Health care has expanded from the bedside to the desktop in the form of secure messaging via patient portals. At Kaiser Permanente, for instance, message exchanges increased 10-fold over a decade, comprising 25% of all primary care services.^1,2 Concurrently, the Centers for Medicare & Medicaid Services have adapted billing options for messaging, provided they involve medical decision-making and 5 minutes of clinician time. This has spurred investments in patient messaging services, including using team-based approaches or artificial intelligence (AI) chatbots to draft messages. However, these efforts primarily focus on reducing clinicians鈥� burdens. While addressing clinician workload is vital, opportunities also exist to improve the quality of messaging for patients. We propose a messaging framework that defines quality standards to inform research that could enhance patient-clinician relationships and patient outcomes called CREATE TRUST (Box):

• Correct鈥攑rovide accurate information;

• Referenced鈥攄irect the patient toward quality sources;

• Empathic鈥攅xpress understanding and concern;

• Authentic鈥攁void sounding prewritten or robotic;

• Thorough鈥攁ddress all questions;

• Engaging鈥攄raw the patient into a narrative;

• Tailored鈥攗se clinical history to inform the response;

• Respectful鈥攈onor the patient鈥檚 requests;

• Understandable鈥攁lign with patient鈥檚 health literacy;

• Safe鈥攁ffirm that the patient understands risks; and

• Timely鈥攔espond while the answer is salient.

Substantial investment in training, research, and professional standards have underpinned high-quality in-person clinician-patient communication. Yet, the quality of care provided through messaging is largely unknown. Health systems lack clear quality standards for messaging, such as response times and response rates. Lacking quality standards, some regulations such as mandates for immediate access to test results can confuse patients, increasing the volume of messages.³ Research predominantly adopts a clinician-centered perspective, highlighting messaging as a contributor to burnout. Interventions, as a result, aim at aiding clinicians rather than patients (eg, exploring if billing patients for messages would reduce message volume).⁴

In contrast, the patient鈥檚 perspective has been inadequately addressed by research, health systems, and regulations. Nearly all research considering patients focuses on inequities in access to online communication portals,⁵ rather than the substance of communication. Only 1 study has evaluated the quality and empathy of clinicians鈥� messaging but used generic measures and data from an online question-and-answer service.⁶ Poor responses can harm patients by creating misunderstandings or care delays. Some patients may turn to unreliable messaging platforms, including lay social media, known as crowd-diagnosis, partly because 79% of questions receive responses within 24 hours, often within minutes.⁷

There is a pressing need for a patient-centered framework to guide the development of evidence-based messaging. A theoretical framework can spur research to inform message standards, aid the evaluation of newer messaging tools, help advocate for appropriate clinician support, and influence regulation. The CREATE TRUST framework synthesizes decades of research on communication, concentrating on the clinical substance and style of messaging from the patient鈥檚 perspective. The goal is to simplify the complexity of measuring message quality into a collection of attributes that are discrete and easy to operationalize and evaluate.

Consider a patient inquiring about a recent change in symptoms who asks, 鈥淚s this normal?鈥� While a response of 鈥渃ompletely normal鈥� may be correct, it is insufficient. This response leaves a patient less educated about their health and possibly more likely to schedule unnecessary appointments. Alternatively, we support a higher quality response that incorporates multiple attributes of the CREATE TRUST framework. For patients with multiple chronic conditions, messages are likely more complex. In such instances, being thorough, tailored, understandable, and safe is crucial. Furthermore, given the prevalence of misinformation, it is imperative that these responses are resourced to direct patients toward reliable information.

Researchers can leverage the CREATE TRUST framework to gain insights into the dynamics of communication and inform their study designs. Possible avenues include evaluating the presence of CREATE TRUST attributes, exploring the interrelationships among attributes, and establishing which attributes are predictive of patient outcomes (including satisfaction, treatment adherence, health care utilization patterns, or disease progression). Researchers can also use CREATE TRUST to study patients鈥� perspectives and understand the benefits and pitfalls of each attribute.

Consider AI question-and-answer systems like ChatGPT (OpenAI). They typically use basic response preference via A/B testing, in which users choose between 2 response options to fine-tune models. However, A/B testing lacks a robust theoretical foundation and ignores specific attributes that contribute to preference. In contrast, using the CREATE TRUST framework, developers can evaluate responses across multiple dimensions, ensuring AI systems are rooted in theory.

Health systems and regulators can use CREATE TRUST to monitor communication quality, including across health systems, patient demographics, and clinician types (eg, physicians and support staff). Human annotations of CREATE TRUST attributes can inform the development of machine learning classifiers to automatically score responses en masse, including summary and attribute-specific scores. This would enable real-time monitoring and data-driven decision-making, such as informing health systems鈥� decisions when procuring message support systems. Furthermore, CREATE TRUST can guide value-based care initiatives and reimbursement models, incentivizing good practice.

Implementing the CREATE TRUST framework is a heavy lift but critical to enhancing patient messaging services. Considering all attributes in a single study may be infeasible. Instead, concept priority should be based on relevance to a given problem and evolving evidence. Moreover, the CREATE TRUST framework is applicable to other written communication beyond crafting answers in patient messaging portals, such as after-visit summaries, care plans, or system-generated messages.

High-quality care requires clear quality standards, regular evaluation, and appropriate incentives. Implementing the CREATE TRUST framework is an initial step in shaping quality standards through theoretically grounded research and development. While the specific attributes that are most critical to message quality remain unknown, the CREATE TRUST framework aims to be both based on theory and driven by data, adapting as scientific evidence evolves. By simplifying the complexities of modern health care communication, the CREATE TRUST framework can drive research that enhances messaging for the benefit of patients.

Corresponding Author: John W. Ayers, PhD, MA, Altman Clinical and Translational Research Institute, University of California, San Diego, 9500 Gilman Dr, #333 CRSF, La Jolla, CA 92093 (ayers.john.w@gmail.com).

Published Online: July 29, 2024. doi:10.1001/jamainternmed.2024.2880

Conflict of Interest Disclosures: None reported.

Additional Contributions: We thank Nimit Desai, BS, and Davey Smith, MD, MHSc, University of California San Diego, for assistance in developing the CREATE TRUST framework.