Pregnancy outcomes in females with hidradenitis suppurativa

A pregnancy datamart was constructed from Cerner Health Facts EMR data using pypreg, a python package developed by the author based on a validated pregnancy identification algorithm by Keran Moll. The demographics of the pregnancy datamart were consistent with the demographics of Health Facts overall. Various measures were compared against national data for the United States including rates of cesarean section, 42-day maternal mortality, severe maternal morbidity, preeclampsia, fetal growth restriction, gestational diabetes, and gestational hypertension. These measures were compared using Chi-square and Breslow-Day. No evidence suggested the rates observed in the datamart were substantially different from national figures. Thus, providing a degree of validity to the data. During construction of the datamart, several data quality problems were identified in Health Facts and were addressed. The quality issues included patient identifiers that are defined to be unique being linked to other unique patient identifiers, patient encounters with more than one date of admission, patient encounters being listed at more than one facility, and patient age not changing appropriately over time. A small percentage of the total encounters and patients were recommended to be removed from the general database to avoid unforeseen problems due to their inclusion in research. With the pregnancy datamart established, pregnancy outcomes were studied between the populations with and without hidradenitis suppurativa. Several covariates were selected based on association with inflammatory imbalance or association with condition for use in a multinomial logistic regression. This model type was selected to model all outcomes simultaneously to avoid the problem of multiple testing. Additionally, two models were considered. The first model treated each pregnancy as an independent event, while the second model only looked at the first record pregnancy for each patient. The odds ratios for negative pregnancy outcomes were elevated in both model types. When considering model predictions on counterfactual data and the estimated marginal means, several negative outcomes showed significant increases in risk.

Name thesis:degree_name

Ph.D. (Doctor of Philosophy)

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Computer Science (UMKC)

Year dc:date.issued

2024

Author dc:creator

• Walsh, David Paige, Jr.

Advisors dc:contributor.advisor

• Wyckoff, Gerald J.
• Wang, Shizhen, (Jeff)
• Lee, Yugyung, 1960-
• Monaghan-Nichols, Paula

Handle dc:identifier.uri

https://hdl.handle.net/10355/106958

OAI identifier oai:identifier

oai:mospace.umsystem.edu:10355/106958