Analyzing Patient Questions before and after Injectable Facial Aesthetic Procedures Using Machine Learning.

Abstract

Injectable facial aesthetic procedures such as botulinum toxin chemodenervation and soft-tissue dermal fillers are generally associated with mild and limited adverse effects and high patient satisfaction rates.1 A variety of different factors may affect a patient’s satisfaction, including patient expectations, short-term and long-term aesthetic outcomes, and postoperative complications. Information regarding aesthetic surgical procedures on the Internet has increased rapidly as patients often turn to online resources before consultation with a physician.2 In this study, we analyzed the top viewed preoperative and postoperative patient questions on realself.com regarding injectable facial aesthetic procedures in an effort to better understand the patient concerns expressed on social media. The most viewed questions posted to realself.com about injectable facial aesthetic procedures including botulinum toxin and dermal fillers were collected automatically using a web crawler. Each question was categorized as preoperative or postoperative and then subcategorized based on the topic focus. A machine-learning workflow described by Tseng et al.3 was applied to analyze the questions further. Question text was preprocessed and then grouped via K-means clustering. The most representative questions in each cluster were then determined to produce the most common preoperative and postoperative questions. A total of 2000 questions were collected, including 1500 dermal filler and 500 botulinum toxin questions because of the greater representation of dermal fillers on the site. The overall number and frequency of questions with their primary categories are described in Table 1. There were 1012 preoperative questions (50.6%) and 988 postoperative questions (49.4%). The most common preoperative categories were eligibility for injectables (25.1%) and specific appearance desired (19.8%). The most common postoperative queries concerned appearance (60.6%) and behavior allowed or disallowed (14.6%). (See Appendix, Supplemental Digital Content 1, which shows the most common preoperative and postoperative questions determined by machine-learning analysis, https://links.lww.com/PRS/F604.) Table 1. - Frequency of Primary Question Categories Category No. of Questions Proportion of Questions Preoperative (n = 1012) Alternatives to injectables 23 2.27 Behavior allowed or disallowed 8 0.79 Best product for patient 138 13.64 Cost 39 3.85 Dosage 79 7.81 Duration of results 29 2.87 Eligibility for injectables 254 25.10 Injectables with other procedures or products 30 2.96 Injection location 43 4.25 Miscellaneous 25 2.47 Potential adverse effects 91 8.99 Specific appearance desired 200 19.76 Surgeon recommendation 21 2.08 Technical aspects of procedure 12 1.19 Who can perform injections 20 1.98 Postoperative (n = 988) Appearance 599 60.63 Behavior allowed or disallowed 144 14.57 Miscellaneous 8 0.81 Recovery duration 64 6.48 Reversing results 17 1.72 Symptoms after surgery 156 15.79 The most common preoperative question category was eligibility for injectable procedures, indicating the importance of a thorough individualized discussion based on the patient’s personal factors during the initial consultation.4 The next most common category was obtaining a specific appearance. Before surgery, physicians can increase patient satisfaction rates by encouraging patients to express their aesthetic goals and establishing realistic expectations for specific products.5 The most common postoperative question category concerned patients’ postprocedure appearance, underscoring the significance of thoroughly discussing with patients their anticipated appearance during preoperative planning. The next most common postoperative question category was behavior allowed or disallowed, demonstrating the importance of explaining to patients what behaviors should be done or not done, as well as the duration of time before patients can return to normal activities, to optimize results and improve patient satisfaction. These results were largely reflected by the most common preoperative and postoperative questions determined by machine learning. Our analysis using crowdsourced data from realself.com can provide insight into patient concerns regarding injectable facial aesthetic procedures. Our study revealed several common categories of questions patients have during their preoperative and postoperative course regarding injectable facial aesthetic procedures. The educational handout we developed can be applied to assist physicians in better addressing patient questions and improving patient satisfaction. DISCLOSURE The authors have no financial interest to declare in relation to the content of this article.