PPG Biomarkers for Early Sepsis Screening: Signal Features, Evidence & Clinical Potential

Sepsis kills approximately 11 million people annually worldwide and remains the leading cause of death in hospitalized patients, yet the average time from sepsis onset to clinical recognition is 4-8 hours, during which mortality increases by 7.6% per hour of delayed treatment. Photoplethysmography offers a compelling platform for continuous, non-invasive monitoring of the microvascular and autonomic dysfunction that precedes clinical deterioration in sepsis. While PPG cannot replace blood cultures and lactate measurements for sepsis diagnosis, PPG-derived biomarkers may serve as an early warning system that triggers earlier clinical evaluation.

This article reviews the pathophysiology connecting sepsis to PPG signal changes, the specific biomarkers under investigation, the clinical evidence, and the path toward implementation. For background on PPG signal acquisition and processing, see our introduction to PPG technology.

Sepsis Pathophysiology and PPG-Relevant Changes

Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection (Singer et al., 2016, JAMA; Third International Consensus Definitions for Sepsis, Sepsis-3; doi:10.1001/jama.2016.0287). The pathophysiological cascade produces several changes directly measurable by PPG.

Microvascular Dysfunction

Microvascular dysfunction is a hallmark of sepsis pathophysiology and is increasingly recognized as a driver of organ failure independent of macrocirculatory parameters (blood pressure, cardiac output). In sepsis, inflammatory mediators including nitric oxide, prostacyclin, and endothelin disrupt normal arteriolar vasomotion, the rhythmic contraction and relaxation of small arteries that distributes blood flow across capillary networks.

The result is heterogeneous perfusion: some capillary beds are over-perfused while others are completely shut down. This maldistribution persists even when systemic blood pressure and cardiac output are restored through fluid resuscitation and vasopressors, explaining why macrocirculation-targeted resuscitation strategies have failed to improve outcomes in clinical trials (ProCESS, ARISE, and ProMISe trials).

PPG sensors positioned at the finger, toe, or ear measure perfusion in exactly these peripheral microvascular beds, making them inherently sensitive to the microvascular dysfunction that characterizes sepsis. The PPG signal amplitude, morphology, and variability all reflect the state of local microvascular perfusion and vasomotor function.

Autonomic Nervous System Disruption

Sepsis profoundly disrupts autonomic regulation of the cardiovascular system. Early sepsis is characterized by sympathetic activation (tachycardia, vasoconstriction) as compensatory mechanisms engage. As sepsis progresses, autonomic uncoupling occurs: the heart becomes unresponsive to autonomic inputs, baroreflex sensitivity decreases, and the normally complex, fractal-like pattern of heart rate variability collapses toward a monotonous rhythm (Goldstein et al., 1998, doi:10.1097/00003246-199802000-00040).

This loss of complexity, measurable through entropy analysis and fractal dimension calculations of PPG-derived pulse intervals, often precedes clinical deterioration by hours. The biological basis is multifactorial: inflammatory cytokines directly depress cardiac autonomic neurons, nitric oxide impairs sympathetic neurotransmission, and organ dysfunction disrupts the afferent limbs of autonomic reflexes.

Hemodynamic Instability

Sepsis-induced vasodilation (primarily through excessive nitric oxide production) reduces systemic vascular resistance, initially causing hyperdynamic circulation (high cardiac output, bounding pulses) that evolves into cardiovascular collapse as myocardial depression develops. These hemodynamic phases produce distinct PPG waveform patterns: initially increased pulse amplitude and prominent dicrotic notch during the hyperdynamic phase, followed by dampened amplitude, loss of dicrotic notch, and narrowed pulse pressure during cardiovascular decompensation.

PPG-Derived Biomarkers for Sepsis Detection

Perfusion Index (PI)

The perfusion index, calculated as (AC/DC) x 100 where AC is the pulsatile PPG component amplitude and DC is the non-pulsatile component, is the most studied PPG biomarker in sepsis.

He et al. (2013) studied 46 septic patients after resuscitation and found that nonsurvivors had a significantly lower PI; PI predicted ICU mortality with an accuracy similar to arterial lactate, and a PI below 0.2 was associated with poor outcome (doi:10.1186/cc12788).

Heart Rate Variability Derived from PPG

PPG-derived pulse rate variability (PRV) closely approximates ECG-derived HRV under resting conditions and captures the autonomic dysfunction characteristic of sepsis.

Time-domain metrics: SDNN (standard deviation of NN intervals) decreases from normal values of 100-150 ms to below 50 ms in septic patients. RMSSD (root mean square of successive differences), reflecting parasympathetic modulation, drops below 20 ms. Continuous HRV monitoring can show these changes before clinical diagnosis: in 14 bone marrow transplant patients who developed sepsis, Ahmad et al. found a 25% drop in wavelet HRV from baseline an average of 35 hours before sepsis was diagnosed (Ahmad et al., 2009; doi:10.1371/journal.pone.0006642).

Frequency-domain metrics: High-frequency power (0.15-0.4 Hz) decreases, reflecting parasympathetic withdrawal. The LF/HF ratio initially increases (sympathetic activation) then decreases as autonomic uncoupling progresses. Pontet et al. (2003, doi:10.1016/j.jcrc.2003.08.005) found that among 46 septic patients without organ dysfunction at ICU admission, those who later developed multiple organ dysfunction already had significantly lower low-frequency power and RMSSD, with low-frequency power the best predictor.

Nonlinear metrics: Sample entropy of pulse intervals, which quantifies signal complexity, decreases significantly in sepsis. Lake et al. (2002, doi:10.1152/ajpregu.00069.2002) studied 89 neonatal ICU admissions with 21 sepsis episodes and found that sample entropy falls before clinical signs of sepsis. Detrended fluctuation analysis reveals loss of the normal 1/f fractal scaling pattern, with the short-term scaling exponent ($\alpha_1$) increasing above 1.5 during sepsis.

Pleth Variability Index (PVI)

The PVI, quantifying respiratory-induced variation in PPG pulse amplitude, reflects intravascular volume status and vascular tone, both of which are deranged in sepsis. PVI is calculated as $((\mathrm{PI}_{\max} - \mathrm{PI}_{\min}) / \mathrm{PI}_{\max}) \times 100$ over the respiratory cycle.

In mechanically ventilated patients, a PVI above 14% predicted fluid responsiveness with 81% sensitivity and 100% specificity in 25 patients under general anesthesia (Cannesson et al., 2008; doi:10.1093/bja/aen133), providing non-invasive guidance for volume resuscitation in sepsis. However, PVI reliability decreases in spontaneously breathing patients due to irregular respiratory patterns and in patients with arrhythmias. Several studies have demonstrated PVI's utility in the early hours of sepsis resuscitation, where it guides fluid management comparable to invasive stroke volume variation monitoring.

PPG-Derived Respiratory Rate

Respiratory rate extracted from the PPG signal through amplitude modulation, frequency modulation, and baseline wander analysis provides continuous respiratory monitoring. Tachypnea (respiratory rate > 22 breaths per minute) is a component of the qSOFA sepsis screening score and is often the earliest clinical sign of sepsis.

Bergese et al. (2017, doi:10.1213/ANE.0000000000001852) validated a pulse-oximetry-derived respiratory rate against clinician-reviewed capnography in 26 healthy volunteers and 53 hospitalized low-acuity patients, finding a mean bias of 0.18 breaths per minute with a precision of 1.65 breaths per minute (95% limits of agreement of about -3.8 to +4.0 breaths per minute in the hospitalized patients). While less precise than capnography, continuous PPG-derived respiratory rate monitoring on general wards, where capnography is rarely available, could detect respiratory decompensation hours earlier than intermittent nursing assessments.

Pulse Wave Morphology Features

Beyond simple amplitude metrics, detailed analysis of the PPG pulse waveform shape provides additional sepsis-relevant information:

Dicrotic notch index: The prominence of the dicrotic notch reflects arterial compliance and vascular tone. In sepsis, loss of normal vasomotor regulation causes the dicrotic notch to flatten or disappear.

Pulse wave area ratios: The ratio of systolic to diastolic area under the PPG pulse curve changes with alterations in peripheral resistance and arterial compliance. This ratio increases in early (hyperdynamic) sepsis and decreases in late (hypodynamic) sepsis.

Second derivative analysis: The acceleration plethysmogram (APG), computed as the second derivative of the PPG signal, produces characteristic wave components (a, b, c, d, e) whose ratios reflect vascular aging and stiffness. The b/a ratio and the aging index ((b-c-d-e)/a) are altered in sepsis, with studies showing significant differences between septic and non-septic ICU patients (AUROC 0.69-0.74).