Network-model Health Maintenance Organization Research Articles

Primary care provider perceptions of barriers to and facilitators of colorectal cancer screening in a managed care setting.

Colorectal cancer (CRC) screening tests (e.g., fecal occult blood testing [FOBT], flexible sigmoidoscopy [FS], etc.) are underused. Primary care providers (PCPs) play a critical role in screening, but barriers to and facilitators of screening as perceived by PCPs in managed care settings are poorly understood. The objectives of the current study were to describe current CRC screening practices and to explore determinants of test use by PCPs in a managed care setting. In 2000, a self-administered survey was mailed to a stratified, random sample of 1340 PCPs in a large, network model health maintenance organization in California. The survey response rate was 67%. PCPs indicated that 79% of their standard-risk patients were screened for CRC. PCP-reported median rates of recommendation for the use of specific screening tests were 90% for FOBT and 70% for FS. In logistic regression models, perceived barriers to the use of FOBT and FS included patient characteristics (e.g., education) and PCP-related barriers (e.g., failure to recall that patients were due for testing). Perceived facilitators of the use of FOBT and FS included interventions targeting certain aspects of the health care system (e.g., reimbursement) and interventions targeting certain aspects of the tests themselves (e.g., provision of evidence of a test's effectiveness). Assignment of high priority to screening, integrated medical group (as opposed to independent practice association) affiliation, and the proportion of patients receiving routine health maintenance examinations were positively associated with reported test use. CRC screening tests appear to be underused in the managed care setting examined in the current study. The perceived barriers and facilitators that were identified can be used to guide interventions aimed at increasing recommendations for, as well as actual performance of, CRC screening.

Prevalence and characteristics of individuals with chronic kidney disease in a large health maintenance organization

This study is designed to estimate the prevalence of and gain further insight into the characteristics of the chronic kidney disease (CKD) population in a large US health maintenance organization (HMO) to better understand the CKD population in the United States overall. Analyses were performed using data from a staff and network model HMO in the southwestern United States with more than 150,000 members per year during 1994 to 1997. The estimated prevalence of CKD in the HMO population varied from 0.4[percnt] to 7.1[percnt], depending on the definition of CKD used. Regardless of the definition, CKD was more common in men compared with women and in patients with diabetes mellitus and[sol ]or hypertension. Applying the age- and sex-specific prevalence rates in the HMO to the US population in 1990, we estimate there were approximately 9.1 million Americans with at least one elevated sex-specific creatinine (Cr) value and approximately 4.2 million Americans with at least two elevated Cr values separated by 90 days or greater, a more rigorous definition of CKD. From these results, it is apparent that there are a large number of patients in the United States with CKD. Most have not been identified because screening for CKD generally is not performed. Considering the high prevalence of CKD and the high cost and clinical morbidity associated with end-stage renal disease (ESRD), it is clear that CKD is an important public health problem. Early identification of patients with CKD would allow treatment that could slow the progression to ESRD, improve clinical outcomes, and constrain the growth of costs in the ESRD program. The time has come for a structured public and professional educational program to address this serious condition.

Organizational and Financial Characteristics of Health Plans

Primary care performance has been shown to differ under different models of health care delivery, even among various models of managed care. Pervasive changes in our nation's health care delivery systems, including the emergence of new forms of managed care, compel more current data. To compare the primary care received by patients in each of 5 models of managed care (managed indemnity, point of service, network-model health maintenance organization [HMO], group-model HMO, and staff-model HMO) and identify specific characteristics of health plans associated with performance differences. Cross-sectional observational study of Massachusetts adults who reported having a regular personal physician and for whom plan-type was known (n = 6018). Participants completed a validated questionnaire measuring 7 defining characteristics of primary care. Senior health plan executives provided information about financial and nonfinancial features of the plan's contractual arrangements with physicians. The managed indemnity system performed most favorably, with the highest adjusted mean scores for 8 of 10 measures (P<.05). Point of service and network-model HMO performance equaled the indemnity system on many measures. Staff-model HMOs performed least favorably, with adjusted mean scores that were lowest or statistically equivalent to the lowest score on all 10 scales. Among network-model HMOs, several features of the plan's contractual arrangement with physicians (ie, capitated physician payment, extensive use of clinical practice guidelines, financial incentives concerning patient satisfaction) were significantly associated with performance (P<.05). With US employers and purchasers having largely rejected traditional indemnity insurance as unaffordable, the results suggest that the current momentum toward open-model managed care plans is consistent with goals for high-quality primary care, but that the effects of specific financial and nonfinancial incentives used by plans must continue to be examined.

Risk and protective factors associated with screening for complications of diabetes in a health maintenance organization setting.

To identify risk and protective factors associated with screening for complications of diabetes, we performed a cross-sectional study of 3,612 diabetic members enrolled in CaliforniaCare, a large network-model health maintenance organization (HMO). We used the Health Plan and Employer Data Information Set (HEDIS) 3.0 technical definition to identify all members (aged > or = 31 years) receiving any diabetes medication(s) during a 12-month period. Using a telephone survey instrument, identified members were interviewed about their diabetes care and screening, patient, and provider history. Survey data were supplemented with HMO claims and demographic information. Multivariate analysis was performed to identify demographic, clinical, and utilization characteristics that affect the odds of diabetic members receiving annual retinal examination, foot examination, and HbA1c testing. While results varied by screening category, the odds of obtaining screening were higher for diabetic members who were older, spoke English, received diabetes nutrition counseling, visited a diabetes specialist physician, belonged to a diabetes association or support group, used insulin, performed glycemic level self-examination at least once a day, and had higher overall prescription drug use (suggesting higher comorbidity). Since this study is a cross-sectional review, these results do not imply a cause-and-effect relationship between dependent and independent variables. Results of this study suggest barriers, risks, and protective factors associated with screening for complications of diabetes. Diabetic members who do not possess these characteristics may be at increased risk.

Effect of a Large Managed Care Program on Emergency Department Use: Results From the CHAMPUS Reform Initiative Evaluation

Study Objective: To determine the effect of the CHAMPUS Reform Initiative (CRI) on emergency department use and charges, and to ascertain whether any reductions were concentrated among repeat users of the ED, those with less serious ED diagnoses, or those with selected chronic medical conditions. Methods: Participants were approximately 1.2 million beneficiaries of the Civilian Health and Medical Program of Uniformed Services (CHAMPUS) residing within either 11 military hospital catchment areas in California and Hawaii (“demonstration areas”) or 11 matched control areas in other parts of the United States. Under CRI, participants were offered a choice of the standard CHAMPUS indemnity plan, a Preferred Provider Organization– type plan, or a network-model Health Maintenance Organization plan. Beneficiaries were encouraged to use alternatives to the ED for nonemergency conditions. Visits to civilian EDs during two 12-month periods, before and after institution of CRI, were compared. Results: Under CRI, the number of CHAMPUS ED visits decreased by approximately 40% relative to the control, and allowed charges fell by almost 50%. Relative reductions in ED use under CRI were seen among both frequent and infrequent users of the ED. ED case-mix severity increased modestly relative to control (+3.5% versus +.9%). ED use among patients with diabetes, hypertension, and asthma fell sharply in the demonstration areas (by 14% to 41%) but rose in control areas (by 4% to 9%). Conclusion: In one of the largest managed care demonstrations ever conducted, a nonintrusive use management program and improved access to outpatient care appeared to reduce ED use, allowed charges, and costs to the government. Reductions in ED use were concentrated to some extent among repeat users and patients with less severe illnesses and were effected without capitation of provider groups or strict gatekeeping requirements. [Kravitz RL, Zwanziger J, Hosek S, Polich S, Sloss E, McCaffrey D: Effect of a large managed care program on emergency department use: Results from the CHAMPUS reform initiative evaluation. Ann Emerg Med June 1998;31:741-748.]

Are health risks related to medical care charges in the short-term? Challenging traditional assumptions.

This study examines the association between self-reported health risks and short-term use of medical resources. Cross-sectional study comparing responses from a mailed survey with medical encounter and expenditure data. Bivariate analyses and multivariate linear and logistic regressions controlling for age, gender, and health status were conducted to illustrate the relationship between selected risk factors and resource use. A group network model health maintenance organization (HMO) in Minneapolis, Minnesota. The study population was comprised of a randomly selected sample of nonsenior adults (18 to 64 years old, n = 3825) and seniors (65 years and older, n = 1955) who were enrolled in an HMO. The response rate was 72%. Five independent variables: smoking, alcohol use, obesity, lack of physical activity, and unhappiness were derived from a questionnaire designed to assess health status and health risks. Outcome measures were medical care charges and use of hospital services. Obesity, physical inactivity, and unhappiness were related to higher charges among seniors. These relationships dissipated to some degree after controlling for age, gender, and health status. Unexpected differences were found for alcohol use among both age groups, with those considered to be "not at risk" accruing higher charges and demonstrating a higher likelihood of using inpatient hospital services than those defined to be at risk. Bivariate and multivariate regression results demonstrated that traditional risk factors are weak and inconsistent predictors of short-term medical charges. Charging smokers and other high risk individuals a higher annual insurance premium than is offered for those at low risk raises questions about fairness when such risk factors are not necessarily good predictors of short-term resource use. The rationale for insurance cost-shifting should be supported with reliable data connecting risks to charges.

Evaluation of Three Methods for Improving Mammography Rates in a Managed Care Plan

Mammography has been shown to reduce breast cancer mortality among women 50 and older. Although mammography rates are increasing nationally, this effective screening tool remains underused. This study was conducted among 395 women who were members of a network model health maintenance organization (HMO) in Philadelphia in order to determine which of three methods was most effective in increasing mammography rates: (1) a birthday card reminder only (the standard method); (2) a personalized letter from the medical director and materials promoting mammography; and (3) a multicomponent phone call incorporating a reminder, counseling, and scheduling of appointments. An additional goal was to determine whether the interventions were more or less effective depending on a woman's readiness to get a mammogram, as measured by stage of change. Eligible women were randomized into one of three treatment groups described earlier. Mammography rates were calculated on the basis of a claims review and follow-up phone interviews after a period of six months. Women who received the telephone intervention were most likely to obtain a mammogram (28%); followed by the group that received the birthday card only (15%), and those who received the mailed intervention (9%). These results indicate that a multicomponent phone intervention is significantly effective in promoting mammography in managed health care plan members. An analysis by women's stage of change found a difference in the effectiveness of the three interventions among contemplators only.

Taking health status into account when setting capitation rates. A comparison of risk-adjustment methods

Objective. —To compare performance of different health status measures for risk-adjusting capitation rates. Design. —Cross-sectional study. Health status measures derived from 1 year were used to predict resources for that year and the next. Setting. —Group-network health maintenance organization in Minnesota. Participants. —Sample of 18- to 64-year-old (n=3825) and elderly (aged ≥65 years; n=1955) members enrolled in a network-model health maintenance organization in Minnesota. Main Outcome Measures. —Total expenditures in the year concurrent with the health status survey (July 1991 through June 1992) and total expenditures in the year following the survey (July 1992 through June 1993). Results. —Capitation adjustment based on demographic measures performed least well. Both self-reported health status measures and diagnoses predicted future expenditures twice as well as demographics. When predicting costs for groups of patients rather than individuals, the demographic model worked well for average groups but tended to overpredict healthier groups and underpredict sicker groups. Ambulatory Care Groups based on diagnoses performed better than self-reported health status both in the retrospective models and across healthier and sicker groups. Conclustion. —Without risk adjustment, capitation rates are likely to overpay or underpay physicians for certain patient groups. It is possible to improve prediction using health status measures for risk adjustment. When selection bias is suspected and administrative data are available, we recommend a risk-adjustment method based on diagnostic information. If diagnostic data are not available, we recommend a system based on simple self-reported measures, such as chronic conditions, rather than complex functional status measures.

Taking health status into account when setting capitation rates: a comparison of risk-adjustment methods

To compare performance of different health status measures for risk-adjusting capitation rates. Cross-sectional study. Health status measures derived from 1 year were used to predict resources for that year and the next. Group-network health maintenance organization in Minnesota. Sample of 18- to 64-year-old (n=3825) and elderly (aged > or = 65 years; n=1955) members enrolled in a network-model health maintenance organization in Minnesota. Total expenditures in the year concurrent with the health status survey (July 1991 through June 1992) and total expenditures in the year following the survey (July 1992 through June 1993). Capitation adjustment based on demographic measures performed least well. Both self-reported health status measures and diagnoses predicted future expenditures twice as well as demographics. When predicting costs for groups of patients rather than individuals, the demographic model worked well for average groups but tended to overpredict healthier groups and underpredict sicker groups. Ambulatory Care Groups based on diagnoses performed better than self-reported health status both in the retrospective models and across healthier and sicker groups. Without risk adjustment, capitation rates are likely to overpay or underpay physicians for certain patient groups. It is possible to improve prediction using health status measures for risk adjustment. When selection bias is suspected and administrative data are available, we recommend a risk-adjustment method based on diagnostic information. If diagnostic data are not available, we recommend a system based on simple self-reported measures, such as chronic conditions, rather than complex functional status measures.

Quality Assurance in Capitated Physician Groups

To describe quality assurance (QA) programs implemented by capitated physician groups; to measure their relative emphasis on monitoring of overuse compared with underuse and monitoring and improving preventive services compared with chronic disease care; and to examine how group characteristics influence QA activity. Cross-sectional questionnaire. A large network-model health maintenance organization in California (133 contracting physician groups). Ninety-four physician groups (71%) caring for 2.9 million capitated patients. Self-reported use of quality monitoring and improvement methods. All capitated physician groups conducted some QA. Groups' QA programs monitored areas subject to overuse, such as cesarean delivery and angioplasty rates, more than areas subject to underuse, such as childhood immunization rates and performance of retinal examinations for diabetic patients (64% vs 43%, P<.001). They monitored underuse of preventive services more than follow-up services for chronic diseases (54% vs 31%, P<.001). Groups also used reminders for preventive services more than they monitored follow-up services for chronic diseases (26% vs 15%, P<.01). Physician group characteristics independently associated with higher overall QA activity were greater number of years in existence, higher profitability, and capitated care penetration. Capitation places a large share of responsibility for QA in the hands of physician groups, but not all aspects of QA are being equally addressed. The emphasis on overuse may result from financial incentives inherent in capitation, while the focus on preventive services may stem from lack of adequate quality measurement tools for monitoring chronic disease care. Further research efforts should address how capitated physician groups might expand their QA programs to include monitoring of underuse, especially for patients with chronic disease.

Managed care and capitation in California: how do physicians at financial risk control their own utilization?

To describe the structure and range of utilization management methods initiated by physicians in response to capitation. Cross-sectional questionnaire. A large network-model health maintenance organization (133 contracting physician groups) in California. 94 (71%) physician groups caring for 2.9 million capitated patients. Self-reported use of five major utilization management methods. All physician groups reported using gate-keeping and preauthorization for certain referrals or tests. Most also used profiling of utilization patterns (79%), guidelines (70%), and managed care education (69%). Most physician groups asked gatekeepers to submit preauthorization requests for specialty referrals and restricted patient self-referral. For example, 60% of groups required preauthorization for an internal medicine subspecialty referral, and 7% allowed patient self-referral. Most groups also asked gatekeepers to obtain preauthorization for many tests (for example, 95% for magnetic resonance imaging and 53% for pulmonary function tests). Preauthorization requests were denied infrequently (less than 10% of the time) by more than 75% of groups. Of the 54 groups reporting utilization profiles to their physicians, 61% never adjusted for case-mix among patients and more than 60% suggested practice changes to their physicians based on utilization. Fewer than 35% of the groups used written guidelines for expensive tests that required preauthorization (such as angiography). Physicians are responding to capitation by using utilization management techniques, some at early stages of development, that were previously used only by insurers. This physician-initiated management approach represents a fundamental transformation in the practice of medicine.

The Effect of an Education and Feedback Intervention on Group-Model and Network-Model Health Maintenance Organization Physician Prescribing Behavior

The authors evaluated the effect of an educational and feedback intervention on H2-blocker prescribing patterns and determined, if such effects differed for network- versus group-model health maintenance organization (HMO) physicians and in academic versus nonacademic settings. Physicians were randomized to receive an educational memorandum alone or combined with feedback regarding their individual prescribing behavior. The memo suggested preferred use of an H2-blocker (cimetidine) that would be less expensive to the HMO. Prescribing was monitored during the 6 months before and after the intervention. The study was undertaken at the primary care practices of a mixed group- and network-model university-affiliated HMO. Thirty group-model (at two academic and four nonacademic sites) and 33 network-model (all in full-time private practice) primary care physicians participated in the study. The analysis utilized weighted and unweighted analysis of covariance of the change in physicians' cimetidine-prescribing rates between the baseline and study periods. A significant response to the intervention was noted among academic and nonacademic group-model HMO physicians, but not among network physicians (adjusted mean absolute prescribing changes of +9.9% and +8.9% versus -2.8%, P = .02). There was no difference in prescribing change based on type of intervention (education versus feedback). The authors conclude that a simple passive educational intervention can be effective at changing group-model HMO physician behavior.

The Effect of an Education and Feedback Intervention on Group-Model and Network-Model Health Maintenance Organization Physician Prescribing Behavior

The Effect of an Education and Feedback Intervention on Group-Model and Network-Model Health Maintenance Organization Physician Prescribing Behavior

Effect of pharmaceutical formulation of diltiazem on the utilization of medicaid and health maintenance organization services

Hypertension is a silent disease, the full seriousness of which becomes evident only when the sequelae of long-standing high blood pressure are manifest as congestive heart failure, angina, myocardial infarction, cerebral vascular accident, or end-stage renal disease. A major barrier to the management of hypertension is the extent to which patients comply with treatment. We report the findings of retrospective analyses designed to determine the relationship between dosage formulation and administration schedule for the calcium channel antagonist diltiazem, and the utilization of health care services within the state of South Carolina's Medicaid program and a network-model health maintenance organization (HMO). The research objectives were threefold: (1) for patients prescribed 180 mg/d of diltiazem, compare financial outcomes of patients receiving three dosage formulations: 60 mg three times daily (TID), immediate-release (IR) diltiazem; 90 mg two times daily (BID), sustained-release (SR) diltiazem; or 180 mg once daily (QD), controlled-delivery (CD) diltiazem; (2) for patients prescribed 240 mg/d of diltiazem, compare financial outcomes of patients receiving three dosage formulations: 60-mg four times daily (QID) diltiazem IR, 120-mg BID diltiazem SR, or 240-mg QD diltiazem CD; and (3) determine the effect of the medication possession ratio (MPR), an index of compliance, on health service utilization. Multivariate regression analysis was used to determine the incremental influence of selected demographic characteristics, use of medical services 6 months before the diagnosis of hypertension, and the prescribed formulation of diltiazem on health service expenditures 1 year after the diagnosis of hypertension. Results indicate that patients initially prescribed a formulation and regimen facilitating once-daily administration have the fewest number of changes in antihypertensive pharmacotherapy, require fewer concomitant antihypertensives, achieve the highest MPR for antihypertensive pharmacotherapy, and record a decrease in the use of and expenditures for physician and hospital services ( P ≤ 0.05) under Medicaid and within the HMO.

Antihypertensive pharmacotherapy: Economic outcomes in a health maintenance organization

Hypertension is a silent disease, the full seriousness of which becomes evident only when the sequelae of long-standing high blood pressure manifest as congestive heart failure, angina, myocardial infarction, cerebral vascular accident, or end-stage renal disease. A significant barrier to the management of hypertension is the extent to which patients comply with the treatment regimen. We conducted a retrospective analysis to determine the relationship between dosing schedules for antihypertensive pharmacotherapy, regimen compliance, and the utilization of health care services. Information regarding health care service utilization was derived from the computer archive of a network-model health maintenance organization system serving 400,000 beneficiaries. The retrospective archive facilitated the abstraction of data regarding disease-specific health care service utilization. Patient-level paid-claims data for the period January 1, 1990, through December 31, 1993, were abstracted for patients who began antihypertensive therapy with either atenolol, captopril, oral clonidine, transdermal clonidine, diltiazem, enalapril, metoprolol, prazosin, terazosin, or verapamil. Multivariate regression analysis was used to determine the incremental influence on health care expenditures 1 year after diagnosis for hypertension of selected demographic characteristics, use of medical services before diagnosis for hypertension, initial antihypertensive medication, and the number of prescribed maintenance medications for diseases other than hypertension. Separate multivariate analyses were performed to determine the influence of the medication possession ratio (MPR), an index of regimen compliance with antihypertensive therapy, on health care expenditures for hypertension. Our results indicate that patients initially prescribed antihypertensive medication administered once weekly require the fewest number of changes in therapeutic regimen, require far less concomitant pharmacotherapy for blood pressure control, achieve the highest MPR for antihypertensive pharmacotherapy, and require fewer physician, hospital, and laboratory services than patients prescribed antihypertensive pharmacotherapy requiring twice daily, and in some instances once daily, administration. An increasing MPR was found to be correlated with an increase in expenditures for antihypertensive therapy and a decrease in financial commitments for physician, hospital, and laboratory services.

Assessing the needs of women and clinicians for the management of menopause in an HMO.

To determine the perceived needs of perimenopausal women regarding the management of menopause and the resource needs of the clinicians who treat them. A large staff and group network model health maintenance organization (HMO) in New England. A random sample of 790 perimenopausal women aged 45-60 years who were members of the HMO in 1991, and a random sample of 180 clinicians in internal medicine, family practice, and obstetrics/gynecology practicing in the HMO during 1991. Mailed surveys of women and clinicians were designed to assess possible needs and attitudes that could lead to the improvement of care for menopausal women. The chi-square test was used to determine differences in perceived needs and satisfaction levels among women with differences in self-reported menopausal status. The Kruskal-Wallis one-way analysis of variance and the Mann-Whitney U test were used in the clinician survey to test for differences among specialties and between genders. The key findings include that: 1) most (81%) of the women wanted to see a woman clinician, 2) many (50%) were interested in a menopause support group, 3) 30% reported that their care for menopause had been fair to poor, 4) only 55% of the primary care specialists (including internal medicine and family practice) reported high confidence in their abilities to treat menopause, compared with 68% of the obstetric/gynecology clinicians, and 5) 56% of the clinicians surveyed said that support from the HMO to their practices for the treatment of menopause was fair to poor. There is an opportunity for better care for perimenopausal women as reported by two sources, HMO clinicians and members. To provide this care, clinicians may need explicit guidelines as well as administrative supports such as educational materials and specialty access. Since the capability for menopausal care from clinicians in obstetrics/gynecology is perceived to be higher than that from primary care clinicians, an opportunity for cross-specialty collaboration and training may exist.

Comparison of the Quality of Ambulatory Care for Fee-for-Service and Prepaid Patients

To determine whether the quality of care for common ambulatory conditions is adversely affected when physicians are provided with incentives to limit the use of health services. Retrospective cohort study over a 2-year period. Four group practices that cared for both fee-for-service patients and prepaid patients within a network model health maintenance organization (HMO). Equal numbers of prepaid (HMO) and fee-for-service patients were selected by randomly choosing medical records from each group practice: 246 patients with chronic uncomplicated hypertension and 250 women without chronic diseases who received preventive care. Adequate hypertension control was defined as a mean blood pressure of less than 150/90. Adequate preventive care was defined as the provision of blood pressure screening, colon cancer screening, breast cancer screening, and cervical cancer screening within guidelines recommended by the 1989 U.S. Preventive Services Task Force. Resource use was measured by the annual number of visits and tests. The adjusted relative odds of HMO patients having controlled hypertension, compared with fee-for-service patients, were 1.82 (95% CI, 1.02 to 3.27). The relative risks of HMO patients receiving preventive care within established guidelines were 1.19 (CI, 0.93 to 1.51) for colon cancer screening, 1.78 (CI, 1.11 to 2.84) for annual breast examinations, 1.75 (CI, 1.08 to 2.84) for biannual mammography, and 1.35 (CI, 1.13 to 1.60) for Papanicolaou smears every 3 years. Prepaid patients had visit rates that were 18% to 22% higher than those of fee-for-service patients. In the type of network model HMO we studied, the quality and quantity of ambulatory care for HMO patients was equal to or better than that for fee-for-service patients. In this setting, the incentives for physicians to limit resource use may be offset by lack of disincentives for HMO patients to seek care.

Patterns of stability and change in health use among elderly people. Do service systems leave an imprint on behavior?

Stability and change in patterns of health service use over a 3-year period were determined for a sample of elderly people in an urban area who claimed one of five types of health service provider as a primary source of health care--a hospital, a private physician, a network model health maintenance organization (HMO), a hospital-based group practice program (G-HMO), or a preferred provider organization (PPO). Despite certain differences in use rates for individual services, the total volume of ambulatory service use was equivalent for all five groups as was the relative rank order of use of specific ambulatory services for four of the five groups. People who claimed a hospital as their primary care source had the most unique use patterns over a full range of health care services, characterized by extremely low rates of physician visits and the highest rates of visits to hospital outpatient clinics across three time periods. G-HMO members used health-related services more frequently than did all others. PPO members, at baseline, had lower rates of total and mean hospital days than other source group members except hospital users. People who changed principal source of care during the study period were most likely to report a hospital as their care source initially. Although there is much consistency in hospital and ambulatory use across groups, the persistence of certain use patterns for members of some groups suggests that health care systems can leave an imprint on the health service use of people for whom they provide regular care.