Deriving A Novel Health Index Using A Large-Scale Population Based Electronic Health Record With Deep Networks

Deriving A Novel Health Index Using A Large-Scale Population Based Electronic Health RecordWith Deep Networks

Table of Contents




Abstract

Health indexes are useful tools for monitoring the health condition of a population and can be used to guide the healthcare policy of governments. However, most health indexes are constructed by using statistical methods to summarize recent adverse events (e.g., mortality). Information from these tools may reflect merely the impact of prior health policy holistically and can hardly indicate the most recent dynamics and its impact on future health conditions. As the advancements in medications and medical techniques rapidly evolve, there is a need for new health indexes that can reflect the most recent predictive health condition of a population and can easily be summarized with respect to any sub-population of interest. In this work, we develop a novel health index by using deep learning techniques on a large-scale and longitudinal population-based electronic health record (EHR). Three deep neural networks (DNN) models were trained to predict 4-year event rates of mortality, hospitalization, and cancer occurrence at an individual level. The Platt calibration approach was used to transform DNN output scores into estimated event risks. A novel health index is then constructed by weighted scoring these calibrated event risks. This individual-level health index not only provides a better predictive power but can also be flexibly summarized for different regions or sub-populations of interest – hence providing objective insights to develop precise personal or national policy beyond conventional health index.

  •  Non-Controlled Indexing

    • weighted scoring,
    • DNN,
    • Platt calibration approach,
    • cancer occurrence,
    • hospitalization,
    • mortality,
    • EHR,
    • deep learning,
    • medical techniques,
    • medications,
    • health condition monitoring,
    • deep networks,
    • predictive health condition,
    • a deep neural network,
    • health policy,
    • electronic health record,
    • large-scale population,
    • health index
  • Controlled Indexing

    • calibration,
    • cancer,
    • electronic health records,
    • health care,
    • learning (artificial intelligence) ,
    • neural nets,
    • patient monitoring

Introduction

Severe health-related events (such as mortality, hospitalization, or cancer occurrence) significantly impact healthcare costs and the lives of patients. A key function of governments is to develop informed healthcare policies to prevent these events and improve their citizen’s health. Population-based health indexes are useful tools for monitoring the health of a society and are used to assist healthcare policymakers to better understand the current health conditions [1, 2]. Health indexes are quantifiable evidence to objectively describe the health conditions of a population. In the past, researchers usually use a survey-based methodology to collect event statistics that generalize to the target population. There are several health indices that have been developed for these purposes, i.e. the Health Status Index[3] and the Healthcare Quality and Access Index [1]. Each of these health indicators has already been used for evaluation, comparison, resource allocation, and decision making [4]. Although health indexes have been extensively used by the governments to make health-related policy, traditional health index metrics, i.e. annual mortality data [5], may only reflect the summary impact of the past health policies which have been in place for several years. However, the rapid progress of medicines and medical technologies in this era has affected healthcare costs and public health dynamically over time. Information from these tools is therefore no longer a sensitive measurement when being used to help adapt the current health policy for the future [6]. It is reasonable to assume that these global health indicators suffer from major limitations when using it in the current complexly evolved society. A health index consists of future health events prediction may better reflect the current health status of a population and could offer better predictive decision support for policymakers. However, the challenge in reliably predicting health events has hindered the development of such a prediction-based health index.

Artificial intelligence (AI) techniques, e.g., those based on deep learning algorithms, have already demonstrated impressive power for predicting clinical events in several works [7]. Deep learning can model complex non-linear relationships between predictive variables without prior statistical assumptions [8]. Our recent works have shown that by using deep networks on large-scale EHRs, it can achieve a higher disease predictive accuracy than other machine learning methods [9, 10]. Deep learning has also been successfully applied in disease identification and outcome prediction in conventionally challenging clinical prediction tasks, such as young stroke prediction [11]. Furthermore, population-based EHRs are collected non-obtrusively in a large-scale long-term follow-up manner that includes several important yet diverse aspects of health-related information at an individual level. These characteristics make population-based EHR an especially valuable data source for constructing a prediction health index with deep learning techniques.

In this work, we propose a novel health index developed by using a deep learning technique with a large-scale population-based EHR. The health index incorporates 3 important health predictive indicators (mortality, hospitalization, and cancer occurrence). There are 4 steps to develop such a health index: (1) training 3 DNN models to predict the 4-year event rate of mortality, hospitalization, and cancer occurrence for each individual, (2) using the Platt calibration approach to transform the DNN outputs into estimated 4-year event risks, (3) calculating the individual health index value by scoring and weighing the impact of each indicator, (4) summarizing health index for the selected population (e.g., people lives in different regions) to provide intuitive insights (e.g., using map data visualization) for government to develop health improvement policy. Importantly, since the method is based on individual-level prediction, one can flexibly derive relevant index for different sub-population of interest for the policymaker to assess the societal health condition with variable granular precision. In this study, we detail our approach and show that the method indeed better predicts the risk of events than traditional health indexes.

Discussion

The deep learning method has repeatedly achieved impressive predictive and recognition results across a variety of AI tasks in recent years [7, 8]. In this work, we used the DNN model to construct a novel health index that can perform better risk estimation than the traditional surveillance model in constructing health index. This approach makes the estimation of the real-time health condition of the population more accurately and arguably provide better-informed analytics for policymaker with its future predictability. The similar validation results on TE-2003 and TE-2007 demonstrated that the superior predictive ability of the DNN model changes little over time. To the best of our knowledge, this population-based health index is one of the first studies for applying deep learning techniques for developing a health index based on predictive risk assessment. Our results show that DNN algorithms can reliably estimate health event risk in different age range by using information from the EHR source. As the use of novel clinical intervention strategies and medication changes over time, the health condition of the population also changed rapidly, especially in the modern era[6]. While the traditional surveillance models may reflect the holistic results of the medical care system several years ago, our novel DNN based health index estimated the health condition by predicting the future and therefore reflect the impact of the current medical care system. Health policymakers may find this innovative index potentially very useful as it incorporates several health dimensions and reflects the result of the current medical care system with its predictability into the future. Moreover, these health indicators are constructed from routinely collected data of the Taiwan health care system. Ideally, this approach can even provide a real-time individual-level health index for any specific person(or any sub-populations/region-specific cohort of interest) to evaluate and manage their health condition.

Conclusion

In this evaluation of applying the DNN strategy in using EHRs for constructing a novel health index, we demonstrate that this method can achieve better performance for further risk estimation than traditional approaches. This work presents one of the first methods in applying DNN to achieve health condition recognition for a specific population. Further prospective research is necessary to determine the feasibility of applying this novel health index in real-world practice and to see whether such a DNN based health index could improve the health care system and assist health care policymaking for the general population.

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

Deriving A Novel Health Index Using A Large-Scale Population-Based Electronic Health RecordWith Deep Networks

Bibliography

author

C. -Y. Hung, H. -Y. Chen, L. J. K. Wee, C. -H. Lin and C. -C. Lee,

Year

2020

Title

Deriving A Novel Health Index Using A Large-Scale Population-Based Electronic Health RecordWith Deep Networks

Publish in

2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Montreal, QC, Canada, 2020, pp. 5872-5875,

Doi

10.1109/EMBC44109.2020.9176454.

PDF reference and original file: Click here

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.

Website | + posts

Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.

Website | + posts

Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.