We express our gratitude to Sven Vrielinck from Ghent University for his timely and comprehensive delivery of data. Additionally, we extend our thanks to Jean-Paul Sanderson from the Université Catholique de Louvain for his collaboration in providing the more recent data. Further acknowledgment is directed towards Lise Bevernaegie, Harmony Brulein, Janna Dinneweth, Philippe Paeps who contributed to the digitalization of the data. This paper is part of the project titled “How inequality kills. Two centuries of social and spatial disparities in all-cause and cause-specific mortality in Belgium (1800-2025)”. The project (EOS number 40007537) has received funding from the FWO and F.R.S.-FNRS under the Excellence of Science (EOS) programme.

Introduction

1Tuberculosis (TB) is an infectious disease caused by Mycobacterium tuberculosis, typically transmitted through the inhalation of aerosol particles. Despite its ability to infect various organs, pulmonary TB is the predominant form of the disease [Toujani et al., 2015; Mackenbach, 2020; WHO, 2021]. Common symptoms of pulmonary TB include a chronic cough, fatigue, fever, and weight loss [Reid & Garrett, 2018; Mackenbach, 2020; WHO, 2021] . Historically, TB has been known under various names, such as phthisis or consumption. Many colloquial names are also met until the 20^th century, such as the white plague and the King’s evil. In Belgium, TB was named in historical sources as “tuberculose,” “consumption,” “peste blanche,” “tering” and “phthisis.” We here focus on pulmonary TB, the most prevalent form of the disease and the most significant over time [Wood & Sheldon, 1997]. TB has gone with humankind for at least 8000 years. Its incidence fluctuated, but it only became a major human threat in the early 19^th century with the advent of the Industrial Revolution in Western Europe [Davies, 2005] . TB caused millions of deaths in Europe, especially among the poor. At its peak around 1850, TB was the leading cause of death in many industrialized countries, accounting for 15-20 % of all deaths [Kochi, 1994; Bynum, 2012]. The rise of TB in 19^th century Europe is closely tied to the Industrial Revolution, along with large-scale urbanization, poverty, and overcrowding [Glaziou et al., 2018; Mackenbach, 2020]. Improvements in nutrition, living and working conditions, public health and sanitation interventions, as well as a better understanding of the disease, were the key factors driving the decline of TB in the 20^th century [Dye et al., 2009; Glaziou et al., 2018; Mackenbach, 2020]. Given that today an estimated quarter of the world population carries M. tuberculosis, and at least 15 million have active disease, TB remains a global concern. While the current prevalence is influenced by issues like multi-drug resistance and HIV/AIDS, TB continues to be strongly associated with spatial and socio-economic inequalities, making its historical study relevant for modern times as well [WHO,2021]. Since the 1990s, TB has reappeared in industrialized countries, an increase associated to rising prevalence of homelessness, drug use, economic inequality, comorbidity with HIV/AIDS and deteriorating health structures, making the population more susceptible [Kochi, 1994; Mackenbach, 2020; WHO,2021].

2Until the early 20^th century, infectious diseases were responsible for over half of all deaths in Belgium [Devos, 2006]. The specifics of how the mortality transition unfolded in Belgium are not well known [Asaria et al., 2012; Eggerickx et al., 2020]. The primary goal of this study is to describe the spatial and temporal evolution of TB mortality in Belgium from 1889 to 1991, using a unique dataset compiled and digitized by the authors of this paper. This dataset includes data never analysed before, offering a unique perspective on the historical trajectory of TB mortality. We assess the spatial and temporal heterogeneity of prevalence at different time points and according to sex.

Historical determinants of spatial and temporal differences of tuberculosis mortality in Western Europe

3Scholars have evoked various determinants for the spatial and temporal differences of TB mortality in industrialized countries during the 19^th and 20^th centuries. Here, we review the most discussed proximate and distant risk factors of TB incidence and mortality (Figure 1).

4One of the main drivers commonly associated with a higher incidence of TB is urbanization and high population density. As large-scale industrialization arose in many European nations in the 19^th century, it was often accompanied by unorganized urbanization. Unsanitary, overcrowded housing, along with poverty, constituted an optimal environment for the transmission of TB, especially among the labouring populations, who lived in particularly poor conditions [Newsholme, 1906; van Helden, 2003; Glaziou et al., 2018; Mackenbach, 2020]. These factors, as well as an inadequate public health response, were responsible for the high incidence of various infectious diseases in large cities at the time [Gyselen, 1994; de Vries et al., 2014]. In England, a quarter of the urban TB deaths between 1860 and 1890 were associated with poor housing and the high population density around factories [Wood & Shelton,1997]. Industrialization grew substantially in Belgium in the 1840s and displayed all these features. During the Industrial Revolution, many people migrated to urban areas seeking better opportunities but often found themselves in low-paying jobs or facing unemployment, which exacerbated poor living conditions and urban overcrowding [Leboutte et al., 1998].

5Working conditions during the Industrial Revolution were especially a key factor of TB mortality for women who worked in large numbers in the textile industries [Eggerickx et al., 2020]. Poor working conditions in factories, such as inadequate ventilation, and lack of food and light can affect resistance to TB [Heaf, 1947]. Employment in very dusty conditions such as those experienced by grinders, sand workers, and coal heavers may have indirectly promoted the transmission of TB through coughing and sneezing from dust, especially in the absence of efficient personal protection equipment, and may also have deteriorated lung health [Heaf, 1947; Davies, 2005; Mackenbach, 2020].

6Poor nutrition may reduce immunity and affect TB mortality [Farmer, 2000; Davies, 2005]. Malnutrition and undernutrition may further lead people to develop TB or other infectious diseases, leading to comorbidities [Macallan, 1999; Gupta et al., 2009]. Before the advent of antituberculosis chemotherapy in the 1950s, a combination of a nutritious diet and rest was used to treat TB [Gupta et al., 2009; Glaziou et al., 2018]. People of lower socio-economic status also tended to have less access to information about prevention and treatment [Lönnroth et al., 2009; Odone et al., 2014].

7TB was also prevalent in rural areas. In general, people in rural areas had less access to health services and poor relief compared to those in urban centers [Lönnroth et al., 2009; Dye & Williams, 2010]. Higher female mortality is often associated with discriminatory practices in nutrition and healthcare that favoured men. An excess of female mortality was for example observed in rural regions of England, which may have accelerated the development of TB [Lönnroth et al., 2009; Dye & Williams, 2010]. Due to their unprivileged social conditions, women were more vulnerable to succumbing to infectious diseases [Eggerickx & Tabutin, 1994; Davies, 2005; Reid & Garrett, 2018].

8Crises often worsen poverty and poor socio-economic status which created a favourable context for TB. In Belgium, a textile and agricultural crisis in the 1840s and 1850s led to a general impoverishment of the rural population, especially in the northern region of Belgium [Eggerickx et al., 2020]. During war times, there was widespread impoverishment of the population and poorer nutrition, leading to an increase in TB mortality [Glaziou et al., 2018].

9Other lifestyle factors can also affect TB mortality. Smoking and alcoholism affect individuals’ immune systems and pulmonary health and make them more susceptible to the disease [Davies, 2005]. The age structure of the population can also increase overall mortality. Biological immunity varies with age, and the duration of exposure to disease risk increases over time. During the late 19^th century in Belgium, children and adolescents aged 5 to 19 years demonstrated increased biological susceptibility to tuberculosis (TB) [Devos, 1996]. Similarly, the elderly are more susceptible to TB due to factors such as immunosenescence, increased individual susceptibility, and a higher likelihood of residing in communal settings, which elevate the risk of infection [Caraux-Paz et al., 2021]. The behavioural and social habits of different age groups, such as communal living, also impact their risk of TB exposure [Davies, 2005; Dye & Williams, 2010]. This underscores the significant influence of age on vulnerability to the disease during this era.

10After the peak in the mid-19^th century, TB started to decline progressively in Western Europe before the advent of an effective drug treatment. A range of factors contributed to this progressive decline, ranging from factors affecting exposure to host susceptibility and access to care [Eggerickx et al., 2017; Mackenbach, 2020].

11The decline in TB mortality is part of the broader context of the mortality transition, characterized by a long-term shift in the causes of mortality since the mid-18^th century. Infectious diseases were gradually replaced by degenerative diseases (cancers, cardiovascular diseases, diabetes, etc.) and social diseases (suicide, smoking, alcoholism, etc.) [Picheral, 1989; Eggerickx et al., 2017; Mackenbach, 2020]. This phase of the mortality transition is commonly called the "age of declining pandemics" and was accompanied by increased life expectancy and decreased general mortality [Meslé & Vallin, 2000; Eggerickx et al., 2017]. In Belgium, the gradual decline in infectious diseases was only interrupted by a few epidemic episodes such as cholera (1866), the Spanish flu (1919), and during the two World Wars [Eggerickx et al., 2017; Devos,2006]. TB followed the mortality transmission with a progressive decline thanks to improvements in living and working conditions, medical knowledge and health infrastructures and policies addressing socioeconomic determinants.

12Improvements in living and working conditions, specifically more spacious houses and better-ventilated rooms with fewer occupants likely contributed to preventing contagion [Dye et al., 2009; Mackenbach, 2020]. Belgium, like many European countries, experienced a general rise in living standards in the 19^th century with improved housing, as well as higher real incomes, and lower food costs, also creating less favourable conditions for TB spread and severity [Wilson, 2005; Odone et al., 2014; Mackenbach, 2020]. However, as demonstrated by the Hungarian experience, where TB continued to be prevalent despite rising living standards, other factors were likely needed to support the decline [Wilson, 2005].

13The decline of TB in Europe is also commonly associated with improvements in medical knowledge. The description of Mycobacterium tuberculosis (then called tubercle bacillus), the pathogenic agent of TB discovered by Koch in 1882, established the contagious nature of TB [Gyllepsie, 2006; Frith, 2014]. A better understanding of the transmission led to designing and implementing extensive public hygiene plans dedicated to fighting TB. Such public health policies were key to the decline of TB, involving preventive measures such as identifying vulnerable populations and isolating patients in sanatoria and other healthcare institutions [Lefebvre & Falzon, 2008; Wanlin, 2018; Mackenbach, 2020]. Measures implemented in England and the USA coincided with an accelerated reduction in TB cases. Infrastructures were sometimes insufficient to accommodate all cases, these facilities played a role in curbing TB spread by isolating numerous infectious patients from the general population [Wood & Sheldon, 1997].

14The more systematic use of antibiotics and vaccination from 1950 onwards amplified the decline of infectious diseases, leading to a shift towards degenerative diseases [Meslé & Vallin, 2000]. However, an effective anti-tuberculosis drug was not discovered until 1947 [Levy, 2012; Glaziou et al., 2018]. Streptomycin became available only in the 1950s and made a small contribution to TB decline, except in countries with high incidence [Wilson, 2005; Glaziou et al., 2018]. In Belgium, TB vaccination was not as effective as in other countries due to the indifference and skepticism of medical practitioners, and no systematic vaccination was implemented [Wanlin, 2018]. Finally, the decline of TB could also be associated with the natural selection of individuals with stronger resistance to TB, reducing the susceptibility of the population [Davies, 2005].

15Despite the prominent effect of antibiotic treatment, numerous studies emphasized the role of interventions addressing the primary social determinants linked to TB. Addressing issues such as poverty, poor nutrition, housing quality, and, in recent times, HIV, diabetes, drug abuse, and smoking, can play a role in decreasing both the incidence and mortality rates of TB [van Helden, 2003; Levy, 2012; Glaziou et al., et al., 2018].

16This study aims to describe the evolution of TB and identify the variations across time and space between 1889 and 1991 at the municipality level in Belgium. We examine four research hypotheses explaining the spatiotemporal differences in TB mortality between municipalities. We hypothesize that:

• People living in urban centers are more likely to die of TB in relation to higher population density and poorer living conditions.

• There are substantial regional differences in TB mortality between Flanders and Wallonia related to their distinct socio-economic trajectories.

• The presence or absence of health infrastructure such as sanatoria and dispensaries in the municipalities affects TB mortality.

• TB mortality differs according to sex due to the different working conditions of men and women as well as the evolution of social norms and behavioural habits specific to each sex.

Sources and methods

Data and studied area

17Our research on TB mortality focuses on Belgium between 1889 and 1991. Before 1995, Belgium is divided into two regions: Wallonia in the south and Flanders in the north gathering nine provinces: Hainaut, Namur, Luxembourg, Liège, Brabant, Antwerpen, West-Vlaanderen, Oost-Vlaanderen and Limburg (Figure 3).

Table 1 - Overview of the data and data sources

Data	Years	Source
Municipality Level
TB deaths	1889-1890-1891 1905-1909-1910 1929-1930-1931 1947-1948-1949	Le Mouvement de la Population et de l’État civil – Belgian State Archives
	1969-1970-1971 1989-1990-1991	Cause of death registers – Statistics Belgium (STABEL)
Population by age	1890 1910	Published census records – LOKSTAT (0-14; 15-54; 55+)
	1930	Data not available
	1947	Data unusable
	1970 1990	Census – Statistics Belgium (STATBEL) (0-14; 15-24; 25-34; 35-44; 45-54; 55-64; 65-74; 75+)
National level
Mortality scheme National level TB deaths and population by age in Belgium	1890 1910	Le Mouvement de la Population et de l’État civil – Belgian State Archives
	1935-1994	Statistiques des causes de Décès – STATBEL (1963 ;1970 ; 1975, 1989 and 1994)
Deaths from infectious diseases	1851-1900	Exposé de la situation du Royaume publié par le Ministre de l’Intérieur – LOKSTAT.
	1857-1869	Documents statistiques publiés par le Département de l’Intérieur – LOKSTAT.
	1870-1978	Annuaire Statistique de Belgique – LOKSTAT.

(a) TB mortality and overall mortality at the national level

(b) TB mortality and overall mortality at the municipality level

21TB deaths and other demographic information at the municipality level came from” Le Mouvement de la Population et de l’État Civil” and causes of death registries of STATBEL. These archives include TB deaths by sex and for the entire population of Belgian municipalities. They also include the population by municipalities, which was used to compute the TB death ratio. From 1889 to 1947, deaths were recorded at the place of occurrence (de facto), including medical establishments such as sanatoria and dispensaries, and from 1969 to 1991, deaths were recorded at the place of residence (de jure). As data on specific causes of death at the municipality level are only available from 1886 for TB, the analysis period is limited to 100 years, from 1889 to 1991.

(c) Age structure at the municipality level

Hybrid aggregation

23To mitigate the impact of an outlier event during a specific year, limit the time necessary for data input, and account for gaps in data, we focused on six periods, each representing an average of three years and spaced approximately 20 years apart, between 1889 and 1991 (Table 2).

Table 2 - Periods of analysis

Years	Period
1889, 1890 and 1891	1889-1891
1905, 1909 and 1910	1905-1910
1929, 1930 and 1931	1929-1931
1947, 1948 and 1949	1947-1949
1969, 1970 and 1971	1969-1971
1989, 1990 and 1991	1989-1991

24In this study, we aggregated the data into an artificial unit to mitigate the effects of changing municipality boundaries and to smooth fluctuations associated with small populations. This approach follows the methodology proposed by Costa et al. (2021), which involves merging old municipalities into larger spatial units, keeping urban centers with more than 10000 inhabitants in 1890 unchanged, and merging all other towns and villages according to the current administrative divisions of Belgium. This method produces clearer maps and reduces sharp changes related to small numbers while keeping the distinction between urban and rural areas. The data for hybrid, aggregated, municipalities were formed using current boundaries excluding urban centers and computed by summing data from old, unaggregated, municipalities.

Table 3 – Number of municipalities per period (1889 to 1991)

Period	Number of municipalities
1889-1891	2596
1905-1910	2629
1929-1931	2671
1947-1949	2670
1969-1971	2359
1989-1991	581

25We used 627 hybrid municipalities: 87 urban centers (unchanged with old municipalities borders) and 540 spatial units of villages and smaller towns resulting from the aggregation of municipalities specific to each period (Table 3). During the first two study periods, municipalities of the eastern (German-speaking) cantons were not yet part of Belgium.

26From 1929-1931, the addition of the municipalities of the eastern cantons resulted in 638 municipalities: 87 urban centers and 551 other spatial units of villages and smaller towns.

27From 1989 to 1991, the current 581 municipalities were used without distinctions between urban centers and other municipalities. The distinction between urban centers and rural surroundings is in many cases no longer possible.

28Unaggregated municipalities were mostly historical urban centers (Namur, Tournai, Brugge, Gent, Hasselt) and towns of the industrial belt (Mons, Charleroi, and Liège). Towns along the canal between Charleroi and Brussels and from Brussels to the port of Antwerp in Flanders were also kept intact (Figure 4).

Methods

29In this article, we used a three-step procedure.

30At the national level, we computed the percentage of the deadliest infectious diseases (causing more than 10000 deaths in at least one year) estimated from 1851 to 1978. This allowed us to assess the importance of TB among all causes of death. For the same period, we computed the evolution of TB and TB-associated deaths (other names and forms of TB) to estimate TB identification in Belgium.

31The age structure of a population can affect the observed mortality. Age can influence the risk of death biologically (through variations in immunity) and by cumulative exposure to risk factors [Ahmad et al., 2007]. Considering age structure is thus crucial when studying mortality, as age structures can vary by location and over time [Ahmad et al., 2007; Robson et al., 2007]. We calculated the Death Rate (DR_{Age n}) of TB by age groups, for 1890, 1910 and every five years from 1935 onwards to 1994 (n corresponding to a specific age group). We computed the DR_{Age n} (1) for the population as a whole, by age category and by sex.

32At the municipality level, TB mortality was analyzed using the mean Crude Death Rate (CDR) for each period of three years (period x). CDR, a straightforward measure of TB mortality, accounts for the total population, making it useful for initial assessments and comparisons across municipalities due to its simplicity and ease of calculation (2). We used CDR in the indirect standardization and to map the periods 1929-1931 and 1947-1949.

33However, because CDR does not adjust for age or other demographic factors, it is essential for further analysis to standardize the rates to enable more accurate comparisons. Due to the unavailability of age-specific data at the municipal level, direct standardization was not feasible. Therefore, we used a reference TB mortality rate to calculate the Indirect Standardized Death Rate (ISDR) for more appropriate comparisons of TB mortality across populations and periods (Figure 5).

34We first used the DR_Age n as a national reference for TB mortality by age groups to perform indirect standardization. For the periods of 1889 to 1910, we used the DR_Age n of 1890 and 1910. For 1969-1971 and 1989-1991, we used the DR_Age n of 1970 and 1990.

35To compare the observed deaths (TB) by age with a reference population and neutralize the differences due to age structure, we estimated the Expected deaths of TB by age groups (Expected deaths (TB)_Ag n) using population data by age groups for the years 1890, 1910, 1970, and 1990, multiplied by the age-specific DR_Age n (3). It should be noted that age categories available for the whole study period vary, with three age categories for periods 1889-1891 and 1905-1910 (0-14; 15-54; and 55+), and eight for periods of 1969-1971 and 1989-1991 (0-14; 15-24; 25-34; 35-44; 45-54; 55-64; 65-74; 75+).

36Standardized Death Rates (SDR) were estimated for each period and by sex, considering expected deaths and observed deaths related to TB (4). Municipalities with an SDR higher than one have a higher TB mortality than expected, while municipalities with an SDR lower than one have lower mortality than the standard mortality scheme. This indicator accounts for the effect of age structure on TB mortality.

37Finally, we obtained the ISDR by multiplying SDR by the mean CDR for each municipality from 1889 to 1991 (5). The ISDR enables the comparison of TB mortality controlled for the age structure of populations across different periods.

38We illustrate the methodology and indicators with a random municipality, Aaigem (Flanders) (Table 4 & Table 5). The DR_Age n (1) was estimated based on national data, while the Expected deaths (TB)_Age n (3), the mean CDR (2), the SDR (4) and the ISDR (5) were measured based on municipality-specific data.

39Since the SDR for Aaigem from 1889 to 1891 was bigger than one, the TB mortality is higher than the general mortality rate for the year 1890. CDR and ISDR values are not identical, indicating that TB mortality in this municipality is twice the expected values based on age structure standardization for 1890 (Table 5).

40For 1929-1931 and 1947-1949, ISDR could not be estimated, and CDR was used instead for those periods. Both CDR and ISDR were mapped for each period and by sex using hybrid municipality boundaries, except for the last study period of 1989-1991. Spatiotemporal disparities in TB mortality were assessed between regions using an unpaired t-test to compare average TB mortality rates.

41We analyzed the spatiotemporal structure of TB mortality at the municipality level, assessing both long-term and short-term perspectives using the relative differences in TB mortality between the selected periods. Negative values indicate a decrease in the number of deaths over time, while positive values indicate an increase. The long-term perspective (1889-1891 to 1989-1991) aims to identify major trends and temporal fluctuations in TB mortality from the end of the 19^th century, when TB was a leading cause of death, to the end of the 20^th century, when TB prevalence had significantly decreased. The short-term perspective (1929-1931 to 1947-1949) was selected to offer a detailed insight into the relative differences in TB mortality over the period preceding the fastest acceleration of TB decline after the 1950s.

42Finally, we investigated sex disparities in TB mortality over time. Due to the inability to perform age standardization for the periods of 1929-1931 and 1947-1949, the analysis of TB mortality by sex relied on the CDR to enable comparisons across the six periods. Independent paired t-tests were conducted to compare the mean TB mortality between men and women at the municipality level. Furthermore, we mapped TB mortality by sex for periods where the mean difference was statistically significant.

Results

Temporal trends: tuberculosis at the national scale from 1851 to 1978

43There was a gradual reduction in the percentage of infectious diseases among all causes of death in Belgium over the study period (Figure 6). Infectious diseases were a major cause of death around the mid-19^th century, accounting for 40 to 65 % of deaths. The share of deaths from infectious diseases decreased progressively from the late 19^th century, dropping below 10% after 1950. Enteritis, bronchitis, pneumonia, and TB were endemic causes of death in Belgium throughout this period. TB mortality remained high throughout the 19^th century, peaking at 19478 deaths in 1865, before starting to decrease slowly in the 20^th century, from nearly 20% of all deaths in 1851 to almost 0% in 1978. Cholera and smallpox were primarily epidemic diseases. Major outbreaks of cholera occurred in 1854-1855 (7552 deaths), 1859 (5812 deaths), and especially 1866 with 43345 deaths. Two significant outbreaks of smallpox took place in 1865 (5809 deaths) and 1871 (21315 deaths).

44Figure 7 focuses on the causes of death associated with tuberculosis, phthisis, pleurisy, or chronic chest diseases, according to the International Classification of Diseases (ICD - Table B, Appendix). The classification of TB evolved between the mid-19^th century and the late 20^th century, with diagnosis and terminology becoming more accurate over time, enabling better differentiation between respiratory TB, meningitis TB, and other forms of TB.

45Tuberculosis exhibited a consistent trend with other infectious diseases, showing a general decline, from 17730 deaths in 1851 to 296 in 1978. However, this decrease was not uniform over time. Several rebounds in TB mortality incidence can be observed. TB and associated diseases accounted for 15000 and 20000 annual deaths until the early 20^th century. The sharp drop in deaths in 1903 did not result from a specific event but was related to the introduction of the first International Classification of Disease in Belgium, which involved a change in the recording of deaths associated with pulmonary issues. The category “Phthisis and other chronic diseases” was too broad and led to an overestimation of deaths caused by TB.

46Two outbreaks of TB occurred during World War I and World War II. After World War II, the decline in TB mortality accelerated.

47The DR of TB from 1935 to 1990 indicated a major shift in the primary population affected by TB in the 20^th century (Figure 8). The DR of TB showed an overall decrease, except during World War II when TB mortality increased before accelerating its decline post-war. TB mortality was influenced by the age structure of the population, with age groups most affected by TB changing over the study period. From 1935 to 1950, individuals aged 15 to 54 bore the heaviest burden of TB mortality. After 1950, the burden of mortality gradually shifted to the elderly.

48TB mortality was influenced by the age structure of the population, with age groups most affected by TB changing over the study period. From 1935 to 1950, individuals aged 15 to 54 were the main population at risk of TB mortality. After 1950, the burden of mortality gradually shifted to the elderly.

Spatial and temporal trends in tuberculosis mortality at the municipality scale

Spatial trends

49At the regional level, there is an important shift in TB mortality. At the end of the 19^th century, Flanders had higher mortality rates than Wallonia (Figure 9a). This situation reversed after the 1950s, with higher TB mortality in Wallonia (Figures 9e-9f). Comparing the mean CDR at the municipality level between Flanders and Wallonia reveals statistically significant differences during all periods, except for 1905-1910 and 1947-1949 (Student t-test, p-value < 0.05; Table 6). In 1969-1971, higher TB mortality was particularly prevalent in the western part of the industrial belt of Wallonia (Figure 9e).

Table 6 - Mean CDR (‰) between Flanders and Wallonia from 1889 to 1991(*: significative unpaired t-test (p-value<0.05))

Period	Mean CDR Flanders	Mean CDR Wallonia
1889-1891	1.881*	1.664*
1905-1910	0.919	0.955
1929-1931	0.659*	0.568*
1947-1949	0.349	0.410
1969-1971	0.040*	0.073*
1989-1991	0.006*	0.010*

Sources: Algemeen Rijksarchief Brussel, Archief van het Nationaal Instituut voor de Statistiek: beweging van de burgerlijke stand en loop der bevolking, 1841-1976; Causes of death registries, STATBEL; https://statbel.fgov.be/​nl/​themas/​bevolking/​sterfte-en-levensverwachting/​doodsoorzaken#figures

50TB mortality rates at the municipality level gradually declined between 1889 and 1949, from a maximum of 17.755 deaths per 1,000 inhabitants at the end of the 19^th century (Figure 9a) to 2.125 at the end of the 20^th century (Figure 9e). The prevalence of municipalities with the highest values consistently decreased over the study period. The decline in TB mortality accelerated after 1950, with maximum mortality rates dropping from 9.070 in 1969-1971 (Figure 9e) 2.125 in 1989-1991 (Figure 9f).

51No significant spatial trend was detected in TB mortality at the municipality level. ISDR and CDR were heterogeneous across Belgium, although some urban centers exhibited higher values than their suburbs. For instance, the urban centers of Brugge (Flanders) and Tournai (Wallonia) showed a distinct TB mortality compared to their surroundings until 1947-1949 (Figures 9a-9d). Several municipalities had higher TB mortality than their neighbours such as Merksplas in 1905-1910 (Figure 9b) and Colfontaine in 1969-1971 (Figure 9e).

52Some areas exhibit consistently lower tuberculosis mortality rates compared to the rest of the country throughout the entire analysis period, such as the majority of municipalities in the southern part of Luxembourg province and a substantial portion of Namur province. From the 20th century onwards (Figure 9b), municipalities in West Flanders also display low mortality rates. Municipalities exhibiting lower TB mortality could potentially be explained by their rural character and higher environmental quality.

53For each period, we identified the three municipalities with the highest CDR of TB (Table C, Appendix) (Figure 9). Most of these municipalities were in Flanders during the first four periods, and in Wallonia after 1950. Several of these municipalities had collective structures such as health establishments (sanatoria/dispensaries): Duffel (1905-1910), Malle (1929-1931), Damme (1947-1949), and Stoumont (1969-1971 and 1989-1991). Merksplas was hosting a prison (1905-1910 and 1929-1931), and a coal mine was in Blégny (1889-1891). Such infrastructures could explain the high TB mortality observed locally.

54Maps of the SDR for the periods 1889-1891, 1905-1910, 1969-1971, and 1989-1991 distinctly identify regions of both excess and reduced TB mortality (Figure A, Appendix). The elevated mortality rates observed in Flanders at the end of the 19^th century and in the Walloon industrial belt during 1969-1971 are corroborated, with specific hotspots, such as Merksplas during 1905-1910, being clearly delineated.

Temporal trend

55TB mortality declined across Belgium over the past century (Figure 10). However, the rate of decline is spatially heterogeneous, with several municipalities experiencing a sharper reduction in TB mortality. The largest drops in TB deaths occurred in the municipalities of Affligem in West Flanders and Blégny. The decline in Blégny could be associated with the closure of the coal mine in 1982.

56The decrease in TB mortality was not uniform, and some municipalities still experienced increase in TB mortality during the periods 1947-1949 and 1929-1931 (Figure 11). The largest increases occurred in Bonheiden, south of Antwerp, where a sanatorium was established in 1933, and in Baelen, east of Liège, where an iron ore mine was operating. The largest decreases were observed in Merksplas, Malle, and Tessenderlo, which had collective structures that could affect TB mortality. The sanatorium in Tessenderlo was however destructed during war time in 1942, which could explain the decline in TB mortality between 1929 and 1949.

Sex differences

57There are regional disparities among the sexes between Wallonia and Flanders (Table 7). TB mortality for women was significantly higher in Flanders (except 1905-1910) than in Wallonia until 1929-1931 For men TB mortality was significantly higher in Flanders at the end of the 19^th century. After 1949 there was a shift with higher TB mortality for women and men in Wallonia than in Flanders.

Table 7 – Mean CDR (‰) of TB by sex and regions from 1889 to 1991 in Belgium. (*: significative unpaired t-test between regions according to the sex (p-value<0.05))

Period	Mean CDR Men Flanders	Mean CDR Men Wallonia	Mean CDR Women Flanders	Mean CDR Women Wallonia
1889-1891	1.977*	1.609*	1.928*	1.773*
1905-1910	0.872	0.936	0.870*	1.006*
1929-1931	0.524	0.572	0. 604*	0.490 *
1947-1949	0.437	0.451	0.295	0.254
1969-1971	0.055*	0.117*	0.012*	0.025*
1989-1991	0.008*	0.016*	0.003	0.007

Sources: Algemeen Rijksarchief Brussel, Archief van het Nationaal Instituut voor de Statistiek: beweging van de burgerlijke stand en loop der bevolking, 1841-1976; Causes of death registries, STATBEL; https://statbel.fgov.be/​nl/​themas/​bevolking/​sterfte-en-levensverwachting/​doodsoorzaken#figures

58TB mortality by sex is temporally coherent with the national trend but municipalities with the highest TB mortality are not necessarily the same between men and women.

59From 1889 to 1949 (Figures B & C, Appendix), no strong spatial patterns in TB mortality in men and women were observed. A cluster of higher ISDR values for men around Leuven and Maaseik in eastern Flanders is visible in 1889-1891. Additionally, some municipalities in the western part of the industrial belt in 1947-1949 had higher mortality for men. This may be due to mining activity in the Charleroi and Mons Basin, which employed many male workers.

60Stronger spatial differences in the rate of death due to TB in men were found for the period 1969-1971, with a higher incidence of TB for men (Figure 12a) in the industrial belt of Wallonia and along the canal between Charleroi and Brussels. No such patterns were observed for women (Figure 12b). The western part of the industrial belt did not continue to show sex differences in mortality from 1989 to 1991 (Figure 12c), with no significant clusters present.

61The mean CDR at the municipality level between women and men was not statistically different before 1947-1949 (Paired t-test, p-value < 0.05; Table 8 and Figure 12), although the mean was consistently slightly higher for men than women. The mean CDR at the municipality level between women and men is statistically different and higher for men after World War II with the numbers gradually more divergent as time progresses. This is particularly noticeable for the periods 1947-1949 and 1969-1971, where we observe a significant difference: the CDR for men was twice that of women, suggesting increasing sex inequalities.

Table 8 - Mean CDR (‰) of TB by sex from 1889 to 1991 in Belgium. (*: significative paired t-test (p-value<0.05) comparing both groups)

Period	Mean CDR Men	Mean CDR Women
1889-1891	1.792	1.754
1905-1910	0.962	0.912
1929-1931	0.621	0.607
1947-1949	0.479*	0.281*
1969-1971	0.096*	0.011*
1989-1991	0.012*	0.004*

Sources: Algemeen Rijksarchief Brussel, Archief van het Nationaal Instituut voor de Statistiek: beweging van de burgerlijke stand en loop der bevolking, 1841-1976; Causes of death registries, STATBEL; https://statbel.fgov.be/​nl/​themas/​bevolking/​sterfte-en-levensverwachting/​doodsoorzaken#figures

Discussion & conclusion

Main findings

62We conducted a descriptive analysis of the spatiotemporal evolution in TB mortality in Belgium from 1889 to 1991 using “Le Mouvement de la Population et de l’État Civil” data. Our results highlight the variation in TB mortality across space and time at the national and the local scale and considering the historical context, as well as between regions and sexes.

63We observed a general decline in TB mortality over time, disrupted by the two World Wars. The progressive decrease of TB from the 19^th to 20^th century is commonly associated with a combination of factors: improved living conditions (better nutrition, better ventilated and less crowded housing), public health policies to limit the spread of TB and improve patient outcomes, and medical progress including effective treatment against TB [Wanlin, 2018; Mackenbach, 2020]. However, our results highlight spatiotemporal distinctions between municipalities, suggesting an uneven effect of these factors on the decline of TB mortality in Belgium.

64The disruptions observed during both world wars could be explained by the consequences of war: the disorganization of everyday life, destruction, occupation of the healthcare institutions, overcrowding due to civilian and troop movements, food rationing, and a lack of hygiene, leading to increased precarity. Furthermore, the capacity of dispensaries and sanatoria became insufficient for isolating infectious patients [Wanlin, 2018].

65The evolution of TB mortality in Belgium is coherent with the mortality transition for infectious diseases. TB’s decline occurred rather late, starting only in the second half of the 19^th century. This delay was due to industrialization and urban concentration, which worsened living conditions [Meslé et Vallin, 2000]. At the municipality level, high TB mortality in industrial areas occurred well after 1950, suggesting that the mortality transition did not necessarily coincide across all regions in Belgium or for all conditions.

66We also highlighted a shift in the age structure of the populations most affected. In 1935, TB mortality was higher among young adults, with most deaths occurring among the youngest. This trend could be linked to their social behaviour, such as becoming independent, moving out of their homes, and living in communal settings, which could have increased their risk of TB infection [Springett, 1950; Davies, 2005]. This was especially true in urban areas where high-density living conditions facilitated the spread of the disease [Dye & Williams, 2010]. After 1945, TB mortality progressively increased among older age groups, and by 1950, most deaths were observed among the elderly. Older individuals with lower immunity are more susceptible to TB, so that as young people got healthier and older people gradually made a larger fraction of the population, they made off most of the susceptible [Davies, 2005; Dye & Williams, 2010]. This shift could also be attributed to significant changes in the working environment, leading to a reduction in occupational risks to lung health and an increase in the age of the population exposed to such risks [Eggerickx et al., 2020]. This change in the age distribution of TB deaths could also be explained by the effects of the two World Wars and by generational effects, with poorer living conditions for those born at the end of the 19th century and the beginning of the 20^th century. Evaluating age structure is thus crucial when studying TB mortality over the long term.

Research hypotheses

67Our first hypothesis posited that people living in urban centers had a greater chance of dying from TB due to higher population density and poorer living conditions. From 1889 to 1949, TB mortality was notably higher in some urban centers (e.g., Tournai, Bruges) compared to their surroundings. However, TB mortality in Belgium did not differ in most urban centers compared to their surroundings, despite the well-documented association between urbanization and TB. Between 1969 and 1991, higher TB mortality rates were observed in the industrial belt and along the canal between Charleroi and Brussels. These rates might be linked to poor working conditions in factories with inadequate ventilation and lots of dust, which can impact immunity [Heaf, 1947; Davies, 2005; Mackenbach, 2020]. The decline of the steel industry after World War II [Buyst, 2009], removed risk factors but also likely contributed to the impoverishment of these areas. This rise of TB mortality in this area could also be explained by generational and catch-up effects. During the period 1969-1971, the elderly were the most affected by TB. It can be hypothesized that many of these individuals had worked in coal mining and industrial sectors before the Second World War and had been subjected to more challenging working conditions However, it is important to note that the urban status of a municipality alone may not be sufficient to identify factors such as poverty and working conditions, highlighting the necessity of using multivariate models in ongoing analyses to better understand these relationships.

68Our second hypothesis suggested that there were substantial regional differences in TB mortality between Flanders and Wallonia related to their distinct socio-economic evolutions. TB mortality was significantly higher in Flanders from the end of the 19^th century to the mid-20^th century. The pattern changed in the second half of the 20^th century, when higher rates of TB mortality were found in Wallonia, particularly in the industrial belt during the period 1969-1971.

69These regional changes can be associated with the different economic and social conditions in the two regions, as well as their divergent trajectories. Flanders was impoverished until the 1960s, undergoing industrialization after the collapse of the rural linen industry in the 19^th century. This industrialization was accompanied by the ’agricultural invasion’—the arrival of cheap grain from the New World—leading to mass unemployment, except in Antwerp, which benefited from the development of the port and a rail connection to the coal and steel basins in Wallonia. Conversely, Wallonia experienced considerable prosperity due to the growth of industrialization tied to the development of the coal and steel sectors, establishing Wallonia as a leading industrial hub in Europe until the early 20^th century [Buyst, 2009]. However, this progress exacted a heavy toll on the health of the workforce and led to a notable deterioration of socio-economic conditions following the decline of the steel industry [Buyst, 2009]. Differences in health-promoting behaviours and infrastructure, such as a more developed health infrastructure in Wallonia [Eggerickx & Tabutin, 1994], and less prevalent breastfeeding in Flanders may have further contributed to the contrast. In Flanders, less prevalent breast feeding associated with poorer water quality used in infant food may have increased susceptibility to TB by reducing host immunity [Van Rossem et al., 2018].

70In the early 20^th century socio-economic differences between Flanders and Wallonia had diminished, even though the provinces of Hainaut and Liège remained the wealthiest, followed by Antwerp. The mechanization of the textile industry, the development of coal mining in Limburg, and the industries of the ports of Antwerp, Ghent, and Zeebrugge all contributed to this change. At the same time, the first signs of coal mine exhaustion appeared in Wallonia, which also suffered greatly during World War II. Following the war, Wallonia failed to reconstruct its industrial sector robustly. Beyond the industrial core, the provinces of Namur and Luxembourg, which focused on timber harvesting and agriculture, did not significantly contribute to economic expansion. Eventually, from the 1950s onwards, Flanders developed oil refineries in relation to major sea harbours and associated canals, while the coal and steel industry in Wallonia gradually declined [Buyst, 2009]

71Our third hypothesis proposed that the presence or absence of health infrastructure such as sanatoria and dispensaries in the municipalities affected the mortality of TB. In our study, the highest TB mortality rates could be associated with the existence of various collective structures. Some were associated with health care, such as sanatoria and dispensaries where TB patients were diagnosed, isolated, and treated. Others, like Merksplas, housed prisons where conditions were particularly favourable to TB transmission. These municipalities create local outliers in mortality that require separate consideration in the analyses. Currently, data on the location of TB healthcare facilities and the periods they operated is not yet comprehensively available.

72Following the devastation of TB in the 19^th century, public health policies were introduced to limit the spread of TB. During the first half of the 20^th century, private initiatives and public interventions favoured creating and extending clinics, sanatoria, preventoria, dispensaries, and other public and private institutes to help control the disease. In 1919, a comprehensive health plan was implemented, focusing on public health with a dual approach. Firstly, prevention and diagnostics were enhanced through public awareness campaigns, home visits by nurses for surveying and educating, widespread screening, and early detection of diseases. A central bureau was established to increase awareness among the population. Secondly, efforts to curb the spread of TB included the isolation of contagious individuals, treatment within medical facilities, and post-treatment monitoring [Wanlin, 2018].

73Our last hypothesis was that the mortality of TB differed according to sex due to the different working conditions of men and women as well as the evolution of social norms and behavioural habits specific to each sex in Belgium. Our study highlighted differing patterns of TB mortality depending on sex. This became significant in the second half of the 20^th century with higher mortality for men, especially in the western part of the manufacturing belt of Wallonia. This difference could be explained by men’s and women’s specific employment and associated health impacts [Eggerickx, & Tabutin,, 1994; Van Rossem et al., 2018]. Excess mortality in men in Wallonia may be associated with nefarious working conditions in the industrial belt. After World War II, sociocultural and sanitary conditions during pregnancy improved for women, while men continued to be exposed to unhealthy work conditions as well as the increased prevalence of smoking and drinking [ Lönnroth et al., 2009; Eggerickx et al., 2020]. Smoking increases susceptibility to dying from TB [Davies, 2005].

74We did not find any significant sex differences in TB mortality before 1947-1949, even though several studies have shown that women were more susceptible to TB than men during the 19^th and early 20^th centuries. Eggerickx and Tabutin [1994] found an excess of mortality for women in Belgium analyzing mortality data by sex in 1890 for the 41 districts and 9 provinces of the country. Potential explanations for higher female mortality include the position of women in the 19^th century and the prevailing cultural norms regarding men and women. Due to discriminatory practices in nutrition and healthcare access that favoured men, women were more vulnerable to infectious diseases. [Eggerickx & Tabutin, 1994; Davies, 2005; Reid & Garrett, 2018]. Higher susceptibility of young women to TB could also be linked to pregnancy [Davies, 2005]. In Flanders, excess mortality among women was linked to their employment in the textile industry, where they worked in poorly ventilated environments, received low wages, and experienced malnutrition, all factors contributing to increased TB incidence among women [Eggerickx & Tabutin, 1994]. The intensity and nature of labour undertaken by young women in industrial settings, particularly in rural and textile districts, were identified as significant factors contributing to the higher mortality rates observed among females [Devos, 1996].

Limitations and perspectives

75Several factors need to be considered to critically evaluate our results. First, our study analyses a period covering the progressive decline of TB during the 20^th century but not the peak of TB in the mid-19^th century. This choice was made because there are no older data available at municipality level. Therefore, we miss here the rise of TB in Belgium in the early 19^th century. Also, using a three-year average involves smoothing potential TB outbreaks. We opted for a compromise: being able to establish spatial trends using the detailed data of “Le Mouvement de la Population et de l’État Civil” while keeping the digitization – done manually – manageable.

76Handling historical data requires caution due to the potential biases in estimating death rates, particularly over longer periods, resulting from misclassification or misdiagnosis [Zens & Peart, 2003]. In the Belgian municipalities, although deaths from infectious diseases have been recorded since 1851, the information is incomplete, potentially biasing our results [Eggerickx & Tabutin, 1994]. This issue is exacerbated by the historically inconsistent diagnosis of TB [Wanlin, 2018]. Cases of TB were often misclassified as bronchitis or chronic pneumonia, and census-taking in dispensaries and sanatoria was sometimes imprecise, leading to under-recording of TB deaths [Neven & Oris, 1995].

77Another issue with historical data is the distinction between de jure and de facto deaths. Until the 1950s, deaths in “Le Mouvement de la Population et de l’État Civil” were recorded based on the place of occurrence (de facto). In contrast, the population census in Belgium was based on the place of residence (de jure) [Van Rossem et al., 2018]. Consequently, the ISDR and CDR indicators could be underestimated or overestimated depending on the municipality. This distinction may explain the high TB mortality in municipalities with medical facilities such as dispensaries and sanatoria. To derive an unbiased TB mortality indicator, the deaths of non-residents recorded in these facilities should be excluded [Van Rossem et al., 2018]. The change in reporting methods in 1969 could have also biased the results, but the reduced number of TB-related deaths after 1950 minimizes its impact.

78The absence of age structures for the periods 1929-1931 and 1947-1949 affected our analysis of TB mortality. Results for those periods must be interpreted cautiously as they did not include an age standardization, especially considering the shift in the age distribution of deaths as described at the national level. Variations in age structures across time further complicate our interpretations. Moreover, the change in age groups between the 1889-1891/1905-1910 periods and the 1969-1971/1989-1991 periods may affect the accuracy of the DR and therefore the relevance of the ISDR index. The age structure was less precise at the end of the 19^th century, so the comparison between years may be biased.

79Finally, we chose to use a hybrid scale of analysis to map TB mortality in Belgium. Even if this method allows us to reduce the number of municipalities, enhancing the clarity and interpretation of TB mortality, it must be used cautiously, as aggregating data in such a manner is anachronistic and leads to a loss of precision. However, spatial heterogeneity can occur at various scales, such as contrasts between rural and urban areas or different economic sectors, often identifiable with the advantage provided by the hybrid scale without relying on former municipality levels.

80In conclusion, even if historical data must be interpreted cautiously, our analysis provides an interesting novel approach using a hybrid spatial aggregation of municipalities to describe and better understand TB mortality in Belgium. However, there are still missing clues to explain spatiotemporal differences between municipalities belonging to the same districts, provinces, and regions. Therefore, an explanatory analysis considering drivers of TB such as poverty, socio-economic status, urbanized areas, and working conditions but also the absence/presence of healthcare institutes (sanatoria, dispensaries,) may provide further understanding of the trajectory of TB in Belgium from 1889 to 1991.