ML20133C790

TMI Nuclear Accident & Effect on Surrounding Population. Reprints of Medical Journal Excerpts Encl

ML20133C790
Person / Time
Site:	Three Mile Island
Issue date:	01/31/1984
From:	Tokuhata G PENNSYLVANIA, COMMONWEALTH OF
To:
Shared Package
ML20132B167	List: ML20132B163 ML20133C709 ML20133C724 ML20133C733 ML20133C747 ML20133C751 ML20133C767 ML20133C782 ML20133C790 ML20133D391 ML20133D591 ML20133D652 ML20133E380
References
FOIA-85-285 NUDOCS 8507200567
Download: ML20133C790 (39)

Text

, . - ._

_ e ... \ -. 00* Q ' ' ,h y THREE MILE ISLAND NUCLEAR ACCIDENT AND ITS EFFECT ON THE SURROUNDING POPULATION t

I George K. Tokuhata, Dr.P.H., Ph.D.

Director.

Division of Epidemiology Research Pennsylvania Department of Health and Professor of Epidemiology and Biostatistics Graduate School of Public Health The University of Pittsburgh January, 1984 8507200567 PDR 850524 FOIA LODOE85-PR5 PDR

r. \-

THREE f1ILE ISLAND NUCLEAR ACCIDENT AND ITS EFFECTS ON THE SURROUNDING POPULATION

Introduction:

On the morning of ifarch 28, 1979, a series of "unlikely events" at the Three Mile Island (TMI) nuclear plant led to a loss-of-coolant accident which became the most serious accident yet to occur in commercial nuclear power generation (1). For several hours after the reactor first tripped, the reactor core was allowed to overheat. Up to 10 million curies of rad-ioactivity have been estimated to have escaped into the atmosphere during a tense week of worldwide concern over the fate of the nuclear plant and its surrounding population (2-3).

The maximum possible dose to a hypothetical person standing unprotected anywhere along the border of the plant site for the duration of the accident was estimated as no more than 100 millirems (4), the approximate equivalent of one year natural background radiation in the area. The average likely dose to persons living within 5 miles of the plant was estimated at 9 milli-rems (5). At these low doses of radiation, no major health effects on the exposed population can be expected. The long-term health effects from the THI radiation exposure to the more than 2,164,000 persons living within 50 miles of the plant at that time was projected as one excess cancer death over the lifetimes of these residents. The total number of excess health effects from TMI radiation, including all cases of cancer (fatal and nonfatal) and genetic ill health to all future generations, was estimated as two (4).

Despite these radiation estimates and learned opinions of several technical groups, including those from government, industry, national labor-atories and universities, substantial amount of anxiety was created and resul-tant apprehension remained in the area. The public questioned the validity of the estimated radiation dose to local residents and also the health risk

r. t-2-

from that dose. The apprehension was due, in part, to the fact that radi-ation is invisible and its effects potentially pernicious. It was felt that even nonlethal doses are capable of causing immediate detrimental effects, especially on the unborn and the very young, as well as latent cancers and other chronic conditions. Many local residents actually be-lieved that they received vary high doses of radiation and some of them

, in fact developed a " radiation syndrome," a form of iatrogenic disease.

Health authorities in both the Commonwealth of Pennsylvania and the Federal government agreed that, because of the confusion and uncertainty surrounding the TMI accident from the beginning and because the nuclear accident was the first of its kind, the exposed population should be fol-lowed and studied for many years in order to monitor any possible changes in health status (6-7). Also, because of the high levels of psychological distress experienced by the local residents during the crisis period and the likelihood of distress continuing over the many years needed to clean up the damaged reactor, psychological health and its sequelae were perceived as im-portant outcomes to monitor independently of the issue of radiation exposure.

Psychological Stress and Health:

Stress is an organismic state that can contribute, under the proper circumstances, to changes in body function, which, if intense or chronic, may lead to disease. In other words, stress can trigger a multiplicity of organ-ismic reactions, some of which may contribute to illness, while others may result in normal adaptive responses.

Psychophysiological studies (8) indicate that life situations which threaten the security of the individual would evoke attempts at adaptive be-havior and also evoke significant alterations in the function of most bodily tissues, organs and systems. These physiological changes, in turn, will lead to a lowering of the body's resistance to disease. It is assumed that certain

is 1*

events require more intense and prolonged coping efforts than do others.

The greater the strains on the coping mechanisms, the more likely that an inadequate or inappropriate response will be utilized, thus eliciting idiosyncratic or pathological physiological reactions.

It is also important to recognize that an understanding of a life event's impact must take into account the physical susceptibility of the individual, t.ne meaning of an event, the person's ability to cope with a variety of stresses and the individual's social support network. With the exception of extreme and sudden life-threatening situations, no raw stimulus is a universal stressor. The true consequences of stress arise from the manner in which the organism responds to the presumed danger. It is the way in which the organism handles perceived stressors - the defenses it mobilizes and the alarm reactions ignited - that constitutes the true nature of the stress (9).

The psychosomatic approach (10), on the other hand, identifies certain personality type and life history that would make them more vulnerable to certain diseases. Whenever a stimulus is perc'.ived to threaten a fundamental human need, the stress response also will be inititated. Imagination can ,

produce its own stressors and prompt a neuroendocrine-autonomic response that itself poses a real threat to the organism.

Stress can cause disease by lowering or exaggerating the immune response (11), creating endocrine problems through either hypoactivity or hyperactivity (12), altering the balance of autonomic control, resulting in changes in the cardiovascular, respiratory, secretory, and visceral system (13), altering sleep patterns, with attendant impact on protein metabolism, hormone secre-tion and other vegetative functions (14), and by affecting the functions of the brain itself, which can have profound impact upon health through a variety of mechanisms, including changes in eating and health habits, such as exercise,

A

?* \*

drug, alcohol or cigarette consumption (15). Numerous studies have shown that the pituitary-adrenal axis may be activated or inhibited by fear, anger, rage, pain or adverse environmental conditions.

The stress associated with the TMI nuclear accident cannot be considered as a single unique experience because the prolonged recovery period following the accident gives rise to numerous additional stressors. However, it is unlikely that any given psychological stressor will be etiologically specific for any given disease. The important point is that a range of health outcomes, both mental and physical, need to be assessed in studies of stress or disaster since certain individuals may be more susceptible to health sequelae than others.

There are a number of studies in humans which have found an association between prenatal anxiety / stress and gestational, perinatal, and developmental pathology including complications of pregnancy (16-17) and infant growth anddevelopment(18). These findings suggest a number of practical and scientific questions to be addressed within the context of the TMI Health Effects Research Program. The first is whether or not the local-population, including pregnant women, as a whole experienced any detectable stress effects. The second question concerns factors which render individual preg-nant women particularly vulnerable to stress effects. As reviewed earlier, stress may be associated with morbidity only in the absence of supportive interpersonal relations.

While the specific mechanism of stress induced morbidity is not yet fully understood, there may be several different explanations with respect to pregnancy outcome; e.g., stress-anxiety induced changes (a) in maternal behavior, such as increased smoking, drinking or medication while pregnant, (b) in obstetric practice, such as increased prescription of analgesics and psychotropic drugs or use of special diagnostic procedures, (c) in maternal-

.. i infant bonding and child-rearing practices, and (d) in the hypothalamic-adrenocortical mechanism (19).

Subsequent to the TMI nuclear accident, the Pennsylvania Department of Health developed a comprehensive plan for a variety of epidemiologic and sociological studies designed to assess the impact, both immediate and long-term, of the accident upon the local population. Some of the short-term studies are still in progress while long-term followup studies are being planned. Investigators of other organizations and institutions have also conducted short-term studies. In this paper, findings from certain major studies are briefly summarized in three categories: (o) psycho-behavioral studies,(b) physical health studies, and (c) long-term epidemiologic surveillance.

Psycho-Behavirral Studies:

Although no imediate radiation health effects were recognized during the nuclear accident, and probably no delayed or late radiation health effects are to be expected, what emerged from this experience was that the major health effect of the accident appears to have been on the mental health of the people living in the region of TMI and of the workers at the nuclear power plant.

There was immediate mental distress produced by the accident among cer-tain groups of the general population living within 20 miles of Three Mile Island (20). The highest levels of distress were found among adults living within 5 miles of TMI, or those with preschool children; and among teenagers living within 5 miles of TMI, those with preschel siblings, or whose families left the area. Workers at the TMI nuclear plant experienced more distress than workers at the Peach Bottom nuclear plant in Pennsylvania which was studied for comparison purposes. The level of distress was higher among the nonsupervisory employees and stress continued in the months following the

accident.

Health-related behavioral studies conducted by the Pennsylvania De-partment of Health in collaboration with the Hershey Medical Center (21) indicated that persons who are younger, more educated, married and female were especially distressed during the crisis. The greater responsiveness of younger, married persons was probably due to their concerns about the effects of radiation on their present and future children and, since radia-tion effects often have a long latency, concerns about their own future health. .However, these demographic variables did not relate to changes in the level of distress over time. People who actively coped had high distress during the crisis and tended to maintain that distress over time. Persons with poor menta' or physical health tended to have high distress scores and to maintain their distress over time.

The number of persons with severe distress dropped shortly after the crisis, but between 10% and 20% of local residents residing close to TMI remained distressed nine months after the crisis. Persons residing close to TMI used more alcohol, tobacco, sleeping pills and tranquilizers during the two week period immediately following the crisis than before, but the ,

use of these substances which were mediated through coping with the crisis situation did not persist beyond that time.

The October 1980 survey conducted by the Pennsylvania Department of Health in collaboration with the Hershey Medical Center indicated that the l

l level of anxiety and stress declined more among residents within 5 miles of

! TMI than among those living more than 40 miles away. Thus, 18 months after the accident, the previously significant differences in stress-related

, symptoms, both behavioral and somatic which existed between the close and the far groups were no longer present. However, differences still persisted f

through October,1980 as far as perceived threat of TMI and attribution of the recognized symptoms to TMI were concerned.

. o.

An in-depth epidemiologic study of psychological impact in a more psychiatric context was conducted by Bromet at the Western Psychiatric institute (22). Her study covered three selected "high risk" groups in che THI area, namely, (a) TMIemployees,(b) mothers with preschool children, and (c) mental health clinic patients. People residing around the undamaged nuclear plant at Shippingport in western Pennsylvania were used as controls for comparison. One year after the accident, the condition of psychiatric outpatients near TMI did not differ significantly from that of counterpart in the control group. She also found that TMI workers ex-perienced only slightly higher rates of clinical depression and anxiety as compared with Shippingport workers. But, mothers of preschool children living within 5 miles of TMI suffered far more anxiety and depression than did mothers living near Shippingport.

Bromet also found that mothers who evacuated in the height of the accident had more distress one year later than mothers who did not evacu-ate. Mothers living within 5 miles of TMI reported more distress symptoms than mothers living farther away from the plant. It was concluded that manifestations of clinical levels of mental health effects occurred pri-marily during the 2-month period after the accident, but sub-clinical levels of symptomatology were elevated as late as one year following the accident.

There was evidence that social support bore an important relationship to these symptoms. Bromet's findings support a view that the burden of the stress was determined more by the actual experience, such as actual living in the vicinity of TMI, rather than by the perception of the stressful situ-ation.

Related to the psychological stress caused by the TMI accident was crisis evacuation during the accident by local residents. Although the level of radiation exposure was minimal, a substantial number of residents in the

vicinity of the TMI plant left the area primarily because of their per-ception of imminent danger associated with radiation. The Governor of Pennsylvania advised pregnant women and small children to evacuate. Within hours of the Governor's advisory and with mounting media coverage of the accident, which was often confusing, mass evacuation occurred. Some 64%

of the population in the 5-mile area left their homes some time during the nuclear crisis. It is important to document individual evacuation as it can be related to estimating radiation exposure and the future health effects studies.

A total cross-sectional population census conducted by the State Health Department supported by the Federal Center for Disease Control and Bureau of the Census shortly after the accident within five miles of the plant revealed that evacuation behavior was related to several demographic variables.

Specifically, more younger people evacuated and for longer periods than older people. More females evacuated than males. The more educated and white collar workers evacuated somewhat more than the less educated and blue collar workers. The strongest predictor of evacuation was the presence of one or more preschool children in the household. Distance of residence from the ,

damaged plant was inversely correlated with the decision to evacuate. There were no major differences in the pattern of evacuation between medical per-sonnel and other residents in the same community, i.e. , nurses and young women behaved similarly while physicians and middle-age men were alike in their evacuation behavior.

Radiation Exposure and Health Risks:

Nuclear accidents, such as the 1979 episode at TMI, are potentially harmful to health if the amount of ionizing radiation absorbed by humans is substantially high. However, whether health is affected by exposure at the low levels characteristic of natural background radiation is a matter of con-

jecture. Observations at. higher radiation intensities have implied, but are difficult to measure, that the risk of certain health effects may be increased even at the lowest dose levels. These effects may include any one or combination of the following: (a) damage to genes and chromosomes, or mutagenic effects, (b) damage to the growth and development of the embryo and fetus, or teratogenic effects, and (c) damage to cells that increases the risk of their forming cancer, or carcinogenic effects (23-24).

However, since health effects of radiation at the levels of natural background cannot be distinguished individually from similar effects pro-duced by other causes, the effects of low-level radiaticn are estimated only by extraporation from observations at higher radiation doses and dose rates, based on tentative assumptions about the relevant dose-effect rela-tionships. In the present state of our knowledge, such estimates must be regarded as highly uncertain at best (25-26).

The accidental radiation received by people residing in the vicinity of Three Mile Island (TMI) came almost entirely from xenon-133 (half-life, 5.3 days), xenon-135 (half-life, 9.2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br />), and traces of radioactive iodine (principally iodine-131, half-life, 9.0 days), which escaped intermittently from the plant as gases (27-28). These radioactive gases followed prevailing winds and increased the level of ionizing radiation along their path. However, the increase was short-lived because xenon dispersed rapidly and because radioactive iodine was present only in barely detectable amounts. No release of long-lived fission products, such as strontium-90, cesium-137, and pluton-ium-239, was detected.

Based on the available measurements, it is estimated that the maximum cumulative whole-body gamma radiation dose to anyone off site was less than 100 mrem, that the average cumulative dose to those withi 10 miles of the plant was approximately 8 mrem, and that the average cumulative dose to those

within 50 miles of the plant was about 1.5 mrem. Because these estimates make no allowances for shielding, they are generally considered to repre-sent over-estimates (27-28). Additional exposure of the population came from the beta radiation dose to the skin and from the inhalation dose to the lung.

It is estimated that the total dose to the skin could have been much larger than the whole-body gamma dose by a factor of 3 to 4 if the protective effects of shelter and clothing are neglected (27). The inhala-tion dose is estimated to have constituted no more than 3% to 7% of the dose to the whole body.

The risk of cancer is generally assumed to be increased by low-level radiation, but it is clear from observations at intermediate-to-high dose levels that the risk may vary depending on the type of cancer in question, age at the time of irradiation, the quality of radiation, and other factors.

According to a linear, nonthreshold extrapclation model, with no allowance ft alogical repair at low doses and low dose rates, cancer risks are regarded by many experts as being likely to overestimate the risks of low-level radiation. For this ret:on, some experts prefer a linear-quadratic model, which yields risk estimates that tend to be 25%-50% smaller (29-30). .

If these risk coefficients are applied to the population of about 2.2 million people residing within 50 miles of Three Mile Island, they predict a lifetime risk of less than one extra fatal cancer and less than one extra nonfatal cancer.

It is generally assumed that irradiation can cause genetic damage in human germ cells that is transmissible to future generations in the form of various inherited diseases. It has been estimated that the incidence of genetic abnormalities in humans would be doubled by a dose of 20 rem - 200 rem (25-26) and, that the number of descendants of the population within 50 miles of TMI who are likely to be affected by genetic disorders resulting from

the TMI accident would be approximately one.

The risks of teratogenic effects of radiation on the human embryo and fetus are more difficult to estimate, owing to the paucity of relevant data.

The evidence at hand implies, however, that the risks of such effects are smaller per unit dose than are the risks of carcinogenic and mutagenic effects (25-26). On this basis, it may be inferred that such effects are unlikely to result from the TMI accident in view of the small magnitude of the radiation dose.

Physical Health Studies:

Although increased risks of cancer, birth defects, and genetic abnorm-alties are potential long-term consequences of low-level irradiation, few if any such effects of the TMI accident are likely to be observed, because the collective dose of radiation received by the population within a 50-mile radius of the plant was so small.

In order to evaluate the potential effect of radiation and/or acute stress upon the reproductive process, an epidemiologic study was conducted to determine whether the incidence of spontaneous abortion was greater than expected near the Three tiile Island nuclear plant during the months following the March 28, 1979 accident. All persons including those who were pregnant living within five miles of TMI were registered shortly after the accident, and information on pregnancy at the time of the accident was collected.

After one year, all pregnancy cases were followed up and outcomes ascertained.

Using the life table method, it was found that, given pregnancies after four completed weeks of gestation counting from the first day of the last menstrual period, the estimated incidence of spontaneous abortion (miscarriage before completion of 16 weeks of gestation) was 15.1 percent for women pregnant at the time of the Till accident. Combining spontaneous abortions and stillbirths (delivery of a dead fetus after 16 weeks of gestation), the estimated incidence l

. a.

was 16.1 percent for pregnancies after four completed weeks of gestation.

Both incidences are comparable to baseline studies of fetal loss, indi-cating that the effect of the TMI accident upon spontaneous abortion was negligible, if any.

The crisis at Three Mile Island presented a natural experiment in disaster response, although this disaster was substantively different from any before it. Not only was this the first'to involve a nuclear plant, but no one was bodily hurt, no property outside the nuclear facility was physically damaged and, it is generally believed, no excess deaths or illness will be detected as a result of the accident. Nevertheless, a disaster situation was experienced psychologically and emotionally by the nearby population.

A study was conducted by the Pennsylvania Department of Health to determine the effect of the 1979 nuclear accident at Three Mile Island on residential mobility and subsequent population composition. The entire population living within five miles of TMI was registered shortly after the accident and traced one year later to identify movers. The results of this analysis showed that the rate at which people moved remained the same the ,

year after the accident as before, and that approximately 15% of those who moved (changed address) gave TMI as the main reason for their decision to move. The study also found that those moving because of TMI had attributes highly associated with mobility in general. When those attributes were con-trolled in analysis, attitudes about TMI were virtually the same among movers and nonmovers. On the other hand, demographic characteristics of new people f

moving into the area were not different from those who had moved out. How-ever, attitudes about TMI were significantly more positive among the newly moved-in people than among the moved-out people.

l Probably the most important study developed shortly after the accident k

was to determine if the TMI nuclear accident has had any measurable impacts upon pregnancy outcome and infant health in the vicinity of the damaged nuclear reactor. -The embryo, the fetus and the infant are highly sensitive to environmental insults, such as ionizing radiation and maternal psycholog-ical stress, depending upon the severity or intensity of the insults, the mode of exposure, and the gestational / post-natal age at exposure.

A carefully designed retrospective coh' ort study of pregnancy outcome was initiated in August, 1979. This study covered all pregnant women re -

siding within a 10-mile radius of the TMI plant, who gave births between March 28, 1979 and March 27, 1980. This study cohort consisting of some 4,000 deliveries was compared with a control cohort of another 4,000 deliver-ies which took place during the immediately following one year period for women who also resided in the same 10-mile area communities.

Measures of adverse pregnancy outcome investigated were: fetal deaths (stillbirths with 16-week or more gestation including abortions after 16-week gestation), neonatal deaths (deaths within 28 days postpartum), hebdomadal deaths (deaths within seven days postpartum), perinatal deaths (combined measure of fetal and neonatal deaths), prematurity (gestation less than 37 ,

weeks), immaturity (birth weight 2,500 grams or less), congenital anomalies (one or more developmental defects observed at birth), and low Apgar score (less than seven at one minute of delivery).

Since there are numerous factors other than radiation and stress that are known or suspected to influence the course of pregnancy and fetal outcome, it is important that the influences of such factors be considered. Detailed data on these factors have been collected, inicuding maternal characteristics (sociodemographic, behavioral, and medical-obstetric histories), health care provider characteristics, and prenatal care attributes. The influences of all these factors were taken into account when maternal stress and/or radiation

exposure were related to any of the various pregnancy outcome measures under study.

11aternal stress during and immediately following the TMI accident has been measured by overt personal statements of " anxiety-fear" as experienced and reported by individual pregnant women, and by actual stress-coping patterns described, such as taking extra medications (tranquilizers, sleeping pills, anti-hypertensive preparations, etc.) because of anxiety and fear.

Maternal radiation exposure during the 10-day crisis following the nuclear accident has been estimated by the Department of Radiation Health, University of Pittsburgh. For this purpose, already documented, reliable thermoluminescent dosemetry (TLD) and other source data including time-de-pendent dose-rate distribution compiled by government and non-government agencies were used to estimate maximum possible and most likely doses, to each individual pregnant woman, of whole-body gamma, thyroid doses to the mother and the fetus as well as combined gamma and beta doses to the skin.

For estimating maximum possible doses the evacuation factor was not considered, but for determining most likely doses this factor was taken into account, i.e. , those who evacuated during the accident were assigned smaller doses depending upon when and how long evacuation took place on an individual basis.

When pregnancy outcome measures were compared between the exposed study cohort and the unexposed control cohort, no significant differences were noted for any of the various outcome measures under study indicating that the impact of the TMI nuclear accident upon pregnancy outcome was negligible, if any.

After adjusting for the influences of the many maternal and provider character-istics described earlier, the incidences of fetal and neonatal mortalities, congenital anomalies, prematurity, immaturity, and of 1ow Apgar score within the study cohort were not significantly different from those within the con-trol cohort.

A separate analysis of the comprehensive data by multivariate logistic analysis indicated that neither radiation exposure nor psychological stress as such was significantly correlated to the incidence of fetal-neonatal mortality, congenital anomalies, prematurity, imaturity or low Apgar score within the exposed study cohort.

It should be noted, however, that the excess medication taken by those '

pregnant women v .o were severely stressed during and/or shortly after the accident was significantly correlated to the incidence of low Apgar score which was measured at one minute postpartum, and to the incidence of imma-turity, i.e. , the risk of low birth weight. This was interpreted to mean that the one-minute Apgar scores among newborns were significantly influenced by maternal excess medication of tranquilizers, sedatives, and anti-hyper-tensives which was mediated through the accident-caused stress and anxiety.

Our data also indicated that the low Apgar score at 5 minutes postpartum was not significantly correlated to the same maternal excess medication while pregnant. This may suggest that the low Apgar score is a negative, but only a very short-term prognostic indicator with probably minimal clinical significance. However, the stress-mediated low birth weight can be a poten -

tially significant long term health effect which requires special attention.

Apart from the above observations on pregnancy outcome, there was one other potentially important observation to be made particularly with respect to the effect of radioactive iodine upon thyroid function among newborn in-fants. Since State Health Department initiated a statewide screening program for congenital hypothyroidism in mih1978, the available data were analyzed l

in relation to the March 28, 1979 nuclear accident.

During the March 28, 1979 - March 27, 1980 period, only one case of

! congenital hypothyroidism was identified within a ten-mile radius of TMI l

among approximately 4,000 newborn infants. This incidence rate is well within

s.

a normal range of expectation.

An apparent clustering of seven cases of congenital hypothyroidism reported in Lancaster County during 1979 presented serious interests among epidemiologists and was subjected to a special in-depth analysis and inves-tigation because of physical proximity of the county and timing of the TMI nuclear accident. From this investigation the following diagnostic and

~

epidemiologic features emerged: (a) One of the seven cases identified was reported prior to the TMI accident, thus cannot be related to the nuclear ~

accident. (b) One with severe multiple central nervous system anomalies was born three months after the accident; this case is unlikely to have been associated with the TMI accident because of the late gestation period of the fetus when exposed to the accident, and also of coexisting other developmental anomalies which are unlikely to be related to radiation. (c)

One case was of dysgenesis, representing one of discordant Amish twins, thus, non-supportive of the etiology secondary to radiation exposure. (d) One case of dyshormonogenesis from an Amish family where the condition (lack of enzyme to synthesize thyroxine) was inherited from the parents. (e) Another case of dysgenesis in whom the thyroid glands were displaced from the normal, position.

(f) For the remaining two cases thyroid scan,was not conducted, thus, exact diagnostic entity remains unknown.

Having completed detailed diagnostic analysis and epidemiologic assessment of all cases reported in Lancaster during 1979, it was concluded that reported cases of congenital hypothyroidism were not related to the TMI nuclear accident, i.e., these types of anomalies are not expected to have resulted from direct or indirect exposure of the fetus to radiciodine. This conclusion was also supported by an independent Hypothyroidism Investigative Committee organized by the State,iiealth Department, which included expertise in the fields of I

epidemiology, pediatric endocrinology, obstetrics, medical genetics, biostatis-tics, and radiation physics.

Apart from the incidence analysis described above, there was also an important biological consideration with respect to radiation in relation to congenital hypothyroidism.

First, after March 28 through December 31, 1979, no single case of con-genital hypothyroidism was reported in Dauphin, Cumberland, Perry, Northum-berland, Juniata, Snyder, Mifflin, and Union Counties, the areas downwind (N, NW, NNW) from the Three Mile Island during the first 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> of the accident, when probably the largest amount of radioactive releases took place, thus the largest amount of contamination including 1 131 ,

Second, the maximum combined (inhalation and ingestion) human thyroid dose of radioactive iodine in the vicinity of the TMI following the March 28, 1979 accident through April 1979 is estimated to be 7.5 mrad (Editorial:

Annals of Internal Medicine, Vol. 91, No. 3, September 1979). At least 1,000 times greater thyroid doses (i.e. , 7.5 rads) would be required to have signi-ficant acute damages to the thyroid glands; however, even at this dose level, many of the damaged cells may be repaired. Based on the experiences of the Marshallese exposed to fresh radioactive fallout and atomic bomb victims, it is considered likely that as much as 50 rads to 100 rads fetal thyroid doses.

would be necessary to cause irreversible tissue damages, such as congenital hypothyroidism and/or thyroid cancer. Acknowledging the fact that the fetal thyroid is much more sensitive to radiciodine than is the maternal thyroid (a conservative upper bound estimate is that the thyroid dose to a fetus may be as high as ten times the maternal thyroid dose), the maximum likely fetal thyroid dose of approximately 75 mrad and the maximum possible fetal thyroid dose of 190 mrad to 200 mrad in the vicinity of the damaged nuclear plant are still far too small to have caused congenital hypothyroidism.

In any epidemiological investigation of possible " cluster" of a disease or morbid condition, it is important to recognize the technical difficulty

m

. e-and methodological limitations associated with such investigations. It is the overall consistent pattern of observations that provi. des useful clues for conclusion, rather than a single isolated change or difference, which in most cases occurs without substantive epidmeiologic significance. This is particularly true when relatively small populations are being studied.

One may or may not find a " statistically significant" change, difference, or clustering in morbid rates in an area depending upon how such population is delineated geographically and/or temporally. It is equally important that investigators carefully exa:nine the observed relationships. and deter-mine if such relationships are consistent with the known biological theory or orientation, which is based on the previous studies and experiences.

Our conclusions regarding congenital hypothyroidism around the TitI nuclear plant have been based on both the overall pattern of epidemiologic observa-tions and in reference to existing scientific knowledge.

Long-Term Epidemiologic Surveillance

, TMI Population Registry:

Within three months after the March 1979 nuclear accident, a cross-sectional population census of some 36,000 persons living within 5 miles of .

the plant was undertaken jointly by state and federal governments (31).

The information collected through the census provided baseline data for fu-ture epidemiologic studies of possible health effects of the TMI accident.

The data base, known as the Tl1I Population Registry, is comprised of demo-graphic characteristics on each resident and a brief medical history of cancer diagnoses, thyroid disorders, prior radiation therapy and exposure to ionizing radiation on the job. Smoking histories were also included for teenagers and adults. In addition, each person's daily travel in and out of the 5-mile area during the 10-days after the accident was recorded so that TMI-related radiation doses could be estimated from the already documented time-place l

dependent radioactivity distribution in the area. After two months of data collection, the TMI Population Registry was considered to be 95 per-cent complete in coverage. For each resident included in the Registry, two radiation dose estimates (maximum possible and most itkely) were given with respect to wholebody gamma and thyroid tissue respectively. Living status and whereabout of the registrants are updated annually for future contacts.

TMI Mother-Child Registry:

Within five months following the TMI accident, a carefully designed retrospective cohort study of pregnancy outcome was initiated (32). This study included two separate cohorts, the exposed study group and the unex-posed control group, all residing within 10 miles from the damaged nuclear plant. In each group there were approximately 4,000 mother-child pairs which constitute the TMI Mother-Child Registry. For each registered pair, de-tailed information regarding maternal characteristics and perinatal character-istics of the index infant were recorded. For the exposed study pairs estimated radiation doses (wholebody gamma and thyroid tissue) and the proxy measure of maternal stress during and shortly after the accident were docu ,

mented on an individual basis, which can and will be related later to the various measures of possible long-term health effects. The TMI Mother-Child Registry includes 94% of all eligible cases of pregnancy in the area and provides the necessary baseline data for long-term epidemiologic studies.

Living status and whereabout of all registrants are updated annually in preparation for such studies.

Objectives of Long-Term Studies:

The aim of the TMI Health Effects Research Program is to provide factual information based on such studies which are epidemiologically sound and/or sociologically justified with respect to possible health effects of the TMI

accident upon local residence. Based on the available TMI radiation ex-posure data and from the previously reported epidemiologic studies of low dose radiation, major adverse health effects from the Tlil accident are not expected. Although this may provide assurance to many people at potential risk, the assurance is only as good as the radiation data itself, which has become a subject of debate. There is also a possibility that psychological stress from the accident and its aftermath, which has been well documented, will cause some adverse health effects among the TMI residents.

Although the effect of psychological stress is difficult to predict, these public health concerns should be addressed. We are taking a precaution-ary route by carefully documenting both the exposed population and its health experiences after the nuclear accident. The already established TMI popula-tion Registry and the TMI Mother-Child Registry will provide reliable data bases for long-term followup studies of the health effects (physical, psycho-logical and behavioral), if any, from the TMI nuclear accident for both the general population and for the special cohort of pregnant women and their in-utero exposed children. Causes of death and cancer diagnoses will be routinely ascertained by linkage to the State mortality and cancer incidence, fil' es . Data for other physical, psychological and behavioral health indices will be collected every five years, on the basis of a random sample through prospective followup surveys for both cohorts.

• Regardless of the results of a variety of short-term and long-term studies undertaken, the primary mission of the Till Health Research Program is to fullfill the need to respond to the much publicised, potentially im-portant public health concerns. Because of the uniqueness of the TMI nuclear accident, thus its historical significance, as well as the scientific need to document health effects of very low dose radiation in humans, the rare opportunity presented by the TMI nuclear accident should not be lost in the pur-suit of these important epidemiologic studies.

References

1. Kemeny, J.G. (Chairman): Report of the President's Commission on the Accident at Three Mile Island. U.S. Gov. Printing Office, Washington, D.C., 1979.

2. Nuclear Regulatory Commission Special Inquiry Group: Three Mile Island, A Report to the Commissioners and to the Public. Vol. 1. U.S. Nuclear Regulatory Commission, Washington, D.C.,1980.

3. Woodward, K.

Assessment of Offsite Radiation Doses from Three Mile Island Unit 2 Accident. Pickard, Lowe, and Garrick (Consultants to Metropolitan Edison Corp.), 1979 (TDR-TMI-116, revision 0, August 31).

4. Ad Hoc Population Dose Assessment Group: Population Dose and Health Impact of the Accident at the Three Mile Island Nuclear Station. U.S.

Govern. Printing Office, Washington, D.C.,1979.

5. Gur, D., et al. Radiation Dose Assignment to Individuals Residing Near the Three Mile Island Nuclear Station. Department of Radiation Health, University of Pittsburgh, PA 1983.

6. Committee on Federal Research in the Biological Effects of Ionizing Radiation: Followup Studies on Biological and Health Effects Resulting -

'from the Three Mile Island Nuclear Power Plant Accident of March 28, 1979.

NIH Pub. No. 79-2064. U.S. Dept. of Health, Education and Welfare, Washington, D.C., 1979.

7. Tokuhata, G.K. Three Mile Island Health Effects Research Program.

Proceedings of the 56th Meeting, Pennsylvania Adademy of Science, April, 1980.

8. Rahe, R., et al.

Social Stress and Illness Onset. J. of Psychosomatic Research 8,:35, 1964.

9. Zegans, L. Stress and the Development of Somatic Disorders. In:

Goldberger and Breznitz (eds.). Handbook of Stress: Theoretical and Clinical Aspects. New York: Free Press,1982.

10. Dunbar, H. Psychosomatic Diagnosis, New York: Hoeber,1954.

11. Stein, M. , Keller, S. , and Schleifer, S. The Hypothalamus and the Immune Response. In: Weiner, Hofer, and Stunkard (eds.) Brain, Behavior and Bodily Disease. New York: Raven, 1981.

12. Lipton, M. Behavioral Effects of Hypothalamic Polypeptide Hormones in Animals and Man. In: Sachar (ed.) Hormones, Behavior and Psychopathology. New York: Raven, 1976.

13. Lisander, B. Somato-autonomic Reactions and Their Higher Control.

In: Brooks, Koizumi, and Sato (eds.) Integrative Functions of the Autonomic Nervous System. New York: Elsevier,1979.

14. Weitzman, E., Boyar, R., Kapen, S., and Hellman, L. The Relationship of Sleep and Sleep Stages to Neurcendocrine Secretion and Biological Rhythms in Man. Recent Progress Hormene Research 31_:399, 1975.

15. Antelman, S. and Caggiula, A. No rephi nep h ri ne-dopami ne Interacti on s and Behavior. Science 195:646, 1977.

16. Nuckolls, K.B. Psychological Assets, Life Crisis and the Prognosis of Pregnancy. Amer. J. Epid. 9_5:431, 5 1972.

17. Morishima, H.0. The Influence of flaternal Psychological Stress on the .

Fetus. Amer. J. Obs. and Gyn. 131:286, 1978.

18. Barlow, S. M. Delay of Postnatal Growth and Development of Offspring Produced by flaternal Restraint Stress during Pregnancy in the Rat.

Teratology M:211, 1978.

19. Smith, D.J. Modification of Prenatal Stress Effects in Rats by Adrenalectomy, Dexamethasone, and Chlorpromazine. Physiology and Behavior H:461, 1975.

20. Dohrenwend, B.P., Dohrenwend, B.S., Kasl, S.V. and Warheit, G.J. " Technical Staff Analysis Report on Behavioral Effects" In: Report of the Public Health and Safety Task Force to the President's Commission on the Accident at Three Mile Island, U.S. Govern. Printing Office, Washington, D.C.,1979.

21. Houts', P.S., M, iller, R.W., Tokuhata, G.K. et al. Health-Related Be-havioral Impact of the Three Mile Island Nuclear Incidence. A Report Submitted to the TMI Advisory Panel on Health Research Studies, The Pennsylvania Department of Health. Part I (April 1980), II (Nov. 1980),

and III (May 1981).

^

22. yBromet, E. and Dunn, L ' Mental Health of Three Mile Island Residents.

Psychiatric Epidemiology Program, Western Psychiatric Institute and Clinic, University of Pittsburgh, 1982. 55 0\ G23-11 X'

23. Upton, A.C. Effects of Radiation on Man. Ann. Rev. Nucl. Sci. 18_:495f 1968.

24. Upton, A.C. Radiation Injury: Effects, Principles, and Perspectives, University of Chicago Press, Chicago, Ill. 1969.

25. National Academy of Sciences Advisory Committee on the Biological Effects of Ionizing Radiation. National Academy of Sciences, National Research Council, Washington, D.C., 1972.

26. United Nations Scientific Committee on the Effects of Atomic Radiation.

Sources and Effects of Ionizing Radiation. Report to the General Assembly. United Nations, New York, N.Y. ,1977. ,

27. Battist, L. , Buchanan, F. , Congel, H. , Peterson, C. , Nelson, M. , and Rosenstein, M. Population Dose and Health Impact of the Accident at Three Mile Island Nuclear Station. Preliminary Estimates for the Period March 28, 1979 through April 7,1979. U.S. Nuclear Regulatory Commission, Washington, D.C., 1979.

28. Gerusky, T. Three Mile Island: Assessment of Radiation Exposures L and Environmental Contamination. Ann. N.Y. Acad. Sci., 1981.

l 29. Upton, A. Radiobiological Effects of Low Doses: Implications for Radiological Protection. Radiat. Res. 71,:51, 1977.

l i

__ _ w

4_

30. National Academy of Sciences Advisory Comittee on the Biological

. Effects of. Ionizing Radiation. National Academy of Sciences, Washington, D.C., 1980.

31. Goldhaber, M.K.,'et al. The Three flile Island Population Registry.

Public Health Reports 98_(6):603,1983.

32. Tokuhata, G.K. Pregnancy Outcome Around Three Mile Island. Pennsylvania Department of Health, Division of Epidemiology Research, Harrisburg, Pennsylvania, 1981.

I

s .,

p. .a....  ; -

m1 2 F/ .s l so.e., = sm v.t a. w t nwt i,ss c:eaum sm.m Pwed m ihe U s A. Pusamon Prums Ltd.

, HEALTH-RELATED ECONOMIC COSTS OF THE THREE-MILE ISLAND ACCIDENTt TEH-WEI HU Institute for Policy Research and Evaluation. The Pennsylvania State University, University Park, PA 16802, U.S.A.

.- and

, KENNETH S. SLAYSMAN Depc.rtment of Economics, Providence College, RI 02906, U.S.A.

(Received 26 April 1983)

Ahmerset-On 28 March 1979, a nuclear power station at Three-Mile Island frMI) near Harrisburg.

Pennsylvania, had a major breakclown. Dunns the two. week period of the accident, about 150,000 residents were evacuated for reasons associated with safety and health. Many residents dunas and after the accident, regardless of whether they left or stayed, made mental and physical adjustments due to this accident. This paper is to estimate the economic costs incurred by individuals or commumties as a result of a change in physacal or mortal health status and/or a change in health care services due to the TMI accident. The findings indicate that strer. symptoms caused by the accident did affect the health-related behaviors of area residents.

Of the costs examined, the economic costs of work days lost and physician visits are the largest cost items.

There were some increases in consumption of alcohol, casarettes, and tranquilisers immadi=tely following the acendent.

t. Drraoo0CTION of the current study is to estimate the health-related On 28 March 1979, a nuclear power station at Three. economic costs of the accident. All these costs are Mile Isitad (TMI) near Herrtsburg, Pennsylvania, . estimated from survey data sources.

had a major breakdown. A meltdown rnd subsequent In this study, health-related economic costs are radioactive discharge from the nuclear plant would defmed as the economic costs incurred by individuals have had catastrophic consequences on the lives and or communities as a result of a change in physical or properties ofindividuals in the surrounding area. Dur- mental health status and/or change in health care ing the two-week period of the accident, many resi- services due to the TMI accident. Although changes in dents were vacated for safety and health reasons. agricultural production, business activities, the hous-Many industrial and business establishments were ing market, and tourism may have direct economic also closed. This accident had various efects on local consequenas in the area, changes in the physical and communities, as well as the country as a whole. The mental health status of the local populations will also nationwide impact on the energy industry has been produce economic consequences to their commu-widely pubiacized and is well known; however, perhaps nities. These fatter economic consequenas are: (1) the the most important efects of the accident were on the possible increase in demand for health care services residents of the area surrounding TMI. Possible efects which may increase health cast expenditures and utili-included an escalation in the cost of electricity, zation of health care resources, (2) the possible in.

channes in agricultural production, business activities, crease in raorbidity which may increase a worker's the housing market, the tourism industry, and the absenteeism or other forms of loss of worker's pro-physical and mental health status of the people in the ductivity, and (3) the possible increase in consumption area. ,

of alcohol, cigarettes, sleeping pills, and other tran-The Goverr:or's Office of the Commonwealth of quilizers during and after the accident. Estimation of

, Pennsylvania has issued a report which examines the these costs will provide policymakers with needed in-l socioeconomic impact of the accident [1]. This report formatica regarding household responses and the re-estimated a monetary loss of $7.7 m in the value of lated economic costs of these adjustments in the local

! production in manufacturing industries, 574.2 m in communities.

business sales i(non-manufacturing industries, and In Section 2, if e data sources used in the study will 50.25-50.50m in the agricultural sector for a total be described. Section 2 will provide an analysis of the monetary loss to industry of about $82 m. The efect efects of the TMI accident on physician visits, work l

on the housing market w.;re negligible [2]. The house- days lost, and alcohol and cigarette consumption.

hold economic coste, t,f evacuation were estimated at Concluding remarks are contained in Section 4.

' 56 m, excluding fl e $1.2 m insurance reimbursement f3]. These co6t sci nates do not include the possible 104TA DESCRIFrlON t

costs relating to changes in the physical and mental Two separate data collections were made in July

1 calth status of the people in the TMI area The focus 1979 by the Chilton Research Services of Radnor, l Pennsylvania, under aeparate contracts from The Pennsylvania State University and the Nuclear Regu-tThis study was funded by the Division of Epi. latory Commission. A telephone survey using random l demiological Research. Pennsylvania Department of digit dialing was employed, with no overlapping of Health, under service contract 641996. responden:s between the two surveys.

183

\m

! 1,

< s

184 T.-W. Hu and K. S. StAYsMAN The Pennsylvanis State University (PSU) data con- mean education of the heads of households is about tained 691 households; 5% of households within the high schoollevel. )

5-mile radius of TMI were randomly selected (there In these two samples, about 20% of the households I l

are 14,300 households in the 5-mile ring). Obviously, have an income ofless than $10,000. About 40% of the this was the most intensely afected area within the household have an income between 10,000 and TMI region. The data included responses to extensive $20,000, while about 20-25% of the households have questions about the social and psychological eNects of an income between 20,000 and $30,000. The charac-the accident, the costs of evacuation, and health care teristics of marital status and age of the heads of utilization patterns during the two-week period of the households are also very similar in these two samples.

accident and a two-week period of the accident and a Table I shows that according to the PSU data,60*/. ,

of households within the 5-mile ring evacuated during H two-week period in July 1979. The data contain re-spondents' files as well as information on the individu- the two-week TMI period. The NRC data show 16%

of households evacuated within the $5-mile ring; haw-als within the respondents' households. A follow-up ever, a detailed breakdown indicates that the evacua-f survey of 400 households within the PSU data set was ~

conducted in January 1980. tion rate was 63% within 0-5 miles, and declined to 49, The Nuclear Regulatory Commission (NRC) data 32 and 5% for 5-10 miles, Il-15 miles, and beyond contain 1503 households covering an area 0 -55 miles 15-miles, respectively [3]. Table 2 provides summary from TMI, with heavier sampling within the 15-mile information about the size of population and the per-  %

ring than beyond the 15-mile ring. These data include cent of evacuation within the 15-mile ring.  :

responses concerning the evacuation decision, evacua. ,,

tion costs, views on nuclear power plant installation, 3. HOUSrJIOt.D ECONOMIC CO6TS OF _

social and psychological effects of the axident, but no HEALM-RELATED BEHAVIOR information on health care utilization. The im- Regardless of whether a household decided to evac-portance of these data is the cross-sectional informa- uste or not to evacuate during the TMI accident, '

tion (variations in distance away from TMI) which household members in the TMI area may have felt allows comparisons to be made in terms of the evacua- tension, confusion, or stress due to the TMI accident.

tion decision, costs of evacuation, and the efects of Not all households experienced these mental phenom-the TMI accident on mental and psychological behav- ena, and even if they did, the stresses may not have led iors. A detailed discussion of the sampling design and to physical or behavioral responses. However, some weighting procedures is presented in a report prepared household members may have experienced physical symptoms or behaved differently as a result of the  ;-

by the Social Impact Research, Inc. [4].

In order to understand the sociodemographic char- TMI accident. Physical symptoms included stomach j acteristics of the samples in these two data sources, trouble, headache, abdominal pain, diarrhea, etc. Be-they are briefly discussed and compared in this section. havioral symptoms included insomnia, excessive Table I provides sociodemographic information on sweating, loss of appetite, irritability, etc.

the total sample of both data sets. As shown in Table The basic analytical framework for estimating the I, these two data sources are comparable, since they health-related economic costs of the TMI accident is are both randomly selected from the area.The average illustrated in Fig.1. As shown in Fig. I, the TMI household size is about three persons. During the accident as well as a set of socioeconomic and de-two-week TMI accident period, a little over 4% of the mographic factors may have affected the stress level of households in the area included pregnant women. The individuals in the TMI area. The change in stress Icvel Table 1. Selected sociodemographic characteris.ics of households. The Pennsylvania State University data (within the 5-mile ring) and the Nuclear Regulatory Commission Data (within the SS-mile ring) ' ,

P9U FRC Variables sample sempga sine of household (no. cf persong) 3.1 3.2 Ps esnant woman in the household (in percent) 4.s 4.3 Mean education of the heads of households (in years)1t.8 12.4 Occupations (in percent) 32.1 White collar 1s.1 s1. r 37.4 slue collar 24.2 30.2 other Family income before taxes (in percent) 23.3 less than $10,000 20.2 41.8 42.7

$10,000 - $19,999 21.2

$20,000 - $29.999 23.8 7.9 10.4

1. 30,000 and above 2.4 No answer 6.3 83.9 82.1 Male as the head of household (in percent) 43.s Mean ase of the heads of households (in years) 42.4 Marital statust 71.2 Married 74.3 14.7 14.8 Divorced, separated, or Widowed 14.0 single 11.0 Lef t the area durins two-week TML (in percent) 60.0 16.0 I

%9I sample sine

a. %

f. .E Health-related econosuc costs cf the Three-Ele Island accident 185

~

Tath 2. getsmased number of hama and population evacussed dunas the two week penod of TMI accident withis the l$-aule ring. 28 March-10 Apn! 10 Neusehold s reputatten I

reecent ortottaated No.3 g Estimated No.4 Males Ivacuated Universe Evacuattena Universe of evacuattens b5 63 14.300 9,000 4 ,470 26.100 4-13 49 40,161 19,700 124,500 41,000 11-15 32 72,262 72,262 209,560 67,000  ;

-- 1 fetal .. .. $1.s00 .. 154,100 I

Notes: obtained from t'e NRC data.

I This flaure was supplied by the renasylvania Department of Nealth, based on their TMI eensue.

this figure is obtained f ree the product of the percent of evacuation and the universe in the area. Estimated figures are rounted to hundreds.

'This figure is obtained f ree the product of the househo1J mise and the number er avecoated households.

l TM1 Accioent ]

, Socioeconomic

' Demographic ractors streat I I 4 .

4

+ + +

Change in Change in Change in Consumption health Care Productivity Alcohol, Cigarettes, services Tranquilisers I I v

l u. lth-selated Econcate costs l t

Fig. I Framework for estimating health-related economic costs of the TMI accident.

together with a set of socioeconomic and demographic As discussed earlier, the NRC data cover an area factors may, in turn, have induened (1) the demand 0-55 miles from the TMI facility. Distance from TMI for health care services (number of physician visits), is used as a control variable to test whether the acci.

(2) the charage in productivity (work days lost and dent increased the tension and stress of the TMl area slower pace of work), and (3) the changes in con- residents. Two forms of dependent variables are sumption of alcohol, cigarettes, and tranquilizers. The measured-the number of psychosomatic symptoms a resources expended for these three types c*r effects respondent reported having during the two-week in-constitute the health related econonne costs of the teniew penod and a dummy variable for those who TMI accident. had any orm of the behavioral symptoms. Three time-periods are used:(1) the two-week period aner the 3.1 Atletionships hetween TM/ medstress accident (23 March 1979);(2) a two-week period four To measure and test these possible eTects of the months aRer the accident (July 1979); and (3) a two TMI accident, one has to Arst establish the linkage week period ten months aAer the accident (January between the effects of the TMI acculent and the stress 1980). In addition to the TMI acculent itself, the level of area rendents. Once the effects of TMI 'on independent variables that snay affect a person's psy.

' stress is certain, a statistical model will be used to chosomatic and behavioral symptoms are: age, sex, examine the resultant stress on three types of health- education, marital status, occupation, income, and related behavior. A much detailed analysis on the pregnancy status. Few non white residents are in the relationship between the TMI accident and the psy- sample so the race variable is not included.

chological effects on the area residents has been exam. Since the dependent variables are expressed in ined by another study (5]. This section is simply to either discrete numbers or dichotomous (0-1) form, quantify the relationships between TMI and stress. tobit and probit techniques should be used to obtain

l 13 T. W. Hu and K. S. Saatsnam the maximum bkelihood estimation. This study used To examine the duration of these stress effects, the ordinary least squares for it is easy to estimate January 1930 survey data are used to examine the and the ht is easy to interpset. The drawbacks efects of stress level in March on stress level in July, of the ordiasey least-equares techmque are minimiaod and the efects of stress levels in March and July on because of the relatively large sample siae and stress level in January. Table 4 presents the regression sulEcient non-aeros in the dependent variable, results based on the PSU data within the 5-mile ring, Table 3 shows that during the TMI pened the since only this data set has the longitudinal informa-distance variable had the most statistacally signi6 cant tion on stress level. The results indicate that the efect on residents' psychosomatic and behavioral earlier stress level has a signincant eNect on later symptoms. The deleted distance category is 15 nules stress level, with the magnitude of these efects about beyond the TMI area. Residents within the 15-mile 0.30-0.50. In other words, one stress symptom in an ring were about 20*4 rnore hkely to have behavioral . early penod would contribute to about 0.30-0.50 symptoms than those beyond the 15-mile ring. The symptoms in the later penod. The numerical mas-efects on psychotomatic symptoms are very similar. nitudes in Tables 3 and 4 are used to calculate the The efects of TMI were still apparent during the July efect of stress level on health-related behaviors.

penod but to a lesser degree Residents within the Based on these findings, the analysis of the efects of 5-mile ring and those in the 10-15 mile ring showed stress on the health care services and health-related stronger stress symptoms than the reendents in the behaviors can be linked to the TMI accident.

5-10 mile ring. Enamination of other socio-demographic variables suggest that the higher the level of education, the lees the degree of strees, that 3.2 Relationships between stress and physician lisirs males felt less stress than females, and that pregnant The NRC survey data did not collect physician women had higher levels of stress. visit information. Therefore, the analysis of the eNect Table 3. Relationships between TMI acadent and stress levels 1 March July Yariables Psychesonatic sehavioral Psychosomatic sehavioral intercept .264*** .448*** .965*** .354***

(.115) (.064) (.169) (.045) a8e .013 . 013 . 01s .013

(.017) (.009) (.024) (.009) 0-5 miles .225*** . 38 5m .2188** .115 m

(.061) (.034) (.089) (.034) 6-10 alles .179*** .201*** .099 .036

(.055) (.031) (.080) (.030) 11-15 males .270*** .197*** .248*** .068***

(.054) (.030) (.080) ( . 030 )

married .010 .060 .012 .012

(.074) (.041) (.109) (.042) divorced, separated. . 021 .0J2 .00% .0004 widned (.085) (.0*8) (.126) (.048) blue collar . 081 .037 .060 . 014 '

(.060) (.034) (.089) ( .034 )

white collar .013 .011 . 017 .01"

(.053) (.029) (.078) (.010) [

education . 014 .014 . 080** .013

(.028) (,015) (.041) (.015) ramily stae .0007 .004 - .0458' . 005

(.016) (.009) (.024) (.009) incese .021 .020*** .018 . 018***

(.015) (.004) (.021) (.008) presnant .158* .005 441?** .06)

(.099) (.055) (.145) (.055) male .0111*** .110*** .225*** .065***

(.045) (.025) (.066) (.025) sample eine 1356 1356 1356 1356 2

n .038 .066- .038 .031 F statistic 4.06 7.31 4.07 3.34 isote t Values in the parentheses are standard errors of coefficients

• • • Indicates the 11 level of significance (one-tailed test)

• • Indicates the SI level of s18nificance (one-tailed test)

• Indleases the 10% level or s18atiscance (one-takled test)

Health-related econosuc costs of the Three-Mile Island accident 187 Table 4 Relationships between stress level and physician visits March and July 1979 March July tatercept . 052 . 062

(.051) (.063) male . 026 . 032

(.019) (.024)

+

ase .007 .012

(.007) (.009) presnant .541*** .253***

'* (.074) (.092)

. married .034 .039

(.032) (.040) separated, divorced, widowed . 026 .104

(.034) (.047) chronic ailments .0688** .012

(.020) (.025) faatly sise . 008 . 004

(.007) (.009) education .020 .013

(.013) (.014) income .0046 . 0003

(.011) (.014) behavioral symptoms .01s .069**

(.025) (.029)

1. of psychosomatic evaptoms .031*** . 036**

in March (.013) (.017) f of psychosomatic symptome .. .056***

in July (.013) work in the marker .016 .049*

(.023) (.028) sample size 639 639 R .132 .0911 r statistic 7.96 4.82 Notes Values in the parentheses are standard errors of coef ficients

• • • Indicates the is level of significance (ene. tailed teet)

L ** Indicates the 5s level of significance (one-tailed test)

• Indleates the los level et significance (oce-tailed test) of stress of physician visits rely solely on the PSU accident and January 1980. Since few respondents in data. the survey were admitted to the hospital, only the Residents in the TMI area may have sought med- physiciar. visits will be used as dependent variables.

ical help during the TMI accident for two possible In reality, people who are under mental stress are reasons. (1) Psychological and emotional tension may rarely admitted to the hospital; more often they seek

. have led people to consult with a doctor or health help from a physician. A demand function for physi-

! professional in order to reduce their anxiety or to cian visits usually includes income, pfice of a visit, gain medical opinions about the possible efects of insurance coverage, and a set of sociodemographic

- radiation. In the PSU survey, 73 out of 692 re- variables. In this study, information on the price of I spondents indicated that they did consult with doc. physician visits was not available. However, this is a l

tors and health professionals to make them less tense relatively small geographic area and there are only a f during the two-week period of the accident. (2) few physicians in the area. A separate physician fee Physical discomfort resulting from psychological and survey indicated that in this area fees for a routine i emotional disturbances during the TMI accident may visit range from 12 to 514, a rather narrow variation.

i have prompted a physician visit. It is diflicult to make Insurance coverage is approximated by the oc-accurate judgments as to wh;ch visits to physicians cupation of the respondent. In addition to age, sex, were due to the TMI accident even from the patient's educational level, and marital status, both psycho-point of view. Therefore, a model of the demand for somatic and behavioral symptoms are included in the physician services is used to measure the efects of equation. A self-valued health status of the re-stress on physician visits. The dependent variable is spondent and the existence of chronic ailments are the number of physician visits during the two-week included as state of health variables. In the survey, a period of the accident, the two-week period in July, question was asked about whether a respondent will and total number of physician visits between TMI usually consult with a physician when discomfort

188 T. ~!. Hu and K. S. SLAYsb4M occurs. This variable b included to re6ect the prefer- greater demand f:r physician visits. Also, t.s ex-ence for a doctor visit. pected, pregnant women and older persons visit Table 5 presents the regression results for total physicians more often than everyone else, number of physician visits between the TMI accident and January 1980. This equation includes more infor- 3.3 Relationship berween stress and work days lost mation about the respondents

• health status and in the January 1980 survey, respondents were preferena for physician visits. The results show that asked about the number of work days lost for health both behavioral symptoms and psychosomatic symp- reasons for the period from Labor Day 1979 to toms have statistically signif) cant positive efects en January 1980. On the average, individuals lost 3.2 physician visits. One additioaal behavioral symptom work days.

in the study period increases physician visits by 0.13, The factors that afect the number of work days while one additional psyclosomatic symptom in- lost include the status of health of the respondent, creases physician visits by 0.07. age, education, sex, marital status, pregnancy status, Table 5 also indicates that an obviously healthier preference for doctor visits, and status of labor person makes less physician visits, that individuals market participation. For the purpose of this in-with chronic ailments prefel to consult with physi- vestigation, behavioral and psychosomatic symptoms cians and make more physician visits. Pregnant were also included in the analysis. The regression women made four more visits than non-pregnant analysis will be able to introduce these confounding women during the period. Males made less physician variables as well as the stress variables, measured visits, and older persons made more physician visits. during and after the TMI accident. Thus, indirectly, Neither income nor education shows signi6 cant efect the efect of stress may measure the efect of the TMI on physician visits. It also appears that people with accident on work days lost, holding other socio-any one or more of the chronic aliments have a demographic variables the same.

Tabk 5, Relationship between stress and physician visits since TMI accident intercept 2.454***

(.836) health .56 7***

(.140) preference for physician .908***

visite (.122) male .350*

(.207) age .196***

(.079) married .286

(.385) separated, divorced, widowed .609

(.460) fasily size .099

(.068) education .090 '

( .137) incone .067

(.123) chronic ailments .512***

(.206) pregnant 4.259***

(.711) thehavioral symptoms .801***

(.215) total f of pavchoncearic .092**

svmotens since TM1 (.041) farmer, homemaker, vnemployed. .156 retired, other (.236) sample stae 179 R .411 F statistic 18.15 tie t e Values in the parentheses are standard errors of coefficients

• • • Indicates the is level of significance (one-tailed test)

• • Indicates the 5s level of significance (one-tailed test)

• Indicates the 101 level of signiricance (one-tailed test)

.. t Health-related economic costs of the Three-Mile Island accident 189 Tcbic 6 shows the regression results for the gumber individuals who w:rk ia the market may have fringe of work days lost. As in the demand for physician benents that compensate them for their work days visits, stress has a statistically signi6 cant positive lost so that there is an incentive to take time oE for efect on the number of work days lost. Additionally, health reasons.

the pregnancy variable had a greater effect on the Another aspect of the change in the productivity of number of work days lost. .

labor is the possible change in work habits in relation An interesting result of this analysis is that people to the TMI accident. Accordint *.o the January 1980 who have a preference for visiting a physician when follow-up survey,10 out of 404 respondents indicated they are ill also have a larger demand for time off having difficulty concentrating in the work place and work. This suggests that tirne off work and physician five reported working at a slower pace, for a total of visits are complements in the production of health. about 3.7% of the respondents. Table 7 presents the The alternative explanation would be that physician regression estimates for the relationship between the visits and time must be combined in fixed proportions behavioral and psychosomatic symptoms and change in order to produce health. Thus, a certain amount of work habits. Both variables suggest statistically of time must be used in conjunction with a physician significant efects on work habits.

visit to produce health.

Finally, individuals who work in the labor market 3.4 Relationship between stress and alcohol and ciga-have more work days lost than individuals who do rette connanprion not work in the labor market. There are two possible There is a tendency for people under stress or explanations. One is that individuals who do not anxiety to increase their consumption of alcohol or work in the labor market may always report zero cigarettes or to take tranquilizers or sleeping pills.

days lost from work for health reasons. Alternatively, According to the PSU survey,63 respondents indi.

Table 6. Relationships between stress and work days lost, September 1979-January 1980 intercept -2.317 (5.972) preference for doctor visits 2.1778**

(.471) market workers 5.443***

(1.686) health -1.433 (1.004) chronic ailments .778 (1.471) age .178

(.567) education .110

(.979) income .613

(.681) n.e l e -2.050 (1.480) marrted 1.319 (2.753) divorced, separated, widowed -1.524 (3.140) tamily sise 229

(.485) pregnant 21.090***

(5.084) behavtoral symptoms 1.215 (1.535) 8 of psychosomatic symptoms .951***

(March. July. January) (.290) sample sise )?9 R .1501 F statistic 4.59 l

heet Values in the parentheses are standard errors or coefficients

• Indicates the 1% level of significance (one-tailed test)

• • Indicates the 5s level of significance (one-tailed test)

• Indicates the ICI level of significance (one-tailed test)

.? ..

190 T.-W. Hu and K. S. 51avsw cated they drank more alcoholic beverages than usual respondents who took sleeping pills during the ques-during the accident period. This represents about tionnaire interview time (July). Similarly, there were 13.4% of the 469 regular consumers of alcoholic 60 respondents who took tranquilizers during the beverages. On the average, the regulars had four TMI accident compared to 32 respondents who took additional servings a day in any one day during the them during the questionnaire interview period.

period. In general,55*/ eof the people drank beer,15% These figures indicate that about twice as many drank wine, and 30*/e drank liquor. In the January people took sleeping pills or tranquilizers during the 1980 fouow up interview, however, only 7 out of 404 TMI accident as at other times. The consumption of respondents indicated that they increased their con- sleeping pills and tranquilizers in January 1980 was sumption of alcoholic beverages in relation to the about 22 and 31 out of 404 respondents, respectively.

TMI accident. Tables 8 and 9 provide the regression for the Within the 5-mile radius,43% of the respondents relationship between the consumption of sleeping were regular smokers, consuming about one pack (20 pills, tranquilizers, alcohol, and cigarettes and the cigarettes) a day. Thirty two percent of regular smok- stress levels of the TMI residents, during March 1979 ers indicated that they increased their smoking during and January 1980. Other sociodemographic variables the two-week TMI accident. These respondents are also included in the three equations. Table 8 smoked an additional ten cigarettes each day during indicates that both psychosomatic and behavioral the accident period. In the January 1980 foUow-up symptoms have statistically signiScant positive effects interview,32 out of 404 respondents indicated that on the taking of sleeping pills and smoking cigarettes they increased smoking in relation to the TM1 acci- during the two weeks after the sccident. In addition dent. the psychosomatic symptoms have significant posi-During the two-week period of the TMI accident, tive effects on the taking of tranquilizers and the 51 respondents took sleeping pills compared to 24 behavioral symptoms have effects on drinking. These Table 7. Relationslups between stress and work habits, January 1980 trouble concentrating Variables at Work Work slower latercept .045 .040

(.054) (.036) sen .0004 .019

(.017) (.011) age .009 .004

(.007) (.004) income .012 .012*

(.010) (.007) married .002 .009

(.032) (.021) i divorced, separated, widowed .015 .018 l

(.037) (.075) j- family saae .014*** .002

(.006) (.004) t educatson .004 .009 .

l (.071) (.008)

{ pregnant .012 .0006

(.063) (.04) white cellar .046 .04***

(.027) (.018) blue co1Lar .002 .304

(.021) (.01-)

see the iM1 plant iros work .016 .004

(.016) (.010) behavioral symptoma .013 .021*

(.018) (.012) psychosomatic symptoms .009 .001 l (.003)*** (.002) sample slae 360 380 R .06 .05 F statistic 1.82 1.63 Note: Values in the pare theses are standard errors of coefficients

! *** Indicates the is level er significance (one-tailed test) l ** Indicates the 5s level of alanificance (one-tailed test) j

• Indicates the los level of significance (one-tailed test) l

t Health-related econonuc costs of the Three. Mile Ist;nd accident 191 Table 8. Relationslup between stress and consumption of sleeping pills, tranquihrers. alcohol and cigarettes. March 1979 sleeping Pills f ranq uilise r s Alonel Cigarette, Variables (Dummy Variatle) (Dumv Variable) (a of servin m iin v bert inte rc ept .012 .010 . 16 ..t i

(.05) (.05) (.2 0) (2.3.e male .016 - 034* .1.. .Me t.02) s.022) .1377 #1.0$s age .01) .01.* - .-o**e .;,p ...

(.008) (.904) ..v.; s.39s income .0Wic) - .06 .Oda ,. 3 5

(.013) (.013) (.967 *.als married .0002 .01* .0 6 .; .

• 2 s.0.) (.030) t.173) e l.en divorced, separated, widowed .03 .075* .v67 2.32

(.04) (.0=.; (.206s t2.v;s iamily saae .009 .01e** .903 .;J t . %d) (.006) f.037) 4. 39) pregn4nt . 01. .12. .6 -e.6o**

i.06.) (.040s (..Os 43.93 waste .ollar .015 .016 .220 -1.07 4.J33) (.33*) (.191) 1.57;

>14e sallar .0J1 .006 .Os4 .e77 4.026e (.027; (.129) it.Ja) pse w .ce.at;. sympt ma . 012 * * .0et*** .060 . 0 =*=

t%s r t n> 4. ..) (.01): (.07) (. ave be t.a v io r a ; avmptems .072*** .0 0 .236** J.ss***

(Ma rc o s t . s2 8 ; ( 029) (.134) (1.30) sa ; . size 93) 93) 63) 63)

R 06 .112 .056 759 F statisti. ..?O 7.13 3.21 9.:3 18 ate: Values in the parentheses are standard errors of coefficients

• • • Indicates the 11 level of algnificance (one-tailed test)

• • Indicates the 51 level of significance (one-tailed test)

• Indicates the 10% level of significance (one-tailed test) same patterns of significance exist for the January the local market. Costs can be estimated for the period as well. except that the alcohol consumption period from Apsil 1979 to January 1980, based on the was not sigmfcantly affected by the stress lesci(see three interval time period. Table 10 provides the cost Table 9). estimates of the changes m health < elated behavior due to the TMI accident within the 5-mile ring, since 3.5 Estimates of health-related economic costs data on the health-related behaviors are limited to Empirical estimations of the effects of the TMI within the 5-mile ring.

accident on stress and the effects of stress on healt- Based on the results in Table 4. about 50'e of the related behaviors base been established in the pre- stress symptoms existing in January 1980 may have sious four sections. The estimation of economic costs been affected by the symptoms which existed in of these behaviors will rely on the magnitude of these March or July 1979. Therefore, if one takes 50*e of changes. The procedure used to estimate these costs the stress level in March 0.ll2 (0.255 2) for psycho-is as follows: somatic symptoms, and 0.240 (0.481/2) far behav-ioral symptoms, and multiplies by the coefficients for (A Stress level due to TMI) x (A Health related psychosomatic and behavioral symptoms in Table 5.

behavior due to stress) x (size of the population in the results indicate that the additional amount of a given area) x (unit cost of the health-related stress lesel due to the TMI accident increases physi.

behavior). cian visits by 0.4 [0.1125 x 0.084) + 0.240 x 0.128)]

for an average individual living 0-5 mi!cs from TMI Changes in stress level due to the TMI accident can during the ten month period. With 14.300 households be obtained from regression coefficients in Tables 3 in the 0-5 miles the total number of additional and 4. Changes in health-related behavior due to physician visits becomes $720 (14.300 x 0.4) for the stress can be obtained from regression coefficients in ten-month period. Given a cost of 515 per physician Tables 5-9. Size of the population of the 0-5 mile ring sisit, the total cost for the entire penod would be is presented in Table 2. Unit costs of the health- 385.800.

rehted behaviors are obtained from prevailing fees in Similarly, the effects of psychosomatic stress on the

192 T.-W. Hu and K. S. 51.AY$44AN Table 9. Relationslup betweer r*ress and consumpoon of sleeping pills, tranquilizers, clcohol f nd cigarettes January 1980 Variables Sleeping Pills Tranquilisers Alcohol Cigarettes (Dumme Vert *>le) (tsummy Verlable) (e of Servings) (* Packs) intercept . 114e . 114 .024 .132

(.069) (.08) (.037) (.084) male .014 .0 34 .019 .016

(.024) (.029) (.013) (.03) age .026*** .014 .006 .032***

- (.009) (.01) (.005) (.01) income .012 .016 .0008 .0*7***

(.015) (.017) (.C08) (.018) married .0) .026 .019 . 057 t.05) ( . 0 54 ) (.024) (.056) divorced, separated. .01 .077 .031 .026 sidowed (.05) (.061) (.028) (.065) family size .006 .016* .001 .007

(.008) (.009) (.004) (.01) education .004 .001 .009 .044.e.

(.016) (.01) (.009) (.02) pregnant .008 .068 .01 .09

(.09) (.099) (.05) (.11) white collar . 005 .026 .003 .039

(.04) (.044) (.02) (.047) blue collar .029 . 301 .001 .026

(.029) ( . 034 ) (.015) (.036) behavioral symptoms .051*** .037 .01 .079***

(.026) (.029) (.013) (.031) psychosomatic symptoms .017*** .019*** .003 .014***

(.005) (.005) (.003) (.006) sample else - 379 379 379 379 R .112 . 0a#4 .024 .112 F statistic 3.85 3.15 0.76 3.86 Bete Values in the parentheses are standard errors of coefficients

• • • Indicates the 11 level of significance (one-tai'ad teet)

• • Indicates the SI level of significance (one-talked test)

• Indicates the ICI level of signitteance (one-tatted test)

Table 10. Changes in hen:th-related behavior due tn TMI,0-5 mile rics two weeks after the TMI accident at d two weeks between July 197bhnuary 19R0 (additional qusetitws)

April 197/i July 1979-January 1980 Quantitya Costa b Cuantity

• Costab physician visite 290 $ 4.350 235 8 3,525 volk days lost (days) 8.870 . 266.100 8,597 257.100 sleeping pills (tablets) 298 36 153 18 tranquilisers (tablets) 802 9e 171 12 alcohol (servings) 560 336 -- --

cigarettes (pack) 1.900 950 1.753 876 Total 5271.668 3261.531 Notes: *rhe quantities are derived from the product of the three dis-tance regression coef ficient in Table 3 (March) in relation to psychosomatic stress and the regression coefficients in Tables 4. 6. and 8.

b ihe unit costs were $15 for each physician vis&t. 430 for each work day lost. 8.12 for each sleeping pill or tranquiliser, and 5.70 for one alcoholic beverase serving. end $.50 for one pack of cigarettes.

,, 1

. Health-related economic costs of the Three-Mile Island acculent 193 work days lost duries the September 19794anuary and psychosomatic symptoms of peopic in the area.

1980 pened was the 1.34. Statistical results indicate This study examines three types of health-related eco-that about 32% of the stress level in March could nomic costs of the TMI accident: (1) the increase in afect the stress status during the last quarter of the health care services,(2) the increase in work days lost, year, as shown in Table 4. Table 3 also shows that and (3) the increase in consumption of sleeping pills, there were 0.22 additional stress symptoms for 0-5 tranquilizers, alcohol, and cigarettes. Two data sets mile resulents due to the TMI accident. Thus, the are used to examine these efects-the NRC data and resultant number of work days lost for 14,300 house. the study survey (PSU data).

holds becomes 1349 days (1.34 x 0.32 x 0.22 x The findings indicate that stress symptoms caused 14,300). If one extends this estimate for the period of by the TMI accident did afect the health-related be-Apnl 1979 to August 1979 (an additional 5 months), haviors of area residents. Based on regression analysis, it would become about 2698 (1349 x 2) for the entire it is estimated that the cost of the changes in health-ten-month period after the accident. At 530 for each related behaviours was about 5178,419 for a period of work day lost, the total amount would be $80,940. 10 months within a 5-mile ring of TMI. Of the costs Similar estimation techniques are used for esti- examined, the economic costs of work days lost and mating the additional consumption of sleeping pills, physician visits are the largest cost items. The results tranquilizers, alcohol, and cigarettes. Actual quan- also show that there are significant efects of the stress tities are also shown in Table 10. The total cost of on the increased consumption of cigarettes, alcohol,

- sleeping pills, tranquilizers, alcohol, and cigarettes and tranquilizers after the accident.

are about 511,679.

The monetary valuation of the health-related eco-

! nomic costs of the TMI acculent indicates that the "" " "C""

increase in physician visits and work days lost are the I* C#ownor's 00lce, Pennsylvania. Report of the Gorcr-two largest health-related cost items. The total mon- ** C'""'*" on nree Mile Islauf. Commonwealth of Pennsylvania, Hamsburg, Philadelphia (1980).

etary costs for residents within the 5-mile ring are 2. J. P. Nelson, Three Mile Islad aM residential propeny about $178,419 for the ten-month period. Compared values: empirical analyes and policy implications. Work-to the economic costs of evacuation and the costs to ins Paper, Dept. of Econ., The Pennsylvania State Uni-the industrial and utility sector, the health-related versity, University Park,(tuly 1980).

economic costs appear rather trivial. However, from 3. T. W. Hu, Neal:A-Atlased Econonue Costs of the nree-the survey data, it was obvious that the TMI-related Mdc Isised Accident. Institute for Policy Research and stress was real and did exist. Thus, these monetary Evaluation, The ."-- A__5 tate University, Univer-

- cost estimates of the health-related behaviors pro- sity Park,(March 1980).

vided in this study may be conservative figures. 4. C. E Myan, nw Mac Isw Telephone hy, Pre-linunary Report on Proceares andFt sdings. Social impact 4, coNct.tlDtNG RDtARES Research, Inc. Seattle, Washington (Sept.1979).

5. P. S. Houts, R. Miller, G. K. Tokuhata and K. S. Ham, This study has hypothesired that the accident at - Neofth.Actored Acharierallspect of the Arce Mdc Is-TMI afected area residents' mental status, if not their lauf Nuclear Incident. Hershey Medical Center Hershey, physical status. Mental status refers to increased stress Philadelphia (Nov.1980).

l-

[cs48)s Aarunraar og og n p mum mi'l(?

aprint Utilization of Medical Care following the Three Mile Island Crisis PETER S. HouTs, PHD, TEH WEl Hu, PHD, RUGH A HENDERSON, MD, MPH, PAut. D. CLEARY, PHD, AND GEORGE TOKUHATA, DRPH, PHD Abstract. Four studies are reported on how utilization of indicated only slight increases in utilintion rates during the year pnmary health care was afected by the Three Mile Island (TMI) following the crisis. One study found that persons who were upset crisis and subsequent distress experienced by persons living in the during the crisis tended to be high practice utilizers both before and vicinity of the plant. The studies concerned: 1) Blue Ooss-Blue after the crisis. Dese results suggest that, while patterns of physi-Shield records of claims by primary care physicians in the vicinity of cian utilization prior to the TMI crisis predicted emouonal response TMl; 2) utilization rate s in a family practice located near the facility; during the crisis, the impact of the TMI crisis on subsequent

3) interviews with persons living within five miles of TMI following physician utiliraron was small. (Am / Public #calth 1984; 74:140-the crisis; and 4) responses to a questionnaire by primary care 142.)

physicians practicing within 25 miles of TMI. All four studies Introduction December 1978 and April through December 1979 from the

$56 primary care physicians practicing in the five counties There is substantial evidence that psychological distress is related to physician utilization.' However, less is known 5.urr unding TMI. The sample included 287 family practi-about the conditions that affect that relationship, such as tioners,152 internists, 78 gynecologists, and 39 pediatn-cians. Distance from TMI was used as a proxy measure for whether there is a threshold level of distress which must be reached before utilization is affected, and whether certain stress, based on the finding that both symptom reporting and attitudinal concern about TMI for nine months following the kinds of stress situations are more likely to lead to increased ensis were elevated within a 15-mile radius of the facility but utilization than others. The crisis at Three Mile Island (TMI) dropped sharply beyond that distance > The limitation of in March and April of 1979* and related events during the following year provide an opportunity to study such ques-these data is that visits due to stress are likely to be underrepresented m Blue Cross / Blue Shield claims, since tions. Four studies of physician utilization in the vicinity of they may not result in procedures that are chargeable to Blue TM. I during the year following the crisis are summarized in Cross or Blue Shield.

facts about the TMl crisis provide context for the Table I shows a drop in t,he number of procedure claims from 1978 t 1979 I r physicians practicing both withm and studies reported here. First, the releases of radiation during beyond 15 mdes from TMI, but the drop, although not the crisis were quite small** and, therefore, very unlikely to statistically significant, was smaller within 15 miles of TMI have any measurable physical health impact on the popula.

than beyond. Additional analyses were conducted to deter-tion. Second, levels of distress, as evidenced by both attitudinal concern and symptom reporting, were reported mine if an effect of TMI would be evident for the diffe:ent types f primary care physicians. The results were consist-among persons living in the immediate vicinity of TMl for at ent with those reported here with no statistically sigrifican, least a year following the crisis >-5 These long-term distress levels near TMl were in the "high normal" rather than effects that could be attnbuted to the TMI crisis noted.

"pathologica!" range as indicated by Symptom Check List 90 scores?3 The question addressed in the studies reported Family and Community Medicine Study here wts whether these, persistent, low Icvels of psychologi- The second study concerns frequency of visits to physi-cal distress resulted m increased utilization of prirnary care cians by a sample of patients of the Department of Family physicians (family practitioners, pediatricians, obstetricians, and Community Medicine. at the M.S. Hershey Medical and internists) by persons in the immediate vicinity of TMI. Center (8% miles from TMI) during the year before and the year following the crisis. The 498 subjects in this study had l Blue Cross / Rime Shield Study been patients in the practice for at least a year prior to the I . TMl crisis, had visited the practice two to six weeks after the I The first study

• analyzed procedure claims submitted to crisis and had filled out a brief questionnaire in which they the Pennsylvania Blue Cross / Blue Shield from April through reported their age, sex, education level, and rated how upset

! they were during the TMl crisis on a six-point scale ranging

'The "acensent" besan on March 2s,1979 at 4 00 a m.

from **not upset" to " extremely upset." This rating scale

,,, 'i,,N"i o" ,,ii,"*n was subsequently used in surveys conducted by the Pennsyl-vania Department of Health where it was shown to be Address repnet requests to Peter S. Houts. PhD, Department of Behav- correlated with concerns about the safety of TMI, with iorn Science, honiytvama State Universary, Conese of Medicine, Hershey, attributing of symptoms to TMI, and with distance from TMI PA 17033. Dr. Cleary is with the Department of Social Medacme and Health Pohey, Harvard Medical School; Dr. Henderson is wph the Department of Mp & ,

TW h m & hm M Fanuly and Community Medicine, Pennsylvansa State tJniversity College of these studies to correlate with Langner scale scores?

Medacine; Dr. He is with the Department of Econormes. Pennsyivania State Results of the Hhrshey study are shown in Table 2. The Unnersity; and Dr. Tokuhata is with the Divisaan of Epidenuoiosy. Pennsyl- comparison of post-TMI utilization rates of high and low E wasY and pi)(Nb$tYMy3 '

distressed patients, controlling for utilization rates prior to TMI, age, sex and education of the respondents, is not C 1964 Amencan Journal of Pubhc Health Coho %54 51.50 statistically significant. However, an examination of the d 140 AJi H February 1984, Vol. 74, No. 2 N~_

y.

MEDICAL CARE UTILIZATION FOLLOWING TMI TASLE 1--40een Number of Stue Crees Cloisused Procedures por Phyel- TASLE 2-4Aemi Number of Pract6ce Visite before end rfter the TMI cien per RAonth Pre and Post Tiel Catete for Primary Care Catete, Femity and CL rei eedicine t Practice, M.S. Her-

~ , " " 4 Practicing in the Five Coundes Surrounding Ttfi shey tendical Center Mean Nurnber of Cimmed Pebert Retings of Procedures per Month Upset dunng TMI 12 rnonths 12 rnonths Cnsis N before TMI efter TMI Apre- Apre-Decernber December Extremely, very or 1978 1979 somewhat upset 344 4.10 4 63 A little or not upset 154 3 55 3 91 Disorencet .55' .72*

W prececing wehen Diserence controk 15 rrwee of TMl(N = 295) 112.5 110.7 ling W e nurnber Physicians prececing beyond 89 8 of vielts before 15 rruise of TMI(N = 261) 93 2 TMI. ego, sex Deserence** 19.3 20 9 and educationtt .65 Deserence controsing for nutreer of clams Apre-December 1978tt 45 *p < 05 3Dreerences m ponert utAreton roies tested utermg taest.

• • Deserences m patiert utaronon roies post Tui careceing for ponent utersten pre-

i. agt mm. ano eoucaton or ine respanoeni usermo enews or covenanc=

,',',Ji,,,,,,,,,,,,,,,,,,,,,,,,,,,,,

-o rer.e.m - - re ~ t4e.t.

441979 phyacien morevy retos tested controemg tot 1978 monmly roles uter *g enspree of cronanos or of certain types of patients may have been affected by the crisis and, as with the results reported here, no statistically means shows a trend in the predicted direction with greater significant effects of the crisis could be detected.

utilization by the upset as compared to non-upset groups post TMI, controllirg for pre-TMI utilization rates. It can Population and Physician Surveys also be noted that both high and low distressed groups showed an increase in utilization after TMI, but this is In this section, results of three interview surveys will be difficult to interpret since the sample was taken from patients discussed. Two of these were telephone surveys of persons v./aiting to see physicians which would normally include living within five miles of TMI, conducted by the Pennsylva-many patients who were increasing their utilization. There nia Department of Health in July 1979 and January 1980.' In were, however, two interesting statistically significant find- these surveys, adult heads of households were asked if they ings indicating that patients who were high utilizers of the had visited physicians since the TMI crisis and,if so,if the practice both before as well as after the crisis period tended visit was because of Three Mile Island? The third survey was to rate themselves as upset during the crisis. This suggests a questionnaire mailed to primary care physicians in the fis e-that persons with a pattern of high practice utilization county area surrounding Three Mile Island.* Physicians responded emotionally to the crisis to a greater degree than were asked questions about how their practices had been did those who were low practice utilizers. affected by the TMI crisis. The strength of these studies is Sample bias is an important limitation of this study and that they report the perceptions of the persons who receive should be considered in interpreting its findings. The sample and provide medical care. In the case of the two telephona included only patients who had been in the practice over one surveys, a high response rate (82 per cent) was achieved.

year, who used the practice during the period immediately The timitations of these studies include the possibility of following the crisis. and who agreed to fill out the question. distorted recall, the fact that data were obtained only for naire. Because the sample was of consecutive users, it is adults in the telephone surveys, and that the return rate was likely to include a disproportionate number of high practice only 32 per cent in the physician survey. Therefore, both utilizers which is indicated by the fact that the study group response bias and sample bias are possible.

averaged over four visits to the practice per year while the In the July 1979 survey, fewer than I per cent of c.verage number of visits to family practices nationally is respondents (two out of 692) said they had visited physicians estimated to be 2.55.' Since the sample included only during the two-week period of the crisis because of symp-patients who used the practice for at least a year before the toms associated with TMI. However, in January 1980, 9 study period, it is also biased toward long-term utilizers. months after the crisis,8 per cent of respondents said they Other biases are indicated by comparing the age and educa- had visited physicians at some time because of the crisis.

tion levels of persons in this sample with those of the adult Results of the physician survey indicated that primary population within 25 miles of TMI as estimated from ran- care physicians were aware of only a small effect of the TMI domized telephone surveys conducted by the Nuclear Regu- crisis on their patients' visiting patterns. As shown in Table latory Commission shortly after the crisis.' The mean ages of 3, within the 25-mile radius, primary care physicians report-sample members was 43.4 as compared to 44.9 for the NRC ed an average of only 4.9 patients each who believed that sample, and the mean years of schooling was 13.5 as their visits were due to TMI. Furthermore, the physicians compared to 12.5 for the NRC group. Therefore, this sample felt that only one of those patients (23.6 per cent) were is probably slightly younger and more educated than the caused by TMI-related stress. The most frequently cited population as a whole, ne strengths of this study, on the symptoms which the physicians felt may have been due to other hand, are that objective records of utilization were TMl included stomach trouble, headaches, difficulty in used and that pre TMI baseline rates were included in the sleeping, and loss of appetite. These are among the symp-analyses. Additional analyses of patient records were carried toms that were found, in telephone surveys, to be elevated out to determine whether visits for certain types of problems near TMI.'

AJPH Fetruary 1904, Vol. 74 No. 2 141

. 4 4 HOUTS, ET AL TAst2 3-PersepneneomsPennerycm Phye6esenepresencineomhin finding suggests that high utilizers of medical services may as nee es m of Heer noir Penents were Asocess by vie be especially sensitive to stressful situations. Such sensitiv-ity may be one factor in the reported relationship between Average number of patente (per physcan) who behewed visa is due psychological distress and physician utilization.'

to swees from TMI 49 A practical implication of these fmdings is for planning Aversos por mnt C4 these patents' visits lhet physcans ther* ra due services that may be needed in dealing with similar events in to evees trom TMI 23 6 the future. The fmdings presented here suggest that there a-'-w sympeome pnyeoens nonced may be due to Tui (per cent of physonne repareng each symptom) will be only small increases in. demand for medic. al services saamach wome 12 following a crisis where there is little direct physical harm to Headache 5 persons and where long-term distress is well below the level Troute shaping 5 characteristic of mental patients."

Loos at apposte 3 REFERENCES I. Mechanic D: FEccts of psychological distress on perceptions of physical Discussion health and use of medical and psychiatnc facihtses. J Human Stress 1978 Despite the limitations of the studies presented here, the 2 PopuWoon Dose Assessment Group: Population Dose and Health results are strikingly consistent: all indicate only small, if Impact of the Accident at Three Mdc Island Nuclear Station. Washmgton, any, increases in utilization of primary physician services DC: US Govt Pnntmg Omce.1979.

L Houts P.Goldhaber M; Psychological and social effects on the population following the TMI accident. Since magnitud.e of the effects. surroundmg Three Mde Island after the nuclear accident on March 28.

observed were small, the lack of statistical sigmficance may 1979. In: Mgundar S tedt Energy Environment and the Economy.

be the result of type 11 error. In view of this, the consistency Pennsylvama Academy of Sciences.1981.

of findings across the studies is especially important. Patient 4. Bromet E: Dree Mde Island: Mental Health Findmst Pittsburgh:

Western Psychiatnc Institute and Chmc.1980.

self-reports yielded the largest estimates of utilization due to 5. Baum A: Catchel R. Flenung R. Lake C: Chrome and acute stress TMI (8 per cent). Two other investigators have reported ,,sociated with the Three Mde Island accident and decontamination:

similar self. report findings for physician utilization in the prehminary andmss of a longitudmal study. Report submitted to the vicinity of TMI. Kast cf al."' reported that approximately 7 Nuclear Regulatory Commission.1981.

6. Hu T. Slaysman L. Ham K. Yoder Y: Health-related economic costs of per cent of nuclear workers at TMI said they had seen a physician because of the accident, and Dunn" reported that 's,",*,$g','$ ,'f'ARePort submitted to the Pennsylvania 8 per cent of her sample of mothers of young children living 7. Houts P. Mdier R.Tokuhata G. Ham K: Health-related behavioralimpact withN 10 miles of TMI said they had visited a primary care of the Three Mde Island nuclear meident. parts I and ll Report submitted to the TMI advisory Panel on Health Research Studies of the Pennsylva-ph. ysician because of TMI. However, attributing a physician .. ma Department of Health,1980.

veut to TMI does not necessarily mean that the visit would

8. US Department of Health and Human Services: Health-t*mted States.

not have occurred in the absence of an accident. This was 19nt. DHHS Pub. No (PHS) 821232. December 1981. Tabic 36, p 159.

suggested by the results of the physician survey where less 9. Flynn C: nree Mile Island Telephone Survey Prehmmary Report on Procedures and Findmgs. Washmgton. DC: US Nuclear Regulatory than one-fourth of the visits which patients attributed to TMI Comnuss.on.1979.

were felt by physicians to be due o TMI 10. Kast S.Chisholm R. Eskenazi B: The impact of the accident at Three Mdc The low levels of long-term distress near TMI were liiand on ne behavior ana wep-bems of nuclear workers. part I.

below the range characteristic of mental patients." It may perceptions and evaluations. tchavioral responses. and work-related be that distress must be in the mental patient range for a *"Sd" *"d f"h"80 A'n J Pubhc Health 198h 75:472-483.

II. Dunn L: Comparatsve study of the mertal heahh of mothers with young considerable period of time in order to have a marked effect chAiren near Three Mde liland nudear reactor: imtial nadmst I.:np.ib-on physician utilization. Furthermore, as Gortmakern has hshed Master's nesis, Graduate School of Pubhc Hech. Umscruty of pointed out, the linkage between stress and health care Pittsburgh,1990.

12. Gor maker S. Eckenrode J. Gore S' Stress and utihzation of heahh utilization depends on the measures and time periods in. '" """ ' ' ' '" "

volved. For example, the length of the time periods studied 'g',$'h'n;T25 3s" may have obscured short-term effects of the crisis on health care utilization.

The one statistically significant fmdmg was in the Fam-

, ACKNOWLEDGNENTS ily and Community Medicme study indicating that persons The authors wish to acknowledge the assistance of Robert Mdler. PhD.

with a pattern of high practice utilization tended to be more Manlyn Goldhaber. MPH and Robert Cnst. PliD. in both data analpis and upset during the crisis than low practice utilizers. This edironal aisistance.

H2 AJPH Fetwuary 1984, Vol. 74. No. 2 l

c