Outpatient Benzodiazepine Prescribing, Adverse Events, and Costs

Objectives: The objectives of this preliminary study were to identify a cohort of patients receiving outpatient prescriptions for a class of medications, benzodiazepines, that are known to increase the risk of adverse events, and to analyze the temporal association between outpatient benzodiazepine usage and inpatient and outpatient injury-related health care encounters for this cohort. Methods: As part of a larger research program on high-risk medications and patient injuries, we identified 17,558 patients receiving benzodiazepine outpatient prescriptions at one Veterans Health Administration (VHA) hospital system, with 9,304 individuals more than 59 years old. Adverse outcomes of interest, viz., inpatient or outpatient health care encounters coded as injuries while using benzodiazepines, were analyzed. Direct medical costs for inpatient stays and average costs for outpatient visits were obtained from cost extracts from the VHA Decision Support System. Modified Beers criteria (Zhan et al., JAMA 2001;286(22):2823–9) for potentially inappropriate medications in the elderly, irrespective of dose, were applied to three years of outpatient prescription data for the cohort of patients more than 59 years old. More than 1 million outpatient prescriptions were analyzed by Zhan’s modified Beers inappropriateness categories, namely, always avoid, rarely appropriate, and some indications. Results: For the 17,558 patients receiving outpatient benzodiazepines, we identified 297 inpatient injury admissions and 2,977 outpatient injury encounters for 1,352 patients that occurred while using benzodiazepines at the time of the injury. Over $3 million dollars in direct medical costs were associated with these injuries. Conclusions: Pharmacy Benefit Management data linked with clinical administrative data can be used to identify evidence of adverse events (patient injuries) linked to potentially inappropriate prescribing patterns in elderly outpatients. Introduction Certain medications have been identified as risk factors for fall-related injuries, adverse drug events, motor vehicle accidents, and increased hospitalizations, all of which result in potentially preventable health care utilization and costs. Older persons are particularly at risk for injuries associated with the use of certain medications, including antidepressants, antihypertensives, barbiturates, sedative hypnotics, anxiolytics, and combinations Advances in Patient Safety: Vol. 1 186 of these medications. Beers and other researchers have used explicit criteria developed by experts, including geriatricians, pharmacologists, and others, to identify potentially inappropriate drug prescribing for the elderly. 18, 21 Zhan, using a modified Delphi method, further categorized Beers’ 1997 list of drugs into appropriateness categories of “always avoid,” “rarely appropriate,” and “some indications,” irrespective of dosage. 18 The Zhan categorizations could be considered conservative, as they do not include the impact of drug dosage, drugdrug, or drug-disease interactions, which could expand the list. While another study used some of these factors to develop other potentially inappropriate prescribing criteria, we used Zhan’s more conservative approach for this preliminary study. Benzodiazepines (BZDs) are generally acknowledged as a class of medications that are an independent risk factor for fall-related and other serious injuries in community dwelling elders. 23–28 Two BZDs are on Zhan’s list, and her expert panel considered long-acting BZD use by the elderly to be inappropriate, as have others. Nevertheless, benzodiazepines continue to be disproportionately prescribed for older adults, and may be prescribed for long periods of time. 28, 30 In our study population, BZDs were widely prescribed during the study period. Preliminary analyses of administrative datasets for the hospital found that as many as one in five outpatients received at least one prescription for a BZD during the study period. There were four objectives of this study. The first was to identify BZD outpatient prescriptive patterns in a cohort of patients at one Veterans Health Administration (VHA) hospital system. The second was to analyze the temporal association between outpatient BZD usage and inpatient and outpatient injuryrelated health care encounters for this cohort. The third objective was to analyze the direct medical costs associated with those injury-related health care encounters. The fourth objective was to apply Zhan’s modified Beers’ criteria to those patients in the cohort who were 60 years of age or older to identify other potentially inappropriate outpatient medications they were receiving. The unique contribution of this preliminary study was a demonstration of the potential for enhancing patient safety by our methodology of linking outpatient drug usage of certain high-risk medications with actual adverse outcomes. Materials and methods Sources of data A datamart was created using data extracted from the administrative datasets for a VHA hospital system, which included its medical center and associated outpatient hospitals and community-based outpatient clinics. VHA outpatient prescription data from the Pharmacy Benefit Management (PBM) system was extracted for three calendar years (1999–2001). The PBM database contained information on the strength of the drug, prescribed daily amount, fill date, quantity supplied, and a unique patient identifier. Using the PBM data, we Benzodiazepine Prescribing and Adverse Events 187 identified a cohort of all patients receiving outpatient BZD prescriptions to analyze their BZD usage for the 3-year period. Using the patient identifier, the pharmacy data was combined with health care utilization data. The inpatient and outpatient health care encounter data were extracted from the centralized VHA National Patient Care Database, which included information on patient demographics and International Classification of Diseases-9th Revision-Clinical Modification (ICD-9-CM)-coded diagnoses in different datasets. The VHA Decision Support System (DSS) cost extracts provided information on treatment costs. These extracts from administrative datasets were linked to the PBM data to create a master dataset for our analyses. There were 142,204 outpatient BZD prescriptions for 17,558 unique individuals. Over the same time period, there were over 1 million other prescriptions (nonBZDs) for those unique patients, which we could search for other potentially inappropriate prescriptions. Injury identification We identified a cohort of all patients receiving outpatient BZD prescriptions to analyze their BZD usage for the 3-year period. The cohort’s inpatient and outpatient injury-related health care encounters were identified using the ICD-9CM codes for injuries and poisonings (800–999). The administrative datasets we used did not permit us to obtain the actual occurrence dates of the injuries. The administrative data has the dates of associated health care encounters for those injuries. Injuries were identified by ICD-9-CM injury codes for inpatient admissions and outpatient clinic visits. Even though we defined an injury within the ICD-9-CM range of 800–999, injuries for certain types of codes were excluded from our analyses. The Clinical Classification Software (CCS) of the Agency for Healthcare Research and Quality (AHRQ) was used to aggregate the injury codes into homogenous diagnosis groups. The CCS categories 237 (complication of device, implant, or graft) and 238 (complications of surgical procedures or medical care) were excluded from our analyses because they are iatrogenic injuries and historically have not been included in injury studies (e.g., the Centers for Disease Control and Prevention’s [CDC] E-coding matrix for injuries). The remainder of the CCS injury categories, accidental injuries and poisonings, were included in our analyses and are consistent with types of accidental injuries that have been related to BZD use. Finally, CCS category 227 patients (spinal cord injuries) were also excluded from the analysis due to a coding anomaly. Consultations with clinicians and medical record coders suggested that almost all spinal cord injury patients in our dataset were being treated for followup care rather than for their original spinal cord injury. The health care encounters coded for an injury may, or may not, reflect treatment for an incident injury. Treatment for an injury may include inpatient and outpatient phases, with multiple injury-coded health care encounters associated with an episode of care. The primary and secondary diagnosis fields in both inpatient and outpatient datasets were examined for injury codes. In order to Advances in Patient Safety: Vol. 1 188 temporally associate an individual’s outpatient BZD usage with health care encounters, we linked drug fill dates with injury-coded health care encounters. The resulting dataset of injury-coded encounters was compared with fill dates of BZDs and parsed down to only those encounters with an injury code for injuries that occurred while receiving BZDs. Injury costs Two different sources of data were used to identify costs associated with patient health care encounters coded for an injury. The direct costs of hospitalizations were obtained from the National Data Extract (NDE) of the DSS financial reporting datasets at the Austin Automation Center (AAC), a Federal data center with the Department of Veterans Affairs. The outpatient costs per visit were based on the facility’s outpatient average costs per CCS category injury visit obtained from the AAC’s CCS summaries for 2002. Other potentially inappropriate medications Our final objective in this study was to apply Zhan’s modified Beers criteria to a subset of the cohort in a two-step manner. Using the 17,558 patients receiving outpatient BZD prescriptions, we identified those patients who were 60 years or older at the time of their last outpatient visit (n = 9,304). Using Zhan’s criteria, we searched for other potentially inappropriate prescriptions, organized into the three categories of “always avoid,” “rarely appropriate,” and “some indications.” For simplicity, and following Goulding, we will refer to Zhan’s list of potentially inappropriate medications as ZL. Analysis The data in this study were clustered outpatient pharmacy data with 142,204 BZD and 1 million non-BZD prescriptions for 17,558 unique patients. The analysis consisted of descriptive statistics for the types of injuries and associated direct medical costs by CCS categories and settings of care. Separate frequency distributions for the subset of patients who were 60 years of age and older were generated for BZD prescriptions and for ZL prescriptions. The ZL prescriptions were further analyzed by unique patients and by the average prescriptions per unique patient. All analyses were conducted with Statistical Analysis System (SAS) version 8.2. This study was reviewed by and received all necessary approvals from the Institutional Review Board. Results Based on a temporal alignment of BZD usage and health care utilization, 1,649 unique patients were identified as having inpatient or outpatient health care treatment with an ICD-9-CM coded injury code while being prescribed outpatient BZDs. Two hundred ninety seven patients had more serious injuries that required hospitalization. We analyzed the direct medical costs by CCS injury categories Benzodiazepine Prescribing and Adverse Events 189 associated with inpatient stays (Table 1) and average outpatient visit costs of 1,352 patients (Table 2). The direct medical cost associated with inpatient stays for patients who were on BZDs at the time of their injuries, from Table 1, was approximately $2.89 million for 297 unique patients. Including outpatient costs associated with injuries while receiving BZDs (Table 2) for 1,352 unique patients raises the total health care costs to approximately $3.3 million. Hospitalizations and associated costs by CCS injury code in the primary or secondary category for patients not on BZDs at the time are presented in Table 3. Patients in this group may or may not have ever received outpatient BZDs, but there was no BZD outpatient prescription temporally associated with their health care encounter. Thus, this set of patients injured while not receiving BZDs represents a facility-level comparison group for the types and health care costs of those injured on BZDs (Tables 1 and 2). We should note that the costs associated with hospitalizations for these patients are based on average costs of inpatient discharges for the hospital for the time period FY1999–2001. Finally, the outpatient and inpatient grouping in Table 3 was based on calculating a weighted mean of the average outpatient visit cost and average inpatient discharge cost. Table 1. CCS injuries and costs for inpatient hospitalizations while on benzodiazepines, all ages CCS class Frequency Average cost $ Total cost $ 225 Joint injury 5 6,195 30,975 226 Fracture hip 17 11,405 193,885 228 Fracture skull & face 6 4,160 24,960 229 Fracture arm 17 6,521 110,857 230 Fracture leg 18 10,943 196,974 231 Other fracture 21 10,569 221,949 232 Sprain 18 15,021 270,378 233 Intracranial injury 31 7,413 229,803 234 Crush injury 8 4,744 37,952 235 Open wound head 16 16,499 263,984 236 Open wound extremity 27 4,150 112,050 239 Superficial injury 25 3,995 99,875 240 Burns 3 6,911 20,733 241 Poison psychotropic 17 2,482 42,194 242 Poison other medication 35 6,010 210,350 243 Poison nonmedication 5 31,513 157,565 244 Other injury 28 23,780 665,840 ALL CCS 297 $2,890,310 *Clinical Classification Software Advances in Patient Safety: Vol. 1 190 Thus, the outpatient and inpatient combined costs for each CCS class was less than an inpatient stay, as outpatient visits are generally less costly. As with patients who were treated for an injury while on BZDs, we did not attribute the costs of an episode of care to a set of inpatient and/or associated outpatient visits in this preliminary study. Table 2. CCS injuries and average costs for outpatient visits while on benzodiazepines, all ages CCS class Cost per visit ($) Total number of visits Total costs over 36 months ($) 225 Joint injury 162.12 95 15,401 226 Fracture hip 82.05 50 4,103 228 Fracture skull & face 237.83 82 19,502 229 Fracture arm 102.48 316 32,384 230 Fracture leg 129.50 302 39,109 231 Other fracture 90.66 146 13,236 232 Sprain 129.62 369 47,830 233 Intracranial injury 122.23 248 30,313 234 Crush injury 231.05 14 3,235 235 Open wound head 206.20 306 63,097 236 Open wound extremity 175.97 200 35,194 239 Superficial injury 108.75 308 33,495 240 Burns 146.83 58 8,516 241 Poison psychotropic 109.03 5 545 242 Poison other medication 82.29 79 6,501 243 Poison nonmedication 99.13 95 9,417 244 Other injury 125.98 304 38,298 Overall totals 2,977 $400,176 N = 1,352 unique patients *Clinical Classification Software Benzodiazepine Prescribing and Adverse Events 191 Table 3. Inpatient discharges and outpatient visits with an injury for patients NOT using benzodiazepines at the time of injury health care encounter Outpatient and inpatient Inpatient discharges n % Avg cost* n % Avg cost 225 Joint injury 614 3.07 378 14 1.15 9,644 226 Fracture hip 323 1.62 5,324 123 10.07 13,847 228 Fracture skull & face 233 1.17 1,036 35 2.87 5,552 229 Fracture arm 1,911 9.56 510 73 5.98 10,773 230 Fracture leg 1,410 7.06 770 94 7.70 9,742 231 Other fracture 552 2.76 1,522 87 7.13 9,171 232 Sprain 4,941 24.73 176 24 1.97 9,702 233 Intracranial injury 926 4.63 2,530 294 24.08 7,707 234 Crush injury 76 0.38 1,793 30 2.46 4,189 235 Open wound head 1,801 9.01 324 39 3.19 5,631 236 Open wound extremity 1,139 5.70 633 74 6.06 7,210 239 Superficial injury 2,838 14.20 256 86 7.04 4,967 240 Burns 359 1.80 324 24 1.97 2,802 241 Poison psychotropic 24 0.12 2,473 19 1.56 3,095 242 Poison other medication 347 1.74 835 105 8.60 2,569 243 Poison nonmedication 257 1.29 425 23 1.88 3,740 244 Other injury 2,230 11.16 218 77 6.31 2,795 TOTAL 19,981 $11,535,505 1,221 $ 9,034,449 * Weighted mean of average outpatient visit cost and average inpatient discharge cost †Average cost per inpatient discharge over 3 years (FY1999–2001) Table 4. Frequency of prescribed benzodiazepines by strength (1999–2001), age 60+ Drug name mg Frequency Percent Alprazolam 0.25 6,827 11.59 Alprazolam 0.5 7,046 11.97 Alprazolam 1 2,283 3.88 Chorazepate 3.75 36 0.06 Chorazepate 7.5 43 0.07 Chlordiazepoxide 5 625 1.06 Chlordiazepoxide 10 1,977 3.36 Chlordiazepoxide 25 803 1.36 Clonazepam 0.5 3,168 5.38 Clonazepam 1 2,088 3.55 Diazepam 2 1,393 2.37 Advances in Patient Safety: Vol. 1 192 Table 4. Frequency of prescribed benzodiazepines by strength (1999–2001), age 60+, cont. Drug name mg Frequency Percent Diazepam 5 8,266 14.04 Diazepam 10 1,959 3.33 Lorazepam 0.5 242 0.41 Lorazepam 1 1,230 2.09 Lorazepam 2 194 0.33 Oxazepam 10 5,078 8.62 Oxazepam 15 5,098 8.66 Oxazepam 30 1,052 1.79 Temazepam 15 5,731 9.73 Temazepam 30 3,668 6.23 Triazolam 0.25 75 0.13