Evaluating the impact of local tracing partnerships on the performance of contact tracing for COVID-19 in England

TL;DR

This paper introduces a novel Bayesian method for evaluating intervention impacts using time-series data, applied here to assess local tracing partnerships' effect on COVID-19 contact tracing in England, revealing small overall benefits with significant variability.

Contribution

The paper develops a new Bayesian approach for causal inference with mixed outcomes, enhancing efficiency and uncertainty quantification, and applies it to evaluate local tracing partnerships in England's COVID-19 response.

Findings

Local tracing partnerships had a small positive impact on contact tracing effectiveness.

Significant heterogeneity in effects across different units and over time.

The proposed method effectively handles mixed outcome types and provides comprehensive uncertainty estimates.

Abstract

Assessing the impact of an intervention using time-series observational data on multiple units and outcomes is a frequent problem in many fields of scientific research. In this paper, we present a novel method to estimate intervention effects in such a setting by generalising existing approaches based on the factor analysis model and developing a Bayesian algorithm for inference. Our method is one of the few that can simultaneously: deal with outcomes of mixed type (continuous, binomial, count); increase efficiency in the estimates of the causal effects by jointly modelling multiple outcomes affected by the intervention; easily provide uncertainty quantification for all causal estimands of interest. We use the proposed approach to evaluate the impact that local tracing partnerships (LTP) had on the effectiveness of England's Test and Trace (TT) programme for COVID-19. Our analyses suggest that, overall, LTPs had a small positive impact on TT. However, there is considerable heterogeneity in the estimates of the causal effects over units and time.