Can program enrollees with specific illnesses or medical conditions be identified in the Medicaid Analytic eXtract (MAX) data?
Users must remember that the MSIS and MAX data are extracts from state Medicaid data systems used for eligibility and claims payment administrative systems. As such, these data do not directly identify persons with specific illnesses or medical conditions. Instead, users should develop a set of criteria to identify a catchment group of individuals who may have the illness or medical condition. The criteria may include specific diagnoses, procedures, prescription drugs or services that were provided to the person. Users should be careful to develop the criteria in a way that casts a “broad net” of possible individuals for the study population. The idea is to develop criteria that include all possible individuals, even those who may marginally meet the criteria. Once a set of initial criteria is developed and run against the data, users should examine profiles of the selected individuals to determine if the criteria should be refined to remove some individuals from the study population.
For criteria related to diagnosis codes, users should consider: ICD-9-CM
• With regard to diagnoses, users should consider whether they want to select persons based on all diagnoses reported or just on the principal diagnosis, given that states may vary in how they determine which diagnosis to report as the principal diagnosis.
• Users should be aware of the potential for under-reporting of some diagnoses.
• Users may want to use the MAX Validation Tables to examine the percent of records containing primary diagnosis codes of length 3, 4 or 5.
• Users should be aware that diagnosis codes for persons in nursing facilities and other institutions may not be up-to-date or complete.
For criteria related to inpatient hospital procedures, users should consider that states are not required to use a single standard coding system. Different systems (e.g. ICD-9-CM and CPT-4) may be in use even within a single state.
For criteria related to outpatient procedure (service) codes, users should know that some states are required to use standardized coding (Level 1 – CPT codes and Level 2 – HCPCS codes) for most ambulatory services. States also use Level 3 state-defined (state-specific) codes in both standard and non-standard formats. There are two standard formats: (1) ANNNN where A has an alpha value of W-Z and N is numeric or (2) AANNN where A is alpha and N is numeric. However, states sometimes use non-standard formats for state-specific codes. States also use UB-92 codes for services billed on UB-92 forms (e.g. hospital outpatient services). The procedure (service) code modifier may be useful to provide more information about services provided that relate to a procedure (e.g. assistance in surgery). However, users should exercise care not to over-count services delivered. For example, there may be a single surgery with more than one service record for the surgery (e.g. a physician record for the surgery and a second record for assistance in that surgery).
Data users may also want to consider the use of prescribed drugs as indicators of selected conditions. Users should note that injectable drugs that are administered by a medical provider will be found in the MAX OT file.
Using claims-based criteria to identify a catchment group in this way has some important limitations for some groups of Medicaid enrollees. For persons enrolled in prepaid managed care plans, reporting of “encounter” data reporting from most plans has been judged to lack the completeness and consistency to support a wide variety of research activities at this time. Also, claims data for dual Medicaid and Medicare enrollees may be missing diagnosis and/or procedure codes. This may result from Medicare’s role as the primary payer of care for dual enrollees. In addition, details on service utilization may not be available for all women receiving prenatal care. Users should review the response to the Question “Can I obtain a complete view of all health services delivered to Medicaid enrollees by using the MAX data?” for more details.
Even with regard to the same illness or medical conditions, not all users may agree on the criteria. For example, not all researchers studying diabetes will want to select the exact same set of diagnosis codes. For some types of research, users may want to consider if they want to identify a study population by defining a sentinel event (e.g. diagnosis of asthma by a physician) and then identify a “window” of time before or after the event to examine services delivered. Users should also consider whether they simply want to examine services for only the illness or medical condition (e.g. diabetes care) or whether they want to examine all services provided to individuals who have the illness or medical condition (e.g. all services to persons who have diabetes). These considerations may result in different record selection criteria based on the user’s specific research questions.
Users should be reminded that MSIS and MAX data contain records of services provided to Medicaid enrollees up to the amount, duration and scope of coverage provided by each State Medicaid agency. If the services were covered by another payer (e.g. Medicare, Veteran’s Administration, out-of-pocket, block grants, or Ryan White coverage for HIV/AIDS patients), some services used by Medicaid enrollees with the target illness (or condition) may not be included in the MAX data. Therefore, some Medicaid enrollees with the target illness (or condition) may not be identified as having the target illness (or condition) by using the MAX data.
Users should also be aware that identifying persons with some illnesses and medical conditions may be fairly direct and the catchment group may include nearly all of the target population. For others, such as persons with HIV/AIDS, identification of the entire population (or even a representative subpopulation) may be much more difficult.