Toward Prevention of Adverse Events Using Anticipatory Analytics

Cite as:

Joseph Norman, Amir Akhavan, Chen Shen, David Aron, Luci Leykum, and Yaneer Bar-Yam, Toward prevention of adverse events using anticipatory analytics, Progress in Preventive Medicine (June 2020), doi: 10.1097/pp9.0000000000000029.

Download PDF

Abstract

Introduction: 

Electronic Medical Records provide new opportunities for studying the historical condition and dynamics of individual patients and populations to enable new insights that may lead to improved care and treatment. Diabetes is a prime target for new analyses as it is a chronic condition that affects 1 in 10 of the U.S. adult population and causes substantial disability and loss of life.

Methods: 

We take typical physiological measures from 3 healthcare appointments of 1,711 diabetic patients and extract combined measures that capture the overall conditions of patients and the structure of the population. Further, we examined the dynamics of individual patients across appointments in this combined measure space and examined regions associated with variability in clinical measures.

Results: 

Our results suggest that the dynamics of standard measures may aid evaluation of the risk of adverse events, and their utility should be tested in medical trials.

Conclusions: 

Dynamic variability of vital signs and standard measures may reflect a loss of homeostasis, associated physiological instability, and potential for adverse events that can be estimated using the proposed method.

Fig. 1. Scatter plot showing individual appointments as points in the 2 combined measure dimensions that capture the most variation across the population. Areas with particular clinical measure signatures are circled. A, Anomalously low blood pressure, accompanied by low to normal LDL and A1c values. B, Young adults with very high LDL and/or Hemoglobin A1c values. C, Overweight with high blood pressure and raised LDL and A1c values. D, Very high blood pressure yet normal BMI and other values. E, Oldest members of the population, who have normal values overall. Of note is the relative sharpness of the boundary of E with the unpopulated region compared with the rest of the periphery. BMI indicates body mass index; LDL, low-density lipoprotein.