Estimating the impact of COVID-19 on older people in LMICs: the PICHM calculator

Sep 26, 2020 | All posts

By Peter Lloyd-Sherlock, from the University of East Anglia, with the PICHM Expert Consortium (

Developing a simple tool to predict COVID-19 deaths


On 13 March one of the first publications about COVID-19 in low and middle-income countries (LMICs) set out the huge threat to older adults (Lloyd-Sherlock et al, 2020). This was just two days after the World Health Organisation official had declared a global pandemic. Many early predictions about the course of the pandemic, such as its potentially devastating impact on sub-Saharan Africa, have been wide of the mark. Tragically, other predictions have been largely vindicated, including our prediction older people in LMICs would “bear the brunt” of global COVID-19 deaths. This claim was based on the following simple facts:
case fatality is strongly associated with old age,
almost 70% of the global population aged 60+ live in LMICs,

and LMICs have fewer resources and infrastructure than high-income ones to control infection risk or to effectively treat cases.

As March progressed and the pandemic spread, the authorship team and a wider network of academics and NGOs rapidly launched a range of activities to raise the global alarm. We recognised the danger of taking for granted that the “facts” of the pandemic (as set out in our paper) would frame COVID-19 policy priorities. Past experiences with other health issues had demonstrated that global health policy often marginalises older people’s interests (Lloyd-Sherlock et al, 2016).

One initial response was to develop a simple tool to predict COVID-19 mortality in different countries by age group, the
Potential Impact of COVID-19 on Human Mortality (PICHM) calculator.

This simple tool takes case fatality rates (CFRs) for different age groups in countries for which data were available (initially China and Italy) and applies them to the age compositions of different countries based on varying levels of infection across the national population. In Brazil, for example, the PICHM predicts 61,440 COVID-19 deaths among people aged 80 or more, based on Chinese CFR data and a 10% national infection rate (Table 1).

Table 1.
PICHM estimates of potential COVID-19 deaths by age group, Brazil.

Age Group


Case fatality rate (China scenario) Total population in age group Potential deaths 10% infection scenario Potential deaths 25% infection scenario
0-49 0.2 158,280,767 33,426 83,566
50-59 1.3 24,421,202 24,013 60,032
60-69 3.6 16,896,562 60,831 152,078
70-79 8.0 8,801,551 70,089 175,222
80+ 14.8 4,159,027 61,440 153,600

As well as providing an interactive data visualization, enabling users to apply the methodological approach to all countries and world regions for different scenarios, we published a series of national and sub-national PICHM reports. These started with a key message that:

“The data presented in this report are not predictions about what will happen in this country. They set out potential scenarios based on hypothetical levels of COVID19 infection and on hypothetical rates of mortality for those who become infected. Whether this country sees higher or lower rates of infection and mortality largely depends on actions that can be taken by government, other organisations and the general public to contribute towards these twin objectives.”

We viewed these resources as a quick way to draw policy-maker attention to the disproportionate impacts the pandemic was likely to have on older people in all countries, including LMICs where population ageing was not advanced. We had no illusions about the crude nature of our tool, and looked forward to being able to develop more sophisticated epidemiological models once better data were available.

Trying to improve on the PICHM
By May we started to consider developing a more sophisticated version of the PICHM. For example, there are now sufficient data to differentiate CFRs by sex as well as age group. This could have important effects on predictions for LMICs where sex ratios at older ages are quite variable. In India for example, there are more older men than women (Government of India, 2018). To date, however, it has not been possible to build a new version of the PICHM. In part, this has been due to a lack of resources (all work to date has been unfunded and members of the PICHM team were required to resume other duties). Another reason for not enhancing the PICHM was we assumed it would quickly be superseded by better alternatives, and that real data would quickly become more abundant. Yet neither has happened.

At first sight, it would seem we now have much more data on cases and mortality, including for many LMICs. In theory, these should support more realistic CFR scenarios than the initial PICHM ones based on China and Italy. Instead, there is growing awareness that estimating CFRs from LMIC data is very problematic, since coverage of testing (and hence reporting of cases) is extremely incomplete. A new study of Brazil found reported case fatality rates across different regions varied by a factor of more than ten (Baptista and Queiroiz, 2020). Indeed, the original CFRs from China and Italy are likely to be very inflated, due to a failure to detect asymptomatic and less severe cases, as well as misdiagnosis.

As well as problems with data on cases, there are problems with age-disaggregation. The new Global Platform website provides information about available data and will seek to update it over time. A recent review of the quality and availability of data for COVID-19 cases and mortality in LMICs found that many countries do not publish any data by age group and, for those that do, there is evidence of substantial under-reporting (Lloyd-Sherlock et al, in press). It concludes that the state of the data for age groups is so poor that they may do more to obscure than support our understanding of the pandemic.

Revisiting the PICHM: does it still have value?

Given these problems with the “real numbers”, it is worth returning to the PICHM, to compare its estimates to reported data and to ask whether it may still have something to contribute to planning and predicting the course of the pandemic. 

Staying with the example of Brazil, what do the official data show?

As of 18 September Brazil had reported 4.5 million confirmed cases of COVID-19, representing 2.15% of the total population, and a total of 136,000 COVID-19 deaths. But there is growing evidence that these official data substantially under-represent the true toll of the pandemic. For example, data on excess mortality for parts of Brazil indicate that less than a third of COVID-19 deaths are being officially recorded as such (López-Calva, 2020). The PICHM predicts a total of 249,799 COVID-19 deaths based on a Chinese CFR and a 10% national infection rate (which was in reality much higher national infection rate due to missed cases). In the light of excess mortality data, this would appear to be as plausible, if not more, as the official numbers.

Although the Federal Ministry of Health does not publish data by age group, a study of deaths up to 30 June reports that 72% occurred among people aged 60 or more (Baptista and Queiroiz, 2020). This compares to our tool’s prediction that older people would account for 77% of deaths.
There is growing evidence that older people are especially affected by under-reporting of COVID-19 mortality (Nogueirai et al., 2020; Magnani et al., 2020), which is a plausible explanation for their different shares of official mortality (72%) and our own prediction (77%).

PICHM resources on our website include subnational data for 22 countries. These are based on subnational data on population size and age composition. This can potentially offer a more specific planning tool to factor in the uneven geographical spread of the pandemic within different countries. In the case of Brazil, for example, some states, such as Ceará in the northeast, have been affected earlier and to a greater extent than others.

By 12 September Ceará’s Health Secretariat had reported 8,759 COVID-19 deaths (Governo do Estado do Ceará, 2020). This compares to 10,635 applying our 10% infection rate China CFR scenario (Table 2). Of those deaths, 58.8% were aged 70 or more, and 89.1% were aged 50 or more. The PICHM predicts people aged 70 or more would account for 53.8% of deaths, and those aged 50 or more 85.9%.

Table 2.
PICHM estimates of potential COVID-19 deaths by age group, Ceará State, Brazil.

Age Group


Case fatality rate (China scenario) Total population in age group Potential deaths 10% infection scenario Potential deaths 25% infection scenario
0-49 0.2 7,638,238 1,502 3,755
50-59 1.3 930,897 1,209 3,023
60-69 3.6 599,636 2,159 5,397
70-79 8.0 361,840 2,881 7,204
80+ 14.8 195,199 2,884 7,209

Final thoughts

The PICHM has many weaknesses and limitations. In its current form, it is based on CFR scenarios that are entirely notional. These CFRs that are likely to be exaggerated due to large numbers of unreported cases in Italy and China. Also, they lack precision since they do not factor in other important effects such as sex. These weaknesses could be partly, but not entirely, rectified were the resources available.

For Brazil and for the State of Ceará, the overall mortality predictions of the PICHM based on a Chinese CFR scenario and a 10% infection rate are broadly compatible with reported data. However, it is likely that official data substantially under-estimated mortality and the degree of underestimate will only become evident of or when excess mortality data are available.

Rather than predicting the absolute numbers of potential COVID-19 deaths, the main usefulness of the PICHM is to show the likely distribution of deaths across age groups. For Brazil and for the State of Ceará, these predictions have so far been broadly compatible with official reports. Nationally, the PICHM predicts a somewhat higher share of deaths among people aged 60 and over compared to official data. A plausible explanation is that the risk of misclassifying COVID-19 deaths is higher for older people than for other age groups.

Therefore, despite its evident weaknesses and limitations, the PICHM may be less redundant than we had anticipated two or three months ago and will continue to be useful until the quality of data and reporting from LMICs dramatically improves. PICHM predictions on the share of deaths among older people appear to be broadly plausible, and are essential, since no other global source currently provides data by age group.
This starkly contrasts with efforts to provide global sex-disaggregated data on COVID-19As far as we are aware, no other organisation or academic network is seeking to generate comparable resources that include age data. This lack of interest both reflects and feeds back into the neglect of older people in global health policy. The Global Platform will do what it can, but this is almost entirely unfunded work.


E.Baptista and B.Queiroiz (2020) Regional Covid-19 Mortality in Brazil by Age. file://ueahome/eresssf1/d025/data/Downloads/Baptista_Queiroz_Covid_Age.pdf 

Government of India, Central Statistics Office (2018) Women and Men in India (A statistical compilation of Gender related Indicators in India) 2018.

Governo do Estado do Ceará, 2020, Secretaria da Saúde (2020) Boletim Epidemiológico. Doença pelo novo Coronavírus (COVID-19) Nº43, Ceará – 17/09/2020.

P.Lloyd-Sherlock, S.Ebrahim, M.McKee and M.Prince (2016) Institutional ageism in global health policy. BMJ 2016;354:i4514.

Lloyd-Sherlock, P.G., Ebrahim, S., Geffen, L. & McKee, M. (2020). Bearing the brunt of covid-19: older people in low and middle income countries. BMJ 368, m1052.

P.Lloyd-Sherlock, L.Sempe, M.McKee and A. Guntupali (in press) Problems of data availability and quality for Covid-19 and older people in low and middle-income countries. The Gerontologist.

L.López-Calva (2020) A greater tragedy than we know: Excess mortality rates suggest that COVID-19 death toll is vastly underestimated in LAC. UNDP–excess-mortality-rates-suggest-t.html.

Magnani, C., Azzolina, D., Ferrante, D., Gallo, E. & Gregori, D. (2020). How Large Was the Mortality Increase Directly and Indirectly Caused by the COVID-19 Epidemic? An Analysis on All-Causes Mortality Data in Italy. International Journal of Environmental Research and Public Health. 17(10):3452.

Nogueirai, P.J., Furtado, C., Vaz Carneiro, A., De Araújo Nobre, M. & Nicola, P.J. (2020). Excess Mortality Estimation During the COVID-19, Pandemic: Preliminary Data from Portugal. Acta Medica Portuguesa. 33(6):376-383.