Variation in the Treatment of Children by Primary Care Physician Specialty FREE

THERE IS a large literature documenting the differences between the approaches of general pediatricians and family practitioners in the care of children; however, results have been conflicting. Studies fall into 3 broad categories: (1) studies indicating that family practitioners use a more aggressive or resource-intensive approach than their colleagues in pediatrics^{1 - 4} ; (2) studies indicating that general pediatricians use a less aggressive or resource-intensive approach than family practitioners^{5 - 7} ; and (3) studies indicating differences in the style, rather than the quantity, of treatment provided by general pediatricians and family practitioners.^{8 - 10}

The conflicting results may stem from 2 methodological limits in much of this literature — limited generalizability and limited explanatory potential. First, many of the studies comparing general pediatricians and family practitioners involve restricted samples. This research tends to focus either on specific geographic regions or on unique patient groups such as those with a particular kind of insurance,^{6 ,8} while the studies that involve nationally representative samples have focused on only a few conditions. Second, the research generally focuses on clinical rather than structural determinants of practice.^{6 ,11} Few efforts have been made to understand how practice and community characteristics affect the practice patterns of general pediatricians and family practitioners. In this article, we attempt to address the gaps in the literature. We use a nationally representative sample of general pediatricians and family practitioners to examine the role of physician specialty in determining reactions to 6 diverse vignettes designed to have multiple appropriate treatment plans, as well as examining how practice and community characteristics affect treatment plans.

The data for this study were drawn from the Community Tracking Study (CTS) Physician Survey, a telephone survey of licensed physicians that was funded by the Robert Wood Johnson Foundation (Princeton, NJ) and conducted by the Center for Studying Health System Change (Washington, DC). Interviews with 12 385 physicians were completed between August 1996 and August 1997. The sample frame for the survey was developed by combining lists of physicians from the American Medical Association and the American Osteopathic Association. The CTS survey includes physicians from 60 specific sites and a nationally representative sample. The response rate for primary care practitioners was 65.4%. To be interviewed, the physician had to be practicing at least 20 hours per week. The CTS project is described in detail elsewhere.^{11 - 12} Data weights allow for nationally representative descriptive statistics. The analysis in this article is limited to the 3002 primary care physicians (1735 general pediatricians and 1267 family practitioners) who responded to specific survey questions about the treatment of children. The relevant subsample is nationally representative.

As mentioned earlier, this article assesses physicians' responses to 6 vignettes designed to have multiple appropriate treatment outcomes. There is some debate as to the validity of physicians' responses to vignettes as a measure of actual practice patterns. While it is reasonable to believe that there is some deviation between self-reports and actual clinical behavior, multiple studies indicate that self-reports are a valid alternative to more accurate, but more disruptive or expensive measures.^{13 - 14}

The first vignette describes an otherwise healthy 10-year-old boy with long-term primary enuresis, and repeatedly negative urinalysis and cultures, and who has failed fluid restriction and environmental interventions. Physicians are asked for what percentage of such patients they would recommend desmopressin acetate.

The second hypothetical patient is an otherwise healthy 10-year-old boy whose parent calls to report a 2-day history of a temperature of 38.3°C, sore throat, nasal stuffiness, and no other signs or symptoms. Physicians are asked for what percentage of such patients would they recommend an office visit.

The third model patient is an otherwise healthy 10-year-old girl with a 3-day history of a temperature of 38.6°C, productive cough, tachypnea, and rales at the right base. She is taking fluids, is uncomfortable, but not in acute distress. Physicians are asked for what percentage of such patients would they recommend a chest x-ray.

The fourth model patient is a 4-year-old boy with eczema and seasonal asthma who has been managed with intermittent oral steroids and bronchodilators. The frequency of asthma attacks is increasing despite prophylactic use of inhaled steroids. Physicians are asked what percentage of such patients would they refer to an allergist for evaluation.

The fifth model patient is a 24-month-old girl with a history of 6 episodes of suppurative otitis media during the last year. The child was treated with antibiotics, with complete clearing. After her fifth episode, she was placed on prophylactic antibiotics, but had a recurrence that again responded completely to antimicrobials. She is in good health otherwise and has normal hearing. Physicians were asked what percentage of such patients would they refer to an ear, nose, and throat specialist.

The sixth model patient is an otherwise healthy full-term 6-week-old infant with a temperature of 38.3°C. Physicians are asked for what percentage of such patients would they recommend a sepsis workup, including at least a complete blood cell count, sterile urine, and blood cultures.

In addition to years of experience, age, and sex, we adjust for 7 physician characteristics: (1) board certification in a primary specialty, (2) whether a physician is a graduate of a foreign medical school, (3) principal specialty, (4) number of hours spent during the previous week in a medically related activity, (5) number of hours spent during the previous week in direct patient care, (6) number of weeks spent practicing in the previous year, and (7) income.

There are 3 categories of practice characteristics in this analysis: (1) practice ownership (ie, full, partial, or nonowner); (2) practice type (ie, solo practice or partnership, group practice of 3 or more physicians, health maintenance organization [HMO], hospital-based practice, medical school–based practice, or other); and (3) source of practice revenue (ie, percentage of practice revenue from Medicaid, percentage of practice revenue from Medicare, percentage of practice revenue from capitated managed care, percentage of practice revenue from all types of managed care).

There were 4 items in this group: (1) perceived change in the complexity of patients treated by respondents without referral; (2) perceived complexity of patients treated without referral even though referral is appropriate; (3) perceived change in referral rates; and (4) estimated percentage of patients for whom respondents serve as a gatekeeper.

Item 1 reads, "During the last 2 years, has the complexity or severity of patients' conditions for which you provide care without referral to specialists increased a lot, increased a little, stayed same, decreased a little, or decreased a lot?"

Item 2 reads, "In general, would you say that the complexity or severity of patients' conditions for which you are currently expected to provide care without referral is much greater than it should be, somewhat greater than it should be, about right, somewhat less than it should be, or much less than it should be?"

Item 3 reads, "During the last 2 years, has the number of patients that you refer to specialists increased a lot, increased a little, stayed the same, decreased a little, or decreased a lot?"

Item 4 reads, "Some insurance plans or medical groups require their enrollees to obtain permission from a primary care physician before seeing a specialist. For roughly what percent of your patients do you serve this role?"

There are 10 categories of community attributes in our analysis. The first 5 categories address health resources in the county where the physician practices: (1) active physicians per 100 000 residents; (2) short-term hospital beds per 100 000 residents; (3) members in traditional HMOs, or HMOs that deliver health services through a physician group that is controlled by the HMO unit; (4) members in the independent practice association (IPA) (HMOs, or HMOs that contract directly with physicians in independent practices, and/or with associations of physicians in independent practices, and/or with one or more multispecialty group practices); (5) members in other HMOs, or HMOs that contract with 2 or more independent group practices and HMOs that use a combination of model types.

The remaining community attributes are demographic and financial measures of the county in which the physician practices: (1) population density, (2) per capita income, (3) persons per capita living below the poverty line, (4) high school graduates per capita, and (5) college graduates per capita.¹⁵

This analysis has 2 phases. We begin by comparing general pediatricians' and family practitioners' responses to the ordinary least-square models. These models identify the portion of the difference between general pediatricians and family practitioners that can be attributed to or explained by specific characteristics. We used SUDAAN (Research Triangle Institute, Research Triangle Park, NC) software to account for the complex sampling design of the Community Tracking Study. In analyses not shown, we used the SAS (SAS version 7.0; SAS Institute Inc, Cary, NC) system to examine the effect of disproportionate lumping of responses to our outcome variables. We specified a series of accelerated failure time models, which parallel the ordinary least-squares models presented in this article. The results of the 2 analyses are essentially equivalent, so we assume that disproportionate lumping does not jeopardize our estimates.

The first model considers the relationship between specialty and treatment at the bivariate level. The second model assesses whether differences between general pediatricians and family practitioners can be attributed to physician characteristics such as years of experience and board certification. In the third model, we add practice characteristics, and assess whether treatment differences by specialty can be explained by practice characteristics such as the percentage of revenue attributed to managed care. In model 4, we add measures of referral patterns. Finally, in model 5, we control for community effects by adding dummy variables for sample sites. All analyses are completed with the version 8 of the stand-alone SUDAAN system.

Table 1 compares family practitioners and pediatricians on the covariates described earlier. Two conditions must hold for a group of covariates to play a significant role in explaining the sex differences noted earlier. First, there must be a significant difference between family practitioners and pediatricians with regard to the particular characteristic. Second, the covariate must play a significant role in predicting the 2 outcomes. Thus, Table 1 identifies those covariates, which might play a role in explaining specialty differences.

As illustrated in Table 1, relative to their colleagues in family practice, on average, pediatricians in this sample are slightly less experienced, but significantly more likely to be board certified than their colleagues in family practice. Pediatricians are also significantly more likely than family practitioners to have graduated from a foreign medical school and to be female. Family practitioners spend significantly more time doing medically related work and providing patient care. Not surprisingly, given their greater work effort, family practitioners have a higher average income than pediatricians. Relative to pediatricians, family practitioners are more likely to be full owners of their practices, but less likely to be partial owners. Family practitioners are also more likely to work in solo practice or partnership, but less likely to work in group practice or in HMO environments.

As expected, Table 1 suggests that pediatricians' practices receive a significantly larger portion of total revenues from Medicaid, and family practitioners' practices receive a larger portion of total revenues from Medicare. Pediatricians' practices also receive significantly more revenue from managed care.

There are some differences in the communities where family practitioners and pediatricians work. Pediatricians work in areas with a larger number of active physicians and short-term hospital beds per 100 000 people. As suggested earlier, pediatricians are significantly more likely than family practitioners to work in large metropolitan areas. The communities served by pediatricians have a slightly higher per capita income and a slightly lower number of persons per capita living below the line of poverty. They also have a slightly higher percentage of high school graduates.

Family practitioners elected a more active treatment approach for all but 1 of the 6 model patients. Family practitioners more frequently recommended desmopressin for patients with enuresis (58.9% vs 37.4%); an office visit for more patients with fever and nasal stuffiness (69.8% vs 63.3%); a chest x-ray for patients with a productive cough, tachypnea, and rales at the right base (51.3% vs 47.7%); referral of the 4-year-old patient with eczema and seasonal asthma to an allergist (64.3% vs 59.3%); and referral of patients with otitis media to an ear, nose, and throat specialist (49.7% vs 33.9%). However, family practitioners were significantly less likely than pediatricians to order a sepsis workup for a full-term 6-week-old infant with a temperature of 38.3°C (66.4% vs 81.1%) (Table 2).

Available covariates account for little of the variation in the treatment of all of the 6 model patients. In fact, results suggest suppression effects in several of the models. That is, controlling for available covariates actually exacerbates the specialty difference for 4 of the 6 model patients, thereby causing negative explanatory effects. Physician attributes, practice characteristics, referral patterns, and community attributes account for −6.7% of the specialty difference in the treatment of primary enuresis; −11.2% of the specialty gap in treatment of sore throat, fever, and nasal stuffiness; −91.3% of the specialty gap in treatment of productive cough, fever, and tachypnea; −7.8 of the specialty gap in treatment of eczema and seasonal asthma; 7.3% of the specialty gap in treatment of suppurative otitis media; and 14.8% of the specialty gap in treatment of a 6-week-old infant with a temperature of 38.3°C (Table 3).

In analyses not shown, we found that the only other covariate to have a consistently significant and sizable association with treatment decisions is years of experience. Each additional year of experience corresponds with: (1) a lower tendency to use desmopressin for an otherwise healthy 10-year-old boy with primary enuresis; (2) a greater tendency to request an office visit for a child with nasal stuffiness and a low fever; (3) a greater tendency to recommend a chest x-ray for a child with productive cough and rales at the right base; and (4) a greater tendency to refer a child with eczema and asthma to an allergist. The only outcome that is not significantly associated with years of experience is referral to an ENT specialist for a 24-month-old child with suppurative otitis media.

In this nationally representative study, physician specialty was a strong independent predictor of physicians' responses to 6 model pediatric patients with diverse medical problems. The extent of the specialty effect varies significantly, from a difference of 5.4% for referral of a child with eczema and asthma to an allergist, to 22.9% for the use of desmopressin for an otherwise healthy 10-year-old boy with primary enuresis. Furthermore, except for the workup of a 6-week-old infant for sepsis, family practitioners were more likely to choose a more resource-intensive approach compared with general pediatricians.

These findings are consistent with much of the literature on specialty-related differences in the treatment of specific pediatric problems.^{5 - 8} Unlike previous research, however, this analysis suggests that, in general, there is a tendency for family practitioners to elect a more resource-intensive approach across a diverse range of treatment outcomes. We also find that the extent of this trend depends greatly on the type of problem a patient may have.

Despite previous research, however, the tendency of family practitioners to elect a more resource-intensive treatment approach is somewhat unexpected. Given the distinct histories and philosophies of the 2 specialties, it is surprising that it is family practitioners, rather than general pediatricians, who are generally more aggressive in their treatment choices. The focus of family medicine on holistic approaches to health would lead one to expect that family practitioners adopt a less resource-intensive approach to health care. Consistent with this supposition are many studies that have shown that family physicians take a less active approach than general internists when treating adults.^{16 - 19}

Several studies attribute specialty-related differences in resource utilization across practices to differences in patient mix.^{20 - 22} (These studies suggest that physicians seeing patients with more complex problems use more resources.) This research is not directly relevant to our analysis since the physicians in the present study were responding to the same set of hypothetical patients. Nevertheless, because research indicates that pediatricians see patients with more complicated cases,^{23 - 25} one might expect that general pediatricians would use more resources than family practitioners, rather than less.

Differences in the patients of general pediatricians and family practitioners may in fact explain the scenario in which general pediatricians opted for a more resource-intensive approach (ie, the treatment of a 6-week-old infant with a temperature of 38.3°C). Data from the 1997 National Ambulatory Medical Care Survey²⁶ indicates that family and general practitioners provide 22.0% of all visits for children between 0 and 17 years of age, but only 16.6% of visits for children younger than 3 years. There is also some evidence that family practitioners play an even less critical role in the care of infants. In a study of 2 multispecialty primary care practices covering more than 45 000 patient visits, Onady²⁷ found that pediatricians handled all newborn nursery and pediatric admissions. Similarly, 2 studies comparing the treatment of children in rural hospitals indicated that family practitioners and pediatricians had similar case mixes for all groups of children except neonates.^{22 ,26} Melzer et al²⁸ found that pediatricians care for significantly more neonates weighing less than 2500 g, and Bertonlino and Gessner²⁹ found that pediatricians care for significantly more newborns with complications. Thus, it is possible that family practitioners are less likely than pediatricians to recommend a sepsis workup for the 6-week-old infant because they prefer, in general, to refer these types of patients.

Our relative inability to explain the difference between general pediatricians and family practitioners with physician characteristics, practice characteristics, and referral patterns is also noteworthy. We suspect that the practice differences stem from systematic differences in experience, knowledge, and/or confidence. This interpretation is derived from research suggesting that physicians facing more uncertainty tend to use more resources.²³ Although the vignettes in this analysis were designed to have multiple appropriate treatment plans, general pediatricians may have approached the model patients with greater confidence stemming from their greater experience dealing with children's health problems.

Finally, this analysis indicates that few variables beyond physician characteristics are consistently associated with the treatment patterns. Taken together, these findings raise questions about how much physicians should focus their areas of expertise. Historically, advocates of health care reform have stressed the need to generally refocus the health care system on primary care, so as to increase the coordination of patients' treatments and to ensure that patients receive comprehensive, cost-effective services.^{24 - 25} Yet, there is a growing body of literature indicating that specialists manage certain adult conditions in a more knowledgeable and efficient manner than general internists.^{30 - 32} This article adds to this literature by indicating that there may be benefits to having a more narrow focus of expertise.

These findings were limited by our inability to assess several factors that include, but that are not limited to, the exact profiles of physicians' patient panels and the extent to which physicians know their patients over time. As suggested earlier, the study is also limited by its focus on vignettes rather than actual practice patterns. We reiterate, however, that vignettes have been shown to be a valid method of assessing the quality of care delivered by physicians.³³ Finally, we note that our study assesses physicians at only one point in time, and thus, it cannot address the consistency or evolution of physicians' practice patterns.

In conclusion, although most general pediatricians consider themselves to be primary care providers, it is clear that their scope of practice is more focused than that of family practitioners. This research seems to suggest that additional specialization by pediatricians translates into a unique practice style that may ultimately result in more efficient or effective care under certain circumstances. What is not addressed by this analysis, however, is whether the exclusive focus on children, rather than families, causes pediatricians to offer less holistic or comprehensive services than those provided by family practitioners. Further research on the relative benefits of pediatrics and family medicine might help to clarify the most appropriate roles for both specialties within the greater health care system.

Accepted for publication July 26, 2002.

This research was supported by a grant from the Academy for Health Services Research and Health Policy, Washington, DC.

Corresponding author and reprints: Ann K. Boulis, PhD, Department of Sociology, University of Pennsylvania, 3718 Locust Walk, Philadelphia, PA 19104 (e-mail: ann@pop.upenn.edu).

There is a significant literature documenting differences in how general pediatricians and family practitioners treat children. However, that literature yields inconsistent results and fails to identify the proximate variables that ultimately cause specialty-related differences. We believe that the inconsistencies in the literature exist primarily because previous research has focused on unique populations and a limited range of clinical issues. In this study, we attempt to assess whether specialty-related differences in the treatment of children demonstrate consistent trends, and to explain why such specialty-related differences exist. This research begins to assess whether the primary care physician specialty relates to the quality and efficiency of care offered to children.