Impact of large scale, multicomponent intervention to reduce proton pump inhibitor overuse in integrated healthcare system: difference-in-difference study

Abstract

Objective: To determine how a large scale, multicomponent, pharmacy based intervention to reduce proton pump inhibitor (PPI) overuse affected prescribing patterns, healthcare utilization, and clinical outcomes.

Design: Difference-in-difference study.

Setting: US Veterans Affairs Healthcare System, in which one regional network implemented the overuse intervention and all 17 others served as controls.

Participants: All individuals receiving primary care from 2009 to 2019.

Intervention: Limits on PPI refills for patients without a documented indication for long term use, voiding of PPI prescriptions not recently filled, facilitated electronic prescribing of H2 receptor antagonists, and education for patients and clinicians.

Main outcome measures: The primary outcome was the percentage of patients who filled a PPI prescription per 6 months. Secondary outcomes included percentage of days PPI gastroprotection was prescribed in patients at high risk for upper gastrointestinal bleeding, percentage of patients who filled either a PPI or H2 receptor antagonist prescription, hospital admission for acid peptic disease in older adults appropriate for PPI gastroprotection, primary care visits for an upper gastrointestinal diagnosis, upper endoscopies, and PPI associated clinical conditions.

Results: The number of patients analyzed per interval ranged from 192 607 to 250 349 in intervention sites and from 3 775 953 to 4 360 868 in control sites, with 26% of patients receiving PPIs before the intervention. The intervention was associated with an absolute reduction of 7.3% (95% confidence interval -7.6% to -7.0%) in patients who filled PPI prescriptions, an absolute reduction of 11.3% (-12.0% to -10.5%) in PPI use among patients appropriate for gastroprotection, and an absolute reduction of 5.72% (-6.08% to -5.36%) in patients who filled a PPI or H2 receptor antagonist prescription. No increases were seen in primary care visits for upper gastrointestinal diagnoses, upper endoscopies, or hospital admissions for acid peptic disease in older patients appropriate for gastroprotection. No clinically significant changes were seen in any PPI associated clinical conditions.

Conclusions: The multicomponent intervention was associated with reduced PPI use overall but also in patients appropriate for gastroprotection, with minimal evidence of either clinical benefits or harms.

Fig 1

Percentage of patients receiving proton pump inhibitor (PPI) prescriptions (top) and percentage of time with co-prescribed PPI in patients appropriate for gastroprotection (bottom) per 6 month period before and after implementation for Veterans Integrated Service Network 17 (VISN 17) and all other Veterans Affairs (VA) sites (controls), with fitted lines showing predicted percentages based on difference-in-difference regression models. For both outcomes, fitted lines were generated using predicted values for each time interval and site using difference-in-difference regression models after exclusion of data from 1 year implementation period (2013-14). Table 2 shows significant absolute changes associated with VISN 17 intervention in two outcomes based on models and their 95% confidence intervals, which exclude zero. Predictors included time (centered at implementation period), post-implementation period indicator, VISN 17 indicator (intervention), and interaction of VISN 17 by post-implementation. Robust standard errors were used to account for heteroscedasticity. For top panel, patients were included in numerator if they had received any PPI prescription dispensed by VA health system. Denominator included all patients who had ≥2 primary care visits during 2 years before interval. For bottom panel, patients were considered appropriate for gastroprotection during time that they had medication possession of ≥2 antithrombotic drugs with at least daily dosing (including anticoagulants, aspirin, and P2Y12 inhibitors) or an antithrombotic drug together with a non-steroidal anti-inflammatory drug with at least daily dosing. Patients were considered to have gastroprotection during time in which they had medication possession of a PPI with at least daily dosing. Only medications dispensed by VA were included

Fig 2

Incidence rate of hospital admission for acid peptic disease during person time when patients aged ≥65 filled prescriptions for ≥2 anticoagulant, antiplatelet, or non-steroidal medications, per 6 month period before and after implementation for Veterans Integrated Service Network (VISN) 17 and all other Veterans Affairs (VA) sites (controls), with fitted lines showing predicted incidence rates based on difference-in-difference regression model. Regression model was fitted to data after exclusion of implementation period data and was used to estimate absolute change in outcomes associated with VISN 17 intervention. No evidence for VISN 17 effect during post-implementation period was seen as indicated by interaction term of post-implementation by VISN 17 not being significant (see table 3). Hospital admissions for bleeding and non-bleeding acid peptic disease were included in outcome (supplement 1). Data were drawn from both VA health system and Centers for Medicare and Medicaid Services data to capture hospital admissions that occurred in VA health system and at outside facilities