Biology Reference
In-Depth Information
Each panel in Figure 6-3 shows 4096 consecutive RRIs, approximately 20
to 30 minutes of heart beat data. The data in panel A, representing
normal HRV, were recorded well before any signs of sepsis occurred. As
expected, the length of the RRIs oscillated around 360 to 370 milli-
seconds, with the persistent small changes showing the accelerations
and decelerations characteristic of normal, healthy variability in heart
rhythm.
The data in panels B and C were recorded 18 and 9 hours before an
acute clinical deterioration, and 30 and 21 hours before death,
respectively. The clinical differential diagnosis included septic shock,
and blood cultures were positive for coagulase-negative Staphylococcus,
and a urine culture grew E coli. The abnormal records in panel B
display no small or frequent changes; rather, there is a long-lived
baseline of much reduced variability—the length of the RRIs is generally
constant, changing slowly from 390 milliseconds in the beginning
of observation to 410 milliseconds toward the end. Near the end of
panel B, there is a large unexpected deceleration. In panel C, HRV is
also reduced, the length of the RRIs fluctuates slowly around
350 milliseconds, and the plot is punctuated intermittently by large
decelerations.
To rule out the possibility that the record in Figure 6-3 may be
atypical, we investigated hundreds of HR records of premature infants
when healthy or before sepsis. Given that the heart of a neonate beats
more than 100 times per minute, the size of HR records is
overwhelming—more than 150,000 beats per baby per day. In most of
the records, we found evidence that the abrupt clinical deteriorations
were preceded by up to 24 hours by the abnormal HR characteristics
of reduced baseline variability and subclinical, short-lived decelerations
in HR.
At this point, our hypothesis seemed plausible: the individual records
appeared to confirm a characteristic HR abnormality, manifested as a
reduced baseline variability and the presence of transient decelerations
was detectable early in the course of neonatal sepsis and SIRS. The
question then was: How do we quantify these changes? None of the
decelerations would have triggered an alarm for low HR, which is
conventionally set for 100 beats per minute (RRI 600 msec). Moreover,
neither the mean RRIs of the three data sets (369, 398, and 362 msec) nor
the standard deviations (SDs; 3.7, 5.0, and 7.4 msec) would identify the
abnormal records. Thus, we needed to design new quantitative
measures accurately depicting our observations. We arrived at two
types of measures: (1) Time-independent, based on properties of the
statistical distribution of RRIs; and (2) time-dependent measures
of temporal regularity. We describe these measures in the next
sections.
