Matlack, Meghan
(2022)
Municipality Level Dengue Risk Prediction Modeling in Brazil and its Impacts for Future Public Health Interventions.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Current prediction models for dengue risk are restricted to country-wide estimates or are insufficient at accounting for localized variations in outbreak risk. These models focus primarily on climate and large-scale factors that may not reflect true risk across municipalities or neighborhoods, and do not account for other determinants of health that have also been previously correlated with risk of dengue. We hypothesized that widespread municipality-level dengue outbreak forecasting would have the potential to better capture small-scale transmission dynamics and provide more accurate local outbreak predictions. We built several boosted regression tree (BRT) models to predict outbreak risk for 167 municipalities in Pernambuco, Brazil. Using training data from 2010-2015, we utilized climate, prior surveillance data, and social and environmental indicators of health as features for classification of regions as low or high risk for dengue in 2016. Our best model included real-time temperature and precipitation, lagged climate effects and prior surveillance data. This model gave a training AUC of 0.963 and a testing AUC of 0.939, with a total of 1,842 correct observations from 2,004 predictions. Additionally, this model successfully predicted 74.8% of high risk classifications, a marked improvement from previous iterations. Quantification of predictor associations through univariate and multivariate regression analyses revealed correlations fairly consistent with our BRT results. In general, we saw that inclusion of most socio-environmental predictors had minor influence over BRT predictions, and were not statistically significant in correlation with increased dengue risk when compared to other predictors. We can conclude from this dissertation research that a BRT approach is effective for modeling dengue transmission dynamics and can successfully predict high risk dengue regions using relevant climatic factors as well as prior case data, though more research is needed to establish strong socio-environmental patterns with small-scale dengue outbreak risk. Predictive models can serve as in-depth complementary tools to current dengue surveillance systems by providing useful information on upcoming outbreaks and by estimating where most cases are most likely to occur prior to peak transmission. Further development of these models can also provide the insight necessary to restructure current vector control policies and strengthen existing dengue intervention practices.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
10 May 2022 |
| Date Type: |
Publication |
| Defense Date: |
14 April 2022 |
| Approval Date: |
10 May 2022 |
| Submission Date: |
29 April 2022 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
155 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Environmental and Occupational Health |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
infectious disease, vector-borne disease, predictive modeling, climate change, public health informatics, environmental health |
| Related URLs: |
|
| Date Deposited: |
10 May 2022 18:46 |
| Last Modified: |
10 May 2022 18:46 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/42837 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}