1 Department of Economic History, Lund University, 5220-07 Lund, Sweden. E-mail: tommy.bengtsson{at}ekh.lu.se
2 Department of Community Medicine, Lund University, Sweden. E-mail: martin.lindstrom{at}smi.mas.lu.se
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Methods Longitudinal demographic and socioeconomic data for individuals and household socioeconomic data from parish registers were combined with local area data on food costs and disease load using a Cox regression framework to analyse the 5580 year age group mortality (number of deaths = 1398).
Results In a previous paper, the disease load experienced during the birth year, measured as the infant mortality rate, was strongly associated with old-age mortality, particularly the outcome of airborne infectious diseases. In the present paper, this impact persisted after controlling for variations in food prices during pregnancy and the birth year, and the disease load on mothers during pregnancy. The impact on mortality in later life stems from both the short-term cycles and the long-term decline in infant mortality. An asymmetrical effect and strong threshold effects were found for the cycles. Years with very high infant mortality, dominated by smallpox and whooping cough, had a strong impact, while modest changes had almost no impact at all. The effects of the disease load during the year of birth were particularly strong for children born during the winter and summer. Children severely exposed to airborne infectious diseases during their birth year had a much higher risk of dying of airborne infectious diseases in their old age.
Conclusions This study suggests that exposure to airborne infectious diseases during the first year of life increases mortality at ages 5580.
Accepted 11 November 2002
The decline in old-age mortality in Sweden and other Western countries started in the mid-19th century, several decades later than the decline in infant mortality.1,2 Old-age mortality is determined by a number of factors, some of which relate to economic and social conditions during the old-age period of life (period determinants). Others are related to the long-term effects of conditions in utero or during early childhood (cohort determinants).
Period determinants of old-age mortality include long-term changes in hygiene and public health,3 and nutrition and standards of living in general,4 but also short-term insecurity in access to food, etc.5,6
Kermack, McKendrick, and McKinley proposed the cohort explanation in 1934. They studied age-specific mortality in England, Wales, Scotland, and Sweden. Their conclusion was that reductions attained at any particular time in the death rates of the various age groups depended primarily on the individuals date of birth, and only secondarily on the current year. The essential effects on health and survival among adults and older people were mainly caused by beneficial effects and improvements achieved in these birth cohorts during childhood several decades earlier.7 This life-course perspective has been given more attention in recent years.8,9 The plausible causal relationship between early life experiences and old-age mortality has been discussed, particularly in relation to intrauterine cellular development and cellular development during early childhood. Robert Fogel has proposed several plausible causal mechanisms that connect malnutrition, whether due to a lack of nutrients or increased demands as a result of disease, in utero and during early life to chronic diseases in later life.10 The propositions regarding the effects of conditions before birth are supported by the work of Barker, among others, who suggested that the preconditions for coronary heart disease, hypertension, stroke, diabetes, and chronic thyroiditis are initiated in utero without becoming clinically manifest until much later in life.11,12 In contrast, Fridlizius suggested that the genesis of disease in later life could be due to exposure to certain infectious diseases, such as smallpox, in the first 5 years after birth, resulting in reduced immunity to other diseases throughout life and, consequently, a higher general risk of other infectious diseases in later life.13 This notion is supported by some studies that have suggested that lower respiratory tract illness in the first year of life is associated with later cough, phlegm, and impaired ventilatory function, independently of smoking and social class. Illness after the first year of life was not associated with any risk, which supports the idea of a critical period of influence for infection.14,15 The hypothesis implies that factors other than nutrition were important early life determinants of mortality in later life, because the outcome of smallpox infection as well as some other important infectious diseases is almost completely unrelated to the nutritional status of the infected individual.16 It also implies that it is not the conditions in utero that are important but rather the situation during infancy and early childhood. Fridlizius does not, however, discuss the possible effect on later health of the nutritional loss due to having experienced a non-nutritional disease or any infectious disease. Thus the cohort hypothesis put forward so far involves whether low nutritional intake or increased nutritional demands due to disease affect mortality in later life and whether the conditions in utero or during the first years of life are of the most importance. The problem in historical analysis has been that these four hypotheses have never been confronted with each other.
In a previous paper, Bengtsson and Lindström investigated these cohort hypotheses for four parishes in southern Sweden, 17661894.17 They used a multivariate Cox regression model on longitudinal demographic data for individuals combined with household socioeconomic data, local area data on food prices, and disease load. While the focus was on cohort hypotheses, the analyses included both variables measuring period short-term economic stress and trends and cohort variables measuring the disease load on mothers during pregnancy and children during the first years of life and access to nutrition in early life. The hypothesis that access to nutrition was of primary importance was not supported by the results. Furthermore, no effects of conditions in utero were found. In contrast, the disease load experienced during the year of birth showed a consistent impact on mortality in later life, particularly on the outcome of airborne infectious diseases during old age. However, the previous study has some important limitations that we deal with in this paper. Firstly, different early-life indicators of stress were analysed in separate multivariate models, one at a time, due to technical constraints. Secondly, effects of long-term trends and short-term fluctuations in disease load during early life, threshold effects, and the patterns of disease load during the first year of life were not specified. Thirdly, possible seasonal effects of the date at birth were not taken into account. This is of potential importance as both diets and the exposure to disease vary seasonally. Mortality peaked in the winter season during the JanuaryApril period and was at its lowest from July to October.18,19 It has also been shown that the season of birth has a strong influence on life expectancy.20 The purpose of this paper is to eliminate these limitations so as to allow us to better understand the causal mechanisms of the previously observed effects of conditions in early life on old-age mortality, in particular on mortality from airborne infectious diseases.
Computer software development now allows us to analyse four or more time-varying community covariates simultaneously, which means that we can estimate the net effects of the disease load during the first year of life (infant mortality rate) and the disease load on the mother during pregnancy (measured indirectly by the crude death rates among adults aged 2050 years), and the food prices during pregnancy and the first year of life will also be analysed as plausible determinants of old-age mortality. We are particularly interested in whether the large annual variations of food prices influence mortality in laterlife as one would expect it to reflect changes in nutrition. Variation in food prices also has a strong influence on current mortality. The effects of trends in infant mortality as opposed to fluctuations (cycles) from year to year on old-age mortality will be investigated by decomposing the infant mortality rate into two components using a Hodrick Prescott filter with a filtering factor of 100 (for a discussion of and references to this method, see ref. 21). The same filter is used to divide the crude death rates among adults aged 2050 years into trend and cycle components. Furthermore, the cycle component of the infant mortality rate is divided into five categories to enable us to estimate potential threshold effects, and thus some years with a very high disease load have a proportionately stronger effect on old age mortality than other years. Furthermore, causes of death are identified for such years. The problem of seasonality is dealt with by including a season dummy (winter, spring, summer, autumn) as a covariate in the multivariate model, and also by analysing the season effect separately in four independent multivariate models. In addition, effects of sex, birth year, birthplace, present place of residence, present socioeconomic status (SES) (freeholders/crown tenants, tenants on nobility-owned land, semi-landless/crofters, landless), and current access to food (rye prices) are included in the model, based on both the variables available in this data collection17 and in previous literature.19
![]() |
Material and Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
Demographic data from Sweden dating from 1749 and onwards are generally highly valid and reliable both at the aggregated and individual level. The data are of much higher quality than, for example, those for England22Secondly, we have both census type information and records about migration and thus know the population at risk. Thirdly, we have occupational information and data on farm types and sizes. Fourthly, we have cause-of-death records. The quality of the death records in the parishes of this study is high, as indicated by the ratio of male to female births, the proportion of stillbirths, and the proportion of deaths during the first month.23 The validity and reliability of historical data on causes of death in Sweden have been thoroughly discussed elsewhere.17 In summary, changes in the nomenclature of medical diagnoses (in 1774, 1802, 1811, 1821, 1831, 1873, and 1891) and the succession of individual clergymen in the parishes seem to have had no effect on the validity and reliability of the medical diagnoses made by the clergymen.24 Furthermore, some diseases, especially infectious diseases such as smallpox, were identified as the same diagnostic entities (based on distinct symptoms) and denoted by the same names as today.17 The aetiology of infectious diseases, the most common cause of death during the whole study period, was not known; proper diagnosis was impossible, except for symptomatic diagnoses, until the end of the 19th century.25 The system remained basically nosological during the whole period.24
The social structure of the agricultural sector is often difficult to analyse as differences in wealth between the various categories of farmers and occupations are unclear and subject to change with the passage of time. Data from land registers on types of tenure must therefore be combined with information from poll tax records concerning farm size in order to arrive at a better understanding of the social structure. From these sources we can conclude that the nobility was a rather small group and it has therefore been excluded from our sample. The peasants were divided into two categories: freeholders, tenants on crown land, and tenants on church land constituting the first group, while tenants on nobility-owned land constitute the second one. We only include peasants with farms larger than 1/16 mantal in these two categories as it has been argued that peasants with smaller farms could not support themselves. A mantal was not a measure of the actual size of the farm but a tax-assessment unit based on potential production. A third group, which we label semi-landless, includes farmers with land smaller than 1/16 mantal and crofters. The fourth group is the landless workers.
In estimating the parameters of the models, we use event-history analysis with time-varying external covariates, which makes it possible to run regressions on the change of life status, i.e. dying or giving birth to a child, measuring the effects of different explanatory variables (or covariates) on the hazard of the event. More specifically, we use the Cox proportional hazards model, which does not require specification of the underlying hazard function. The main interest in this case is to estimate the impact of different covariates on the hazard of death. The aggregated indicator of the food prices is included in the regressions as a communal, or external, covariate.26,27 This means that the aggregate economic information is used as a time-varying covariate common to all individuals in the risk set at each point in calendar time.27 Aggregated indicatorsthe infant mortality rate (Figures 2 and 3
), the crude death rate at ages 2050 years (Figure 4
), and the food prices (Figure 5
)are also used as fixed community covariates. The value of the community covariate is then shared by all individuals with the same birth year or with the same year of conception, depending on which indicator it is. The infant mortality rate during the year at birth is shared by all individuals with the same birth year to indicate the disease load on infants. The crude death rate at ages 2050 during the year of conception is shared by all individuals with the same year of conception to indicate the disease load on the mothers during the pregnancy. Food prices are likewise fixed to reflect the access to food during both the fetal stage and the birth year. We use the logarithm of local rye prices as an indication of the availability of food because grain dominated the diet and rye was the major crop. This indicator has been used in many previous studies and has been shown to have high validity.27 The software program used is called MLIFE and was developed by Professor Göran Broström of Umeå University and Lund University. It is a GNI licence program and has specific features to facilitate the use of time-varying and fixed community covariates.28
|
|
|
|
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
The effect of the disease load during the year of birth is strong and highly significant. A one-unit change in the disease load increased the mortality risk by approximately 50%, from an average of 0.042 to about 0.06. The effects of birth year and SES are also significant. Being born one year later is associated with a 0.3% lower death risk. The finding regarding SES seems at first paradoxical as the semi-landlesscrofters and people with very small farmshave lower death risks than the better-off freeholders and tenants. They have, on the other hand, lower mortality than the landless, which is what one would expect from an economic point of view. A possible explanation is that the size and composition of the semi-landless group changes over time, partly as a result of the agricultural reforms at the beginning of the 19th century. We have therefore estimated the model similarly to the one shown in Table 1 with controls for various time periods. While we are not reporting these details in this paper, we did find that the results are stable throughout the 19th century. Thus, the lower mortality for the semi-landless group relative to the better-off groups is not a result of its increase in size in a period with lower mortality but is genuine. A possible explanation of the lower mortality among the semi-landless is the fact that they often lived in cottages outside the villages and that they were therefore less exposed to infectious diseases. Furthermore, we did not find effects of gender, the time of the year when the birth occurred, which parish the elderly lived in, and whether people were living in their native parish or not.
In as much as the disease load during the birth year had such a strong effect on mortality at later ages, we have tried to specify it better. The infant mortality rate between 1686 and 1839, when the elderly people were born, fluctuated substantially. Figure 2 shows that the infant mortality rate changes annually, but also a secular trend. The estimated effect shown in Table 1
could either be a result of the short-term changes or the long-term ones. This infant mortality rate was therefore divided into two components using a Hodrick Prescott filter with 100 as the filter term, which resulted in the short-term component called cycle, shown in Figure 3
, and the long-term component shown as the broken line in Figure 2
. For a discussion of the method in this context see ref. 21. Table 2
shows that not only the long-term development of the infant mortality rate but also annual changes influenced mortality in later life.
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
The results agree to some extent with those of Fridlizius,13 who proposed that adult and old-age mortality was mainly affected by the disease load during the first 5 years of life. Fridlizius also suggested that exposure to especially infectious diseases, e.g. smallpox, but also, to some extent, other infectious diseases during the first years of life would affect survival during adulthood through an irreversible damaging effect on the immunological system. There is, however, no medical evidence to account for such a causal mechanism. Furthermore, our findings regarding the long-term component of the infant mortality influence on old-age mortality are in agreement with those of Fridlizius, since he filters out the short-term components by estimating the effects of the disease load during the first 5 years of life. The short-term fluctuations in infant mortality are also of great importance during the first year of life, as shown in this paper.
The influence of the short-term variations in infant mortality on old-age mortality was mostly due to peaks in infant mortality caused by different infectious diseases, particularly by smallpox and whooping cough. These infectious diseases are so virulent that they must have penetrated the entire area. Smallpox is one example. It was mainly a childhood disease during the 18th century.29,30 Approximately 95% of all deaths due to smallpox in Sweden during the 18th century occurred under age 10.29,31 The 18th century patterns of total childhood dominance in smallpox mortality indicate that most of the adult population during this period had already been exposed to smallpox as children but had survived. Those cohorts exposed to such infectious diseases during infancy may also be much more susceptible to high morbidity and mortality rates even in old age as an effect of this exposure than cohorts exposed to epidemics of smallpox and other infectious diseases later in childhood. The causal biological mechanisms in early life that might explain these significant associations are only partly understood, although long-term effects on morbidity and mortality seem to have some support in the literature.32 The pathways that have been discussed mostly concern infectious diseases of the respiratory tract. Respiratory infections, atopy, reversible airway obstruction, chronic mucus hypersecretion, and irreversible airflow obstruction are interconnected by a complex web of associations and putative causal relationships.8 Respiratory infectious diseases in infancy (ages 01 year) have been suggested to be one cause of a chronic wheezing tendency,33,34 chronic cough and phlegm,33,35 irreversible impaired ventilatory function,33,36 and the related mortality.37 Two historical cohort studies in England suggest that respiratory infectious diseases in the first year of life are associated with later cough, phlegm, and impaired ventilatory function, independently of smoking habits and social class. Infectious respiratory diseases after the first year of life did not seem to be a risk, which supports the idea of a critical period of influence for infection. Mortality from chronic respiratory diseases was also associated with early bronchitis and pneumonia.14,15 These studies have been interpreted as evidence of persistent lung damage from respiratory infectious diseases during the first year of life.8
This study deals with exposure to smallpox and other airborne infectious diseases during the latter part of the 18th century and whether this exposure during infancy affected mortality later in life. The results and conclusions of this study could most probably be generalized to other geographical areas and countries than Sweden, and also to the 20th century. As smallpox was eradicated in human populations as late as in the late 1970s, long-term effects of smallpox on morbidity and mortality are still highly interesting and will remain so for many years to come.33
Several studies on more recent 20th century data series show that there is a strong association between low SES and both mortality and general practice consultations for adult respiratory disease.38 Although most of these socioeconomic differences are probably due to socioeconomic differences in smoking behaviour, independent associations between SES and symptoms of mucus hypersecretion39,40 and ventilatory function41 have been demonstrated. A significant socioeconomic gradient in bronchitis mortality was observed before the occurrence of socioeconomic differences in smoking behaviour.42 However, the present study on an older historical data series does not demonstrate the same results. We find no effects of rye prices experienced in early life, an indicator of economic hardship, the consequences of which varied greatly in different socioeconomic strata, on old-age mortality. In contrast, highly contagious infectious diseases such as smallpox, which affected different socioeconomic strata similarly, had this effect on old-age mortality.
This study suggests that airborne infectious diseases are important for the causal mechanisms linking infant mortality to old-age (5580 years) mortality. The variations in infant mortality that affected old-age mortality were mainly caused by both trends and short-term cycles in infant mortality from airborne infectious diseases. Furthermore, old-age mortality was mainly affected by these cohort mechanisms through an increase in old-age mortality from infectious diseases.17 The combination of these two patterns may seem paradoxical, as exposure to airborne infectious diseases during the first year of life might be expected to result in lesser susceptibility to such diseases in later life because of stronger immunity. However, the results may instead indicate a more general vulnerability throughout life caused by weakening effects in other aspects of the airborne infectious diseases of infancy. In fact, Bengtsson has also shown that the negative impact is not found only at ages 5580, but also at ages 2055.43
![]() |
Acknowledgments |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 Perrenoud A. The mortality decline in a long term perspective. In: Bengtsson T, Fridlizius G, Ohlsson R (eds). Preindustrial Population Change. Lund: Studentlitteratur, 1984, pp. 4169.
3 Easterlin RA. How beneficient is the market? A look at the modern history of mortality. European Review of Economic History 1999;3: 25794.
4 McKeown T. The Modern Rise of Population. London: Edward Arnold, 1976.
5 Bengtsson T, Ohlsson R. Levnadsstandard och mortalitet i Sverige 17501860. Meddelande från ekonomisk-historiska institutionen 37. Lund University, 1984.
6 Bengtsson T, Ohlsson R. Age-specific mortality and long-term changes in the standard of living: Sweden, 17511859. Eur Journal of Population 1985;1:30926.[ISI]
7 Kermack WO, McKendrick AG, McKinley PL. Death rates in Great Britain and Sweden: some regularities and their significance. Lancet 1934;226:698703.[CrossRef]
8 Kuh D, Ben-Shlomo Y. A Life Course Approach to Chronic Disease Epidemiology. Oxford: Oxford University Press, 1997.
9 Preston SH, Hill ME, Drevenstedt GL. Childhood conditions that predict survival to advanced ages among African-Americans. Soc Sci Med 1998;47:123146.[CrossRef][ISI][Medline]
10 Fogel RW. The relevance of Malthus for the study of mortality today: Long-run Influences on health, mortality, labour-force participation and population growth. In: Lindvall K, Landberg H (eds). Population, Economic Development, and the Environment: The Making of Our Common Future. Oxford and New York: Oxford University Press, 1994, pp. 23184.
11 Barker DJP. Mothers, Babies, and Disease in Later Life. London: British Medical Journal Publishing Group, 1994.
12 Barker DJP. Fetal origins of coronary heart disease. BMJ 1995;311:17174.
13 Fridlizius G. The deformation of cohorts: nineteenth century mortality in a generational perspective. Scandinavian Economic History Review 1989;37:317.
14 Barker DJP, Godfrey KM, Fall C, Osmond C, Winter PD, Shaheen SO. Relation of birthweight and childhood respiratory infection to adult lung function and death from chronic obstructive lung disease. BMJ 1991;303:67175.[ISI][Medline]
15 Shaheen SO, Barker DJP, Shiell AW, Crocker FJ, Wield GA, Holgate ST. The relationship between pneumonia in early childhood and impaired lung function in late adult life. Am Rev Respir Dis Crit Care 1994;149:61619.
16 Rotberg RI, Rabbs TK. Hunger and History: The Impact of Changing Food Patterns and Society. Cambridge, London: Cambridge University Press, 1985.
17 Bengtsson T, Lindström M. Childhood misery and disease in later life: the effects on mortality in old age of hazards experienced in early life, southern Sweden, 17601894. Popul Stud 2000;54:26377.[ISI]
18 Berg FT. Årstidernas inflytelse på dödligheten. Statistisk Tidskrift 1879;7:87121.
19 Wrigley EA, Schofield R. The Population History of England 15411871: A Reconstruction. Cambridge: Cambridge University Press, 1981.
20 Gavrilov LA, Gavrilova NS. Season of birth and human longevity. Journal of Anti-aging Medicine 1999;2:36566.
21 Bengtsson T, Dribe M. New evidence on the standard of living in Sweden during the 18th and 19th centuries: long-term development of the demographic response to short-term economic stress among landless in Western Scania. In: Allen RC, Bengtsson T, Dribe M (eds). New Evidence of Standards of Living in Preindustrial Europe and Asia. (Forthcoming).
22 Wrigley EA, Davis RS, Oeppen JE, Schofield RS. English Population History from Family Reconstitution 15801837. Cambridge: Cambridge University Press, 1997.
23 Bengtsson T, Dribe M. Fertility Response to Short-Term Economic Stress: Deliberate Control or Reduced Fecundability? Lund: Department of Economic History, 2002. (Unpublished).
24 Bengtsson T. Mortality and causes of death in Västanfors Parish, Sweden, 17001925. In: Brändström A, Tedebrand LG (eds). Society, Health, and Population During the Demographic Transition. Umeå: Umeå University, 1988.
25 Ackerknecht EH. A Short History of Medicine. Baltimore and Boston: Johns Hopkins University Press, 1982.
26 Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. New York, Chichester, Brisbane, Toronto: John Wiley and Sons, 1980.
27 Bengtsson T. Combined time-series and life-event analysis: the impact of economic fluctuations and air temperature on adult mortality by sex and occupation in a Swedish Mining Parish, 17571850. In: Reher DS, Schofield R (eds). Old and New Methods in Historical Demography. Oxford: Oxford University Press, 1993.
28 Broström G. MLife. Reference Manual. Umeå: Umeå University, 2000.
29 Sköld P. The Two Faces of Smallpox: A Disease and its Prevention in Eighteenth and Nineteenth Century Sweden. 1996. Umeå: Umeå University. The Demographic Data Base.
30 Smith JR. The Speckled Monster: Smallpox in England, 16701970, with Particular Reference to Essex. Chelmsford: Essex Record Office, 1987.
31 Riley JC. Rising Life Expectancy: A Global History. Cambridge: Cambridge University Press, 2001.
32 Razzell P. The Conquest of Smallpox. London: Caliban Books, 1977.
33 Samet JM, Tager IB, Speizer FE. The relationship between respiratory illness in childhood and chronic air-flow obstruction in adulthood. Am Rev Respir Dis 1983;127:50823.[ISI][Medline]
34 McConnochie KM, Roghmann KJ. Bronchiolitis as a possible cause of wheezing in childhood: new evidence. Pediatrics 1984;74:110.[Abstract]
35 Colley JRT, Douglas JWB, Reid DD. Respiratory disease in young adults: influence of early childhood lower respiratory tract illness, social class, air pollution, and smoking. BMJ 1973;3:19598.[Medline]
36 Britten N, Davies JMC, Colley JRT. Early respiratory experience and subsequent cough and peak expiratory flow rate in 36 year old men and women. BMJ 1987;294:131720.[ISI][Medline]
37 Barker DJP, Osmond C. Childhood respiratory infection and adult chronic brochitis in England and Wales. BMJ 1986;293:127175.[ISI][Medline]
38 Strachan DP, Epidemiology: a British perspective. In: Calverley P, Pride N (eds). Chronic Obstructive Pulmonary Disease. London: Chapman and Hall, 1995, pp. 4768.
39 Dean G, Lee PN, Todd GF, Wicken AJ, Sparks DN. Factors related to respiratory and cardiovascular symptoms in the United Kingdom. J Epidemiol Community Health 1978;32:8696.[Abstract]
40 Respiratory Diseases Study Group of the College of General Practitioners. Chronic bronchitis in Great Britain. BMJ 1961;ii:97378.
41 Cox BD. Blood pressure and respiratory function. In: Cox BD, Blaxter M, Buckle ALJ et al. The Health and Lifestyle Survey. Preliminary Report of a Nationwide Survey of the Physical and Mental Health, Attitudes and Lifestyle of a Random Sample of 9003 British Adults. London: Health Promotion Research Trust, 1987, pp. 1733.
42 Registrar-General for England and Wales. Decennial Supplement for 1921. Occupational Mortality. London: HMSO, 1931.
43 Bengtsson T. Adult mortality in rural Sweden 17601895. Eurasian Project of Population and Family History. Working Paper Series. Kyoto: JRCJS, 1997;8.