Weight Trajectories During Inpatient Treatment for Anorexia Nervosa: A Dynamic Time Warp Analysis

Area of Science:

• Psychiatry and Behavioral Sciences
• Data Science and Analytics

Background:

• Restoring weight is crucial for treating anorexia nervosa (AN).
• Previous research identified links between weight gain patterns and AN treatment outcomes, but lacked consensus on profile classification.
• Heterogeneity in weight gain profiles may arise from temporal variations, or 'warping,' in the recovery process.

Purpose of the Study:

• To identify distinct weight gain trajectories in patients with AN during inpatient treatment.
• To apply a novel non-parametric approach accounting for temporal distortions (warping) to analyze weight gain patterns.
• To determine if identified weight gain profiles predict treatment outcomes.

Main Methods:

• Employed time series clustering with dynamic time warping (DTW) on a sample of 518 AN patients.
• Utilized within-person body-mass-index gain (Δ BMI) as the primary metric for cluster identification.
• Characterized identified clusters by admission psychopathology scores and analyzed their association with clinical outcome changes.

Main Results:

• Identified four distinct weight gain clusters: initial gain (n=76), continuous gain (n=329), initial loss and recovery (n=70), and weight loss (n=43).
• Clusters differed significantly in admission BMI, psychopathology scores, and treatment duration.
• Cluster assignment was found to predict treatment outcomes, indicating the clinical relevance of these profiles.

Conclusions:

• A novel DTW-based approach identified outcome-predictive weight gain profiles in a large AN sample.
• The findings offer a more nuanced understanding of weight gain heterogeneity compared to prior studies.
• These elaborated profiles can inform individualized AN treatment strategies and optimize resource allocation.