Public health resilience.

Health crises are challenging because the normal channels of communication and ways of reporting have often broken down under novel and unexpected circumstances.  The most critical need in such situations is accurate, timely, and relevant information. Area is a mobile application designed specifically for bio surveillance and health crisis resiliency.  The app helps health workers identify and obtain the most critical information to assess and mitigate health risks; distribute these data in a secure and curated way, and optimize cross-organizational resource allocation.  Area provides situation awareness, comprehension, and projection of relevant on-the-ground information.  In short, Area meets the most pressing need in health crisis management. The Area app combines state-of-the-art social networking technology with advanced, proven algorithms for assessing risk and managing distributed resources. Intuitive user experience design ensures that the Area app is easy to understand and to use, even for non-technical personnel.  The Area app is designed for application in all phases of public health or humanitarian crisis management, from Early Detection/Onset to Response/Relief and Recovery/Transition.

 

Mobile Survey
Data Collection

Enables clear understanding of health crises based on crowd sourcing and mobile surveys and uses Value of Information (VOI) for real-time responsive mobile data collection needs and direction.

Library of Sampling
Methodologies

Deploys standard and novel sampling methodologies such as respondent-driven sampling, two-cluster sampling, and stratified sampling.

Uncertainty
Visualization
 

Conveys compact and scalable visualizations that show multiple dimensions of uncertainty around cumulative cases, population base rates and cultural factors.

Task Management
Service

Allocates tasks to users based on capabilities, locality and availability.

Problem

Random sampling techniques suffer from a serious, sometimes critical, shortcoming: A sample is only statistically random with reference to the sampling population. These non-respondents are often a substantive proportion of the target population and their absence threatens statistical reliability. Being able to reduce this error would dramatically increase the reliability and accuracy of any survey effort.  

Solution

Respondent driven sampling (a.k.a. snowballing) is a form of network sampling which collects “waves” of data. We implemented a mobile app version of this sampling technique into Area so that research assistants could quickly master data collection. The mobile app provides substantially easier data management, and virtually eliminates transcription errors.

 

We demonstrated the effectiveness of our mobile application and RDS sampling by replicating major findings from a portion of a large-scale, well-known survey named the Los Angeles and Neighborhood Survey (LA FANS). We were able to accurately replicate a portion of this survey in 5 % of the time and at less than 0.5 % of the cost.

 

Furthermore, our replication proved to have dramatically lower proportions of non-respondents – less than 20 % throughout the process – than did the conventional survey, which had non-response rates of almost 50 %. These composite non-response rates resulted from an almost 50 % non-response rate in the initial wave, which reduced to less than 15 % for higher-order waves. Table 1 below shows the relative time and cost data.



Money Invested

Time Invested

LA FANS

$3,000,000

 

6000 Hours

Area

$11,000

 

300 Hours

Area % 

0.4%

 

5%

Table 1. Conventional (LA FANS) and Respondent Driven Sampling (RDS) Results.