Anticipatory care and predictive analytics; sensemaking in the emerging world of 'big data'

Abstract

Introduction: Multiple dynamics of internal and external network interactions result in the emergent observable state of a person’s health and care. A tsunami of Big Data predictions are developing to identify specific interventions to improve health. Anticipatory care on the other hand, can be understood, using Rosen's model – to recognize our innate nature to anticipate health journeys with multiple ways of knowing including biopsychosocial domains. Theory/Methods: Monash Health is the largest public hospital and community care system in Victoria. Over 12,000 patients have a Big Data prediction of 3+ admission per year, with >30% potentially avoidable. The challenge for Monash Health remains how to improve their readmission. rates. Monash Health implemented the Patient Journey Record System (PaJR), a biopsychosocial telehealth tool to understand and manage potentially avoidable hospitalizations in a pilot service of 300 patients. Telecare guides regularly converse with “at risk” individuals to track their concerns and self-perceived health. The main journey characteristics of 3 randomly selected cases from PaJR, were descriptively analysed to examine the interface between prediction and anticipation. Results: Three cases, monitored for >25 calls over 6/12, were 1 male with gastrointestinal, alcohol, mild frailty and family problems, 1 male with moderate frailty, COPD and social isolation and 1 female with COPD, social isolation and moderated frailty. All had lower socioeconomic status. ANOVA (analysis of variance) demonstrated statistically significant differences among trajectories of ’Self Rated Health (SRH)’, ’Health/Care Change’, ’Medication/Drug/Alcohol Change’ and ’Support Structure Change’ within each and among all cases. Only ’Health/Care Change’ was consistently high and not statistically significantly different across all journeys (p>0.5). Only 2 patients had an admission. On examination of each trajectory, no admission was definitely predicted with both false negatives and positives. Admission rates reduced over time. Care responses fluctuations were continual. Discussion: Prediction would seem to be an imprecise activity in a patient with multiple unstable trajectories. It was not clear how, when and what root causes triggered each admission where they occurred, from the PaJR data, although Big Data analytics predicted the occurrence of >1 admission in the time period. Each person’s health and biopsychosocial circumstances were unique. Admission rates reduced over time and high levels of healthcare change suggested anticipatory care was enacted in response to the monitoring data. Conclusion: Three case studies of patients, with predicted high readmission risk, demonstrated journeys with fluctuating — ’SRH’, ’Medication/Drug/Alcohol Change’ and ’Support Structure Change’ and admission rates. All reported fluctuating ’Health/Care Change’. Anticipatory care may be informed, but not replaced by Big Data and journey analytics. Lessons learned: Close monitoring of highly variable journeys identified by Big Data predictions should enable anticipatory care. each trajectory in each setting Limitations: Case studies are not externally generalisable. Disease management was not a focus of the monitoring activity. Suggestions for future research: Research into the improvement in the nexus of Big Data predictive analytics and anticipatory care in the quest to intervene in the emergent observable state of a person’s health to improve health and better utilize health care resources.

Keywords

• big data
• anticipatory care
• potentially avoidable hospitalizations