Gender Differences in Patient Perceptions of Physicians' Communal Traits and the Impact on Physician Evaluations

Informant characteristics influence Clinical Dementia Rating Sum of Boxes scores-based staging of Alzheimer's disease

The Clinical Dementia Rating Sum of Boxes (CDR-SB) is a staging scale for Alzheimer's disease (AD)1 and is commonly used as an outcome in clinical trials2. It relies on information provided by the patient and an informant3. The CDR-SB should reflect only the patient's disease severity. However, we explored whether informant characteristics were associated with CDR-SB scores because that association might introduce bias in Alzheimer's disease research. We found that the CDR-SB was 0.20 higher when informants were female, 0.39 higher when the informant was a patient's child and 0.18 lower if the relationship was other than spouse or children. Regarding the frequency of contact, CDR-SB scores were 0.38 higher when contact was at least once a week, 0.65 higher when daily and 0.57 higher when living with the patient. Our analysis results suggest that informant characteristics can modify the CDR-SB scores and might introduce bias into Alzheimer's disease trials and research.

Assessment of Navy Provider Knowledge of Women's Mental Health Services, Treatment Practices, and Training

INTRODUCTION

As female active duty populations increase in all military environments, it is critical that women's health be addressed in a comprehensive manner. The study's results will be utilized to assist Navy health care leaders in addressing female force readiness policies, treatment gaps, and training specific to women's mental health.

MATERIALS AND METHODS

In total, 212 active duty participants were recruited from the Navy's mental health specialties. The survey was hosted on the U.S. Government's MAX.gov survey website and received Institutional Review Board and Survey Review Board approval per Navy and Defense Health Agency requirements. Participants recorded their experiences with various patient presentations using Likert scale assessments, indicated their comfort in prescribing medication to patients with 11 distinct presenting concerns, and responded to six questions regarding their training and clinical experience in the field of women's mental health.

RESULTS

Differences were noted for provider gender, treatment setting, patient sex, provider rank, and years of independent practice. Female providers were more likely than males to report that their female patients presented with 15 of the 21 measured issues. Providers located at MTFs were significantly more likely than providers in operational billets to report female patients presenting with certain conditions and reported being more comfortable prescribing medication. Eighty percent of respondents authorized to prescribe medicine rated themselves as very or extremely comfortable prescribing medications to their patients for all specified conditions except two: women who are breastfeeding and women who are pregnant. Senior officers reported the most comfort prescribing medication to women who are planning to become pregnant and women who have experienced perinatal loss. Only a minority of providers (20%, female; 33%, males) reported receiving women's mental health education during their training. Of those who did receive training, it was limited to post-partum and pregnancy. Most participants (93%) agreed that women's mental health should be incorporated into training programs for military providers.

CONCLUSIONS

This exploratory study highlights that provider variables impact assessment and treatment of and for patients. The study highlights the interplay of gender, treatment setting, experiences, and level of comfort are associated with provider assessment of presenting concerns. The authors hope this study will help in prioritizing women's mental health practices, mental health training, and research, and in informing policy and decision-making.

Classification of Patients' Judgments of Their Physicians in Web-Based Written Reviews Using Natural Language Processing: Algorithm Development and Validation

BACKGROUND

Patients increasingly rely on web-based physician reviews to choose a physician and share their experiences. However, the unstructured text of these written reviews presents a challenge for researchers seeking to make inferences about patients' judgments. Methods previously used to identify patient judgments within reviews, such as hand-coding and dictionary-based approaches, have posed limitations to sample size and classification accuracy. Advanced natural language processing methods can help overcome these limitations and promote further analysis of physician reviews on these popular platforms.

OBJECTIVE

This study aims to train, test, and validate an advanced natural language processing algorithm for classifying the presence and valence of 2 dimensions of patient judgments in web-based physician reviews: interpersonal manner and technical competence.

METHODS

We sampled 345,053 reviews for 167,150 physicians across the United States from Healthgrades.com, a commercial web-based physician rating and review website. We hand-coded 2000 written reviews and used those reviews to train and test a transformer classification algorithm called the Robustly Optimized BERT (Bidirectional Encoder Representations from Transformers) Pretraining Approach (RoBERTa). The 2 fine-tuned models coded the reviews for the presence and positive or negative valence of patients' interpersonal manner or technical competence judgments of their physicians. We evaluated the performance of the 2 models against 200 hand-coded reviews and validated the models using the full sample of 345,053 RoBERTa-coded reviews.

RESULTS

The interpersonal manner model was 90% accurate with precision of 0.89, recall of 0.90, and weighted F1-score of 0.89. The technical competence model was 90% accurate with precision of 0.91, recall of 0.90, and weighted F1-score of 0.90. Positive-valence judgments were associated with higher review star ratings whereas negative-valence judgments were associated with lower star ratings. Analysis of the data by review rating and physician gender corresponded with findings in prior literature.

CONCLUSIONS

Our 2 classification models coded interpersonal manner and technical competence judgments with high precision, recall, and accuracy. These models were validated using review star ratings and results from previous research. RoBERTa can accurately classify unstructured, web-based review text at scale. Future work could explore the use of this algorithm with other textual data, such as social media posts and electronic health records.

"Women's work": Gender and the physician workforce

Historically, the physician professional identity and the organizational structure of Western medicine have been defined by masculine norms such as authority and assertiveness. The past five decades have seen a rapid shift in the demographics of attendees as medical schools, with equal numbers of women and men matriculants for nearly twenty years. Gender as a social, cultural, and structural variable continues to influence the physician workforce. The entry of women into medicine, has had far reaching effects on the expectations of patients, the interactions of physicians with other members of the healthcare team, and the delivery of care. Redefining the culture of medicine to accommodate the diversity of the modern workforce may benefit all physician and improve the delivery of healthcare.

Language Patterns in Letters of Recommendation for Residency Applicants in Otolaryngology

OBJECTIVE

Assess for gender and race patterns in agentic and communal language used in letters of recommendation for Otolaryngology-Head and Neck Surgery (OHNS) residency applicants.

STUDY DESIGN

Retrospective content analysis.

SETTING

Applications from OHNS applications at a single training institution for the 2019 and 2020 match cycles.

METHODS

A total of 2283 letters of recommendation for 611 OHNS applicants were analyzed. Applicant and letter writer gender, applicant race and ethnicity, and applicant characteristics including United States Medical Licensing Examination® Step 1 score, research productivity, and medical school rank were extracted. Agentic and communal word use from the letters of recommendation was compared across applicant and writer characteristics using multilevel negative binomial regression modeling.

RESULTS

Letter writers use a greater rate of agentic terms when describing applicants who self-identify as Asian (incidence rate ratio [IRR] = 1.16, p < .01) or "Other/not reported" (IRR = 1.23, p < .01) as compared to white applicants. Further, standardized letters of evaluation had significantly more communal language and less agentic language. Although there was an increase in communal language in letters for female applicants compared to male applicants, these gender differences disappeared in the multivariate model.

CONCLUSION

Multivariate analysis demonstrated no significant gender-based patterns in the communal or agentic language in letters of recommendation for OHNS residency applicants. However, letters for applicants identifying as Asian or "other/not reported" had more frequent use of agentic terms. Future studies should investigate other components of residency applications to assess how gender and race bias might unfairly influence an applicant's chances at a given program.

Quantifying EHR and Policy Factors Associated with the Gender Productivity Gap in Ambulatory, General Internal Medicine

BACKGROUND

The gender gap in physician compensation has persisted for decades. Little is known about how differences in use of the electronic health record (EHR) may contribute.

OBJECTIVE

To characterize how time on clinical activities, time on the EHR, and clinical productivity vary by physician gender and to identify factors associated with physician productivity.

DESIGN, SETTING, AND PARTICIPANTS

This longitudinal study included general internal medicine physicians employed by a large ambulatory practice network in the Northeastern United States from August 2018 to June 2021.

MAIN MEASURES

Monthly data on physician work relative value units (wRVUs), physician and practice characteristics, metrics of EHR use and note content, and temporal trend variables.

KEY RESULTS

The analysis included 3227 physician-months of data for 108 physicians (44% women). Compared with men physicians, women physicians generated 23.8% fewer wRVUs per month, completed 22.1% fewer visits per month, spent 4.0 more minutes/visit and 8.72 more minutes on the EHR per hour worked (all p < 0.001), and typed or dictated 36.4% more note characters per note (p = 0.006). With multivariable adjustment for physician age, practice characteristics, EHR use, and temporal trends, physician gender was no longer associated with productivity (men 4.20 vs. women 3.88 wRVUs/hour, p = 0.31). Typing/dictating fewer characters per note, relying on greater teamwork to manage orders, and spending less time on documentation were associated with higher wRVUs/hour. The 2021 E/M code change was associated with higher wRVUs/hour for all physicians: 10% higher for men physicians and 18% higher for women physicians (p < 0.001 and p = 0.009, respectively).

CONCLUSIONS

Increased team support, briefer documentation, and the 2021 E/M code change were associated with higher physician productivity. The E/M code change may have preferentially benefited women physicians by incentivizing time-intensive activities such as medical decision-making, preventive care discussion, and patient counseling that women physicians have historically spent more time performing.

Gender Disparity in Teaching Evaluations of Pediatric Faculty by Residents

OBJECTIVES

Despite their overrepresentation, female physicians continue to have lower rates of promotion compared with male physicians. Teaching evaluations play a role in physician advancement. Few studies have investigated gender disparity in resident evaluations of pediatric faculty. We hypothesized that gender disparities in resident evaluations of faculty exist and vary across subspecialties and primary work environments.

METHODS

Pediatric faculty institution-specific evaluations completed by residents from January 1, 2015, to March 9, 2020, were obtained from a single academic center. Mean ratings of faculty performance were compared by gender using a Wilcoxon 2-sample test.

RESULTS

Fifteen-thousand one-hundred and forty-two evaluations (5091 of male faculty and 10 051 of female faculty) were included. Female faculty were rated higher in overall teaching ability (female = 4.67 versus male = 4.65; P = .004). There was no statistical difference in the mean ratings of male and female faculty in the inpatient setting, whereas outpatient female faculty were rated higher in overall teaching ability (female = 4.79 versus male = 4.73; P = .005). For general pediatric faculty, females received higher ratings for overall teaching ability (female = 4.75 versus male = 4.70; P < .001). By contrast, there was no difference in ratings of subspecialty pediatric faculty.

CONCLUSIONS

Pediatric female faculty were statistically rated higher than male faculty in overall teaching ability, although these findings may not be educationally significant. The difference was driven by evaluations in the outpatient setting and for general pediatricians. This study is one of the first in pediatrics adding to the continued investigation of gender disparities in academic medicine.

Patient-Centered Insights and Biases Regarding Cardiologists Via Online Review Platform Analysis

Background Online cardiologist reviews, such as those on the Yelp website, are a frequently used method for patients to find a cardiologist. It remains unknown how bias may influence such reviews. Our objectives for this study were to (1) determine which cardiologist- or practice-related factors influence the overall rating of cardiologists and patient satisfaction and (2) discover any associations between sex and race with the overall rating of cardiologists or with cardiologist- or practice-related factors. Methods and Results Cardiologist Yelp reviews from practices in the United States from 2007 to 2020 were analyzed. A total of 563 reviews were coded for positive and negative themes. Binary logistic regression was used to determine whether certain factors increased the likelihood of high ratings. Chi-squared tests were used to determine associations between sex and race with certain factors and overall cardiologist ratings. Cardiologists were more likely to receive higher ratings when reviewers noted the characteristics of competency/knowledge base and thoroughness, positive interactions with staff, and when the cardiologist's name was mentioned in the review. Negative interactions with staff were associated with lower ratings. Female cardiologists received lower ratings and more negative mentions of cardiologist-patient communication than expected. White and Black cardiologists received lower ratings than expected compared with other racial groups. Conclusions Patient-perceived cardiologist competency, thoroughness, and positive staff interactions were associated with positive reviews in online assessments. Sex and racial differences were also found. Further research must be done to confirm these findings and to understand the association of online reviews with clinical care and patient outcomes.

Gender Bias in Risk Management Reports Involving Physicians in Training - A Retrospective Qualitative Study

OBJECTIVE

Gender bias, which contributes to burnout and attrition of female medical trainees, may manifest as disparate workplace evaluations. Here, we explore gender-based differences in perceived competence and professionalism as described in an institutional electronic risk management reporting system.

DESIGN

In this retrospective qualitative study, recurring themes were identified from anonymous entries reported to an electronic institutional risk management database from July 2014 to July 2015, and from July 2019 to July 2020 using inductive methods. This electronic system is often used by hospital staff to document complaints against physicians under the pretext of poor patient care, regardless of whether an adverse event occurred. Two individuals independently coded entries. Themes were determined from event indicator codes (EIC) using Delphi methodology and compared between gender and specialty using bivariate statistics.

SETTING

A multi-center integrated healthcare delivery system.

PARTICIPANTS

Risk management entries pertaining to physician trainees by hospital staff as written submissions to the institution's electronic risk management reporting system. Main outcomes included themes defined as: (1) lack of professionalism (i.e., delay in response, attitude, lack of communication), (2) perceived medical error, (3) breach of institutional protocol.

RESULTS

Of the 207 entries included for analysis, 52 entries identified men (25%) and 31 entries identified women (15%). The gender was not available in 124 entries and, therefore, categorized as ambiguous. The most common complaint about men involved a physician-related EIC (n = 12, 23%, EIC TX39) and the most common complaint about women involved a communication-related EIC (n = 7, 23%, EIC TX55). Eighty-eight (43%) entries involved medical trainees; 82 (40%) involved surgical trainees. Women were more often identified by their name only (n = 8, 26% vs. n = 3, 6%; p < 0.001). This finding was consistent in both medical (n = 0, 0% vs. n = 5, 31%; p < 0.001) and surgical (n = 2, 7% vs. n = 3, 25%; p = 0.006) specialties. In entries involving women, a lack of professionalism was most frequently cited (n = 29, 94%). Entries identifying medical errors more frequently involved men (n = 25, 48% vs. n = 7, 23%; p = 0.02).

CONCLUSIONS

Gender-based differences exist in how hospital staff interpret trainees' actions and attitudes. These differences have consequences for training paradigms, perceptions of clinical competence, physician burnout, and ultimately, patient outcomes.

Gender Bias in YouTube Videos Describing Common Urology Conditions

OBJECTIVE

To study implicit and explicit gender biases in YouTube videos describing common urologic conditions based on language patterns, speaker gender, and speaker profession.

METHODS

Using a Boolean search, the top 30 videos for benign prostatic hyperplasia (BPH), kidney stones, urinary tract infections (UTIs), overactive bladder (OAB), erectile dysfunction (ED), and pelvic organ prolapse (POP) were retrieved. Using the Linguistic Inquiry and Word Count program (LIWC) software, video transcripts were analyzed for 16 word categories and compared by speaker gender and urology topic to assess for bias.

RESULTS

OAB and POP had the least view counts and subscribers; kidney stone and ED videos had the most. Student education channels were more likely to feature male than female speakers (19 male vs. 6 female, P=0.01). A significant difference was noted between speaker gender in BPH (25 male vs. 4 female, P<0.001), OAB (4 male vs. 22 female, P<0.001), and POP (6 male vs. 23 female, P<0.001) videos. When examining linguistic patterns with the LIWC program, female speakers were more likely to mention personal concerns and use tentative words when speaking alone compared to males.

CONCLUSIONS

Gender bias exists in YouTube videos concerning common urologic conditions. We must be mindful of how information is distributed in order to minimize the perpetuation of gender stereotypes that are common in medicine. Awareness of these patterns and biases should encourage Urologists to proactively consider how they present themselves and how they reference the conditions they present in social media outlets.

Gender Difference in Teaching Evaluation Scores of Pediatric Faculty

OBJECTIVE

To evaluate associations between faculty gender and milestone-based teaching assessment scores assigned by residents.

METHODS

We performed a retrospective cohort study of milestone-based clinical teaching assessments of pediatric faculty completed by pediatric residents at a mid-sized residency program from July 2016 to June 2019. Assessments included 3 domains (Clinical Interactions, Teaching Skills, Role Modeling/Professionalism) comprised of a total of 11 sub-competency items. We used multilevel logistic regression accounting for repeat measures and clustering to evaluate associations between faculty gender and assessment scores in the 1) top quartile, 2) bottom quartile, or 3) top-box (highest score). Findings were adjusted for faculty rank and academic track, and resident year and gender.

RESULTS

Over 3 years, 2889 assessments of 104 faculty were performed by 91 residents. Between assessments of women and men faculty, there were no significant differences in the odds of receiving a score in the top quartile for the 3 domains (Clinical aOR 0.99, P = .86; Teaching aOR 0.99, P = .93; Role Modeling aOR 0.87, P = .089). However, assessments of women were more likely to receive a score in the bottom quartile in both Teaching (aOR 1.23, P = .019) and Role Modeling (aOR 1.26, P = .008). Assessments of women also had lower odds of receiving the highest score in 6 of 11 sub-competencies.

CONCLUSION

Results suggest that gender bias may play a role in resident assessments of pediatric faculty. Future studies are needed to determine if findings are replicated in other settings and to identify opportunities to reduce the gender gap in pediatric academic medicine.

Gender-Based Linguistic Analysis of Pediatric Clinical Faculty Evaluations

BACKGROUND

Gendered stereotypes are embedded in the culture of medicine. Women are stereotypically expected to act collaboratively and less assertively, while men are expected to act with authority and power. Whether gender-biased language is expressed in academic pediatric teaching evaluations is unknown.

OBJECTIVE

Determine whether stereotypic gender-based linguistic differences exist in resident evaluations of pediatric faculty.

METHODS

We performed a retrospective cross-sectional study of clinical faculty evaluations by pediatric residents in a single program from July 2016 to June 2019. Using Linguistic Inquiry and Word Count, responses to 2 open-ended questions were analyzed for stereotypic language. Categories were reported as a percent of total words written. Comparisons between gender groups were conducted using nonparametric Wilcoxon rank sum tests. Rates of word use within each category were analyzed using logistic regression where faculty and resident gender were included as predictor variables.

RESULTS

A total of 6436 free-text responses from 3218 unique evaluations were included. As hypothesized, evaluations of women faculty were less likely than those of men to include certain agentic language like power (odds ratio [OR] 0.9, P < .001) and insight (OR 0.9, P < .001), and research words (OR 0.6, P = .003). As expected, evaluations of women were more likely to include grindstone words, like "hardworking" (OR 1.2, P = .012). Contrary to our hypothesis, women received fewer teaching words like "mentor" (OR 0.9, P = .048) and communal words like "friendly" (OR 0.6, P = .001).

CONCLUSION

Certain stereotypic language was demonstrated in clinical teaching evaluations of pediatric faculty. These findings should be further examined to improve gender inequities in academic pediatrics.

Sex Differences in the Care Experiences of Patients Hospitalized Due to Ischemic Heart Disease in Alberta, Canada

BACKGROUND

Women with heart disease experience disparities in the diagnosis, treatment, and management of their condition. However, it is unknown whether these sex differences exist with respect to in-hospital patient experience. We examined the comprehensive experience of patients hospitalized due to ischemic heart disease (IHD) across Alberta, Canada, according to sex.

METHODS

Patients completed a modified version of the Canadian Patient Experiences Survey-Inpatient Care (CPES-IC) within 6 weeks of discharge. We examined 37 questions, including 33 regarding specific care processes and 4 global rating scales. Survey responses were reported as raw "top-box" percentages, that is, the most-positive answer choice to each question. Odds and corresponding 95% confidence intervals of women reporting a top-box response were then calculated for each question, while controlling for demographic and clinical factors.

RESULTS

From April 2014 to March 2020, a total of 5795 surveys (1612 women, 4183 men) were completed. Taking the survey margin of error into account, women had lower top-box percentages on 26 of 37 questions. Similar results were obtained for the adjusted odds of reporting a top-box response. Women did not have a higher percentage of top-box responses on any of the questions studied.

CONCLUSIONS

This study is a Canadian first, which stratified the experiences of hospitalized patients living with ischemic heart disease according to sex. Our results highlighted important sex differences. Future research to understand the mechanisms associated with these observed sex differences in patient-reported experiences is warranted.