Objective. To examine the changes in identification of pediatric psychosocial problems from 1979 to 1996.
Research Design. Comparison of clinician-identified psychosocial problems and related risk factors among large primary care pediatric cohorts from 1979 (Monroe County Study) and 1996 (Child Behavior Study). Data were collected from clinician visit questionnaires developed originally for the 1979 study.
Setting. Private practice offices of 425 community-based pediatricians and family practitioners across both studies.
Patients. We enrolled all children from 4 to 15 years of age who presented for nonemergent services in primary care offices. The 1979 study included 9612 children seen by 30 clinicians and the 1996 study included 21 065 children seen by 395 clinicians.
Selection Procedure. Each clinician enrolled consecutive eligible patients for both studies.
Measurements and Results. From 1979 to 1996, clinician-identified psychosocial problems increased from 6.8% to 18.7% of all pediatric visits among 4- to 15-year-olds. We found increases in all categories of psychosocial problems, except for mental retardation. Attentional problems showed the greatest absolute increase (1.4%–9.2%) and emotional problems showed the greatest relative increase (.2%–3.6%). The use of psychotropic medications, counseling, and referral also increased substantially. In particular, the percentage of children with Attention deficit/hyperactivity problems receiving medications increased from 32% to 78%. These increases in psychosocial problems were associated with increases in the proportions of single-parent families and Medicaid enrollment from 1979 to 1996. Changes in clinician characteristics did not appear to be the source of increases in clinician diagnoses of psychosocial problems.
Conclusions. Substantial increases in the identification of psychosocial problems in primary care paralleled demographic changes in children presenting to primary care offices and in the larger population.
Community studies of adult populations have reported a gradually rising prevalence of mental disorders, especially depression.1,,2 The growing prevalence of adult mental disorders has greatly affected primary care practice because persons with these disorders are often treated in primary care settings. In addition, persons with mental disorders use more general medical services than persons without these disorders.3,,4
We lack similar data on changes in the prevalence of child mental disorders specifically or pediatric psychosocial problems defined more generally. However, community epidemiology studies suggest that both child mental disorders and psychosocial problems are common.5 Some investigators report mental disorder prevalence rates of 17% to 20% in community samples. Similarly, primary care clinicians state that 15% to 20% of children in their practices have psychosocial problems that require intervention or monitoring.6
These estimates of the prevalence of childhood psychosocial problems are greater than those obtained in the first large study of clinician-identified child psychosocial problems. The Monroe County Study (MCS)7 of 1979 reported data on the prevalence of psychosocial problems and management for 18 000 children from 30 pediatric offices in and around Rochester, New York. At that time, clinicians identified psychosocial problems in only 6.8% of all visits for children 4 to 15 years old.
Since the MCS, only 2 major studies have examined the identification of psychosocial problems in primary care settings for school-aged children, what Haggerty8 described as the “new morbidity.” Costello et al9 examined rates of clinician identification of child mental disorders in a metropolitan health maintenance organization in Western Pennsylvania in 1986. Horwitz et al10 conducted a study using a cohort of pediatricians around Yale University in 1990. Both studies compared clinician assessment of psychosocial problems with psychiatric diagnostic instruments and behavior checklists. Costello et al9 reported only on pediatric identification of psychiatric disorders and found that pediatricians identified mental disorders in 4% of visits for patients 8 to 16 years old. In contrast, Horwitz et al10 employed a broad definition of psychosocial problems and found that clinicians identified approximately 20% of all children 8 to 16 years old with psychosocial problems.
These studies were important descriptions of clinician identification of psychosocial problems. However, the differences in study design between Costello's9 study on the one hand and the MCS7 on the other mean that one cannot compare the identification and management of psychosocial problems by Costello9 with those obtained in the MCS, performed almost 20 years ago. In contrast, Horwitz's study10 used a definition of psychosocial problems similar to the MCS, but found a much higher rate of psychosocial problems. Whether the discrepant findings are attributable to different instruments, the use of single cities in each study, clinician characteristics, or to actual changes in the occurrence of psychosocial problems over time, is not certain.
Therefore, we designed an assessment of a national sample of primary care office visits to examine changes in clinician identification and management of child psychosocial problems since 1979. Our design was closely modeled on the MCS of 1979. We hoped to learn a) if rates of psychosocial problem identification and treatment had increased over time, and b) if they had increased, what portions of these increases might be associated with patient or clinician factors. We hypothesized that psychosocial problem identification and treatment rates would be higher in our sample than in the MCS. We also predicted that the prevalence of single-parent and Medicaid households would be much higher in our sample. Finally, we collected data on clinicians in the Child Behavior Study (CBS) sample to examine whether provider training, attitude and age were associated with interphysician variation in identification or treatment rates.
We obtained the original data from the MCS (1979)7from the University of Rochester and combined it with data from the CBS.6 The CBS was supported by NIMH and conducted in the Pediatric Research in Office Settings network (PROS) and the Ambulatory Sentinel Practice Network (ASPN) during 1995, 1996, and the first part of 1997.
The MCS was conducted in Rochester, New York, and surrounding towns. In 1979, approximately 730 000 persons lived in this area, including approximately 210 000 children. Seventy-four pediatricians practiced in the Monroe County Study area. The MCS researchers believed that these pediatricians provided >80% of all pediatric care in 1979. The MCS stratified the primary care pediatricians by practice type (solo, group, health center), from which systematic samples (every third in each category) were selected for participation. Only 7% (3 of 41) of the physicians asked to participate refused because of lack of interest in the study with the other 8 not participating attributable to either staff changeover or previous participation in the study by another clinician in their own practice. Overall, 30 of the 74 Monroe County clinicians participated in the study.
Each participating clinician reported on all consecutive eligible children enrolled over a 2-month period. Eligible children and adolescents included those from 0 to 18 years presenting to the clinician's office for nonemergent care with a parent or guardian. Overall, the 30 clinicians provided data on >21 000 visits by >18 000 children. For comparability, we report only the data on the first visits by the 9612 children 4 to 15 years old because the CBS did not re-enroll children for subsequent visits once they had participated in the study.
Physicians were recruited into the MCS through the University of Rochester with the endorsement of the Monroe County Pediatric Society. Participating physicians and their office coordinators received instruction either directly during a seminar with the consultant psychiatrist and study staff (16 clinicians) or a videotape of this seminar (14 clinicians). Clinicians filled out the Physician Visit Record with information on Medicaid status, the reason for the visit, diagnosis, and management of psychosocial problems. The office coordinators usually provided the demographic information.
The CBS was conducted in PROS11 and ASPN,12 2 large practice-based primary care research networks. PROS is a pediatric network that was established in 1986 and currently comprises >1500 clinicians from >480 practices in all 50 states and the Commonwealth of Puerto Rico. ASPN is a family medicine network that was established in 1978 and currently consists of 148 practices, with approximately 750 clinicians from 43 states and 6 Canadian provinces. Eighty-nine percent of PROS clinicians are pediatricians, 10% are nurse practitioners, and 1% are physician assistants. Eighty-five percent of ASPN clinicians are family physicians, 7% are nurse practitioners, and 8% are physician assistants. ASPN also collaborated with 2 regional networks to expand the number of participating family physicians. The characteristics of the Wisconsin Research Network and the Minnesota Academy of Family Physicians Research Network are similar to those of ASPN and contributed 38 and 24 participating clinicians, respectively. Recruitment of clinicians into the study has been described fully elsewhere.6
Clinicians were recruited from network practices that a) had previously completed PROS or ASPN research studies, b) were not participating simultaneously in other major studies, or c) expressed an interest in this study. This study included 395 clinicians representing 204 practices in 44 states, the Commonwealth of Puerto Rico, and 4 provinces in Canada. Participating clinicians were 43 years of age on average, 50% female, and completed their training 16 years before participating in the study. Pediatricians were 66% of the clinicians, while family practice physicians made up 26%. The other clinicians (7%) were physician assistants and nurse practitioners.
Previous research from both ASPN and PROS confirms the similarity of patients, clinicians, practices and clinical behaviors of physicians participating in primary care network studies with those identified in national samples.13–16 A survey conducted as part of the CBS17 showed no difference in demographic factors or practice characteristics among participating pediatricians and a random sample of primary care pediatricians from the American Academy of Pediatrics (AAP). AAP pediatricians, however, had minimally higher rates of patients with either private insurance or no insurance. As expected, both the MCS and CBS samples were representative of patients seen in private practice settings. Thus, minority and inner-city populations are underrepresented as compared with the US population.
Each participating clinician reported on a consecutive sample of approximately 55 children 4 to 15 years old, presenting for nonemergent visits in the presence of a parent or caregiver. Children were enrolled only once. Ninety-one percent of eligible children across all sites participated. We compared participating with nonparticipating children and detected no differences in age or gender. Children in the western United States, however, were slightly more likely to participate. We obtained results on 22 059 visits. Among those visits, 994 (4.5%) had inadequate or missing data sufficient to preclude further analyses, resulting in a study sample of 21 065 visits.
Network coordinators and staff recruited clinicians. Practices received training materials for the study including videotaped and written instructions. Clinicians completed the Clinician Visit Questionnaire with information on insurance status, reason for the visit, diagnosis, and management of psychosocial problems.
MCS and CBS Measures
In the MCS, clinicians indicated identification of a psychosocial problem by answering “yes” to this question: “Regardless of the purpose of this visit, in your opinion, does this patient currently have a behavioral, emotional or school problem, treated or untreated?” For the CBS, a focus group of clinicians in 1994 modified this question. Clinicians identified psychosocial problems by responding positively to the question, “Is there a new, ongoing, or recurrent psychosocial problem present?” We defined psychosocial problems as any mental disorders, psychological symptoms or social situations warranting clinical attention or intervention. Because clinicians in the later study may have included family or social problems, we examined the number of children with only family problems noted. The number was extremely small and did not change the overall results of the study. Clinicians in both studies coded severity as mild, moderate, or severe.
For both studies, clinicians identified the type of psychosocial problem(s) present using the World Health Organization classification scheme.18 On the recommendation of our clinician advisors, we changed the category “Hyperkinesis” to “Attention deficit/hyperactivity problems” (AHPs). Clinicians could, and often did, check more than one category of problems for both studies.
Clinicians in the MCS and the CBS reported on the reason for the visit (acute or chronic medical concerns, psychosocial concerns, or well-child care and preventive services) and whether the patient was their primary care patient, or a patient in their practice typically followed by another clinician.
For the MCS, physicians or staff members provided all of the patient demographic information from their records, including household structure, race, gender, and patient date of birth. In the CBS, patient race and household structure were reported by parents or guardians on the Parent Questionnaire. Parents reported both on race and ethnicity. Household structure was classified using the methods used in the MCS.
Data Management and Analyses
Dr Klaus Roghmann, the original statistician for the MCS, helped us obtain the MCS data tapes from the University of Rochester. All CBS data were sealed and submitted without identifiers to data entry sites where forms were visually inspected. Data were double-entered and transmitted to the University of Pittsburgh for analyses.
We calculated proportions of visits for identification, impairment and treatment for all patient visits among youth 4 to 15 years old and calculated χ2 tests to compare proportions. We estimated mixed logistic regression equations with random effects for the practices to calculate the distribution of risk factors predicting clinician identification. We included a random effect for the practices because of the sampling technique used for both studies. The term mixed refers to the use of both random and fixed effects in the regression equations. In these regressions, the fixed effects are the variables such as age and season of visit. We used the GLIMMIX (SAS Institute, Cary, NC) procedure in SAS (Version 7) to estimate the random effect. Classical regression techniques assume that selected samples are uncorrelated as well as being representative of the population being studied. Due to the fact that patients were recruited from clinician offices, this assumption is not valid because we expect that identification and treatment decisions are correlated within the same practice, but varied among practices. This variation tends to understate the significance levels of statistical tests for regression coefficients, leading to too-frequent rejections of null hypotheses of no effect.
Did Clinicians Identify More Psychosocial Problems in 1996 Than in 1979?
Table 1 reports a substantial difference in clinician-identified psychosocial problems from 1979 to 1996. The overall identification rate more than doubled (6.8%–18.7%) and a similar increase can be seen in each of the problem categories except mental retardation. The largest absolute percentage change was in AHPs (called hyperkinetic in the MCS), which increased from 1.4% in 1979 to 8.5% in 1996. The change in emotional problems from .2% to 3.2% represented the largest relative increase.
There are several possible explanations for this marked increase in clinician identification of psychosocial problems. The comparison of identification rates in Table 1 rests on the generalizability of the findings from the primary care offices in the MCS to other primary care offices in the United States in 1979. If Monroe County was similar enough to the rest of the United States with respect to primary care in 1979, then the results in Table 1 suggest that there has been a substantial increase in the identification of psychosocial problems in the United States in the intervening years. We considered several possible explanations for the increase using the available data.
Is the higher rate of identification attributable to CBS clinicians knowing their patients better than MCS clinicians?
Were the MCS clinicians different from the clinicians who participated in the CBS in a way that would have lead them to identify fewer children with psychosocial problems in 1979? Could this difference be related to changes in training or geographic location?
Can the increase in clinician identification of psychosocial problems be attributed to differences in patient demographic and risk factors between the MCS patients and the patients in the CBS?
Is the higher identification rate attributable to a basic change in the acceptance of treatment for AHPs by clinicians, parents, and teachers?
Did CBS Clinicians Know Their Patients Better Than MCS Clinicians?
There was a substantial change in familiarity between pediatric patients and their clinicians from 1979 to 1996. More than 94% of children in both studies were seen in their usual practice. However, 43% of the CBS children were seen by a clinician other than their primary care provider, whereas only 12% of the MCS children were seen by someone other than their primary care provider. Previous work by Horwitz et al10 and Kelleher et al6 showed that if the patient was not one of the clinician's usual patients, the probability that the clinician would recognize a psychosocial problem was diminished. Had the CBS clinicians been as familiar with their patients as the MCS clinicians appeared to be, the identification rate for the CBS might have been even higher. We note, however, that seeing their own primary care provider in the MCS did not have a statistically significant effect on identification after controlling for patient risk factors.
Were the MCS Clinicians Different From Clinicians Participating in the CBS?
We considered 2 possible differences between the 2 sets of clinicians that could have resulted in fewer children being identified as having psychosocial problems in 1979.
The first was that clinician training might have been different for clinicians in the CBS when compared with the MCS. Because we did not have data on the training of the MCS clinicians, we compared identification rates within the CBS sample to assess whether any possible changes in clinician training may have been important. We hypothesized that if changes in training mattered, younger clinicians and those with more residency or school training in behavioral or psychiatric issues would have higher identification rates. We found few differences in identification across clinician characteristics. Clinician gender, region, and amount of specialty training had no detectable effect on identification in the CBS.6 Nor were there any systematic differences in identification rates among clinicians who finished training in different decades. The highest rates were for those who had trained in the 1970s, followed by those who had trained before 1970. We conjectured that older clinicians in this sample tended to have slightly older patients who, in turn, were more likely to have psychosocial problems. However, the correlation between clinician age and patient age was only .04. Alternatively, older clinicians may have been more adept at identifying or discussing psychosocial issues.
Second, we considered that the MCS clinicians may have been different either in attitude toward treating psychosocial problems or in some other way that was not measured. Kelleher et al,6 using preliminary data from the CBS alone, found that physician beliefs about treatment effectiveness had no impact on the identification of psychosocial problems after controlling for patient demographics and problem severity.
Could there be some other unmeasured difference between the MCS and the CBS clinicians? We note that some of the clinicians who participated in the MCS also participated in the CBS 17 years later, however, the MCS sampled >40% of all the pediatricians in Monroe County and the CBS sampled only 15% of all Monroe County pediatricians. We repeated our analyses between the MCS and the CBS using only Monroe County samples from both studies. The different identification rates of Monroe County clinicians in 1979 and 1996 are shown in Table 1. The differences between identification rates among Monroe County clinicians in the 2 time periods were virtually as large as those between all CBS clinicians and the MCS clinicians. The small apparent difference in identification of psychosocial problems between Monroe County clinicians in 1996 and other CBS clinicians was not statistically significant in an F-test (P = .22), which adjusted for the correlation among patients, even without controlling for patient risk factors that are discussed next.
Did Patient Demographic and Risk Factors Differ Between the MCS and the CBS?
Again we examined 2 possible categories of differences between 1979 and 1996 which may have had an impact on the identification rate of psychosocial problems. Of primary concern was the possibility of a divergence in demographic structure between Monroe County and the rest of the country. To evaluate the extent to which this could be possible, we conducted a number of comparisons.
First, we considered whether Monroe County was demographically comparable to the nation as a whole in 1979. The population figures for 4- to 15-year-old children from the 1980 Census indicate that there was little or no difference between Monroe County and the United States with respect to gender or the percentages of European-Americans and African-Americans. The Hispanic population was underrepresented in Monroe County relative to the United States. Although the percent of single-parent families with children <18 years old was 16.8 in the United States versus 18.7 in Monroe County, the percent of these families below the poverty line was a little lower in Monroe County relative to the United States (38.0% vs 40.3%).
This comparability to national demographic figures persists today with Monroe County showing comparable increases in the proportion of residents living in poverty and <18 years old. Although Monroe County had slightly fewer minority residents than the country as a whole, this number increased over the past decade.
Next we considered the possibility that children in the CBS were at greater risk of being identified as having a psychosocial problem than the children in the MCS.
Based on the χ2 tests summarized in Table 2, children were more likely to be recognized as having a psychosocial problem if they were male, older, enrolled in Medicaid, or not living with both parents. From 1979 to 1996, the proportion of children seen for primary care visits who were not living with both parents increased from 15% to 25%. These results were identical to increases reported in US Census Bureau data.19–27 In both studies, children not living with both parents were more likely to have a psychosocial problem. In addition, the percentage of children enrolled in Medicaid almost tripled during the interval between the 2 studies.
Mitigating these more adverse conditions in 1996 were the small differences in age and gender between the MCS and the CBS. Children in the CBS were slightly more likely to be female and younger despite the identical age ranges chosen for comparability between the studies. This potential bias means that the identification rates in the CBS are somewhat lower compared with the MCS than they would have been had these 2 factors been the same in the 2 samples.
We also examined χ2 tests for each sample separately to see whether the predictors of clinician identification differed between the MCS and the CBS. The only results that differed between the studies were the effects of different seasons and whether the patient saw his or her primary care clinician, which was noted previously.
We used a mixed logistic regression analysis with both the common and divergent sets of patient risk factors described above to summarize their effect on the chance of being identified with a psychosocial problem. We included separate regressors for the MCS and the CBS for each of the divergent factors by crossing them with (0,1) indicators for participation in each of the studies. The divergent factors included the different seasons, whether or not the clinician reported the child as their patient, and whether or not the visit was reported by the clinician as a well-child visit. The remaining common factors that have 1 coefficient for both studies were: not living with both parents, Medicaid enrollment, age, sex, and minority status. The regression results are reported in the first half of Table 3. The most important demographic factors appeared to be Medicaid enrollment and male sex, which increased the likelihood of identification, followed by cohabitation with both parents, which decreased it. Provider familiarity with patient was a factor for the CBS patients, but not for those in the MCS. Patient age and minority status also appeared to have small effects on clinician identification. The one source of comparison that we did not have, but would certainly have had an independent impact on the results, was the parents' perception of whether or not there was a problem. This information was not a part of the MCS.
We combined these various risk factors in a summary score that was the estimated probability of being identified based on the mixed logistic regression. For example, a child with a risk score of 20% would have had a combination of risk factors leading to a 1 in 5 chance of being identified with a psychosocial problem.
Figure 1 illustrates how the distribution of risk factors has changed from 1979 to 1996 by presenting separate estimated distributions for each of the studies using these summary risk scores. The dark area under the MCS curve contains the MCS children who have the highest risk scores, those in the 95th percentile or above. The MCS children at this 95th percentile of risk scores had more than a 1 in 6 (or 17%) chance of being identified as having a psychosocial problem.
The hatched area under the CBS curve represents the proportion of children in the CBS who fell above the risk score that defined the 95th percentile in the MCS. This risk score was reached by the 60th percentile and above in the CBS sample. Put another way, if we definedhigh risk in terms of the prevalence of risk factors in 1979, there was an eightfold increase in the percentage of children athigh risk between 1979 and 1996. Thus, Fig 1 indeed shows that risk factors were more prevalent in 1996 (ie, the mean risk score for the CBS of 19% is higher than the MCS 7%). However, the separate distributions also indicate that based on the high-risk standard of 1979, many more children in the CBS would have been considered at high risk for being identified by clinicians as having a psychosocial problem.
Can the Differences Be Explained by AHPs?
If there was a fundamental shift in the awareness and acceptance of treatment for AHPs by clinicians, parents, and teachers, then it is possible that clinicians identified more psychosocial problems in children because parents were more willing to bring children with AHPs to primary care clinicians for treatment. This is, after all, one of the possible causes for the increased resource use by children with psychosocial problems in primary care. Evidence of such a shift can be found in the different percentages of children with AHPs who received medications during the 2 separate periods. Nearly 78% of children with AHPs in the CBS received psychotropic drugs, whereas only 32% of MCS children with AHPs received them.
An assessment of psychosocial problems independent of the clinicians, such as the parent-reported reason for the visit or a behavior inventory, would have afforded us a more careful test of this hypothesis, but no such information was collected for the MCS. We can shed some light on the question by considering the likelihood of identification for the rest of the sample, excluding those identified as having AHPs. The second set of columns in Table 3 shows the results of a regression similar to the one performed on the full sample from both studies. In this regression children with AHPs in the CBS and the MCS have been removed. The results are not changed dramatically for the risk factors in the regression, except for male sex, some seasonal variations, and well visits (all of which could be expected considering the drop of AHPs from the regression sample) and a reduction in the importance of the clinician's familiarity with the patient.
Clearly, the increase in AHPs in primary care had a significant impact on the overall increase in the identification of psychosocial problems between 1979 and 1996. Without AHPs the rate of identification among CBS patients was nearly double that of MCS patients, instead of almost triple for all psychosocial problems.
The rapid growth of clinician-identified psychosocial problems among children and adolescents presenting for primary care visits is consistent with the increase in parent-reported behavior problems in national surveys.28,,29 Data from community-based studies of the prevalence of psychiatric disorders using structured diagnostic interviews also indicate a trend toward increased rates of emotional and behavioral disorders in children.5 This broad range of evidence suggests that behavioral and emotional problems have increased among children in the United States during the past 18 years.
Some portion of the increase in clinician-identified psychosocial problems might be related to changes in clinicians' perceptions of psychosocial problems. Clearly, clinician training has changed markedly for both pediatricians and family practitioners since the MCS. Primary care residency programs now include more ambulatory training and more psychiatric and behavioral pediatric and developmental training. However, neither our own studies6 nor others10 have found associations between enhanced training at these gross levels and better identification of psychosocial problems in primary care.
Patient and family characteristics have changed dramatically since 1979. Demographers have noted a marked increase in the number of children living in poverty during these years, both in numbers and as a share of their population. Only 16% (10.4 million) of children <18 years old were living in poverty in 1979, whereas 21% (14.6 million) were so situated in 1996.19–27 Poverty is a known risk factor for the development of behavior problems and emotional and behavioral disorders in childhood, although we do not fully understand the specific mechanisms by which poverty affects children and their development.30 For example, how these demographic changes relate to the increased prevalence of parental substance abuse, parental depression, and residential instability—all risk factors for psychosocial problems among youth—is not fully elucidated.1,,231–33
The increase in single-parent households from 1979 to 1996 was associated with the increased rate of clinician-identified psychosocial problems. Children from single-parent households are more likely to develop behavior problems.31 In both the MCS and the CBS, children from single-parent households were roughly twice as likely to be identified with psychosocial problems, although some component of this may reflect additional attentiveness by clinicians to children at risk.
Limits of This Report
Caution is advised in the interpretation of our study findings because of several design limitations. First, the studies had slightly different data collection procedures. The principal differences were the sampling of clinicians for participation and the number of patients enrolled per clinician. Both studies included volunteer clinicians engaged in office-based primary care practice. However, the early study (MCS) used a stratified sampling scheme while the later study (CBS) included all clinicians eligible from a large national primary care research network. The early study (MCS) also used an enrollment strategy that included many patients per clinician during a 2-month period, while the later study (CBS) enrolled a smaller target number of patients per clinician over an unspecified time period, but many more clinicians. There was also a minor difference between the studies in the wording of the question regarding the identification of a psychosocial problem. Some of the clinicians16 in the early study were trained in a seminar while the remainder14 and all of the clinicians in the later study were trained by videotape.
Besides these design characteristics, some variables could only be assessed superficially in studies that gather data from clinician reports. For example, only 1 indicator of socioeconomic status, Medicaid enrollment, was common to both studies and eligibility for this program expanded in the late 1980s through 1996 to include families with somewhat higher income. Finally, because clinician report was the sole data source for the early study, no verifying information on parent or child perceptions of psychosocial problems was available.
Our finding of a dramatic increase in the rates of clinician-identified childhood psychosocial problems over the last 2 decades raises questions about a decline in the well-being of some children in the United States. The apparent association of this trend with parallel adverse changes in childhood poverty and the proportion of single-parent households suggests that the means to prevent childhood psychosocial morbidity lies beyond the primary health care system, and hence beyond the scope of this article.
Even though clinicians have no direct influence over the socioeconomic status of their patients, our findings suggest a need to restructure outpatient primary care for children and adolescents. Clinicians in 1996 reported that almost 19% of all pediatric visits involved a child or adolescent with psychosocial problems requiring attention or intervention. As predicted by Haggerty8 25 years ago, psychosocial problems are the most common chronic condition for pediatric visits, eclipsing asthma and heart disease. Moreover, they are among the most disabling of pediatric conditions, with mental health symptoms accounting for fully one-third of all school days missed by adolescents.34 Psychosocial problems are becoming the centerpiece of pediatric primary care for school-aged children and pediatrician surveys suggest that they are among the most time-consuming and frustrating problems to deal with in routine practice.35
The use of brief and rare office visits to clinicians for school-aged children has not been documented to be effective in preventing or managing psychosocial problems. Instead, primary care clinicians should forge partnerships with mental health professionals for direct initial assessments and management of psychosocial problems, as described by Pincus.36 The use of parent advocates and home visitors for partial support and behavioral interventions may also be more effective than current practices.37 Group well-care and related support groups may allow better exchange of parenting information and social interventions that have longer lasting effects.38 Recent efforts to develop primary care-friendly classification systems for psychosocial problems and related training materials may help if incorporated into ongoing education and payment systems.39
Although these novel approaches to improving mental health services in primary care are laudable, several trends in the organization and financing of care pose a threat to greater integration of medical and mental health services.40 In particular, managed care incentives that encourage clinicians to see greater numbers of patients per day or patients to change clinicians more often will decrease recognition and treatment in primary care.6 Similarly, the increasing use of behavioral health carveouts may diminish the willingness of mental health providers to work creatively with primary care clinicians. We discuss these changes elsewhere.41
Clearly, prevention and treatment of mental health problems of children must involve the multidisciplinary efforts of medical, mental health, social service, and education professions and even insurers because the roots of children's emotional problems are multifactorial. Attempts to solve these problems solely in the medical sector have not been successful, nor could they be. A coordination of efforts of all those interested in the well-being of children is essential to reverse this disturbing trend. Although the specific choice of treatment structure and process will be related to the capabilities of clinicians and facilities as well as the needs of a given population, the need to change can no longer be denied, nor solutions delayed.
This study was supported by a grant from the National Institute of Mental Health (Grant No. MH50629, Principal Investigator: Kelleher) and the Health Resources and Services Administration Maternal and Child Health Bureau (Grant No. MCJ-177022) and the Staunton Farm Foundation of Pittsburgh.
We wish to acknowledge the contributions of the Monroe County Study Investigators including Irving D. Goldberg, MPH; Klaus J. Roghmann, PhD; Thomas K. McInerny, MD; and Jack D. Burke, Jr, MD, MPH. Dr Roghmann was especially helpful in obtaining the 1979 data.
- Received July 28, 1998.
- Accepted July 13, 1999.
Reprint requests to (K.J.K.) 3510 Fifth Ave, Suite 100, Pittsburgh, PA 15213. E-mail:
- MCS =
- Monroe County Study •
- CBS =
- Child Behavior Study •
- PROS =
- Pediatric Research in Office Settings (network) •
- ASPN =
- Ambulatory Sentinel Practice Network •
- AAP =
- American Academy of Pediatrics •
- AHPs =
- Attention deficit/hyperactivity problems
- ↵Haggerty RJ, Roghmann KJ, Pless IB. Child Health and the Community. New York, NY: John Wiley and Sons; 1975
- Costello EJ,
- Edelbrock C,
- Costello AJ,
- Dulcan MK,
- Burns BJ,
- Brent D
- Horwitz SM,
- Leaf PJ,
- Leventhal JM,
- Forsyth B,
- Speechley KN
- Wasserman RC,
- Slora EJ,
- Bocian AB,
- et al.
- Wasserman RC,
- Croft CA,
- Brotherton SE
- ↵American Academy of Pediatrics, Division of Child Health Research. Periodic Survey of Fellows #32. Elk Grove Village, IL: American Academy of Pediatrics, Division of Child Health Research; 1995
- ↵US Bureau of the Census. General Social and Economic Characteristics: New York. Washington, DC: 1980. Census of Population, Publication No. PC80-1-C34:1980
- ↵US Bureau of the Census. General Population Characteristics: United States Summary. Washington, DC: 1980. Census of Population, Publication No. PC80-1-B1:1980
- ↵US Bureau of the Census. General Social and Economic Characteristics: United States Summary. Washington, DC: 1980. Census of Population, Publication No. PC80-1-C1:1980
- ↵US Bureau of the Census. General Population Characteristics: New York. Washington, DC: 1980. Census of Population, Publication No. PC-1-B34:1980
- ↵US Bureau of the Census. General Population Characteristics: United States. Washington, DC: 1990. Census of Population, Publication No. CP-1-1:1990
- ↵US Bureau of the Census. General Population Characteristics: New York. Washington, DC: 1990. Census of Population, Publication No. CP-1-C34:1990
- ↵US Bureau of the Census. Social and Economic Characteristics: United States. Washington, DC: 1990. Census of Population, Publication No. CP-2-1:1990
- ↵US Bureau of the Census. Social and Economic Characteristics: New York. Washington, DC: 1990. Census of Population, Publication No. CP-2-34:1990
- ↵Lamison-White L. Poverty in the United States: 1996. Current Population Reports, Series P60-198. Washington, DC: US Government Printing Office; 1997
- ↵Zill N, Rogers C. Recent trends in the well-being of children in the United States and their implications for public policy. In: Cherlin AJ, ed. The Changing American Family and Public Policy. The Changing Domestic Priorities Series. Washington, DC: Urban Institute Press; 1988:31–115
- ↵Coiro MJ, Zill N, Bloom B. Health of our nation's children: National Center for Health Statistics. Vital Health Stat. 1994;10
- Watson JE,
- Kirby RS,
- Kelleher KJ,
- Bradley RH
- Bailey V,
- Graham P,
- Boniface D
- ↵Taylor JA, Davis RL, Kemper KJ. Health care utilization and health status in high-risk children randomized to receive group or individual well child care. Pediatrics. 1997;100(3). URL: http://www.pediatrics.org/cgi/content/full/100/3/e1
- Kelleher KJ,
- Wolraich ML
- Kelleher KJ,
- Scholle SH
- Copyright © 2000 American Academy of Pediatrics