Missouri in the 21^st century

Missouri is the 21^st largest state in area in the U.S. and consists of 114 counties and St. Louis City (hereafter included as one of 115 counties). It lies in the middle of the United States, close to the geographic center of the continental U.S. and containing the mean center of the present U.S. population [31]. The state has two major Metropolitan Statistical Areas (MSAs)–St. Louis on the east central border and Kansas City on the west central border, along with two other MSAs with populations over 200,000, Springfield in the south-central and Columbia in the central region of the state [32]. It also has three other sizable MSAs—Joplin in the southwestern corner of the state, Jefferson City in the central region, and St. Joseph along the northwestern border [32]. Outside of these places the state is largely rural. Except for Columbia and Jefferson City in the central part of the state, all of these population centers were evident in the early part of the 20^th century, when the influenza pandemic of 1918–20 spread through American communities.

The U.S. Census Bureau defines an urban county as one in which 50% or more of the county’s population lives in either “urbanized areas”, incorporated places of 50,000 or more people, or “urban clusters,” incorporated places adjacent to urbanized areas and with a population between 2500 and 50,000 [1]. Using this definition, 32 of the 115 counties in Missouri are currently classified as urban, while 83 are rural. About 1/3 of the total population currently lives in rural areas [33]. All counties that were urban in 1918 are still urban, and the majority of counties that were rural in 1918 are still classified as rural in 2020 by the U.S. Census Bureau (S1 Table). In addition, in the 100 years between the pandemics, there has been no change in the number or boundaries of the state’s counties.

The 2020 Missouri population was 6,154,913 persons, which ranks it 19th in the U.S.; females make up 50.6% of the population [34]. Over half of Missouri residents live in the St. Louis or Kansas City MSAs. In comparison to the U.S. as a whole, Missouri’s rate of population growth is lower (0.2% vs. 0.4%) and the percentage of high school graduates is higher (91% vs. 88.9%), but the percentage with at least a bachelor’s degree is lower (30.7% vs. 33.7%). The state also has substantially less racial and ethnic (especially Hispanic) diversity (78.7% vs. 59.3% White, non-Hispanic; 4.7% vs. 18.9% Hispanic), a lower percentage of foreign-born (4.2% vs. 13.6%), a smaller household size on average (2.5 vs. 2.6 persons per household), and lower median household income ($61,043 vs. $69,021). The age-sex distribution is similar to that of the rest of the U.S. [34]. The primary industries in the state at the present time are health care and social assistance, retail trade, and manufacturing [35].

As is true for most of the U.S., transportation occurs primarily by personal car. Missouri is bisected by two major east-west interstates (I-70 and I-44) and one north-south interstate (I-35). Passenger trains, which carry only a small number of passengers, run east-west between St. Louis and Kansas City and southwest to northeast from Kansas City to Chicago. No significant human travel occurs by water along the Missouri or Mississippi Rivers.

The distribution and nature of health care at the present time is quite different from that present in the early 20^th century. Between 10% and 20% or more of working age residents in almost all Missouri counties are uninsured [36]. A significant percentage of the rural counties do not have affordable health clinics, and approximately 11% of Missouri’s total population live in a county with no hospital; for these persons the closest hospital is 14–52 miles away (average 29 miles) [37]. Missouri also has one of the lowest percentages of health care needs met in the entire country [38]. Access to care varies considerably across the state, however. As a consequence, health screenings and preventive care as well as obtaining timely needed medical care for trauma, serious illnesses, and chronic conditions vary dramatically among the counties.

Missouri during the 1918–20 influenza pandemic

The information used to describe the demographic structure of Missouri in 1918 was collected from the 1910 Missouri census as this document is the closest census prior to the pandemic. The U.S. Census Bureau only produces decennial censuses for each state and despite an exhaustive search, information could not be found on sociodemographic characteristics of the state during the intercensal years. The 1920 Census was closer in time to the pandemic, but it was not chosen to represent the pandemic time period as it reflects major sociodemographic changes caused by the 1918–20 influenza pandemic itself and World War I. A comparison of data from the 1910 and 1920 censuses indicates that, although there were changes during the decade, they were moderate in scope. For example, the overall population increase between 1910 and 1920 was 110,720 people (3.4%), but this was half as large as the increase between 1900 and 1910, a time period that was not influenced by major events such as WWI and a global influenza pandemic that were present in the later decade [39]. Cities were also increasing in size at the expense of rural areas, but as was the case for the total population, these changes were relatively slow between 1910 and 1920 and the rate was lower than what was observed between 1900 and 1910. These and other changes indicate that the sociodemographic characteristics measured in the 1910 census, although not a perfect reflection of those characteristics in 1918, are a reasonable proxy for 1918.

In 1910, Missouri was the 7^th most populated U.S. state, with a total of 3,293,335 inhabitants. The sex distribution in 1910 was slightly different from the 2020 population—48.75% of the population was female [40]. The county divisions were the same in 1910 as they are at present, although the census did not designate counties as urban or rural. That designation was reserved for communities, which were deemed to be urban if they were incorporated and had over 2500 inhabitants. The census did indicate the percentage of state and county residents that lived in communities designated as urban, however. Over half (57.5%) of the population lived in communities with fewer than 2,500 inhabitants or in other smaller rural landscapes, making the state substantially more rural in 1910 than in 2020.

The state was less racially diverse in 1910 than in 2020–95% of the population identified as White (native and foreign-born) [40]. Black individuals constituted 4.8% of the population, and the remaining 0.2% of the population’s makeup consisted of Chinese, Japanese, Indian, and other ethnicities. The majority of Missouri’s non-White population lived in urban areas, a pattern that is still true today, although a substantial minority (about 1/3) lived in rural areas.

In 1918, agricultural employment was most common; 78.6% of Missouri land consisted of farmland [40, 41]. The manufacturing industry and trade employment were the next most common, likely due to those occupations’ support of the agricultural industry [41]. Mining was also an important source of employment, especially in the Southern parts of the state [42]. While agricultural employment often was relatively solitary, manufacturing and mining usually involved enclosed workspaces, which would facilitate the spread of diseases such as influenza.

Ownership of automobiles increased dramatically during the years before the 1918–20 influenza pandemic [43], but the main mode of transportation was by railroad. Railroads in Missouri were not used extensively by the public during 1917–1920, however, as they were appropriated by the federal government for World War I efforts [44].

In early 20^th century Missouri, it was already understood that urban and rural communities faced different sets of challenges and levels of health disparities [45]. Both hospital beds and physicians were disproportionately clustered in higher population counties [46, 47]. Thus, some aspects of the heterogeneity observed today were present in 1918.

It is more difficult to assess other domains by which health care is usually considered. Health insurance was not widely available in 1918, and most health spending was out of pocket. Charity homes could provide care to specific populations (e.g., unwed mothers or the elderly), but these were privately run, and little information is available about them. Additional health care resources included state-run inpatient facilities for high-need populations, such as epileptics, people with severe mental health needs, or people with tuberculosis [48].

Health care quality may be even more difficult to assess than capacity. According to early 20^th century State Board of Health reports, the primary concern was how best to implement sanitary and hygienic measures [45]. Control of infectious disease was the main priority, but available tools still focused on non-pharmaceutical measures like cleanliness and food and water safety. Although the state had a laboratory used for pathogen identification, it was not able to provide timely diagnoses of viral diseases, such as influenza. So, while there were advances at the state laboratory in evidence-based protocols for clinical diagnosis, treatment, and control of many infectious diseases, implementing these advances in treating patients at point of care was limited.

This comparison illustrates that many aspects of life in Missouri in 1918 were similar to those in 2020, but there were also significant differences. Similarities include the high proportion of rural land, the geographic locations of urban centers, and relative homogeneity of the population across counties in terms of the age, sex, and racial and ethnic distribution. Differences include better roads and nearly universal access to vehicles, resulting in greatly increased mobility in the 21^st century. Although the number and locations of urban areas are similar, their resident populations are much greater today, increasing the risks of person-to-person transmission of infectious diseases. The predominant occupations have changed significantly, with marked declines in mining and agricultural activities and increases in health professions, the retail sector, and manufacturing. These changes likely have had effects on underlying risk factors for a variety of health conditions. Access to health care during both time periods was heterogeneous, with substantially greater availability and access in urban counties than in rural counties. Treatment of infectious diseases in 1918 was much less sophisticated than at present. This suggests that the outcomes for those who were afflicted with influenza in 1918 may have been more similar throughout the state due to fairly equal treatment regimens, while outcomes in 2020 may exhibit more obvious differences between residents of rural vs. urban counties due to variability in accessing appropriate treatments. In light of these comparisons across the last century, we turn to the two pandemics bounding that century and discuss their impacts overall, across different regions of the state, and in urban vs. semirural vs. rural counties.

Selection of sociodemographic variables

Data on 20–30 different sociodemographic variables were initially collected for each time period. To ensure comparability of analyses of both pandemics, however, data used in the study were limited to a smaller set of variables that adequately measured similar population characteristics.

The variables used in analyses by county included total population size, proportion male, ratio of people aged 20–44 to those aged 65 and older, proportion White, literacy rate in 1910 and high school graduation rate in 2020, average value of a farm, and number of primary care physicians per 100,000 persons. Proportion Hispanic was included in analyses of the 2020–21 COVID-19 pandemic, but in 1910 Hispanic proportions were very low and data were unreliable, so the variable was omitted from analyses of the influenza pandemic. Variables chosen for the analyses were minimally correlated with one another. The values used for the chosen variables are provided in S1 File.

Several challenges arose during data collection, especially for data from the early 20^th century. For example, the 1910 Missouri census had to be used to describe the demographic structure of Missouri in 1918 as this document is the closest census prior to the pandemic. Sources other than the 1910 Missouri census, with state-level information on income, education, or living standards could not be identified.

County-level data were especially sparse for the early 20^th century. Population numbers were provided for some age categories, but at the county level these were not the standard 5-year age categories used in demographic data today and these sometimes changed from census to census.

It is probable that the 1910 population and death counts for non-White residents were underreported as they did not have the same rights or treatment as White individuals. In addition, regulations requiring death registration were just becoming widespread at the time of the pandemic—Missouri was first included in the U.S. registration area in 1911 [49], so deaths in all age groups, and especially young children, were likely under-recorded. This problem was probably worse for underrepresented groups [50, 51].

Polk’s Medical Register [47] was used to provide data on health care access during the early 20^th century. These data are imperfect estimates, as all types of health care providers were conflated into the umbrella terms of physician or surgeon. No documents describing household-level economic variables, such as per capita income by county were available; the only economic information of that type uncovered was the average value of all farm property per county provided by the 1910 Missouri Census Supplement [40].

The opposite problem occurred with the collection of sociodemographic data during the 2020–21 COVID-19 pandemic—there exists an overabundance of all types of data. The goal of identifying a common set of variables from both time periods as well as minimizing the correlations among variables motivated the final choice of variables used in the analyses.

Using sociodemographic data to aggregate Missouri’s counties

To enable the comparisons of the influenza and COVID-19 pandemics, two methods were used to aggregate the state’s counties: a) distribution based on the urban and rural proportions in their respective populations, and b) sociodemographic similarity (S1 Table). The criteria used for both classification schemes did not include any specific disease-related input and thus the classifications provide independent foundations to use in evaluating the spread of each virus—influenza or SARS-CoV-2.

The RSU (rural/semirural/urban) classifications for both 1918 and 2020 were based on the urban proportion of a county’s population reported in the 1910 and 2010 censuses and the present definition of “urban” used by the U.S. Census Bureau (a county with over 50% of the population living in urbanized areas or urban clusters) [1]. The 2010 census was used because criteria for determining the rural/urban classifications for the 2020 census were being revised and the new designations were not available at the time the analyses in this paper were completed. Within the rural category we have defined a subgroup, semirural, to include those counties in which over 50% but less than 100% of the residents live outside of urbanized areas or urban clusters. Those counties with a totally rural population are classified as “rural”.

County-level sociodemographic variables were used to group counties into four distinct regions in both 1918 and 2020. In each time period, Ward clustering was performed on these variables using the ClustGeo package in R [52]. This method assigns cluster membership by minimizing the Euclidean distance between county sociodemographic variables and results in clusters of counties that are sociodemographically similar. To account for geographic similarities, county centroids were used to introduce a weighted contiguity constraint. The constraint provided a means of selecting the amount of influence physical proximity had on the sociodemographic homogeneity of each cluster. Selecting a reasonable weighting of the physical space dimension relative to the feature space was accomplished by choosing a value that minimized the reduction in sociodemographic similarity while maximizing the influence of spatial proximity. Using a Ward dendrogram of all possible cluster divisions based on sociodemographic features as well as input from project collaborators, four clusters were selected for both time periods.

Once the clusters were defined, a random forest algorithm [53] was used to assess the relative importance of each individual sociodemographic variable for determining group/cluster membership. The goal in evaluating individual variable importance was to provide additional interpretability when assessing differences in disease progression between the four county clusters. Descriptive labels were used to characterize each of the regions for a particular time period, but it is important to note that the clustering procedures were done independently on the variables from each time period, with the result that the county composition of each region, although similar, is not identical for the two periods.

Disease data from the 1918–20 influenza pandemic

The primary source of data on influenza mortality used in this project is the Missouri Digital Heritage death certificate database [22], which contains all death certificates held by the state and filed between 1910 and 50 years before the present (a rolling boundary that changes each year). Birth date, death date, age, sex, and other demographic data were collected on all people who died from pneumonia and influenza in all 115 Missouri counties during 1918, 1919, and 1920. Missouri became a death registration state and reported mortality statistics to the U.S. Census Bureau beginning in 1911. Thus, the state is listed in the U.S. Mortality Statistics reports [e.g., 49] after this time. Data in these reports provide information only at the state level and occasionally for selected cities, however, so data culled from death certificates, which provide death statistics at the county level, have been used for most analyses in this paper.

Additional information on the 1918–20 influenza pandemic was collected from newspaper reports, State Board of Health Biannual reports, and other ephemera found in archives. These information sources provide illustrative details about control measures and the social climate that existed in Missouri during the pandemic period, but rarely provide comprehensive quantitative data on epidemic outcomes.

Disease data on the 2020–21 COVID-19 pandemic

Both county-level case and mortality data dated between March 7, 2020 and March 9, 2022 were obtained directly from the Missouri Department of Health and Senior Services (MODHSS) or by using their public facing dashboard. The data capture the first 4 waves of the pandemic (through the first omicron wave). Subsequent to this time, reliable comprehensive reporting and publication of data was affected by the increasingly greater availability of vaccination and self-testing and the waning of the first omicron wave.

Analysis of the mortality and morbidity data

Results are presented below for morbidity (COVID-19 only) and mortality patterns (both pandemics) by RSU status and sociodemographic region. Because some counties had low numbers of deaths, especially at the time of the influenza pandemic, all analyses of death data aggregated the death counts according to the groups defined by RSU status or the sociodemographic cluster analysis. Data were also not available for age-adjustment during the 1918–20 pandemic, so to maintain comparability of both pandemics, crude mortality rates were calculated (referred to as mortality rates in subsequent sections). Cumulative mortality and morbidity patterns were calculated directly from the relevant daily death or case data. Counties were grouped as previously described, and cumulative deaths per 100,000 persons in these aggregated regions was used to compare regional mortality. Time series of deaths/100,000 persons or cases/100,000 persons were used to visually compare the county groupings.

Estimation of excess all-cause mortality

The mortality patterns associated with both the 1918–20 influenza and 2020–21 COVID-19 pandemics in the state were quantified by estimating monthly all-cause excess mortality rates per 10,000 population. County-level monthly mortality figures are not easily accessible for either pandemic; thus, excess all-cause mortality estimates were calculated only for the state as a whole. For the influenza pandemic, total monthly deaths for 1911–21 were extracted from the Missouri Digital Heritage database [22]. Annual population figures were published in a report from the U.S. Bureau of the Census [54]. For the COVID-19 pandemic, monthly all-cause death totals for 2011–21 were reported in the Monthly Vital Statistics reports of the Missouri Department of Health and Senior Services [55]. Population estimates for those years were derived from U.S. Census data and were downloaded from https://www.macrotrends.net/states/missouri/population. Complete COVID-19 mortality data were not available for 2022, so excess all-cause mortality during the COVID-19 pandemic was determined only through the end of 2021. All-cause excess mortality rates were found by computing the mortality rate above a seasonal baseline of expected mortality rates in the absence of influenza or COVID-19 activity.

The time series of monthly all-cause deaths during each pandemic was first examined to estimate the timing of pandemic influenza or COVID-19 activity. The months that were determined to be associated with pandemic activity were excluded so that a baseline non-pandemic mortality rate could be modelled. The baselines for the two pandemics and their 95% confidence intervals were estimated by fitting cyclical Serfling regression models to all-cause deaths for the total populations in non-pandemic months, with separate models used for each pandemic, as done in previous studies [56–59]. Temporal trends and harmonic terms for seasonality were included in these linear regression models [60]. Periods of pandemic activity were then defined for influenza as the months in 1918 through 1920 where the observed total all-cause mortality rate per 10,000 population was greater than the upper 95% CI of the baseline. The same strategy was used to determine periods of COVID-19 pandemic activity.

For both pandemics, the excess mortality rate was defined as the difference between the observed and model-adjusted baseline mortality rates for each month within a pandemic wave. The excess death rates were summed across the pandemic months from 1918 through 1920 to get the overall pandemic excess mortality during the influenza pandemic; excess death rates were summed across the pandemic months in 2020 and 2021 for the COVID-19 pandemic, in line with previous studies [56–59]. All excess mortality analyses were completed using SAS 9.4.

The rural/semirural/urban (RSU) classification and sociodemographic clustering of the state’s counties

The proportion of residents living in rural areas in most Missouri counties decreased with economic development and population growth between 1910 and 2010. The RSU groups remained stable for about half of the counties (n = 61) and all urban counties in 1910 were still classified as urban in 2010. The remaining changes in RSU classification were as follows: semirural in 1910 to urban in 2010 (n = 18), rural in 1910 to semirural in 2010 (n = 30), and rural in 1910 to urban in 2010 (n = 6) (S1 Table).

Fig 1 illustrates the location within the state of the sociodemographic clusters for the two pandemic time periods; the designated cluster for each county is also indicated in S1 Table. In both time periods, the counties were separated into four groups. In 1910 the southern half of the state separated into distinct southeastern and southwestern clusters, while the other two clusters consisted primarily of central and northern counties (Fig 1A). The two counties in the southwestern part of the state that fell into the central group included the cities of Springfield and Joplin (SW portion of the state), both of which were urban counties.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Classification of Missouri counties into four regions based on sociodemographic characteristics.

(A) The 1910 classification. (B) The 2020 classification. Source: Authors’ original figure. Data for the base map is in the public domain and available from the State of Missouri Data Portal at https://data.mo.gov/Demographics/Missouri-County-Boundaries-Map/n34b-fwqr. The data made available here has been modified for use from its original source, which is the State of Missouri. THE STATE OF MISSOURI MAKES NO REPRESENTATIONS OR WARRANTY AS TO THE COMPLETENESS, ACCURACY, TIMELINESS, OR CONTENT OF ANY DATA MADE AVAILABLE THROUGH THIS SITE. THE STATE OF MISSOURI EXPRESSLY DISCLAIMS ALL WARRANTIES, WHETHER EXPRESS OR IMPLIED, INCLUDING ANY IMPLIED WARRANTIES OF MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE. The data is subject to change as modifications and updates are complete. It is understood that the information contained in the Web feed is being used at one’s own risk.

https://doi.org/10.1371/journal.pone.0290294.g001

The clustering in the 21^st century is broadly similar, but has some distinct differences. The urban grouping in 2020 is mostly limited to the highly urbanized parts of the state and does not consist of geographically contiguous counties (Fig 1B). The counties containing Springfield and Joplin are not included in the urban group in 2020, most likely due to broad similarities in the chosen sociodemographic characteristics across the southern half of the state. The other feature that is distinctive about the 2020 regional grouping is that the Bootheel region in the southeastern corner of the state forms its own sociodemographic cluster. This region is known to be socially, culturally, and economically distinct from much of the state [61], and it is significant that the variables we chose capture the uniqueness of the region.

For both time periods, the random forest algorithm identified average farm value, proportion White, and literacy as the most significant sociodemographic variables that determined cluster membership. Two additional variables, proportion Hispanic and age ratio had an appreciable influence on the 2020 county clusters as well.

General pattern of the 1918–20 influenza pandemic in the State of Missouri

The pattern of spread of the global 1918–20 influenza pandemic has been characterized as a series of three waves occurring during an approximate one-year period in 1918 and 1919 [62, 63]. A growing body of recent research suggests that many regions, including the State of Missouri, experienced a substantial fourth wave of the pandemic in 1920 [58, 64–71].

The first wave in Missouri (Spring 1918) was unremarkable, with an all-cause excess mortality rate of 1.98 (95% CI 0.07, 3.88) per 10,000 population (Figs 2 and S1). By the start of the second wave, it was recognized that the pandemic was more serious than normal and newspapers and public health authorities warned of increased severity, leading to mandatory social distancing and closures of many non-essential businesses in major cities [19]. This fall wave was bimodal with peaks in October and December 1918 and highly significant excess mortality of 12.69 deaths per 10,000 persons (95% CI 10.79, 14.59) in October and 12.34 per 10,000 persons (95% CI 10.44, 14.25) in December. The ending of the wave is unclear due to the presence of a small third wave in the early spring of 1919 (Figs 2 and S1). Thus, our all-cause excess mortality analyses lump waves II and III into the time period from September 1918 to April 1919. In a fourth wave of the pandemic (November 1919-April 1920), an additional and substantial peak in influenza mortality occurred; in fact, this mortality peak was higher than that of the second 1918–19 wave in over 25% of Missouri’s 115 counties. The excess all-cause mortality estimates for the fourth peak were highly significant in February 1920, with a rate of 10.63 deaths per 10,000 persons (95% CI 8.71, 12.54). Overall, the all-cause excess mortality during the 8 months of Wave II/III and 6 months of Wave IV was 51.6 deaths per 10,000 persons (95% CI 40.2, 63.1) and 6 months of that 14-month period exhibited statistically significant excess mortality. This level of excess mortality is substantially higher than that reported by Hoffman and Fox [24], who used a different calculation method, but the pattern of mortality observed in their data [24, Fig 2] is that same as that presented here (Figs 2 and S1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. All-cause excess deaths in Missouri during the 1918–20 influenza pandemic.

Waves occurred over the following spans: I—early 1918 (through April), II/III—September 1918-April 1919, IV—November 1919-April 1920. Dashed boxes show pandemic periods. Note: During Wave I of the 1918 influenza pandemic there was only a single month with statistically significant excess mortality; thus, this wave is not indicated as a pandemic period.

https://doi.org/10.1371/journal.pone.0290294.g002

The timing and severity across time of the pandemic in Missouri was somewhat different from that of the United States as a whole. Missouri’s death rate from pneumonia and influenza (P&I) per 10,000 population was substantially lower in 1918 (47.7 vs. 58.7 deaths per 10,000) and somewhat lower in 1919 (20.6 vs. 22.2 deaths per 100,000 population), but substantially higher in 1920 (26.1 vs. 20.6 deaths per 10,000 population) [72]. These and the excess mortality illustrated in Fig 2 confirm the severe impact of the pandemic in 1920 in comparison to the rest of the U.S. In fact, of all the death registration states reporting in 1921, only Colorado experienced more P&I mortality in 1920 than did Missouri.

The Missouri COVID-19 Pandemic (Mar 2020-Mar 2022)

The first case of COVID-19 was reported in Missouri on March 7, 2020 with the first death reported on March 22, 2020. From this beginning to the end of data collection (March 9, 2022), Missouri experienced four waves of COVID-19, most of which seem to be closely associated with the circulation of new strains of the SARS-CoV-2 virus. During the time period covered by the present study, the state experienced an early, small wave between March and June 2020. This wave peaked in April 2020 with an all-cause excess mortality rate of 0.77 deaths per 10,000 people (95% CI 0.03, 1.52). A large second wave occurred between July 2020 and January 2021; it peaked in December 2020 with all-cause excess mortality of 4.06 deaths per 10,000 people (95% CI 3.31, 4.82). Missouri experienced a brief respite from the pandemic between February and June 2021, but from July through October 2021 there was a third wave. Deaths in this wave peaked in mid to late August 2021 (all-cause excess mortality of 3.18 deaths per 10,000 people, 95% CI 2.40, 3.95). A fourth wave occurred between November 2021 and the end of data collection with peak mortality in early February 2022 (Figs 3 and S1). Excess all-cause mortality calculations could not be made for the fourth wave due to a lack of available data, but it was remarkable for much higher case numbers than any of the previous waves, but lower case fatality rates than the second wave.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. All-cause excess deaths in Missouri during the 2020–21 COVID-19 pandemic.

The following waves can be observed: I—April-June 2020, II—July 2020-January 2021, III—July 2021-October 2021, IV—November 2021-end of data collection (shown only through December 2021; see S1 Fig for remaining months). Dashed boxes show pandemic periods.

https://doi.org/10.1371/journal.pone.0290294.g003

The time span of the pandemic covered by our data divides naturally into two phases, from April 2020 through January 2021 (Waves I and II) and from July-December 2021 (Waves III and IV). During the 10 months of Phase I, the all-cause excess mortality was 18.0 deaths per 10,000 people (95% CI 11.3–24.8); 9 of these 10 months exhibited statistically significant all-cause excess mortality. The all-cause excess mortality for the 6 months of Phase II was 10.2 deaths per 10,000 people (95% CI 5.5–14.9) and all 6 months exhibited statistically significant all-cause excess mortality. Combining these two phases leads to an overall estimate for the all-cause excess mortality of 28.3 deaths per 10,000 people (95% CI 16.8–39.8).

Subsequent waves with lower case numbers and markedly lower mortality than the fourth wave have occurred since data collection for the present study ended. Interested readers can learn more about Missouri’s COVID pandemic on the CDC COVID Data Tracker.

RSU comparisons throughout the two pandemics

Daily and cumulative mortality curves for the RSU groups during the 1918–20 influenza pandemic are illustrated in Fig 4. Fig 4A indicates that the pandemic occurred in a similar pattern in all three regional groupings, with some temporal separation during the 1920 wave. This final epidemic peak occurred first in the urban RSU group, around February 1, followed a week later in the semirural RSU group. The rural RSU group peaked about two weeks after the urban peak. The similar patterning of the pandemic in all regions is also indicated in Fig 4B, which exhibits parallel trajectories in the crude mortality rates per 100,000 persons throughout the 1918–20 pandemic, especially during and after the rapid rise of the second wave in October 1918. The rural RSU group exhibited the lowest cumulative crude mortality rate throughout the pandemic, with the urban RSU group consistently at the highest cumulative crude mortality rate. Fig 4A and 4B also indicate that urban counties experienced substantially more mortality early in the pandemic than did non-urban counties. A high 1918–20 influenza mortality rate in the cities was also noted by Hoffman and Fox [24].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Daily and cumulative influenza mortality curves among groups defined by rural status, 1918–20.

(A) Daily influenza deaths per 100,000 people. (B) Cumulative influenza deaths per 100,000 people. All curves are 7-day rolling averages of the data within each county cluster. Pandemic peaks begin when the curves rise most steeply.

https://doi.org/10.1371/journal.pone.0290294.g004

During the COVID-19 pandemic between March 2020 and December 2021, early introduction of the virus into the urban RSU group is apparent, as it shows a quicker rise in both cases and deaths (Figs 5 and 6). The higher mortality rate in the urban RSU group during the first six months of the pandemic is especially marked (Fig 6). Through most of the time period included in this study, the rural RSU group exhibited fewer cases per 100,000 persons (Fig 5A), but its cumulative mortality rates were similar to those seen in the urban RSU group from December 2020 through July 2021 (Fig 6A) before rising above the urban RSU group. Mortality rates in the semirural RSU group occurred at rates similar to the rural RSU group through June 2021, but after that they increased markedly in comparison to the urban and rural RSU groups. A comparison of cumulative case rates with cumulative mortality rates, especially between November and the end of the study period (Figs 5B and 6B), indicates that the semirural RSU group had similar case rates to the urban RSU group, but much higher mortality rates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Daily and cumulative COVID-19 case curves among groups defined by rural status, March 2020-December 2021.

(A) Daily cases per 100,000 persons. (B) Cumulative cases per 100,000 persons. All curves are 7-day rolling averages. The early peak in rural cases in June 2020 was likely due to a mass testing event within one or more large businesses.

https://doi.org/10.1371/journal.pone.0290294.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Daily and cumulative COVID-19 mortality curves among groups defined by rural status, March 2020-December 2021.

(A) Daily deaths per 100,000 persons. (B) Cumulative deaths per 100,000 persons. All curves are 7-day rolling averages.

https://doi.org/10.1371/journal.pone.0290294.g006

Analysis by sociodemographic region

The timing of the major peaks during the 1918–20 influenza pandemic was similar in all sociodemographic regions (Fig 7A), although the parallel trajectories observed in the RSU curves (cf. Fig 4) do not become apparent until after the end of the third wave in late Spring 1919 (Fig 7B). The later development of parallel trajectories may be related to a moderate peak in January 1919 observed only in the Southeast region (Fig 7A). The cumulative mortality curves separate into two distinct groups, Central/Southeast and North/Southwest (Fig 7B). The latter two regions are similar to each other throughout the pandemic; the Central and Southeast regions are distinct from one another until the rise of the second wave in October 1918. The Southeast region is aligned with the North and Southwest regions prior to this time while the Central region has substantially higher cumulative mortality than all other parts of the state. The mortality for both the Central and Southeast regions rises very quickly at the beginning of the second wave, and the two regions become similar to one another by the beginning of November 1918, although the cumulative mortality in the Central region remains higher than what was observed in the Southeast region throughout most of the pandemic.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Daily and cumulative influenza mortality curves among sociodemographic clusters, 1918–1920.

(A) Daily deaths per 100,000 persons. (B) Cumulative deaths per 100,000 persons. All curves are 7-day rolling averages.

https://doi.org/10.1371/journal.pone.0290294.g007

The Central region defined for the influenza pandemic contains all the urban areas present in the state and is clearly the most densely populated region. This is likely a major reason for the higher cumulative mortality levels throughout the pandemic. This higher mortality is especially apparent during the mild first wave when the Central region was distinct from all other regions.

All regions experienced significant peaks in influenza mortality rates in 1920, with Kansas City and St. Louis experiencing the highest rates per 100,000 persons (Fig 7A). The impact of the 1920 wave is also visible in the cumulative mortality curves as a rapid rise beginning in February 1920 and a continued positive slope after that rise, especially in the Central and Southeast regions (Fig 7B).

To further examine the potential associations between sociodemographic characteristics and the severity of the influenza pandemic, the important sociodemographic variables identified using the random forest algorithm mentioned previously (average farm value, proportion White, and literacy) were compared between the two groups, Central/Southeast and North/Southwest. Uncorrected, one-sided T-tests show that the Central/Southeast clustered counties had significantly lower mean literacy rates and proportion White (p ≈ 0 for both variables). Average farm values for these combined regions are not significantly different at the .05 level (p = 0.06). See the S2 Fig, for boxplot comparisons of these variables.

Mortality during the 2020–21 COVID-19 pandemic shares some features with the 1918–20 influenza pandemic, including higher mortality in urban areas at the beginning of the pandemic and exceptionally high mortality in the southeastern section of the state (in this case, only in the four counties furthest to the south and east–the “Bootheel” counties) (Fig 8). The North and South counties had nearly identical mortality patterns throughout the pandemic, as was the case during the 1918–20 influenza pandemic, although in that pandemic, most counties in the southeastern quadrant were clustered together with the Bootheel counties.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Daily and cumulative COVID-19 mortality curves among sociodemographic clusters, March 2020-December 2021.

(A) Daily deaths per 100,000 persons. (B) Cumulative deaths per 100,000 persons. All curves are 7-day rolling averages.

https://doi.org/10.1371/journal.pone.0290294.g008

Case numbers per 100,000 population during the COVID-19 pandemic indicate that the Bootheel region was hit much harder by the pandemic than other regions when corrected for differences in population size. As Fig 9 shows, both the daily incidence curves and the cumulative case rates in this region were substantially higher than observed in all other regions throughout much of the time period included in this study.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Daily and cumulative COVID-19 cases among sociodemographic clusters, March 2020-December 2021.

(A) Daily cases per 100,000 persons. (B) Cumulative cases per 100,000 persons. All curves are 7-day rolling averages.

https://doi.org/10.1371/journal.pone.0290294.g009

To further examine the potential associations between sociodemographic characteristics and the severity of the 2020–21 COVID-19 pandemic in the Bootheel, the four counties in that region were aggregated and compared to the combined North, South, and Urban counties. Mean comparisons were conducted using simple, one-sided T-tests on these two sets of counties. The Bootheel-clustered counties have significantly lower literacy levels (p ≈ 0) and proportion White (p = 0.01) than the rest of the state. The average farm value for Bootheel counties is significantly higher (p = 0.001). The proportion of Hispanics and age ratios are not significantly different. It should be noted that the Bootheel cluster includes only a small number of counties (4 out of 115 total Missouri counties) which lends less credibility to these mean comparisons due to possible undetected issues with distributional assumptions required for the statistical tests. However, as the small sample does not appear to be heavily skewed (see S3 Fig), there is still reasonable evidence to support the objective validity of these conclusions.

The analyses of data from both pandemics indicates that the southeastern part of the state experienced more severe pandemic outcomes during both pandemics, but this impact appears to have become more localized to the Bootheel region rather than the entire southeast in the 21^st century. As discussed above, the most important explanatory variables, literacy, average farm value, and proportion White, were the same in both time periods, supporting the conclusion that negative outcomes during pandemics may result from long-standing disparities in education and factors, such as generational poverty and limited access to resources, that may vary differentially among racial and ethnic groups.