*3.2. Performance Comparisons: Compensatory Reserve versus Vital Signs*

Clinical measurements that inform and change the medical management of critically injured and sick patients should demonstrate high diagnostic accuracy. One approach to assess diagnostic accuracy includes direct comparisons of sensitivity and specificity across various monitoring capabilities. Table 2 presents such comparisons for the prediction power of standard vital signs and hemodynamic measurements for the onset of decompensated shock from data generated from LBNP experiments. The measurement of compensatory reserve displayed by far the greatest sensitivity, indicating its superior ability to correctly predict the onset of decompensated shock. Similarly, a greater specificity generated from the measurement of compensatory reserve indicated its superiority compared to the other vital signs and hemodynamic measures in the ability to identify patients who will not experience decompensated shock. The higher specificity of CRM reflects the failure of standard vital signs and hemodynamic measures alone to recognize the difference between individuals who are 'good' compensators from those who are 'poor' compensators [18,21–23,31,56–59]. Perhaps most striking is that standard vital signs and hemodynamic measurements have consistently been shown to lack sufficient accuracy as diagnostic tools to provide reliable clinical information [18,23,38,39,54,60,61]. In contrast, the ability of CRM to provide early reliable information with acceptable diagnostic accuracy is reflected by it being the only measurement with a Youden's J index above the discriminative threshold value of 0.5 that confirms a useful clinical result [33,62] (Table 2).


**Table 2.** Sensitivity, specificity and Youden's J index of traditional vital signs and hemodynamic responses for prediction of the onset of decompensated shock secondary to progressive central hypovolemia. Modified from Convertino et al. [14,22,23,25].

Note: For Youden's Index, a value of 1 represents a perfect diagnostic test, while a value of 0 represents a test with poor diagnostic accuracy. Stroke volume (SV), systolic, diastolic and mean blood pressures were measured by finger photoplethysmograpy; heart rate (HR) was measured by standard electrocardiogram; cardiac output was calculated as SV times HR; Pulse pressure variability and SpO<sup>2</sup> was measured with standard pulse oximetry; SmO<sup>2</sup> was measured with near-infrared spectroscopy; compensatory reserve was measured by pulse oximetry.

The performance of standard vital signs and hemodynamic measurements to provide an early and accurate prediction for onset of decompensated shock can also be assessed with comparisons of sensitivity and specificity calculated using the Area Under the Curve (AUC) Receiver Operating Characteristic (ROC) statistical analysis. Figure 4 provides ROC AUC comparisons of CRM with various hemodynamic (top panel), metabolic (middle panel), and autonomic cardiac (bottom panel as represented by metrics of heart rate variability and complexity) responses. The ROC AUC data in Figure 4 are based on human data generated from experimentally-controlled progressive reductions in central blood volume using the LBNP hemorrhage model [38,39,54,56,57,63]. Similar results have been reported from experiments involving controlled hemorrhage in humans [25,61,64]. These latter data corroborate the results presented in Table 2 that arterial waveform feature analysis provides a monitoring technology with the greatest ability for early and accurate prediction for the onset of decompensated shock.

Optimal management of significant traumatic hemorrhage and other compromising clinical conditions is often delayed by failure to recognize a medical crisis due to the current reliance on traditional vital signs and/or other standard physiological measures that represent a limited assessment of a totally integrated compensatory response [22,24–29,54,61,64]. In this regard, the value of monitoring the arterial waveform morphology for early detection of a clinical crisis using a CRM algorithm has been well documented during actual controlled human hemorrhage in the laboratory setting [14,22,25,38–42,44,50,52,53,61,64,65], and translated to early recognition of hypovolemia and hypotension when used by first responders during simulated emergencies training exercises [66,67], and in hospital critical care settings [20,21,43,45–47,49,51,60,68–72]. The comparative data regarding sensitivity, specificity and diagnostic accuracy of various monitoring technologies presented in this review provide compelling support for the notion that the development of wearable sensors must include an ability to capture analog signals that allow for continuous real-time analysis of changes in features of the analog arterial waveform. It should be recognized that a functional FDA-cleared monitoring system with the CRM algorithm integrated into a standard finger pulse oximeter has been developed and tested [20,22,30,69]. However, such technology has proven to provide limited information to the clinical caregiver about patient status because of unstable positioning and movement artifact. In this regard, we use the following sections of this review to emphasize the need for developing

new wearable sensor technologies that can be integrated with the established CRM algorithm in order to advance vital sign monitoring for emergency critical care.

**Figure 4.** ROC AUC comparisons for prediction of onset of decompensated shock between measures of compensatory reserve (CRM) and standard vital signs (**top** panel), metabolic metrics (**middle** panel), and autonomic nervous system responses measured by indices of heart rate variability (HRV) and complexity (HRC) (**bottom** panel). SBP, systolic blood pressure; HR, heart rate; PI, perfusion index; PPV, pulse pressure variability; RR, respiratory rate; StO<sup>2</sup> , tissue oxygen saturation; EtCO<sup>2</sup> , end-tidal carbon dioxide; RRISD, R-to-R interval standard deviation; HF, high frequency; LF, low frequency; RMSSD, root mean square standard deviation; pNN50, percentage of RRI that vary by at least 50 ms; DFA detrended fluctuation analysis; SampEn, sample entropy; StatAv, stationarity.
