Precise control balances epidemic mitigation and economic growth

COVID-19 has caused over 635 million cases and 6.5 million deaths globally, prompting widespread use of nucleic acid testing and behavioral restrictions. China’s Health Code system, a QR-based digital tool linking test results, travel, contact history, and residence to categorize risk (green/yellow/red), has been widely adopted to identify exposure and manage access. While Health Codes potentially enable real-time, accurate detection with minimal disruption, their overall effectiveness and economic impact remain understudied. The authors classify control strategies into No control, Precise control (quarantining Health Code red/yellow individuals), and Extreme control (broad lockdown with only essential personnel active). They translate Health Code functionality into two policy levers: nucleic acid positive detection rate (r1) and quarantine rate (r2). The study aims to: (1) evaluate effectiveness of different control strategies in curbing spread; (2) provide guidance on short- and long-term measures from a Health Code perspective; and (3) determine the turning point/end of control for the epidemic under Precise control.

The paper situates its work among global policy experiences and models of epidemic control. It notes examples akin to Extreme control (e.g., national lockdowns in Europe) and No control tendencies in parts of the Middle East due to delayed measures, and regionally tiered approaches in Italy resembling Precise control. Prior modeling studies analyzed COVID-19 dynamics and control strategies (SEIR-type models; optimal control; effects of quarantine/isolation; contact tracing combined with distancing; and network-based analyses of NPIs). Research on China’s Health Code largely discusses its governance role and ancillary analyses (e.g., vaccine effectiveness) rather than quantifying Health Code’s direct impact on infections and the economy or its capacity limits. This study addresses that gap by quantifying Health Code-enabled Precise control impacts, thresholds, and policy design, including limits by R0 and economic trade-offs.

The authors develop an extended SEIR-type compartmental model tailored to Omicron’s characteristics and policy interventions, termed the SLAIRD model: Susceptible (S), Latent (L), Asymptomatic (A), Infectious (I), Recovered (R), and Death (D). Additional intervention states include O (untracked but of concern), T (tested positive, not yet quarantined), and Q (quarantined). Transitions incorporate: infection via β (related to R0 and recovery γ), progression from L to A/I at rate α, possible return from L to S with probability ξ, recovery and immunity waning (R→S) at rate p with immunity duration φ, and mortality among I at rate d. Health Code effects are parameterized by r1 (positive detection rate via NAT) and r2 (quarantine rate for detected/exposed individuals), determining flows into T and Q. The model equations (provided in the paper) include discrete-time updates for each compartment, tracking quarantined and tested subgroups (A^Q, A^T, I^Q, I^T, L^Q) and aggregate transitions to R and D. Policy scenarios: (a) No control—no additional testing/quarantine beyond symptomatic detection; (b) Precise control—quarantine individuals with red (confirmed) or yellow (high-risk contacts) codes, governed by r1 and r2; (c) Extreme control—quarantine all non-essential personnel (~90% of population). Assumptions include fixed population N, constant R0 within a wave, and compliance among controlled individuals. Economic modeling: Using Shanghai’s historical GDP growth (1980–2019) and population, the study estimates GDP per capita and total GDP over the study period under each policy, accounting for reductions in the workforce due to symptomatic infection and quarantine, and incorporating sectoral remote work rates (China, 2022). Prevention and treatment costs include per-capita quarantine costs, nucleic acid testing costs, and medical treatment costs for asymptomatic and symptomatic cases. Policy shift analysis: Simulates initial Extreme control starting on day 15 (after 14 days of free spread), then switching to Precise control at different times. The trigger is the proportion of abnormal codes (red+yellow) at the switch, assessing whether peak recurrence occurs, total cases, and duration of Extreme control saved. NAT allocation strategies: Compares priority versus random allocation of daily NAT quotas M. Priority targets Health Code yellow groups (LT, LQ) first; random spreads tests across the population. The study computes positive detection rates at specific time points (day 15, peak day, last day) and cumulatively across the wave. Intervention limits: Defines controllability thresholds where Precise control yields accumulated cases ≤10% of No control and ≤1.5% of total population. It explores the maximum R0 suppressible under varying r1 and r2, and evaluates outcomes when R0 exceeds this maximum. Fitting and validation use data from Shanghai (current daily), New York and Los Angeles (new daily), with low forecast errors reported. Robustness checks vary recovery ratio γ, assess NAT errors (reducing r1) and Health Code assignment delays, and test model applicability to Alpha (R0≈2.5) and Omicron BA.5 (R0≈18.6).

- Precise control (Health Code–assisted) achieves a strong balance: reduces deaths by 97% and increases GDP by ~1% compared to No control or city shutdown (Extreme control). - Economic outcomes: Total GDP over the study period was $186.17B (Precise) versus $185.23B (No control) and $80.39B (Extreme control). The similar GDP between Precise and No control reflects trade-offs between quarantine-induced telework losses (Precise) and infection-induced workforce loss (No control). - Cost structure: Extreme control minimizes total prevention cost via suppressed transmission; Precise control incurs higher quarantine costs ($3.81B) but avoids massive treatment costs. No control incurs very high treatment costs ($55.96B). Comprehensive economic analysis shows Precise control increases total prevention cost by only ~15% versus Extreme control while yielding much higher GDP. - Infection dynamics: Precise control yields comparable reductions in infections and deaths to Extreme control while minimizing disruption to productive life; No control produces the highest infections and deaths. - Policy timing: Switching from Extreme to Precise control when abnormal-code proportion reaches ~0.19% (about day 24 from policy initiation) avoids peak recurrence and can save ~30 days of Extreme control within a two‑month wave. - Testing allocation: Prioritizing NAT to Health Code–identified high-risk (yellow) groups outperforms random allocation across day 15, peak day, last day, and cumulatively, especially when NAT resources are limited. - Control limits: Even at maximal r1 and r2, the highest suppressible R0 is ~16.5. r2 (quarantine) generally has stronger preventive effect than r1 (detection) at equal intensities. For R0>16.5, control effectiveness declines rapidly; beyond a threshold, cases cannot be suppressed to target levels. - End-of-control scenarios: As COVID-19 evolves toward higher R0 and lower fatality, Precise control’s value diminishes. The epidemic’s “end” under control arises either when virulence becomes low (fatality comparable to flu ~0.1%) or when infectivity exceeds the controllable R0 threshold (~16.5). - Robustness/sensitivity: Health Code tracking can reach >90% of target groups within weeks. NAT errors and Health Code assignment delays increase infections (up to ~2x in worst cases) and costs but still outperform No control; GDP remains comparable to No control under errors/delays though costs rise. Model fits Alpha well but not BA.5, consistent with limited control capacity for very high R0.

The findings show Health Code–enabled Precise control can simultaneously mitigate infections/deaths and preserve economic activity, addressing the study’s central question on balancing public health and growth. Comparable infection control to Extreme control, coupled with far better GDP, supports using targeted rather than blanket restrictions where feasible. The quantified thresholds guide practical policy: begin with short Extreme control to curb early exponential growth, then shift to Precise control once abnormal-code prevalence is low enough (~0.19%) to avoid rebound. Prioritizing NAT to yellow-coded contacts maximizes detection efficiency under constrained resources; simulations identify test quotas beyond which returns diminish, improving resource allocation. The analysis of control limits by R0 clarifies when Precise control ceases to be effective (R0>~16.5), informing when to pivot from community-wide digital control to alternative measures (e.g., medical interventions). These insights generalize to evolving variants: as infectivity increases and virulence declines, the marginal benefit of wide-scale Health Code–based control wanes, suggesting time-bound, staged use. Overall, the results are relevant for urban policymakers seeking scalable, cost-effective strategies that adapt to epidemic phase, resource constraints, and variant characteristics.

This work introduces a Health Code–integrated SLAIRD model to quantify how Precise control compares with No control and Extreme control across epidemiological and economic dimensions. Precise control substantially reduces mortality (~97%), maintains or slightly improves GDP versus No control, and avoids the severe economic contraction of Extreme control while incurring manageable prevention costs. It provides actionable guidance: early short Extreme control followed by Precise control at a data-driven threshold (abnormal-code ≈0.19%), priority NAT for high-risk contacts, and calibrated r1/r2 settings. It also identifies a fundamental limit—diseases with R0 above ~16.5 cannot be effectively suppressed by Precise control—signaling when to transition away from community-wide Health Code measures. Future research should incorporate fine-grained contact networks and individual-level mobility to better capture spatio-temporal transmission, refine cost functions, and further address privacy-preserving implementations to enhance public trust and compliance.

- Model simplifications: Population categories lack individual identity and explicit contact networks, limiting accurate spatio-temporal transmission modeling, especially for close contacts. - Parameter uncertainty: Several parameters are literature-derived or fitted and may not reflect exact real-world values; results emphasize trends over precise magnitudes. - Applicability limits: The framework fits Alpha well but not BA.5, consistent with limited capacity when R0 exceeds ~16.5; effectiveness declines for highly infectious variants. - Operational sensitivities: NAT errors and Health Code assignment delays increase infections and costs; although performance remains better than No control, practical implementation quality matters. - Privacy and data constraints: Health Code use involves sensitive personal data; societal concerns and regulatory constraints may affect deployment and data granularity, potentially weakening Precise control effectiveness.