Missing Prescribed Date in 2019 TAF Pharmacy File

Background

The TAF Pharmacy files (TAF RX) are a CMS Research Identifiable File (RIF) that contains pharmacy prescription fill and over the counter (OTC) drug information covered by Medicaid. They are sourced from the T-MSIS, which is a repository of service utilization and reimbursement records from state Medicaid agencies. This file contains multiple record types including fee-for-service, when the state reimburses providers for services they provide, and managed care, when the state works with a health plan to administer Medicaid services. It serves as a useful research file for researchers interested in Medicaid prescriptions.

Prescribed Date is a variable that describes the date that a drug, device, or other supply was prescribed to a Medicaid beneficiary by a provider (1). It can be useful in conjunction with other variables, such as Prescription Fill Date, to understand delays in care. This variable must have a valid calendar date on every TAF RX record, so it would be unusual to have missing values(1). This article quantifies the missingness of Prescribed Date in 2019 TAF RX data at a national and state level and stratifies by various record and beneficiary characteristics.

Methodology

This article uses data from the 2019 100% TAF Pharmacy File with no modifications or cleaning applied, including the retention of managed care and non-prescription service tracking records. These records tend to have a reduced amount of information available compared to fee-for-service records. However, they were retained to inform the overall shape of missingness and for additional stratification.

This article reports counts at the line level as prescribed date is reported at the line level. Any reference to a “record” is at the line level.

Missingness is defined as having a null date value. Enrollment in managed care plan was determined using TAF RX, not the TAF Demographic and Eligibility file. There is a variable that indicates if a record is associated with managed care.

To comply with CMS’s cell suppression policy, some table values have been suppressed or altered. Any values between 1 and 10 have been suppressed and can be identified as “<11”. The total missing prescribed date records exclude counts from Illinois, the only state to have counts less than 11. This suppression is not expected to meaningfully affect overall patterns described in this article.

Results

Table 1 describes the total count of Prescribed Date missingness and the total count of all line records. Approximately 7% of all line items contain a missing prescribed date.

Table 1. 2019 TAF RX Line Record Counts

Table 1 caption: Total number of missing prescribed date records compared to all TAF RX records. The starred value is the sum of all states that have counts equal to or greater than eleven.

Table 2 breaks down the total missing prescribed date records by state. Missingness is associated with twenty-four states at varying degrees. Rhode Island and South Carolina are missing prescribed dates in all RX records. The remaining states have varying degrees of missingness in comparison to their total RX records.

Table 2. Missing Prescribed Date Record Counts by State, 2019

Table 2 caption: The total number of records with missing prescribed date by state. The denominator is the state’s TAF RX record count in 2019. The final row (“Total”) replicates the information in Table 1.

Table 3 further stratifies the information in Table 2 by managed care enrollment. Around 60% of all missing prescribed date records are associated with managed care enrollment. States are not all impacted equally. Eleven states have approximately 95% or more of their missing prescribed date records associated with managed care enrollment. Nine states have less than 1% related to enrollment in a managed care plan. The remaining four states are somewhere in the middle.

Table 3. Missing Prescribed Date Record Counts by State & Managed Care Enrollment, 2019 

Table 3 caption: This table describes the total number of missing prescribed date by state and managed care enrollment. The denominator is the total number of missing prescribed date by state.

Additional Characteristics of Records with Missing Prescribed Date and NOT associated with a Managed Care Plan

Among missing prescribed date records that were not associated with a managed care plan, Illinois and Arkansas were disproportionally associated with service tracking records. All of Illinois’s records were associated with a service tracking record. Approximately 10% of Arkansas’s records were associated with a service tracking record.

For these records, we also reviewed the following characteristics and did not find any correlation with missing prescribed dates: dual eligibility, compound drug status, age groups, and eligibility groups. Values cannot be reported due to small cell sizes.

Discussion

We anticipate that almost all records in the TAF RX require a prescription, including instances where OTCs are present(2). This, coupled with the variable being a T-MSIS requirement(3), makes it unusual that around 7% of all 2019 TAF RX records are missing a prescribed date.

We subset missing prescribed date records by state, as suggested when working with Medicaid data(4), and found that this concern occurs in 24 states. Thus, this is a state-specific concern and state-level differences should be considered when using this variable.

These records were further subset by managed care enrollment and service tracking claims. Reporting requirements for managed care plans may differ by plan and state(5). Prescribed Date could be one of these reporting differences as eleven states (45%) have missing date records correlated with managed care enrollment. Further analysis by plan could provide more insight into missingness, but researchers may have issues with small cell sizes.

Service tracking records represent lump-sum payments, which could include drug rebates(6). We anticipate that these types of records have mostly missing data. All of Illinois’s missing prescribed dates were associated with a service tracking record. However, due to the complexity of Illinois Medicaid data, it is not indicative that service tracking records are the sole reason for missing prescribed date.

We did not identify other potential correlating reasons for missingness among some common beneficiary or drug characteristics. Additional work focused on state-specific policies could provide some insight into prescribed date missingness.

Limitations

This article has multiple limitations. First, this study only looks at one year of data. It does not capture temporal changes in prescribed date reporting. These findings do not describe a persistent pattern. Second, although multiple beneficiary and claim characteristics were reviewed, the analysis does not offer an exhaustive list of potential missingness correlations. There can be other factors, such as state-specific policies, which could speak to missingness. Third, the interpretation of unaltered TAF Illinois records should be approached with caution. Illinois records should not be considered complete claims(5). Additionally, this article does not address what service tracking records look like for TAF RX data. Finally, as the data became further segmented, produced results were not aligned with CMS’s cell suppression policy and could not be reported.

Conclusions

It can be difficult to identify the root cause of anomalies in administrative claims data. For missing prescribed date records, the findings suggest that the missingness is state-specific and, for some states, correlated with managed care enrollment. However, checking for common administrative or eligibility statuses did not yield any additional information for remaining states. Researchers are encouraged to review prescribed date missingness at a state-level, including supplementing with sensitivity analyses and looking for potential state-specific policies or recommendations.