Publications

This study uses Medicaid drug utilization data to describe reimbursement and market share of long-acting insulins before and after approval of new insulin products and to estimate savings associated with a “biosimilar” for insulin glargine approved in 2015.

BACKGROUND & AIMS: Despite increased risks for adverse effects in patients with cirrhosis, little is known about opioid prescriptions for this population. We aimed to assess time trends in opioid prescribing and factors associated with receiving opioids among patients with cirrhosis.

METHODS: Among Veterans with cirrhosis, identified using national Veterans Health Administration data (2005-2014), we assessed characteristics of patients and their prescriptions for opioids. We calculated the annual proportion of patients receiving any opioid prescription. Among opioid recipients, we assessed prescriptions that were long-term (>90 days' supply), for high doses (>100 MME/day), or involved combinations of opioids and acetaminophen or benzodiazepine. We evaluated patient characteristics independently associated with long-term and any opioid prescriptions using mixed-effects regression models.

RESULTS: Among 127,239 Veterans with cirrhosis, 97,974 (77.0%) received a prescription for an opioid. Annual opioid prescriptions increased from 36% in 2005 to 47% in 2014 (P < .01). Among recipients of opioids, the proportions of those receiving long-term prescriptions increased from 47% in 2005 to 54% in 2014 (P < .01), and19%-21% received prescriptions for high-dose opioids. Prescriptions for combinations of opioids and acetaminophen decreased from 68% in 2005 to 50% in 2014 (P < .01) and for combinations of opioids and benzodiazepines decreased from 24% to 19% over this time (P < .01). Greater probability of long-term opioid prescriptions was independently associated with younger age, female sex, white race, hepatitis C, prior hepatic decompensation, hepatocellular carcinoma, mental health disorders, nicotine use disorders, medical comorbidities, surgery, and pain-related conditions.

CONCLUSION: Among Veterans with cirrhosis, 36%-47% were prescribed opioids in each year. Mental health disorders and hepatic decompensation were independently associated with long-term opioid prescriptions.

BACKGROUND: Treatment by high-opioid prescribing physicians in the emergency department (ED) is associated with higher rates of long-term opioid use among Medicare beneficiaries. However, it is unclear if this result is true in other high-risk populations such as Veterans.

OBJECTIVE: To estimate the effect of exposure to high-opioid prescribing physicians on long-term opioid use for opioid-naïve Veterans.

DESIGN: Observational study using Veterans Health Administration (VA) encounter and prescription data.

SETTING AND PARTICIPANTS: Veterans with an index ED visit at any VA facility in 2012 and without opioid prescriptions in the prior 6 months in the VA system ("opioid naïve").

MEASUREMENTS: We assigned patients to emergency physicians and categorized physicians into within-hospital quartiles based on their opioid prescribing rates. Our primary outcome was long-term opioid use, defined as 6 months of days supplied in the 12 months subsequent to the ED visit. We compared rates of long-term opioid use among patients treated by high versus low quartile prescribers, adjusting for patient demographic, clinical characteristics, and ED diagnoses.

RESULTS: We identified 57,738 and 86,393 opioid-naïve Veterans managed by 362 and 440 low and high quartile prescribers, respectively. Patient characteristics were similar across groups. ED opioid prescribing rates varied more than threefold between the low and high quartile prescribers within hospitals (6.4% vs. 20.8%, p < 0.001). The frequency of long-term opioid use was higher among Veterans treated by high versus low quartile prescribers, though above the threshold for statistical significance (1.39% vs. 1.26%; adjusted OR 1.11, 95% CI 0.997-1.24, p = 0.056). In subgroup analyses, there were significant associations for patients with back pain (adjusted OR 1.25, 95% CI 1.01-1.55, p = 0.04) and for those with a history of depression (adjusted OR 1.28, 95% CI 1.08-1.51, p = 0.004).

CONCLUSIONS: ED physician opioid prescribing varied by over 300% within facility, with a statistically non-significant increased rate of long-term use among opioid-naïve Veterans exposed to the highest intensity prescribers.

It is unknown to what extent rising drug costs are due to inflation in the prices of existing drugs versus the entry of new products. We used pricing data from First Databank and pharmacy claims from UPMC Health Plan to quantify the contribution of new versus existing drugs to the changes in costs of oral and injectable drugs used in the outpatient setting in 2008-16. The costs of oral and injectable brand-name drugs increased annually by 9.2 percent and 15.1 percent, respectively, largely driven by existing drugs. For oral and injectable specialty drugs, costs increased 20.6 percent and 12.5 percent, respectively, with 71.1 percent and 52.4 percent of these increases attributable to new drugs. Costs of oral and injectable generics increased by 4.4 percent and 7.3 percent, respectively, driven by new drug entry. The rising costs of generic and specialty drugs were mostly driven by new product entry, whereas the rising costs of brand-name drugs were due to existing drug price inflation.

BACKGROUND: Mitigating the risks of adverse outcomes from opioids is critical. Thus, the Veterans Affairs (VA) Healthcare System developed the Stratification Tool for Opioid Risk Management (STORM), a dashboard to assist clinicians with opioid risk evaluation and mitigation. Updated daily, STORM calculates a "risk score" of adverse outcomes (e.g., suicide-related events, overdoses, overdose death) from variables in the VA medical record for all patients with an opioid prescription and displays this information along with documentation of recommended risk mitigation strategies and non-opioid pain treatments. In March 2018, the VA issued a policy notice requiring VA Medical Centers (VAMCs) to complete case reviews for patients whom STORM identifies as very high-risk (i.e., top 1% of STORM risk scores). Half of VAMCs were randomly assigned notices that also stated that additional support and oversight would be required for VAMCs that failed to meet an established percentage of case reviews. Using a stepped-wedge cluster randomized design, VAMCs will be further randomized to conduct case reviews for an expanded pool of patients (top 5% of STORM risk scores vs. 1%) starting either 9 or 15 months after the notice was released, creating four natural arms. VA commissioned an evaluation to understand the implementation strategies and factors associated with case review completion rates, whose protocol is described in this report.

METHODS: This mixed-method study will include an online survey of all VAMCs to identify implementation strategies and interviews at a subset of facilities to identify implementation barriers and facilitators. The survey is based on the Expert Recommendations for Implementing Change (ERIC) project, which engaged experts to create consensus on 73 implementation strategies. We will use regression models to compare the number and types of implementation strategies across arms and their association with case review completion rates. Using questions from the Consolidated Framework for Implementation Research, we will interview stakeholders at 40 VAMCs with the highest and lowest adherence to opioid therapy guidelines.

DISCUSSION: By identifying which implementation strategies, barriers, and facilitators influence case reviews to reduce opioid-related adverse outcomes, this unique implementation evaluation will enable the VA to improve the design of future opioid safety initiatives.

TRIAL REGISTRATION: This project is registered at the ISRCTN Registry with number ISRCTN16012111 . The trial was first registered on 5/3/2017.

OBJECTIVES: To evaluate how changes in generic drug prices and the incidence of abrupt price increases varied with the number of manufacturers supplying each drug.

STUDY DESIGN: Analysis of 2005 to 2016 monthly wholesale acquisition costs (WACs) and University of Pittsburgh Medical Center Health Plan counts of pharmacy claims for National Drug Codes (NDCs) for generic drugs.

METHODS: Each year, NDCs were categorized according to the number of manufacturers offering each combination of active ingredient and dosage form: 1 to 3, 4 to 7, and more than 7. For every month from January 2006 to January 2017, we estimated the 12-month change in WAC (eg, 12-month change in January 2006 was calculated as the difference in WAC between January 2006 and January 2005, divided by the WAC in January 2005), before and after weighting each NDC by counts of pharmacy claims. We evaluated the proportion of NDCs that had large price increases, greater than 20%, 50%, 100%, and 500% within a year.

RESULTS: Before 2010, price changes were higher for drugs supplied by a lower number of manufacturers; however, after 2010, prices increased sharply, and drugs supplied by 4 to 7 manufacturers showed increases similar to or higher than those supplied by 1 to 3. In 2013, prices increased by an average of 29% for drugs supplied by 1 to 3 and 4 to 7 manufacturers, and 10% for more than 7. Price changes increased after weighting by counts of pharmacy claims, demonstrating that price increases disproportionately affected widely used drugs. The proportion of NDCs from drugs supplied by 1 to 3 manufacturers that doubled in price within a year was 3.6 times higher in 2012 to 2015 than in 2005 to 2009 (4.6% vs 1.3%, respectively).

CONCLUSIONS: Increases in generic drug prices are concerning because they affected widely used drugs and suggest that generic drug prices may be increasingly insensitive to competition.

OBJECTIVE: To characterize primary care physicians' (PCPs') perceptions of the reasons patients receive opioid medications from both VA and non-VA healthcare systems.

DESIGN: Qualitative.

SETTING: Department of Veterans Affairs (VA).

PARTICIPANTS: Forty-two VA PCPs who prescribed opioids to at least 15 patients and who practiced in Massachusetts, Illinois, or Pennsylvania.

METHODS: Thirty-minute, semistructured telephone interviews were conducted in 2016, addressing topics regarding PCPs' experiences and perspectives on patients who use both VA and non-VA healthcare systems to obtain prescription opioids. The analysis focused on two questions: attributes that PCPs believe characterize dual-use patients and reasons that PCPs believe patients obtain opioids from both VA and non-VA sources.

RESULTS: PCPs identified multiple attributes of, and reasons for, patients obtaining opioid medications from both VA and non-VA healthcare systems, including pain issues, opioid misuse, having healthcare managed through multiple healthcare systems, and transferring care between systems. More than half of the PCPs identified addiction and diversion as key attributes and reasons why patients obtain prescription opioids from multiple sources. PCPs also identified several behavioral and psychological factors as attributes of these patients.

CONCLUSIONS: PCPs within the VA have varying perceptions of patients obtaining opioid medications from multiple healthcare systems, with pain complaints and opioid misuse as the primary themes. This knowledge about PCPs' perceptions can be incorporated into interventions to better manage pain and prescription opioid use by VA patients.

IMPORTANCE: Current approaches to identifying individuals at high risk for opioid overdose target many patients who are not truly at high risk.

OBJECTIVE: To develop and validate a machine-learning algorithm to predict opioid overdose risk among Medicare beneficiaries with at least 1 opioid prescription.

DESIGN, SETTING, AND PARTICIPANTS: A prognostic study was conducted between September 1, 2017, and December 31, 2018. Participants (n = 560 057) included fee-for-service Medicare beneficiaries without cancer who filled 1 or more opioid prescriptions from January 1, 2011, to December 31, 2015. Beneficiaries were randomly and equally divided into training, testing, and validation samples.

EXPOSURES: Potential predictors (n = 268), including sociodemographics, health status, patterns of opioid use, and practitioner-level and regional-level factors, were measured in 3-month windows, starting 3 months before initiating opioids until loss of follow-up or the end of observation.

MAIN OUTCOMES AND MEASURES: Opioid overdose episodes from inpatient and emergency department claims were identified. Multivariate logistic regression (MLR), least absolute shrinkage and selection operator-type regression (LASSO), random forest (RF), gradient boosting machine (GBM), and deep neural network (DNN) were applied to predict overdose risk in the subsequent 3 months after initiation of treatment with prescription opioids. Prediction performance was assessed using the C statistic and other metrics (eg, sensitivity, specificity, and number needed to evaluate [NNE] to identify one overdose). The Youden index was used to identify the optimized threshold of predicted score that balanced sensitivity and specificity.

RESULTS: Beneficiaries in the training (n = 186 686), testing (n = 186 685), and validation (n = 186 686) samples had similar characteristics (mean [SD] age of 68.0 [14.5] years, and approximately 63% were female, 82% were white, 35% had disabilities, 41% were dual eligible, and 0.60% had at least 1 overdose episode). In the validation sample, the DNN (C statistic = 0.91; 95% CI, 0.88-0.93) and GBM (C statistic = 0.90; 95% CI, 0.87-0.94) algorithms outperformed the LASSO (C statistic = 0.84; 95% CI, 0.80-0.89), RF (C statistic = 0.80; 95% CI, 0.75-0.84), and MLR (C statistic = 0.75; 95% CI, 0.69-0.80) methods for predicting opioid overdose. At the optimized sensitivity and specificity, DNN had a sensitivity of 92.3%, specificity of 75.7%, NNE of 542, positive predictive value of 0.18%, and negative predictive value of 99.9%. The DNN classified patients into low-risk (76.2% [142 180] of the cohort), medium-risk (18.6% [34 579] of the cohort), and high-risk (5.2% [9747] of the cohort) subgroups, with only 1 in 10 000 in the low-risk subgroup having an overdose episode. More than 90% of overdose episodes occurred in the high-risk and medium-risk subgroups, although positive predictive values were low, given the rare overdose outcome.

CONCLUSIONS AND RELEVANCE: Machine-learning algorithms appear to perform well for risk prediction and stratification of opioid overdose, especially in identifying low-risk subgroups that have minimal risk of overdose.