Key Steps to Developing Personalized Adherence Communications

Co-authored by: Rick Watts and Debbie Fields

Introduction

Personalization is tailoring communications such that it evokes positive experiences and resonates with an individual based on their personal preferences and circumstances. It can be powerful in influencing human behavior. Personalized medication adherence strategies such as text messaging interventions have demonstrated success and improved patient adherence to prescription therapy. Advancements in computational technologies and availability of diverse datasets could enable organizations to deliver personalized patient communications through a variety of delivery channels at scale and positively impact medication adherence. In this article we outline key process steps that healthcare organizations can execute to deliver personalized communications to the patient population they service.

Step One: Data Enrichment

Foundational Data

Access to rich and high-quality patient medication history that is uniquely attributable to individual patients across the population at scope is the bare minimum requirement and the foundation. To ensure available and compiled data is actionable, organizations must also establish and execute a robust data management program.

Data Enrichment by Extension

No single organization has access to all necessary data required to create and inform a 360-degree patient view. Healthcare organizations must study their own data to identify gaps and procure the data required from external organizations or federal and state agencies to inform personalization. Integrating diverse data sources through robust data management and interoperability systems is crucial to inform a comprehensive 360-degree patient view. The key data sources contributing to this holistic view of a patient include: electronic health records, practice management data, pharmacy data, patient portals, mobile apps, claims and billing data, health information exchanges, genomic data, social determinants of health (#SDoH) data, and remote monitoring and telehealth data.

Data Enrichment by Inquiry

In addition to extending data by reference, valuable patient insights can be gleaned through inquiry methods such as survey instruments to compile psychographic and behavioral patient data. Additional inquiry tools can include: interviews, observational studies, website analytics, social media, and purchased services.

Step Two: Personas & Theme Development

Patient personas are representations of groups of patients within the population that share common attributes, needs, and behavioral characteristics such as motivations, attitudes, and pain points. Themes are the engagement modes and methods of interaction most conducive to a particular persona (e.g., attitude, phrasing, propositions, and delivery channel preferences). Using dominant predictors of non-adherence, expert staff—such as data scientists, psychologists and other clinicians—create the patient personas and align them to corresponding themes for use in training Machine Learning and Large Language Models.

Personas together with themes serve as analytical instruments to identify patient cohorts within a population and also can dictate the engagement strategies specific to each identified cohort.

Step Three: Computational Analysis

Machine Learning Models

#MachineLearning models are adept at analyzing large datasets to unravel hidden data patterns, trends, and correlations. Techniques such as supervised learning, unsupervised learning, and deep learning can empower organizations and help them glean meaningful and actionable insights. Machine learning classification or clustering algorithms can segment a population at scale based on developed personas and also inform creation of new and distinct patient personas.

Large Language Models (LLMs)

Large language models can be trained using identified themes and sub-themes to generate delivery channel specific messaging content. LLMs can be continuously trained using reinforcement learning (RL) and real-time patient feedback to optimize output.

Implementation of artificial intelligence tools can effectively deliver personalized patient experience to positively impact patient outcomes at scale.

Step Four: Personalized Communications

Personalized messages are perceived by patients to be specifically relevant to them. Additionally, tailoring the delivery timing and channel to their lifestyle and communication preferences makes patients feel catered to and better understood. Altogether, personalization creates a statistically higher chance the patient will read the content, act on it, and remain engaged.

Conclusion

High-quality baseline data that is further enriched through inquiry and procurement of additional datasets from external sources is fundamental. A dedicated interdisciplinary team consisting of data scientists and clinicians is essential to develop and implement medication adherence strategies. Advancement in artificial intelligence techniques can deliver personalized adherence messaging to affect positive adherence outcomes in a large population.

Given the advancement in computational technologies coupled with relative availability of pertinent diverse patient datasets, the vision of delivering artificial intelligence enabled medication interventions to a population at scale is no longer futuristic. However, foundational building blocks such as data enrichment and data management activities must be put in place and actively nurtured to help us realize this vision.

Let us know your thoughts and perspectives on what needs to be done to further move the patient adherence needle through harnessing the power of artificial intelligence technologies.