Files and links (1)
Schwartz_dissertation983.35 kB
Embargoed Access, Embargo ends: 06/29/2028
Dissertation
Understanding the epidemiology of Lyme disease through multiple data sources
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Spring 2026
DOI: 10.25820/etd.008443
Abstract
Background: Lyme disease (LD) is the most commonly reported vector-borne disease in the United States. The number of cases continues to grow, and risk is expanding outward from historically endemic areas, highlighting the importance of surveillance. While entomological surveillance provides important information regarding upstream risk, human surveillance data provides information regarding disease incidence and clinical features of illness. These data streams are essential for a comprehensive LD surveillance system and can be utilized for evaluation of interventions and risk communication. Unfortunately, it is not well understood how entomological measures translate into human risk at small spatial scales, complicating evaluation of environmental interventions and risk communication. Additionally, traditional surveillance is burdensome to conduct and changing surveillance practices over time have complicated comparison across states and years. Using claims data to supplement traditional surveillance data is a less burdensome alternative; however, validation studies of these methods have been limited. The primary objectives of this dissertation were to advance LD surveillance methodology and bridge these identified gaps using alternative methods to traditional surveillance.
Methods: To explore the relationship between entomological risk measures and LD risk, we repurposed data collected from a previously conducted reservoir targeted vaccine study with the goal of exploring the relationship between a tick infection density measure and LD risk using logistic regression. To evaluate use of claims data algorithms to supplement surveillance, we utilized an existing electronic health record data source in Iowa, a state with emerging LD incidence. Our goal was to first determine the positive predictive value (PPV) of an outpatient LD algorithm applied to University of Iowa Health Care data through chart abstractions of a sample of 100 LD diagnosis events. We then evaluated a disseminated LD algorithm to evaluate its ability to classify LD diagnoses into disseminated and non-disseminated. Chart abstractions were performed and PPVs and negative predictive values (NPV) were calculated.
Results: With respect to our evaluation of the relationship between entomological measures and LD risk, we found that mean site-specific coefficients varied across sites with coefficients ranging from -0.90 to 0.47. With respect to evaluation of the LD algorithm applied to UIHC data, we found a broad PPV which included confirmed, probable, and suspected cases of 80.0% (95% CI: 70.8-87.3) and a narrow PPV which included confirmed and probable cases of 41.0% (95%CI: 31.3-51.3). With respect to our evaluation of a disseminated algorithm, we found an overall broad PPV of 72.6% (95% CI:61.6-82.8) and a narrow PPV of 67.7% (95% CI: 56.5-78.7). We found an overall NPV of 79.0% (95% CI:67.9-89.3).
Conclusions: This dissertation explored how complementary alternative surveillance approaches could be utilized to fill current gaps left by current practices. Although we did not identify a consistent relationship between an entomological risk measure and LD risk at small spatial scales, these results hint at additional complexity in the relationship, which should be explored further. We evaluated the performance of an outpatient LD claims algorithm in Iowa and found that it performed reasonably well. Additionally, we found that the disseminated algorithm can identify disseminated LDDEs from claims data with moderate accuracy. These results indicate that claims data could be used to supplement traditional surveillance practices in Iowa.
Details
- Title: Subtitle
- Understanding the epidemiology of Lyme disease through multiple data sources
- Creators
- Amy Schwartz
- Contributors
- Ryan Carnahan (Advisor)Christine Petersen (Advisor)Natoshia Askelson (Committee Member)Grant Brown (Committee Member)Jonathan Platt (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Epidemiology
- Date degree season
- Spring 2026
- DOI
- 10.25820/etd.008443
- Publisher
- University of Iowa
- Number of pages
- xii, 118 pages
- Copyright
- Copyright 2026 Amy Schwartz
- Language
- English
- Date submitted
- 04/20/2026
- Description illustrations
- Illustrations, graphs, charts, tables
- Description bibliographic
- Includes bibliographical references (pages 111-118).
- Public Abstract (ETD)
Lyme disease (LD) is the most common vector-borne disease in the United States, and risk is expanding into new areas. The growing burden of LD highlights the importance of monitoring trends in new cases and risk. However, our ability to monitor LD is hindered by some challenges. First, we do not have a good understanding of how the number of infected ticks on small properties is related to human LD risk. Second, traditional methods to track LD are extremely burdensome, which has led to changing practices over time. Healthcare claims data could supplement these traditional methods. However, only a few groups have evaluated these claims-based methods.
To address these challenges, we first used ecology data to study the relationship between the number of infected ticks on individual properties and the number of newly infected dogs. The included dogs lived and exercised on the properties where ecology data was collected. Next, we evaluated an LD claims algorithm and a disseminated LD claims algorithm to determine their accuracy in Iowa.
We did not identify a consistent relationship between a measure of infected ticks and incident LD in dogs. We found that both algorithms had moderate accuracy.
Our findings suggest that the relationship between the number of infected ticks and LD risk may be complicated by other factors related to individual property characteristics. Our results also support the use of claims algorithms to supplement LD monitoring in Iowa; however, it is important to be aware of the accuracy and limitations of these methods.
- Academic Unit
- Epidemiology
- Record Identifier
- 9985177273502771
Metrics
1 Record Views