1.1. COVID-19 Pandemic and Reactions by Society
Coronavirus disease 2019 (COVID-19), a respiratory disease caused by a novel coronavirus that initially emerged in the city of Wuhan at the end of 2019 [1
], has quickly spread all over the world. Because of its novelty, which means the lack of specific medicine for it, the dominant countermeasures are isolation and supportive medication. This indicates that a large number of hospital beds will be needed. Thus, control strategies such as early diagnosis, isolation, and hospitalization are essential. A lack of strategy can lead to the collapse of the healthcare system if it is overwhelmed by patients.
In handling the complex COVID-19 transmission processes in the population and the effects of societal factors, the idea to use system dynamics, describing complex social systems as a collective set of mathematical equations, was drawn based on some considerations. First, a stock-flow model in system dynamics adequately describes population transition, including delay in time course. Moreover, the effects of social factors on disease transmission can be mathematically modeled with minimal complexity. Second, a causal loop diagram, also used in system dynamics, can describe feedback systems in society, which is important in social reaction.
System dynamics have often been used in health systems, and in addressing health problems such as obesity, diabetes, hypertension, mental health, mortality, smoking, infectious diseases, injury by violence, respiratory diseases, substance abuse, disability, quality of life, and maternal and child birth complications [2
]. As an infectious disease, human immunodeficiency virus (HIV) transmission has been studied. Batchelder et al. conceptualized the effects of social and ecological conditions affecting women at risk of HIV [3
]. Weeks et al. studied the effects of the HIV test and treatment care continuum on community viral load [4
]. None of these studies addressed the quantitative features of disease transmission. On the other hand, a subpopulation or sector frame-based quantitative model has been used as a subcomponent of the model, such as population change over time in an obesity study [5
] or demographics of the elderly in a health care systems study [6
]. Based on these, the use of a population-based stock-flow model to describe disease transmission was thought to be feasible.
This study primarily aimed to clarify the effect of intervention by a modeling approach. In addition, the study sought to explore social factors of the effectiveness of the control of new infectious diseases. As an overall approach, a COVID-19 epidemic case in Japan in spring 2020 was analyzed using system dynamic modeling.
1.2. Previous Studies and Research Questions
This study examines four research questions. (1) Is a modeling approach effective in overcoming the lack of information (actual number of infected patients)? (2) What are the most important measures to prevent the spread of infection? (3) What are the factors that influence infection among societal factors (in demography and behavior)? (4) What is important in the future response to new infectious diseases?
There are several studies dealing with measures against COVID-19. Yan et al. summarized various countermeasures by different authorities [1
]. Summarized recommendations are mainly on behavior (washing hands, keeping rooms ventilated and sanitized, wearing masks, avoiding social activities, staying away from crowded areas, and observing social distancing). Dickens et al. used an agent-based model to test the effectiveness of home-based and institutional isolation. The analysis clarified the usefulness of institutional containment and risks of home-based isolation [7
]. Gerli et al. investigated the lockdown effort of European countries, and pointed out the importance of timeliness of lockdown [8
Still, no studies encountered the ambiguity of the COVID-19 situation in Japan. Making the best use of the flexibility and simplicity of system dynamics, this study aims to grasp the whole picture of the COVID-19 outbreak in Japan using abundant information on demography and behavior. To detect potential regional differences, three regions that have enough confirmed cases and have urban cities were analyzed. They were Tokyo (the capital), the Osaka prefecture, and the Hokkaido prefecture. To avoid complexity, prefectures that have satellite cities were not analyzed.
1.3. Analytical Approach Applied in This Study
As analytical approaches, the following five analyses were performed. First, the effects of medical countermeasures were illustrated using causal loop analysis. Second, a stock flow model describing the mass of infected population was developed to analyze the dynamics of infection. Third, the effectiveness of actions preventing the saturation of medical capacity was tested by simulation. Fourth, the relationships between transmission reduction efficiency and regional differences in social factors were explored. Finally, the effects of social factors on disease preventive behavior were analyzed using causal loop analysis to provide suggestions for a sustainable society beyond new infectious diseases.
The rapid spread of COVID-19 is threatening health systems with capacity challenges. The United States, with the largest number of patients as of March 2020, seems to be challenged by a healthcare capacity problem [23
]. The inpatient bed occupancy rate varies by region, with the highest being 79% (in Maryland, May 2020) [24
]. The highest intensive care unit (ICU) bed occupancy rate is 84% (in the District of Columbia, May 2020).
Japan, with only 7.3 beds per 100,000 inhabitants [25
], is one of the countries that suffer from hospital bed shortage [26
]. As of April 2020, there were 12,500 beds for novel infectious diseases nationwide, while there were 10,000 patients in Japan [27
]. Although 31,383 hospital beds were ensured by 21 May 2020 [28
], health systems are still at risk. Some local governments even have plans to provide care to low-risk patients in hotels. Under this condition, a control strategy to reduce the peak number of patients remains important.
To reduce the transmission as an effort to delay and lower the epidemic peak, governments imposed restrictions on movement in local communities. Many countries, such as China, Italy, the US, and the UK, locked down their cities to prevent the spread of the disease. In Japan, a state of emergency was declared on 7 April 2020, giving authorities the power to enforce stay-at-home orders and to close businesses. Although Japanese authorities did not describe the countermeasures as a lockdown, prefectural authorities asked people to refrain from traveling across prefectures, unnecessarily going out, and to stay away from public gatherings [29
]. In addition, all schools were closed. Initially, this affected the capital, Tokyo, and six other prefectures (Saitama, Chiba, Kanagawa, Osaka, Hyogo, and Fukuoka). Subsequently, it was expanded nationwide on 16 April 2020.
Information on the effectiveness of these interventions is warranted, but the whole context of what is happening is not well understood because of the limited testing capacity. No one knows the actual number of patients infected. To overcome this, the use of a structured model with an apparent/inapparent infection ratio and efficiency of virus testing was considered.
This study primarily analyzed COVID-19 transmission dynamics and the effects of initial measures by governments. A special feature of the current model is that it includes symptom rate, virus testing capacity, and hospital capacity, which were major concerns in the early phase of the COVID-19 outbreak. The current model heavily depended on demographic data and has fewer accompanying variables in comparison to prior system dynamics studies [4
]. Consideration of more detailed variables, especially health-protective behaviors such as the practice of hygiene or physical distancing measures, may help identify important factors in basic societal systems regarding disease prevention. In addition, the association of human activity and temperature or humidity is possible [30
]. Specific research on each component to provide detailed information is warranted.
In the causal loop diagram, the importance of reducing contacts was highlighted. A strategy to obtain social immunity by allowing infection is theoretically possible; however, realistically, allowing infection leads to an increase in deaths through an increase in disease transmission and inadequate medication. Stock and flow analysis confirmed that an increase in infections overwhelms healthcare systems.
The stock-flow model adequately described the dynamics of the COVID-19 outbreak in three Japanese regions. Baseline isolation effect in the early phase was negligible in Tokyo, little in Osaka, and considerable in Hokkaido. Primarily, this could be related to population density as supported by transmission theory [21
] and observations in the United States [30
]. The hypothetical mechanism for the associations between population density and transmission proposed by Rubin et al. is increased droplet transmission and potentially airborne transmission in close proximity [30
]. After the state of emergency declaration in April, transmission efficiency of the disease markedly decreased to 17–31% of the baseline. In this case, the disease was primarily controlled by national and local government interventions. The most important measure was the reduction of contacts in the early phase of the outbreak by national and local governments.
Attempts to build a causal loop diagram for interrelationship analysis revealed that no self-strengthening dynamics were noticeable in the society. This indicates that interventions by the government were essential in the meantime. As a potential reinforcing loop, a loop with new business practice and awareness raising regarding physical distancing and hygiene measures was hypothesized. Any other well-recognized component did not construct any reinforcing loops. Further investigation to confirm the self-strengthening dynamics, beginning with new business practice, and efforts to strengthen such dynamics are warranted for a sustainable society.
The strength of this study is the use of SD techniques. Stock-flow modeling is relatively simple, but it was effective in showing the overall dynamics of virus transmission when virus testing was inadequate. Stock-flow modeling also enabled estimation of the impact of interventions. Further simulation is possible for virus testing efficiency, hospital capacity, and a new medicine.
The limitations of this study are as follows. The stock-flow model utilized simple arithmetic operations and described the average dynamics of a population. This does not adequately describe the probability process that should be demonstrated by more complex models or multi-agent models. The model was constructed based on the fundamental monitor and control strategy in Japan but detailed approaches may have slightly changed over time based on local government’s policy. In addition, model validity was not fully investigated although basic validity, such as dimension consistency and consistency of predicted and real positives, was checked. The approach of leaving one parameter as endogenous made extensive validation somewhat challenging. This could be overcome by comparing multiple regions as external validation; however, no other region in Japan has enough patients to be used in building models with comparative accuracy. Nevertheless, the model sufficiently described the outbreak of COVID-19 in three Japanese regions and was useful in describing the early phase of the outbreak. A more precise investigation should be conducted in the future for the development of science. The basic structure of the current stock-flow model reflects Japanese national response strategy for COVID-19 to limit virus testing to patients with obvious symptoms for better use of diagnostic resources. This makes comparing the effectiveness of measures across countries difficult, which is an important theme with this new disease. Nevertheless, this approach enabled determination of the possible effects caused by saturation of virus testing, which was important in analyzing the effectiveness of measures undertaken by Japanese authorities. Further analysis using newly collected epidemic data and more detailed social activity data is warranted in the future.