fbpx Q&A: What COVID-19 policies should governments impose? Comparing 10 interventions, weighing cost and quality-of-life trade-offs | Harvard Kennedy School

Soroush Saghafian.Assistant Professor of Public Policy Soroush Saghafian, whose academic interests include applying operations research methods to health care management, has been working with the government of Bahrain to analyze the effectiveness of policies to address the coronavirus. Saghafian, who is a faculty affiliate of the Harvard PhD program in health policy and the Harvard Center for Health Decision Science, cautions that no one policy is best in all cases—and that governments must weigh cost and quality of life considerations. However, he says that closing businesses like cinemas and gyms for four months could be one of the most low-cost and effective measures. Saghafian shared analysis that he and his former PhD student (now an assistant professor of statistics and family medicine at Michigan State University) Alireza Boloori conducted in an HKS faculty working paper, “COVID-19: What Intervention Policies Are Most Effective? A Brief Report Using Data from Government of Bahrain.”


Q: You have worked with the government of Bahrain to analyze data on what intervention policies to address COVID-19 might be most beneficial, comparing 10 possible policies (including no intervention). How did you determine this set of interventions?

We determined these policies based on the needs of the government of Bahrain (which has one of the highest per capita COVID-19 testing programs in the world) and after direct conversations with them. However, these policies are common policies such as the closure of schools, malls, and retail stores, or the suspension of cinemas, gyms, home delivery, and takeaway food that are also pursued in other countries. As such, many of our findings might be useful for other countries that want to know the answer to a simple but important question: which intervention policies, if imposed, will be most effective?


Q: Can you talk about which interventions are predicted to be most beneficial and why?

We compare intervention policies by predicting their impact on different outcomes. To this end, we use a compartmental analytical model known as SEIRS, which considers subpopulations such as susceptible, exposed, infected, and recovered people. Our model is calibrated with the data that we have received from the government of Bahrain. However, we have also augmented our data by combining it with data from other sources, including from the Centers for Disease Control and Prevention (CDC), the World Health Organization (WHO), other countries, like China, as well as some related studies in the literature. To compare different policies, we take into account their predicated impact on (a) the number of people susceptible, exposed, infected, and recovered, (b) required hospital resources (common hospital beds, hospital ICU beds, and number of ventilators), and (c) the number of deaths. Finally, we compare the impact of these policies by performing cost-effectiveness analysis, in which we measure the impact of these policies simultaneously in terms of costs and citizens’ overall quality of life.

As is expected, one might choose to impose a different policy depending on which measure an individual considers to be most important. There is typically no single policy that can dominate others in terms of all measures. Furthermore, the effectiveness of some policies depends on the time period during which they are imposed. For instance, it matters how long we keep schools closed. Nevertheless, we see that generally there is a tradeoff in cost versus quality of life (measured by a metric known as quality-adjusted life years (QALY)): as the duration of imposed policies increases so do both costs and improvements in a population’s overall quality of life. However, our results indicate that policies such as suspension of cinemas, gyms, home delivery, and takeaway food for about 120 days are great options in that, with a relatively low-cost impact, they can yield significant improvements in citizens’ overall health.


Q: Did anything surprise you in your analysis?

Some of the surprising findings are related to the differences between what we observe from data in Bahrain versus those reported in the United States. For instance, the average length of stay of an infected patient in Bahrain is about 11.5 days while in the United States it is about eight days (for a patient using a common hospital bed). Similarly, the average rate of hospitalization in Bahrain is about 6 percent but, in the United States it is about 25 percent.


Q: Are there any conditions specific to the context of Bahrain that you had to take into account? Or are the recommendations for Bahrain widely applicable elsewhere?

There are both differences and similarities. While some of the baseline rates differ from country to country, the main insights we have gained could be useful for other countries as well. To ensure this, we have also used data from other sources and countries including data from the CDC, WHO, and China.


Q: Are there any next steps following from your report? Will the government of Bahrain plan on implementing any of the interventions?

We have communicated our main results to them. Based on our results and recommendations, we do hope that they take correct actions. We are still in the midst of collaboration, and this continued collaboration will enable us to ensure that this will happen.

I would like to especially thank Hamad Faisal Almalki (Bahrain’s undersecretary of national economy, Ministry of Finance and National Economy), Manaf Husain Alsabbagh (Bahrain’s director of the planning and economic studies directorate, Ministry of Finance and National Economy), and their team members for making their data available to us. I am also thankful to my former PhD student (now a professor of statistics and family medicine) Alireza Boloori who performed analysis in a timely manner.