The state of data science and machine learning in healthcare
As healthcare continually advances through digitalization and digital transformation, it’s become one of the best equipped industries to maximize usage of data science and machine learning. Since 2015, venture capital investments in AI companies in healthcare have seen a 22X increase in Europe alone (McKinsey).
Data science and machine learning are transforming healthcare across several verticals, from patient care to pharmaceuticals and more. But scaling the impact of data science in healthcare requires careful consideration of many challenges, including compliance, data governance and oversight, data culture, and the availability of data skills.
The opportunity for data science and machine learning in healthcare
Today, the healthcare space is ripe for machine learning and data science due to the high volume of healthcare data and the many applicable use cases for public health outcomes. According to Statista, the amount of data generated yearly in the global healthcare industry is around 2,314 exabytes (1 exabyte = 1B gigabyte), marking a 15X increase in the amount of data generated in global healthcare since 2013.
There are countless benefits for society in terms of improved population health outcomes. Deloitte estimates that efficiencies gained through data science and machine learning use cases can save between 380,000 to 403,000 lives in Europe alone. Data science and machine learning can be integrated throughout the patient journey from prevention and early detection, to diagnosis, to treatment and care management.
For example, individuals can leverage wearables and personalized applications for early disease detection and prevention, or experience lower wait times with deep learning powered medical imagery analysis. The research and development space is also gearing up to create massive gains in health outcomes, from automated drug discovery to DeepMind’s progress on protein folding with its AlphaFold algorithm.
Healthcare providers across verticals can also realize massive gains in terms of cost savings and improved efficiencies. For instance, operationalizing data science and machine learning can save the European healthcare system between 170.9 to 212.4 billion euros a year (Deloitte). Efficiency gains can run the gamut from pharmaceutical companies improving supply chain flows, to insurance providers optimizing commercial spend by predicting customer churn, to increased productivity through workflow automation across several verticals.
Data science and machine learning use cases in healthcare
Data science and machine learning use cases can improve health outcomes for individuals and automate time-consuming administrative tasks for healthcare professionals.
Appointment management: With the use of machine learning and rule-based artificial intelligence, healthcare providers can optimize for patient outcomes and alleviate resource mismanagement with automated appointment management.
Early diagnostics and prevention: Health monitoring applications and wearables leverage machine learning and descriptive analytics to provide important insights over various aspects of an individual’s health. These tools can empower individuals to make data-driven decisions around their health, and diagnose potential diseases early on.
Patient triage: By using machine-learning powered symptom checker applications, healthcare providers can triage patients based on need and urgency. This can lead to massive decreases in wait times for patients and massive efficiencies gained for healthcare providers.
Medical imaging and diagnostics: Arguably one of the most important use-cases for data science and machine learning in healthcare (McKinsey), medical imaging and diagnostics promises a wide range of gained efficiencies and improved health outcomes across the board. With the use of deep learning, healthcare providers can automate workflows and deliver value to patients faster.
Business Processes and Management
As in any industry, healthcare applications for data science and machine learning span a range of use cases for improved operational efficiency and customer experience.
Robotic process automation: Using a combination of machine-learning and rule-based artificial intelligence, healthcare providers across verticals can streamline workflows and digitalize processes.
Customer churn: Insurance providers can use machine learning to predict which customers will churn, which can help them retain customers and optimize marketing spend.
Chatbots: With the use of chatbots, healthcare providers from hospitals to insurance agencies can deliver better customer service and faster time to value for healthcare consumers.
Business Intelligence: Business intelligence combines business analytics, data manipulation, and visualization to help organizations make more data-driven decisions. By leveraging data insights, healthcare providers can gain more visibility into financial operations, automate compliance reporting, and much more (Villanova University).
With AI-powered drug discovery and improved supply chain management, pharmaceutical companies can leverage data science and machine learning to provide more value for individuals.
Drug discovery: According to the AI Index Report 2021, AI-powered drug discovery startups received the most private AI investments across all industries. The promises of ML-based drug discovery are starting to bear fruit, and may result in massive gains in population health outcomes.
Supply chain planning: The use of data science and machine learning in supply chain planning can reduce time to production, reduce wait times for vaccine deliveries, and optimize supply chain spend for pharmaceutical companies.
Forecasting excellence: Using simple to complex forecasting tools, pharmaceutical companies can leverage population health data to forecast supply and demand for certain medications, as well as optimize business processes across the board for finance, marketing, sales, and more.
Improving clinical trial processes: The use of data science and wearables can reduce risks for patients by automatically monitoring and flagging any adverse effects during clinical trials. Moreover, machine-learning powered applications can speed up clinical trials by assessing patient eligibility, pre-screening, and randomization.
Challenges and risks to operationalizing data science and machine learning in healthcare
Data Quality and Infrastructure
A significant challenge for any organization trying to operationalize and scale data science and machine learning is enabling a modern, secure, centralized, and discoverable data infrastructure (DataCamp). The issue is especially true for healthcare organizations where data is still being digitized, and building large datasets is hindered by a lack of data interoperability and misalignment on quality between healthcare providers (McKinsey).
Compliance and governance
Since healthcare organizations collect highly valuable and sensitive data, governance and compliance are central aspects of operationalizing data science and machine learning in healthcare (Collibra). While regulations differ between regions (McKinsey), the healthcare industry contains one of the most complex regulatory landscapes. For example, the Health Insurance Portability and Accountability Act (HIPAA) in the United States dictates national standards for protecting and managing patients’ health data.
These types of health-specific regulations are only one aspect of the complex regulatory landscape for the healthcare industry. Data protection laws such as the General Data Protection Regulation (GDPR) in the European Union and the California Consumer Protection Act (CCPA) create additional complexity for healthcare organizations when building and linking disparate datasets for data science and machine learning applications. As such, creating strong data governance and compliance is essential to operationalizing data science and machine learning in healthcare.
Enabling data infrastructure and reaching consensus on regulatory frameworks are essential to operationalizing data science and machine learning in healthcare. But the most significant threat to the adoption of these technologies in healthcare is the data skills gap. In fact, according to research by Qlik, when ranking the data literacy of various industries, the healthcare industry is the least performing one. The lack of data skills within healthcare organizations hinders the utilization of data science and machine learning across verticals and limits the ability of organizations to create buy-in around data initiatives.
For example, frontline health workers need to have basic AI literacy to be able to understand and interact with machine-learning based systems and applications (McKinsey). Managers and leaders at pharmaceutical or insurance companies need to understand what’s possible with data science and machine learning, so they can drive data initiatives and contribute to growing a data-driven culture (DataCamp).
Merely hiring skilled data workers isn’t enough. There’s a shortage of talent that understands the complexities of healthcare and a shortage of data science talent in general (Forbes).
A lack of data literacy is the single largest enemy we are fighting. As the world becomes increasingly driven by data, it’s an even bigger challenge. Everyone has to understand the basics, and we have to be able to convey that in a way that is intuitive and fun
How data training solves the healthcare industry’s data challenges
Addressing the data skills gap is foundational to operationalizing data science and machine learning in healthcare. According to the World Economic Forum, digital and data training in healthcare and pharmaceuticals can boost global GDP by more than 400 billion dollars by 2030. Creating a continuous learning culture is imperative to build the next generation of leaders in healthcare (McKinsey), who will combine data and biomedical sciences to scale and operationalize the value of data science and machine learning in healthcare.