Type 2 diabetes mellitus (T2DM) is associated with cardiovascular complications, including cardiovascular autonomic neuropathy. Measures derived from heart rate variability (HRV) and cardiorespiratory interactions, quantified through frequency response function (FRF) and impulse response (IR) metrics, reflect different aspects of autonomic regulation. A study published in JMIR Diabetes evaluated whether these physiological domains, individually or in combination, could distinguish individuals with and without T2DM using machine learning classifiers.

Electrocardiogram and respiratory signals from two PhysioNet datasets were used to derive spectral HRV indices, FRF metrics characterizing frequency-specific respiratory-cardiac transfer properties, and causal IR metrics capturing time-domain responsiveness to respiratory inputs. Logistic regression and support vector machine (SVM) classifiers were assessed under NearMiss-1 undersampling and Synthetic Minority Oversampling Technique balancing strategies.

IR features frequently produced comparatively strong standalone performance. With logistic regression under NearMiss, IR achieved a mean accuracy of 0.770 (SD 0.179), a precision of 0.783 (SD 0.217), a recall of 0.900 (SD 0.224), and an F1-score of 0.798 (SD 0.140). Under NearMiss, combined HRV+FRF features yielded the highest observed performance (accuracy 0.830 [SD 0.172], recall 0.933 [SD 0.149], F1-score 0.853 [SD 0.145]; SVM RBF). Under the Synthetic Minority Oversampling Technique, HRV+IR showed the strongest combined performance (accuracy 0.700 [SD 0.128]; F1-score 0.691 [SD 0.097]), while IR alone retained higher recall (0.950, SD 0.112) and F1-score (0.708, SD 0.038).

These findings indicate that performance varied according to feature domain and sampling strategy. Larger datasets are needed to assess generalizability and clinical relevance.