Photo Credit: JohnnyGreig

The following is a summary of “Applying Mobility Prediction Models to Real World Patients with Major Amputations,” published in the March 2025 issue of the Journal of Vascular Surgery by O’Banion et al.

Outcome prediction models are increasingly utilized to guide clinical decision-making and patient counseling following major amputation (MA). However, their applicability across diverse patient populations remains uncertain. This study aimed to evaluate the accuracy of existing mobility prediction models in a real-world population characterized by socioeconomic disadvantage. A retrospective analysis was conducted on patients who underwent MA due to peripheral arterial disease between 2016 and 2022. Patients who were non-ambulatory prior to amputation or had a contralateral amputation were excluded. 

Three established prediction models were assessed: AmpPredict, which forecasts one-year post-amputation mobility; AMPSIMM, which predicts the degree of prosthetic-assisted mobility at one year based on Veterans Affairs (VA) data; and a model derived from the Vascular Quality Initiative (VQI) database, also designed to predict one-year mobility. The predicted mobility outcomes from these models were compared against actual patient outcomes. A total of 126 patients met inclusion criteria, with a demographic composition of 71% male and 60% non-white, and a mean state Area Deprivation Index of 9/10, indicating significant socioeconomic disadvantage. The study cohort exhibited baseline characteristics that were markedly different from the original derivation cohorts of these models. 

The actual one-year post-amputation mobility rate was 43%. Among the 38 patients for whom AmpPredict estimated a ≥70% probability of mobility, only 45% achieved mobility. Similarly, of the 101 patients classified as having a “high” probability of mobility (≥71%) by the VQI model, only 48% regained mobility. A substantial discrepancy was noted between AmpPredict and VQI predictions for individual patients, with an average absolute difference of 36% (range: 1–81%). Furthermore, AMPSIMM predicted that 87% of patients would achieve community ambulation in one year, whereas only 32% did so (Sensitivity: 91%, Specificity: 14%, Positive Predictive Value: 33%, Negative Predictive Value: 79%). Overall, all three models significantly overestimated post-amputation mobility in this patient cohort. These discrepancies may stem from differences in demographic and clinical characteristics between the study population and the populations used to develop the models. The findings underscore the importance of exercising caution when applying predictive models to populations that differ substantially from the original derivation cohorts. This study highlights the need for validation of predictive tools in broader, socioeconomically diverse populations to ensure accurate and equitable clinical decision-making.

Source: jvascsurg.org/article/S0741-5214(25)00612-3/fulltext