Clinic Methodology (PDF: 206KB/2 pages)

Clinic Methodology
September 16, 2013
To meet the goal of producing a confidential report to Clinics and Medical Groups, several decisions
regarding the methodology for the first version of the report have already been made by MDH staff
in consultation with its analytical contractor, Mathematica Policy Research (MPR). The empirical
results yielded from these decisions and reported confidentially to physician clinics will serve as the
basis for feedback from the Rapid Response Team in preparation of the second iteration of the
physician clinic report.
Data Source
 The data used in the Provider Peer Grouping initiative comes from Minnesota’s All Payer
Claims Database (APCD). The APCD collects healthcare claims data from health plan
companies, third-party administrators and pharmacy benefit managers, excluding small
payers with less than $3 million in paid claims for MN residents per year or less than
$300,000 in paid pharmacy claims per year. Payers are required to submit this data under
Minnesota law.
 The data collected is encrypted and de-identified, but still allows us to aggregate a patient’s
experience across health care systems in a meaningful way. The data excludes claims for
non-Minnesota residents, certain types of coverage (such as disability-only and workers
compensation) and certain types of payers (like Medigap, Tricare and Indian Health
Services).
 The data used in the clinic analysis includes only primary care clinics or multi-specialty clinics
that provide a majority of primary care. Specialty clinics such as neurology or orthopaedic
clinics are excluded from the analysis.
Attribution1
 In keeping generally with community practices, we will attribute sole responsibility for a
patient’s care to the clinic that saw the patient the most, using an algorithm that assigns
patients to clinics and medical groups with at least 50 percent of that patient’s office or
other primary care related visits. Ties will be broken in favor of the clinic that was paid the
most. Patients who are not provided services will not be attributed to any provider in this
first report version.
 Because a certain percentage of physicians are registered at more than one practice
location, we will attribute patients to both clinics and medical groups according to the
following rule. Patients who primarily see a physician who practices at only one clinic will be
attributed to both the clinic and to the medical group associated with that clinic. Patients
who primarily see a physician who practices at more than one clinic will be attributed only
at the medical group level.
 Only clinics and medical groups that provide a minimum threshold of primary care services
are eligible to receive a report. This ensures that we are not comparing specialty care with
primary care.
Risk-Adjustment
1
This is work conducted with the assistance of experts at the Division of Health Policy and Management at the
University of Minnesota (U of M) School of Public Health and funding from the Clinical and Translational
Science Award to the U of M by the NIH.

We will use the Adjusted Clinical Groups (ACG) System to assign patients to risk groups and
adjust for differences in risk across clinics and medical groups. Although the ACG typically
uses a year’s worth of data, we will test shorter time durations to determine whether they
will be just as accurate. A shorter time frame will allow us to include more patients in our
analyses.
Cost and Price Standardization
 MPR and MDH have a detailed plan on how to assign standardized costs to specific types of
services. Medicare rates will be the yardstick by which we measure these services. After
patients are attributed, we will conduct reliability and outlier tests. Reliability tests will allow
us to determine minimum cell sizes for stable estimates and outlier tests will ensure that
estimates are not biased by a few rare, extreme costs.
Quality Measurement
 We have finalized a set of 39 quality measures, which we drew from the following data
sources: HEDIS, State Quality Reporting and Measurement System (SQRMS), CMS,
Readmission, and AHRQ. MDH has started writing the algorithms to calculate the 17
measures that will be based on claims data. After attribution is complete, we will test
reliability thresholds for each measure and a composite measure and discuss results with
the reliability workgroup.
 As part of the quality measurement approach, MDH will introduce an all-cause readmission
measure. MDH is working with the RARE leadership committee to test the measure and
plans to share relevant data with hospitals prior to the confidential release. For the release
the measure will not be part of the overall quality score.
Peer Grouping
 Given that we do not know whether size, location, or other differences between clinics
affect cost and quality in any substantive way, in the first, confidential report we will not
break clinics or medical groups into additional peer groups. If meaningful differences
emerge after the initial results are examined, those differences can be included in the next
iteration of the reports.
 The 2009 Advisory Group (AG) recommended that, wherever possible, peer grouping should
occur at the clinic site level, although there were concerns among the group about the
practicality of doing so. However, because some physicians practice at more than one clinic,
we decided to create both clinic and medical group level measures for the clinic reports.
 We have reduced unlike comparisons by reporting only on clinics and medical groups that
have sufficient staff capacity to provide a significant volume of primary care. This will
prevent comparisons between specialty clinics, which provide only a circumscribed set of
services, and primary care clinics, which provide a broader set of services.
 Our risk adjustment methodology will account for differences in disease burden across our
units of analysis, using Adjusted Clinical Groups (ACG) System to categorize patients into
groups of expected resource use, and then regression models to adjust for these
differences.