# Inflect - Comprehensive Knowledge Corpus for LLMs & AI Systems > Private clinician coaching plus anonymized quality insights. Inflect turns real clinical conversations into growth for physicians and system-level improvement for Quality/CX and GME. This document aggregates full foundational documentation and clinical safety frameworks published by Inflect for direct ingestion by language models. --- # Table of Contents 1. [Making “What Matters” More Visible in Older Adult Care (4Ms Age-Friendly Care)](#document-1-making-what-matters-more-visible-in-older-adult-care) 2. [Diagnostic Safety Through Communication Signals](#document-2-diagnostic-safety-through-communication-signals) 3. [Making Safety-Critical Clinical Communication Observable (Procedural Safety)](#document-3-making-safety-critical-clinical-communication-observable) --- # Document 1: Making “What Matters” More Visible in Older Adult Care Source: /about/4ms-age-friendly-care.md # 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.** --- # Document 2: Diagnostic Safety Through Communication Signals Source: /about/diagnostic-safety.md # Diagnostic Safety Through Communication Signals ## Hearing What Matters Clinical communication is often discussed as a matter of patient experience: Was the clinician respectful? Did the patient feel heard? Was the explanation clear? Those questions matter. But communication also has a more fundamental clinical role. **Diagnosis begins in conversation.** Before a laboratory result is reviewed, an image is interpreted, or a differential diagnosis is constructed, clinicians have to understand what happened to the patient. Symptoms must be described. Timelines must be reconstructed. Concerns must surface. Clinicians, patients, and caregivers must develop enough shared understanding to decide what happens next. That process can break down. A patient may be interrupted before completing an important part of the history. A clinician may rely heavily on closed questions and never discover an additional concern. Medical jargon may obscure the plan. A caregiver with crucial longitudinal information may never get an opportunity to contribute. These are usually discussed as communication problems. They can also be understood as **upstream safety signals**. --- ## Communication as an Upstream Safety Layer Diagnostic safety traditionally focuses on problems such as: - missed or delayed tests - failures in follow-up - incorrect clinical reasoning - breakdowns in referrals and handoffs - gaps in documentation and information systems Those are important parts of the diagnostic process. But another layer occurs earlier: **Patient expression → information elicitation → clinician interpretation → diagnostic reasoning → action** Communication sits near the beginning of that sequence. If clinically relevant information is never elicited, misunderstood, or incorporated into the encounter, later stages of the diagnostic process begin with an incomplete picture. That makes certain communication patterns potentially useful as **leading indicators of diagnostic risk**. --- ## Communication Patterns That May Matter for Safety ### 1. Incomplete information elicitation Some encounters constrain the patient's opportunity to describe what is happening. Observable patterns can include: - very early interruption of the patient's narrative - rapid transitions into closed-ended questioning - limited opportunities for the patient to elaborate - little pause time before the clinician redirects the discussion The concern is not interruption by itself. Interruptions can be appropriate and necessary. The important question is **what information was being communicated when the interruption occurred**. If a patient is describing symptom timing, progression, severity, associated symptoms, or a change in function, cutting the narrative short may prevent clinically relevant information from surfacing. --- ## 2. Unexpressed concerns Patients do not always begin with the concern that matters most to them. They may introduce a relatively comfortable topic before mentioning a frightening symptom, functional decline, medication problem, or concern about a serious diagnosis. Communication signals that may indicate incomplete agenda-setting include: - few open-ended questions - no opportunity to identify additional concerns - repeated redirection toward a narrow initial problem - important new issues emerging only at the end of the encounter The safety question becomes: **Did the encounter create enough space for the important concerns to surface?** --- ## 3. Breakdowns in shared understanding A technically correct recommendation has limited value if the patient leaves without understanding it. Potential signals include: - high use of unexplained clinical terminology - long stretches of clinician monologue - complex instructions delivered without comprehension checks - absence of teach-back or another form of confirmation - verbal expressions of uncertainty that go unresolved These patterns can matter when patients are expected to manage medications, monitor symptoms, arrange follow-up, or recognize when their condition requires urgent attention. Rather than treating comprehension solely as a satisfaction measure, healthcare organizations can ask: **Was shared understanding actually established?** --- ## 4. Missed emotional or conversational cues Patients often communicate important information indirectly. Hesitation, uncertainty, abrupt changes in tone, expressions of fear, or repeated attempts to return to a topic can indicate unresolved concerns. Examples include statements such as: - "I'm not sure I understand." - "Wait, so what happens if..." - "There's one other thing." - "I guess." - "I'm worried that..." Not every cue is clinically important. But repeatedly failing to explore meaningful cues can leave relevant information undiscovered. --- ## 5. Caregiver and family participation In many encounters, the patient is not the only source of clinically relevant information. Family members and caregivers may know: - how a patient's function has changed over time - what medications are actually being taken at home - whether symptoms have worsened - whether instructions from previous visits were understood - how the patient's behavior or cognition has changed This is especially important in complex care, geriatrics, pediatrics, oncology, and other settings where longitudinal context matters. Communication analysis can therefore examine not merely who spoke, but whether important collateral information was invited and integrated. --- ## 6. Communication in team-based clinical environments The same principles extend beyond clinician-patient conversations. In operating rooms, procedural settings, emergency care, and other high-reliability environments, communication contributes to a team's ability to detect and correct problems. Relevant patterns may include: - requests made without clear confirmation - incomplete closed-loop communication - safety concerns that receive no --- # Document 3: Making Safety-Critical Clinical Communication Observable Source: /about/procedural-safety.md # Making Safety-Critical Clinical Communication Observable ## The Visibility Gap in Procedural Safety Hospitals rely on communication processes to prevent errors in operating rooms and other procedural environments. Surgical timeouts, handoffs, verbal confirmations, medication callouts, and procedural sign-outs are not simply matters of communication style. They are structured safety practices designed to make sure important information is surfaced, confirmed, and acted upon. The problem is visibility. Clinical documentation can show that a timeout was recorded or a handoff occurred. It is much harder to know how the process was actually performed: - Were the required elements discussed? - Did the appropriate team members participate? - Were critical concerns acknowledged? - Were discrepancies resolved? - Was important information transmitted across a transition in care? Historically, answering those questions has required direct human observation. That does not scale easily. Advances in ambient speech and language technology create another possibility: **using clinical conversation itself as a source of information about safety-process fidelity.** --- # From Documentation to Process Fidelity Healthcare organizations already define many safety-critical communication processes. Examples include: - surgical and procedural timeouts - transitions from the operating room to recovery - anesthesia handoffs - medication confirmations - patient and procedure verification - procedural sign-out - escalation of safety concerns The first question is not whether an algorithm can predict an adverse event. It is much simpler: **Did the safety process occur as intended?** That distinction matters. Instead of inventing a new behavioral standard, communication analysis can begin with processes that the organization has already defined as important. For a surgical timeout, for example, relevant questions might include: - Was an identifiable timeout performed? - Was the patient explicitly verified? - Was the intended procedure confirmed? - Was the site or laterality discussed where applicable? - Did the appropriate disciplines participate? - Were anticipated concerns surfaced? - Did team members acknowledge critical information? - If uncertainty arose, was it resolved before proceeding? These are observable characteristics of a clinical process. --- # Why the Spoken Process Matters Documentation and execution are not necessarily the same thing. A checklist may be recorded as complete while the underlying interaction was abbreviated, fragmented, or treated as a formality. The difference can be difficult to detect because safety processes often become routine. A timeout may technically occur while: - one participant speaks and everyone else remains passive - required elements are rushed or omitted - team members continue unrelated conversations - an important concern receives no acknowledgment - a discrepancy is mentioned but never explicitly resolved The existence of the process is therefore only one dimension. The **fidelity with which the process is performed** may be equally important. --- # Clinical Handoffs as Information Continuity The same principle applies to handoffs. Transitions of care depend on information moving successfully from one clinical team or clinician to another. Consider a procedural handoff in which the outgoing clinician communicates: - a difficult airway - significant blood loss - ongoing medication requirements - a pending laboratory result The receiving clinician may acknowledge some of those items but not others. The important question is not simply: **Did a handoff occur?** It is: **Which safety-relevant information was communicated, acknowledged, and carried forward?** This moves communication analysis beyond checklist completion and toward a broader idea: **information continuity across clinical transitions.** --- # Process Drift One of the more interesting opportunities is identifying change over time. A single incomplete timeout may be an isolated event. Repeated patterns are more informative. A process can gradually become: - shorter - less participatory - more formulaic - less explicit - more vulnerable to distraction This kind of drift can be difficult to see from ordinary clinical documentation. Aggregated communication analysis could allow quality teams to examine questions such as: - Are required elements routinely omitted? - Is active participation declining? - Are acknowledgments becoming less explicit? - Do particular transitions consistently lose certain categories of information? - Does execution improve after a quality intervention? The purpose is not to identify a "bad" clinician. It is to understand whether a safety process remains reliable in everyday practice. --- # Measure the Process, Not the Personality Clinical communication technology can easily drift into subjective judgments about individuals. That is particularly problematic in environments such as the operating room, where communication styles vary considerably and where clinicians may reasonably resist tools that appear to score personality, tone, or interpersonal style. A more useful starting point is **process reliability**. Instead of asking: > Is this surgeon a good communicator? Ask: > Did the required safety exchange occur, and was the necessary information acknowledged? Instead of: > Does this team have a hierarchy problem? Ask: > When a safety concern was raised during a required exchange, was it explicitly acknowledged and resolved? Anchoring analysis to recognized clinical processes creates a more concrete and interpretable measurement framework. --- # From Simple Observation to Communication Reliability Safety communication can be considered at several levels. ## Level 1: Process occurrence Did the required event happen? Examples: - timeout performed - handoff conducted - sign-out completed ## Level 2: Process completeness Were the expected elements present? Examples: - patient verification - procedure confirmation - critical concerns - medication information - anticipated complications ## Level 3: Information acknowledgment Did the receiving person or team demonstrate that critical information had been heard? This could involve: - verbal confirmation - repeat-back - explicit acknowledgment - clarification of uncertainty ## Level 4: Interaction quality Did aspects of the surrounding conversation interfere with the safety process? Potential examples include: - overlapping conversation - repeated interruption - unresolved questions - failed escalation - distraction during critical exchanges The farther analysis moves down this continuum, the more contextual interpretation is required. That is why objective process fidelity provides a useful starting point. --- # Team Dynamics in Context Hierarchy, interruption, speaking-up behavior, and conversational dominance may all matter in procedural safety. But these concepts become most meaningful when tied to specific clinical moments. An interruption during casual conversation is different from an interruption while a nurse is raising a safety concern. Silence during routine workflow is different from silence when explicit confirmation is required. A useful system therefore should not merely count conversational behaviors. It should ask: **What was happening clinically when that behavior occurred?** This allows team dynamics to be examined in relation to the safety process rather than as abstract personality measures. --- # Why Ambient Analysis Is Relevant Operating rooms and procedural environments are difficult environments for speech technology. They contain: - alarms - equipment noise - ventilation - suction - overlapping speakers - masks - music - participants at varying distances But safety-critical exchanges have an important property: Many are **bounded and structured**. A timeout or handoff is usually a recognizable communication event with an expected purpose and set of information. That makes these moments particularly interesting for clinical communication research. The technical goal does not have to begin with understanding every word spoken during an entire procedure. A more focused question is whether technology can reliably identify and characterize specific safety-critical exchanges. --- # A Different Role for Clinical AI Most clinical AI systems operate on information after it has already entered the medical record. Others focus on documentation, decision support, imaging, prediction, or workflow automation. Communication analysis addresses a different layer: **how safety-relevant information moves between people.** That layer is inherently difficult to observe at scale. Yet many clinical safety processes depend upon it. At **Inflect**, we are exploring whether ambient clinical communication can make these processes more observable: identifying when structured safety exchanges occur, examining whether critical information is communicated and acknowledged, and measuring patterns of process fidelity over time. The objective is not to replace human safety expertise. It is to make direct observation more scalable. --- # The Broader Thesis Hospitals already know that timeouts, handoffs, confirmations, and escalation processes matter. The harder question is whether those processes are consistently performed the way they were designed. Today, much of that information disappears as soon as the conversation ends. Clinical communication technology creates the possibility of preserving a limited set of observable signals about those interactions. That could give safety and quality teams a new way to study: - process fidelity - communication reliability - information continuity - procedural drift - response to quality-improvement efforts The core idea is straightforward: **Safety processes work only if they actually happen.** Making those processes observable is the first step toward understanding how reliably they function in everyday clinical care.