HPH metrics allow comparative risk assessments for active healthcare workers and general populations during crises. As shown, these metrics can vary enormously. They can reach extreme values in both relative and absolute terms during an outbreak like Western African Ebola where largely unprotected healthcare workers in understaffed facilities were exposed far more than the average citizen. They can be very high in absolute terms for healthcare workers who persevere offering their services in war-torn areas. Conversely, for COVID-19, HPH(r) in high-income countries was very low and HPH(a) was negative, suggesting that physicians or even all healthcare workers in such countries had much lower risk of COVID-19 deaths and lower excess deaths overall than the general population that they served. This privilege may not exist nevertheless in less developed countries.

Prior experience from past crises can be extrapolated to new, similar emerging crises. New data collected also in real time can help recalibrate HPH metrics, monitor their evolution and assess the impact of interventions. E.g. an Ebola outbreak may have by default a higher death rate among healthcare workers than the average person in the community. However, this grim situation can be ameliorated with higher availability of protective measures for healthcare workers, vaccination (given the current availability of apparently effective vaccine options [26]), and processes that minimize unnecessary exposures.

During wars, a main problem is the attacks on healthcare professionals that constitute an organized program of terror [7,15]. In Syria, most healthcare worker deaths were even caused by the government and its allies [15]. It is important to monitor atrocities, identify perpetrators, and intervene to avert further escalation. Accurate data collection in war conditions becomes very difficult and it may also be subverted, suppressed, or fabricated. However, some patterns of excessive risk for targeted healthcare workers may be common across very different countries and types of war conflicts. A literature of such observations has been amassed from conflicts in other areas such as Sudan, Pakistan, Burma, Iraq, Afghanistan, Nepal, Yemen, and Ukraine [27–35].

Some calculated HPH metrics were extreme, highlighting the inordinate risks that healthcare workers are occasionally exposed to. The highest absolute death risk (92 per 1000, i.e. a literal decimation) and HPH(a) (91 per 1000) was seen in the Sierra Leone Ebola outbreak. Sierra Leone had the lowest availability of healthcare workers (<0.04 per 1000 population) [14]. If data on the number of healthcare workers are accurate, the volume of Ebola cases handled per worker was also higher than elsewhere, leading to devastating outcomes for this highly vulnerable workforce. More generally, professionals working in understaffed healthcare systems may face disproportional risk, besides the burnout that is highly prevalent even in more developed countries [36]. In a vicious circle, understaffed systems have great difficulty to retain healthcare workers and witness large attrition rates [37]. Qualified physicians are sparse in these settings and newly trained physicians massively immigrate [38,39].

Conversely, for some types of crises, healthcare workers may be at lower risk than the people they serve (8,18,21). Their absolute risk may be small or even non-existent. COVID-19 was a classic example in this regard. When the pandemic started there was widespread fear of decimated healthcare workers, overrun hospitals and system collapse. This outlook contributed major psychological distress to healthcare workers [40]. Fear of COVID-19 was a major contributing theme for healthcare workers’ intention to quit [41]. The same outlook probably also facilitated the adoption of aggressive population-wide policy measures [42,43]. Nevertheless, in most developed countries, physicians and even other healthcare workers may have suffered minimal excess deaths, if any. Kiang et al. found far lower excess deaths in physicians versus the general population even after adjusting for age and gender [8]. Therefore this “privileged” low-risk status was conferred primarily by factors other than just demographics, e.g. a healthy worker effect [44], better health/lifestyle choices [45] and (by definition) higher socioeconomic status. Higher socioeconomic status strongly correlates with higher life expectancy in general [46] and low socioeconomic position was associated with 5-fold higher COVID-19 deaths [47]. Healthcare workers often had also better access to and use of effective treatments and vaccines. Finally, many, perhaps most, healthcare workers were infected in the community rather than in occupational settings; seroprevalence rates between healthcare workers and the general population usually did not markedly diverge [48].

Physicians had minimal excess deaths even in USA that had the highest excess deaths proportion among non-elderly individuals than any other country with reliable death registration [49]. Ontario (that had much lower general population excess deaths than USA), physicians also had ∼8 times lower COVID-19 deaths than the general population. Across Canada, until Jan 14, 2022, healthcare workers accounted for only 46 of 30,756 reported COVID-19 deaths [50]. While healthcare worker definitions are not sufficiently consistent in these data to allow accurate calculations, HPH(r) in Canada may have been even lower than Ontario. For countries (such as in the Nordic Region) and provinces that had even lower excess deaths of even a death deficit in the pandemic versus the pre-pandemic period [49], healthcare workers may have fared even better. In Finland, data on all healthcare workers use an expanded definition of health and social work and 8% of the total population are included in this occupational bracket; among them the risk of COVID-19 death in 2020-21 was 10 times lower than the general population [21]. In British Columbia, a province of 5 million inhabitants, during 2020-2021 not a single healthcare worker died of COVID-19 among ∼11 thousand infected [39]. Conversely, this privileged status may not exist for some low-pay, disadvantaged healthcare workers in some developed countries and for the majority of healthcare workers in less developed countries. There are large uncertainties in death ascertainment in less developed countries [51]. However, their healthcare workers probably suffered the same fate as the general population, or even worse.

Making inferences on healthcare workers from the general population experience becomes misleading if HPH is ignored. A WHO report covering COVID-19 deaths until May 2021 [52] assumed that healthcare workers had the same COVID-19 death rate as the general population in each country. It thus estimated they must have suffered 115,500 COVID-19 deaths or even 179,500 after adjusting for under-reporting – versus 6,633 officially reported deaths. However, most of these speculated deaths were imputed in developed countries; as shown here, modeling on the general population may have led to major (5-10-fold) overestimation. The same report [52] estimated 39,875 COVID-19 deaths among healthcare workers with a different method, assuming 0.4% infection fatality rate, which is probably also substantially exaggerated for an active, non-elderly population [53].

Death risk estimates and perceptions can be markedly inflated if retired healthcare workers are included in the calculations. This was a major misleading factor in early pandemic media stories and registries of physician deaths. Deaths pertaining mostly to elderly retirees were extrapolated to the active workforce. The most dramatic early data on physician deaths came from Italy, where the national registry included retirees. Few of these deaths reflected active physicians [54]. As a group, retirees largely followed the fates of high-risk general population subsets. Actually, often nursing homes were even disproportionately hit in the early pandemic [55] and high-risk people were unfortunately less protected than low-risk individuals [56]. Misclassifications of retiree deaths as active occupational risk gave rise to stunning misunderstandings, ranging from passionate editorials by esteemed opinion leaders [57] to extremely exaggerated estimates of fatality rates [58]. E.g., a systematic review of healthcare worker deaths in 2020 [58] found 37.2% fatality rate among those >=70 years old – massively inflated due to missed infections and inclusion of retirees. Infection fatality rates in active practitioners in this age range were probably 10-50 times lower [59]. In the examples analyzed in the current paper, contamination with retiree data is probably low. However, if any such misclassification ensued, then HPH is even lower than the low values obtained here.

Several limitations regarding HPH metrics should be discussed. Calculating HPH metrics is limited by data inaccuracy in outcomes and populations at-risk. For several crises, even deaths are extensively miscounted [6]. Even to obtain a simple trend, the reliability of the data needs to be verified. Healthcare professionals may have health-related data of better quality and lower missingness than the average person. This possibility should be examined on a case-by-case basis in different circumstances. The ability to have a biological diagnosis of an infection or even a clinical one may be different in the general population than in healthcare workers. This may be particularly true in developing countries, but it may also be seen in developed countries, especially in settings and populations with limited resources and suboptimal access to healthcare. In war-torn countries, population-at-risk numbers may be unknown and volatile, affected by refugee status, displacement and incarceration. In each case, one should scrutinize carefully the quality and reliability of previously available data, identify gaps and major sources of bias in data collection and reporting and try to fill some of these gaps or use more appropriate adjustments and correction factors in the future. Given the difficulty to correct fragmented data, HPH metrics should be interpreted with sufficient uncertainty surrounding them.

Finally, different healthcare workers may have different HPH metrics. Here, for illustrative purposes only age strata and direct patient care were explored, but sometimes additional stratifications may be of interest. More major differences may arise if the population served is stratified. E.g., in developed countries children and adolescents typically had a substantial death deficit during the COVID-19 pandemic versus pre-pandemic years [17,49]. It is unknown whether pediatricians suffered any excess deaths of had an equally large or even larger death deficit. Conversely, for professionals serving nursing home facilities during the pandemic, their gap versus the nursing home residents was much wider. Also for outbreak- and war-stricken countries, HPH may vary in different parts of the country, with highest values at crisis epicenters.

Allowing for these caveats, HPH may be a useful concept to measure, monitor and compare across settings and both within and between different crises. It also offers a reminder of the sacrifices demanded of healthcare workers worldwide, often under very adverse circumstances.