JAMIA: Auto omits can be detected in medication lists
In the August issue of the Journal of American Medical Informatics Association, study authors Sharique Hasan, PhD, of the Graduate School of Business at Stanford University in Stanford, Calif., and George T. Duncan from Carnegie Mellon University in Pittsburgh, found that collaborative filtering can detect omissions in medication lists, which is analogous to an item that a consumer might purchase from a retailer’s product list.
Online retailers use collaborative filtering to suggest relevant products to consumers using retrospective purchasing data. Similarly, patient information in EMRs combined with artificial intelligence, can enhance medication reconciliation, the authors wrote.
Collaborative filtering can detect omissions using a machine-learned approach that evaluated its effectiveness using medication data from three long-term care centers. The authors also proposed using decision-theoretic extensions with the methodology to incorporate medical knowledge into the recommendations.
In their research, Hasan and Duncan identified the missing medication in the top 10 list using collaborative filtering about 40 percent to 50 percent of the time, and the therapeutic class of missing drugs 50 percent to 65 percent of the time.
Researchers said collaborative filtering helped reconcile medication lists, complementing existing process-driven approaches. However, one size does not fit all, and the authors suggested giving context (e.g., types of patients and drug regimens), as well as consequence (e.g., the impact of omission on outcomes).
Some ways of reconciling medication list discrepancies focus on improving organizational processes, improving responsibilities, better communication and increased access to timely and relevant patient information. A common approach uses reconciliation forms which ask patients about their medication history, allergies and other relevant health information along the care-giving process. The medical staff makes sure these forms are complete and verified. Form-based interventions improve medication list accuracy, and can reduce medication errors by 70 percent and ADEs by 80 percent, the authors reported.
Recent advances in IT have also aided the medication reconciliation effort. EMRs, prescribing systems and computerized physician order entry systems can store medication information in a structured, easily accessible format. Clinical decision support modules with preprogrammed rules alert prescribers about potential adverse interactions. However, the effectiveness of these alerts depends on the accuracy of the patient information in the reconciliation process.
In addition to traditional hospital information systems, technologies and modules have been designed for medication reconciliation. One module electronically incorporates a three-step reconciliation process. In the first step, the system compiles a list of medications prescribed and recorded by the physician in the EMR. The physician or nurse conducting the reconciliation is required to ask the patient about compliance with the drug regimen.
Next, the system asks the nurse to enter any medications that the patient is taking, but which are not recorded in the EMR, including drugs prescribed outside the current setting, herbal supplements and over-the-counter medications. This information is obtained through a patient self-report, an effort that has been prone to errors and discrepancies. After the electronic verification step, the clinician then clarifies and reconciles the list with current orders.
Finally, linking health information across organizations can improve medication reconciliation efforts. Healthcare organizations are currently focusing on connecting databases between clinical settings, pharmacies and insurers. This can improve medication reconciliation, but is still limited by the information stored in the database.
As such, current approaches to medication reconciliation focus on procedural and organizational issues that provide structure within which verification, clarification and reconciliation can effectively occur.
Collaborative filtering makes inferences or predictions about the information of other entities. With the growth of the internet and emergence of databases of user purchases, online retailers are increasingly using collaborative filtering to predict and suggest products that an individual may be interested in based on aggregated information from users with similar preferences.
The central piece to medication reconciliation is the patient's list of medications. This list is a set of entities, where each entity represents a drug. The most granular drug entity is its brand name with an associated dose and route (e.g., Tylenol Oral Tablet 325 mg).
Often, knowledge of a patient's true medication list is incomplete. Prescribers may only see a partial list of drugs because of a failure to record a previously prescribed medication, or the unintentional or intentional omission of a drug during patient self-report.