Telemedicine: reasoning without an in-person exam

The consultation begins without the handshake, without the smell of the room, without the walk to the examination table. On the screen, you see the face, hear the voice and, often, lose the complete physical examination. Telemedicine does not eliminate clinical reasoning; it redistributes what you can observe, what you need to ask and what you should schedule in person. This text deals with that redistribution: how to think through hypotheses and management when the in-person examination is absent or incomplete.

NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.

The topic is neutral and international. Pandemic experiences accelerated its use in dozens of countries; the cognitive logic is the same: less examination data, more weight on history, risk triage and the follow-up plan. We link to empathy in the consultation, differential with clues, calibration and diagnostic pause where appropriate, without repeating those h2s.

Key points

• Telemedicine reduces information from the physical examination and from context; reasoning must compensate with a structured history, red flags and explicit criteria for going in person.
• Classic reviews find stronger evidence of diagnostic agreement in areas such as dermatology and psychiatry; in other specialties the evidence is more heterogeneous.
• Clinical judgment in virtual care for complex chronic disease requires time, a script and awareness of the modes of reasoning (abduction, deduction, induction).
• Audio and video quality, lighting and the patient's environment are part of the clinical data, not a minor technical detail.
• Training with incomplete clues (as in NEXO) brings the mental habit closer to the teleconsultation: deciding the next step without the complete examination in front of you.

What changes when the examination is missing

In person, you collect history and examination in parallel: the walk, the color, the respiratory effort, the smell, the posture. On a video call, part of this arrives distorted or does not arrive at all. Reasoning must make explicit what is missing. Instead of assuming "normal examination because I saw no abnormality", record "physical examination incomplete due to modality".

A well-conducted teleconsultation is not a compressed in-person visit. It is another design: pre-data (form, home measurements, photos), a focused encounter, a post-plan (when to return, what to watch for, when to go to the emergency department). Authors who discuss diagnostic quality in the era of telemedicine emphasize the pre- and post-encounter, not only the minutes of video (discussion on diagnostic quality in primary care and telemedicine).

Management reasoning (deciding on treatment, follow-up, tests and resource allocation) gains prominence when the definitive diagnosis does not close in the virtual encounter. Accepting calibrated uncertainty and combining a safety plan is management, not weakness.

What the evidence says about diagnosis

An older systematic review on the effectiveness of telemedicine for diagnostic and management decisions concluded that the strongest evidence was in psychiatry and dermatology; in other areas there were fewer high-quality studies, despite wide use (Hersh et al., 2002). Decades later, the map has expanded, but the lesson remains: agreement with in-person care is not uniform across specialties and tasks.

Reviews on observer agreement in diagnosis by videoconference found support especially in dermatology and care signals in specialties where fine physical examination weighs more (Martin-Khan et al., 2011). In telepsychiatry, a recent meta-analysis suggested high agreement for several diagnoses and scales, with nuances by type of disease and modality (review telepsychiatry vs in person).

The message for the generalist clinician in training: do not generalize "telemedicine gets it right just the same" or "telemedicine is unsafe". Ask: which task? Which specialty? Which signal quality? Which plan if the uncertainty remains?

Dimension	In person	Typical teleconsultation	Implication for reasoning
Physical examination	Extensive	Limited / guided	Name the gaps; ask for possible maneuvers
Sensory context	Rich	Filtered by the camera	Ask what the camera does not show
Useful time	Variable	Risk of technical loss	Pre-data and script
Follow-up	Return scheduled	Must be explicit	Written safety plan
Diagnostic agreement	Reference	Varies by area	Do not extrapolate dermatology to everything

History, red flags and plan

With less examination, the history carries more weight. Structure it: onset, course, factors that improve and worsen, associated symptoms, medications, allergies, past history, relevant social context. Then go through the red flags of the condition (neurological, cardiovascular, infectious, etc.) without turning the consultation into a punitive interrogation.

Ask for what the patient can show: a counted respiratory rate, edema with a sock mark, a rash with a photo under good light, blood pressure and glucose readings if there is a validated device. Treat home numbers as hypotheses, not as laboratory truth.

Always close with a plan: what to do in the next hours or days; warning signs for seeking emergency care; whether and when there will be a virtual or in-person return; what remains open. This closing is applied management reasoning.

Reasoning: abduction, deduction, induction

In an essay on clinical judgment in virtual care for complex chronic disease, the author argues that the reduction of information pushes even experienced specialists toward more fragile modes of inference if the encounter is too short (Enhancing clinical judgement in virtual care, 2021). Abduction (generating the best explanation in the face of the unfamiliar), deduction (applying a rule to a case) and induction (adjusting over time with new data) must be conscious.

In teleconsultation practice: if the condition is familiar and stable, protocol-guided deduction may suffice, with a follow-up plan. If the condition is confusing or new, state the abduction ("the best explanation now is X, but Y is not excluded") and reserve induction for the return with new data. Insufficient time pushes toward early closure: the same enemy of in-person reasoning, now with less examination to correct the course.

Eva and Norman, in classic texts on reasoning and error, recall that dual process and context matter (Eva, 2005). In telemedicine, the context includes the modality. Treating the modality as invisible is bias.

Technical limits that become error

Poor audio, pixelation, backlight, a patient in motion, a companion talking over them, a connection that eats five minutes of a short schedule: all of this reduces clinical data. Platforms designed for social meetings were not designed for high-fidelity diagnostic observation; discussions of diagnostic quality during the pandemic pointed to exactly this risk.

The clinician must decide early: does the technical quality allow continuing safely, or should the encounter become triage for in-person or emergency care? Insisting on diagnosing with a useless signal is a process error. Saying it out loud to the patient ("I cannot assess respiratory effort with this image; we need an in-person assessment today") is clinical reasoning, not technological failure.

Greenhalgh and colleagues mapped the advantages and limits of online consultations (the VOCAL study and related work), including questions of suitability, relationship and workflow (VOCAL, BMJ Open 2016). The persistent lesson: the modality is a clinical choice, not an administrative default.

When to convert to in-person

Guiding criteria (adapt to local rules and to the specialty):

1. A warning sign that requires an examination or intervention the screen does not allow.
2. Diagnostic uncertainty with relevant risk if the dangerous hypothesis is correct.
3. A technical failure that prevents minimal assessment.
4. A patient or caregiver unable to provide reliable data at a distance.
5. A condition that has worsened since the last virtual contact without a clear explanation.

Converting is not a defeat. It is the telemedicine equivalent of the diagnostic pause: stopping before closing the case with insufficient data. See also the logic of the diagnostic pause.

Document the gap and the reason for the conversion. This protects the patient and the clinical trail. "Patient advised to go to urgent care for chest pain with dyspnea and video inadequate to assess effort" is better than a generic "return if worse".

How to train this reasoning

Students rarely have formal discipline in "reasoning without examination". It can be simulated. Ask a colleague to describe only what a camera would see; generate hypotheses and the next step; only then receive the full physical examination and see what would change. Practice video-guided maneuvers (where to place the hand, how to show the throat) in a supervised setting.

Use vignettes with missing data on purpose. Training differential with clues is a close relative: incomplete information, deciding the next step, calibrating confidence. Calibration matters even more when the examination is gone: high confidence without data is a risk.

Empathy on screen requires explicit effort: look at the camera, validate the complaint, avoid typing in long silence without warning. The post on empathy helps; in a teleconsultation, empathy also means stating the limits of the modality.

Mental checklist before a virtual consultation: goal of the encounter; prior data read; what absolutely must be seen or heard; plan if the technique fails; conversion criteria. Two minutes of this preparation reduce dangerous improvisation.

During: narrate what you are doing ("I will ask you to move closer to the light"); ask for confirmation of understanding; avoid piling up jargon. After: send short written instructions; confirm a channel for questions; record gaps.

Ethical and legal limitations vary by country (prescription, privacy, consent, jurisdiction). This text does not replace local regulation. Before practicing, know the rules of your jurisdiction and your institution.

How NEXO helps

NEXO presents fictional cases with a realistic clinical presentation and up to six clues. The habit of closing a hypothesis and a next step with incomplete information is structurally similar to a teleconsultation. Use the case of the day to rehearse: what you would ask further; what would require an in-person examination; what safety plan you would write.

There is no promise that playing cases improves outcomes in telemedicine. There is alignment of method: tolerating uncertainty, asking for the right clue, avoiding early closure. NEXO Researcher and NEXO Closer programs go deeper into method and communication; limits in Economy.

For teachers: virtual OSCE stations, role-play of technical failure and of conversion to in-person teach what a lecture on "telemedicine is the future" does not teach. Assess the process (did the student name gaps? did the student make a safety plan?) beyond the diagnostic label.

For curriculum managers: including telemedicine only as an access tool without teaching reasoning under reduced data builds fragile habits. The minimum is supervised practice with feedback on gaps and conversion.

Relation to diagnostic error: less examination increases the chance of a miss if the clinician does not compensate. The answer is not to ban the modality; it is to use it where the task fits and convert where it does not. Reviews of diagnostic effectiveness have pointed to this heterogeneity for decades (Hersh 2002).

In complex chronic disease, virtual follow-up can shine in fine-tuning and adherence, as long as there are periodic in-person anchors and clear criteria for worsening. The essay on virtual clinical judgment emphasizes time, a multiprofessional team and a prior script as protectors of reasoning (PMID 33559390).

Avoid two rhetorical extremes. Extreme one: "everything can be virtual". Extreme two: "nothing is safe at a distance". Adult clinical reasoning chooses the modality as it chooses a test: by the value of the information and the risk of error.

When training alone, record (with consent in a real teaching context) or take notes after the consultation: what I could not examine; what I asked the patient to show; what remained open; what I would convert today if I could go back in time. This short diary is metacognition applied to the modality.

Finally, remember that the patient also reasons. Explain why you are asking for an in-person visit. Transparency reduces the feeling of rejection ("the doctor did not want to see me") and turns the conversion into a shared plan.

There is also the reasoning about who is in the room. A helpful companion can complete the story; a dominant companion can silence the patient. On screen, explicitly ask to hear the patient. In pediatrics and geriatrics, the dynamics of another person's voice are even more critical.

Medications in a teleconsultation deserve a label read via camera when there is doubt, and careful reconciliation. Look-alike name errors do not decrease at a distance; sometimes they increase with poor audio. Ask the patient to show the box, not just cite it from memory.

Remote mental health has growing evidence of diagnostic agreement in several conditions, but risk, suicide and the need for a local emergency network require an explicit protocol before the encounter. Again: seek the help needed and use local channels when risk is high. This blog does not list phone numbers.

Teaching: ask the resident, at the end of each supervised teleconsultation, to state in one sentence the hypothesis, the main examination gap and the trigger for conversion. Three elements. Feedback on these three elements teaches the skeleton of remote reasoning better than generic speeches about innovation.

Future research still needs more high-quality designs outside the areas already studied. Meanwhile, clinical prudence is modular: use the modality where the task fits; convert where the risk of the gap outweighs the benefit of access. Access without diagnostic safety is not a victory.

If you are setting up a service, define menus of complaints suitable and unsuitable for teleconsultation, train the triage team, and audit conversions and sentinel events. Individual clinical reasoning does not hold up a fragile institutional design on its own.

For the student who only sees in-person care during the internship: ask for shifts or electives that include supervised telehealth. Without exposure, the first remote job becomes learning under fire. With reflected exposure, it becomes competence.

Also integrate what you already know about heuristics. Anchoring on the initial complaint can be stronger in a teleconsultation because there is less examination to challenge the anchor. Force a minimum written differential before closing. The post on heuristics helps name the mechanism.

Consider the continuum of modalities: asynchronous message, phone, video, in-person, emergency. Each step adds data and cost. Moving up or down a step is a clinical decision. Moving up too early wastes resources; moving up too late delays care. Mature telemedicine reasoning is, to a large extent, choosing the step.

Asynchronous (dermatology photo, questionnaire) works well when the task is standardized and the risk is low. Video helps when interaction and dynamic observation matter. In-person remains necessary when examination, procedure or severity demand it. Teaching this continuum avoids the false dichotomy of "virtual versus real".

Privacy: ask the patient whether they are in a place where they can talk; headphones, a closed room, companions. Sensitive data on a bus speakerphone is a risk. This is also clinical data: if the patient cannot speak freely, the history becomes biased.

Bias and equity: those without broadband, a smartphone or privacy at home are left out. Telemedicine widens access for some and narrows it for others. When interpreting evidence of diagnostic agreement, remember who was included in the studies. Digitally excluded populations do not appear in the success rates.

For day-to-day reasoning, a minimal script fits on a card: (1) confirm identity and location; (2) reason and perceived severity; (3) targeted history; (4) what to ask the patient to show; (5) red flags; (6) hypothesis and uncertainty; (7) plan and triggers; (8) documentation of the modality. Repeating the script until it becomes a habit reduces omission under pressure.

When the case is stable chronic follow-up, the teleconsultation may be the right step and periodic in-person visits the anchor. When the case is new and undifferentiated, the bar for staying only virtual rises. Undifferentiated plus risk equals presence, until proven otherwise.

Always close the learning loop: if you converted and the in-person visit changed everything, note what the screen did not show. If you did not convert and there was an adverse event, review without blaming only the technology. Process analysis improves the next encounter.

This post does not teach how to operate specific software or how to bill. It teaches the cognitive skeleton. Software changes; the question "what data am I missing and what do I do with that gap?" remains.

In a remote interprofessional team, state roles: who collects signs, who decides on conversion, who documents. Role ambiguity is a latent error. Remote clinical reasoning is also team reasoning.

If you are on the teaching side and only assess factual knowledge about telemedicine, you are measuring the wrong step of Miller's pyramid. Ask for performance: a simulated station with video failure, a reticent patient, a poorly lit rash. Observe whether the student names the gap. That observation is worth more than an MCQ about the definition of telehealth.

Frequently asked questions

Does telemedicine replace the physical exam?

No. In many encounters the exam stays incomplete. The reasoning must name the gap and decide whether the gap is acceptable for the task.

In which areas is agreement usually better?

Classic reviews point to more strong evidence in dermatology and psychiatry; generalize with caution to other tasks.

What should I do if the connection drops in the middle of a serious case?

Treat it as an assessment failure. Direct the patient to an emergency channel according to the situation and document the interruption.

How can I train before remote care?

Simulations with incomplete information, video role-play and cases with limited clues. NEXO is one of the possible formats.

Can I close a diagnosis by phone alone?

It depends on the task, the risk and local rules. The higher the risk of a dangerous hypothesis, the greater the need for in-person data.

How does NEXO fit into this?

As training in decision-making under incomplete clues, complementary to supervised teleconsultation practice.

Final checklist

• [ ] Exam gaps named
• [ ] Red flags asked about
• [ ] Technical quality assessed
• [ ] Explicit safety plan
• [ ] Clear conversion criteria
• [ ] Written guidance for the patient
• [ ] Record of the modality and its limits

Summarize the virtual encounter in four lines in the medical record: modality and technical quality; hypothesis with degree of uncertainty; exam gaps; plan and triggers for conversion or emergency. Four lines train the thinking and protect continuity. Colleagues who read the record later need to understand what the screen did not show.

Avoid empty innovation jargon. Talking about disruption does not replace red flags. Patients notice when the consultation is rushed because of the platform. If the virtual schedule is too saturated for a minimal history, the problem is in service design, and the clinician needs to flag it, not just "be more efficient".

Finally: well-used telemedicine expands care. Poorly used telemedicine spreads false certainty. The switch between the two positions is explicit reasoning about missing data. This post exists to make that switch visible in the training of students and clinicians in training.

Keep the working phrase: modality is a clinical choice. Every time you accept the default without asking whether the data is enough, you outsource the reasoning to the schedule. Every time you name the gap and act on it, you practice Medicine at a distance with the same rigor as at the bedside. Rigor does not live in the stethoscope; it lives in honesty about what the stethoscope (or the camera) managed and failed to show.

If there is only one practice to take from here: when ending any teleconsultation, say out loud (or write down) the main gap and the conversion trigger. Two seconds. This micro-ritual is more valuable than memorizing definitions of telehealth. Repeat it until it becomes a professional reflex.

With this reflex, distance stops being a dangerous shortcut and becomes a setting for conscious decisions. That is what we separate here from the hype: not the promise that the screen solves everything, but the method for knowing when it is enough and when it is not.

Play the case of the day

Open the case of the day and ask: what would I still need to examine in person before closing? That question is the muscle of the teleconsultation.

References

1. Hersh W, et al. A systematic review of the efficacy of telemedicine for making diagnostic and management decisions. J Telemed Telecare. 2002. PubMed 12217102.
2. Enhancing clinical judgement in virtual care for complex chronic disease. J Eval Clin Pract. 2021. PubMed 33559390.
3. Greenhalgh T, et al. Virtual online consultations: advantages and limitations (VOCAL) study. BMJ Open. 2016. PubMed 26826147.
4. Martin-Khan M, et al. Observer agreement during medical specialist diagnosis using videoconferencing (review). J Telemed Telecare. 2011. DOI 10.1258/jtt.2011.101113.
5. Diagnosis and symptom assessment in telepsychiatry vs face-to-face: systematic review and meta-analysis. 2025. PMC12405819.
6. Ensuring primary care diagnostic quality in the era of telemedicine (discussion). PMC9746257.
7. Eva KW. What every teacher needs to know about clinical reasoning. Med Educ. 2005. PubMed 15612906.
8. Miller GE. The assessment of clinical skills/competence/performance. Acad Med. 1990. PubMed 2400509.
9. Epstein RM. Assessment in medical education. N Engl J Med. 2007. PubMed 17251535.
10. Norman GR, Eva KW. Diagnostic error and clinical reasoning (classic context of the field). Med Educ. 2010 (see related citations in the diagnostic error field).
11. Page G, Bordage G. Key features approach (context of clinical decision assessment). Acad Med / Med Educ (key features literature).
12. van der Vleuten CPM, Schuwirth LWT. Assessing professional competence. Med Educ. 2005 (assessment context).