I Am Still Trying to Change It

SCHOLARLY ESSAY · DIGITAL HEALTH, EMBODIMENT, AND HUMAN AGENCY

Digital Self-Advocacy, Embodied Difference, and the Ethics of the Longitudinal Record

I am still trying to change it.

Abstract

Digital access to clinical notes, laboratory results, imaging reports, and medical literature is changing the patient’s role in health care. This development is commonly described as patient engagement or empowerment, terms that understate its epistemic significance. A patient can now construct a longitudinal record that preserves relationships among findings, interventions, symptoms, and unanswered questions across a health system organized principally around discrete encounters. Such a record may reveal clinically important continuity, expose contradictions, and improve the precision of patient–clinician deliberation. It may also preserve errors, encourage spurious causal inference, intensify illness surveillance, and expose sensitive information to systems outside traditional health-privacy protections.

This essay examines those possibilities through a reflexive clinical narrative involving pronounced inward knee alignment present since childhood, an atypical gait, spinal curvature reported as relatively stable across available imaging, reduced bone density, degenerative spinal findings, and a later history of surgically induced malabsorption with nutritional deficiencies. The chronology excludes several simplistic explanations but does not establish a single causal model. Structural lower-limb alignment may influence gait and compensatory posture; pelvic or spinal asymmetry may alter apparent limb mechanics; nutritional compromise may affect bone and muscle without explaining a childhood deformity; and several conditions may coexist without belonging to one unifying pathology.

The essay proposes patient-held continuity infrastructure as a model for technologically assisted self-advocacy. Its function is not autonomous diagnosis but the preservation of provenance, chronology, uncertainty, and clinically testable questions. Its legitimacy depends upon task-specific artificial-intelligence governance, professional verification, privacy safeguards, equitable access, and a disability-informed distinction between care and normalization. Medicine supplies measurement and differential diagnosis; narrative medicine restores temporality; philosophy clarifies epistemic authority; psychology explains the capacities required for sustained care; disability studies interrogates compulsory normality; and Christian disability theology offers one tradition-specific account of dignity independent of cure. The resulting model is neither passive patienthood nor algorithmic sovereignty, but disciplined participation in the interpretation and care of a body no isolated encounter can fully contain.

Method and Limits

This is an interdisciplinary critical essay organized around a reflexive clinical narrative. It is not a case report, systematic review, diagnostic assessment, or substitute for individualized medical care. The personal history identifies the problem under examination; it does not establish the cause of any finding. Exact dates and potentially identifying clinical details are intentionally omitted. Relative chronology is retained only where sequence materially constrains causal reasoning—for example, the fact that a visible lower-limb difference existed before an adult surgical procedure associated with later nutritional risk.

The empirical discussion draws from official clinical positions, government regulations and data, systematic reviews, randomized trials, observational research, and peer-reviewed clinical literature. The conceptual argument engages narrative medicine, philosophy of medicine, disability studies, and Christian disability theology. These fields do not supply interchangeable forms of proof. Medical research addresses measurement, association, intervention, and prognosis. Philosophy analyzes knowledge and authority. Psychology studies distress, behavior, and adaptation. Disability studies examines the relation between embodiment and social environments. Theology contributes a normative interpretation of dignity, dependence, and healing within a specific religious tradition.

The essay uses three evidentiary categories. A documented fact is contained in a clinical source or directly established by the history presented. A general proposition is supported by research but may not apply to the individual case. A case-specific hypothesis is a plausible relationship that remains unverified. Preserving these distinctions is not preliminary housekeeping; it is the central intellectual discipline of digitally assisted self-advocacy.

---

I. The Body Made Public

Since childhood, my knees have bent markedly inward. I have generally concealed their shape beneath pants and remained conscious of how I walk. Such consciousness can be dismissed as insecurity, as though the relevant problem exists only in perception. That interpretation fails because the gait is not entirely private. It is visible to other people.

A stranger recently asked why I walked so strangely. The question did not appear intentionally cruel. It seemed to have escaped before the speaker considered whether my body was available for public explanation. The absence of malice did not make the encounter inert. It continued after the conversation ended, repeating itself until I formed an answer:

I am still trying to change it.

The sentence is true, but ethically unstable. “It” might mean alignment, gait, pain, mobility, muscular control, confidence, medical understanding, or the social exposure produced by visible difference. Wanting to move with less pain is not identical to wanting to become less noticeable, although both motives may coexist. Nor is the distinction easily purified. Decisions about bodily change are made within environments that reward some appearances and movements while marking others as unusual.

Visible-difference research documents experiences that include staring, name-calling, unsolicited questions, concealment, and appearance-related distress. The evidence also resists a simple geometric theory in which psychosocial burden is determined by the objective extent of bodily difference. Subjective appraisal, social response, biography, support, and perceived control influence adjustment, and the intervention literature remains methodologically limited (Norman and Moss 2015).

The stranger’s question therefore belongs in the clinical narrative, although it supplies no evidence about anatomical causation. It identifies a consequence: the body has become a site of involuntary explanation. The patient must manage both the mechanics of movement and the social meaning assigned to visible movement.

Rosemarie Garland-Thomson’s concept of misfitting is useful because it locates disability neither wholly inside the body nor wholly inside an abstract society. A misfit occurs in the relation between a particular embodiment and an environment unable or unwilling to sustain it. The same body may fit in one context and misfit in another; the relationship is material, spatial, temporal, and political rather than a fixed personal essence (Garland-Thomson 2011).

Narrative medicine adds a different claim. Rita Charon defines narrative competence as the capacity to recognize, absorb, interpret, and respond to accounts of illness. Its purpose is not to displace biomedical evidence with personal meaning but to improve the range and accuracy of what becomes clinically knowable. Arthur Frank likewise treats illness narratives as ethical acts through which embodied persons interpret experiences that institutions may otherwise reduce to coded findings (Charon 2001; Charon 2006; Frank 2013).

“My knees are bent,” “my gait is unusual,” “my spine is curved,” and “my bone density is low” name observations or findings. They do not determine how those findings relate, which are causing present functional burdens, what is modifiable, or what forms of change are worth pursuing. The emotional account cannot answer those questions. Neither can an imaging report viewed in isolation. The problem becomes properly specified only when lived chronology and disciplined measurement are allowed to correct one another.

II. Medicine Encounters Visits; the Patient Carries Time

Modern medicine is organized through specialization. A radiologist interprets images acquired under particular technical conditions. An orthopedic clinician assesses alignment. A spine specialist considers curvature and degeneration. A physical therapist observes movement. A metabolic-bone specialist evaluates densitometry and secondary causes. A bariatric team monitors nutritional consequences. A primary-care clinician may be expected to coordinate these domains within limited appointment time and across records that remain institutionally fragmented.

Fragmentation is not necessarily evidence of neglect; it is an ordinary consequence of distributed expertise. The problem is that the patient’s body is continuous while the system’s knowledge of it is episodic.

The patient occupies a distinctive but nonsovereign epistemic position. I do not possess the diagnostic training of my clinicians. I do know which visible difference was present in childhood, which concerns emerged after surgery, what I have attempted, what has remained unexplained, and which outcomes matter to me. That knowledge is indispensable and fallible. Memory can distort chronology, privilege dramatic events, and create causal relationships retrospectively. Records can correct memory, yet records also contain copied assumptions, inconsistent terminology, omissions, technical artifacts, and mistakes.

The proper contrast is therefore not clinical expertise versus patient truth. It is a system of partially overlapping, differently fallible forms of knowledge.

A patient-held digital record can strengthen that system when it preserves distinctions among five layers:

Layer	Function	Principal risk
Source record	Preserves reports, results, notes, images, and medication history	Source errors acquire authority through repetition
Patient chronology	Records sequence, change, and intervention history	Memory is mistaken for measurement
Patient observation	Preserves symptoms, function, priorities, and lived consequences	Subjective experience is either overgeneralized or dismissed
Interpretive layer	Records clinicians’ assessments and research-based explanations	Interpretation is mistaken for established fact
Hypothesis register	Lists unresolved relationships and discriminating questions	Plausibility is converted into diagnosis

This architecture is more demanding than accumulating PDFs. Its value comes from provenance: the ability to identify where a statement originated and what epistemic status it possesses.

In the United States, the HIPAA Privacy Rule generally gives individuals access to protected health information in designated record sets maintained by covered entities or their business associates, subject to limited exceptions. These records can include clinical notes, laboratory reports, billing information, and imaging. The rule does not require a provider to create an explanation or analysis that does not already exist, and separately maintained psychotherapy notes are generally excluded from the access right (U.S. Department of Health and Human Services).

Technical infrastructure has also expanded. In 2024, approximately nine in ten surveyed U.S. hospitals reported enabling patient access through an application programming interface, seven in ten reported standards-based access, and roughly two-thirds reported some capability for patients to submit data electronically. These figures describe hospital capability, not seamless patient experience; adoption varied by hospital size, system affiliation, and electronic-record vendor (ASTP/ONC).

Patient-level data reveal the remaining fragmentation. In 2024, 59 percent of surveyed individuals reported having multiple portals or online medical records, while only 7 percent reported using an application that consolidated information from multiple sources. Technical access has expanded faster than practical integration (ASTP/ONC).

Open-notes research illustrates both promise and limits. In a large survey across three health systems, many respondents who read ambulatory notes reported that access helped them remember care plans, prepare for visits, participate actively, and feel more in control. The response rate was 21.68 percent, however, and the responding population was disproportionately older, White, and college educated; these limitations constrain generalization (Walker et al. 2019).

Transparency also exposes inaccuracies. In a survey of patients invited to review notes, approximately one in five respondents reported finding a mistake, and 40 percent of those patients considered the mistake serious. Reported errors involved diagnoses, history, medications, examinations, results, laterality, and—in a small number of cases—information apparently belonging to another patient. These were patient-reported perceptions rather than independently adjudicated error rates, but the findings demonstrate why access can serve patient safety as well as autonomy (Bell et al. 2020).

A longitudinal archive is therefore neither a private copy of medical truth nor a decorative expression of engagement. It is a reconciliation instrument. It allows source records, patient memory, professional judgment, and emerging evidence to remain comparable without becoming indistinguishable.

III. What the Available Evidence Establishes

The personal evidence can be summarized without exact dates.

The pronounced inward appearance of my knees and the atypical gait have existed since childhood. Spinal imaging documents curvature and degenerative changes; comparison with earlier imaging has reportedly shown no dramatic interval change. Bone imaging and densitometry have raised concern about reduced bone density. I later underwent a malabsorptive bariatric procedure and have subsequently had documented nutritional deficiencies. The causal relationships among these findings remain unresolved.

This relative chronology excludes at least one simplistic account. An adult surgical procedure cannot by itself explain a lower-limb difference already present in childhood. That does not determine whether the original alignment was idiopathic, developmental, metabolic, physeal, rotational, neuromuscular, post-traumatic, or associated with another process. It establishes a boundary around one causal claim.

The reported stability of spinal imaging is reassuring within similarly bounded limits. Comparable serial imaging showing little change reduces concern about rapid progression during the interval examined. It does not establish when the curvature began, whether it is structural or compensatory, or whether it explains the gait. Apparent differences in spinal and limb alignment can also be influenced by positioning, projection, rotation, flexion, measurement technique, and interobserver variability. Stability narrows one uncertainty without resolving the system.

Bone-density terminology requires comparable discipline. “Diffuse osteopenia” on a conventional radiograph is a qualitative observation and is not equivalent to a densitometric diagnosis. For males younger than fifty, the International Society for Clinical Densitometry recommends Z-scores rather than T-scores. A Z-score of −2.0 or lower is described as “below the expected range for age,” and osteoporosis should not be diagnosed in this population from bone mineral density alone. Interpretation also depends on fracture history, secondary causes, anatomical site, scanner quality, and the technical comparability of serial studies (ISCD 2023).

The history of biliopancreatic diversion with duodenal switch is relevant to metabolic bone health but cannot serve as a universal explanation. A systematic review found substantial long-term nutritional-deficiency burdens after duodenal switch, with considerable heterogeneity among studies (Nakanishi et al. 2024).

A ten-year randomized comparison of duodenal switch and Roux-en-Y gastric bypass provides stronger but narrower evidence. The trial enrolled sixty people with baseline body mass indexes between 50 and 60. At long-term assessment, bone mass was lower after duodenal switch, vitamin deficiencies were more common, and four of twenty-nine participants in the duodenal-switch group developed severe protein-calorie malnutrition, three of whom underwent revisional surgery. The study was small, and its population and surgical configuration cannot be assumed to represent every person who has undergone the procedure (Salte et al. 2024).

The responsible individual-level account remains plural. A longstanding lower-limb alignment difference may influence movement. A structural or functional limb-length difference may or may not exist. Pelvic obliquity may be primary, compensatory, or absent. The spinal curve may be structural, compensatory, or partly independent. Adult malabsorption may contribute to bone or muscle vulnerability without explaining the childhood appearance. Degenerative imaging findings may be symptomatic, contributory, or incidental. Several processes may interact; others may coexist without a common origin.

The desire to connect the findings is understandable. It is not evidence that they form one disease.

IV. Alignment, Movement, and the Limits of Anatomical Intuition

The ordinary expression “knock knees” conceals several different phenomena. Structural genu valgum concerns osseous alignment in the coronal plane. Dynamic knee valgus describes a movement pattern that may include hip adduction or internal rotation, knee abduction, tibial rotation, altered foot mechanics, or combinations of these. Apparent valgus can also be influenced by rotational anatomy, flexion, pelvic position, stance, or soft-tissue contours. The distinction is essential because an intervention that changes dynamic control cannot be presumed to change mature bone geometry.

Clinical deformity analysis asks not only whether the knees appear close together but where the alignment originates and how it behaves under load. Potentially relevant components include distal femoral and proximal tibial orientation, joint-line obliquity, femoral and tibial rotation, foot progression, pelvic position, contracture, muscular control, and compensatory posture.

Full-length standing anteroposterior radiography is a principal method for assessing coronal lower-limb alignment because it permits measurement of the mechanical axis from the hip to the ankle and helps localize femoral and tibial contributions. Isolated knee images may not establish whole-limb mechanical alignment, and positioning or limb rotation during acquisition can materially alter measurements (Luís and Varatojo 2021).

Static imaging answers a static question. Walking is dynamic. Two people with similar radiographic alignment may use different movement strategies because gait also reflects pain, strength, balance, range of motion, neurological control, fatigue, footwear, learned compensation, and speed. Instrumented gait analysis can quantify joint angles, temporal-spatial parameters, ground-reaction forces, and kinetic patterns, but its value depends on whether the findings are expected to alter management. Description is not the same as causal identification.

Limb length presents another source of ambiguity. A structural discrepancy reflects differences in bony length. A functional or apparent discrepancy may arise from pelvic obliquity, spinal posture, contracture, joint position, or foot mechanics. Clinical and imaging methods vary in validity, and a lift may be inappropriate where the apparent difference is compensatory rather than anatomical (Alfuth, Fichter, and Knicker 2021; Vogt et al. 2020).

The language of a “kinetic chain” can be useful but is often oversimplified. It suggests that one defective link transmits dysfunction linearly through neighboring joints. Human movement is better understood as adaptive control across multiple anatomical segments, sensory inputs, tissue constraints, and learned strategies. A compensation may preserve balance while increasing load elsewhere. Anatomical adjacency makes a relationship plausible; it does not establish its direction.

The same principle governs orthoses. “Orthotic” may describe a foot insert, heel lift, ankle-foot orthosis, brace, or another device with a specific objective. A device may redistribute pressure, accommodate asymmetry, influence distal motion, improve stability, or reduce symptoms. It should not be represented as a generic means of correcting mature whole-limb deformity. The pertinent questions are: What measured impairment is being targeted? What outcome should change? Over what period? What compensatory or adverse effects will be monitored?

Physical therapy also requires an impairment-specific rationale. It may improve strength, balance, endurance, motor control, movement confidence, gait efficiency, and pain. Its expected effects depend on what is modifiable: hip or trunk control, quadriceps capacity, ankle mobility, balance, deconditioning, fear avoidance, or dynamic movement. Structural permanence does not imply functional futility, but functional improvement should not be described as skeletal correction.

The clinical task is therefore to identify which features are structural, dynamic, compensatory, symptomatic, progressive, and modifiable. “Can this be fixed?” is too coarse a question. The more useful inquiry is: Which outcome can plausibly be changed, by which mechanism, with what burden and uncertainty?

V. From Access to Epistemic Agency

Technology changes the patient’s position when it permits preparation before an appointment, comprehension during it, and verification afterward.

A patient can learn the difference between a Z-score and a T-score, structural and functional discrepancy, or static and dynamic valgus. That knowledge does not confer diagnostic authority. It permits better questions: Which hypothesis is this test evaluating? What measurement would distinguish the possibilities? Does the intervention target pain, structure, movement, future risk, or appearance? What result would change management? How will benefit be assessed?

This capacity is best described as epistemic agency: meaningful participation in practices through which knowledge is produced, criticized, and applied.

Miranda Fricker’s account of epistemic injustice identifies wrongs committed against people specifically in their capacity as knowers. Testimonial injustice occurs when prejudice improperly reduces a speaker’s credibility. Hermeneutical injustice occurs when unequal access to shared interpretive resources prevents an experience from becoming adequately intelligible (Fricker 2007).

Clinical settings can produce both. A patient may know that a gait difference is longstanding but be unable to express why that chronology matters. Another may use precise terminology and still be discounted as anxious, difficult, psychiatrically labeled, or “internet informed.” Technology can help close a vocabulary gap; it cannot abolish prejudice, hierarchy, time pressure, or institutional power. It may intensify those forces when an AI-generated account is treated as evidence that the patient has become fixated on an implausible explanation.

Epistemic agency therefore requires calibrated language:

This finding is documented.
This relationship is supported in general populations.
Its relevance to my case has not been established.
This hypothesis would require a particular examination or measurement.
This study’s population differs materially from me.
This part of the decision depends upon my goals rather than anatomy alone.

Such language neither capitulates to authority nor claims authority the patient does not possess. It creates an auditable interface between forms of knowledge.

Shared decision-making provides a clinical structure for that interface. The Agency for Healthcare Research and Quality defines it as collaboration in which evidence, professional knowledge, and the patient’s values, goals, preferences, and circumstances jointly inform care (AHRQ).

Patient decision aids offer the strongest adjacent evidence. A 2024 Cochrane review included 209 studies involving 107,698 participants across seventy-one decisions. The aids generally improved knowledge, accuracy of risk perceptions, and alignment between informed values and choices. They did not remove uncertainty or guarantee superior clinical outcomes (Stacey et al. 2024).

This matters when a patient asks for the “percentage of impact” associated with an option. A probability is clinically meaningful only when its endpoint, population, comparator, time horizon, uncertainty interval, and applicability are specified. “Success” may mean radiographic correction, reduced pain, improved walking, satisfaction, avoidance of later surgery, or a composite outcome. Technology can help expose those distinctions. It must not generate a precise individualized percentage from evidence incapable of supporting one.

VI. Artificial Intelligence and the Manufacture of Coherence

Artificial intelligence should be evaluated by task rather than by category. Organizing records is not the same as triage; translating an imaging term is not diagnosis; generating questions is not selecting treatment. Each task has a different consequence of error and therefore requires a different evidentiary threshold.

Task	Potential value	Characteristic failure	Required safeguard
Chronology construction	Organizes dispersed records	Incorrect temporal order or omitted source	Source links and human verification
Report explanation	Makes technical language usable	Loss of nuance or false reassurance	Original text displayed beside explanation
Question generation	Improves appointment preparation	Anchoring on one proposed cause	Competing hypotheses and uncertainty prompts
Trend detection	Identifies repeated or changing values	Noncomparable measurements treated as a trend	Verification of method, site, units, and reference range
Evidence synthesis	Compares studies and options	Population evidence individualized without justification	Applicability and certainty assessment
Triage	May support navigation	Delayed or inappropriate care	Validated workflow with conservative escalation
Diagnosis or treatment recommendation	May support clinicians in bounded settings	Hallucination, omission, bias, or automation dependence	Clinical validation, professional oversight, and accountability

One of generative AI’s principal advantages is cognitive compression. It can convert a long archive into a navigable timeline, compare terminology, or separate findings from hypotheses. Compression is also deletion. A summary necessarily decides what matters. It can omit a negative result, misorder events, treat copied chart language as independent confirmation, or connect findings because the relationship produces a persuasive narrative.

The central risk is therefore not only fabricated information. It is premature coherence: an elegant explanation assembled from individually real facts whose causal relationship has not been demonstrated.

A 2026 preregistered randomized study illustrates the difference between model performance and human–model performance. Leading language models identified relevant conditions in 94.9 percent of complete test scenarios when evaluated directly. Yet 1,298 members of the public using those systems identified relevant conditions in fewer than 34.5 percent of cases and selected appropriate dispositions in fewer than 44.2 percent, performing no better than participants using conventional sources. The study concerned triage scenarios rather than record organization, but it demonstrates that model benchmarks cannot substitute for evaluation of the full human–AI interaction (Bean et al. 2026).

A pragmatic 2026 trial in sixteen Kenyan primary-care facilities tested a different use case: an electronic-record-integrated language-model decision-support tool used by clinical officers. Among 9,691 enrolled patients, the primary fourteen-day treatment-failure outcome did not differ significantly between groups. The intervention did improve several documentation measures. Some reviewed outputs were rated unsafe or inappropriate, and the trial was not powered to establish equivalence for rare serious harms. Workflow integration and clinician oversight made the system more governable; they did not guarantee improved patient outcomes (Agweyu et al. 2026).

The appropriate conclusion is neither technological triumph nor rejection. Systems must be evaluated in the exact work they perform, with real users, representative data, meaningful outcomes, and analysis of downstream errors.

WHO guidance on large multimodal models emphasizes autonomy, safety, transparency, accountability, equity, and governance. NIST’s Generative AI Profile similarly treats risk as contextual and lifecycle-dependent, requiring organizations to govern, map, measure, and manage risks rather than rely on a single accuracy metric (WHO 2024; NIST 2024).

For personal medical use, AI should behave less like an oracle than an auditor. It should preserve source provenance, distinguish extraction from inference, retain contradictions, state what information is absent, generate more than one causal possibility, and identify the evidence required to discriminate among them. The more consequential the task, the less acceptable unsupervised fluency becomes.

VII. Privacy, Inequality, and the Labor of Self-Advocacy

A comprehensive personal archive concentrates highly sensitive information. Its usefulness and its exposure increase together.

HIPAA protection does not follow health information everywhere it travels. HIPAA principally governs covered entities and business associates. Many general-purpose AI systems, consumer applications, and data platforms operate outside that relationship. The Federal Trade Commission’s revised Health Breach Notification Rule addresses vendors of personal health records and related technologies that are not covered by HIPAA, requiring notification after certain breaches. Breach notification is not equivalent to the clinical privacy, use limitations, or fiduciary expectations patients may assume (FTC 2024).

Before placing records into a digital service, the patient should understand whether information is retained, used for model improvement, reviewed by humans, shared with affiliates, processed internationally, exportable, correctable, or permanently deletable. Encryption and authentication matter, but governance also concerns purpose: who may use the information, for what reason, and under what obligation.

A locally controlled encrypted archive is not equivalent to placing a full medical history into a consumer chatbot. Convenience can obscure a transfer of custody.

The model also creates distributive problems. Maintaining a sophisticated record requires devices, broadband access, literacy, time, medical vocabulary, executive capacity, and confidence in institutional interaction. Portal and application use remains uneven, and clinician encouragement materially influences whether patients engage with available tools.

Patient-generated data can improve communication and monitoring, but systematic reviews identify persistent problems involving integration, data quality, clinical usefulness, volume, responsibility, and workflow. Evidence of direct improvement in clinical outcomes remains limited and heterogeneous (Demiris et al. 2019).

The burden is not abstract. Records must be located, downloaded, labeled, compared, corrected, and updated. Appointments must be scheduled. Recommendations must be implemented. Depression, anxiety, pain, fatigue, ADHD, cognitive overload, disability, finances, and work can disrupt every stage.

A health system must not make technological fluency a new criterion of deservingness. A patient who maintains a rigorous archive should be able to use it. A patient unable or unwilling to do so remains entitled to synthesis, explanation, and competent care. “Patient activation” must not become a euphemism for transferring coordination failures from institutions to patients.

The archive should also be clinically legible rather than maximal. Its appointment-facing layer should answer a limited set of questions: What is the current concern? What has been objectively established? What changed? What remained stable? Which interventions have been attempted? What is the functional burden? Which hypotheses remain unresolved? What decision is now required?

A complete archive may contain hundreds of pages. Effective advocacy often requires presenting one.

VIII. Mental Health and the Capacity to Persist

Self-advocacy is not only a knowledge practice. It is an expenditure of psychic and executive energy.

Knowing that an appointment matters does not create the capacity to schedule it. Understanding supplementation does not guarantee sustained adherence. Recognizing that physical therapy may help does not eliminate fatigue, shame, executive dysfunction, cost, or discouragement. Complex care requires repeated action while outcomes remain uncertain.

Patient-activation research describes differences in knowledge, confidence, skills, and willingness to manage health. A meta-analysis of activation interventions in chronic disease found improvements across several physiological, behavioral, and psychosocial outcomes, but study quality and effects varied (Lin et al. 2020). Activation should therefore be treated as modifiable and context dependent, not as a moral trait.

My own experience is that improved mental health has changed the future tense of care. When a long future becomes emotionally credible, preventive effort becomes intelligible. Appointments, nutritional treatment, rehabilitation, and investigation become investments in a self one expects to continue inhabiting.

This relationship must not be inverted into blame. A patient who cannot maintain a digital archive has not proved that they value life less. Psychological distress, executive dysfunction, illness severity, economic constraint, and poor system design may all reduce the capacity to act.

Self-compassion offers a more defensible motivational premise than humiliation. Observational research associates self-compassion with lower perceived stress, health-promoting behaviors, and better self-reported physical health, but the evidence does not establish a simple causal pathway from self-compassion to medical outcomes (Homan and Sirois 2017).

Shame says: Change this body so that it may become acceptable.

Self-compassion says: Because this body already warrants care, determine what might improve function, reduce preventable harm, or enlarge the life available to it.

The second proposition may support surgery, therapy, orthoses, nutritional treatment, accommodation, observation, or no structural intervention. It does not predetermine the choice. It changes the moral premise from which the choice is considered.

IX. Disability, Cure, and the Ethics of Change

Medicine can measure whether an intervention changes alignment, pain, gait, bone density, or fracture risk. It cannot demonstrate that a person becomes more worthy when a measurement improves.

Disability studies complicates the assumption that bodily correction and human flourishing are interchangeable. Garland-Thomson’s relational account of misfitting shows that disability emerges through encounters between bodies and environments. Alison Kafer’s political and relational model rejects a future in which disability appears only as a problem to be eliminated, while also resisting the opposite assumption that disabled people must reject treatment in order to possess politically legitimate identities (Kafer 2013).

This framework makes the sentence “I am still trying to change it” more difficult and more honest. What suffering belongs to the body? What suffering is imposed by social judgment? What intervention expands function or chosen possibility? What intervention primarily makes the body pay the price of other people’s intolerance?

There may be no pure motive. A person can seek less pain and also want strangers to stop noticing. They can pursue structural correction without believing that disabled bodies lack value. They can accept an enduring difference without converting acceptance into a prohibition against treatment.

Christian disability theology contributes one account of why dignity cannot depend upon cure. Nancy Eiesland’s The Disabled God interprets the resurrected, wounded Christ as a challenge to theological traditions that imagine divinity through bodily invulnerability. Her argument is not a generalized celebration of wounds; it is a political and ecclesial claim that disabled people belong as authoritative members of communities that have often treated them as objects of pity, healing, or inspiration (Eiesland 1994).

Deborah Beth Creamer’s limits model approaches the problem differently. Rather than dividing humanity into a limited disabled minority and an unrestricted normal majority, she treats limitation, dependence, and vulnerability as ordinary features of embodiment. The model does not erase disability-specific exclusion; it challenges the fantasy that independence and unrestricted capacity define normal personhood (Creamer 2009).

Christian Scripture can support this account only if used with interpretive restraint. John 9 rejects the immediate attribution of congenital bodily difference to personal or parental sin. First Corinthians 12 redirects honor toward members treated as weaker or less presentable. The resurrection narratives preserve continuity between wounded and risen embodiment. None of these passages independently constitutes a modern disability theory, and Christian healing traditions have often pressured disabled people to become evidence of cure. Their constructive value depends upon being read alongside that history.

The language of stewardship is equally ambivalent. It can describe attentive responsibility toward health; it can also sanctify optimization and imply that illness reflects deficient management. A disability-informed stewardship must include dependence, accommodation, rest, refusal, uncertainty, and the judgment that an intervention’s burdens exceed its likely value.

Care responds to dignity already possessed. Treatment may serve that dignity; it does not create it.

X. Patient-Held Continuity Infrastructure

The personal digital record becomes defensible when governed as continuity infrastructure rather than as a private diagnostic engine.

Its purpose is to preserve the longitudinal body across fragmented care. Seven principles follow.

Provenance

Every extracted statement remains traceable to its source. Original reports are never silently overwritten by summaries.

Temporal integrity

Sequence is preserved without claiming precision not present in the evidence. Exact dates may be withheld from publication while remaining available in the private clinical archive.

Evidentiary status

Documented facts, patient observations, professional interpretations, population evidence, and hypotheses remain distinguishable.

Comparative validity

Trends are asserted only when the measurements are technically comparable. Changes in anatomical site, scanner, units, reference ranges, posture, or imaging technique may produce apparent change without biological change.

Task-bounded AI

Systems receive defined work—retrieval, translation, reconciliation, or question generation—rather than open authority to determine what is wrong.

Clinical legibility

What is presented during an appointment is limited to the issue, material facts, functional consequences, unresolved questions, and decision at hand.

Noncompulsory participation

The patient who cannot or does not wish to maintain such infrastructure remains entitled to competent care. The archive is a support, not an entrance examination.

This model distributes responsibility without dissolving it. The patient contributes continuity, lived consequences, corrections, priorities, and informed questions. Clinicians contribute examination, measurement, differential diagnosis, treatment expertise, and professional accountability. Records provide durable but imperfect evidence. Research constrains general claims. Technology supports retrieval, comparison, and translation. No component becomes trustworthy by declaring itself complete.

XI. The Discipline of Not Knowing

Several explanations remain plausible.

Perhaps the lower-limb alignment contributes to the gait and to compensations above or below the knees. Perhaps a structural or functional limb-length difference affects pelvic position. Perhaps the spinal curve changes apparent mechanics. Perhaps rotational anatomy alters the visual alignment. Perhaps a device could improve a measured distal component. Perhaps physical therapy could reduce the functional burden without changing bone geometry. Perhaps adult nutritional compromise has affected bone or muscle around a much older developmental pattern. Perhaps some findings are causally related and others are not.

Plausibility is not confirmation.

This is where artificial intelligence is most useful and most dangerous. It can retain several hypotheses, identify missing evidence, and formulate questions. It can also satisfy the desire for closure by converting anatomical possibility into narrative inevitability.

A trustworthy system must preserve the unresolved.

The available spinal comparison does not indicate dramatic recent change. That is good news. It does not establish origin.

The bone findings warrant serious attention in the context of malabsorption and nutritional deficiency. They do not explain a childhood deformity.

The gait is observable and the knee appearance longstanding. Neither fact identifies the precise anatomical source or guarantees that changing one measurement would normalize movement.

The stranger’s question was real, as was the distress it produced. Neither determines the medically appropriate intervention.

Technology has not solved my body. It has helped me represent the problem more faithfully. I can separate documentation from inference, preserve results across specialties, detect contradictions, understand enough terminology to ask which measurements matter, and refuse false precision. I can ask whether an intervention targets structure, pain, movement, risk, or appearance.

I can also be wrong.

That possibility is not a defect in self-advocacy. Mature advocacy does not make the patient immune to contradiction; it creates conditions in which disagreement can be tested against evidence rather than settled through hierarchy alone.

Conclusion: The Right to an Intelligible Body

A person should not require professional medical training to preserve the continuity of their own health. Nor should that continuity be surrendered entirely to institutions whose records divide the body into visits, specialties, and incompatible systems.

Digital access permits a different form of participation. Records can be gathered. Technical language can be translated. Findings can be compared across time. Questions can be prepared before appointments rather than improvised under pressure. Errors can be noticed. Options and probabilities can be interrogated.

These capabilities reduce interpretive friction while introducing new hazards: privacy exposure, false coherence, information overload, compulsive surveillance, clinician burden, and unequal access. Technology does not eliminate the politics of credibility. It creates another arena in which credibility is produced and contested.

My body remains partly unexplained. That is not permission to invent an answer. It is a reason to improve the method of inquiry.

The method begins with the body made public: a gait noticed by a stranger before it has been fully interpreted by medicine. It continues through source records, comparative measurement, constrained hypotheses, and clinical dialogue. It is sustained by enough psychological hope to treat the future as worth protecting. It is bounded by an ethic in which the pursuit of function does not become compulsory normalization. Within a Christian framework, it is deepened by the conviction that dignity precedes symmetry, independence, health, and cure.

I am still trying to change it.

The object of change, however, is no longer only the visible body. I am trying to change the conditions under which it becomes intelligible: scattered results into usable continuity, embarrassment into inquiry, uncertainty into discriminating questions, and technological access into disciplined participation.

Some physical change may be possible. Some may not. Some may not justify its cost. The ethical achievement is not guaranteed correction. It is the recovery of sufficient authority to ask, with scientific seriousness and without self-contempt, what form of care this body actually requires.

The body does not need to become typical before it becomes intelligible.

It does not need to become corrected before it becomes worthy of care.

And it does not need a perfectly unified explanation before its history deserves to be held together.

---

References

Agweyu, Ambrose, et al. 2026. “Generative AI-Enabled Clinical Decision Support System in Primary Care: A Pragmatic, Cluster-Randomized Trial.” Nature Medicine. doi:10.1038/s41591-026-04503-6.

Alfuth, Martin, Patrick Fichter, and Axel Knicker. 2021. “Leg Length Discrepancy: A Systematic Review on the Validity and Reliability of Clinical Assessments and Imaging Diagnostics Used in Clinical Practice.” PLoS ONE 16 (12): e0261457. doi:10.1371/journal.pone.0261457.

Bean, Andrew M., et al. 2026. “Reliability of LLMs as Medical Assistants for the General Public: A Randomized Preregistered Study.” Nature Medicine 32: 609–615. doi:10.1038/s41591-025-04074-y.

Bell, Sigall K., et al. 2020. “Frequency and Types of Patient-Reported Errors in Electronic Health Record Ambulatory Care Notes.” JAMA Network Open 3 (6): e205867. doi:10.1001/jamanetworkopen.2020.5867.

Charon, Rita. 2001. “Narrative Medicine: A Model for Empathy, Reflection, Profession, and Trust.” JAMA 286: 1897–1902. doi:10.1001/jama.286.15.1897.

Charon, Rita. 2006. Narrative Medicine: Honoring the Stories of Illness. New York: Oxford University Press.

Creamer, Deborah Beth. 2009. Disability and Christian Theology: Embodied Limits and Constructive Possibilities. Oxford: Oxford University Press.

Demiris, George, et al. 2019. “Patient Generated Health Data Use in Clinical Practice: A Systematic Review.” Nursing Outlook 67 (4): 311–330. doi:10.1016/j.outlook.2019.04.005.

Eiesland, Nancy L. 1994. The Disabled God: Toward a Liberatory Theology of Disability. Nashville: Abingdon Press.

Frank, Arthur W. 2013. The Wounded Storyteller: Body, Illness, and Ethics. 2nd ed. Chicago: University of Chicago Press.

Fricker, Miranda. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.

Garland-Thomson, Rosemarie. 2011. “Misfits: A Feminist Materialist Disability Concept.” Hypatia 26 (3): 591–609. doi:10.1111/j.1527-2001.2011.01206.x.

Homan, Kristin J., and Fuschia M. Sirois. 2017. “Self-Compassion and Physical Health: Exploring the Roles of Perceived Stress and Health-Promoting Behaviors.” Health Psychology Open 4 (2). doi:10.1177/2055102917729542.

International Society for Clinical Densitometry. 2023. “Official Adult Positions.” ISCD.

Kafer, Alison. 2013. Feminist, Queer, Crip. Bloomington: Indiana University Press.

Lin, Mei-Yu, et al. 2020. “Effects of Patient Activation Intervention on Chronic Diseases: A Meta-Analysis.” Journal of Nursing Research 28 (5): e116. doi:10.1097/JNR.0000000000000387.

Luís, Nuno Marques, and Rui Varatojo. 2021. “Radiological Assessment of Lower Limb Alignment.” EFORT Open Reviews 6: 487–494. doi:10.1302/2058-5241.6.210015.

Nakanishi, H., et al. 2024. “Evaluation of Long-Term Nutrition Outcomes After Duodenal Switch: A Systematic Review and Meta-Analysis.” The American Surgeon. PMID 37694730.

National Institute of Standards and Technology. 2024. Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. NIST AI 600-1. doi:10.6028/NIST.AI.600-1.

Norman, Alyson, and Timothy P. Moss. 2015. “Psychosocial Interventions for Adults with Visible Differences: A Systematic Review.” PeerJ 3: e870. doi:10.7717/peerj.870.

Salte, Odd Bjørn Kjeldaas, et al. 2024. “Ten-Year Outcomes Following Roux-en-Y Gastric Bypass vs Duodenal Switch for High Body Mass Index: A Randomized Clinical Trial.” JAMA Network Open 7 (6): e2414340. doi:10.1001/jamanetworkopen.2024.14340.

Stacey, Dawn, et al. 2024. “Decision Aids for People Facing Health Treatment or Screening Decisions.” Cochrane Database of Systematic Reviews, CD001431. doi:10.1002/14651858.CD001431.pub6.

Vogt, Bernd, et al. 2020. “Leg Length Discrepancy—Treatment Indications and Strategies.” Deutsches Ärzteblatt International 117: 405–411. doi:10.3238/arztebl.2020.0405.

Walker, Jan, et al. 2019. “OpenNotes After 7 Years: Patient Experiences with Ongoing Access to Their Clinicians’ Outpatient Visit Notes.” Journal of Medical Internet Research 21 (5): e13876. doi:10.2196/13876.

World Health Organization. 2024. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multi-Modal Models. Geneva: World Health Organization.