BACKGROUND: The Centers for Disease Control and Prevention (CDC) provides an annually updated conversion reference table which lists, by National Drug Code (NDC), morphine milligram equivalents (MME) conversion factors and drug classifications for certain drugs. The Tennessee Controlled Substances Monitoring Database (CSMD) contains NDCs for every prescription, but relies upon the CDC table for MME conversion factors and drug classifications, which are necessary for prescription drug overdose analyses. NDCs may exist in the CSMD but not in the CDC table. To determine how this may affect potential for misinformed analyses in pharmacoepidemiology studies, our objectives were to quantify the number of NDCs that could not be classified and determine the impact of incomplete classification overall and in an ongoing pharmacoepidemiology study of Neonatal Abstinence Syndrome (NAS).
METHODS: We utilized TSQL queries on a SQL Server 2016 database containing CSMD information and the CDC table, limiting CSMD information to prescriptions since 2012 when reporting became mandatory. First, we determined number of NDCs 1) prescribed and 2) matched to the CDC table from the result of a left join from the CSMD NDC table to the CDC table on NDC. Next, we determined the percentage of all prescriptions which could be matched to the CDC table by adding an inner join from the CSMD prescription table to the result on CSMD NDC. Lastly, we filtered to only prescriptions used in the NAS study to calculate each of the variables above.
RESULTS: Since 2012 (n= 91,314,324 prescriptions), 13,213 unique NDCs were prescribed. 4,242 NDCs were in the CDC table and 8,971 NDCs were not, respectively representing 86.5% and 13.5% of prescriptions. When limited to the NAS study (n=96,640 prescriptions), 938 unique NDCs were prescribed. 778 NDCs were in the CDC table and 160 NDCs were not, respectively representing 97.2% and 2.8% of prescriptions.
CONCLUSIONS: Generally, the potential effect of missing NDCs, which leads to unknown MME calculations and drug classification, is greatly reduced as a result of the NDCs in the CDC table accounting for a majority of prescriptions in CSMD. However, for specific studies, the potential is determined by the proportion of prescriptions whose NDCs are represented in the CDC table. Therefore, prior to conducting pharmacoepidemiology studies using Prescription Drug Monitoring Programs, quantification of prescriptions with missing NDC can help to determine the potential for misinformed analyses.