Heart rate variability index in trauma patients.

BACKGROUND: Heart rate variability (HRV) changes often reflect autonomic dysfunction with high sensitivity, but the specificity is also low. There are several different methods for measuring HRV, but interpretation is often complex, and the units are not interchangeable. For these reasons, HRV monitoring is not routinely used in many clinical situations. We hypothesized that the specificity of HRV as a screening tool for trauma patients could be improved by controlling some of the confounding influences using multiple logistic regression.
METHODS: A prospective observational trial with waiver of consent was performed in 243 healthy student volunteers and 257 trauma patients, in the resuscitation bay and intensive care units of a Level I trauma center, who received computed axial tomography (CT) scans of the head as part of the initial work up. Electrocardiogram results were recorded for 5 minutes. HRV was defined by SD of normal R-R intervals (SDNN5) and by root mean square of successive differences of R-R intervals (RMSSD5). A head CT scan was considered positive (+) if there were abnormalities in the parenchyma (diffuse axonal injury or contusion), vasculature (intraparenchymal, subdural, or epidural hemorrhage), and/or structural or bony components (fractures of the face or cranium).
RESULTS: In volunteers, SDNN5 was 73 +/- 15 (M +/- SD) milliseconds, compared with 42 +/- 22, 31 +/- 19, 28 +/- 17, and 12 +/- 8 milliseconds in, CT(-) patients with no sedation (n = 82), CT(-) with sedation (n = 60), CT(+) with no sedation (n = 55), and CT(+) with sedation (n = 60), respectively. The differences between trauma, sedation, and CT categories were significant (all p < 0.001). RMSSD5 differences were similar and also highly significant (all p < 0.001). For both SDNN5 and RMSSD5, in each category, there was wide overlap in the range of values, and strong inverse correlations with heart rate (all p < 0.001). Using multiple logistic regression in a subset with no missing data (n = 194), an index was derived from ln(SDNN5) adjusted for six confounding factors. With a negative predictive value held constant at 0.90, compared with ln(SDNN5) alone, the stepwise addition of heart rate, sedation, age, gender, and blood pressure progressively improved the specificity of the HRV index from 0.56 to 0.77, positive predictive value from 0.55 to 0.68, and efficiency from 0.68 to 0.80. This index was then normalized (0-100 scale) for ease of interpretation.
CONCLUSIONS: (1) Several factors alter HRV in patients; (2) when HRV was indexed for some of these factors, its specificity and efficiency were improved for predicting a discrete pathologic state in trauma patients, i.e. (+) or (-) cranial CT scans; (3) the algorithm can incorporate other factors to further refine the diagnostic and/or prognostic ability of HRV as a noninvasive clinical tool; (4) this concept should be applicable to any other HRV measurement technique or outcome.