Pronósticos de la dinámica no estacionaria de hiper-superficies Gaidai multidimensionales en biosistemas: un enfoque...

It is of practical interest to forecast future mortality rates from cancer, cardiovascular, and diabetes disorders in any geographical area of interest at any given temporal return periods. The current study utilized a state-of-the-art multi-failure-mode Gaidai hypersurface spatiotemporal prognostics concept to analyse a raw/unfiltered clinical surveillance dataset.

A multicentre, population-based, biostatistical methodology has been applied to 685-dimensional (685D) biosystems, comprising annual death rates from cancer, cardiovascular, and diabetes disorders across 195 nations worldwide. This case study benchmarks a recently developed multimodal methodology for assessing the reliability of biosystems. It is especially suited to multi-regional public health & bio-environmental systems, monitored over a representative period. Extensive regional/spatial dimensions are not always effectively handled by existing statistical schemes when dealing with spatiotemporal observations of multidimensional, multi-regional phenomena. Extending Extreme Value Theory (EVT) statistics from univariate (1D) to bivariate (2D) settings poses challenges. The 1D EVT is not directly extendable to the 2D case, let alone to the challenges posed by system dimensionality beyond 2D.

The proposed methodology can effectively address this challenge. An advocated multimodal bio-risk evaluation methodology can be applied across a variety of public health prognostics contexts, using available raw clinical field/survey data. Projected 15- and 100-year death rates along with 95% Confidence Intervals (CI) are reported.

The findings of this case study may have practical implications for digital health and multimodal prognostics tools in public health systems, where big data (e.g., raw clinical observations) needs to be efficiently analysed. In assessing epidemiological risks, biosystem failure probabilities are typically low (e.g., less than 10^-6). Presented the multivariate bio-reliability concept, shown to provide enhanced accuracy in bio-risk estimates, when the underlying data sample is of limited size.