Alternating quarantine for sustainable epidemic mitigation

The study addresses how to mitigate COVID-19 spread while sustaining socio-economic activity in the absence of pharmaceuticals. Traditional non-pharmaceutical interventions (NPIs) such as prolonged lockdowns and mobility restrictions, while epidemiologically effective, have severe and potentially unsustainable socio-economic and psychological costs. The authors propose and evaluate an alternating quarantine (AQ) strategy in which households are partitioned into two cohorts that alternate weekly between activity and home quarantine. The research aims to test whether AQ can reduce transmission comparably to population-wide quarantine while maintaining approximately 50% economic activity, and to explore how AQ’s weekly rhythm can synchronize with SARS-CoV-2’s incubation and presymptomatic infectious period to isolate invisible spreaders. The work compares AQ against other 50%-capacity strategies (intermittent quarantines applied to the whole population and half quarantine via selective shutdown) and against varying levels of population-wide quarantine, assessing epidemiological outcomes and implementation robustness.

The paper situates AQ within the broader literature on NPIs, mobility restrictions, and social distancing policies implemented worldwide during COVID-19. It references work on lockdown effects on epidemic dynamics, estimation of transmission parameters, presymptomatic and asymptomatic transmission, incubation period distributions (including Weibull characterizations), and digital/organizational interventions. It contrasts AQ with previously proposed intermittent cyclic exit strategies (e.g., 4:10 or similar periodicity) and with strategies focusing on selective or population-wide quarantine. Studies on household secondary attack rates, hospital capacity constraints, and the disproportionate risk to vulnerable populations are cited to motivate model parameters (e.g., in-house transmission) and complementary measures (testing, protection of vulnerable groups).

Modeling framework: The authors simulate an agent-based, time-resolved social contact network of approximately N ≈ 10^4 individuals organized into M = 4 × 10^3 households. Social interactions are represented by two networks: (1) an external network A_ij (out-of-home interactions at work/school/public places) with degree distribution P(k), and (2) a household network B_ij composed of isolated cliques with household size distribution P(m) derived from empirical data. Links in A and B switch between active and idle states at 15-minute resolution over T = 150 days to capture temporal collocation; infection can occur only when links are active. External links are active mainly during daytime; household links at night and during quarantine.

Transmission parameters: Each network has two parameters: average daily active time T1 (A) and T2 (B), and per-interaction transmission probabilities p1 (A) and p2 (B), with realistic assumptions T2 > T1 and p2 > p1 due to closer in-house contact. Network density (mean external degree ⟨k⟩) and household size (⟨m⟩) also shape transmission.

Disease natural history: Individuals transition through SARS-CoV-2/COVID-19 states: Susceptible (S), Exposed (E), Presymptomatic (PS), Asymptomatic infectious (AS), symptomatic states (IM mild, IS severe, IC critical), Hospitalized (H), Ventilated (V), Recovered (R), Deceased (D). Average timelines are: incubation ≈ 5 days; infectiousness begins ≈ 3 days post-exposure for symptomatic and ≈ 4 days for asymptomatic; presymptomatic infectious window ≈ 2 days before symptom onset; upon symptom onset, the case and cohabitants isolate (external links shut). A fraction (~30%) are asymptomatic and continue infectiousness until recovery. Inter-individual variability in transition times is modeled by distributions P1(t′) (e.g., Weibull) and P2(t) for exposure-to-infectiousness and symptom onset times.

Observable parameters: The simulations focus on two emergent measures linked to model parameters (P(k), P(m), T1, T2, p1, p2): (i) epidemic growth rate β obtained from early exponential growth I(t) ≈ e^{βt} (I = IM + IS + IC), and (ii) in-house infection rate α, the fraction of transmissions occurring via household links, α = θ_h/θ_tot, where θ_h counts B_ij transmissions.

Calibration and scenarios: β is empirically estimated by fitting early unmitigated growth in 12 countries (pre-NPI window), yielding β ≈ 0.26 day^-1 on average and corresponding R ≈ 2.4. Simulations explore multiple scenarios: worst-case β ≈ 0.26; intermediate β ≈ 0.21 (≈20% reduction, reflecting hygiene/masks/contact avoidance); best-case β ≈ 0.15 (≈40% reduction). α is varied from negligible (α ≈ 0) to high (α ≈ 0.32–0.34), consistent with 20–40% household secondary attack probabilities. External network A is primarily an Erdős–Rényi graph with ⟨k⟩ = 15; supplementary analyses consider scale-free networks.

Mitigation strategies tested: (1) Full Quarantine (FQ): deactivate all A_ij (perfect population-wide quarantine) leaving only household transmission; (2) Alternating Quarantine (AQ): partition households into two cohorts alternating weekly between activity (both A and B active) and home quarantine (A inactive, B active) with bi-weekly cycle; (3) Intermittent Quarantine (IQ): entire population alternates between active and quarantine phases in unison, examined with a 7:7 cycle to match AQ; (4) Half Quarantine (HQ): 50% of population remains continuously active, 50% continuously quarantined. Additionally, Population-Wide Quarantine (PWQ) levels η = 50–75% are compared to AQ.

Outcome measures: Epidemic trajectories I(t), residual mortality ΔD = D_S(∞) − D_FQ(∞) (avoidable mortality beyond FQ), peak hospitalization H_peak = max_t H(t), peak ventilation V_peak = max_t V(t), compared against an estimated average national hospital capacity of ≈ 3 × 10^-3 of the population. Simulations average over >20 stochastic realizations. Robustness to limited compliance is tested by introducing a fraction f of continuously active individuals (defectors/essential workers) from S, E, IAS, or mild IM states, exempting severe/critical cases.

Implementation considerations: Practical partitioning at household level, flexibility to accommodate employer/individual needs without enforcing exact 50:50 split, social compliance via organizational liability (schools/workplaces), supply preparation, support networks, and a cohort-status mobile app for access control. Alternative AQ cycles (e.g., 5 workdays + 9 quarantine days with weekend PWQ) are discussed.

• AQ consistently outperforms IQ and HQ across a wide range of scenarios (β ≈ 0.26, 0.21, 0.15; α from ~0 to ~0.32–0.34), yielding lower infections, reduced residual mortality ΔD, and lower H_peak and V_peak, often bringing hospital demand within capacity (~3 × 10^-3 of population).
• AQ approaches the effectiveness of the idealized full quarantine (FQ) while maintaining ~50% socioeconomic activity. Under more favorable β (e.g., β ≈ 0.21 or 0.15), AQ’s performance moves even closer to FQ, with dramatic reductions in mortality and hospitalizations.
• The multiplicative mitigation effect of AQ arises from dual partitioning: (i) halving interacting population (cohort separation) and (ii) halving interaction time (weekly alternation), yielding an approximate fourfold reduction in transmission while preserving 50% activity.
• AQ’s weekly cycle synchronizes with SARS-CoV-2 time scales (~5-day incubation; presymptomatic infectiousness), effectively quarantining invisible spreaders during the week following potential exposure.
• Compared to population-wide quarantine (PWQ), AQ delivers mitigation roughly equivalent to a 70% lockdown (η ≈ 70%) despite 50% continuous activity, which is near the practical upper bound for realistic PWQ implementation.
• AQ remains robust under imperfect compliance, tolerating up to f ≈ 0.2 (≈20%) continuously active defectors/essential workers before significant performance degradation is observed.
• Without interventions (β ≈ 0.26, α baseline), projected mortality reaches ~3% and hospitalization/ventilation needs surpass average national capacity; AQ substantially reduces these burdens.
• High in-house transmission (α up to ~0.32) degrades all quarantine-based strategies but AQ still outperforms IQ and HQ and remains close to FQ relative performance.
• Calibration from 12 countries yields mean early growth rate β ≈ 0.26 day^-1 and implied R ≈ 2.4, consistent with independent estimates.

The findings demonstrate that alternating quarantine directly addresses the core challenge of COVID-19 control: substantial presymptomatic/asymptomatic transmission that undermines traditional isolation of symptomatic cases. By splitting the population into household-based cohorts and alternating weekly between activity and quarantine, AQ both reduces contact opportunities and aligns the quarantine week with the likely presymptomatic window following potential exposure, thereby removing many invisible spreaders from active social mixing. This dual mechanism produces a compounded reduction in transmission compared with strategies that only halve population density (HQ) or only halve active time (IQ). Consequently, AQ maintains economic activity at ~50% while achieving epidemiological outcomes comparable to much stricter population-wide quarantines (≈70% lockdown), significantly lowering avoidable mortality, hospital and ventilator demand. The robustness to moderate non-compliance (up to ~20%) and the ability to combine with other NPIs (masking, hygiene, testing, protection of vulnerable populations) further enhance AQ’s practicality and impact. These results suggest AQ can serve either as a primary mitigation approach or as an exit strategy from strict lockdown, mitigating resurgence risks while enabling socio-economic continuity.

This work introduces and evaluates an alternating quarantine (AQ) policy that partitions households into two cohorts alternating weekly between activity and quarantine. Through network-based, time-resolved simulations calibrated to early COVID-19 growth rates and realistic disease timelines, AQ consistently outperforms comparable 50%-capacity strategies (IQ, HQ) and approaches the efficacy of full quarantine, while maintaining socio-economic continuity. AQ’s advantages stem from dual partitioning (people and time) and synchronization with SARS-CoV-2’s incubation and presymptomatic infectious periods. The strategy is robust to moderate levels of non-compliance and achieves mitigation comparable to ~70% population-wide lockdown at substantially lower social and economic cost. Future work should refine AQ periodicity for different pathogens, integrate large-scale testing and targeted protection of high-risk groups, reduce household transmission (e.g., via isolation facilities), and adapt implementation to sector-specific constraints and network structures.

• Idealized assumptions: Full Quarantine (FQ) is a theoretical upper bound; real-world PWQ cannot reach 100% due to essential services. AQ simulations assume clear household partitioning and strict adherence during quarantine weeks.
• Parameter uncertainty: Precise values of β and α are emergent from model parameters and cannot be perfectly controlled; disease timelines and distributions vary across individuals and contexts, potentially affecting synchronization benefits.
• Model structure: External network modeled primarily as Erdős–Rényi with ⟨k⟩ = 15; while scale-free cases were explored, real contact networks may be more complex. Household transmission parameters (p2, T2) and secondary attack rates vary across settings.
• Compliance and exemptions: While robustness to up to ~20% defectors is shown, higher non-compliance, heterogeneous adherence, or sector-specific constraints (irreplaceable personnel) may reduce effectiveness.
• In-house transmission: Elevated household transmission (high α) can limit the benefits of any quarantine-based strategy, including AQ. Practical measures to reduce intra-household spread are necessary for maximal impact.
• Generalizability: Results are calibrated to COVID-19 early dynamics; applying AQ to other pathogens requires retuning the cycle to disease-specific time scales and validating with pathogen-specific data.