Bottom-up approach to build a ‘precision’ risk factor classification for diabetic foot ulcer healing. Proof-of-concept

Aims: Diabetic foot ulcers (DFU) have a complex multifactorial pathophysiology. It is crucial to identify essential prognostic variables to streamline therapeutic actions and quality-of-care audits. Although SINBAD and University of Texas (UT), the most frequently used prognostic classification systems, were prospectively validated, not all individual parameters were shown to have consistent associations with healing. In this study, we used a bottom-up approach relying on robust methods to identify independent predictors of DFU healing.

Methods: 1,664 DFU patients were included by 34 Belgian diabetic foot clinics (DFCs). Twenty-one patient- and foot-related characteristics were recorded at presentation. Predictors of healing were identified using multivariable Cox proportional hazard regression. Multivariable models were built using backward regression with multiple imputation of missing values and bootstrapping.

Results: Five essential independent variables were identified: presentation delay, history of minor amputation, ulcer location, surface area and ischemia. This 5 variable-model showed a better performance compared to models based on existing classification systems.

Conclusions: A bottom-up approach was used to build a prognostic classification for DFU healing based on large databases. It offers new insights and allows to tailor the classification to certain clinical settings. These 5 parameters could be used as a 'precision classification' for specialized DFCs.