3rd Annual Population Health Sciences Poster
Session
Poster Abstracts
Student posters that are eligible for the Kit Allen Memorial
Scholarship are denoted with an ►.
Poster #01: The Appropriateness of Assuming a Lognormal Distribution when Modeling Health Care Costs
Poster #02: Chlamydia trachomatis and Neisseria gonorrhoeae infections among female and male Wisconsin Family Planning Clinic clients: prevalence of infection and performance of selective screening criteria
Poster #03: Students Teaching and Reaching Students: Planting Seeds for Careers in Public Health
►Poster #04: Use Patterns of Internet-Based Systems for Smoking Cessation and Smoking Outcome
Poster #05: Wisconsin Community Health Report Cards—2003 Edition
Poster #06: Relation between Sleep Apnea and Metabolic Syndrome in a Community-Based Sample
Poster #07: Oral Contraceptive Use, Reproductive Factors and Colorectal Cancer Risk: Findings from Wisconsin
►Poster #08: Self-Reported Habitual Snoring Reliability in the Wisconsin Sleep Cohort Study
Poster #09: Validation of an Environmental Tobacco Smoke Exposure Questionnaire
Poster #10: Vaccination levels in older adults in Beaver Dam, WI
►Poster #11: The Success of Collaboration: Wisconsin’s Diabetes Quality Improvement Project, 1999-2001
Poster #12: Relationship Between Dietary Patterns and Age-Related Maculopathy in the Carotenoids in Age-Related Eye Disease Study (CAREDS), an ancillary study of the Women's Health Initiative
Poster #13: New Mothers Survey, July 2003
►Poster #14: The Newborn Lung Project Statewide Cohort Study
Poster #15: Exercise Induced Wheezing and Lung Function in Very Low Birth Weight Children at ages 8 – 10 with the Newborn Lung Project Regional Cohort
►Poster #16: Factors responsible for observed changes in Chlamydia trachomatis test positivity
►Poster #17: Psychological traits and preference for control over health care decisions
►Poster #18: Cancer care capacity and out-of-area care utilization among urban and rural women with breast cancer in Wisconsin
►Poster #19: Wrong site surgery: the use of human factors system analysis and process analysis in the outpatient setting
Poster #20: Geographic and Age-Specific Variation of Pedestrian Injuries in New York City: Implications for Prevention
Poster #21: A Comparison of Epidemiological Indices of Association for Setting MCAT and GPA Thresholds in Medical School Admissions
►Poster #22: Progress in Reducing Mortality among Wisconsin Residents, 1980-2000: Rates Decline, but Black-White Disparities Increase
►Poster #23: Trends and predictors of blood pressure among children and adults during the first 10 years of type 1 diabetes
►Poster #24: The Effect of Data Aggregation in the Estimation of Health Care Costs
►Poster #25: School achievement and its predictors in a regional cohort of Very Low Birthweight children
Poster #26: Relationships of body fat level and distribution to age-related maculopathy in the Carotenoids in Age-Related Eye Disease Study (CAREDS), an ancillary study of the Women’s Health Initiative
Poster #27: Association of allergic sensitization (atopy) with hepatitis A virus exposure in an adult population-based survey: hygiene hypothesis or a confounding effect of age?
►Poster #28: Does Diagnosis by Screening Mammography Lead to a Gain in Life Expectancy for Women with Breast Cancer and if so How Much?
Poster #1
The Appropriateness of Assuming a Lognormal
Distribution when Modeling Health Care Costs
MA Rosenberg,
University of
Wisconsin-Madison,
Madison, WI.
Purpose: To illustrate possible errors from assuming
cost data are lognormally distributed.
Background: The lognormal distribution, characterized by
its skewness and long-tail, has become the choice to
model health care costs, as the log of a lognormal random variable has a normal
distribution and then can be easily modeled in any statistical computing
package. Yet there are many distributions
that have similar shapes to the lognormal distribution. What is the impact of assuming the log of
costs is normal when it is not, and when other similar distributions may have
better fit?
Method: Five hundred
data points are simulated from 3 distributions, Weibull,
Gamma, and Burr, with comparable means and variances. Parameters of a lognormal
distribution were fit to these data using the Maximum Likelihood method.
Results: Figure 1 show
data for the Weibull simulation on a log scale, together
with an overlay of a normal density function using the estimated parameters.
The normal distribution fits the log of the data well.
However Table 1 shows that the fitted lognormal distribution yields a
larger mean, much larger standard deviation and skewness,
and smaller mode than the true distribution on the original scale. Figure 2 shows a graphical representation of
Table 1. These relationships are similar for the Burr and Gamma simulations.
Conclusions: Fitted means and standard deviations in
modeling costs are sensitive to the assumed distribution. Always assuming lognormality
can lead to erroneous estimates of mean costs and standard deviations of costs,
and can bias cost-effectiveness analyses.
Exploration of other long-tailed distributions is appropriate.
Poster #2
Chlamydia trachomatis and Neisseria gonorrhoeae infections
among female and male Wisconsin Family
Planning Clinic clients: prevalence of infection and performance of selective
screening criteria
Arcari CM1,2, Pfister J2,Amsterdam L3, Vaughn
M3, Rombca J4, McDonald R2, Davis JP1,3
1University of Wisconsin,
Madison, WI; 2Wisconsin State Laboratory of Hygiene, Madison, WI; 3Wisconsin
Division of Public Health, Madison, WI; 4Planned Parenthood of Wisconsin,
Milwaukee, WI, USA
Objective:
We conducted a study to determine the current prevalence of Chlamydia
trachomatis (CT) and Neisseria gonorrhoeae
(NG) infections in females and males attending family planning clinics (FPC) in
Wisconsin and to evaluate the effectiveness of existing
selective screening criteria (SSC).
Methods: Sexually active female and male clients
seeking services during June 2001 - February 2002 at ten selected FPCs were invited to participate. During the study period the patient services
protocol at these clinics was changed from selective to universal screening of
all participants for CT and NG infection.
Cervical swab specimens were collected from female participants
scheduled to receive a pelvic exam; urine specimens were collected from all
other females and male participants.
Specimens were tested for CT and NG by a Strand Displacement
Amplification assay (Becton Dickinson ProbeTec). Standard treatment and follow-up guidelines
were used for all participants testing positive for one or both infections. The
sensitivities of existing SSC were calculated using the results of universal
testing as the gold standard.
Results: 6,585 (81.2%) of eligible female clients and
550 (75.8%) of eligible male clients participated. Symptomatic illness or sexual contact with a
person with an STD was the reason for the clinic visit for 14.1% of females and
45.0% of males. The prevalence of CT
infection was 6.5% in females and 18.7% in males, and prevalence of NG
infection, 1.8% in females and 8.6% in males. The sensitivities of current SSC
for CT infection were 83.8% in females and 93.2% in males, and for NG
infection, 89.2% in females and 100% in males.
Conclusions: Existing
SSCs for these infections perform well in FPC
settings. Evaluation of existing SSC can help focus limited STD testing
resources. The impacts of STD screening
programs and changing cultural practices on rates of STDs make it necessary to
evaluate SSC.
Poster #3
Students Teaching and Reaching Students: Planting
Seeds for Careers in Public Health
*CM Arcari, NL Lee (Johns
Hopkins
Bloomberg
School of Public
Health, Baltimore, MD
21205).
PURPOSE: Despite efforts to increase minority representation in
public health, there has been relatively little success in enhancing diversity
among the student population in schools and programs of public health. A pilot
program, Students Teaching and Reaching Students (STARS), was implemented in a
public high school in Baltimore, Maryland, to spark the interest of students in considering
careers in public health and epidemiology. METHODS: The authors created a
series of three 45-minute modules for epidemiology graduate students to teach
to 9th graders. The modules introduce basic epidemiologic concepts using
age-relevant examples. The majority of the students in the targeted high school
are African American and planning to attend college. RESULTS: STARS reached
nearly 500 students during four 3-day teaching sessions from January 2001-June
2002. Based on evaluations from three sessions (N=365), 64% of the students
were African American, 98% considered attending college, and 55% were
interested in science. When asked how much they learned from STARS, 67%
answered a lot and 17% were interested in learning more. 21% responded that
they were somewhat interested, 6% interested, and 3% very interested in a
career in epidemiology or public health. CONCLUSION: The success of the STARS
pilot program is reflected in the number of students who expressed an interest
in epidemiology and public health as a career, in light of the fact that the
vast majority of students were not previously familiar with these careers.
STARS can be implemented throughout the country to widen the pipeline of
minority students from high school to college and into graduate programs of
public health.
►Poster #4
Use Patterns of Internet-Based Systems for Smoking
Cessation and Smoking Outcome
Mark E. Zehner, B.S.*, Devayani
S. Pophali, M.S., Stevens S. Smith, Ph.D., Michael C.
Fiore. M.D., M.P.H., Timothy B. Baker, Ph.D., David H.
Gustafson, Ph.D., University of
Wisconsin
Aim: The goal of this study is to determine benefits of
using interactive health communications (IHCs)
designed for smoking cessation by examining program use patterns of smokers
given access to the Comprehensive Health Enhancement Support System (CHESS)
Quitting Smoking For Life website.
Design: As part of a clinical trial designed to
evaluate the efficacy of the CHESS Quitting Smoking for Life website,
participants were assigned to receive brief treatment (BT) or BT plus access to
CHESS for up to 90 days (BT+CHESS). BT
consisted of bupropion SR pharmacotherapy for 9
weeks, three counseling sessions and a quit guide booklet. Use patterns were
analyzed by examining instances of distinct page views per participant. Smoking
status was verified by exhaled CO.
Results: The sample (N=213) was primarily white (78%);
20% were African-American. Mean (SD) age
was 40.0 (11.8) years; 53% were male; mean years of smoking = 22.4 (SD=11.3)
and current cigarettes/day = 21.5 (SD=9.3).
The median (22 views) was used to group CHESS users into low and high
use groups. At 12 weeks post-quit, 41%
of high use participants were abstinent compared to only 16% of low use and 21%
of BT participants, chi-square(2)=9.3, p=0.01.
Conclusion: Participants with greater CHESS activity had
greater success in achieving 3-month abstinence. IHCs hold promise
for providing access to treatment information and support during a quit attempt
and increase the likelihood of successfully quitting.
Supported
by NIH Center Grant #1P51-CA84724-03.
CORRESPONDING AUTHOR: Mark
Zehner, CTRI, University of Wisconsin Medical School, 1930
Monroe Street, Madison, WI 53711-2027, USA.
Poster #5
Wisconsin Community Health Report Cards—2003
Edition
Paul E. Peppard, David Kindig,
Amanda Riemer, Elizabeth Dranger, Patrick Remington
Wisconsin Public Health and Health Policy Institute,
Department of Population Health Sciences,
University of
Wisconsin-Madison.
The
Wisconsin Public Health and Health Policy Institute presents
the first annual Wisconsin County Health Rankings. The Rankings represent the
ongoing development of a method to summarize and compare population health at
the county level using publicly-available data. The framework underpinning this
effort is based on a model of population health improvement. This indicates
that our health outcomes and their distribution across the population are
produced by a set of health determinants, which in turn are influenced by
policies and interventions which enhance or limit the determinants.
The
Rankings compare counties on indices of population health in an attempt to
stimulate discussion within the Wisconsin health policy and public health communities. Specific
intended functions of the report include: 1) a synthesizing-integrating function—by
succinctly summarizing population health measures at the Wisconsin county
level; and 2) a comparative function—by explicitly contrasting population
health indicators among Wisconsin counties.
The
Rankings are hierarchal: at the highest level, counties are ranked on global
measures: A) health outcomes, and B) health determinants. The health outcomes
are a combined measure of two items: 1) mortality (years of potential life lost
prior to age 85 years—YPLL-85) and, 2) self-reported general health status. The
health determinants are subdivided into four major components: 1) access to
health care; 2) health behaviors; 3) socioeconomic correlates of health; and 4)
the physical environment. In total, 18 different health measures are used to
express the determinants components. Individual health measures that comprise
these components (e.g., smoking prevalence, a health behavior) were explicitly
chosen to reflect the majority of the State’s Health Priority areas.
The
data for the Rankings come from a variety of sources including: the Centers for
Disease Control and Prevention’s Behavioral Risk Factor Surveillance System and
WONDER mortality data bases; the State of Wisconsin’s Department of Health and Family Service’s Family
Health Survey and Wisconsin Interactive Statistics on Health; and the U.S. 2000 Census. These data were used to make
county-level estimates. Weighted sums of the county-level estimates were
combined to create scores for health outcomes, determinants and their
components. These scores were then used to rank counties’ multi-dimensional
population health.
Poster #6
Relation between Sleep Apnea and Metabolic Syndrome in
a Community-Based Sample
Nieto FJ, Peppard P, Shankar A, Austin D, Young TB.
Clinical and epidemiologic
studies suggest a causal link between sleep apnea and cardiovascular disease,
hypertension, and insulin resistance but the mechanisms for these associations
remain uncertain, particularly in view of the strong relation of obesity with
all these conditions. To explore the relation
between sleep apnea and the metabolic syndrome we used prospective data from
the Wisconsin Sleep Cohort, a community-based sample of 1,049 middle-age men
and women who have undergone up to four overnight examinations at four year
intervals between 1989 and 2002.
Metabolic syndrome was defined based on the National Cholesterol
Education Program criteria. Sleep apnea
was defined based on polysomnographic determination
of the apnea-hypopnea index (AHI, average number of
apneas/hypopneas per hour of sleep). Baseline prevalence of the sleep apnea
(AHI>5/hour) was 46% among individuals with metabolic syndrome (n=433)
compared to 17% among those without MS (n=716), p<0.001. Among participants without baseline metabolic
syndrome, sleep apnea was associated with incident metabolic syndrome in
age-sex-adjusted analyses [odds ratio (OR), 2.5, 95% confidence limits (CL),
1.4-4.4). Further adjustment for
baseline BMI decreased this estimate but a significant association was still
present (OR 1.9, 95% CL, 1.2-2.9). These results are compatible with the
hypothesis that sleep apnea is associated with the pathogenic precursors of
insulin resistance and the metabolic syndrome.
This relation does not appear to be totally explained by the common
occurrence of obesity in association with these conditions.
Poster #7
Oral Contraceptive Use, Reproductive Factors and
Colorectal Cancer Risk:
Findings from Wisconsin
Hazel Nichols; Amy Trentham-Dietz;
John Hampton; Polly Newcomb
Epidemiological
similarities between colorectal and breast cancer have stimulated speculation
on the role of reproductive hormones in colorectal cancer etiology. To evaluate
the association of oral contraceptive use, reproductive factors, and colorectal
cancer, we conducted a population-based case control study. Female Wisconsin residents aged 20-74 years with a new diagnosis of colon (N=1,122) or
rectal (N=366) cancer were reported to the statewide tumor registry. Similarly-
aged female controls (N=4,297) were randomly selected from driver's license
files and Medicare beneficiary lists.
Oral contraceptive use and other risk factor information was collected through structured telephone interviews. Odds
ratios (ORs) and 95% confidence intervals (CI) were
estimated using conditional logistic regression and were adjusted for known
risk factors.
A
modest inverse association was observed between oral contraceptive use and
colon cancer risk (OR=0.9; CI: 0.7, 1.1). Premenopausal
status (OR=0.6; CI: 0.4, 0.9) and age at first birth older than the median (23
years) (OR=0.8; CI: 0.7-1.0) were associated with decreased risk of colon
cancer.
Oral
contraceptive use did not appear to be associated with rectal cancer risk,
except among recent users. Use of oral
contraceptives within the past 14 years was associated with reduced rectal
cancer risk (OR=0.5; CI: 0.3, 1.0). Rectal cancer risk appeared to decrease
with increasing parity (p=0.05). Women with 5 or more births had 0.7 times the
risk of rectal cancer (CI: 0.4, 1.0) compared to nulliparous
women.
Despite
numerous investigations, the association between reproductive factors, oral
contraceptive use and colorectal cancer has remained uncertain. Decreased bile acid secretion, shifts in
endogenous hormone levels following pregnancy, and the down-regulation of
insulin-like growth factor (IGF) are plausible biological mechanisms for
reductions in colon and rectal cancer risk.
These findings support differential roles of oral contraceptive use and
reproductive factors in colon and rectal cancer etiology.
►Poster #8
Self-Reported Habitual Snoring Reliability in the Wisconsin Sleep
Cohort Study
Swain, R. Austin, D. Young, T. Dept. of Population
Health Sciences, University of
Wisconsin-Madison
Introduction: Epidemiology studies of the correlates and
adverse health effects of sleep-disordered breathing commonly rely on
self-reported snoring frequency as the key variable. Classification of chronic
snoring status based on this one-time measure may introduce error due to the
subjective nature of the data as well as possible changes in snoring frequency
over time. Limited data indicate that
short-term reliability is moderate at best.
We investigated the consistency of self-reported snoring frequency over
3 -24 months to determine the reliability of habitual snoring status.
Methods: Our
sample consisted of 1197 men and women, ages 30-65, enrolled in the Wisconsin
Sleep Cohort. Participants completed a
mailed questionnaire at baseline that included the following question: “According to what others have told you,
please estimate how often you snore ”. Semi-quantitative response categories were:
1) Never; 2)Rarely
-only once or a few times ever ; 3) Sometimes -a few nights per month; 4) At
least once a week, but pattern may be irregular; 5)Several (3 to 5) nights per week;
6) Every night or almost every night.
The same question was asked again at the participant’s overnight
laboratory study. For this analysis,
categories were combined to define snoring status as follows: non-snoring =
category 1 or 2, some snoring= category 3 or 4, and habitual snoring = category
5 or 6. The time intervals between the
mailed survey and data collected by questionnaire in the laboratory were less
than 3 months for 73 participants, 3-6 months for 178 participants, 6-12 months
for 379 participants and 12-24 months for 567 participants.
Results: Of
participants reporting habitual snoring at baseline, 88%, 84%, 86% and 86%
reported habitual snoring at the 3-, 6-, 12-, and 24-month time intervals. Over 99% (80/81) of the participants who were
discordant on habitual snoring shifted to the some snoring category at
follow-up. The % concordance, however, for non-snoring was lower, with 36%,
50%, 66% and 62% of non-snorers also reporting never or rarely snoring at the
3-, 6-, 12, and 24-month time intervals.
Of those discordant on non-snoring, approximately 75% reported some snoring follow-up and
approximately 25% reported
habitual snoring at follow-up.
Conclusion:
Habitual snoring status based on self-reported frequency of snoring appears to
be a reliable indicator for at least 2 years post survey, with nearly all
habitual snorers reporting at least some snoring at any follow-up. The lesser reliability of self-reported
non-snoring is likely to be due in part to the expected increase in prevalence
of snoring with age, but may also reflect errors in the subjective report.
Research support: NIH grants HL 62252 and
AG14124
Poster #9
Validation of an Environmental Tobacco Smoke Exposure
Questionnaire.
DM Nondahl,* CR Schubert, KJ Cruickshanks.
University of
Wisconsin,
Madison, WI 53726.
Exposure to environmental
tobacco smoke (ETS) has been associated with a variety of negative health
consequences among nonsmokers. Researchers often obtain self-report data
on ETS exposure via questionnaire. In this study, serum cotinine levels are compared with self-reported ETS
exposure to assess the validity of the data obtained by self-report. As
part of the population-based Epidemiology of Hearing Loss Study in Beaver Dam,
Wisconsin, self-reported data on ETS were obtained at the
five-year follow-up examination (1998-2000; n=2800). Serum cotinine levels were obtained on 643 of these participants
(341 males, 302 females, ages 53-75). Among
nonsmoking participants with serum cotinine
measurements, about one fourth (24.2%) reported exposure to ETS within the past
24 hours, but 56.0% had detectable levels of serum cotinine.
Serum cotinine levels increased with reported number
of hours of ETS exposure in the past 24 hours (F-test for trend: p <
.0001). Most (95.3%) participants who reported being smokers had serum cotinine levels > 15 ng/mL,
compared to only 2.7% of nonsmokers. Usual ETS exposure among nonsmokers
was assessed with a brief questionnaire quantifying exposure at work, at home,
and in social settings and then classified into three levels of exposure.
Those reporting no/little exposure had a geometric mean cotinine
level of 0.06 ng/mL (95%CI
0.05, 0.07), compared to 0.14 (95%CI 0.09, 0.22) and 0.63 (95%CI 0.36,
1.10) for those reporting moderate and high ETS exposure, respectively (F-test
for trend: p < .0001.) These preliminary results suggest that the
questionnaire can be used to distinguish relative levels of exposure to ETS.
Poster #10, Location #19
Vaccination levels in older adults in Beaver Dam, WI.
*CR Schubert, DM Nondahl, KJ Cruickshanks,
BEK Klein, R Klein (University of
Wisconsin,
Madison,
WI 53726)
It is recommended that all
persons age 65 years and older get an annual influenza vaccination and a
one-time pneumococcal vaccination. The purpose of
this research was to determine the level of vaccination and factors associated
with vaccination in an older population.
Subjects were participants in the 7-year interview of the Epidemiology
of Hearing Loss Study, a population-based study in Beaver Dam, WI. Participants (n=2433) aged 55-99 years (mean
of 71.3 years) of age at time of the interview, were asked if they had a flu
shot in the last 12 months and if they ever had the vaccination for pneumonia.
Preliminary data indicates that vaccination levels for participants 65 years
and older were 71.5% (69.3-73.7, 95% Confidence Interval (CI)) for influenza,
64.3% (62.0-66.7%, 95% C.I.) for pneumococcal, and 55.5 % (53.1- 58.0%, 95% C.I.) received
both. There was no difference between
men and women in the rate of vaccination.
Factors associated with receiving the influenza vaccine included age
(Odds Ratio (O.R.) = 1.10, 1.01-1.20 95% C.I.), having seen a doctor within the
last year (O.R.=2.56, 1.83-3.58 95% C.I.), having a
chronic disease (O.R.=1.47,1.17-1.84 95% C.I.), and being retired/homemaker
(vs. working full or part-time)(O.R.=1.59, 1.16-2.17 95% C.I.). Factors associated with receiving the pneumococcal vaccine were similar. While vaccination rates for this population
were fairly good, of those 65 years and older with a chronic disease, 23.1 %
had not received an influenza vaccine and 29.2 % had not received a pneumococcal vaccine.
Efforts are still needed to increase vaccination levels of older adults,
especially those in the highest risk groups.
►Poster #11
The Success of Collaboration: Wisconsin’s Diabetes
Quality Improvement Project, 1999-2001
Effie Siomos, Jennifer Camponeschi,
Robert Stone Newsom, Patrick Remington
Objective: To monitor trends in the quality of diabetes
care and collaborate on quality improvement efforts among Wisconsin HMOs in
Wisconsin.
Methods:
Each year from 1998 to 2001, all Wisconsin HMOs were invited to submit HEDIS® Comprehensive Diabetes
Care measures and since 2000, selected cardiovascular care measures. HMOs reported data collection methods, sample
size, eligible population, and percent of individuals receiving each HEDIS®
measure. Confidentiality was maintained,
yet each HMO could compare their performance to other organizations. Aggregate
percentages were calculated as weighted averages to total eligible population
of diabetics in each organization.
Results: Fifteen HMOs participated in all 3 years,
between1999-2001, and collectively improved on 4 of the 6 HEDIS® Comprehensive
Diabetes Care measures (LDL-C screening improved by 14%, LDL-C control by 22%,
nephropathy monitoring by 15%, and poorly controlled A1c
by 11%). Some measures demonstrated great variability between HMOs (nephropathy
monitoring varied from 44% to 98% in 2001), while other measures showed little
variability (one/more A1c varied from 80% to
97%). Between
2000-2001, the HMOs in Wisconsin collectively improved on all of the HEDIS® selected
cardiovascular measures. For all but one
of the diabetes and cardiovascular measures, Wisconsin HMOs perform above
regional and national averages.
Conclusions: Most diabetes-related clinical testing
continues to improve in Wisconsin HMOs.
Variation in some measures demonstrates potential for continued
improvement. This project demonstrates
that HMOs will collaborate with multiple partners, the state health department,
and a university, on quality improvement projects, and these collaborations are
successful in promoting better diabetes care in Wisconsin.
Poster #12
Relationship Between Dietary Patterns and Age-Related
Maculopathy in the Carotenoids in Age-Related Eye Disease Study (CAREDS), an
ancillary study of the Women's Health Initiative
SM Moeller1, LF Tinker2, B Blodi3, R Chappell4, C
Ritenbaugh5, JA Mares3
1Nutritional Sciences and
Ophthalmology & Visual Sciences, University of WI-Madison, 610 N. Walnut
St.; 1071 WARF; Madison, WI 53726-2336
2Fred
Hutchinson Cancer Research Center, Seattle,
WA
3Ophthalmology and Visual
Sciences, University of WI-Madison, Madison, WI
4Statistics and Biostatistics
& Medical Informatics, University of WI-Madison, Madison, WI
5Kaiser
Permanente Center for Health Research, Portland, OR
Purpose: To examine the relationship between overall
dietary patterns and prevalence of age-related maculopathies
(ARM) in older women.
Methods: Women’s Health Initiative Observational Study
participants aged 50 y+, at 3 sites, who reported high or low intakes of lutein and zeaxanthin on their
baseline food frequency questionnaire (1994-1998) were recruited approximately
5 years later (2001-2003) to participate in the Carotenoids in Age-Related Eye
Disease Study (n=1512). Stereo fundus photographs were obtained and graded using the
Wisconsin Age-Related Maculopathy Grading System. Cluster analysis was used to identify the
predominant dietary patterns in these women.
Results: Seven dietary pattern groups were identified,
with comparatively high intakes of: 1) alcohol (13% of energy); 2) low-fat
dairy (18% of energy); 3) refined grains (18%); 4) whole grains (16%); 5) red
meat (15%); 6) fruit and vegetables (31%); and 7) desserts and sweets
(23%). After adjusting for age and
energy intake, the fruit and vegetables group had significantly higher intakes
of carotenoids, including lutein
and zeaxanthin, than the other groups. The lowest intakes of carotenoids
and micronutrients, and the highest fat intakes, were seen in the red meat and
desserts groups. There was some
suggestion of a difference in the prevalence odds of early ARM between the
pattern groups (p=0.09), with the fruit and vegetables group appearing the
least protective. After excluding women
with a prior diagnosis of AMD (n=110), this association no longer approached
significance (p=0.20). Diet patterns
were not associated with any specific lesions of early ARM or late ARM. Further adjustment for smoking, family
history, high dose antioxidant supplement use, hormone replacement therapy,
waist-to-hip ratio, or history of cardiovascular disease (late ARM only) did
not significantly alter the odds ratios.
Conclusion: The self-selected dietary patterns of older
women five years in the past do not appear to be strongly associated with
prevalence of ARM. Further work is
needed to determine if dietary intakes further in the past, or specifically
prescribed diets, may be related to ARM.
Support: T32 DK07665, EY0
13018
Poster #13
New Mothers Survey, July 2003
Lynda Knobeloch, Ph.D, Marty Kanarek, Ph.D, Gemma Gliori
The Wisconsin Department of
Health and Family Services (DHFS) initiated a pilot intervention program aimed
toward educating women of childbearing age to the risks of methylmercury
(MeHg) toxicity.
Details regarding fish and seafood contamination as well as resulting
health affects were illustrated in multilingual posters, brochures, fact cards,
growth charts and other items. These
outreach materials were sent to WIC clinics, local health departments and
physician offices (primarily family practice, OB/Gyn
and pediatric offices) to display and distribute to women visiting their
facilities. The New Mothers survey was
the first evaluation tool implemented to assess the effectiveness of this
intervention program. 1,000 women that
gave birth between June 1 and June 7, 2003 were randomly selected for the survey and were sent questionnaires
through the mail. The questionnaire was
designed to assess estimates of the number and types of fish meals respondents
were eating per month, their knowledge of fish/seafood contamination and
details regarding exposure to advisories or outreach materials. There was a 74% response rate to the survey
with 726 completed questionnaires, 253 non-responses and 21 non-samples. On average, women were eating 2.56 fish meals
per month that consisted mostly of canned tuna and frozen fish. Approximately 1/3 of participants were aware
that older fish and fish that eat other fish had higher levels of mercury then
younger, plant eating fish. 46.7% of
women had at least some awareness of sport fish advisory guidelines provided to
anglers in Wisconsin while 13% of women recalled seeing the posters and
pamphlets provided by DHFS.
►Poster #14
The Newborn Lung Project Statewide Cohort Study
Aggie Albanese, Kathleen Madden, Mona Sadek-Badawi, Mari Palta
The Newborn Lung Project
Statewide Cohort Study is to continue our long-standing research in the
evolving risk factors and outcomes of bronchopulmonary
dysplasia (BPD) and other neonatal conditions of very
low birth weight (VLBW) infants and children. This study will document all the
VLBW births in Wisconsin occurring in the 2003 and 2004 calendar years. The
study includes 16 level III NICUs (Neonatal Intensive
Care Units) in Wisconsin and one in Duluth,
Minnesota, for births in far northern Wisconsin. Nurses at the hospitals collect baseline and
clinical information on the infants who are at or below 1500 grams at birth.
The collected information will examine the currently proposed hypothesis of the
cause of BPD and implement and validate new diagnostic criteria for the
severity of BPD, developed at a recent NIH workshop. A comprehensive follow-up
will be done of all children at age 2, with a parent interview. At age 4, lung
measurements will be collected using Impulse Oscillometry
(IOS) testing.
Poster #15
Exercise Induced Wheezing and Lung Function in Very
Low Birth Weight Children at ages 8 – 10 with the Newborn Lung Project Regional
Cohort.
Kathleen Madden, Mari Palta,
Mona Sadek-Badawi, Aggie Albanese, Christoper Green for the Newborn Lung Project.
Very low birth weight (VLBW)
infants are known to suffer from respiratory problems related to prematurity. BPD (bronchopulmonary
dysplasia) is the presence of radiographic changes in
the lungs and the need for supplemental oxygen at 1 month of age. One part of
this study intended to find and evaluate a reliable lung function test, obtain
information on wheezing, exercise induced wheezing, and lung function in VLBW
children and controls and relate those outcomes to the incidence of BPD for
children at 8 – 10 years of age.
►Poster #16
Factors responsible for observed changes in Chlamydia
trachomatis test positivity.
Pfister JR1, RA McDonald1, C Arcari2, L Amsterdam3, M
Vaughn3.
1Wisconsin State Laboratory
of Hygiene, and 2Department of Population Health Sciences, University of
Wisconsin-Madison; 3Wisconsin Department of Health and Family Services,
Division of Public Health, Madison, WI.
Background:
Systematic monitoring of the proportion of positive laboratory tests
(positivity) among clients screened at sentinel provider sites is frequently
used to evaluate the success of Chlamydia trachomatis (Ct) control programs.
Observed temporal changes in test positivity are often attributed to changes in
the prevalence of infection in the underlying population. However, other
factors may be responsible for these observed changes in Ct positivity.
Objective:
To assess possible explanatory factors for the increase in Ct positivity
observed in two sequential prevalence studies conducted in Wisconsin.
Methods:
Data from two Ct prevalence studies in 1997 and 2001 employing universal
screening in the same five Wisconsin family planning clinics were evaluated. To estimate
the true prevalence in the study populations, the observed test positivity was
adjusted for the sensitivity and specificity of the enzyme immunoassay (EIA)
used for the 1997 study and the nucleic acid amplification test (NAAT) used for
the 2001 study. Direct standardization was used to further adjust test
positivity for differences in age distribution, client risk profiles, and
clinic mix.
Results:
Unadjusted Ct positivity among females attending these five family planning
clinics increased from 3.9% using EIA in 1997 to 6.0% using NAAT in 2001
(p<0.01). Compared to the 1997 study population, a higher proportion of
study subjects in 2001(73.7% versus 44.6%, p<.001) were enrolled at the two
urban clinics with the highest positivity. The age distribution and risk
factors for Ct infection were similar in both studies. When the differences in
testing methodology and clinic mix were taken into consideration through
statistical adjustment, the estimated true prevalence rates were 4.7% for 1997 and 4.9% for 2001
(p=.78).
Conclusions: The
observed increase in Ct test positivity between 1997 and 2001 in the same five Wisconsin family planning clinics may be attributed to a change in laboratory
testing methodology and differing proportions of testing in high versus low
prevalence clinics, rather than a true increase in prevalence.
Implications for Programs, Policy, and/or Research: Changes in test positivity rates identified through
screening programs and research must be interpreted with caution. Observed
changes in positivity may reflect changes in testing practices and composition
of study populations rather than changes in true prevalence.
►Poster #17, Location #32
Psychological traits and preference for control over
health care decisions
Kathryn Flynn, Maureen Smith
Little
is known about the importance of long-standing psychological traits in
preferences for control over healthcare decisions. If preferences are related
to underlying psychological traits, the effectiveness of attempts to empower
patients through increased shared decision making (e.g., decision aids) may be
limited because some patients cannot be motivated to participate.
Data
are from the first three random replicates of the 2003 Wisconsin Longitudinal
Study, a 1/3 random sample of graduates from Wisconsin high schools in 1957 (N=10,317) who have been followed via telephone
and mail surveys for 46 years. The dependent variable representing preference
for control over decision-making was, “The important medical decisions should
be made by my doctor, not by me,” coded on a five-point scale from “strongly
agree” to “strongly disagree.” Psychological traits, collected in 1992,
included the five-factor model of personality (extraversion, openness,
conscientiousness, neuroticism, agreeableness) and Ryff’s
six-factor model of psychological well-being (autonomy, environmental mastery,
personal growth, positive relations, purpose in life, self-acceptance).
Multinomial logistic regression adjusted for potential confounders including
patient, provider, and relationship characteristics.
Of
1279 respondents aged 63-65, 24% preferred doctor control over important
medical decisions, 20% were neutral, and the remaining 56% preferred personal
control. After adjustment, increasing openness (OR=1.10, 95%CI=1.03-1.17),
autonomy (OR=1.09, CI=1.03-1.16), and personal growth (OR=1.10, CI=1.03-1.18)
were associated with preference for personal control compared to doctor
control. Other variables significantly associated with preference for personal control
included female gender, increased education, and lack of chronic joint
problems. Autonomy and self-acceptance were significantly associated with
neutral preference for control compared to doctor control.
Respondents varied in their preferences for control
over healthcare decisions. For the near elderly, psychological traits measured
a decade previously strongly predicted decision making preferences, supporting
future investigations of the role of long-standing psychological traits in
empowering individual patients for shared decision making.
►Poster #18, Location #35
Cancer care capacity and out-of-area care utilization
among urban and rural women with breast cancer in Wisconsin
Authors: Strombom IM, Smith MA, Trentham-Dietz
A, Remington PL and Newcomb PA
Research objective: Treatment for breast cancer is well-established and involves
comprehensive services offered predominantly at approved cancer centers. Not
all women have an approved cancer center in their residential area, and no
population-based studies have examined whether a lack of comprehensive cancer
care affects the decision to seek treatment outside a woman's residential
area. We evaluate the association between a woman's local cancer care
capacity and utilization of cancer treatment outside the woman's residential
area in rural and urban Wisconsin.
Methods: We use data on breast
cancer cases diagnosed between 1987and 1994 (n=6,909). Hospital data was
aggregated to the level of Dartmouth Health Service Area (HSA). The
dependent variable was cancer care within the woman's residential HSA.
Cancer care capacity was measured by the number of hospitals with approved
cancer centers in the HSA. Multivariable logistic regressions were
stratified by rural/urban, adjusted for confounders, and accounted for
clustering within HSAs.
►Poster #19, Location #36
Wrong site surgery: the use of human factors system
analysis and process analysis in the outpatient setting
Kara Schultz, Pascale Carayon, Lynn Jenkins, and Sue Cota
Wrong site surgery is an
issue of increasing importance in patient safety, especially in outpatient
settings where the number of surgeries performed and the time pressures
involved continue to increase. Most attempts to prevent the occurrence of wrong
site surgeries tend to address specific aspects of the system without
consideration of the overall work system and interactions between elements of
the system. The use of a human factors system analysis can be used to better
understand how elements of a work system combine and interact to contribute to
breakdowns in the system. This analysis can further be used in efforts to
optimize the work system with the goal of preventing or detecting system
breakdowns before they result in serious events occurring. Highlights of a
human factors system analysis of a wrong site surgery case study in an
outpatient setting illustrate how different work system elements can contribute
to a wrong site surgery occurring.
Poster #20, Location #38
Geographic and Age-Specific Variation of Pedestrian
Injuries in New York City:
Implications for Prevention.
Janine R. Clemmons, Jane A. McElroy, Maureen S.
Durkin.
Introduction: Pedestrian injuries are a leading public
health problem in urban communities, with children and elderly people at
greatest risk. The purposes of this
study were to describe the spatial distribution of traffic collisions resulting
in pedestrian injuries in New York City, determine whether areas of greatest risk are the same for children
and for elderly persons, and draw implications for planning effective
interventions to prevent pedestrian injuries.
Methods: Data on traffic collisions occurring in
New York City during a 10-year period, 1991-2000, were obtained
from the New York State Department of Motor Vehicles computerized Police
Accident Reports (MV-104). 123,037
pedestrians were injured during this time in the 5 counties of New York City. Each
collision was geocoded to an intersection or
mid-block site. ArcView
3.3 and ArcGIS 8.2 software were used for mapping and
exploration of spatial patterns. A
kernel density estimator, CrimeStat, was used to
interpolate the density of crash sites across the continuous surface.
Results: 24,381 children under the age of 13 years and
13,571 persons over the age of 64 years sustained pedestrian injuries in
New York City during the 10-year study period. Two thirds of the injuries occurred at
intersections. Collisions resulting in
pedestrian injuries occurred in a non-uniform manner throughout the City. The highest risk areas included the south Bronx
and east Harlem for children, and lower east Manhattan and central Queens for persons
over 64. Analysis of smaller areas
within the high risk areas revealed specific intersections at greatest
risk.
Conclusions: Pedestrian collision sites are not uniformly
distributed throughout New
York City, and
the areas at greatest risk vary for children and the elderly. Information on the spatial distribution of
pedestrian collision sites is a powerful tool for mobilizing local communities
and agencies to implement effective injury prevention interventions.
Poster #21, Location #39
A Comparison of Epidemiological Indices of Association
for Setting MCAT and GPA Thresholds in Medical
School Admissions
Mark A. Albanese, Ph.D., Philip Farrell, MD, Ph.D.,
Susan L. Dottl, Ph.D.
University of Wisconsin
Medical
School,
Madison,
Wisconsin
Recent Supreme Court rulings
are likely to increase the complexity of the admissions processes for
professional schools, raising awareness that such processes cannot be
formulaic. In 2001, Dr. Jordan Cohen,
President of the AAMC, called for medical schools to consider using an MCAT
threshold to eliminate high risk applicants from consideration and then to use
non-academic qualifications for further consideration. This approach would seem to be consistent with
the recent Supreme Court ruling.
Research to support such an approach has been reported in many different
ways and over various time periods, making comparability problematic.
Objective:
This timely study uses various statistical indices to determine optimally
discriminating thresholds for
Medical College Admission Test (MCAT) sub-score and undergraduate
science grade point average (USGPA) use in admissions in terms of their
sensitivity and specificity for determining USMLE Step 1 first time failure.
Design: Data for medical school entering classes of
1992-1998 (N=752) from a large Midwestern medical school are used to develop
guidelines for cut-scores that optimize discrimination between students who
pass and do not pass Step 1.
Results: Plots of risk differences, odds-ratios,
sensitivity and specificity provided the most discriminating information for
setting threshold cut-scores.
Compensatory versus non-compensatory cut-score procedures produced
inconclusive results.
Conclusions: Rational and defensible intellectual
achievement thresholds that comply with recent Supreme Court decisions can be
set based upon information readily available in the application materials and
is achievable in an educational setting, where much uncontrolled variability
exists in both the subjects and the measures used for evaluation.
►Poster #22, Location #41
Progress in Reducing Mortality among Wisconsin Residents,
1980-2000:
Rates Decline, but Black-White Disparities Increase
Elizabeth A. Dranger, Patrick Remington, MD, MPH, Paul
E. Peppard, PhD
Purpose: To
assess progress towards two overarching public health goals—improvement in
length of life and reducing health disparities.
Methods: Age
specific mortality rates in Wisconsin from 1980 to 2000 were obtained from the U.S. Centers
for Disease Control and Prevention WONDER database. Rates for each age group were gathered for
the entire Wisconsin population and for black and white subgroups. Trends in mortality rates were plotted,
change in mortality rates was estimated, and the number of “lives saved”
annually from the 1980-1984 to 1996-2000 was calculated. In addition, black versus white rate ratios
were calculated at both the beginning and the end of the time period to determine
trends in black-white mortality disparities.
Results:
Mortality is decreasing in Wisconsin
in every age group. The largest relative
improvements in mortality rates occurred among infants < 1 year (-30%),
children 1-14 years (-27%), and adults 45-64 years (-23%). Comparatively little progress was seen among
adults 25-44 years (-5%), and those 85 years and older experienced a slight
increase in mortality (+0.5%). Lives
saved in every age group correspond to the reduction in mortality in that age
group. During 1996-2000, approximately
5,000 fewer deaths occurred each year than expected based on mortality rates
from 1980-1984.
Conclusion:
Despite progress towards increasing length of life, progress towards
eliminating disparities was not seen over the 1980 to 2000 time period, rate
ratios increased in every age group of the population.
►Poster #23, Location #42
Trends and predictors of blood pressure among children and
adults during the first 10 years of type 1 diabetes.
K.K. Danielson, H. Zhang, M. Palta,
T. LeCaire, C. Allen, and D.J. D’Alessio
for the Wisconsin Diabetes Registry Project
Type 1 diabetes is a known
risk factor for hypertension. However,
there is limited information on changes in blood pressure during the first 10
years of diabetes duration. A Wisconsin population-based incident cohort with type 1 diabetes was followed for
up to 10 years duration from 1987 to 2001 to examine, 1) whether blood pressure
levels differed from individuals of similar age in the general population, and
2) the effects of duration and glycemic control on
blood pressure in early diabetes. Data
were available on 525 individuals from physical exams conducted at baseline
(4-6 months), 4, 7, and 9 years duration, phone questionnaires at enrollment,
semiannual self-administered questionnaires, and medical records. Total glycosylated
hemoglobin was determined from blood samples collected up to three times each
year. Age-specific mean systolic and
diastolic blood pressure were similar in individuals
with diabetes and a general population sample from the National Health and
Nutrition Examination Survey II.
Multivariable random-effects modeling showed that diastolic blood
pressure increased significantly with duration for all age groups except young
adults, and systolic blood pressure increased significantly with duration in
the youngest age group. Glycosolated hemoglobin was not related to systolic or
diastolic blood pressure. These data
suggest that there are small, non-detectable differences in blood pressure
between individuals with and without type 1 diabetes during the first 10 years,
but that diabetes does effect blood pressure via
duration, though the effect is not uniform across all ages.
►Poster #24, Location #43
The Effect of Data
Aggregation in the Estimation of Health Care Costs
Henry J. Henk
Department of Population
Health Sciences, University of Wisconsin-Madison
Introduction: Economic evaluation
is increasingly common in clinical trials. Often, individuals’ health care costs are not
observed in these trials, rather health care cost estimates require
that monetary values be assigned to measured resource utilization (generically
referred to as “resource costing”).
The level of detail at which resource use is measured can bias
cost estimates. Using hospital inpatient data, this study illustrates
the effect level of aggregation of utilization data can have on cost
estimation.
Methods: Resource use data was collected as part of Project
CARE, a randomized clinical trial testing the cost effectiveness of
a depression management program among adult patients with depression
and a history of high medical utilization.
This study uses 3 years of hospitalization data (1996 –1998)
from Project CARE and price data from the Hearth Care Utilization Project’s
(HCUP) Nationwide Inpatient Sample (NIS) consisting of a sample of 1,000+
U.S. hospitals. Cost-to-Charge
ratios were applied and only those patients 18-64 year old were selected.CARE
hospital days were cost at four aggregation levels: (1) single price
per hospital day, (2) price per days by medical or surgical admission,
(3) price per day by Major Disease Classification (MDC), and price per
day by DRG.
Results:
Average hospital costs were: $4,053 (method 1), $4,986 (method
2), $4,685 (method 3), and $5,482 (method 4).
Conclusions: The level of detail at which resource use is
measured can affect the results of economic evaluation when resource
use rather than costs are observed.
Costing hospital days using DRG level data resulted in mean costs
35% larger than methods using a fixed per diem price.
►Poster #25
School achievement and its predictors in a regional
cohort of Very Low Birthweight children
Erika Hagen, Mari Palta, Aggie Albanese, Mona Sadek-Badawi
Background:
Studies following children born in the 1970s and early 1980s have demonstrated
that children born very low birthweight (VLBW, <
1500g) have worse school outcomes than their peers. However, little is known about the school-age
outcomes of VLBW children born during the late 1980s and early 1990s, and
earlier-life predictors of school-age success have not been investigated in
children born during this time period.
Objectives:
To determine how the 4th grade academic achievement of a regional cohort of
VLBW children compares to that of the general 4th grade population, and to
determine whether success can be predicted by neonatal and early childhood
variables.
Sample:
Cohort of all VLBW neonates admitted 8/1/88 - 6/30/91
to 6 regional Neonatal Intensive Care Units, covering a contiguous area of WI
and IA. This study analyzes the standardized
test scores of children who attended public school in Wisconsin and teacher ratings for all cohort children. Controls for the test scores analyses are all
children who took the 4th grade tests in the school districts represented by
our cohort. Classroom controls were
selected for a teacher rated comparison group.
Measures: The Wisconsin Knowledge and Concepts Exam
(WKCE) categorizes children into 4 proficiency levels
for 5 subject areas: reading, language, math, science, and social studies. Teacher ratings on the Teacher Report Form
(TRF) provide an overall academic T score, based on a 5-point rating system for
4 subject areas: reading, math, science, and social studies.
Analysis:
Chi-square tests for trend were performed to compare WKCE levels between VLBW
and controls. Mean scores were used to
compare VLBW and controls for the TRF.
Ordinal logistic and linear regression was used to predict WKCE scores
and TRF outcomes, respectively, based on indicators at the neonatal and early
childhood time points.
Results:
Math scores on the WKCE differed significantly, with VLBW children performing
less well than population controls. No
significant differences were found for the other 4 subjects. For math scores, birthweight
(odds ratio, OR=1.14 per 100 gm), social function at age 5 (1.04 per 10
points), socioeconomic status (OR=1.87 per SD), and family relative
socioeconomic status (OR=1.52 per SD) were all independent predictors of
success. VLBW and controls had significantly
different academic T scores on the TRF.
Neonatal respiratory score, social function at age 5, SES, and family
relative SES were each independent predictors of the TRF academic T score.
Conclusion:
VLBW children perform significantly less well than population controls on
standardized tests and in teacher ratings.
While socioeconomic variables were the most significant predictors of
school success, neonatal and early childhood variables also predict success.
Poster #26
Relationships of body fat level and distribution to
age-related maculopathy in the Carotenoids in Age-Related Eye Disease Study
(CAREDS), an ancillary study of the Women’s Health Initiative
TL LaRowe1,2, JA Mares1,2, RB
Wallace,3 K Gehrs4, and R Chappell5 for the CAREDS Group of Investigators.
1University of Wisconsin-Madison,
Department of Ophthalmology & Visual Sciences
2 Interdepartmental Graduate Program in Nutritional Sciences
3 University
of Iowa, Department of Epidemiology
4 University
of Iowa, Department
