Abstract
Objective: To measure the effects of the regional Tier system on the COVID-19 pandemic in the UK between the first and second national lockdowns, prior to the emergence of the B.1.1.7 variant.
Design: A modelling study combining estimates of the real-time reproduction number $R_t$ (derived from UK case, death, and serological data) with publicly available data on regional non-pharmaceutical interventions. We utilized a Bayesian hierarchical model with latent factors to account for national trends alongside subnational Tier effects.
Results: Nationally, transmission increased between July and late September. Immediately prior to the Tier system, $R_t$ averaged 1.3, declining to 1.1 two weeks later. Tier 1 had negligible effects, while Tiers 2 and 3 reduced transmission by 6% and 23%, respectively. If every LTLA had entered Tier 3, 93% would have suppressed their epidemics, compared to 29% in reality.
Conclusions: Interventions at least as stringent as Tier 3 are required to suppress transmission, particularly in the face of more transmissible variants.
Introduction
Following the first national lockdown in 2020, the UK government transitioned to a localized "Tier" system in October 2020. This strategy aimed to apply restrictions to Lower Tier Local Authorities (LTLAs) based on local transmission intensity. Evaluating these policies is complex because transmission rates are influenced by broad "secular trends"—national behavioral shifts and mandates like the "rule of six"—that occur independently of local Tiers. This study aims to disentangle these national trends from the specific impacts of Tiers 1, 2, and 3.
Methodology
The researchers employed a two-stage Bayesian hierarchical model to analyze data from 310 LTLAs between July 1st and November 5th, 2020. First, $R_t$ was estimated by back-calculating from observed deaths, cases, and antibody surveys to account for infection-to-outcome lags. Second, a latent factor regression was applied to these $R_t$ values.
As shown in the model specification above, latent factors were used to represent shared national changes in transmission. By factorizing the secular trend into basis functions, the model could "soak up" broad national noise, leaving a cleaner signal to measure the specific multiplicative effect of the Tiers.
Key Results
The model revealed that transmission began to plateau in late September, before the Tier system was officially introduced. This highlights the importance of accounting for secular trends to avoid overestimating policy effectiveness. The specific impacts of the Tiers were found to be:
- Tier 1: Showed a negligible effect (Coefficient = 1.0), likely because its core components were already practiced nationally.
- Tier 2: Provided a modest ~6% reduction in $R_t$ (Coefficient = 0.94).
- Tier 3: Provided a substantial ~23% reduction in $R_t$ (Coefficient = 0.77).
A counterfactual analysis indicated a significant gap in suppression: while only 29% of LTLAs achieved $R_t < 1$ under the mixed-tier approach, an estimated 93% would have seen their epidemics shrinking if Tier 3 had been applied universally.
Discussion and Limitations
The relatively small effect sizes for Tiers 1 and 2 suggest they were insufficient to suppress transmission. Only Tier 3 demonstrated the "punch" necessary to bring $R_t$ below 1.0. These results are associative; because interventions were bundled into Tiers, the model cannot isolate the impact of specific rules, such as closing hospitality versus banning indoor mixing.
Furthermore, this study concluded just before the Alpha (B.1.1.7) variant became dominant. Given that newer variants are more transmissible, the interventions evaluated here would likely be even less effective today. The findings underscore that high-intensity interventions are essential for epidemic control until widespread population immunity is achieved.
Summary updated