Eighteen percent of school-aged children seen in primary care settings are identified by clinicians as having a psychosocial problem. 1,2 Another 5% to 10% have significant symptoms by parent report but are not identified by clinicians. Treatment and management of these children varies considerably. 3 Improved understanding of factors that account for such treatment variations could lead to more effective interventions. 4 Many studies have considered the effect of physician gender on health care variation. Male and female physicians differ on several characteristics: female physicians express more favorable attitudes towards psychosocial factors in patient care, 5 they spend more time with patients, 6,7 they are more likely to discuss psychosocial issues, and they give more encouragement and reassurance. 7,8 Treatment differences have also been demonstrated, primarily for preventive care 6,9,10 and especially for adult female conditions. 9–13 Several studies suggest that female physicians provide more counseling and testing but are less likely to prescribe medications or perform procedures. 6,14–16 However, physician gender differences have not been noted in the quality of care (other than provision of preventive care). 11,17 The few existing studies that have examined physician gender effects on psychosocial care for adults have found no consistent effects. 16,18–20
Differences between male and female physicians may represent true gender differences in how physicians practice. However, other patient and practice environment factors may also contribute to these differences. Patients who are female and more prevention-oriented may self-select to female physicians. Female physicians also tend to work in more structured practice settings, 21 in which organizational aids for preventive care (such as electronic reminders and standardized medical record forms) are frequently used. Only a few studies that showed physician gender differences in care have controlled for these factors. 6,22 No studies have considered whether differences in the patient-provider relationship might account for treatment variations. Most of the existing work on physician gender effects has been done on adult care, and we found no previous studies comparing female and male physicians’ management of psychosocial problems in children.
The Child Behavior Study gave us a unique opportunity to look at the effect of physician gender on psychosocial care for children. The Child Behavior Study collected data on the recognition and treatment of childhood psychosocial problems in nearly 20,000 visits by children to 366 pediatricians and family physicians across the country. Importantly, the CBS captured data on several physician, patient, and practice variables not available in other studies of physician gender. Our goals in this analysis were (1) to determine whether male and female primary care physicians differed in demographics, training, attitudes towards psychosocial care, and type of patients seen and (2) to examine whether male and female physicians differed in identification and treatment of childhood psychosocial problems. We hypothesized that female physicians would identify more psychosocial problems in children and would be more likely to manage psychosocial problems with counseling rather than medication.
The Child Behavior Study was conducted by Pediatric Research in Office Settings (PROS) 23 and the Ambulatory Sentinel Practice Network (ASPN), 24 2 large practice-based primary care research networks. Recruitment of PROS and ASPN clinicians into the study has been described fully elsewhere. 2 This study included 401 pediatric and family practice clinicians in 44 states, Puerto Rico, and 4 Canadian provinces. For this analysis, we excluded 29 nonphysicians and 6 physicians who enrolled children incorrectly. 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. 25–28
Children ages 4 to 15 years who presented for nonemergency care accompanied by a parent or primary caregiver who spoke English or Spanish were eligible for the study. Visits for procedures only were excluded. Children were eligible for the study for only 1 visit. Data were collected on 22,059 visits. Exclusions for visits with inadequate or missing data (909), visits for providers who enrolled subjects incorrectly (85 children), and visits to nonphysicians (1,102), yielded a final group of 19,963 visits to 366 physicians, including 172 women and 194 men.
Institutional review boards affiliated with PROS, ASPN, and the University of Pittsburgh approved procedures and consent forms. Study procedures have been described in detail elsewhere. 2 Physicians completed a brief survey about themselves, their practice characteristics, and their attitudes toward psychosocial care before enrolling subjects in the study. Before the visit, consenting parents completed a brief questionnaire assessing demographics, child and family functioning, and behavior problems. After the visit, the clinician completed a questionnaire describing identification and management of psychosocial problems.
Individual and Practice Characteristics.
Physicians reported information on age, gender, specialty, year when training was completed, completion of a residency rotation or fellowship in mental health care or child development, and region of the country. Practice factors included structure (multi-specialty group vs. single-specialty group or solo practice), degree of managed care penetration in the practice, and availability of mental health services on site. For physicians participating in the PROS network, additional information was available on race and practice ownership. Pediatricians were considered to be owners if they reported that more than half of their income derived from practices in which they were the owner, a partner, a shareholder, or on track to partnership.
Attitudes Toward Psychosocial Problems.
Physicians completed the Physician Belief Scale (PBS), developed by Ashworth et al 29 and modified by McLennan et al. 30 The Belief and Feeling Subscale contains items that measure provider attitudes about treatment of psychosocial problems. The Burden Subscale contains items such as, “One reason that I do not consider information about psychosocial problems is the limited time I have available.” Respondents rated their views on a 5-point Likert scale, ranging from “disagree” to “agree.” Lower scores represent higher psychosocial orientation.
Parents reported patient age, gender, race, and ethnicity. Race was coded as African American/black versus other. Ethnicity was coded as Hispanic origin or non-Hispanic origin (regardless of race reported). Parental education was classified as the highest level achieved by either parent (high school or less, any college, or any post college). Marital status was categorized as children living with married parents versus all others.
Insurance status was reported by the person most knowledgeable about billing or insurance in the office and included both payer (coded as Medicaid vs. commercial) and type (coded as managed care vs. fee-for-service). Separate variables identified the uninsured and those with Canadian coverage.
Child/Family Symptoms and Functioning.
Parent-Reported Behavior Problems.
Parents completed the Pediatric Symptom Checklist (PSC), a brief symptom list for primary care. 31,32 The original PSC consists of 35 items that are summed to provide an overall parental impression of psychosocial functioning. The PSC has been demonstrated to be a valid and reliable measure for diverse youth. 31,33–36 Gardner et al 37 have developed and validated a briefer version of the PSC, including 17 items and 3 subscales that identify as internalizing, externalizing, and attentional problems. 38 The sensitivity of the subscales ranges from 77% to 87% with specificity of 68% to 80%.
Family functioning was measured by use of the Family Apgar, a 5-item instrument designed to measure adult satisfaction with family support. 39 Lower scores indicate low satisfaction.
Parents indicated whether children had functional limitations due to physical health problems by using a 10-item scale adapted from the RAND child health measure. 40
Primary Care Responsibility.
The physician was considered to have primary care responsibility for the patient if he or she identified the child as “my patient.”
Reason for Visit.
The reason for the visit was coded on the basis of the physician report of whether the visit was for well-child care or preventive services, psychosocial concerns, or acute or chronic medical concerns.
Length of Visit.
Physicians recorded the time spent with each patient by choosing from 8 categories representing 5-minute intervals that ranged from 0 to 5 minutes to >36 minutes.
Other Visit Characteristics.
Visits were classified by season of the year because school referrals prompted a number of visits. To control for a possible Hawthorne effect, we included a dummy variable that identified subjects recruited early in the study period.
Recognition and Management of Psychosocial Problems.
Identification of a Psychosocial Problem.
Physician identification of a psychosocial problem was an affirmative response to the question, “Is there a new, ongoing or recurrent psychosocial problem present?” Psychosocial problems were defined as any mental disorders, psychological symptoms, or social situations warranting clinical attention.
Physicians characterized the type of psychosocial problem using 11 categories from Horwitz 1 modified through focus group discussions with clinicians and pilot testing. These 11 categories were later grouped into several categories following the model of Achenbach. 41 Externalizing problems included behavior/conduct problems. Internalizing problems included adjustment reaction/reaction to stress and anxiety or sadness. Attentional/hyperactivity problems formed the third category. Physical manifestations (sleep, enuresis, or eating problems), learning disabilities, childhood psychosis, chemical abuse/dependency, and mental retardation were grouped as other problems. Physicians noted whether the problem was new or a previously recognized problem.
Physicians indicated the treatment provided at that visit, including whether psychotropic medications were prescribed and whether the patient was referred for mental health treatment. Counseling at that visit was coded “yes” if the physician reported that he/she gave education about the problem/treatment or recommended behavioral interventions or if the patient received counseling by someone else at the visit.
Physicians reported whether they believed the family agreed with the physician about the nature of the child’s problem and the appropriate treatment.
Data Management and Analysis
Data for the present article were collected from October 1994 through June 1997. In this analysis, children are clustered by physician. We used the cluster option in Stata 42 to account for clustering of observations by physicians. Our first goal in the present study was to compare differences between male and female physicians with regard to physician, patient, and visit characteristics (Tables 1, 2, and 3). We used SAS version 6.12 to calculate unadjusted percentages or means and unadjusted (bivariate) P values.
Our second goal was to show how the effect of physician gender on identification and management of psychosocial problems changed after other patient, provider, and system variables were considered. First, we present unadjusted data showing the proportion of cases in which the physician identified a psychosocial problem, and among children with an identified problem, the proportion who received counseling, medications, or a referral in the office. Second, we calculated mean identification rates for male and female physicians (assuming all patient and physician characteristics were held constant) by estimating a multivariate model and obtaining predictions of the dependent variable (assuming all physicians saw the same set of patients) (Tables 4 and 5). To do this, the physician gender variable was recoded to 0 to obtain predictions for male physicians and recoded to 1 to obtain predictions for female physicians. The dichotomous variables were modeled using logit analysis. A thorough description of these econometric methods can be found elsewhere. 43,44 The models were estimated using Stata version 6.0 with the Huber/White heteroskedastic consistent estimator of the variance/covariance matrix with a correction for intracluster correlation within physicians. 42,45,46 Statistically significant differences in adjusted means were determined by use of the t statistic for physician gender from the multivariate models.
For the analysis examining whether the physician identified any psychosocial problems at this visit, the entire sample of children was included (n = 19,963). The analyses that considered what type of problem was identified (attentional/hyperactivity, internalizing, externalizing, and other) were limited to children who were identified as having any psychosocial problem (n = 3,797; presented in Table 4). The regressions that looked at treatment were limited to children who were identified by the physician as having a psychosocial problem and were not already receiving services (n = 2,468; presented in Table 5). We limited the sample in this way because we believed that physicians may view treatment decisions differently for children already receiving mental health services. Because nearly all the psychotropic drug prescriptions observed in the present study were given to children with attentional problems, we limited the analysis of medication use to children identified as having attentional/hyperactivity problems (n = 1,146; presented in Table 5).
The following covariates were included in these regressions: patient demographics (age, race, gender, parent education, parent marital status, and insurance); symptoms and functioning (family functioning, PSC internalizing score, PSC externalizing score, PSC attention score, number of clinician identified problems, whether a problem was previously recognized, and any health limitations); physician attitudes toward psychosocial care (Belief and Burden Subscales); practice characteristics (high percentage of managed care patients, region of the county, and availability of on-site mental health provider); visit characteristics (primary care relationship, season, length of visit, type of visit, and indicator for early enrollment into study); and physician characteristics (age, years of training completed, specialty, and practice setting and completion of residency or fellowship in mental health development). In the regressions that predicted identification of psychosocial problems and type of problem, number of clinician-identified problems and whether the problem had been previously recognized were dropped because these were closely related to the dependent variables. In the regression that predicted counseling, we did not include the length of the visit because counseling visits were on average 10 minutes longer than other visits.
We used a Bonferroni technique to handle the problem of multiple comparisons. Specifically, we identified that set of regression covariates as statistically significant in which the sum of the P values for the t tests of the regression weights was 0.05, summing across the identification and type of problem regressions as 1 group and across the treatment regressions as a second group.
The 366 participating clinicians enrolled on average 55 consecutive children. Ninety-one percent of eligible children across all sites participated. No differences in age or gender were detected in a comparison of participating with nonparticipating children. Children in the western US were slightly more likely to participate.
Bivariate Comparisons of Male and Female Physicians
Tables 1 through 3 present bivariate analyses comparing male and female physicians with regard to background, training and beliefs; patient demographics and health characteristics; and characteristics of the visit. Female physicians differed from male physicians in age, discipline, recency of training, and practice setting (Table 1). Data available only for pediatricians showed that female physicians were less likely to own their practice (37.6% vs. 62.4%) but did not differ in race. Female physicians viewed treatment of psychosocial problems as less burdensome compared with male physicians.
Female and male physicians also differed in the types of patients they saw. Children seen by female physicians were younger and more often female, minority, enrolled in public and managed care insurance, and not living with married parents (Table 2). Children seen by male physicians had more parent-reported psychosocial symptoms; however, children seen by female physicians were more likely to have a health limitation.
Male physicians more often reported having primary care responsibility for their patients. Among children with identified psychosocial problems, male physicians were more likely to report that parents agreed with their assessment of the child’s problem and the planned treatment (Table 3). Children seen by female physicians were more likely to be seen for well-child visits and less likely to be seen for psychosocial problem visits. Female physicians spent more time with their patients. Male and female physicians also differed in the season of the visit.
Management of Psychosocial Problems
In Tables 4 and 5, we compared male and female physicians with regard to identification and management of childhood psychosocial problems. Contrary to our expectations, unadjusted data suggested that female physicians were less likely to identify psychosocial problems compared with male physicians (18.2% for female physicians vs. 19.7% for males;P = 0.008). After controlling for physician, patient, and visit characteristics, physician gender was no longer significant (adjusted identification rates, 19.4% for female and 18.7% for male physicians;P = 0.510). Table 6 lists factors significantly related to identification: patient demographic characteristics, symptoms, and visit characteristics (especially type of visit and primary care relationship) had the strongest significant relationship.
Next, we considered whether male and female physicians differed in the kinds of problems they identified. In unadjusted data, female physicians were less likely to identify children as having an attentional problem (44.6% vs. 54.3%), more likely to identify an internalizing problem (35.1% vs. 29.5%), and more likely to identify other psychosocial problems (52.9% vs. 45.6%) compared with male physicians. However, as the adjusted means in Table 4 illustrate, these differences disappear after controlling for physician, patient, and visit characteristics. In each of these regressions, child age, gender, psychosocial symptom, and type of visit had the most consistent significant relationships with the type of problem identified (see Table 6 for more detail).
We expected female physicians to be more likely to give counseling compared with male physicians. Although the unadjusted data did not support this finding, they did suggest other differences between male and female physicians in treatment decisions. In unadjusted data, female physicians were more likely to refer patients to specialty mental health care (24.2% vs. 19.8%;P = 0.008). Female physicians were also less likely to prescribe psychotropic medications for children with attentional problems (53.4% vs. 64.5%;P = 0.001). Again, these observed differences disappeared after controlling for other factors (23.9% vs. 20.1%, P = 0.113 for referral and 58.7% versus. 61.3%, P = 0.392 for medications;Table 5). As shown in Table 7, factors most strongly and consistently related to treatment choices were visit characteristics, such as whether the condition was a previously recognized problem, whether the visit was for a psychosocial problem versus well-care or physical health problem, and whether the physician perceived family agreement with care.
In the present study of primary care management of child psychosocial problems, female clinicians expressed more positive beliefs about psychosocial care and had longer visits. Although unadjusted data presented differences between male and female physicians in the management of childhood psychosocial problems, these differences disappeared in analyses controlling for several physician, patient, and visit characteristics.
We have confidence in these findings because our study group was large and was drawn from primary care practices across North America. Male and female physicians who participated in this practice-based research network study may be more similar in how they identify and treat psychosocial problems than male and female physicians who did not participate in these studies. However, physicians in the present study were similar to national samples of physicians in demographic and practice characteristics. 47 The rate of identification and treatment of psychosocial concerns in the present study was similar to the rate reported in Horwitz et al, 1 the only comparable study.
We found that when male and female physicians see similar children in a similar context, the physicians managed the children in a similar way. Male and female physicians in the present study differed substantially in the kind of patients they saw, the type of visits they provided, and their relationship with their patients. In the unadjusted data, male and female physicians appeared to differ in their treatment patterns because they saw different patients in different types of visits and had different physician-patient relationships.
Patients seen by female versus male physicians differed in age, gender, and severity of illness. In particular, we observed that older children were more likely to see physicians of the same gender. Because attentional problems comprise the majority of psychosocial problems diagnosed in pediatric visits and these problems are 3 times more commonly diagnosed among boys as among girls, selection (either by parents, children or practices) to same-gender physicians meant that male physicians saw more boys. Thus, the higher unadjusted rate of psychosocial problem identification by male physicians is not surprising. In the adjusted data, it is child gender (as well as age and severity and other visit characteristics) rather than physician gender that explained the identification of psychosocial problems. For this reason, our findings underscore the importance of considering patient case-mix (including gender and type of problem) in comparisons of male and female physicians’ treatment patterns.
Different Relationships and Reasons for Visit
Compared with female physicians, male physicians more often saw patients for psychosocial problem visits and less often for well-child visits. Male physicians also were more likely to assert primary care responsibility for their patient and to believe that the family agreed with their diagnosis and treatment plan. Along with patient characteristics (such as age and symptom severity), these factors were significantly related to treatment choices and perhaps account for the nonsignificance of physician gender in use of medications, referral, and counseling. Why do male and female physicians differ in the kinds of visits and the relationships they develop with patients and families? Gender differences in practice characteristics and work hours may explain these differences in doctor-patient relationship and visit type. Female primary care physicians tend to work fewer hours than men (49 vs. 59 hours in 1 recent survey). 48 In addition, female physicians are more likely to work in larger practices and as salaried employees, 21 as we found among physicians in this sample. Hence, women may more often work in settings in which responsibility for specific children is shared by the clinic rather than assigned to a physician.
Although male and female physicians in our study expressed different attitudes about the identification and treatment of psychosocial problems, they nevertheless differed very little in how they managed these problems. External practice constraints such as practice structure and resources probably limit the degree to which physicians can act upon beliefs about psychosocial care. The average visit length in the present study was 12 minutes for pediatricians and 9 for family physicians. Productivity expectations and tight schedules may limit physicians’ ability to delve into complicated issues. Moreover, several factors may discourage physicians from identifying concerns that are hard to treat in the primary care practice. These include the lack of mental health resources, physicians’ lack of knowledge about such resources, and concerns about their quality. Families’ concerns about the stigma associated with mental health diagnoses also limit physicians’ willingness to broach these concerns. In addition, aspects of the practice environment may limit the identification and treatment of psychosocial problems: for example, some practices assign cases to the first available physician that day rather than to a continuity provider, thus reducing opportunity for the physician to learn about patients. Such policies may have a greater effect on female physicians, especially doctors who work part-time or in multiple locations. Future studies should investigate how such concerns affect care for psychosocial problems.
Although male and female physicians differ in attitudes toward psychosocial care, they were similar in their identification and treatment of such problems among children. Improving management of psychosocial problems depends on identifying modifiable factors that affect diagnosis and treatment; the present work suggests that characteristics of the practice environment, the patient-physician relationship, and patient self-selection deserve more research.
The authors wish to acknowledge the contributions of the Pediatric Research in Office Settings (PROS) network of the American Academy of Pediatrics, Elk Grove Village, Illinois; the Ambulatory Sentinel Practice Network (ASPN), Denver, Colorado; the Wisconsin Research Network (WReN), Madison, Wisconsin; and the Minnesota Academy of Family Physicians Research Network (MAFPRN), St. Paul, Minnesota.
1. Horwitz SM, Leaf PJ, Leventhal JM, et al. Identification and management of psychosocial and developmental problems in community-based, primary care pediatric practices. Pediatrics 1992; 89: 480–485.
2. Kelleher KJ, Childs GE, Wasserman RC, et al. Insurance status and recognition of psychosocial problems: A report from PROS and ASPN. Arch Pediatr Adolesc Med 1997; 151: 1109–1115.
3. Gardner W, Kelleher K, Wasserman R, et al. Primary care treatment of pediatric psychosocial problems: A study from PROS and ASPN. Pediatrics. 2000; 106: e44.
4. National Advisory Mental Health Council’s Clinical Treatment and Services Research Workgroup. Bridging science and service. Rockville, Md: National Institute of Mental Health; 1998.
5. Maheux B, Dufort F, Béland F, et al. Female medical practitioners: More preventive and patient oriented? Med Care 1990; 28: 87–92.
6. Bensing JM, Brink-Muinen AVD, deBakker DH. Gender differences in practice style: A Dutch study of general practitioners. Med Care 1993; 31: 219–229.
7. Roter D, Lipkin M, Korsgaard A. Sex differences in patients’ and physicians’ communication during primary care medical visits. Med Care 1991; 29: 1083–1093.
8. Bernzweig J, Takayama JI, Phibbs C, et al. Gender differences in physician-patient communication: Evidence from pediatric visits. Arch Pediatr Adolesc Med 1997; 151: 586–591.
9. Kreuter MW, Strecher VJ, Harris R, et al. Are patients of women physicians screened more aggressively? A prospective study of physician gender and screening. J Gen Intern Med 1995; 10: 119–125.
10. Franks P, Clancy CM. Physician gender bias in clinical decision making: Screening for cancer in primary care. Med Care 1993; 31: 213–218.
11. Hall JA, Palmer H, Orav EJ, et al. Performance quality, gender, and professional role: A study of physicians and nonphysicians in 16 ambulatory care practices. Med Care 1990; 28: 489–501.
12. Lurie N, Slater J, McGovern P, et al. Preventive care for women. Does the sex of the physician matter? N Engl J Med 1993; 329: 478–482.
13. Seto TB, Taira DA, Davis RB, et al. Effect of physician gender on the prescription of estrogen replacement therapy. J Gen Intern Med 1996; 11: 197–203.
14. Frank E, Harvey LK. Prevention advice rates of women and men physicians. Arch Fam Med 1996; 5: 215–219.
15. Rosen MP, Davis RB, Lesky LG. Utilization of outpatient diagnostic imaging. J Gen Intern Med 1997; 12: 407–411.
16. Banazak DA. Late-life depression in primary care: How well are we doing? J Gen Intern Med 1996; 11: 163–167.
17. Ruch-Ross HS, O’Connor KG. Immunization referral practices of pediatricians in the United States. Pediatrics 1994; 94: 508–513.
18. Morabia A, Fabre J, Dunand J-P. The influence of patient and physician gender on prescription of psychotropic drugs. J Clin Epidemiol 1992; 45: 111–116.
19. Robbins JM, Kirmayer LJ, Cathebras P, et al. Physicians characteristics and the recognition of depression and anxiety in primary care. Med Care 1994; 32: 795–812.
20. Badger LW, Berbaum M, Carney PA, et al. Physician-patient gender and the recognition and treatment of depression in primary care. J Soc Serv Res 1999; 25: 21–39.
21. Randolph L, Seidman B, Pasko T. Physician characteristics and distribution in the US. Chicago: American Medical Association; 1995.
22. Britt H, Bhasale A, Miles DA, et al. The sex of the general practitioner: A comparison of characteristics, patients, and medical conditions managed. Med Care 1996; 34: 403–415.
23. Wasserman RC, Slora EJ, Bocian AB, et al. Pediatric Research in Office Settings (PROS): A National Practice-Based Research Network to Improve Children’
s Health Care. Pediatrics 1998; 102: 1350–1357.
24. Green LA, Wood M, Becker L. The Ambulatory Sentinel Practice Network: purpose, methods and policies. J Fam Pract 1984; 18: 275–280.
25. Green LA, Hames CG, Nutting PA. Potential of practice-based research networks: Experiences from ASPN. J Fam Pract 1994; 38: 400–406.
26. Nutting PA. Practice-based research networks: Building the infrastructure of primary care research. J Fam Pract 1996; 42: 199–203.
27. Green LA, Miller RS, Reed FM, et al. How representative of typical practice are practice-based research networks? A report from the Ambulatory Sentinel Practice Network Inc (ASPN). Arch Fam Med 1993; 2: 939–949.
28. Wasserman RC, Croft CA, Brotherton SE. Preschool vision screening in pediatric practice: A study from the Pediatric Research in Office Settings (PROS) Network. Pediatrics 1992; 89: 834–838.
29. Ashworth CD, Williamson P, Montano D. A scale to measure physician beliefs about psychosocial aspects of patient care. Soc Sci Med 1984; 19: 1235–1238.
30. McLennan JD, Jansen-McWilliams L, Comer DM, et al. The Physician Belief Scale and Psychosocial Problems in Children: A report from PROS and ASPN. J Dev Behav Pediatr 1997; 20: 24–30.
31. Jellinek MS, Murphy JM, Robinson J. The Pediatric Symptom Checklist: Screening school-age children
for psychosocial dysfunction. J Pediatr 1988; 112: 201–209.
32. Jellinek MS, Murphy JM, Burns BJ. Brief psychosocial screening in outpatient pediatric practice. J Pediatr 1986; 109: 371–378.
33. Little M, Murphy JM, Jellinek MS, et al. Screening four- and five-year-old children
for psychosocial dysfunction: A preliminary study with the Pediatric Symptom Checklist. J Dev Behav Pediatr 1994; 15: 191–197.
34. Murphy JM, Ichinose C, Hicks RC. Utility of the Pediatric Symptom Checklist as a psychosocial screen in EPSDT. J Pediatr 1996; 129: 864–869.
35. Murphy JM, Arnett HL, Bishop SJ, et al. Screening for psychosocial dysfunction in pediatric practice: A naturalistic study of the pediatric symptom checklist. Clin Pediatr (Phila) 1992; 31: 660–667.
36. Murphy JM, Reede J, Jellinek MS, et al. Screening for psychosocial dysfunction in inner-city children: Further validation of the pediatric symptom checklist. J Am Acad Child Adolesc Psych 1992; 31: 1105–1111.
37. Gardner WP, Murphy M, Childs G, et al. The PSC-17: A brief pediatric symptom checklist with psychosocial problem subscales: A report from PROS and ASPN. Amb Child Health 1999; 5: 225–236.
38. Loney J, Milich R. Hyperactivity, inattention, and aggression in clinical practice. Adv Dev Behav Pediatr 1982; 3: 113–147.
39. Smilkstein G. The family APGAR: A proposal for a family function test and its use by physicians. J Fam Pract 1978; 6: 1231–1239.
40. Eisen M, Ware JE, Donald CA, et al. Measuring components of children’
s health status. Med Care 1979; 17: 902–921.
41. Achenbach TM. Child behavior checklist. Burlington, Vt: University of Vermont; 1988.
42. StataCorp. Stata Statistical Software, release 6.0. College Station, Tex: Stata Corp; 1999.
43. Greene WH. Econometric Analysis. 2nd ed. Englewood Cliffs, NJ: Prentice Hall; 1993.
44. Maddala GS. Limited-dependent and qualitative variables in econometrics. New York, NY: Cambridge University Press; 1993.
45. Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability
. Berkeley, Calif: University of California Press; 1967:221–233.
46. White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 1980; 48: 817–830.
47. American Academy of Pediatrics. Periodic survey of Fellows #32. Chicago: Division of Child Health Research; 1995.
48. Xu G, Rattner SL, Veloski JJ, et al. A national study of the factors influencing men and women physicians’ choices of primary care specialties. Acad Med 1995; 70: 398–404.