What Complexity Science Predicts About the Potential of Artificial Intelligence/Machine Learning to Improve Primary Care.

Primary care physicians are likely both excited and apprehensive at the prospects for artificial intelligence (AI) and machine learning (ML). Complexity science may provide insight into which AI/ML applications will most likely affect primary care in the future. AI/ML has successfully diagnosed some diseases from digital images, helped with administrative tasks such as writing notes in the electronic record by converting voice to text, and organized information from multiple sources within a health care system. AI/ML has less successfully recommended treatments for patients with complicated single diseases such as cancer; or improved diagnosing, patient shared decision making, and treating patients with multiple comorbidities and social determinant challenges. AI/ML has magnified disparities in health equity, and almost nothing is known of the effect of AI/ML on primary care physician-patient relationships. An intervention in Victoria, Australia showed promise where an AI/ML tool was used only as an adjunct to complex medical decision making. Putting these findings in a complex adaptive system framework, AI/ML tools will likely work when its tasks are limited in scope, have clean data that are mostly linear and deterministic, and fit well into existing workflows. AI/ML has rarely improved comprehensive care, especially in primary care settings, where data have a significant number of errors and inconsistencies. Primary care should be intimately involved in AI/ML development, and its tools carefully tested before implementation; and unlike electronic health records, not just assumed that AI/ML tools will improve primary care work life, quality, safety, and person-centered clinical decision making.

Universal health care - A matter of design and agency?

Universal health care (UHC) is primarily a financing concern, whereas primary health care (PHC) is primarily concerned with providing the right care at the right time to achieve the best possible health outcomes for individuals and communities. A recent call for contributions by the WHO emphasized that UHC can only be achieved through PHC, and that to achieve this goal will require the strengthening of the three pillars of PHC - (a) enabling primary care and public health to integrate health services, (b) empowering people and communities to create healthy living conditions, and (c) integrating multisectoral policy decisions to ensure UHC that achieves the goal of "health for all." "Pillars" - as a static metaphor - sends the wrong signal to the research and policy-making community. It, in fact, contradicts the WHO's own view, namely that there is "the need to strengthen comprehensive primary health care systems based on local priorities, needs and contexts … [that are] co-developed by people who are engaged in their own health." What we really need to develop PHC as the basis to achieve the goal of UHC is a dynamic agency to drive a "system-as-a-whole framework" that simultaneously takes into account finance, individual, and local needs. Health systems are socially constructed organizational systems that are "functionally layered" in a hierarchical fashion - governments and/or funders at the top-level not only promote the goals of the system (policies) but also constrain the system (rules, regulations, resources) in its ability to deliver. Hence, there is a need to focus on two key system features - political leadership and dynamic bottom-up agency that maintains everyone's focus on the goal to be achieved, and a limitation of system constraints so that communities can shape best adapted primary care services that truly meet the needs of their individuals, families, and community.

Potentially preventable hospitalizations-The 'pre-hospital syndrome': Retrospective observations from the MonashWatch self-reported health journey study in Victoria, Australia.

RATIONALE, AIMS, AND OBJECTIVES: HealthLinks: Chronic Care is a state-wide public hospital initiative designed to improve care for cohorts at-risk of potentially preventable hospitalizations at no extra cost. MonashWatch (MW) is an hospital outreach service designed to optimize admissions in an at-risk cohort. Telehealth operators make regular phone calls (≥weekly) using the Patient Journey Record System (PaJR). PaJR generates flags based on patient self-report, alerting to a risk of admission or emergency department attendance. 'Total flags' of global health represent concerns about self-reported general health, medication, and wellness. 'Red flags' represent significant disease/symptoms concerns, likely to lead to hospitalization. METHODS: A time series analysis of PaJR phone calls to MW patients with ≥1 acute non-surgical admissions in a 20-day time window (10 days pre-admission and 10 days post-discharge) between 23 December 2016 and 11 October 2017. Pettitt's hypothesis-testing homogeneity measure was deployed to analyse Victorian Admitted Episode/Emergency Minimum Datasets and PaJR data. FINDINGS: A MW cohort of 103 patients (mean age 74 ± 15 years; with 59% males) had 263 admissions was identified. Bed days ranged from <1 to 37.3 (mean 5.8 ± 5.8; median 4.1). The MW cohort had 7.6 calls on average in the 20-day pre- and post-hospital period. Most patients reported significantly increased flags 'pre-hospital' admission: medication issues increased on day 7.0 to 8.5; total flags day 3, worse general health days 2.5 to 1.8; and red flags of disease symptoms increased on day 1. These flags persisted following discharge. DISCUSSION/CONCLUSION: This study identified a 'pre-hospital syndrome' similar to a post-hospital phase aka the well-documented 'post-hospital syndrome'. There is evidence of a 10-day 'pre-hospital' window for interventions to possibly prevent or shorten an acute admission in this MW cohort. Further validation in a larger diverse sample is needed.

Improving quality in a complex primary care system-An example of refugee care and literature review.

RATIONALE, AIMS AND OBJECTIVES: Applying traditional industrial quality improvement (QI) methodologies to primary care is often inappropriate because primary care and its relationship to the healthcare macrosystem has many features of a complex adaptive system (CAS) that is particularly responsive to bottom-up rather than top-down management approaches. We report on a demonstration case study of improvements made in the Family Health Center (FHC) of the JPS Health Network in a refugee patient population that illustrate features of QI in a CAS framework as opposed to a traditional QI approach. METHODS: We report on changes in health system utilization by new refugee patients of the FHC from 2016 to 2017. We review the literature and summarize relevant theoretical understandings of quality management in complex adaptive systems as it applies to this case example. RESULTS: Applying CAS principles in the FHC, utilization of the Emergency Department and Urgent Care Center by newly arrived refugee patients before their first clinic visit was reduced by more than half (total visits decreased from 31%-14% of the refugee patients). Our review of the literature demonstrates that traditional algorithmic top-down QI processes are most often unsuccessful in improving even a few single-disease metrics, and increases clinician burnout and penalizes clinicians who care for vulnerable patients. Improvement in a CAS occurs when front-line clinicians identify care gaps and are given the flexibility to learn and self-organize to enable new care processes to emerge, which are created from bottom-up leadership that utilize existing interdependencies and interact with the top levels of the organization through intelligent top-down causation. We give examples of early adapters who are better applying the principles of CAS change to their QI efforts. CONCLUSIONS: Meaningful improvement in primary care is more likely achieved when the impetus to implement change shifts from top-down to bottom-up.

Capitated Telehealth Coaching Hospital Readmission Service in Australia: Pragmatic Controlled Evaluation.

BACKGROUND: MonashWatch is a telehealth public hospital outreach pilot service as a component of the Government of Victoria's statewide redesign initiative called HealthLinks: Chronic Care. Rather than only paying for hospitalizations, projected funding is released earlier to hospitals to allow them to reduce hospitalization costs. MonashWatch introduced a web-based app, Patient Journey Record System, to assess the risk of the journeys of a cohort of patients identified as frequent admitters. Telecare guides call patients using the Patient Journey Record System to flag potential deterioration. Health coaches (nursing and allied health staff) triage risk and adapt care for individuals. OBJECTIVE: The aim was a pragmatic controlled evaluation of the impact of MonashWatch on the primary outcome of bed days for acute nonsurgical admissions in the intention-to-treat group versus the usual care group. The secondary outcome was hospital admission rates. The net promoter score was used to gauge satisfaction. METHODS: Patients were recruited into an intention-to-treat group, which included active telehealth and declined/lost/died groups, versus a systematically sampled (4:1) usual care group. A rolling sample of 250-300 active telehealth patients was maintained from December 23, 2016 to June 23, 2019. The outcome-mean bed days in intervention versus control-was adjusted using analysis of covariance for age, gender, admission type, and effective days active in MonashWatch. Time-series analysis tested for trends in change patterns. RESULTS: MonashWatch recruited 1373 suitable patients who were allocated into the groups: usual care (n=293) and intention-to-treat (n=1080; active telehealth: 471/1080, 43.6%; declined: 485, 44.9%; lost to follow-up: 178 /1080, 10.7%; died: 8/1080, 0.7%). Admission frequency of intention-to-treat compared to that of the usual care group did not significantly improve (P=.05), with a small number of very frequent admitters in the intention-to-treat group. Age, MonashWatch effective days active, and treatment group independently predicted bed days. The analysis of covariance demonstrated a reduction in bed days of 1.14 (P<.001) in the intention-to-treat group compared with that in the usual care group, with 1236 bed days estimated savings. Both groups demonstrated regression-to-the-mean. The downward trend in improved bed days was significantly greater (P<.001) in the intention-to-treat group (Sen slope -406) than in the usual care group (Sen slope -104). The net promoter score was 95% in the active telehealth group compared with typical hospital scores of 77%. CONCLUSIONS: Clinically and statistically meaningful reductions in acute hospital bed days in the intention-to-treat group when compared to that of the usual care group were demonstrated (P<.001), although admission frequency was unchanged with more short stay admissions in the intention-to-treat group. Nonrandomized control selection was a limitation. Nonetheless, MonashWatch was successful in the context of the HealthLinks: Chronic Care capitation initiative and is expanding.

Health and Disease-Emergent States Resulting From Adaptive Social and Biological Network Interactions.

Health is an adaptive state unique to each person. This subjective state must be distinguished from the objective state of disease. The experience of health and illness (or poor health) can occur both in the absence and presence of objective disease. Given that the subjective experience of health, as well as the finding of objective disease in the community, follow a Pareto distribution, the following questions arise: What are the processes that allow the emergence of four observable states-(1) subjective health in the absence of objective disease, (2) subjective health in the presence of objective disease, (3) illness in the absence of objective disease, and (4) illness in the presence of objective disease? If we consider each individual as a unique biological system, these four health states must emerge from physiological network structures and personal behaviors. The underlying physiological mechanisms primarily arise from the dynamics of external environmental and internal patho/physiological stimuli, which activate regulatory systems including the hypothalamic-pituitary-adrenal axis and autonomic nervous system. Together with other systems, they enable feedback interactions between all of the person's system domains and impact on his system's entropy. These interactions affect individual behaviors, emotional, and cognitive responses, as well as molecular, cellular, and organ system level functions. This paper explores the hypothesis that health is an emergent state that arises from hierarchical network interactions between a person's external environment and internal physiology. As a result, the concept of health synthesizes available qualitative and quantitative evidence of interdependencies and constraints that indicate its top-down and bottom-up causative mechanisms. Thus, to provide effective care, we must use strategies that combine person-centeredness with the scientific approaches that address the molecular network physiology, which together underpin health and disease. Moreover, we propose that good health can also be promoted by strengthening resilience and self-efficacy at the personal and social level, and via cohesion at the population level. Understanding health as a state that is both individualized and that emerges from multi-scale interdependencies between microlevel physiological mechanisms of health and disease and macrolevel societal domains may provide the basis for a new public discourse for health service and health system redesign.

Resilience and health (care): A dynamic adaptive perspective.

This special forum on resilience explores particular worldviews of resilience-clinical, psychosocial, sociological, complexity science, organizational, and political economy through eight papers. This forum aims to open up the wealth of understandings and implications in health care by taking a transdisciplinary overview.

Quantifying resilience of humans and other animals.

All life requires the capacity to recover from challenges that are as inevitable as they are unpredictable. Understanding this resilience is essential for managing the health of humans and their livestock. It has long been difficult to quantify resilience directly, forcing practitioners to rely on indirect static indicators of health. However, measurements from wearable electronics and other sources now allow us to analyze the dynamics of physiology and behavior with unsurpassed resolution. The resulting flood of data coincides with the emergence of novel analytical tools for estimating resilience from the pattern of microrecoveries observed in natural time series. Such dynamic indicators of resilience may be used to monitor the risk of systemic failure across systems ranging from organs to entire organisms. These tools invite a fundamental rethinking of our approach to the adaptive management of health and resilience.

Resilience, health perceptions, (QOL), stressors, and hospital admissions-Observations from the real world of clinical care of unstable health journeys in Monash Watch (MW), Victoria, Australia.

RATIONALE, AIMS, AND OBJECTIVES: Monash Watch (MW) aims to reduce potentially preventable hospitalisations in a cohort above a risk "threshold" identified by Health Links Chronic Care (HLCC) algorithms using personal, diagnostic, and service data. MW conducted regular patient monitoring through outbound phone calls using the Patient Journey Record System (PaJR). PaJR alerts are intended to act as a self-reported barometer of stressors, resilience, and health perceptions with more alerts per call indicating greater risk. AIMS: To describe predictors of PaJR alerts (self-reported from outbound phone calls) and predictors of acute admissions based upon a Theoretical Model for Static and Dynamic Indicators of Acute Admissions. METHODS: Participants: HLCC cohort with predicted 3+ admissions/year in MW service arm for >40 days; n = 244. Baseline measures-Clinical Frailty Index (CFI); Connor Davis Resilience (CD-RISC): SF-12v2 Health Survey scores Mental (MSC) and Physical (PSC) and ICECAP-O. Dynamic measures: PaJR alerts/call in 10 869 MW records. Acute (non-surgical) admissions from Victorian Admitted Episode database. ANALYSIS: Logistic regression, correlations, and timeseries homogeneity metrics using XLSTAT. FINDINGS: Baseline indicators were significantly correlated except SF-12_MCS. SF12-MSC, SF12-PSC and ICECAP-O best predicted PaJR alerts/call (ROC: 0.84). CFI best predicted acute admissions (ROC: 0.66), adding CD-RISC, SF-12_MCS, SF-12_PCS and ICECAP-O with two-way interactions improved model (ROC: 0.70). PaJR alerts were higher ≤10 days preceding acute admissions and significantly correlated with admissions. Patterns in PaJR alerts in four case studies demonstrated dynamic variations signifying risk. Overall, all baseline indicators were explanatory supporting the theoretical model. Timing of PaJR alerts and acute admissions reflecting changing stressors, resilience, and health perceptions were not predicted from baseline indicators but provided a trigger for service interventions. CONCLUSION: Both static and dynamic indicators representing stressors, resilience, and health perceptions have the potential to inform threshold models of admission risk in ways that could be clinically useful.

General practice work and workforce: Interdependencies between demand, supply and quality.

BACKGROUND: General practice is regarded as central to the Australian health system. However, issues affecting the general practitioner (GP) workforce have been focused mainly on remuneration, numbers and distribution. The focus is shifting to how best to enable GPs to deliver effective, efficient and equitable care. OBJECTIVE: The aim of this paper is to identify important elements, dynamics and interdependencies that influence GPs' work and their ability to continually improve outcomes for individuals and communities. DISCUSSION: Most important problems are multifaceted and cannot be reduced to a simple, single solution. Influence diagrams capture the interdependent domains that affect general practice, such as the variations in patients' needs in the community and the impact of disadvantage and care expectations on outcomes. Identifying interrelationships between key domains should capture the dynamics that 'feed the problem'. Finding 'best possible solutions' to improve interdependent system problems and avoid the inherent risk of unintended failures requires an ongoing mix of qualitative and quantitative modelling.

Anticipatory Care in Potentially Preventable Hospitalizations: Making Data Sense of Complex Health Journeys.

Purpose: Potentially preventable hospitalizations (PPH) are minimized when adults (usually with multiple morbidities ± frailty) benefit from alternatives to emergency hospital use. A complex systems and anticipatory journey approach to PPH, the Patient Journey Record System (PaJR) is proposed. Application: PaJR is a web-based service supporting ≥weekly telephone calls by trained lay Care Guides (CG) to individuals at risk of PPH. The Victorian HealthLinks Chronic Care algorithm provides case finding from hospital big data. Prediction algorithms on call data helps optimize emergency hospital use through adaptive and anticipatory care. MonashWatch deployment incorporating PaJR is conducted by Monash Health in its Dandenong urban catchment area, Victoria, Australia. Theory: A Complex Adaptive Systems (CAS) framework underpins PaJR, and recognizes unique individual journeys, their dependence on historical and biopsychosocial influences, and difficult to predict tipping points. Rosen's modeling relationship and anticipation theory additionally informed the CAS framework with data sense-making and care delivery. PaJR uses perceptions of current and future health (interoception) through ongoing conversations to anticipate possible tipping points. This allows for possible timely intervention in trajectories in the biopsychosocial dimensions of patients as "particulars" in their unique trajectories. Evaluation: Monash Watch is actively monitoring 272 of 376 intervention patients, with 195 controls over 22 months (ongoing). Trajectories of poor health (SRH) and anticipation of worse/uncertain health (AH), and CG concerns statistically shifted at a tipping point, 3 days before admission in the subset who experienced ≥1 acute admission. The -3 day point was generally consistent across age and gender. Three randomly selected case studies demonstrate the processes of anticipatory and reactive care. PaJR-supported services achieved higher than pre-set targets-consistent reduction in acute bed days (20-25%) vs. target 10% and high levels of patient satisfaction. Discussion: Anticipatory care is an emerging trajectory data analytic approach that uses human sense-making as its core metric demonstrates improvements in processes and outcomes. Multiple sources can provide big data to inform trajectory care, however simple tailored data collections may prove effective if they embrace human interoception and anticipation. Admission risk may be addressed with a simple data collections including SRH, AH, and CG perceptions, where practical. Conclusion: Anticipatory care, as operationalized through PaJR approaches applied in MonashWatch, demonstrates processes and outcomes that successfully ameliorate PPH.

What matters in "multimorbidity"? Arguably resilience and personal health experience are central to quality of life and optimizing survival.

RATIONALE: Much is written about "multimorbidity" as it is a difficult problem for health systems, as it reflects a complex phenomenon unique to each individual health journey and health service context. This paper proposes the adoption of 2 constructs or knowledge streams into mainstream "multimorbidity" care which are arguably most important to person-centered care-personal health perceptions and resilience. ANALYSIS: "Multimorbidity" is the manifestation of multiple nonlinear physical, psychosocial, and environmental phenomena in an individual health journey. Multimorbidity encompasses very stable states for the most part together with highly unstable phases that are difficult to manage. Averting or controlling the underlying loss of resilience in instability can be challenging without early warning signals pointing towards tipping points. Monitoring resilience and early warning signals for tipping points is new to health care. Yet what should we monitor in the complexity of multimorbidity? There are multiple and competing health service features and biometrics that can be measured. However, an expanding of literature endorses importance of simply asking a person about their self-rated health in order to provide predictions of their resilience and survival. Interoception, exemplified as self-rated health, arises from internal neurocognitive self-monitoring functions of different internal and external phenomena. Interoception is being to be recognized as predictors and barometers of resilience and survival. CONCLUSIONS: Two phenomena of human systems-interoception and resilience-can guide care in the complex nature of multimorbidity in unstable health journeys and should be incorporated into clinical practice.