Understanding the Twentieth-Century Decline in Chronic Conditions among Older Men Dora L ...

21 Pages·2008·0.55 MB·English

Understandingthe Twentieth-CenturyDecline inChronicConditionsamongOlder
Men
DoraL.Costa
Demography,Vol.37,No.1.(Feb.,2000),pp.53-72.
StableURL:
http://links.jstor.org/sici?sici=0070-3370%28200002%2937%3A1%3C53%3AUTTDIC%3E2.0.CO%3B2-S
DemographyiscurrentlypublishedbyPopulationAssociationofAmerica.
YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat
http://www.jstor.org/about/terms.html.JSTOR'sTermsandConditionsofUseprovides,inpart,thatunlessyouhaveobtained
priorpermission,youmaynotdownloadanentireissueofajournalormultiplecopiesofarticles,andyoumayusecontentin
theJSTORarchiveonlyforyourpersonal,non-commercialuse.
Pleasecontactthepublisherregardinganyfurtheruseofthiswork.Publishercontactinformationmaybeobtainedat
http://www.jstor.org/journals/paa.html.
EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprinted
pageofsuchtransmission.
TheJSTORArchiveisatrusteddigitalrepositoryprovidingforlong-termpreservationandaccesstoleadingacademic
journalsandscholarlyliteraturefromaroundtheworld.TheArchiveissupportedbylibraries,scholarlysocieties,publishers,
andfoundations.ItisaninitiativeofJSTOR,anot-for-profitorganizationwithamissiontohelpthescholarlycommunitytake
advantageofadvancesintechnology.FormoreinformationregardingJSTOR,[email protected].
http://www.jstor.org
TueApr120:12:152008
UNDERSTANDING THE TWENTIETH-CENTURY DECLINE IN CHRONIC

CONDITIONS AMONG OLDER MEN*
DORA L. COSTA
I argue that the shiftfrom manual to white-collarjobs and re- medical procedures may improve health, it may do so only
duced exposure to infectious disease were important determinants for those who would have survived in any event while keep-
ofdeclines in chronic disease rates among older men from the early ing alive, in a poor state of health, those who would have
1900s to the 1970s and 1980s. The average decline in chronic respi-
died. The latter effect may even offset or overwhelm the
ratory problems, valvular heart disease, arteriosclerosis, and joint
former. Lifestyle changes such as cessation of smoking or
and back problems was about 66%. Occupationnl shifts accounted
the change to a low-fatdiet may account for some of the re-
for 29% of the decline; the decreased prevalence of infectious dis-
cent improvements; however, the increase in smoking from
ease accounted for 18%; the remainder are unexplained. The dura-
tion of chronic conditions has remained unchanged since the early the late 1900s through the first half of the twentieth century
1900s, but when disability is measured by difJiculty in walking, men probably contributed to increased morbidity rates.
with chronic conditions are less disabled now than they were in the Other factors that could explain the long-run decline in-
past. clude reductions in occupational injuries, a decline in the use
of salt and smoke as food preservatives, increases in nutri-
tional intake during the growing years, and declines in child-
T h eh ealth of the elderly in the United States has improved hood infectious diseases. As pointed out by Preston (1993),
remarkably during the twentieth century. Costa (1998) finds those who reached age 90 in 1992 were born in 1902, when
that in the 1930s, rates of blindness among those aged 65 to life expectancy at birth was only 49 years and the burden of
84 were four times as high as in the 1990s. Fogel and Costa infectious disease was exceptionally heavy. An ongoing se-
(1997) find that men older than 64 in 1910 were much more ries of studies by Barker and his colleagues (Barker 1992,
likely to suffer from heart, respiratory, musculoskeletal, and 1994) links many of the degenerative conditions of old age
digestive disorders than their counterparts today. Costa and to exposure to infectious disease, malnutrition, and other
Steckel (1 997) document that weight adjusted for height for types of biomedical and socioeconomic stress in utero and in
older men has increased from dangerously low levels at the the first year of life. The authors argue that the focus of re-
turn of the century. search should be shifted from adult risk factors such as obe-
This improvement in health has not necessarily been sity, smoking, and lack of exercise to early childhood fac-
continuous. Costa and Steckel (1997) report that chronic dis- tors. These studies, however, have been criticized on several
ease rates at older ages were higher for cohorts born between grounds, including the confounding of stress at birth and in
1840 and 1849 than for those born between 1820 and 1829. early infancy with stress at young and late adult ages because
Cycles may be present in more recent years as well. Although of the persistence of biomedical and socioeconomic risk fac-
clinicians' reports document steady improvements in health tors (Paneth and Susser 1995).
since the 1970s (Waidmann, Bound, and Schoenbaum 1995), In this paper I investigate the risk factors for heart dis-
self-reported health declined during the 1970s but increased ease, respiratory disorders, and musculoskeletal problems in
during the 1980s (Chirikos 1986; Colvez and Blanchet 1981; the first cohort to reach age 65 in the twentieth century by
Crimmins 1990; Cutler and Richardson 1997; Manton, using a longitudinal data set on Union Army veterans. In
Corder, and Stallard 1997; Poterba and Summers 1987; many ways the data are ideally suited to a study of the impact
Verbrugge 1984). Nonetheless, the overall trend appears to of environmental stress throughout life. Because infant mor-
be one of improving health. tality varied greatly across U.S. counties during this cohort's
Several environmental factors may account for the long- early childhood, this variation allows me to investigate the
term decline in chronic disease rates among older Americans impact of the disease environment in the county of enlist-
and for cycles in health. Although the increased efficacy of ment. These men were exposed to a variety of stresses, par-
ticularly from infectious diseases, while in the army as young
adults; therefore I can investigate the impact of exposure to
'Dora L. Costa, Massachusetts Institute of Technology, Department of
disease at young adult ages on later life outcomes. I can also
Economics, E52-274C, 50 Memorial Drive, Cambridge, MA 02142 and
investigate the impact of socioeconomic factors at young
National Bureau of Economic Research; E-mail [email protected]. I gratefully
acknowledge the support of NIH Grants AG12658 and AG10120 and of the adult and older ages on chronic disease rates at older ages.
Russell Sage Foundation through their Visiting Scholar program. I have ben- Because medical care in the past was ineffective at best,
efited from the comments of Robert Fogel, Matthew Kahn, Robert Mitten- I can examine whether there is any evidence that in the past,
dorf, Louis Nguyen, James Poterba, Sven Wilson, two anonymous referees,
marginal survivors who might have been kept alive today
and participants at the 1998 NBERIDAE Summer Institute and at UC
Berkeley's brown bag seminar in demography. died, and that men with specific chronic conditions were less
Demography, Volume 37-Number 1, February 2000: 53-72
54 DEMOGRAPHY,VOLUME 37-NUMBER 1, FEBRUARY 2000
disabled than their counterparts today. Riley (1989) argues and the three major chronic disease categories examined in
that although incidence rates for specific conditions may this paper: heart, respiratory, and musculoskeletal disorders.
have declined, conditional on having a disease, those with These are reviewed briefly in this section (for more thorough
the disease may be more disabled now than they were 100 discussions see Elo and Preston 1992; Stein 1990).
years ago. In contrast, Cutler and Richardson (1997) find that The link between infectious disease and rheumatic heart
since the late 1970s, self-reported health of those with spe- disease, a heart disorder common in most developing coun-
cific chronic conditions has improved. tries and in the United States at the beginning of the century,
The findings will help us understand why the health of is particularly well established. Acute rheumatic fever in-
different cohorts has been changing in the United States, and volves the joints, the heart, and subcutaneous tissue, and re-
therefore will improve our forecasts of health trends. These sults in damage to the heart valves. Today this disease oc-
forecasts will permit us to assess policies postponing the nor- curs most frequently in children age 6-1 5 following group A
mal age of retirement for Social Security and to reduce the streptococcal upper respiratory tract infection. In the nine-
fiscal deficits in Medicare and Medicaid. The findings also teenth century, however, it occurred more frequently among
will help us formulate public policies for countries undergo- older persons: 25% of first attacks occurred between ages 20
ing an epidemiological transition. The infectious diseases and 25, and 49% between ages 30 and 39 (Kiple 1993:971).
that were prevalent in the United States in the nineteenth cen- Individuals who have had acute rheumatic fever become par-
tury are still as common, and as deadly, in many developing ticularly likely to experience recurrences of attacks follow-
countries today. Controlling infectious disease therefore may ing the streptococcal infection. Crowding associated with
be the appropriate health response to lower chronic disease poor living conditions is the major predisposing condition
rates. Furthermore, if ill health in older populations arises for the increased risk of streptococcal infection and acute
from their experiences in early childhood or as young adults, rheumatic fever.
those years may the appropriate time at which to target health Other infectious diseases that can affect cardiac function-
interventions. ing include late-stage syphilis, measles, and typhoid fever. In
The findings also have implications for theories of the a sample of typhoid patients, electrocardiogram assessments
epidemiological transition (Fries 1980; Omran 1971). If in- showed that 12% had cardiac involvement (Khosla 1981 ).
fectious diseases contribute to chronic disease, then the link Electrocardiogram assessments of Nigerian children with
between infectious and chronic diseases may be much closer measles showed that 35% had some abnormality (Olowu and
than previously thought. Although lifestyle changes and im- Taiwo 1990).
proved medical care allow for the compression of morbidity There is also increasing evidence of a link between in-
"
in these theories, chronic diseases are generally viewed as fectious disease and atherosclerosis. Several studies have de-
part of the natural process of aging and thus as limits to life tected chlamydia pneumoniae, a bacterium that causes acute
expectancy; this limit is now generally set at age 85 (Fries upper and lower respiratory infections, in atherosclerotic le-
1989). Recent research, however, finds little evidence for a sions from coronary and carotid arteries, in abdominal aortic
genetically determined life span; at the very least, it suggests aneurysms, and in sclerotic and aortic valves. Infections are
that such a limit would be well above age 85 (McGue, known to influence lipid metabolism, and may play some
Vaupel, and Holm 1993; Wilmoth and Lundstrom 1996). Pre- role in the atherosclerotic process itself by inducing damage
dictions by Manton, Stallard, and Tolley (1991) suggest that and inflammation in the vascular endothelium in the pres-
if people changed their health behaviors in optimal ways, ence of hypercholesterolemia (see reviews by Lindholt et al.
then life expectancies might reach the 90s or the 100s. Re- 1999; Valtonen 1991 ;Wong, Gallagher, and Ward 1999). In
ductions in other risk factors for chronic disease, such as ex- older work, Buck and Simpson (1 982) found a high correla-
posure to infectious disease, therefore may play a role in ex- tion, in the United States, between atherosclerotic heart dis-
tending life expectancy and may explain the declines in ease at older ages and diarrheal deaths from birth to age 20.
older-age mortality observed since the 1970s. They speculated that infection facilitates the production of
I begin by discussing the relationship between early-life autoimmune complexes that promote the later development
conditions and chronic disease. Next I describe the data and of atherosclerotic lesions. The fatty streaks that are the pre-
present some results on trends in chronic conditions. Then I cursors of these lesions may already be present in children
examine the relationship between early-life conditions and by age 10.
the probability of chronic disease, how life expectancy for Infectious disease is not the only early-life factor that
individuals with particular diseases has changed over the may affect heart functioning. Barker (1992, 1994) finds in
course of the century, and whether there is any evidence that longitudinal data that undernutrition in utero and in the first
persons with a given chronic condition are more disabled year of life, as proxied by anthropometric measures of the
now than they were in the past. I end with a discussion of the child, increases risk of coronary heart disease and stroke.
implications of the findings. Chronic obstructive lung disease is another important
cause of adult morbidity and mortality in developing coun-
EARLY-LIFE CONDITIONS AND CHRONIC DISEASE tries. Researchers have hypothesized that lower respiratory
The medical and epidemiological literature provides many tract infections in childhood may aid in the development of
examples of the possible links between early-life conditions chronic obstructive lung disease in later life, particularly dur-
THE TWENTIETH-CENTURY DECLINE IN CHRONIC CONDITIONS
ing early infancy, when children's lungs are undergoing de- young ages is correlated positively with genetic susceptibil-
velopmental change. In a longitudinal study of British men, ity to developing chronic disease at older ages, then, because
Barker (1992, 1994) found that bronchitis, pneumonia, and fewer genetically "frail" individuals survive to old age, the
whooping cough before age 5 are associated with reduced morbidity rate of such a cohort may be lower than in a co-
mean forced expiratory volume in one second (FEV1) at ages hort in which more genetically frail individuals survive. In
59 to 70. Follow-ups of disadvantaged children in South Af- addition, cohorts who survive infectious disease may acquire
rica revealed that abnormalities in lung function persisted partial or complete immunity and therefore may have lower
years after the contraction of pneumonia (Wesley 1991). mortality rates.
In this group of children and in most developing coun-
tries, measles is a major cause of acute lower respiratory in- DATA
. .
fection (Markowitz and Nieburg 1991). Autopsies and radio- The data used in this DaDer are drawn from the records of the
graphs show that measles bronchopneumonia results in bron- Union Army pension program. (See the data appendix for
chiolar obstruction, distension of airways, and a thickening availability of data.) This pension program was the most
of the peribronchial walls (Jean et al. 1981). Among Civil widespread form of assistance to the elderly before Social
War soldiers, measles was followed by complications such Security: It covered 85% of all Union Army veterans by 1900
as chronic bronchitis, pneumonia, pleurisy, chronic diarrhea, and 90% by 1910. The program benefited an estimated 25%
and general debility (Cliff, Haggett, and Smallman-Raynor of the population older than 64, whether as a couple consist-
1993:105). ing of the former soldier and his wife, the single or widowed
Previous tuberculous disease also may lead to respira- veteran, or the veteran's widow (Costa 1998: 160). The pro-
tory abnormalities. One-third of typhoid fever patients also gram began in 1862, when Congress established the basic
suffer from cough; this suggests that typhoid fever also may system of pension laws, known as the General Law pension
be a likely candidate for respiratory distress in later life. system, to provide pensions to both regular and volunteer re-
Other viral infections, such as pulmonary syphilis, may play cruits who were severely disabled as a direct result of mili-
a role as well. tary service. (For a history of the Union Army pension pro-
Lung diseases and respiratory symptoms resulting from gram, see Costa 1998: 197-212.)
occupational exposure to dust, fumes, or gases include The Union Army pension program became a universal
asthma. chronic bronchitis. chronic air flow limitation. and disability and old-age pension program for veterans with the
tuberculosis: Workers in mining, steel foundries, tool grind- passage of the Act of June 27, 1890, which specified that any
ing, glassmaking, metal casting, and stone polishing are par- disability entitled the veteran to a pension. Even though old
ticularly prone to occupational lung disease. Farmers are also age was not recognized by statute law until 1907 as sufficient
affected because they inhale organic dust from moldy plant cause to qualify for a pension, in 1890 the Pension Bureau
materials and from animal waste, hair, and feathers. instructed the examining surgeons to grant a minimum pen-
Joint and back pains remain common afflictions in de- sion to all men at least 65 years of age unless they were un-
veloped countries and are recognized increasingly as sub- usually vigorous. Veterans, however, had every incentive to
stantial burdens in developing countries (Muirden 1995). Al- undergo a complete examination because those with a severe
though rheumatic fever, particularly in adults, is often ac- chronic condition, particularly if it could be traced to war-
companied by arthritis, this manifestation is generally con- time experience, were eligible for larger pensions. The sur-
sidered to be of relatively short duration; it is benign in chil- geons rated the severity of specific conditions according to
dren, but painful in adults. In rare instances, however, a re- detailed guidelines provided by the Pension Bureau.
sidual nonprogressive athropathy develops (Katz 1977). The Union ~ r pemn sio~n program generated copious
Musculoskeletal symptoms are common with many viral in- records. Pension applications included detailed medical ex-
fections, as well as with syphilis and malaria. Although in aminations, both for men whose pension application or bid
most cases these symptoms-fade away after a short period, for a pension increase was rejected and for men whose ap-
permanent joint damage can occur from diseases such as tu- plications were accepted. These records currently are being
berculosis, vaccinia, and gonorrhea, among others (Von collected and linked to the 1900 and 1910 censuses and to
Hartmann 1974). military service records. The 1900 and 1910 censuses pro-
In addition to disease, occupational stress is another im- vide occupational information; the military service records
portant determinant of musculoskeletal capacity; even in contain information on stress at young adult ages, such as
modern, developed-country populations, this is lower among prisoner of war status, whether the soldier was ever dis-
men in physical work than among men in mental or mixed charged for disability, and illnesses such as measles, diar-
work. In addition. the rate of deterioration in musculoskel- rhea, tuberculosis, typhoid, rheumatism, acute respiratory in-
etal capacity is greater among men in physical work (Nygbrd, fections (e.g., pneumonia, bronchitis), and malaria, as well
Luopajarvi, and Illmarinen 1988). as war injuries. Infant mortality rates in county of enlistment
Thus far I have stressed the positive relationship be- are available from the 1850 United States Census of Mortal-
tween early-life conditions and chronic disease at older ages, ity and from the 1855 New York State Census of Mortality
but the relationship could be negative as well. If genetic sus- (see the data appendix for further details). These mortality
ceptibility to death from infectious disease or other insults at rates are based on reports to the census enumerator, and
DEMOGRAPHY,VOLUME 37-NUMBER 1, FEBRUARY 2000
therefore are biased downward. The bias, however, is un- geon ever noted a respiratory, heart, or musculoskeletal con-
likely to differ markedly across counties. dition, this detailed information is used in the analysis. Thus,
The records used in this paper represent a 49% sample for heart disease, the physician described pulse rate charac-
of the final sample that will be available. They pertain to men teristics; whether a murmur was present and (if so) its tim-
in 45 companies from Ohio, 5 1 from New York, 37 from Illi- ing, type, and location, and which valves were involved;
nois, and 28 from Pennsylvania. The sample is restricted to whether the murmur was accompanied by a thrill; and
native-born men linked to the 1900 census and aged 50 to 64 whether there was enlargement, edema, cyanosis, dyspnea,
in 1900.1 About 10% of these men were not yet collecting a atherosclerosis, or impaired circulation. Respiratory exami-
pension in 1900, either because their applications had been nations included reports of respiratory sounds such as mur-
rejected or because they had not yet applied for a pension. A murs, rales, crepitus, vocal fremitus, and ronchae, as well as
record of a surgeons' examination is available for 93% of all reports of decreased breath sounds. Descriptions of rheuma-
men who received a pension in 1900. Men for whom a sur- tism included the location of the rheumatism and whether
geons' examination is missing tended to be those who en- pain, tenderness, swelling, or crepitation was associated with
tered military service at a late age and received a pension on the joint.
the basis of age. Although I assume that these men did not Because the surgeons did not base their diagnoses on a
suffer from the specific chronic conditions that I examine, recruit's wartime ex~eriencest.h e correlation between illness
the results are virtually identical when these men are omitted while in the army and a chronic condition at later ages is
from the sample. unlikely to be spurious.3 Disease rates are based on an exam-
Men who entered the Union Army were probably ining surgeon's ever having noted a condition. Prevalence
healthier than the population as a whole. An examination of rates for 1910 may be underestimated because men who
men who were rejected for military service suggests that qualified for a pension on the basis of age alone, as did many
mean height for the population was about 0.18 inch less than in 1910 , underwent fewer surgeons' examinations than their
the mean for the recruits. Once men entered the service, the counterparts who qualified on the basis of health problems.
farmers, who were the better-nourished segment of society, I compare the Union Army data with random samples of
were more likely to die because they lacked immunities to the non-institutionalized, white population drawn from the
common camp diseases such as measles and typhoid (Lee 1959-1962 National Health Examination Survey and the
1997). Men who survived the war, however (regardless of 1971-1975, 1976-1980, and 1988-1994 National Health and
occupation), were only 0.02 inch shorter than all recruits at Nutritional Examination Surveys (see the data appendix for
enlistment; this suggests that the war itself induced minimal references.)4 These surveys included medical examinations;
survivorship selection on the basis of height and hence on although not strictly comparable across all years, these ex-
earlv net nutritional status. aminations yield descriptions and diagnoses that can be com-
Increased exposure to disease probably left men in worse pared with those of physicians working under contract for
health than when they entered the army. By age 50, however, the Pension Bureau.5
even men who had grown up in rural areas and had not served The symptoms, signs, and conditions that I examine did
had probably been exposed to as many infectious diseases as not require any diagnostic equipment that was unavailable to
veterans because of increased migration. Although little is late-nineteenth-century physicians. These symptoms, signs,
known about the experience of Union Army veterans from and conditions are valvular heart disease (murmurs involv-
the time when they left the service until they appear on the ing the mitral and aortal valves); congestive heart failure
pension rolls, several tests indicate that this sample is repre- (contemporaneous occurrence of edema, cyanosis, and dysp-
sentative of the general population before the war in regard to nea); arteriosclerosis (detected by feeling whether the arter-
wealth and circa 1900 in regard to mortality experience.2 ies had hardened); irregular pulse; adventitious sounds (mur-
The examining surgeons' re. show height, weight, murs, rales, crepitus, vocal fremitus, and ronchae) or de-
-.,D
pulse rate, respiration, and temperature at every examination.
The physicians also noted general appearance, including gait
and ability to walk. They described each condition; if a sur- 3. I found the same correlations between chronic disease at older ages
and illness while in the army for men who claimed that their disabilities
were traceable to the war and for men who did not make this claim. Unfor-
1. I imposed these restrictions to ensure the availability of occupa- tunately the sample of men who had been examined and whose claim for a
tional information for all men circa 1900 and the availability of epidemio- pension was rejected was too small to be used in an analysis.
logical information on the county in which they grew up. 4. The bias introduced from an examination of non-institutionalized
2. Among all adult males age 20 and over in the households to which men only should be small. In 1900 institutionalization rates for men older
recruits were linked in the 1860 census, mean wealth was similar to that than 64 were just under 3%; by 1990 they were just under 4%. Increases in
found in a random sample. This suggests that military service was not very institutionalization rates for women were much higher (Costa 1998: 11 3).
selective of men of lower socioeconomic status. In fact, 95% of the sample 5. In previous work (e.g., Fogel and Costa 1997), broad categories of
were volunteers. Cohort life expectancies of veterans who reached age 60 conditions in a small sample were compared with those of veterans in the
between 190 1 and 1910 resemble those found in genealogies. Also, the dis- National Health Interview Survey. Because of sample size considerations, I
tribution of deaths from specific causes for all veterans who died between do not limit myself to veterans in NHES and NHANES. When World War I1
1905 and 19 15 does not differ significantly from the distribution of expected veterans and nonveterans in the 1983-1988 National Health Interview Sur-
number of deaths from those causes in the death-registration states in 1910 veys were compared, no differences in prevalence rates for specific chronic
(Costa 1998: 197-2 12). conditions were detected.
THE TWENTIETH-CENTURY DECLINE IN CHRONIC CONDITIONS
creased breath; dullness of chest on percussion; joint prob- etal pathologies relating to general activity, including degen-
lems; and back problems. erative joint disease (osteoarthritis), nonarthritic joint
changes resulting from habitual postures, and fractures aris-
TRENDS ing from traumas.
Table 1 displays prevalence rates for Union Army veterans The declines in disease rates seen in Table 1 are readily
aged 50-64 in 1900 and for the same men aged 60-74 10 reconciled with changes in cause-of-death patterns observed
years later, as well as point prevalence rates for white men during epidemiological transitions. Information on cause of
in the same age groups in recent health surveys. The preva- death simply may be a poor indicator of morbidity rates in a
lence of respiratory findings declined by 67 to 71%, and regime in which infectious diseases are dominant. Fifty-two
would have diminished further if not for the increase in percent of Union Army veterans aged 50 to 64 in 1900 with
smoking. The prevalence of irregular pulse rates, murmurs, valvular heart disease died of heart disease, as did 52% of
and valvular heart disease declined by 90%; of arteriosclero- those with an arrhythmia and 56% of those with arterioscle-
sis, by 53 to 72%. Joint and back problems declined by 20 to rosis. In contrast, 67% of men of the same age with valvular
44%. The prevalence of respiratory findings and arterioscle- heart disease in the 1971-1975 survey died of heart disease,
rosis increased more sharply with age at the beginning of the as did 69% of those with an arrhythmia and 75% of those
century than it does today. with arteriosclerosis. Had 9% of Union Army veterans not
How reliable are these estimates of declines in preva- died from pneumonia, bronchitis, and influenza, 4% from in-
lence rates? A sizable fraction of the decline in irregular fectious disease, and 13% from genito-urinary disorders,
~ulsreates and murmurs mav reflect the more careful exami- cause-of-death patterns might well have been more similar
nations of surgeons accustomed to direct observation, as well to those observed today."
as looser definitions of irregular pulse. The examining sur-
geons' use of hard arteries as a detection criterion provides DETERMINANTS OF CHRONIC CONDITIONS
evidence of peripheral arteriosclerosis; this may be evidence I estimate the impact of early- and late-life stress on chronic
of atherosclerosis (cholesterol and fatty plaques in the blood) conditions among native-born Union Army veterans aged 50
or other diseases such as diabetes mellitus or systemic or lo- to 64 in 1900 and 60 to 74 in 1910 by a series of probit equa-
cal inflammation. A more precise diagnosis of congestive tions in which the dependent variables are equal to l if the
heart failure might include not only edema, cyanosis, and veteran had one of the following specific chronic problems:
dyspnea, but also cardiomegaly; it might exclude coexisting decreased breath or adventitious sounds, dullness of chest on
respiratory infection and asthma. This definition reduces the percussion, joint problems, back problems, valvular heart
prevalence of congestive heart failure to 2.0% among men disease, congestive heart failure, and arteriosclerosis. The
aged 50-64 and to 6.1% among men aged 60-74. In an ex- typical probit equation that is then estimated is
amination of the determinants of congestive heart failure,
these two definitions yield very similar findings. I focus on
the more inclusive definition because of considerations of where I = 1 if a veteran had the specific chronic condition,
sample size. @() is a standard normal cumulative distribution function,
Cumulative ~revalencer ates calculated from the Union and X is a vector of control variables. The effect of a unit
Army sample are not strictly comparable to the point preva- change in one of the independent variables on the probabil-
lence rates estimated from recent surveys for two reasons: A ity of having a given chronic condition is given by the par-
condition would be more likely to be noticed in multiple ex- tial derivative of the probit function P with respect to that
aminations, and men would be more likely to develop a con- independent variable. For example, if the dependent vari-
dition between the time of the last examination and either able is whether a veteran had decreased breath or adventi-
1900 or 1910. If the latter is true, however, then prevalence tious sounds, a 0.096 value of aP/ax for wartime respiratory
rates in the Union Army sample will be underestimated. Yet infection implies that having had a respiratory infection
even if the former is true, it seems unlikely that such a large during the war increases the probability of decreased breath
decline in prevalence rates could be explained by definitional or adventitious sounds by 0.096.
biases alone. Thus, although the estimated prevalence rates The control variables are derived from the 1900 census,
may not be precise indicators of true prevalence rates, the the pension records, and the military service records. They
prevalence rates still were probably much higher in the consist of variables proxying for environmental stress before
Union Army sample than they are today. joining the army, such as dummy variables indicating size of
Analyses of skeletal remains confirm the high preva- city of enlistment and whether the infant mortality rate in
lence rates for joint and back problems (where the latter county of enlistment was high. (A dummy variable fits the
present the problem of being symptoms, not signs). This sug- specification better than a continuous variable for infant mor-
gests that motivations such as pension money are not the pri- tality, perhaps because the measure is noisy.) The control
mary explanation for the high prevalence in the Union Army variables also include dummy variables such as whether the
sample. Larsen et al. (1995), for example, report that in a
mid-nineteenth-century frontier cemetery in the Midwest,
6. Men who died during the influenza epidemic (4.1% of the sample)
eight out of 11 adult skeletons exhibited one or more skel- were excluded from these numbers.
58 DEMOGRAPHY, VOLUME 37-NUMBER 1, FEBRUARY 2000
TABLE 1. PREVALENCE RATES IN A COHORT OF WHITE MEN AGED 50-64 IN 1990 AND 60-74 IN 1910, AND CO-
HORTS AGED 50-64 AND 60-74 IN 1959-1962,1971-1980, AND 1988-1994
1900/1910 1959-1962 1971-1980 1988-1994
Ages 50-64
Decreased breath or adventitious sounds

Never smoked

Former smoker

Current smoker

Dullness chest

Joint problem

Pain/tendemess/swelling specified

Back problem

Irregular pulse

Murmur

Thrill

Valvular heart disease (mitral or aortic origin murmurs)

Congestive heart failure (edema, cyanosis, and dyspnea)

Arteriosclerosis

Ages 60-74
Decreased breath or adventitious sounds

Never smoked

Former smoker

Current smoker

Dullness chest

Joint problem

Painltendemessiswelling specified

Back problem

Irregular pulse

Murmur

Thrill

Vzlvular heart disease (mitral or aortic origin murmurs)

Congestive heart failure (edema, cyanosis, and dyspnea)

Arteriosclerosis

Notes: See the discussion in the text on the comparability of prevalence rates between 190011910 and later years. All prevalence rates are weighted
by the age distribution in 190011910. Prevalence rates are calculated from the Union Army data, the 1959-1962 National Health Examination Survey,
and the 1971-1975, 1976-1980, and 1988-1994 National Health and Nutrition Examination Surveys. Prevalence rates in the Union Army sample are
calculated for both the native and the foreign born. Questions on arthritis are not comparable across all years. In 1971-1975 and 1976-1980 physicians
were asked to note tenderness, swelling, deformity, and pain on motion in the shoulder, elbow, wrist, lnetacarpophalangeal, proximal interphalangeal,
and distal interphalangeal joints, ankle, feet, knees, and hips. In 1988-1994 the examiners were instructed to report only on the wrist, each finger of the
hand, the toes, and the knee for those older than 59. This procedure apparently increased overall finding rates.
veteran ever suffered from measles, typhoid fever, diarrhea, tism; these suggest that 40% of all cases of rheumatism were
tuberculosis, acute respiratory infections (e.g., pneumonia, acute and were caused mainly by rheumatic fever. Some of
bronchitis), malaria, and rheumatism while in the army. the chronic cases, however, may have been prolonged acute
These reports, which were generated either from hospital rheumatic fever (Bollet 1991).Other measures of stress while
stays or from roll call reports, are extremely terse and may be in the army are whether the veteran was ever injured, was
vague in their diagnosis (e.g., "fever"). Very few men are ever discharged for disability, or was ever a prisoner of war.
identified specifically as having suffered from rheumatic fe- Several potential biases are likely to be introduced from
ver while in the army; thus rheumatism while in the army the use of wartime measures of stress. Infectious diseases
may indicate either an underlying chronic condition, trau- acquired in the army may have been more severe than those
matic arthritis, or a viral infection accompanied by rheumatic acquired in civilian life because repeated rapid passage in-
symptoms. The reports submitted by camp doctors to the sur- creased the virulence of infectious agents. Contemporary ob-
geon general distinguish between acute and chronic rheuma- servers noted that measles among Civil War troops was a
THE TWENTIETH-CENTURY DECLINE IN CHRONIC CONDITIONS
much severer infection than witnessed in the civilian popu- measles while in the army. These infectious diseases had a
lation (Cliff et al. 1993:106). Yet if survival rates among substantial impact: Having had only one of these diseases
Civil War soldiers with infectious diseases were lower than increased the probability of respiratory problems by about
in the civilian population, the impact of infectious disease 0.1. Syphilis while in the army was a statistically significant
on older-age chronic disease rates will be underestimated.' and substantial positive predictor of these two chronic symp-
Furthermore, the effect of infectious disease will be un- toms as well. A high infant mortality rate in county of enlist-
derestimated because many men undoubtedly experienced ment increased the probability of dullness of chest in 1900
infectious disease either before they entered the army or af- by 0.04 when regiment fixed effects were excluded from the
ter they left it. If men who caught infectious diseases while regression. When state-of-enlistment fixed effects were ex-
in the army were inherently frailer (an unlikely supposition, cluded as well, it was a significant, positive predictor of de-
given the rates of exposure in army camps), then I will be creased breath or adventitious sounds.
overestimating the impact of infectious disease on later mor- Men who had been farmers and laborers at enlistment
bidity. Yet because I am observing survivors, and therefore were more likely than men in other occupations to suffer
perhaps the hardiest men, I may consequently underestimate from respiratory difficulties both in 1900 and in 1910. By
the effect of infectious disease. 19 10, men who had been artisans and laborers in 1900 were
A different type of bias is likely to arise in the case of more likely to exhibit decreased breath or adventitious
the war-injured. Men who were injured in the war tended to sounds. Artisans who were millers, stone masons, carpenters,
be taller (suggesting better net nutritional status during their and painters, all occupations involving exposure to dust and
growing years); they also tended to come from urban areas fumes, had higher rates of respiratory disorders than did arti-
(suggesting that they already had been exposed to many in- sans who were blacksmiths, machinists, or coopers: 24.9%
fectious diseases). Despite their longer years of army service, of artisans in high-exposure occupations in 1900 suffered
they were less likely to develop tuberculosis, a disease likely from decreased breath or adventitious sounds, compared with
to strike those of poor net nutritional status. If men who had 16.2% of artisans in low-exposure occupations. This differ-
-
been injured in the war were initially healthier and remained ence is statistically significant at .lo.
healthier because they had been exposed to milder forms of Prevalence rates for dullness of chest would have been
infectious disease rather than the severe camp forms, then 31% lower in 1900 and 57% lower in 1910 if the occupa-
being injured in the war might well be a negative predictor tional distribution circa 1900 had been that of older men in
of chronic disease. 1990 and if the occupational distribution at enlistment had
Additional control variables are occupation at enlistment been that of younger men in 1940. Prevalence rates for de-
and occupation circa 1900 (derived from the 1900 census for creased breath and adventitious sounds would have been
men in the labor force and from the pension records for men 2 1% lower in 1900 and 12% lower in 19 10, thus accounting
out of the labor force). Occupation will proxy for socioeco- for 31% of the decline in decreased breath or adventitious
nomic status and for exposire to occupational hazards re- sounds among men aged 50 to 64 and 18% of the decline
spectively at younger and at older ages. Occupations are among men aged 60 to 74.
grouped into professional or proprietor, farmer, artisan, and Marginal effects for probits for valvular heart disease
laborer. Size of city of residence in 1900 proxies for expo- and irregular pulse in 19 10 are given in Table 4. (Results for
sure to infectious disease at older ages. Medium-sized cities 1900 were very similar to those for 1910 and therefore are
had not yet initiated sanitary reforms, whereas large cities only discussed in the text.) Measles, typhoid fever, and
had done so; therefore exposure to disease in medium-sized rheumatism while in the army increased the probability of
cities should be greater (Preston and Haines 1991:98). valvular heart disease by about 0.08 in 1900 and 0.10 in
State of enlistment, age, marital status, state of enlist- 1910. Rheumatism while in the army increased the prob-
ment dummy variables, and regiment dummy variables are ability of an irregular pulse by 0.1 both in 1900 and in
also control variables. I include these two sets of dummy 1910; typhoid fever increased the probability by 0.06 in
variables to control for omitted state characteristics and for 1900 and 0.11 in 1910; acute respiratory infection by 0.05
omitted regimental experiences such as battlefield stress. in 1900 and 0.09 in 1910; tuberculosis by 0.14 in 1910; and
Because regiments were drawn from the same geographic measles by 0.09 in 1910. Men who enlisted in cities with
areas, however, regiment fixed effects may reflect county- populations below 25,000 were less likely to have devel-
of-enlistment fixed effects. oped valvular heart disease in 1900, but not in 191 0. When
The marginal effects for probits for respiratory problems state-of-enlistment fixed effects are omitted from the re-
in 1900 and 19 10 are shown in Tables 2 and 3. For both dull- gressions, infant mortality in county of enlistment becomes
ness of chest and decreased breath or adventitious sounds, a significant predictor of an irregular pulse.
statistically significant positive predictors in all four regres- Table 5 presents marginal effects for probits for con;
sions were an acute respiratory condition, tuberculosis, and gestive heart failure and for arteriosclerosis in 1910. (I do
not examine these conditions in 1900 because they were
found in so few men.) Only rheumatism while in the army
7. The findings also underestimate the impact of infectious disease in
was a statistically significant predictor of congestive heart
a modem population, in which survival rates of men with infectious disease
are higher. failure. This finding is consistent with those of Wilson et
60 DEMOGRAPHY, VOLUME 37-NUMBER 1, FEBRUARY 2000
TABLE 2. MARGINAL EFFECTS FOR PROBITS FOR RESPIRATORY PROBLEMS. MEN AGED 50-64 IN 1900"
Dullness Chest Decreased BreathlAdventitious Sounds
-ap -ap -ap -ap
Mean ax SE ax SE ax SE ax SE
Dullness Chest 0.093 - - - -
Decreased Breath/
Adventitious Sounds 0.237 - - - -
High Infant Mortality
County 0.121 0.035* 0.019 0.029 0.020 0.040 0.027 0.052' 0.019
City of Enlistment
2 50,000 0.069 - - - -
< 50,000 and 2 2,500 0.321 0.048' 0.024 0.037 0.025 0.041 0.037 0.037 0.025
< 2,500 0.610 0.007 0.022 -0.003 0.023 -0.016 0.037 -0.003 0.023
Wartime Conditions
Diarrhea
Respiratory infection
Tuberculosis
Measles
Typhoid
Malaria
Syphilis
Rheumatism
Injury
Prisoner of War 0.106 0.007 0.017 -0.001 0.016 0.050' 0.028 0.048' 0.028
Discharged for
Disability 0.197 0.006 0.013 0.006 0.012 0.017 0.021 0.017 0.021
City of Residence, 1990
Top 10 city 0.046 -0.038 0.020 -0.036 0.019 -0.086* 0.034 -0.088* 0.034
Medium-sized city 0.109 4.011 0.015 -0.011 0.014 0.006 0.026 -0.001 0.026
Small citylnoncity 0.845 - - - -
Occupation c. 1900
Professional/proprietor 0.2 12 - - - -
Farmer 0.410 0.005 0.014 0.005 0.013 0.03 1 0.023 0.036 0.023
Artisan 0.148 0.004 0.018 -0.007 0.017 0.011 0.028 0.007 0.028
Laborer 0.230 -0.007 0.015 -0.008 0.014 0.022 0.025 0.021 0.026
Occupation at Enlistment
Professionallproprietor 0.066 - - - -
Farmer 0.629 0.047* 0.022 0.005 0.013 0.080' 0.034 0.074* 0.034
Artisan 0.144 0.019 0.031 -0.001 0.017 0.066 0.043 0.064 0.044
Laborer 0.161 0.069 0.038 0.059' 0.037 0.1 13** 0.046 0.103* 0.046
Married in 1900 0.861 0.037** 0.011 0.035** 0.011 0.02 1 0.022 0.022 0.022
Age in 1900 57.479 0.002 0.001 0.002 0.001 0.004* 0.002 0.004 0.002
Regiment Fixed Effects Yes Yes
Pseudo-R2 0.10 0.06
- -
Notes: 2,972 observations. The derivative is evaluated at the sample means and, for dummy variables, is for a discrete change of the dummy variable from 0 to
1. All regressions include state-of-enlistment fixed effects.
Wependent variable is equal to 1 if problem by 1900.
THE TWENTIETH-CENTURY DECLINE IN CHRONIC CONDITIONS 61
TABLE 3. MARGINAL EFFECTS FOR PROBITS FOR RESPlRATORY PROBLEMS. MEN AGED 60-74 IN 1910"
Dullness Chest Decreased BreathIAdventitious Sounds
- - - -
Mean ax SE ax SE ax SE ax SE
Dullness Chest 0.113 - - -
Decreased Breath1
Adventitious Sounds 0.378 - - -
High Infant Mortality
County 0.127 0.027 0.023 0.021 0.024 0.0 15 0.036 0.0 13 0.040
City of Enlistment
2 50,000 0.062 - - - --
< 50,000 and 2 2,500 0.323 0.103 0.041 0.088** 0.044 0.096* 0.047 0.061 0.052
< 2,500 0.6 15 0.039 0.034 0.027 0.037 0.021 0.047 -0.004 0.053
Wartime Conditions
Diarrhea
Respiratory infection
Tuberculosis
Measles
Typhoid
Malaria
Syphilis
Rheumatism
Injury
Prisoner of War 0.108 0.000 0.021 -0.005 0.021 0.017 0.037 0.0 18 0.039
Discharged for Disability 0.195 -0.008 0.015 -0.011 0.015 -0.007 0.028 -0.007 0.029
City of Residence, 1900
Top 10 city 0.043 -0.019 0.033 -0.020 0.035 -0.014 0.057 -0.005 0.059
Medium-sized city 0.105 -0.009 0.020 -0.012 0.019 0.013 0.036 0.008 0.036
Small citylnoncity 0.852 - - - -
Occupation c. 1900
Professionallproprietor 0.221 - - - -
Farmer 0.418 0.002 0.0 17 0.003 0.018 0.040 0.031 0.045 0.032
Artisan 0.137 0.003 0.023 -0.002 0.023 0.097** 0.040 0.094* 0.041
Laborer 0.224 0.001 0.020 0.004 0.020 0.071* 0.035 0.079* 0.036
Occupation at Enlistment
Professional/proprietor 0.067 - - - -
Farmer 0.647 0.140** 0.035 0.136** 0.034 0.070 0.046 0.080t 0.047
Artisan 0.130 0.151* 0.086 0.158* 0.089 0.022 0.054 0.044 0.056
Laborer 0.156 0.263** 0.097 0.250** 0.097 0.133* 0.056 0.119* 0.057
Married in 1900 0.873 0.050** 0.014 0.050** 0.014 0.076* 0.031 0.076* 0.03 1
Age in 1910 67.191 0.001 0.002 0.001 0.002 0.005 0.003 0.004 0.003
Regiment Fixed Effects No Yes No Yes
Notes. 2,044 observations. The derivative is evaluated at the sample means and, for dummy variables, 1s for a discrete change of the dummy variable from 0 to
1. All regressions include state-of-enlistment fixed effects.
"cpcndent variable is equal to 1 if problem by 1910.
tp< .lo; *p< .05; **p< .01