Journal

of

Public]

Health
The vexing problem of air pollution and its effects is examined retrospectively in this study. The authors report on the findings of illness or lack of illness among those who were involved in the acute episode of air pollution in Donora, Pa., in 1948, and endeavor to interpret their findings.

A FOLLOW-UP OF DONORA TEN YEARS AFTER:

METHODOLOGY AND FINDINGS

Antonio Ciocco, D.Sc., F.A.P.H.A., and Donovan J. Thompson, Ph.D., F.A.P.H.A.

I NTEREST in the health aspects of air pollution continues to mount in this country as well as elsewhere throughout the world. The deleterious effects on health of the acute episodes of air pollution occurring in the Meuse Valley, London, and in Donora, Pa., are well documented,1-3 if not completely understood. The long range effects on the health of a population exposed intermittently to varying levels of air pollution are far less well documented. Moreover, the problems of measurement and the difficulties encountered in analyses, when one undertakes to elucidate the relations between continued exposure to air-borne pollutants and health in large population groups, suggest that the efforts and experience of many investigators will be required before these
FEBRUARY, 19961

relations will be understood. Accepting this point of view it seems appropriate to report the findings of analyses of data from follow-up studies undertaken in Donora and Pittsburgh.

Material and Method Detailed descriptions of the several surveys providing the data on which this report is based have been published elsewhere.3-5 A brief summary of the Donora survey will be helpful to an understanding of the material to be presented here. Following the acute episode of air pollution in Donora in October, 1948, the Public Health Service conducted a community survey on a systematic onethird sample of the households of
155

I)onora listed in the 1947 files of the Tax Assessor of the borough.3 Data were obtained from a responsible pers>on in the household on all members of the household concerning:
(a) such demographic characteristics as age, sex, occupation, residence; ( p) experience of specific coni(litions suich past as asthma, sinlUsitis, heart disease, chronlic b)ronchitis, tubercuilosis; (c the occturrence of syrn1)toms suich as cotugh, nausea, vomiting, headache, and smarting of the eyes during the October episode; (d) the persistence of these symptoms; (e) physician and hospital services received for the ailments.

under "chronic illness." An "all illness" classification has been defined to indlude persons reporting "acute illness" and/or "chronic illness" in 1948.

By this survey information was obtained on 4.092 residents of Donora. This group constitutes the part of the 19418 Donora population which was again surveyed early in 1957. By personal interview and mailed questionnaires, slightly more than 99 per cent of the 4,092 persons were traced to 1957. In the follow-up, information was obtained on the 1957 prevalence and 1948-1957 incidence of certain clhronic conditions as well as on certain cardiorespiratory symptoms, hospitalization and medical services received, place and cause of death, residence, occupation, smoking history, and other related information. It should be emphasized that the 1957 interviewers were not aware of the responses to health questionls which had been made in 1948. In the 1948 survey of Donora the main question asked was: "Were you affected by the smog of October 28-31?" If the answer was "Yes," the persons involved were placed in the class: "acute illness." Those for whom the answer was "No" were classified under "no acute illness." The information on past experience obtained in 1948 permits additional classifications of the illness status of the study group for 1948. Persons reporting a history of asthma, heart disease, chronic bronchitis, tuberctilosis, or sinusitis have been classified
156

1, mortality dulring the eighlt aIl(l one-lhalf years following the episode among- those who reported "acute illniess" in the Public Healthl Service survey is compared with the mortality among those who did not become ill. The age adjusted cumulative rates* summarize in a single figure the relationship consistently observed for each of five broad age groups investigated. As may be observed from the figure, persons who became acutely ill during the episode differ from those who did not become ill in their subsequent mortality experience. The difference appears in the first year following the episode and is maintained throughotit the period of follow-up. There is a stiggestion in the figure that the differenitial rate of mortality in the two groups begins to disappear about midway in the follow-up period. A similar differentiation of the acute illness and no acute illness groups is possible with respect to indexes of morbidity. D)ata were obtained in 1957 on heart disease, asthma, high blood pressure, hardening of the arteries, chronic bronchitis, chronic sinus trouble, anid kidney trouble. Table 1 includes results which are typical of our findings concerning these diseases. In the uppermost portion of the table, it will be noted that reported heart
* If Puj is the cumulative age-specific mortality rate to year j for males of age i who reported acute illness and ni the number of males in age group i in the total 1948 sample, the age adjusted cumulative rate to year j for acutely ill males has been computed as MPiindlni with the summation taken over the five age groups considered. Similarly, the

Findings Ill Figulre

rate for males with no acute illness and both female rates were adjusted to the age distribution of the total male and female samples, respectively.

VOL. 51, NO. 2, A.J.P.H.

A FOLLOW-UP OF DONORA, PA.

disease prevalence in 1957 among surviving members of the 1948 study group is higher among those acutely ill in 1948 than for those with no acute illness, with a single exception. A similar relation holds for the 1957 prevalence of one or more of all the chronic conditions about which inquiry was madle, as may be noted in the second section of the table. In the third sectionI of Table 1 a summary measure of morbidity has been calculated. This measure combines a number of individual measurements of morbidity each showing, generally, the same contrasting pattern for the "acute illness-no acute illness" categories. From the total study group we have isolated those persons who:
1. died during the intervening eight and onehalf years since the episode; 2. reported in 1957 a disablement, one or more of the chronic diseases, and/or one or more of the symptoms about which inquiry was made, and/or;

3. are currently under a physician's care for any reason.

These persons are labeled Type S persons, and Table 1 shows the number of such persons per 1,000 population by age, sex, and illness status during the episode. In each age-sex category the ill group contains a higher proportion of Type S persons than the not ill group. If the severity of the acute illness (numbers in parentheses) is taken into consideration the differences, with one exception, are much larger. Finally, at the bottom of the table we have included the incidence of reported heart disease during the period 1948-1957. In calculating the incidence, the denominator includes only those persons who did not report heart disease in 1948, while the numerator is the sum of those persons who died during the interval from heart disease and those persons alive in 1957 who reported the disease. Without exception

13
1,2 0 l l 0 10

0~ 0 G1)
a)
4-

9 8 7
6

Acute Illness

'No Acute Illness

>L E
0

5 4 3 2

Acute Illness

6 1948 49 50 51 52- 53 54 5 56 57 1948 49 50 51 52 53 54 55 56 57 Figure 1-Age-Adjusted Cumulative Mortality Rates for Persons 20 Years of Age or Older; Donora, 1948-1957

FEBRUARY, 1961

157

Table 1-Morbidity per 1,000 Population from Specified Conditions Classified by Age, Sex, and Illness Status During the 1948 Episode; Donora, Pa.
Male
Age in 1948 (years) Acute Illness No Acute Illness

Female
Acute Illness No Acute Illness

Under 20 20-34 35-49 50-64 65 and over

Number of persons reporting heart disease in 1957 per 1,000 23 7 15

19 31 110 86
11 17 49

33

5 72 93 156

10 56 72 143

Under 20 20-34 35-49 50-64 65 and over

Number of persons reporting one or more of the conditions, heart disease, asthma, high blood pressure, hardening of the arteries, chronic bronchitis, chronic sinus trouble, or kidney trouble, in 1957, per 1,000 108 57 96 47 82 168 135 100 212 151 140 275 312 393 238 195 469 133 467 343

Under 20 20-34
35-49 50 and over

Number of Type S persons per 1,000* 345 (714) 200 (636)t 136 453 (529) 351 (294) 285 597 (639) 430 (696) 347 772 (892) 643 707 (833)

168 338 503 729

Number of persons not reporting heart disease in 1948 who have since died from this cause or report heart disease in 1957, per 1,000 Under 20 20-34 35-49 50-64 65 and over
*

15 19 40 139 373

11

23 128 172
as severe.

16 10 50 122 259

10 10 39 119 354

See text for definition. t Rate per 1,000 for those whose actite illness wvas classified

the acutely ill

to
or

greater

persons have rates equal than those not acutely ill

in 1948.

From the foregoing findings it is clear that the subsequent health of persons who became acutely ill during the episode differs from the group which did not become ill. However, before these facts are hastily interpreted a more searching analysis is required. The U. S. Public Health Service
158e

analysis of 1949 pointed out that a higher proportion of persons with prior cardiorespiratory illness became acutely ill during the episode than persons with no prior chronic illness of this kind. In Table 2 the mortality is analyzed taking into consideration both the prior chronic illness about which a direct query was made in 1948 and the acute response to the episode. There is not complete consistency withVOL. 51. NO. 2. A.J.P.H.

A FOLLOW-UP OF DONORA, PA.

in each age-sex subdivision of Table 2, but in general the rates for persons with acute illness only, line (b), are only slightly higher than those with neither acute or chronic illness, line (a). Furthermore, the persons reporting a prior history of chronic illness with or without acute illness during the episode have had the highest subsequent mortality, lines (c) and (d). Although the numbers on which these rates are based are in some instances quite small, the conclusion seems warranted that the previously noted differences in mortality between the acutely ill and the not acutely ill are sharply reduced if the comparison is limited to those persons reporting no prior chronic illness in 1948. It should be noted, however, that in the age groups 20-49, among both males and females who re-

ported chronic illness in 1948, those who became acutely ill during the episode have had a higher mortality than those who were not acutely ill, lines (c) and (d). The causes of death in the several illness groups are also of interest with respect to possible interpretations of previous findings. Death certificate information was obtained for 290 of the 314 deaths occurring in the study population. Table 3 presents the observed and expected deaths from specific causes among persons in the specified sickness classes. In calculating the expected mortality from specific causes of death, we applied to persons in each age group who reported a stated illness in 1948 the appropriate cause mortality rate of the total sample. Since the numbers become small the causes of death were

Table 2-Mortality Rates per 1,000 Popuulation, Donora, 1948-1957, by Illness Status in 1948, Age and Sex
Males
Age (years)
20-34

In 1948 Illness Reported

MIortality
Rate 7 8 114 28 30 41 48 94 53 153 171 323 265 212 455
400 579 679 510

Number 278 134 36 44 214 166 169 21 53 243

137 170 31 68 269 44 55 19 28 102

Females Mortality Number Rate

(a) (b) (c) (d) (e)

None No chronic-acute Chronic-no acute Chronic-acute All illness

286 167 44 48 259 164 142 33 58 233

129 123 19 53 195
44 50 13 21 84

7
12
42 15

35-49

(a) (b) (c) (d) (e)

None No chronic-acute Chronic-no acute Chronic-acute All illness None No chronic-acute Chronic-no acute Chronic-acute All illness None No chronic-acute Chronic-no acute Chronic-acute All illness

30 42 61 138 69
109 138 263 245 179

50-64

(a) (b) (c) (d) (e)

(a) (b) (c) (d) (e)

65 and over

432 480 538 381

464

FEBRUARY, 1961

159

Table 3-Relationship of Reported Illness to Stated Cause of Deaths in Persons 21 Years of Age and Over, Observed and Expected Deaths; Donora, 1948-1957
Cause of Death Cardiovascular Neoplasms (140-205)* Other (410-456) of Persons Ohserv ed Expected Observed Expected Observed Expected
Number

Illness

Reported in 1948
All illness

Sex

MI F
NI F

796 747
509 468

20 17
12 12
1 1

19.2 14.6
11.7 8.9

59 44
25 22

51.2 39.8
30.3 24.1

43 28
21 11

41.8 24.2
25.2 14.7

\Aute illness

only
Iteart disease
Asthma

Al F

46 72 16 29
177

1.7 1.8
1.8 0.8
1.1

12 13
12 4
16 13 18 20
if Death.

4.9 5.4
5.5 2.4
12.1 8.7

7 6

3.8 3.2

Nt F
it F
NI F

2 0
6

:3
1
12

1.1 1.4
9.7

Pneumonia an (I bronchitis
Not ill

146
584 585
to the

3 9 7

3.1
9.8 9.4
1List of (Catises

11
21 11

5.2
22.2

25.7 24.2

14.9

Numbers int parentheses refer

Intternationaiirl

into only, three categorie(s: nleoheart disease, and "other." Thedata show that a higher mortality thani expected occurs prinicipally in the case
grouiped

plasms,

of cardioN-ascular deaths amonig persons r epor-ted to have had heart disease. asthma, or pnteuimoiiia and bronchitis inl 1918. Observed mortality among per(

sons

rep)orting only

actute illness in

1948

is aettuallY letss than expectationi for car(liovascular and "other" deaths. A comparison of the current prevaltence of heart disease, asthma, and arthritis or rheumatism among survivors of the 1918 episode and three other poptulation groups is presented in Table 1. The iniformation for the 1957 resi(lents of Donora. for the Arsenal Health D)istrict of Pittsburgh. and for Butler Couity. Pa.. was obtained from surveys coniducted in these areas employing essentially ideintical questions and technics to those used in the Donora follow-up.
160

W;hile -ariations in the rates of thle three illnesses considered are presetlt in the fouir coluimns of the table. there is little evidence to stuggest that sturvivors of the 19-18 Donora episode differ from cuirrenit residents of the borouTJ(1918 residents diltuted by personis moxing inito the, borough since 1918) or from personis residing in the two othiler nearlby communtiities. as far as pr(e valeice of these (liseases is con(ern('d. The findings presented thus far havre demonstrated that differences int sud)sequient health experience do exist between persons who became acutely ill dlurin, the episode and those who did nlot. but. in general. this difference appears to be due principally to health conditions existing prior to the episode. A question of paramount interest is. "Did illness experience vary with expostire to pollutants?" Lacking a direct meastire of the expostire of each Donora
VOL. 51, NO. 2, A.J.P.H.

A FOLLOW-UP OF DONORA, PA.

perlson to air-borne pollutants before, during. or following the 1948 episode,

a search for indirect measures of exposure was undertaken. The period of residence in the Donora environment. the place of residence in the community, the occupation and place of work, and other similar variables suggest themselves as bases for classifications which might reasonably be expected to exhibit variation in illness rates. Our examiniation of this question, while admittedly incomplete and of necessity crude, has produced little that would suggest that variations in these environmental factors are associated with health effects which are measurable in terms of the availal)le data. Information was also obtained in 1957 concerning the smoking histories of the

survivors of the 1948 episode. Our investigation of the relation of "personal air pollution" caused by smoking to the illness reaction both during and subsequent to the 1948 episode may be useful as an illustration of the kinds of problems encountered. Among the persons surviving in 1957 who could be personally questioned concerning their smoking habits (approximately 80 per cent of the total study group), there is no evidence that those who smoked tobacco in any form prior to October, 1948, became ill during the episode at a higher rate than those who did not smoke. The 1948-1957 illness experienice of male smokers when contrasted with the experience of males who have never smoked, is in the direction of higher rates of illness for the

Table 4-Prevalence per 1,000 Population of Specified Conditions in Donora, Pittsburgh, and Butler County, Pa., by Age
l)isease or (ondition
Age 0-14 15-24 25-34 35-44 45-54 55-64 65 and over 0-14 15-24 25-34 35-44 45-54 55-64 65 and over

Survivors of 1948 1957 Donora Donora Survey* Sample

4 7 8 18 51 55
99

Arsenal Health Butler County, District, 1952 Pa., 1954

10 11 6 18 29 59 6 10 19 17 55 86
97

Heart disease

3 6 12 8 30 68
90

77
6 6 6 21 9 24
33

Asthma

7 12 15 13 24 20
51

9 11 15 13 19 43
44

19 12 11 17 19 54
29

Arthritis or rheumatism
35-44 45-54 55-64 65 and over
*

NMale Female
64 98 211
205

MIale Female
37 98 209
175

MIale Female
37 48 91
172

61 208 254

97 233 229
268

71 121 158
264

Male Female 39 85 129 70 130 204

213

270

315

Age

as of 1948.

FEBRUARY, 1961

161

Table 5-Illness Status of Nonsmoking Males 12 to 20 Years of Age in 1948, and Cumulative Frequency with Which the Smoking Habit Was Acquired, by Specified Age Between 1948 and 1957 Illness Status in 1948
Ill Not ill
Nonsmokers Who Began to Smoke Cigarettes by Specified Age (Cumulated rate per 100 nonsmokers in 1948) Number 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
44

100

0 0

0 0

6 11 19 25 30 39 46 50 57 62 62 64 72 77 0 4 9 16 20 27 33 36 40 44 48 51 51 58

group which has smoked. With few exceptions the males who became ill during the episode and who smoked before the episode have the highest rates, while the nonsmokers who did not become ill during the episode have the lowest rates. No consistent patterns could be discerned among the females. In general, the findings for the males are similar to the linkage between smoking and various kinds of illness reported from the many recent investigations of this association.-67 A simple interpretation of these findings for the males would be that personal air pollution arising from smoking contributes in an additive way to the subsequent illness differentials already demonstrated between the persons who became ill during the 1948 episode and those who did not. That such an interpretation may not be warranted can be demonstrated directly from these data. Two questions can be raised concerning this interpretation of the smoking-sickness association: First, since prior illness has been shown to be related to the subsequent mortality and morbidity differentials between the group who became acutely ill and those who did not become ill, what relation exists between smoking history and illness prior to 1948? Second, did the acquisition of the smoking habit differ according to past illness experience? Unfortunately, since the smoking question was asked only of persons
162

surviving in 1957, we cannot shed any light on the first question. With respect to the second question, the data in Table 5 relate illness status, as reported in 1948, to the subsequent smoking experience of a group of young males. Among the individuals interviewed in 1957, there are 144 males who were between the ages of 12 and 20 years in 1948 and who stated that they had not begun to smoke prior to the survey date in 1948. The cumulative frequency with which the smoking habit was acquired by specified age has been estimated separately for the males who became acutely ill in 1948 and those who were not reported ill at the time. Although the numbers in Table 5 are too small to warrant definitive conclusions, the calculations clearly indicate that males who became ill in 1948 have since acquired the smoking habit more frequently and at a more rapid rate than those who were not reported ill at the time. Similar calculations for young females showed no consistent pattern, probably because of the small number of this group who acquired the smoking habit in the period considered. These findings, together with those of Fisher8 and the considerations of Berkson,9 emphasize the need to investigate how the smoking habit is acquired, and by whom, in order to understand the meaning of the observed relationships between smoking and morbidity or mortality.
VOL. 51, NO. 2, A.J.P.H.

A FOLLOW-UP OF DONORA. PA.

Discussion The essential findings of this investigation are that persons who reported acute illness at the time of the smog episode have demonstrated subsequently higher mortality and prevalence of illness than the other persons living in the community at that time. Furthermore, persons who complained of more severe acute illness in 1948 demonstrate greater subsequent morbidity and mortality than persons with mild complaints. There is some evidence in these data that this greater morbidity and mortality is related to the cardiorespiratory system. When persons who reported chronic cardiorespiratory illness which antedated the 1948 episode are removed from the comparisons, the differentials in subsequent illness experience between the acutely ill and those who did not experience illness are narrowed considerably. Finally, no substantial or consistent relations between environmental variables and health experience could be demonstrated. In terms of the study of air pollution and its relationship to health and disease, these findings raise three central questions:
1. To what extent is the subsequent health experience of those persons who became ill during the episode related to complaints first arising during the episode or to complaints present before the episode? 2. To what extent is the greater morbidity and mortality a direct consequence of (a) a short-term massive exposure to air pollutants, (b) a continued exposure both before and after the episode, and (c) a combination of (a) and (b)? 3. To what extent are the manifestations of sickness specifically related to the particular pollutants in the Donora air?

In discussing the first question, the association reported in 1949 in the U. S. Public Health Service analysis between a prior history of cardiorespiratory conditions and acute illness during the episode continues into the period of follow-up considered here. Our data
FEBRUARY. 1961

show, however, that even the persons who denied a history of heart disease before October, 1949, but became ill during the episode, have had a higher subsequent morbidity rate than persons of the same age and sex who did not become ill. Further, though the difference is less marked, mortality in the former group is also slightly higher than in the latter. Taking these findings at face value, it would seem that subsequent health experience is related to complaints first arising during the 1948 episode. The basic assumption in this inference is that persons who denied a history of cardiorespiratory conditions actually did not have them, and conversely, that those who reported them did. Our experience in other surveys indicates that there is considerable under-reporting of impairments in interview surveys.10'11 If any significant amount of such understating of previous illness did exist in the survey of 1948, the meaning of the findings might be that the discomforts of the smog episode served only to bring out or emphasize already existing impairments. With respect to the second question, our analysis has attempted to construct classifications of environmental variables presumably related to exposure and to compare variation in these classifications with the illness history during and after the episode. From this analysis we had hoped to obtain an indication of the relative significance of the acute episode and the chronic exposure over the years. However, no clear relationship between illness and environmental variables has been established. The comparison of chronic disease prevalence among 1957 residents of Donora and survivors of the 1948 study group might reasonably have been expected to show a difference if the episode alone had any long-term effects on the subsequent health of those exposed. No such difference could be demonstrated. Among persons reporting a history of
163

prior cardiorespiratory illness, those between the ages 21-50 show a differential subsequent mortality depending upon their reaction to the smog episode. This finding is pertinent to both our first and second questions. In relation to question 1, the differential mortality in the age group 21-50, if real, would indicate that the reaction to the smog episode played a role even among those whose impairment antedated the episode. This role may have been one of aggravating the existing cardiorespiratory condition or simply one of labeling those cases which were the more severe or disabling. The failure to find a mortality difference in the age group 50 and older between the group who became ill during the episode and those who did not, again if real, is disturbing. If either the "labeling" or the "aggravating" hypothesis is correct, we would have expected to find a mortality differential in this age group. That we did not find a differential may be a chance occurrence, or it may be that our time period is long enough to give competitive causes of death a chance to operate and thus obscure the picture. In regard to the third question, although the Donora data do point to some effects on the cardiorespiratory system, they do not provide the means for relating particular pollutants to specific symptoms. Without such specific information it is difficult, if not impossible, to distinguish between those persons whose disease conditions conceivably could be due to exposure to air pollution from persons whose disease

conditions are due to other factors. In this same connection, the lack of knowledge as to what syndrome(s) appears most pertinent for air-pollution health studies is a vexing problem for all researchers in this area. Therefore. although we know that an answer to this question is vital for the understanding of not only the Donora experience, but also of the relationship of air pollution to health in general, we find no way in which the analysis of the data now in our possession can provide an answer.
REFERENCES
1. Roholm, K. Oit the Cause of the Fog Catastrophe in the Valley of the Mettse, December, 1930. Hosbtitalti(i. 79, 1936. 2. Logan, W. P. D. Fog attd Mortality. Lancet 256:78, 1949. 3. Schrenk, H. H.; Heittiaitn, H.; Clayton, G. D.; and Cafafer, W. M. Air Pollution in Donora,
4.

Pennsylvania. Putb. Health Bull. No. 306, 1949. Thompson, D. J. Mortality, 1948-1957, and Morbidity, 1957, Amottg Persons Residing in Donora, Pennsylvania, During the Smog Episode of October, 1948. Mimeographed Report to Air Pollution Medical Program, PHS. 5. Thompson, D. J., an(d Ciocco, A. Sickness, Change
of Residence and Death. Brit. J. Prev. & Social Med. 12:172, 1958. Doll, R., and Hill, A. B. Lung Cancer and Other Causes of Death in Relation to Smoking: A Second Report on the Mortality of British Doctors. Brit. M. J. 2:1071, 1956. Hammond, E. C., and Horn, D. Smoking and Death Rates-Report on Forty-Four Months of Followv-lp of 187,783 Men. J.A.M.A. 166:1159, 1958. Fisher, R. Lttng Cancer and Cigarettes? Natuire 182:108, 1958. Berkson, J. Smoking and Lung Cancer: Some Observations on Two Recent Reports. J. Am. Statist. A. 53:28, 1938. Tauber; J., and Thompson, D. J. The Prevalence of Diseases of the Heart in an Urban Center as Estimated from Medical Examination of a Probability Sample. J. Chronic Dis. 6:595, 1957. Thompson, D. J., and Tauber, J. Household Survey, Individual Interview and Clinical Examination to Determine Prevalence of Heart Disease. A.J.P.H. 47:1131, 1957.

6.

7.

8.
9.

10.

11.

Pollution Medical Program, Public Health Service, U. S. Department of Health, Education, and Welfare.

This paper was presented before a Joint Session of the Conference of Municipal Public Health Engineers, the Conference of State Sanitary Engineers, and the Engineering and Sanitation, Occupational Health, and Statistics Sections of the American Public Health Association at the Eighty-Seventh Annual Meeting in Atlantic City, N. J., October 22, 1959. This work was supported in part by a research contract with the Air

Drs. Ciocco and Thompson are associated with the Department of Biostatistics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pa.

164

VOL. 51, NO. 2, A.J.P.H.