The Health Big Data - Driven Complex Adaptive System - Study of Systemic Impact on Health
Ref:- J. Bircher and E. G. Hahn “Applying a complex adaptive system's understanding of health to primary care”, F1000 Research,; 5: 1672 (2016) www.ncbi.nlm.nih.gov/ mc/articles/PMC5043445/
Health as a Complex Adaptive System
“Health as a complex adaptive system (CAS) … is based on five components (a-e). Humans like all biological creatures must satisfactorily respond to (a) the demands of life. For this purpose they need (b) a biologically given potential (BGP) and (c) a personally acquired potential (PAP). These properties of individuals are embedded within (d) social and (e) environmental determinants of health. Between these five components of health there are 10 complex interactions that justify viewing health as a CAS.
System Dynamics Studies - Systems Information-theoretic View
The Health System, in general population health system is represented as a physical system like an engine, but in particular in terms of information flow as mutual information and associated probabilities, a systems information-theoretic view*of an engine.
To the above the theory of Dirac Notation - a framework designed to present Quantum Mechanics, especially the duals (Dirac dualization) is employed. It Dirac Notation provides the basis for canonical representations of elements probabilistic knowledge expressed in a comprehensible semantic way.
The BioIngine an ensemble of machine learning algorithms designed as an overarching system of applications, employs Dirac Notation to create and represent the elements within the Complex Adaptive System bringing the ensemble into a knowledge representation network as an inference network , so that the detected relationships can be described and quantified in terms of their predictive power in computation, i.e. in terms of a degree of utility in addressing determinants, outcomes, equity etc.
The Big Data Challenges
Sources of knowledge related to medicine are diverse and interoperability in medicine and related disciplines is still rather poor (e.g. the 2010 Report of the President's Council of Advisors on science and Technology).
How can we integrate potential determinants from many data structure sources to bring epidemiology, social and economic analytics, and Precision Medicine under one unified interoperable canonical knowledge framework, in order to better drive communities with best possible goal health outcomes?
Not everything is certain, and everything potentially influences everything else.
Even if we capture the knowledge elements as building blocks of inference in unified form, the world is at very least better modeled by a general graph of probabilistic knowledge representation with many interdependencies, but how can we escape that?
Can we find a system natural to Evidence Based Medicine for physician and patient guidance without the all the unnatural graph structure limitations and related independency assumptions of previous inference methods?
Social, economic, educational and Health determinants are a mix of cause and effect. Which is which? Which are both?
How can we acquire ability to impact evidence-driven Legislative Policies on the local Health outcomes so as to integrate diverse social and economic factors that are bidirectional and may involve cyclic relationships?
How can we engage the community and administrators via Deep Learning Epidemiology?
How can we provide a probabilistic semantics that is intrinsically easy to understand, provide explanations, or curate by its relations to human thinking and natural (particularly subject-verb-object) languages.
Results Employing The BioIngine
The BioInginewas developed to meet pressing needs, as a means of
a) extracting and combining clinical and biomedical knowledge in universal interoperable form
b) from a variety of both structured(spreadsheet-like) and unstructured (natural language text) sources
c) with extracts rendered as elements of probabilistic knowledge in canonical form in a Knowledge Representation Store for inference
d) based directly on the already long and widely accepted computational and notational standards in physics and theoretical chemistry established by Paul Dirac in the 1930s
e )but now emphasizing the Lorentz rotation of quantum mechanics, as Schrödinger’s wave mechanics based on the imaginary number i, to one based on hyperbolic imaginary number h rediscovered in various guises by Dirac
f) so that a totality of knowledge, probabilistic as well as sure, may be combined by otherwise direct application of quantum mechanics with empirical probabilities for automated analysis, inference, prediction, decision support and risk assessment.
Sources - 22 Peer-Reviewed Publications as the Basis of the BioIngine– see Section 3.c, plus established sparse data methods proven for years in bioinformatics, generalized.
Result 1:- Equity is NOT Equality
Health Equity vs Equality: - https://publichealthonline.gwu.edu/blog/equity-vs-equality/
Many definitions converge to the idea that equity is the determinant as recognizing equal rights and making attempts to implement that, while equality is a hoped-for or actual outcome.
CAS allows to study and explore in particular how equity in the data relates to all the above and population health.
–Equity, though often thought of as the desire for and implementation of means to obtain equality, appears to resemble equality more closely in terms of the measures on which it is based in the given population data.
–Irrespective of that, however, it is clearly desirable to increase the scores for equity in future forms of the kind of data used in this study.
–How equity and/orequality impacts and is impacted by all the above factors in the previous slide is a matter of considerable current interest as discussed in a recent Scientific American. 319:5, November, (2018)*
Equity Shows Negative Correlation with Population Health
The results on the above plots depicts perhaps a controversial view, since equity shows negative correlation/association with Population Health
It appears either "Equity" is badly defined, or attempts to improve it are not producing "Equality", or both!...or it is a statistical effect such as Simpson’s Paradox
Result 2 : Equity and Equality are Duals
Though a negative correlation of equity with other factors including education is controversial for the US, it is not surprising.
The negative relationship between equity and education that correlates via economy and other economy-related factors such as health, but also may have several plausibleexplanations.
Whatever the explanations, the OECD (Organization for Economic Collaboration And Development) in its publication “Educational Opportunity for All” (2017) rated the US as performing poorly compared with a basket of primarily European quite prosperous counties
Results : Summary of the Observations Made - Equity vs Equality Dilemma Continues.
- Economy, Infrastructure, Education, and Food & Nutrition show significant classical correlations and data mining associations with Population Health.
- Prediction studies (good sensitivity and specificity in predicting high Population health show that Economy , Infrastructure, Education are best predictors.
- Public Safety, Community Vitality, and Environment are weaker influencers but they do show up as significantly associating in data mining.
- Housing does not show up as a strong correlate in classical correlation studies nor in data mining.
- Equity shows up as having persistent negative correlations and negative (less than random) associations with Population Health in data mining. However, records where equity is high and population healthis high providing that in particular two or more of Community Vitality.
- The Population Heath versus Education correlation is very significantly enhanced for schools with good health education programs, but they only highlight the dilemma with equity.
