Introduction
Oregon Saludable: Juntos Podemos
Methods
Site Location Identification Through Optimization
Assumptions for Objective Selection
Specification of Algorithmic Function
Model Parameterization
Leveraging Algorithm-Provided Sites in the Community Partner Site Selection Process
This model focuses entirely on geographic locations, and has no knowledge of access limitations, or other factors that would impact site selection. Sites nearby each proposal should also be considered as alternatives, as they may provide practical benefits with negligible impact on the optimization objective. It is also important to note that the geographic center of census tracts may not accurately represent the population centers.
SARS-CoV-2 Testing Events
Predictive Validity Evaluation
Measures
Analytic Strategy
Results
Community Partner Feedback on Algorithm Proposed Sites
Sites | Count | |
---|---|---|
Proposal used as site | C2_4, C3_1, C8_2 | 3 (9%) |
Proposals with sites located nearby with a 5-min drive | C1_2, C1_5, C2_5, C2_1, C2_2, C2_6, C3_2, C4_1, C4_2, C6_1, C6_2, C7_3, C8_1, C9_1, C9_4, C9_6 | 16 (48%) |
Proposals with sites located nearby with a 5–10-min drive | C1_4 C2_3, C7_ | 3 (9%) |
Proposals with sites with > 10-min drive | C1_3, C4_4, C7_2, C7_1, C7_4 | 5 (15%) |
Rejected proposals | N/A | 6 (18%) |
County | Reasons for moving sites |
---|---|
1 | The CBO and CPH provided site recommendations in the communities identified by the algorithm. Reasons for not using proposals included sites being unwilling to hostesting or insufficient facilities. Two local Catholic churches who collaborated closely with OSJP’s partner CBO and the local school districts provided the final sites. |
2 | The CBO and CPH provided site recommendations in the communities identified by the algorithm. One proposal was retained, but was located in a community where the CBO was uncomfortable operating. Reasons for site changes included strong CBO relationships at other sites and feedback that proposals were not frequented by community members. We partnered with local schools and one Catholic church. |
3 | At the beginning of the project, CPH had agreed to collaborate on providing testing services, however, became unresponsive. Furthermore, the CBO communicated that they would not support events as sufficient testing existed in the county. We selected one site at a Catholic church to coincide with the end of Spanish mass. |
4 | We were unable to identify a CBO partner in County 4. The CPH was a willing collaborator but had a lack of knowledge about the Latinx community. Reasons for site changes included insufficient facilities and two churches that were unwilling to provide on-site testing. One clinic was adopted by the suggestion of the CPH. Other sites were selected with the local school districts and a park proximal to a proposal. |
5 | Members of the research study team were in County 5 and had strong working knowledge of the community and strong collaboration with the local CBO and CPH Latinx outreach team. Sites were pre-determined by these partners without algorithm use. Sites were at schools and Catholic churches with Spanish mass. |
6 | The CBO partner and CPH in County 6 provided site recommendations in the communities identified by the algorithm. Reasons for not using site-specific proposals included insufficient facilities and inability to determine whether an operable site existed. Final sites were a community center recommended by the local CBO and a school that was contacted after a CBO proposed church was unable to collaborate. |
7 | The CBO and CPH provided limited site recommendations in the communities identified by the algorithm. Our study team identified that the proposals had either insufficient facilities or were at sites that were not interested in collaborating. Final sites were comprised of schools and Catholic churches with Spanish mass. |
8 | The CPH provided site recommendations in the communities identified by the algorithm. One proposal at a school was retained. One proposal was rejected because of access difficulties, and was described by partners as a highly rural, White community. The remaining proposal was replaced by a proximal site. |
9 | The CBO and CPH provided specific site recommendations. Additionally, a study team member had strong ties to the Latinx community. Only some sites were in the communities identified by the algorithm. Reasons for site changes included insufficient facilities and inaccessibility. Final sites were comprised of one school, a community college, four places of employment, and one community events center. |
Analysis of Testing Utilization by Drive Time
Fixed effect | IRR | [0.78 – 1.55] |
Average drive time | 0.66*** | [0.55 – 0.79] |
Latinx populace | 1.27* | [1.01 – 1.60] |
Nativity | 0.87 | [0.72 – 1.06] |
Income inequality | 0.93 | [0.79 – 1.10] |
Median age | 1.20* | [1.00 – 1.44] |
Vaccine event | 2.35*** | [1.64 – 3.36] |
Promotores event | 2.14*** | [1.45 – 3.16] |
Other partner events | 1.31 | [0.76 – 2.24] |
Mexican Consulate event | 8.09*** | [3.57 – 18.37] |
Time varying covariate | ||
Lag county cases PC | 2.00*** | 1.60 – 2.48 |
Lag vax coverage | 0.87 | 0.74 – 1.02 |
Model fit | ||
Log likelihood | − 891.16 | |
AIC | 1822.32 | |
BIC | 1862.82 |