tion at their disposal for identifying candidates for intervention, with claims data
and notification requirements providing only partial insights into what was done to
the patient and even less insight into the patient’s underlying clinical condition and
future need for care. The diversity in health status and care-seeking behavior implies
that intervention programs should be tailored to the distinct needs of patients with
catastrophic conditions, those suffering from chronic illnesses, those with acute but
temporary needs for services, and those who need to be reminded to take advantage
of preventive and primary care services.
Health plans are at the center of the flow of reimbursement claims and can ob
-
serve patterns of physician, hospital, laboratory, and pharmaceutical care better
than anyone else in the fragmented delivery system. If successfully captured,
cleaned, integrated, warehoused, analyzed, and interpreted, claims data consti
-
tute a powerful tool for the identification of patients with current and future
needs, the comparison of existing patterns of care with clinical guidelines, the
stratification of providers according to quality and efficiency, the underwriting
and pricing of insurance products for particular purchasers, and myriad other
uses. All of the major health plans have invested heavily in data and analytic sys-
tems in the effort to develop and exploit this asset. However, claims data as a
means for identifying disease management candidates suffer from important limi-
tations that are yielding only slowly to efforts at refinement. The data are incom-
plete, often lacking claims for services provided by carved-out pharmaceutical
benefit managers (PBMs), carved-out managed behavioral health organizations
(MBHOs), and capitated physician practices. Diagnostic coding is weak on physi-
cian claims, clinical values are missing for tests performed in small local laborato-
ries, and the many tests and procedures done during a hospital stay may be rolled
up into a terse account. Claims often come in through multiple transmission sys-
tems and reside on multiple computers that communicate with each other imper
-
fectly, especially for health plans that have grown through mergers and acquisi
-
tions. Claims are entered into the data warehouse only when they are paid;
complex and contested claims must wait for the resolution of disputes over cover
-
age, coding, pricing, and medical necessity.
n
Intervention. The principles underlying intervention are the same as those un
-
derlying identification. The skewed distribution of spending, opposition from phy
-
sicians, skepticism among patients, and diversity among purchasers drive health
plan intervention programs to focus their attention on a small number of conditions,
limit resources, measure impacts, and structure programs into a core set of services
with buy-up options. The salient feature of most health plan initiatives is their em
-
phasis on changing the behavior of patients, rather than that of physicians. Efforts to
change physicians’ behavior require expensive peer-to-peer contact by health plan
medical directors, and they run the risk of further antagonizing the targeted practi
-
tioners. Patient-oriented programs are managed by nurses and nonclinical staff, tar
-
get chronic conditions where self-care is important, and are consistent with the
Managed Care
HEALTH AFFAIRS ~ Web Exclusive W4-271