Heart-rate variability and SIDS. Examination of heart-rate patterns using an expert system generator

In a prospective, population-based study, HRV was analyzed from 24-hr tape recordings made on 16 full-term and one preterm infant who had subsequently suffered SIDS and compared to similar data on 23 control infants (n of recordings, 44). In the SIDS group, heart rate was higher, and overall and beat-to-beat HRV (CV, CVS, respectively) were lower, than in the controls, but not significantly. Respiratory rate and respiratory HRV (by spectral analysis) were similar in both groups. Assuming that cardiorespiratory mechanisms of SIDS are multifactorial, we expected that several subgroups would be detected in both test groups. Therefore, the average data for each recording were subsequently examined by means of an expert system generator (ExTran, Intelligent Terminals Ltd., Edinburgh, UK). By rules induced with 25 nodes, the following results were obtained: 16/44 recordings were diagnosed as SIDS on the basis of (1) respiratory rate (RR) less than 33 and CV less than 3.46% (n = 8); (2) RR greater than 33, CVS less than 2.18%, and BW greater than 3,520 g (n = 4); and (3) RR greater than 33, CVS less than 2.18%, BW less than 3,520 g, HR greater than 136, and CV greater than 1.89% (n = 4). Seventeen of 44 were considered as non-SIDS when (1) RR was 33-47.4, CVS greater than 2.18%, and RSA less than 74.3 and (2) RR greater than 33, CVS less than 2.18%, BW less than 3,520 g, and HR less than 142. The remaining 11 cases required more complicated rules in order to be classified. This study shows that although the trend of increased HR and decreased HRV in the SIDS cases was statistically non-significant, an expert system program may be helpful in defining decision rules to identify cases of SIDS on the basis of cardiorespiratory data.