We recommend that users infer service in a long-term care facility on a monthly basis (for one or more of the long-term care TOS values), as follows: (1) If there is one (or more) MAX records for long-term care services for the individual, covering any number of days in the month, assume that the individual resided in that type of institution in the month, and (2) If there were no MAX records for long-term care services for the individual, for any days in the month, assume that the individual did not reside in that type of institution in the month. For a user who plans to conduct analyses spanning several months or the entire year, we recommend that the users separate long-term care residents into two groups:
• The first group could be described as continuous or “fully” institutionalized persons. These are enrollees who had records for long-term care services covering every month of Medicaid enrollment during the year.
• The second group could be described as discontinuous or “intermittently” institutionalized persons. These are enrollees who had records for long-term care services covering at least one month but fewer than the total number of months of Medicaid enrollment during the year.
An alternative approach to that presented above, might be to count the number of Medicaid covered days, for the type of institution(s), in a month (or year) and compare that day count to the number of days the individual was eligible for Medicaid in a month (or year). This approach is NOT valid. For example, a user might be inclined to say that an individual is: (1) “fully” institutionalized if the number of institutional days is greater than or equal to the number of days of eligibility, (2) “intermittently” institutionalized if the number of institutional days is greater than zero but less than the number of days of eligibility, or (3) “non-institutionalized” if the number of institutional days is less than or equal to zero. Medicaid long-term care records, as reported in the MAX data, CANNOT support this type of categorization for enrollees. Day counts may vary because an individual has leave days (days the person is not in the institutional facility for a variety of reasons), acute days (days the person has an acute episode involving a hospital stay), or institutional days covered by other payers (e.g. Medicare, private pay, or long-term care insurance). In addition, institutions may submit bills in other ways that create problems in counting institutional days. A hypothetical example may illustrate the problem. Say that a patient was eligible for Medicaid from March 1 through March 31, was institutionalized in facility X from March 5 continuously through March 31 and received facility-based ancillary services (e.g. physical therapy) between March 10 and March 31. The facility may submit two bills: (1) For the per-diem amount, spanning 26 days, and (2) for the ancillary services, spanning 22 days. The total day count for the two claims would be 48, greater than the number of days in the month. If the day field were edited to be less than or equal to 31 days, a data user might assume (incorrectly) that the person was “fully” institutionalized for all of March. There is no sure way to determine which bill was for the per-diem and which bill was for ancillary services. Because of this, there is no reasonable way to resolve this dilemma. So, determining institutional status on the basis of day counts is not advised, because it may not produce valid and consistent results.