IANPHI Hosts COVID-19 Webinar with IHME on Pandemic Forecasts

In partnership with the Institute for Health Metrics and Evaluation (IHME), IANPHI held a webinar on June 3 to discuss IHME’S COVID-19 pandemic forecasts with NPHIs and key stakeholders. The webinar focused on the evolution of the IHME COVID-19 model (currently using the multi-stage hybrid modeling framework; soon to be expanded to its “third generation model” – the RKCS-SEIR hybrid model), key data sources incorporated into the model, the utilization of the model for policy planning, as well as the implications of the COVID-19 forecasts for national authorities and national public health institutes (NPHIs).

Dr. Jean-Claude Desenclos, scientific director and deputy to the director general at Santé publique France, and secretary general of IANPHI, moderated a discussion with Dr. Christopher J. L. Murray, director of IHME, professor and chair of the department of Health Metrics Sciences at the University of Washington. 

Dr. Murray’s presentation for this webinar is available here.

Based at the University of Washington in Seattle (USA), and with many of its collaborating institutions and partners already among the 100+ IANPHI members, IHME is an independent population health research institute working to provide impartial, real-time and comprehensive scientific evidence on global health trends, its determinants and the performance of health systems through an ongoing production of the Global Burden of Disease (GBD) study.

IHME currently has COVID-19 forecasts at the national level for 56 countries and subnational-level for seven countries. These publicly available forecasts provide data on daily and cumulative COVID-19 deaths, infections/new cases, testing rates, and hospital resource use. IHME will shortly release the third generation of its model (RKCS-SEIR hybrid model) for all locations. The model’s statistical algorithm builds on internal performance evaluations since March 2020. The new model will produce forecasts for all countries worldwide and will strengthen future predictions. With the new iteration of the model, time window will be extended each month allowing for four-month predictions, on average. 

Some of the key questions that the model seeks to answer include the number of COVID-19 deaths likely to occur, the peak of daily deaths, the impact of changes in mobility across countries, the variations between trends and patterns among countries, as well as what hospital and ICU bed shortages for COVID-19 patients are likely to be faced. The model relies on epidemiological and health system outputs, and produces a reference forecast (what is most likely to happen), while incorporating new data and evidence as they become available.

IHME forecasts draw from a range of data sources, including but not limited to the John Hopkins University COVID-19 Resource Center, American Hospital Association, Our World in Data, government websites, the World Health Organization (WHO), the Organization for Economic Co-operation and Development (OECD), EUROSTAT, the statistical office of the European Union Eurostat; published studies, as well as through local and national data provided by IHME’s key country and regional collaborators. In many cases, data sets are supplemented by government data sources to ensure forecasting is as accurate and up-to-date as possible. Using multiple data sources enables IHME to produce models for countries where there may be an under-reporting of deaths and testing. 

Where possible, IHME consults national authorities or key GBD collaborators for feedback on forecasts. IHME’s partnership activities have become increasingly important elements of COVID-19 activities. These dialogues, including with national authorities, can be important in verifying the accuracy of data. Furthermore, they enable models to include up-to-date surge capacity implementation in national systems that may not have been captured by international monitoring.

IHME has evaluated performance comparisons of its COVID-19 forecasts with other publicly available models. IHME has produced relatively fewer prediction errors at four weeks and has remained largely prudent when providing forecasts for countries in sub-Saharan Africa. There are as of yet few accurate forecasts available for this particular region. 

IHME integrates various determinants into its forecasts, from seasonality to social distancing mandates implemented by governments. By integrating external predictors, IHME is able to adapt its models to evolving seasonal and policy environments. In particular, pneumonia deaths by week, mobility, mask use, testing per capita and population density are among the variables currently integrated into IHME’s modeling. Pneumonia admissions by seasons appear to have a strong statistical relationship with COVID-19 mortality rates. Testing rates, and by extension tracing and isolation, also seem to have a statistical significant relationship with COVID-19 transmission rates. Although, there are significant variations of testing rates within and between regions across the world.

National governments are the drivers of a number of determinants involved in IHME’s modeling. The early easing of social distancing mandates is likely to have important statistical relationships with COVID-19 transmission and mortality rates. IHME would be interested in further exploring the statistical significance of individual mandates, such as partial business and school closures and restrictions of group gatherings. Where countries are already lifting certain mandates, this data is already integrated in the IHME model. 

IHME’s COVID-19 forecasting tool can be of particular relevance for anticipating equipment and infrastructure needs as well as for studying policy scenarios. Since March, hospitals and primary care institutions across the U.S. have been using IHME’s COVID-19 forecasts to plan for a surge in demand for hospital and ICU beds, ventilators, and personal protective equipment. The U.S. government has been using the model to inform the nationwide mandates on social distancing; and IHME is in regular contact with the White House COVID-19 Taskforce. 

IHME’s COVID-19 model and forecasts have also been used by the European Commission – via its ‘clearing house for medical equipment’ to allocate personal protective equipment and medical equipment resources across the EU member states. In addition, the model results have been shared amongst a number of EU Member States’ stakeholders. 

Looking to future perspectives, IHME predicts that transmission rates are likely to rise as mobility increases in many parts of the world. For the institute, the increasing testing rates and seasonality effects are currently the two largest factors contributing to falling transmission rates in certain countries. Based on the statistical significance observed by IHME between pneumonia admissions, seasonality and COVID-19 deaths, the Northern Hemisphere may be set to experience increased growth in the spread of COVID-19 from September. 

For certain parts of the Southern Hemisphere, the factors of large populations, high density rates and low testing rates may also lead to a significant increase in transmission rates. In particular for sub-Saharan Africa, there are concerns that low testing rates and low case and death ascertainment may be masking the extent of the epidemic. IHME will shortly be publicly releasing its global COVID-19 forecasts.

Through its work in the Global Burden of Disease estimation, IHME aims to integrate the overall health loss as a result of COVID-19 and the associated policy measures taken across countries. The GBD 2020 will seek to highlight the effects of lockdowns on deferred and ‘missed’ healthcare interventions and on mental health, in particular.

For further information on the IHME COVID-19 model and forecasts, please contact covid19@healthdata.org. 

The latest IHME’s COVID-19 resources and projections can be found on their website: https://covid19.healthdata.org/projections. 

To learn more about the IHME COVID-19 model, you can visit this FAQ page: http://www.healthdata.org/covid/faqs

For more IANPHI COVID-19 resources, visit our COVID-19 resource page and our COVID-19 webinar resource page.