Survivorship models for estimating the risk of decompression sickness.

Several approaches have been used for modeling the incidence of decompression sickness (DCS) such as Hill's dose-response and logistic regression. Most of these methods do not include the time-to-onset information in the model. Survival analysis (failure time analysis) is appropriate when the time to onset of an event is of interest. The applicability of survival analysis for modeling the risk of DCS is illustrated by using data obtained from hypobaric chamber exposures simulating extravehicular activities (n = 426). Univariate analysis of incidence-free survival proportions were obtained for Doppler-detectable circulating microbubbles (CMB), symptoms of DCS and test aborts. A log-linear failure time regression model with 360-min half-time tissue ratio (TR) as covariate was constructed, and estimated probabilities for various TR values were calculated. Further regression analysis by including CMB status in this model showed significant improvement (p < 0.05) in the estimation of DCS over the previous model. Since DCS is dependent on the exposure pressure as well as the duration of exposure, we recommend the use of survival analysis for modeling the risk of DCS.