Online patient reviews of dermatologic surgeons differ by gender, study finds

Online patient reviews have become a powerful force in modern medicine, shaping how prospective patients choose their doctors and how physicians build—or lose—their reputations. Now, a new study suggests that the language patients use to describe their dermatologic surgeons differs in striking ways depending on the surgeon’s gender, adding fresh evidence to a growing body of research indicating that gender bias quietly permeates the digital评价 ecosystems of healthcare.

A team of researchers at The Ronald O. Perelman Department of Dermatology at New York University Grossman School of Medicine set out to determine whether the publicly available reviews of dermatologic surgeons reflect the same gendered patterns documented in other medical and surgical specialties. Their findings, published as a research letter in Archives of Dermatological Research, reveal measurable differences in how patients evaluate male and female dermatologic surgeons online—differences that persist even in a field where women now make up a substantial majority of practitioners.

The research team, led by co-first authors Derek Maas and Archie Spindler, with Kristen I. Lo Sicco and Maressa C. Criscito serving as co-senior authors, drew on a well-established body of prior work. Previous studies have shown that physician rating websites are widely consulted by patients before making appointment decisions, and that the content of reviews varies by physician gender. A 2025 analysis published in JAMA Network Open found that patient perceptions of physicians’ interpersonal and technical skills differed systematically by physician gender in online reviews, while other work has shown that patients writing about female physicians more frequently mention communication style, empathy and appearance-related attributes, whereas reviews of male physicians more often emphasize technical competence and confidence. In surgery specifically, a large 2023 study in JAMA Surgery examining long-term postoperative outcomes among patients undergoing common surgeries found that surgeon sex was associated with differences in outcomes—adding complexity to the question of whether patient perceptions track objective performance or reflect ingrained stereotype.

Against this backdrop, the NYU team focused on dermatology, a specialty that occupies a unique position at the intersection of medicine, aesthetics and surgery. Dermatologic surgeons perform procedures ranging from Mohs micrographic surgery for skin cancer to cosmetic interventions such as laser resurfacing, fillers and excisions. Because these procedures directly alter appearance, patient satisfaction is influenced not only by clinical outcomes but also by subjective expectations about beauty, trust and bedside manner—territory in which gendered assumptions may be especially potent.

To conduct the analysis, the researchers collected publicly available online patient reviews of dermatologic surgeons. Rather than relying solely on crude metrics such as star ratings, the team employed a hybrid methodology combining human judgment with modern natural language processing. Trained human coders hand-labeled a subset of reviews according to the themes and attributes they contained, and those hand-coded data were then used to train RoBERTa—a robustly optimized, transformer-based language model that represents a widely used evolution of the BERT architecture. RoBERTa, first described by researchers in 2019, learns contextual word representations by masking and predicting tokens across massive text corpora, and when fine-tuned on a labeled dataset it can classify incoming text with high accuracy. This approach allowed the investigators to scale their qualitative content analysis across the full review corpus while maintaining consistency in how themes were defined and applied.

The use of machine learning in this context illustrates a broader trend in health services research: the pairing of qualitative content analysis—long the gold standard for understanding what patients actually say—with supervised learning algorithms capable of processing thousands of documents. In their ethics declarations, the authors note that no human participants were involved and that informed consent was not required, since the analysis was confined to publicly posted content. All findings were derived from publicly available data, and the analytic materials may be shared upon reasonable request to the corresponding author.

The study’s central finding is that the content of online reviews differs by the gender of the dermatologic surgeon being reviewed. While the specific dimensions of difference echo patterns identified in prior physician review research, their presence in dermatologic surgery is notable for several reasons. Dermatology is a specialty in which female physicians are heavily represented, meaning that gendered review patterns have the potential to affect a large share of the workforce. Moreover, because online ratings increasingly influence patient flow, referral patterns and even institutional hiring and promotion decisions, systematically different review content could translate into unequal reputational capital for male and female surgeons performing comparable work at comparable quality.

The implications extend beyond reputation. A growing literature suggests that patient perceptions, though subjective, are not trivial. They influence adherence, satisfaction and the likelihood of seeking timely care. They also feed back into the professional ecosystem: surgeons with stronger online profiles may attract more patients, more complex cases and more opportunities for academic advancement. If reviews of female surgeons disproportionately emphasize interpersonal attributes while reviews of male surgeons emphasize technical ones—or if similar behaviors are described in different terms depending on the surgeon’s gender—the cumulative effect may be a subtle but persistent penalty or premium attached to gender rather than skill.

The new study also sits within a larger scientific conversation about whether gender differences in surgical outcomes and perceptions reflect performance or perception. The 2023 JAMA Surgery study by Wallis and colleagues, which followed patients undergoing common surgical procedures, reported associations between surgeon sex and long-term postoperative outcomes, fueling debate about whether such differences stem from practice patterns, patient mix, communication styles or bias in how care is delivered and evaluated. Studies of online reviews add a complementary layer: they measure not what surgeons do, but what patients perceive and choose to record. Both layers matter, because perception drives demand even when objective outcomes are equivalent.

The NYU researchers’ methodological choices deserve attention from anyone interpreting the results. Hand-coding ensures that categories such as “interpersonal skill,” “technical skill,” “wait time,” “office staff” and “appearance-related outcomes” are defined by humans with domain expertise rather than by opaque statistical associations. Fine-tuning RoBERTa on those human judgments allows the labeled categories to be applied at scale, reducing coder fatigue and inter-rater drift. The approach is not without limitations—language models can inherit biases present in training labels, and online reviewers are not a representative sample of all patients—but the combination of human curation and machine scale has become a standard for extracting signal from the sprawling, unstructured text that patients generate.

The authors received no external financial support for the research, and the disclosed conflicts of interest relate to unrelated consulting and investigator relationships in dermatology therapeutics. The study was received in February 2026, revised in July, and accepted in August before publication on 9 September 2026 in volume 318 of the journal.

For dermatologic surgeons, the message is twofold. First, online reviews are not gender-neutral mirrors of care; they are texts shaped by the same social expectations that shape face-to-face interactions. Second, because these texts increasingly function as public scorecards, surgeons and the institutions that train and employ them may need to contextualize review data rather than take it at face value. For patients, the findings are a reminder that the adjectives attached to a doctor’s name online may say as much about cultural expectations of gender as about the doctor.

For researchers, the study opens several avenues. Extending the RoBERTa-based classification to other surgical specialties, to multiple review platforms and to longitudinal data could reveal whether gendered language patterns are strengthening or fading as generations of patients and physicians change. Linking review content to objective outcome measures—complication rates, revision rates, patient-reported outcome instruments—could test whether perception gaps correspond to performance gaps or exist independently of them. And methodologically, comparing multiple large language models and coding schemes would help establish how robust the observed differences are to analytic choices.

What is already clear is that the digital waiting room—the constellation of stars, comments and ratings that patients scroll through before booking—reflects social reality as much as clinical reality. In dermatologic surgery, where the product is the patient’s own visible skin, that reflection may be sharper than anywhere else in medicine. This study provides a quantitative, machine-assisted reading of that reflection, and it suggests that even in one of medicine’s most female-dominated surgical fields, the way patients talk about their doctors still depends on who the doctor is.