Introducing the MN APCD Public Use Files
User Webinar
Thursday, April 28, 2016
10:30-11:30 AM
Linda Green
Vice President, Programs
Freedman HealthCare
1
Agenda
History and Background of the PUFs
10 minutes
Currently Available PUFs
35 minutes
5 minutes
Access and Resources for PUF Users
Q&A
© 2016 Freedman Health Care, LLC
10 minutes
2
HISTORY AND BACKGROUND
3
Overview of the MN APCD
The Minnesota All Payer Claims Database (MN APCD) is the
foundation for the Public Use Files (PUFs)
• Established in 2008 by MN Legislature to increase transparency
• Analytic database of health care claims data for 2009-2015
• For 2013 (the baseline year for the Public Use files):
• 4.2 million Minnesota covered lives (public and private)
• 175 million medical claims
• 65 million prescription drug claims
Includes:
• Information on publicly- and privately-insured residents who received
care in Minnesota
• Information about diagnoses, procedures, duration of treatment
• De-identified demographic information (age, gender, geography)
• High-level health plan product information
• Payer, provider, and payment information
© 2016 Freedman Health Care, LLC
4
Overview of the MN APCD
The Minnesota Department of Health (MDH) uses the MN
APCD to analyze variation in cost, quality, utilization, and
disease burden; as well as conduct authorized evaluation
activities
Prior to submission, direct identifiers are encrypted using
a standardized, one-way process that does not allow reidentification of individuals
Does not include information on uninsured populations or
those covered through Tricare, Indian Health Services, or
Veterans Affairs Department
© 2016 Freedman Health Care, LLC
5
Development of the Public Use Files
2014: The legislature created an APCD Advisory Group to inform future
uses of the MN APCD. During the Advisory Group’s discussions, there
was broad support for creation of a widely accessible file known as a
“Public Use File”
2015: The MN State Legislature acted on the Advisory Group’s
recommendations and directed MDH to develop Public Use Files from
the MN APCD with emphasis on an iterative PUF development process
that engages users at all stages. This approach included:
• Reconvening the MN APCD Advisory Group to discuss various aspects of the PUFs,
including principles and options for guardrails that would protect the identities of
patients, providers and payers
• Provided for a deliberative process that would begin with a first release of data by March
1, 2016 and continue with public and user feedback on the nature of the files through
June 30, 2019
© 2016 Freedman Health Care, LLC
6
Required Characteristics of the Public Use Files
The MN State Legislature further directed that Public Use Files from the
MN APCD meet the following:
•
•
•
•
•
•
Summary data must be made available to the public at no (or minimal) cost
Available for web-based download by June 2019
Data about patients, providers and payers requires continuing protection
MDH should update the files at least annually
Documentation should clearly explain the data’s characteristics
The MN APCD should be the PUFs’ sole data source
© 2016 Freedman Health Care, LLC
7
What are the MN APCD Public Use Files?
Summarized, aggregated claims data from the MN APCD
Provides data for insured Minnesota residents who received health care
services
Considered public data and freely accessible to the public
Structured to provide meaningful information in a manageable file size while
protecting privacy
Focused on three themes:
• Health Care Services – the volume of health care services used by
Minnesotans
• Primary Diagnoses – the distribution of primary diagnoses among
Minnesotans who received health care services
• Health Care Use – the various categories of health care service use
among Minnesotans
© 2016 Freedman Health Care, LLC
8
What do the PUFs retain from the MN APCD?
What do they lose?
Similarities
• Based on the same set of records as the MN APCD
• Subject to the same identity protections for individuals
• Subject to the same rigorous quality assurance processes
Differences
• PUFs do not provide identifiable information on providers or
payers
• Some small segments of the population included in the MN
APCD are rolled up in the PUFs to prevent re-identification;
some are removed entirely
• PUFs are aggregated and therefore do not provide the same
level of analytic flexibility as a larger, more detailed dataset.
© 2016 Freedman Health Care, LLC
9
Value of the PUFs
The PUFs can describe:
• Disease prevalence
• Disease prevention
• Chronic conditions
• Service utilization
• Access to care
• Health system cost trends
What can this data contribute?
• Offer a high-level view of state trends
• Support the state’s commitment to “open data”
• Provide users with access to reliable population health data
that they can use to improve health, reduce costs, and
enhance patient experience
• Expand the potential impact of the MN APCD by broadening
the research community
© 2016 Freedman Health Care, LLC
10
CURRENTLY AVAILABLE PUBLIC
USE FILES
11
Common Features among Current PUFs
Contains medical claims data from CY 2013
• Does not include pharmacy claims
Aggregates by geography and age
• Geography aggregated into 3-digit ZIP codes
• Age aggregated into 3 age groupings
• Children and Youth: <18 years
• Adults: 18-64 years
• Older Adults: 65+ years
• Generally results in a total of 48 segments (16 ZIP codes x 3
age groups)
• Small ZIP codes are sometimes blended to cause redaction
© 2016 Freedman Health Care, LLC
12
Common Features among Current PUFs
Protects against re-identification
• Data rolled up into higher level of aggregation where necessary
due to small cell contents
• Records with <11 patients are “redacted” and rolled up into a larger
group
• Redaction occurs at geography level first, then age group
• If the field still has <11 patients after 2 rounds of redaction, it is not
included in the published PUF
• PUFs do not include records associated with <20 health care
providers or <4 distinct payers
• Redaction standards will differ by file because data
completeness varies across dimensions
• All files available in Excel and CSV; largest unzipped file is 4MB
© 2016 Freedman Health Care, LLC
13
PUF Derivation Process
How a health care transaction becomes a published record:
Consolidation – combining multiple versions of a claim that
may be submitted by a single insurance company or other payers
Deduplication – resolving all duplicate records from multiple
payers into a single, de-duplicated claim record
Aggregation – grouping and summarizing multiple de-duplicated
claim records into the PUF format, with each line of data
representing all records for each combination of PUF data elements
Redaction – protecting potentially re-identifiable information first
by eliminating geographic detail, then age group detail, and as a last
resort, removing records completely – with the goal of retaining as
much data as possible in the PUF while providing necessary identity
protections to patients, providers and payers
© 2016 Freedman Health Care, LLC
14
Data Redaction Example: Primary Diagnoses, 2013
Start
Total medical claims transactions
submitted to MN APCD
End
211.8 million
Consolidation
184.2 million
Deduplication
176.7 million
Aggregation
Claim Records:
175.2 million
Total Dollars:
$25.9 billion
Redaction
Claim Records:
174.8 million
Total Dollars:
$25.7 billion
Claim Records:
99.7%
Total Dollars:
99.5%
Percent of aggregated data
remaining in published PUF
Between deduplication and aggregation, certain records are removed from the dataset due to disqualifying factors such as miss ing service or
revenue codes, inaccurate or untranslatable information, or negative paid amounts.
The level of redaction of patient, provider, or payer information varies across dimensions within each PUF, causing some of the 48 segments to
have richer data than others.
For more information, see the Derivation (http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/diagnosesderivation.pdf) and
Summary Statistics (http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/diagnosesstatistics.pdf) documents for each PUF.
© 2016 Freedman Health Care, LLC
15
Primary Diagnoses PUF
Shows the distribution, frequency,
and cost of certain conditions
diagnosed during a health care visit
Diagnostic information comes from
the primary diagnosis recorded on
the medical claim
Uses ICD-9 diagnosis codes (9th
revision), summarized at the 3-digit
level (describes disease type or
organ system)
Shows total distinct member count
and total paid amount associated
with each diagnosis code
Overall Data Completeness of Primary Diagnoses PUF
Number of
Claim
Records
Total Dollars
Spent
Raw PUF
(pre-redaction)
175 million
$26 billion
Published PUF
(post-redaction)
175 million
$26 billion
Percent of raw data remaining
in Published PUF
99.7%
99.5%
Users may obtain ICD-9 code descriptions from the
Centers for Disease Control
(http://www.cdc.gov/nchs/icd/index.htm)
© 2016 Freedman Health Care, LLC
16
Primary Diagnoses PUF, cont’d
This PUF can help users answer the following questions:
• What is the frequency of a diagnosis reported by providers to the health
insurance company?
• How does the distribution of patients with certain conditions vary by age and
geographic area of residence?
• How much was spent on care for insured patients, on average, for the visit or
procedure under a specific primary diagnosis?
• How many patients received a particular diagnosis?
• Among all patients that received a particular diagnosis, what was the average
total amount paid?
Example: In 2013, there were 13,085 diagnoses of Streptococcal
sore throat and scarlet fever (ICD-9-CM 034) among children and
youth under the age of 18 in Minneapolis (Zip Code 554). The
combined spending for patient care related to the diagnoses was
about $2.6 M.
© 2016 Freedman Health Care, LLC
17
Primary Diagnoses PUF, cont’d
The data in this file is very complete overall, but data richness may vary across
dimensions, due to the level of redaction in certain regions or age groups:
Primary Diagnoses: Total Paid (in millions)
Region
Age Group
Raw PUF
(pre-redaction)
Published PUF
(post-redaction)
Percent of raw data remaining
in Published PUF
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Age <18
Age 18-64
Age 65+
$710
$3,020
$1,455
$710
$3,020
$1,455
100%
100%
100%
Northeast (ZIP 556)
Northeast (ZIP 556)
Northeast (ZIP 556)
Age <18
Age 18-64
Age 65+
$4
$32
$32
$3.2
80%
$29
91%
$29
91%
% of total dollars remaining in
published PUF after redaction: 99.5%
Users should compare the PUF data to the Summary Statistics
(http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/diagnosess
tatistics.pdf) provided online to determine whether the data are rich enough for
their specific purposes.
© 2016 Freedman Health Care, LLC
18
Health Care Utilization PUF
Shows the distribution of
utilization and total amount paid
across different types of health
care service use categories
Categories include office/clinic
visits, nursing facilities, hospital
admission and ambulance
Full list of utilization categories is
included in the Data Dictionary
(http://www.health.state.mn.us/
healthreform/allpayer/publicusefil
es/datadictionaryutilization.pdf)
Utilization information is derived
from the Place of Service and
Type of Bill codes reported on
institutional claims
© 2016 Freedman Health Care, LLC
Overall Data Completeness of Health Care Utilization PUF
Number of
Claim
Records
Total Dollars
Spent
Raw PUF
(pre-redaction)
176 million
$26 billion
Published PUF
(post-redaction)
176 million
$26 billion
Percent of raw data remaining in
Published PUF
99.98%
99.98%
19
Health Care Utilization PUF, cont’d
This PUF can help users answer the following questions:
• In what type of settings did patients receive care?
• What are patient characteristics (e.g. counts by age and geographic area of
residence) for settings?
• How much total health care spending was there for health care delivered in
particular settings?
• How many patients received care in a particular setting?
• Among all patients that received care in a particular setting, what was the
average total amount paid?
Example: In 2013, there were 1,389,505 encounters in a clinic or
office setting among children and youth under the age of 18 in
Minneapolis (ZIP Code 554). The combined spending for patient care
related to these visits was about $181 M.
© 2016 Freedman Health Care, LLC
20
Health Care Utilization PUF, cont’d
The data in this file is very complete overall, but data richness may vary across
dimensions, due to the level of redaction in certain regions or age groups:
Health Care Utilization: Total Paid (in millions)
Region
Age Group
Raw PUF
(pre-redaction)
Published PUF
(post-redaction)
Percent of raw data remaining
in Published PUF
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Age <18
Age 18-64
Age 65+
$722
$3,030
$1,464
$722
$3,030
$1,464
100%
100%
100%
Northeast (ZIP 556)
Northeast (ZIP 556)
Northeast (ZIP 556)
Age <18
Age 18-64
Age 65+
$6
$36
$35
$5.5
92%
$36
100%
$35
100%
% of total dollars remaining in
published PUF after redaction: 99.98%
Users should compare the PUF data to the Summary Statistics
(http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/utilizations
tatistics.pdf) provided online to determine whether the data are rich enough for
their specific purposes.
© 2016 Freedman Health Care, LLC
21
Health Care Services PUF
Shows the volume and distribution of
health care services used
Data provided at service code level
(procedure and revenue codes)
Shows counts of individuals and services,
plus total dollars spent
Overall Data Completeness of Health Care Services PUF
Number of
Claim
Records
Total Dollars
Spent
Raw PUF
(pre-redaction)
176 million
$26 billion
Published PUF
(post-redaction)
165 million
$22 billion
Percent of raw data
remaining in Published PUF
93.4%
86.4%
Users may obtain procedure and revenue code descriptions from the
following sources:
• CPT codes through the American Medical Association (http://www.amaassn.org/ama)
• HCPCS codes through CMS
(https://www.cms.gov/Medicare/Coding/MedHCPCSGenInfo/index.html)
• Revenue codes through the American Hospital Association (http://www.aha.org/)
© 2016 Freedman Health Care, LLC
22
Health Care Services PUF, cont’d
This PUF can help users answer the following questions:
• In the aggregate, what procedures were delivered in a calendar year to
patients in certain age and geographic combinations? How did this vary
across the state?
• What is the age, gender, and geographic area of residence of patients
using different types of services?
• How much, on average and in total, was spent by insurers and patients
for a particular service?
• How many times was a particular procedure performed over the course of
a year?
Example: In 2013, there were 121,332 claims for 25-minute
outpatient office visits (CPT/HCPCS code 99214) among
65,074 children and youth under the age of 18 in Minneapolis
(ZIP code 554). The combined spending for outpatient care
related to these visits was around $14.6M.
© 2016 Freedman Health Care, LLC
23
Health Care Services PUF, cont’d
Data richness varies across dimensions, due to the level of redaction in
certain regions or age groups:
Health Care Services: Record Count
Region
Age Group
Raw PUF
(pre-redaction)
Published PUF
(post-redaction)
Percent of raw data remaining
in Published PUF
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Minneapolis (ZIP 554)
Age <18
Age 18-64
Age 65+
5,622,573
21,516,983
11,336,562
4,949,611
20,518,867
11,117,313
88%
95%
98%
Northeast (ZIP 556)
Northeast (ZIP 556)
Northeast (ZIP 556)
Age <18
Age 18-64
Age 65+
43,556
21,433
49%
195,778
161,493
82%
217,152
184,979
85%
% of claims records remaining in
published PUF after redaction: 93.4%
Users should compare the PUF data to the Summary Statistics
(http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/ser
vicesstatistics.pdf) provided online to determine whether the data are
rich enough for their specific purposes.
© 2016 Freedman Health Care, LLC
24
ACCESS AND RESOURCES FOR
USERS
25
Accessing the PUFs
PUF request process designed to provide streamlined access while simultaneously
building a user community
PUF Data Request Form
(http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/requestform.pdf)
available for download on MN APCD website
(http://www.health.state.mn.us/healthreform/allpayer/publicusefiles/request.html)
• Form asks the user to provide contact information and acknowledge the Data
Disclaimer
• The MN APCD team at MDH will coordinate with the user to transmit the requested
file(s)
Building a user community will:
• Allow MDH to provide technical assistance to users
• Keep users updated on future PUF releases
• Share findings and lessons learned among users
• Find out what users want from future PUFs
MDH will reach out to the user community with questions and requests for feedback
© 2016 Freedman Health Care, LLC
26
Resources for Users
PUF documentation available on MN APCD website
(http://www.health.state.mn.us/healthreform/allpayer/publicus
efiles/index.html)
• Overview of the PUFs
• A detailed description of each PUF, including:
• Data dictionary – describes each data element included in the PUF
• Summary statistics – shows summary totals from the pre-redaction PUF,
for purposes of comparison to the final published file
• Derivation document – describes the process through which the PUF was
developed from the MN APCD
Check the website for future PUF updates
If you have questions or would like to join the MN APCD
mailing list, email [email protected]
© 2016 Freedman Health Care, LLC
27
Future Directions for the PUFs
The current PUFs will be updated at least annually
MDH is exploring options for expanding the PUFs
to include additional years of data and additional
topics of interest for users
User feedback will inform future PUFs
Survey will be sent to known users
Frequently Asked Questions document will be
posted and updated
© 2016 Freedman Health Care, LLC
28
QUESTIONS OR COMMENTS?
[email protected]
© 2016 Freedman Health Care, LLC
29