Spillover Effects of Restrictive Drug Formularies: A Case Study of PacifiCare in California

January 15, 2005
Y. Richard Wang, MD, PhD

Mark V. Pauly, PhD

Volume 11, Issue 1

Background: A restrictive drug formulary may influence howphysicians treat other, unaffiliated patients, a phenomenon knownas the "spillover effect." In a previous study we found significantspillover effects from Maine's Medicaid formulary.

Objective: To determine whether similar spillover effects existfor private insurers with less restrictive formularies and less dominantmarket presence.

Study Design: We treated PacifiCare's formulary changes forproton pump inhibitors (PPIs) as a natural experiment and studiedwhether these changes spilled onto non-PacifiCare patients inCalifornia. Rabeprazole and pantoprazole are the newly preferredPPI products.

Methods: We analyzed the physician-level before-and-afterchanges in prescribing of rabeprazole and pantoprazole forPacifiCare and non-PacifiCare patients. We also estimated theeffect of PacifiCare share of practice on spillover effects using linearregressions.


Results: The number of rabeprazole and pantoprazole prescriptionsincreased simultaneously for non-PacifiCare patients and theincrease was positively associated with PacifiCare share of practice.For non-PacifiCare prescriptions, a 10% increase in PacifiCareshare of practice led to a 3.3% share increase for rabeprazole and a1.6% share increase for pantoprazole, respectively (both < .001).

Conclusions: PacifiCare's PPI formulary changes generated significantspillover effects onto non-PacifiCare patients in California.

(Am J Manag Care. 2005;11:24-26)

A restrictive drug formulary may influence howphysicians treat other, unaffiliated patients, aphenomenon known as "spillover effects. "Aphysician faces a mix of drug formularies from healthplans and may not realistically have time to keep upwith them. Consider the simple case in which all of aphysician's patients are either on a single restrictive formularyor have an open formulary with no restrictions.If the physician prefers to use only 1 of a number oftherapeutically similar products for all patients, therestrictive formulary will spill into the open-formularyplan. In other words, patients with an open formularymay only be prescribed products preferred by therestrictive formulary. Of course, other concerns mayoverride the physician's desire for convenience, forexample, if drugs have significant differences in clinicalappropriateness or if some patients have to pay significantlymore out of pocket.

We recently demonstrated that spillover effects existfor Maine's Medicaid program, which adopted a pantoprazole-only, closed formulary for the proton pumpinhibitor (PPI) class.1 The introduction of the Medicaidformulary was associated with a simultaneous increasein pantoprazole prescriptions for other third-party payersand cash-paying patients in Maine, and the increasewas positively associated with Medicaid share of aphysician's practice.

It is unknown whether similar spillover effects existfor private plans that often have less dominant marketpresence and less restrictive formularies. In otherwords, private plans can hardly mimic Medicaid's 29%market share in Maine or its achievement of more than80% compliance within 2 months. We chose to studyPacifiCare, one of the largest private health plans inCalifornia and accounting for about 9% of total PPI prescriptions.In addition to its size, PacifiCare was knownfor its ability to ensure formulary compliance. An industrysurvey of pharmaceutical executives in 2001 ratedPacifiCare second only to Kaiser in "ability to impact Rxproduct utilization and market share."2


Table 1 describes PacifiCare's formulary changes forthe PPI class in 2000 and our natural experiment studydesign (baseline and study period). We confirmed theformulary changes through 2 independent databases—InfoScan Formulary by MediMedia USA (Yardley, Pa)and Formulary Focus by IMS Health (PlymouthMeeting, Pa). Briefly, omeprazole and lansoprazole wereon formulary before rabeprazole was added in June2000; omeprazole and lansoprazole were subsequentlydropped in September 2000; and pantoprazole,launched in May 2000, wasadded to the formulary inOctober 2000. In later regressionanalyses, we designatedMarch 2000 to May 2000 asthe baseline period andDecember 2000 to February2001 as the study period. Thelatter choice was made toavoid potential biases associatedwith the launch ofesomeprazole in March 2001,although the results were qualitativelysimilar using alternativestudy periods.

Similar to our previous study,1 we used IMS Health'sXponent PlanTrak database (IMS Health, PlymouthMeeting, Pa) to track physician-level prescribing patternsin California from October 1999 to September2001. We used the term "physician" liberally andincluded a small number of nurse practitioners andphysician assistants. To examine how PacifiCare shareof a physician's practice affected formulary complianceand possible spillover effects, we focused on 10 472physicians who were the top 20% in PPI volume or jointlyaccounted for 85% of all PPI prescriptions. We categorizedthese physicians into 4 groups according topercentage of practice from PacifiCare during the 24-month period: 0%, less than 10%, 10% to 20%, or morethan 20%. Mail order prescriptions (less than 7% of total)were excluded as we could not designate them asPacifiCare or non-PacifiCare prescriptions.

In addition, we used linear regressions to quantifythe effect of PacifiCare share of practice on spillovereffects or the change of rabeprazole or pantoprazoleshare of non-PacifiCare prescriptions during the studyperiod. To reducemeasurement errorsfor product share, theregression sample wasfurther restricted to7345 physicians with30 or more non-PacifiCare PPI prescriptionsduring boththe baseline and studyperiods.


Figure 1 showsmonthly changes inthe combined rabeprazoleand pantoprazole share of PacifiCare PPI prescriptionsor formulary compliance across 3 physician groupswith differing percentages of practice from PacifiCare.Formulary compliance was similar after formularychanges but diverged with time, and was higheramong groups with more practice from PacifiCare.Interestingly, the launch of esomeprazole in March2001 led to a small, one-time drop in compliance acrossall groups.

Figure 2 shows monthly changes in the combinedrabeprazole and pantoprazole share of non-PacifiCarePPI prescriptions or so-called spillover effectsacross 4 physician groups differing in PacifiCareshare of practice. Spillover effects appeared to haveincreased with PacifiCare share of practice. Forexample, in September 2001, the combined rabeprazoleand pantoprazole share of non-PacifiCare PPIprescriptions ranged from 18% among physicians withno practice from PacifiCare to 40% among physicianswith more than 20% of practice participating fromPacifiCare.



Linear regression results in Table 2 confirm thatPacifiCare share of practice was a significantly positivepredictor of change in rabeprazole or pantoprazoleshare of non-PacifiCare prescriptions. Forexample, a 10% increase in PacifiCare share of practiceled to a 3.3% share increase for rabeprazole anda 1.6% share increase for pantoprazole respectively(both < .001). Despite the fact that Medicaid had apantoprazole-only formulary in Maine and PacifiCarehas a rabeprazole-and-pantoprazole formulary inCalifornia, the estimated spillover effects for pantoprazolewere similar between PacifiCare (0.16) andMedicaid (0.14).1 The 2 values were small because ofthe effects of unobserved variables. Using alternativestudy periods led to qualitatively similar results (datanot shown).


We found thatPacifiCare's new 2-PPIformulary spilled ontonon-PacifiCarepatients in Californiaand that spillovereffects increased withPacifiCare share ofpractice. These resultsverified the robustnessof our previous Mainestudy.1 Spillover effectslikely exist for allrestrictive formularies(of private or publichealth plans) and inother therapeuticclasses, although the extent ofspillover probably depends onformulary restrictiveness andproduct interchangeability.

Although such spillovereffects may enhance therapeuticcompetition and facilitatethe dissemination ofcost-effective drugs throughoutthe whole healthcare system,physicians and patientsshould be better informed offormulary design so as to optimizeindividual patient treatment.Spillover effects areespecially relevant to the newMedicare drug benefit; should Medicare allow prevalentuse of restrictive formularies, the size of Medicareimplies substantial spillover from restrictive Medicareformularies onto non-Medicare patients.


Although funded by AstraZeneca Pharmaceuticals, this study does notrepresent the view of AstraZeneca Pharmaceuticals. We wish to thank Y. AileenLin, MS, for help with data analysis, 2 anonymous reviewers at this journal forhelpful comments, and Lisa Croll, BA, for administrative assistance.

From the Public Policy Department, AstraZeneca Pharmaceuticals, Wilmington, Del(YRW); and the Leonard Davis Institute of Health Economics (YRW, MVP) and theDepartment of Health Care Systems (MVP), University of Pennsylvania, Philadelphia, Pa.

This study was funded by AstraZeneca Pharmaceuticals.

Address correspondence to: Y. Richard Wang, MD, PhD, Public Policy Department,AstraZeneca Pharmaceuticals, FOC 3 CE 317, 1800 Concord Pike, Wilmington, DE 19850-5437. E-mail: y.richard.wang@astrazeneca.com.

Am J Manag Care.

1. Wang YR, Pauly MV, Lin YA. Impact of Maine's Medicaid drug formularychange on non-Medicaid markets: spillover effects of a restrictive drug formulary.2003;9:686-696.

2. Health Industries Research Center. National accounts: marketing challengesand opportunities for the mid 2000s. Santa Cruz, Calif: Health IndustriesResearch Center; Fall/Winter 2001. Available at: www.hire.com. AccessedDecember 1, 2004.