The importance of sustained compliance with physical distancing during COVID-19 vaccination rollout

The study addresses how declining compliance with physical distancing during COVID-19 vaccination rollouts impacts SARS-CoV-2 transmission and hospitalisations. In late 2020–2021, more transmissible variants (Alpha, Delta) emerged as vaccines were introduced, and evidence indicated decreasing adherence to distancing, especially among vaccinated individuals. The research question is under which conditions vaccination, coupled with behavior change (reduced compliance), could transiently worsen epidemic outcomes compared to no vaccination, and how targeted interventions might mitigate these effects. The purpose is to provide qualitative guidance on vaccination-behavior interplay for public health policy during rollouts, focusing on cumulative hospitalisations at three and six months. The importance lies in informing whether and how to maintain physical distancing measures during vaccine deployment to avoid transient increases in infections and hospitalisations, especially with more transmissible variants and variable vaccine effectiveness against infection.

The paper situates its work within evidence that non-pharmaceutical interventions (NPIs), including physical distancing, reduced transmission in early pandemic waves, but alone were insufficient for eradication, necessitating vaccination. It reviews clinical and real-world estimates of vaccine effectiveness against infection for mRNA and adenoviral vaccines (e.g., Pfizer-BioNTech and Moderna 80–95% against original/Alpha variants; reduced to ~64% against Delta), highlighting variants’ impact on effectiveness. Behavioral studies from the Netherlands indicated declining compliance post-vaccination, with motivations including reduced perceived risk or access to venues (QR codes), aligning with the Health Belief Model where perceived susceptibility drives protective behavior. Prior modeling showed feedback between epidemic dynamics and behavior can alter disease spread and that relaxing compliance past thresholds increases cases and hospitalisations. The paper also references contemporaneous modeling indicating that vaccination benefits can be offset by behavioral relaxation or premature NPI lifting, and comparative observations (e.g., Cyprus vs. Malta) where stricter measures during rollout correlated with better outcomes.

Design: A deterministic compartmental SEIR model extended to include vaccination and dynamic compliance behavior, applied over a six-month horizon.

Population structure and infection dynamics: Compartments include susceptible (S), exposed/latent (E), infectious (I), recovered (R). Average latent period 4 days, infectious period 7 days. No births/deaths within time horizon; recovery confers permanent immunity. Three virus scenarios: original (R0≈2.5), Alpha-like (R0≈3.75), Delta-like (R0≈4.92), differing only by transmission probability per contact.

Vaccination: All-or-nothing mechanism. Individuals in any infection state can be vaccinated at rate ν, but only those susceptible at vaccination obtain sterilizing immunity with probability ω (efficacy), moving to V. Others move to corresponding vaccinated-but-susceptible or infected states (SV, EV, IV, RV) and are epidemiologically similar to non-vaccinated counterparts. Vaccination confers immediate effect. Vaccine efficacy ω varied 0.4–1.0 in analyses; main figures often examine 0.60 and 0.91. Vaccine uptake rates span a slow-to-fast range calibrated to early rollout data (Belarus slow; Israel fast), yielding ~10% to ~60% vaccinated at six months if sustained.

Behavioral compliance: Non-vaccinated individuals switch between non-compliant (U) and compliant (Uc) states with respect to distancing. Compliant individuals have reduced contact rate r1·c relative to non-compliant c (0≤r1≤1). Compliance acquisition rate ψc is proportional to incidence (δ·α·[E+Ec+EV]), reflecting response to reported cases. Compliance loss rate μ increases linearly with vaccination coverage: μ(t)=μ0+μ1·V(t)/N. Only non-vaccinated can be compliant; vaccinated individuals move to a permanently non-compliant class with higher contact rate r2·c (r2≥1), reflecting near pre-pandemic contacts. Baseline parameters include c=8.8 contacts/day; r1=0.34 (compliant contacts ≈2.8/day); r2=1.5. Pre-pandemic average contacts ≈14.9/day. Effective reproduction number at start calibrated to Rt≈1.1 (Netherlands, Nov 2020). Initial compliant fraction 65% (Netherlands behavioral survey).

Calibration and initial conditions: Calibrated to Netherlands Nov 2020. Infectious prevalence I+Ic=112,435 (RIVM). Daily incidence ≈16,062 (from 7-day infectious period). Seroprevalence-based recovered proportion set to 8%. Population N≈17,000,000. Initial values: S=5,412,160; E=22,487; I=39,352; R=476,000; Sc=10,051,156; Ec=41,762; Ic=73,082; Rc=884,000; vaccinated compartments start at 0. Sensitivity analyses varied initial conditions and epidemiological parameters.

Hospitalisation mapping: Probabilities of hospitalisation among infected: original 4%; Alpha 5.6% (1.4× original); Delta 10% (1.85× Alpha). For vaccinated, protection against hospitalisation: 95% (original) and 85% (Alpha, Delta), yielding hospitalisation probabilities among vaccinated infections: 0.2% (original), 0.84% (Alpha), 1.55% (Delta). Prevalence of hospitalised at a time computed as proportion of currently infectious multiplied by corresponding probabilities.

Outputs: Prevalence of infected (E+I) and hospitalised, and relative difference in cumulative new hospitalisations (vs. no-vaccination) at 3 and 6 months. Also attack rates by vaccination status.

Interventions: Three compliance-targeted strategies: (1) Non-vaccinated: maintain average compliance duration at pre-vaccination level (30 days) despite rising coverage (i.e., reduce μ’s growth). (2) Vaccinated: prevent increase in contacts post-vaccination (set r2=1, equal to non-compliant non-vaccinated). (3) Combined: apply both. Additionally, a centralized lockdown trigger: when prevalence of infectious cases exceeds threshold θ (50–500), contacts for all (including vaccinated) reduced to CL=3 contacts/day until prevalence falls below threshold, then reverts.

Sensitivity analyses: Extensive multivariate sensitivity over vaccine efficacy, uptake rate (translated to coverage at 3 and 6 months), behavioral parameters (δ, μ0, μ1), contact rates (c, r1, r2), latent/infectious durations, initial conditions, and lockdown thresholds.

• Declining compliance during vaccination can transiently increase infections and hospitalisations above a no-vaccination scenario, particularly when vaccine efficacy against infection is low and variants are more transmissible. Faster vaccination under such conditions can initially worsen outcomes compared to slower rollout due to accelerated compliance loss and higher contacts among vaccinated.
• With high vaccine efficacy (e.g., ~91%), vaccination rapidly reduces prevalence and cumulative hospitalisations below no-vaccination, with larger benefits for faster uptake.
• Compliance dynamics: Under fixed incidence (~16,062/day), fast vaccination led to compliant population dropping to ~54% at 3 months and ~32% at 6 months; slow vaccination retained ~89% and ~84% compliance at 3 and 6 months, respectively.
• Contribution by vaccination status: With low efficacy (60%) and fast uptake, vaccinated individuals can account for a substantial share of infections; for a Delta-like variant, more than one-third of infections can be among vaccinated, implying many infections are less likely to lead to severe outcomes.
• Hospitalisation probabilities used: 4% (original), 5.6% (Alpha), 10% (Delta) in general; among vaccinated infections: 0.2% (original), 0.84% (Alpha), 1.55% (Delta), indicating strong decoupling of infections from severe disease in vaccinated groups.
• Threshold behavior: For each time point (3 and 6 months) and variant, there exists a vaccine-efficacy/coverage boundary above which cumulative hospitalisations are reduced vs. no-vaccination. The minimal efficacy required to avoid excess hospitalisations decreases with time and increases when rollout is slower.
• Interventions:
  • Targeting non-vaccinated compliance (keeping compliance duration at 30 days) consistently reduces cumulative hospitalisations versus no intervention, lowers the efficacy threshold required to outperform no-vaccination, and reduces excess when present.
  • Targeting vaccinated contacts (preventing contact increase post-vaccination) yields mixed short-term results for the original variant under slow rollout (can shift transmission to non-vaccinated), but improves outcomes under fast rollout and for more transmissible variants; benefits are more apparent at six months than at three.
  • Combined intervention performs best, decreasing both excess hospitalisations and the efficacy threshold across rollout speeds and variants.
  • Lockdown trigger intervention (contact reduction to 3/day when infectious prevalence exceeds a threshold) prevents excess infections in both short and long term, with relatively low sensitivity to trigger threshold; largest relative decreases occur with fast rollout and high efficacy.
• Long-term perspective: Even when short-term infections may exceed no-vaccination with low efficacy and fast rollout, over >1 year horizons vaccination confers substantial infection prevention benefits.

The findings demonstrate that the interaction between vaccination rollout and behavioral responses can qualitatively change epidemic trajectories. The central research question—whether declining compliance during rollout can lead to worse short-term outcomes than no vaccination—is affirmed under specific conditions: low efficacy against infection, rapid rollout, and higher variant transmissibility. The model explains that rising vaccination coverage can reduce perceived susceptibility, increasing contacts both among vaccinated (higher r2) and by accelerating compliance loss among non-vaccinated (μ increasing with coverage), which together can transiently raise transmission. These results underline the significance of maintaining or actively supporting compliance during rollout to ensure that vaccination benefits are realized promptly, especially before reaching high coverage. Interventions tailored to non-vaccinated individuals (sustaining compliance duration) and to vaccinated individuals (limiting post-vaccination contact increases) can mitigate adverse transients; their relative advantage depends on variant transmissibility, vaccine efficacy, and rollout speed. The lockdown-trigger strategy is an effective backstop to prevent surges but has social and economic costs. Practically, the study argues for behaviorally informed public health messaging and policies during rollouts to prevent temporary rebounds and to protect healthcare capacity. The qualitative insights generalize beyond SARS-CoV-2 to future pandemics featuring similar behavior-disease feedbacks.

The study contributes a socio-epidemiological modeling framework that couples vaccination with dynamic compliance to physical distancing, showing that behavior-vaccination feedback can transiently negate expected benefits of rollout under certain conditions. Key contributions include: identification of parameter regimes (vaccine efficacy, uptake speed, variant transmissibility) where transient excess hospitalisations can occur; demonstration that sustaining compliance—especially among non-vaccinated—reduces necessary vaccine efficacy thresholds and improves outcomes; and evidence that combined compliance-targeted interventions or lockdown triggers can prevent excess infections and hospitalisations during rollout. For future research, incorporating age structure, contact heterogeneity, waning immunity, and refined (data-driven) behavioral response functions would enable more quantitative predictions and tailored policies. Extending the framework to immune-escape variants and partial protection against susceptibility/transmissibility would also enhance applicability.

• Simplified population structure: no age stratification or contact heterogeneity; limits quantitative accuracy and ignores targeted vaccination strategies’ effects.
• Fixed NPIs/lockdown baseline assumed over six months; real-world policies and contacts may change, potentially accelerating compliance decay and contact rates.
• Constant vaccination rates over time; actual rollouts vary and often accelerate.
• Behavioral functions (compliance acquisition and loss) modeled as linear monotonic functions of incidence and vaccination coverage; true responses may be nonlinear and context-dependent.
• All-or-nothing sterilizing immunity assumption for vaccine protection against infection; real vaccines confer partial reductions in susceptibility/transmissibility and wane over time, which were not modeled.
• No explicit modeling of waning immunity (natural or vaccine-induced), reinfections, or mortality.
• Calibration primarily to Netherlands in late 2020; generalizability is qualitative.
• Assumes vaccinated individuals move to a permanently higher-contact class; actual behavior is heterogeneous and policy-dependent.