The target hospitals for this study are the three Bethesda hospitals. In order to accomplish this work, we will first develop a system to securely manage communication between the target hospitals and a consortium of Pharmacy Benefit Managers (PBMs) who carry the prescription dispensing records of interest. Under appropriate confidentiality agreements and security protection, this system will request the medication profile for patients entering care at the target hospitals from this consortium to assist patient care. ? ? The need for information about medication is especially pressing during disasters, when patients are cared for by unfamiliar providers who do not have adequate time to spend with each patient. This study was motivated by the need to obtain a medication history during disasters, without needing much personnel time. Our hypothesis is that when a patient has medication information in the PBM consortiums databases, that information will provide a better medication history obtained from that source more complete and more precise than the corresponding information collected manually from the patient. We also hypothesize that the PBM medication history will be obtained more quickly and with less effort than the manual history. The value of such data for a population will depend upon the proportion of patients in the population who have medication information within the consortium database The PBM consortium will have no information about patients without insurance and those with insurance that is not processed by the PBM consortium. We will set a binary variable to identify patients who had no data in the PBM consortium and will collect demographic, administrative variables (arrival mode), and insurance class on all patients. Then we will model the existence of PBM consortium data on these attributes. We will use this model to predict the proportion of patients with data in the PBM based on the above mentioned PBMs and to assess policies that might eliminate this gap. ? ? The drug history collected automatically by the PBM, the one collected manually by the hospitals, the patient demographics and other patient characteristics will be linked and then de-identified. The research will be performed only on the de-identified databases. The research will only involve data that is collected for routine care purposes, and there will be no intervention. ? ? We will make most of the comparisons between the two sources of medication history (PBM consortium versus direct collection from the patient) via computer methods. Different kinds of preparatory work are needed for the two different sources of drug history information. The information from the PBMs comes with codes but those codes will have to convert to a generic clinical drug code (Rx.terms/Rx.Norm). The frequency and duration information in the PBM medication records will also be coded. The drug information taken directly from the patient (or their family) will come from the hospitals as a free text drug name and free text instructions. The free text drug names will have to be converted from text, to Rx.terms codes, via free text processing. In the hospital history we expect to see classes, .e.g. sulfa med, water pill, or blood pressure med, instead of drug names in some cases, and we will classify them as such. We will count the number of distinct entries in each of the sources. We will assess completeness based on the percent of the total number of distinct drugs per patient that are found in both sources and the ones that are found only in one source. We will assess specificity by counting the number of entries that are recorded as classes rather than drug names, and those that do not include information about route, strength and frequency respectively.? ? (This work has funding from the Navy to cover contractor and non federal salaries)
