Background
After decades of influenza circulation with relatively low virulence following the 1918 influenza pandemic [
1‐
5], the 1957–58 H2N2 pandemic spread to more than 20 countries in less than 4 months and caused almost 60,000 excess deaths in the United States from September 1957- March 1958 [
6‐
8]. The 1957 influenza pandemic has been associated with an average respiratory excess death rate of 1.9 per 10,000 during 1957–1959 [
9]. Moreover, the impact of this pandemic was moderate relative to the 1918 pandemic, but about 10 times greater than the 2009 A/H1N1 influenza pandemic [
9].
With the possible exception of persons older than 67 years, individuals had no prior exposure to the A/H2N2 virus, and therefore had no previous immunity to this virus, resulting in about a quarter of the United States population becoming infected [
6,
7]. The widespread effects of recent influenza pandemics have emphasized the importance of understanding historical pandemics in order to prepare for and mitigate future outbreaks. A better understanding of the age, seasonal, and transmissibility patterns of previous influenza pandemics may help public health officials prepare for challenges that we may face during future influenza pandemics. Although past studies have quantified the mortality burden of the 1957–58 influenza pandemic in the United States [
8,
10‐
12], little is known about the transmission and mortality characteristics of this pandemic at small spatial scales. Here we aimed to quantify age-specific mortality rates and transmissibility patterns of the 1957–1958 pandemic in Maricopa County using publicly available data comprising a long series of detailed mortality records during 1954–1961.
Originating from the Kweichow province of China in late February 1957, the virus induced fever, sore throat, headache, malaise, and myalgia [
13]. The pandemic reached the United States in late May 1957 and the first Asian influenza outbreak in the United States occurred in early June in Newport, Rhode Island [
10]. However, the first serologically confirmed case of the H2N2 virus in Arizona was not reported until September 23, 1957 [
14].
It has previously been reported that the Mountain region of the United States, which included Arizona, did not experience a second wave of high mortality in early 1958 [
10]. However, there is evidence for variation in the excess mortality rates and temporal patterns for different cities and states within the same geographic region [
10]. By using publicly available archival death certificates from the Arizona Department of Health Services, we set out to quantify the age-specific seasonal patterns, mortality rates, and transmissibility patterns of the pandemic in Maricopa County and compare our local estimates with those previously derived at the national level.
Discussion
By analyzing primary data from archival death certificates from 1954 to 1961 and archival newspaper articles, we found that Maricopa County exhibited low mortality impact associated with the 1957 influenza pandemic, compared with other regions of the United States. In the United States, excess mortality values for all-ages from pneumonia and influenza deaths as well as from all-cause deaths were greatest from October-December 1957, compared to January-March 1958 and January-March 1960 [
11]. From September 1957 to March 1958, the US had a 4.5 (per 10,000) absolute all-cause excess mortality value for all ages and a 1.17 (per 10,000) absolute influenza-pneumonia excess mortality value for all ages [
10]. However, in Maricopa County, the absolute respiratory excess mortality for all-ages was greatest (1.8 per 10,000) during the 1959–1960 wave and excess mortality peaked in the first week of May 1960. While some age groups did have extremely mild excess-mortality during October 1, 1957-March 31, 1958 or during October 1, 1958-June 30, 1959, there was little overall evidence for herald waves in 1957 or 1958, based on respiratory deaths. This is consistent with the less pronounced mortality from January-March 1958 and the higher excess mortality from 1959 to 1960 in the Mountain region compared to other regions of the U.S. [
11]. Absolute all-cause excess mortality (per 10,000) in Maricopa County was greatest from October 1, 1958-June 30, 1959, for all-ages (0.64) and for the elderly (≥65) (3.50). However, absolute excess-mortality from all-causes was minimal throughout all predicted waves. Interestingly, there was also some excess-mortality from respiratory or all-causes during the week of June 2, 1957 or during June 29, 1958-July 20, 1958 for some age groups, but not overall. We cannot rule out the potential contribution of high temperature in the region to excess mortality during these summer months. It has also been suggested that influenza epidemics may coincide with rainy reasons in the tropics due to indoor crowding [
16]. However, influenza transmission in Maricopa County seems to be more efficient in cold, dry conditions with low air pressure. Soebiyanto et al. also demonstrated that influenza cases in Maricopa County do not seem to be associated with rainfall [
41].
The virus seemed to have been introduced in Maricopa County relatively late, between August 23, 1957 and September 23rd, 1957 [
14,
35]. The U.S. had seen its first case by June and the first epidemic in the U.S. occurred in early August in the southeast of the U.S. [
6,
13]. However, the first case of the H2N2 virus in Arizona was not confirmed until September 23, 1957 and transmission reached epidemic levels in Arizona on September 25, 1957 [
14,
38]. One of the major factors for the emerging community epidemics in the fall of 1957 was the opening of schools around September [
6]. Mortality began to rise during the fourth week of October 1957, about 3 weeks after the rest of the U.S. [
11]. Mortality in Maricopa County peaked in the third week of November 1957, about 2 weeks after other regions of the U.S. [
11]. According to newspaper reports, influenza incidence in Maricopa County was still rising rapidly on September 26, 1957 and seemed to continue to rise at least through November 1, 1957 [
39,
42]. A rise in mortality can follow a rise in acute respiratory illness incidence by as much as 3–4 weeks [
11]. This lag between morbidity and mortality may be because the 1957 influenza virus generally affected high school age adolescents first, followed by elementary school students, and the adult population last. [
13]. Incidence in the United States was especially high in those between 5 and 19 and lowest in those 65 and older. However, mortality was highest in those 65 and older [
13]. Our results confirm that age-patterns in Maricopa County were similar to those from the rest of the United States, with the excess-mortality concentrated in the elderly (≥65). While incidence may have been high in children and adolescents, this age group experienced minimal excess-mortality, avoiding the effects of an over-reactive immune system and the over-production of cytokines theorized for the high mortality rates of young adults reported during the 1918 pandemic [
43]. Instead, younger individuals may have transmitted the virus to the elderly after a couple of weeks of high incidence in school-aged populations. While individuals >67 years of age may have had antibodies for the 1957 H2N2 virus due to a possibly related pandemic in 1889–90, individuals ≥65 years of age also had a high-risk of death from influenza in 1957 due to cardiovascular disease and bronco-pulmonary co-morbidities [
8,
44].
From October 1957-March 1958, excess deaths from all-causes in the United States were greatest in the elderly (≥65) and demonstrated a U-shaped age-pattern (high mortality in infants and elderly with low mortality in young adults). While all-cause excess deaths were concentrated in those 65 and older, there were no all-cause excess deaths in infants (≤1 year) and low all-cause excess deaths in children (1–14 years), avoiding the U-shaped age pattern in the United States from January-March 1960 [
11]. With the exception of excess-mortality rates from respiratory causes during October 1, 1958-June 30, 1959, excess-death rates in Maricopa County were greatest in the elderly and had minimal values in younger age groups. While other periods demonstrated no excess-mortality in young children (≤5), the 1959–1960 wave had a very slight elevation in excess-mortality for the age group. However, the excess-mortality in those less than 5 years did not approach that of the elderly, as is common in a traditional U-shaped age-pattern. Therefore, the age-pattern did not truly resemble a U-shaped curve.
For both all-causes and respiratory illnesses, the standardized mortality ratios were minimal for most age groups throughout all three waves. Most likely due to crowding in schools, the standardized mortality ratio peaked (4.06) in young children and adolescents (5–14 years) from October 1, 1957-March 31, 1958, based on mortality rates of respiratory deaths. This is consistent with what was reported by Dauer: the epidemic in September 1957 began in high schools and colleges and moved into elementary schools and pre-school children. However, in the United States from October-December 1957, the standardized mortality ratio was highest (~2.25) in those 30–39 years old, perhaps due to proximity in the workplace [
10]. Additionally, it is important to note that although the standardized mortality ratio was elevated for children and adolescents during the 1957 wave, the excess-mortality rate was minimal for the same age group and time period. However, these results are not contradictory. While the baseline mortality was a fraction of the observed mortality during the 1957 wave, the difference between the values was negligible for young children and adolescents. This may be due to a relatively low baseline mortality in those aged 5–14, compared to other age groups. In contrast, the higher excess mortality rate and the low SMR seen in those ≥65 during the 1959 wave may be due to a higher baseline mortality for the elderly, when compared to other age-categories.
Although the respiratory excess-mortality rates during the 1957 and 1959 waves both disproportionally affected the elderly, there was a shift to greater excess mortality in the latter wave. A previous study showed that excess-mortality during the 1959 wave was concentrated in the elderly and approached the excess-mortality rate of the 1957 wave [
11]. While excess-mortality may have increased between waves in Maricopa County, the possibility that the 1959 wave could have been due to a different influenza strain cannot be ruled out. This study however did not address evidence demonstrating that the 1959–1960 dominant strain in Maricopa County was the 1957 pandemic strain.
To estimate the baseline mortality from a non-epidemic period, this study utilized mortality data from January 3, 1954 to June 30, 1957. Baseline periods vary in length between studies and longer periods may be used for country-wide analyses. Using a longer period to estimate baseline mortality may have increased the accuracy of the Serfling model. However, a three-year period has been previously used to estimate the epidemic threshold of smaller populations, such as cities or counties, in which there is reduced variation [
2].
Our calculation of excess mortality is not exempted of limitations. In particular, due to lack of laboratory confirmation, our excess mortality approach would not have been unable to distinguish elevation in mortality rates associated with other causes and coinciding with the pandemic period. Our approach for calculating excess mortality was relatively simple, due to lack of contemporaneous virological surveillance. Moreover, by grouping deaths into all-cause mortality and respiratory mortality categories, the study prioritized sensitivity over specificity. Because influenza deaths are often attributed incorrectly, we believe that categorizing deaths into all-causes and respiratory causes provides conservative estimates of excess mortality.
Reproduction numbers were relatively similar between waves, assuming mean generation intervals of 3 or 4 days that follow exponential or fixed distributions. In the United Kingdom, R
0 was estimated to be 1.7–1.8 for an infectious period of 2 days and 1.5–1.6 if the infectious period was 1.5 days [
45]. Thus, the Maricopa County mean reproduction number of 1.08–1.11, using 3 or 4 day generation intervals and exponential or fixed distributions, was substantially lower than that observed in the UK. However, the reproduction number is preferably calculated from case incidence curves rather than time series of deaths. Consequently, it is likely that our R
0 estimates could be slightly underestimated. Nevertheless, the lower reproduction number observed in Maricopa County in the 1957–1958 pandemic wave was most likely not due to the public health interventions put in place in the fall of 1957. While closing schools can reduce the effects of an epidemic by 22 %, when the R
0 is low (≤1.8) [
45], the state health department seemed to implement few non-pharmacological mitigation strategies, and instead urged residents to receive an influenza vaccine and communicated the symptoms of Asian flu [
46,
47]. While Valley of the Sun School closed for about 5 days on September 30th, when absences reached 39 % of enrollment, most schools remained opened [
48]. Absences reached 25 % of enrollment in Glendale schools, 20 % at St. Francis Xavier School, and 25 % at Tempe High School [
48]. However, absences in Phoenix elementary schools in the first week of October were only 13 % of enrollment [
49].
While some have theorized that the 1918 and 1957 differed in virulence, the differences in the rates of severe disease are likely due to medical and public health advances. In 1918–19, almost all of the well-observed influenza deaths were due to bacterial infections in the lungs [
50]. Similarly, in 1957, deaths were often associated with bacterial pneumonia and staphylococcal infections [
51]. However, unlike the 1918 pandemic, secondary bacterial infections were partially controlled through antimicrobials in 1957 [
7]. The bacterial infections that did result in death were generally multidrug resistant [
52]. 1957 was the first time a pandemic virus was available for laboratory analysis and the first time that an influenza vaccine became available [
7,
50,
52]. Although 22,017 vaccine doses were reportedly allocated to Arizona, newspaper reports suggest that many of these doses had not been received by November 1957 [
38,
53] casting doubt on the explanation that low mortality rates in Maricopa County might be explained by vaccine administration.
Climate and its effects on virus survival and transmission may be a possible explanation for the lower mortality rates and reproduction numbers observed in Maricopa County during the expected waves. Influenza virus survival is optimal at low temperatures, low sunlight, and low absolute humidity [
16]. However, while humidity was generally low in Maricopa County, temperature and sunlight during the winter were not low, compared with other regions of the U.S. These environmental conditions may have contributed to the decreased reproduction number and mortality rates, as it would have been more difficult for the virus to survive transmission between hosts.
Conclusions
By using primary data from archival death certificates from 1954 to 1961 to quantify the age, seasonal, and transmissibility patterns of the second influenza pandemic of the 20th century, this study confirmed that Maricopa County largely avoided the effects of the 1957 pandemic. Compared to other regions of the United States, Maricopa County had few excess deaths due to the 1957 influenza pandemic. Although results varied between age groups, the 1957 pandemic in Maricopa County was characterized by a mild wave from October 1, 1959 to June 30, 1960, when there were 16.59 absolute excess-deaths due to respiratory causes per 10,000 population in the elderly (≥65 years), the age group most affected. However, the standardized mortality ratio peaked (4.06) in children and young adolescents (5–14 years) from October 1, 1957-March 31, 1958, based on mortality rates of respiratory deaths. Transmissibility was greatest during the same 1957–1958 period, when the mean reproduction number was a low 1.08–1.11, using 3 or 4-day generation intervals and exponential or fixed distributions.
Through analyzing archived newspaper articles, the low mortality and transmissibility rates recorded in Maricopa County were most likely not due to public health interventions or vaccine distribution. While there is much unknown about climate and virus survival, the environmental conditions in Maricopa County may have prevented high transmission and excess-mortality rates. By analyzing historical data of different regions, researchers can better understand how mortality and transmission rates are related to different environmental conditions and public health interventions, providing important lessons to optimize current country-level preparedness and control plans.