key: cord-0259863-ts3llerc authors: Wang, Qiang; Shi, Naiyang; Huang, Jinxin; Cui, Tingting; Yang, Liuqing; Ai, Jing; Ji, Hong; Xu, Ke; Ahmad, Tauseef; Bao, Changjun; Jin, Hui title: Effectiveness and cost-effectiveness of public health measures to control COVID-19: a modelling study date: 2020-03-23 journal: nan DOI: 10.1101/2020.03.20.20039644 sha: 5729b586bc1957db0b1bdd33bdf1acae0819236a doc_id: 259863 cord_uid: ts3llerc Background The Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) was first reported in China, which caused a respiratory disease known as Coronavirus Disease 2019 (COVID-19). Since its discovery, the virus has spread to over 100 countries and claimed more than 4000 deaths. This study aimed to assess the effectiveness and cost-effectiveness of various response public health measures. Method The stochastic agent-based model was used to simulate the process of COVID-19 outbreak in scenario I (imported one case) and II (imported four cases) with a series of public health measures, involving the personal protection, isolation-and-quarantine, gathering restriction, and community containment. The virtual community was constructed following the susceptible-latent-infectious-recovered framework. The epidemiological and economic parameters derived from the previous literature and field investigation. The main outcomes included avoided infectors, cost-effectiveness ratios (CERs), and incremental cost-effectiveness ratios (ICERs). The sensitivity analyses were undertaken to assess uncertainty. Findings In scenario I and II, the isolation-and-quarantine averted 1696 and 1990 humans infected respectively at the cost of US$12 428 and US$58 555, both with negative value of ICERs. The joint strategy of personal protection and isolation-and-quarantine could avert one more case than single isolation-and-quarantine with additional cost of US$ 166 871 and US$180 140 respectively. The effectiveness of isolation-and-quarantine decreased as lowering quarantine probability and increasing delay-time. Especially in scenario II, when the quarantine probability was less than 25%, the number of infections raised sharply; when the quarantine delay-time reached six days, more than a quarter of individuals would be infected in the community. The strategy including community containment could protect more lives and was cost-effective, when the number of imported cases was no less than 65, or the delay-time of quarantine was more than five days, or the quarantine probability was below 25%, based on current assumptions. Interpretation The isolation-and-quarantine was the most cost-effective intervention. However, personal protection and isolation-and-quarantine was the optimal strategy averting more infectors than single isolation-and-quarantine. Certain restrictions should be considered, such as more initial imported cases, longer quarantine delay-time and lower quarantine probability. As of March 13, 2020, about 80 824 cases of coronavirus disease 2019 have been identified in China. 1 The global number of reported cases of COVID-19 has surpassed 110 000 and the confirmed cases of COVID-19 have been reported in more than 100 countries. 2 As date, the 21th century has witnessed several large-scale outbreaks of infectious diseases caused by coronaviruses. The cases infected with COVID-19 were significantly higher than ones infected with Severe Acute Respiratory Syndrome (SARS) and Middle East Respiratory Syndrome (MERS). The statistic China, restrictions on gathering referred to the restriction of crowd-gathering activities, especially catering and entertainment. The enforcement of community containment was a restriction on the movement of people throughout the community, minimizing human contact. 17 The incubation period and serial interval came from the estimation of Chinese Center for Disease Control and Prevention (CDC) and Guangdong Provincial CDC in the field work, 18, 19 and were considered fitting to the gamma distribution in the model. 3 The parameter of distance transmission probability has been reported in previous study. 6 The protective effectiveness of personal physical interventions derived from the cluster randomized controlled trial. 20 In our study, we converted odds ratio (OR) of handwashing and mask-wearing into the relative risk (RR), and calculated the (1-RR)/RR as the personal protection effectiveness. 21 In the model, we set the probability and delay-time for isolation and quarantine. The isolation delay-time meant that the time of dealing with patients lagged behind the time of infection onset, and the quarantine delay-time meant that the time of handling close contact lagged behind the time of exposing. Initially, we assumed that the index case (initial imported case) would be 100% isolated with no time delay (infecting others and isolation were carried out within the same day and infecting others preceded isolation). The quarantine probability was 100% and delay-time was two days. In the sensitivity analysis, the probability of quarantine of close contacts was set from 25% to 100% and the delay-time was from zero day to six days. The economic data derived from the field work and previous literature (table 1). The cost of personal protection included masks and handwashing (water and soap). The price of the mask was US$0·14 each and we assumed that two masks were used per person per day. 22 Given the soap using, the cost of handwashing per person per day was calculated as the formula provided in the previous study: 23 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 where the costpp = cost of hand washing, f= times of hand washing per day, and we set to six, v= volume of hand washing per time, and we set to 1000c.c/ml, Cwater = water cost per liter, and was US$0·00041, Csoap = cost of soap, and was US$2·85, t = the number of days soap available, and we set to 60. We assumed the day of personal protection was equal to the time from the day first case occurred to the last case recovered in the area plus 14 days. The cost of cases included the direct medical cost and indirect cost. We searched the cost of SARS patients to estimate the COVID-19 cases. In Guangzhou, China, the average hospitalization cost per patient was US$2900, and the average hospital stay was 17 days. 24 The average hospitalization cost achieved US$10 000 in Canada, 25 which was higher than that in China. 24 We estimated the average medical cost of US$6500 for COVID-19 patient. Referring to human capital approach in disease burden, 26 we estimate that the indirect cost of infected patient using per capita disposable income (PCDI)/365·25* (hospitalization days added rest days). The average rest days were estimated to seven days. We assumed that the cost of isolation would be included in the cost of hospitalization. The cost of quarantine of close contacts included direct and indirect parts. The cost of quarantine (accommodation and surveillance daily) per day was US$50 for each close contact. Similar to human capital approach in disease burden, 26 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Main health benefits of our study were avoided infections conducting measures versus no-interventions. The cost-effectiveness ratios (CERs) and incremental costeffectiveness ratios (ICERs) were calculated as the main cost-effectiveness outcomes. We calculated the CERs for interventions through cost divided by humans protected (uninfected) . The ICERs were calculated as the difference in the total costs between the intervention cohorts and non-intervention cohorts, divided by the difference in the total avoided infection. Positive ICERs showed the incremental costs required for avoiding 1 infected person. Negative ICERs indicated that intervention results in fewer costs while avoiding infected people than no intervention. The strategy was considered to be cost-effective if ICERs were lower than three times of per capita GDP. In 2018, the per capita GDP in China was US$9595. 28 We did not discount the cost because of the short One-and-two-way sensitivity analyses were performed to explore impact of the parameters in the range to test the robustness of the findings, including the epidemiological characteristics, interventions implement, and economic parameters. . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020. . Introduction of one case, each strategy could avoid the number of infectors and be costeffective compared with no intervention (table 2) . The isolation-and-quarantine was the most cost-effective intervention, avoiding 1696 cases and saving US$11 515 944 (ICERs < 0). The most protective single strategy was community containment, which avoided one more case than the isolation-and-quarantine at the additional US$549 186. Among the joint strategies, there was the lowest ratio of cost-effectiveness for the program A (CERs= 90 US$/ per human protected). The program A could avert one more infector comparing to single isolation-and-quarantine. In scenario II (table 3) , compared with no intervention, personal protection or gathering restriction was not cost-effectiveness (ICERs > three times of per capita GDP). The isolation-and-quarantine was still the most cost-effective, avoiding 1990 cases and saving US$13 372 397 (ICERs< 0). Compared with the isolation-and-quarantine, community containment could avoid one more case with the additional US$600 044. Among the joint strategies, there was the lowest ratio of cost effectiveness for the program A (CERs= 121 US$/ per human saved). Similarly, the program A versus single isolation-and-quarantine could avert one more infector. The number of infectors depended on transmission constant in scenario I (appendix 4 ). Varying the transmission constant from the 0·25 to two, the isolation-and-quarantine was the most cost-effective single intervention, and program A was the most cost-effective joint intervention. . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020. . https://doi.org/10. 1101 /2020 The number of imported cases was a key parameter influencing the effectiveness and cost-effectiveness analysis. There were not significantly differences in effectiveness between the program A and C, when the imported cases were set to ten or 20 (figure 1a and appendix table 5). When the imported cases were no less than 50, the program C including community containment could effectively decrease the infectors than program A including isolation-and-quarantine, but the former was not cost-effective. The CERs of interventions increased significantly as the increase of imported cases (figure 2a). The threshold analysis showed that program C became cost-effective (ICERs< three times of per capita GDP) comparing to program A when initial cases increased to imported 65 cases (appendix table 6). The isolation delay-time did not contribute to the spread of infections in scenario I (figure 1b). The increase of isolation delay time, however, caused a significant increase in the number of infections in scenario II. When the isolation delay of four index cases reached four days, there were more than 15 humans being infected, which was three times as the one without isolation delay. The CERs of interventions increased as the increase of the isolation delay-day (figure 2b). The program A dominated the program C in scenario I and II within the sensitivity analysis of isolation delay-time (appendix table 7) . The effectiveness of isolation-and-quarantine was sensitive to the low quarantine probability. When the tracing probability of close contact was reduced to 25%, the number of people infected increased significantly, especially in the scenario II ( figure 1c ). In scenario I and II, the effectiveness of outbreak controlling was close between program A and C when the probability of tracing above 50% (appendix table 8 and table 9 ). The CERs decreased as the increase of quarantine probability, and was most unstable when the quarantine probability was 25% (figure 2c). In scenario I, the program C was not cost-effective comparing to program A. The ICERs of program C was close to three times of per capita GDP when the quarantine probability was 25% in scenario II. The is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020. . https://doi.org/10. 1101 /2020 threshold analysis showed that program C became cost-effective (ICERs< three times of per capita GDP) comparing to program A when quarantine probability was below 28% (appendix table 10). Varying the quarantine delay time from zero day to four days, it had little influence on averting infected cases (figure 1d). When the tracing delay-time of close contacts was extended to six days, the number of people infected increased significantly (appendix table 11 and table 12 ). In scenario II, when quarantine delay-time reached six days, there were likely more than 500 humans being infected, accounting for a quarter in the space. The CERs of interventions was unstable when the quarantine delay-time was no less than five days (figure 2d). Comparing with program A, the program C was costeffective when the delay-time more than five days in scenario I and four days in scenario II respectively (ICERs< three times of per capita GDP). Varying the cost of patient from US$2900 to US$10 000, the CERs of interventions increased and ICERs of interventions comparing to the non-intervention decreased (appendix table 13 and table 14) . The most cost-effective strategy was isolation-andquarantine in scenario I and II. In scenario I, the effectiveness of outbreak controlling was not sensitive to the transmission constant and quarantine probability (appendix table 15). When the transmission constant was set to two, the outbreak could be controlled by the 25% probability quarantine. However, as the transmission constant increased in scenario II, the control of outbreak required higher quarantine probability. When the quarantine probability was 25% and transmission constant was two, it was likely about a quarter of people would be infected in scenario II (appendix table 16). The program A dominated the program C in the scenario I and II in general. When the transmission constant was above one and the quarantine probability was below than 25%, the . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020. . https://doi.org/10. 1101 /2020 probability could accelerate the outbreak of COVID-19. The effectiveness and cost-effectiveness of interventions were sensitive to the initial imported cases. The increase of imported cases could lead to the increase of risk of COVID-19 infection, even conducting the strict interventions. We suggested that the infectors avoided by isolation-and-quarantine and community containment were not significantly when the imported the cases below 20. When the imported cases reached 50, community containment could avoid more cases significantly. The strategy including community containment was cost-effective when imported cases reached 65, the 3·25% of the community population (2000 humans). The current article found that the initial number of cases had an effect on the effectiveness of interventions. 30 The choice of optimal strategy depended on the setting parameter of interventions. We compared the strategy of personal protection and isolation-and-quarantine (program A) with strategy of personal protection and community containment (program C). Generally, program A was cost-effective versus program C. however, the program C was cost-effective at the 25% probability and more than two quarantine delay-days, or 50% probability and no less than five quarantine delay-days in the sporadic outbreak area. The program C would dominate the program A at the 25% quarantine probability or quarantine delay-time was more than three days in the cluster area. The effectiveness of isolation and contact tracing was associated with the extent of transmission before symptom onset. 31 The proportion of asymptomatic infection would contribute to the outbreak of COVID-19, 30 which was consistent with our findings. In our study, the community containment would be more efficient and cost-effective when the quarantine delay-time was more than latent period. We suggested that increase of is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 There were some limitations in the study. First, COVID-19 was recently emerged disease first reported in Wuhan, China, therefore the availability of epidemiological data is insufficient. We set the study parameters referring to the existing published epidemiological studies and adopted the Gamma distribution to some of the parameters, which could improve the precision of estimate. Second, the cost of societal interventions was difficult to estimate. In our study, human capital approach was borrowed which might more conservatively estimate the cost. The cost of the disease would also increase, if according to the actual situation in Wuhan, China. Third, our model simulated a local area with 2000 humans, which may result in limited extrapolation ability. Finally, the simplification of the model will have some biases compared with the real situation, because the flow of people will be affected by many factors. In the sporadic and cluster outbreak area, the isolation-and-quarantine was the most cost-effective intervention. The personal protection and isolation-and-quarantine was the optimal joint strategy averting more cases than single isolation-and-quarantine. Rapid and effective isolation and quarantine could control the outbreak of COVID-19. The strategy including community containment could be more effective and costeffective when low probability and long delay of implements of interventions or much imported cases. . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted March 23, 2020 . . https://doi.org/10.1101 /2020 Economic evaluation of the routine childhood immunization program in the United States