Applying Latent Profile Analysis to CLEFT-Q Data: A Novel Approach to Understanding Patient-Reported Outcomes in Cleft Care

PURPOSE: Patient-reported outcomes (PROs) play a crucial role in evaluating the quality of care for patients with cleft lip and/or palate (CL/P). CLEFT-Q, a validated PRO instrument, provides valuable insights into patients’ perceptions of their appearance, function, and psychosocial well-being. However, translating CLEFT-Q data into actionable clinical insights remains challenging due to its high-dimensional structure. This study employs latent profile analysis (LPA), a statistical clustering method commonly used in psychology and personality research, to categorize CLEFT-Q responses into meaningful patient phenotypes. LPA has been increasingly applied in healthcare research, including geriatric quality-of-life and healthcare worker resilience studies, to identify subpopulations within heterogeneous datasets that would otherwise be missed using conventional analysis. By leveraging LPA, this study aims to facilitate a deeper understanding of patient experiences. METHODS: Following institutional review board approval, all CLEFT-Q responses collected at a tertiary pediatric craniofacial center from September 2021 to March 2024 were analyzed. Patients aged 8 to 29 years with complete CLEFT-Q responses were included. LPA was conducted in R on 11 CLEFT-Q domains to identify distinct response profiles. Clinical characteristics and profile stability over time were assessed. RESULTS: Responses grouped into six general profiles: two Universally Affected (UA1-Moderate, UA2-Severe), two primarily focused on Facial Appearance (FA1-Moderate, FA2-Severe), one primarily with Social, School, and Speech issues (SSS), and one Minimally Distressed (MD). Most patients fell into moderate or benign profiles, FA1 (148, 21.7%), UA1 (176, 25.8%), and MD (174, 25.6%), with smaller proportions in severe, FA2 (35, 5.1%) and UA2 (77, 11.3%), or specific, SSS (71, 10.4%), profiles. Patients in the more severe profiles were older (FA2: 14.9 years, UA2: 14.2 years) compared to the cohort average of 13.0 years (ANOVA p<0.001) and more likely to have complete cleft lip (UA2: 83.1%, cohort average: 79.1%, ANOVA p<0.001). Patients in the MD profile were younger (12.0 years vs. cohort average of 13.0 years) and more likely to have cleft palate only (MD: 43.7%, cohort average: 28.9%, ANOVA p<0.001). Among patients with multiple CLEFT-Q responses, 46.1% remained in the same profile over time without surgical intervention, suggesting moderate profile stability. The FA2 (60.0%), UA1 (55.8%), and MD (51.4%) profiles were the most stable, while FA1 (33.3%) and SSS (29.4%) demonstrated greater variability. Patients undergoing surgery exhibited a lower profile stability (38.2%), with procedures such as rhinoplasty leading to more favorable profile shifts. For instance, rhinoplasty patients commonly transitioned from universally affected profiles to those primarily affected by facial appearance concerns, reflecting improved psychosocial function. CONCLUSIONS: This study demonstrates the utility of LPA in distilling high-dimensional CLEFT-Q data into clinically relevant patient phenotypes. LPA enables the identification of nuanced patient subpopulations that may be overlooked in conventional analyses, offering a data-driven approach to cleft care. The identified profiles correlate with key clinical factors such as age, cleft severity, and surgical history, providing a framework for more personalized patient care. As PROs gain prominence in cleft care, integrating profiling methodologies like LPA may enhance the translation of patient-reported data into meaningful clinical decision-making.