Probabilistic modeling of hypobaric decompression sickness

• Main Author: Conkin, Johnny
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01-01-1994
• Subjects: Anatomy & physiology; Animals; Biophysics; Biostatistics; Physiology
• ISBN: 9798209356127

I did probabilistic modeling on data from decompressions in altitude chambers as a means to develop safe strategies for decompression of aviators and astronauts. Two kinds of probability models, mechanistic and empirical, were fitted to the data I collected from 86 reports on 406 decompressions of groups of people (131,022 total exposures). From consideration of a simple mass balance of evolved gas, I formulated mechanistic models that serve as expressions of decompression sickness (DCS) "dose". Doses are statistically optimized to the DCS or venous gas emboli (VGE) dichotomous response data using the Hill equation and maximum likelihood. Each dose necessarily represents an isoincidence-isovolume isopleth when displayed on a P2 vs. P1N$\sb2$ plot. I also performed multivariate logistic regressions to provide empirical models which used up to 19 covariates. The major findings are: (a) the best mechanistic model for each of the DCS or VGE response categories is an expression for DCS dose that is related to the volume of nitrogen released, expressed at ambient pressure, from the tissue, (b) it is not appropriate to predict symptom-free ascents for persons in altitude chambers with an extrapolation of the Critical Bubble Volume Hypothesis, and (c) variables about the type and intensity of exercise, time at altitude, gender, and the grade of VGE contribute significantly to the multivariate logistic regressions. The models can be used to calculate the probabilities of three categories of DCS symptoms, or the probability of the occurrence of VGE, given information about the planned decompression.