0%
Dr. Brenda McGrath is a biostatistician at OCHIN. She is on the OCHIN Research Data Science and Research Analytics team and co-leads the Quantitative Science Core.
Brenda is committed to fostering interdisciplinary collaborations by providing comprehensive biostatistical support throughout all stages of the scientific process. Her central mission is to advance health equity by providing biostatistical expertise for research projects led by OCHIN or in partnership with external organizations.
At OCHIN, Brenda has served as site principal investigator, co-investigator, and biostatistician on several investigations. She supports proposal development of new grant applications and advises on queries related to feasibility, experimental design, and statistical approach, thereby enabling her to apply her specific expertise in the use of real-world data for health research. Additionally, Brenda provides mentorship to a team of research analysts and other investigators on statistical methodology.
Brenda’s primary interests are in harnessing electronic health records for population-based research, with a focus on causal inference, multi-level analyses, survival analysis, and cluster-randomized trial designs. She is passionate about team science and uses her expertise as part of the larger scientific team to propel project execution. Brenda has engaged in research across diverse substantive areas, such as health care quality assessment, cardiometabolic diseases, healthy aging, and implementation of clinical decision support in community health centers.
OCHIN electronic health records include critically important data for health in underrepresented populations that experience health care inequity. Brenda’s role at OCHIN includes providing insights for investigators related to the appropriateness of these real-world data for scientific inquiry. Moreover, Brenda has provided biostatistical support on investigations that have utilized OCHIN electronic health records to help advance health equity. For example, her collaborative contributions to this work included multilevel models and data visualization. These investigations include topics such as community health center response to COVID-19, cardiovascular disease, and equitable access to health care. Electronic health records are not created for research, and thus, many complexities must be considered when using them for such purposes. While working with one of the largest and most robust electronic health records available, from the Veterans Health Administration (VHA), Brenda developed expertise in the analysis of electronic health records and has disseminated methodologies for electronic health record data. For example, in a study examining the differential use of LOINC codes to identify laboratory measures across the VHA hospitals over time, they found that their use was inconsistent and could not be relied on for complete data extraction. Brenda and team also developed an algorithm to process hospital visits that involved multiple unit stays, transfers, or out-of-VHA care into an extended hospitalization. The findings from these suitability studies highlight the intricacies of electronic health records and the importance of quality assurances when analyzing these data. Brenda has demonstrated experience using multi-level (or mixed effects) modeling to properly account for the nesting of patients in health care facilities or repeated measures on individuals. In a study examining both the hospital-level variation in statin discontinuation and new antipsychotic use after hospitalization, a multi-level model with hospital random effects was used and showed that there was a significant variation across hospitals. Similarly, another multi-level analysis was used to create a risk adjustment model with patient-level fixed effects and hospital-level random effects to examine hospitals with both significantly higher and lower risk-adjusted mortality rates. Another study examining the impact of decreased handgrip strength on activities of daily living limitations and time to mortality used mixed effects models to account for the repeated measures on individuals over multiple years. Although there are many benefits to using large-scale databases for observational studies, there are some limitations to consider. In particular, when assessing outcomes or differential treatment effects in observational studies, there is a strong potential for bias because individuals were not randomly allocated to treatment groups. One of the statistical approaches to account for this challenge is to use propensity score matching or weighting for attempting to replicate a randomized experiment. Brenda has documented expertise in using propensity score matching and other causal inference methods in epidemiological studies.
Research using OCHIN electronic health record data
Suitability of using electronic health records for clinical research
Multi-level modeling
Causal inference for epidemiological studies
OCHIN Connections is a monthly newsletter featuring the latest OCHIN news and perspectives supporting our mission to drive health equity.