Sunday, April 12, 2020
Effective modelling and good political judgement needed
Throughout this crisis all the governments have insisted that they are following the best scientific advice in the decisions they are making, however as David McCoy, a professor of Global Public Health and director of the Centre for Public Health at Queen Mary University of London, points out in this excellent article, technical evidence can only take us so far.
He says that there is a non-scientific element to decision-making which involves choosing between competing demands and needs in society, determining what is ethical and moral, and balancing challenges that are current and immediate with those that will only emerge in the future. For example, a model that incorporates value judgments is needed to balance the direct, visible and dramatic harms of Covid-19 with the more indirect, chronic and hidden social and economic harms of lockdown.
He illustrates this by pointing out that countries that lived through the experience of Sars and Mers appear to have been better prepared to deal with Covid-19 than those that didn’t. And in Canada, the success of British Columbia in controlling Covid-19 compared with other provinces has been partly attributable to the combined ingredients of political experience and good judgment.
He alludes to Professor Neil Ferguson who has been described as being “more influential right now than any politician”. Ferguson leads a team of modellers at Imperial College London whose mission has been to predict the pattern and health impact of coronavirus transmission in the UK under different scenarios, and thereby help government make the right decisions about how best to protect the population.
Andrew Marr described Ferguson as being the most influential person in the country, and while that may have ben rhetorical, McCoy argues that it hints at the perception that epidemiological modelling has played a disproportionately influential role in shaping the UK’s strategy and plans. The rest of his article and the conclusions therein are worth quoting in full:
Others have pointed out that some of the earlier modelling work had gaps that have adversely affected policy decisions. These include not accounting for the potential role of testing and contact tracing in mitigating the epidemic; not modelling the impact of sick patients on intensive care beds and other parts of the health system; and not modelling the epidemic across different regions of the UK.
It is not clear what caused these omissions. Perhaps they were the result of blindspots that persisted due to the lack of external scrutiny and challenge. In the case of the sidelining of testing and contact tracing, perhaps it was because the Scientific Advisory Group for Emergencies (Sage) had concluded that these measures were not worth while – a view that is understandable if Sage also saw the building up of herd immunity through natural infection as part of its strategy.
There has been lots of speculation as to why the UK delayed lockdown compared with other countries; why it was slow to increase viral testing capacity; and why the NHS was left so poorly prepared. Now is not the time to look backwards. Neither would I want to place any criticism on modellers. Rigorous modelling is done by clever modellers; but the design and testing of models, including the assumptions built into models and the questions that modellers have been asked to help answer, is a collective responsibility.
Modelling is vital and could assist greatly in going forward. In particular, two things should be urgently modelled. The first is to model the impact of different strategies that will help us avoid the harms of draconian, population-wide lockdown. This should include modelling the costs and benefits of weekly viral testing of the whole population (not 100,000 tests a day, but 10 million tests a day), or a more modest but greatly expanded testing regime, combined with aggressive case detection and contact tracing, coupled with continued physical distancing and improved hygiene. The former idea may sound radical and extreme, but is a serious proposal that could be implemented if there was sufficient political will.
Second, we must model (and then keep monitoring) the health and socioeconomic impacts of both the virus and the control measures, with short-, medium- and long-term timeframes, and with a social lens to look at how the virus has impacted differentially across society. For some reason, there was no modelling of the socioeconomic impact of lockdown. This was an omission – and as we enter the beginning of what could be a global economic recession, we could see thousands of lives cut short from this collateral damage.
Finally, from a policymaking perspective, there are two things needed. First, we need greater transparency and participation to allow effective scrutiny and challenge from scientists and public health specialists who are more independent of government. This will avoid the danger of groupthink and blindspots. Second, we must better acknowledge the limitations of evidence-based policy and recognise the importance of expertise, experience and sound judgment.
Some sage advice here.
He says that there is a non-scientific element to decision-making which involves choosing between competing demands and needs in society, determining what is ethical and moral, and balancing challenges that are current and immediate with those that will only emerge in the future. For example, a model that incorporates value judgments is needed to balance the direct, visible and dramatic harms of Covid-19 with the more indirect, chronic and hidden social and economic harms of lockdown.
He illustrates this by pointing out that countries that lived through the experience of Sars and Mers appear to have been better prepared to deal with Covid-19 than those that didn’t. And in Canada, the success of British Columbia in controlling Covid-19 compared with other provinces has been partly attributable to the combined ingredients of political experience and good judgment.
He alludes to Professor Neil Ferguson who has been described as being “more influential right now than any politician”. Ferguson leads a team of modellers at Imperial College London whose mission has been to predict the pattern and health impact of coronavirus transmission in the UK under different scenarios, and thereby help government make the right decisions about how best to protect the population.
Andrew Marr described Ferguson as being the most influential person in the country, and while that may have ben rhetorical, McCoy argues that it hints at the perception that epidemiological modelling has played a disproportionately influential role in shaping the UK’s strategy and plans. The rest of his article and the conclusions therein are worth quoting in full:
Others have pointed out that some of the earlier modelling work had gaps that have adversely affected policy decisions. These include not accounting for the potential role of testing and contact tracing in mitigating the epidemic; not modelling the impact of sick patients on intensive care beds and other parts of the health system; and not modelling the epidemic across different regions of the UK.
It is not clear what caused these omissions. Perhaps they were the result of blindspots that persisted due to the lack of external scrutiny and challenge. In the case of the sidelining of testing and contact tracing, perhaps it was because the Scientific Advisory Group for Emergencies (Sage) had concluded that these measures were not worth while – a view that is understandable if Sage also saw the building up of herd immunity through natural infection as part of its strategy.
There has been lots of speculation as to why the UK delayed lockdown compared with other countries; why it was slow to increase viral testing capacity; and why the NHS was left so poorly prepared. Now is not the time to look backwards. Neither would I want to place any criticism on modellers. Rigorous modelling is done by clever modellers; but the design and testing of models, including the assumptions built into models and the questions that modellers have been asked to help answer, is a collective responsibility.
Modelling is vital and could assist greatly in going forward. In particular, two things should be urgently modelled. The first is to model the impact of different strategies that will help us avoid the harms of draconian, population-wide lockdown. This should include modelling the costs and benefits of weekly viral testing of the whole population (not 100,000 tests a day, but 10 million tests a day), or a more modest but greatly expanded testing regime, combined with aggressive case detection and contact tracing, coupled with continued physical distancing and improved hygiene. The former idea may sound radical and extreme, but is a serious proposal that could be implemented if there was sufficient political will.
Second, we must model (and then keep monitoring) the health and socioeconomic impacts of both the virus and the control measures, with short-, medium- and long-term timeframes, and with a social lens to look at how the virus has impacted differentially across society. For some reason, there was no modelling of the socioeconomic impact of lockdown. This was an omission – and as we enter the beginning of what could be a global economic recession, we could see thousands of lives cut short from this collateral damage.
Finally, from a policymaking perspective, there are two things needed. First, we need greater transparency and participation to allow effective scrutiny and challenge from scientists and public health specialists who are more independent of government. This will avoid the danger of groupthink and blindspots. Second, we must better acknowledge the limitations of evidence-based policy and recognise the importance of expertise, experience and sound judgment.
Some sage advice here.
