Clinical and Financial Outcomes Associated With a Proton Pump Inhibitor Prior-Authorization Program in a Medicaid Population

January 15, 2005
Thomas Delate, PhD
Thomas Delate, PhD

,
Douglas E. Mager, BS
Douglas E. Mager, BS

,
Jagat Sheth, PhD
Jagat Sheth, PhD

,
Brenda R. Motheral, PhD
Brenda R. Motheral, PhD

Volume 11, Issue 1

Objective: To examine the clinical and financial outcomesassociated with a proton pump inhibitor (PPI) prior-authorizationpolicy.

Study Design: Interrupted time-series analyses of antisecretoryprescription drug claims. Separate 6-month retrospective cohortanalyses were conducted to estimate the clinical and financialeffects of the policy.

Patients and Methods: More than 1.2 million Medicaidenrollees, with subgroup analyses of 5965 continuously eligible,potential antisecretory medication users. Measures included antisecretorydrug expenditures, proportions of patients with at least 1gastrointestinal diagnosis and gastrointestinal-related ambulatoryand inpatient medical service visit, and subsequent gastrointestinal-related and total medical service expenditures.

P

P

Results: There was a 90.9% decrease in PPI per-member-per-monthexpenditures and a 223.2% increase in histamine2-receptorantagonist (H2A) per-member-per-month expenditures in themonth immediately following the implementation of the policy (< .001 for both). A greater proportion (80.7%) of prior-authorizationeligible enrollees who received a PPI had at least 1 diagnosisfor a gastrointestinal condition than enrollees who received an H2A(64.1%) or no antisecretory drugs (48.4%) (< .001 for both). Two-part,finite mixture regression analyses indicated that the enrolleeswho received an H2A or no antisecretory drugs were no more likelyto have incurred greater total medical care expenditures thanenrollees who received a PPI.

Conclusion: Prior authorization for PPIs had the effect of reducinguse of high-cost PPIs, while encouraging use of lower costingH2As without evidence of adverse medical consequences.

(Am J Manag Care. 2005;11:29-36)

Increasing enrollment in Medicaid, combined withrising drug and health service costs, has led to largeincreases in Medicaid expenditures.1,2 Approximately20% of states' budgets are devoted to financingfor Medicaid, and expenditures for Medicaid are amongthe fastest growing items in the federal and state budgets.3 On average, fee-for-service Medicaid programshave seen rises of 19.6% per year in ambulatory prescriptiondrug expenditures between fiscal years 1998and 2001.4 Therefore, Medicaid programs are takingassertive measures to manage drug expenditures.5-8

The Omnibus Reconciliation Act of 1990, as amendedin 1993, allows prior authorization (PA) as a meansof drug cost containment.9 Prior authorization restrictsthe use of specific medications by requiring an advanceapproval by the Medicaid program or its agent for thedrug before dispensing to qualify for reimbursement.10High-cost drugs that have a history of inappropriate useand effective drugs for which there are lower costingtherapeutic equivalents typically are placed in PA programs.11 Prior authorization is designed to allow patientaccess to essential pharmacotherapies while promotingcost-effective prescription drug use.

States may require PA for any drug 6 months afterFood and Drug Administration marketing approval.11 Itis estimated that more than 40 states and the District ofColumbia have some type of drug PA policy.4 Notwithstandingthe widespread use of these programs, limitedempirical evidence exists on their effects. The availableevidence suggests that such programs reduce drugexpenditures without incurring increased medical servicesuse10; however, only 3 retrospective evaluations ofMedicaid PA programs were found in the peer-reviewedliterature.11-13 As PA programs expand to include otherclasses of drugs, research is needed to evaluate the clinicaland economic effects of the programs.

Proton pump inhibitors (PPIs) are an example of atherapy class that is a candidate for PA. Proton pumpinhibitors are indicated for short-term therapy of acuteupper gastrointestinal (GI) disorders (eg, peptic ulcersand esophagitis), pathologic gastric hypersecretory conditions(eg, Zollinger-Ellison syndrome), and maintenancetherapy with healed ulcers and erosiveesophagitis.14 However, an estimated 30% to 70% ofpatients with painful reflux symptoms do not have GIconditions with sustained tissue damage.15 An alternatetreatment for nonacute hypersecretory conditionsexists in lower-cost histamine2-receptor antagonists(H2As).16,17 To channel potential PPI users to less expensiveH2A alternatives when clinically appropriate (eg,dyspepsia and reflux without complications), the fee-for-service Medicaid program implemented a PA programthat used diagnosis-and risk-based criteria18-23 toestablish medical necessity (eg, erosive or ulcerative GIconditions) for approval of PPIs. During the 12 monthsbefore the implementation of the PPI PA policy, PPIsaccounted for 5.6% ($45.5 million) of the net pharmacyexpenditures and ranked first in expenditures among alldrug therapy classes for the Medicaid program (data notshown).

This study used population-based, retrospective, longitudinalanalyses of pharmacy and medical claims toevaluate the effects of the Medicaid program's PA policyfor PPIs. The objective of the study was to examinethe effectiveness of the PPI PA program from the perspectiveof the Medicaid program payer11,24-26 in regardsto (1) its effect on antisecretory drug (ie, PPIs andH2As) expenditures and use, (2) the channeling ofPPIs to those patients demonstrating a medical needfor a PPI, and (3) the incurrence of unintended medicalconsequences.

METHODS

Prior-Authorization Policy

Beginning on February 1, 2002, the Medicaid programrequired that PA be obtained for all PPI prescriptions(esomeprazole magnesium, lansoprazole, omeprazole,pantoprazole sodium and rabeprazole sodium) to qualifyfor reimbursement. Diagnosis-and risk-based clinicalcriteria were established for the PA program.18-23Prescribers and pharmacies participating in theMedicaid program were sent notification of the PArequirement before the implementation of the policy. Inaddition, notifications of the impending requirementwere appended as a banner message to Medicaid programremittance advices sent to prescribers and pharmacies.When a patient presented at the pharmacy witha PPI prescription without first receiving PA approval,pharmacists were alerted during the prescription adjudicationprocess that authorization was required.Prescribers or pharmacists submitted PA requests bymail, fax, or toll-free telephone call to the Medicaid program'spharmacy benefit manager agent, and theserequests were initially reviewed by a PA representative.Requests that were not approved during the initialphase were then reviewed by a pharmacist. Theresponse was handled at the point of service for telephonecalls, within 24 hours for a faxed request, orwithin 48 hours for a mailed request. For those patientrequests that were denied, the prescriber was madeaware of the 2 levels for appeals. The PA program wasmodified on April 1, 2002, because of concernsexpressed by prescribers and patient advocates regardingthe GI-related and other medical conditions forwhich initial therapy with PPIs was approvable.

Helicobacter pylori

Authorizations were valid for 6 months. Patientsrequesting a prescription for esomeprazole or rabeprazolewere required to document therapeutic nonresponsewith at least a 2-week course of 1 of the 3 preferred PPIs(lansoprazole, omeprazole, or pantoprazole). In addition,the PA program incorporated step therapy, wherebypatients with nonerosive hypersecretory conditions wererequired to document therapeutic nonresponse with atleast a 2-week course of an H2A (cimetidine, ranitidine,famotidine, or nizatidine) before approval for a PPIwould be granted. Patients with a infection were granted a 1-month PPI supply for usewith antibiotic eradication therapy. The Medicaid programpaid an administrative fee of $20 per PA request.

Study Design and Data Sources

To compare the change in net expenditures (includingthe administrative costs of the PA program) for antisecretorydrugs, this study used the pharmacy benefitmanager's ambulatory pharmacy reimbursement claimrecords for the Medicaid program for the 12 monthsbefore (hereafter referred to as the preperiod) and 12months after (hereafter referred to as the post-period)the implementation of the PA policy (February 1, 2002)(Figure 1). Antisecretory prescriptions were convertedto 30-day supplies (hereafter referred to as claims). Thepharmacy benefit manager's pharmacy edit, PA requestrecord, and eligibility information electronic files wereused to identify continuously eligible patients (hereafterreferred to as subjects) who attempted to have a PPIprescription filled from April 1, 2002, through May 31,2002 (hereafter referred to as the index period) andtheir subsequent PA approval or denial status. Thesedates were chosen because the modified criteria thatwere put in place on April 1, 2002, were, at the time ofthe manuscript preparation, the current criteria usedby the Medicaid program and because they allowed foran adequate follow-up with the data available.14,27Patients without continuous eligibility were not includedbecause these patients would not have complete datato allow for an adequate assessment of their health conditionsand medical service use.

To assess the channeling of PPIs and possible unintendedmedical consequences, subjects' ambulatoryand inpatient medical service claim records wereobtained from MEDSTAT, the Medicaid program's medicalclaims database manager. These medical claimswere examined to assess GI-related diagnoses andexpenditures in the 6 months before (hereafter referredto as baseline) and after (hereafter referred to as followup)subjects' first attempt to have a PPI prescriptionfilled (hereafter referred to as the index date) (Figure 1).

Subject Groups

All Medicaid Enrollees.

Because the PA was appliedto all enrollees in the Medicaid program, the base-caseanalysis consisted of all persons enrolled in theMedicaid program at any time during the preperiod andthe postperiod. The base case provided a broad assessmentof the effects of the PA policy change by allowingfor the adjustment of per month expenditures based onincreases or decreases in the enrollee population (ie,the prevalence of use of antisecretory drugs was notassumed to be static).

Antisecretory Medication Users.

To assess drugchanneling and possible unintended medical consequences,subjects were categorized by cohorts who,subsequent to their index date, (1) had a pharmacy claimfor a PPI (hereafter referred to as PPI users), (2) had apharmacy claim for an H2A (hereafter referred to as H2Ausers), or (3) had no antisecretory drug pharmacy claim(hereafter referred to as nonusers). The latter 2 subjectgroups served as natural comparators to examine theeffects of the policy change on medical care use. Theresearch was performed in accord with the principlesoutlined in the Declaration of Helsinki and recentlyapproved Health Insurance Portability and AccountabilityAct regulations regarding use of personal healthinformation for program evaluation.

Study Outcomes

This study evaluated prescription expenditures anduse and ambulatory services (including physician, otherhealthcare professional, emergency department visits,and GI-related screenings) and inpatient care (includinglong-term care) events and expenditures for the managementof GI-related and all health conditions. Theanalysis was restricted to drugs, services, and facilitycharges for which the Medicaid program made at least apartial payment. Expenditures reported were net paymentsfor the Medicaid program. These are credible outcomesassessable with data available in a retrospectiveanalysis of a PA program.10

Pharmacotherapy.

Antisecretory drugs were identifiedin the pharmacy claim records using generic productindicator codes beginning with "4927" for PPIs and"4920" for H2As.28 Expenditures (ie, [prescription drugcost plus tax plus administration fee plus applicable PAadministration fee] minus copay) and claims for thesedrugs were tabulated separately on a per-member-per-month(PMPM) basis.

Ambulatory Care.

International Classification of Diseases,

Ninth Revision, Clinical Modification (ICD-9-CM)

ICD-9-CM

The ambulatory services andevents were identified by the place of service identifiercodes not assigned to inpatient admissions. Gastrointestinal-related diagnoses and procedures were identifiedwith anddiagnosis related grouping codes (appendix availablefrom the author). Gastrointestinal-related emergencydepartment visits were identified with andCurrent Procedural Terminology codes. Ambulatory GI-relatedscreening visits were identified with CurrentProcedural Terminology codes. A comprehensive definitionof GI-related procedures was used because physiciansand other healthcare professionals might havetreated hypersecretory-related GI conditions during thecourse of visits for other GI problems11 and becauseMedicaid diagnostic data for ambulatory visits may beincomplete for enrollees with dual eligibility in Medicaidand Medicare.29 Gastrointestinal-related expendituresfor ambulatory service visits (including screenings)included only payments for study procedures; GI-relatedexpenditures for emergency department visits includedpayments for all services rendered on the day of the visit.

Inpatient Care.

ICD-9-CM

The inpatient care services andevents and long-term care services were identified bythe place of service identifier codes assigned to inpatientadmissions and long-term care, respectively.Gastrointestinal-related inpatient and long-term carediagnoses and services were identified with ,diagnosis related grouping, and Uniform Bill (UB-92)revenue codes (appendix available from the author).Gastrointestinal-related expenditures for inpatient andlong-term care included payments for all services renderedduring the facility stay associated with the claim.

Analysis

Interrupted time-series analysis with adjustment forthe time trend and effect of the policy implementationwas used to assess the immediate effect of the PA programon antisecretory drug expenditures and use.30 The4 time-series analyses consisted of the 24 PMPM valuesfor the PPI and H2A expenditure and claim outcomevariables for the entire Medicaid program's population.

To assess the channeling of PPIs, logistic regressionanalysis controlling for sex, age in years, race (white vsnonwhite), and Chronic Disease Score, an adjuster forhealth status across time that affects healthcareexpenditures and use,31 was used to compare the proportionsof subjects between the subgroups who had atleast 1 GI-related diagnosis, condition, or screeningduring baseline. Gastrointestinal-related diagnoses werecategorized based on erosive or ulcerative and symptomaticconditions (appendix available from the author).

To assess the incurrence of possible unintended medicalconsequences, logistic regression analysis controllingfor sex, age in years, race, and Chronic Disease Score wasused to compare the proportions of subjects between thesubgroups who had received at least 1 GI-related ambulatoryor inpatient visit or screening during follow-up. Two-part,finite mixture modeling for zero-inflated costdata,32,33 controlling for possible explanatory variables,was used to assess across subgroups during follow-up (1)the probabilities of GI-related or any medical serviceevents and (2) GI-related and total net medical costs.Typical of medical costs, the distributions of the costswere strongly right-skewed, with sizeable proportions ofzero-cost observations. Zero costs were replaced by $0.50,and logarithmic transformationsof the cost values were performed.Explanatory variablesincluded a dummy variable indicatingthe presence or absenceof baseline net medical costs,the logarithm of the baseline netmedical costs, sex, age in years,race, and Chronic Disease Score(appendix available from theauthor).

Incomplete diagnostic andmedical procedure data forMedicare beneficiaries receivingMedicaid benefits (ie, dualenrollees) possibly may biasthe results of Medicaid claimsdata analyses.29 To eliminatethis source of possible bias,sensitivity analysis was performedusing only the datafrom subjects who were not eligible forMedicare benefits.

RESULTS

Enrollment in the Medicaid program populationincreased from 1 142 866 to 1 324 643enrollees during the 24-month study period.Midpoint characteristics indicated a primarilyyoung, female, nonwhite, and nondisabledpopulation (Table 1). Subject groups generallywere comparable and were predominantlyolder, female, and disabled. However, mostPPI users were white, while most H2A usersand nonusers were nonwhite.

P

There were 377 574 PPI and 238 157 H2Aclaims during the preperiod and 105 745 PPIand 467 047 H2A claims during the post-period.Proton pump inhibitor drug expendituresdecreased (from $44.1 million to $13.2 million,including the cost of the administration ofthe PA program) and H2A drug expendituresincreased (from $6.0 million to $13.5 million)from the preperiod to the postperiod. Time-seriesanalyses showed a decrease in the rateof PPI PMPM claims (92.2%) and expenditures(90.9%) and an increase in the rate of H2APMPM claims (98%) and expenditures (223.2%)in the month immediately following the implementationof the PA policy (< .001 for all)(Figure 2). The mean PMPM expenditures forantisecretory drugs decreased 49.9%, from$3.44 in the preperiod to $1.74 in the postperiod,realizing the Medicaid program a netexpenditures decrease of $23.4 million.

P

P

Pharmacy records indicated that 7966 ofthe Medicaid program's enrollees attempted tohave a PPI prescription filled during April andMay of 2002. Of these enrollees, 74.9% werecontinuously eligible during baseline and follow-up. Approximately 50% of these continuouslyeligible enrollees did not attempt to gothrough the PA process. Of those who did notgo through the PA process, 56.5% were lateridentified as having a claim for an H2A, while43.5% did not have a prescription claim for anyantisecretory drug although they could havereceived an H2A without going through the PAprocess. Among enrollees who went throughthe PA process, 95.1% were approved. Of those whowent through the PA process and were denied, 63.7%had a claim for an H2A.Proton pump inhibitor users were more likely to havehad at least 1 diagnosis for a GI condition and GI-relatedscreening during baseline than H2A users and nonusers(< .001 for both) (Table 2). Proton pump inhibitorusers were more likely to have had at least 1 GI-relatedinpatient and ambulatory (including emergency departmentvisits) medical service event during follow-up comparedwith nonusers (< .05), and there was nostatistical difference compared with H2A users (Table 3).No evidence of an increase in the use of GI-relatedscreenings from baseline to follow-up was found amongH2A users (8.3% vs 7.3%) and nonusers (8.5% vs 5.6%).Removing Medicaid and Medicare dual enrollees' datafrom the analyses provided similar results.

P

P

P

P

P

P

P

P

The Medicaid program's net paymentsfor GI-related and total medicalservice visits were $13 558 872and $37 899 979, respectively, duringbaseline vs $12 428 928 and$37 424 469, respectively, duringfollow-up for the subjects. Comparingbaseline with follow-up,increases in the mean net paymentsfor GI-related and total medicalservice visits were not found for H2Ausers or nonusers (Table 4). Finitemixture regression analysis of thefollow-up indicated that, comparedwith PPI users, H2A users (< .01 and> .05) and nonusers (< .001 and< .05) were no more likely to havehad at least 1 GI-related and anymedical service visit, respectively(appendix available from theauthor). Similarly, H2A users (< .01and > .05) and nonusers (< .001and < .001) were no more likely tohave had greater GI-related and total medical servicevisit expenditures, respectively, during follow-up comparedwith PPI users. Removing Medicaid and Medicaredual enrollees' data from the analyses provided similarresults.

COMMENT

Prior authorization for PPIs in this Medicaid populationhad the effect of reducing the use of high-cost PPIs,while encouraging increased use of lower costing H2As,leading to a decrease of $23.4 millionin annual net expenditures forantisecretory drugs. Requiring administrativeapproval for PPI reimbursementdid not increaseambulatory or inpatient medicalservices use for the management ofGI-related conditions and total medicalcare expenditures among thoseenrollees who did not receive a prescribedPPI. If enrollees who did notreceive a PPI had more severe GI-relateddisease, they might havebeen more vulnerable to adverseeffects of the policy change. Ourdata indicate that this PA programachieved its desired effects ofdecreasing the Medicaid program'spharmacy expenditures for antise-cretory drugs by channeling PPIs tothose patients with demonstratedmedical need. Sensitivity analysisexcluding Medicaid and Medicaredual enrollees provided additionalevidence to support these findings.

The retrospective database analysisdesign of this study precluded theassessment of GI disease-specificclinical (eg, esophageal healing andsymptom relief) and humanistic (eg,quality of life and satisfaction) outcomes.34 In theory, PA programs thatdirect prescribers to adhere to evidence-based clinical practice shouldresult in positive clinical and humanisticoutcomes.10 Our data suggestthat, if any negative outcomesoccurred and affected overall health,they were not sufficient enough toincrease the use of medical servicesby the patients affected by this policy change. In addition,acceptability for this PA program among stakeholdersmay have been enhanced by the state'sextending the PA requirement to its beneficiaries outsideof the Medicaid program.6

We were unable to measure the effect of the policyon prescribers' time spent when managing the PAadministrative process instead of on other clinical oradministrative tasks. However, participating prescribersin the Medicaid program were made aware ofthe PA policy through 2 channels, and they are compensatedfor in toto episodic care of the Medicaid program'spatients. Although we could not compare theunderlying trend in unmanaged Medicaid expendituresfor PPIs with the examined Medicaid program's, it isreasonable to expect that PPI expenditures would havecontinued to increase or, at a minimum, remain constantduring the study period in the absence of the PAprogram. Although no antisecretory prescriptionclaims were identified for the nonusers subgroup, theseenrollees could have obtained antisecretory drug samplesfrom their physician or purchased out-of-pocketprescription PPIs or H2As or over-the-counter H2As orother antacids.35 We were unable to examine the possibleeffect of the use of these drugs. However, the availabilityof these therapies provides patients withtherapeutic options with which they can manage theirhealth conditions outside of Medicaid. In addition, suchuse would not have affected the Medicaid program'spharmacy expenditures. Restricting the study toMedicaid enrollees may have limited its generalizabilityto other patient populations. However, given that anestimated 30% to 70% of patients with painful refluxsymptoms do not have GI conditions with sustained tissuedamage,15 positive outcomes should be realizedacross patient populations when limiting PPI prescriptionsto those with a demonstrable medical need.

To our knowledge, this is the first study of the effectsof PA to include in the analysis the administrative costsof the PA program.36 Despite the inclusion of thesecosts, our findings are still in agreement with otherstudies3,11-13 of pharmacy reimbursement restrictions,which found substantial reductions in drug expenditureswithout concomitant increases in other medicalexpenditures.

As Medicaid expenditures continue to rise, states arefaced with the options of restricting Medicaid eligibilityor benefits or eliminating other public programs toattain budget neutrality. The challenge for Medicaidprograms has been to craft healthcare policies that provideaccess to essential therapies while concomitantlyapplying monetarily responsible control over theirhealthcare budgets. Since completion of data collectionfor this study, the introduction of over-the-counterPrilosec37 and the stabilization of supply and price forgeneric prescription of other omeprazole products38have presented new opportunities for managing the PPIbenefit. Limiting PPI formularies to generic-only productsor offering coverage for over-the-counter Prilosecare 2 benefit management strategies now available. Inaddition, the inclusion of over-the-counter and genericPPI products and H2As in PPI PA programs can furtherdirect patients and prescribers to cost-effective, appropriatetherapy options.

Managed care organizations and other healthcareplan sponsors have been challenged to control risingdrug costs while not denying or limiting access to thosemedications that improve clinical and humanistic outcomes.From a policy perspective, PA offers a strategy toreduce expenditures for PPIs without incurring adversemedical consequences. Plan sponsors that have implementedPPI PA programs with appropriate diagnosis- andrisk-based clinical criteria should be encouragedthat they have moved in the right direction to cost-effectivelymanage rising drug costs.

Acknowledgments

We thank Kathleen Fairman, MA, for her analytic assistance andEdmond Weisbart, MD, for his clinical expertise.

From Outcomes Research, Express Scripts, Inc, Maryland Heights, Mo.

This study was funded by Express Scripts, Inc.

Address correspondence to: Thomas Delate, PhD, Clinical Rx Research, KaiserPermanente Centretech Parkway, Aurora, CO 80011. E-mail: tdelate@express-scripts.com.

1. Kaiser Commission on Medicaid and the Uninsured. Medicaid and prescriptiondrugs. October 2002. Available at: http://www.kff.org/medicaid/1609-index.cfm. Accessed June 23, 2003.

2. Kaiser Commission on Medicaid and the Uninsured. Medicaid: fiscal challengesto coverage. May 2003. Available at: http://www.kff.org/medicaid/upload/14341_1.pdf. Accessed June 23, 2003.

Health Serv Res.

3. Cromwell DM, Bass EB, Steinberg EP, et al. Can restrictions on reimbursementfor anti-ulcer drugs decrease Medicaid pharmacy costs without increasing hospitalizations?1999;33:1593-1608.

4. Kaiser Commission on Medicaid and the Uninsured. States strive to limitMedicaid expenditures for prescribed drugs. February 2002. Available at:http://www.kff.org/medicaid/4030-index.cfm. Accessed June 23, 2003.

NHPF Issue Brief.

5. Gencarelli DM. Medicaid prescription drug coverage: state efforts to controlcosts. 2003;790:1-17.

6. Kaiser Commission on Medicaid and the Uninsured. Prior authorization forMedicaid prescription drugs in five states: lessons for policy makers. April 2003.Available at: http://www.kff.org/medicaid/4094-index.cfm. Accessed June 23, 2003.

Health

Care Financ Rev.

7. Gondek K. Prescription drug payment policy: past, present, and future. 1994;15:1-7.

Health

Aff (Millwood).

8. Soumerai SB, Ross-Degnan D. Experience of state drug benefit programs. 1990;9(3):36-54.

9. Social Security Act, §1927(d)(5).

J Manag Care Pharm.

10. MacKinnon NJ, Kumar R. Prior authorization programs: a critical review of theliterature. 2001;7:297-302.

N Engl J Med.

11. Smalley WE, Griffin MR, Fought RL, Sullivan L, Ray WA. Effect of a prior-authorizationrequirement on the use of nonsteroidal anti-inflammatory drugs byMedicaid patients. 1995;332:1612-1617.

J Res Pharm Econ.

12. Kotzan JA, McMillan JA, Jankel CA, Foster AL. Initial impact of a Medicaid priorauthorization program for NSAID prescriptions. 1993;5:25-41.

J Res Pharm

Econ.

13. Kotzan JA, Jankel CA, McMillan JA, Foster AL, Myers S. Initial impact of aMedicaid maintenance dose program for H2 antagonist prescriptions. 1993;5:43-58.

Best Pract Res Clin Gastroenterol.

14. Sachs G, Shin JM, Vagin O, et al. Current trends in the treatment of upper gastrointestinaldisease. 2002;16:835-849.

JAMA.

15. Shaheen N, Ransohoff DF. Gastroesophageal reflux, Barrett esophagus, andesophageal cancer: scientific review. 2002;287:1972-1981.

Am J Med.

16. Harris RA, Kupperman M, Richter JE. Prevention of recurrences of erosiveesophagitis: a cost-effectiveness analysis of maintenance proton pump inhibition.1997;102:78-88.

N Engl J Med.

17. Vigneri S, Termini R, Leandro G, et al. A comparison of five maintenancetherapies for reflux esophagitis. 1995;333:1106-1110.

JAMA.

JAMA.

18. Soll AH, Practice Parameters Committee of the American College ofGastroenterology. Consensus conference: medical treatment of peptic ulcer disease:practice guidelines [published correction appears in 1996;275:1314].1996;275:622-629.

Am J Gastroenterol.

19. DeVault KR, Castell DO, Practice Parameters Committee of the AmericanCollege of Gastroenterology. Updated guidelines for the diagnosis and treatmentof gastroesophageal reflux disease. 1999;94:1434-1442.

Gastroenterology.

20. Marks RD, Richter JE, Rizzo J, et al. Omeprazole versus H2-receptor antagonistsin treating patients with peptic stricture and esophagitis. 1994;106:907-915.

Digestion.

21. Dent J. Long-term aims of treatment of reflux disease, and the role of non-drugmeasures. 1992;51(suppl 1):30-34.

Am J Gastroenterol.

22. Sampliner RE, Practice Parameters Committee of the American College ofGastroenterology. Practice guidelines on the diagnosis, surveillance, and therapyof Barrett's esophagus. 1998;93:1028-1032.

Helicobacter pylori

Aliment Pharmacol Ther.

23. Yousfi MM, el-Zimaity HMT, al-Assi MT, Cole RA, Genta RM, Graham DY.Metronidazole, omeprazole and clarithromycin: an effective combination therapyfor infection. 1995;9:209-212.

N Engl J Med.

24. Soumerai SB, Ross-Degnan D, Avorn J, McLaughlin TJ, Choodnovoskiy I.Effects of Medicaid drug-payment limits on admissions to hospitals and nursinghomes. 1991;325:1072-1077.

N Engl J

Med.

25. Soumerai SB, Avorn J, Ross-Degnan D, Gortmaker S. Payment restrictions forprescription drugs under Medicaid: effects on therapy, cost, and equity. 1987;317:550-556.

N Engl J Med.

26. Soumerai SB, McLaughlin TJ, Ross-Degnan D, Casteris CS, Bollini P. Effects ofa limit on Medicaid drug-reimbursement benefits on the use of psychotropic agentsand acute mental health services by patients with schizophrenia. 1994;331:650-655.

JAMA.

27. Tamblyn R, Laprise R, Hanley JA, et al. Adverse events associated with prescriptiondrug cost-sharing among poor and elderly persons. 2001;285:421-429.

Generic Product Identifier Number.

28. First DataBank (Medispan). Indianapolis,Ind: First DataBank (Medispan); 2002.

Am J

Epidemiol.

29. Ray WA, Griffin MR. Use of Medicaid data for pharmacoepidemiology. 1989;129:837-849.

J Clin Epidemiol.

30. Ward MM, Leigh JP. Pooled time series regression analysis in longitudinalstudies. 1993;46:645-659.

Finite Mixture Models.

31. McLachlan GJ, Peel D. New York, NY: John Wiley &Sons Inc; 2000.

Med Care.

32. Clark DO, Von Korff M, Saunders K, Baluch WM, Simon GE. A chronic diseasescore with empirically derived weights. 1995;33:783-795.

JBES.

33. Duan N, Manning WG, Morris CN, Newhouse JP. A comparison of alternativemodels for the demand for medical care. 1983;1:115-126.

Milbank Q.

34. Soumerai SB, Ross-Degnan D, Fortess EE, Abelson J. A critical analysis of studiesof state drug reimbursement policies: research in need of discipline. 1993;71:217-252.

Med Care.

35. Martin BC, McMillan JA. The impact of implementing a more restrictive prescriptionlimit on Medicaid recipients. 1996;34:686-701.

Am J

Manag Care.

36. Lexchin J. Effects of restrictive formularies in ambulatory care settings. 2001;7:69-76.

37. Food and Drug Administration Web site. FDA approves Prilosec OTC to treatfrequent heartburn. June 20, 2003. Available at: http://www.fda.gov/bbs/topics/news/2003/NEW00916.html. Accessed October 20, 2003.

38. Food and Drug Administration Web site. FDA approves generic omeprazole:first marketing of cost-saving drug product to begin November 1, 2002. Availableat: http://www.fda.gov/bbs/topics/NEWS/2002/NEW00848.html. Accessed October20, 2003.