# Making “What Matters” More Visible in Older Adult Care ## From Conversation to Age-Friendly Care The Age-Friendly Health Systems framework organizes care for older adults around four domains: - **What Matters** - **Medication** - **Mentation** - **Mobility** The framework is powerful because it asks clinicians to look beyond individual diagnoses and consider whether care reflects the priorities and circumstances of the person receiving it. But the four domains are not equally easy to observe. Medication can often be represented as a list. Mobility can be assessed through function, falls, or activity. Mentation can be evaluated using established screening tools. **What Matters is different.** It is often conversational. A patient may say: - “I want to stay in my own home.” - “I don’t want to end up in the hospital again.” - “I need to be able to get to the bathroom by myself.” - “I’m worried this treatment will leave me too tired to take care of my husband.” These statements can carry enormous clinical significance. Yet they do not naturally fit into conventional documentation systems. --- # The Problem with “What Matters” “What Matters” is difficult to capture because it is rarely a simple field. It may be: - explicit or implied - stable or changing - expressed by the patient or a caregiver - embedded in a story rather than stated as a goal - revealed through hesitation, tradeoffs, or concern A patient may never say: > “My primary healthcare goal is maintaining independence.” They may instead say: > “I really don’t want my daughter to have to move me out of the house.” Those statements can mean roughly the same thing clinically, but only one resembles structured data. This creates a gap between what is discussed in the room and what remains available to the care team afterward. --- # Much of “What Matters” Is Already in the Conversation One response to this problem would be to ask clinicians and patients to complete more forms. That may add structure, but it also adds work. Another possibility is to ask whether the information is already present in ordinary clinical conversation. Often, it is. Patients talk about: - independence - family responsibilities - fear of hospitalization - treatment burden - functional goals - side effects they are unwilling to tolerate - activities they want to preserve - worries about cognition or mobility Caregivers add another layer of context: - “She has fallen twice this month.” - “He stopped taking that medication.” - “She is having a much harder time living alone.” - “He cares more about staying alert than eliminating the pain completely.” These are not merely conversational details. They can shape what constitutes a reasonable care plan. --- # What Makes a “What Matters” Signal? Not every personal statement needs to become clinical data. The interesting challenge is identifying the parts of conversation that reveal goals, preferences, constraints, and tradeoffs. Several kinds of signals may be especially useful. ## Direct goals and preferences These are explicit statements such as: - “I want to stay independent.” - “I would rather avoid surgery.” - “I want to be able to attend my granddaughter’s wedding.” ## Tradeoffs Patients frequently reveal priorities by describing what they are willing or unwilling to tolerate. For example: - “If it makes me that sleepy, I don’t want to take it.” - “I’d rather have some pain than feel confused all day.” - “Staying home matters more to me than going through another hospitalization.” Tradeoffs often provide richer information than a generic preference field because they reveal how a patient weighs competing outcomes. ## Functional priorities Goals are often expressed through ordinary activities: - gardening - walking - driving - cooking - attending religious services - living independently - caring for a spouse These activities can make abstract treatment decisions concrete. ## Caregiver observations Family and caregivers may provide longitudinal information that does not emerge in a short visit. Their perspective can be especially important when cognition, mobility, medication management, or functional decline is changing over time. --- # “What Matters” Changes Another reason “What Matters” is difficult to represent is that it is not static. A patient's priorities can change because of: - disease progression - new symptoms - treatment side effects - a hospitalization - changes in cognition - changes in living situation - caregiver burden - changes in prognosis The useful question is therefore not simply: **Was “What Matters” ever documented?** It is: **Do we still know what matters to this patient now?** That creates a longitudinal problem as much as a documentation problem. A health system may know what a patient valued six months ago without knowing whether those priorities still hold today. --- # From Capture to Alignment Capturing patient goals is only useful if those goals can inform care. The more important question is whether the plan aligns with what the patient has said matters to them. Suppose a patient repeatedly emphasizes maintaining alertness and independence. A treatment plan that substantially impairs either may still be clinically appropriate. But the tradeoff should be recognized and discussed. That is the distinction between **recording preferences** and **using preferences in decision-making**. Communication technology may eventually make this easier by preserving important statements from the encounter and making them easier to revisit. The goal is not to convert a patient's life into another checklist. It is to reduce the chance that meaningful context disappears after the conversation ends. --- # The Role of Caregivers Older adult care is frequently distributed across a network of people. A physician may see the patient for twenty minutes. A daughter, son, spouse, or other caregiver may observe: - day-to-day function - medication use - falls - confusion - appetite - sleep - transportation problems - ability to manage activities at home That means “What Matters” can sometimes be relational rather than individual. The patient's own goals remain central, but caregiver input may reveal practical constraints or changes that are essential to understanding those goals. Clinical communication systems should therefore be able to distinguish among different voices rather than treating the encounter as a simple clinician-patient exchange. --- # The Limits of Structured Documentation Traditional clinical documentation is necessarily selective. It is designed to preserve information needed for care, billing, continuity, and legal documentation. But not everything clinically meaningful fits naturally into a conventional note. A sentence such as: > “Patient wants to remain independent.” captures something important. But it loses much of the surrounding context: - Why is independence important? - What does independence mean to this patient? - Which activities are most important? - What tradeoffs are they willing to make? - Has that priority changed? Conversation often contains this context before it is compressed into documentation. That makes the original interaction an important source of information in its own right. --- # A Measurement Problem If “What Matters” is important to age-friendly care, health systems face a practical measurement challenge. They may want to know: - How often are patient goals meaningfully discussed? - Are goals revisited over time? - Are caregivers included when appropriate? - Are tradeoffs explicitly explored? - Does the eventual plan reflect the priorities expressed in the encounter? These questions are difficult to answer through retrospective chart review alone. The medical record may show whether a field was completed. It may not show whether the underlying conversation was substantive. That is where conversational analysis becomes interesting. --- # Beyond Checkbox Completion There is an important distinction between **presence** and **quality**. A clinician can technically ask: > “What matters most to you?” and receive a superficial response. Another clinician may never use that exact wording but spend several minutes exploring the patient's goals, fears, responsibilities, and treatment preferences. A useful measurement approach therefore should not depend solely on whether a particular phrase occurred. It should look for evidence that the interaction surfaced meaningful information about the patient's priorities. That requires context. --- # A Longitudinal View The most useful application may not be analyzing one visit in isolation. Imagine being able to see how a patient's stated priorities evolve across encounters. For example: ### January > “I want to keep working and stay active.” ### April > “The fatigue is starting to make work difficult.” ### August > “I care more about feeling well enough to spend time with my family than continuing this treatment at the highest dose.” Each statement changes the context in which future care decisions are made. A longitudinal view could make changes like these easier for clinicians and care teams to recognize. --- # Where Conversational AI Fits Most ambient clinical technology focuses primarily on converting speech into documentation. That is a valuable use case. But clinical conversation contains information that extends beyond what belongs in a traditional note. At **Inflect**, we are exploring whether conversational AI can help identify and preserve signals related to the 4Ms, particularly the parts of **What Matters** that are easily lost when an encounter is compressed into conventional documentation. That includes questions such as: - Was an important goal expressed? - Was a treatment tradeoff discussed? - Did a caregiver provide relevant context? - Did the patient's priorities appear to change? - Was the care plan discussed in relation to those priorities? The purpose is not to replace clinician judgment or turn personal preferences into an automated score. It is to make clinically meaningful context easier to see. --- # What Needs to Be Validated This field requires careful empirical work. Important research questions include: - Can conversational systems reliably identify expressions of patient goals and preferences? - How accurately can they distinguish meaningful “What Matters” content from ordinary conversation? - Do clinicians agree with the system's interpretation? - Can relevant changes be tracked longitudinally? - Does making this information more visible change clinical decision-making? - Does it improve patient or caregiver understanding? - Does it affect adherence, utilization, or other outcomes? These questions should be tested rather than assumed. The promise is not that conversational AI automatically creates age-friendly care. The promise is that it may make an important but poorly structured part of care more observable. --- # The Broader Idea Age-friendly care asks a deceptively simple question: **What matters to this person?** Healthcare systems have become very good at storing diagnoses, laboratory values, medication lists, and procedures. Patient priorities are harder. They are contextual. They change. They emerge through conversation. And they are often distributed across patients, clinicians, and caregivers. That makes them difficult to reduce to a checkbox. It also makes them worth preserving. The opportunity for conversational intelligence is not to automate the philosophy of age-friendly care. It is to make more of the information behind that philosophy visible. **“What Matters” should not disappear when the conversation ends.**