Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
VALUEIN
HEALTHVO
LUME
14N
UMBER
3M
AY2011
PAGES
A1-A232
ELSEVIER
ISPOR 16th Annual International Meeting Research AbstractsMay 21–25, 2011
Baltimore, MD, USAResearch Podium AbstractsResearch Poster Abstracts
IN HEALTH
Volume 14 Number 3 May 2011 ISSN 1098-3015
www.ispor.org
EDITORIAL BOARDCo-Editor-in-ChiefMichael Drummond, PhDUniversity of YorkHeslington, York, [email protected]. Daniel Mullins, PhDUniversity of MarylandBaltimore, MD, [email protected]
CO-EDITORSChris Bingefors, PhD, MScUppsala UniversityUppsala, [email protected] Brennan, PhDUniversity of SheffieldSheffield, [email protected] Doshi, PhDUniversity of PennsylvaniaPhiladelphia, PA, [email protected] Fehnel, PhD, MARTI Health SolutionsResearch Triangle Park, NC, [email protected] P. Geisler, MD, MPHWing Tech, Inc.Blue Hill, ME, [email protected] Greenberg, PhDBen-Gurion University of the NegevBeer-Sheva, [email protected] Hornberger, MD, MSCedar Associates, LLCMenlo Park, CA, [email protected] Kauf, PhDUniversity of FloridaGainesville, FL, [email protected] G. Liu, PhDPeking University, Guanghua School ofManagementPR [email protected] Lloyd, DPhil, BScOxford Outcomes, Ltd.Oxford, [email protected] Manca, PhD, MScUniversity of YorkYork, [email protected] Naughton, PhDWake Forest University School of MedicineWinston-Salem, NC, [email protected] Scuffham, PhD, BAGriffith University - School of MedicineQueensland, [email protected] L. (Hans) Severens, PhDErasmus UniversityRotterdam, The [email protected] (Tina) Shih, PhD, MSUniversity of ChicagoChicago, IL, [email protected] S. Skjoldborg, PhD, MAEli Lilly Denmark A/SCopenhagen, [email protected]
EDITORIAL ADVISORY BOARDMarc L. Berger, MDIngenix Life SciencesNew York, NY, [email protected] Briggs, DPhilUniversity of GlasgowGlasgow, [email protected] Pete Fullerton, PhDStrategic Pharmacy InnovationsSeattle, WA, [email protected] Paul Gagnon, PhDPittstown, NJ, [email protected] Glick, PhDUniversity of PennsylvaniaPhiladelphia, PA, [email protected] Husereau, MSc, BScUniversity of OttawaOttawa, ON, [email protected] KindUniversity of YorkHeslington, York, [email protected] C. Langley, PhDUniversity of MinnesotaWoodbury, MN, [email protected] Lapuerta, MDBristol-Myers SquibbPrinceton, NJ, [email protected] Levy, PhDDalhousie UniversityHalifax, NS, [email protected] E. Marx, PharmD, MSAbbott LaboratoriesAbbott Park, IL, [email protected] A. Matuszewski, MS, PharmDFirst DataBank, Inc.South San Francisco, CA, [email protected] F. McGhan, PharmD, PhDUniversity of the Sciences in PhiladelphiaPhiladelphia, PA, [email protected] J. Milne, PhDUniversity of AucklandAuckland, New [email protected] A. Morris, PhDLouis A. Morris & AssociatesDix Hills, NY, [email protected] J. Neumann, ScDTufts—New England Medical CenterBoston, MA, [email protected]
Mark Nuijten, PhD, MD, MBAArs Accessus MedicaDorpsstraat, The [email protected] L. Pashos, PhDUnited BioSource CorporationLexington, MA, [email protected] Revicki, PhDUBCBethesda, MD, [email protected] A. Schulman, MDDuke Clinical Research InstituteDurham, NC, [email protected] Seibert, MD, MPH, MSc, ScDUniversity of Health Sciences,Medical Informatics & TechnologyHall i.T., [email protected] Sullivan, PhDUniversity of WashingtonSeattle, WA, [email protected] C. Weinstein, PhDHarvard School of Public HealthBoston, MA, [email protected]
MANAGEMENT ADVISORY BOARDShelby Reed, PhD, RPh (Chair)Duke UniversityDurham, NC, [email protected] KindUniversity of YorkHeslington, York, [email protected] Yang, PhDSeoul UniversitySeoul, South [email protected]
EDITORIAL OFFICEManaging EditorStephen L. PrioriISPORLawrenceville, NJ, [email protected] AssistantDanielle MrozISPORLawrenceville, NJ, [email protected]
Podium Session I
A1 Cancer Outcomes Research: CN1–CN4
A1 Comparative Effectiveness Research: CO1–CO4
A2 Effects of Drug Management Programs on Patients: DM1–DM4
A3 Employee Health & Productivity Outcomes Research: OR1–OR4
A4 Case Studies in Addressing Selection Bias: SB1–SB4
Podium Session II
A5 Research on Methods: Cost-Effectiveness Analysis: CE1–CE4
A6 Research on Methods: Database Analysis: DS1–DS4
A6 Drug Use and Patient Safety: DU1–DU4
A7 Infection Outcomes Research: IN1–IN4
A8 Impact of Medication Compliance: MC1–MC4
Podium Session III
A9 Research on Methods: Economic Evaluations: EE1–EE4
A10 Evolving Concepts in Outcomes Research: EV1–EV4
A10 Analysis of Medicare Policy and Resource Use: MD1–MD4
A11 Examining the QALY: QA1–QA4
Poster Session IHealth Care Use & Policy Studies
A12 Health Care Use & Policy Studies – Consumer Role in Health Care: PHP1–PHP5
A13 Health Care Use & Policy Studies – Diagnosis Related Group: PHP6
A13 Health Care Use & Policy Studies – Drug/Device/Diagnostic Use & Policy: PHP7–11, PHP13–PHP20, PHP22–PHP28
A17 Health Care Use & Policy Studies – Equity and Access: PHP29–PHP31
A17 Health Care Use & Policy Studies – Formulary Development: PHP32–PHP33, PHP35
A18 Health Care Use & Policy Studies – Health Care Costs & Management: PHP36–PHP38, PHP40–PHP57,PHP59–PHP68
A23 Health Care Use & Policy Studies – Health Care Research & Education: PHP69–PHP82
A26 Health Care Use & Policy Studies – Health Technology Assessment Programs: PHP83–PHP85
A27 Health Care Use & Policy Studies – Population Health: PHP86
A27 Health Care Use & Policy Studies – Prescribing Behavior & Treatment Guidelines: PHP87–PHP90
A28 Health Care Use & Policy Studies – Quality of Care: PHP91–PHP93
A28 Health Care Use & Policy Studies – Regulation of Health Care Sector: PHP94–PHP100
A30 Health Care Use & Policy Studies – Risk Sharing/Performance-Based Agreements: PHP101–PHP105
A30 Health Care Use & Policy Studies – Conceputal Papers: PHP106–PHP114
VOLUME 14 NUMBER 3 MAY 2011
ISPOR 16th ANNUAL INTERNATIONAL MEETING RESEARCH ABSTRACTS
Disease-Specific Studies
A32 Cardiovascular Disorders – Clinical Outcomes Studies: PCV1–PCV8, PCV10–PCV31
A38 Cardiovascular Disorders – Cost Studies: PCV32–PCV36, PCV38–PCV58
A43 Cardiovascular Disorders – Patient-Reported Outcomes & Preference-Based Studies: PCV59–PCV78
A46 Cardiovascular Disorders – Health Care Use & Policy Studies: PCV79–PCV108
A52 Cardiovascular Disorders – Research on Methods: PCV109–PCV112
A53 Sensory Systems Disorders – Clinical Outcomes Studies: PSS1–PSS6
A54 Sensory Systems Disorders – Cost Studies: PSS7–PSS14
A56 Sensory Systems Disorders – Patient-Reported Outcomes & Preference-Based Studies: PSS15–PSS22
A57 Sensory Systems Disorders – Health Care Use & Policy Studies: PSS23–PSS26
A58 Sensory Systems Disorders – Research on Methods: PSS27–PSS30
A59 Systemic Disorders/Conditions – Clinical Outcomes Studies: PSY1–PSY7
A60 Systemic Disorders/Conditions – Cost Studies: PSY8–PSY29
A64 Systemic Disorders/Conditions – Patient-Reported Outcomes & Preference-Based Studies: PSY30–PSY47
A68 Systemic Disorders/Conditions – Health Care Use & Policy Studies: PSY48–PSY52, PSY54–PSY74
A72 Systemic Disorders/Conditions – Research on Methods: PSY75–PSY81
A74 Urinary/Kidney Disorders – Clinical Outcomes Studies: PUK1–PUK6
A75 Urinary/Kidney Disorders – Cost Studies: PUK7–PUK17
A77 Urinary/Kidney Disorders – Patient-Reported Outcomes & Preference-Based Studies: PUK18
A77 Urinary/Kidney Disorders – Health Care Use & Policy Studies: PUK19–PUK21
A78 Urinary/Kidney Disorders – Research on Methods: PUK22–PUK24
Poster Session IISelected Health Care Treatment Studies
A79 Medical Device/Diagnostics – Clinical Outcomes Studies: PMD1–PMD10
A80 Medical Device/Diagnostics – Cost Studies: PMD11–PMD27, PMD29
A84 Medical Device/Diagnostics – Patient-Reported Outcomes & Preference-Based Studies: PMD30–PMD31, PMD34
A85 Medical Device/Diagnostics – Health Care Use & Policy Studies: PMD35–PMD39
A85 Medical Device/Diagnostics – Research on Methods: PMD40–PMD41
A86 Surgery – Clinical Outcomes Studies: PSU1–PSU5
A87 Surgery – Cost Studies: PSU6–PSU11
A88 Surgery – Patient-Reported Outcomes & Preference-Based Studies: PSU12–PSU15
A89 Surgery – Heath Care Use & Policy Studies: PSU16–PSU19
A89 Surgery – Research on Methods: PSU20–PSU25
Disease-Specific Studies
A91 Diabetes/Endocrine Disorders – Clinical Outcomes Studies: PDB1–PDB18
A94 Diabetes/Endocrine Disorders – Cost Studies: PDB19–PDB36
A97 Diabetes/Endocrine Disorders – Patient-Reported Outcomes & Preference-Based Studies: PDB37–PDB49
A100 Diabetes/Endocrine Disorders – Health Care Use & Policy Studies: PDB50–PDB71
A104 Diabetes/Endocrine Disorders – Research on Methods: PDB72–PDB76
V A L U E I N H E A L T H 1 4 ( 2 0 1 1 )
A105 Individual’s Health – Clinical Outcomes Studies: PIH1–PIH7
A106 Individual’s Health – Cost Studies: PIH8–PIH21
A109 Individual’s Health – Patient-Reported Outcomes & Preference-Based Studies: PIH22–PIH32
A111 Individual’s Health – Health Care Use & Policy Studies: PIH34–PIH40
A112 Individual’s Health – Research on Methods: PIH41–PIH46
A113 Infection – Clinical Outcomes Studies: PIN1–PIN8
A115 Infection – Cost Studies: PIN9–PIN17, PIN19–PIN35
A120 Infection – Patient-Reported Outcomes & Preference-Based Studies: PIN36–PIN41
A121 Infection – Health Care Use & Policy Studies: PIN43–PIN45, PIN47–PIN52
A123 Infection – Research on Methods: PIN53
A123 Muscular-Skeletal Disorders – Clinical Outcomes Studies: PMS1–PMS7
A124 Muscular-Skeletal Disorders – Cost Studies: PMS8–PMS9, PMS11–PMS36
A129 Muscular-Skeletal Disorders – Patient-Reported Outcomes & Preference-Based Studies: PMS37–PMS47
A132 Muscular-Skeletal Disorders – Health Care Use & Policy Studies: PMS48–PMS54, PMS56–PMS58,PMS60–PMS69
A136 Muscular-Skeletal Disorders – Research on Methods: PMS70–PMS74
A136 Respiratory-Related Disorders – Clinical Outcomes Studies: PRS1–PRS8
A138 Respiratory-Related Disorders – Cost Studies: PRS9–PRS21
A140 Respiratory-Related Disorders – Patient-Reported Outcomes & Preference-Based Studies: PRS22–PRS31
A142 Respiratory-Related Disorders – Health Care Use & Policy Studies: PRS32–PRS42
A144 Respiratory-Related Disorders – Research on Methods: PRS43–PRS46
Poster Session IIIResearch on Methods Studies
A145 Research on Methods – Clinical Outcomes Methods: PRM1–PRM3
A146 Research on Methods – Cost Methods: PRM4–PRM9
A147 Research on Methods – Databases & Management Methods: PRM10–PRM11
A147 Research on Methods – Modeling Methods: PRM12–PRM14
A148 Research on Methods – Patient-Reported Outcomes Studies: PRM15–PRM30, PRM32–PRM35
A152 Research on Methods – Statistical Methods: PRM36–PRM38
A152 Research on Methods – Study Design: PRM39
A152 Research on Methods – Conceptual Papers: PRM40–PRM49
Disease-Specific Studies
A154 Cancer – Clinical Outcomes Studies: PCN1–PCN24
A159 Cancer – Cost Studies: PCN25–PCN43, PCN46–PCN50, PCN52–PCN78
A168 Cancer – Patient-Reported Outcomes & Preference-Based Studies: PCN79–PCN96
A172 Cancer – Health Care Use & Policy Studies: PCN97, PCN99–PCN125
A177 Cancer – Research on Methods: PCN126–PCN138
A180 Gastrointestinal Disorders – Clinical Outcomes Studies: PGI1–PGI5
A181 Gastrointestinal Disorders – Cost Studies: PGI6–PGI16
A183 Gastrointestinal Disorders – Patient-Reported Outcomes & Preference-Based Studies: PGI17–PGI24
A185 Gastrointestinal Disorders – Health Care Use & Policy Studies: PGI25–PGI26
V A L U E I N H E A L T H 1 4 ( 2 0 1 1 )
A185 Mental Health – Clinical Outcomes Studies: PMH1–PMH18
A188 Mental Health – Cost Studies: PMH19–PMH44
A193 Mental Health – Patient-Reported Outcomes & Preference-Based Studies: PMH45–PMH57
A196 Mental Health – Health Care Use & Policy Studies: PMH58–PMH80
A200 Mental Health – Research on Methods: PMH81–PMH85
A201 Neurological Disorders – Clinical Outcomes Studies: PND1–PND9
A203 Neurological Disorders – Cost Studies: PND10–PND27
A207 Neurological Disorders – Patient-Reported Outcomes & Preference-Based Studies: PND28–PND50
A211 Neurological Disorders – Health Care Use & Policy Studies: PND51–PND59
A213 Neurological Disorders – Research on Methods: PND60–PND65
A215 DISCLOSURE INFORMATION
A223 AUTHOR INDEX
V A L U E I N H E A L T H 1 4 ( 2 0 1 1 )
Editorial Office. Value in Health, ISPOR, 3100 Princeton Pike, Suite 3E,Lawrenceville, NJ 08648.
ISPOR Office. Marilyn Dix Smith, RPh, PhD, Executive Director, 3100 PrincetonPike, Suite 3E, Lawrenceville, NJ 08648. Tel: (609) 219-0773, Fax: (609) 219-0774, E-mail: [email protected], Web site: http://www.ispor.org/valueinhealth_index.asp.
CUSTOMER SERVICE (orders, claims, online, change of address): ElsevierHealth Sciences Division, Subscription Customer Service, 3251 Riverport Lane,Maryland Heights, MO 63043. Tel: (800) 654-2452 (U.S. and Canada); (314)447-8871 (outside U.S. and Canada). Fax: (314) 447-8029. E-mail:[email protected] (for print support);[email protected] (for online support). Address changesmust be submitted four weeks in advance.
YEARLY SUBSCRIPTION RATES: United States and possessions: Individual$250.00; Institution $535.00. All other countries (prices include airspeeddelivery): Individual $250.00; Institution $535.00. Single Issues $67.00. Toreceive student/resident rate, orders must be accompanied by name ofaffiliated institution, date of term and the signature of program/residencycoordinator on institution letterhead. Orders will be billed at the individual rateuntil proof of status is received. Current prices are in effect for back volumesand back issues.
Further information on this journal is available from the Publisher or from thisjournal’s website (http://www.elsevier.com/locate/jval). Information on otherElsevier products is available through Elsevier’s website (http://www.elsevier.com).
Author inquiriesFor inquiries relating to the submission of articles (including electronicsubmission where available), visit http:/www.elsevier.com/authors. The site alsoprovides the facility to track accepted articles and set up e-mail alerts to informyou of when an article’s status has changed, as well as detailed artworkguidelines, copyright information, frequently asked questions, and more. Pleasesee Information for Authors for individual journal. Contact details for questionsarising after acceptance of an article, especially those relating to proofs, areprovided after registration of an article for publication.
English language help service: Upon request, Elsevier will direct authors toan agent who can check and improve the English of their paper (beforesubmission). Please contact [email protected] for further information.
Reprints. For queries about author offprints, e-mail [email protected]. To order 100 or more reprints for educational, commercial, orpromotional use, contact the Commercial Reprints Department, Elsevier Inc.,360 Park Avenue South, New York, NY 10010-1710. Fax: (212) 462-1935; e-mail [email protected]. Reprints of single articles available online may beobtained by purchasing Pay-Per-View access for $36 per article on the journalwebsite http://www.elsevier.com/locate/jval.
Copyright © 2011, International Society for Pharmacoeconomics and OutcomesResearch (ISPOR). Published by Elsevier Inc.
This journal and the individual contributions contained in it are protected undercopyright by International Society for Pharmacoeconomics and OutcomesResearch, and the following terms and conditions apply to their use:
PhotocopyingSingle photocopies of single articles may be made for personal use as allowedby national copyright laws. Permission of the Publisher and payment of a fee isrequired for all other photocopying, including multiple or systematic copying,copying for advertising or promotional purposes, resale, and all forms ofdocument delivery. Special rates are available for educational institutions thatwish to make photocopies for non-profit educational classroom use.
Permissions may be sought directly from Elsevier’s Rights Department inOxford, UK: phone (215) 238-7869 or �44 (0) 1865 843830, fax �44 (0)1865 853333, e-mail [email protected]. Requests may also becompleted online via the Elsevier homepage (http://www.elsevier.com/locate/permissions).
In the USA, users may clear permissions and make payments through theCopyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923,USA; phone: (978) 750-8400, fax: (978) 750-4744, and in the UK through theCopyright Licensing Agency Rapid Clearance Service (CLARCS), 90 TottenhamCourt Road, London W1P 0LP, UK; phone: (�44) 20 7631 5555; fax: (�44) 207631 5500. Other countries may have a local reprographic rights agency forpayments.Derivative WorksSubscribers may reproduce tables of contents or prepare lists of articlesincluding abstracts for internal circulation within their institutions. Permission ofthe Publisher is required for resale or distribution outside the institution.Permission of the Publisher is required for all other derivative works, includingcompilations and translations.Electronic Storage or UsagePermission of the Publisher is required to store or use electronically anymaterial contained in this journal, including any article or part of an article.Except as outlined above, no part of this publication may be reproduced, storedin a retrieval system or transmitted in any form or by any means, electronic,mechanical, photocopying, recording or otherwise, without prior writtenpermission of the Publisher.Address permissions requests to: Elsevier Rights Department, at the fax ande-mail addresses noted above.Abstracting and IndexingValue in Health is indexed in Index Medicus/MEDLINE, Current Contents/Social &Behavioral Sciences, SciSearch/SCI Expanded, Social Sciences Citation Index,International Pharmaceutical Abstracts, Embase/Excerpta Medica, andPsychInfo/Psychological Abstracts, and Journal Citation Reports/Science Edition(Thomson ISI).NoticeNo responsibility is assumed by the Publisher or the International Society forPharmacoeconomics and Outcomes Research for any injury and/or damage topersons or property as a matter of products liability, negligence or otherwise,or from any use or operation of any methods, products, instructions or ideascontained in the material herein. Because of rapid advances in the medicalsciences, in particular, independent verification of diagnoses and drug dosagesshould be made.Although all advertising material is expected to conform to ethical (medical)standards, inclusion in this publication does not constitute a guarantee orendorsement of the quality or value of such product or of the claims made of itby its manufacturer.
PC
lgyy6iptTpwTPu
CacetMcce1fadwPehtS11eSst
CTL
V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
avai lable at www.sc iencedirect .com
journal homepage: www.elsevier .com/ locate / jva l
ABSTRACTS
UN
aetictjfapecrtEmrhdnFpct
CPAV
ocp
Apv1pGscli
PC
ISPOR 16th ANNUAL INTERNATIONAL MEETING RESEARCH ABSTRACTS
ODIUM SESSION I:ANCER OUTCOMES RESEARCH
CN1MONOTHERAPY OF ANDROGEN DEPRIVATION THERAPY VERSUS RADICALPROSTATECTOMY AMONG VETERANS WITH LOCALIZED PROSTATE CANCER: ACOMPARATIVE EFFECTIVENESS ANALYSIS OF RETROSPECTIVE COHORTSLiu J, Shi L, Sartor OTulane University, New Orleans, LA, USA
OBJECTIVES: There is no consensus regarding the optimal treatment for localizedprostate cancer. This study aimed to examine the comparative effectiveness ofmonotherapy of either primary androgen deprivation therapy (PADT) or radicalprostatectomy (RP) in terms of overall survival rate. METHODS: Male patients withocalized prostate cancer were identified in the Veterans Affairs Veterans Inte-rated Service Network 16 data warehouse (January, 2003-June, 2006), with oneear baseline and at least 3-year follow-up (till 06/2009). Eligible patients (18-75ears old) had no other cancer history and used PADT or monotherapy of RP withinmonths after the first diagnosis of prostate cancer. The overall survival from
nitiation of index treatment was analyzed using Kaplan-Meier method and Coxroportional hazard regression, adjusted for age, race, marital status, insuranceype, cancer stage, Charlson comorbidity index, alcohol and tobacco use. RESULTS:he age was 66.2(6.07) [Mean(SD)] years in 211 PADT patients, 59.9(6.15) in 215 RPatients. During the follow-up of 4.2(0.95) years, the cumulative incidence of deathas 29 (13.74%) among PADT patients and 6 (2.79%) among RP patients (p�0.001).he overall 3-year survival rate was 89.57% in PADT and 98.60% in RP (p�0.001).atients who received PADT had almost 4 times as high mortality risk as thosesing RP (HR �3.820, 95% CI�1.483 to 9.845, p�0.006). CONCLUSIONS: Overall sur-
vival rate following RP among localized prostate cancer patients was significantlyhigher than that after PADT, controlling for other covariates. More research amonga larger population with longer follow-up are warranted to confirm this finding.
CN2ESTIMATED EFFECTS OF THE NATIONAL BREAST AND CERVICAL CANCEREARLY DETECTION PROGRAM ON CERVICAL CANCER MORTALITYEkwueme DU1, Uzunangelov V2, Hoerger T2, Saraiya M1, Miller J1, Hall I1, Benard V1,Royalty J1, Li C11Centers for Disease Control and Prevention, Atlanta, GA, USA, 2RTI International, ResearchTriangle Park, NC, USA
OBJECTIVES: The National Breast and Cervical Cancer Early Detection Program (NB-CEDP) is the largest organized cancer screening program for low-income, un-insurednd under-insured women in the United States. The program’s effectiveness in in-reasing the life expectancy of participating women has never been measured. Westimated the benefits of NBCCEDP-funded cervical cancer screening (Program) inerms of life-years (LYs) saved compared to No Program and No Screening scenarios.
ETHODS: Based on an existing model developed by Myers et al., we constructed aervical cancer simulation model by modifying the age and screening schedule of theohort to reflect screening frequency for NBCCEDP participants from 1991-2007. Westimated screening habits in the absence of the program based on data from the990-2005 National Health Interview Survey. We performed Markov cohort analysisor each age in the 18-64 range and calculated an overall weighted average using thege distribution at first NBCCEDP Pap test for screening. Weighted averages were pro-uced for three scenarios – women receiving testing from the NBCCEDP (the Program),omen receiving testing from alternative sources in the absence of the program (No
rogram), and women receiving no testing at all (No Screening). We compared LYstimates for 69,100 women detected with human papillomavirus infection, low-and-igh-grade squamous intraepithelial lesions or cervical cancer under the program to
he counterfactual of having their disease undetected under No Program and Nocreening scenarios. RESULTS: From 1991-2007, we estimate that the Program added0,369 LYs to the total lifespan of tested women when compared to No Program, and01,509 LYs when compared to No Screening. Furthermore, the Program prevented anstimated 325 cervical cancer deaths relative to No Program, and 3,825 relative to Nocreening. CONCLUSIONS: These estimates suggest that NBCCEDP cervical cancercreening may have reduced mortality among medically underserved women inhe United States.
N3HE VALUE OF RESEARCH FOR ERCC1 TESTING IN STAGE I NON-SMALL CELL
UNG CANCER
Roth J1, Carlson JJ2, Steuten L3, Veenstra D1
Copyright © 2011, International Society for Pharmacoeconomics and Outcomes Research (
1University of Washington, Pharmaceutical Outcomes Research and Policy Program, Seattle, WA,SA, 2University of Washington, Seattle, WA, USA, 3University of Twente, Enschede, Theetherlands
OBJECTIVES: To assess the value of additional research for ERCC1 expression test-ing to guide adjuvant chemotherapy decisions in fully resected Stage I non-smallcell lung cancer (NSCLC). METHODS: We refined a previously-developed decision-
nalytic model to estimate the expected value of perfect information (EVPI) andxpected value of sample information (EVSI) for two treatment strategies: 1) ERCC1esting to inform adjuvant chemotherapy decisions, with ERCC1� patients receiv-ng no chemotherapy and ERCC1- patients receiving chemotherapy; 2) standardare, with all patients receiving no chemotherapy. Model parameters and uncer-ainty ranges were derived from a retrospective analysis of the International Ad-uvant Lung Cancer Trial, published literature, and government sources. The af-ected population was derived from SEER incidence estimates, and examined overdiscounted 10-year time horizon. RESULTS: At a willingness-to-pay of $150,000er quality-adjusted life year, ERCC1 and standard care strategies resulted in av-rage net-benefits of $630,500 and $625,200, respectively. The ERCC1 and standardare strategies produced greater net-benefit in 64% and 36% of 10,000 simulations,espectively. The average net-benefit difference was $14,000 in simulations wherehe standard care strategy was optimal. With an affected population of 233,825;VPI was $1.2 billion. Preliminary estimates suggest an EVSI of approximately $20illion at plausible sample sizes. CONCLUSIONS: Considerable value could be
ealized through additional research to reduce uncertainty about the comparativeealth outcomes of ERCC1 and standard care strategies. The EVPI of $1.2 billion wasriven by the large 10-year affected population, probability that ERCC1 testing isot the optimal strategy, and consequences of selecting the non-optimal strategy.orthcoming results will enable estimation of the expected net-benefit of sam-ling, which compares the EVSI of various study designs and sample sizes to theost of conducting such studies. These findings can assist stakeholders in priori-izing funding for ERCC1 research relative to alternative research investments.
N4ALONOSETRON VERSUS OTHER 5-HYDROXYTRYPTAMINE3 RECEPTORNTAGONISTS FOR PREVENTION OF CHEMOTHERAPY INDUCED NAUSEA ANDOMITING AMONG MEDICARE PATIENTS WITH CANCER
Craver C1, Gayle J1, Balu S2, Buchner D21Premier, Inc., Charlotte, NC, USA, 2Eisai, Inc., Woodcliff Lake, NJ, USAOBJECTIVES: To assess the rate of uncontrolled chemotherapy induced nausea andvomiting (CINV) associated with palonosetron initiation versus other 5-hydroxytryptamine3-receptor antagonists (5-HT3-RAs) among Medicare patients with cancer
n chemotherapy (CT) treatment in a hospital outpatient setting. METHODS: Medi-are patients with a cancer diagnosis initiating CT and anti-emetic prophylaxis withalonosetron (Group 1) and other 5-HT3-RAs (Group 2) for the first time (index date)
between April 1, 2007 – March 31, 2009 were identified from the Premier Perspectivedatabase. Inclusion criteria were no evidence of nausea and vomiting, CT, and anti-emetic medication in the 6-month pre-index date period and 36-consecutive monthsof data submission. A negative binomial distribution generalized linear multivariateregression model estimating the rate of CINV events on CT emetogenicity and cyclematched groups in the follow-up period (first of eight CT cycles or six months post-index date) was developed after adjusting for several demographic and clinicalvariables. RESULTS: Of 4799 identified patients, 962 initiated palonosetron (Group 1;20.1%). Group 1 patients were significantly younger [70.4 (SD: 9.3) versus 71.6 (9.0)years; p�0.0001], comprised more females [52.9% versus 48.6%; p�0.0001], less African
mericans [8.7% vs. 11.3%] and more Hispanic patients [6.3% versus 2.5%]; all�0.0001, more highly and moderately emetogenic CT [33.6% versus 20.7% and; 47.3%ersus 40.3%, respectively; p�0.0001], and more lung and breast [30.9% vs. 24.9% and2.3% vs. 9.6%, respectively; p�0.0001]. In the follow-up period, the regression modelredicted a 11.8% decrease in the CINV events per CT cycle for Group 1 patients versusroup 2 patients; p�0.05. CONCLUSIONS: In this retrospective hospital outpatienttudy, after matching for CT emetogenicity and cycle and adjusting for other potentialonfounders, Medicare patients with cancer initiated on palonosetron were moreikely to experience a significantly lower rate of CINV events per CT cycle versus thosenitiating other 5-HT3-RAs.
ODIUM SESSION I:OMPARATIVE EFFECTIVENESS RESEARCH
CO1COMPARATIVE EFFECTIVENESS ANALYSIS OF TNF BLOCKERS IN RHEUMATOIDARTHRITIS (RA) PATIENTS IN A REAL-WORLD SETTING
1 2 3 4 4 5 6
Bonafede RP , Pearson D , Babich J , Chastek B , Becker L , Watson C , Chaudhari S ,Harrison DJ7, Gandra SR7ISPOR). Published by Elsevier Inc.
http://dx.doi.org/10.1016/j.jval.2011.02.010
T
E
il
atEtfbibmtrfa
CPL
A2 V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
1Providence Arthritis Center, Portland, OR, USA, 2Ventura County Medical Center, Ventura, CA,USA, 3Carolina Bone & Joint, Charlotte, NC, USA, 4i3 Innovus, Eden Prairie, MN, USA, 5Amgen,Inc., Newbury Park, CA, USA, 6KForce Clinical Research, Tampa, FL, USA, 7Amgen, Inc.,
housand Oaks, CA, USAOBJECTIVES: To evaluate effects of dose escalation on clinical outcomes of RApatients initiating TNF-blocker treatments in community practice. METHODS:TNF-blocker-naïve adult RA patients initiating etanercept, adalimumab, or inflix-imab (index) between July 1, 2005 and May 31, 2008 with �12 months’ enrollmentpost-index were identified from the Ingenix database. Patients receiving �9months TNF-blocker treatment or diagnosed with psoriasis, psoriatic arthritis, an-kylosing spondylitis, or Crohn’s disease were excluded. Rates of dose escalationusing 3 different methods were calculated using claims data. Participating physi-cians provided de-identified charts. Each chart was reviewed by 4–6 clinical rheu-matologists to evaluate and agree on overall clinical change from baseline to thevisit closest to 1 year post-index (12�3 months). Multivariate models comparedchange in clinical outcomes and dose escalation rates, controlling for differencesamong etanercept, adalimumab, and infliximab patients at index. RESULTS: Over-all, 715 etanercept, 501 adalimumab, and 393 infliximab patients were identifiedfrom claims; 141 etanercept, 115 adalimumab, and 104 infliximab patients hadevaluable charts. Patient characteristics were similar among the claims and charts.Regardless of dose escalation method used, significantly fewer etanercept-treatedpatients had dose escalations (1.8%, 5.2%, 6.7%) than patients treated with adali-mumab (9.8%, 8.6%, 10.4% respectively) or infliximab (50%, 31%, 34% respectively)(p�0.05 for all comparisons). After treatment initiation, 86% of etanercept-treatedpatients had “much better” or “better” clinical outcomes at 12�3 months, versus82% of adalimumab patients and 78% of infliximab patients. Multivariate analysesshowed significantly fewer dose escalations in etanercept patients (p�0.05), withno significant difference in clinical change score between etanercept patients andadalimumab (p�0.22) or infliximab (p�0.07) patients. CONCLUSIONS: This studyshowed dose escalation in fewer etanercept than adalimumab or infliximab patients,but similar improvements in clinical outcomes for all 3 treatments, indicating thathigher dose escalation rates may not be associated with better clinical outcomes.
CO2REAL-WORLD COST-EFFECTIVENESS ANALYSIS OF CANCER DRUGS:COMPARATIVE EFFECTIVENESS RESEARCH USING RETROSPECTIVE CANADIANREGISTRY DATA BEFORE AND AFTER DRUG APPROVALKhor S1, Krahn M2, Hodgson D3, Bremner K4, Luo J5, Hoch J11Cancer Care Ontario, Toronto, ON, Canada, 2Toronto Health Economics and TechnologyAssessment (THETA) Collaborative, Toronto, ON, Canada, 3Princess Margaret Hospital, Toronto,ON, Canada, 4University Health Network, Toronto, ON, Canada, 5The Institute for Clinical
valuative Sciences, Toronto, ON, CanadaOBJECTIVES: Using linked administrative databases from Ontario, our study exam-ned the “real world” cost, effectiveness and cost-effectiveness of Rituximab in diffuse-arge-B-cell lymphoma. METHODS: Patients were defined as those who had a diagno-
sis of diffuse-large-B-cell lymphoma according to ICD-O histology classificationbetween January 1997 and December 2007 and received either CHOP (cyclophospha-mide, doxorubicin, vincristine and prednisone) or R-CHOP (CHOP plus Rituximab) asfirst line treatment. We used a historical cohort design to compare the overall survival,toxicity profiles, direct costs, and cost-effectiveness of CHOP before Rituximab wasapproved (pre-era CHOP) with R-CHOP after Rituximab approval (post-era RCHOP).R-CHOP and CHOP patients were hard-matched on age, and then subsequentlymatched on propensity scores by use of a 1:1 matching algorithm. Propensity scoreswere calculated from demographic and clinical history information. We estimatedresource use and direct medical costs using the linked administrative data. To analyzecensored cost data, we employed and compared different methods, including thesimple non-adjusted average, the Kaplan-Meier sample average estimator, inverseprobability weighting estimator, Pfeifer and Bang’s estimator (2005) and Basu’s two-part estimator (2010). RESULTS: A total of 1131 matched pairs of patients were eval-uated. 3-year overall survival was significantly improved in the post-era RCHOP groupcompared to pre-era CHOP (69% [95%CI 66-71] vs 59% [95%CI 56-62]; Klein testp�0.001). Groups did not differ in the frequency of adverse events, but 3-year directcost was significantly higher in the post-era RCHOP group. The incremental cost-effectiveness ratio varied depending on the method employed. CONCLUSIONS: Thisstudy illustrated how different methods can be applied to observational data to esti-mate costs and cost-effectiveness. The results from this study can be compared tothose from clinical trials and economic models. This will help drug decision-makerscalibrate healthcare policies while helping researchers evaluate assumptions madeand methods used in economic models.
CO3PROJECT LIBRA: A NEW ANALYTIC TOOL FOR COMPARATIVE EFFECTIVENESSANALYSES OF MULTIPAYER CLAIMS DATABASESMark T1, Pepitone A2, Hatzmann M1, Navathe A3, Goodrich K3, Chang S11Thomson Reuters, Washington, DC, USA, 2Thomson Reuters, Santa Barbara, CA, USA,3Assistant Secretary for Planning and Evaluation, Washington, DC, USAOBJECTIVES: The project aimed to develop a secure, interactive tool to enableresearchers to perform comparative effectiveness studies and other types of re-search on a multipayer claims database with reduced need for complicatedprogramming. METHODS: A common data model, through which multiple datasources are standardized and linked via common data structures and vocabularies,was established. It was used to format five administrative databases: the MedicareChronic Condition Warehouse, the Thomson Reuters MarketScan® Medicaid Mul-tistate, Medicare Supplemental, and Commercial databases, and the Healthcare
Cost and Utilization Project National Inpatient Sample database. A web-based Us-er-Interface was developed that captures the logic typically required by CER meth-
ods and capitalizes on the longitudinality of administrative data. Tools were de-veloped to allow users to search taxonomies to select particular drugs, diagnoses,or procedures by typing in substrings of the numeric codes or textual descriptions.The tool allows researchers to apply enrollment and demographic constraints andcreate variables. CER studies were conducted including a comparison of atrial fi-brillation treatment with rate or rhythm control medications. RESULTS: The tool
llowed users to quickly define a study sample. Flow diagrams graphically illus-rated the attrition of the sample size and visualization of treatment and outcomes.mbedded SAS procedures enabled reporting and analysis of comparison popula-ions. The analyses revealed a higher rate of coronary artery disease and heartailure prior to drug initiation among the amiodorone versus the calcium channellocker population and a higher rate of post-drug initiation acquired hypothyroid-
sm, and pulmonary disease among the amiodorone versus the calcium channellocker population. CONCLUSIONS: New data designs and software analytic toolsay allow claims databases to be more efficiently leveraged. The tool developed for
his project has the following advantages: 1) allows for a substantial portion of theesearch exploration, hypothesis generation; and statistical analysis to be per-ormed in real-time using a web-based interface; 2) improves the speed of research;nd 3) allows access to a multipayer database.
O4OTENTIAL COST SAVINGS FROM COMPARATIVE EFFECTIVENESS RESEARCH:ESSONS FROM COURAGE STUDY
Bonakdar tehrani A, Howard DEmory University, Atlanta, GA, USAOBJECTIVES: During the debate over health reform, comparative effectiveness re-search was touted as a relatively painless approach to reducing costs. A compara-tive effectiveness study of two treatments will find either that the costlier treat-ment is more effective or is not more effective than a less expensive alternative.Studies that report negative results have the potential to reduce costs, but only iffindings affect clinical practice. One concern is that the same factors that promoterapid adoption of new therapies in the U.S. may retard the abandonment of exist-ing technologies found to be ineffective. METHODS: The COURAGE trial found thatoptimal medical therapy is as effective as percutaneous coronary intervention (PCI)for patients with stable angina. PCI refers to stenting and angioplasty. The trial waspublished and widely publicized in early 2007. We evaluate trends in PCI volumepre- and post-COURAGE by indication using 1) 100% samples of outpatient andinpatient discharge data for California, Florida, New Jersey, and Maryland, 2) a 100%sample of discharge data for Veteran’s Administration hospitals and 3) data from aproprietary cardiac catheterization laboratory registry at 15 hospitals. RESULTS:Between the fourth quarter of 2006 and the fourth quarter of 2007, PCI volume inCalifornia, Florida, New Jersey, and Maryland among patients without serious cor-onary disease declined from approximately 17,000 to 13,000 procedures (an 18%decline). There was only a 5% decline among patients with unstable angina, whowere not included in COURAGE. We found similar patterns in the other datasets.CONCLUSIONS: Publication of the COURAGE trial had an impact on PCI volume.Many patients with stable angina continue to receive PCI. The results are consis-tent with the view that as long as the health system is configured around proce-dural-based medicine, the impact of trials which find that medical therapy is aseffective as invasive procedures will be modest.
PODIUM SESSION I:EFFECTS OF DRUG MANAGEMENT PROGRAMS ON PATIENTS
DM1IMPACT OF A PHARMACY REFILL MANAGEMENT SYSTEM ON OUTCOMES INEND STAGE RENAL DISEASE (ESRD) PATIENTSRubin JL1, Wilson SM1, Golomb J21DaVita Clinical Research, Minneapolis, MN, USA, 2DaVita Rx, San Bruno, CA, USAOBJECTIVES: In dialysis patients, bone and mineral (phosphorous, calcium) andregulatory hormones (parathyroid hormone (PTH)) become dysregulated, increas-ing risk of fractures, cardiac events and death. First line treatment is a low phos-phorus diet and prescription phosphate binders. We examined the impact of a refillmanagement system (RMS) - which helps patients proactively manage their refillsusing predictive algorithms and refill reminders for prescriptions - on serum phos-phorus, calcium and PTH in patients prescribed phosphate binder monotherapy.METHODS: Data from a large dialysis organization were used to identify dialysispatients prescribed monotherapy phosphate binder between 1/1/2008-9/30/2010with at least 6 months of follow-up. Patients enrolled in the RMS were 1:1 propen-sity score matched to patients not enrolled utilizing age, race, gender, dialysisvintage, body mass index, baseline laboratory values (albumin, calcium, Kt/V,phosphorus, PTH, normalized protein catabolic rate), Charlson comorbidity score,and other comorbid conditions commonly associated with ESRD. The matchedcohorts were compared on the percent meeting guideline ranges for phosphorus(3.5-5.5 mg/dL), corrected calcium (8.4-9.5 mg/dL) and PTH (150-600 pg/mL). Valueswere assessed over the 6-months following the first phosphate binder prescription.Differences between groups were tested using chi-square for proportions.RESULTS: 3,247 RMS patients met the inclusion criteria and were matched 1:1 to acohort of non-RMS patients. There were no significant differences between thegroups on any baseline variables. Patients enrolled in the RMS were more likely tobe in target range over the 6 month period on all 3 measures: phosphorus (58.0% vs55.1%); corrected calcium (74.3% vs. 69.1%) and PTH (80.5% vs. 77.2%), compared topropensity matched controls. All differences were significant at the p�0.05 level.
CONCLUSIONS: Results indicate that participation in a pharmacy RMS is associ-ated better laboratory outcomes for dialysis patients.
A
MIc
t7rtpd
paCt
P
B
A3V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
DM212-MONTH OUTCOMES OF A PHARMACIST-PROVIDED TELEPHONEMEDICATION THERAPY MANAGEMENT (MTM) PROGRAMMoczygemba L1, Barner JC2, Gabrillo E31Virginia Commonwealth University, Richmond, VA, USA, 2University of Texas at Austin,
ustin, TX, USA, 3Scott & White Health Plan Prescription Services, Temple, TX, USA
OBJECTIVES: Determine if Medicare Part D beneficiaries who received telephoneTM services had:1) Decreased medication/health-related problems (MHRPs); 2)
mproved medication adherence; and 3) Decreased total Part D drug costs whenompared to a control group. METHODS: Part D beneficiaries from a Texas health
plan participated. The Andersen Model was the theoretical framework. Indepen-dent variables were: predisposing factors (age, gender, and race); and need factors(number of medications and chronic diseases and medication regimen complexity(MRC). The health behavior (intervention) was MTM utilization. Outcomes werechange (from baseline to 12-month follow-up) in: 1) Number of MHRPs; 2) Medica-tion adherence measured by the medication possession ratio (MPR); and 3) Totaldrug costs. Descriptive and inferential statistical analyses were conducted.RESULTS: The intervention (n�60) and control (n�60) groups were not statisticallydifferent regarding age (71.2�7.5 vs.73.9�8.0), medications (13.0�3.2 vs.13.2�3.4),chronic diseases (6.5�2.3 vs.7.0�2.1) or MRC [(21.5 (range 8–43) vs.22.8 (range9–42.5)], respectively. The majority (51%) were male in the intervention group butonly 28% were male in the control group (p�0.009). At baseline, 4.8� 2.7 (interven-ion group) and 9.1� 2.9 (control group) MHRPs were identified and 2.2�2.0 and.3�3.0 MHRPs remained at the 12-month follow-up, respectively. Multivariateegression revealed that MHRPs decreased significantly (p�0.0120) among the in-ervention group when compared to the control group. There were no significantredictors of change in MPR. Total drug costs (change from baseline to follow-up)ecreased by $588�$2,086 in the intervention group and increased by $207�$1,752
in the control group. A t-test indicated the cost difference between the 2 groups wassignificant (p�0.03), but the multivariate regression did not indicate significantpredictors. CONCLUSIONS: A telephone MTM program positively impacted MHRPs.Unadjusted cost comparisons also showed cost savings among the interventiongroup. Future research should focus on understanding predictors that impact ad-herence and cost-related MTM outcomes.
DM3IMPACT OF MONTHLY PRESCRIPTION CAP ON MEDICATION PERSISTENCYAMONG PATIENTS WITH DIABETES, HYPERTENSION, OR HYPERLIPIDEMIAWang CC, Wei D, Farley JUniversity of North Carolina at Chapel Hill, Chapel Hill, NC, USA
OBJECTIVES: To evaluate the effect of a policy implemented in the Louisiana Med-icaid program which capped reimbursement to eight prescriptions per member permonth on medication persistency in patients with diabetes, hypertension, orhyperlipidemia. METHODS: A pre-post non-equivalent comparator group designwas applied using Medicaid claims data from 2001-2003 for Louisiana (LA) andIndiana (IN) to identify patients with the specified conditions and their medicationpersistency. Persistency was defined as the number of days to discontinuationwhich was identified as a gap in treatment 30 days or longer. To capture pre-intervention trends in medication persistency, we compared “pre-policy” cohortsin LA and IN followed for ten months prior to policy adoption (March 3, 2002 toDecember 31, 2002) to policy cohorts followed ten months after policy adoption(March 3, 2003 to December 31, 2003). All incident cohorts were identified using asix-month washout period. We used Cox-proportional hazard models to compareddiscontinuation rates in LA and IN across the pre-policy and policy period cohorts.RESULTS: After adjusting for patient characteristics and comorbid conditions, nosignificant differences in persistency were found prior to policy implementationbetween LA and IN for any of the three chronic conditions. In the post-policyperiod, all cohorts had significantly lower persistency in LA than in IN. Patients inLA with diabetes and hypertension were 1.38 (p� 0.03) and 2.00 (p�0.01) timesmore likely to discontinue their medications at day 30 of the follow-up, respec-tively. The hazard ratios declined to 1.21 and 1.67 for diabetes and hypertensionpatients respectively after 260 days. The hazard ratio of discontinuation for pa-tients with hyperlipidemia in LA was constantly 1.31 (p�0.01) across the follow-upperiod. CONCLUSIONS: Patients with chronic conditions subject to medicationcaps may be vulnerable to medication discontinuation. Policy makers need to con-sider carefully when implementing such policies on patients with chronic condi-tions.
DM4EVALUATION OF CLINICAL LABORATORY-PHARMACY LINKAGE DECISIONSUPPORT IN THE USE OF POTASSIUM SUPPLEMENTSYu S, Galanter W, Lin FJ, Lambert BUniversity of Illinois at Chicago, Chicago, IL, USA
OBJECTIVES: Clinical decision support (CDS) has been utilized to link laboratoryand pharmacy data to optimize medication therapy. This study aimed to evaluatethe effect of synchronous and asynchronous CDS in inpatients receiving potassiumsupplements. METHODS: The synchronous and real-time asynchronous lab-phar-macy CDS was implemented in our 450-bed academic teaching hospital in June2003. Non-hemolyzed serum potassium ([K�]) values and medication orders for
otassium supplements, 12 months prior to and after CDS implementation, werenalyzed. A Cox proportional hazards model was constructed to assess the effect ofDS in improving clinicians’ response time at the presence of hyperkalemia. Po-
assium �5.0 mEq/L and �5.4 mEq/L were used to define high normal and elevatedhyperkalemia. Response time was measured from the time of the first instance of
hyperkalemia to cancelation of the medication order. Response time was censoredat the time of a consequent normal potassium or patient discharge. RESULTS: Inthe pre-CDS period, 12.5% (1439/11512) of the potassium supplement orders werefollowed by at least one abnormal serum potassium value; whereas in the post-CDSperiod, 10.96% (1206/11005) of the orders were alerted by CDS. While controlling forpatient characteristics, posting time of the hyperkalemic results, and severity ofhyperkalemia (5.0 mEq/L�[K�]�5.4 mEq/L vs. [K�]�5.4 mEq/L), potassium supple-ment orders were cancelled more rapidly after CDS implementation (hazard ra-tio[HR]�1.14, p�0.007). Moreover, clinicians responded sooner if the result withhyperkalemia was posted during 6am-noon vs. midnight-6am (HR�1.36, p�0.001),while the response time was longer if the result was posted during noon-6pm(HR�0.73, p�0.004) or 6pm-Midnight (HR�0.57, p�0.001). Patient age, sex, race, andthe severity of hyperkalemia had no significant effect on clinicians’ response time.CONCLUSIONS: The synchronous and asynchronous real-time lab-pharmacy link-age decision support helped clinicians to manage potassium supplements in amore timely manner in patients with high normal or elevated potassium.
PODIUM SESSION I:EMPLOYEE HEALTH & PRODUCTIVITY OUTCOMES RESEARCH
OR1THE ASSOCIATION BETWEEN SELF-PERCEIVED COGNITIVE DIFFICULTIES ANDLEVEL OF DEPRESSION AMONG EMPLOYEES WITH CURRENT DEPRESSIONLawrence C1, Roy A2, Harikrishnan V2, Yu S2, Dabbous OH21Xcenda AmerisourceBergan Consulting Services, Palm Harbor, FL, USA, 2Takedaharmaceuticals International, Inc., Deerfield, IL, USA
OBJECTIVES: Many facets of job performance may be impaired by depression. Im-paired performance by depressed employees may be attributed to self-perceivedcognitive difficulties. The goal of the current study was to assess self-perceiveddeficits in cognition experienced by employees with depression. METHODS: Indi-viduals (�18 years of age) employed full-time with diagnosed depression (exclud-ing bipolar disorder) completed a Web-based computer-generated 25-minute sur-vey in February 2010 (study population identified by Harris InteractiveTM). Thepatient survey used the Perceived Deficits Questionnaire (PDQ) to assess self-per-ceived difficulties in memory, attention, planning and organization, and concen-tration using a 0-20 scale, where higher scores indicate greater impairment. ThePatient Health Questionnaire (PHQ-9) was used to assess depression severity. Theimpact of depression on the PDQ scores was assessed using a trend test based on ananalysis of covariance with age, gender, and PHQ-9 score as independent variables.RESULTS: A total of 1051 employees were included in the analysis (58% female,mean age 47 yrs, and 38% held professional employment). PHQ-9 scores indicated423 (40.25%) employees with no depression symptoms, 319 (30.35%) with mild de-pression, 166 (15.79%) with moderate depression, 82 (7.80%) with moderately se-vere depression, and 61 (5.80%) with severe depression. PDQ scores showed thatperceived cognitive functioning worsened progressively with increasing severity ofdepression symptoms (p�0.0001). PDQ scores showed the most impairment in theattention/concentration and planning/organization scales in the severely de-pressed (12.26 and 12.25, respectively) compared with non-depressed subjects (4.45and 3.75, respectively). CONCLUSIONS: In full-time employees experiencing de-pression, self-perceived cognitive difficulties worsened with increasing severity ofdepressive symptoms.
OR2ASSESSING THE RELATIONSHIP BETWEEN MEDICATION ADHERENCE ANDEMPLOYEE PRODUCTIVITYLoeppke R1, Haufle V2, Jinnett K31U.S. Preventive Medicine, Inc., Brentwood, TN, USA, 2Alere, Rosemont, IL, USA, 3Integrated
enefits Institute, San Francisco, CA, USA
OBJECTIVES: This study examined how adherence to prescribed medications forspecific chronic conditions affects lost work time and on-the-job performance in anemployed population. METHODS: Patients aged 18 to 64, with at least one of fourconditions – CAD, hypertension, Type II diabetes, or depression – were identifiedfrom claims data for employees of five employers. Patients had also completed ahealth and productivity assessment (HPA). We conducted descriptive and regres-sion analyses of the merged claims (12-months pre-HPA) and HPA data to assesswhether medication adherence, comorbidity, health risks and other factors wereassociated with workplace productivity. Adherence measures included MedicationPossession Ratio (MPR) and a dichotomous measure that assigned “adherent” pa-tients with an MPR greater than 80%. RESULTS: The study population showed highrates of adherence: mean MPR ranged from 0.70 to 0.79 across conditions. Betteradherence was associated with fewer lost hours due to absence for CAD patients onstatin therapy and diabetes patients on statin or ACE/ARB therapy. For Diabetespatients on insulin or like treatment, those with MPRs of 0.80 or higher had higherjob performance and fewer total work hours missed than individuals with MPRslower than 0.80. Unexpectedly, better adherence by CAD patients on anti-platelettherapy was associated with poorer on-the-job performance. The presence of co-morbid diseases and higher health risk status was associated with increased ab-sence and poorer job performance across the four conditions. Those in the profes-sional/executive job class typically had less absence from work but lower jobperformance. CONCLUSIONS: Despite high rates of adherence in this population,significant medication treatment effects were observed, as were consistent effectsof comorbidities, health risks and job class on lost work time. Better targeting of
treatment by job class for individuals with high health risks and comorbidities mayhelp reduce health risks and improve productivity outcomes.
1
st(t(crictaacimbegs8pdbpccb
OAU
soeTvmsaoupett
S
seuipMgas((mppta2Us4zta
SISI
L
sttmadftdsu
A4 V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
OR3THE DIRECT AND INDIRECT COSTS ASSOCIATED WITH HYPOGONADISMAMONG PRIVATELY-INSURED EMPLOYEES IN THE UNITED STATESKaltenboeck A1, Foster S2, Thomas N2, Ivanova J1, Diener M1, Bergman R1, Birnbaum H3,Swindle R2
Analysis Group, Inc., New York, NY, USA, 2Eli Lilly and Company, Indianapolis, IN, USA,3Analysis Group, Inc., Boston, MA, USAOBJECTIVES: Compare direct and indirect (workloss) costs between privately-in-ured U.S. employees with hypogonadism (HG) and demographically matched con-rols without HG. METHODS: Male employees, ages 35-64, with �2 HG diagnosesICD-9-CM: 257.2x) or �1 HG diagnosis and �1 claim for testosterone therapy be-ween 1/1/2005-3/31/2009 were identified from a privately-insured claims databaseN�12,000,000). The index date was defined as the most recent HG diagnosis withontinuous eligibility �1 year before (baseline period) and 1 year after (study pe-iod). Employees with HG were matched 1:1 on age, region, employment status, andndex year to controls without HG. Descriptive analyses compared demographicharacteristics, comorbidities, resource utilization, direct costs (reimbursementso providers for medical services and prescription drugs) and indirect costs (dis-bility and medically-related absenteeism) inflated to $2009. Multivariate analysesdjusting for baseline patient differences were used to estimate risk-adjustedosts. RESULTS: 4,269 HG employees, mean age 51, with matched controls metnclusion criteria. Compared with controls, HG employees had higher baseline co-
orbidity rates: hyperlipidemia (50.2% vs. 25.3%), hypertension (37.7% vs. 21.1%),ack/neck pain (32.0% vs. 15.7%), and HIV/AIDS (7.1% vs. 0.3%) (all p�0.0001). HGmployees had higher study period rates of inpatient stays (10.8% vs. 5.2%), Emer-ency Department visits (27.5% vs. 16.3%), outpatient visits (100.0% vs. 76.7%), pre-cription medication use (95.7% vs. 68.3%), and higher mean workloss days (19.3 vs..8) (all p�0.0001). HG employees compared with controls had higher mean studyeriod direct ($10,914 vs. $3,823) and indirect costs ($3,204 vs. $1,450); HG-relatedirect costs were $832. HG employees’ costs remained higher after adjusting foraseline differences (direct: $9,291 vs. $5,248; indirect: $2,729 vs. $1,840) (all�0.0001). CONCLUSIONS: Employees with HG had higher comorbidity rates andosts compared with demographically matched controls. Given the low HG-relatedosts, the main driver of overall costs among HG patients may be their comorbidityurden.
R4SSOCIATIONS BETWEEN JOBLESSNESS AND ALL-CAUSE HEALTH SERVICESTILIZATION IN DIABETIC WORKING AGE ADULTS IN THE UNITED STATES
Davis-Ajami ML1, Nahata M1, Seiber E1, Reardon G2, Balkrishnan R31Ohio State University, Columbus, OH, USA, 2Informagenics, LLC, Worthington, OH, USA,3University of Michigan, Ann Arbor, MI, USAOBJECTIVES: To assess associations between joblessness and all-cause emergencydepartment (ED), hospitalization, outpatient and office-based health services uti-lization in US diabetic working-age adults. METHODS: This retrospective longitu-dinal panel design used nationally-representative 2001 – 2007 pooled public do-main complete panel data from the Medical Expenditure Panel Survey (MEPS).Eligible MEPS respondents aged 24–59 years with an ICD-9-CM diabetes diagnosticcode “250”, were included. Those with pregnancy diagnostic codes, seasonal jobstatus, or prescribed insulin were excluded. Subjects reporting an employmentstatus as “not employed with no job to return to” were classified as jobless. UsingMEPS weights to account for the complex survey design, logistic regression modelsestimated associations between joblessness and the likelihood of utilization. Neg-ative binomial regression models assessed number of utilizations. The Taylor lin-earization method estimated variance. RESULTS: 2,678 subjects (means: age 48.7years [SD 0.28], BMI 31.5 [SD 0.30], Charlson Comorbidity Index 0.369 [SD 0.12]) meteligibility criteria. Compared to those employed, joblessness significantly in-creased the odds for all-cause ED utilization 64% (OR 1.64, p� 0.007) outpatientvisits 46% (OR 1.46, p� 0.011) and office-based visit 45%, (OR 1.45, p� 0.009). Job-lessness was associated with higher logs of expected counts for ED visits (�� 0.43,p� 0.005) outpatient visits (�� 0.49, p� 0.002), and office-based visits 41% (�� 0.41,p�0.000). The following covariates showed significant (p�0.05) associations acrossED, hospitalization, outpatient and office-based utilization sectors: family size, age,the Charlson Comorbidity Index, and the presence of one or more diabetes relatedcomplication. Hispanic ethnicity was associated with fewer ED visits (�� �0.51, p �0.001), and fewer hospitalizations (�� �0.41, p � 0.026) than other ethnicities.Though non-significant, compared to uninsured individuals, private and publichealth insurance coverage also showed increased odds and number of utilizations.CONCLUSIONS: Joblessness was significantly associated with increased all-causehealth services utilization.
PODIUM SESSION I:CASE STUDIES IN ADDRESSING SELECTION BIAS
SB1COMPARISON OF DIFFERENCE-IN-DIFFERENCE, PROPENSITY SCORE MATCHINGAND INSTRUMENTAL VARIABLES IN ESTIMATING COST DIFFERENCESBETWEEN TWO COHORTSCao Z, Song XThomson Reuters, Cambridge, MA, USAOBJECTIVES: Endogeneity is a common problem in retrospective claims data stud-ies because patients in claims data were not assigned to treatment by randomiza-tion. Propensity score matching (PSM), instrumental variables (IV), and difference-in-difference (DID) have been used to control for selection bias in evaluating the
impact of treatment on outcome measures. This study compares the estimatedincremental costs between typical and atypical antipsychotic medication users in
ps
patients with schizophrenia using these three methods. METHODS: Patients ages18-64 years old with at least one prescription of typical or atypical antipsychoticmedication and at least one diagnosis claim of schizophrenia (ICD-9-CM diagnosis295.xx) within 90 days of the antipsychotic medication were identified in Market-Scan® Multi-State Medicaid Database 2002–2009. The index date was the first pre-cription date of antipsychotic medication. All patients had �12-month continu-us enrollment prior to and post the index date. Outcomes were total all-causexpenditures and psychiatric-related expenditures during the 12-month follow up.he incremental costs associated with the use of typical antipsychotic medicationersus atypical medication was estimated using six regression models–three esti-ation methods (GLM, IV, and DID) on two samples (matched and non-matched
ample). IV was prescribing physicians’ characteristics. RESULTS: A total of 447typical and 4544 typical antipsychotic users met all study criteria, with a mean agef 38.9. Five out of the six models (with the exception of the DID model on thenmatched sample) did not find a significant relationship between types of anti-sychotic medications and total costs, and five out of the six models (with thexception of the IV model on the full unmatched sample) found atypical medica-ion users had significantly higher psychiatric-related costs than typical medica-ion users. CONCLUSIONS: The PS-based approach combined with the DID or IV
methods may be better than each approach alone.
SB2ZEROS AND NON-REPORTED HEALTH CARE AND WORKPLACE PRODUCTIVITYDATA: AN APPLICATION OF TWO-STAGE ESTIMATION TECHNIQUESMEASURING INPATIENT COSTS AND ABSENTEEISM ASSOCIATED WITH LOWBACK AND NECK PAINSimons WR1, Chow W2, Biondi D2, Benson C2, Kim M21Global Health Economics & Outcomes Research Inc., Summit, NJ, USA, 2Ortho-McNeil Janssencientific Affairs, LLC, Raritan, NJ, USA
OBJECTIVES: Abundance of zero values is commonly observed in cost data result-ing in skewed distribution. This analysis measured the inpatient cost and work-place absenteeism associated with low back and neck pain and demonstrated theconsequences of ignoring zeros in inpatient cost and unreported absenteeism.METHODS: We used employer-based claims from the Thomson Marketscan© Re-earch Database (2007), a database representing approximately 100 payers of insuredmployees containing health and productivity management (HPM) and health caretilization data. Adult insured employees with continuous eligibility in 2007 were
ncluded. The ICD-9 codes identified medical conditions including low back and neckain without (nociceptive pain, NOCI) or with a neuropathic component (mixed pain,IXED). Ordinary least squares (OLS) and two-stage Tobit analyses evaluated the mar-
inal inpatient costs while OLS and Heckman’s Selection Bias (HSB) were applied tobsenteeism data. Estimated inpatient costs and absenteeism using OLS versus two-tage techniques were compared. RESULTS: A total of 2,046,332 employeesmale�59.2%; mean age 40.2�11.6 years) were analyzed. Hypertension (9.8%), NOCI9.5%), diabetes (3.7%), MIXED (3.0%) and depression (1.1%) were the most prevalent
edical conditions among these employees. 1,976,952 (96.6%) employees had no in-atient episodes, thus, with no inpatient costs. Mean inpatient cost for the entire studyopulation was $537.45 (median�$0) versus $15,851.93 (median� $8,302.20) amonghose with inpatient episodes. The incremental inpatient costs associated with MIXEDnd NOCI were $1,333.02�26.67 and $328.36�15.63 using OLS versus $2,478.97 [95%CI:,148.50 – 2,811.16] and $1,242.41 [95%CI: 1,020.10 – 1,469.18] using the two-stage Tobit.nreported absenteeism occurred in 80% of the employees. Annual absenteeism as-ociated with MIXED and NOCI using OLS were 5.25�0.21 and 4.06�0.35 compared to5.92�1.06 and 16.33�2.01 hours using the HSB technique. CONCLUSIONS: Ignoringeros in cost data and unreported absenteeism may result in substantial underes-imation of inpatient cost and workplace absenteeism associated with low backnd neck pain.
B3NNOVATIVE DESIGN FOR A COMPARATIVE EFFECTIVENESS STUDY OFCHIZOPHRENIA TREATMENTS: ANALYSIS OF RECORD REVIEW DATANCORPORATING RANDOMIZATION AND PROPENSITY SCORE MATCHING
McCarrier KP1,, Durkin MB2, Dirani R2, Markowitz M2, Slabaugh SL2, Martin ML11Health Research Associates, Inc., Seattle, WA, USA, 2Ortho-McNeil Janssen Scientific Affairs,LC, Titusville, NJ, USA
OBJECTIVES: Abstraction of hospital records is currently underway at inpatientpsychiatric facilities across the United States to facilitate a large comparative ef-fectiveness study with the following goals: 1) to observe and describe re-hospital-ization patterns among patients with schizophrenia, and 2) to compare re-hospi-talizaton outcomes between patients receiving paliperidone palmitate and thosereceiving oral atypical antipsychotics. This abstract is intended to describe theinnovative design of this study. DESIGN/METHODS: This naturalistic record reviewtudy incorporates several novel design elements and a unique two-phase abstrac-ion process. In the first phase, all patients with a qualifying inpatient hospitaliza-ion for schizophrenia are identified and basic demographic, clinical, and treat-
ent data is abstracted. From this pool of potentially-eligible patients, two groupsre identified; 1) patients discharged on paliperidone palmitate, and 2) patientsischarged on oral atypical antipsychotics. Random samples of patients are drawnrom each of these groups and designated for full data collection in phase two. Inhe second review phase, these designated records are further abstracted to collectetailed demographic variables, hospitalization and treatment history, conditioneverity, comorbid conditions, and discharge characteristics. These variables aresed to model propensity scores for receipt of the target drug, and identify two
ropensity-matched cohorts for the subsequent comparative effectiveness analy-is. Pilot testing at three hospitals has confirmed the availability of key data ele-
dipicmp
SADS
dntotm
£swi0Cvtet
CAT
wawagpts
CCG
fiptzHtefipdy
A5V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
ments and supports the feasibility of this approach. CONCLUSIONS: This novelesign seeks to address the bias inherent in observational research through the
mposition of randomization. By separating data collection into a preliminaryhase collecting only variables needed for treatment identification and random-
zation and a separate full review of only these randomly-selected patient records,hart abstraction burden is minimized. Furthermore, the use of propensity scoreatching to create two matched cohorts for comparison allows greater control of
otential confounding in analyses of treatment effect.
B4METHODOLOGY FOR ASSESSING TREATMENT EFFECT IN THE PRESENCE OFISEASE SEVERITY AND COMORBIDITY IN RETROSPECTIVE OBSERVATIONALTUDIES
Kiri VAPAREXEL International, Uxbridge, London, UKOBJECTIVES: There are many examples in health outcomes research where inad-equate control for comorbidity influence has resulted in effect estimates con-founded by disease severity. Selection bias is a common feature of data from rou-tine healthcare setting where the decision to give a particular drug to a patient witha given disease is generally based on patient characteristics, including diseasecondition. Thus, failure to properly control for the bias could result in false associ-ations. Propensity scores methodology is commonly used despite its limitationsbecause of its potential for minimising the association between exposure and con-founding factors. We describe a methodology for assessing drug effect in longitu-dinal data that minimises confounding by disease severity generally associatedwith observational studies. METHODS: For a particular outcome of interest, weobtain the profiles of rates ratios from two sets of matched cohorts. In set A, pa-tients with disease X are compared with others free of X in the periods prior to andpost diagnosis of X. In set B which involves only patients with disease X, thoseexposed to treatment Y are compared with those unexposed to the drug in theperiods prior to and post exposure. The two sets of profiles are then assessed usingsimple regression over the respective periods. In effect, we attempt to disentanglethe disease and treatment effects. Data from the UK GPRD are used to assesspossible association between a particular outcome and treatment in COPDRESULTS: We found evidence of association between the outcome and COPD butnone for the drug. CONCLUSIONS: The profile approach utilizes the data collectedover the disease natural history and exposure history to assess the relationshipsbetween the outcome and both the disease and treatment. This is a key strengthoften ignored when results are reported as point estimates. By design, it also mi-nimises the effect of selection bias.
PODIUM SESSION II:RESEARCH ON METHODS: COST-EFFECTIVENESS ANALYSIS
CE1COST-EFFECTIVENESS SENSITIVITY ANALYSIS METHODS: A COMPARISON OFONE-WAY SENSITIVITY, ANALYSIS OF COVARIANCE, AND EXPECTED VALUEOF PARTIAL PERFECT INFORMATIONCampbell J1, McQueen RB1, Libby A1, Briggs A21University of Colorado, Aurora, CO, USA, 2University of Glasgow, Glasgow, UKOBJECTIVES: Advanced sensitivity methods including value of information were
eveloped to quantify overall decision uncertainty and to assess the cost-effective-ess of additional research that would reduce that uncertainty. Our objective waso compare the information gained by utilizing three alternative sensitivity meth-ds with increasing complexity: a simple one-way sensitivity analysis; probabilis-ic analysis of covariance (ANCOVA); and expected value of partial perfect infor-
ation (EVPPI) of input parameters. METHODS: We replicated and expanded apublished HIV/AIDS cost-effectiveness Markov model (zidovudine vs. zidovudineplus lamivudine in the UK) using TreeAge®. Health states included three HIV/AIDSstates and death. Our outcome of interest was the incremental net monetary ben-efit (INMB) assuming a willingness-to-pay of £20,000/QALY. We generated one-wayand probabilistic sensitivity analyses of the INMB using published input parameteruncertainties. One-way sensitivity analysis identified the 10 most influential pa-rameters. A total of 10,000 Monte Carlo draws were used to estimate the ANCOVAresults from the same ten parameters. EVPPI for each of the same ten parameterswas estimated specifying 1000 inner and 1000 outer Monte Carlo draws. We rankedthe parameters based on their influence on variation for each sensitivity methodand compared them using Spearman’s rank correlation. RESULTS: Mean INMB was9694 in favor of combination therapy. The two most influential inputs were theame across all methods, contributed 78% of variation in outcome (ANCOVA), andere the only inputs with non-zero EVPPI values. The rank order for the top ten
nputs from all methods was similar (correlation�0.99 for one-way vs. ANCOVA,.70 for one-way vs. EVPPI and 0.70 for ANCOVA vs. EVPPI, all p-values � 0.05).ONCLUSIONS: The correlation was significant between one-way and more ad-anced sensitivity analyses. Although each method provides unique information,he additional resources needed to generate advanced analyses should be weighed,specially when the outcome decision uncertainty and therefore value of informa-ion is low.
E2NOVEL WAY OF ESTIMATING COST-EFFECTIVENESS RATIOS FROM CLINICAL
RIALS WITH MISSING DATA: A SIMULATION STUDYGagnon DD1, Engelhart L21Thomson Reuters, Santa Barbara, CA, USA, 2DePuy, Inc., Raynham, MA, USA
OBJECTIVES: In a simulated dataset, evaluate incremental cost-effectiveness ratios(ICERs) adjusted for covariates and missing data using three different regression
Ca
models. The regression parameter of interest is the incremental net monetarybenefit (INMB). Models are ANCOVA, mixed effects (ME), and joint mixed effectsand log time-to-dropout (joint ME), a selection model. METHODS: Traditional cost-effectiveness analysis (CEA) uses the incremental cost-effectiveness ratio (ICER), ameasure with statistical issues and limitations with missing data. Regression anal-ysis can estimate incremental net monetary benefit (INMB) and avoid these statis-tical issues while adjusting for covariates as well as missing data. The cost-effec-tiveness acceptability curve (CEAC) generated from a family of these regressionscan identify an ICER adjusted for the factors included in the INMB regressions (theICER is the point on the CEAC where the probability of being cost-effective is 50%).Data were simulated to include missing at random (MAR) and missing not at ran-dom (MNAR). Simulated treatment effect provided a “true” INMB for model evalu-ations that included bias (absolute difference from “true”), precision (ratio of vari-ances), and CEACs with willingness-to-pay (�) values from $0 to $100K. RESULTS:The ANCOVA and ME models produced the least biased estimates. At � � $50K, bias
as $1.3K, $1.4K, and 2.3K, and precision was 1.27, 0.90, and 1.24 for ME, ANCOVA,nd joint ME, respectively. The joint ME model performed best when missingnessas high. CONCLUSIONS: Once the CEACs had been generated, deriving ICERs
djusted for covariates and missing data from those CEACs based upon INMB re-ressions proved easy and feasible. The models used in this simulation analysiserformed differently under alternative missingness conditions and were sensitiveo nonresponse mechanisms. All estimates were poor when missingness was high;uggesting prevention of missing data should be a goal of research.
E3OST-EFFECTIVENESS ANALYSIS AND BUDGET IMPACT ASSESSMENT: ARAPHICAL WAY TO COMBINE THE TWO FOR THE AID OF DECISION-MAKERS
Paulden M, Pham BUniversity of Toronto, Toronto, ON, Canada
OBJECTIVES: Cost-effectiveness analysis (CEA) has traditionally been seen as ameans of satisfying a specific and explicit social objective subject to a fixed budgetconstraint. As a result, existing CEA methods largely ignore budget impact consid-erations in health systems where budgets are not fixed. In particular, none of thetraditional methods of presenting results (such as the cost-effectiveness plane,ICER tables and CEAC graphs) can be used to summarize the results of a CEA andbudget impact assessment simultaneously. Our objective was to develop such amethod in a manner which is meaningful to decision makers. METHODS: Wepresent a novel way of combining cost-effectiveness and budget impact consider-ations into a single graph. To do this, we disaggregate the incremental costs of thenew technology into those which fall on the health budget and displace othertechnologies (resulting in forgone health) and those which lead to an expansion ofthe health budget (resulting in a net budget impact). The incremental health ben-efit of the technology and any forgone health are combined to give the net healthbenefit of the technology, which is plotted against the net budget impact. RESULTS:Our method clearly reveals the trade-off between the cost-effectiveness andbudget impact of the technology in question. This trade-off is simultaneously re-vealed across a range of plausible values of the cost-effectiveness threshold.CONCLUSIONS: Decision makers who are concerned with both the cost-effective-ness and budget impact of new technologies have tended to consider each of theseseparately, with the inherent trade-off between the two blurred in the process. Ourproposed method makes this trade-off explicit and does so across a range ofthreshold values, enabling analysts to provide meaningful information to decisionmakers while respecting decision makers’ authority in determining the appropri-ate threshold to use.
CE4USING DYNAMIC TRANSMISSION MODELS TO ESTIMATE THE COSTEFFECTIVENESS OF VACCINES: FOUR DIFFERENT METHODS AND THEIRRELEVANCE FOR DECISION MAKERSMauskopf J1, Talbird SE1, Standaert B21RTI Health Solutions, Research Triangle Park, NC, USA, 2GlaxoSmithKline Biologicals, Wavre,Belgium
OBJECTIVES: To compare the differences in the methods used to estimate thecost-effectiveness of vaccination programs using the clinical outcomes from dy-namic transmission models. METHODS: A targeted electronic literature search oftitle words in the PubMed databases was performed to identify studies publishedsince 2000 that included a description of the methods and presentation of results ofcost-effectiveness analysis of vaccination programs based on data from dynamictransmission models for any infectious disease. Further studies were identified inthe bibliographies of the initial set of papers. RESULTS: Information was abstractedrom 29 papers presenting cost-effectiveness analyses of vaccination programs fornfluenza, HPV, varicella virus, pertussis, meningococcal meningitis, rotavirus, H.ylori, and hepatitis A. Both cohort and population-based estimates of cost-effec-iveness were presented. The population-based estimates had variable time hori-ons from 1 year for influenza or pertussis (the steady state year) up to 100 years forPV, varicella, and meningococcal vaccination. All cohort analyses used a lifetime
ime horizon. Four method types for the estimation and presentation of a cost-ffectiveness ratio were identified: 1) average population values (costs and bene-ts) over a long time horizon assuming a continuing vaccination program (20 pa-ers) 2) average population values over a long time horizon assuming a limiteduration vaccination program (1 paper) 3) population values for the steady-stateear only (1 paper) and 4) cohort values with a lifetime horizon (7 papers).
ONCLUSIONS: The variability of the estimation framework (population or cohort)nd time horizon used as well as the variability in other input parameters observed
C
Ciatu
dgvvp
oooii
rp
PD
A6 V A L U E I N H E A L T H 1 4 ( 2 0 1 1 ) A 1 – A 2 1 4
in the review illustrate the problems that may be encountered in comparing cost-effectiveness estimates of different vaccination programs among themselves aswell as with other prevention or treatment interventions.
PODIUM SESSION II:RESEARCH ON METHODS: DATABASE ANALYSIS
DS1INTEGRATING DATA SOURCES TO CONDUCT COMPREHENSIVE ONCOLOGYBASED OUTCOMES RESEARCHAlbright F1, Bollu V2, Kuo KL3, Raimundo K1, Barney R3, Stenehjem D3, Brixner D31University of Utah College of Pharmacy, Salt Lake City, UT, USA, 2Novartis Pharmaceuticals
orporation, East Hanover, NJ, USA, 3University of Utah, Salt Lake City, UT, USAOBJECTIVES: Individual data sources contain non-integrated data componentsneeded to assess outcomes, resource use, and costs in cancer patients. This workdescribes methodology to integrate disparate electronic data sources in chronicmyelogenous leukemia (CML) patients with a common identifier (CI). METHODS: A
ML Patient cohort from the Huntsman Cancer Institute was created by extractingnformation across the Utah Cancer Registry; the Utah Population Database (UPDB);nd the Enterprise Data Warehouse, including Cerner inpatient and EPIC ambula-ory care clinic data. Medication use was from inpatient medication orders. Anique patient index identifier linked disparate records. RESULTS: A total of 602
patients were identified by ICD-9 diagnosis code for CML (250.1, 205.10-12) from1995 through 2009, median age � 51, 42.6% female. Of these 598 (99.3%) were linkedto the UPDB and 245 had a state death certificate. Charlson Comorbidity Index (CCI)analysis (�/� 90 days) identified 232 (38.5%) subjects with a score of zero, 199(33.1%) with 1-3, 99 (16.4%) with 4-6, 47 (7.8%) with 7-9 and 25 (4.2%) with a score of10-17 (median�2, mean� 2.6, and SD� 3.1). Inpatient admission data was availablefor 380 (63.1%) patients, with a total of 267 CML related drug orders. Procedureswere observed for 531 (88.2%) patients. Lab results were available for 564 (93.7%)subjects. Of those, BCR/ABL biomarker results were available for 210 (37.2% of alllab results) patients. CONCLUSIONS: Integrating data across different data sourcesin an academic health care center with a National Comprehensive Cancer Networkhospital can provide comprehensive health care data. This methodology may in-fluence the evolution of electronic health records, as a data resource tool for out-comes data, resource use and cost utilization across complex disease states such asCML. Future research will expand on drug data sourcing and evaluate the medicalrecord notes to evaluate CML specific outcomes.
DS2A VALIDATION STUDY OF ALGORITHMS FOR IDENTIFYING METASTATICBREAST, LUNG, OR COLORECTAL CANCER IN ADMINISTRATIVE CLAIMS DATAWhyte JL1, Engel-Nitz NM2, Teitelbaum A2, Gomez Rey G2, Kallich J11Amgen, Inc., Thousand Oaks, CA, USA, 2i3 Innovus, Eden Prairie, MN, USAOBJECTIVES: In cancer research using claims data, identifying metastases is oftenessential yet difficult. The objective of this study was to examine the validity ofalgorithms identifying metastatic breast (BC), lung (LC), or colorectal (CRC) cancerin healthcare claims data. METHODS: A proprietary clinical cancer database con-taining physician-reported clinical data on patients with BC, LC, or CRC betweenJanuary 1, 2007 and March 31, 2010, was linked to claims data. Inclusion requiredhealth plan enrollment � 3 months prior to the initial clinical cancer diagnosis
ate. Un-validated claims algorithms from previous research were identified. Aeneric metastatic algorithm with all metastatic ICD-9 codes and tumor-specificariations of the algorithm were assessed for validity. The algorithms’ validityersus the clinically reported metastases was tested using sensitivity, specificity,ositive predictive value (PPV), and negative predictive value (NPV). RESULTS: Of
14,480 patients in the database, 4631 BC (mean age 53.6 yr), 2449 LC (mean age 62.9yr), and 2058 CRC patients (mean age 58.3 yr) met inclusion criteria. Metastases atdiagnosis were recorded in 8.0% (371) of BC, 49.2% (1204) of LC, and 25.7% (528) ofCRC patients. The tumor specific algorithm for identifying metastatic BC had 53.2%sensitivity and 98.6% specificity; PPV and NPV were 77.6 and 95.8. The lung-specificalgorithm had 55.2% sensitivity and 85.3% specificity; PPV and NPV were 81.0 and62.6. Similarly, the CRC-specific algorithm had 59.4% sensitivity, 89.8% specificity,with PPV 72.9 and NPV 82.7. The generic algorithm had lower specificity and highersensitivity for all 3 cancers and a significantly lower PPV for breast cancer.CONCLUSIONS: Specificity, but not sensitivity, was high for all tumor-specific al-gorithms. Although not tested, better sensitivity might be gained by includingchemotherapy in the algorithms for some tumor types.
DS3AVAILABILITY OF LABORATORY RESULTS DATA IN A CLAIMS DATABASE INTHE UNITED STATESHorne LN, Ming EE, Doyle cAstraZeneca Pharmaceuticals LP, Wilmington, DE, USAOBJECTIVES: To describe the frequency of available laboratory results data in acommercial healthcare database, among patients who are being treated for diabe-tes or dyslipidemia and who have have at least one documented laboratory resultfor a hemoglobin a1c (HbA1c) or lipid test. METHODS: The source population wasadults continuously enrolled in a large U.S. health plan during 2009 with at leastone CPT code for a HbA1c or lipid test. Laboratory results data were consideredavailable if LOINC codes or free text identified a result recorded within �/� 3 days
f the CPT claim date. The final study cohort included only patients with at leastne result available. We calculated the 1-year person-level percent of the numberf tests ordered in 2009 that had results available. Results for each test were strat-
fied by whether the patient received an antidiabetic or antidyslipidemia drug dur-ng the same year. RESULTS: Overall, a result was available for 41% of HbA1c tests
and 42% of lipid tests. Persons with at least one prescription claim for an antidia-betic or antidyslipidemia drug had more frequent tests recorded during the studyperiod (HbA1c: mean 4.5 with drug, 2.0 without drug; Lipid: mean 3.9 with drug, 2.0without drug). However, results were less likely to be consistently available amongtreated patients: 44% of those treated (among whom 70% of tests had results), and39% of those not treated (among whom 83% of tests had results) had any resultsavailable. CONCLUSIONS: While laboratory data may enhance studies conductedin administrative claims databases, results may be inconsistently available. In thisstudy, among treated patients, 44% had any laboratory results recorded, for whomresults were missing approximately 30% of the time. An evaluation of the com-pleteness of laboratory data prior to any study is feasible and may help understandany potential bias.
DS4BURDEN OF PROOF. . .PROOF OF PRINCIPLE: REPLICATION QUANTIFICATION,REPLICATION AND VALIDATION. . . STANDARDS OF EVIDENCE IN OUTCOMESRESEARCH SURROGATE ENDPOINTS FOR ALL-CAUSE MORTALITYSimons WRGlobal Health Economics & Outcomes Research Inc., Summit, NJ, USA
OBJECTIVES: To demonstrate replication of the quantification of relationships be-tween surrogates and endpoints as well as reconciliation with previous epidemio-logical studies; original studies for heart rate as a surrogatefor all-cause mortality,pain management and gastrointestinal adverse events, and treatment for diabetesand HbA1c and HbA1c and complications. METHODS: For heart rate, three epide-miological studies from three countries using a Weibull survival analysis and Gen-eralized Estimating Equations were used; namely, the Coronary Artery SurgeryStudy (CASS), the Copenhagen City Heart Study (CCHS) and the General Practitio-ner Research Network (GPRN). These equations reproduced a meta-regression andmeta-analysis of all available placebo-controlled clinical trials with heart rate as aprognostic factor for all-cause mortality. For pain, data consisted of 2005 HealthCare Utilization Project (HCUP) and Premier. Logistic regressions were used to ob-tain evaluate and compare odds-ratios. In diabetes, Generalized Estimating Equa-tions (GEE) allowing serially correlated behavior with repeated HbA1c reading atvariable frequencies and durations between their measurement. RESULTS: Heart isconsistently prognostic for all-cause mortality. Moreover, its quantification is con-sistent, 0.00694 (P�0.001) in CASS and 0.00683 (P�0.001) in CCHS (1981-1983) and0.00717 in CCHS (1991-1993) with the Weibull. With the GEE, the coefficient is 0.0268(P�0.006) in GPRN, 0.0249 (P�0.008) in the meta-regression of controlled clinicaltrials, and 0.01595 in the GEE with CCHS data. All three equations reproduced thepublished clinical trials with odd-ratios within 1/100ths.Conditional odds-ratioswere replicated in measure between the two datasets for fecal impaction, post-operative illeus, other bowel obstruction, vomiting and abdominal pain. The dia-betic equations were replicated exactly in 3 countries, treatment and HbA1c andcomplications with coefficients within 1/100th in patients with newly diagnosedT2DM. CONCLUSIONS: These are three studies where the quantification of theelationship between a surrogate and and endpoint have beed replicated withrecision and subsequently applied to clinical trials.
ODIUM SESSION II:RUG USE AND PATIENT SAFETY
DU1COMPARATIVE SAFETY OF STIMULANT AND ATOMOXETINE ASSOCIATEDWITH THE RISK OF SUBSTANCE USE DISORDER AMONG ADOLESCENTS WITHATTENTION-DEFICIT/HYPERACTIVITY DISORDERBhattacharjee S1, Chen H2, Bhatara V3, Aparasu RR21West Virginia University, Morgantown, WV, USA, 2University of Houston, Houston, TX, USA,3University of South Dakota Sanford School of Medicine, Sioux Falls, SD, USA
OBJECTIVES: This study compared the risk of developing substance use disorder inchildren with Attention-Deficit/Hyperactivity Disorder (ADHD) utilizing stimulantand atomoxetine. METHODS: This study involved retrospective, propensity scorematched cohort assessing the risk of developing substance use disorder amongstimulant and atomoxetine users with ADHD using the IMS LifeLink Health PlanClaims Database
