Watching the Contagion with AI: Applications in Epidemiology
Artificial intelligence (AI) is undoubtedly one of the most remarkable transformative technologies of the decade. It has improved productivity in the healthcare sector and made advancements in precision community health and digital epidemiology.
Epidemiology is the study of the distribution and determinants of disease health and disease present in a particular population. Public health policies, decision-making, and evidence-based practice are based on the risk factors of disease and targets for preventive healthcare identified by this science branch.
One of the crucial features of epidemiology is data analytics. AI improves the computational power in the presence of enormous data, thus stretching the epidemiologists’ abilities to model and predict disease patterns, outbreak, effective treatment, etc. Due to each case’s varying needs, there is a need to develop new computational modeling methods to combat and alert devastating crises like this present COVID-19 pandemic.
Some of the applications of AI in public health epidemiology are electronic health data (EHR) based prediction of hospital readmissions, de-identifying EHRs, and analysis of social media activities. AI techniques such as machine learning, artificial neural networking, and statistical learning are often used for these purposes.
Digital surveillance of infectious diseases can be based on EHR data, social media, or even the data regarding climatic changes. For example, the climate data generated by the US National Aeronautics and Space Administration (NASA) predicts the occurrence of infectious disease epidemics and outbreaks of influenza or malaria by considering factors such as temperature, average monthly rainfall, etc. using ML and ANN.
Startups in this arena
Canada based Bluedot had promptly identified risks from the Ebola outbreak in West Africa in 2014, predicted the spread of Zika virus to Florida six months before official reports in 2016, and warned the world of undiagnosed pneumonia in Wuhan on December 31, weeks before it was declared a pandemic. BlueDot was able to accurately predict more than 80% of cities to import the novel coronavirus.
Malaysia-based AIME uses state-of-the-art Big Data Analytics to predict disease outbreaks. Their Airbo (AI for Arboviral Diseases) can predict disease outbreaks one month in advance with 80% accuracy. Although it started by predicting the 2016 Zika outbreak, it has now developed into a full-fledged platform to manage many mosquito-borne diseases like Dengue and Chikungunya.
IBM Watson Health provides companies and epidemiologists better information about the disease progression and its economic impact using Explorys data set and analytics solution. It claims that this information will enable the company to identify better treatment and populations that would benefit most. The company’s machine learning model has been trained on around 50 million ambulatory and inpatients using electronic medical record systems. For example, IBM Watson has collaborated with Smart Analytics to treat more than 6,500 psoriasis patients using IBM Explorys using predictive analytics.
Real-World Analytics of Saama Technologies mine data which could be useful in monitoring huge populations during clinical trials. It can also predict disease incidence and disease prevalence using machine learning. The company claims that the machine learning model behind the software was trained on over a billion patients’ EHR data. Thus, the algorithm is used to identify a drug’s efficacy, predict treatment patterns, the ideal duration of treatment, etc. Using the ML-based algorithms developed by Saama, Pharmacyclics has been able to maximize its clinical data in treating cancers and other autoimmune diseases.
Amadeus by Orion Health is a population health management and precision medicine software. It assists healthcare organizations in handling enormous volumes of data to predict and identify risk factors in a population with machine learning. This helps make quick and informed decisions about healthcare policies. This software’s machine learning model has been trained on insurance claims, EHR, social, and behavioural data, which correlates to the risk factors.
HealthMap is a startup by Boston Children’s Hospital that leverages AI to scan social media, internet queries, news clippings, and other data for signs of disease outbreaks. Its ProMED-mail sends notifications about disease outbreaks prediction in real-time with the help of automated classification and visualization.
Despite being automated, the initial process of training the machine with relevant information and correlations is tedious and time-consuming. Again, in case of a rapidly spreading disease outbreak, finding the exact connection between online activities and the disease can be challenging. For example, Google Flu Trends by Google from 2009 to 2015 to track the prevalence of flu in America worked with good accuracy initially by predicting the tallies about two weeks in advance but later on the system failed as it overestimated the prevalence of the disease. This happened as the machine was not retrained to make allowance for the people’s search behavior and misdiagnosed the searches for news on flu as a sign of infection.
What’s the future?
AI in epidemiology can help identify the emerging risk from unfortunate pandemics if any in the future, notifying government and hospitals, and identify epicenters of the pandemic. The most used application of AI in epidemiology is predictive analytics. It required a vast number of patient data, which sometimes is not easy to obtain. Predictive analytics involves AI algorithms that use historical data to predict future outcomes.
Having said all this, AI can’t replace direct medical testing and surveillance in any way. But it can help government and healthcare policymakers take appropriate decisions such as identifying hot spots to direct more testing. The role of data mining and machine learning in epidemiology is only going to be more significant with time. As medical information gets digitized, AI will continue to convert it into usable data, making epidemiological predictions a lot more manageable and efficient.