Steep increases in health care costs over the past decade have stimulated demands for cost containment, particularly among large corporate purchasers of group health insurance. Numerous strategies for controlling health care costs have been marketed under the concept of managed care. Data are lacking regarding the impact of these market-driven health care strategies on children, particularly the care of the neonate.

Academic health science center hospitals, including children's hospitals and their associated medical school faculty, often are depicted as being unable to deliver cost-effective health care.¹ Such institutions are deemed costly because of high rates of uncompensated care and cost shifting, personnel-intense and technology-oriented delivery systems, and their central role in development of innovative (and expensive) new medical techniques. Patient care by medical school physicians is viewed as costly, and possibly inefficient, in comparison to the private sector because of participation of physicians in training, detailed laboratory testing, and unwillingness to consider approaches that may compromise perceived standards of care. The high cost of neonatal intensive care is of particular concern. Despite the rhetoric surrounding these issues, little objective information is available concerning the cost of neonatal care provided by academic physicians compared with physicians associated with managed care organizations.

Indicators most commonly used to analyze resource utilization are hospital charges and length of hospital stay (LOS). Recent publications have emphasized LOS as the primary determinant of cost of neonatal care^2,3 because of its relationship to low birth weight. Reduced costs have been demonstrated in association with nursery strategies that reduce LOS.^4,5

It was the purpose of our study to determine whether neonatal care by academic physicians was more costly than that delivered by other neonatologists. To gain further insight into this complex issue, we examined resource utilization by two groups of neonatologists practicing in the common setting of a university children's hospital.

METHODS

Study Setting

Two groups of neonatologists provide primary care for neonates in this setting. An academic group (group A) consists of 23 full-time faculty neonatologists from the Department of Pediatrics, Baylor College of Medicine. A private provider group (group B) consists of three neonatologists associated with a large managed care organization.

Neonatologists in both physician groups initially place most infants 31 weeks gestation on mechanical ventilation and administer prophylactic surfactant (Survanta, Ross Laboratories, Columbus, OH). Subsequently, infants are weaned from positive pressure support as soon as pulmonary maturation and control of breathing allow. Faculty neonatologists or postdoctoral fellows from group A are present in the neonatal intensive care unit on a 24-hour basis, whereas group B does not provide round the clock in-house coverage. As a result, group A neonatologists participate in the early level III care of most infants 31 weeks gestation belonging to group B.

Data Source and Study Population

Data Analysis

Three levels of analysis were performed. First we compared multiple clinical variables among patients receiving care from the two physician groups. Table 1 depicts results of these comparisons. Stepwise linear regression analysis was then used to assess factors determining LOS. Results are summarized in Table 2.

Table 1. Study Population Descriptive Data* [View Table]

Table 2. Factors Predicting Length of Hospital Stay [View Table]

Finally, analysis of covariance was used to compare physician groups, although adjusting for differences in important clinical variables between the two physician groups and the effect of interaction among the variables. Variables analyzed included birth weight, gestation, sex, race, admission age, death, age at death, antenatal steroids, congenital anomalies, congenital heart disease, congenital diaphragmatic hernia, hyaline membrane disease, transient tachypnea, persistent pulmonary hypertension, congenital pneumonia, positive pressure ventilation, air leak, early sepsis, all sepsis, bronchopulmonary dysplasia, necrotizing enterocolitis, patent ductus arteriosus (PDA), cryosurgery, patent ductus arteriosus surgery, necrotizing enterocolitis surgery, level of care, intraventricular hemorrhage, hydrocephalus, seizures, and physician group. Clinical variables influencing LOS at a P  .20 or less were identified and used to develop a regression model testing for influence and interaction between these variables and the physician groups. This allowed LOS between provider groups to be adjusted for mean differences in both patient risk factors and assessment of interaction between provider groups and risk factors.

RESULTS

Patients of group A were older at admission (P = .006) than those of the other group and exhibited characteristics of a sicker population. Group A patients included more outborn infants (P < .001), level III patients (P < .001), infants with sepsis (P = .018), infants requiring mechanical ventilation (P < .001), and patients with PDA (P = .008). Group B had a higher percentage of infants with transient tachypnea of the newborn (P = .003). There were no significant differences in mortality, age at death, occurrence of air leak, bronchopulmonary dysplasia, or grade 3 to 4 intraventricular hemorrhage.

Mean patient charges were $104 571 ± $119 179 for group A and $97 236 ± 101 566 for group B (P = .527). Average LOS was 34.8 ± 36.1 days for group A and 36.0 ± 33.9 days for group B (P = .752). Regression analysis revealed that charges were primarily a function of length of hospital stay (R² = .95, P < .001). We therefore analyzed further for factors predicting length of hospital stay (Table 2).

Among all patients, factors significantly associated with LOS were level of care (P < .001), birth weight (P = .034), and gestation (P = .002). Influence of physician provider group did not reach statistical significance (P = .085). Among level III infants >31 weeks gestation, however, physician group was a significant determinant of LOS (P = .02). Among these patients, LOS for group A was 23.4 ± 24.2 days vs 38.9 ± 33 days for group B (P < .005, t test). Provider group was also a significant determinant of LOS among level II patients (P = .048), where group A LOS was 16.9 ± 13.8 days vs 22.3 ± 23.4 days for group B (P = .067).

Differences in risk and acuity of illness between the two patient groups tended to bias the comparison of LOS. These factors included whether the patient died, the need for mechanical ventilation, the level of care required, gestational age, antenatal steroid usage, and the presence of the following diagnoses: congenital anomalies, primary pulmonary hypertension, congenital pneumonia, necrotizing enterocolitis, sepsis, patent ductus arteriosus, and intraventricular hemorrhage. Correcting for these variables demonstrated that the LOS for group A was on average 7.8 days shorter than group B. The difference between physician groups depended on the complexity of the case. For example, LOS for group A was approximately 6 days shorter than that of group B when none of these risk factors were present and as much as 29 days shorter for patients who had all of these risk factors.

Using the average difference in LOS corrected for risk factors, we calculated the fiscal impact of a shorter hospital stay among patients cared for by group A. The average per diem charge for this patient population was $2915. Had group A, with a shorter stay of 7.8 days, provided care for all of the 137 patients of physician group B, a $3 114 969 reduction in net hospital charges might have been realized.

DISCUSSION

The authors realize the limitations of retrospective review. However, our data were drawn from a relatively large sample size. Certain group differences in clinical variables favored a longer LOS for group A, which had a significantly higher proportion of outborn patients and significantly more level III patients requiring mechanical ventilation. High risk infants born in the community and transferred to a level III perinatal center require neonatal intensive care more frequently and have greater morbidity than those born in tertiary centers.⁸ These outborn patients would be expected to have longer LOS and higher resulting costs.⁹ The larger proportion of deaths in group A, although not significant, tended to shorten LOS for that group. However, the adjusted difference in LOS persisted despite this variable.

Objective data regarding the effect of prepaid and other managed health care strategies on cost and delivery of services to children are scarce. In a prospective, randomized study Valdez et al¹⁰ reported no difference in total expenditures for care of children in a prepaid group health plan as compared with those in two fee-for-service plans. To our knowledge, no such studies have been done in the setting of acute neonatal care nor are we aware of published studies comparing resource utilization parameters in similar neonatal populations managed by different physician groups in the same hospital setting.

In our study, physician provider group was an important predictor of length of hospital stay. The data do not support the hypothesis that care by academic neonatologists is less efficient or more costly than that provided by other neonatologists. When adjusted for patient risk factors, the patients of full time faculty neonatologists in this hospital setting had shorter hospital stays than those of neonatologists associated with a managed care organization. This was particularly evident among infants older than 31 weeks gestation and those admitted for level II care. This distribution is not surprising because group A physicians participated in the early care of most group B infants younger than or 31 weeks.

Our data also reveal the many confounding variables that must be considered in obtaining an accurate view of patterns of resource utilization among physicians and identify potential pitfalls involved in so-called outcomes management analysis being used with increasing frequency by hospitals and insurance companies today. Unless appropriately adjusted for risk and patient acuity, such analysis may present a very distorted view of resource utilization by physicians caring for high-risk patients.

CONCLUSIONS