Patient-reported outcome measures (PROMs) are instruments that can capture patients’ health status directly through digital administration [1-3]. Patients vary in their skills, knowledge, preferences, and motivations to complete PROMs, as well as their ability to interpret and use the resulting patient-reported outcomes (PROs) for health management. However, PROMs and PROs are not designed to accommodate these differences, often excluding certain populations [4]. This includes older adults, nonnative speakers, individuals with poor health, those lacking social support, people in less privileged socioeconomic positions, or those with low health literacy—defined as “the degree to which individuals can obtain, process, and understand the basic health information and services they need to make appropriate health decisions” [4-6].

Communication is crucial for achieving the clinical benefits of PROMs, for instance when health care professionals (HCPs) discuss PROs with patients. However, the communication processes involved in using PROMs in clinical practice are underexplored [7-9]. We therefore argue that improving communication processes can enhance PROMs implementation. Specifically, personalized communication (information that is customized to the individual) is essential to address the diverse information needs, preferences, and capacities of a broad population [10]. However, personalizing information manually is labor-intensive and costly, which may explain its limited application to PROMs.

To address this, we propose leveraging existing digital technologies to streamline personalized communication for PROMs. Tools such as data-to-text generation, multimodal communication, and conversational agents can offer innovative solutions to improve communication and, in turn, support the broader implementation of PROMs in clinical practice. This paper begins by summarizing communication processes for PROMs in clinical practice and identifying their shortcomings. These shortcomings highlight the limitations of a one-size-fits-all approach, which fails to meet the needs of a diverse patient population. We propose solutions rooted in personalized communication and suggest several digital technologies to support the implementation of these strategies in clinical practice.

From many available frameworks, we use the framework of Lasswell [11] to outline key aspects of communication for PROMs in clinical practice: who communicates what, in what form, to whom, and to what effect (Figure 1). We distinguish 4 communication processes, that are PROMs invitation, PROMs completion, PRO presentation, and PRO-based health management.

First, HCPs or administrative staff (“who”) invite patients (“to whom”) verbally or by email (“in what form”), including the goal and practical guidance on how to complete PROMs (“what”) [11,12]. This aims to engage patients to complete PROMs (“to what effect”).

Second, PROMs (“what”) are administered through online portals (“in what form”) for patients (“to whom”) to complete (“to what effect”).

Third, HCPs (“who”) should consistently discuss PROs (“what”) with patients (“to whom”) during consultations (“in what form”) [12,13], providing insights into their health and health-related quality of life (HRQoL; “to what effect”) [12]. Alternatively, patients can occasionally access their PRO reports directly via electronic portals (“in what form”), bypassing HCPs.

Fourth, HCPs (“who”) can translate the PROs into actionable health management information (“what”) to aid patients’ (“to whom”) health management (“to what effect”). Similarly, patients may independently access this information (“what”) through electronic portals (“in what form”), bypassing HCPs.

The shortcomings of each communication process are detailed below, along with solutions based on personalized communication and digital technologies to support implementation. Table 1 summarizes the identified shortcomings, proposed solutions, and relevant digital technologies.

With increasing pressure on health care, digital strategies for patient self-management are expected to play a larger role [50,51]. Paradoxically, those who would benefit the most from enhanced self-management support (such as vulnerable populations) are often the ones facing barriers to participation. For instance, lower health literacy is associated with worse health outcomes and healthcare service utilization [52], while lower digital health literacy correlates with lower overall survival rates in cancer patients [53]. While PROMs have the potential to improve symptom management, HRQoL, patient-HCP communication, and patient satisfaction [7,12,13,54,55], vulnerable patients may currently lack access to these benefits. By implementing personalized communication strategies and leveraging digital technologies, we believe PROMs can become more accessible and easier to complete, thereby increasing completion rates. Furthermore, making PROs easier to interpret can improve their usability and value for patients, ensuring that PROMs can more effectively support patient care, particularly for those currently left out.

Despite promising advances, the use of the digital applications described here is not without risks. Data-to-text algorithms, particularly those based on data-driven approaches, present certain challenges. Traditionally, data-to-text systems relied on hand-crafted rules, which were difficult to scale. However, recent advances in machine learning, driven by increased computing power, have streamlined this process, enabling scalability and efficiency [18,56]. Large language models, such as exemplified by ChatGPT, are arguably the culmination of this approach. While these models generate fluent text, they are not always accurate and may contain harmful biases. Large language models rely on massive volumes of human-authored texts, typically sourced from the internet, raising concerns about the exclusion of underrepresented populations (eg, those with a small presence on the internet) and the inclusion of biased or toxic language [57,58]. Therefore, rule-based NLG might be preferable for converting PROs into text, offering more control and reliability. Similarly, while advances in machine learning have enhanced conversational agents with increased dialogue complexity, they introduce risks in natural language understanding, response generation, and patient interpretation, necessitating rigorous monitoring and validation [59]. The safety of conversational agents for patients remains inadequately evaluated [32]. We emphasize that these digital applications require further exploration and testing, despite their clear promise.

In addition to personalized communication, patient-centered communication could improve the use of PROMs in practice [21]. Currently, patients face issues with wording, length of questionnaires, and formatting of PROMs, which can hinder their understanding of questions and answer choices [15,34,39]. Over half of the commonly used cancer-related PROMs to not follow best-practices for plain language use [60]. To improve completion rates, existing PROMs should be revised to use simple, concrete language, avoiding jargon and negations, and eliminate redundant or distracting information [15,61]. Instructions should also be simplified [61-63] and structured with clear orientation statements (eg, “First I will explain the goals of PROMs, then I will show you what the questions are about”) to aid understanding [61,62]. Resources like the “health literacy universal precautions toolkit” can help assess readability and convert medical terms into plain language [63].

To reduce patient burden, PROM should be relevant for individuals [21]. Computer adaptive testing can ensure follow-up questions are tailored to each patient, rather than asking the same set for everyone. The formatting of PROMs should prioritize clarity, with large fonts, clear headings, and ample white space [15,61]. Patient-preferred modifications include removing or altering items, using visuals, simplifying language, and adjusting layouts for better comprehension [29]. Using user-centered design can help meet the diverse needs of patients, considering factors such as clinical conditions, culture, languages, and literacy levels [4]. This approach should also explicitly involve individuals with low literacy skills or learning disabilities, as they are often excluded from PROMs development [64]. By doing so, we can ensure that PROMs are accessible to a wide range of patients, including those with disabilities [20].

Training HCPs about the goals, tasks, and responsibilities related to PROMs is essential for motivating their engagement with PROMs [65,66]. In general, HCPs should discuss or, at a minimum, briefly mention PROs with patients, even when no new clinical information arises from the data [13]. HCP engagement plays a significant role in patients’ motivation to complete PROMs [67]; failing to follow up on PROMs may negatively affect patients’ HRQoL [68]. Further research is needed to assist HCPs in interpreting PROs and converting them into actionable steps for health management [3,13]. That said, some patients may avoid engaging with detailed PRO data [36], highlighting the need for tailoring information to meet patients’ individual preferences for how they receive and process information. Although digital technologies can support communication around PROMs in clinical practice, they may not be able, now or ever, to replace the HCP in addressing patients’ HRQoL.

Implementing the proposed technologies in clinical practice might present some practical hurdles. For instance, many of the identified applications have not been seamlessly integrated into electronic health records (EHRs) [47,69]. Future health care systems will require EHRs that can safely incorporate digital applications. However, models could be trained without sharing data from EHRs, using techniques like federated learning [70]. Furthermore, trust, which is crucial for patient adherence [71], is challenged by the perceived “black box” nature of AI models. Therefore, it is essential to have human oversight of AI outputs and to maintain transparency about data sources and methods [59]. Both HCPs and patients should become more AI-literate, and have a basic understanding of the capabilities and limitations of AI models [70]. To ensure that readers can assess the credibility of generated texts, it is essential to explain that texts come from a computer model, and not a fellow human [57].

The successful implementation of PROMs in clinical practice involves complex interactions between care processes, technological development, and human behavior [66]. Often, PROMs implementation is led by HCPs or researchers, potentially resulting in a lack of recognition of these complex interactions. This paper was written by a multidisciplinary team with expertise in PROMs, implementation science, health communication, and digital technologies. We have provided a comprehensive overview of key communication processes around PROMs. Future work could deepen this assessment by studying the effects of context and noise in communication, which are aspects currently missing from the application of the Lasswell model [11]. With ongoing advancements in AI, which may help mitigate the risks associated with the proposed digital applications, we envision a future where personalized communication strategies and digital tools can overcome current implementation challenges, ensuring that PROMs are applied inclusively to benefit all patients.