Snowball
An integrated cloud platform for respondent-driven sampling
The project
Facing uncontained spread of the SARS-CoV-2 infection in North Carolina, the Snowball study was designed around the urgent need to diagnose cases, calculate the prevalence of the disease, and understand its spread or distribution. Simple random sampling techniques typically used to ascertain distribution and prevalence of disease were not practically feasible due to multiple unknowns around transmission, limited testing capacity, and an inability to obtain sufficient sample size.
The Snowball study’s approach proposed utilization of a form of network-targeted sampling design, respondent-driven sampling (RDS), which leverages effort on the part of index or “seed” cases to recruit contacts for participation. Simply, seed cases are provided unique one-time-use coupons for testing that they can distribute to contacts. The contacts may decide they want to ‘redeem’ the coupons for COVID-19 testing by presenting to a research coordinator, enrolling in the study, and obtaining their test. This sampling method can therefore “snowball” out into a community.
The Product
Critical to the realization of the Snowball study, Crucible developed an integrated cloud platform to manage the recruitment, enrollment, and management of study candidates and their data. As part of every phase of the study’s design, proposal, and implementation, Crucible has been able to frequently iterate on the product in order to respond to ever-evolving needs of a project being done in unprecedented circumstances.
Snowball v 1.x is designed to work within the current Duke Clinical Research Model. The Snowball platform links eligible participants identified through positive lab results in the EHR at Duke Hospitals to REDCap consent and electronic data capture (EDC), to Snowball which delivers, stores, and connects “coupons” across participants.
Snowball provides the integrated data for the data management team to extract and analyze in the statistical tool of their choice. Amazon Web Services is the cloud provider for v 1.0 and, through the use of a virtual private cloud, is approved by Duke Health Security Office to store PHI. Since the initial delivery, the product and study teams have worked closely to refine the operational workflow.
Snowball GR
As part of our work on the Snowball project, Crucible has developed an open source general release version of the Snowball platform. This customizable platform can be configured and deployed by any development team, offering the flexibility for utilization in future pandemics or any other situation where respondent-driven sampling will be employed.
Snowball GR acts as a standalone application out of the box, but also offers options for integration with the deploying organization’s existing EHR systems and preferred survey tools (or just configure the included SurveyJS template).
NOTE: This is a publicly available demo instance. Do not upload PHI or private data.
Features
- GUI dashboard for managing study enrollment and participation
- Integrated connection to the Duke Electronic Health Record for seed identification
- Automated seed case exclusion based on protocol standards
- Daily delivery of seed case reports for cohort identification, supporting manual inclusion/exclusion by investigator
- Unique human-readable coupon generation for seed enrollment
- Coupon expiration and validation tracking
- Integration with online REDCap participant consent form and questionnaire
- Adjustable peer coupon quantity and alternative email option
- Options for both participant self-enrollment and operator-assisted enrollment
- Supports communication with participants via both email and SMS
- CRM logs to support high-touch approach for increasing study enrollment
- Encrypted data storage on Amazon Document DB
- Identity access management utilizing Shibboleth authentication
- Automatic user logoff, ensuring protection of PHI
- Generation of multiple downloadable reports, including integrated lab results data, for study team utilization and analysis
- Error monitoring, logging, and alerts
Structure
Snowball v 1.x is built to enable identification of and contact with people who many have been exposed to SARS-C0V-2 (COVID-19). The application sends coupons (unique four-word combinations) to invite seed cases (identified via diagnosis in DUHS EHR data and matching study-specific inclusion criteria). After enrollment via coupon redemption, coupons are distributed to study-qualifying peers identified by the seed participant.
Snowball v 1.x is architected in AWS Virtual Private Cloud (Amazon VPC) with separate subnets for different application tiers and private (back-end) subnets for application and database. This Multi-AZ architecture helps ensure high availability. The application stack uses Amazon VPC security groups for Amazon EC2 instances and load balancers, which limit access to only necessary services. Data is encrypted in transit from its entry point via the Duke F5 firewall, through to the application running in AWS Fargate.
Integrated emergency procedures are implemented to watch errors in real time. Amazon CloudWatch sends notification via email to the product support team to immediately resolve urgent, critical application issues.
