected benefit of policies that limit access to benzodiazepines is a subsequent decrease in the incidence of falls and hip fractures associated with benzodiazepine use among elderly persons (9). However, to date, no data demonstrate this effect from such policies. We evaluated whether a statewide policy that suddenly decreased benzodiazepine use by more than 50% among elderly persons decreased the incidence of hip fracture in this group (5). Our hypothesis, like that of policymakers (9) and other researchers (10), was that the sudden, large, sustained decrease in benzodiazepine prescribing in New York would result in a decrease in the incidence of hip fractures, particularly among those at highest risk for hip fractureswomen who use benzodiazepineswhereas rates of hip fracture would not decrease in New Jersey. Methods Design Using a longitudinal, controlled, quasi-experimental design (11, 12), we compared monthly rates of benzodiazepine use and hazard ratios for hip fracture among elderly Medicaid enrollees in New York (the intervention state) and New Jersey (the comparison state) during the 12 months before and the 21 months after implementation of the New York TPP. Cohort The study sample consisted of elderly Medicaid recipients (65 years of age on 1 January 1988) in New York and New Jersey. Eligible enrollees received benefits under the Aid to the Permanently and Totally Disabled or the Old Age Assistance programs. We required cohort members to have been enrolled in Medicaid for at least 10 of 12 months in the year before the TPP, with no period of residence in a long-term care facility. Because the New York State Medicaid program is about 4 times larger than that of New Jersey and the New York State Medicaid office preferred not to provide a large data set on its entire population, the New York data are derived from a 25% random sample of Medicaid enrollees. Data Sources Medicaid enrollment files were used to obtain complete and reliable monthly data on patient age and sex (13). Reimbursement claims from pharmacies provide valid measures of prescription medications dispensed to Medicaid patients that have been shown to be internally consistent and stable over time for both individual drugs and broader therapeutic classes (1315). Medicaid and Medicare claims for acute care inpatient services, which contain admission and discharge dates and multiple discharge diagnoses, were used to identify hip fractures. Benzodiazepine Use We linked Medicaid medication claims to historically complete National Drug Code Files (16) to identify dispensed benzodiazepines. At least 1 dispensed benzodiazepine (alprazolam, chlordiazepoxide, clonazepam, clorazepate, diazepam, flurazepam, halazepam, lorazepam, oxazepam, prazepam, quazepam, temazepam, triazolam) defined a benzodiazepine recipient in a given month. We defined benzodiazepine recipients and nonrecipients in the year before the TPP as persons who received or did not receive, respectively, at least 1 dispensed benzodiazepine in that period. Hip Fracture Similar to other studies (1, 2, 17), we defined hip fractures on the basis of Medicaid or Medicare claims for acute care hospitalizations that lasted longer than 1 day (to avoid misclassifying admission to the emergency department to rule out hip fracture as admission for hip fracture treatment) and involved a primary discharge diagnosis of hip fracture (International Classification of Diseases, 9th Revision, Clinical Modification codes 820.xx) or a secondary diagnosis of hip fracture with a primary diagnosis of fracture of other and unspecified parts of the femur (code 821.xx) or fracture of unspecified bones (code 829.xx). We defined eligible hip fractures as first hip fractures that occurred during the study period. We excluded hip fractures that occurred on days when patients were not enrolled in Medicaid. The hospital admission date was used as the date of the hip fracture. Statistical Analysis We used SAS software, version 8.02 (SAS Institute, Inc., Cary, North Carolina), for all analyses (18). We first compared baseline demographic characteristics (age, sex, Medicaid eligibility category) in the intervention and control cohorts. We then used segmented regression analyses to estimate the relative change in benzodiazepine use after the TPP compared with before, taking into account pre-TPP benzodiazepine use and trend (12). Using life-table estimates, we calculated crude hip fracture hazard rates by sex, state, and pre-TPP use of benzodiazepines (19). We fit extended piecewise Cox models (20), which allow for fixed and time-dependent covariates and for different hazard ratios in different time segments, to determine whether hip fracture hazard rates differed between states before the New York TPP was implemented and whether post-TPP hazard ratios between states differed from pre-TPP hazard ratios. In separate models for women and men, we modeled time in months to first hip fracture, censoring patients when they disenrolled from Medicaid for the remainder of the study period. Each model contained both states and the pre- and post-TPP periods and controlled for age and Medicaid eligibility category (Aid to the Permanently and Totally Disabled or Old Age Assistance). We assumed that the hazard ratios for hip fracture in New York compared with New Jersey were constant within the pre- and post-TPP segments, given the same age and Medicaid eligibility category. We used the Wald chi-square statistic to test whether the hazard ratios for hip fracture changed significantly from before to after the TPP, comparing New York and New Jersey enrollees. The institutional review board of Harvard Medical School exempted the study from human subjects review. Role of the Funding Sources The study sponsors were not involved in the design and conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript. The authors had full access to the study data and were responsible for statistical analysis, reporting of data, and manuscript submission. Results Study Cohorts Table 1 shows characteristics of the cohorts of elderly Medicaid enrollees in New York (n= 51529) and New Jersey (n= 42029) in 1988. Compared with New Jersey enrollees, New York enrollees were older and a greater percentage was eligible for benefits under the Old Age Assistance Program. Sex distribution (77% women) was the same between states. Overall, 23% of New York enrollees and 24% of New Jersey enrollees received at least 1 dispensing of benzodiazepine in 1988. Table 1. Age, Sex, and Eligibility Categories of Medicaid Enrollees in 1988 in New York and New Jersey Table 2 shows the number of benzodiazepine recipients and nonrecipients before and after the TPP, by sex and state. During the 33 study months, 11221 (21.8%) Medicaid enrolle