About
Andrea J. Cook, Ph.D., is a Senior Investigator in the Biostatistics Unit at Kaiser Permanente Washington Health Research Institute (GHRI) and an Affiliate Professor in the Department of Biostatistics at the University of Washington. At GHRI, she is involved in seven studies, including four clinical trials focused on alternative medicine, cardiovascular care, and medical error disclosure, as well as three observational studies examining obesity's relationship to food and the environment, the CDC's Vaccine Safety Datalink, and the FDA's Mini-Sentinel project on post-surveillance monitoring of vaccine and drug safety. Her current research includes statistical methodology and application for spatial cluster detection and methods for group sequential monitoring of post-marketing surveillance data. Dr. Cook completed her Ph.D. in Biostatistics at Harvard University in June 2005 under the supervision of Dr. Yi Li. Her dissertation encompassed broad topics such as spatial statistics, survival analysis, and longitudinal data analysis. Prior to her doctoral studies, she earned a Bachelor of Science degree in Applied and Computational Mathematical Sciences (ACMS) from the University of Washington.
Research topics
- Medicine
- Internal medicine
- Demography
- Political Science
- Surgery
- Environmental health
- Computer Science
- Gerontology
- Public relations
- Library science
- Biology
- General surgery
- Psychiatry
- Food science
- Physics
Selected publications
Statistics in Medicine · 2026-02-27
article1st authorCorrespondingRisk differences allow decision makers to easily estimate the excess safety risk associated with a medical product relative to the potential benefits. However, in post-market observational surveillance studies that actively monitor (e.g., sequentially over time) for safety risk of new medical products, available methods target a relative measure (e.g., odds ratio and relative risk), which can be especially unstable in the rare event setting. These studies are typically conducted within distributed healthcare networks (e.g., Food and Drug Administration [FDA] Sentinel and Centers for Disease Control [CDC] Vaccine Safety Datalink) with patient-level data protected behind firewalls, but sharing of aggregate, deidentified data for centralized analyses. We propose an inverse probability of treatment weighting (IPTW) method that uses site-specific propensity scores to estimate site-specific risk differences that are combined to create an overall stratified risk difference estimate. This method is tailored to the rare event setting and requires minimal data sharing. The stratified IPTW approach is then extended to the active post-market surveillance setting by incorporating group sequential monitoring boundaries using a novel permutation approach. A simulation study is conducted to evaluate the performance of the new methods relative to two centralized analysis approaches, and the methods are applied to a safety surveillance study comparing the risk of febrile seizure between two vaccines using FDA Sentinel Data from three healthcare organizations.
Integrative Medicine Research · 2026-04-01
articleOpen accessSenior authorAcupuncture has been shown to be an effective treatment for chronic low back pain (cLBP). The BackInAction trial was conducted to inform CMS/Medicare acupuncture coverage decision specifically for older adults with cLBP. CMS coverage of acupuncture for cLBP among older adults specifies dose parameters despite gaps in research of optimal session-dose for such treatment. This is a secondary analysis of data from the BackInAction trial to report fidelity of the acupuncture intervention and to explore the association between the number of acupuncture sessions received (session-dose) and reductions in functional disability (RMDQ), pain intensity, and on a composite pain interference/intensity measure (PEG) at 3-months post-randomization. The majority of acupuncture recipients (82%) received the minimum clinical threshold of ≥8 sessions in the first three months with almost half meeting or exceeding the CMS 12-session threshold. Higher session-dose was associated with significant disability-related reductions and improvements on other pain outcomes. Pain intensity was better reduced with 12 or more sessions than with 8-11 sessions. These positive outcomes were found to increase monotonically with increasing session-dose. These findings provide support for 12-sessions within three months being more effective for reducing pain and functional disability than a dose of less than 8 sessions. Overall, study findings suggest that for those 65 and older with cLBP optimal outcomes may require at least 8 but preferably 12 or more acupuncture sessions. ClinicalTrials.gov Identifier: NCT04982315
BMC Medical Research Methodology · 2026-01-12
articleOpen accessEvaluating heterogeneity of treatment effects (HTE) across subgroups is common in both randomized trials and observational studies. Although several statistical challenges of HTE analyses including low statistical power and multiple comparisons are widely acknowledged, issues specific to clustered data, including cluster randomized trials (CRTs), have received less attention. For testing interactions in linear mixed-effects models (LMM), Barr et al. (2013) suggested that: random slopes for interaction terms should be studied. In this paper, we explore the impact of model misspecification, including generalized LMM (GLMM) with or without random slopes, and provide recommendations for conducting inference for HTE across subgroups in CRTs. We conducted a simulation study to evaluate the performance of common analytic approaches for testing the presence of HTE for continuous, binary, and count outcomes: generalized linear mixed models (GLMM) and generalized estimating equations (GEE) including interaction terms between treatment and subgroup. Several simulation scenarios covered broad range of scenarios in CRTs, for example, small to a large number of clusters, small to moderate cluster-specific random slopes for subgroup. The performance metric was the empirical type I error rate compared to a nominal level. We applied the analytical methods to a real-world CRT using the count outcome utilization of healthcare from the motivating Primary Care Opioid Use Disorder treatment (PROUD) trial. We found that standard GLMM analyses that assume a common correlation of participants within clusters can lead to severely elevated type 1 error rates of up to 47.2% compared to the 5% nominal level if the within-cluster correlation varies across subgroups. A maximal GLMM, which allows subgroup-specific within-cluster correlations, achieved the nominal type 1 error rate, as did GEE (though rates were slightly elevated even with as many as 50 clusters). Applying the methods to the real-world CRT, we found a large impact of the model specification on inference. We recommend that HTE analyses using the maximal GLMM account for within-subgroup correlation to avoid anti-conservative inference. For Wald t-testing of HTE in small sample clusters, appropriate small sample correction methods should be considered based on the outcome data type.
Pain · 2026-04-15
article1st authorCorrespondingABSTRACT: Remote cognitive behavioral therapy-based interventions for chronic pain (CBT-CP-based) have consistently shown modest pain benefits. With remote treatment availability accelerating and increasingly tight healthcare resources, clinical decision-makers need to understand when working with a therapist may be beneficial vs a largely self-directed online program. Using data from a large-scale 3-arm pragmatic trial, we assessed the role of patient session adherence and potential moderators of intervention pain severity effectiveness for 2 remote CBT-CP-based approaches (online program [painTRAINER] and therapist-delivered telephone or video program [health coach]) and usual medical care. Of the prespecified moderators (demographics [sex, age, race/ethnicity, rurality], social determinants of health, and clinical variables [comorbid depression and/or anxiety, multiple types of chronic pain]), the health coach program was more effective than painTRAINER at 3 months among men and for those who screened positive for depression. No other factors moderated intervention pain severity effects at 3 or 12 months. The health coach program had higher session adherence than painTRAINER (70.4% vs 47.8%). Among intervention completers, pain severity outcomes were similar between CBT-CP-based interventions (Adjusted relative risk [95% confidence interval]: 0.99 [0.85-1.16] and 0.93 [0.82-1.05] at 3 and 12 months, respectively), suggesting that both interventions may be helpful for those with a variety of demographic and clinical characteristics if adherence is achieved. Participant engagement is critical in optimizing outcomes for online programs, but these findings suggest flexibility in the specific modality for delivering remote CBT-CP based on patient preference and healthcare system capacity that may enhance scalability and patient access to care.
Open MIND · 2026-01-01
articleSupplementary Material 1.
Figshare · 2026-01-01
otherOpen accessAbstract Background/Aims Evaluating heterogeneity of treatment effects (HTE) across subgroups is common in both randomized trials and observational studies. Although several statistical challenges of HTE analyses including low statistical power and multiple comparisons are widely acknowledged, issues specific to clustered data, including cluster randomized trials (CRTs), have received less attention. For testing interactions in linear mixed-effects models (LMM), Barr et al. (2013) suggested that: random slopes for interaction terms should be studied. In this paper, we explore the impact of model misspecification, including generalized LMM (GLMM) with or without random slopes, and provide recommendations for conducting inference for HTE across subgroups in CRTs. Methods We conducted a simulation study to evaluate the performance of common analytic approaches for testing the presence of HTE for continuous, binary, and count outcomes: generalized linear mixed models (GLMM) and generalized estimating equations (GEE) including interaction terms between treatment and subgroup. Several simulation scenarios covered broad range of scenarios in CRTs, for example, small to a large number of clusters, small to moderate cluster-specific random slopes for subgroup. The performance metric was the empirical type I error rate compared to a nominal level. We applied the analytical methods to a real-world CRT using the count outcome utilization of healthcare from the motivating Primary Care Opioid Use Disorder treatment (PROUD) trial. Results We found that standard GLMM analyses that assume a common correlation of participants within clusters can lead to severely elevated type 1 error rates of up to 47.2% compared to the 5% nominal level if the within-cluster correlation varies across subgroups. A maximal GLMM, which allows subgroup-specific within-cluster correlations, achieved the nominal type 1 error rate, as did GEE (though rates were slightly elevated even with as many as 50 clusters). Applying the methods to the real-world CRT, we found a large impact of the model specification on inference. Conclusions We recommend that HTE analyses using the maximal GLMM account for within-subgroup correlation to avoid anti-conservative inference. For Wald t-testing of HTE in small sample clusters, appropriate small sample correction methods should be considered based on the outcome data type.
Figshare · 2026-01-01
articleOpen accessSupplementary Material 1.
Figshare · 2026-01-01
otherOpen accessAbstract Background/Aims Evaluating heterogeneity of treatment effects (HTE) across subgroups is common in both randomized trials and observational studies. Although several statistical challenges of HTE analyses including low statistical power and multiple comparisons are widely acknowledged, issues specific to clustered data, including cluster randomized trials (CRTs), have received less attention. For testing interactions in linear mixed-effects models (LMM), Barr et al. (2013) suggested that: random slopes for interaction terms should be studied. In this paper, we explore the impact of model misspecification, including generalized LMM (GLMM) with or without random slopes, and provide recommendations for conducting inference for HTE across subgroups in CRTs. Methods We conducted a simulation study to evaluate the performance of common analytic approaches for testing the presence of HTE for continuous, binary, and count outcomes: generalized linear mixed models (GLMM) and generalized estimating equations (GEE) including interaction terms between treatment and subgroup. Several simulation scenarios covered broad range of scenarios in CRTs, for example, small to a large number of clusters, small to moderate cluster-specific random slopes for subgroup. The performance metric was the empirical type I error rate compared to a nominal level. We applied the analytical methods to a real-world CRT using the count outcome utilization of healthcare from the motivating Primary Care Opioid Use Disorder treatment (PROUD) trial. Results We found that standard GLMM analyses that assume a common correlation of participants within clusters can lead to severely elevated type 1 error rates of up to 47.2% compared to the 5% nominal level if the within-cluster correlation varies across subgroups. A maximal GLMM, which allows subgroup-specific within-cluster correlations, achieved the nominal type 1 error rate, as did GEE (though rates were slightly elevated even with as many as 50 clusters). Applying the methods to the real-world CRT, we found a large impact of the model specification on inference. Conclusions We recommend that HTE analyses using the maximal GLMM account for within-subgroup correlation to avoid anti-conservative inference. For Wald t-testing of HTE in small sample clusters, appropriate small sample correction methods should be considered based on the outcome data type.
2025-03-07
book-chapter1st authorCorresponding2025-03-07
book-chapter1st authorCorresponding
Recent grants
Pragmatic Trial of Acupuncture for Chronic Low Back Pain in Older Adults
NIH · $9.1M · 2019–2025
Frequent coauthors
- 104 shared
Diana L. Miglioretti
- 77 shared
Karen J. Sherman
Kaiser Permanente Washington Health Research Institute
- 68 shared
Robert Wellman
Kaiser Permanente Washington Health Research Institute
- 64 shared
Daniel C. Cherkin
Kaiser Permanente Washington Health Research Institute
- 54 shared
Beverly B. Green
Kaiser Permanente
- 54 shared
Jennifer C. Nelson
- 45 shared
Diana S. M. Buist
Menlo School
- 42 shared
Karla Kerlikowske
San Francisco VA Health Care System
Education
- 2005
PhD, Biostatistics
Harvard University
- 2003
MS, Biostatistics
Harvard School of Public Health
- 2001
BS, Applied Computational Mathematical Sciences
University of Washington
Awards & honors
- Featured Faculty and alumni elected 2026 Fellows of the Amer…
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Andrea Cook
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup