This is a guest article by Vanhishikha Bhargava from Experfy.
According to McKinsey, machine learning and artificial intelligence in pharma and medicine are going to revolutionize the industries to help them make better decisions, optimize innovations, improve the efficiency of clinical and research trials, and provide for new tools for physicians, consumers, regulators, and even insurers. In fact, they estimate the technology could generate up to $100B in value annually.
Simply put, the more insights that data science and AI bring to the table based on biomedical data, the faster the medicine and pharma industry grows.
Trending AI Articles:
1. Cheat Sheets for AI, Neural Networks, Machine Learning, Deep Learning & Big Data
2. Data Science Simplified Part 1: Principles and Process
3. Getting Started with Building Realtime API Infrastructure
4. How I used machine learning as inspiration for physical paintings
Data protection in medicine and pharma
The amount of data in the healthcare industry knows no bounds. All the data accumulation by companies and hospitals are done during commercial researches, health outcomes over weeks, months and years, research and development projects, and clinical studies in pharma. But for decades, data analytics has been a customarily manual task for healthcare professionals.
In fact, the biggest challenge in the medicine and pharma industry has been data sharing and regulation. Drug manufacturers actively collaborate with tech companies, researchers, startups and more, thus sharing the data of millions of people. When Royal Free hospital partnered with DeepMind to develop an app for kidney injury detection, it shared the personal data of 1.6 million people with Google’s subsidiary. A subsequent investigation ruled that the hospital failed to comply with the then governing Data Protection Act 1998, now replaced by the GDPR. The problem is that currently no official regulatory framework exists to make the adoption of AI in pharma widespread.
However, some businesses manage to implement governance control for such collaborations, making a positive impact on the healthcare industry. The advancements in healthcare using data science have been previously covered in a dedicated article. Here we’ll look at the main applications in pharma.
Machine learning-driven innovation in medicine and pharma
1. Faster and Better Diagnosis
There are cases in which a patient goes undiagnosed for an extremely long period. They are not able to find the right treatment and continually struggle with multiple medical therapies to find a solution to an inaccurately identified problem. The biggest challenge here being the lack of ability to pull in past records and a medical trial for the patient.
David Talby, CTO Pacific AI and an Experfy Expert says, “With machine learning, hospitals and even pharmaceutical companies can draw up medical patterns of a patient based on specific criteria — like symptoms, medications, data from wearable devices, labs, etc. The information can then be used for a timely and better diagnosis, tracking progression, and recommending personalized treatments. For example, Kaiser Permanente improved hospital patient flow by combining natural language understanding from free-text medical notes with structured data to predict which patients will require hospitalization and for how long.”
Google has developed an algorithm that helps in identifying cancerous tumors on mammograms, using machine learning. Stanford is actively using the technology to identify the different types of skin cancer.
2. Pharmaceutical recommendations
Knowing a patient’s history and early detection of disease, medical professionals can recommend the right treatment and put a patient on the right path sooner. But what the data also enables, is letting pharmaceutical companies run targeted campaigns to promote medications and treatments, or make recommendations backed by data that could help build awareness amongst undiagnosed patients.
This doesn’t just help pharmaceutical companies increase their sales but can also support the identification of individuals at risk due to early detection of disease symptoms via the campaigns.
3. Health outcomes
The patient journey is what makes medical treatments more effective. It refers to the process of tracking how a patient suffering from a disease is responding to medication or different lines of therapies. This data is then used by medical professionals to predict health outcomes for a positive impact on the patient.
Machine learning helps create treatment pathways for patients with even the rarest of diseases, tracking their response to every little change in medication, to help optimize their journey, increasing their comfort on their way to desired health outcomes.
4. Physician trends
AI can also help medical organizations and pharmaceutical companies map physician trends. This could include the number of times a certain treatment path was chosen to treat a disease or medication was being recommended to patients in a specific area. The data doesn’t just help analyze medical practices, but also assist in understanding common needs of patients based on where they were and the environment they were exposed to.
The data, in this case, has also been used to conduct extensive commercial market research in the medicine and pharma industry, using Associative Rules Mining.
5. Risk monitoring
Data science can help gather critical patient information in real-time and respond proactively to symptoms to prevent an event from occuring. Risk-based monitoring has been used in association with sensors and electronic data capturing devices, such as oximetry, ECG, and others to acquire data and detect suspicious changes in a patient’s vital signs.
Let’s take for instance a blood pressure monitor. A machine learning algorithm could be trained to recognize critical events based on perturbations of a patient’s to prevent adverse health outcomes with timely interventions.
6. Physician matching and automation
As mentioned before, the health and pharma industries have massive databases — of physicians across all departments and patients suffering from various diseases. With machine learning applied over these data sets, you could match physicians to patients quickly instead of using general categorization to choose a doctor to treat a certain disease or provide for a treatment path.
The richer the data sets, the more relevant the matching will be, leading to patients getting access to the right physicians and treatments in a timely manner.
7. Social Media Analytics & Influencer Mapping
Pharmaceutical companies have been leveraging the influence of experienced physicians and researchers to find more patients on a global scale — for adoptions of new drugs or for clinical trials. But today, artificial intelligence enables them to measure the influence these physicians have with more meaning and value by quantifying the success of a campaign.
Pharmaceutical companies can use machine learning for influencer marketing by mapping the right physician for their campaign needs. The criteria could be the topics they extensively discuss or write about, their experience, or others. This will enable companies to isolate and reach out to a relevant target audience.
As Single Grain correctly highlighted in the following infographic, — using machine learning and AI you go beyond the superficial layer of the “number of followers”, and dig deeper into the hidden layers to predict the outputs.
8. Recruitment for Clinical Trials
Nearly 80 percent of clinical research and trials either fail to finish on time or get delayed by six months or more. The reason being that 85 percent of these trials fail to retain enough patients, with an average churn of around 30 percent.
With machine learning and AI, healthcare companies can extract pertinent EMR information to sift through physician notes efficiently and effectively. The data collected can then be used to identify appropriate patients for clinical trial enrollments. Even during the span of the trial, the technology can be used to predict patient churn using real-world evidence (RWE) from their medical history, giving the companies a buffer to find replacements.
9. Manufacturing Optimization
Statistical analysis has remained at the core of ensuring greater product quality and maintaining minimal consumer risk. The data helps engineers understand why and how a manufacturing process can be optimized to yield an expected quality with a known certainty.
Simply put, statistical analysis helps to ensure that the best of practices are followed in the manufacture of pharmaceutical products and medical devices, for consumer safety. Along with machine learning, pharma companies can improve their manufacture efficiency, product yield and cost, and final product quality.
10. Digital Health, Digital Medicines, and Patient Safety
Digital Health is an emerging trend in the healthcare industry. Both pharma and tech companies have been developing mobile applications that can keep track of health-related parameters, record drug prescriptions, improve compliance, and remind patients about upcoming medical appointments. Digital medicines are pharmaceuticals which combine drugs with an ingestible sensor used to monitor whether the patient correctly followed medical advice.
These systems are not only designed to record and store data but also to share information with health-care professionals, reducing communication asymmetries between patients and physicians.
Machine Learning will play a big role in the development of more effective Digital Health applications, for example, by incorporating models that can send alerts and information at the right time to the right person. Ultimately this will result in increased patient safety.
Adoption of AI in medicine and pharma
While the application of data science and AI in the medicine and pharma industry looks promising, there are still challenges that are yet to be addressed.
Experfy’s online course instructor on Machine Learning Assisted Clinical Medicine, Damiano Fantini says, “First, there are important ethical considerations to take into account when collecting, storing, processing, and analyzing biomedical data from human subjects. Second, despite recent efforts to standardize data storage and communication protocols (especially in fields such as medical imaging and electronic health records), data variety is still a problem.
“In addition, machine learning is highly dependent on qualitative information. But when it comes to pharma, the industry requires some element of human intuition as well.”
Sometimes, making the right diagnosis takes more than just a set of rules. It requires experience in addressing certain symptoms and health outcomes over the years. Similarly, the choice of the best treatment for a patient requires information about the most recent developments in the field, current clinical trials, and recent scientific advances.
For these reasons, successful healthcare ML applications will require medical experience to be fully engaged. Damiano further highlights that “machine learning with human moderation is the key to leveraging the best of both worlds in making the medicine and pharma industry more effective.”
— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —
Vanhishikha Bhargava is the Content Marketer at Experfy, a Harvard-incubated startup enabling on-demand consulting and upskilling platform for companies of the future.
Want to write an article for our blog? Read our requirements and guidelines to become a contributor.
Learn how AI impacts other industries:
8 Real-Life Applications of Artificial Intelligence in eCommerce
AI in Insurance: Business Process Automation Brings Digital Insurer Performance to a New Level
7 Ways Airlines Use Artificial Intelligence and Data Science to Improve Operations
Data Science and AI in the Travel Industry: 12 Real-Life Use Cases
How to Implement Data-Driven Decision Making in School Education
7 Ways Data Science Is Reshaping Healthcare
5 Ways AI Will Transform the Logistics Industry