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.

Limited Evidence of Shared Decision Making for Prostate Cancer Screening in Audio-Recorded Primary Care Visits Among Black Men and their Healthcare Providers.

Prostate-specific antigen (PSA)-based prostate cancer screening is a preference-sensitive decision for which experts recommend a shared decision making (SDM) approach. This study aimed to examine PSA screening SDM in primary care. Methods included qualitative analysis of audio-recorded patient-provider interactions supplemented by quantitative description. Participants included 5 clinic providers and 13 patients who were: (1) 40-69 years old, (2) Black, (3) male, and (4) attending clinic for routine primary care. Main measures were SDM element themes and "observing patient involvement in decision making" (OPTION) scoring. Some discussions addressed advantages, disadvantages, and/or scientific uncertainty of screening, however, few patients received all SDM elements. Nearly all providers recommended screening, however, only 3 patients were directly asked about screening preferences. Few patients were asked about prostate cancer knowledge (2), urological symptoms (3), or family history (6). Most providers discussed disadvantages (80%) and advantages (80%) of PSA screening. Average OPTION score was 25/100 (range 0-67) per provider. Our study found limited SDM during PSA screening consultations. The counseling that did take place utilized components of SDM but inconsistently and incompletely. We must improve SDM for PSA screening for diverse patient populations to promote health equity. This study highlights the need to improve SDM for PSA screening.

Deciding Whether to Take Sacubitril/Valsartan: How Cardiologists and Patients Discuss Out-of-Pocket Costs.

Background Out-of-pocket costs have significant implications for patients with heart failure and should ideally be incorporated into shared decision-making for clinical care. High out-of-pocket cost is one potential reason for the slow uptake of newer guideline-directed medical therapies for heart failure with reduced ejection fraction. This study aims to characterize patient-cardiologist discussions involving out-of-pocket costs associated with sacubitril/valsartan during the early postapproval period. Methods and Results We conducted content analysis on 222 deidentified transcripts of audio-recorded outpatient encounters taking place between 2015 and 2018 in which cardiologists (n=16) and their patients discussed whether to initiate, continue, or discontinue sacubitril/valsartan. In the 222 included encounters, 100 (45%) contained discussions about cost. Cost was discussed in a variety of contexts: when sacubitril/valsartan was initiated, not initiated, continued, and discontinued. Of the 97 cost conversations analyzed, the majority involved isolated discussions about insurance coverage (64/97 encounters; 66%) and few addressed specific out-of-pocket costs or affordability (28/97 encounters; 29%). Discussion of free samples of sacubitril/valsartan was common (52/97 encounters; 54%), often with no discussion of a longer-term plan for addressing cost. Conclusions Although cost conversations were somewhat common in patient-cardiologist encounters in which sacubitril/valsartan was discussed, these conversations were generally superficial, rarely addressing affordability or cost-value judgments. Cardiologists frequently provided patients with a course of free sacubitril/valsartan samples without a plan to address the cost after the samples ran out.

How can healthcare organizations improve cost-of-care conversations? A qualitative exploration of clinicians' perspectives.

OBJECTIVES: Clinicians increasingly believe they should discuss costs with their patients. We aimed to learn what strategies clinicians, clinic leaders, and health systems can use to facilitate vital cost-of-care conversations. METHODS: We conducted focus groups and semi-structured interviews with outpatient clinicians at two US academic medical centers. Clinicians recalled previous cost conversations and described strategies that they, their clinic, or their health system could use to facilitate cost conversations. Independent coders recorded, transcribed, and coded focus groups and interviews. RESULTS: Twenty-six clinicians participated between December 2019 and July 2020: general internists (23%), neurologists (27%), oncologists (15%), and rheumatologists (35%). Clinicians proposed the following strategies: teach clinicians to initiate cost conversations; systematically collect financial distress information; partner with patients to identify costs; provide accurate insurance coverage and/or out-of-pocket cost information via the electronic health record; develop local lists of lowest-cost pharmacies, laboratories, and subspecialists; hire financial counselors; and reduce indirect costs (e.g., parking). CONCLUSIONS: Despite considerable barriers to discussing, identifying, and reducing patient costs, clinicians described a variety of strategies for improving cost communication in the clinic. PRACTICE IMPLICATIONS: Health systems and clinic leadership can and should implement these strategies to improve the financial health of the patients they serve.