Artificial intelligence and machine learning can be used to analyse both structured and unstructured data to support medical professionals in decision making, and related policy decisions by providing a holistic view of a particular situation.
With an increase in number of users adopting technologies like artificial intelligence (AI), cloud-computing, the Internet of Things (IoT), the data is rapidly increasing in size and complexity. This is driving the growth of data analytics in various industries. Also, it is having a profound impact on healthcare services where massive data from areas such as pharmaceuticals, medical bills and patients’ health history can be used for prediction and prevention instead of treatment and response.
In a recent development, IIIT-Delhi signed a Memorandum of Understanding (MoU) with the Lords Education and Health Society (Wish Foundation) with an aim to carry on research on health data analytics. For the coronavirus outbreak too, institutions such as World Health Organization (WHO) are utlilising data analytics and machine learning (ML) to track and predict its spread so that better decisions can be made.
According to a report from Research and Markets, global Big Data analytics market size in healthcare is expected to reach US$ 105.08 billion at a CAGR of 12.3 per cent during the forecast period of 2019 to 2027.
In hospitals and other treatment centres, electronic health records (EHRs) are created in order to have separate comprehensive digital records for every patient containing such information as medical history, test reports, allergies and the like. These can be updated by a doctor even when the patient visits a different hospital and some new data needs to be added.
A large amount of this data consists of medical imaging reports, which can be analysed through algorithms that identify patterns in pixels based on previously available images to help with diagnosis. Last year, Google revealed its plans to provide search functionality to aid in performing tasks such as data entry and billing in these systems.
Apart from previous health data for statistics, real-time information related to vitals can be obtained from wearables and other health devices for population health management through predictive analytics. The IoT makes it possible to collect information using sensors in health equipment for remote monitoring. This is useful in identifying and prioritising people more at risk early on. The data integrated from various sources needs to be accurate for correct analysis.
What are we trying to achieve
Collecting and analysing large datasets and images through data analytics solution saves time by quickly filtering huge amounts of data to find accurate and customised solutions for tough problems that are beyond the scope of researchers and medical staff. Benefits are multifaceted and inter-related. Some of the major ones are discussed next.
Improved patient satisfaction and engagement
The aim of data analytics in any field is to make better decisions. As medical data gets analysed, better decision making and informed strategic planning by health professionals leads to improved patient satisfaction and recovery. Tedious and monotonous operations, such as making appointments, get streamlined. Researchers can analyse success rate of results from different treatments and filter out the ones that have better health outcomes. In the past, University of Florida made use of Google Maps and public health data to prepare neighbourhood-level hotspot maps for efficient delivery of primary care in most areas.
Healthcare dashboards can be used by healthcare professionals to monitor key performance indicators of patient statistics in real time. Health tracking devices like fitness bands make patients more aware about their health so that they take the necessary precautions and medications to prevent their problem from escalating. When a patient’s parameters are alarming, the system immediately alerts the doctor to take necessary action(s) to deal with the situation.
Reduction in expenses
Big Data can be used to predict the admission rate of patients using ML. This is done by evaluating information including the number of patients present at a time, the costs incurred, total time of stay, among others to get an overall view of the process. This can help in understanding the right amount of hospital and clinic staff required so that the patients’ care is proper and quick, with lower wait times and without unnecessary costs.
By leveraging population health data and identifying people that need more attention, patients can be kept away from hospitals. This reduces hospital expenses due to reduction in readmission rates.
Preventing fraud in insurance companies
Insurance companies provide high cover for small premium and are prone to fraud, such as in personal injury claims. Fraudsters often receive money into their accounts through identity theft. In other instances, they claim the records to be from leading hospitals and display more costly treatments than actually provided.
Using data to validate a pre-payment and analytics to perform checks against public record databases ensures the validity of information provided to insurers. After verifying medical records, claims can be processed quickly so that genuine cases get resolved easily and treatment institutions are paid faster.
Besides efficient insurance claims processing, payers can use predictive analytics to determine at-risk claims. They can also find the best providers for specific health conditions, enabling patients to get better returns on their claims.
Technologies to achieve the feat
Formulation of a well developed data analytics solution is crucial for preventive, predictive, precise and personalised healthcare. Cloud services can economically and easily manage huge volumes of data and share information across different systems, making data analytics possible.
For scalable implementation, it is always a good idea to process all the information to extract the important data while not missing out on any detail, which can be then integrated into the system to gain meaningful insights. This is where the idea of decentralised processing, that is, edge cloud computing comes into play. Unlike traditional cloud, the information is stored locally, increasing network performance and reducing response times.
AI and ML can be used to analyse both structured (such as ontologies) and unstructured data to support medical professionals in decision making, and related policy decisions by providing a holistic view of a particular situation. The Pittsburgh Health Data Alliance started working with Amazon Web Services and utilising its ML research program to boost innovation and revolutionise disease treatment with data last year.
Apache spark is a popular framework with built-in fault tolerance for iterative and interactive processes in Big Data analysis. It is written in Scala and can be integrated with Hadoop. PySpark API allows using spark framework with Python, a popular language among developers for ML due to its various advantages like flexibility, robustness and ease of implementation.
For predictive modeling, the inputs and outputs for the model can be discrete or continuous, and are decided based on the type of problem. Neural networks are trained with past data available to understand the patterns and trends in the gathered datasets, which is known as data mining. Based on similarities in characteristics with the data of former patients, new ones can be more effectively treated.
It is important to assess the performance of a model to understand the improvements that can be made.
Impact and challenges
One of the best ways to understand the impact of health data analytics is from the response of Taiwan to coronavirus pandemic (COVID-19). According to a Stanford health report, the state integrated its national health insurance database with its immigration and customs database to create Big Data for analytics. QR code scanning and online reporting were also allowed.
This allowed identification of susceptible cases who had recent travel history to high-risk areas and were visiting clinics with symptoms through real-time tracking. People who had travelled to high-risk areas were tracked through their phones to make sure they stayed at home during the quarantine period. This helped them nip the disease in the bud.
Healthcare data involves numerous complexities that need to be tackled. For instance, medical data comes from different sources and has to comply with regulations set by different state governments and other administrative departments. There has to be a common standard when it comes to sharing patient information.
In 2019, a former patient at University of Chicago Medical Center filed a lawsuit against the institution’s partnership with Google to improve predictive analysis, stating that it violated Health Insurance Portability and Accountability Act (HIPAA) passed in 1996 in the US by recording the entry and exit dates of patients. The lawsuit was dismissed later on.
Technically too, developing infrastructure to interface datasets from different data providers on such a large scale is not easy. Security loopholes in the system can have drastic implications. So, advanced security measures like firewalls and encryption algorithms are essential to avoid risks and maintain the trust of patients.
Due to such issues, cloud alternatives are considered a safe option to reduce vulnerability by many companies. Corporate Health International (CHI) is one such organisation that is utilising the computational efficiency provided by hardware and software stack from Intel. Servers running Intel Xeon are used for both data processing and AI development.
In spite of realising its potential, healthcare organisations in many countries are still not actively using analytics due to these issues. Developments are ongoing to resolve such problems and avail the untapped benefits from analytics that are widespread in other industries already.