{"id":"f391e266-0eba-4dc3-b893-650a0e62e237","arxiv_id":"1908.05261","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A two-community model shows high mobility from a high-risk area into a sufficiently safe area can push the global reproduction number below one, while a cordon sanitaire around the high-risk area can increase total final epidemic size.","lead":"An epidemic model with two connected communities indicates that quarantining a high-risk area, a cordon sanitaire, can raise the total number of infections, while letting high-risk residents spend time in a safer neighbor can reduce the overall outbreak. This matters because it challenges a standard containment policy and specifies exactly when mobility restrictions help or backfire.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central result hinges on infected residents remaining as mobile as healthy ones; if sick individuals stay home, the mobility threshold for control may vanish.","rationale":"The paper is a theoretical exploration, not a data-fitting exercise, and its qualitative thresholds are interesting. The reader's CONDITIONAL verdict is appropriate. My stress-test identifies the same load-bearing assumption: constant, state-independent mobility. The central counterintuitive result—that allowing high-risk residents into a low-risk community can reduce total epidemic size—only operates if infectious individuals actually spend substantial time in the low-risk community. For EVD this is empirically questionable, and the paper does not discuss it. This is a correctness risk rather than an internal inconsistency: within the stated model the results likely follow, but the policy conclusion depends on an unvalidated behavioral assumption. The omitted SI compounds the problem by preventing independent verification of R0 and final-size computations, but the state-dependent mobility issue is the more substantive challenge to the central claim. Since the reader already recommended a conditional verdict, my assessment does not change that verdict; it sharpens the condition that must be addressed: either justify state-independent mobility for the disease of interest or show robustness to reduced mobility of infected individuals.","tokens_in":11813,"tokens_out":7215,"duration_ms":78794,"concrete_test":"Modify the Lagrangian model so mobility is state-dependent: let infected/infectious HRC residents spend p^I_12 = q p^S_12 in the LRC, with q in [0,1], where q=1 is the paper's assumption and q=0 means infected individuals never leave the HRC. Fix R01=2.45, R02=0.9, N1=N2, t2=0, and recompute the R0(t1)=1 threshold and total final size as functions of q. If, for any plausible q (e.g., q<=0.5), there is no t1<1 with R0<1, or no mobility level that lowers total final size below the cordon-sanitaire baseline, then the central policy conclusion fails. If the threshold persists for small q, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim—that unrestricted HRC-to-LRC mobility can reduce total final epidemic size and even drive global R0 below 1—rests on a model in which the residency-time matrix P=(p_ij) is constant over time and identical across disease states (Section 2). The reported threshold t1^+≈0.8 means every high-risk resident, including every infected and infectious one, spends about 80% of their time in the low-risk community. This is not a harmless simplification: the mechanism is precisely that infectious HRC residents generate fewer secondary cases while in the LRC. For EVD, symptomatic patients are bedridden, hospitalized, or isolated, so their realized mobility is far below that of healthy residents. If infected individuals are even modestly less mobile than the general population, the export-of-cases mechanism weakens, the R0(t1)=1 curve shifts upward, and the existence of an attainable t1^+<1 is not guaranteed. The paper never flags this dependence; the abstract states the conclusion without the qualifier that infected individuals move as much as healthy ones. For R01=2.45 and R02=0.9, the threshold is near t1=0.8; reducing infected mobility to, say, 50% of healthy mobility may require t1≈1 or make R0<1 unattainable. A supporting concern is that the equations defining R0(P) and final size appear only in the omitted SI, so the threshold values cannot be independently checked from the arXiv text. The regime analysis in Fig. 3 also shows the abstract's 'reduces overall epidemic size' is not universal (e.g., R02>1.45), but the dominant unresolved issue is state-dependent mobility of infected individuals.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a two-community epidemic model with a Lagrangian mobility formulation, in which individuals from a high-risk community (HRC) and a low-risk community (LRC) spend fixed proportions of time in each community. The model is calibrated to the 2014 West African EVD outbreak (R01 = 2.45) and used to compare final epidemic sizes under a cordon sanitaire (no mobility) versus various levels of one-way or two-way mobility. The central claims are that unrestricted one-way mobility from the HRC to the LRC can reduce the total final epidemic size below the cordon-sanitaire baseline when the LRC is sufficiently safe, and that sufficiently high mobility from the HRC can bring the global basic reproductive number R0(P) below one, thereby controlling the outbreak. The paper also identifies two mobility thresholds, t1- (mobility level that outperforms the cordon sanitaire) and t1+ (mobility level that drives R0 below one), and examines how these thresholds depend on the LRC risk level R02, the LRC mobility t2, and the population density ratio N1/N2. The conclusions are framed as general insights for when mobility restrictions are beneficial or harmful.","tokens_in":12045,"tokens_out":6122,"duration_ms":54850,"significance":"If the main results hold, they would provide a formal counterexample to the conventional view that restricting mobility from a high-risk to a low-risk community reduces overall epidemic burden, and they would offer an explanation for the observed counterproductive effects of cordons sanitaires in some outbreaks. The paper's use of a residence-time (Lagrangian) framework is a useful alternative to Eulerian metapopulation models, and the explicit thresholds t1- and t1+ are falsifiable predictions that can be tested with parameter estimates. However, the central quantitative claims rest on assumptions that are currently not verifiable from the preprint: the model equations, R0 derivation, and final-size formulas are relegated to an omitted SI appendix, and the main control result assumes that infected individuals remain as mobile as healthy individuals, an assumption that is not flagged or tested. The internal consistency of one stated result (R0<1 despite R02>1) is also questionable. The paper has merit as a conceptual contribution, but the strength of the conclusions currently exceeds what the available evidence supports.","major_comments":[{"comment":"The residency-time matrix P = (p_ij) is assumed to be constant over time and identical for all residents, with no distinction between susceptible and infected individuals. The main control result, including the threshold t1+ ≈ 0.8 reported in Section 3.1 and the abstract's claim that high mobility can control an outbreak, depends on infected and infectious HRC residents spending up to 80% of their time in the LRC. For EVD, symptomatic patients are typically bedridden, hospitalized, or isolated, so their realized mobility is far below that of healthy residents. The paper never flags this dependence, and the abstract states the conclusion without this qualifier. This assumption is load-bearing: if infected individuals are even modestly less mobile than healthy ones, the export-of-cases mechanism weakens and t1+ may become unattainable. Please either relax this assumption in the model or provide a sensitivity analysis showing that the qualitative results persist when infected individuals have reduced mobility, and state the assumption explicitly in the abstract and conclusions.","section":"Section 2"},{"comment":"The manuscript states that 'Detailed model formulation, computation of the community-specific and global basic reproductive numbers obtained using the next generation approach, as well as community-specific and global final epidemic size, can be found in the SI appendix.' However, the SI appendix is not included in the arXiv version. Consequently, the equations defining R0(P), the final-size formulas, and the numerical thresholds in Figures 1–8 cannot be independently checked. For a modeling paper in which all quantitative conclusions flow from these derivations, this is a central omission. The model equations and the R0 derivation must be included in the main text or a complete supplement must be provided with the submission.","section":"Section 2"},{"comment":"The statement that 'high mobility by itself can lead to a global basic reproductive number below the critical threshold, even when R02 is slightly greater than one' is inconsistent with the model assumption in Section 2 that Community 2 is unable to support an outbreak in isolation (R02 < 1). Moreover, for a nonnegative next-generation matrix, the spectral radius (the global R0) is at least the spectral radius of each diagonal block, so if R02 > 1, the global R0 cannot be less than one regardless of mobility. This claim is either a typographical error (perhaps 'slightly less than one' was intended) or it indicates a problem with the R0 computation in the SI. Please correct the statement and check the level curves in Figures 4 and 5 against this bound.","section":"Section 4"}],"minor_comments":[{"comment":"The introduction states that 'our results hold for a range of disease types,' but all quantitative results use EVD parameters only. Please either provide an additional disease example or soften this claim.","section":"Section 1"},{"comment":"The thresholds t1- and t1+ are defined in the text, but the notation is introduced after the phrase 'two empirical thresholds' in a way that may be confusing. Consider defining t1- and t1+ explicitly before the first use in the description of Figure 1.","section":"Section 3.1"},{"comment":"In the left panel of Figure 1, the threshold values t1 ≈ 0.45 and t1 ≈ 0.8 are described in the text but are not marked on the figure. Adding vertical dashed lines at these values would improve readability.","section":"Figure 1"},{"comment":"The left panel of Figure 4 has y-axis labels that appear to contain numerical values such as '0.0234192' instead of R02 values, which makes the level curve difficult to interpret. Please regenerate the figure with clear axis labels.","section":"Figure 4"},{"comment":"In the discussion of Figure 7, the text refers to 'same populations densities' and 'identical populations densities' in the caption, but the figure description is not fully legible. Please ensure the figure caption specifies the population density ratio clearly.","section":"Section 3.3"},{"comment":"There are inconsistencies in the use of subscripts for t1 and t2 (e.g., t−1 vs. t1−). Please use consistent notation throughout.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and important question, but the review process is hampered by the absence of the SI appendix, which contains the model equations, the R0 computation, and the final-size formulas. The internal inconsistency regarding R02>1 and R0<1 may reflect a deeper issue in the next-generation matrix derivation, and the infected-mobility assumption could undermine the main conclusion for EVD. I recommend requesting the complete SI and a careful revision of the relevant statements before the paper can be considered for publication. The topic is within the journal's scope, and the conceptual contribution is potentially valuable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this paper says a cordon sanitaire can make an epidemic worse, and that letting people move from the high-risk to a low-risk area can actually reduce total cases and even drive R0 below one. That's a genuinely counterintuitive result, and it's worth reading. But the quantitative claim depends on an assumption the paper doesn't flag: infected residents are presumed to move exactly as much as healthy ones. For Ebola, that's a hard sell.\n\nWhat's new: the Lagrangian residency-time model itself is the authors' own prior framework, and Ref 22 already showed cordons can fail to minimize final size. This paper goes beyond that by systematically mapping thresholds t1^- and t1^+, including two-way mobility and density ratios. The regime plots (Figs 3 and 7) give a clear, useful picture of when a cordon helps, hurts, or is irrelevant. That is a real contribution, and the policy discussion is sensible.\n\nThe soft spots, in order of size. First, the arXiv version omits the SI containing the model equations, R0 computation, and final-size formulas. That means no one can verify the numbers or the threshold values from the preprint. I'd ask for the SI or put the equations in the main text. Second, and more substantive: the residency-time matrix P is constant in time and identical across disease states. The mechanism that makes mobility work is that infectious HRC residents spend up to 80% of their time in the LRC and generate fewer secondary cases there. For EVD, symptomatic patients are bedridden or isolated; they don't travel. If infected mobility is even half that of healthy people, the R0=1 curve shifts upward and the t1^+ threshold may not be attainable. The paper never mentions this, and the abstract states the conclusion without that qualifier. I think the qualitative result—cordon can increase total size—is probably robust, but the 'mobility can control the outbreak' claim is quantitatively fragile until this is tested. Third, the abstract says unrestricted mobility reduces overall epidemic size, but the text itself shows this only holds for R02 below about 1.45; above that, all mobility increases total size. The abstract is overbroad, though the discussion does qualify.\n\nOverall: the paper is clear, the question is important, and the extension is real. It deserves a serious referee, but the authors need to provide the SI and address state-dependent mobility before the result is established. I'd bring it to reading group to kick around the policy implications.\n\nRecommendation: send to peer review, with the expectation that the mobility-of-infected assumption is the main substantive revision.","headline":"A clearly argued extension of prior work showing when mobility controls backfire, but the headline claim rests on an unflagged assumption that infected people stay as mobile as healthy ones, and the SI is missing.","tokens_in":12666,"tokens_out":7644,"would_cite":true,"duration_ms":65215,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that unrestricted mobility from a high-risk to a low-risk community can reduce overall epidemic size, while a cordon sanitaire can increase it.","keywords":["mobility restrictions","cordon sanitaire","basic reproduction number","final epidemic size","Lagrangian residency-time model","Ebola","two-community epidemic model","travel bans"],"falsifier":"The claim would be falsified by an observed epidemic in two neighboring communities with R01 ≈ 2.45 and R02 ≈ 0.9 in which the total final number of cases under unrestricted mobility from the high-risk to the low-risk community is larger than the total under a full cordon sanitaire, contrary to the predicted threshold behavior.","tokens_in":11546,"feed_emoji":"🦠","tokens_out":5536,"duration_ms":58149,"temperature":0.7,"pith_summary":"The paper asks when mobility restrictions between a high-risk and a low-risk community actually reduce the final size of an epidemic. Using a two-community epidemic model in which people spend fractions of their time in each community, it finds that sealing off the high-risk community can make the total epidemic larger, while allowing high-risk residents to move freely into a safer community can make it smaller. The key is that infected people generate fewer secondary cases when they are in the low-risk, better-resourced community, so exporting infection can lower overall transmission even as it raises cases in the safer community. This matters because cordons sanitaires and travel bans are common policy tools, and the model gives concrete conditions under which they backfire.","feed_headline":"Lockdowns can worsen epidemics; open borders can shrink them","feed_subtitle":"A two-community model finds mobility thresholds that push the global reproduction number below one.","key_machinery":"The central object is a Lagrangian residency-time matrix P = (p_ij), where p_ij is the constant average fraction of time a resident of community i spends in community j, with i, j ∈ {1, 2}. This matrix converts two coupled communities into a single integrated epidemic system. Disease risk is summarized by community-specific basic reproduction numbers R01 > 1 and R02 < 1, and the global basic reproduction number R0(P) is computed using the next-generation matrix method. Two emergent thresholds carry the argument: t1^- (the HRC mobility level needed to push total final size below the cordon sanitaire baseline) and t1^+ (the HRC mobility level needed to make R0(P) = 1). The mechanism is that moving infectious people from high-transmission, resource-poor settings to low-transmission, resource-rich settings reduces the average number of secondary cases per infected person, shifting infections into the LRC while reducing the combined total.","core_discovery":"For an epidemic centered in a high-risk community (HRC) with local basic reproduction number R01 > 1, the total final epidemic size is a non-monotonic function of mobility from the HRC to a low-risk community (LRC). With one-way mobility (t2 = 0), low mobility levels (t1 < 0.45) raise the total final size above the cordon sanitaire baseline, moderate mobility (t1 > 0.45) lowers it below that baseline, and high mobility (t1 > 0.8) can drive the global basic reproduction number R0(P) below one, ending the outbreak even when the LRC's local R02 is slightly greater than one. Conversely, mobility from the LRC into the HRC can undo this benefit: when R02 ≈ 0.9 and t2 rises above about 0.23, no amount of HRC mobility can bring the global R0 below one. The thresholds depend on the risk levels in each community and on their relative population densities, with a very safe LRC (R02 < 0.35) making any HRC mobility beneficial and an unsafe LRC (R02 > 1.45) making the cordon sanitaire the better strategy.","pith_inferences":["If policy is set by each community independently, the cordon sanitaire remains locally attractive because it protects the low-risk community's own residents, even when it is globally harmful; a formal game-theoretic extension would likely predict over-restriction unless incentives are aligned.","The mechanism depends on infected, symptomatic people continuing to travel at the same rate as healthy people; behavioral data on sick individuals reducing travel would require lowering the mobility thresholds or could invalidate the control result in practice.","The same residency-time framework could be turned into a real-time decision tool: if local R0 values can be estimated from surveillance data during an outbreak, the thresholds t1^- and t1^+ give an immediate criterion for whether to relax or tighten travel restrictions.","Because the results hold for a range of disease types and depend only on risk differentials and density ratios, the qualitative conclusion that open mobility can outperform quarantine is likely to extend to other pathogens with similar transmission routes, though the specific thresholds would need recalibration."],"forward_implications":["When one-way mobility from the high-risk to the low-risk community is high enough, the global basic reproduction number can fall below one, meaning an ongoing outbreak is driven to extinction without any mobility ban.","A cordon sanitaire is beneficial only within specific risk windows: it helps when the low-risk community is unsafe (R02 > 1.45) but is the worst possible policy when the low-risk community is very safe (R02 < 0.35).","Mobility from the low-risk community into the high-risk community can cancel the benefit of HRC mobility, so travel advisories need to account for two-way movement, not just outflow from the outbreak zone.","Relative population density shifts the mobility thresholds: a larger high-risk population relative to the low-risk population lowers the mobility level needed for disease control, while a larger low-risk population raises it.","If the low-risk community's healthcare or sanitary conditions improve, the mobility threshold needed to control the outbreak falls, implying that investments in the safer community can substitute for movement restrictions."],"supporting_citations":[{"why":"Supplies the Lagrangian residency-time epidemic modeling framework on which the two-community model is built.","marker":"[30]"},{"why":"Earlier work showing cordons sanitaires do not always minimize total final epidemic size, which this paper extends by mapping when mobility helps or hurts.","marker":"[22]"},{"why":"Provides the temporal effective reproduction number estimates for the 2014 West Africa Ebola outbreak used to calibrate R01 = 2.45.","marker":"[16]"},{"why":"Reviews EVD transmission dynamics and control, supporting the local reproduction number values and disease parameters.","marker":"[32]"},{"why":"Estimates the reproduction number of Ebola during the 2014 West Africa outbreak, used alongside other calibration data.","marker":"[33]"},{"why":"Defines and computes the basic reproduction ratio in heterogeneous populations, providing the next-generation method used for R0(P).","marker":"[34]"},{"why":"Gives the next-generation matrix conditions for reproduction numbers and sub-threshold endemic equilibria used to determine when R0(P) < 1.","marker":"[35]"}],"fun_headline_variants":["Cordon sanitaire paradox: it can increase total cases","Open borders from hot zones can cut epidemic toll","When mobility bans backfire and fuel outbreaks","Sealing a hotspot may worsen the epidemic","Travel from high-risk areas can stop an outbreak"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The model assumes the fraction of time each resident spends in the other community is constant and identical for everyone, including infected and infectious residents; if sick people stop traveling once they feel ill, the mechanism that makes open borders reduce total cases weakens or disappears.","fun_headline_variants_meta":{"raw":{"variants":["Cordon sanitaire paradox: it can increase total cases","Open borders from hot zones can cut epidemic toll","When mobility bans backfire and fuel outbreaks","Sealing a hotspot may worsen the epidemic","Travel from high-risk areas can stop an outbreak"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000655,"raw_usage":{"total_tokens":3067,"prompt_tokens":1082,"completion_tokens":1985,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":698,"completion_tokens_details":{"reasoning_tokens":1914}},"tokens_in":698,"tokens_out":1985,"duration_ms":21273,"temperature":1.0,"reasoning_tokens":1914,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:19:52.009329+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The claim would be falsified by an observed epidemic in two neighboring communities with R01 ≈ 2.45 and R02 ≈ 0.9 in which the total final number of cases under unrestricted mobility from the high-risk to the low-risk community is larger than the total under a full cordon sanitaire, contrary to the predicted threshold behavior.","supporting_citations":[{"cited_title":"Sis and sir epidemic models under virtual dispersal.The Bulletin of Mathematical Biology, DOI: 10.1007/s11538-015-0113-5, 2015","cited_arxiv_id":null,"evidence_quote":"Supplies the Lagrangian residency-time epidemic modeling framework on which the two-community model is built."},{"cited_title":"Assessing the efficiency of movement restriction as a control strategy of ebola","cited_arxiv_id":null,"evidence_quote":"Earlier work showing cordons sanitaires do not always minimize total final epidemic size, which this paper extends by mapping when mobility helps or hurts."},{"cited_title":"Temporal variations in the effective reproduction number of the 2014 west africa ebola outbreak.PLoS currents, 6, 2014","cited_arxiv_id":null,"evidence_quote":"Provides the temporal effective reproduction number estimates for the 2014 West Africa Ebola outbreak used to calibrate R01 = 2.45."},{"cited_title":"Transmission dynamics and control of ebola virus disease (evd): a review.BMC medicine, 12(1):196, 2014","cited_arxiv_id":null,"evidence_quote":"Reviews EVD transmission dynamics and control, supporting the local reproduction number values and disease parameters."},{"cited_title":"Estimating the reproduction number of ebola virus (ebov) during the 2014 outbreak in west africa.PLoS currents, 6, 2014","cited_arxiv_id":null,"evidence_quote":"Estimates the reproduction number of Ebola during the 2014 West Africa outbreak, used alongside other calibration data."},{"cited_title":"Diekmann, J","cited_arxiv_id":null,"evidence_quote":"Defines and computes the basic reproduction ratio in heterogeneous populations, providing the next-generation method used for R0(P)."},{"cited_title":"reproductionnumbersandsub-thresholdendemicequilibria for compartmental models of disease transmission.Math","cited_arxiv_id":null,"evidence_quote":"Gives the next-generation matrix conditions for reproduction numbers and sub-threshold endemic equilibria used to determine when R0(P) < 1."}],"review_version":1}