Technology, Innovation & Evaluation
Call for Papers
Health Care Information Systems:
Imminent Solutions or Distant Hope
Chair: Stephen Parente, Ph.D., M.P.H.
Sunday, June 6 • 5:00 p.m.-6:30 p.m.
• Bridging Terminology and Classification Gaps between
Patient Safety Information Systems: A Comparison of Two
Coding Schemas for Near Misses and Adverse Events
Andrew Chang, J.D., M.P.H., Laurie Griesinger, M.H.S, Peter
Pronovost, M.D., Ph.D., Jerod Loeb, Ph.D.
Presented by: Andrew Chang, J.D., M.P.H., Director, Center
for Patient Safety Research, Joint Commission on
Accreditation of Healthcare Organizations, One Renaissance
Boulevard, Oakbrook Terrace, IL 60181; Tel: 630.792.5967; Fax:
630.792.4967; E-mail: anchang@jcaho.org
Research Objective: To examine the semantic relationships
between the coded data fields of an incident reporting tool
and the classification nodes of a patient safety event taxonomy
that has mapping facilities to natural language expressions
and coding schemas.
Study Design: The elements of the Joint Commission’s
Patient Safety Event Taxonomy (PSET) were mapped to the
coded fields of the Johns Hopkins Intensive Care Unit Safety
Reporting System v. 2.0 (ICUsrs) of 1248 reported events, and
compared for homogeneity. PSET encompasses 5
primary/root classifications: 1) Impact – the outcome or effect
of an incident; 2) Type – the perceptible or visible process or
structure that failed; 3) Domain – the characteristics of the
setting where an incident occurred and the type of individuals
involved; 4) Cause – the factors and agents that lead to an
incident; and 5) Prevention and Mitigation – the interventions
to reduce the incidence and effects of adverse outcomes.
Thirty-four sub-classification nodes of the PSET data
framework (with known high face and content validity) were
compared to corresponding event-related data fields in ICUsrs
on a 5-point Likert scale where 1=unmatched, 2=extrapolated,
3=related, 4=synonymous, and 5=identical. The frequency of
each Likert item was found by analysing the ratings assigned
to the 34 sub-classifications. The overall degree of similarity
between the PSET primary classifications and the ICUsrs fields
was reported by averaging the scores of the sub-classifications
within each of the root classifications.
Population Studied: ICU patients
Principal Findings: Of the 75 coded fields in ICUsrs
containing event-related information, 46 (61%) fields mapped
to PSET and 29 (39%) were unmapped. Overall, the
terminology and classification nodes of PSET showed
acceptable homogeneity with the mapped coded data fields of
ICUsrs, although some fields differed significantly. Among the
most frequently coded fields that mapped to PSET(n=34),
4(12%) were identical, 10(29%) synonymous, 5(15%) related,
4(12%) had to be extrapolated, and 11(32%) were unmatched.
Average rating by primary classification was 2.8 (Impact);
1.8(Type); 2.8(Domain); 3.6(Cause); and 2.7(Prevention).
Because events are multidimensional, the deconstructed
components(fields) were not always mutually exclusive when
mapped. Patient characteristics varied widely between coding
schemes.
Conclusions: Results suggest that standardization of patient
safety event data may not be as simple as proffered by the
2003 Institute of Medicine (IOM) report, Patient Safety:
Achieving a New Standard of Care. Integrating local specialtybased, focused reporting programs with each other or to a
national standard demands a unified viewpoint for
understanding the reported data, and creation of maps and
transformations for reconciling disparate data schemas.
Implications for Policy, Delivery or Practice: Given the
multiple efforts by states and others to develop reporting
forms, tools that translate data into a common platform are
important. As a benchmark tool, PSET explores the use of
explicitly coded terminologies and classifications to map the
entire structure of alternative data frameworks. This work
demonstrates the utility of using PSET to structure patient
safety information searching; to structure the presentation of
the retrieved information in a consistent manner; and to map
related information for further search and retrieval. We believe
that this overall approach of explicit linking of information via
PSET provides a potentially powerful capability for common
data exchange among dissimilar reporting systems.
• Use of e-Health Services (1999 – 2002): Mountains or
Molehills?
Vicki Fung, B.A., Jie Huang, Ph.D., Robert Miller, Ph.D.,
Eduardo Ortiz, M.D., M.P.H., Joseph Selby, M.D., M.P.H.,
John Hsu, M.D., M.B.A., M.S.C.E.
Presented by: Vicki Fung, B.A., Division of Research, Kaiser
Permanente, 2000 Broadway, Oakland, CA 94612; Tel:
510.891.3527; Fax: 510.891.3606; E-mail: vicki.fung@kp.org
Research Objective: Using the internet to deliver healthrelated services (e-Health) holds promise for improving the
quality and efficiency of health care for patients, yet there is
little quantitative data on e-Health use. We evaluated the use
of four e-Health services over a four-year period.
Study Design: We conducted a longitudinal study of e-Health
use by members of a large, prepaid integrated delivery system
(IDS) between January 1999 and December 2002. We
classified the available e-Health services into two types:
transactional services (drug refills and appointment
scheduling) and care-related services (advice on medical and
drug-related concerns). We then determined the number and
proportion of members who used e-Health services between
1999 and 2002. Using a generalized linear model approach,
we also assessed trends in the use of the two service types
over time.
Population Studied: The number of IDS members increased
over the study period from 3,213,571 (1999) to 3,482,511
(2002); In the year 2002, 59.4% were white, 14.3% Hispanic,
13.5% Asian, 8.2% Black, and 4.7% Other Race/Ethnicity; 24%
of members were age 17 years or less and 12% of members
were age 65 years or older. In addition, 11% of members had
Medicare; 2% had Medicaid; and 87% had commercial
insurance.
Principal Findings: The number of members who registered
for access to e-Health services increased from 51,536 (1.6%) in
1999 to 324,522 (9.3%) in 2002; the number who used any e-
Health service at least once increased from 13,261 (0.4% of all
members) to 117,174 (3.4% of all members). In 2002, 56,768
members (1.6% of all members) used the prescription drug
refill service and 83,634 members (2.4% of all members) used
the appointment scheduling service, compared with 10,759
members (0.3% of all members) who used the medical advice
service and 3,094 members (0.1% of all members) who used
the drug-advice service. The difference between the number
of users of transactional services and of care-related services
increased from 4,387 to 126,549 users per year (1999-2002),
which was statistically significant in the multivariate models
(p<0.0001). Subjects most likely to use e-Health services
tended to be between 30-64 years old, female, white, have a
regular primary care physician (PCP), live in a high SES
neighborhood, and have a high level of comorbidity.
Conclusions: While use of all e-health services is growing
rapidly, patients are not using all services equally. Use of carerelated services appears to lag significantly behind use of
transactional services, and the gap continues to increase.
Implications for Policy, Delivery or Practice: Although eHealth services have tremendous potential for improving the
delivery of health care, only a small percentage of health care
consumers currently are using transactional services, and far
fewer are using care-related services. Additional efforts to
develop and improve access, ease of use, utility, and
knowledge about such services could result in substantial
benefits in terms of access, efficiency, and quality of care.
Primary Funding Source: AHRQ
• Handheld Computers as Technology Innovation in
Clinical Practice
Ann McAlearney, Sc.D., Sharon Schweikhart, Ph.D., Mitchell
Medow, M.D., Ph.D.
Presented by: Ann McAlearney, Sc.D., Assistant Professor,
Health Services Managment and Policy, The Ohio State
University, 1583 Perry Street, Atwell 246, Columbus, OH
43210-1234; Tel: 614.292.0662; Fax: 614.438.6859; E-mail:
mcalearney.1@osu.edu
Research Objective: In the context of developing a digital
patient care environment, handheld computers stand out as
an emerging, yet soon-to-be essential, building block. These
small devices are functionally flexible, customizable, openplatform computers and, unlike personal computers and
workstations, they are easily portable to facilitate their use at
the point of care—an important feature for physicians. Our
research objective was to examine organizations’ and
physicians’ perspectives regarding the use of handheld
computers in clinical practice.
Study Design: An extensive qualitative study including seven
case studies in healthcare organizations and eight physician
focus groups. Key informant interviews were conducted at
organizations with both organizational informants and
physician informants. Focus groups were conducted at both
healthcare organizations and at national meetings of
physicians. All interviews and focus groups were transcribed
and analyzed using both deductive and inductive methods.
Population Studied: Our seven case studies were conducted
at a variety of healthcare organizations and included academic
medical centers, community hospital systems, independent
practice associations, and community hospitals. A total of 67
key informants were interviewed across the organizations,
including both organizational and physician representatives.
Our focus groups included both users and non-users of
handheld computers, and represented 54 physicians. These
respondents included both generalists and specialists, and
both physicians in training--residents and fellows--and those in
full-time clinical practice.
Principal Findings: Use of handheld computers in clinical
practice can help physicians to increase productivity, improve
interactions with patients, and enhance overall ability to be a
better doctor. However, device limitations and personal
issues both create barriers to physician use. Physicians report
concern about device security and reliability, but are
particularly concerned about their likelihood to focus on
details rather than the bigger clinical picture, and the
possibility that they will become dependent and over-reliant
on handheld computers as a substitute for their own clinical
thinking. Organizations have the opportunity to leverage use
of handheld computers as a stepping stone to other clinical
information technologies by helping to develop user comfort
with the devices. Further, organizations can promote greater
use of handheld computers by providing training, advice, and
dedicated support for physician users.
Conclusions: Organizations view clinical use of handheld
computers as a relatively inexpensive option for physician
users to develop confidence with clinical information
technologies. Physicians expect these devices to become
more useful, and most are interested in leveraging handheld
computer use. Developing strategies to support, promote,
and accommodate physicians’ use of handheld computers
can help both organizations and doctors improve the quality
of care and service to their patients.
Implications for Policy, Delivery or Practice: By
understanding what physicians need and expect from such
devices, organizations can make appropriate investment
decisions in personnel and support to successfully leverage
their investment and maximize the potential for the use of
handheld computers and other clinical information
technologies in medical practice.
Primary Funding Source: Center for Health Management
Research
• Bar Code Point of Care Medication Administration
Systems for the Prevention of Inpatient Medication Errors
Julie Sakowski, Ph.D., Jeff Newman, M.D., M.P.H., Tom
Leonard, R.N., M.P.A., Tim Schiro, PharM.D., Susan Colburn,
R.N., Beverly Michaelsen, M.B.A.
Presented by: Julie Sakowski, Ph.D., Senior Health Services
Researcher, Sutter Health Institute for Research and
Education, 345 California Street, Suite 2000, San Francisco,
CA 94104; Tel: 415.296.1808; E-mail:
sakowsj@sutterhealth.org
Research Objective: Health information technologies are
being explored as tools to reduce medical errors and increase
patient safety. One of these technologies is an electronic
point of care medication administration system that utilizes
bar codes to electronically track medications being
administered. If the drug, dosage, or timing of the medication
being administered does not coincide with that has been
ordered for a patient, the system provides a warning message
to nurse or other health care practitioner attempting to
administer the drug. The purpose of this study is to examine
the effectiveness a bar code point of care system (BPOC) in
reducing medication administration errors and to assess the
burden it places on users.
Study Design: This is a descriptive study examining the
warnings generated by a BPOC system implemented in a
community hospital and the user responses those warnings.
Errors prevented were defined as drug administrations that
were discontinued as a result of a system-generated warning.
BPOC system output data was collected through retrospective
audits of the warning and error logs. Supplemental drug order
and administration information was gathered through chart
reviews.
Population Studied: Attempted medication administrations
for adult patients admitted to a general medical/surgical unit
at a community hospital.
Principal Findings: In a sample of over 26,000 medication
administrations, 13,000 warnings were generated by the
BPOC system, including multiple warnings for some
individual administrations. We identified 300 possible
medication errors that were prevented by the BPOC system.
Preliminary reviews indicate that a small percentage of the
errors prevented potentially would have been clinically
significant. The most common types of medication
administration errors prevented with this BPOC system found
in our sample were dose early errors (45%), no order in the
system (33%), and order discontinued or expired (19%).
Conclusions: Our findings are consistent with the limited
existing literature that indicates that BPOC systems are useful
in preventing medication administration errors. However, the
frequency of warnings that need to be assessed and require a
response places a burden on users.
Implications for Policy, Delivery or Practice: BPOC and
other health information technologies are being proposed to
help reduce preventable medical errors, but there is little
empirical evidence on the true impact of these technologies in
actual practice. Extrapolating the medication errors prevented
experience in our sample to the volume of inpatient
medication administrations nationwide suggests the potential
for substantial clinical benefits from BPOC. Evidence on the
effectiveness of these technologies will aid future adoption
decisions and help guide development of the next generation
of patient safety enhancing technologies. We recommend that
provider organizations considering early adoption of BPOC
systems take into account the need for continual refinement
and optimization of these systems and dedicate the resources
necessary to undertake these developmental activities.
• Influences of Social Factors on Computer Use by Rural
Primary Care Physicians
Timothy Weddle, BSIOE, M.B.A., M.A., Kenneth Johnson,
Ph.D.
Presented by: Timothy Weddle, BSIOE, M.B.A., M.A.,
Graduate Student, Sociology, Loyola University Chicago, 6525
N. Sheridan Road - DH 9th Floor, Chicago, IL 60626; Tel:
312.208.0533; Fax: 773.508.7099; E-mail: tweddle@luc.edu
Research Objective: Computer use by physicians has long
been identified as a major potential contributor to improved
medical care quality. While there is mounting evidence that
increased computer use improves medical quality, few
physicians actually utilize computers to their full potential.
This study looks for factors within the social environment of
physicians’ medical practice that systematically produce and
reproduce the gap between desired and actual physician
information system usage. The study focuses on rural primary
care physicians (PCPs) who practice in a variety of practice
sizes and are removed from the pressures of using computers
in response to managed care.
Study Design: Experiences and opinions about computer use
were obtained from a target population of primary care
physicians in 20 mid-west rural counties using a two-stage
data collection design. The first stage involved semi-structured
interviews of approximately 45-minutes with PCPs in 5
counties. The interview transcriptions were analyzed to
identify common patterns. The interview participation rate was
33%. The second data collection stage consisted of a mail
survey to PCPs in 15 additional rural counties. The mail survey
responses were analyzed using simple tabulation. Factor
analysis was used to identify groupings of responses. The
survey response rate was 67%.
Population Studied: The target population consisted of
primary care physicians (family medicine, general practice,
internal medicine, pediatrics, and OB/GYNE) who practice in
20 mid-west rural counties.
Principal Findings: Most physicians reported using a
computer to retrieve information, but with the exception of
writing email to other physicians, only a very small minority of
physicians entered patient-related information. Physicians in
small groups see a greater need then physicians in large
groups to improve their use of computers. Physicians with 10
to 20 years of experience see less of a need for improved
computer use than younger or older physicians. The single
most influential social factor in computer use was clarity of
the social meaning of the computer within a particular
medical practice’s setting.
Conclusions: Rural PCPs are not resistant to computers per
se but rather they are resistant to technology that takes them
out of their routines. These findings indicate that the barriers
to improved computer use, and improved medical quality, are
largely social rather than individual. The development of
physician computer technology, the use of computers by
physicians, and the definition of what it means to be a
physician vacillate within a struggle to clarify the social
meaning of computers in medicine.
Implications for Policy, Delivery or Practice: The results of
the study are of interest to policy makers, providers and
organizations interested in quality of care issues, and to
organizations involved in the implementation of physician
computer systems. For instance, national policies may be
needed to assure that the definition of being a physician
includes at least a minimum requirement to provide patient
information via computer, while protecting physicians from
being required to overuse computers. Or, for instance, special
strategies may be needed to protect small physician practices
from information system mandates that disproportionately
disadvantage them vis-à-vis larger group practices.
Primary Funding Source: Self Funded Dissertation Research
Project
Call for Papers
Technology Assessment: Identifying Value in Innovation
Chair: Kathryn McDonald, M.M., B.S.
Monday, June 7 • 8:30 a.m.-10:00 a.m.
• Health Insurance Coverage and Access to Technologies:
The Case of Insurance Mandates for the Treatment of
Infertility
M. Kate Bundorf, M.P.A., M.P.H., Ph.D., Melinda Henne,
M.D., Laurence Baker, Ph.D.
Presented by: Melinda Henne, M.D., Fellow, Center for
Health Policy/Health Research and Policy, Stanford University,
117 Encina Commons, Room 210, Stanford, CA 94305-6019;
Tel: 650.723.1164; Fax: 650.723.1919; E-mail:
mhenne@stanford.edu
Research Objective: The rapid development of new
technologies during the last two decades has led to dramatic
improvements in the treatment of infertility. Correspondingly,
over the past 17 years, 15 states have passed significant
legislation mandating that health insurers offer or provide
coverage for infertility treatments. Although advances in the
treatment of infertility have clearly benefited many women and
their families, they have also raised important questions
regarding the effects of infertility treatment on the health of
mothers and babies and the appropriate role of public policy
in providing access to these treatments. In this project, we
empirically examine the effects of regulations mandating the
insurance coverage of infertility treatments on the use of
infertility services and birth outcomes.
Study Design: We combine information about the
characteristics of mandates in different states with secondary
data on birth outcomes from the National Vital Statistics
System and the process of care from the Society of Assisted
Reproductive Technologies (SART). We compare changes
over time in states that adopted mandates with changes in
states that did not. Using the birth data from 1985 to 1995, we
estimate linear probability models including indicators of state
mandate status and state and year identifiers. The dependent
variables in these models are measures of outcomes including
birth and multiple birth rates. We also analyze trends in the
use of reproductive technologies from 1990 to 1999 to
determine whether differences in changes in the rate of use of
these technologies are consistent with our findings.
Population Studied: Women age 20-49 from 1985 to 1995.
Principal Findings: We find that the states that adopted
regulation requiring comprehensive coverage of infertility
treatments have significantly higher rates of infertility
treatments and a higher number of births and multiple births
from assisted reproductive technology per capita. The effects
of the insurance mandates differ, however, by age group.
While the mandates had no effect on women 20-29, among
women 30-39, comprehensive mandates were associated with
increases in rates of births, twins and high-order multiple
pregnancies. Among women 40-49, in contrast,
comprehensive mandates were associated with higher rates of
births and twins, but not triplets.
Conclusions: Our results suggest that mandates for
comprehensive insurance coverage of fertility services have
increased the utilization of services for the treatment of
infertility. Increased access to these services, however, had
both benefits and costs in terms of outcomes. While
comprehensive mandates were associated with increased
birth rates, particularly among women age 30 and over, they
were also associated with increases in higher risk multiple
births, particularly among women aged 30-34.
Implications for Policy, Delivery or Practice: While
insurance coverage mandates may improve access to new
technologies, this access may have unintended effects. In the
case of the treatment of infertility, the benefits of higher birth
rates came at a cost of higher rates of multiple births, which
are at higher risk for complications.
Primary Funding Source: Iris M. Litt, M.D. Fund
• A Cost-Effectiveness Analysis of Left Ventricular Assist
Devices as Destination Therapy for End-Stage Heart
Failure
David Samson, B.A., Alan Garber, M.D., Ph.D., Gillian
Sanders, Ph.D., Naomi Aronson, Ph.D.
Presented by: David Samson, B.A., Associate Director,
Technology Evaluation Center, Blue Cross and Blue Shield
Association, 1310 G Street, N.W., Washington, DC 20005; Tel:
202.626.4835; Fax: 845.462.4786; E-mail:
david.samson@wro.bcbsa.com
Research Objective: This project compared the costeffectiveness of left-ventricular assist devices used as
permanent implants (as destination therapy) and optimal
medical management for patients with end-stage heart failure
who are not candidates for cardiac transplantation. Leftventricular assist devices (LVADs) augment impaired cardiac
pumping ability.
Study Design: This cost-effectiveness analysis used qualityadjusted life-years (QALYs) as the metric used for
summarizing the effectiveness of health strategies, taking into
account all healthcare costs. A simple Markov model was
used that included two health states, alive and dead. The
Randomized Evaluation of Mechanical Assistance for the
Treatment of Congestive Heart Failure (REMATCH) trial
compared LVADs with optimal medical management in 129
patients. The REMATCH trial reported a 48% reduction in the
risk of death for LVAD patients. The alive state was weighted
for quality (utility) by New York Heart Association (NYHA)
class categories (I/II or III/IV), which classify functioning with
heart failure. Monthly additions to costs were derived from
the probability of surviving, the probability of being
rehospitalized versus being an outpatient and the associated
costs of each. The analysis assumed that the average cost of
implanting the LVAD is approximately $270,000, the average
cost of being rehospitalized for 1 month is about $40,000 and
the cost of 1 month of outpatient care is around $1,700. It
was also assumed that all patients would be dead at the end
of 3 years and the cost of rehospitalization and outpatient care
would be the same for LVAD and optimal medical
management. Sensitivity analyses were performed on
multiple variables including: utilities, costs, probabilities of
rehospitalization, discount rates, and the relative survival
experience of LVAD and optimal medical management.
Population Studied: This analysis applies to patients meeting
the REMATCH trial selection criteria. These end-stage heart
failure patients may be excluded from heart transplantation
due to advanced age (over 65 years), or other major
comorbidities such as insulin-dependent diabetes mellitus
and chronic renal failure.
Principal Findings: The incremental cost effectiveness ratio
(ICER) is the increase in cost required to gain 1 QALY, by
using LVAD compared with optimal medical management. In
the baseline cost-effectiveness analysis, the ICER was
approximately $800,000. The sensitivity analysis based on
uncertainty about the relative survival of LVAD and optimal
medical management showed that the ICERs could vary
between $500,000 per QALY and $1,400,000 per QALY.
Results also appeared to be highly influenced by the cost of
initial hospitalization for LVAD implantation. ICERs of
$500,000/QALY or less could only be achieved by making
quite improbable assumptions that the cost of LVAD
implantation is half the average cost or less and that LVADs
achieve survival past 3 years in a substantial proportion of
patients, well beyond the survival observed in the REMATCH
trial.
Conclusions: These findings indicate that use of LVADs as
destination therapy for end-stage heart failure in patients
ineligible for heart transplantation exceeds common standards
of cost-effectiveness.
Implications for Policy, Delivery or Practice: Use of LVADs
in this setting raises economic and societal issues of resource
allocation for end-of-life care.
Primary Funding Source: Blue Cross and Blue Shield
Association
• The Cost-Effectiveness of RSV Prophylaxis: Using
Decision Analysis to Build a Better Guideline
Melony Sorbero, Ph.D., M.S., M.P.H, Nahed ElHassan, M.D.,
M.P.H., Timothy Stevens, M.D., Caroline Hall, M.D., Andrew
Dick, Ph.D.
Presented by: Melony Sorbero, Ph.D., M.S., M.P.H, Associate
Scientist, RAND Corporation, 201 North Craig Street, Suite
202, Pittsburgh, PA 15213; Tel: 412.683.2300; Fax:
412.683.2800; E-mail: Melony_Sorbero@rand.org
Research Objective: Premature infants are at high risk for
respiratory syncytial virus (RSV) related hospitalization.
Palivizumab, a humanized monoclonal antibody, is approved
for RSV prophylaxis. The objectives of this project are to
evaluate the cost-effectiveness of current American Academy
of Pediatrics (AAP) guidelines for the use of palivizumab in
premature infants without chronic lung disease (CLD) and
identify more cost-effective alternative guidelines.
Study Design: We constructed a decision analytic model from
the societal perspective to evaluate the cost-effectiveness of
palivizumab relative to not using a prophylaxis. Two versions
of the model were constructed (with and without asthma). An
association between severe RSV infection and childhood
asthma has been identified, but causality has not been
established. Models including asthma utilize semi-Markov
processes to allow the risk of asthma to vary as the child ages.
The literature indicates no increased mortality due to RSV
infection; therefore, cost-benefit analyses were performed on
models without asthma. The models combined published
data about the risk of RSV hospitalization by gestational age,
the efficacy of palivizumab, national data on the costs of RSV
hospitalizations, palivizumab injections, emergency room
visits, and hours missed from work by parents for office and
emergency room visits and hospitalizations. Costs were
adjusted to 2002 dollars. Drug wastage was incorporated into
the models. In models including the risk of asthma, the
impact of asthma on costs and quality of life were
incorporated. Future benefits and costs were discounted at
3%. Simulations modifying the current AAP guidelines were
performed to identify alternative policies with improved costeffectiveness.
Population Studied: Seven hypothetical cohorts of premature
infants without CLD born at 26-32 weeks’ gestation and
discharged from the neonatal intensive care unit (NICU) at 36
weeks post-conceptual age. Month of discharge was equally
distributed throughout the year.
Principal Findings: The expected costs of the current AAP
guidelines exceeded the expected costs of not using the
prophylaxis for all gestational ages. When the risk of asthma
was included in the models, the incremental cost-effectiveness
of palivizumab remained excessive and exceeded $1,000,000
for many of the gestational ages. While the elimination of drug
wastage improved the incremental cost-effectiveness, it did
not approach a recently suggested guideline of $200,000 for
any of the gestational ages. Within each gestational age, the
cost-effectiveness of the use of palivizumab varied widely by
month of discharge from the NICU. Simulations indicate
revising AAP recommendations to limit the use of
palivizumab to infants born at 28 weeks’ gestation or less for
their first RSV season only. This combined with younger age
cutoffs at the start of the RSV season would result in a costeffective guideline if drug wastage can be avoided.
Conclusions: Current AAP recommendations are not costeffective. Simulations with decision-analytic models can be
used to develop cost-effective guidelines for the use of
palivizumab in infants without CLD.
Implications for Policy, Delivery or Practice: AAP
recommendations for palivizumab use should be revisited.
Delivery settings to minimize drug wastage, such as providing
injections on a specific day each month during RSV season,
should be explored to improve the cost-effectiveness of the
prophylaxis.
• Increasing Health Care Costs: The Price of Innovation?
Barry Friedman, Ph.D., Claudia Steiner, M.D., M.P.H.,
Roxanne Andrews, Ph.D., Herbert Wong, Ph.D.
Presented by: Claudia Steiner, M.D., M.P.H., Senior Research
Physician, Agency for Healthcare Research and Quality, 540
Gaither Road, Rockville, MD 20850; Tel: 301.427.1407; Fax:
301.427.1430; E-mail: csteiner@ahrq.gov
Research Objective: Health care spending in the US has
accelerated since 1998, along with resource cost of inpatient
care. Although inpatient costs do not increase as rapidly as
other components, inpatient costs are the largest single
component of health care costs. Our objective was to
determine what categories of treatments or patient
characteristics contributed most to the growth of inpatient
costs from 1993 to 2001, and to the acceleration after 1998.
Changes in medical technology that favor more expensive over
less expensive treatments, or use of new expensive
technologies could be the major contributor to the rising cost
of inpatient care. The relative influence of other factors, such
as easing of managed care restrictions, more severe inpatient
case mix, or a more "defensive" style of medical practice, are
also considered.
Study Design: A Clinical Classification System (CCS) with 267
disease categories was used to categorize all principal
diagnosis codes. Charges from discharge records were
discounted to cost using hospital accounting reports from the
Centers for Medicare and Medicaid. Changes in inpatient
costs were determined for the overall time period 1993 to
2001, and for the two relevant time periods: 1998-2001 and
1993-1998. The contribution of a CCS category to the overall
increase in inpatient costs was determined as the product of
the initial share of total costs in a base year multiplied by the
percentage increase between the base and ending year. The
CCS categories were rank ordered by their contribution to the
national increase of inpatient costs for 1993-2001. Changes in
severity of illness scores within category (using the APR-DRG
system), average age, length of stay and volume of discharges
were investigated.
Population Studied: This study uses data from the 1993 2001 Nationwide Inpatient Sample (NIS) of the Healthcare
Cost and Utilization Project. The NIS is the largest, all-payer
discharge abstract data base, produced through a FederalState-Industry partnership. Approximately 7 million discharges
for 1000 hospitals in the NIS can be weighted to annual
discharges in US community hospitals.
Principal Findings: There were an estimated 37.2 million
discharges in US community hospitals in 2001 with an
estimated total cost of $213.9 billion. The number of
discharges increased by only 0.86% per year from 1993 to
2001, while cost per case increased by 1.6% per year. Both of
these rates were substantially higher after 1998 (2.2% and
3.3% respectively). Interestingly, the market basket index for
hospital input costs did not accelerate after 1998.
The top 50 CCS categories contributed 95% of the overall
increase in inpatient costs. The top disease category, heart
attack, contributed 6.7% of the national increase in inpatient
costs. Two of the top 5 categories had an increase of cost
more than 3 times the national rate of increase: back disorders
and heart dysrhythmias. Dysrhythmia cases and cost per case
rose considerably more than average. This condition has
documented improvements in technology. The increased cost
in back disorders was associated with a relatively high increase
in cost per case. Published reports have pointed to increased
uses of certain technologies even when the value has been
questioned.
Comparing increases after 1998 to the previous 5 years, 6 of
the top 50 disease categories had a significant change in
growth of admissions (non-specific cheat pain, pregnancy,
anemia, abdominal pain, benign neoplasm and connective
tissue diseases), while 29 categories showed significant
change in growth of cost per case. Examples are cited showing
many of these disease categories involve technology change,
either introduction of a new medical innovation, or changes in
practice favoring a more expensive technology. Overall,
average length of stay declined more rapidly before 1998 than
after (-16.5% vs. -3.6%). Most of the increase in average age
occurred before 1998 (4.8% vs. .5%). Severity of illness
declined in most categories after 1998.
Conclusions: Our study disaggregates inpatient care allowing
attention to inpatient conditions where costs are accelerating.
Our study demonstrates that inpatient costs accelerated due
to the number of admissions for selected conditions and cost
per case beyond the input price index, but not due to changes
in age, length of stay or severity of illness.
Implications for Policy, Delivery or Practice: In an era of
eased managed care restrictions, new technologies and
increased use of existing technologies appear to be quite
important in the treatment of conditions that contribute the
most to increasing inpatient costs. Some of these technology
uses deserve more scrutiny for cost-effectiveness and patient
outcomes. Personnel costs are a general contributor to
increasing cost per case, but not to the recent acceleration. In
specific patient categories, avoidable costs due defensive
practices or inadequate preventive outpatient services warrant
further testing and concern due to avoidable hospitalization
costs.
Primary Funding Source: AHRQ
• Unequal Utilization of New Technologies by Race:
Adjusting for Geography in the Use of TUNA and TUMT
among Medicare Beneficiaries
Xinhua Yu, MB, Ph.D., Alexander McBean, M.D., M.S., Debra
Caldwell, M.S., Janet Anderson
Presented by: Xinhua Yu, MB, Ph.D., Research Associate,
Division of Health Services Research and Policy, University of
Minnesota, MMC 79 Mayo, 420 Delaware Street, S.E.,
Minneapolis, MN 55455; Tel: 612.624.1411; Fax: 612.378.4866;
E-mail: yuxx0131@umn.edu
Research Objective: Many studies have documented lower
rates of surgical procedures among Black Americans than
Whites. Little is known about the availability and accessibility
of new technologies. Two new surgical procedures,
transurethral microwave thermotherapy (TUMT) and
transurethral needle ablation of prostate (TUNA) have recently
been introduced as alternatives for the traditional surgical
treatment of benign prostate hypertrophy (BPH), transurethral
resection of prostate (TURP). This study compares the
availability and accessibility (utilization) of these three
procedures among elderly black Medicare beneficiaries with
white beneficiaries. Rates of TURPs were included to provide
information on the usual variation in BPH surgical treatment
by race.
Study Design: Data source: Medicare administrative claims
data from 1999-2001. Institutional and physician claims for
BPH were extracted from Center for Medicare and Medicaid
Services’ (CMS) National Claim History repository. National
race-specific age-adjusted procedure rates were computed.
Geographic areas (counties) were identified where the new
technologies were available. The same rates were calculated
for residents of these counties and which had more than 10
black males beneficiaries using generalized estimate equation
models to adjust for geographic variation.
Population Studied: Black and White male Medicare
beneficiaries age 65+ years of age with a diagnosis of BPH
who had a TUMT, TUNA, and TURP procedure during 19992001 (N=140,207). Those who were enrolled in managed care
or who had ESRD or prostate cancer during the study years
were excluded.
Principal Findings: During 1999-2001, the national ageadjusted rates for TUMT, TUNA, and TURP among Whites
were 5.7, 1.7, and 40.7 per 10,000, respectively. Among Blacks,
the rates were 2.6, 0.9 and 36.6 per 10,000. Residents of 631
counties had TUMT available. TUNA was available in 406
counties. The ‘TUMT-available’ counties accounted for 57% of
the White male Medicare elderly enrollees and 69% of the
Black enrollees. The ‘TUNA-available’ counties had 46% of the
Whites and 57 % of the Blacks. In these counties, after
adjusting for age and geographic variation, Whites had higher
procedure rates per 10,000: 6.9 for TUMT, and 2.6 for TUNA,
than Blacks: 4.0 for TUMT, and 1.7 for TUNA. The rate ratios
(W/B) were 1.73 (95%CI: 1.52-1.96) for TUMT, and 1.51
(95%CI: 1.18-1.95) for TUNA. For comparison, the adjusted
rates for TURP were 41.8 per 10,000 for Whites and 38.1 per
10,000 for Blacks, and the rate ratio was 1.10 (95%CI: 1.001.20).
Conclusions: Interestingly, the new procedures for BPH were
more available to a greater percentage of Blacks than Whites.
However, after adjusting for the availability of the new
procedures, elderly Black Medicare beneficiaries were less
significantly likely to receive the new BPH procedures than
Whites, while the utilization of TURP did not differ
significantly between Blacks and Whites.
Implications for Policy, Delivery or Practice: One of the two
major goals of Healthy People 2010 is to reduce racial
difference in health care. This study demonstrates although
there is no known greater effectiveness using the new
procedures, these services are more frequently used on White
beneficiaries. Differences by race in the use of new
technologies with proven greater effectiveness should be
monitored.
Primary Funding Source: CMS
Related Posters
Poster Session B
Tuesday, June 8 • 7:30 a.m.-8:45 a.m.
• Is Less Intensive Treatment Really Cost Effective?
Understanding Common Biases
Jeremy Bray, Ph.D., Katherine Harris, Ph.D., Gary Zarkin,
Ph.D., Debanjali Mitra
Presented by: Jeremy Bray, Ph.D., Senior Research
Economist, Behavioral Health Economics Program, RTI
International, 3040 Cornwallis Road, Research Triangle Park,
NC 27709; Tel: 919.541.7003; Fax: 919.541.6683; E-mail:
bray@rti.org
Research Objective: The proposed project uses simulation
methods to illustrate for a non-technical audience the
misleading inferences that can result from failing to account
for client heterogeneity in studies of the cost effectiveness of
substance abuse treatment. In particular, we show how
ignoring client heterogeneity can lead to the systematic biases
that understate the cost effectiveness of resource intensive
modalities (i.e., inpatient treatment). We address two
common ways in which client heterogeneity goes
unmeasured: (1) through the strict criteria governing
participation in clinical trials, which leads to the exclusion of
the most severely impaired clients and (2) through the
inability to observe key clinical characteristics in readily
available forms of observational data (i.e., administrative,
survey).
Study Design: We generate hypothetical data describing the
effects and costs of two alternative treatment modalities on
prototype categories of clients. We use standard costeffectiveness methodologies (e.g. ordinary least squares,
propensity score methods) to analyze our hypothetical data in
two ways: (1) excluding the most severely impaired category of
clients from our data to simulate clinical trial conditions and
(2) ignoring key clinical characteristics that determine category
membership to simulate observational study conditions. We
illustrate the effect of ignoring patient heterogeneity
by comparing the results of our two analyses with what we
know to be the true cost effectiveness.
Population Studied: Our simulations are designed to
reasonably represent the treatment assignment and outcomes
process for individuals with substance use disorders who are
candidates for both multiple treatment modalities.
Principal Findings: Under the assumption that inpatient care
improves compliance for severely addicted patients who
would otherwise drop out of outpatient care, we illustrate how
two common study designs lead to biases
against finding inpatient care cost effective. The first case
occurs in non experimental settings when analysts are unable
to fully observe and incorporate information on patient
severity into their models. The second care occurs in
experimental settings where the most severely addicted
patients are excluded, either by design or implicitly.
Implications for Policy, Delivery or Practice: Over the past
two decades, economic considerations have played an
increasingly large role in determining third-party coverage for
substance abuse treatment. Over this time period, there has
been a steady decline in the use of inpatient treatment
modalities based on the perception of poor economic value.
At the same time, the adoption of clinically effective
treatments increasingly requires evidence of the cost
effectiveness of new treatments relative to standard practice.
A number of new approaches to the screening and treatment
of substance use disorders have shown promise in clinical
trial settings. Because coverage decisions will be made on the
basis of cost effectiveness analysis, it is important that they
provide valid information.
• Does Mammographic Detection of DCIS Reduce Invasive
Breast Cancer?
Andrew Dick, Ph.D., Kathryn Sattelberg, B.A., Melony Sorbero,
Ph.D., M.S., M.P.H., Jennifer Griggs, M.D., M.P.H.
Presented by: Andrew Dick, Ph.D., Assistant Professor,
Community and Preventive Medicine, University of Rochester
School of Medicine and Dentistry, Box 644, Rochester, NY
14642; Tel: 585.275.3276; Fax: 585.461.4532; E-mail:
Andrew_Dick@URMC.rochester.edu
Research Objective: The diagnostic incidence of ductal
carcinoma in situ (DCIS) has increased over 500-fold in the
last 3 decades, largely because of the increased use of
mammography, with which about 85% of DCIS cases are
detected. It is not clear, however, whether detection and
treatment of DCIS prevents development of invasive breast
cancer (IBC), calling into question the role of mammography
in both outcomes and efficiency. This study investigates the
link between DCIS and IBC and characterizes the natural
history of progression from DCIS to IBC.
Study Design: We identified all cases of DCIS (n = 1,157) and
IBC (n = 9,072) in Monroe County, NY from 1985 - 2002 using
the Monroe County Tumor Registry. We linearly extrapolated
and interpolated zip code level population totals for women
age 35 – 79 by year from 1990 and 2000 census data. We then
generated incidence rates for DCIS and IBC by zip code and
year from 1985 - 2002. To investigate the impact of DCIS
detection on IBC incidence, we estimated regression models
in which IBC rates were specified as flexible functions of time
and lagged DCIS rates. We allowed for IBC rates to be affected
by lagged DCIS rates (each of the five prior years) because of
the possibility that transit time from in situ to invasive disease
is heterogeneous. We included zip code-level fixed effects to
eliminate spurious correlation due to unobservable zip-code
factors that may drive both DCIS and invasive cancer rates.
Thus, the relationship between invasive cancer and lagged
DCIS is identified by intra-zip code variations in DCIS and
invasive cancer rates over time.
Population Studied: Women ages 35-79 who resided in
Monroe County, NY, from 1985 until 2002.
Principal Findings: Bivariate analysis shows that DCIS and
IBC rates are positively correlated. Multivariate results,
however, indicate that there is a strong negative relationship
between lagged DCIS diagnosis rates and IBC rates. The
model predicts that 33% (p = .014) of diagnosed DCIS cases
would have been diagnosed as IBC cases in two years, a total
of almost 50% (p = .06) in the first three years combined, and
a total of 72% (p = .046) in the first 5 years combined. The
sum of the effects of lagged DCIS rates from years 1 through 5
is significantly different from 0 (p = .046). The relationships
become stronger when socioeconomic and demographic
controls are included in the model.
Conclusions: Results indicate that DCIS diagnoses reduce
future IBC diagnoses. This suggests that the natural history of
DCIS is progression to IBC at high rates within two to five
years. Because recurrence rates following treatment of DCIS
are low, the detection of DCIS prevents many cases of invasive
breast cancer.
Implications for Policy, Delivery or Practice: Our results
underscore the importance of mammographic screening for,
detection of, and treatment of DCIS. By detecting DCIS, the
widespread use of mammography prevents many cases of
invasive breast cancer.
Primary Funding Source: NCI
• Using Encoded Guidelines to Improve Screening of Adult
Patients at High Risk of Latent Tuberculosis Infection
Sheri Eisert, Ph.D., Andrew Steele, M.D., M.P.H., Art
Davidson, M.D., M.P.H., Nedra Garrett, M.S., Patricia Gabow,
M.D., Eduardo Ortiz, M.D., M.P.H.
Presented by: Sheri Eisert, Ph.D., Director of Health Services
Research, Health Services Research, Denver Health, 777
Bannock, MC 8701, Denver, CO 80204; Tel: 303.436.4072;
E-mail: sheri.eisert@dhha.org
Research Objective: To determine the impact of automated
computer-generated alerts and web-based documentation
within an electronic medical record (EMR) on screening rates
for the latent tuberculosis infection (LTBI) in accordance with
the CDC guidelines.
Study Design: Non-randomized, pre- (4-month) and post- (3
month) intervention study from October 2002 through April
2003. The number of patients appropriately screened for LTBI
was assessed, before and after implementation of a
computerized decision support system in two outpatient
clinics providing a majority of care to underserved patients in
the Denver area.
Population Studied: All adult patients who registered at two
outpatient clinics in Denver were eligible for the intervention.
All provider staff, including physicians, allied health providers
(nurse practitioners, physician assistants), resident
physicians, nurses, and medical staff assistants, participated
in the study.
Principal Findings: There were 14,044 patient registrations by
8,463 patients during the 7-month study period. The average
age of these patients was 49; 64% were female, 71% were
Hispanic and 50% were uninsured. 73% percent of the
patients had at least one CDC risk factor for LTBI. High-risk
country of birth was a risk factor for 39% of patients, and 49%
had at least one clinical risk factor. Of these, the most
common were diabetes (23%), hematological disorders (17%)
and alcoholism (13%). Because the large number of patients
with risk factors for LTBI would overwhelm the clinic’s capacity
to screen and treat patients according to the recommended
guidelines, we focused our post-intervention efforts on
patients who were 18-40 years of age and born in a high-risk
country. Of the 4,135 patients registering during the postintervention period, 610 met the criteria for screening. Based
upon data entered into the online documentation tool,
providers complied with the CDC guideline in 105 patients
(17%). Among the 67 patients that had a PPD placed and
returned to have it read, 30 (45%) were positive. Based upon
a separate chart review, the compliance with the LTBI
screening guideline improved from 8.9% pre-intervention to
25.2% post intervention (183% increase) (p<0.001).
Conclusions: The majority of patients presenting to our
clinics had at least one risk factor for LTBI, which has
substantial public health implications. This project
demonstrated the success of a safety-net healthcare provider,
a federal public health agency, and an IT developer/vendor to
collaborate and encode clinical guidelines within a clinical
decision support system, resulting in a significant increase in
identifying and screening high-risk patients for LTBI.
Implications for Policy, Delivery or Practice: Use of
computerized decision support systems can significantly
increase screening of patients at high risk for LTBI and other
important conditions, such as diabetes and hypertension.
Despite the potential benefits of identifying and treating these
patients, more work is needed to understand the net benefit of
these systems, including their impact on clinical workload and
staffing, which can have significant implications in care
settings that are already overburdened and underfunded.
Primary Funding Source: AHRQ
• Oregon Health Services Utilization of GIS
Mary Ann Evans, Ph.D., M.P.H., Lynne Pettit
Presented by: Mary Ann Evans, Ph.D., M.P.H., Lead Research
Analyst, Health Systems Planning Office, Oregon Department
of Human Services, 800 NE Oregon Street, Portland, OR
97232; Tel: 503.731.4017; Fax: 503.731.4078; E-mail:
maryann.evans@state.or.us
Research Objective: Oregon Health Services Utilization of
GIS
Geographic information system (GIS) software is a powerful
tool for assessment, decision-making, and information
sharing. While GIS technology has been around for a number
of years, it more recently became available in a desktop
version with affordable access to geographical boundary files
and street addresses. Free and low cost geocoding services
simplify converting street addresses or zip codes to latitude
and longitude codes for placement on maps.
GIS provides a platform for the analysis of health data in
relationship to population demographics, socioeconomic
factors, surrounding social and health services, and the
natural environment. GIS is suitable for analyzing
epidemiological data and revealing trends and patterns of
relationships that may be more difficult to discover in tabular
or text formats. GIS helps policy makers to easily visualize
existing health and social resources and to more effectively
target resources to address health issues. It is an effective tool
for advocacy, informing the public, and generating action by
decision makers.
Oregon Department of Human Services (ODHS) houses
diverse datasets including vital statistics, child welfare, seniors
and persons with disabilities, Medicaid, TANF, WIC and
others. GIS technology facilitates visual analyses of client
characteristics and multiple services that clients receive.
Information displayed in a map resonates with many
stakeholders. They can ‘see’ the data in a sort of gestalt and
are then more interested in examining complementary text,
tables, and charts.
ODHS uses GIS for emergency preparedness planning,
evaluating racial and ethnic health disparities, and tracking
geographical trends in demography, morbidity, mortality,
poverty, education, and insurance. Maps of hospital
locations, emergency medical and public health departments
and services, emergency personnel, and road maps facilitate
emergency planning.
GIS maps assist analysis of trends and facilitate
communication with key stakeholders. For example, mapping
of health professional shortage areas, medically underserved
areas, medically underserved populations, and J1 Visa provider
placements provides a visual picture of areas with unmet
health care needs. Maps that display density of race and
ethnic populations by state, county, and census tract help
identify diverse population areas. GIS information has been
leveraged in grant applications to target research and health
service programs.
We would like to share samples of health service GIS maps to
demonstrate the usefulness of geographical information and
to promote the use of GIS in public health assessment,
research, and health care utilization. We are prepared to
provide a brief overview and demonstration of GIS software,
share resource information for the location of boundary files
and answer group questions. We are also interested in
discussing the adoption of GIS software, the Oregon
geospatial data warehouse and workgroup, training resource
requirements for personnel, and outreach efforts for the use of
GIS information by managers, administrators, and community
groups.
Primary Funding Source: Oregon State Government
• Funding Surgical Device Trials: Perspectives of
Parkinson’s Disease Researchers
Dorothy Vawter, Ph.D., J. Eline Garrett, J.D., Karen Gervais,
Ph.D., Angela Witt Prehn, Ph.D., Timothy McIndoo, M.A.,
Raymond De Vries, Ph.D., Thomas Freeman, M.D.
Presented by: J. Eline Garrett, J.D., Associate, Health Policy
and Public Health, Minnesota Center for Health Care Ethics,
601 25th Avenue, S., Minneapolis, MN 55454;
Tel: 651.690.7719; E-mail: garrette@stolaf.edu
Research Objective: As part of a larger inquiry regarding
surgical research, we explored PD researchers’ experiences
with and perspectives on securing funding for surgical device
trials or SDTs.
Study Design: Forty-eight North American surgical
researchers in PD participated in one-hour focus groups or
interviews in November 2002 and May 2003.
Population Studied: Most participants were neurologists or
neurosurgeons working at academic health centers or AHCs.
Principal Findings: Participants named funding as a major
barrier to quality SDTs. They identified multiple funders:
government, device manufacturers, third-party payers,
foundations, their institutions, themselves, and patients.
Except for foundations and government, funding research is
not these funders’ priority; instead they seek to avoid paying
and shift costs to each other. Participants stated that the US
offers a less robust infrastructure for SDTs than for drug
research.
Participants agreed that the exorbitant costs of SDTs make
government funding critical, especially for long-term,
multicenter, controlled trials. They applauded the large-scale
VA-NIH trial of deep brain stimulation for PD, but lamented
the limited funds allocated to answer other outstanding
questions about DBS and other surgical devices. Some found
NIH-supported Clinical Research Centers of great value.
Overall, they expressed frustration with the fairness of the NIH
grant process and funding decisions for SDTs.
According to participants, device manufacturers fund fewer
studies and subsidize fewer types of expenses than
pharmaceutical manufacturers. Participants reported varying
support from industry and acknowledged that industry’s
priority is generating revenue, not advancing science. Some
asserted that industry should fund more research on a
broader range of topics, though they recognized that once a
device is FDA-approved, industry often has few incentives to
invest in research.
Though insurers officially exclude coverage for unproven
interventions, participants frequently relied on insurers to fund
DBS research and other SDTs. Only a few prospectively
engaged insurers to support studies; most sought coverage
patient by patient. They reported varied coverage practices by
private insurers and regional Medicare carriers; confusion
about coding and potentially fraudulent coding; no coverage
for placebo studies; incentives to offer procedures off-protocol
prematurely; study delays; and inequitable patient access to
studies.
Some AHCs and hospitals fund small trial expenses and allow
clinicians to undercharge for research activities. Increasingly,
though, institutions are disallowing studies or off-label use of
surgical devices when the value is uncertain, funding
unreliable, external monitoring requirements too expensive, or
indemnification unavailable.
Conclusions: Participants characterized funding efforts for
SDTs as time-consuming, unclear, unpredictable, and legally
and financially risky. Their accounts of piecemeal funding from
multiple sources far outnumbered instances of deliberate
decisions by funders to collaborate in supporting timely,
quality studies. These findings reinforce the work of our
public/private task force to recommend ways to support
deliberate, collaborative funding arrangements.
Implications for Policy, Delivery or Practice: Concerned that
risky medical devices are routinely diffused without data
registries, standardization, or compelling data--resulting in
lost research opportunities and exposure of patients and
society to uncharted risks and unjustified expenses-participants appealed for a more coherent infrastructure to
fund quality SDTs. Collaborative SDTs require collaborative
funding arrangements.
Primary Funding Source: NIH grant NS40883-02
• Making the Patient Encounter More Productive: Fast
Screening in Primary Care
Rachel Hess, M.D., Melissa McNeil, M.D., M.P.H., ChungChou Joyce Chang, Ph.D., Cindy Bryce, Ph.D.
Presented by: Rachel Hess, M.D., Fellow, General Internal
Medicine, University of Pittsburgh, 230 McKee Place Suite
600, Pittsburgh, PA 15213; Tel: 412.692.2025; Fax:
412.692.4838; E-mail: hessr@upmc.edu
Research Objective: To facilitate the collection of screening
information (e.g., age, marital status, education level, tobacco
and alcohol use, social support, medical comorbidities, and
health related quality of life (HRQOL) via RAND-36) directly
from patients and create a de-identified research database, we
designed the Functional Assessment System Tablet (FAST).
The FAST uses touch-screen data entry and a wireless webbased interface to facilitate real time data collection from
patients and reporting to clinicians. We hypothesized that the
system would be well received and easy to use.
Study Design: 54 consecutive patients presenting to the
University of Pittsburgh General Internal Medicine Outpatient
practice completed the FAST on 1 of 3 tablet computers. These
represented all patients seen by 2 physicians, each with 2
exam rooms and a shared medical assistant (MA), on
implementation days. Upon completion, patients responded
to the question: “Did you have trouble using the computer to
answer these questions? (0=not at all difficult, 1=some
difficulty, 2=a lot of difficulty). Additional feedback was
solicited from both patients and staff regarding the FAST.
Ease of use data was analyzed using Fisher´s exact and
Student´s t-test and patient comments were summarized to
evaluate the content validity of the questions and make
appropriate changes to the final instrument. We used staff
comments to evaluate impact on patient flow.
Population Studied: Patients seen in a General Internal
Medicine outpatient practice.
Principal Findings: During initial testing, no patients refused
to complete the FAST. Patient characteristics included: 93%
women; average age of 48 years (range 19-76); 58% with at
least a college degree; 53% married or in a committed
relationship; 38% reported a diagnosis of depression; 32%
reported a diagnosis of arthritis; RAND-36 mental health
composite (MHC) of 43±12.7; and physical health composite
(PHC) of 45±12.6. Patient commented on phrasing of
questions and consistency of response anchors. Overall,
patients found it easy to use (94% “not at all difficult”, 6%
“some difficulty”, 0% “a lot of difficulty”). We were unable to
detect any significant association between difficulty using the
tablet computer and: gender; educational attainment; MHC;
or self-reported depression or arthritis. Older age, lower
(worse) PHC, and more pain were associated with reporting
some difficulty using the tablet computer (1-sided pvalues=0.04, 0.002, and 0.02 respectively). Based on
investigator observation and discussions with physicians and
staff, the FAST did not disrupt patient flow and required 15
minutes to complete.
Conclusions: Computerized systems used to collect general
screening information (such as the FAST) can be designed to
enhance patient and staff acceptance and patient´s perceived
ease of use. The primary factors affecting ease of use are
patient age, PHC, and self-report of pain at the time of
physician visit. The FAST facilitates the real time use of
standardized screening instruments, such as the RAND-36, in
general medical care.
Implications for Policy, Delivery or Practice: Use of
technology, such as the FAST, can facilitate both patient care
and outcomes research.
• Evaluating the Potential Impact of Pharmacogenomics
on the Pharmaceutical Industry and Regulatory Policy
Amalia Issa, Ph.D.
Presented by: Amalia Issa, Ph.D., Health Services, UCLA
School of Public Health and RAND Division of Health
Sciences, 650 Charles Young Drive South, Box 951772, Los
Angeles, CA 90095-1772; E-mail: aissa1@ucla.edu
Research Objective: Few developments in health care have
greater impact on patients, providers, payers and health
systems than new pharmaceuticals. Pharmacogenomicsbased drugs have the potential to reduce adverse drug
reactions, optimize therapeutic efficacy, and streamline the
drug development process. Recently, the United States Food
and Drug Administration (FDA) issued draft guidance related
to the submission by industry to the FDA of
pharmacogenomics data. Few attempts in the literature
consider the possible scenarios that might confront regulatory
policy as pharmacogenomics-based drugs are increasingly
incorporated into the regulatory process. At the same time,
the pharmaceutical and biotechnology industries’ efforts to
market pharmacogenomic drugs will likely be significantly
impacted by the FDA’s current efforts and future regulatory
policy. The objectives of this paper are: (1) to analyze the key
drivers and barriers to the development and marketing of
pharmacogenomics-based drugs; and (2) to explore the
potential challenges and opportunities for the industry as a
result of developments in regulatory policy.
Study Design: The analysis is based upon a review of the
literature limited to the last 20 years and regulatory
documents as well as discussion with key stakeholders.
Databases searched included EMBASE and Medline and key
terms used included medical subject headings (MeSH) as well
as pharmacogenomics, regulatory policy, federal regulations,
drug development process, pharmaceutical industry and FDA.
Several vignettes are presented to illustrate the types of
possible issues and scenarios that might confront regulatory
policy makers and the pharmaceutical and biotechnology
industries in the next few years.
Principal Findings: A number of factors including medical
demand, access, cost-effectiveness, costs and complexity of
accompanying diagnostic tests, will likely drive the extent to
which various pharmacogenomics-based drugs become
incorporated into routine clinical practice. Current trends
indicate that the pharmaceutical and biotechnology industries
are concerned about the effects of future regulatory policies on
potential market segmentation and R & D.
Conclusions: Although pharmacogenomics promises
numerous benefits, there is still much that is unclear about
the use of pharmacogenomics in clinical practice ranging from
basic science and clinical applications through health
economics to social and policy effects.
Implications for Policy, Delivery or Practice: Despite the
advances made to date, pharmacogenomics remains laden
with scientific, technical and regulatory concerns.The key
challenge for the FDA and other policymakers is the
development of policy under the purview of uncertainty on a
number of different fronts. Nevertheless, it is important to
anticipate the challenges and issues in order for policy
development to keep pace with the science of
pharmacogenomics. Consideration of the key drivers and
barriers that will likely impact regulation of
pharmacogenomics-based drugs is important for future
research aimed at informing clinical and policy decisions and
understanding the implications of this relatively new
technology for the industry and eventually health systems.
Primary Funding Source: AHRQ
• Estimating the Breakeven Price for a New Technology:
Case Study of Detecting Organ Rejection in Cardiac
Transplantation Patients
Mahmud Khan, Ph.D., Jon Kobashigawa, M.D., Eric Marton,
M.D., Mandeep Mehra, M.D., Howard Eisen, M.D., Richard
Culbertson, Ph.D.
Presented by: Mahmud Khan, Ph.D., Professor, Health
Systems Management, Tulane University, 1440 Canal Street,
#1900, New Orleans, LA 70112; Tel: 504.584.1979; Fax:
504.584.3783; E-mail: khan@tulane.edu
Research Objective: To estimate the break-even market price
at which the policy makers will be indifferent between the
current practice (using biopsies) and the new molecular
testing for detecting acute rejection of organs after
transplantation. This breakeven price not only guides the
policy decisions, it can also tell the technology firms the
maximum price market will be willing to pay for the services.
Study Design: Standard production function approach has
been used to estimate the resource requirements for the
production of an endomyocardial biopsy. It is assumed that
the production function is a fixed coefficient Leontieff type.
Assuming that there is no misuse of inputs in the process, the
cost of producing one biopsy can be derived. The cost of
managing the side effects of each biopsy and quality of life
related costs can be added together to find the total cost of a
biopsy. The new technology, molecular testing, will reduce the
number of biopsies and given the number of tests needed in a
year, the breakeven cost of the new test can be derived. The
production function will also indicate the current utilization
pattern of inputs for conducting biopsies and how the
financial allocation among the inputs may change due to the
introduction of the new procedure.
Population Studied: The study is based on the data collected
from one large cardiac transplant facility in the USA. During
the years 2000-2002, the facility performed more than one
thousand transplantations including 152 heart transplants.
Individual patient level data including the insurance status of
all allograft recipients were collected. The use of real resources
for conducting biopsies were obtained from hospital records
for the post-transplantation year. The payments received by
the facility from various sources were also obtained to
examine the potential impact of the new technology on thirdparty payers and individual patients.
Principal Findings: Introduction of the new procedure should
reduce the number of biopsies from 13 to 6.5 during the first
year after transplantation. Complication rate of
endomyocardial biopsy during the first year after
transplantation is about 30% (25% minor and 5% major
complications). With the molecular test based monitoring, the
complication rates should reduce to about 16% (13% minor
and 3% major). 28% of all transplant patients of the center
during the study period was insured by Medicare and 57% by
region’s top six commercial health plans. The full-cost of
hospital services for biopsy was estimated at $4,628 and the
total cost of a procedure becomes $5,548 with Medicare-level
fee for physician and the pathologist. Using the cost of biopsy
only (without considering the cost of complications), the
breakeven price for the new molecular testing becomes $4,508
per assay. The average total reimbursement for biopsy from
private insurance agencies was $3,900 for this health center
and this reimbursement level implies a breakeven price of
$3,169. The breakeven price will be even higher if the cost of
complication is included. Since the reimbursement for a
biopsy is lower than the cost of producing it (excluding the
cost of physician’s and pathologist’s time), hospitals can save
about $11,738 per case by adopting this new mode of
monitoring acute rejection after transplantation. However,
payments for physicians and pathologists will decline by about
$3,899 and $3,490 per patient respectively due to lower
demand on their time inputs. Patients would also save an
average of $2000 in out-of-pocket expense during the first
year. The new procedure saved 450 hours (56 working days) of
physician time and 200 hours of laboratory time for the study
hospital.
Conclusions: The lower bound of the breakeven price for the
new non-invasive procedure is estimated at about $3,170
without considering the cost of managing the biopsy
complications. If the cost of producing the test plus the
normal profit is less than this amount, the implementation of
this new management protocol at a high-volume cardiac
transplant center will result in significant cost savings for
payers, providers and the patients. Sensitivity analyses
indicate that the cost savings hold true for wide variations in
the payer mix and reimbursement patterns. Another
advantage of the new protocol is the reduction in physician
time requirement and increased hospital capacity. Patients
get additional benefits in terms of improvements in quality of
life due to non-invasive nature of the new test.
Implications for Policy, Delivery or Practice: The cost of
producing a molecular test is likely to be less than the lower
bound of the breakeven price. Therefore, production of the
new test should be profitable in the market place and the
adoption of the technology will improve overall benefits in the
society.
Primary Funding Source: This research is not funded but
data were collected through private sector funding.
• The Effect of Clinical Trial Participation on Prescription
Drug Utilization
Meredith Kilgore, MSPH, Ph.D., Dana Goldman, Ph.D.
Presented by: Meredith Kilgore, MSPH, Ph.D., Assistant
Professor, Health Care Organization and Policy, UAB School
of Public Health, 1665 University Boulevard, RPHB 330,
Birmingham, AL 35294-0022; Tel: 205.975.8840; Fax:
205.934.3347; E-mail: mkilgore@uab.edu
Research Objective: Several studies have found that
participation in clinical trials is associated with a small but
significant increase in the costs of routine care for trial
participants. Most studies have omitted prescription drug
utilization and costs, which may be of particular importance to
individuals deciding whether to join research studies. Our
objective is to ascertain the effect of cancer clinical trial
participation on prescription drug utilization and costs.
Study Design: We used a retrospective case-control design to
compare prescription drug utilization and costs for patients
participating in cancer clinical trials to a matched cohort of
patients treated for similar cancer diagnoses. Matching was
made based on trial eligibility criteria and propensity scores
were used to mitigate differences between the trial
participants and non-participants. Information on drug
utilization was obtained from telephone surveys for cancer
patients identified by participating providers. To minimize
recall bias, respondents were asked to recall prescription
drugs used only in the six months prior to the interview and
were provided with drug reminder cards in advance of the
survey. Costs were estimated based on average observed
treatment costs from a large pharmacy transactions database.
Principal analyses were conducted using weighted least
squares and generalized linear models to assess the effects of
trial participation on total costs and out-of-pocket
expenditures for prescription drugs. Control variables included
insurance coverage, provider characteristics, patient
characteristics (e.g. age, gender, income, health locus of
control responses), cancer diagnosis and comorbid
conditions.
Population Studied: The sampling frame for selecting clinical
trial participants (cases) included a census of patients
participating in NCI-sponsored treatment trials between
October 1, 1998 and December 31, 1999. Non-participants
(controls) received treatment for the same cancers at the
same providers as trial participants.
Principal Findings: Using weighted least squares to adjust for
covariates, trial participation was associated with $131 in
additional costs for prescription drugs, or 47% increase in
costs compared with non-participants (p < 0.012). When a
generalized linear modeling approach was used to assess the
potential for differential costs to increase as baseline expected
costs rose, no such non-linear effect was detected. Although
total drug costs increased with trial participation, this did not
result in any statistically significant increase in out-of-pocket
drug expenditures for trial participants.
Conclusions: Participation in cancer clinical trials is
associated with an increase in total drug costs, but the
magnitude of the difference is not large in comparison overall
cancer treatment costs and the effect on out-of-pocket
expenditures for patients participating in trials is negligible.
Implications for Policy, Delivery or Practice: The additional
costs for prescription drugs associated with trial participants
is statistically significant but not so large as to pose an
economic burden in excess of the value of the information
derived from rigorous research on new therapies. These
findings hold from the perspectives of both health systems
and patients.
Primary Funding Source: NCI, National Science Foundation
• The Determinants of CT and MRI Utilization in Taiwan
Pei-Tseng Kung, Sc.D., Wen-Chen Tsai, Dr.PH, Chih-Liang
Yaung, Ph.D.
Presented by: Pei-Tseng Kung, Sc.D., Assistant Professor,
Health Administration, Taichung Healthcare and
Management University, No. 11 Ln., 16 Sec. 3 Chungching
Road, Taya, Taichung, 42805; Tel: 886.4.25603149; Fax:
886.4.25603149; E-mail: ptkung@seed.net.tw
Research Objective: Magnetic resonance imaging (MRI) and
the computerized tomography scanner (CT) are high-tech
equipment with the highest utilization and expenditures in
Taiwan. This study investigates the factors that affect MRI and
CT utilization, such as whether supply-side factors have more
impact than demand-side factors, and the effect of utilization
control policies.
Study Design: This study used the nationwide CT and MRI
claim data for the years 1998 to 2001 from the Bureau of
National Health Insurance (BNHI) in Taiwan. Multiple
regression analysis was the statistical method used to
determine the relative factors that influence CT and MRI
utilization.
Population Studied: This study used Taiwan's NHI
nationwide medical claim data, which comprised the medical
care services for all insured patients or 97% of Taiwan's
population (approximately 23 million people), for the years
1998 to 2001, to determine CT and MRI utilization. We also
used the Health Care Regions (HCRs) as the observation unit
for further analyses. The HCRs are the living areas mapped
out by the DOH in which the health care needs of residents
can be conveniently met. This national health care net
encompasses all 21 cities and counties and consists of 17
HCRs, with each HCR containing 1 or 2 cities or counties in
Taiwan, much like the Metropolitan Statistical Area (MSA) in
the United States.
Principal Findings: CT-population ratio, hospital-based
physician-population ratio, female ratio, pediatric (less than 14
years) population ratio, and family income significantly
influenced CT utilization. MRI-population ratio and hospitalbased physician-population ratio, female ratio, and family
income significantly influenced MRI utilization. However, the
proportion of the aged (greater than 65 years) did not
significantly influence CT or MRI utilization after controlling
other factors. The rates of CT and MRI utilization for the
outpatient and inpatient settings and the repeated uses of CT
or MRI have decreased significantly after a utilization review
was implemented by BNHI, which implies that CT and MRI
have been overused.
Conclusions: Numeral changes in supply-side factors
significantly affect CT/MRI utilization, and supply-side factors
impact on CT/MRI utilization more than demand-side factors.
The results show that the utilization instant report policy
effectively decreases repeated uses of CT/MRI.
Implications for Policy, Delivery or Practice: Proposed the
suitable method for the control and management of high-tech
medical utilization.
Primary Funding Source: Taiwan Department of Health
• AskCHIS: Evaluating the Effectiveness of an Online
Query System for Health Survey Results
Jeff Luck, M.B.A., Ph.D., Eric Mindel, E. Richard Brown, Ph.D.,
Wei Yen, Ph.D., Rong Huang, M.S., Charles DiSogra, Ph.D.,
M.P.H.
Presented by: Jeff Luck, M.B.A., Ph.D., Associate Professor,
Department of Health Services, UCLA School of Public Health,
Box 951772, Los Angeles, CA 90095; Tel: 310.206.4744; Fax:
310.206.4722; E-mail: jluck@ucla.edu
Research Objective: To assess user experience with AskCHIS,
an online query system that provides interactive access to
health survey data, and to identify enhancements that could
improve its usability and functionality and expand the user
audience.
Study Design: An online survey was developed to: assess the
usability of AskCHIS; understand the determinants of regular
or infrequent use; and identify and prioritize future system
enhancements. An initial e-mail solicitation in November
2003 was followed by a reminder 1 week later. Data from
system registration and usage tracking were also analyzed.
Population Studied: AskCHIS is a web-based information
system providing interactive, real-time access to results of the
2001 California Health Interview Survey (CHIS), a telephone
survey of 55,000 California households. AskCHIS users can
select any combination of variables, and automated
algorithms calculate accurate estimates and standard errors
from the complex stratified survey design. Results are
provided in tables or graphs.
AskCHIS is free of charge and publicly available at
www.chis.ucla.edu. Registration is required to access to the
system. All 2286 registered users were solicited to complete
the online questionnaire. The 24% of registered users who
had not successfully completed any queries received a
separate questionnaire.
Principal Findings: AskCHIS attracted an average of 151 new
users per month during 2003, with users most often learning
about the system through word of mouth. Registered users
have logged in to AskCHIS an average of 6.7 times,
completing an average of 21.1 queries each.
The 593 user survey responses received constitute a response
rate of 29.2% of registered AskCHIS users with valid e-mail
addresses. Respondents and nonrespondents had similar
demographic characteristics, although respondents tended to
be more regular system users.
The majority of respondents work in universities or research
organizations (31.7%) or government agencies (28.2%). On
average, they report “significant experience” in data analysis.
Respondents indicate that AskCHIS is easy to use (2.2/5, i.e.,
average rating of 2.2 on a 5-point Likert scale where 1 is “very
easy”), and that the results they obtain from it are useful
(1.8/5). Sixty-two percent of respondents have used other
online query systems, and find AskCHIS somewhat easier to
use (2.5/5). The most commonly reported uses of AskCHIS
results were research, program design/evaluation, and
grant/proposal writing. Respondents expressed clear
preferences about desired improvements: easier specification
of the query population was most desired, followed by 3-way
tables and variable recoding. The feature most desired by
zero-query users was more or different variables.
Conclusions: AskCHIS has become widely used in its first full
year of operation, even without dedicated marketing or
advertising. Users return to use the system regularly, and
open-ended survey responses indicate that many find the
results highly useful. However, most current users are fairly
sophisticated in data analysis, and so usability enhancements
such as a simplified query building interface could expand the
user audience.
Implications for Policy, Delivery or Practice: Online query
systems can make results of complex health surveys available
to a large audience, who can use the data to plan programs
and influence policy. Usability enhancements that attract
more community users could further democratize access to
health data.
• Development and Evaluation of an Integrated Clinical
Data System for Continuing Care Retirement Communities
Sue Nonemaker, R.N., M.S., John Morris, Ph.D.
Presented by: Sue Nonemaker, R.N., M.S., Senior Research
Associate, Research and Training Institute, Hebrew
Rehabilitation Center for Aged, 1200 Centre Street, Boston,
MA 02131; Tel: 717.235.5509; Fax: 617.363.8936; E-mail:
nonemaker@mail.hrca.harvard.edu
Research Objective: 1. To develop and implement a
standardized clinical data system across the various levels of
care in CCRCs.
2. To establish a National CCRC Data Consortium using
standardized individual-level data to: 1)improve individual
resident's health and well-being; 2) support benchmarking and
quality improvement efforts; 3) inform health planning; and 4)
facilitate research on the impact of health promotion
interventions on residents' function, health status, and wellbeing.
Study Design: Descriptive
Population Studied: Residents of Continuing Care
Retirement Communities (CCRCs) and other supportive
housing environments
Principal Findings: This session describes the development
and testing of a standardized clinical data system to support a
variety of service and programmatic needs within Continuing
Care Retirement Communities (CCRCs). Standardized data
includes measures of functional and cognitive status, medical
and other health conditions, psychosocial and emotional wellbeing, service use, medication use, and sociodemographic
data. The standardized data system has already been
implemented in approximately 50 CCRCs, and is expected to
be used by 200 communities by 2005. A National CCRC Data
Consortium has been established to provide CCRCs with data
to inform service and programmatic needs, as well as the
capacity to engage in research.
Conclusions: A reliable and valid data system has been
developed and is currently being implemented in a
geographically and socioeconomically diverse sample of
CCRCs. The establishment of a National CCRC Data
Consortium will provide CCRCs with the capacity to engage in
applied research.
Implications for Policy, Delivery or Practice: This initiative
provides an opportunity to implement an electronic clinical
record using common assessment items across the various
levels of care found in Continuing Care Retirement
Communities. It also provides the infrastructure to conduct
applied research in CCRC practice settings, and will serve as a
laboratory for studying the population that most closely
resembles the soon-to-be retiring Baby Boomers.
Primary Funding Source: CMS, private foundation
• Detection of Pancreatic Malignancy with Positron
Emission Tomography: A Meta-Analysis
Lori Orlando, M.D., David Matchar, M.D., FACP
Presented by: Lori Orlando, M.D., Internal Medicine Fellow in
Health Services Research, Duke University and the Durham
VA, Box 152, Building 6, 508 Fulton Street, Durham, NC 27705;
Tel: 919.970.7140; Fax: 919.416.5839;
E-mail: lorlando@mebtel.net
Research Objective: Pancreatic cancer, which has a 5-year
survival rate of only 3%, is the fourth most common cause of
cancer deaths in the U.S. Several factors contribute to this
high mortality rate, including limitations of current diagnostic
imaging modalities. Some investigators have conducted small
studies comparing the use of positron emission tomography
with a fluoro-deoxy-glucose radiotracer, FDG-PET, to
conventional imaging with computed tomography, CT, for the
detection of pancreatic malignancy. We performed a metaanalysis of these studies in order to better understand and
refine the operating characteristics for FDG-PET.
Study Design: Articles were identified through a MEDLINE
search using the keywords: tomography, emission-computed,
positron emission tomography, fdg-pet, and pancreatic
neoplasms, as well as through a bibliography review. Studies
were included if they were published in an English peer
reviewed journal and reported primary data on at least 12
human subjects. Of the 66 abstracts initially identified, 55
underwent full text review and 21 met the inclusion criteria. Of
these 21, two were excluded as duplicate populations. We
conducted the meta-analysis using the diagnostic testing
meta-analytic software, Meta-Test 0.6. To detect publication
bias and heterogeneity we used a summary measure, the logodds ratio, to create a Galbraith plot and perform a chi square
test of homogeneity. Sensitivity analyses were performed
based upon predefined criteria.
Population Studied: Although many of the studies did not
provide characteristics of their patient population, patients
were referred for FDG-PET scanning for one of the following
three reasons: referral to a surgical center for pancreatic
resection, clinical signs or symptoms concerning for
pancreatic disease, or an abnormal conventional imaging
study.
Principal Findings: The pooled sensitivity and specificity for
FDG-PET were 88% with a 95% CI 84%-90%, and 79% with a
95% CI 72%-85%, respectively; while for conventional imaging
the pooled sensitivity was 79% with a 95% CI 70%-86%, and
specificity was 63% with a 95% CI 51%-74%. Summary ROC
curves were calculated and the area under the curve for FDGPET was 0.94 and for CT was 0.79. The studies were
homogenous and no publication bias was detected. During
sensitivity analyses the pooled test characteristics were similar
to the overall pooled sensitivity and specificity except for two
sub-groups: those with a quality score less than 4 and those
who were referred for FDG-PET regardless of the results on
conventional imaging. These two sub-groups both had higher
pooled estimates than the overall average, but were composed
of only 3 studies each.
Conclusions: Our results suggest that the diagnosis of
pancreatic malignancy may be enhanced by adding FDG-PET
to the diagnostic work-up, although the overlap in confidence
intervals prevents a clear conclusion from being drawn. In our
analysis no one sub-group benefited more from FDG-PET
than another; however we were limited by a lack of
information on the performance of FDG-PET with regards to
lesion size and false negative CTs.
Implications for Policy, Delivery or Practice: Research in
this area has some clear deficiencies, including a failure to
report lesion size, clearly define patient populations, and
recruiting methods. New research should focus upon
identifying the operating characteristics of FDG-PET for
patients with false negative CTs, lesions < 2cm, and low
probabilities of malignancy. Once these questions have been
addressed the role for FDG-PET in detection of pancreatic
malignancy will become much clearer.
Primary Funding Source: AHRQ
• Why Is It Challenging to Regulate New Technologies?
The Example of Regulation of Pharmacogenomics by the
United States Food and Drug Administration
Kathryn Phillips, Ph.D., Stephanie Van Bebber, M.Sc.., David
Veenstra, PharmD, Ph.D.
Presented by: Kathryn Phillips, Ph.D., Associate Professor of
Health Economics and Health Services Research, School of
Pharmacy and Institue for Health Policy Studies, University of
California San Francisco, 3333 California Street, Suite 420, Box
0613, San Francisco, CA 94143; Tel: 415.502.8271; Fax: (415)
502-0792; E-mail: kathryn@itsa.ucsf.edu
Research Objective: New technologies offer the promise of
better health, but may contribute to increased health care
costs. New technologies also require reconsideration of
appropriate regulations in order to maximize benefits while
minimizing risks, to provide appropriate incentives, and to
protect the public’s health.
The mapping of the Human Genome portends rapid
development of new technologies – and more regulatory
challenges. One of the most imminent advances emanating
from genomics is the individualization of drug therapy based
on genetics (pharmacogenomics). Promises of
pharmacogenomics are fewer adverse drug reactions, more
effective pharmaceuticals, and lower drug development and
health care costs. Yet, the advent of pharmacogenomics
presents challenges to regulators, in particular the FDA who
must balance its mandate of protecting the health and safety
of the public while creating incentives to realize the promises
of pharmacogenomics.
Our objectives are to: (1) identify key pharmacogenomicsrelated issues faced by the FDA and (2) discuss how these
issues illustrate broader challenges of regulating new
technologies.
Study Design: We obtained data from three sources: (1) our
work as FDA advisors (2) our work consulting with biotech
companies on the development of new PGx technologies, and
(3) literature review (based on 74 articles and reports).
Principal Findings: We identified three primary challenges:
1. “You can’t regulate what you don’t understand, but you
can’t understand it until you regulate it”.
Regulatory agencies inexperienced with new genomic
technologies (e.g. microarrays) must still work quickly to
evaluate and regulate them. Thus, the FDA is developing
approaches to encourage voluntary submission of
pharmacogenomic data so that the agency can gain
experience evaluating it. In addition, new technologies often
require new approaches to regulation. In the case of
pharmacogenomics, the regulation of combination
drug/diagnostic products presents a challenge since
historically the industry and the FDA have been organized
around either drugs or diagnostics.
2. “Who will evaluate new technologies?”
The US does not have a cohesive or comprehensive approach
to evaluation of new technologies. No single agency evaluates
technologies (e.g. the now disbanded Office of Technology
Assessment). Since the FDA, unlike in other countries, does
not consider costs in drug and diagnostics approval, the value
(cost-effectiveness) of new pharmacogenomic-based
interventions and whether they will be reimbursed are
determined by a hodgepodge of agencies (e.g. CMS) and
private insurers.
3. “Who will pay and who will gain from new technologies and
their regulation?”
New technologies often have unknown costs and benefits.
This is particularly true for pharmacogenomics; for example,
we found only eight published cost-effectiveness analyses of
pharmacogenomics. Although one reason that few
pharmacogenomics interventions have been evaluated is that
clinical use is currently limited, new technologies must be
evaluated before they become widely disseminated if society is
to maximize their benefits and costs. The costs and benefits of
the regulations themselves also have to be considered in order
to provide appropriate incentives for industry.
Conclusions: This review identifies issues faced by the FDA
with respect to pharmacogenomics, which the FDA has begun
to address through a variety of initiatives.
Implications for Policy, Delivery or Practice: The evaluation
of new technologies often requires new perspectives,
resources, and expertise. However, there are numerous
challenges in evaluating new technologies, which are amply
illustrated by the example of pharmacogenomics. Health
services researchers have an important role in ensuring that
new technologies are appropriately evaluated.
Primary Funding Source: NCI, United States Food and Drug
Administration
• An Empirical Analysis of Technology Adoption among
Florida Hospitals
Edward Schumacher, Ph.D., Gary Fournier, Ph.D.
Presented by: Edward Schumacher, Ph.D., Associate
Professor, Health Care Administration, Trinity University, 715
Stadium Drive, San Antonio, TX 78212; Tel: 210.999.8137; Fax:
210.999.8108; E-mail: eschumac@trinity.edu
Research Objective: This paper focuses on the effects of
technology adoption on hospital revenues. We examine the
hypothesis that hospitals experience a prestige effect as a
result of their choice of technology set. This effect reflects the
reward or penalty to the hospital according to the diversity of
specialized services it provides relative to local rivals. If
patients use the overall array of technology available at a
hospital as a signal of the quality of a hospital, or if physicians
have a preference for higher quality facilities and can direct the
choice of the patient, then the prestige effect takes the form of
a positive demand shift in the hospital’s demand function.
Thus, hospitals may have an incentive to invest in
technologies that are not financially sustainable in and of
themselves, but generate sufficiently large demand spillovers
to make the investments worthwhile.
This paper searches for evidence of this type of behavior
Study Design: We first examine specific episodes of entry into
specialized service markets and show that the revenues of the
entrant do not equal the pre-entry revenues of the incumbent
for many years after the entry. We then turn to a composite
picture of the overall technology. For each hospital we
construct an aggregate technology index following Baker and
Spetz (1999). Then, for hospitals in the same county we relate
each individual hospital’s revenues as well as the average rival
hospital’s revenues to a relative measure of technology.
Because both equations may be affected by the same
unobserved variables, the method of Seemingly Unrelated
Regressions (SUR) is used to allow cross-equation covariance
in the error terms in the pair of equations. The aggregate
effects sought are indicative of "prestige" and represent
aggregate spillovers owing to technology-induced goodwill for
the hospital, relative to other locally available choices.
Population Studied: The data used in the study are from
Worksheet A of the Annual Uniform Financial Report for the
years 1985 to 2000. This report is required by the Florida
Agency for Health Care Administration to be filled out for all
hospitals in the state of Florida. The data contain information
on hospital type and location, input variables such as
employment, as well as costs, quantity, and revenue generated
from specific services and specialties.
Principal Findings: Our findings show that when a hospital
has a relative technology advantage over its rivals, overall
revenues are higher for that hospital, and the revenues of rival
hospitals are diminished. These findings are consistent with
prestige effects and suggest there are aggregate spillovers
owing to technology-induced goodwill for the hospital.
Conclusions: Our findings suggest that a hospital’s decision
to adopted specialized technologies is not based solely on the
financial viability of that particular service. It appears that
hospitals experience aggregate spillovers owing to technologyinduced goodwill for the hospital relative to other locally
available choices. Hospitals that adopt new technologies
ahead of their rivals receive a relative demand shift due to the
prestige associated with the new technology. Thus services
that may not make a profit as an independent service may
make sense to a hospital if this demand shift is large enough
to make up for the losses.
Implications for Policy, Delivery or Practice: These results
have implications for understanding the role of technology in
hospital competition. They suggest that merger guidelines
and certificate of need legislation may play a role in
maintaining an efficient level of technology in hospitals.
• Integration of New Technology: Lessons From
Cholecystectomy Care Patterns
Elmer L. Washingtion, M.D., M.P.H., Jay Shen, Ph.D.
Presented by: Jay Shen, Ph.D., University Professor, Health
Administration, Governors State University, One University
Parkway, University Park, IL 60466; Tel: 708.534.3144; Fax:
708.235.2197; E-mail: j-shen@govst.edu
Research Objective: The Institute of Medicine report,
“Crossing the Quality Chasm: A New Health System for the
21st Century” identifies integration of new technology into
routine medical practice as one of the many challenges facing
health care professionals. This study analyzes patterns of care
with respect to laparoscopic versus open cholecystectomy to
assess the timeliness and appropriateness of integration of
this new technology into routine health care delivery.
Study Design: We compared trends over time of utilization of
laparoscopic cholecystectomy as a proportion of all
cholecystectomies for 1995 and 2000. We also analyzed
trends of length of stay, cost, and mortality.
Population Studied: Using the National Inpatient Sample, we
identified 70,020 and 67,222 discharges in 1995 and 2000,
respectively.
Principal Findings: Although our data demonstrated that
laparoscopic cholecystectomy has increasingly become the
favored procedure with percentages of the laparoscopic
version increasing from 67.6% of all cholecystectomies in
1995 to 73.5% in 2000, we found this percentage to still be
lower than what would be expected based upon vastly
favorable outcome measures. While several outcome
measures showed some improvement for both procedures
over time, these measures showed consistently superior
results for laparoscopic cholecystectomy than for open
cholecystectomy during each year studied. The average length
of stay was 3.8 days for laparoscopic procedures in 1995
compared to 10.4 days for open procedures in 1995. In 2000,
the average length of stay for laparoscopic procedures was 4.0
days compared to 9.5 days for open procedures. Rates of
infection in 1995 were 9.5% for open procedures and 3.3% for
laparoscopic procedures. In 2000, rates of infection were
9.0% for open procedures and 2.9% for laparoscopic
procedures. Death rates decreased from 4.3% to 3.9% in
2000 for open procedures while showing no statistically
significant change for laparoscopic procedures (0.4% in 1995
and 0.6% in 2000).
Conclusions: Based upon the evidence showing vastly
superior outcomes associated with laparoscopic
cholecystectomy as compared to open cholecystectomy, the
open procedure should be relegated to circumstances
requiring conversion after a failed attempt at the laparoscopic
procedure and to situations where the laparoscopic procedure
is clinically contraindicated.
Implications for Policy, Delivery or Practice: Public policy
should ensure that providers of care are responsible for timely
integration of new technology when such technology has a
demonstrably superior track record of performance. Further
research should be conducted to determine an appropriate
percentage of all cholecystectomies that should be performed
laparoscopically as a benchmark. Barriers to adopting new
technology should be identified and removed including
financial disincentives and unnecessarily burdensome
approval processes.
Primary Funding Source: AHRQ
• Information Technology in the Chiropractic Office
Monica Smith, DC, Ph.D.
Presented by: Monica Smith, DC, Ph.D., Associate Professor,
Palmer Center for Chiropractic Research, 741 Brady Street,
Davenport, IA 52803; Tel: 563.884.5173; Fax: 563.884.5227;
E-mail: monica.smith@palmer.edu
Research Objective: Chiropractors were surveyed to assess
their use of PC and web-based technologies to manage the
business and clinical aspects of chiropractic practice
Study Design: Pilot survey of 105 “key informants” of the
chiropractic profession, 45% response rate. Followed by
survey of 3 separate samples of chiropractors: simple random
samples of U.S. and Pennsylvania chiropractors, and sample
of Kentucky DCs stratified by health workforce shortage areas.
Response rates 20-30%.
Survey items queried practice’s market service area; personal
and practice demographics; computer technology and internet
use within the practice; and factors associated with clinical
decision-making such as use of various information resources.
Population Studied: Samples of U.S. chiropractors: National
N=66,790, n=400; Pennsylvania N=3,390, n=250; Kentucky
N=636, n=150.
Principal Findings: Results from all surveys indicate
moderate to high use of electronic technologies for business
aspects of practice. Use of automation in clinical applications,
or of automated systems that link between billing and clinical
charts, was far lower. Clinical decision making primarily based
on textbook and peer reference, and majority of respondents
have never used clinical support software packages and only
sometimes use internet resources to assist in clinical decisionmaking. While less than 40% of DCs use electronic claims
processing, those that do automate this function, do so for
most of their claims. Overall, most computer systems are
personally selected by the practicing DC (78%) rather than a
practice management group (14%), and most respondents
plan to increase automation within the next 2 years.
Most chiropractic practices serve health workforce shortage
areas.
Conclusions: Converging imperatives increase pressure for
chiropractors to incorporate more electronic information
technologies use into clinical and business aspects of practice.
HIPAA rules will eventually include mandates for small service
providers such as chiropractors. Initiatives to improve quality
of care encourage providers such as chiropractors to use
current evidence-based best practice standards, increasingly
accessible via internet or electronic formats. Current policy
mandates, e.g. DVA Health Care Programs Enhancement Act
of 2001, accelerate inclusion of chiropractors into DVA and
military medical centers and clinics. Managed care
organizations, hospital & ambulatory clinic systems, and
multi-disciplinary physician practices increasingly use clinical
information systems, and successful integration of
chiropractic providers into extant health care systems requires
that today’s DC be information literate and computer savvy.
Clinical informatics curricula in chiropractic education could
better prepare chiropractic students for health system
integration, and could further support other initiatives in
chiropractic education such as evidence-based care and
student-directed learning.
As an important part of the healthcare workforce in rural and
underserved areas, The Health Care Safety Net Improvement
Act now makes chiropractors eligible to participate in the
National Health Service Corps program. This further
necessitates that chiropractic educational and research
institutions reach out to assist and improve chiropractic
health care delivery in rural and underserved areas, through
both postgraduate and undergraduate efforts.
Implications for Policy, Delivery or Practice: Federal and
state initiatives advancing the development of telehealth
technologies to assist healthcare providers in rural and
underserved areas should also consider the integral role of
nonmedical providers such as chiropractors in total healthcare
delivery.
• A Web-Services Framework For Health Care
Emile Soueid, B.S., M.S., Ph.D., ECE
Presented by: Emile Soueid, B.S., M.S., Ph.D., ECE, President,
WirelessTies LC, 2502 Montclaire Circle, Weston, FL 33327;
Tel: 954.294.8405; E-mail: soueid@wirelessties.com
Research Objective: This presentation examines the role of
web services technologies in providing a basis for a costeffective, secure and reliable end-to-end Internet health care
infrastructure, and in establishing a framework for
implementing uniform security and privacy solutions for
health care nationwide.
Starting with the HIPAA Administrative Simplification
requirements (AS) as a guide and primary reference, this
presentation addresses the following:
•
The distinct advantages of web services (a) based on
their technical merit; (b) considering the favorable market
drivers in health care and other verticals; (c) the virtual
ubiquity of standards-based products.
•
The “ongoing” challenges to the effective adoption of
web services in the health care community, such as the need
for further standardization of vocabularies, tools and business
process definitions aimed at improving interoperability within
that community.
•
The early adopters of web services in the health care
market, market trends, and examples of health care
establishments that have successfully implemented web
services.
• Strategies and Barriers in Sharing Knowledge: Caring for
Clients through Methods of Best Practice
Dorian E. Traube, CSW, Dorian Traube, CSW, Jennifer Vick
Bellamy, LCSW, Sarah Elizabeth Bledsoe, M.S.W.
Presented by: Dorian Traube, CSW, Student, Columbia
University School of Social Work, 506 W. 122nd Street #64,
New York, NY 10027; Tel: 212.678.6234; E-mail:
det17@columbia.edu
Research Objective: While there is a call for Evidence Based
Practice (EBP) in Social Work and mental health services,
there has been a range of criticisms about the process of
translation and implementation of research findings into
viable methods of practice. These barriers range from an
egregiously long amount of time between research
development to implementation of cutting edge practices to a
veritable lack of support and training for community
practitioners. There is a growing body of evidence on
efficacious programs, but there is not a substantial body of
work addressing the general adaptation of these lessons for
use in the field of mental health services.
Study Design: Presenter will outline methods for this
comprehensive review of published literature, gray literature,
and qualitative findings. A sample of current proposed
models of conceptualization and implementation of EBP will
be compared and contrasted. This presentation will also
include a review of the current barriers from the federal, state,
agency, practitioner, and client perspectives. Strategies at the
micro, mezzo, and macro levels improving implementation as
described by experts in the field will also be outlined.
Population Studied: Experts in the Field of Evidence Based
Practice were surveyed as well as a comprehensive review of
the current EBP literature.
Principal Findings: While there is an abundance of
information about the barriers to implementing EBP, little has
been developed in terms of strategies for implementing EBP
in an efficacious manner.
Conclusions: The authors identify the current barriers in
implementing EBP and offer possible solutions to these
barriers. Some of the current efforts toward the
implementation of EBP will be discussed to present an
overview of the activities in the field as well as a resource
guide for practitioners and service providers.
Implications for Policy, Delivery or Practice: The authors
claim that without the common use of EBP, practitioners may
harm patients. In addition, in the future, insurance
companies may not cover treatments that are not
substantiated with an efficacious base. Presenters will make
policy and practice recommendations
Primary Funding Source: NIMH
• Quality Transformation: Implementation of a
Computerized Order Set
Ann McAlearney, Sc.D., Sofia Veneris, M.H.A.
Presented by: Sofia Veneris, M.H.A., Administrative Fellow,
Columbus Children's Hospital, Inc., 700 Children's Drive,
Columbus, OH 43210-1234; Tel: 614.438.6869; Fax:
614.293.7710; E-mail: VenerisS@chi.osu.edu
Research Objective: Creating an environment focused on
quality requires a transformation of the way in which
healthcare is delivered. Columbus Children’s Hospital, Inc.-CCHI--has attempted this transformation by implementing a
Computerized Physician Order Entry--CPOE--System that
includes disease-specific order sets. The research objective of
this project was to assess order set use and to identify factors
impacting the use of three order sets: asthma, communityacquired pneumonia--CAP, and post-appendectomy.
Study Design: Eligible patients for each order set were
identified from the CCHI Decision Support System. These
patients were matched with data from the Eclypsis CPOE
system to determine order set use. Order set use rates by
month were calculated for each order set and significance of
trends was tested using linear regression. Potential
predictors of order set use studied included: stage of
implementation, patient age, patient race, payor, admit day,
admit time, emergency department--ED--admission, and
length of stay. Relationships were tested using chi-squared
analysis and multivariate logistic regression.
Population Studied: The study population includes patients
admitted to CCHI between November 1, 2001 and November
30, 2003. The study included 529 asthma patients defined by
a primary ICD-9 diagnosis code of 493.xx, excluding those
admitted to the pediatric intensive care unit, 277
appendectomy patients defined by ICD-9 procedure codes of
47.09 or 47.01, and 210 CAP patients defined by a primary
ICD-9 diagnostic code of 486.xx. Patients with cystic fibrosis
were excluded from analysis because a CF-specific pneumonia
order set is in place.
Principal Findings: No single pattern is seen in the uptake of
this new technology. The asthma order set shows a linear
trend of increasing use after implementation, but no such
trend is seen in appendectomy or CAP. Order set use levels
are also inconsistent. Six months after implementation, the
asthma order set was used in 90 percent of admissions and
the post-appendectomy order sets were used in 70 percent of
admissions. Use of the CAP order set, however, did not reach
20 percent. Variables associated with order set use varied by
set. Use of the asthma order set was associated with weekend
admission, ED admission, older age and later stage of order
set implementation. Use of the post-appendectomy order set
was associated with ED admission and with longer stays.
None of the studied predictors impacted use of the CAP order
set.
Conclusions: There is no single pattern of uptake of order
sets in the early implementation period. Predictors of use also
vary by set.
Implications for Policy, Delivery or Practice: The goal of
order set implementation is to reduce unnecessary practice
variation while promoting best practices. This goal is best
met when order sets are used consistently within their target
populations. This evaluation shows that uptake trends in
order set use vary by set as do predictors of use. Healthcare
systems looking to implement computerized order sets must,
therefore, consider the specific factors that may impact the
use of each set rather than relying on a one-size-fits-all roll-out
strategy.
• AntiTachycardia Pacing and Cardiac Events in the
MADIT-II Trial
Wai Shun Wong, M.D., Arthur Moss, M.D., Robert Goldberg,
Ph.D., N.A. Mark Estes, M.D., Robin Ruthazer, M.S., Ira
Wilson, M.D., M.Sc.
Presented by: Wai Shun Wong, M.D., Clinical Care Research
Fellow, Medicine/Division of Clinical Care Research, TuftsNew England Medical Center, 750 Washington Street, Box
#63, Boston, MA 02111; Tel: 617.636.8894; Fax: 617.636.0525;
E-mail: wwong@tufts-nemc.org
Research Objective: Previous studies have suggested that
defibrillation may cause myocardial damage and cardiac
desynchronization in patients implanted with implantable
cardiac defibrillators (ICDs), potentially contributing to new or
worsened heart failure. Antitachycardia pacing (ATP) is a new
technology used to terminate ventricular tachycardia by
overdrive pacing. The aim of this study was to determine
whether the presence of ATP improved clinical outcomes by
reducing the incidence of defibrillations.
Study Design: Using the Multicenter Automatic Defibrillator
Implantation Trial II (MADIT-II) database, we evaluated
patients who had prior myocardial infarction with ejection
fraction = 30% and received an ICD. The ATP function was
either turned on (ATP-on) or off (ATP-off) at baseline in trial
participants at the physicians’ discretion. Our primary
composite study outcome was time to death or first
hospitalization for congestive heart failure. We used a Cox
proportional hazard model stratified by the different study
centers to examine the relationship between ATP status and
the composite outcome. Episodes of ventricular tachycardia
terminated with ATP or defibrillation therapy were analyzed in
the two study groups.
Population Studied: We studied patients who enrolled in the
Multicenter Automatic Defibrillator Implantation Trial II
(MADIT-II). 76 hospital centers participated in the trial, 71 in
the United States and 5 in Europe.
Principal Findings: Of the 720 patients who received ICDs,
425 (59%) patients had ATP-on. Compared with patients with
ATP-off, patients with ATP-on were slightly older (65 vs. 64
years), had more often received revascularization procedures
(53% vs. 40% Coronary Artery Bypass Grafting and 63% vs.
55% Percutaneous Coronary Intervention), were more often on
amiodarone (9% vs. 5%), more often had dual-chambered
devices (55% vs. 36%), had a higher prevalence of left-bundlebranch block (24% vs. 16%), and had longer QRS intervals
(13ms vs. 12ms) (all p<0.05). During an average follow-up of
20 months, 28% of patients in the ATP-on group developed
the composite study outcome compared to 31% of patients in
the ATP-off group. After adjustment for differences in baseline
characteristics and other potential confounders and after
stratifying by study centers, the composite outcome in
patients with ATP-on vs. ATP-off was similar (HR: 0.95, 95%
CI 0.61 – 1.46, p=0.80). Analyses that used time to death or
time to first hospitalization for congestive heart failure as
separate outcomes revealed similar findings. Of the 286
episodes of ventricular tachycardia in patients with ATP-on
during the follow-up period, 69% were successfully terminated
with ATP and 31% required defibrillation therapy. In patients
with ATP-off, there were 217 episodes of ventricular
tachycardia, all successfully terminated with defibrillation
therapy.
Conclusions: In this analysis, ATP-on was not associated with
improved outcomes. However, ATP therapy terminated a
majority of ventricular tachycardia episodes that would have
potentially ended with defibrillation therapy, thereby
supporting the importance of this therapeutic modality in
reducing the frequency of defibrillator shocks in patients
implanted with ICDs.
Implications for Policy, Delivery or Practice: This may
promote the introduction of ICDs which solely have the
defibrilation therapy. They are as effective as ICDs that can
deliver both ATP and defibrillation therapy in saving lives from
malignant heart arrhythmias but at a significantly reduced cost
to society. When we realize how many patients have such
devices implanted, the implications of a less-expensive device
are tremendous.
Primary Funding Source: AHRQ