Roughly 1.5 million preventable adverse drug events occur in the US annually at a cost in excess of $4 billion. Many of these events result from inappropriate prescribing and efforts to improve prescribing practices would have a major impact on outcomes. This is particularly true for the many older adult veterans who suffer from multiple comorbid conditions and require complex medication regimens. The ability to improve prescribing depends on valid methods for measuring medication appropriateness.
Aim 1. Contrast incidence and prevalence rates of selected prescribing quality indicators.
AIM 2. Identify patient, prescriber, and system variation in the selected prescribing quality indicators.
AIM 3. Compare the performance of prevalence and incidence-based prescribing quality indicators in predicting adverse outcomes using traditional multivariable regression-based methods.
AIM 4. Determine the impact of selection bias on the relationship between selected prescribing quality indicators and adverse health outcomes using instrumental variable methods.
Aim 1 used national VA data from 2003 to 2012 to determine the prevalence and incidence of a selected panel of prescribing quality indicators, including drugs to avoid, drug-drug interactions, therapeutic duplication, and noncompliance. Aim 2 identified sources of variation in prescribing quality indicator violations, with variables selected according to a conceptual framework based on the Donabedian quality model of structure, process and outcome. Aim 3 compared the performance of prevalence and incidence-based prescribing quality indicators in predicting adverse outcomes, adjusting for patient complexity using traditional multivariable regression and propensity scores approaches. Aim 4 determined the impact of selection bias on the relationship between prescribing quality indicators and adverse health outcomes using instrumental variable methods.
Aim 1: The prevalence of potentially inappropriate prescribing according to Zhan criteria decreased from 20.8% to 17.7% from 2003 to 20012, and incidence decreased from 9.8% to 7.9% during this period. Similar findings were seen with other potentially inappropriate medication lists, including the 2003 Beers criteria and the HEDIS indicator for high risk medications in the elderly (HRME). These findings suggest that VA prescribers have become less likely to write new prescriptions for potentially inappropriate medications, which has led to an overall decrease in prevalence. In contrast, the point prevalence of therapeutic duplication increased from 3.3% to 5.9% from 2003 to 2012.
Aim 2: Patient and facility level factors associated with the use of potentially inappropriate medications according to HRME criteria were examined in two models, one predicting prevalent prescribing and the other predicting incident prescribing. Patient variables associated with increased risk for HRME use in both models included: younger age, female sex, greater number of primary care visits, and a greater number of unique prescribing clinicians. Having a visit with a geriatrician or clinical pharmacist was associated with lower rates of HRME use in both models. Of facility-level characteristics, the only variable that was consistently associated with HRME use was a greater proportion of rural residents among patients receiving primary care at that facility. However, rural residence was not significantly associated with HRME use at the patient level. The only variable that showed substantially different associations in the two models was number of drugs, which was highly correlated with prevalent HRME use, but was uncorrelated incident HRME use. This finding has important implications in determining whether prevalence-based versus incidence-based indicators of prescribing quality are more useful for specific applications. The fact that prevalence-based indicators tend to be highly correlated with number of drugs leads to a number of methodological and interpretation problems that are not readily solvable. Thus, the lack of association with the incidence-based HRME indicator is a highly desirable property for many applications.
Aim 3: Using national VHA administrative data, veterans age >= 65 years in FY09 with regular VHA medication use and >= 1 primary care visit in the prior year were selected. The exposed group was comprised of new HRME starters; defined as a first prescription in FY09 preceded by 365 days with no HRME use. Controls were patients with no HRME use in FY09 where a random non-HRME prescription date served as the index date. Adverse outcomes included emergency department visit, inpatient hospitalization, or death within 90 days of index, using VHA plus merged Medicare data. The unadjusted frequency of adverse outcomes was 27.9% vs 16.2% in the HRME and control groups, respectively. New HRME exposure remained significantly associated with adverse outcomes after adjustment (OR: 1.67; 95% CI: 1.64, 1.71). This finding persisted across sensitivity analyses: 30-day adverse events (OR: 1.79; 95% CI: 1.74, 1.84), VHA events only (OR: 2.15; 95% CI: 2.10, 2.20), and substitution of non-HRME new drug starters as controls (OR: 1.67; 95% CI: 1.63, 1.70). In a separate analysis, we compared prevalent HRME users to non-user controls (OR: 1.34; 95% CI: 1.32, 1.35).
Aim 4: Preliminary results are pending at this time.
We anticipate the results of this study will have three specific impacts. First, our findings provide a detailed characterization of prescribing quality in VA across multiple clinical domains, over time, and between regions and facilities. Second, we have established a methodological framework for examining the validity of prescribing quality in VA, with a particular emphasis on incident prescribing events, as opposed to prevalent events. Most importantly, these findings were used to inform the development of a healthcare informatics tool called the Triage Algorithm for PACT Pharmacy Services (TAPPS). TAPPS was designed to help PACT pharmacists efficiently target primary care patients most likely to benefit from clinical pharmacy series. The effectiveness of this tool is currently being studied in an HSR&D funded Merit grant.
- Abrams TE, Lund BC, Alexander B, Bernardy NC, Friedman MJ. Geographical diffusion of prazosin across Veterans Health Administration: Examination of regional variation in daily dosing and quality indicators among veterans with posttraumatic stress disorder. Journal of rehabilitation research and development. 2015 Sep 1; 52(5):619-27.
- Lund BC, Carrel M, Gellad WF, Chrischilles EA, Kaboli PJ. Incidence- Versus Prevalence-Based Measures of Inappropriate Prescribing in the Veterans Health Administration. Journal of the American Geriatrics Society. 2015 Aug 1; 63(8):1601-7.
- Lund BC, Schroeder MC, Middendorff G, Brooks JM. Effect of hospitalization on inappropriate prescribing in elderly Medicare beneficiaries. Journal of the American Geriatrics Society. 2015 Apr 1; 63(4):699-707.
- Patterson BJ, Kaboli PJ, Tubbs T, Alexander B, Lund BC. Rural access to clinical pharmacy services. Journal of the American Pharmacists Association : JAPhA. 2014 Sep 1; 54(5):518-25.
- Lund BC, Kaboli PJ. Adverse Outcomes of High Risk Medication in Older Adult Veterans. Paper presented at: VA HSR&D / QUERI National Meeting; 2015 Jul 9; Philadelphia, PA.
- Lund BC. Rural Access to Clinical Pharmacy Services. Poster session presented at: AcademyHealth Annual Research Meeting; 2014 Jun 9; San Diego, CA.
- Lund BC, Ernst ME. The Comparative Effectiveness of Chlorthalidone versus Hydrochlorothiazide Veterans: Nurse Attitudes and Performance. Poster session presented at: VA HSR&D / QUERI National Meeting; 2012 Jul 16; National Harbor, MD.