Table 2. Studies evaluating the use of HRV in adult sepsis.
| Author of study | Year of study | Study population | Number of HRV parameters analysed | Relevant HRV parameters | Conclusions | Limitations | ||
|---|---|---|---|---|---|---|---|---|
| Time domain | Frequency domain | Non-linear | ||||||
| Samsudin et al. (42) | 2018 | 214 | 22 | Mean-NN | DFA-α2 | The Singapore ED Sepsis (SEDS) model for mortality, which incorporates respiratory rate, systolic blood pressure and two HRV parameters (mean-NN and DFA-α2), performed best in predicting for 30-day in-hospital mortality (IHM) and adverse events with a ROC curve of 0.78, compared to an AUC of 0.70, 0.70 and 0.56 by qSOFA, NEWS and MEWS score respectively | Single-center study in Singapore | |
| Chiew et al. (43) | 2019 | 214 | 22 | SDNN, RMSSD, TINN | LF, HF | DFA-α2, approximate entropy, SD1, SD2 | The gradient boosting model performed best with a ROC curve of 0.35, compared to the SEDS (0.22) model, qSOFA (0.21), NEWS (0.28) and MEWS (0.25) score in predicting for 30-day IHM. Top predictors for 30-day mortality included temperature, detrended fluctuation analysis (DFA) a-2, heart rate, Glasgow Coma Scale (GCS) score and approximate entropy. DFA-α2 is the most important HRV parameter in predicting for 30-day IHM | Single-center study in Singapore. Identical database employed by Samsudin et al. |
| Chen et al. (44) | 2007 | 81 | 10 | RMSSD | LF, HF, LF/HF ratio | Patients who eventually developed septic shock within 6 hours of presentation were found to have an increased RMSSD and HF and decreased LF and LF/HF ratio. Among the HRV parameters analysed, a raised RMSSD [median =0.78 (4.2–8.7), P<0.01] may be best at predicting for impending septic shock | Single-center study in Taiwan | |
| Bonjorno et al. (45) | 2019 | 60 | 14 | RMSSD | SD1 | HRV measures, specifically a RMSDD threshold of 10.8 ms were optimal at discriminating survivors and non-survivors with sepsis with a mean survival time difference of 9.9 days | Small sample size. Single-center study in Brazil. Performed in intensive care unit setting. Findings may be confounded by medications influencing the autonomic nervous system (e.g., sedatives, vasopressors) | |
| Barnaby et al. (46) | 2002 | 15 | 7 | LFnu, LF/HF ratio | All patients who survived or did not require ventilatory or hemodynamic support had a normalised LF (LFnu) values greater than 0.5 or LF/HF ratios less than 1.0. LFnu correlated with increased illness severity as calculated using APACHE II (r=20.67, r2=0.43) and SOFA (r=20.80, r2=0.64) and accounted for 40–60% of the variance in illness severity scores in patients presenting with sepsis | Small sample size. Single-center study in the United States | ||
| Barnaby et al. (47) | 2018 | 466 | 1 | LF/HF ratio | LF/HF ratio <1 was only 34% sensitive (95% CI, 19–53%) in identifying patients who required critical care or died within 72 hours of presentation. A SOFA score of ≥3 or LF/HF ratio of <1 are insufficient predictors of morbidity and mortality in sepsis | Single-center study in the United States. Only study to define the endpoint within 72 hours of presentation | ||
| Pong et al. (48) | 2019 | 364 | 22 | DFA-α2, SD 2 | A combination model incorporating best-performing clinical and one HRV parameter (SD2) performed best with a ROC curve of 0.91, compared to the NEWS (0.70), MEWS (0.61), qSOFA (0.70), SOFA (0.74), APACHE II (0.76)and MEDS (0.86) in predicting for 30-day IHM. Among the HRV parameters, DFA-α2 had the strongest predictive value as a rapid triage tool in septic patients | Single-center study in Singapore. Study limited to patients triaged to PACS 1 to 2. PACS 3 to 4 were excluded (PACS1 = critically ill, PACS2 = non-ambulant, PACS3 = ambulant, PACS4 = non-emergencies). 22.4% of septic patients were excluded due to ECG readings unsuitable for HRV analysis | ||
HRV, heart rate variability; DFA-α2, de-trended fluctuation analysis alpha-2; ROC, receiver operating characteristics; SDNN, standard deviation of all NN-intervals; RMSSD, square root of the mean of the sum of the squares of differences between adjacent NN intervals; TINN, baseline width of the minimum square difference triangular interpolation of the highest peak of the histogram of all NN intervals; LF, low frequency; HF, high frequency; SD1/SD2, Poincare section; qSOFA, quick Sequential Organ Failure Assessment; NEWS, National Early Warning Score; MEWS, Modified Early Warning Score; APACHE, Acute Physiology And Chronic Health Evaluation; MEDS, Mortality in Emergency Department Score; PACS, patient acuity category scale.