Introduction Despite much research, interventions to improve medication adherence report disappointing and inconsistent results. Tailored approaches that match interventions and patient determinants of non-adherence were seldom used in clinical trials. The presence of a multitude of theoretical frameworks and models to categorize interventions and patient determinants complicated the development of common categories shared by interventions and determinants. We retrieved potential interventions and patient determinants from published literature on medication adherence, matched them like locks and keys, and categorized them according to the Theoretical Domains Framework (TDF). Methods We identified the most relevant literature reviews on interventions and determinants in a pragmatic literature search, extracted all interventions and determinants, grouped similar concepts to umbrella terms and assigned them to TDF categories. All steps were finalized in consensus discussion between the authors. Results Sixteen articles (5 with determinants, 11 with interventions) were included for analysis. We extracted 103 interventions and 42 determinants that we divided in 26 modifiable and 16 unmodifiable determinants. All interventions and modifiable determinants were matched within 11 categories (Knowledge; Skills; Social/professional role and identity; Beliefs about capabilities; Beliefs about consequences; Intentions; Memory, Attention and decision processes; Environmental context and resources; Social influences; Emotion; and Behavioral regulation). Conclusion In published trials on medication adherence, the congruence between interventions and determinants can be assessed with matching interventions to determinants. To be successful, interventions in medication adherence should target current modifiable determinants and be tailored to the unmodifiable determinants. Modifiable and unmodifiable determinants need to be assessed at inclusion of intervention studies to identify the patients most in need of an adherence intervention. Our matched categories may be useful to develop interventions in trials that investigate the effectiveness of adherence interventions.

Generative artificial intelligence (gAI) tools and large language models (LLMs) are gaining popularity among non-specialist audiences (patients, caregivers, and the general public) as a source of plain language medical information. AI-based models have the potential to act as a convenient, customizable and easy-to-access source of information that can improve patients’ self-care and health literacy and enable greater engagement with clinicians. However, serious negative outcomes could occur if these tools fail to provide reliable, relevant and understandable medical information. Herein, we review published findings on opportunities and risks associated with such use of gAI/LLMs. We reviewed 44 articles published between January 2023 and July 2024. From the included articles, we find a focus on readability and accuracy; however, only three studies involved actual patients. Responses were reported to be reasonably accurate and sufficiently readable and detailed. The most commonly reported risks were oversimplification, over-generalization, lower accuracy in response to complex questions, and lack of transparency regarding information sources. There are ethical concerns that overreliance/unsupervised reliance on gAI/LLMs could lead to the “humanizing” of these models and pose a risk to patient health equity, inclusiveness and data privacy. For these technologies to be truly transformative, they must become more transparent, have appropriate governance and monitoring, and incorporate feedback from healthcare professionals (HCPs), patients, and other experts. Uptake of these technologies will also need education and awareness among non-specialist audiences around their optimal use as sources of plain language medical information. Plain Language Summary: More and more people are using special computer programs called artificial intelligence (AI) or large language models (LLMs) to find and get medical facts in simple words they can understand. This can help people take better care of themselves, learn about their health, and talk with their doctors. We found that AI/LLMs generally provided correct and helpful information to people. However, there could also be a risk of incorrect or unreliable information in certain situations if the question is complex. This can cause harm to people if they use this information to make their own medical decisions. Also, gAI/LLMs provide human-like responses, which make people trust them more than they should. There could be a risk that people may share their medical information with AI/LLMs, which could get into the wrong hands. To make sure these programs really help people, they need to be clear about how they work, and they must have good rules to follow and take advice from doctors and patients to improve performance. People also need to be trained on how to best use these AI tools to find easy-to-understand and reliable medical information. It is important for doctors, patients and other health workers to help make sure the AI is producing reliable and understandable medical information for patients.

Background: The description of adherence based on medication refill histories relies on the estimation of continuous medication availability (CMA) during an observation period. Thresholds to distinguish adherence from non-adherence typically refer to an aggregated value across the entire observation period, disregarding differences in adherence over time. Sliding windows to divide the observation period into smaller portions, estimating adherence for these increments, and classify individuals with similar trajectories into clusters can retain this temporal information. Optimal methods to estimate adherence trajectories to identify underlying patterns have not yet been established. This simulation study aimed to provide guidance for future studies by analyzing the effect of different longitudinal adherence estimates, sliding window parameters, and sample characteristics on the performance of a longitudinal clustering algorithm. Methods: We generated samples of 250–25,000 individuals with one of six longitudinal refill patterns over a 2-year period. We used two longitudinal CMA estimates (LCMA1 and LCMA2) and their dichotomized variants (with a threshold of 80%) to create adherence trajectories. LCMA1 assumes full adherence until the supply ends while LCMA2 assumes constant adherence between refills. We assessed scenarios with different LCMA estimates and sliding window parameters for 350 independent samples. Individual trajectories were clustered with kml, an implementation of k-means for longitudinal data in R. We compared performance between the four LCMA estimates using the adjusted Rand Index (cARI). Results: Cluster analysis with LCMA2 outperformed other estimates in overall performance, correct identification of groups, and classification accuracy, irrespective of sliding window parameters. Pairwise comparison between LCMA estimates showed a relative cARI-advantage of 0.12–0.22 (p < 0.001) for LCMA2. Sample size did not affect overall performance. Conclusion: The choice of LCMA estimate and sliding window parameters has a major impact on the performance of a clustering algorithm to identify distinct longitudinal adherence trajectories. We recommend (a) to assume constant adherence between refills, (b) to avoid dichotomization based on a threshold, and (c) to explore optimal sliding windows parameters in simulation studies or selecting shorter non-overlapping windows for the identification of different adherence patterns from medication refill data.

Background: Cystic fibrosis (CF) is a life-shortening genetic condition that usually affects several organs and involves significant treatment burden. Adherence to medication is important for successful CF management. Objective: To describe medication adherence according to age, therapeutic class, and pharmaceutical form in adults and children followed in four regional CF centers in France. Methods: We conducted a cross-sectional study with non-transplanted patients followed in two adult and two pediatric centers during 2015 who were covered by the French National Health Insurance (NHI). Sociodemographic, clinical, hospitalization, and prescription data were collected from patient medical records. Medication dispensations were extracted from the regional French NHI database. Adherence was calculated over 12 months using continuous medication availability (CMA) accounting for dose adjustments and hospitalizations. Drug-specific CMA was computed in R with the AdhereR package for each medication prescribed more than 3 months, which was averaged to obtain a composite CMA score (cCMA) for all treatments and per therapeutic class as well as pharmaceutical form for each patient. Results: A total of 228 patients were included. The number of chronic medications increased with age (r=0.50, p<0.001): a median of 7 medications per patient were prescribed. The mean±SD cCMA was significantly different between age groups (p=0.0098): it was 0.71±0.20 for the 0–5 years age group, 0.73±0.16 for 6–11 years, 0.64±0.17 for 12–17 years, 0.57±0.23 for 18–25 years, and 0.65±0.20 for the over 25 years age group. cCMA varied significantly according to pharmaceutical forms: the mean±SD cCMA was 0.70±0.21 for oral medications and 0.54±0.28 for inhaled medications (p<0.001). Conclusion: This study suggests that adherence to medication regimens in CF patients remains suboptimal and varies substantially between age groups and pharmaceutical forms. These variations in adherence should be considered when developing effective strategies to improve adherence.