Aims: Cardiogenic shock (CS) is a critical condition resulting from severe cardiac dysfunction, often following myocardial infarction, advanced heart failure, or major surgeries. A key precursor to CS is hemodynamic instability (HI), marked by hypotension and compensatory tachycardia, which, if undetected, can rapidly escalate to multi-organ failure and increased mortality. Traditional HI monitoring relies on intermittent blood pressure (BP) measurements, causing detection delays of 15-35 minutes. This study proposes an alternative approach by identifying ECG-derived biomarkers that signal impending HI. By mapping subtle ECG variations to BP fluctuations, we aim to develop an early warning system for timely intervention, enhancing patient outcomes in critical care. This study explores ECG-derived biomarkers as an early warning system for HI. Methods: This study utilized MIMIC-III Clinical Database for Electronic Health Records and the MIMIC-III Waveform Database for ECG signals. From clinical data, we identified ICU patients aged over 45 who developed CS during hospitalization and did not survive. After reviewing 100+ discharge summaries, 15 cases were confirmed, with 6 (4 males, 2 females) exhibiting concurrent hypotension and tachycardia. From the waveform database, we extracted linear and non-linear features from both the time and frequency domains from 10, 5, 3, and 2-minute ECG segments. Results: A Random Forest (RF) classifier trained on the most relevant nonlinear features (Entropy and Fractal Dimension) detects HI onset with 94% accuracy and F1 score using an optimal 2-minute ECG window. To further validate these findings, we calculated the AUC (Area Under the Receiver Operating Characteristic curve) values of the RF model for different window segments. Using 5-fold cross-validation, the model achieved mean AUC of 0.99 ± 0.01 for the 2-minute windows. Conclusion: Nonlinear features in short 2-minute ECG segments effectively capture rapid, complex changes, enabling early detection of hemodynamic instability for real-time ICU monitoring.