The Autonomous Pharmacy: Applying AI and ML to Medication Management Across the Care Continuum
Ken applies artificial intelligence (AI) and machine learning (ML) solutions to medication management.
Every year, almost half a trillion dollars are spent on medications in the United States. This includes prescription drugs (and its supply chain), pharmacy benefit managers (PBMs), pharmacies, distributors, and providers dispensing and administering the medications. One out of every seven dollars we spend on healthcare is related to medications.
However, about 20-30 percent of prescriptions are not even filled. Many patients are not retrieving or using their pharmacy medications. The overall cost of medication non-adherence is $300 billion, according to the New England Healthcare Institute.
A million emergency department (ED) visits a year occur from adverse drug effects in outpatient settings. Of those, about 1 out of 8 result in hospitalization.
How the healthcare industry is directly affected:
Legislation has been passed to regulate and reform our healthcare system. As a result, we’ve created a burden — a 3,000 percent growth in administrators. The growth of physicians has remained relatively flat, however. It costs the U.S. $15 billion a year on quality reporting.
This weighs down not just on physicians and administrators, but also on pharmacists. 76 percent of pharmacists’ time is spent time on administrative, manual tasks, according to the American Society of Healthcare Pharmacists (ASHB) Full disclosure: Ken sits on the board.
Proposed solution to disrupted, wasted healthcare: artificial intelligence and automation.
The autonomous Pharmacy Advisory Board looks to transform the pharmacy care delivery model through the use of technology. This includes automation, data analytics, robotics, artificial intelligence and machine learning.
The result: a fully autonomous pharmacy.
How it’s done: by applying automation and robotics to the healthcare system. The ASHB is gaining intelligence so that pharmacy processes are performed more efficiently, safely, with better clinical outcomes. We can also better comply with regulations, and ultimately improve population health.
Three use cases that can benefit from AI and ML:
Use case #1: drug shortages
We currently have over 100 drug shortages in the United States (a bipartisan bill addressing these shortages has currently been submitted to the Senate).
Although the common definition of a shortage is when demand exceeds supply, the ASHP defines it as when a drug manufacturer can attest that there is an actual problem with supply and, as a result, is not able to meet the demand.
Impact of drug shortages:
- Hard dollar and soft dollar costs. A hard dollar cost is buying more expensive substitutes. Brand names cost more than generics. Soft dollar costs include solutions research, and involve a good deal of coding, analysis and planning.
- Risk of patient safety. Medication errors can cause confusion and delays.
Solutions using ML:
- Proactive identification of risk, pre-shortage. Identify and apply risk scores on drugs that are based on multiple factors.
- Confronting the shortage when it actually occurs. Using algorithms to identify the therapeutic alternatives. This could all be automated; caregivers can be notified.
Use case #2: drug diversion
This is a polite term for theft, when a drug is diverted away from its intended use and from the patients who need them.
According to a Protenus 2019 Drug Diversion Report:
- 47.2 million doses were diverted in 2018, up 126 percent vs. 2017
- $454 million worth of drugs lost
- 94 percent of diversion incidents involved opioids. In fact, diversion is a major cause of the opioid crisis.
Who is likely diverting?
- 15 percent of pharmacists
- 10 percent of nurses
- 8 percent of physicians
Use AI and ML for diversion monitoring and detection:
- More comprehensive data matching makes it more difficult for diverters to avoid detection.
- Actionable insights to identify risk and a clear path to reconcile discrepancies.
- Minimize false positives. Not intended to create a Big Brother environment, but to draw an unbiased, non-judgemental conclusion that is based on data.
Use case #3: Medication adherence
- Identify patients who are at high-risk of non-adherence (the ones most likely not to pick up or use their prescriptions).
- Identify patients who may cooperate with interventions. Provider organizations and retail pharmacies can optimize their interventions by using prescriptive, rank-ordered, preferred, and recommended intervention lists.
Longitudinal and large data sets can evaluate whether these algorithms and models are effective and successful. The goal is to measure the outcomes over time for large populations.
The bottom line: applying AI and ML may solve significant issues concerning medication management and use.
- Overarching vision and ultimate goal: the autonomous pharmacy
- Real work is currently being done
- The ultimate test of value and the path to refinement: outcomes. We can actually test whether these outcomes are effective or not.