Using CCW Data for Prevalence Studies of Alzheimer’s Disease and Related Dementias (ADRD) Presented by Galina Khatutsky, Edith G. Walsh and Nancy T. McCall RTI International James E. Leonard and Wendy Funk Kennell and Associates, Inc . AcademyHealth Annual Meeting Chicago, Illinois June 29, 2009 www.rti.org RTI International is a trade name of Research Triangle Institute Study Consultants • Malaz Boustani, MD, MPH, Indianapolis Discovery Network for Dementia, Healthy Aging Brain Center at Wishard Senior Care, IU School of Medicine, Regenstrief Institute, National Institute on Aging • Christopher Murtaugh, PhD, Center for Home Care Policy and Research, Visiting Nurse Service of New York. www.rti.org Study Goals • Document issues in studying ADRD using various data sources, focusing on Medicare data (literature review) • Test the usability and explore new opportunities available in the Medicare Chronic Condition Warehouse (CCW) for the study of ADRD – The CCW supports analyses combining Medicare claims and clinical assessment data from home health and nursing facilities, resulting in a unique and powerful resource for ADRD analyses www.rti.org Chronic Condition Data Warehouse (CCW) • • • A data source well suited for studying ADRD and other conditions; summary flags for 21 chronic conditions Medicare data for institutional and non-institutional services for FFS Medicare beneficiaries. Files linked on beneficiary level: continuum of care across settings – Medicare beneficiary claims (final action) – Medicare enrollment – Clinical and in-person assessments for • Home Health recipients • Nursing Facility residents • Inpatient Rehab residents – Medicare Current Beneficiary Survey (MCBS) • Funded by CMS and available to researchers www.rti.org CCW: Overview • Database developed in response to Medicare Modernization Act – Calendar year 1999 forward • Enhanced 5% Sample (1999-2004) – Once in, always in, regardless of changes in Medicare Health Insurance Claim Number (HICN) • Beginning in 2005, CCW contains 100% data, – Enhanced 5% data can still be obtained • Significant advantage over routine Medicare 5% sample data for longitudinal analyses www.rti.org Chronic Conditions in CCW • • • • • • • • • • • • • • • • • • • • www.rti.org Alzheimer’s Disease Alzheimer’s Disease and Related Disorders or Senile Dementia Acute Myocardial Infarction Ischemic Heart Disease Osteoporosis RA/OA (Rheumatoid Arthritis/Osteoarthritis) Stroke / Transient Ischemic Attack Female Breast Cancer Colorectal Cancer Prostate Cancer Lung Cancer Endometrial Cancer Atrial Fibrillation Cataract Chronic Kidney Disease Chronic Obstructive Pulmonary Disease Depression Diabetes Glaucoma Heart Failure Challenges estimating the future disease burden of ADRD • ADRD prevalence is generally under-reported • Published rates vary significantly due to different methodologies and data sources • There are challenges to identifying ADRD from Medicare claims or any other single data source • Each data source offers unique information and identifies a different population subset www.rti.org Identifying Persons with ADRD Using Medicare Claims Strengths • Available for entire FFS Medicare population across care settings • Useful for calculating prevalence of treatment for a condition (claims-based treated prevalence) Challenges • • • • Selecting appropriate diagnostic codes Determining look-back periods Selecting appropriate Medicare claims files Avoiding potential bias – Undercounting if treatment is for other conditions and ADRD diagnosis is not recorded during episode – Overcounting if an ADRD diagnostic code is a “rule out” www.rti.org Studying ADRD Using the CCW • Appropriate for ADRD research due to combination of claims data and in-person clinical assessments for – Medicare beneficiaries in nursing facilities: Minimum Data Set (MDS) – Medicare beneficiaries receiving home health: Outcome and Assessment Information Set (OASIS) • Allows diagnostic validation across data sources www.rti.org Research Questions • What is the prevalence in calendar year 2005 of Alzheimer’s Disease and related dementias (ADRD) in the Medicare population comparing – Fee-for-service (FFS) claims data? – Claims plus in-person clinical assessment data from • Nursing facilities (MDS)? • Home health (OASIS)? • How much agreement is there between claims-based and assessment-based diagnoses of ADRD? • What is the value in using assessment data for prevalence estimates? www.rti.org Medicare Claims • Claims-based Alzheimer's’ Disease and Related Dementias (ADRD) flag – 3 year look back period – At least one claim from hospital inpatient, hospital outpatient, physician, skilled nursing facility or home health files – Any diagnosis within a claim, not restricted to primary diagnoses • Diagnostic flags are based on diagnosis ICD-9 codes; determined by CMS together with a panel of experts – ADRD ICD-9 codes: 331.0; all other dementia codes in the 290, 331, and 294 series; and 797 www.rti.org Nursing Home Minimum Data Set (MDS) • Allows prevalence estimates for those without acute care Medicare claims related to ADRD • Completed routinely, regardless of payor (Medicare, Medicaid or private pay) and includes: – ADRD diagnosis entered by RNs • Check-off items for AD or other dementias – Spaces to enter additional ICD-9 diagnostic codes – Measure of cognitive function: Cognitive Performance Scale (CPS) www.rti.org Variables derived from the Nursing Home Minimum Data Set (MDS) • Diagnoses (ADRD) collected for nursing facility residents • Cognitive Performance Scale (0 -6) 0 = intact 1 = borderline intact 2 = mild impairment 3 = moderate impairment 4 = moderately severe impairment 5 = severe impairment 6 = very severe impairment – validated MDS measure of cognitive impairment commonly used to assign residents into easily understood cognitive performance categories • Details on CPS are available in Morris, Fries, Morris et.al., "MDS Cognitive Performance Scale" J. Gerontology: Medical Sciences 49(4):M174-M182 (July) 1994. www.rti.org Home Health Outcome and Assessment Information Set (OASIS) • Allows prevalence estimates for those with Medicare home health utilization • Completed at each start and resumption of care episode, and at follow-up • Diagnoses and associated ICD-9-CM codes on each home care episode entered by RNs or therapists – Measures of cognitive function – ADRD may be undercounted • ADRD diagnosis is rarely a reason/focus for home health episode • Persons with ADRD diagnosis may not be considered as “improving” and so ADRD diagnosis may not be recorded www.rti.org Variables derived from Outcome and Assessment Information Set (OASIS) • Diagnoses of dementia • Behavioral symptoms of cognitive impairment – Memory deficit: failure to recognize familiar persons/places, inability to recall events of past 24 hours, memory loss significant enough that supervision is required – Impaired decision making: failure to perform usual activities of daily living (ADL) or instrumental activities of daily living (IADL), inability to appropriately stop activities, safety jeopardized through actions www.rti.org Study Sample (n=88, 476) • Medicare beneficiaries in the CMS Chronic Condition Data Warehouse (CCW) – Enhanced 5% Medicare sample AND – Medicare Part A and Part B, full FFS coverage AND – MDS clinical assessment present AND – OASIS clinical assessment present – Reference year 2005 – Look back period 3 years www.rti.org Analysis Plan • Compare ADRD prevalence for persons with all 3 types of records using: – Diagnosis only (identified using claims, Nursing Home Assessments (MDS), Home Health Assessments (OASIS), or any combination of these sources) – Diagnosis or behavioral/cognitive impairment (identified in MDS or OASIS) www.rti.org ADRD Prevalence for Study Sample Varies By Data Source Combinations (percent) • Claims-based diagnoses 40.1 • MDS-based diagnoses • MDS-based diagnosis or CPS>1 6.5 39.0 • OASIS-based diagnoses • OASIS-based diagnoses or symptoms of cognitive impairment 2.0 • All sources combined 54.3 www.rti.org 17.1 Agreement Between Data Sources in Identifying ADRD Good Agreement Claims/All other Sources KAPPA Coefficient 0.720 Moderate Agreement Claims / both MDS sources 0.507 Claims / MDS Cognitive Performance Scale 0.489 Fair Agreement Claims / both OASIS Sources 0.252 Claims / OASIS Behavioral Symptoms 0.240 Slight Agreement Claims / MDS Diagnosis 0.181 Claims /OASIS Diagnosis 0.056 www.rti.org Limitations • These multiple data sources are only available for the most impaired subsets of the Medicare populationusers of home health and nursing facility care • Some of the lack of agreement may relate to disease progression and when in the 3 year time period the Medicare claims, OASIS or MDS use occurred • Measures of cognitive impairment may indicate conditions other than ADRD www.rti.org Conclusions • Each individual source of data appears to underestimate the prevalence of ADRD • Medicare claims data appear to capture about ¾ of beneficiaries with ADRD • Diagnoses from nursing facility and home health data appear limited – Perhaps due to lack of incentive to list a complete set of diagnoses www.rti.org Conclusions • Addition of cognitive impairment measures from home health or nursing facility assessment data substantially increases prevalence estimates within each data source • Addition of nursing home and home health assessment data produces an 35% increase in prevalence over claims-only • Combination of MDS diagnoses and CPS produces prevalence estimates comparable to claims-only • Agreement between all other data sources combined and claims based identification of ADRD is good, however each separate source compared with claims has moderate agreement at best • A combination of data sources seem to be the best way to create a more accurate estimate of prevalence: CCW provides an effective way to conduct studies comparing data sources or to maximize identification and track beneficiaries over time http://www.resdac.umn.edu/CCW/data_available.asp www.rti.org