healthcare charges and use in commercially insured children enrolled in managed care health plans in...

Maternal and Child Health Journal, Vol. 4, No. 1, 2000

Healthcare Charges and Use in Commercially InsuredChildren Enrolled in Managed Care Health Plans inWashington State

Charles Maynard, PhD,1,2,3 Scott Ramsey, MD, PhD,1 Thomas Wickizer, PhD,1 andDouglas A. Conrad, PhD1

Objective: To determine the relative importance of enrollee, physician, medical group, andhealthcare plan characteristics as determinants of healthcare use and expenditures in commer-cially insured children �18 years of age enrolled in managed care health plans. We focusedon the effects of age and benefit level, the two most important predictors of cost and utilizationin our study of adults. Methods: This study included 67,432 commercially insured childrenwho were between 1 and 18 years of age, and were cared for by 790 primary care physicians,who practiced in 60 medical care groups in Washington State. Plan enrollment and utilizationdata for 1994 were linked to a survey of medical care groups contracting with three managedcare health plans. Benefit level for each enrollee was defined as low, medium, or high and wasbased on cost sharing by the health plan for hospitalization, outpatient care, and emergencydepartment services. The three outcome measures included estimated total per member peryear charges, number of ambulatory visits, and hospital days. Results: In multivariate analysis,enrollee age was the most important determinant of total charges, with younger childrenincurring higher charges and utilization. For children 5 years and younger, mean total permember per year charges were $617 in the low-benefit category and $878 in the high category(p � .0001). These differences were less apparent for children 6–12 years ($355 versus $420,p � .012), and were not statistically significant for children 13 years and older ($503 versus$552, p � .14). The annual number of visits increased with benefit level for children of allages. Conclusions: Enrollee age and benefit level were the most important determinants ofhealthcare use and expenditures in children enrolled in managed care health plans.

KEY WORDS: Children; expenditures; utilization; managed care; healthcare plans; health servicesresearch; multivariate statistical methods; commercial insurance.

INTRODUCTION

Very little is known about how the change fromfee-for-service to managed care has influenced thedelivery of health services for children, particularly

1Department of Health Services, University of Washington, Seat-tle, Washington.

2Health Services Research and Development, Department of Vet-erans Affairs, Seattle, Washington.

3Correspondence should be directed to Charles Maynard,PhD, Alcohol and Drug Abuse Institute, 1107 NE 45th, #205,Seattle, Washington 98105; e-mail: [email protected].

29

1092-7875/00/0300-0029$18.00/0 2000 Plenum Publishing Corporation

those who are commercially insured (1–3). However,the degree to which providers assume financial riskfor services they provide can influence not only qual-ity of care, but access and utilization as well (1). Ourprevious study of primary care physician compensa-tion and healthcare expenditures in adults enrolledin managed care health plans found that physiciancompensation method was not significantly relatedto use and charges of health services per person,but that enrollee, physician, and health plan benefitfactors were the major determinants of cost and utili-zation (4). Most of our knowledge of the effect ofmanaged care on children has come from demonstra-

30 Maynard, Ramsey, Wickizer, and Conrad

tion projects comparing a Medicaid health mainte-nance organization (HMO) with fee-for-service Med-icaid (3). In 1994, children comprised nearly one thirdof the HMO enrollment, but only 29% of the UnitedStates population (1).

In contrast to individual patient and family fac-tors, little is known about how physician, medicalgroup, and healthcare plan characteristics affecthealthcare charges and utilization in children, partic-ularly those in managed care plans. Female pediatri-cians appear to spend more time with their patientsthan their male counterparts; in addition, parentsseem to be more satisfied with the care provided byfemale pediatricians (5). A randomized trial of fee-for-service versus staff model HMO found that chil-dren with cost-sharing fees-for-service plans hadfewer medical contacts and received fewer preventiveservices than those assigned to the HMO (6).

Previous studies have identified age and race askey predictors of utilization, with younger childrenusing more services and Black children using fewerthan expected services (7). Other studies have shownthat psychosocial factors also affect the utilization ofservices. In Canada, where universal health insur-ance eliminates insurance barriers to obtaininghealthcare, children’s physical health status as per-ceived by their parents was associated with frequencyof use (8). In a study of 5- to 11-year-old childrenenrolled in a single HMO, child health need andmaternal patterns of healthcare use distinguishedhigh from low users of care (9). In addition, children’suse of services appears to remain relatively constantover the childhood years; that is, those who start outto be high users of service remain so at the end ofthe childhood years (10).

This brief review of the literature indicates thatlittle is known about how managed care influencesthe cost and utilization of health services for children,particularly those enrolled in private managed careplans. Given the large number of children coveredin private or non-Medicaid HMOs, it is valuable toknow more about how managed care influences thedelivery of services to children.

Since enrollee age was the most important pre-dictor of cost and utilization in our study of adults,we examined whether enrollee age, defined by threecategories, was associated with other enrollee, physi-cian, medical group, or healthcare plan characteris-tics, and if it was associated with expenditures andutilization. Second, following the focus on compensa-tion method in our study of adults, we determinedwhether physician compensation was associated with

expenditures and utilization in children. Finally, wedetermined by using multivariate statistical methodsthe relative effects of enrollee, physician, medicalgroup, and health plan characteristics on healthcareutilization and expenditures in children. Addition-ally, given that age and benefit level were the mostimportant predictors of charges and the number ofannual visits in the adult study, we assessed the rela-tionship between benefit level and these two depen-dent variables for the three different childhood agecategories.

Although specialty care is considered in thisstudy, the major focus is on the primary care of chil-dren, defined by the American Academy of Pediatricsas ‘‘accessible and affordable, first contact, continu-ous and comprehensive, and coordinated to meetthe health needs of the individual and family beingserved’’ (11).

METHODS

Patient, Physician, and Medical Group Samples

A detailed description of the study methodologyhas been published (4). This study included 67,432children, who were between the ages of 1 and 18years and were continuously enrolled in one of threemanaged care organization (MCO) health plans dur-ing calendar year 1994. The study sample did notinclude newborns or children under the age of 1 year.Children covered by publicly funded programs suchas Medicaid were not included in the data providedby the MCOs. The children in this study were caredfor by 790 primary care physicians who were identi-fied by the MCOs as having children assigned totheir panels.

A separate survey of clinic practices with respectto physician compensation and health plan paymentwas mailed to 76 medical groups in Washington Stateand was completed by 62, a response rate of 82% (4).There were 2 medical groups whose physicians didnot have children in their panels; therefore, this re-port considers 60 medical care groups in which these790 primary care physicians practiced.

The medical groups in this study were identifiedby four MCOs that agreed to participate in the studyand provide individual enrollee data. We had exten-sive meetings and negotiations with the MCOs toensure that the identification of enrollees and physi-cians was complete and that all relevant utilizationdata were obtained. These MCOs included a large

Healthcare Charges and Utilization in Children 31

staff model HMO, a major network model HMO, apreferred provider organization (PPO), and a groupmodel HMO. In the staff model HMO, physiciansare employed by the HMO and have an exclusivecontract with the MCO. In the network arrangement,the MCO establishes a contract with the individualmedical group. The PPO, which has elements of bothHMO and traditional fee-for-service, has arrange-ments with both medical groups and individual physi-cians not in medical groups, but typically does notimpose actuarial risk on providers. With the excep-tion of the group model HMO, the plans assigned aprimary care physician to each enrollee. Since thegroup model HMO did not do this, enrollees fromthis MCO were not included in this analysis.

Dependent Variables

Patient-level charges and utilization data ob-tained for 1994 contained all utilization, includingspecialty care. There were three measures of cost andutilization in this study. First, per member per year(PMPY) charges included primary and specialty careas well as hospital services, but did not include phar-macy or dental claims. PMPY charges were deter-mined by assigning resource-based relative valuescale weights to each ambulatory service or proce-dure and then multiplying them by $46.23, the aver-age dollar conversion factor for commercial healthinsurance plans in Washington State in 1994. Hospitaldays were converted to dollar equivalents by firstapplying all-payer diagnosis-related group weights toeach discharge, and then dividing by the mean lengthof stay for that discharge. This number was thenmultiplied by the dollar payment per hospital dayamong commercially insured patients in WashingtonState in 1994. Second, the total number of ambulatoryvisits was defined as those to primary care physicians,specialists, and other providers. A third dependentvariable was the annual number of days hospitalized.Finally, children who had no healthcare utilizationduring the year were identified.

Independent Variables

Enrollee Characteristics

Items of interest for individual enrollees wereobtained from MCO enrollment files and includedage and gender. Enrollee age was categorized as (1)

1–5 years, (2) 6–12 years, or (3) 13–17 years. Benefitlevel was determined by measuring the extent of cov-erage for inpatient hospitalization, hospital outpa-tient or physician office care, and emergency depart-ment visits. Depending on the health plan share ofpayment, coverage for each of the three componentswas scored as 0 (low), 1 (medium), or 2 (high). Add-ing the three scores resulted in an overall level ofcoverage defined as low (0–3), medium (4), or high(5–6).

The network HMO and the PPO had literallyhundreds of options with respect to type and extentof coverage, so it is difficult to characterize benefitsbeyond this simple measure. We did not assess cover-age for the screening, dental, and vision services cov-ered under the Medicaid Early Periodic ScreeningDiagnostic Treatment (EPSDT) program. It is likelythe range and types of services in our study were notas extensive as those offered by the EPSDT program.In addition, Medicaid does not require copays,whereas copays are important in all commercial plansin this study.

Physician Characteristics

Age, gender, and medical specialty were ob-tained from the three health plans, the WashingtonState Medical Association, the American MedicalAssociation’s Physician Masterfile, and the Directoryof Medical Specialists. Self-reported specialty wasdefined as pediatrics, family practice, internal medi-cine, or in very few cases, general practice.

Medical Group Characteristics

Details of variables contained in the medicalgroup practice survey have been published (4).Briefly, information about group size, location in theSeattle Metropolitan Area, multispecialty status, uti-lization management protocols, physician compensa-tion method, and distribution of revenues from allhealth plans was obtained.

For both fee-for-service and at-risk managedcare contracts, a utilization management score wascalculated with 1 point assigned for doing each ofthe following: (1) preauthorization for referral to spe-cialists; (2) preauthorization for hospital admissions;(3) preauthorization for outpatient procedures; (4)concurrent management of specialty referrals; (5)concurrent review of hospital stays; (6) information


feedback to primary care physicians; (7) clinicalguidelines; (8) case management for high cost, cata-strophic episodes of care; and (9) internal utilizationreview by physician peers.

Physician compensation was categorized as (1)salary only; (2) �50% salary-based; (3) productiononly; (4) �50% production-based; or (5) other. Pro-duction means that individual physicians were com-pensated on the basis of their individual work effort,generally measured as fee-for-service equivalent ser-vices provided to patients.

Health plan payment, measured by the percent-age of the group’s revenues, was categorized as (1)full-risk capitation (group at risk for utilization ofprimary care, specialty, hospital, and ancillary ser-vices utilization); (2) professional capitation (at riskfor primary care and specialty services only); (3) pri-mary care capitation (at risk for primary care servicesonly); (4) fee for service plus a withhold (withholdonly returned if utilization of services is below certaintarget amount); and (5) fee-for-service without risk(not subject to withhold or other form of risk ac-count).

Health Care Plan Characteristics

Based on information provided by the healthcare plan, we determined whether the health planpayment put the medical group at financial risk (yes/no) for utilization of all healthcare services.

Statistical Methods

First, we compared enrollee, physician, medicalgroup, and healthcare plan characteristics by catego-ries of enrollee age with the chi-square statistic forcategorical variables such as enrollee gender and one-way analysis of variance for continuous variables suchas physician age. To compare measures of utilizationand expenditures by categories of enrollee age, weused the chi-square statistic and one-way analysis ofvariance, and for assessing the association betweenphysician compensation method and the dependentvariables of interest, we used one-way analysis ofvariance.

Analyzing the relative effects of enrollee, physi-cian, medical group, and healthcare plan characteris-tics is complicated by differing units of analysis. Be-cause of this, mixed model analysis of variance inwhich the individual physician and medical group

were modeled as random effects was employed (4,12, 13). In this model, the standard errors of theregression coefficients are adjusted for the nesting ofphysicians within medical group and of individualenrollees within physicians’ panels. Three separatemodels, using independent variables listed in TableI, were constructed with dependent variables PMPYtotal charges, number of ambulatory visits, and num-ber of hospital days. In order to normalize the distri-butions of the dependent variables, the natural loga-rithm of the PMPY charges and hospitals days wasemployed, as was the square root of the number ofambulatory visits. Finally, in order to assess the effectof benefit level on PMPY charges and the annualnumber of visits within the three categories of age,the analysis of variance test for linearity was used.This test is appropriate for independent variablessuch as benefit level that have an inherent order.

RESULTS

Bivariate Results

Enrollee, Physician, and Medical GroupCharacteristics

Table I shows the distribution of the four levelsof variables by the three categories of patient ageand for children of all ages. Of the 67,432 childrenin this study, 44% were in the 6- to 12-year age group;the mean age was 10.1 � 4.5 years, with a median of10. The distribution of girls was similar across thethree age categories, while a slightly higher propor-tion of the youngest age group was in the high-bene-fit category.

The average age of the 790 primary care physi-cians was 43 � 7 years with a range of 28 to 75 years.The specialty distribution of the physicians, 30% ofwhom were women, was 68% family practice, 19%pediatrics, and 13% internal medicine, with less than1% general practice. There was no association be-tween physician specialty and gender. When viewedat the enrollee level, these results were somewhatdifferent, as seen in Table I. Children in the youngestage group were more likely to have women and pedi-atricians as primary care physicians.

Of the 60 medical groups in this study, 48% hadbetween 3 and 5 physicians, 25% between 6 and 19,and 27% had 20 or more physicians. Multispecialtyclinics accounted for 30% of the groups, 65% ofwhich were located in the Seattle Metropolitan Area.


Table I. Enrollee, Physician, and Medical Group Characteristics by Enrollee Age

Age (years)

All ages �5 6–12 13–1767,432 13,517 29,503 24,412 p

EnrolleeFemale (%) 48 48 48 48 .44

Plan benefit level �.0001Low (%) 30 30 30 30Medium (%) 50 48 50 51High (%) 20 23 20 19

PhysicianAge (years) 46 � 8 45 � 8 46 � 8 46 � 7 �.0001Female (%) 27 31 26 26 �.0001Speciality �.0001

Pediatrics (%) 47 58 52 34Family practice (%) 52 41 47 63Internal medicine (%) 1 �1 1 3

Medical group �.0001Size

3–5 (%) 3 4 3 36–19 (%) 3 4 3 3�20 (%) 94 92 93 94

Multispecialty (%) 94 93 94 95 �.0001Extensiveness of utilization 6.6 � 1.6 6.6 � 1.7 6.6 � 1.6 6.6 � 1.5 .40

management (managed care)Extensiveness of utilization 5.1 � 2.5 5.0 � 2.6 5.1 � 2.5 5.1 � 2.4 .015

management (fee-for-service)Compensation method �.0001

Salary only (%) 85 82 85 86�50% salary-based (%) 3 4 3 2Production only (%) 6 6 6 5�50% production-based (%) 6 7 6 6

Located in Seattle MSAa (%) 93 93 93 93 .03Medical group at risk by health 96 95 96 97 �.0001

plan payment (%)Managed care organization �.0001

Staff model (%) 84 80 85 87Major network (%) 13 17 13 11Preferred provider (%) 3 4 2 2

a MSA � Metropolitan Statistical Area.

The distribution of physician compensation methodacross groups was 45% production only, 18% �50%production-based, 17% salary only, 17% �50% salary-based, and 3% other. In this study, 35% of the groupshad contracts with two or more of the MCOs. Whenviewed at the enrollee level, findings for medical caregroups were quite different, since the largest groupshad the vast majority of enrollees. Over 90% ofchildren were in large groups, the vast majority be-ing multispecialty. The predominant compensationmethod was salary only.

Healthcare Plan Characteristics

Given that the HMOs in this study were vastlydifferent types of organizations, there were plan dif-

ferences with respect to enrollee, physician, and med-ical group characteristics. While enrollee age andgender were similar across groups, benefit levels werehighest in the staff model HMO, where 23% of enroll-ees had high levels of benefits. In the network HMO,only 7% had high benefit levels and in the PPO, only1% had high benefit levels. Low benefit levels wereevident in 24% of enrollees in the staff model HMO,in 56% of the network HMO, and in 98% of the PPO.In general, physician age and specialty were similaracross plans, although the PPO had a higher propor-tion of female physicians.

Differences among the plans were most apparentwith respect to medical group characteristics. Almostall of the groups in the staff model and PPO werelocated in the Seattle Metropolitan Area, while 40%


of groups in the network model were located outsidethe area. With respect to group size, 98% of groupsin the staff model were large multispecialty groups;the network model was intermediate, and the PPOhad a much lower proportion of large multispecialtygroups. With few exceptions, virtually all physiciansin the staff model HMO were salaried, whereas pro-duction-based methods were dominant in the net-work model and the PPO. The employment of utiliza-tion management practices was similar across plans.Finally, 100% and 90% of the groups in the staffand network models, respectively, were at risk forservices, whereas in the PPO, no groups were at riskfor services provided.

Utilization and Expenditures

The mean total PMPY charges were $499 � 2001(Table II) with a median of $209 and a maximum ofalmost $200,000. The mean annual number of ambu-latory visits was 4.2 � 4.9 with a median of 3 and amaximum of 219. Only 1.6% of enrollees were hospi-talized; for these children, the mean length of staywas 4.8 � 10.2 days and the median was 2 days.There were 9621 enrollees (14%) who incurred nohealthcare charges during the year; 91% of these chil-dren were 6 years and older. Charges and utilizationvaried considerably by patient age. Expenditureswere highest for the very young and lowest for theintermediate age group. The proportion of enrolleesnot receiving care also varied by age, with only 6%in the youngest age group not incurring any charges.

Physician Compensation

Given the interest in physician compensation,we examined the dependent variables by physiciancompensation method. There was no statistically sig-nificant association between compensation method

Table II. Estimated Annual Mean Charges and Utilization of Health Services by Enrollee Age

Age (years)

All ages �5 6–12 13–1767,432 13,517 29,503 24,412 p

PMPY charge ($) 499 � 2001 716 � 3018 380 � 1523 522 � 1780 �.0001Number of visits 4.2 � 4.9 62 � 6.0 3.4 � 4.3 4.0 � 4.7 �.0001Hospital days per 1000 77 � 1431 119 � 2217 42 � 856 99 � 1430 �.0001Hospitalized (%) 1.6 2.7 1.1 1.6 �.0001No visits (%) 14.3 6.4 17.0 15.3 �.0001

and per member per year charges, annual number ofvisits, or annual number of days hospitalized (Ta-ble III).

Summary

It should be recognized that given the large sam-ple size, even small differences among age categoriesproduced results that were highly statistically signifi-cant. This is particularly true for the findings in TableI. The key age-related differences were that childrenin the youngest age group were more likely to beseen by pediatricians and had higher PMPY charges,visits, and hospital days than did children in the othertwo groups. Overall, children in this study averaged10 years of age, were equally distributed with respectto gender, and were seen in large multispecialtygroups by pediatricians and family practice physi-cians. Overall median charges were about $200 peryear and the overall rate of hospitalization was low.

Multivariate Results

Enrollee age was by far the most important pre-dictor, such that younger patients had higher levelsof charges and utilization (Table IV). Benefit levelwas the second most important predictor, such thatcharges and use were greater with increasing benefitlevel. Female enrollees also had lower total per mem-ber per year charges (p � .007), although gender wasnot significantly associated with differences in thenumbers of annual visits or hospital days.

In general, physician characteristics were not as-sociated with any of the measures of charges andutilization. The one exception was that expendituresper enrollee were significantly higher for female pri-mary care physicians than male physicians (p � .007).Physician specialty was not associated with cost or uti-lization.


Table III. Estimated Annual Charges and Utilization of Health Services by Compensation Method

Per member per Number Hospital daysCompensation method year charges ($) of visits (per 1000)

Salary only (n � 57,073) 496 � 2022 4.2 � 4.8 78 � 1469�50 salary plus other (n � 1952) 568 � 2483 4.3 � 4.7 67 � 866Production-based only (n � 3854) 524 � 1899 4.3 � 6.0 69 � 1267�50% production-based plus other (n � 4290) 480 � 1540 4.1 � 4.8 76 � 1295Other (n � 263) 531 � 1217 4.6 � 8.4 23 � 212p .96 .48 .32

Consistent with results from bivariate analysis,there was no association between physician compen-sation method and the dependent measures, althoughthere was a borderline statistically significant findingof increased visits for production-only compensation(p � .06). This finding was of minimal importance,given that p values for other compensation categorieswere not statistically significant. Other medical groupvariables, including multispecialty status, extensive-ness of utilization management, group at risk, andclinic location in the Seattle Metropolitan Area werenot statistically significant. There was, however, astatistically significant relationship between MCOtype and the dependent variables in that both the

Table IV. Predictors of Annual Charges and Utilization

Per member per year charges Number of visits Hospital days

Predictors t p t p t p

EnrolleeAge �37.04 .0001 �56.16 .0001 �2.17 .03Gender �2.72 .007 1.44 .15 �0.60 .55Benefit level 9.71 .0001 10.40 .0001 1.06 .29

PhysicianAge �1.34 .18 �1.31 .19 1.72 .08Gender 2.68 .007 2.80 .005 1.56 .12Family practice �0.73 .47 0.31 .76 0.17 .87Pediatrics �0.42 .67 0.51 .61 0.21 .84Internal medicine �0.60 .55 0.41 .68 0.03 .97

Medical groupSalary plus �0.12 .90 1.51 .13 0.39 .69Production only 0.13 .90 1.90 .06 �0.20 .84Production plus �0.39 .69 1.38 .17 �0.89 .38Other 0.33 .74 1.41 .16 �0.45 .65

Multispecialty clinic �1.23 .22 0.74 .46 �1.20 .23Extensiveness of utilization management 0.29 .77 0.16 .87 1.42 .15

(managed care)Extensiveness of utilization management �0.48 .63 �0.44 .66 �1.06 .29

(fee-for-service)Seattle Metropolitan Area 0.20 .84 0.24 .81 0.90 .37Medical group at risk by health plan payment �0.24 .81 �0.66 .51 1.51 .13

Managed care organizationStaff model 1.88 .06 3.00 .003 �1.46 .14Major network 4.55 .0001 4.47 .0001 �1.62 .11

staff and major network models had higher annualcharges and visits than the PPO.

Given the importance of enrollee age and benefitlevel, we examined annual expenditures and visits bycategories of these two variables (Figs. 1 and 2). Asseen in Fig. 1, the association of cost and benefit levelwas most apparent in the youngest age group. Annualcharges were $617 in the low category and $878 inthe high one; differences were less apparent in the6- to 12-year-old group, and were not statisticallysignificant in the oldest age group. For all three agecategories, there was a statistically significant associa-tion between the number of annual visits and benefitlevel (Fig. 2). Again, the differential was most appar-


Fig. 1. Estimated annual total per member per year charges bycategories of enrollee age and benefit level. The p values are byone-way analysis of variance test for linearity. The associationbetween benefit level and total per member per year charges wasstatistically significant for children �5 years of age (p � .0001)and children 6–12 years (p � .012), but not for children 13–17years (p � .14).

ent in the youngest age group; the mean number ofannual visits was 6.6 in the high group and 5.8 in thelow one.

DISCUSSION

The purpose of this study was to assess the rela-tive importance of enrollee, physician, medical group,and healthcare plan characteristics with respect tohealthcare expenditures and use for children in threemanaged healthcare plans. As was the case in ourstudy of adults, enrollee age was the primary determi-

Fig. 2. Annual number of visits by categories of enrollee age andbenefit level. The p values are by one-way analysis of variancetest for linearity. The association between benefit level and annualnumber of visits was statistically significant for all age categories(p � .0001).

nant of healthcare charges and utilization for childrencontinuously enrolled in these three plans in Wash-ington State in 1994. Plan benefit level was of second-ary importance in the multivariate statistical models.Children 5 years of age and younger were more oftenseen by pediatricians, had higher annual charges, andhad more visits than did children in the two olderage groups. In this study, annual charges increasedwith benefit level for children in the youngest andmiddle age groups, but not the oldest one. However,the positive relationship between number of visitsand benefit level was apparent for children of allages. Only 1.6% of children were hospitalized duringthe year.

Contrary to the results of our adult study (4),female enrollees had lower charges than males. Al-though children 5 years and younger were more likelyto be seen by pediatricians, physician specialty wasnot associated with charges or utilization. However,female physicians were more likely to have enrolleeswith higher charges and utilization. Medical groupcharacteristics including physician compensationmethod were not associated with annual total PMPYcharges, number of visits, or number of days hospital-ized. However, PMPY charges and the annual num-ber of visits did vary by healthcare plan, with higherexpenditures seen in the staff and major networkmodels than the PPO.

These results are consistent with findings fromour earlier study that examined adults 18 years andolder (4). In that study, it was hypothesized thatphysician compensation method was an importantdeterminant, with physicians who were compensatedon a production basis being more likely to haveenrollees with higher utilization. However, in thatstudy, contrary to our expectations, there was noassociation between compensation method and utili-zation. While there was a borderline associationbetween production only and annual number ofvisits in the current study, there were nonsignificantfindings for �50% production, �50% salary, andother methods of compensation. Production-onlycompensation was not associated with charges ornumber of hospital days.

In this study, 9621 enrollees (14%) receivedno healthcare services during the year; 91% of thesechildren were 6 years of age and older. This findingcould be due to a number of factors; for example,missing data, the lack of outreach by plans orproviders, or limited benefits for preventive care.It is also possible that these very healthy childrenhad no need for services during the year.


This study of utilization was a snapshot anddid not collect information to track patterns ofutilization or expenditures over time. A longitudinalstudy of utilization in children in two prepaid grouppractices reported that many more children stayedat the same level of use over a 6- to 10-year periodthan was expected by chance (8). In contrast, thereis evidence that the daily practice of primary carehas changed considerably from 1979 to 1994 (14).Findings from the National Ambulatory MedicalCare Surveys, which may provide perspective tothe results of our study, suggest that the numberof primary care visits increased during the 15-yearperiod, while the average age of children decreasedand ethnic diversity increased. There is furtherevidence that the average visit duration increasedwith more preventive services provided. The admin-istrative data for our study did not allow us todistinguish between primary and specialty care forpatients with primary care physicians, so it was notpossible to readily identify preventive care providedby the primary care physician.

Utilization data for newborns and childrenunder the age of 1 year were not obtained from theMCOs, since these children were not continuouslyenrolled for the year. Pregnancy and childbirth havebeen found to be the largest source of medicalexpenditures for children from 0 to 21 years (15).While childbearing in the teenage years is relativelycommon in this country, in this group of children,there were only 12 hospitalizations for young womengiving birth. If hospitalizations for newborns hadbeen included in the length-of-stay analysis, thehospitalization rate would have increased, andthe mixed model analysis of variance results forPMPY charges and length of stay may have beendifferent.

Another limitation of this study concerns gener-alizability of results. The findings of this study do notapply to children in the Medicaid program or childrenwithout health insurance. The children in our studyhad access to care for the full year and were commer-cially insured; yet, in the United States in 1993–94,13% of children did not have health insurance (16).Given that most children in this study were enrolledin a large staff model HMO, these results may applyto similar HMOs across the country. Another limita-tion was that enrollee race was not reported by theMCOs.

A major limitation of this study was that 85% ofthe enrollees were from the staff model HMO inwhich physicians were salaried. This overwhelming

effect may have prevented the detection of statisti-cally significant differences with respect to physiciancompensation and other medical group characteris-tics. The staff model HMO included only three medi-cal groups that were under direct control by theHMO. Another 12 medical groups had contractualarrangements with the HMO. We attempted to ad-dress this deficiency by employing random effectsregression that accounts for the nesting of enrolleeswithin physician panels and physicians within medicalgroups. Also, plan was entered as a covariate in themultivariate model. In addition, analyses were runat the level of the medical group and produced resultssimilar to those obtained in the enrollee-level analy-sis. These measures represent our best effort to ad-dress this problem, which nevertheless remains themajor limitation of this report.

Despite these limitations, this study has severalstrengths. First, it collected not just individual pa-tient data, but also characteristics of physicians,medical care groups, and healthcare plans. Theability to measure benefit level was a particularstrength. Second, it used a multivariate statisticaltechnique that adjusted for the nesting of physicianswithin medical care groups and enrollees withinphysicians’ panels. This statistical method is particu-larly useful for datasets that incorporate severaldifferent levels of analysis. Third, with the exceptionof pharmacy and dental data, utilization data con-tained both primary care and referred services.Finally, because of extensive discussions and negoti-ations with the MCOs, we are confident that allchildren enrolled in their respective healthcareplans, as well as all primary care physicians caringfor them, were included in the enrollee and physi-cian files provided by the MCOs.

It is clear that there is much to learn abouthow managed care influences the delivery of healthservices to children, particularly those in commer-cially insured programs. In this study, age andbenefit level were the primary determinants ofhealthcare charges and utilization. It is not surprisingthat health plan was also important, given that thethree health plans in this study were vastly differentorganizations. Physician and medical group charac-teristics added very little in the way of explanation,but this was not surprising given that the samepattern was observed in our study of adults. Forboth children and adults enrolled in managed carehealth plans in Washington State, age and benefitlevel were the primary determinants of total charges,number of visits, and days hospitalized in 1994.


ACKNOWLEDGMENTS

The authors gratefully acknowledge the supportof a grant from the Robert Wood Johnson Founda-tion, as well as the helpful comments of the editorand three anonymous reviewers.

REFERENCES

1. Hughes DC, Luft HS. Managed care and children: An over-view. Future Child 1998;8:25–38.

2. Szikagyi PG. Managed care for children: Effect on access tocare and utilization of health services. Future Child 1998;8:39–59.

3. Bergman DA, Homer CJ. Managed care and quality of chil-dren’s health services. Future Child 1998;8:60–75.

4. Conrad DA, Maynard C, Cheadle A, Ramsey S, Marcus-Smith M, Kirz H, Madden CA, Martin D, Perrin EB, WickizerT, Zierler B, Ross A, Noren J, Liang S. Primary care physiciancompensation method in medical groups. JAMA 1998;279:853–58.

5. Bernzweig J, Takayama JI, Phibbs C, Lewis C, Pantell RH.Gender differences in physician–patient communication: Evi-dence from pediatric visits. Arch Pediatr Adolesc Med 1997;151:586–91.

6. Valdez RB, Ware JE, Manning WG, Brook RH, Rogers WH,Goldberg GA, Newhouse JP. Prepaid group practice effectson the utilization of medical services and health outcomes forchildren: Results from a controlled trial. Pediatrics 1989;83:168–80.

7. Fried VM, Makuc DM, Rooks RN. Ambulatory health carevisits by children: Principal diagnosis and place of visit. VitalHealth Stat 1998;13:1–23.

8. Woodward CA, Boyle MH, Offord DR, Cadman DT, LinksPS, Munroe-Blum H, Byrne C, Thomas H. Ontario ChildHealth Study: Patterns of ambulatory medical care utilizationand their correlates. Pediatrics 1988;82:425–34.

9. Riley AW, Finney JW, Mellits ED, Starfield B, Kidwell S,Quaskey S, Cataldo MF, Filipp L, Shematek JP. Determinantsof children’s health care use: An investigation of psychosocialfactors. Med Care 1993;31:767–83.

10. Starfield BH, van den Berg BJ, Steinwachs DM, Katz HP,Horn SD. Variations in utilization of health services by chil-dren. Pediatrics 1979;63:633–41.

11. American Academy of Pediatrics. The medical home state-ment addendum: Pediatric primary health care. Am AcadPediatr News 1993;9:7.

12. Whiting-O’Keefe Q, Simborg D. Choosing the correct unitof analysis in medical care experiments. Med Care 1984;22:1101–14.

13. Williams PT, Fortmann SP, Farquhar JW, Varady A, MellenS. A comparison of statistical methods for evaluating riskfactor changes in community based studies: An example fromthe Stanford Three-Community Study. J Chronic Dis 1981;34:565–71.

14. Ferris TG, Saglam D, Stafford RS, Causin N, Starfield B,Culpepper L, Blumenthal D. Changes in the daily practice ofprimary care for children. Arch Pediatr Adolesc Med 1998;152:227–33.

15. Miller TR, Lestina DC, Galbraith MS. Pattern of childhoodmedical spending. Arch Pediatr Adolesc Med 1995;149:369–73.

16. Newacheck PW, Stoddard JJ, Hughes DC, Pearl M. Healthinsurance and access to primary care for children. N Engl JMed 1998;338:513–19.

healthcare charges and use in commercially insured children enrolled in managed care health plans in...

Documents