The case definitions below represent the first, provisional set of case definitions used by this system. These case definitions, and the resulting estimates are subject to change as we continue to validate and improve the case definitions and methods. We published the methods for developing our case definition in JAMA Open Network on September 3, 2024.
For questions or to access public use data, contact us at:
Includes people with dementia-specific diagnosis codes on at least two separate dates over a three-year period.
Includes people with dementia-specific diagnosis codes on only one date over a three-year period.
Includes people with dementia-specific diagnosis codes on at least one date over a three-year period. Inclusive of all people in the likely and highly likely categories
Includes people who did not have a dementia-specific diagnosis code, but did have one or more diagnosis codes that may indicate dementia (including mild cognitive impairment, age-related physical debility, or other degenerative disease of the nervous system), or a prescription drug indicated for dementia without a corresponding dementia diagnosis code, observed over a three-year period.
The sum of all persons with dementia and possible but nonspecific diagnoses.
The case definitions above are considered provisional (version 1). Case definitions used by the Dementia Datahub may change based on the results of ongoing validation research, changes in ICD10 coding, or other reasons.
To develop provisional case definitions to identify dementia and possible dementia in administrative claims and encounter data we reviewed existing diagnostic and drug code case definitions and categorized them based on their frequency of inclusion in prior research. With expert input, we removed certain codes from the case definitions, and used other data in Medicare records to compare people in each category to those in other categories and to people with no evidence of dementia or possible dementia.
While these case definitions are derived from a comprehensive review of previously published case definitions, they have not been validated for accuracy. The Dementia DataHub project is currently conducting a validation analysis to assess the sensitivity and specificity of these provisional case definitions for identifying people with dementia as determined by dementia assessment tests and through clinical observation. Based on the results of this study, these provisional case definitions may change.
We strongly encourage system users to carefully read the system's publication describing how these provisional case definitions were developed and the evidence to support them
The “Highly Likely Dementia” category includes people with dementia-specific diagnosis codes on at least two separate dates over a three-year period. Dementia-specific is defined as a clinical diagnosis code that has either been used previously in 10 or more algorithms to identify dementia, or that is a newer code that is in the same category of diagnoses that have been used more than 10 times. Codes that have not been used more than 10 times but that are in the same category as codes that have been, are codes that should not be used for reimbursement purposes, but sometimes appear in claims. An example is G31.0, “Frontotemporal dementia”.
We base our determination of “Highly Likely” on the fact that people in this category have received these diagnoses on two separate dates, and on the fact that Medicare claims evidence shows that people in this category are older, more frail, more likely to reside in long term care facilities, and more likely to die in a given year, than people in the “Likely” or “Possible” categories.
The “Likely Dementia” category includes people with the same dementia-specific diagnosis codes as in the “Highly Likely” category, but who received these diagnoses on only one date over a three-year period. We are less confident that these people have dementia as defined by a dementia assessment instrument or as defined by a clinician than people in the “Highly Likely” category. Medicare claims evidence shows that people in this category are older, more frail, more likely to reside in long term care facilities, and more likely to die in a given year, than people in the “Possible” category, but less likely than those in the “Highly Likely” category.
The “Possible Dementia” category includes people who did not have a dementia-specific diagnosis code but did have one or more diagnosis codes that were included by some previous research studies and/or the CMS Chronic Condition Warehouse to identify Alzheimer’s Disease Related Dementias. This category includes people who have been diagnosed with mild cognitive impairment, age-related physical debility, other degenerative disease of the nervous system, or a prescription drug indicated for dementia without a corresponding dementia diagnosis code, observed over a three-year period. Based on Medicare claims data evidence, people in this category are older, more frail, more likely to reside in long term care facilities, and more likely to die in a given year than people without any of the diagnoses used in this system, but less likely than people in the “Highly Likely” or “Likely” categories. Some people in this category may have dementia, but many may not, and may never progress to dementia in their lifetime.
The sum of all persons in all three categories.
Users should note that it is impossible for administrative claims and encounter data to correctly identify all people correctly with dementia in administrative claims data. Major sources of error include coding uncertainty and underdiagnosis.
Coding Uncertainty: Clinicians may code observations of clinical signs of dementia differently based on many factors. Further, the sensitivity and specificity of administrative records in detecting dementia as defined by a physician or an assessment instrument is not known. The Dementia Datahub is currently conducting research comparing the system’s provisional case definitions to dementia assessment instruments used in the Health and Retirement Survey. Other studies are comparing our provisional case definitions to physician determinations of dementia conducted as part of the RUSH University Memory and Aging Project. We expect this research to lead to at least some changes in our case definitions in future years of the system.
Underdiagnosis: This system is not capable of detecting and does not include people with dementia who have not received a diagnosis or prescription drug code listed below in their Medicare claims or encounters records. Research suggests that more than half of those people considered to have dementia based on assessment or ascertainment tests were undiagnosed, and that non-Hispanic Black and Hispanic people are most likely to have a missed or delayed diagnosis. Therefore, the results from this system should be interpreted only as those people currently identified with dementia or possible dementia in Medicare, and not as a true accounting of all people in Medicare with dementia or possible dementia because many cases are unrecognized.
Codes Used for “Highly Likely” and “Likely” Categories. A “Highly Likely” categorization requires the use of these codes on two or more different days over a three-year time period for prevalence or over 1 year for incidence. A “Likely” categorization requires the use of one or more of these codes on one day only.
Codes used to identify people with “Possible Dementia”. These people may possibly have dementia, may be experiencing normal age-related memory loss, or they may not have any signs of dementia at all.
Any utilization of the prescription drugs Rivastigmine, Galantamine, Memantine, Donepezil, or Tacrine without any diagnosis code previous listed. List of NDC codes below. The system classifies people who received one of these drugs but who did not have a diagnosis code listed in the table as “Possible Dementia” cases.
