Our aim was to compare access to effective care among elderly Medicare patients in a Staff Model and Group Model HMO and in Fee-for-Service (FFS) care.
We used a retrospective cohort study design, using claims and automated medical record data to compare achievement on quality indicators for elderly Medicare recipients. Secondary data were collected from 1) HMO data sets and 2) Medicare claims files for the time period 1994–95. All subjects were Medicare enrollees in a defined area of New England: those enrolled in two divisions of a managed care plan with different physician payment arrangements: a staff model, and a group model; and the Medicare FFS population. We abstracted information on indicators covering several domains: preventive, diagnosis-specific, and chronic disease care.
On the indicators we created and tested, access in the single managed care plan under study was comparable to or better than FFS care in the same geographic region. Percent of Medicare recipients with breast cancer screening was 36 percentage points higher in the staff model versus FFS (95% confidence interval 34–38 percentage points). Follow up after hospitalization for myocardial infarction was 20 percentage points higher in the group model than in FFS (95% confidence interval 14–26 percentage points).
According to indicators developed for use in both claims and automated medical record data, access to care for elderly Medicare beneficiaries in one large managed care organization was as good as or better than that in FFS care in the same geographic area.
Over 32 million elderly Americans are enrolled in the Medicare health insurance program. Since 1982, the Health Care Financing Administration has offered a managed care option, with capitated payment, to Medicare enrollees. Currently nearly 15% of Medicare enrollees are members of managed care plans nationwide. 
The economic incentives of capitated payment have raised concerns that managed care organizations that use capitation might limit provision of effective services as a way of trimming costs. [2-5] Public concerns about restriction of service in managed care organizations have led to proposed legislation for a "patient's bill of rights." once again being considered in the US Senate.  It is becoming clear, however, that "managed care" is not a single entity, and that a wide range of economic incentives and disincentives may be created in various forms of managed care,  each likely to have different kinds of influence on clinician behavior. [9-11] The proximity of risk to the clinician or clinician group has an especially strong effect on behavior.  Systems maintained by managed care organizations to assist clinicians in providing quality care also differ widely, and may strongly influence performance.
Although reduced access to care is a theoretical risk of capitated payment, relatively few studies have addressed whether this concern is borne out in practice. Perhaps because of the variation in managed care arrangements, the results of studies have been mixed.  If managed care presents barriers to access to effective care, the elderly would be especially vulnerable because they have a higher burden of illness than the rest of the population. We therefore created indicators in several domains for which it was known that care improved outcomes. We then used these indicators to compare health care received by elderly Medicare beneficiaries in three kinds of payment arrangements: a capitated, staff-model health maintenance organization (HMO) and a partially capitated group-model HMO, both parts of the same managed care organization, and in fee-for-service (FFS) care in the same geographic area.
The managed care plan participating in this study, Harvard Pilgrim Health Care (HPHC), is the largest HMO in New England, and is a not-for profit entity. It was begun in 1981 as Harvard Community Health Plan, a Staff Model HMO that at the time of this study comprised about 300,000 members at fourteen health centers in the greater Boston area. Harvard Pilgrim Health Care had grown by merger to include, among its divisions, one made up of non-exclusive medical groups; at the time of this study the Group Model HMO cared for about 186,000 members in eastern Massachusetts.
For both divisions, the Medicare enrollees eligible for this study included all those age 65 and older who were members for part or all of the period January 1, 1994 through December 31, 1995. Almost 10,000 members from the Staff Model and 5,000 in the Group Model met these criteria. Members who switched from one division to the other during the study period (497 members, 2 percent of the combined samples) were not included in the study. Samples were defined separately for each indicator (see Tables and 1 Indicator Definitions).
Staff model HMO
The Staff Model division of HPHC provided primary care-based, multi-specialty care. The organization has cared for aged Medicare patients in risk and other commercial plans since the mid-1980's. All primary care providers were salaried and there were no specific economic incentives (i.e. no bonuses of any kind) for individual providers during the study period.
Information on all ambulatory services, including visits, diagnoses, procedures and tests, was maintained electronically for all members as part of an Automated Medical Record System (AMRS.) These data have been used in previous studies. [15-17] The Staff Model HMO contracted with area hospitals for inpatient services. For hospitalization and other outside utilization, the HMO maintains claims data using ICD-9 codes; this data structure contains up to six diagnoses and up to 3 surgical procedures per claim.
Group model HMO
The Group Model division of HPHC included 16 medical groups in 1993 that contracted with the HMO on a non-exclusive basis, i.e. they were exclusive with regard to other managed care plans, but not with respect to indemnity insurance. Most were primary care groups but a small number were multispecialty groups. Payment of groups was based on various forms of capitation; the most common arrangement was for groups to be capitated for both primary and specialty outpatient care, and for hospitalization risk to be shared in a wider pool. Capitation rates and details such as loss-limit provisions varied from group to group. Almost all groups comprised between 5 and 10 physicians.
Data from the Group Model were obtained from claims that were submitted to the HMO to document care for calculations pertinent to loss-limit provisions, even though payments were not based on the claims themselves. Data files contained information on services provided in the offices of the primary care clinicians, on outpatient services not provided by the primary care provider, (i.e. referrals), and on hospitalizations. Diagnoses and procedures were coded using ICD-9 and CPT-4 codes.
Demographic and insurance coverage information on members in the study were obtained from HMO enrollment files.
Medicare beneficiaries were included in this study if they were: age 65 and older as of January 1, 1994; had both Part A and B coverage for at least 12 consecutive months during 1994 and 1995; were not enrolled in an HMO during these 12 consecutive months; and resided within the HMO catchment area (see below).
We matched Medicare members enrolled in the HMO with those in fee-for-service by area of residence. First, we constructed a list of the zip codes in which eligible HMO members resided during the study period. We constructed a list of Medicare fee-for-service beneficiaries residing in the comparison area consisting of all contiguous zip codes in which at least one HMO Medicare beneficiary resided. The resulting sample included 339,627 people.
Data for the fee-for-service sample were obtained from an enrollment file for demographic information, the MedPAR (Medicare Provider Analysis and Review) file for hospitalization claim data, Part B physician/supplier claims file, and Hospital Outpatient Department claims file.
We developed performance measures that met several or all of the following criteria: the disease or condition was relatively common and potentially severe; a clinical service that could affect the health outcome of the patient was identifiable; the performance of the indicated service depended in part on the role of a physician or health system and not only on the patient; service performance was potentially subject to financial incentives; it was possible from other research to assume that the indicated service would on average improve outcome; and data to construct the indicator were currently or potentially available.
Our indicators assessed access to health care in three domains (Table 1): 1) preventive care, including services that provide future, not current, benefit; 2) diagnosis-specific care, examining treatment for acute conditions or episodes of disease; and 3) chronic disease care, including secondary prevention. The selection of preventive care indicators was guided by the generic applicability of these maneuvers in this age group, making it unnecessary to define a special population for whom the procedure would be indicated. The indicators of timely follow-up after hospitalization (diagnosis-specific care) were selected based on the seriousness of the medical conditions;  the time interval between hospital discharge and outpatient visit were determined by a panel of physicians.
Table 1. Medicare Performance Indicators
A panel of national experts in access to quality health care, convened by the Health Care Financing Administration in March of 1996, suggested additions to our original indicator set. We then selected those that best met the structured evaluation criteria, including at least two indicators in each of the domains. Table 1 shows the indicators organized by domains. More complete definitions of inclusion criteria and coding, including specific AMRS, ICD-9 and CPT codes used, have been published elsewhere  and are summarized in 1 Indicator Definitions.
We compared the three samples by age category using the chi-square test. For annual indicators, we averaged utilization rates for 1994 and 1995; the denominators were individuals eligible for each calendar year. We calculated rates in three age strata: age 65–74, 75–84, and 85 years and older. We also calculated age-adjusted rates standardized to the fee-for-service sample. Analyses were performed using SAS. We conducted two-way comparisons of age-adjusted rates across the three samples (Staff Model, Group Model, and fee-for-service) using the chi-square test.