Practice Title: A language-matching model to improve equity and efficiency of COVID-19 contact tracing
Department: Santa Clara County Public Health Department
Size: Large (Population of 500,000+ people)
State: California
Summary of Practice:
Overview. The County of Santa Clara Public Health Department (PHD) partnered with Stanford University to improve the efficiency and equity of COVID-19 contact tracing by using a machine learning algorithm to match the predicted language needs of COVID-19 cases with the language skills of contact tracers. A randomized controlled trial to pilot the intervention showed it reduced time to case completion by 14 hours and improved community engagement with contact tracing. The findings were recently published in the Proceedings of the National Academy of Sciences (Lu et al. 2021. “A Language Matching Model to Improve Equity and Efficiency of COVID-19 Contact Tracing.” PNAS.)
PHD Description. Santa Clara County (SCC) is home to 1.9M residents in the San Francisco Bay Area, including the City of San Jose. Covering nearly 1,300 square miles, its population ranges from high-tech Silicon Valley to agricultural production. Roughly 37% of the population are immigrants; 25% are Latinx; 39% are Asian; and over half of households speak a language other than English at home. Along with five other Bay Area counties, SCC was the first in the nation to issue a shelter-in-place order in response to the COVID-19 pandemic.
Goal. PHD has engaged in extensive efforts to address substantial inequities in COVID-19 case, testing, and vaccination rates. While 25% of the population is Latinx, at the beginning of this collaboration 56% of cases were Latinx. Early in the pandemic, PHD created one of the largest contact tracing operations in the country, with nearly 1,000 contact tracers at its peak, maximizing bilingual staff where possible. But no information about client language needs were transmitted in real-time, causing significant pain points with contact tracing to reach vulnerable populations. The goal of this partnership was to overcome language barriers in contact tracing in order to enhance equity through improved service to disproportionately impacted communities.
Implementation and Steps. PHD partnered with Stanford University’s RegLab to develop a system to reduce language barriers by (a) merging administrative data (e.g., census records) with laboratory reports, (b) using a machine learning algorithm to predict Spanish language needs of cases in real-time as laboratory results arrived, (c) embedding this predictive model into high volume contact tracing by preferentially assigning cases predicted to speak Spanish to a specialty team with predominantly native Spanish-speaking contact tracers, and (d) demonstrating the effectiveness of the practice in a randomized controlled trial, comparing bilingual contact tracing against the status quo use of simultaneous interpreter services.
Public Health Impact. This low-touch intervention resulted in (a) significant time savings, shortening the time from opening of cases to completion of the initial interview by nearly 14 hours and increasing same-day completion by 12%, and (b) improved community engagement, reducing the refusal to interview by 4%. These findings have important implications for how to use data science and machine learning to reduce social disparities in COVID-19; improving equity in healthcare access and quality of service delivery; and, more broadly, leveling language differences in public health services.
A language-matching model to improve equity and efficiency of COVID-19 contact tracing
Category
Infectious Disease Prevention and Control