I am a Professor in the Department of Biostatistics and Medical Informatics and the Department of Population Health Sciences in the School of Medicine and Public Health at the University of Wisconsin-Madison. I have an affiliate appointment in the Department of Statistics.
I grew up in Duluth, Minnesota. I graduated from East High School in 1988. I received a BA in Mathematics and Economics in 1992 from the University of Minnesota-Duluth and an MS in Statistics in 1994 and a PhD in Statistics with emphasis in Biostatistics in 1998 from the University of Wisconsin-Madison. My PhD advisor was Murray Clayton.
I was a research scientist in the Statistical Data Analysis Center(SDAC) in the Department of Biostatistics and Medical Informatics at the University of Wisconsin-Madison, 1998-2005. I joined the faculty with a joint appointment in the Department of Biostatistics and Medical Informatics and the Department of Population Health Sciences in 2005.
I am an applied biostatistician focusing on problems in clinical and epidemiologic research. Current methodologic areas of interest include (1) multi-state models for incidence, progression and regression of ocular (and other) diseases, (2) small area estimation problems, particularly ranking, (3) spatial and spatio-temporal modeling, particularly cluster detection and high-dimensional variable selection and (4) age-period-cohort modeling.
PhD in Statistics (emphasis in Biostatistics), 1998
University of Wisconsin-Madison
MS in Statistics, 1994
University of Wisconsin-Madison
BA in Mathematics and Economics, 1992
University of Minnesota-Duluth
In regression analysis for spatio-temporal data, identifying clusters of spatial units over time in a regression coefficient could provide insight into the unique relationship between a response and covariates in certain subdomains of space and time windows relative to the background in other parts of the spatial domain and the time period of interest. In this article, we propose a varying coefficient regression method for spatial data repeatedly sampled over time, with heterogeneity in regression coefficients across both space and over time. In particular, we extend a varying coefficient regression model for spatial-only data to spatio-temporal data with flexible temporal patterns. We consider the detection of a potential cylindrical cluster of regression coefficients based on testing whether the regression coefficient is the same or not over the entire spatial domain for each time point. For multiple clusters, we develop a sequential identification approach. We assess the power and identification of known clusters via a simulation study. Our proposed methodology is illustrated by the analysis of a cancer mortality dataset in the Southeast of the U.S.
It is fairly common to rank different geographic units, e.g. counties in the USA, based on health indices. In a typical application, point estimates of the health indices are obtained for each county, and the indices are then simply ranked as if they were known constants. Several authors have considered optimal rank estimators under squared error loss on the rank scale as a default method for general purpose ranking, e.g. situations where ranking units across the full spectrum of performance (low, medium, high) is important. While computationally convenient, squared error loss on the rank scale may not represent the true inferential goals of rank consumers. We construct alternative loss functions based on three components: (1) the inferential goal (rank position or pairwise comparisons), (2) the scale (original, log-transformed or rank) and (3) the (positional or pairwise) loss function (0/1, squared error or absolute error). We can obtain optimal ranks for loss functions based on rank positions and nearly optimal ranks for loss functions based on pairwise comparisons paired with highest posterior density (HPD) credible intervals. We compare inferences produced by the various ranking methods, both optimal and heuristic, using low birth weight data for counties in the Midwestern United States, from 2006 to 2012.
OBJECTIVE: Breast cancer simulation models must take changing mortality rates into account to evaluate the potential impact of cancer control interventions. We estimated mortality rates due to breast cancer and all other causes combined to determine their impact on overall mortality by year, age, and birth cohort. METHODS: Based on mortality rates from publicly available datasets, an age-period-cohort model was used to estimate the proportion of deaths due to breast cancer for US women aged 0 to 119 years, with birth years 1900 to 2000. Breast cancer mortality was calculated as all-cause mortality multiplied by the proportion of deaths due to breast cancer; other-cause mortality was the difference between all-cause and breast cancer mortality. RESULTS: Breast cancer and other-cause mortality rates were higher for older ages and birth cohorts. The percent of deaths due to breast cancer increased across birth cohorts from 1900 to 1940 then decreased. Among 50-year-old women, in the 1920 birth cohort, 52 (9.9%) of 100,000 deaths (95% CI, 9.8% to 10.1%) were attributed to breast cancer whereas 476 of 100,000 were due to other causes; in the 1960 birth cohort, 22 (8.5%) of 100,000 deaths (95% CI, 8.3% to 8.7%) were attributed to breast cancer with 242 of 100,000 deaths due to other causes. The percentage of all deaths due to breast cancer was highest (4.1% to 12.9%) for women in their 40s and 50s for all birth cohorts. CONCLUSIONS: This study offers evidence that advances in breast cancer screening and treatment have reduced breast cancer mortality for women across the age spectrum, and provides estimates of age-, year- and birth cohort-specific competing mortality rates for simulation models. Other-cause mortality estimates are important in these models because most women die from causes other than breast cancer.
PURPOSE: To assess the association between geographic access to mammography facilities and women’s mammography utilization frequency. METHODS: Using data from the population-based 1995-2007 Wisconsin Women’s Health study, we used proportional odds and logistic regression to test whether driving times to mammography facilities and the number of mammography facilities within 10 km of women’s homes were associated with mammography frequency among women aged 50-74 years and whether associations differed between Rural-Urban Commuting Areas and income and education groups. RESULTS: We found evidence for nonlinear relationships between geographic access and mammography utilization (nonlinear effects of driving times and facility density, P-values .01 and .005, respectively). Having at least one nearby mammography facility was associated with greater mammography frequency among urban women (1 vs. 0 facilities, odds ratio 1.26, 95% confidence interval, 1.09-1.47), with similar effects among rural women. Adding more facilities had decreasing marginal effects. Long driving times tended to be associated with lower mammography frequency. We found no effect modification by income, education, or urbanicity. In rural settings, mammography nonuse was higher, facility density smaller, and driving times to facilities were longer. CONCLUSIONS: Having at least one mammography facility near one’s home may increase mammography utilization, with decreasing effects per each additional facility.
PURPOSE: To develop a composite Cancer Burden Index and produce 95% confidence intervals (CIs) as measures of uncertainties for the index. METHODS: The Kentucky Cancer Registry has developed a cancer burden Rank Sum Index (RSI) to guide statewide comprehensive cancer control activities. However, lack of interval estimates for RSI limits its applications. RSI also weights individual measures with little inherent variability equally as ones with large variability. To address these issues, a Modified Sum Index (MSI) was developed to take into account of magnitudes of observed values. A simulation approach was used to generate individual and simultaneous 95% CIs for the rank MSI. An uncertainty measure was also calculated. RESULTS: At the Area Development Districts (ADDs) level, the ranks of the RSI and the MSI were almost identical, while larger variation was found at the county level. The widths of the CIs at the ADD level were considerably shorter than those at the county level. CONCLUSION: The measures developed for estimating composite cancer burden indices and the simulated CIs provide valuable information to guide cancer prevention and control effort. Caution should be taken when interpreting ranks from small population geographic units where the CIs for the ranks overlap considerably.
In regression analysis for spatio-temporal data, identifying clusters of spatial units over time in a regression coefficient could provide insight into the unique relationship between a response and covariates in certain subdomains of space and time windows relative to the background in other parts of the spatial domain and the time period of interest. In this article, we propose a varying coefficient regression method for spatial data repeatedly sampled over time, with heterogeneity in regression coefficients across both space and over time. In particular, we extend a varying coefficient regression model for spatial-only data to spatio-temporal data with flexible temporal patterns. We consider the detection of a potential cylindrical cluster of regression coefficients based on testing whether the regression coefficient is the same or not over the entire spatial domain for each time point. For multiple clusters, we develop a sequential identification approach. We assess the power and identification of known clusters via a simulation study. Our proposed methodology is illustrated by the analysis of a cancer mortality dataset in the Southeast of the U.S.
Purpose: To evaluate the relationship of retinal layer thickness with age and age-related macular degeneration (AMD) in the Carotenoids in Age-Related Eye Disease Study 2. Methods: Total retinal thickness within the macular area, and individual layer thickness was determined for CAREDS2 participants (n=906 eyes, 473 women) from the Women’s Health Initiative using Heidelberg spectral-domain optical coherence tomography (SD-OCT). Mean measurements within the OCT grid were compared across age tertiles (69-78, 78-83, 83-101 years) and AMD outcomes (no AMD, early, intermediate , late AMD). Results: Total retinal thickness in the central circle, inner ring, and outer ring were mean ± standard deviation 277 ± 34 μm, 326 ± 20 μm, and 282 ± 15 μm, respectively. Thickness did not vary by age in the central circle, but decreased with age in the inner and outer circles (p≤0.004). Specifically, ganglion cell (GCL), inner plexiform (IPL), and outer nuclear (ONL) layer thickness decreased with age in the inner and outer rings (p≤0.003). Age-adjusted retinal thickness in all three circles did not vary by AMD outcomes, except that the retinal pigment epithelium (RPE) layer thickness was greatest in eyes with late AMD (p≤0.001). After controlling for age, higher retinal ONL and lower RPE thickness were associated with better best corrected visual acuity. Conclusions: In this cohort of older women, a decrease in peripheral macular thickness was associated with increasing age, particularly in the GCL, IPL, and ONL. The GCL, PRL, and RPE layer contributed to variability in thickness in eyes with AMD. Among all retinal layers, ONL and RPE thickness were associated with visual acuity.
Rationale: Rhinovirus C (RV-C) can cause asymptomatic infection and respiratory illnesses ranging from the common cold to severe wheezing. Objectives: To identify how age and other individual-level factors are associated with susceptibility to RV-C illnesses. Methods: Longitudinal data from the Childhood Origins of ASThma (COAST) birth cohort study were analyzed to determine relationships between age and RV-C infections. Neutralizing antibodies specific for rhinovirus A (RV-A) and RV-C (3 types each) were determined using a novel polymerase chain reaction-based assay. We pooled data from 14 study cohorts in the United States, Finland, and Australia and used mixed-effects logistic regression to identify factors related to the proportion of RV-C versus RV-A detection. Measurements and Main Results: In COAST, RV-A and RV-C infections were similarly common in infancy, while RV-C was detected much less often than RV-A during both respiratory illnesses and scheduled surveillance visits (p<0.001, chi-square) in older children. The prevalence of neutralizing antibodies to RV-A or RV-C types was low (5%-27%) at age 2 years, but by age 16, RV-C seropositivity was more prevalent (78% vs. 18% for RV-A, p<0.0001). In the pooled analysis, the RV-C to RV-A detection ratio during illnesses was significantly related to age (p<0.0001), CDHR3 genotype (p<0.05), and wheezing illnesses (p<0.05). Furthermore, certain RV types (e.g., C2, C11, A78, A12) were consistently more virulent and prevalent over time. Discussion: Knowledge of prevalent RV types, antibody responses, and populations at risk based on age and genetics may guide the development of vaccines or other novel therapies against this important respiratory pathogen.
Objectives: We examined the association of exposure to maternal depression when a child is age 3 with future child problem behavior and investigated whether race/ethnicity is a moderator of this relationship. Methods: We used Fragile Families and Child Well-Being Study data (age 3 N=3288 and 49% Black, 26% Hispanic, 22% non-Hispanic White; age 5 N=3001 and 51% Black, 25% Hispanic, 21% non-Hispanic White; age 9 N=3630 and 50% Black, 25% Hispanic, 21% non-Hispanic White) and ordinal logistic regression to model problem behavior at ages 3, 5, and 9 on maternal depression status at age 3. Results: At age 9, children whose mother was depressed when the child was age 3 were significantly more likely to have higher internalizing (AOR=1.92, 95% CI: 1.42,2.61) and externalizing (AOR=1.65, 95% CI: 1.10,2.48) problem behavior scores. Race/ethnicity did not have moderating effects, potentially due to a limitation of the data that required use of maternal self-reported race/ethnicity as a proxy for child race/ethnicity. Conclusions for Practice: Exposure to maternal depression after the prenatal and perinatal periods may have a negative association with children’s behavioral development through age 9. Interventions that directly target maternal depression during this time should be developed. Additional research is needed to further elucidate the role of race/ethnicity in the relationship between maternal depression and child problem behavior.
Background: Studies indicate that the nasal microbiome may correlate strongly with the presence or future risk of childhood asthma. Objectives: In this study, we tested whether developmental trajectories of the nasopharyngeal microbiome in early life and the composition of the microbiome during illnesses were related to risk of childhood asthma. Methods: Children participating in the Childhood Origins of Asthma study (n=285) provided nasopharyngeal mucus samples in the first two years of life, during routine healthy study visits (2, 4, 6, 9, 12, 18 and 24 months of age) and episodes of respiratory illnesses, which were analyzed for respiratory viruses and bacteria. We identified developmental trajectories of early-life microbiome composition, as well as predominant bacteria during respiratory illnesses, and correlated these with presence of asthma at 6, 8, 11, 13 and 18 years of age. Results: Of the four microbiome trajectories identified, a Staphylococcus-dominant microbiome in the first 6 months of life was associated with increased risk of recurrent wheezing by age 3 years and asthma that persisted throughout childhood. In addition, this trajectory was associated with the early onset of allergic sensitization. During wheezing illnesses, detection of rhinoviruses and predominance of Moraxella were associated with asthma that persisted throughout later childhood. Conclusion: In infancy, the developmental composition of the microbiome during healthy periods and the predominant microbes during acute wheezing illnesses are both associated with the subsequent risk of developing persistent childhood asthma.