4.2.1. Direct Support for the Exposure Hypothesis
Even though a summer wave of influenza is not mentioned in previous reports on the surveys conducted in the United States [10
], the survey data from Bergen, Norway, together with the recognition of higher morbidity in smaller towns compared to more urban areas in Maryland, support the idea that a pandemic wave before the autumn of 1918 may actually have hit the city of Baltimore as well [20
]. This assumption is also supported by other studies showing that a first wave of influenza was present in the Northeastern United States in the spring/summer of 1918, for example, in New York City [30
], and also in Camp Funston in Kansas [11
]. Soldiers in Camp Funston exposed in March seemed less likely to be attacked in the subsequent April and May outbreaks. Black veterans and white veterans in Camp Pike, Arkansas, exposed to the summer wave gained some protection and had much lower morbidity in the fall wave than the newly arrived recruits [12
]. Similar patterns of protection against the 1918 fall wave after being exposed to herald waves are described elsewhere [31
]. The early cases in Kansas led some researchers to believe that the 1918 influenza pandemic first started in the USA and then spread to the rest of the world [2
]. These observations in 1918 also are supported by more recent examples. Among military recruits in Singapore in 2009–2013, prior adenovirus or influenza virus infection conferred cross-protection against subsequent febrile illness episodes relative to prior infection due to other circulating viruses [34
In order to substantiate the Crosby hypothesis that more black people were exposed to the mild summer wave than white people, we need empirical evidence on, for example: (1) Did more black people than white people live in cities, and especially, did more black people than white people live in and were they present in cities exposed to the spring/summer waves? (2) Did more black people than white people live in poor and cramped conditions within their homes and neighborhoods? (3) Did black people have poorer access to sanitation, water, and poorer hygiene and handwashing opportunities than white people? In other words, we need empirical information on these aspects where race may be a proxy for socioeconomic status.
In 1910, over 90 percent of the urban population in the United States consisted of white people [24
]. In 1920, the majority of the Americans still lived in municipalities with no less than 2500 people [27
]. Out of the white and black populations, 27% and 49% lived in urban areas, respectively. Most of the black population lived in the southern states in 1918. We have no information that a summer wave was especially rampant in the South. However, the Great Migration from rural South to northern cities, starting in 1916, meant that a larger share of black people lived in urban areas during the pandemic. Nevertheless, on average, fewer black people than white people lived in high population density urban areas where we can assume that more people were exposed to the spring/summer waves. Despite full segregation of black troops and white troops in US training camps, crowding in black organizations may actually have been less than among the white organizations because the black organizations had fewer individuals. In other words, one reason why influenza morbidity was lower among black troops may be due to the strong segregation.
The black people in the cities in the South often lived in dilapidated, crowded, and segregated housing areas. The Great Migration also left a large portion of the black people in cramped conditions in cities in northern states. One example is the city of Baltimore, where black migrant inner-city neighborhoods were struggling with deterioration and poverty and had less access to sanitation and urban social programs. The segregation created crowded urban ghettoes where tuberculosis became a prominent cause of death [28
]. Individuals with such scarred lungs were likely more susceptible to both influenza and pneumonia and other infectious diseases [35
]. This would also explain why a higher share of black influenza cases developed pneumonia and also the higher lethality from those diseases in 1918. When we take into account that a lower share of black people than white people lived in cities and that individuals of lower socioeconomic status may have poorer immune function due to more psychological stress, increasing their susceptibility of developing influenza given exposure [38
], the relatively consistent result that black people had lower morbidity than white people is somewhat counterintuitive.
4.2.2. Indirect Support to the Exposure Hypothesis
] hypothesis is that exposure to the disease agent early on primes the immune system to respond more effectively to the next insult. Several studies support such suppositions: Hörzer [39
] documented that the first victims
in the city of St. Paul, Minnesota, were reported in impoverished nonwhite areas, while later victims were of higher social status. This is also consistent with the finding for Oslo [40
], where children from the affluent west side had higher morbidity during the major wave than during the early wave because they spent their summer holidays in the countryside where they were not exposed to the early wave and could therefore not acquire immunity to fight the second wave. The history for children on the less affluent east side of Oslo was the opposite. In Bergen, Norway, the first wave during the summer of 1918 hit the poor the hardest, while the rich with less exposure to the first milder wave was harder hit during the fall of 1918. At all socioeconomic status levels, males had the highest morbidity in the summer, while females had the highest morbidity in the fall [41
]. Finally, in both Bergen and the two surveyed regions of Maryland, more females than males just above the age of 20 fell ill [20
]. An explanation for this marked gender difference in both Bergen and Maryland could be that young male workers were more likely to be exposed to influenza during the first pandemic wave than females, who were mostly home-based, leading to males being better protected during following waves [20
Excess mortality among US insurance holders from October to December 1918 was highest for young adults, and by race and gender, mortality was highest for white males, followed by white females, black females, and black males [16
]. Placing these results for insurance holders in the context of the above research for Bergen and Maryland, it is likely that (1) a previously unrecognized pandemic summer wave may have hit the two regions of Maryland in 1918 but also other areas of the USA than New York and Kansas, and (2) that males of both races and black people were more exposed in the spring/summer of 1918, while females and white people with less exposure and acquired immunity during the spring/summer had higher morbidity and mortality in the fall of 1918. This would also be consistent with the assumption of a higher pre-pandemic exposure among black people, which would be one explanation for why white soldiers had higher influenza morbidity than black soldiers [7
]. That positive excess mortality was mainly visible for those aged 20–40 of both races, that it peaked in those around the age of 30, and that excess mortality was negative for those 70+, is generally consistent with prior research [42
]. Why mortality peaked at age 30 but declined into old age is still not clear. Explanations have included cytokine storms (multi organ failure), cardiac stroke and exposure early in life to other strains of influenza [43
]. However, it is unclear why the highgly pathogenic virus should harm white young adults’ lungs and hearts harder than their black counterparts. It has been suggested that cohorts >30 years were protected because they were exposed to H1-like viruses (similar to the H1N1-virus in 1918) prior to the 1889 Russian influenza pandemic [42
]. This prior exposure and subsequent protection may have been stronger for black people aged 30–40 than for their white counterparts, but as we have seen, both races had a decline in their positive excess mortality in their 50s and 60s, and negative excess mortality rates were observed for both races after 70+.
4.2.3. Alternative Hypotheses
In our review, we did not find hypotheses other than the unequal exposure hypothesis suggested by Crosby [2
] to explain the crossover in the role of race in 1918 pandemic outcomes. However, could this finding also be explained by concurrent events or secular changes in the quality of life and/or health care available to the black population? For example, is there any reason to believe that the explanation in the racial crossover in morbidity/mortality was due to any changes, in 1918–1919 or earlier, in how race was associated with disparities in psychological stress, nutritional status, and access to care providers/insurance, and by extension, differential susceptibility and disease severity? In other words, did WWI or other concurrent events lead to a relatively better position for black versus white people? One concurrent event was the Great Migration. However, as we have discussed above, the transition from rural to urban life made those of low socioeconomic status such as the majority of the black people more and not less susceptible to developing disease and to dying from it [35
]. Further, is there any reason to believe that the quality of health care available to and treatment options used by the black and white patients changed in 1918, such that a racial crossover in mortality occurred due to differential consequences after the disease had developed? Because of generally poor living conditions in the South and troublesome conditions in northern cities due to the Great Migration, black nurses and black physicians were pushing for better sanitation laws and housing reforms to clean up the neighborhoods and the tenements [46
]. In addition to this, “health weeks” were organized where black people could learn more about personal hygiene, and how they could avoid getting ill. However, a lack of resources and facilities prevented the efforts black nurses and other volunteers put into this program [46
]. It was almost universal practice that the hospitals would either deny black people admission or they would be placed in a segregated space, like the basement or the attic. This forced the black people to develop their own health organizations, and by 1920, they managed to build several hospitals. Either way, the health care for black people was much poorer than for white people.
Another explanation of our findings might be selection. It can be assumed that the approximately 2 million young adult soldiers who were in France from September to December 1918 were the healthiest, leaving the frail behind [10
]. By the spring of 1918, hundreds of thousands of troops were being deployed overseas monthly, and in November 1918, the draft and mobilization were at full capacity [47
]. Among the medically examined, a larger share of black than white draftees were placed in Class I. Further, among the Class I members, a higher proportion of black individuals ended up in service. This unbalance was probably a consequence of discrimination, but on average, it might be reasonable to believe that this type of selection was stronger for the black than for the white population, leaving a frailer black civilian population behind. However, if this were true, it should have resulted in higher black PI mortality in the fall of 1918, not lower, as was observed. Moreover, the mortality of civilian males (and females) in urban and rural areas in the fall of 1918 in one of the canvassed areas for the surveys, Maryland, follows the same pattern and level of morbidity and mortality by gender as in Bergen in neutral Norway [20
]. If a higher proportion of black than white draftees were sent to the front in Europe, this could potentially explain why morbidity and mortality was higher among the white civilian males in the fall of 1918. However, this was not the case, as half of both races were sent to France [5
]. Finally, the selection hypothesis does not explain the higher white than black female pandemic outcomes in the fourth quarter of 1918, as nearly all soldiers were male.
One intriguing question is why mortality by race turned back to normal patterns exactly in December 1918, after only 2 months of deviation? Was this due to a deeper harvesting of white people than black people due to the higher white mortality in the fall of 1918, leaving more healthy white people at risk of dying during the winter and spring of 1919? The harvesting hypothesis gains some support from a study by Noymer and Garenne [36
] showing that the high 1918 influenza mortality in the US took out those who were supposed to die from tuberculosis, lowering significantly the mortality from that disease in the 1920s. The return of soldiers in the months following the armistice in November 1918 may also explain the return to “normal” racial influenza outcome differences. However, troops from overseas did not return in large numbers until late spring and summer 1919, with the last division arriving in September 1919 [48
]. Moreover, because they were only 3% of American combat forces, black soldiers suffered substantially fewer battlefield deaths and wounds than white soldiers [5
]. Overall, black soldiers from the 92nd and 93rd combat divisions accounted for 773 of the 52,947 battlefield deaths sustained by the AEF in France during the war, less than 2% of all battlefield fatalities. Of American soldiers wounded, 4408 were black and 198,220 were white. White soldiers, therefore, made up nearly 98% of those wounded on the battlefield. The black soldiers probably suffered psychologically from the war as much as the white soldiers. However, the return of more physically wounded white than black soldiers should, everything else being equal, give the black group a mortality advantage; however, in 1919, mortality from PI and all causes were back to the pre-pandemic “normal”, with higher mortality for the black population.
A contemporary theory as to why the black population seemed to be less susceptible to upper respiratory tract diseases such as influenza and polio was that the lining of their noses (nasopharynx) was more resistant to microorganisms [7
]. A more scientific and credible hypothesis, consistent with Crosby’s hypothesis, is that the black population on a general basis were exposed to flu viruses due to their social conditions and therefore had lower susceptibility during both the autumn and the spring/summer waves [46
]. However, prior literature has missed that this also was the case before 1918, which would explain why white soldiers had higher influenza morbidity than black soldiers in 1910–17 [7
]. When black people first got sick, they died more frequently, especially from bacterial complications such as pneumonia, likely because of their poor social conditions, seeking or receiving treatment at hospitals later, and a higher prevalence of concurrent illnesses (malaria and hookworm) compared to the white population [27
A last alternative hypothesis is underreporting of influenza cases and deaths amongst the black population. It is impossible to determine how much of the lower influenza morbidity among the black people in survey data is due to the canvassing of areas with few black people and more complete reporting of white influenza cases [10
]. However, the influenza cases in the surveys did not only include the severe cases brought to a doctor or a hospital. They rather included all cases the black homemaker could remember and report to the (white) enumerators. Thus, there is no reason to believe that morbidity (or mortality) data for black people are biased because cases were not brought to the attention of physicians due to discrimination by white doctors, access to few black doctors, or poor access to health care, health information, health insurance, and poverty. It is nevertheless worth noting that in Charles County, Maryland, where a little more than half of the population canvassed was black, morbidity was higher for both black males and black females compared to that for white males and white females.
One may assume that reporting of influenza cases is complete for both races in the military. However, in the highly segregated training camps in the USA, things were not as equal as believed by some in 1918 [7
]. Black soldiers were usually placed in noncombatant labor battalions and received poorer housing, clothing, and food [5
]. The army only allowed black officers to lead black troops, while white officers commanded both white and black units during the war. Out of 380,000 black soldiers, less than 1% or only 1353 served as officers, and only 1.5% of all military doctors were black [6
]. Black people had poorer access to medical care and inferior care. Some black soldiers may have decided not to report themselves sick with influenza to a (possibly judgmental) white doctor, and independent black observers noted that white medical officers often delayed admitting black people into the hospital for treatment [5
]. This may in sum explain lower hospital admissions for influenza but a higher rate of influenza cases developing into pneumonia in black soldiers [5
]. The data in hospitals in France tell a contrasting story: Influenza morbidity in the calendar year of 1918 was actually higher
for black troops than white troops [8
]. In France, black soldiers were more likely to end up in labor battalions rather than in the trenches to fight. Although this was a consequence of ideas related to white supremacy [5
], perhaps the stigma of black soldiers in the US training camps was less of an issue in France where interracial mingling was more accepted? Official surveys of soldier opinions of returning troops in 1919 support this idea. They show that black but not white soldiers expressed a preference for Europe over the United States in several ways. Black respondents, for example, did not complain about high prices in French stores. Instead, they focused on the fact that they were welcomed by every shopkeeper they encountered [5
]. Finally, a number of diseases had higher pre-pandemic hospitalization rates for black than for white people, including lobar pneumonia and tuberculosis. Although these diseases were more severe than uncomplicated influenza, it shows that discrimination cannot explain racial differences in all causes of hospitalization.
Several major black states in the South were not covered in the National Births and Deaths Registration Area in 1918 [21
]. Under-registration of black people is therefore an issue in the studies using official data for the US to study mortality (as in [22
]). However, even in a major southern city such as New Orleans, white people had higher mortality than black people [23
]. Finally, although insurance data may leave out more black people than white people, underreporting of deaths among black insurance holders is not likely to explain the crossover in mortality by race at end of 1918; why should underreporting only occur in the fall of 1918, and not in the pre-pandemic and post-pandemic phases, where PI mortality was higher for black people?
] suggested that black people due to poor and congested living conditions had higher prior exposure to the spring/summer wave in 1918. The higher morbidity at that time would therefore give the black people immunity, which would give lower morbidity and mortality in the fall. However, most of the black people lived in rural areas in the South, where there is no evidence for a herald wave in the spring of 1918. This would lead to little exposure to influenza before 1918 and to the herald waves in 1918 and thus less immunity and therefore higher morbidity and mortality in the fall of 1918. However, the observed data show the opposite. Another possibility is therefore that the lower black morbidity and mortality in the fall of 1918 occurred because Southern rurality led to a less widespread epidemic, protecting the black people who had low immunity. The black rural isolation may have led to less exposure, but when black people moved to northern cities and the men went off to war, they did get exposed. We can also make similar arguments taking the perspective of the white population. If a larger share was exposed in urban areas before 1918 and in areas that had herald waves, it should lead to higher immunity in white people. Why, then, was white morbidity higher than black morbidity in cities and in the military? Discrimination in the army is not likely to explain all racial differences in morbidity, and discrimination is less likely to be a factor in the lower reported black than white morbidity in surveys of the civilian populations.
Prior studies did not always control for possible confounders. Future studies on morbidity, mortality, and case fatality should therefore search in archives for the primary individual-level data for the military and the household-level data from the surveys undertaken in 1918. The primary level data from the Metropolitan Life Insurance Company may also exist in archives. Another possibility is to study linked census and death certificate data in order to study racial disparities in the time from illness onset to death or mortality during different pandemic waves, with controls for variables such as age, gender, marital status, occupational-based social class, housing standards or factors associated with disparities in exposure, susceptibility, and access to care. One possibility is to use already digitized data from a city such as Saint Paul [39
], or Saint Louis, which had a substantial population of black people in 1918 (8%) [49
]. Such studies would thus contribute to a better understanding as to why there was a crossover in the racial differences in pandemic outcomes that emerged between the third and fourth quarters of 1918, with a return to pre-pandemic greater excesses in black pandemic outcomes in December 1918.