The North Star Ambulatory Assessment (NSAA), which consists of 17 functional items, is foundational for ambulatory DMD care and is regularly conducted per care guidelines and recorded in medical records throughout the United Kingdom, Europe and elsewhere. At too many centers and practices, especially in the United States however, recording of standardized ambulatory assessments in DMD medical records is limited or inconsistent, which hinders retrospective learning from real-word data in this rare, progressive and life-limiting disease. In this study, we evaluate whether consistently recording a concise subset of ambulatory assessments, conducted as part of a complete assessment per care guidelines, could accurately quantify real-world disease progression for Duchenne Muscular Dystrophy (DMD) research. Drawing on data from over 1,800 complete NSAA assessments among 320 patients in the North Star Clinical Network database, we applied machine learning (lasso) to identify item subsets that best predicted total score. Predictions were evaluated in a held-out sample with the standard error of measurement (SEM) of the NSAA total score (previously estimated at 2.7 units) serving as a benchmark for sufficient accuracy. A subset of six NSAA items (hop, jump, rise, climb, walk, stand) predicted total score in validation data with correlation 0.97 and prediction error 1.8 units, less than the SEM. Prediction errors remained consistently below the SEM when stratified by age groups and by NSAA total score. These six items accord with key markers of progression in recent DMD disease models. The full NSAA should be conducted by trained assessors, consistent with care guidelines, so that disease management is not divorced from measurement. When entry of all NSAA items into the medical record is not possible, consistently recording or prioritizing entry of the six items identified here would significantly broaden the foundation for real-world outcomes research in ambulatory DMD.