the importance of machine learning in building public
TRANSCRIPT
The Importance of Machine Learning
in Building Public Health Intelligence
Dr. Kemal N. SiregarSenior Researcher, Health Informatics Research Cluster
Senior Lecturer, Faculty of Public Health Universitas IndonesiaE-mail address: [email protected]
2
We are drowning in information but starved for knowledge
John Naisbitt in his book Megatrends, 1984:
Structure of presentationCurrent Health Information System in IndonesiaIntroduction01
How strategic is the roles of Public Health Intelligence?Roles of Public Health Intelligence02
As to what extent the potential use of machine learning for public health?Potential Use of Machine Learning03
How does the Machine Learning process work?Machine Learning Process04
Experience of FPHUI using Machine Learning in PH Research
05
Conclusions06
Introduction:Current Health Information System (HIS)
in Indonesia
Current HIS in Indonesia
There have been developments in HIS components, but it has not been significant to support national health strategic decision making.
PROGRESS
Progress and Challenges
Limited resources for HIS, low data management capabilities, inadequate data quality and unintegrated data from various sources, no one data warehouse at the national level.
CHALLENGES
Source: Ministry of Health RI, 2017
HIS conditions in Indonesia
Current HIS in Indonesia
Is not ready for use in big data analytics
New research agenda: Potential use of machine
learning for public health
Whereas big data analytics including machine learning,
are needed to build PHI.
How strategic is the roles of Public Health Intelligence (PHI)?
Focus on both systems and processes
Draws on a multiplicity of information sources and types
Acts as a link between data analysis and action
Intelligence is generated through a synthesis of information
Spans the full intelligence ‘cycle’ or ‘continuum’
Many PHI definitions. However, the common features across all definitions are:
How strategic is the roles of Public Health Intelligence (PHI)?PHI in Indonesia must refer to the existing laws and health realities in IndonesiaPHI is to intelligence à Something very strategic
The data obtained must be able to be processed not only into information but further into knowledge and wisdom before being used for decision making.
§ Large-scale or national public health decisions, including to protect
people from infectious diseases that are very dangerous that could
weaken community resilience, for example related to bioterrorism (GHS).
§ Ensuring that every new-born baby is in good health
As to what extent the potentials of machine
learning for public health?
IR 4.0
Benefits of Big Data Analytics for Public Health1) Innovative prevention and health promotion, 2) Health decision making for a better future, 3) Efficiency in health costs
Example Use of Machine Learning in Public HealthAnalysing Facebook statuses and Twitter updates to monitor public attitudes toward online pharmacy, antibiotics, and other health issues. (Isah et al., 2014)
Using health record data to infer mortality, readmissions, length of stay and diagnoses from patients’ health records. (Rajkomar et al., 2018)
Identifying rodent species that can harbour undiscovered zoonotic pathogens using their biological, ecological, and geographical traits. (Han et al., 2015)
Analyzing the content of a tweet to identify whether its author is sick or healthy, and using its location metadata to identify potential disease hotspots. (Sadilek et al., 2012)
How does the Machine Learning process work?
Source: Predictive modeling, UPX academy (Siying Li, Kara Lily, 2018)
As to which extent the Faculty of Public Health University of Indonesia studied the roles of machine learning to be further developed for public health in Indonesia?
Establishment of Health Informatics Research Cluster (HIRC), end of 2017
1
A number of research articles produced during 2018-20192
A review is conducted: • 10 articles selected purposively which were
preliminary studies before using machine learning• 10 articles selected that had used the machine
learning method
3
Overview of the Use of Machine Learning in Public Health Researches at FPH UI, 2018-2019
ParameterPreliminary Studies before Using
Machine Learning in Health Informatics Researches
Use of Machine Learning in Health Informatics Researches
Main Study Results Before Using Machine Learning
As to What Extent the Use of Machine Learning Method
• Searching for factors that need to be asked to determine someone is risky or not
• Prototyping android-based Personal Health Record
• Utilization of sentiment analysis to assess public opinion
• Use decision tree to determine whether someone is at risk or not
• Selection of accuracy test between Naïve Bayesian, Support Vector Machine, Neural Network
• Integration of machine learning into the applications (android- or web-based)
Source: Siregar, Kemal N. et al (2019) New Research Agenda: Potential Use of Machine Learning
Study at the stage of developing idea through literature study although at stage TRL 1, is very important before utilizing machine learning
Overview of the Use of Machine Learning in Public Health Researches at FPH UI, 2018-2019
Source: Siregar, Kemal N. et al (2019) New Research Agenda: Potential Use of Machine Learning
Parameter
Preliminary Studies before Using Machine Learning in
Health Informatics Researches
Use of Machine Learning in Health Informatics Researches
Status of Research Some have been published in national or international journals
Some have been published in national or international journals
Each stage of study in the use of machine learning starting from the initial idea stage, modeling, until getting adequate accuracy for a prediction, deserve to be published both at national and international level.
Overview of the Use of Machine Learning in Public Health Researches at FPH UI, 2018-2019
ParameterPreliminary Studies before Using Machine Learning in
Health Informatics Researches
Use of Machine Learning in Health Informatics Researches
Other Information, Including Support for Research and TRL
Several types of research, namely post graduate research programs, doctoral research programs, where some have received funding from UI.
Several types of research, namely post graduate research programs, doctoral research programs, where some have received funding from UI. Technology Readiness Level from 2 to 6.
Source: Siregar, Kemal N. et al (2019) New Research Agenda: Potential Use of Machine Learning
The prototype of the use of machine learning that was developed is still at an early stage and no one has tested it at the user level (TRL 7 and above)
Conclusions
• Based on the experience of health informatics research at FPH UI in the last 2
years, machine learning methods are very useful for recognizing patterns
(positive or negative) and making predictions of outcomes with high accuracy.
• This research experience
can be the basis for the
future research agenda in
public health field, so that
the machine learning
could be used further in
various management
decision-making,
programs and service
delivery care.
• It could further be used
for national level Public
Health Intelligence,
which is high level
decision making and
very strategic
Acknowledgment
• Appreciation to the Dean and leadership of FPH UI for supporting and launching the establishment of the Health Informatics Research Cluster (HIRC) in early 2018.
• Thank you to the Research Fellows of HIRC who were actively involved in various health informatics research activities during 2018-2019.
• Thank you very much goes to Miss Retnowatiand Miss Berly who were very helpful in preparing this presentation.
Key references
• London Health Observatory. (2006). Mapping health intelligence and its dissemination in London: Summary findings. Retrieved August 7, 2019, from http://www.lho.org.uk/viewResource.aspx?id=10898.
• Fu, P., Luck, J., & Protti, D. (2009). Information systems in support of public health in high-income countries. In R. Detels, R. Beaglehole, M. Lansang, & M. Gulliford (Eds.), Oxford textbook of public health (pp. 395–426). Oxford, England: Oxford University Press.
• Hendropriyono. (2013). Filsafat Intelijen. Jakarta: PT. Kompas Media Nusantara. ISBN 978-979-709-710-3
• Siregar Kemal N, et al. (2019). New Research Agenda: Potentials of Machine Learning for Public Health. Manuscript for The 4th Int’l Symposium on Health Research & 14th Nat’l Congress of IPHA
• Kementerian Kesehatan Republik Indonesia. (2017). Penilaian Sistem Informasi Kesehatandi Indonesia Tahun 2007, 2012, dan 2016 Pusat Data dan Informasi KementerianKesehatan Republik Indonesia.
• Siying Li, Kara Lily. (2018). “Certified fresh” lessons from machine learning. Retrieved from https://www.mawer.com/the-art-of-boring/blog/certified-fresh-lessons-from-machine-learning
Thank you