AI Applications in Medication-Therapy-Management

Josh - Oct 29 '21 - - Dev Community

***Hello everyone. This post is a bit different from what I normally write on dev.to. Please visit my Medium blog where I will start posting more long form discussions such as this one if you find it interesting. Thank you!

In today’s world, pharmaceuticals help manage a wider range of health concerns than ever before. Prescription medications are combined with lifestyle changes to create a care plan to improve patient’s health. However, for many people it can be difficult to connect with the right medical professionals to optimize or receive care. There are programs in place, often required by state funded healthcare entities, that have insurers provide services by pharmacists to members to review and manage their medications. The benefits of these programs can be staggering. Patients that take many medications often have difficulty recalling all of their medications or relaying concerns. The goal of a pharmacist here is to establish a complete medication reconciliation and then optimize each component. Some medications are best used at a certain time of day, cause an interaction with vitamins or supplements, or are not acting as effective as hoped by the physician. By regularly meeting with a pharmacist the patient can get a form of care that often is lacking in a fast paced provider appointment.

The benefits of these programs are pretty apparent when fully utilized. It makes you wonder why this is not available to everyone? Unfortunately healthcare is a very specialized role that requires many years of studying and certification. Healthcare professionals are also expensive to employ and it is not cost effective to offer this service to everyone. This is where AI can help. Most of the users that would benefit from these programs don’t realize that they are available to them. Insurers and health providers have a large amount of patient data that can be utilized effectively. AI can be used to identify patients and points of concern. Over time the models can be enhanced to determine who has the greatest need and allow staff to outreach to this population to see if they would be willing to participate.

To breakdown this system we will first look at who would be eligible for the program. The system would need to utilize machine learning techniques to take large amounts of data and create models on who would benefit from the program the most with the restraints of providers available. For those not familiar, machine learning is a subsection of AI programming that uses modeling techniques to train a program to analyze and make decisions through data. The system could take into account pharmacists who have specializations in fields such as Hepatitis C or Oncology and allocate them for those patients. The models would need to include factors such as comorbidity, medication volume, and total providers currently in their care. Once patients have been identified and scheduled the system can further assist pharmacists by identify talking points that may need addressed. Pre-appointment forms, insurance claim history, and EMRs can provide data on: which providers prescribe specific medication, medication dosing/strength, unaddressed health concerns, etc. Machine learning techniques can tap into drug databases such as Lexicomp to find sub optimal dosing or interactions to discuss in the appointment. Time of day, number of tablets, and cost are inconveniences that can deter patients from being compliant with their medication regimen which may be optimized with Pharmacist intervention.

With any AI application or private health information (PHI) it is important to consider the ethical ramifications of the product and the potential for harm. Starting with the selection process, the AI would be largely in charge of deciding who is to receive services and who will not. Safeguards will need to be in place to try and avoid discrimination. Cultural or systemic limitations may not allow people to submit necessary or accurate information that would trigger an acceptance by the AI. Once patients are in the program there can be additional barriers to understanding and accepting help from a pharmacist. Patients that may speak another language would require an interpreter or provider that can speak the language. Additional cultural components may prevent a patient from accepting care if not presented in a way appropriate for them. Lastly, with any healthcare entity there is always the fear of releasing PHI or violating HIPAA. Measures will need to be taken to ensure that access to PHI is limited only authorized personnel that have a need for it in order to provide services.

As you can see, AI in healthcare has a huge potential to automating processes that can increase access to care. Increasing efficiency can relieve some of the burden that is plaguing the system and decrease overall costs. As with anything there are possible pitfalls that will need to be identified and addressed but this is to be expected with any new system. The possibilities here are exciting and can usher in a new paradigm of healthcare for future generations.

. . . . . . . . . . . . . . . . . . . . .