Cartus, Abigail
(2021)
Severe maternal morbidity: screening and long-term consequences.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
This is the latest version of this item.
Abstract
Severe maternal morbidity (SMM) is an important population indicator of maternal health and health care quality. SMM is often used as a proxy for maternal death in epidemiologic and quality improvement research. This is because it is much more common than maternal death (which is rare in absolute numbers), but shares the same risk factors and etiologies. However, there are two unresolved questions in severe maternal morbidity research, which this dissertation addresses. First, the long-term health consequences of SMM are not well understood. I investigated the association between SMM during the perinatal period and risk of adverse cardiovascular events (heart failure, ischemic heart disease, stroke/transient ischemic attack, and a composite of these three outcomes plus atrial fibrillation) up to 2 years postpartum among deliveries covered by Pennsylvania Medicaid, 2016-2018. I found that SMM is associated with increased risk of adverse cardiovascular events and that elevated risk persists past the traditional end of the postpartum period at 42 days post-delivery. Second, available methods to quantify SMM at the hospital or population level have serious limitations, e.g., identifying a large number of false-positive cases or requiring labor-intensive medical record abstraction, that I attempted to address using ensemble machine learning. To this end, I examined the impact of undersampling, one technique for remedying outcome class imbalance (where non-events outnumber events by a factor of 2:1 or more), on the predictive performance of ensemble machine learning algorithms (SuperLearner). We found that, in a simulated setting with moderate class imbalance, undersampling does not markedly improve the predictive performance of either logistic regression or SuperLearner. We then attempted to use SuperLearner as an alternative to existing screening criteria or medical record review to identify true-positive SMM from a sample of deliveries at Magee-Womens Hospital, 2013-2017. Our SuperLearner algorithms performed better than existing SMM screening criteria on some predictive performance metrics and worse on others, indicating that the choice of SMM screening method involves tradeoffs. This work contributes to improved understanding of maternal health in the United States and points to several future directions for SMM research.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
19 January 2021 |
| Date Type: |
Publication |
| Defense Date: |
23 November 2020 |
| Approval Date: |
19 January 2021 |
| Submission Date: |
30 November 2020 |
| Access Restriction: |
1 year -- Restrict access to University of Pittsburgh for a period of 1 year. |
| Number of Pages: |
147 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Epidemiology |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
severe maternal morbidity, machine learning, epidemiology |
| Date Deposited: |
19 Jan 2021 19:55 |
| Last Modified: |
19 Jan 2022 06:15 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/40020 |
Available Versions of this Item
-
Severe maternal morbidity: screening and long-term consequences. (deposited 19 Jan 2021 19:55)
[Currently Displayed]
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}