Welcome back to the journal review. This is the May twenty-twenty-six issue of the Journal of the American Academy of Dermatology, and we've got four short pieces to get through today — three from the Ethics Journal Club section and one letter-and-response exchange on a surgical device. None of these are original studies with methods and results in the traditional sense, so we'll move through them in their actual form: three "Dear Dr Dermatoethicist" columns and one reply-to-a-letter. Let's get into it.

First up, an ethics column on managing skin cancer in a patient with advanced dementia. The setup is one you've probably lived through clinically: a seventy-seven-year-old man with advanced dementia has a large, ulcerated basal cell carcinoma on the shoulder. He's cognitively unaware of the lesion itself, but he pulls at dressings and becomes agitated during wound care. The question is how to think through this ethically, not just clinically.

The column frames this through the four classic principles — nonmaleficence, justice, autonomy, and beneficence — but the interesting part is how each gets reinterpreted once decision-making capacity is off the table. Justice, they argue, means this patient deserves the same individualized workup and treatment consideration as any other patient, dementia notwithstanding — age or cognitive status alone isn't a reason to default to neglect. Autonomy shifts to what they call autonomy-by-proxy: the power of attorney applying substituted judgment, meaning they try to reconstruct what this specific patient would have wanted based on his prior values and tolerances, and only falling back to a best-interest standard if that history isn't available. And they're careful to note that caregiver burnout and wound-care distress, while real, don't by themselves justify a treatment decision — that only becomes ethically relevant when it starts to affect the patient's own safety, dignity, or comfort.

The clinical hook here is proportionality — weighing benefit against burden. And the column makes a point that's worth sitting with as a Mohs surgeon: for a basal cell carcinoma, which essentially never metastasizes, the long-term benefit of definitive cure may simply be irrelevant if the burden of getting there — sedation, restraint, procedural agitation, and the real risk of precipitating delirium — is high. Delirium in dementia patients is flagged as common, often not fully reversible, and capable of causing lasting functional decline even from outpatient procedures. That's a genuinely useful framing for any of us tempted to push for margin-clear excision reflexively.

Practically, they suggest a stepwise, least-burden-first approach: start with non-invasive palliative measures like hemostatic dressings and topical agents for bleeding and odor control, time wound care around the patient's calmer periods, and treat the whole thing as a time-limited trial with predefined success criteria agreed upon with the proxy. Only if that fails and there's reason to think the patient would have accepted the trade-off do you escalate — local anesthesia for smaller lesions, general anesthesia only if truly necessary for larger ones. And if every option still nets more burden than benefit, the shift to comfort care is the ethically correct move, not a failure of care.

For practice, this isn't practice-changing in the sense of new data, but it's a genuinely useful decision framework to have internalized before you're standing in the room with an agitated patient and a distressed family member — particularly the point that cure is not always the correct endpoint for a low-lethality tumor in a patient with limited life expectancy and high procedural burden.

Second column: the ethics of who actually discloses dermatopathology results to patients — specifically, is it appropriate for medical assistants to be the ones delivering biopsy results.

The tension is structural. The Twenty-First Century Cures Act mandates that clinical information gets released to patients essentially immediately, usually via electronic portal, but pathology reports are often unintelligible without physician interpretation, and dermatologists' time is finite. The column cites survey data to frame the tradeoffs: roughly half of patients want rapid result communication, but under one in ten specifically value the chance to ask questions in that moment — suggesting a real split in what patients actually want. Separately, when dermatologists were asked who they'd prefer to disclose results, more than half wanted to personally deliver malignant results themselves, versus only about a third feeling that strongly about benign ones — and in both cases, an assistant was the next preferred communicator.

There's a nice piece of nuance around delivery format too — patients' preferences shift depending on what's being disclosed. A meaningful chunk preferred voicemail for benign results, but that preference flipped toward wanting an actual phone call once the result was malignant. And this matters clinically because a pigmented lesion clinic study found roughly a third of biopsy patients carry high anxiety both before and after getting their result — the waiting period itself is a distinct source of distress, separate from the diagnosis.

The ethical resolution the column lands on is basically a tiered communication model: benign, straightforward results can reasonably be delegated to trained staff working from dermatologist-created scripts, provided the patient has consented to that format — respecting autonomy while preserving efficiency and access for other patients. But anything malignant or complex — anything life-altering — should come from the dermatologist directly, so questions and psychosocial concerns can be addressed in real time. And they flag that MAs themselves have reported discomfort and feeling unprepared to field follow-up questions, which is itself a reason not to push disclosure downstream without training and support.

Practical takeaway: worth an actual look at your own clinic's disclosure protocol — asking patients up front how they want results delivered, and building explicit scripts and consent processes around who says what for which result type, is a low-cost, high-yield practice change even though nothing here is a randomized comparison.

Third piece is a letter-and-response exchange, not original research — a reply from the original device authors to a prior commentary on their negative pressure drainage device for skin wound management. This is a technical, back-and-forth engineering discussion, and it's brief. The commenting group had proposed a reverse-assembly modification of the original syringe-based drainage device, and the original authors here are responding graciously, essentially negotiating the tradeoffs. Their take: the reverse assembly does offer better stability, but their original configuration allows easier one-handed plunger placement, which matters practically at chairside. They concede that adhesive tape is still needed regardless of assembly method to prevent the plunger from dislodging. On tube caliber, they note a thicker drainage tube suits some wounds better, while their original narrower tube may actually be preferable for small, deep wound configurations — essentially arguing that device configuration should be tailored to wound anatomy rather than standardized. There's no data here, no outcomes — it's a conceptual reconciliation between two groups iterating on a low-cost, self-made device. The only real takeaway for practice is awareness that this device family exists and is being refined in the literature if you're looking for inexpensive negative-pressure options for wound management, with the caveat that tube size and assembly orientation should be matched to the specific wound rather than treated as one-size-fits-all.

Last article: an ethics column on artificial intelligence in dermatologic diagnosis, prompted by a case where an AI triage tool categorized a pigmented lesion as low risk, leading to a six-to-nine month wait for a teledermatology follow-up. By the time the patient was seen, the lesion had grown, and biopsy confirmed melanoma requiring urgent intervention.

The column situates this within what we already know about AI performance in dermatology: strong, fairly consistent accuracy for things like psoriasis and other inflammatory dermatoses using convolutional neural networks, but for melanoma specifically, diagnostic accuracy is described as highly variable and genuinely controversial, heavily dependent on the quality of the training dataset and study design. The core ethical point is that a good statistical performance metric in a validation paper does not translate automatically into clinical safety in your actual patient population.

They bring up liability directly, which is worth flagging for anyone integrating these tools: diagnostic error already accounts for roughly eight to ten percent of dermatology malpractice claims, and AI doesn't remove that risk — it just muddies who's responsible when a misdiagnosis happens, since accountability could plausibly sit with the clinician, the institution that approved the tool, or the algorithm developer, and there's no settled legal precedent yet. There's also the now-familiar equity problem — most training datasets over-represent lighter skin tones, so accuracy predictably drops for patients with skin of color — and a practical blind spot where these tools aren't built to reliably catch atypical presentations of things like contact dermatitis, especially with confounding environmental exposure or infection. And critically, when a tool spits out a categorical answer without a confidence score or differential, it invites anchoring bias — clinicians and trainees stop challenging an output even when the exam in front of them doesn't fit.

Their proposed fix is systems-level, not technological: explicit institutional policy on documenting AI use, defined supervision requirements for trainees using these tools, formal algorithm approval criteria before deployment, and structured AI-literacy and bias-recognition training built into education. The core message is that AI should function strictly as an adjunctive decision aid, never as an autonomous diagnostician, and the clinician retains ultimate responsibility for the outcome regardless of what the algorithm said.

For practice, this is directly relevant if your group is adopting or already using AI-based triage — it's a strong argument for insisting on documented human override authority, confidence-score transparency, and institutional sign-off before any AI tool is allowed to independently gate access to urgent evaluation, rather than treating a "low risk" output as equivalent to a clinical exam.

That wraps our four pieces for this issue — three ethics columns touching dementia care, result disclosure, and AI triage, plus a brief technical exchange on wound drainage devices. Thanks for listening, and we'll see you next month.