Rosemarie Hakim PhD CMS Background 2 Medicare data have been available for research for decades Privacy Act of 1974 allows use of identifiable data for research by a recipient who has provided CMS ID: 778125
Download The PPT/PDF document "Accelerating CMS Outcomes Data to Near R..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Accelerating CMS Outcomes Data to Near Real Time: Challenges & Solutions
Rosemarie Hakim, PhD CMS
Slide2Background
2
Slide3Medicare data have been available for research for decades
Privacy Act of 1974 allows use of identifiable data for research by a recipient who has provided CMS “with advance adequate written assurance that the record will be used solely as a statistical research or reporting record, and the record is to be transferred in a form that is not individually identifiable
”
The Computer Matching and Privacy Protection Act of 1988
allows matching of federal records with non-federal records to produce aggregate statistical data without any personal identifiers
3
Slide4What Works Well Today
4
Slide5Available dataChronic Condition Warehouse (CCW)
A research database that contains 100% Medicare files and.. Medicaid files
Assessment files
Part D Prescription Drug Event data
for Fee-for-service institutional and non-institutional claims
Linked by a unique, unidentifiable beneficiary key allow analysis across the continuum of care
5
Slide6CCW contd.
Plan characteristicsPharmacy characteristicsPrescriber characteristicsFormulary file - beginning with year 2010
CCW data files may be requested for any of the predefined chronic condition cohorts, or users may request a customized cohort(s) specific to research focus areas.
Chronic Conditions Dashboard
6
Slide7CCW conditions Acquired Hypothyroidism
Acute Myocardial InfarctionAlzheimer's DiseaseAlzheimer's Disease, Related Disorders, or Senile DementiaAnemiaAsthma
Atrial Fibrillation
Benign Prostatic Hyperplasia
Cancer, Colorectal
Cancer, Endometrial
Cancer, Breast
Cancer, Lung
Cancer, Prostate
Cataract
Chronic Kidney Disease
Chronic Obstructive Pulmonary Disease
Depression
Diabetes
Glaucoma
Heart Failure
Hip / Pelvic Fracture
Hyperlipidemia
Hypertension
Ischemic Heart Disease
OsteoporosisRheumatoid Arthritis / OsteoarthritisStroke / Transient Ischemic Attack
7
Slide8Medicare – ccw condition period prevalence , 2010
8
Slide9Cardiovascular conditions- Trends
9
Slide10Other data available 10
Master Beneficiary Annual Summary File
Durable Medical Equipment
Medicare-Medicaid Linked Enrollee Analytic Data Source
MedPAR (Hospital and SNF)
Outpatient
Others (see ResDAC.org)
Strengths of CMS Administrative Data11
Clinical validity - accurate and reliable:
Admission and discharge dates, diagnoses, procedures, source of care, demographics, place of residence, date of death,
Link to Other CMS Datasets
Population Coverage
>98% percent of adults age 65 and over are enrolled in Medicare.
> 99% percent of deaths in the US among persons age 65 and older are accounted
> 45 million beneficiaries enrolled in the Medicare program, allowing for detailed sub-group analysis with high statistical power.
Linkage to External Data Sources
:
US Census
Registries
Other providers (e.g. VA, Medicaid)
National death index/State vital statistics
Surveys (e.g. Health and Retirement Study)
Provider Information
Slide12What Is Missing, Broken or Does Not Work Well Today
12
Slide13Reliance on billing codes13
Conditions must be diagnosed to appear in the utilization files
Some diseases (hypertension, depression and diabetes) are underdiagnosed
No information on care
needed
but not provided
Services that providers know will be denied may be not be submitted as bills
Diagnosis information may not be comprehensive enough for detailed analysis
Prevalence may be misinterpreted as incidence
: knowing a person has a chronic disease does not reveal how long they have had the condition or the severity of their condition
The Part D prescription drug event file contains no diagnosis codes
Slide14Reliance on billing codes14
Different care settings use different coding systems for procedures
Inpatient care is coded using ICD-9 procedure codes
Physician/supplier and DME data use CPT and HCPCS codes
Hospital outpatient care is a mix of CPT and revenue center code
No physiological measurements or test results
Not all beneficiaries have Part D coverage
Little information of unknown quality available about managed care enrollees
No information on services for which claims are not submitted (e.g. immunizations provided at Walgreens)
Slide15Other limitations15
Specific programing expertise needed to analyze claims
In most cases, complex statistical techniques needed to correct biases
Propensity scores
Missing data algorithms
Data validation techniques
Severity adjusters
Sensitivity analyses
Complex regressions
Slide16Challenges and solutions
16
Slide17Research Data Time Lag 17
CCW data on 2-year lag for general research community
However – closer to real time data are available
In 6 months
96.7% of inpatient and 96.9% of outpatient claims are complete
How to get closer to real time data
Affordable Care Act allows
qualified entities
to acquire data for the evaluation of the performance of providers of services and suppliers
Data use agreement under a contract with CMS
Slide18Matching Data to Medicare Claims18
Deterministic matching
Use
unique
personal identifiers (UPIs) present in Medicare claims and in registry/trial data
Good
Matching SSNs
Better
Matching SSNs and DOB
Best
Matching SSNs, DOB, gender, and provider
Slide19Matching Data without UPIs19
No unique identifiers in data to be matched to claims
Good results can be obtained using non-unique variables:
DOB or age
Dates (admission, procedure date)
Gender
Hospital
Geographic region
Provider
Diagnosis
Slide20Matching Data without UPIs contd.
20Probabilistic (fuzzy) matching
Uses wide range of potential identifiers
Computes weights based on sensitivity & specificity of identifier
Weights used to calculate the probability that 2 records refer to the same entity
Slide21Matching rates 21
Authors
Data source
Type of matching
Results
St. Peter et al. 2011
Dialysis Clinical Outcomes Revisited (DCOR) Trial/Medicare
Unique identifiers
Nearly 100%
Brennan
et al. 2012
PCI Registry/Medicare
Deterministic
86%
Hammill
et al. 2009
Heart failure registry/Medicare
Deterministic
81%
Hammill
et al. 2009
Hospital HF records /Medicare
Deterministic
91%
Setoguchi
et al. 2012
ICD
Registry/Medicare
Deterministic
61%
Setoguchi
et al. 2012
ICD
Registry/Medicare
Probabilistic
85%
CDC/NCHS
2003-2004
NHANES
/Medicare
Probabilistic
98%
Slide22Short term priorities
22
Slide23Make Good Use of CMS Data 23
Build linking capability into study or registry
Include capability to link to Medicare claims data in informed consent
Plan data collection to include important linking variables
Use data for long term follow up for IDE studies and RCTs
Slide24Make Good Use of CMS Data contd.
24Develop expertise – use of administrative data is increasing
Educational materials on CMS and ResDAC websites
ResDAC gives courses on using CMS data
Develop statistical expertise in using administrative data -
Slide25Long Term Priorities
25
Slide26Health Data Initiatives26
Office of Information Products and Data Analytics (OIPDA)
Develops, manages, uses, and disseminates data and information resources
Goal of improving access to and use of CMS data
Manages the
CMS Data Navigator
- web-based search tool
CMS’ EHR incentive program – encourages data interoperability and development of Health Information Exchanges
Slide27Thank you 27
rosemarie.hakim@cms.hhs.gov
Chronic Conditions Data Warehouse
https://www.ccwdata.org/web/guest/home
ResDAC
http://www.resdac.org/