BACKGROUND/RATIONALE:
Spatiotemporal (ST) analysis has great potential to provide meaningful information about rapidly developing prescribing patterns, and patterns of reductions in prescribing, in real time. The Opioid Safety Initiative (OSI) provides an ideal proof of concept quality improvement venue to test how these methods can be used to assess changes in prescribing patterns through surveillance in real time. The OSI grew out of an accumulating body of evidence that some opioid use is unsafe and may be contributing to harm for our Veteran patients. OSI represents the VA application of this evidence through an organized initiative. Major goals of OSI include: 1) Limiting the use of high-dose opioids; 2) Limiting the co-prescription of opioids and benzodiazepines; and 3) Developing toolkits and other implementation strategies for providing safer alternatives for treating chronic pain.

OBJECTIVE(S):
(1) We will identify Veterans prescribed 200 mg morphine equivalents ("high-dose opioids") and calculate counts by VAMC. We will reconcile these counts with other data sources, including QUERI Substance Use Disorders (SUD) measures and the OSI Dashboard and toolkit implementation from FY2014 through FY2015; (2) We will apply ST methods to describe changes in high-dose opioid prescribing quarter by quarter, starting with FY2014 and continuing through the four quarters of FY2015; (3) We will expand the ST analysis to track other OSI-relevant metrics, such as 100 mg morphine equivalents and both incident prescribing ("new starts") and cessation ("de-prescribing") of co-prescription of opioids and benzodiazepines.

METHODS:
We will link patient, prescriber, and facility data from tables in the Corporate Data Warehouse (CDW) databases on prescribing, along with key facility descriptors from the VHA Complexity Model, and other sources. This approach builds upon the existing methods we have learned in amassing data from previous funded research to examine the spread of second generation antipsychotics (SGAs). We will follow similar thinking in drawing data from the CDW for this study. We then will identify counts by site for patients prescribed 200 mg morphine equivalents ("high-dose opioids") to track short term quarterly trends in response to the OSI (Objective 1). Then we will calculate SpatioTemporal Clusters (STCs) for high dose opioids (Objective 2) and extend to the metrics of lower dose 100 mg morphine equivalents and co-prescription of opioids and benzodiazepines (Objective 3).

All VHA prescriptions for Veterans treated with opioids for non-cancer chronic pain will be included in an analysis of prescribing patterns at the VAMC level. VAMC data will be nested within the VHA Complexity Model level of facilities. After obtaining access to the CDW data, high dose opioid data by site for FY2014 and the first two quarters of FY2015 will be developed and compared to OSI cube data in VHA Support Service Center databases. STC analysis will be generated on this data, and then incrementally, we will test the impact of adding the last two quarters of FY2015. At this point, the 100 mg dose opioids and co-prescription of opioids and benzodiazepines will be added and tested in the same way and STC identified sites will be compared to the VHA Complexity Model, to test what VAMC characteristics are associated with the highest likelihood of developing and sustaining STCs. Initially, we will investigate the STCs by quarter, but we also will test whether variation is great enough to determine STCs by month as well.

FINDINGS/RESULTS:
Pharmacy records and diagnosis codes from CDW for October 1, 2013 through September 30, 2015 have been processed and analyzed. We have identified over one million unique patients (1,031,789) prescribed opioids in the outpatient setting in FY2014 and nearly that many again in FY2015 (916,173). Data have been assembled by site and month, patients were classified by month in terms of high-dose opioid prescribing and co-prescribing of benzodiazepines, and the monthly data were analyzed using our Spatiotemporal modeling methods. A number spatiotemporal clusters were identified for high and low rates of high dose prescribing (>100 and >200 mg morphine equivalents) and co-prescription of opioids and benzodiazepines. For example, we found five areas with high rates of high-dose prescribing (>200 mg) and seven areas with low rates f high-dose prescribing. Clusters varied in terms of their emergence or disappearance over time in FY2014-2015, suggesting the usefulness of this approach in identifying problem areas and monitoring prescribing practices.

IMPACT:
This project demonstrates the use of spatiotemporal (ST) analysis as a method for use in monitoring specific prescribing patterns and in the identification of clusters or "hot spots" of prescribing associated with safety issues or inappropriate practices. For this surveillance demonstration we employ the Opioid Safety Initiative (OSI) as a demonstration of how ST methods can provide valuable information about progress with a high-priority VA initiative. We have evaluated increases and decreases in specific patterns of opioid prescribing using space-time statistics and software.

In addition to guiding OSI implementation, our work will have broader implications. Developing capability to contribute real-time surveillance regarding the uptake of medications will interest QUERI and Pharmacy Benefit Management. Moving the analysis "upstream" is preferable to reacting to prescribing patterns after they occur. In addition, enhanced capability to analyze ST trends of medication prescribing will be instrumental in insights on why differences arise. That is, by identifying ST trends, we will have newfound capability to examine why some sites change practice more rapidly than others.

---

External Links for this Project

NIH Reporter

Grant Number: I21HX001947-01A1
Link: https://reporter.nih.gov/project-details/9077812

---

PUBLICATIONS:

None at this time.

DRA: Substance Use Disorders, Health Systems
DRE: Treatment - Observational
Keywords: Patient Safety, Practice Patterns/Trends, Statistical Methods, Substance Use and Abuse
MeSH Terms: none