STATISTICAL PRESCRIPTION REVIEW
To execute the process of statistical medication prescription review, one may initially map the most fundamental attributes of the prescription under review to a set of predefined classes related to the prescription’s contents. The classes to which the prescription is mapped or assigned include classes for the medication itself (e.g., its various therapeutic classes) and also for the specific drug product to be dispensed (e.g., brand, generic, new to market, mainstay) and the dosage form of the product (e.g., tablet, capsule, injectable). Other prescription details to map and categorize in this same manner include the prescribed dosage (e.g., low, normal, high), the prescribed frequency (e.g., typical, atypical), duration of use (e.g., short-term, long-term), and instructions for use or SIG (e.g., complete, simple, complex, easy to read, hard to read).
One may then continue systematically combining previously collected data elements about the patient and assigning the medication prescription – and the patient for whom it is meant – to other classes using a host of knowledge, facts, and data. This further classification is based on general medication and drug product knowledge in the form of rules and relations, relevant facts about the demographics, current organ function, genetic makeup, medical history, medication-use history, and present medical problems of the patient, along with facts or inferences about the indication and therapeutic goals for the present medication prescription. These items are potentially augmented with objective data about disease, the functional and physiological status of the patient, and by lay and expert judgments gained from the patient and prescriber’s prior use of reliable and valid measurement instruments and questionnaires.
Summarizing, the combination of available medication and drug product knowledge with the contents of the medication prescription and other patient facts and data allows us to classify the medication prescription in many different dimensions.
Once the medication prescription is multiply classified, a variety of corresponding statistical models can be engaged to estimate statistical frequencies and magnitudes of harmful and beneficial effects in each of the relevant medication prescription classes and their combinations.
Finally, these frequencies and their magnitudes can be interpreted, singly or together, using empirically established threshold values for judging both safety and potential efficacy.
To reiterate, the mechanical combining of data elements for classification purposes, and the resulting predictive probabilities, which derive from many empirical statistical relationships or formulae established during prior research, followed by the interpretation of such probabilities alone or together in light of evidence-based numeric threshold values, are the moves that define statistical medication prescription review.
In essence, statistical medication prescription review systematically and algorithmically enables a wide variety of previously studied and relevant features to inform the review of new medication prescriptions in a repeatable and automatic way.
COMPUTERIZED SCORING OF PRESCRIPTIONS
Along with other analytics, statistical medication prescription review involves using data and algorithmic models to compute a variety of scores for individual prescriptions. These scores can provide new prescription-level features for prescription complexity, typicality or atypicality, risk or general concern, and other things.
A reliable and valid medication prescription complexity scoring approach uses the Medication Regimen Complexity Index (MRCI).
Approaches for scoring individual medication prescription typicality or atypicality include applying analytics of past medication prescribing patterns to new medication prescriptions.
Approaches to scoring risk or concern for individual medication prescriptions still need a lot of work. At least one attempt has been made to use the expert judgment of pharmacists to start creating a useful medication-level risk scale. Another approach with some promise is to document the number, types, and criticality of biomarkers that may undergo drug-induced changes subsequent to medication prescriptions. A deep understanding of the degree to which any given prescription could precipitate concerning biomarker changes might result in a useful medication prescription level scale for additional computerized scoring of medication prescriptions.