Chronic respiratory disease among nonsmokers in Hagerstown, Maryland

ENVIRO~MEh’TAL

RESEAHCH

Chronic

4, 213-232

( 1971)

Respiratory

Disease

in Hagerstown, Ill.

Social

Class

and JAMES

Dqmtnxnt The

Chronic

Among

Nonsmokers

Maryland Respiratory

H. STEBHINIX,

DiseaseI,”

JR.”

of Chronic Diseases, School of HtJgicne uncl Public He&h, ]ohns Hopkins Uniuer.rity, Baltimore, Maryland 2120.5 Receiwd

]anrrar!/

18, 197Z

The relationships of respiratory symptoms and pulmonary function to social class in a population of white male nonsmokers are presented. Among the symptoms only tlyspnea (grade 2+) was fonnd to vary in prevalence with social class, being less common among respondents in white collar occupations whose fathers were whitr collar workers. This is consistent with the hypothesis that dyspnea grade 2 is p:imarily a cardiovascular symptom. The FEV,.,, and the PEFR were positively correlated with social class. This effect was of the order of +0.21 liter for the FEV, ,/ and L22 liter/min for the PEFR. LJp\vard social mobility had no effect on the pulmonaq flmction values.

It is clear from a review of the literature of chronic respiratory disease epidemiology that there is an inverse association between chronic respiratory disease and social class. This is the conclusion of studies on mortality, cross-sectional studies of morbidity, and retrospective studies of clinic patients. There is also so~nc cvidencc that downward social mobility may be associated with severe chronic respiratory disease. This relationship is of interest for two reasons. Such a relationship is a clue, however obscure, to the etiology of the disease. Secondly, the social variable must be takc>n into account in any investigation of possible etiological factors associatt~d with social class, such as specific occupations or air pollution c~xposurc~s. In this paper the relationships of social class and ( upward) social mobility to the prc3&nce of standard rc>spiratory symptoms and to pulmonary hmction valuc~s will bc examined. REVIEW

The literature which respiratory disease may

OF

THE

LITERATURE

discusses the relationship conveniently be divided

of into

social studies

class which

to chronic deal with:

’ Read in part hefore the Ninth AMA Air Pollution Medical Kesearch Conference, Denver, July 22, 1968. This paper is alxtracted from a thesis written in partial fulfillment of the requirements of the Johns Hopkins University for the degree of Doctor of Science. ’ Reprint requests to Division of Epidemiology. Mayo Memorial Building, Room 1158, University of hlinnesota, Minneapolis, Minn. 55455. ’ Present address: School of Public Health, Division of EpidemioIoRy, Llniversity of Minnesota, Minneapolis, Minn. 55455. 213

21-I

JAhtES

II.

STERBINGS,

Jlt.

(1) mortality; (2) morbidity in cross-sectional surveys; and ( 3) morbidity in rctrospcctive studies of clinic patients. The morbidity in item 2 may be divided into analysc~s of “bronchitics” arbitrarily defined and analyses of standard respiratory symptoms or pulmonary function values. Studies of acute respiratory disease in childhood ( l-7) will not be discussed here. The \vcight of the literature is in the direction of accepting the hypothesis that tlics prevalence of chronic respiratory disease is inversely related to social class. Ncvcrthclcss, not all findings haves been positive, and this will bc clearly brought out. MOHTALITY

STUnIES

Thcl mortality studies carried out both in Crcat Britain and the United States have>, with 011~~ minor csccption, all been consistent with the hypothesis of an invcrsc social class cflcct. English mortality data from the publications of the Registrar-Central’s office have bcrn reviewed by a number of authors (S-11). These reviews are repetitive, a11 d only one, that of the Scottish Health Services Council ( 10)) will bc summarizccl here. The Registrar-Gcncral publishes statistics by social class, in catcgorics I-V, based on occupation alone. Class I is composed of professionals, III of skilled artisans, V of unskillccl workers. II and IV arc intcrmediatc. In 1949-1953 there was a fivefold range (34-171) between classes T and V in the Standardized hlortalitv Ratio for bronchitis mortality among malts in the age range 20-63. This gradient has bc~n constant since 1921-23. A similar gradient cxistcd among married \vomcn and among both males and marriccl women over 6S years of age. The fact that married women, whose class is judged on the basis of their husband’s occupation, show a relationship of class to bronchitis mortality as stronglv as their husbands suggests that the causative factors are more likely to bc social than occupational. This close association bctwecn bronchitis mortalities of men and married women persists e\en in a detailed study of classes III, IV, and v. Enterlin~~ ( 12) has published standardizccl mortality ratios for selcctccl respiratory discascs (including acute respiratory infections ) by major occupational group for the United States. A strong inverse relationship with OCCIIpatIona1 class, almost as strong as that seen in England and VValcs, is apparent. The percentage of smokers in each group did not appear to account for the obscrvcd differences. Another type of analysis of the relationship between bronchitis mortality and social class was carried out by Buck and Brown (13). This was a regression analysis of bronchitis mortality in males and females on smoke and sulfur dioxide of concentration. persons per acre, and a social index based on the percentage individuals in social class V. Observations were obtained on 214 administrative districts. They concluded that bronchitis mortality (among both males and females) showed a significant positive association not only with smoke and sulfur dioxide concentrations, but also with the social index. Sulfur dioxide and social index together accounted for differences in bronchitis mortality among

CHRONIC

I~ESI’Il
DISEASE.

111

215

and urban districts; differences among London county boroughs, boroughs, Boroughs \vcrc explained by the Social Index alone, and differences among rural districts were not explained. Sto& (14) carried out a similar study, using a social index based sornetinws on classes IV and V and sonictinies on just V in several sets of localities in G-cat Britain. His results were nearly identical to those of Buck and Brown. out an analysis of mortality in 10 Burgess and Shacldick (15) had carried areas within London. Social class, as ~neasured by percentage of the population in classes IV and V, was significantly correlated only with lung cancer mortality iJJ mak. Bronchitis mortality was most strongly associated with the percentagr and this ~vas interpretc~d as suggesting that of the population born iii London, length of exposure to London air was more important than differences in pallution 1~~1s within London in the etiology of bronchitis. It should br Jwtc,d, howcvc~r. that the pcrccntag~ of the population of an arca born in London \vas strongly related to thr percentage in social class IV and V, and may reflect othci social differcnccs. Fairbairn and Reid ( 16) also corwlatcd bronchitis mortality with A Fog Index, persons per acry, rind ptwcmtagc pwwns more than two per rooJn, but not with social iJJclcx, ova 37 areas of Great Britain. Bronchitis mortality corwlated sigJlificaJltly with the Fog Index, but not the other variables, while pulmonary tuberculosis and cancc‘r of the lung shmwcl the rcvc’rsc pattml of association. “Bronchitis wastagcl” (mortality or preniatuw rctirrJncnt) a~nong post office worlwrs showccl even INS rc~latiowhip to thesr particular social indices than total browhitk mortalitv. Crawford et ol. i 17) in an analysis of the effect of water hardness on mortality in 61 bounty boroughs of England and LValcs of population 80,OOOf iJr 1961 fom~l broJwhitis mortality to be most stroJ,gly ( p < 0.001) :wociated with n combitwd social factor score. Air pollution levels \vc’rc not analyzed. Interestingly, \vatclr hardnws was also significantly relatecl to bronchitis mortality ( p < 0.01 ). as \vas latitude ( p < 0.05 ) . Ashl(~>- ( 18) iJ> au analysis of mortality from 1958 to 1963 in .53 county boroughs ill England and Wales also found a significant inverse relationship bc,twr‘c‘n bronchitis mortality in males ancl social class. His social index \ras a compositc~ of populntion density and social class, the latter bring 1~~4 (111 the proportioJ1 of working ~naks in sm~iskillcd or lowc,r positions. This rc>lationship was strongc~r than the positive relationship bctwvem snwlic~ concentrations :u>d tJlal(~ lmmchitis Jnortality. AJJJcrican stutlics which deal with the effect of both social class ancl air palllltion lcvcls d&r from most English studies in that they deal with n~ortality within c~c~rtain nrban awas. 111 :I dctaik3d study of mortahty from disc~ases of th(n respiratory systr,Jn iJ1 NaslJvill~~, TCIIJI.. Zeidberg et (I/. (19) f ound that bronchitis and rmphyserna mortnhtv rates incwased from high to low social class in are:ls of J]>oderate air pollution, alld this \vas consistent when pollution was dcfincd by nny of four 111P;lsur(~s. Otlwr pollL1tioii classes wcrc’ not nJmlyzed. The pattern was c‘ven strongc~r for influcwza amd pneumonia. Pollution levels and social class werca

216

JAMES H. STEBBINGS, JR.

determined by census tract of residence at death, and death rates were adjusted by age, race, and sex. Winkelstein et al. (20, 21) divided census tracts in Buffalo and environs into three air pollution categories by both suspended particulate levels and levels of sulfur oxides, and into five social class categories. Mortality for white males age 50-69 was computed for each air pollution-social class category. A definite inverse relationship of chronic respiratory disease mortality, most pronounced at the lower end of the social class scale, was observed. This relationship held independently of air pollution level. CROSS-SECTIONAL

STUDIES OF MORBIDITY

The bulk of the morbidity surveys have been carried out in England and have often been studies of particular occupations. “Morbidity” here refers to chronic respiratory disease as defined by some arbitrary combination of findings, to positive responses to the standard respiratory questionnaire, or to lower than expected pulmonary function values. First the positive and then the negative findings will be considered. An extensive prevalence survey of chronic bronchitis was undertaken by Ogilvir and Newell (22) in Newcastle-upon-Tyne. They report several findings indicative of a relationship to social class. Locality of housing, but not housing quality itself, was related to bronchitis; bronchitis prevalence was higher in the more densely occupied industrial riverside areas. This finding could also be related to air pollution levels. In the analysis of occupations it was found that “white collar” workers showed a deficit in bronchitics when compared to those in other occupations; analysis by social class indicated a deficit of bronchitics in classes I and II and an excess in V. There was a nonsignificant excess of bronchitics among men with several job changes, and a significant excess among those with extcnsivc unemployment. This latter relationship holds within Social Class III as well as among all classes.No specific occupations were seen to be associatccl with chronic bronchitis, but bronchitis was related to extremes of tempcraturc, drafts, and dusty atmosphere. Among females a social class gradient similar to that for men was found (although overall prevalence was only 17%compared to 30% for men) and no specific effects of occupation were found. Stuart-Harris and Hanley (23) p resent some data on prevalcncc of persistent cough and phlegm plus severe dyspnea in an industrial and a rural population. In the rural population those who had done farming in the past showed a higher prevalence of the symptom complex in the two age groups than those who had done clean work. This relationship was not tested for significance. In the industrial population, those who had done clean work showed a significantly lower prevalence of the symptom complex than es-miners or others who had done dusty work. While these observations were published as relating to OCcupational exposures, they may in fact relate to social class. It should be noted, however, that agricultural workers show a low S.M.R. of bronchitis mortality ( 10). Brown et aZ. (24) in a study of 1054 men in the age group 60-69 from the

CHRONIC

RESPIRATORY

DISEASE.

III

217

lists of II Birmingham general practitioners examined the prevalence of chronic bronchitis by social class and smoking history. No differences in prevalence of bronchitis by social class in this age group were found among nonsmokers, but a twofold excess, not explained by differences in amount smoked, was found in the lower social classes among smokers of cigarettes only. In social classes I and II the prevalence of bronchitis was very little increased among cigarette smokers as a negative finding in that it suggests ( 16.1-20.0X). Th’ 1s might be interpreted that social class per se does not affect the prevalence of bronchitis but, rather, mediates the cffc~ct of smoking. The College of General Practitioners (25) surveyed a sample of individuals on the practice lists of cooperating British practitioners. They fomld that in both males and fcmnles the trend in prevalence of bronchitis was similar to the trend in mortality by social class. In addition, the age-adjusted male bronchitis prevaand present lence rates showed the same trend by social class amon, 0 ex-smokers smokers separately, and possibly among nonsmokers, although the numbers thcrc arc too small for drawing conclusions. This was interpreted as implying that the small diffcrencc~s in smoking habits have no material cffcct upon the social class gradient in chronic bronchitis prevalence. These results are not pcrfcctly consistent with those of Brown et al. (24) abovu. .4 numbrr of papers from The British Medical Research Council’s Pneumoconiosis Research Unit consider the question of social class although the main intcrcst of the group is the health of miners. Higgins et ul. (26) in a population stlldy found that wives of men in dusty occupations (primarily mining and foundary work). when compared with wives of men in nondusty occupations, showc,d the same proportionate cscess of persistent cough and/or sputum as did tlicir men to nwii in nondusty occupations ( ~1.5: 1)) and an even higher proportionate excess of individuals with a history of chest illnesses in the past 3 years. III addition, Higgins and Cochrane (27) found in the Rhondda Fach that wives of miners showed a distinctly higher prevalence of all respiratory symptoms, as well as a lower mean indirect maximum brcathiq capacity than \vi\Tcs of nonminers. Holland et (11. (4, 5) in a study of two areas in a suburb of London, have analyzed the effect of smoking and social class on the age-adjusted prevalence of winter morning cough and phlegm. They studied the parents of newborn children and fomld a steady increase in the prevalence of cough and/or phlegm with lo\vcred social class only among males, and only in one of the two arcas. The finding held for each smokin, CT category. In that the relationship holds for nonsmokers as well, this contradicts the finding of Brown et (II. (24) discussed above. In their analysis of the PEFR the authors fomld no wholly consistent social class gradichnt among fathers, mothers, or siblings, but the data presented suggc‘st that such an cffcct exists among the fathers, both current smokers and nonsmokers. The authors suggest that the effect of social class may as yet be small in this group of relatively young adults. Alternatively, they suggest that the current generation may never develop the marked social class gradient seen in older men as the factors responsible may have diminished in importance. In another study. of school children in Kent. Holland et al. (6) found a sometimes

significant, if slight, superiority in PEFK in children of social classes I and II compared with those in IV and V. Tibblin and Wilhclmsen (28, 29) found in their series of 50.year-old males in Giiteborg that chronic bronchitis occurred significantly more frequently in lowclr income groups than in higher. There was a corresponding rctlationship with education and with housing quality, although the latter di&rcncc was not significant. St&k et (11. (30) in their study of chronic bronchitis in Prague found that, while the frequency of persistent phlegm among their different income groups could bc explained by the frqucwcy of cigarette smoking. the highest income group showed a sharply reduced frccluc>ncy of respiratory infections not explained by their cigarette smoking habits. On this side of the Atlantic, Krinkman and Coates (31) hund bronchitis ( productive daily cough) to bc marts co~~m~on among executives in J)ctroit than among factory workers, but explained it 011the basis of smoking habits. cigarette smoking bciqg more frrqutnt among thtx esccutives. They also suggclst that the more educated were more scllsitivc to their questionnaire. In a later paper ( :32) they present data indicating that the FE\‘, ,, and the maximal midespirator~flow were higher in cxc-cutives than in factory workers of the same age catt>gory. Wyndcr et al. (33) in their study of two groups of malt patients ( IMJIIrespiratory), oiic’ from New York ant1 one from California. nncl ;I group of Sc~venth-Day Adventists, also from California, found that in the Calihm~ia patient group, laborers, both \vith and without industrial q~osurc~ to irritants. show& a highc>r rat<, (54-56X) of pc&stcut cough than did profcxssional and white collar \vorkcrs ( 46%)) and this hight~r ratv was indcpcndcW of thy IIIIIII~IC’P of cigarvttcs smoked. Antlcrson and Larscsn (34) also stlldicd the, frquency of ahsv~lc~s clue to no11communicably rqiratory discascx in grade 1 children in thrre towns in J3ritish Columbia. In 011~~town a consistent social class gradient was found, rrspiratory illness being most common iu upper class male children, and lcast coiiii~~oi~ in lowctr class 1na1~~children. An invcrsc pattern held for females. Entcrlincb (12) in a study of two communities compared coal miners and th(+ wivc>s with other manual workers awl their wivchs. In one community coal miners rcportctl significantly more cough and pc‘rsistent cough, dyspnea gradct 3T. and histories of pncunionia and pleurisy than other manual workers; the wi\-tas of coal miners rrportcd significantly more‘ phlegm and persistcut phlegm, \\:hrclzing in the al~scww of colds, dyspnca grade 3f. and histories of pneumonia than wivc,s of other manual workers. In the second community only a history of among coal miners and their wives, pneumonia was significantly more conimon as was a history of pleurisy among coal miners only. Entcrlinc suggests that “infect” their wives, and that this is as good an rxplanation coal miners may as ~oI~~~~~oIIsocioccononlic and c~n\~ironmental conditions. The, studies dc~scribed above have all yielded SOI~IC positive rclsults; ncgativv findings arc less common, but ththy do exist. Anderson et al. (35) reported the rffcct of four demographic variablrs on tht> pr~;dc~~, adjusted for smoking habits, of chronic rc%spiratory disease. The

220

JAMES

H.

STEBBINGS,

JH.

same institutions as some of the cases, and there was matching on “social status.” For these reasons, the validity of the results is open to question. Fry (41) found that the bronchitics in his practice were mostly of social classes III and IV but felt that this was compatible with the social class distribution of his practice. One interesting retrospective study by Meadows (42) has investigated social class change in 111 male hospital in-patients with chronic bronchitis and an equal number of controls. A lifetime social class history was obtained. The bronchitic patients showed a drift down the social class scale, principally in the 5 years previous to interview; this is presumably due to disability. Meadows also found that fewer bronchitics than control patients attained the social class of their fathers, even in the first half of their working lives. This would precede any significant disability and may indicate an association with downward social mobility. This may be related to the finding of Ogilvie and above, that bronchitics had experienced an excess of Newell (22) d’ lscussed job changes and extensive unemployment. On the other hand, Anderson et al. (35) did not find an excess of job or residence changes among those with chronic respiratory disease in Chilliwack, B. C. Assunling that the control group and the bronchitic group were each representative of their respective base populations in their social class mobility, and recognizing the small numbers involved, Meadows computed the effect on the Registrar-General’s Standardized Mortality Ratios for bronchitis by social class of the observed change. The ratio of class I to V was lowered from 34: 171 to 34: 101 for males. The progression from I to V was no longer smooth, however: the original ratios for classes I to V were 34 : 53 : 98 : 101: 171; the corrected, 34:61: 126: 120: 101. This is an odd effect and could be due either to bias in the Nevertheless, an effect of social class remains. sample, or to small numbers. illness in Australian alcoholics Rankins et al. (43) in a study of respiratory found that the FEV,.,,, expressed as a perccntagc of the predicted value, was (white collar) social group than for greater (87.7%) in smokers of the upper smokers in the lower social group (73%). The d i f f erence was statistically significant. Bronchitic symptoms and abnormal values for the FEV,.,, were more common in the lower social class smokers than in the upper class smokers, but the differences mere nonsignificant. Numbers, however, were small. The authors tentatively concluded that the approximately twofold excess risk of chronic lung disease in alcoholic men was a function of the greater frequency of cigarette smoking among alcoholics. A particularly interesting observation, given an increased frequency of chronic respiratory disease among alcoholics, was an apparent shift of men from skilled to semi- and unskilled blue collar positions as their alcoholism progressed. METIIODOI,OGY

This analysis is based on a population of 410 white male nonsmokers (of cigarettes), aged 35-64, residing in and around Hagerstown, Maryland. All of the respondents had lived outside the county in which Hagerstown is located for at least 20 years of their life. Individuals with a significant history of asthma or of coronary disease are not included with these 410 respondents. The standard

CHRONIC

RESPIRATORY

DISEASE.

221

III

respiratory questionnaire was applied, and the FEV,,, and PEFR measured using the McKesson vitalor and the Wright peak flow meter. The details of the design and execution of the survey are described in previous papers (44, 45). The only elements new to this paper in the way of methodology are those which bear on the determination of social class. The three social class indicators available were income, education, and occupation. These variables have also been combined to form two Index of Social Position Scores. The occupation of the father has also been used; this, in combination with the occupation of the respondent, allows an approach to the question of change in social class. Data collection and coding was designed to allow the construction of an Index of Social Position score similar to that of Ilollingshead and Redlich (46) and to allow cross classification of occupation of respondent and his father after Syme et (11. (47). A d cscription of the scoring of the single variable scales mcl of the two Social Position Scores follows. THE

INCOME

SCALE

Table 1 shows the income categories chosen and the corresponding assigned scores. Incomc rcfcrs to income of the respondent alone, and not to family income (which was found to correlate less well with pulmonary function values). There was originally a category of less than $2.500 per year, but, as it was almost empty, it was combined with the next higher category.

I II c’ome level 12,.500 + 10,000-12,499 7,.500-!J,W’J .5,000-7,499 04,999 Mrlmtion level I + Years graduate school (‘allege gradllat8e I -:I Tears college High school gradllate I-:% Years high school Completed grade school O--B or 7 Years grade srhooP Total ‘I In some systems

grade

school

is completed THE

in seven

EDUCATION

5::

12.9

64 63 11-i 4:s 40 :;s 410

1,i.G

I;,.4 23.1 I 0 .5 !I. x 7.5

years. SCALE

The categories of educational level attained and the corresponding assigned scores are also shown in Table 1. The categories were defined to yield a seven position educational scale very similar to that of Hollingshead and Redlich.

222

JAhlES

H.

STEBBINGS,

JR.

Thcrc arc slight differences between this scale and that of Hollingshead and Redlich. Those who had completed 1 full year of graduate school, rather than those who had received a graduate professional degree, wcrc assigned a score of 1. At the lower end, the Hollingshead and Redlich scale is based on the assunqtion of a grade school (l-6 years), junior high (7-9), and high school (10-12) system rather than a grade scl~~ol ( 1-S years), high school (9-12) system. The catcgorics S~OWJI in Tahlc 1 wcrc dcfincd on the assumption that completion of a stage in schooling is a more important conwpt than the actual number of years in school. THE

OCCUPATIOS

SCALE

I Iollingshcad and Rcdlich used a sc’vcn class occupational scale which is :.I modification of the Alba Edwards system. That scale as presented is insufficientl! well dcfincd for coding, howcvcr, so a similar ‘i-point scale based on that of Warner et al. (LLS), also a modification of the Edwards system, was uwcl. This scnlc involves coding both occupational class (upper middle to lower) ant1 occupational group (professionals, clerks and kinclrcd \vorkcrs, protective and scrvicc workers, etc. ) . Even the system of Warner et (11. was felt to rcquirc some changes. Tlw changes madr followed the usage of Millcr and Swanson (49). Special instructions wcrc written to deal with certain cxtrcmcly variabla groups ovcweprescnted in this study, such as engineers and clcrgymcn. The distribution of respondents 1)~ occupational class scores is sl~own in Table 2. It is not felt that the changes made have in any \vay affected the comparability et (II. While the of the occupational scale used hcrc and the scale of \\‘arncr FVarncr SC&, and the one used here, arc similar to that of IIollingshcxd mcl Rcdlich, thcrc arc distinct diffcwnces. In essence, categories 4-6, and part of 7 in the latter scale are shifted upward one category in the Warner scale. The occupational class scores arc used both as is, and categorizccl into Lvhitcb \Vhen the occupation collar ( 1-3) and l,luc collar (4-7) groups in some analysts.

(1 Cat,egories 1-3 are made up of pwfessionals, proprietors and ma~lagers, businessmen, and and 4 in other categoclerks and kindred workers. Category 3 also inchtdes 2 in the agric1dll1ral and agricldtlwal workers (two irrdiries. Cat,egories 4-7 itlclltde manual, prot,ective nlkd service, viduals).

CHltONIC

HESPIRATORY

DISEASE.

22:3

III

Table 3 shows the occupations of the rcsponclcwts by the occupations of their fntlwrs. A majority of individuals in white collar occupations had fathers in c~itllc~r ldw collar or agricultural occupations. On the other h~lncl, very few incli\-itluals slio~~d don~nmwtl social mobility. ~lr1X.s

OF

SOCIAL

bSlTlON

SCOHE

OlrE

Tll~~ 11~1~~sof Social Position Score of lIollingsl~cat1 rind Rcdlich is obtained as Scorch : 6 14 Residence + 5 X Education +- 9 X Occupation. Education and ocxwlxition liavc bwn scord licw in a siinil;w way, but the r~sidc~nw category tl~~p~ntls 011 ;I tletail(d cw~logical analysis wliidl is not ~ivailnld~~.

INDEX OY SOCIAL POSI~IOS Scow

Two

T11cs II&S of Social Position Score Two lxtter predictecl p~hnonary function wlric~s than any tfvo variabks ant1 as xwdl as c&cation, incoinc~, and owupatiol~ LIW~ as a group of three inclcpcnclcnt \-arial)la For this rc~ason, it was the one chowi to be cat~gorizcd for tabulating rqkxtory symptoms against social class. As no wasonably objccti\~c~ nic~thotl in\dving jutlgecl class wrkl 1)(x used ( as tlow by Hollingshcwl and Rccllich ), divisions wcw madr arbitrarily unckr tlw rc3traint that reasond~ly qua1 proportions of the population fall in each of the classes. For analysis of the prcd~ncc of symptoms thrw catqorirs wc‘w 1lscd: scores from 20-49 (I ), from 50-79 (II), and 80-12’3 (III ). Thescb will 1x1 rc~f~~rrcd to as social class entc~gorics.

224

JAMES H. STEBBINGS, JR. RESULTS

PHEVALENCE

OF RESPIHATORY SYMPTOMS OF THE RESPONDENT

AND SOCIAL CLASS

The relationships of respiratory symptoms and history of respiratory illness to the Index of Social Position Score Two categories are shown in Tables 4 and 5. The only significant relationships are shown by dyspnea grade 2+, which is relatively infrequent in the upper social class, and by a history of hay fever which began in childhood or adolescence. The latter item is most common in the

Social I

class categories

II

III

All

~__

-

No. Cough in a.m., wint,er Cough day/night, wint,er Persistent cough Phlegm in a.m., winter Phlegm day/night, winter Persistent phlegm Periods of cough and phlegm Dyspnea grade 2 + Nasal catarrh, winter Persistent nasal catarrh Total 0 Heterogeneit,y

FI~EQUENCY

76

4 5 6 7 6 6 14 11 42 33

2.9 3.6 4 3 5.0 4.3 4.3 10.0 7.9 30.0 23.6

140 by social

OF HISTORY

No.

95.9

class was tested

‘,‘;

9 4 6 7 5 6 15 32 42 34 155 with

3 8 2.6 3 .9 4.5 3.2 3 .9 9.7 20.6 27.1 21.9 103.2

‘:;,

4 :3 7.8 .7 2 7.0 4 .:3 2. 6 7.8 22.6 27.8 15.7

1X 1x 18 x3 16 I.5 38 69 116 8.5

4.4 4.4 4.4 5.4 3 .9 3.7 9.3 16.8 28.3 20.7

115 105.1

TABLE 5 OF RESPIRATORY ILLNICSS BY SOCL~L WELL NONSMOKERS

Cr,.\ss

No.

by social

No.

table.

33 9 8 17 16 13 8 140

Oi’ ,‘O 23 .6 6.4 5.7 12.1 11.4 9.3 5.7

class was tested

No.

with

19.4 8.4 6.5 9.0 14.2 4.5 8.4

N.8. N.S. N.S. N.S. N.S. N.S. N.S.
101.3

C.\TKGOHY

111 1,1 /o

30 13 10 14 22 7 13 155

410

Higf 1,’

AMO~~(;

class categories

II

I

a Heterogeneity

3 9 6 8 .i :: 9 26 :32 18

‘.‘;m

a 2 X 3 chi-square

Social

Colds go to chest Bronchitis (adult) Viral pneumonia (adult) Sinusitis (early onset) Sinusitis (adult onset) Hay fever (early onset) Hay fever (adult onset) Total

No.

No.

All ci, (3

25 21.7 7 6 7 32 2 8 11.5

a 2 X 3 chi-square

6.1 .i .2 6.1 19.1 1.7 7.0

table.

NC,.

‘; :

88 2!) 24 38 60 23 29 410

21.5 7.1 J-9 , 9.3 14.6 5.4 7.1

N.S. N.S. N.S. N.H. N.S. < 0 0.; N.S.

225

CHRONIC RESPIRATORY DISEASE. III

upper social class. However, the chi-square test by which significance was tested is relatively weak and insensitive to trend. It is desirable to discuss the patterns which appear in the prevalence figures themselves. Cough, particularly during the day or night, might be considered to be more prevalent in the lower social class, but the same is not true of phlegm production, or the occurrence of periods of cough or phlegm, which show no apparelit patterns at all. Dyspnea grade 2+ is significantly less common in the upper social class; this Irold~ for all age groups. It will bc seen below, however, that the true situation is more complicated than this. Nasal catarrh shows no pattern of relationship to social &SS whatsoever in any age group. The item “colds go to chest” aIso shows no relation to social class in any age group. Neither dots a history of bronchitis or viral pneumonia in adult life show any cffrct of social class. Hay fever, beginning in childhood or adolescence, was seen to be significantly more frequent among those in the upper social class and shows a gradient across classes. Sinusitis, beginning in childhood or adolescence, shows exactly the same pattern, although it does not reach significance. Adult onset hay fever and sinusitis, however, tend to show the opposite pattern, being less common in the uppclr social class. As the prcvalencc of all sinusitis or hay fever is remarkably constant by social class, this finding may be an artifact of reporting. Tabulations of symptoms by income and education were also made. While some’ nc:lr significant relationships were found, no consistent pattern emerged, and it was clear that the categorization of the Index of Social Position Scores yielded the only clear patterns. PIIT:~A~~E~CE OF RESP~RATOHY SYMPTOMS

AND CHANCE IX- SOCIAL Ct..xs

A dctailctl analysis of the prevalence of symptoms and a history of respiratory disease by occupation of respondent and his father was carried out. Prevalellcc IYI~CYS wcr(’ computed for CYI& of the occupational categories shown in Table 3, All results were negative except those for dyspnea grade 2+, which are shown in Table 6. This table demonstrates that only one group of individuals shows IOW rates of dyspnea. This is the white collar sons of white collar fathers, and it is true in alI age*groups. All other groups show essentidy the Same rat<, in all age groups. SOCIAL

CLASS EFFECTS

OS PULMONARY

FUNCTION

VALUES

The first step in the analysis of the cflects of social class on pulmonary function VU~UCSwas the addition of the simple social class variables to the basic multiple regression equations. The effects of adding these variables, income, education, and occupation, scored as shown in Tables 1 and 2, are shown in Table 7. Blue collar occupation, as a dummy variable, was also used to test the assumption that any effect of occupation was continuous and not dichotomous. It may be seen that all of these variables significantly affect both measures of pulmonary function. The regression coefficients in all casesare negative, indi-

226

JAMES

White -4ge

and

uccr1pa-

H.

STEBBINGS,

JR.

collar

-

tion of respondent,

XI,.

1)yspnea grade Age 3-64 Whit,e collar ISlae collar Age 35-44 White U&U Blue collar .4ge 4S.54 Whit,e collar Blue collar Age 55-6-l White ~~Jlal Blue colla1

2+

Tot,nl N-.4ge Whit.e collm Blue collar

:&6-l

ti

.i

.i

4

3:;

:;

17.6 ‘LO.3

!I 14

20.0 2X. 6

.-I

19 L’

ti

30.0

‘>

15.4

:;

17.6


N.T. N.T.

I II) 1s

Itegrekou --

“2 14

coefhciellt

s

Proporliori of vsriauce explained

Fil,-LOOi”

-.

FE:V 1.(I Income of rrsp~n~tlcrlt IGlrlr:+tioll O~c~~p:xtiollal clits.ih Blue collar occrlpatitrt( So&l Illdes 1 So&l Indes L’ PElJI1 Il1come of wspo~ldrlrt l*Jdllcatiott Ckcl~pationnl class Blue collar occupation Social Index 1 S;ocGil Index 2

- oJ:~s - .0440 - .0510 - 1070 - ,004 I - 00:x - 9.06s -6.13; --s. 1.5:; -1S.84 - .6”61i I - .5290

‘i,‘J(jz**

.0118 .0140 .011x ,006 1 01.x 0167

10.0:33** s.41:3** 4, “7Sf I , 10.904*** 12.009***

OL’M .Ol!N o’)“‘$ II, ,013s 03.X 029x

11,4:3x*** 0.691** 10.879*** 6.659** 12. .im*** 14 677***

n From mult,iple regression equations which included age, height, alld ilIst rumeltt and iuterviewel efiects. Iudex 1 is weighted combination of education aud ocnlpation; Index 2 also illclttdes income. ” is a dummy variable :Llld cqllals 1 if respotldent’s occrtpat ioll is hllle “Blue collar occupation (*ollar, 11 otherwise. Variables have been tested individually, not added in seqllence. 1’ *. **, nrld *** judicate siglSranre at the o.O.J, 0.01, and 0.001 levels, resprctively.

eating that lower socioeconomic status is associated with lower pulmonary function values. It is clear that the strengths of the effects of the different variables, with the exception of blue collar occupation, are very similar. When the composite Index of Social Position scores are used as independent variables again (Table 7), the proportion of the variance explained becomes noticeably higher. The second Index, incorporating income in addition to education and occupation, is superior to the first, but both are superior to the simple variables taken singly. Two questions arise comeruing thr validity of the ube of these composite scores in the analysis. The first of these is whether or not the use of three independent variables together, income, education, and occupation, might explain more of the variance thau the scores used as single variables. Regression cquations incorporating all three of thesrx variables at once were run; the incrcasc in the proportion of the variance explained by using these three variables rather than the second Index of Social Position was no more than would be cxpccted due to the incrcasc in the number of degrcps of freedom. The scconcl question was that of the linearity of the relationship of the pulmonary function values to the Indices of Social Position. The first approach to this question involved breaking both sets of scores into five classes, yielding four dummy variables for inclusion into the regression equation. This impruvcd the degree of fit only slightly of the FEV,., values. A plot of the regression cocficic~nts of the dummy variables against the class midpoints indicated possible nonlinearity in the cast of the FEV, ,,,, but an almost perfect linear relationship in the cast of the PEFR. As the second Index of Social Position score showed a closer relationship to the pulmonary function values, those scores were divided into nine classes, and furthrr rcgrcssion analysis with eight dummy variables was carried out. A plot of thcsc regression coefficients for both the FEV,,,, and the PEFR against the midpoints of the class intervals showed some variation around the line of regression but did not suggest any alternative to the notion of a linear relationship. Recausc~ no altc~rnativc relationship was indicated in these plots, and because of thfl closely linear relation shown hy the PEFR to the second Index of So&l

Populat.iorl Social

11~irr

20-m 40--.i!I M-79 S&9!) 100-12s

2

No. 8:; 92 I20 S.i :
ilt class “i 20.2 22 4 L’s :; 3.7 7 3

cl~ll~~l:tnt 7 FbV,

.”

.I97

efieel~ PEFR 18.X

0.3)

ti

0 nn - .,IS

$1

0 -9.0 -24.5

a These are the regression mefficients of the appropriate drnnmy variables in a regressioll eqllatiorr also conlaining age, height, and instrlmlellt and int,erviewer effects. .\liddle classes are zero by definition. Other values are relative to these, atld are in liters for FEV, ,,, in liters/min for PEFR.

228

JAMES H. STEBBINGS, JR.

Position when five classes were used, it has been assumed that the assumption of linearity is correct. The addition of a covariance term, social class X age, yielded no improvement in the fit of the regression. The five class intervals and the regression coefficients of the corresponding dummy variables are shown in Table 8. These regression coefficients give a good idea of range of the pulmonary function values by social class. This range is of the order of magnitude of 0.4 liter for the FEV,,, and 45 liters/mill for the PEFR. PULMONARY

FUNCTION

VALUES

IN SOCIAL CLASS

AND CHANGE

In addition to the effect of the social class of the respondent on his pulmonary function values, there is the question of the effect of change in social class. Because data on father’s income was unavailable, and educational levels between generations were thought not to be comparable, only occupational data has been used in this analysis. The first analysis involved the classification of the occupation of the respondent as white collar or blue collar, and that of his father as white collar, blue collar, or agricultural (Table 3). Five dummy variables corresponding to five of the sir possible cross-classifications were defined (50). These yielded estimates of the relative pulmonary function values in each cell and may be seen in Table 9. It is clear that if the respondent is in a white collar occupation neither pulmonary fmlction value, the FEV,.,, or the PEFR, is affected by his family background as dctcrmined by his father’s occupation. The lowered FEV,.,, in blue collar workers also sc’cms to be independent of family background, but there is some indication that peak flow may not be quite so low among blue collar workers whose fathers were white collar workers or in agriculture as it is among those whose fathers were also blue collar workers. Another index of social change, the difference between the occupational scores of the rcspondcnt and his father, was also analyzed. Since this measure is not independent of the occupation of the respondent, the latter effect was removed. The addition of change in occupational class to the occupational class of the

Occupation FEVI

of respondent

White

~~~11s~

Occrlpat,ion Bllle

of father dollar

Agricultrval

.o

Whit,e ~011x1 Bhte collar PEFR White collar Blue u&u

o* - ,167 Oh 6.26

- 0””1Y - .097 1.60 -27.80

0’20 - .12L’ 0, 3’ - 0 !)L’

a E&mated using dummy variables in addition to age, height, and instrument and interviewer effects. Effects are in liters or liters/min. Respondent’s occupation (df = 1) was significant in both cases at t,he 5:’ level. Neit,her for t,he FEV, Onor t,he PEFR did t,he occupations of the respondents’ fathers (@ = 2) nor the interaction terms (df = 2) even approach significance. b This cell is by definition zero, and other values are relat,ive t)o it.

CHRONIC

respondent explained.

does

not

yield

RESPIRATORY

a significant

DISEASE.

increase

229

III

in the proportion

of variance

DISCXJSSION PREVALENCE

OF

RESPIRATORY OF

THE

SYMPTOMS

AND

SOCIAL

CLASS

RESPONDENT

The prevalence of respiratory symptoms and histories of respiratory illness is seen not to depend on the social class of the respondent in this population of nonsmokers. The only significant exceptions to this are dyspnea grade 2+ and a history of hay fever beginning in childhood or adolescence. Below we shall see how the finding concerning dyspnea may be interpreted. These negative findings allow at least three interpretations. One is that there arc in fact no differences in the prevalence of the simple standard respiratory symptoms, particularly cough, phlegm, and periods of cough and phlegm, by social class in the United States after correction is made for smoking habits. This does not csclude the possibility of differences in the frequency of severe symptoms complexes. A second explanation is that the number of individuals with symptoms in this study, particularly of cough and/or phlegm, are too low to allow one to detect the real effect statistically. This is a valid interpretation. A third interpretation is suggested by the findings of Brown et al. (24). This suggests that an effect of social class would be found if cigarette smokers were included in the sample. This means that social class might mediate the effect of cigarette smoking rather than directly reflecting the influence of some other environmental variable associated with social class. THE

PREVALEXCE

OF

RESPIRATORY IN

SOCIAL

SYMPTOMS

AND

CHAXGE

CLASS

The social class of the respondent in adolescence, as measured by the occupation of his father, does not appear to affect the prevalence of respiratory symptoms (other than dyspnea grade 2-t ) except insofar as it is related to the social class of the respondent at time of interview. This is not a strong finding for two reasons: numbers are often too low to permit adequate analysis, and major downward mobility was almost nonexistent. That a significant effect of upward mobility, which was common, was not found dots not imply that no effect of downward mobility exists. It was suggested in a previous paper (44) that dyspnea grade 2+ in this population of nonsmokers might bc more closely associated with cardiovascular disease than with chronic respiratory disease. It should be pointed out here that the prevalence of dyspnea grade 2+ by occupation of the respondent and his father, namely, low in white collar sons of white collar fathers and equally high in all other groups, is very similar to the pattern of coronary heart disease found by Hinkle et al. (51) in Bell System employees. In that population differences in risk of coronary disease appear to exist at the time the men are hired Xl d are not greatly changed thereafter by an individual’s rise in the hierarchy of the organization.

230

JAMES

H. STEBBINGS, JR.

SOCIAL CLASS EFFECTS ON PULMONARY

FUNCTION

VALUES

The finding of a significant effect of socioeconomic status on pulmonary fmlction values in this population of well nonsmokers implies the existence of a factor other than tobacco smoking, and associated with socioeconomic class. which affects pulmonary function values. It will be shown in a future paper that this factor is neither urban residence nor some specific occupational exposure. In fact, the latter explanation is not consistent with the gradient of pulmonary fmlction values among white collar workers rrportcd here. Frequency of acute respiratory illnesses, in childhood or adult life or both, might also be involved. The data presented in Tables 4 and 5 do not support the notion that serious acute respiratory ilhlesscs have been more common in the adult life of individuals in the lower socioeconomic classc~ Pneumonias. other than viral pneumonias in adult life, were recorded so infrequently for either childhood or adult life that they could not even be analyzed. This hypothesis cannot be rejected, however. Questionnaire data on acutca episodes of illnesss in the distant past are open to question. And, of course. there is no data in this study on the frcqucncy of less scvcre acute respiratory disease. However, for this hypothesis to be acccptcd, it would have to be shown to be consistent with two findings: that there are differences in pulmonary function values among white collar workers related to socioeconomic status, and that the socioeconomic status of the respondent at the time of interview, rather than that during his adolescence, was the variable related to his pulmonary function values. The potential relationship of the rather small effect of socioeconomic status on mean pulmonary function values to risk of detectable chronic respiratory disease will be discussed in a future pap”. PLJLMOSARY

FUNCTION

\‘ALUES

ASD CHANGE

IN SOCIAL CLASS

No effect of change in social class on pulmonary fmlction values appeared in this study. This conclusion is probably valid for upward mobility, which domiIt is possible that there was insufficient downward nated in this population. mobility ior a statistically significant effect to appear. ACKNOWLEDCMEXTS This research was wpported 1)~ Public Health Service Grants CRS 47-65 and GRS 35-66. Additional wpport was received from the Washington Comty (Maryland) Tuberculosis Association. The Training Center for Public Health Research in Hagerstown was srlpported 13, PHS Grant CD-00001. Support for the investigator was received from the National IIeart Institnte under Grant TI-HE-5297. Dr. George W. Comstock and Dr. Mary hlonk gave advice and enconragenrent which is gratefully acknowledged. I also thank Dr. Abraham Lilienfeld and Dr. Richard Royall fol their criticism. REFERENCES I. ~IAIINETT, R. F. W., AND hsfAIR. A. Chronic bronchitis and the catarrhal chilcl. Scot. Med. J. 8, 175-184 (1963). 2. DOUGLAS, J. W. B., AND WALLER, R. E. Air pollution and respiratory infection in children. hit. J. Prev. Sot. Med. 20, 1-8 ( 1966).

CHHONIC

5. 6. 7. 8.

9, IO. 1 I. 12. I:].

I 4.

16.

IX. 18.

20.

2 1.

22.

23.

2.1.

25.

26.

III

2:31

ICNOWELDEN,

schoolchildren.

J.,

AXD

Brit.

A. S.. AN) REII), D. D. Air pollution and other local factors in wspiwtoq J. Prcti. Sot. hfetl. 12, 94-103 ( 19.58). CI~~WFOIW, hl. I)., GARDNLI~, hf. J., AN~ ?rlom~rs, J. N. hlortality ant] hart]ncs\ of local \vatcr supplies. Lancet 1, 827-831 ( 1968). .tSHLliY. D. J. B. Environmental factors in the aetiology of lrmg cancer and I~ronchitis. Hrit. .I. Pwo. Sot. Aled. 23, 258-262 ( 1969). ZP:II)IIEIK:, L. I>., 1101mos. R. J. XI., AYD LA’VIIAU, E. The Nash\,ille Air Pollrltion Stntl!,. 1’. hlortality from diseases of the respiratory systenl in relation to air polll~tion. ;\I&. EtttjiWfl. Ile~Ith 15, 214-223 ( 1967). \f'lXKELS7 EI\, W., JH., ct (I/. The relationship of air pollution and economic st,ltrls to total mortality and sclcctetl respiratory system mortality in nwn. 1. Suspentlwl particrilatrs. Arc11. Etiuirot,. flealth 14, 162-171 ( 1967 ), \\'lNKELSlXlN. \V., JH., ct c/l. The relationship of air pollution and economic status to total Inortality and selected respit-atory system mortality in men. II. Oxides of s~llfllr. .bch. Enuirm. Iledth 16, M-305 ( 1968 ). OGIL\.IE, A. G., ASI> NEWELL, 11. J. “Chronic Rronchitis in ~~~\\~castlr-llpo~l-T~~l~.” Livingstone, Edinburgh, 1957. STUART-~IAHHIS, C. Il., ANII HANLEY, T. “Chronic Rronchitis, Emphysenla. ar~l Cur PII~IIN~&.” Wright, Bristol, 1957. BHO\~S, R. (i., MCE;EOW.V, T., AND WHITFIELI), A. C,, \I’, A note on the asso&tion bebee11 smoking and disease in men of the se\,enth tlecade. Brit. J. Prec. Sot. .\fd. 11, 162-164 (1957). College of Genera] Practitioners. Chronic bronchitis in Great Britain. A National Sllrvvy carried ollt by the Respiratory Study (Crotlp of the College of Genera] Practitioners. Brit, Med. J. 2. 973-979 ( 1961). ffI(:GINS, I. T. T., et (12. Population studies of chronic respiratory &ease. .a c()lllp;lristrll of miners, foundry workers, and others in Stately, Derbyshire. Brit. J. Id. Med. 16, 2.55-268 ( 19.59 ).

FAlIIl3AIRN,

disease. Ii.

DISEASE.

A. J. Patterns of respiratory illness in HANDYSIDE, J. Pm. SOC. Md. 21, 7-16 ( 1967). (=oLLEY, J. R. T., AND HOLLAND, W. W. Social and environmental factors in respiratory disease. Arclx. Ent;iron. He&h 14, 157-161 ( 1967). ~[(>LI.ANI), W. W., et ul. Respiratory symptoms and ventilatory function: a fan+’ strldy. Rrit. J. Prea. Sm. Bled. 23, 77-84 (1969). f{()l
3. LUNN, J. E., Sheffield infant 1,

HESI’IHATOHY

Brit.

232 27. 28. 29. 30.

JAMES

H.

STEBBINCS,

JR.

HIG~IXS, I. T. T., AND COCHHANE, A. L. Chronic respiratory disease in a random sample of men and women in the Rhondda Fach in 1958. Brit. J. Ind. Med. 18, 93-102 ( 1961). TIBBLIN, C., AND WILHELMSEN, L. Resultat f&r undersokningen “1913 ars man”: Kronisk bronkit i en stadshefolkning. Lakartidningen 64, 2455-2460 ( 1967). IRNELL, L., AND KIVILO~G, J. Bronchial asthma and chronic bronchitis in a Swedish urban and rural population with special reference to prevalence, respiratory function, and socio-medical condition. Stand. J. Resp. Dis. Suppl. 66, 1968. STANI?K, V., et al. A contribution to the epidemiology of chronic bronchitis. Actu Med.

Scund.179,737-746 (1966). 31. BRINKhIAN, G. L., AND COATES,

E. 0. The prevalence of chronic bronchitis in an industrial population. Amer. Reo. Resp. Dis. 86, 47-54 ( 1962). 32. BRINKILIAX, G. L., AND COATES. E. 0. The effect of bronchitis, smoking, and occupation on ventilation. Am. Reu. Resp. Dis. 87, 684-693 (1963). 33. WYNDER, E. L., LESSON, F. R., AND MANTEL, N. Epidemiology of persistent cough. Amer. Rev. Req. Dis. 91, 679-700(1965). 34. ANDERSON, 11. O., AND LARSEN, A. A. The incidence of illness among young children in two communities of different air quality: a pilot study. Can. A4ed. Ass. J. 95, 893-904 (1966). 35. ANDERSON, D. O., FERRIS, B. G., JR., AND ZICKMASTEL, R. The Chilliwack Respiratory Survey, 1963. Part III. The prevalence of respiratory disease in a rural Canadian town. Can. Med. Ass. J. 92, 1007-1016 ( 1965). 36. ZEIDBERG, L. D., AND PRINDLE, R. A. The Nashville Air Pollution Study. II. Pulmonary anthracosis as an index of air pollution. Amer. J. Ptrb. He&h 53, 185-199 ( 1963). 37. ZEIDBEHG, L. D., PRINDLE, R. A., AND LANDAU, E. The Nashville Air Pollution Study. III. Morbidity in relation to air pollution. Amer. J. Pub. Health 54, 85-97 ( 1964). 38. CREGOHY, J. Occupational factors in the incidence of bronchitis. Trans. Ass. Ind. Med.

Ofj. 5,2-9 (1955). 39. HIGGINS, I. T. T.. et

al. Respiratory symptoms and pulmonary disability in an industrial town. Brit. Med. J. 2, YOP909 ( 1956). 40. OSWALD, N. C., HAROLD, J. T., ASD MARTIN, W. J. Clinical patterns of chronic bronchitis. Luncet 2, 639-643 ( 1953). il. FRY, J. Chronic bronchitis in general practice. Brit. Med. J. 1, 190-194 (1954). 42. MEADOWS, S. H. Social class migration and chronic bronchitis. Brit. J. Prev. Sot. Med.

15, 171-176 (1961). 43.

RANKIN,

44.

Med. ]. Amt. 1, 730-733 (1969). STEBBINGS, J. H., JR. Chronic respiratory

J. G.,

et al. Relationship

between

smoking

and

pulmonary

disease

in alcoholism.

disease among nonsmokers in Hagerstown, Maryland. I. Design of the study and prevalence of symptoms. Environ. Res. 4, 146-162 (1971). 45. STEUIXNGS, J. II., JH. Chronic respiratory disease among nonsmokers in Hagerstown, Maryland. II. Problems in the estimation of pulmonary function values in epidemiological surveys. Environ. Res. 4, 163-192(1971). 46. HOLLINGSHEAD, A. B., AND REDLICH, F. C. “Social Class and Mental Illness.” Wiley, New York, 1958. 47. SYME, S. L., HYMAN, M. M., AND ENTERLINE, P. E. Some social and cultural factors associated with the occurrence of coronary heart disease. .I. Chronic Dis. 17, 277-289

(1964). 48. WARNER,

W. L., MEEKER, M., AND EELLS, K. Occupational composition of social classes. In “Man, Work, and Society” (S. Nosow and W. H. Form, eds.). Basic Books, New York, 1962. 49. MILLER, D. R., AND SWANSON, C. E. “The Changing American Parent.” Wiley, New York, 1958. 50. JOHXSTON, J. “Econometric Methods.” McGraw-Hill, New York, 1963. 51. HINKLE, L. E., JR., et al. Occupation, education, and coronary heart disease. Science 161,

238-246 (1968).