About

Alexander C Allori is an Associate Professor of Surgery and an Associate Professor in Population Health Sciences at Duke University. He is also an Affiliate Faculty Member at the Duke-Margolis Institute for Health Policy. His roles encompass a focus on plastic, maxillofacial, and oral surgery, as well as general surgery. Dr. Allori is involved in various residency programs, including integrated plastic and reconstructive surgery residency programs, and contributes to the DataLab for Clinical Care & Population Health. His work is centered on advancing surgical practices and health policy through research and education, supporting the training of future surgeons and improving patient care.

Research topics

• Surgery
• Medicine
• Nursing
• Anatomy
• Orthodontics
• Operations management
• Medical emergency

Selected publications

• Natural Language Processing for Automated Classification of Cleft and Craniofacial Procedures From Operative Notes: Model Development and Feasibility Study

  JMIR Medical Informatics · 2026-05-11

  articleOpen accessSenior author

  Background: The accurate classification of operative notes is essential for surgical outcomes research; however, CPT code classification is notoriously nonspecific for many procedures. In such situations, the operative note (or "dictation") must be reviewed manually, a process that is labor-intensive and unsustainable. Natural language processing demonstrates tremendous potential for improving the efficiency and accuracy of procedure classification from unstructured operative notes. To date, it remains unexplored whether natural language processing can reliably differentiate between complex, multicomponent procedures, such as those involved in the care of cleft lip or palate and craniofacial anomalies. Objective: This study aims to develop and evaluate a machine learning framework for the automated classification of operative notes for cleft and craniofacial procedures. Methods: This single-institution, retrospective observational study used operative notes from patients undergoing cleft and craniofacial procedures at a single academic medical center from 2016 to 2024. Each note in the database had been manually classified previously. Notes were preprocessed and vectorized using term frequency-inverse document frequency. A One-vs-Rest classification framework with random forest as the base classifier was developed to categorize procedures at 3 levels: primary procedure type (cleft lip repair, alveolar bone grafting, cleft palate repair, velopharyngeal insufficiency correction, rhinoplasty, and other), procedural subtype (primary vs revision), and specific surgical technique used (eg, Fisher, Mulliken, or rotation-advancement technique for cleft lip repair). Each hierarchical level was developed and evaluated using cross-validation. To improve procedural subtype classification for classes with few samples, synthetic notes were added to the dataset. Area under the receiver operating characteristic curve (AUC), an area under the precision-recall curve, micro- and macro-averaged F1-scores, and Hamming loss were used to assess model performance. Results: The dataset comprised 630 operative notes from 311 pediatric patients undergoing cleft and craniofacial procedures between 2016 and 2024, with a mean age of 3.75 (range 0-19) years. The primary classification model achieved strong performance in distinguishing procedure types with an AUC of 0.93 (SD 0.04), area under the precision-recall curve of 0.84 (SD 0.05), micro-averaged F1-score of 0.88 (SD 0.02), a macro-averaged F1-score of 0.84 (SD 0.03), and a Hamming loss of 0.04 (SD 0.01). Secondary classifiers achieved AUC scores of 1.0 (SD 0.00) for cleft lip revision classification but failed to discriminate between alveolar bone grafting primary and revision procedures (AUC 0.49, SD 0.02). Tertiary classifiers for surgical technique identification showed AUC scores of 0.88 (SD 0.03), 0.89 (SD 0.03), and 0.89 (SD 0.09) for cleft lip, cleft palate, and velopharyngeal insufficiency repair techniques, respectively. Conclusions: This pilot study demonstrates that machine learning approaches can automate the classification of pediatric craniofacial operative notes across multiple levels of procedural detail. The implementation of such systems could significantly reduce the administrative burden related to surgical research, operations, and quality improvement.

  PublisherDOI

• Visualizing the Timeline of Care: Development of a Graphical Approach to Better Understanding Complex, Longitudinal Surgical Care of Cleft Lip/Palate

  The Cleft Palate-Craniofacial Journal · 2026-01-19

  articleSenior authorCorresponding

  ObjectiveDevelop and apply novel data visualization techniques to analyze longitudinal cleft surgical care and to identify patterns in treatment timing and procedural burden across 2 cleft teams.DesignRetrospective cohort study of operative data using novel data visualization methods.SettingTwo cleft teams in the United States.Patients and ParticipantsPatients with cleft lip and/or palate with operative clinical encounters between 2018 and 2023. Team A had 228 patients and team B had 355.InterventionsVisualization of primary cleft-related surgical procedures which were cleft lip and palate repair, fistula repair, alveolar bone grafting, correction of velopharyngeal insufficiency, orthognathic surgery, and rhinoplasty.Main Outcome Measure(s)Visual interpretation of surgical timing, frequency, volume, and distribution using novel timelines, stacked-bar charts, and ridgeplots.ResultsTimeline visualizations clarified procedural sequencing and highlighted variation in treatment timing by team and phenotype but were too dense for interpretation for a large volume of patients. Stacked-bar charts illustrated procedural volume but lacked temporal insight. Ridgeplots demonstrated both timing of procedures and aggregate team volume.ConclusionsIndividual patient timelines can effectively depict deviation from "ideal" care protocols, but aggregate data may be best depicted by a ridgeplot. These tools may support quality improvement initiatives by transforming raw data into actionable insights and enhancing multidisciplinary team reflection.

  PublisherDOI

• 44. Describing “Operative Burden” In Long-term Surgical Protocols for Cleft Lip/Palate: A Proof-of-concept Data-visualization Tool

  Plastic & Reconstructive Surgery Global Open · 2025-04-24

  articleOpen accessSenior author

  PURPOSE: Cleft lip/palate repair is classically divided into three phases: the primary phase includes pre-surgical infant orthodontics (PSIO), cleft lip repair at 3 months, and cleft palate repair at 9-12 months; the second phase involves a potential speech operation at 3 years and alveolar bone grafting at 7-9 years; and the third phase includes orthognathic surgery and rhinoplasty around 15 years. Although ideal care pathways are envisioned, actual treatment often deviates from these ideals. Common deviations may include lip and palate revision, lip adhesion, fistula repair, presence/absence of PSIO depending on the practice, and different orthognathic considerations. This project proposes a novel visualization technique to clearly visualize operative burden among patients and teams. METHODS: Various visualization methods, such as flow charts and line graphs, were considered before developing a specific timeline chart. Simulated data for two hypothetical cleft treatment teams—differing in protocols, operative events, complication rates, and compliance—were created in Python (Matplotlib and Altair). The aim was to clearly display and compare the timing and quantity of operative events across teams. Operative events for each patient were placed as dots along a timeline. The initial iteration presents data by date and suggests that Team B performs more operations than Team A, but it is difficult to interpret where these differences exist. To solve this, we converted date to patient age, color-coded procedures, and stratified by different cleft phenotypes. We iterated through multiple designs and present the optimal visualization below. RESULTS: Individual and team differences were more clearly delineated. For example, it was clear that treatment of cleft lip (CL) appears to be similar between teams, but treatment of cleft lip and alveolus (CLA) differs in that Team B is performing more early rhinoplasty (yellow dots). However, it was difficult to quantify large numbers of patients in this layout. Another way to display this data is to use stacked bar charts to succinctly show the average number of interventions per patient for each phenotype. This visualization is ideal for team comparison, as it removes the clutter of individual data points. CONCLUSION: This novel visualization technique allows teams to see the operative burden placed on patients and identify trends in treatment protocols. They can then compare these trends to their ideal protocol and to the protocols of their peers. As teams often follow their patients for years and decades on end, these visualizations can prove crucial to informing cleft care. It will be important to test this visualization technique with real-world data as well as adapt these visuals in a patient facing manner so patients can understand their individual care compared to others. The stacked bar chart aggregation is easily recognizable and understood but it is not able to show the time dimension. We are working to adapt our timeline motif for aggregated statistics which will communicate average number of events and when these events occur.

  PublisherDOI

• Evaluation of PCORnet as an Approach to Accessing Electronic Health Record (EHR) Data for Cleft Outcomes Research: Advantages and Limitations

  The Cleft Palate-Craniofacial Journal · 2025-01-15

  articleSenior authorCorresponding

  Objective: To evaluate the feasibility of using the National Patient-Centered Clinical Research Network (PCORnet ® ) as a source of electronic health record (EHR) data for cleft outcomes research. Design: Exploratory retrospective analysis of multi-year, administrative and clinical, structured data stored in PCORnet. Setting: Academic institution with an ACPA-approved cleft and craniofacial team. Patients/Participants: Encounter-level data pertaining to patients with orofacial clefts treated at this center between 2010 and 2018. Outcome Measures: (1) Ability of PCORnet to report metrics such as the following: number of new and returning patients per year; demographics; phenotype; procedures; readmission or reoperation within 30 days; etc. (2) Accuracy of selected metrics, compared with manual chart review. Results: PCORnet is useful for the calculation of simple process metrics such as patient demographics, phenotype mix, case mix, and number of readmissions. However, as it lacks access to clinical notes, PCORnet alone cannot provide more detailed information. Phenotypic classification (based on ICD codes) and procedural description (based on CPT ® ) are subject to inaccuracy. A 1-2 year delay in data upload to PCORnet may be rate-limiting for certain applications. Multi-institutional queries were feasible. Conclusions: PCORnet does not include all necessary data elements from the EHR. While very convenient for the tabulation of simple process metrics, especially from multiple institutions, supplemental data collection will be required for meaningful cleft outcomes research. Cleft teams whose institutions participate in PCORnet might choose to store the supplemental data as “sidecars” alongside the standard PCORnet database tables, which would allow for future PCORnet queries to be more informative and impactful.

  PublisherDOI

• Race and Gender Bias in Narrative Letters of Recommendation for Plastic Surgery Residency Applicants

  Journal of Surgical Research · 2025-01-07 · 5 citations

  article
  PublisherDOI

• A 5-Step Simplification of the Fisher Anatomical Subunit Cleft Lip Repair

  Plastic & Reconstructive Surgery · 2025-09-03

  article

  SUMMARY: Unilateral cleft lip repair methods have evolved from straight-line repairs to geometric procedures, rotation-advancement, and contemporary hybrid techniques. The Fisher anatomical subunit repair is a versatile, effective, and highly reproducible technique. It utilizes mathematical precision in design, avoids multi-point closure, minimizes scar burden, and does not compromise horizontal lip length for vertical height in various deformities. Despite these benefits, the repair has been criticized for its complexity. We describe a five-step distillation of the repair intended to clarify the procedure and provide adoptees an entry point to the benefits of the technique.

  PublisherDOI

• Discussion: Radical Overlapping Intravelar Veloplasty during Primary Cleft Palate Repair Results in Decreased Secondary Speech Surgery

  Plastic & Reconstructive Surgery · 2025-05-21

  articleSenior author
  PublisherDOI

• A Rating Scale for Obtaining Specific, Actionable Evaluations of Nasolabial Aesthetics after Unilateral Cleft Lip Repair

  Plastic & Reconstructive Surgery · 2025-09-23

  articleOpen access

  BACKGROUND: Surgeons pursuing improvement in the aesthetic outcomes of their cleft lip repairs may benefit from a granular scale evaluating individual objectives of the repair. METHODS: A working group of 9 surgeons convened to develop an assessment scale for nasolabial aesthetics after unilateral cleft lip repair. The group identified objectives of the repair that could be evaluated using two-dimensional facial photographs. Scale items were developed to appraise success or failure in achieving each objective. Scale items were iteratively tested and refined. The scale was subsequently implemented as part of a Continuing Medical Education course that included self-evaluation and peer-to-peer education, culminating in the formation of individual plans for improvement. RESULTS: Twelve distinct objectives of unilateral cleft lip repair were identified, of which 10 could be evaluated using photographs routinely obtained in clinical practice. A comprehensive scale was developed, incorporating these 10 objectives. Each scale item takes the form of a binary (yes/no) question evaluating a specific aesthetic concept, with accompanying reference images. Intrarater reliability for each item ranged from moderate to substantial (kappa value, 0.57 to 0.81). Interrater reliability ranged from fair to substantial (kappa value, 0.27 to 0.81). When implemented in a Continuing Medical Education course, the scale enabled surgeons to identify specific opportunities for improvement in their repair and specific surgical maneuvers to adopt in pursuit of these improvements. CONCLUSIONS: A new scale for evaluating outcomes of unilateral cleft lip repair is presented. The scale provides specific, actionable evaluations for individual objectives of the repair.

  PublisherDOI

• CLEFT-Q SwePsych protocol: A prospective observational study to investigate the psychometric characteristics test-retest reliability, responsiveness, and interpretability of CLEFT-Q

  PLoS ONE · 2025-05-07

  articleOpen access

  OBJECTIVES: Patient perceived benefit of treatment for cleft lip and/or palate is of great importance since it is central to development of cleft care. CLEFT-Q is a cleft-specific questionnaire on health-related quality of life. Test-retest reliability, aspects of responsiveness and interpretability are yet to be established for CLEFT-Q. This study aims to investigate these psychometric characteristics of CLEFT-Q. METHODS: To establish the test-retest reliability of CLEFT-Q, data will be collected repeatedly and independently at approximately 1-week intervals. Inclusion of approximately 50 patients is considered adequate for a test-retest study. To improve the interpretability of CLEFT-Q norm data from a control population of volunteers without a cleft will be collected. A total of approximately 210 participants will be included from schools, high-schools and universities. To test the responsiveness of CLEFT-Q, patients will answer selected subscales of CLEFT-Q, longitudinal anchor questions and perform global ratings of change before and after surgery. To ensure robust results, approximately 50 patients for each type of treatment will be recruited. If CLEFT-Q is found to be responsive, the pre- and postoperative difference in scores of CLEFT-Q will be compared with the change in objective measurements based on assessments by professionals in cleft care obtained in this study. To evaluate interpretability, results will be analysed to investigate the minimal important change using anchor-based, distribution-based and qualitative approach. REGISTRATION DETAILS: This study is registered at ClinicalTrials.gov under the ID 2021-06993-01.

  PublisherDOI

• Exploration of the Utility of the Generic ICHOM Standard Set Measures in Evaluating the Speech of Patients with Cleft Lip/Palate

  Plastic & Reconstructive Surgery Global Open · 2024-01-01 · 3 citations

  articleOpen accessSenior authorCorresponding

  Background: The International Consortium of Health Outcome Measurements (ICHOM) standard set for cleft care appraisal recommends clinicians assess articulation with percentage consonants correct (PCC) and velopharyngeal function with velopharyngeal competency rating (VPC-R). This study explores the utility and limitations of these generic measures in detecting cleft speech sound disorders by comparing them with two cleft-specific speech-rating systems, cleft audit protocol of speech-augmented Americleft modification (CAPS-A-AM) and Pittsburgh weighted speech scale (PWSS). Methods: Consecutive children with repaired, nonsyndromic cleft lip/palate, aged 5 years or older (n = 27) underwent prospective speech evaluations conducted at a single academic institution. These evaluations were conducted, recorded, and evaluated by blinded speech-language pathologists experienced with all tools. Results: When comparing measures of articulation, PCC scores correlated better with scores for relevant subcomponents of CAPS-A-AM than PWSS. When comparing measures of velopharyngeal function, VPC-R scores correlated well with relevant components of both scales. Using a "screening test versus diagnostic test" analogy, VPC-R ratings were 87.5% sensitive and 73.7% specific for detecting velopharyngeal dysfunction according to subcomponents of CAPS-A-AM, and 70.6% sensitive and 100% specific according to subcomponents of PWSS. Conclusions: This exploratory study demonstrates that PCC and VPC-R perform moderately well in detecting articulatory and velopharyngeal dysfunction in patients with cleft lip/palate; however, these tools cannot describe nuances of cleft speech sound disorder. Thus, although PCC and VPC-R adequately track basic minimum outcomes, we encourage teams to consider extending the standard set by adopting a cleft-specific measurement system for further evaluation of the tools.

  PublisherDOI

Frequent coauthors

• Thomas J. Sitzman

  Phoenix Children's Hospital

  180 shared

• Raymond Tse

  177 shared

• Thomas D. Samson

  Penn State Milton S. Hershey Medical Center

  174 shared

• Stephen P. Beals

  Phoenix Children's Hospital

  174 shared

• Damir B. Matic

  Western University

  173 shared

• David M. Fisher

  Hospital for Sick Children

  172 shared

• Ezgi Mercan

  Seattle Children's Hospital

  170 shared

• Babette Siebold

  Seattle Children's Hospital

  169 shared

Labs

• Surgical Education and Activities Lab (SEAL)PI

Education

• Fellowship Training, Pediatric Plastic & Craniomaxillofacial Surgery

  Children's Hospital Boston

  2014

• Plastic Surgery Residency Training, Plastic, Maxillofacial & Oral Surgery

  Duke University Hospital

  2013

• General Surgery Residency Training, General Surgery

  Beth Israel Medical Center

  2010

• Research Fellowship, Institute of Reconstructive Plastic Surgery

  New York University Langone Medical Center

  2008

• Doctor of Medicine (MD)

  University of Texas Medical School at Houston

  2003

• Master of Public Health (MPH), Health Services Organization / Management & Policy Studies

  University of Texas School of Public Health

  1999

• Bachelor of Arts (BA)

  Rice University

  1997