Since time in memoriam has health been regarded as a multidimensional complex-adaptive state (1)—a state that arises from the many non-linear interactions between its macro to nano-level variables (2). People have always experienced their health—as much in the presence as absence of diseases—in some patterned way in the context of their societal and environmental settings (Table 1).

However, the medical discourse for too long has single-mindedly focused on health and disease arising from linear processes rather than entertaining the idea that health and disease arise from systemic interactions amongst their constituent parts. Indeed, a system is not the sum of behaviors of its parts, rather it is the product of the interactions amongst its parts (2). Hence the complex dynamics amongst the—external and internal—parts of a person result in identifiable (and often well-recognized) patterns which are correlated to his/her future trajectories. In turn these patterns govern clinical care and should be the object of future research endeavors.

This paper approaches health and disease understandings as much from a philosophical/theoretical perspective supported by a synthesis of available scientific evidence from across the health sciences fields. The first part briefly introduces the sciences of pattern formation as a way to understand heterogeneous outcomes. It alludes to some of the findings of applying pattern understandings to differentiate distinctive patterns within a disease, and its implication for disease management and outcomes.

While patterns of disease precede the discussion of health patterns the paper's focus is on the complex adaptive patterns of health experiences. Interoception—the ability to sense the internal state of our bodily function—allows us to convey our ever changing health experiences (8). Viewed over time our health experiences lead to well-recognized health patterns. Understanding these health patterns has practical applications to the care of patients, research and health system organization.

Patterns describe similarities and differences in the world over time and space, and are the visible outcomes of emergent self-organization (9). Mathematically self-organization in complex systems is associated with bifurcation, i.e., a complex system reaches a state in space and time where multiple possible solutions are feasible which split the system into new stable states—a phaenomenon widely observed in nature (Figure 1).

Even complex patterns often arise from surprisingly simple mathematical functions—like zn+1= zn2 +c the function of the Mandelbrot set or x_n+1 = rx_n (1− x_n) the function of a logistic map that results in a bifurcation diagram (Figure 2). The logistic map function shows how bifurcation results in both stable variability as well as chaos, and that chaos ultimately emerges again to stability (10, 11). Bifurcation is a common feature of dynamic systems, and, as Prigogine suggested, may well-provide “the physio-chemical basis for understanding pathological behavior and disease” (9, 12). In particular, bifurcation points can emerge from within a possible narrow range of parameter values of a system's agents (9), a point explored later in the paper.

Medical discourse frequently refers to patterns—at the global/policy levels as “life-course patterns of education/work environment/abuse/social class on health/morbidity,” at the community level as “disease patterns are changing in a region/state,” at the disease level as the “pattern of psoriasis,” the “patterns of Crohn's disease” or the “patterns of lung disease in high resolution CT scans,” and at the person level as “patterns of behaviors in eating/exercising/risk taking.” While these descriptive accounts tacitly embrace the heterogeneity of medical phaenomena, it has not inspired the exploration of the underlying principles of such pattern formation.

Clinicians only recently realized that their treatment approaches are suboptimal or fail for failing to consider heterogeneity amongst patients, morphological characteristics of a disease, and drivers of disease dynamics (13–18). Three examples illustrate how non-linear research approaches led to new insights. Cluster analysis has revealed that six variables (glutamate decarboxylase antibodies, age at diagnosis, BMI, HbA1c, β-cell function and insulin resistance) related to 5 distinct diabetes clusters with significantly different disease progression and diabetes complications (nephropathy, retinopathy) (13, 14). Four domains (motor, autonomic dysfunction, rapid eye movement behavior disorder, and cognitive dysfunction) distinguish three Parkinson Disease subtypes with distinct patterns of survival, falls, wheelchair use, onset of dementia and care placement patterns (15). At the microlevel cluster analysis has identified 8 intratumoural subtypes of glioblastoma multiforme based on the patterns of 9 immune markers (18).

Non-linear dynamic models like Catastrophe (from the French term meaning “jump”) Models help us to understand the sudden—often unexpected switch—between various stable states. In general terms Catastrophe Theory states that if we know the number of observed states we also know the number of different control parameters involved (note: a control parameter can be a combination of synergistic factors). The Cusp Catastrophe Model explains the sudden—discontinuous—change between TWO otherwise stable states of a phaenomenon, like experiencing good or poor health, being in a stable or unstable disease state (e.g., heart failure), or being able or unable to maintain a certain behavior (e.g., eating or drinking disorders). Each stable state contains many different sets of control parameter combinations, however, between these two stable states lies a “bifurcation point B” (the point where the variables of the three axes meet) that defines the origin of a highly unstable zone. Here small changes in either control parameter lead to an abrupt shift between the two stable states. Two examples illustrate these non-linear dynamics and their impact on a person's state of well-being: Sudden shifts in the recurrence of alcohol abuse is precipitated by the degree of “risk recurrence” (situational threats to self-efficacy, affective states, stressful life events, loss of social/family support, acute psychological distress) and the degree of “relapse predisposition” (family history of alcoholism, nature and severity of alcoholism, comorbid psychiatric and substance abuse diagnosis) (19), whereas the risk of committing suicide amongst HIV affected patients in China is precipitated by the degrees of “experience of stigma” and the degree of “social capital” (social support, social networks, trust in others) (20).

Understanding the properties behind the pattern formation of disease heterogeneity, and appreciating that small differences in a variable can result in very different disease trajectories is fundamental in our quest to find best possible treatment options as much for the “dis-eases” as the “diseases” of individual patients (Figure 3).

A short detour into the theory of complex adaptive systems. Living systems are structurally bounded, thermodynamically open systems—while they constantly interact with their environment across their boundaries, boundaries also provide the space for their internal (physiological) function. An organism thus is structurally bounded/constraint by its physical embodiment, while simultaneously—at the macro-level—being part of its environment, and—at the micro-level—being the dynamic assembly of its constituent parts. This “multi-level” configuration provides the “necessary space” for the vital physiological functions that keep the organism alive (25). Of note, the function of an organism, while itself bounded, requires the function of its lower level units based on their own rules (25), an observation that has been generalized by Ellis as “top-down causation” in complex adaptive systems (26). Put differently, a complex adaptive system—like an organism—is the product of the interactions of its constituent parts, and its overall properties can neither be inferred from the parts nor are they present in the parts (2).

Back to health. As already alluded to, health is an experiential state of one's being as much in the presence as absence of disease (1), and it arises from our capacity for interoception as the means to convey our ever changing internal state of bodily function (8). As such health—from the old English word “hal”–is the state of being whole, and dis-ease—from the old French “des” and “ease”–describes the state of loss of wholeness. The now objectified use of “disease” relates to—mostly—visible pathologies (1, 27–29).

The discourse so far has outlined that pattern formation can be explained mathematically through often rather simple equations. While health professionals are trained to recognize patterns as a means to diagnose diseases, they have as yet—very rarely—explored the underlying nature and dynamics that leads to these patterns. Furthermore, this discourse has elucidated core features that explain how people experience health as a personal state as much in the presence as absence of identifiable diseases. Basic sciences have shown, firstly, that physiological variables are non-linearly distributed (56), and secondly, that the non-linear networked dynamics across the macro- to nano-levels, largely controlled by the inflammatory cascade, can lead to the described different health states (see examples above).

Likewise, we find patterns in health care delivery. Epidemiological studies have repeatedly shown the pattern of health, dis-ease and disease distribution in the community follows a Pareto (80/20 split) pattern−80% of the community is healthy or healthy enough not to require health care, of the remaining 20%, 80% (i.e., 16% of the community) solely require primary, of the remaining 20%, 80% (i.e., 3.2% of the community) need secondary, and the remaining 20% (i.e., 0.8% of the community) tertiary health care services (57, 58).

Looking further 80% of patient consultations end with no clinical diagnosis; of those with a diagnosis 80% have 20% of all diagnosis (the common ones), whereas 80% of all diagnosis (the rare ones) occur in 20% of all patients (59). Clinical reasoning shows the same pattern, 80% of patients who did not get into a home (instead of a hospital-based) dialysis programme had limited patient education and lack of communication skills (60). Hospital service utilization also showed a Pareto distribution pattern—slightly more than 20% of emergency department presentations were attributable to adverse medication events, and 80% of these events were caused by 20% of all drugs prescribed by community-based physicians (61). And lastly, the 10 most common causes of death affect 80% of all mortals, whereas all other (rarer or rare) causes occur in the remaining 20% (62).

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

The author confirms being the sole contributor of this work and has approved it for publication.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.