Inner TRIM3 Masthead
Home | TRIM3 Navigator | Documentation
Before version 48_0, by default the enrollment decision was performed within the medicaid module (although starting with version 34_1 it was able to override this descision with an out-of-model decision). It wasn't until version 48_3 that an effective method for determining enrollment in an alternative run was added.

The in-model decision was monthly, and the degree to which the decision for any given month of eligibility will be correlated to the decision for other months of eligibility is specified by the rule DegreeOfMonthlyCorrelation.

Eligible Reporters, Cash Recipients, and Other Eligible Non-Reporters

An eligible person's likelihood of being included in the simulated Medicaid/SCHIP caseload depends on whether or not s/he was reported to be enrolled in Medicaid or SCHIP in the CPS interview, whether or not s/he receives cash aid (SSI or AFDC/TANF), and his/her state of residence and "user group" (child, adult, disabled, or elderly).

Eligible persons who are reported to be enrolled in Medicaid or SCHIP according to the CPS interview ("eligible reporters") are almost always included in the simulated caseload for the number of months of reported enrollment. (The exception to this rule is discussed below.)

Eligible individuals who receive cash aid (SSI or AFDC/TANF) in a particular month are always enrolled in Medicaid or SCHIP in that month. This assumes that an individual who receives cash aid would be aware of and would not refuse medical assistance. Information on monthly benefits received is passed to the Medicaid module via the rules SSIBenefitsReceived and AFDCBenefitsReceived.

Additional eligible individuals are included in the simulated caseload probabilistically in order to reach administrative targets, which vary by user group (children, adults, disabled, and elderly) and by state. For each eligible person who is neither a reporter nor a cash recipient, the module obtains the appropriate probability of participation based on his/her user group and state of residence. (See the discussion of adjustment factors and administrative targets below for more details.) If the person's random number for purposes of Medicaid/SCHIP eligibility is less than the probability of participation, the person will be included in the simulated caseload.

Persons who report Medicaid coverage but who seem ineligible for the program in every month are never simulated to enroll. As discussed above, however, some reporters who initially appear ineligible are treated as eligible on the assumption that they are either pregnant or eligible as Medically Needy due to very high medical expenses.)

TRIM3 attempts to make the simulated enrollment decisions consistent among members of a family in a given month. All family members who are eligible for Medicaid and who report Medicaid are generally either enrolled or not enrolled as a group; the same is generally true for all family members who are eligible but who do not report Medicaid coverage. The way that the module accomplishes this is by using the same random number for purposes of Medicaid/SCHIP enrollment for all eligible persons in the same family. The random number that is used for all eligible people in the family is the random number for purposes of Medicaid/SCHIP eligibility that was assigned to the first eligible person in the family.

Adjustment Factors and Administrative Targets

The simulation of the enrollment decision uses adjustment factors to create a simulated "baseline" caseload each year that comes sufficiently close to administrative targets. The enrollment adjustment factors vary by user-group and by state. In addition, the "child" user-group has one set of adjustment factors for SCHIP children and another for non-SCHIP children. Lastly, there is a set of factors which can be applied (in addition to the user-group factors) based on a person's citizenship status, if targets are available by citizenship status.

The adjustment factors are specified by the following state-specific rules:

  • EnrollmentAdjustForChildren
  • EnrollmentAdjustForAdults
  • EnrollmentAdjustForDisabled
  • EnrollmentAdjustForElderly
  • EnrollmentAdjustForCHIP
  • CitizenshipAdjustmentFactors
Average monthly targets for Medicaid/SCHIP enrollment are specified by the following state-specific rules:
  • AvgMonthlyTargetsForChildren
  • AvgMonthlyTargetsForAdults
  • AvgMonthlyTargetsForDisabled
  • AvgMonthlyTargetsForElderly
  • AvgMonthlyTargetsForCHIP
  • CitizenshipTargets

Another set of state-specific rules specifies annual (or "ever-on" targets). (Note that there are no annual citizenship targets.)

  • EnrollmentTargetsForChildren
  • EnrollmentTargetsForAdults
  • EnrollmentTargetsForDisabled
  • EnrollmentTargetsForElderly
  • EnrollmentTargetsForCHIP

An example of how the adjustment process functions: Imagine that for a particular type of user-group in a particular state, the target is 100 enrollees, and (in the month being processed) there are 60 AFDC/TANF and SSI recipients, 30 other eligible reporters, and 50 other persons who are eligible but who did not report Medicaid receipt. The enrollment adjustment factor would be set to .20. The model would assign as enrollees that month all 60 AFDC/TANF and SSI recipients, all 30 other eligible reporters, and 20% (10 out of 50) of the eligible non-reporters.

Adjusting to reach both annual and average-monthly targets

While baseline adjustment factors can be set in order to emphasize closeness to either the average monthly targets or the ever-on targets, ideally the same set of adjustment factors would create a caseload reasonably close to both targets. However, this does not always happen. Instead, what often happens is that reaching annual targets results in too small of an average-monthly caseload -- which is the result of simulating the length of an average enrollee's enrollment spell to be longer than the targets indicate is actually the case (while thoretically the opposite situation could also occur, it has not happened in practice). Therefore, an additional adjustment rule (EnrollmentLagMonths) is available which allows the user to effectively shorten most simulated enrollment spells.

Cases when eligible reporters are not enrolled

As mentioned above, there are some exceptions to the general rule that eligible reporters are always enrolled. It is sometimes the case that, for a particular state and type of enrollee, the number of AFDC/TANF and SSI recipients plus the number of eligible reporters who are not AFDC/TANF or SSI recipients would exceed the targeted number of enrollees. In those cases, only as many eligible reporters are simulated to enroll as is necessary to hit the targets; and none of the other eligible persons (eligible non-reporters) are simulated to enroll. For example, assume that the target in a particular cell is again 100, but this time there are 60 AFDC/TANF and SSI recipients, 50 other eligible reporters, and 50 other persons who are eligible non-reporters. The participation adjustment factor would be set to -.20. This tells the model that 20% of the other eligible reporters (10 out of 50) can not participate. The remaining 80 percent (40 people) would be identified as enrollees along with the 60 AFDC/TANF and SSI recipients. None of the eligible non-reporters would be simulated to enroll. (Note that this is not the same as saying that only 80 percent of all eligible reporters in this cell are simulated to enroll. Many of the AFDC/TANF and SSI recipients who are enrolled with certainty are also Medicaid reporters.)

Note that this aspect of the Medicaid enrollment decision (not enrolling some eligible reporters in some cases) differs from the approaches used in other TRIM3 transfer program modules (which always allow all eligible reporters to participate even if targets are exceeded). Different choices have been made for different modules due to differences in the ways that module results have been used.

Enrollment decision in a non-baseline (i.e. alternative) run

In a non-baseline run, the enrollment decision operates in a way that ensures internal consistency with the enrollment decision that was made in the baseline simulation. If an eligible person is in a family where at least one person was eligible in the baseline, then all eligible people in that family use the same random number to determine their Medicaid/SCHIP eligibility as was used for that family in the baseline simulation. Further, if a person was eligible in the baseline, the module ensures that for each person, the enrollment adjustment factor is taken from the same user group as in the baseline simulation. (Note that changing certain program rules might change an individual's user group from the baseline to the alternative, such as from a child to an adult.)

Ad-hoc modifications to the enrollment decision

Users familiar with the use of "forms" in TRIM can take advantage of two forms in Medicaid named EnrollmentAdjustmentForm and EnrollmentDecisionOverrideForm. The first form is processed for all eligible persons for all months, and allows the user to make a final multiplicative adjustment to the enrollment adjustment factor applied to that person. Using the programming capabilities of "forms", the user can apply different final adjustments based on chracteristics of the person. The second form is also processed for all eligible persons for all months, and allows the user to completely override whatever enrollment decison was made by TRIM. The form must be coded to return a "0" if the person should not enroll, a "1" if the person should enroll, and anything else if the decision made by TRIM should remain unchanged. Note that for both forms, if no form is actually included in the run, its particular ad-hoc modification is not done.

Imputing Prior-Year Information

At certain points in the Medicaid simulation, reference needs to be made to previous months. This presents no problem as long as the previous months do not extend into the year prior to the simulation year (since TRIM only has information about the current year). In cases where prior-year information should (ideally) be used, but in actuality would not have much of an impact on the simulation, simplifying assumptions are made. Currently, the only aspect of the simulation that requires a more sophisticated imputation of prior-year information is the simulation of the "enrollment lag". In order to avoid having TRIM consider every person who is eligible in January to be in their first month of an eligibility spell (which would result in a dearth of early-month enrollments when the lag is being simulated), the model imputes the number of months in the prior year that are part of the current eligibility spell. It does this using a cummulative probability distribution specified by the rule PrevYearMonthsInEligSpellProb.
Small TRIM3 Logo TRIM3 ©2012 | Contact us | Privacy Policy | URBAN INSTITUTE | HHS/ASPE