Objective. To study patterns of referral between primary and specialty care
providers among Medicare beneficiaries and to identify correlates of the p
robability of referral.
Data Sources. The 1992 and 1993 Medicare Current Beneficiary Survey (MCBS),
including associated claims data. MCBS data are linked to the Area Resourc
e File (ARF) and the Physician Identification Master Record (PIMR).
Study Design. This is a retrospective design using cross-sectional descript
ive and multivariate correlational analysis. Estimates are made for two yea
rs. Key variables include two alternative definitions of referrals, patient
socio-demographic and health status, physician characteristics, and county
-level descriptors.
Data Collection. The MCBS is a panel survey of a stratified random sample o
f Medicare beneficiaries begun in 1991. The data are linked to Medicare cla
ims records for survey respondents. The ARF is a health resources data set
that contains more than 7,000 variables at the county level, including info
rmation on health facilities, health professions, services resources and ut
ilization, and socioeconomic and environmental characteristics. The PIMR is
a record of all physicians in the United States and describes their profes
sional characteristics.
Principal Findings. The overall rate of physician referrals in the MCBS, ap
proximately 10 percent, is higher than that found in prior research, as is
the level of self-referral to specialists at about 70 percent. Depending on
the dependent variable definition, between 60 and 85 percent of all Medica
re beneficiaries had at least one referral, and the average number of refer
rals per person per year was greater than two. Referrals show a multi-direc
tional pattern rather than a simple pattern of primary to specialty care, w
ith referrals between primary care physicians, referrals between specialist
s, and referrals from specialty to primary care being not uncommon. Strong
predictors of referral include patient health and patient insurance coverag
e and income. Physician factors do not contribute much to explaining referr
als.
Conclusions, Medicare referral patterns are similar to those found in other
studies. Patient factors appear to be a more important factor in explainin
g referrals than was estimated from prior research. Additional research is
needed to explain the more complex dynamics of referral patterns.