At the End of Every Pandemic: Beginning a Pandemic Playbook to Respond to the Next.

The world was unprepared for COVID-19. Pandemics can unfold quickly; faster than governments can respond, unless they have maintained a realistic pandemic playbook. As the world ahead becomes ever-more complex, such playbook becomes ever-more necessary. This article not only describes the importance of a pandemic playbook but also a system to maintain it. A pandemic playbook both (1) specifies what is needed to respond to a pandemic and (2) provides a lens through which to identify measures that will keep people safe and society secure. The plays in the book are thought-though policies and strategies and corresponding implementation plans. The process of developing a playbook is as important as the product. Any playbook must be fit for purpose in the context of the times in which it is to be used. Above all, it must contain realistic policies and plans that can actually be implemented and can realize their intended effects. Achieving this goal requires (1) repeatedly exercising the playbook so that people know what to do when they need to do it and (2) evaluating results and updating the playbook to keep it relevant and current. Necessarily, to bring ideas alive, this article illustrates them with reference to COVID-19 and earlier pandemics, but it is not intended as a playbook for responding to the next pandemic; nor a postmortem on responses to COVID-19. Instead, it describes actions to take now to be ready when the next global pandemic strikes, so that policy decision-makers will not be lamenting "we should have done that."

The Global Health Security Index: Another Look.

The Global Health Security Index (GHSI) was published in October 2019 (after more than 2 years in preparation); at about the same time that the COVID-19 pandemic (COVID) started. The GHSI was intended to score countries' pandemic preparedness. Within months of the start of the pandemic, articles began to be published that claimed to assess the GHSI's validity. They correlated GHSI scores with countries' COVID per capita death rates. They showed that the better prepared a country, the higher the death rate: a result that was counter to what would have been expected. This article takes another look at the GHSI by exploring the relationship in major European Union countries plus the United Kingdom. The analysis reported here confirms that early on the higher the GHSI score, the higher the COVID per capita death rate (r = 0.52, P < 0.05). But, by the end of 2020, there was no correlation. By July 2021, the correlation was in the expected direction: the higher the GHSI score, the lower the COVID per capita death rate (r = -0.55, P < 0.05); ditto case fatality rate (r = -0.74, P < 0.01). Further, the GHSI was better correlated with excess mortality, the best measure of pandemic impact (r = -0.69, P < 0.01). However, per capita GDP was as good a predictor of excess mortality (r = -0.71, P < 0.01) and the Health System Performance Index of case fatality rate (r = -0.71; P < 0.01). By the end of 2021, the correlation between GHSI scores and COVID per capita death rates had strengthened (r = -0.71; P < 0.01). This exploratory analysis is not intended to produce generalizable conclusions about the effectiveness of countries' COVID pandemic response management, which continues to evolve and hence can only be properly assessed after the pandemic has ended. Nevertheless, the following conclusions would seem to be warranted: 1) there seems to have been a rush to judge, or, at least, to publish, and 2) the validity of any forward looking pandemic preparedness score depends not only on being able to assess countries' capabilities but also being able to forecast what governments will (and will not) do in any given situation, a seemingly quixotic quest.

Machine Learning for Health: Algorithm Auditing & Quality Control.

Developers proposing new machine learning for health (ML4H) tools often pledge to match or even surpass the performance of existing tools, yet the reality is usually more complicated. Reliable deployment of ML4H to the real world is challenging as examples from diabetic retinopathy or Covid-19 screening show. We envision an integrated framework of algorithm auditing and quality control that provides a path towards the effective and reliable application of ML systems in healthcare. In this editorial, we give a summary of ongoing work towards that vision and announce a call for participation to the special issue  Machine Learning for Health: Algorithm Auditing & Quality Control in this journal to advance the practice of ML4H auditing.