- Email: [email protected]
The use of pediatric medical care: A critical review
.IChronDisVol.38.No. II. pp.935-945, 1985
0021-9681/X5 $3.00 +O.OO
Printed in Great Britain. All rights reserved
Copyright
THE USE OF PEDIATRIC MEDICAL A CRITICAL REVIEW SARAH Institution
MCCUE
HORWITZ,*
for Social and
HAL
Policy Studies.
MORGENSTERN~ Yale University.
and
1985 Pergamon Press Ltd
CARE:
LISA
New Haven,
c
F.
BERKMAN-?
CT 06520-7382.
U.S.A
(Received in recked ,form 21 March 1985)
Abstract-The problems of who uses medical services and why, are important ones for investigators interested in studying chronic diseases, particularly if they wish to avoid systematic error when assembling a study population. These issues are important when studying pediatric diseases due to the uneven use of medical services by children and the tendency of lower socioeconomic groups to use hospital facilities rather than private practitioners. In order to address these problems, we must understand why families seek medical care for young children. Utilization research shows that a number of descriptive factors such as child’s age, birth order, parental education, financial resources and perceived symptoms are related to service use. Additionally, psychosocial variables, such as distress, also predict utilization. Neither the descriptive nor psychosocial variables explain much of the variance in utilization. While some of this lack of explanatory power can be attributed to problems in measurement or study design, there are also conceptual and methodological issues that are not addressed in pediatric utilization research. This paper discusses four of these issues. It is our conclusion that two of these problems, the need for a new utilization taxonomy and the ambiguity of cause and effect. can be remedied. While more difficult to address, the inadequate conceptualization of social stress, psychological distress, and social support can be improved. However. measuring health status independently of utilization represents a major methodologic problem for which we currently have no ideal solution.
INTRODUCTION PROBLEMS of who uses medical services and why are important for investigators interested in studying chronic diseases. In order to study etiologic agents in any disease, a critical first step is the assembly of a group of individuals with the disease of interest. Proper subject selection is important to avoid internal validity and generalization problems. However, nonuniformity of symptom patterns and individual variation with respect to symptom recognition and determination for need of medical care, influence who comes into contact with the medical care system and who gets detected for disease. Therefore, various forms of illness behavior influence the selection and classification of subjects in many epidemiologic investigations. The uneven use of the medical care system by children makes recognition of why and how families seek care important to avoid systematic error when studying many common pediatric diseases. Even with improved access to care through government funded health insurance, children from low-income families receive less preventive care and frequently do not have a regular place for obtaining medical care [I, 21. Starfield points out that 25% of children receiving Medicaid use a hospital outpatient clinic, emergency room, or a health center without a particular doctor as their regular source of care [3].
THE
*Postdoctoral Fellow. Institution MH 15783. tAssociate Professor, Department Studies, Yale University.
for Social
and Policy
of Epidemiology
Studies,
Yale University.
and Public Health
935
NIMH
and the Institution
Training
Grant
No.
for Social and Policy
936
SARAH MCCUE HORWITZ ef al.
In contrast to the lack of consistent coordinated care for certain segments of the pediatric population, other children, particularly children from middle and upper-middle class families, receive a wealth of preventive services, usually from private practitioners. Given the uneven use of the medical care system by families and the tendency of lower socioeconomic groups to use hospital facilities rather than private practitioners, the choice of a study population when investigating certain pediatric issues must be carefully addressed. Therefore, we must understand why families seek medical care for young children and how differential utilization influences who is eligible for selection into a study. This paper will review what is known about the characteristics of consumers that infuence the use of pediatric medical services and identify issues that must be addressed for future research to yield a comprehensive picture of the decision process families go through when deciding to seek medical care for young children.
PEDIATRIC
MEDICAL
SERVICES
DETERMINANTS
In general, more is known about the effects on pediatric utilization of variables that are routinely collected in large national surveys (i.e. sociodemographic characteristics, such as child’s age) than is known about sociological or psychosocial factors, such as acute or chronic life stressors. Younger children, particularly infants, are more likely to see a physician in a given period than are older children [4-61. Children from small families and those who are first or second born receive more health care [5,7-IO]. Without controlling for birth order, Robertson et al. [9] found that among families with young children (i.e. less than 4 years old) those with only one child had more medical contacts per child in a one year period. Andersen and Kasper looked at the influence of family size on utilization of physician services and found that infants, regardless of family size, see a physician [8]. However, within all age categories, family size was inversely related to the mean number of physician visits. Tessler found that birth order is significantly related to both total utilization and preventive care independent of family size [IO]. For families of any size, adjusting for eight control variables (e.g. child’s age, mother’s age, number of chronic problems, race, type of medical insurance), first born children received more preventive care. Children whose mothers are white and well-educated have more yearly medical care contacts than children whose mothers are nonwhite and poorly educated [5,8, 10, 111. Similarly, children whose mothers have certain health attitudes and beliefs are more likely to use medical services. Using the Health Belief Model, Becker et al. found that a noncomplier with pediatric medical regimens sees her child as healthy, does not feel the need to take medical advice, tends not to seek care when illness symptoms first appear, and does not see illness as a threat [12]. In a study of 251 low-income mothers who brought a child to a pediatric clinic for treatment, Kirscht et a/. [I 31 concluded that health beliefs, such as efficacy of medical care, interact with situational constraints to determine the action taken in response to illness. Becker et al. [14] found that maternal health attitudes and beliefs were associated with preventive and illness visits but in opposite ways. Women with an active orientation towards health care (i.e. interventionistic and believing illness can be avoided) and those who attributed good health and low illness susceptibility to their children were high users of preventive services but had few acute illness visits. However, mothers who were passive in their health care orientation and those who saw their children as susceptible made few preventive visits but were high utilizers of acute illness care. Children from low-income families receive less care than those from middle or high-income families even with government-funded health insurance [8, 15-18-J. Comparing the actual and anticipated effects of Medicaid on utilization of children’s medical services, Roghmann et al. [19] observed little change in source, frequency, or purpose of care after the introduction of Title XIX. Using data from the 1975 Rochester Community Child Health Survey, Wolfe reports that children who have fewer available family
Use of Pediatric
Medical
Care
937
resources (i.e. working parents, low income, and few community resources) are less likely to receive their care from a private practitioner [17]. Similarly, Bite et al. [18] observed that, while the overall relationship between income and use has diminished, the association still exists for children’s use of medical services. While government-funded health insurance has provided poor families with health care purchasing power, government-sponsored health efforts have had little effect on other access problems. Having a regular source of care encourages the use of health services. independent of need. However, 18% of children from low-income families and 6% from middle-income families have no regular source of care. The problem of physician availability also influences children’s use of health services. While the number of physicians has increased over the past ten years, there are still not enough physicians in areas which the federal government has defined as medically underserved [16]. According to respondents in the National Health Survey, delay in seeking care for pediatric acute conditions is most often due to an inability to obtain an appointment [3]. While perceived health need is consistently related to use, the question of whether perception of health is a predictor of utilization independent of actual health status has not been answered. Andersen and Kasper [8] found that mothers who saw their children as less healthy had more physician contacts. Similarly, Tessler [lo] and Tessler and Mechanic [4] found a significant positive relationship between utilization and number of mother-reported child chronic health conditions. Unfortunately, only the Tessler and Mechanic study used a prospective design. They determined the number of motherreported child chronic health conditions prior to collecting the child utilization data. Consequently, only this study allows us to draw the conclusion that a measure of perceived need predicts use of pediatric health services. We can identify positive and negative correlates of pediatric medical care utilization but these factors do not explain the majority of variation in the use of medical services. Consequently, researchers have started to put more emphasis on the investigation of other types of factors, particularly psychosocial factors, such as stressful life events [20,21]. Initial studies examining stressful life events and their relationship to disease used a retrospective design; individuals were asked about past utilization, life change events, and the distress or readjustment accompanying these events [22]. Recognizing the problems inherent in using retrospective designs to investigate the effects of psychosocial factors (e.g. recall bias), investigators have recently gathered prospective data to examine this relationship 123,241. However, the low incidence of major diseases and the difficulty of obtaining ongoing medical examinations for a study population have forced many investigators to measure minor illness episodes, the detection of which are strongly influenced by illness behavior. Consequently, it is difficult to determine whether people actually become ill after experiencing a number of life events, or rather, whether people seek medical care more readily when distressed [25]. Several studies have examined the use of medical services in emotionally distressed or situationally stressed families [4, 26, 271. Roghmann and Haggerty found that stress, defined as “any event perceived by the family as upsetting” [26, p. 5211 with no illness present. increased utilization of medical services for mothers and children. However, they found that perceived stress, when accompanied by actual illness, decreased utilization for mothers
but increased
utilization
for young
children.
Using a diary method of reporting daily stressful events and symptoms similar to the one used by Roghmann and Haggerty, Gortmaker et al. [27] found that daily stress was related to family utilization, independently of reported symptoms. They also noted that life events recorded prior to the diary reporting period predicted utilization independently of reported symptoms and daily stressors. Unfortunately, neither study assessed the possible impact of the diary method on the outcome under investigation, utilization, although Gortmaker et ul. recognized that this method of reporting might influence use of health services. Tessler and Mechanic [4], looking at pediatric utilization in a prepaid group practice, observed that maternal psychological distress was positively associated with magnitude of
SARAH
938
MCCUE HORWITZ et al.
use of children’s services, although propensity to seek care was more important in the initial decision to seek care. In general, the results of these studies show that psychologically distressed and situationally stressed individuals have a greater tendency to seek medical care, particularly for their young children. The precise dynamics of this phenomenon are not well understood. However, there are at least three alternative explanations for this association among children (Fig. 1). One explanation is that distressed mothers have children who are sick more often or are more severely ill, possibly because of their exposure to stressful family situations. Meyer and Haggerty [28] found an association between acute family crisis and the onset of illness in their study of streptococcal infections. The results of this study are particularly important because illness was measured separately from utilization. A second study by Boyce et al. [24] that differentiated illness from utilization found that a high family life change score was predicitve of illness duration. Beautrais et al. [23] found that family life events were associated with outpatient utilization and hospital admissions for medical problems, such as lower respiratory illnesses, gastroenteritis, burns, and injuries. These results agree with the findings of an earlier retrospective study by Heisel et al. [22] in which four patient groups of children (rheumatoid arthritis patients, general pediatric admissions, surgical patients, and psychiatric patients) had experienced more frequent or more serious life events prior to the onset of their illnesses than had their healthy peers. A second possible explanation for the relationship between distress and utilization is that distressed mothers are more sensitive to their children’s symptoms, compared to nondistressed mothers [29]. In a recent follow-up of the children he studied in 196 1, Mechanic [30] found that several factors influenced the reporting of common physical symptoms including subjective reports of poor physical health and measures of psychological distress. Mechanic postulated that psychological distress may produce discomfort, increase body monitoring, and/or sensitize an individual to physical changes. Thus, the heightened perception of illness through increased attention to symptoms resulting from stressful life events could be another explanation for the increase in medical care utilization among distressed individuals. Third, it is possible that women experiencing psychological distress or confronting stressful situations deal in part with their distress or stress by seeking medical care for their children as a type of social support. This third hypothesis deals with a latent social function of medical care. Latent social functions performed by the medical care system were documented by Shuval in her study of new immigrants to Israel [31]. Looking at those individuals who used twice the average amount of medical services (more than 20 visits yearly), she found that socially isolated immigrants tended to be relatively high utilizers of care, regardless of their morbidity status.
la. b.
2.
3.
FIG.I.Alternate
Family Stress
-
Family Stress
_
Mother’s Distress
_
Mother’s Distress/ Family Stress explanations
__f
Risk of Disease Among Children Increased Severity of Disease Among Children HeightenedSympton Perception Increased Need for Social Support
_
Increased Utilization Increased Utilization
-+
Increased Utilization
-+
Increased Utilization
for the relationship between psychological stress and pediatric utilization.
distress
or situational
Use of Pediatric Medical Care
939
The latent function hypothesis has received little research attention. No investigation in the literature on pediatric utilization has directly tested the hypothesis and only a few studies have evaluated the effects of supportive, nonmedical interventions. In a field experiment designed to reduce children’s stress during hospitalization for tonsillectomy, Skipper and Leonard [32] found that social interaction between nursing personnel and the mothers of young children significantly reduced common physical signs of stress (e.g. elevated temperature, pulse rate, blood pressure, post-operative emesis, disturbed sleep, and long recovery periods) in children. Pless and Satterwhite [33] evaluated the impact of nonprofessional family counselors on families with chronically ill children. Significantly more of the chronically ill children from counseled families showed improvement on psychological scores when compared to control children whose families did not receive the lay intervention. Assuming that the medical care system provides certain latent services to patients, it follows that medical care can serve as a source of support for psychologically distressed families or families who are confronted with stressful life situations, particularly those who do not have strong support available from their social networks. The supportive function of social networks is a relatively new area of concern in psychosocial epidemiology and developed as an important concept due to observations documenting differential host resistance to stressful life events [34-40]. As potentially important as this concept is, there are several major questions that have not been resolved. One question concerns the relationship among support, acute and chronic stressful life changes, and illness. The precise questions of who will become ill, under what conditions. and through what biological pathways, are still to be answered. A second important question concerns the way in which social networks influence utilization. It is unclear whether social networks are related to utilization due to network health beliefs, inadequate network support, or symptom perception [41].
UTILIZATION
RESEARCH
PROBLEMS
In reviewing previous studies that examined possible determinants of pediatric utilization, we find that we cannot predict very well who is likely to use services. For instance, using ten variables Tessler [lo] could explain only ISO,:, of the variation in a measure of total utilization and 8’;;, of the variation in the use of well-child care. Some of this inability to predict pediatric health services use is attributable to conceptual and methodologic problems. including the need for a new utilization taxonomy, the ambiguity of cause and effect. the poor conceptualization of stress and support, and the need to develop a measure of health status independent of utilization. We will discuss each one of these weaknesses separately and, where possible, offer suggestions to remedy the problem.
An important conceptual problem in research looking at the determinants of pediatric utilization is the need for a new utilization taxonomy. Until recently, researchers have viewed pediatric health services use as a single outcome measure, typically the number of physician visits per year. The initial efforts separating physician visits into preventive and illness encounters have shown that utilization must be categorized in different ways, depending on the type of question a researcher is addressing [42]. The best way to classify visits for the purposes of etiologic investigations may not be the best way to classify visits for the study of utilization. For example, Ross and Duff [43] recently suggested that pediatric utilization should be broken down by type of visit (patient initiated or physician initiated) and by type of contact (i.e. preventive or curative). Initial research efforts conceptualizing utilization behavior in nontraditional ways, such as direct vs indirect risk preventive behaviors or telephone contacts vs office visits, have shown that various types of utilization may have different determinants. Roghmann and Hapgerty [26] found that the probability of telephone calls and contacts to outpatient or
940
SARAH MCCUE HORWITZ et al
emergency departments was doubled in situations where stress was combined with illness. However, there was little change in the probability of contact with any other department or the probability of an office visit. We know that medical care vists for acute illness do not occur randomly. Rather, they tend to cluster within specific bouts of illness called illness episodes. We know very little about the factors that influence repeat contacts within an illness episode compared to those factors that predict the number of episodes of illness. It may be that health-related variables, such as the presence of chronic conditions, symptom severity, or symptom recognition, are the important determinants of care seeking for a given health problem. However, once an individual contacts the health care system, system-related and/or psychosocial variables may prove to be more important determinants of service use for the same illness episode. Whether a parent makes two or six contacts with a health care provider for a single episode of otitis media in a 2 year old child may be related less to the child’s physical condition than to the parent’s financial resources or to the presence of a supportive social network. A detailed investigation of episode specific utilization will determine factors important for predicting initial health care contacts and subsequent health services use. Ambiguity of cause and efect Another conceptual deterrent to understanding the factors that influence pediatric health services use is the ambiguity of cause and effect that exists in this research. Some of this ambiguity is due to the use of retrospective rather than prospective designs in assessing utilization predictors. In a retrospective design, an individual is asked about the predictors of interest after the outcome events (utilization) have occurred. Additionally, utilization data has often been collected retrospectively, either through recall by patients or through medical records. With this type of design three types of problems can occur. First, if utilization data are collected retrospectively from medical records, outcome information may not be accurate due to inadequacies of the records. Second, because predictor information is recorded after the utilization period of interest, the outcome variable, utilization, may influence the predictors under investigation. Third, the threat of differential recall is always present in this type of research design. Furthermore, even with a prospective design where information about the predictors of interest is collected at the beginning of the follow-up period and utilization is observed weeks or months after baseline, it is difficult to determine whether changes in certain variables (especially psychosocial variables) preceded or succeeded changes in health status or utilization. This is attributable to the strong feedback system that exists between attitudes and behaviors. Individuals often behave in a way that they think is consistent with what they have said or, conversely, individuals have been known to jusify past behavior by adopting a particular point of view. For example, in a retrospective study, individuals who reported many dental visits saw themselves as less susceptible to tooth decay at interview time [25]. In a prospective study, however, those who saw themselves as less susceptible to tooth decay at baseline subsequently had fewer preventive visits [44]. While this reversal of findings is explainable by attitudinal and motivational theories, it makes the distinction of cause and effect quite difficult. This causal ordering becomes increasingly more complex when assessing daily fluctuations in psychosocial variables. If there is a long lag between the psychosocial event and the outcome of interest (e.g. pediatric utilization), the time ordering is less problematic. However, when the time period between stimulus and response is short, it is often difficult to distinguish the cause from the effect. Using the stress-physical symptoms relationship as an example, it is difficult to ascertain, even with a diary method of reporting, whether stressors or the perception of stress occurred before or after the reporting symptomatology. Robertson et a/. [45] could not estabish a causal sequence for diary data on family illness and social or emotional problems. While illness was correlated with social and emotional problems, neither looking at stressful events in the four days preceding illness nor looking at illness in the four days preceding stressful events produced a clear relationship. It cannot
Use of Pediatric Medical Care
941
be definitely determined whether stress produced symptoms, whether stress heightened sensitivity to existing symptoms, or whether underlying symptoms produced heightened sensitivity to stress. Inadequate conceptualization of stress, psychological distress and social support The relationship of social stressors, psychological distress, and social support to the use of pediatric medical services has received considerable attention. However, these factors are not carefully defined or used in any consistent way in previous investigations. In reviewing earlier work linking psychological distress and utilization, we find that the measures of distress elicited general anxiety-related symptoms. Consequently, previous work did not look at the relationships of other types of psychological distress (e.g. depression) or, more specifically anxiety, and use of pediatric medical services. If the relationship between distress and utilization is caused by anxiety and not other types of psychological maladjustment, then our understanding of children’s utilization determinants will be enhanced if we can establish this association. However, employing a nonspecific term such as psychological distress does not allow the development and testing of hypotheses that will help us to determine why family psychological problems are related to use of children’s services. Similar criticisms can be made about the relationship of social stressors to utilization. Stress has been defined in several ways ranging from family specific definitions (e.g. anything a family finds upsetting) to established lists of stressful life events with standard weights developed from a reference population. We need to understand what aspects of social stressors cause families to use health services for young children. In order to understand the relationship of stress to utilization, we must be able to differentiate whether it is simply an accumulation of disruptive events that prompts use of medical care, or rather, whether it is circumstances that a particular family perceives as upsetting that affects changes in its illness behavior. In addition, we need to know whether social stressors are linked to use of children’s health services in all socioeconomic groups since many of the studies linking stress and utilization have used low or lower middle-income populations. The concepts of social networks and support are not clear-cut nor are they easily measured. Research on the relationship of social networks to health care use has been retarded by imprecise definitions of social network characteristics, nonspecific hypotheses concerning their relationships to utilization, and by a confusion of social support and social networks. Social networks have been defined as an individual’s web of social ties and the quantity, quality and morphology of those ties. Thus, they have certain measurable characteristics (e.g. size and geographical proximity) and functions (e.g. support and role modeling). Social support, on the other hand, is an emotional or instrumental benefit perceived by an individual (i.e. a qualitative dimension of a network). Often. when investigators speak of social support, the assumption is made that all networks are supportive. When investigators make this assumption, they are not allowing all dimensions of networks to be tested [46]. The complexity of social networks is alluded to in the literature; however, little information can be found on what components of networks appear to account for the possible effects on health status and utilization behavior [46,47]. We advocate that three dimensions in the measurement of social networks be included in future investigations of pediatric utilization: (1) structural and interactional properties of the network, such as size, strength of ties, and frequency; (2) a subjective appraisal of perceived network support; and (3) a measure of a mother’s (or family’s) orientation or attitude toward using network resources. Inclusion of network properties provides a structured way to assess the resources available to a person. Once the network properties are determined, a subjective measure of their importance to the family must be determined. Finally. evaluation needs to be made of whether a mother (or family), because of her coping style, will be able to use the support available from network resources. This aspect of support, while receiving some justification in the literature [48-501, has received little empirical attention.
SARAH MCCUE HORWITZ et al.
942
Measurement
qf health status
We need to measure health status so that it can be used as a covariate when assessing the impact of sociodemographic, health attitude, health belief, psychological distress, social stressor, and social network variables on utilization. However, there are several problems associated with the measurement of health status, particularly how to measure health status independently of behavior (utilization) and subject perceptions (self-reports of health status). One attempt to assess health status is the development of different indices based on medical record information. This method allows for the measurement of health status only for children who have used health services and is not independent of utilization behavior. A second way of measuring health status is to use self-reports. Self-reports of children’s health status are gathered in two ways: by global health status ratings, usually done by the child’s mother at baseline or periodically in a follow-up period, and by daily reports of symptoms through the use of diaries. Unfortunately, these methods rely on self-report without corroboration from medical records or physical examination data. Mechanic [51] has noted that global self-reports of health status and the assessment of illness behavior (e.g. symptom recognition and utilization) may not be independent of one another. We can also make the same criticism of the diary method since, like the self-report of global health status, it may not be independent of symptom recognition or utilization. A third approach to measuring health status is through regular examinations by health care providers. This solution is less dependent on utilization than the other approaches but also has important problems associated with it. First, physical examinations of young children rely to a large degree on parent’s reporting of symptoms, and therefore, are somewhat dependent on self-reporting. Second, it is both expensive and impractical to examine a large group of children regularly, assuming families initially agree to such procedures. Third, by introducing such an obvious intervention into families’ lives, one would expect that their utilization behavior might be altered. Evidence of behavioral changes brought on by experimental observational without experimental intervention, the phenomenon [52]. If “Hawthorne Effect,” is a well recognized social-psychological behavior can be influenced by actively engaging subjects in our research, we must consider the potential impact on behavior of more intrusive interventions, such as conducting periodic health examinations to determine health status or having parents keep diaries to assess levels of symptoms or stress. We conclude that assessment of health status is a major methodologic problem in studying the determinants of utilization for which we currently have no ideal solution. The two solutions often employed, self-reports of health status and examinations by health care providers, have serious problems associated with them. Until we can satisfactorily assess health status independently of utilization and in an unobtrusive way, we have no way of testing the effects of other factors on utilization. Unfortunately, given the need to control extraneous factors such as health status when testing alternative hypotheses about utilization behavior, this problem presents a serious deterrent to understanding the determinants of health services use.
SUMMARY
AND
CONCLUSIONS
Even though investigators have expended considerable effort to understand why families use medical care for young children, we cannot predict very well who is likely to use services. We believe that some of this inability to predict the use of services is due to conceptual and methodologic problems present in earlier studies. Two of the problems we have identified, need for a new utilization taxonomy and ambiguity of cause and effect, can be remedied. Initial attempts to disaggregate total physician visits have shown that different variables are related to different types of utilization and that the amount of explainable variation differs according to the type of utilization an investigator chooses to study [4, 10,431. Similarly, while investigators have started to realize that assessing
Use of Pediatric
Medical
943
Care
utilization determinants using a retrospective design can produce ambiguous results, they must also recognize that, even in a prospective design, assessing daily fluctuations in psychosocial variables will produce unclear results particularly when the time period between the social stimulus or stressor and response (utilization) is short. Consequently, it will be necessary to use study designs that allow for a clear time ordering when attempting to evaluate the relationship of psychosocial factors to utilization. Additionally, if the prospective study uses utilization data based on patient perceptions or medical records, then the outcome data will be subject to recall errors and the well-known inaccuracies of medical records. Only a completely prospective study, one that assesses subjects at baseline and collects utilization information at the time of each medical visit, can avoid producing unclear results. The problem of inadequate conceptualization of social stress, psychological distress and social support will be more difficult to address. After many investigations of the relationship of life events to utilization, we still cannot identify what aspects of life events increase utilization, why the association does not hold for all families, or whether this association holds in all socioeconomic groups. Clearly, research which carefully compares standard life events, both as accumulations and as rated by families, with family generated upsetting events in different types of populations will help to clarify the relationship of stress to utilization. In addition, an evaluation of stress in the context of the overall coping strategies of families (i.e. family coping styles, personal and social resources) will aid in our understanding of how stress relates to utilization. Similar comments can be made about the relationship between psychological distress and utilization. We need to identify what kind of distress is related to utilization, whether it is related to all types of utilization or just certain types, and whether its impact is modified by other variables. Social networks and social support are relatively new research concerns. Investigations of the relationship of social networks to health care use have been retarded by imprecise definitions of social network characteristics, nonspecific hypotheses concerning their relationship to utilization, and by a confusion of social support and social networks. We have suggested that three dimensions of social networks be included in future investigations of social networks. In addition, we suggest that alternative hypotheses evaluating the possible ways in which social networks may relate to utilization be tested. Social networks may be related to various types of utilization in more than one way. While the first three problems we have identified can be at least partially resolved with current investigative tools, the measurement of health status independent of utilization represents a major methodologic problem for which we currently have no ideal solution. Until we can measure health status independently of utilization, we have no way of assessing the effect of other factors on utilization. Given the need to control extraneous factors such as health status when testing alternative hypotheses about families’ utilization behavior, this problem presents a serious deterrent to understanding pediatric utilization determinants.
REFERENCES I.
2. 3
Rundall TG. Wheeler JRC: The effect of income on use of preventive care: An evaluation of alternative explanauons. J Health Sot Behav 20: 397406. 1979 Dutton DB: Explaining the low use of health services by the poor: costs, attitudes of delivery systems? Am Sot
Rev 43: 34X-368,
Starfield
B: Special
Pediatrics
4.
1978
responsibilities:
7 I, 433S439.
The role of the pediatrician
and
the goals
of pediatric
education,
19X3
Tessler R. Mechanic D: Factors affecting children’s use of physician services in a prepaid group practice. Med Care 16: 33 46. 1978 5. Slessingcr DP. Tessler RC. Mechanic D: The effects of social characteristics on the utilization of preventive medical services in contrasting health care programs. Med Care 24: 392-404, 1976 6. Ezzati T: Ambulatory care utilization patterns of children and young adults. Vital and Health Statistic Series B(39). Hyattsville. Maryland: National Center for Health Statistics. 1978 1. Haggerty RJ. Roghmann KJ. Pless IB: Child Health and the Community. New York: John Wiley, 1975
944 8.
SARAH MCCUE HORWITZet al
Andersen R, Kasper JD: The structural influence of family size on children’s use of physician services, J Comp Fam Stud 4: 116-130,
9. 10. II.
12. 13. 14. 15.
16.
17. 18. 19. 20. 21.
LS, Kosa J, Alpert JJ ef al: Family size and the use of medical resources. In Family and Fertility. Liu WT (Ed) Notre Dame: University of Notre Dame Press, 1967 Tessler R: Birth order. family size, and children’s use of physician services. Health Services Res 15: 55-62. 1980 U.S. Department of Health and Human Services, Select Panel for the Promotion of Child Health. Better Health for our Children: A National Strategy. Public Health Service Publication No. 79-550-71, Vol. 3. Washington, DC: Government Printing Office, 1981 Becker MH, Drachman RH, Kirscht JR: Motivations as predictors of health behavior. Health Services Rep 87: 852-862, 1972 Kirscht JP, Becker MH, Eveland JP: Psychological and social factors as predictors of medical behavior. Med Care 14: 422431, 1976 Becker MH. Nathanson PA, Drachman RH er al: Mother’s health beliefs and children’s clinic visits: A prospective study. J Commun Health 3: 125-135, 1977 Silver GA: Child Health: America’s Future. Germantown, Maryland: Aspen Systems Corporation, 1978 U.S. Department of Health and Human Services, Select Panel for the Promotion of Child Health. Better Health for Our Children: A National Strategy. Public Health Service Publication No. 79-550-71, Vol. 5. Washington, DC: Government Printing Office, 1981 Wolfe BL: Children’s utilization of medical care. Med Care 18: 11961207, 1980 Bite TW, Eichhorn RL. Fox PD: Socioeconomic status and the use of physician services: A reconsideration. Med Care IO: 261-271, 1972 Roghmann KJ, Haggerty RJ, Lorenz R: Anticipated and actual effects of Medicaid on the medical care pattern of children. N EngI J Med 285: 1053-1057, 1971 Roghmann KL: Models of health and illness behavior. In Child Health and Community. RJ Haggerty, KL Roghmann, IB Pless (Eds). New York: Wiley-Interscience, 1975 Starfield B, van den Berg BJ, Steinwachs DM: Variations in utilization of health services by children. Pediatrics
22. 23. 24. 25. 26. 2-l. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45.
1973
Robertson
63: 633-641,
1979
Heisel JS, Ream S, Raitz R er al: The significance of life events as contributing factors in the diseases of children--II. Behav Pediatr 83: 119-123, 1973 Beautrais AL, Ferguson DM. Shannon FT: Life events and childhood morbidity: a prospective study. Pediatrics 70: 935-940, 1982 Boyce WT. Jensen EW, Cassel J er al: Influence of life events and family routines on childhood respiratory tract illness. Pediatrics 60: 609-615, 1972 Kasl SV, Cobb S: Health behavior, illness behavior, and sick role behavior. Arch Envir Health 12: 246266, 531-541. 1966 Roghmann KL, Haggerty RJ: Daily stress, illness, and the use of health services in young families. Pediatr Res 7: 520-526, 1973 Gortmaker SL, Eckenrode J. Gore S: Stress and utilization of health services: A time series and cross-sectional analysis. J Health Sot Behav 23: 25-38, 1982
Meyer RJ. Haggerty RJ: Streptococcal
infections in families: Factors altering individual susceptibility.
Pediatrics 29: 539-549, 1962 Mechanic D: The influence of mothers on their children’s health attitudes and behavior. Pediatrics 33: 444-453. 1964 Mechanic D: The experience and reporting of common physical complaints. J Health Sot Ftehav 20: 146-155. 1980 Shuval JT: Social Functions of Medical Practice. San Francisco: Jossey-Bass Publishers, 1970 Skipper JK. Leonard RC: Children. stress. and hospitalization: A field experiment. J Health Sot Behav 9: 275-286. 1968 Pless IB, Satterwhite B: Chronic illness in childhood: Selection, activities and evaluation of nonprofessional Family counselors. Clin Pediatr 403407. 1972 Cassel J: The contribution of social environment to host resistance. Am J Epid 104: 107-123, 1976 Cobb S: Social support as a moderator of life stress. Psychosom Med 38: 300-313, 1976 Gore S: The effect of social support in moderating the health consequences of unemployment. J Health Sot Behav 19: 157.-165, 1975 Henry JP. Stephens PM: Stress, Health, and the Social Environment. New York: Springer-Verlag, 1977 Kaplan BH. Cassel JC. Gore S: Social Support and Health. Med Care 15: 47-58. 1977 Nuckolls KB. Cassel J. Kaplan BH: Psychological assets, life crisis. and the prognosis of pregnancy. Am J Epid 95: 431-441. 1972 SOSd R. Kendell J. Klaus M P! nl: The effect of a supportive companion of prenatal problems, length of labor. and mother-infant interaction. N Engl J Med 303: 597-600, 1980 Dimsdale JE, Eckenrodc J. Haggerty RJ E/ al: Role of social supports in medical care. Sot Psychiat 14: 175-180. 1979 Greenlick MR. Larson DR. Mullooly JP (‘I ul: Determinants of utilization of health care systems. In Proc Int Conf on Health Technology. Cullen MF (Ed) San Francisco: Society for Advanced Medical Systems. Operations Research Society of America and Salutis Unitas, 1973 Ross CE, Duff RS: Returning to the doctor: The effect of client characteristics, type of practice and experiences with care. J Health Sot Behav 23: 119-131, 1982 Kegeles SS: Why people seek dental care: A test of a conceptual formulation. J Health Human Behav 4: 149-155. 1963 Robertson LS. Kosa J. Heagarty MC, (‘I u/: Changing the Medical Care System. New York: Praeger Publishers, 1974
Use of Pediatric
46. 47. 48. 49. 50. 51. 52.
Medical
Care
945
Wellman B: Applying network analysis to the study of support. In Social Network and Social Support. Gottlieb BH (Ed) Beverly Hills, California: Sage Publications, 1980 Gourash N: Help-seeking: A review of the literature. Am J Commun Psycho1 6: 413423, 1978 Schaefer C, Coyne JC, Lazarus RS: The health-related functions of social support. J Behav Med 4: 381-407, 1981 Pearlin LI, Schooler C: The structure of coping. J Health !%c Behav 19: 2-21, 1978 Fischer CS: To Dwell Among Friends: Personal Networks in Town and City. Chicago: University of Chicago Press, 1982 Mechanic D: Correlates of physician utilization: Why do major multivariate studies of physician utilization find trivial psychosocial and organizational effects? J Health Sot Behav 20: 387-396. 1979 Franke RH, Kaul JD: The Hawthorne experiments: First statistical interpretation. Am Sot Rev 43: 5% 643. 1978
0021-9681/X5 $3.00 +O.OO
Printed in Great Britain. All rights reserved
Copyright
THE USE OF PEDIATRIC MEDICAL A CRITICAL REVIEW SARAH Institution
MCCUE
HORWITZ,*
for Social and
HAL
Policy Studies.
MORGENSTERN~ Yale University.
and
1985 Pergamon Press Ltd
CARE:
LISA
New Haven,
c
F.
BERKMAN-?
CT 06520-7382.
U.S.A
(Received in recked ,form 21 March 1985)
Abstract-The problems of who uses medical services and why, are important ones for investigators interested in studying chronic diseases, particularly if they wish to avoid systematic error when assembling a study population. These issues are important when studying pediatric diseases due to the uneven use of medical services by children and the tendency of lower socioeconomic groups to use hospital facilities rather than private practitioners. In order to address these problems, we must understand why families seek medical care for young children. Utilization research shows that a number of descriptive factors such as child’s age, birth order, parental education, financial resources and perceived symptoms are related to service use. Additionally, psychosocial variables, such as distress, also predict utilization. Neither the descriptive nor psychosocial variables explain much of the variance in utilization. While some of this lack of explanatory power can be attributed to problems in measurement or study design, there are also conceptual and methodological issues that are not addressed in pediatric utilization research. This paper discusses four of these issues. It is our conclusion that two of these problems, the need for a new utilization taxonomy and the ambiguity of cause and effect. can be remedied. While more difficult to address, the inadequate conceptualization of social stress, psychological distress, and social support can be improved. However. measuring health status independently of utilization represents a major methodologic problem for which we currently have no ideal solution.
INTRODUCTION PROBLEMS of who uses medical services and why are important for investigators interested in studying chronic diseases. In order to study etiologic agents in any disease, a critical first step is the assembly of a group of individuals with the disease of interest. Proper subject selection is important to avoid internal validity and generalization problems. However, nonuniformity of symptom patterns and individual variation with respect to symptom recognition and determination for need of medical care, influence who comes into contact with the medical care system and who gets detected for disease. Therefore, various forms of illness behavior influence the selection and classification of subjects in many epidemiologic investigations. The uneven use of the medical care system by children makes recognition of why and how families seek care important to avoid systematic error when studying many common pediatric diseases. Even with improved access to care through government funded health insurance, children from low-income families receive less preventive care and frequently do not have a regular place for obtaining medical care [I, 21. Starfield points out that 25% of children receiving Medicaid use a hospital outpatient clinic, emergency room, or a health center without a particular doctor as their regular source of care [3].
THE
*Postdoctoral Fellow. Institution MH 15783. tAssociate Professor, Department Studies, Yale University.
for Social
and Policy
of Epidemiology
Studies,
Yale University.
and Public Health
935
NIMH
and the Institution
Training
Grant
No.
for Social and Policy
936
SARAH MCCUE HORWITZ ef al.
In contrast to the lack of consistent coordinated care for certain segments of the pediatric population, other children, particularly children from middle and upper-middle class families, receive a wealth of preventive services, usually from private practitioners. Given the uneven use of the medical care system by families and the tendency of lower socioeconomic groups to use hospital facilities rather than private practitioners, the choice of a study population when investigating certain pediatric issues must be carefully addressed. Therefore, we must understand why families seek medical care for young children and how differential utilization influences who is eligible for selection into a study. This paper will review what is known about the characteristics of consumers that infuence the use of pediatric medical services and identify issues that must be addressed for future research to yield a comprehensive picture of the decision process families go through when deciding to seek medical care for young children.
PEDIATRIC
MEDICAL
SERVICES
DETERMINANTS
In general, more is known about the effects on pediatric utilization of variables that are routinely collected in large national surveys (i.e. sociodemographic characteristics, such as child’s age) than is known about sociological or psychosocial factors, such as acute or chronic life stressors. Younger children, particularly infants, are more likely to see a physician in a given period than are older children [4-61. Children from small families and those who are first or second born receive more health care [5,7-IO]. Without controlling for birth order, Robertson et al. [9] found that among families with young children (i.e. less than 4 years old) those with only one child had more medical contacts per child in a one year period. Andersen and Kasper looked at the influence of family size on utilization of physician services and found that infants, regardless of family size, see a physician [8]. However, within all age categories, family size was inversely related to the mean number of physician visits. Tessler found that birth order is significantly related to both total utilization and preventive care independent of family size [IO]. For families of any size, adjusting for eight control variables (e.g. child’s age, mother’s age, number of chronic problems, race, type of medical insurance), first born children received more preventive care. Children whose mothers are white and well-educated have more yearly medical care contacts than children whose mothers are nonwhite and poorly educated [5,8, 10, 111. Similarly, children whose mothers have certain health attitudes and beliefs are more likely to use medical services. Using the Health Belief Model, Becker et al. found that a noncomplier with pediatric medical regimens sees her child as healthy, does not feel the need to take medical advice, tends not to seek care when illness symptoms first appear, and does not see illness as a threat [12]. In a study of 251 low-income mothers who brought a child to a pediatric clinic for treatment, Kirscht et a/. [I 31 concluded that health beliefs, such as efficacy of medical care, interact with situational constraints to determine the action taken in response to illness. Becker et al. [14] found that maternal health attitudes and beliefs were associated with preventive and illness visits but in opposite ways. Women with an active orientation towards health care (i.e. interventionistic and believing illness can be avoided) and those who attributed good health and low illness susceptibility to their children were high users of preventive services but had few acute illness visits. However, mothers who were passive in their health care orientation and those who saw their children as susceptible made few preventive visits but were high utilizers of acute illness care. Children from low-income families receive less care than those from middle or high-income families even with government-funded health insurance [8, 15-18-J. Comparing the actual and anticipated effects of Medicaid on utilization of children’s medical services, Roghmann et al. [19] observed little change in source, frequency, or purpose of care after the introduction of Title XIX. Using data from the 1975 Rochester Community Child Health Survey, Wolfe reports that children who have fewer available family
Use of Pediatric
Medical
Care
937
resources (i.e. working parents, low income, and few community resources) are less likely to receive their care from a private practitioner [17]. Similarly, Bite et al. [18] observed that, while the overall relationship between income and use has diminished, the association still exists for children’s use of medical services. While government-funded health insurance has provided poor families with health care purchasing power, government-sponsored health efforts have had little effect on other access problems. Having a regular source of care encourages the use of health services. independent of need. However, 18% of children from low-income families and 6% from middle-income families have no regular source of care. The problem of physician availability also influences children’s use of health services. While the number of physicians has increased over the past ten years, there are still not enough physicians in areas which the federal government has defined as medically underserved [16]. According to respondents in the National Health Survey, delay in seeking care for pediatric acute conditions is most often due to an inability to obtain an appointment [3]. While perceived health need is consistently related to use, the question of whether perception of health is a predictor of utilization independent of actual health status has not been answered. Andersen and Kasper [8] found that mothers who saw their children as less healthy had more physician contacts. Similarly, Tessler [lo] and Tessler and Mechanic [4] found a significant positive relationship between utilization and number of mother-reported child chronic health conditions. Unfortunately, only the Tessler and Mechanic study used a prospective design. They determined the number of motherreported child chronic health conditions prior to collecting the child utilization data. Consequently, only this study allows us to draw the conclusion that a measure of perceived need predicts use of pediatric health services. We can identify positive and negative correlates of pediatric medical care utilization but these factors do not explain the majority of variation in the use of medical services. Consequently, researchers have started to put more emphasis on the investigation of other types of factors, particularly psychosocial factors, such as stressful life events [20,21]. Initial studies examining stressful life events and their relationship to disease used a retrospective design; individuals were asked about past utilization, life change events, and the distress or readjustment accompanying these events [22]. Recognizing the problems inherent in using retrospective designs to investigate the effects of psychosocial factors (e.g. recall bias), investigators have recently gathered prospective data to examine this relationship 123,241. However, the low incidence of major diseases and the difficulty of obtaining ongoing medical examinations for a study population have forced many investigators to measure minor illness episodes, the detection of which are strongly influenced by illness behavior. Consequently, it is difficult to determine whether people actually become ill after experiencing a number of life events, or rather, whether people seek medical care more readily when distressed [25]. Several studies have examined the use of medical services in emotionally distressed or situationally stressed families [4, 26, 271. Roghmann and Haggerty found that stress, defined as “any event perceived by the family as upsetting” [26, p. 5211 with no illness present. increased utilization of medical services for mothers and children. However, they found that perceived stress, when accompanied by actual illness, decreased utilization for mothers
but increased
utilization
for young
children.
Using a diary method of reporting daily stressful events and symptoms similar to the one used by Roghmann and Haggerty, Gortmaker et al. [27] found that daily stress was related to family utilization, independently of reported symptoms. They also noted that life events recorded prior to the diary reporting period predicted utilization independently of reported symptoms and daily stressors. Unfortunately, neither study assessed the possible impact of the diary method on the outcome under investigation, utilization, although Gortmaker et ul. recognized that this method of reporting might influence use of health services. Tessler and Mechanic [4], looking at pediatric utilization in a prepaid group practice, observed that maternal psychological distress was positively associated with magnitude of
SARAH
938
MCCUE HORWITZ et al.
use of children’s services, although propensity to seek care was more important in the initial decision to seek care. In general, the results of these studies show that psychologically distressed and situationally stressed individuals have a greater tendency to seek medical care, particularly for their young children. The precise dynamics of this phenomenon are not well understood. However, there are at least three alternative explanations for this association among children (Fig. 1). One explanation is that distressed mothers have children who are sick more often or are more severely ill, possibly because of their exposure to stressful family situations. Meyer and Haggerty [28] found an association between acute family crisis and the onset of illness in their study of streptococcal infections. The results of this study are particularly important because illness was measured separately from utilization. A second study by Boyce et al. [24] that differentiated illness from utilization found that a high family life change score was predicitve of illness duration. Beautrais et al. [23] found that family life events were associated with outpatient utilization and hospital admissions for medical problems, such as lower respiratory illnesses, gastroenteritis, burns, and injuries. These results agree with the findings of an earlier retrospective study by Heisel et al. [22] in which four patient groups of children (rheumatoid arthritis patients, general pediatric admissions, surgical patients, and psychiatric patients) had experienced more frequent or more serious life events prior to the onset of their illnesses than had their healthy peers. A second possible explanation for the relationship between distress and utilization is that distressed mothers are more sensitive to their children’s symptoms, compared to nondistressed mothers [29]. In a recent follow-up of the children he studied in 196 1, Mechanic [30] found that several factors influenced the reporting of common physical symptoms including subjective reports of poor physical health and measures of psychological distress. Mechanic postulated that psychological distress may produce discomfort, increase body monitoring, and/or sensitize an individual to physical changes. Thus, the heightened perception of illness through increased attention to symptoms resulting from stressful life events could be another explanation for the increase in medical care utilization among distressed individuals. Third, it is possible that women experiencing psychological distress or confronting stressful situations deal in part with their distress or stress by seeking medical care for their children as a type of social support. This third hypothesis deals with a latent social function of medical care. Latent social functions performed by the medical care system were documented by Shuval in her study of new immigrants to Israel [31]. Looking at those individuals who used twice the average amount of medical services (more than 20 visits yearly), she found that socially isolated immigrants tended to be relatively high utilizers of care, regardless of their morbidity status.
la. b.
2.
3.
FIG.I.Alternate
Family Stress
-
Family Stress
_
Mother’s Distress
_
Mother’s Distress/ Family Stress explanations
__f
Risk of Disease Among Children Increased Severity of Disease Among Children HeightenedSympton Perception Increased Need for Social Support
_
Increased Utilization Increased Utilization
-+
Increased Utilization
-+
Increased Utilization
for the relationship between psychological stress and pediatric utilization.
distress
or situational
Use of Pediatric Medical Care
939
The latent function hypothesis has received little research attention. No investigation in the literature on pediatric utilization has directly tested the hypothesis and only a few studies have evaluated the effects of supportive, nonmedical interventions. In a field experiment designed to reduce children’s stress during hospitalization for tonsillectomy, Skipper and Leonard [32] found that social interaction between nursing personnel and the mothers of young children significantly reduced common physical signs of stress (e.g. elevated temperature, pulse rate, blood pressure, post-operative emesis, disturbed sleep, and long recovery periods) in children. Pless and Satterwhite [33] evaluated the impact of nonprofessional family counselors on families with chronically ill children. Significantly more of the chronically ill children from counseled families showed improvement on psychological scores when compared to control children whose families did not receive the lay intervention. Assuming that the medical care system provides certain latent services to patients, it follows that medical care can serve as a source of support for psychologically distressed families or families who are confronted with stressful life situations, particularly those who do not have strong support available from their social networks. The supportive function of social networks is a relatively new area of concern in psychosocial epidemiology and developed as an important concept due to observations documenting differential host resistance to stressful life events [34-40]. As potentially important as this concept is, there are several major questions that have not been resolved. One question concerns the relationship among support, acute and chronic stressful life changes, and illness. The precise questions of who will become ill, under what conditions. and through what biological pathways, are still to be answered. A second important question concerns the way in which social networks influence utilization. It is unclear whether social networks are related to utilization due to network health beliefs, inadequate network support, or symptom perception [41].
UTILIZATION
RESEARCH
PROBLEMS
In reviewing previous studies that examined possible determinants of pediatric utilization, we find that we cannot predict very well who is likely to use services. For instance, using ten variables Tessler [lo] could explain only ISO,:, of the variation in a measure of total utilization and 8’;;, of the variation in the use of well-child care. Some of this inability to predict pediatric health services use is attributable to conceptual and methodologic problems. including the need for a new utilization taxonomy, the ambiguity of cause and effect. the poor conceptualization of stress and support, and the need to develop a measure of health status independent of utilization. We will discuss each one of these weaknesses separately and, where possible, offer suggestions to remedy the problem.
An important conceptual problem in research looking at the determinants of pediatric utilization is the need for a new utilization taxonomy. Until recently, researchers have viewed pediatric health services use as a single outcome measure, typically the number of physician visits per year. The initial efforts separating physician visits into preventive and illness encounters have shown that utilization must be categorized in different ways, depending on the type of question a researcher is addressing [42]. The best way to classify visits for the purposes of etiologic investigations may not be the best way to classify visits for the study of utilization. For example, Ross and Duff [43] recently suggested that pediatric utilization should be broken down by type of visit (patient initiated or physician initiated) and by type of contact (i.e. preventive or curative). Initial research efforts conceptualizing utilization behavior in nontraditional ways, such as direct vs indirect risk preventive behaviors or telephone contacts vs office visits, have shown that various types of utilization may have different determinants. Roghmann and Hapgerty [26] found that the probability of telephone calls and contacts to outpatient or
940
SARAH MCCUE HORWITZ et al
emergency departments was doubled in situations where stress was combined with illness. However, there was little change in the probability of contact with any other department or the probability of an office visit. We know that medical care vists for acute illness do not occur randomly. Rather, they tend to cluster within specific bouts of illness called illness episodes. We know very little about the factors that influence repeat contacts within an illness episode compared to those factors that predict the number of episodes of illness. It may be that health-related variables, such as the presence of chronic conditions, symptom severity, or symptom recognition, are the important determinants of care seeking for a given health problem. However, once an individual contacts the health care system, system-related and/or psychosocial variables may prove to be more important determinants of service use for the same illness episode. Whether a parent makes two or six contacts with a health care provider for a single episode of otitis media in a 2 year old child may be related less to the child’s physical condition than to the parent’s financial resources or to the presence of a supportive social network. A detailed investigation of episode specific utilization will determine factors important for predicting initial health care contacts and subsequent health services use. Ambiguity of cause and efect Another conceptual deterrent to understanding the factors that influence pediatric health services use is the ambiguity of cause and effect that exists in this research. Some of this ambiguity is due to the use of retrospective rather than prospective designs in assessing utilization predictors. In a retrospective design, an individual is asked about the predictors of interest after the outcome events (utilization) have occurred. Additionally, utilization data has often been collected retrospectively, either through recall by patients or through medical records. With this type of design three types of problems can occur. First, if utilization data are collected retrospectively from medical records, outcome information may not be accurate due to inadequacies of the records. Second, because predictor information is recorded after the utilization period of interest, the outcome variable, utilization, may influence the predictors under investigation. Third, the threat of differential recall is always present in this type of research design. Furthermore, even with a prospective design where information about the predictors of interest is collected at the beginning of the follow-up period and utilization is observed weeks or months after baseline, it is difficult to determine whether changes in certain variables (especially psychosocial variables) preceded or succeeded changes in health status or utilization. This is attributable to the strong feedback system that exists between attitudes and behaviors. Individuals often behave in a way that they think is consistent with what they have said or, conversely, individuals have been known to jusify past behavior by adopting a particular point of view. For example, in a retrospective study, individuals who reported many dental visits saw themselves as less susceptible to tooth decay at interview time [25]. In a prospective study, however, those who saw themselves as less susceptible to tooth decay at baseline subsequently had fewer preventive visits [44]. While this reversal of findings is explainable by attitudinal and motivational theories, it makes the distinction of cause and effect quite difficult. This causal ordering becomes increasingly more complex when assessing daily fluctuations in psychosocial variables. If there is a long lag between the psychosocial event and the outcome of interest (e.g. pediatric utilization), the time ordering is less problematic. However, when the time period between stimulus and response is short, it is often difficult to distinguish the cause from the effect. Using the stress-physical symptoms relationship as an example, it is difficult to ascertain, even with a diary method of reporting, whether stressors or the perception of stress occurred before or after the reporting symptomatology. Robertson et a/. [45] could not estabish a causal sequence for diary data on family illness and social or emotional problems. While illness was correlated with social and emotional problems, neither looking at stressful events in the four days preceding illness nor looking at illness in the four days preceding stressful events produced a clear relationship. It cannot
Use of Pediatric Medical Care
941
be definitely determined whether stress produced symptoms, whether stress heightened sensitivity to existing symptoms, or whether underlying symptoms produced heightened sensitivity to stress. Inadequate conceptualization of stress, psychological distress and social support The relationship of social stressors, psychological distress, and social support to the use of pediatric medical services has received considerable attention. However, these factors are not carefully defined or used in any consistent way in previous investigations. In reviewing earlier work linking psychological distress and utilization, we find that the measures of distress elicited general anxiety-related symptoms. Consequently, previous work did not look at the relationships of other types of psychological distress (e.g. depression) or, more specifically anxiety, and use of pediatric medical services. If the relationship between distress and utilization is caused by anxiety and not other types of psychological maladjustment, then our understanding of children’s utilization determinants will be enhanced if we can establish this association. However, employing a nonspecific term such as psychological distress does not allow the development and testing of hypotheses that will help us to determine why family psychological problems are related to use of children’s services. Similar criticisms can be made about the relationship of social stressors to utilization. Stress has been defined in several ways ranging from family specific definitions (e.g. anything a family finds upsetting) to established lists of stressful life events with standard weights developed from a reference population. We need to understand what aspects of social stressors cause families to use health services for young children. In order to understand the relationship of stress to utilization, we must be able to differentiate whether it is simply an accumulation of disruptive events that prompts use of medical care, or rather, whether it is circumstances that a particular family perceives as upsetting that affects changes in its illness behavior. In addition, we need to know whether social stressors are linked to use of children’s health services in all socioeconomic groups since many of the studies linking stress and utilization have used low or lower middle-income populations. The concepts of social networks and support are not clear-cut nor are they easily measured. Research on the relationship of social networks to health care use has been retarded by imprecise definitions of social network characteristics, nonspecific hypotheses concerning their relationships to utilization, and by a confusion of social support and social networks. Social networks have been defined as an individual’s web of social ties and the quantity, quality and morphology of those ties. Thus, they have certain measurable characteristics (e.g. size and geographical proximity) and functions (e.g. support and role modeling). Social support, on the other hand, is an emotional or instrumental benefit perceived by an individual (i.e. a qualitative dimension of a network). Often. when investigators speak of social support, the assumption is made that all networks are supportive. When investigators make this assumption, they are not allowing all dimensions of networks to be tested [46]. The complexity of social networks is alluded to in the literature; however, little information can be found on what components of networks appear to account for the possible effects on health status and utilization behavior [46,47]. We advocate that three dimensions in the measurement of social networks be included in future investigations of pediatric utilization: (1) structural and interactional properties of the network, such as size, strength of ties, and frequency; (2) a subjective appraisal of perceived network support; and (3) a measure of a mother’s (or family’s) orientation or attitude toward using network resources. Inclusion of network properties provides a structured way to assess the resources available to a person. Once the network properties are determined, a subjective measure of their importance to the family must be determined. Finally. evaluation needs to be made of whether a mother (or family), because of her coping style, will be able to use the support available from network resources. This aspect of support, while receiving some justification in the literature [48-501, has received little empirical attention.
SARAH MCCUE HORWITZ et al.
942
Measurement
qf health status
We need to measure health status so that it can be used as a covariate when assessing the impact of sociodemographic, health attitude, health belief, psychological distress, social stressor, and social network variables on utilization. However, there are several problems associated with the measurement of health status, particularly how to measure health status independently of behavior (utilization) and subject perceptions (self-reports of health status). One attempt to assess health status is the development of different indices based on medical record information. This method allows for the measurement of health status only for children who have used health services and is not independent of utilization behavior. A second way of measuring health status is to use self-reports. Self-reports of children’s health status are gathered in two ways: by global health status ratings, usually done by the child’s mother at baseline or periodically in a follow-up period, and by daily reports of symptoms through the use of diaries. Unfortunately, these methods rely on self-report without corroboration from medical records or physical examination data. Mechanic [51] has noted that global self-reports of health status and the assessment of illness behavior (e.g. symptom recognition and utilization) may not be independent of one another. We can also make the same criticism of the diary method since, like the self-report of global health status, it may not be independent of symptom recognition or utilization. A third approach to measuring health status is through regular examinations by health care providers. This solution is less dependent on utilization than the other approaches but also has important problems associated with it. First, physical examinations of young children rely to a large degree on parent’s reporting of symptoms, and therefore, are somewhat dependent on self-reporting. Second, it is both expensive and impractical to examine a large group of children regularly, assuming families initially agree to such procedures. Third, by introducing such an obvious intervention into families’ lives, one would expect that their utilization behavior might be altered. Evidence of behavioral changes brought on by experimental observational without experimental intervention, the phenomenon [52]. If “Hawthorne Effect,” is a well recognized social-psychological behavior can be influenced by actively engaging subjects in our research, we must consider the potential impact on behavior of more intrusive interventions, such as conducting periodic health examinations to determine health status or having parents keep diaries to assess levels of symptoms or stress. We conclude that assessment of health status is a major methodologic problem in studying the determinants of utilization for which we currently have no ideal solution. The two solutions often employed, self-reports of health status and examinations by health care providers, have serious problems associated with them. Until we can satisfactorily assess health status independently of utilization and in an unobtrusive way, we have no way of testing the effects of other factors on utilization. Unfortunately, given the need to control extraneous factors such as health status when testing alternative hypotheses about utilization behavior, this problem presents a serious deterrent to understanding the determinants of health services use.
SUMMARY
AND
CONCLUSIONS
Even though investigators have expended considerable effort to understand why families use medical care for young children, we cannot predict very well who is likely to use services. We believe that some of this inability to predict the use of services is due to conceptual and methodologic problems present in earlier studies. Two of the problems we have identified, need for a new utilization taxonomy and ambiguity of cause and effect, can be remedied. Initial attempts to disaggregate total physician visits have shown that different variables are related to different types of utilization and that the amount of explainable variation differs according to the type of utilization an investigator chooses to study [4, 10,431. Similarly, while investigators have started to realize that assessing
Use of Pediatric
Medical
943
Care
utilization determinants using a retrospective design can produce ambiguous results, they must also recognize that, even in a prospective design, assessing daily fluctuations in psychosocial variables will produce unclear results particularly when the time period between the social stimulus or stressor and response (utilization) is short. Consequently, it will be necessary to use study designs that allow for a clear time ordering when attempting to evaluate the relationship of psychosocial factors to utilization. Additionally, if the prospective study uses utilization data based on patient perceptions or medical records, then the outcome data will be subject to recall errors and the well-known inaccuracies of medical records. Only a completely prospective study, one that assesses subjects at baseline and collects utilization information at the time of each medical visit, can avoid producing unclear results. The problem of inadequate conceptualization of social stress, psychological distress and social support will be more difficult to address. After many investigations of the relationship of life events to utilization, we still cannot identify what aspects of life events increase utilization, why the association does not hold for all families, or whether this association holds in all socioeconomic groups. Clearly, research which carefully compares standard life events, both as accumulations and as rated by families, with family generated upsetting events in different types of populations will help to clarify the relationship of stress to utilization. In addition, an evaluation of stress in the context of the overall coping strategies of families (i.e. family coping styles, personal and social resources) will aid in our understanding of how stress relates to utilization. Similar comments can be made about the relationship between psychological distress and utilization. We need to identify what kind of distress is related to utilization, whether it is related to all types of utilization or just certain types, and whether its impact is modified by other variables. Social networks and social support are relatively new research concerns. Investigations of the relationship of social networks to health care use have been retarded by imprecise definitions of social network characteristics, nonspecific hypotheses concerning their relationship to utilization, and by a confusion of social support and social networks. We have suggested that three dimensions of social networks be included in future investigations of social networks. In addition, we suggest that alternative hypotheses evaluating the possible ways in which social networks may relate to utilization be tested. Social networks may be related to various types of utilization in more than one way. While the first three problems we have identified can be at least partially resolved with current investigative tools, the measurement of health status independent of utilization represents a major methodologic problem for which we currently have no ideal solution. Until we can measure health status independently of utilization, we have no way of assessing the effect of other factors on utilization. Given the need to control extraneous factors such as health status when testing alternative hypotheses about families’ utilization behavior, this problem presents a serious deterrent to understanding pediatric utilization determinants.
REFERENCES I.
2. 3
Rundall TG. Wheeler JRC: The effect of income on use of preventive care: An evaluation of alternative explanauons. J Health Sot Behav 20: 397406. 1979 Dutton DB: Explaining the low use of health services by the poor: costs, attitudes of delivery systems? Am Sot
Rev 43: 34X-368,
Starfield
B: Special
Pediatrics
4.
1978
responsibilities:
7 I, 433S439.
The role of the pediatrician
and
the goals
of pediatric
education,
19X3
Tessler R. Mechanic D: Factors affecting children’s use of physician services in a prepaid group practice. Med Care 16: 33 46. 1978 5. Slessingcr DP. Tessler RC. Mechanic D: The effects of social characteristics on the utilization of preventive medical services in contrasting health care programs. Med Care 24: 392-404, 1976 6. Ezzati T: Ambulatory care utilization patterns of children and young adults. Vital and Health Statistic Series B(39). Hyattsville. Maryland: National Center for Health Statistics. 1978 1. Haggerty RJ. Roghmann KJ. Pless IB: Child Health and the Community. New York: John Wiley, 1975
944 8.
SARAH MCCUE HORWITZet al
Andersen R, Kasper JD: The structural influence of family size on children’s use of physician services, J Comp Fam Stud 4: 116-130,
9. 10. II.
12. 13. 14. 15.
16.
17. 18. 19. 20. 21.
LS, Kosa J, Alpert JJ ef al: Family size and the use of medical resources. In Family and Fertility. Liu WT (Ed) Notre Dame: University of Notre Dame Press, 1967 Tessler R: Birth order. family size, and children’s use of physician services. Health Services Res 15: 55-62. 1980 U.S. Department of Health and Human Services, Select Panel for the Promotion of Child Health. Better Health for our Children: A National Strategy. Public Health Service Publication No. 79-550-71, Vol. 3. Washington, DC: Government Printing Office, 1981 Becker MH, Drachman RH, Kirscht JR: Motivations as predictors of health behavior. Health Services Rep 87: 852-862, 1972 Kirscht JP, Becker MH, Eveland JP: Psychological and social factors as predictors of medical behavior. Med Care 14: 422431, 1976 Becker MH. Nathanson PA, Drachman RH er al: Mother’s health beliefs and children’s clinic visits: A prospective study. J Commun Health 3: 125-135, 1977 Silver GA: Child Health: America’s Future. Germantown, Maryland: Aspen Systems Corporation, 1978 U.S. Department of Health and Human Services, Select Panel for the Promotion of Child Health. Better Health for Our Children: A National Strategy. Public Health Service Publication No. 79-550-71, Vol. 5. Washington, DC: Government Printing Office, 1981 Wolfe BL: Children’s utilization of medical care. Med Care 18: 11961207, 1980 Bite TW, Eichhorn RL. Fox PD: Socioeconomic status and the use of physician services: A reconsideration. Med Care IO: 261-271, 1972 Roghmann KJ, Haggerty RJ, Lorenz R: Anticipated and actual effects of Medicaid on the medical care pattern of children. N EngI J Med 285: 1053-1057, 1971 Roghmann KL: Models of health and illness behavior. In Child Health and Community. RJ Haggerty, KL Roghmann, IB Pless (Eds). New York: Wiley-Interscience, 1975 Starfield B, van den Berg BJ, Steinwachs DM: Variations in utilization of health services by children. Pediatrics
22. 23. 24. 25. 26. 2-l. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45.
1973
Robertson
63: 633-641,
1979
Heisel JS, Ream S, Raitz R er al: The significance of life events as contributing factors in the diseases of children--II. Behav Pediatr 83: 119-123, 1973 Beautrais AL, Ferguson DM. Shannon FT: Life events and childhood morbidity: a prospective study. Pediatrics 70: 935-940, 1982 Boyce WT. Jensen EW, Cassel J er al: Influence of life events and family routines on childhood respiratory tract illness. Pediatrics 60: 609-615, 1972 Kasl SV, Cobb S: Health behavior, illness behavior, and sick role behavior. Arch Envir Health 12: 246266, 531-541. 1966 Roghmann KL, Haggerty RJ: Daily stress, illness, and the use of health services in young families. Pediatr Res 7: 520-526, 1973 Gortmaker SL, Eckenrode J. Gore S: Stress and utilization of health services: A time series and cross-sectional analysis. J Health Sot Behav 23: 25-38, 1982
Meyer RJ. Haggerty RJ: Streptococcal
infections in families: Factors altering individual susceptibility.
Pediatrics 29: 539-549, 1962 Mechanic D: The influence of mothers on their children’s health attitudes and behavior. Pediatrics 33: 444-453. 1964 Mechanic D: The experience and reporting of common physical complaints. J Health Sot Ftehav 20: 146-155. 1980 Shuval JT: Social Functions of Medical Practice. San Francisco: Jossey-Bass Publishers, 1970 Skipper JK. Leonard RC: Children. stress. and hospitalization: A field experiment. J Health Sot Behav 9: 275-286. 1968 Pless IB, Satterwhite B: Chronic illness in childhood: Selection, activities and evaluation of nonprofessional Family counselors. Clin Pediatr 403407. 1972 Cassel J: The contribution of social environment to host resistance. Am J Epid 104: 107-123, 1976 Cobb S: Social support as a moderator of life stress. Psychosom Med 38: 300-313, 1976 Gore S: The effect of social support in moderating the health consequences of unemployment. J Health Sot Behav 19: 157.-165, 1975 Henry JP. Stephens PM: Stress, Health, and the Social Environment. New York: Springer-Verlag, 1977 Kaplan BH. Cassel JC. Gore S: Social Support and Health. Med Care 15: 47-58. 1977 Nuckolls KB. Cassel J. Kaplan BH: Psychological assets, life crisis. and the prognosis of pregnancy. Am J Epid 95: 431-441. 1972 SOSd R. Kendell J. Klaus M P! nl: The effect of a supportive companion of prenatal problems, length of labor. and mother-infant interaction. N Engl J Med 303: 597-600, 1980 Dimsdale JE, Eckenrodc J. Haggerty RJ E/ al: Role of social supports in medical care. Sot Psychiat 14: 175-180. 1979 Greenlick MR. Larson DR. Mullooly JP (‘I ul: Determinants of utilization of health care systems. In Proc Int Conf on Health Technology. Cullen MF (Ed) San Francisco: Society for Advanced Medical Systems. Operations Research Society of America and Salutis Unitas, 1973 Ross CE, Duff RS: Returning to the doctor: The effect of client characteristics, type of practice and experiences with care. J Health Sot Behav 23: 119-131, 1982 Kegeles SS: Why people seek dental care: A test of a conceptual formulation. J Health Human Behav 4: 149-155. 1963 Robertson LS. Kosa J. Heagarty MC, (‘I u/: Changing the Medical Care System. New York: Praeger Publishers, 1974
Use of Pediatric
46. 47. 48. 49. 50. 51. 52.
Medical
Care
945
Wellman B: Applying network analysis to the study of support. In Social Network and Social Support. Gottlieb BH (Ed) Beverly Hills, California: Sage Publications, 1980 Gourash N: Help-seeking: A review of the literature. Am J Commun Psycho1 6: 413423, 1978 Schaefer C, Coyne JC, Lazarus RS: The health-related functions of social support. J Behav Med 4: 381-407, 1981 Pearlin LI, Schooler C: The structure of coping. J Health !%c Behav 19: 2-21, 1978 Fischer CS: To Dwell Among Friends: Personal Networks in Town and City. Chicago: University of Chicago Press, 1982 Mechanic D: Correlates of physician utilization: Why do major multivariate studies of physician utilization find trivial psychosocial and organizational effects? J Health Sot Behav 20: 387-396. 1979 Franke RH, Kaul JD: The Hawthorne experiments: First statistical interpretation. Am Sot Rev 43: 5% 643. 1978