How AI and Blockchain Can Combine to Benefit Population Health
By David Campbell
David Campbell is senior developer for Macadamian of Gatineau, Quebec.
The adoption of artificial intelligence (AI) continues to gain momentum as we see how it can augment a healthcare system’s effectiveness. Similarly, blockchain’s potential is very appealing to the healthcare industry for helping to solve the interoperability challenge.
While they have each individually demonstrated their potential to impact the industry, combined together they could greatly benefit population health and transform healthcare.
It seems inevitable that AI will revolutionize healthcare. The potential of AI is massive and our responsibility is to harness its power to maximize its benefits. For instance, how useful would it be for a doctor to compile a list of conditions to which their patient is susceptible to based upon their medical records and cross-referenced with general medical trends? AI can make this happen.
However, before AI can play a full role in healthcare, data collection, transportation, and storage present some complex privacy, integrity, and availability challenges that must be addressed.
Finding data sources is another major hurdle, but with the advent of consumer Internet of Things (IoT) devices, raw data is increasingly available. AI algorithms can use anonymized data from these devices to show general population health trends, but the challenge is mining the huge amount of raw data for useful information with a finite amount of computing power.
Healthcare blockchain represents another source of medical data. The prevalence of these blockchains in the medical domain is increasing because they store transactions in a network of distributed servers, which offers a high degree of availability. This adds protection against network outages and hardware failure. Also, the format of the transactions makes it almost impossible to tamper with the data. Data integrity and accountability are paramount to any healthcare solution.
While the quantity of data does not approach the amount of raw data that can be collected by medical devices, the data received by a medical blockchain is richer.
Using a blockchain solution in an electronic health record (EHR) system allows for the creation of transactions between entities such as patients and medical conditions. In this case, we can think of the diagnosis of a condition as a transaction between a patient and a known condition.
Not only can we store this information as a distributed immutable transaction in a patient record, we can also record the relationship. By updating a patient record using transactions between entities, a graph database can be constructed.
A graph database is a way of storing unstructured data and the relationships amongst the data. For example, if a physician prescribes a drug to a patient, the patient, the doctor, and the drug would be stored along with the relationships amongst the pieces of data. The relationship between the doctor and the patient would be regular doctor / patient or it could be specialist / patient. The relationship between the drug and the doctor would be prescriber.
The graph database can show latent variables, which is information hidden within the data. This can be taken a step further.
One example of a machine learning algorithm that uses graph database to extract and use latent variables is a Bayesian network. A Bayesian network is a graph database built on relationships of cause and effect.
The strength of a Bayesian network is its ability to determine probabilities. When applied to general population health data, it can help make powerful predictions and correlations between seemingly unrelated pieces of information.
For example, smoking has an elevated probability of causing lung cancer. AI can mine data surrounding this relationship from a general graph database using various algorithms. The resulting Bayesian network can be used as a model to predict diagnosis based on the medical history of a patient.
Think about the possibilities where healthcare organizations can leverage the power of these two technologies so that they can find the largest number of common connections such as: if a population is suffering from Condition X and the largest shared connection is prescription to Drug Y, it would be reasonable to investigate whether Drug Y has a side effect that causes or contributes to Condition X.
This only begins to scratch the surface. While there are many obstacles, the potential for AI and blockchain to combine forces is immense and could prove to transform healthcare as we know it.