Background: Cardiac autonomic neuropathy (CAN) is characterized by autonomic neuron dysfunction affecting heart rate regulation, which is often reflected in ECG signal changes. Identifying CAN progression based on ECG changes requires advanced signal complexity analysis due to the slight variation of the signal features across different disease stages. Earlier research focused only on differentiating between healthy and pathological signals rather than disease progression. Hence, this study investigated the relationship between CAN progression and complexity measures using a novel multiscaled modified diffusion entropy analysis (MSMDEA) and multifractal detrended fluctuation analysis (MFDFA). Methods: 15-minute-long ECG signals from participants from the Charles Sturt University Diabetes Complications Screening Group in Australia were classified into normal (N)(n=40), early (E)(n=42), and definite/severe (D)(n=10) CAN stages. MSMDEA and MFDFA were applied to quantify the scaling index (δ) of crucial events identified by the method of stripes and the fractal spectral density exponents (ɑ), respectively. Results: Significant differences in disease progression were observed by comparing MSMDEA δ values across 20 temporal scaling factors between each pair of groups using post hoc analysis (E/D, N/E, and D/N at p<0.05). Similarly, the full ranges of ɑ spectra computed from the MFDFA distinguished the three pairs of ECG signals at a post hoc test statistical significance of p<0.05, supporting the compatibility of the diagnostic methods used. Conclusion: By integrating information across multiple scales, MSMDEA and MFDFA enhance the accuracy and robustness of CAN classification, providing a more comprehensive understanding of the heart's electrical behavior and the inherent multifractal characteristics of physiological processes. Understanding multifractal processes in time series data offers a deeper insight into the underlying complexity and dynamics of the system. In addition, the characteristics of the multifractal properties, provide information on the heterogeneous and scale-dependent behavior within the data, leading to improved modeling, forecasting, and clinical decision-making.