• Writing
  • Books
  • About
  • Menu

For Context

  • Writing
  • Books
  • About

The Grid Paradigm

June 16, 2024

“Good ideas are not adopted automatically. They must be driven into practice with courageous patience.” - Hyman Rickmore

As an integral cornerstone of societal infrastructure, energy systems underpin our daily activities and catalyze the evolution of our civilization. It has long been axiomatic that energy demand and GDP growth are correlated. When economies grow, their energy consumption/demand increases. If access to energy becomes constrained, GDP growth pulls back.

However, during my time as a college student in Austin, I have been witness to, and affected by, two total grid failures, supplemented by numerous power outages over a span of four years.

To understand the broader context of these issues, how they came to be, and the grid inter-workings in general, it is essential to understand the concept of the “three-legged stool” of energy: production, delivery, and consumption. For the system to work effectively, each leg must be robust and in balance.

  • Production: The generation and extraction of energy.

  • Delivery: The moving of power from where it is generated to where it is consumed.

  • Consumption: The demand side of energy, encompassing how energy is utilized by end-users, including households, industries, and services.

Difficulties with the grid predominantly originate from challenges in the delivery. The production aspect is largely influenced by incentive structures associated with renewable & nonrenewable sources, while consumption is driven by demand. These elements significantly impact the grid's proficiency in energy transmission (delivering), but are far from the core of what the grid does.

Breaking Down the Issue

As mentioned, the core issue can be examined using the three-legged stool framework. This reveals a misaligned incentive model, caused by regulations implemented over the past century, and an absence of basic innovation.

Production

The push for clean energy and increased electrification of the world has led to a growth in variable sources of energy.

  • Diversification of Energy (Increased Cost & Volatility): The emphasis on developing and deploying affordable clean energy sources has resulted in technological advancements. These have made variable sources like solar and wind more cost-effective. However, according to IREA, while solar has become the cheapest source of new electricity generation, it has also complicated energy delivery as a result of increased peak-trough volatility. The variability requires more sophisticated grid management and storage solutions for delivery.

Delivery

The process of delivery has become increasingly challenging due to several factors:

  • Aging Infrastructure: Much of the current transmission and distribution infrastructure is outdated and in need of upgrades or replacement. This aging infrastructure is less capable of handling the dynamic demands of modern electricity delivery, especially with the integration of more variable and distributed energy sources.

  • Grid Reliability: The reliability of the grid has been declining, not only because of the growing use of renewables but also due to increased frequency of extreme weather events and insufficient maintenance of distribution systems. These issues have resulted in more frequent and severe outages (U.S. Department of Energy, 2021).

  • Increasing Costs: As the demand for electricity grows, driven by trends such as the electrification of transportation and heating, the costs associated with delivering power to consumers have risen. This is partly because transmission lines must be sized to handle peak loads, even if those peaks occur infrequently.

  • Regulatory and Permitting Delays: Modernizing the grid is hindered by regulatory and permitting challenges. For example, building new transmission lines often faces significant delays due to the need for approvals across multiple jurisdictions and the complexities of land acquisition.

Consumption

As mentioned, there is an increasing effort to electrify as much of the economy as possible. While these changes are crucial for reducing greenhouse gas emissions, they also increase the demand for electricity and the complexity of managing grid loads.

  • Consumption Patterns: As consumption patterns evolve, the grid must adapt to these new loads. For instance, electric vehicles create new demand peaks.

History of the Grid

To better comprehend the issues at hand, it is crucial to review the extensive history of the grid and its influence on its current state. However, because the information is detailed, it may be beneficial to quickly go through it, skim the text, and examine the diagrams.

The historical context of grid information was largely obtained from the excellent four-part substack by Construction Physics, along with other reputable sources.

Early History

In the early 20th century, U.S. electricity was a luxury, used by less than 5% of industries and 8% of homes. Today, it's a widespread utility, with over 89% of the global population having access, increasing to 97% in urban areas. The modern electrical grid's inception can be traced back to Edison's invention of the incandescent light bulb in 1879. This innovation replaced gas lighting, which was labor-intensive and provided insufficient illumination. Early electric lighting technologies, such as arc lamps (1858), were unsuitable for domestic use due to their intense brightness. Edison developed a practical incandescent lamp by heating a resistive material. His design, which reduced copper use in wiring, made electrical lighting affordable, leading to over 1,100 central power stations by 1891.

Currents: AC & DC

Edison's direct current (DC) was inefficient for long-distance power supply due to significant power loss as heat in wires. Just like water in a pipe, where current is the water flow and voltage is the pressure, high current at low voltage leads to substantial losses over distance.

A DC (Direct Current) circuit operates through the interplay of several key components. The battery serves as the power source, with current exiting its positive (+) terminal and returning to its negative (-) terminal. Wires connect the battery to other components, creating a closed loop for current to flow. Within this loop, a resistor often plays a crucial role by limiting the amount of current and protecting other components from excessive flow. The current in a DC circuit follows a consistent direction, moving from the battery's positive terminal, through the wires and components like the resistor, and back to the negative terminal.

  • Resistor (Limits Current): The resistor restricts the amount of current flowing through the circuit, helping to protect other components from excessive current

In contrast, alternating current (AC) offered a solution. AC systems, championed by Nikola Tesla and backed by George Westinghouse, used transformers to step up the voltage for long-distance transmission and then step it down for local use. This allowed power to be transmitted over much longer distances without significant losses, making it more suitable for widespread distribution.

An AC (Alternating Current) circuit is powered by an alternating power source, typically referred to as an AC supply. Unlike DC circuits, the current in an AC circuit periodically changes direction, usually in a sinusoidal waveform. This means the flow of electricity reverses its direction many times per second (e.g., 50 or 60 times per second in most household systems). Wires serve as the current path, connecting the AC supply to other circuit components and forming a loop for the current to travel. A resistor is often included to limit and control the current flow, protecting other components from excessive current. The alternating nature of the current is a defining feature of AC circuits, with electrons moving back and forth rather than in a constant direction. This back-and-forth movement of current allows for efficient long-distance power transmission and is the standard for household and industrial electrical systems worldwide. The ability to easily transform voltage levels using transformers is a key advantage of AC systems, enabling the high-voltage transmission and lower-voltage distribution that characterizes modern power grids.

Despite Edison’s resistance and his campaign against AC’s safety, AC technology eventually prevailed. The ability to convert between AC and DC using devices like the rotary converter, invented by Charles Bradley in 1888, allowed both systems to coexist initially. By 1917, AC had become the dominant method for electrical transmission, with 95% of power stations generating AC.

An AC to DC converter, also known as a rectifier, transforms alternating current (AC) from a power source like a wall socket into direct current (DC) for use in electronic devices. The process begins at the AC input, where the alternating current enters the circuit. A ground connection is incorporated for safety, providing a path for excess charge to dissipate into the earth. The incoming AC then passes through a transformer, which adjusts the voltage to the desired level by either stepping it up or down. Next, a bridge rectifier, typically composed of four diodes, converts the AC into pulsating DC by allowing current to flow in only one direction. However, this pulsating DC isn't smooth enough for most applications, so a filter capacitor is employed to smooth out the voltage fluctuations, acting like a reservoir that fills during voltage peaks and discharges during troughs. The result is a steady DC output suitable for powering various electronic devices and components. This conversion process is crucial for many modern electronics that require DC power but are plugged into AC outlets.

While a comprehensive understanding of AC to DC transformers or the specifics of AC and DC currents might not be necessary unless you're an engineer, it's still important to have a basic grasp of these concepts. These principles are frequently referenced in discussions around electricity, and understanding them can provide valuable context.

The Grid Today & Regulatory Dynamics

Post the 1930s, regulatory dynamics began to shape the formation of the grid. To clarify terminology from the outset, here are all the key regulatory bodies and other noted key players that play a significant role in the grid and its dynamics.

  • Independent System Operators (ISOs): Organizations that coordinate, control, and monitor the operation of the electrical power system within a specific region. They ensure the reliable delivery of electricity, maintain balance between supply and demand, and oversee electricity transactions.

  • Regional Transmission Organizations (RTOs): Entities similar to ISO but typically cover larger geographical areas. They coordinate, control, and manage high-voltage electricity transmission networks across multiple states or countries. They are responsible for ensuring the efficient and reliable delivery of electricity to vast regions.

  • Federal Energy Regulatory Commission (FERC): FERC plays a central role in overseeing the interstate transmission of electricity, natural gas, and oil. It regulates the wholesale electricity markets and ensures fair competition by overseeing the activities of Independent System Operators (ISOs) and Regional Transmission Organizations (RTOs). FERC's regulatory purview includes approving tariffs and setting rates for transmission services, ensuring that these rates are just and reasonable.

  • State Public Utility Commissions (PUCs): At the state level, Public Utility Commissions are responsible for regulating retail electricity rates and services within their respective states. PUCs oversee the operations of investor-owned utilities (IOUs) and, in some states, have a role in guiding the implementation of renewable energy policies and energy efficiency programs.

  • Environmental Protection Agency (EPA): The EPA's regulations impact the energy sector significantly, particularly through policies related to emissions and environmental compliance. Regulations such as the Clean Air Act and the Clean Power Plan (though legally challenged and evolving) influence the types of power generation that are economically viable and encourage shifts towards cleaner energy sources.

  • Department of Energy (DOE): The DOE provides policy guidance, research funding, and support for energy innovation. It plays a critical role in advancing technologies that enhance energy efficiency, grid reliability, and the integration of renewable energy sources.

  • North American Electric Reliability Corporation (NERC): NERC is responsible for developing and enforcing reliability standards for the bulk power system. It ensures that the U.S. electric grid operates reliably and securely, especially as the grid incorporates more variable renewable energy sources.

The Expansion of the Grid (1930 - Present)

By 1930, the U.S. consumed around 114 billion kilowatt-hours of electricity annually, generated by over 40,000 power plants. Coal-fired plants produced 56% of this, while hydroelectric power contributed 34%. The New Deal in the 1930s expanded and regulated electricity production and distribution, notable projects included the Hoover Dam and the establishment of the Rural Electrification Administration (REA). After the 1929 crash revealed the precariousness of utility empires like Samuel Insull's, stringent regulation was imposed on utility holding companies through the Public Utility Holding Company Act (PUHCA) and the Federal Power Act.

The post-World War II period saw a surge in electricity demand and a corresponding expansion in generating capacity. Between 1940 and 1945, U.S. electricity consumption increased by 60%, even as generation capacity only rose by 25%. This period highlighted the efficiencies gained through the interconnection of regional power grids, which not only facilitated capacity expansion but also enhanced the system's resilience to potential disruptions. By the 1960s, the U.S. electric power industry had become the largest by asset value, surpassing even the next two largest industries combined. During this time, the country utilized more than 40% of the world's electric power, despite global advances in electrification. The construction of nearly 80,000 miles of transmission lines from 1950 to 1963 exemplified the scale of infrastructure development required to meet this demand. Most power providers had integrated their systems into a few large, interconnected networks, enhancing operational efficiency and reliability.

However, the industry's rapid growth began to show signs of strain by the late 1960s. The environmental movement, bolstered by the National Environmental Policy Act (NEPA) of 1969, imposed new constraints on infrastructure projects. Additionally, the rising costs and safety concerns associated with nuclear power plants, which had been heavily promoted as a future energy source, further complicated the industry's expansion strategies.

The 1970s marked a period of significant upheaval for the U.S. electricity industry. The 1965 Northeast blackout, which left 30 million people without power for up to 13 hours, exposed the vulnerabilities of a highly interconnected power system. The ambitious Golden Spike Operation of 1967, which attempted to link the U.S. power grid into a single system, ultimately failed due to instability, leading to the removal of AC ties and the adoption of high-voltage DC transmission lines for inter-grid power sharing.

The energy crisis of 1973 exacerbated these challenges, causing demand growth to plummet and electricity prices to skyrocket by nearly 390% in real terms between 1973 and 1983. This period saw the introduction of significant legislative measures aimed at restructuring the industry. The Public Utility Regulatory Policies Act (PURPA) of 1978, part of the National Energy Act, mandated utilities to purchase power from non-utility producers at a rate equivalent to the utility's avoided cost. This policy fostered the emergence of Independent Power Producers (IPPs) and set the stage for greater competition in power generation.

In response to escalating costs and the inefficiencies of the existing system, the Federal Energy Regulatory Commission (FERC) began restructuring the industry in the mid-1980s to introduce competitive market principles. The Energy Policy Act of 1992 further facilitated this transition by creating a new class of power suppliers, the Exempt Wholesale Generators (EWGs), who could sell power without being regulated as utilities. This act also mandated the establishment of Independent System Operators (ISOs) to manage transmission lines and wholesale power markets, aiming to reduce costs and enhance market competition.

The deregulation efforts of the 1990s culminated in a mixed legacy. The FERC's initiatives to create competitive wholesale electricity markets led to the formation of Regional Transmission Organizations (RTOs), which assumed broader responsibilities for system reliability and market operations over wide geographic areas. However, the implementation of deregulation faced significant challenges, most notably in California.

In California, market design flaws and regulatory constraints led to a severe energy crisis in 2000-2001. Wholesale electricity prices soared, driven by market manipulations by companies like Enron, which exploited the system's vulnerabilities. Utilities, unable to pass on the increased costs to consumers due to rate caps, faced financial ruin, resulting in widespread blackouts and a collapse of public confidence in deregulated markets.

Despite these setbacks, deregulation persisted in various forms across the U.S. Today, approximately two-thirds of the country's electricity is supplied through ISO/RTO-managed markets, while the remaining third is provided by traditional vertically integrated (the chain of production to delivery is owned by utility companies) utilities. The industry continues to grapple with the balance between market competition and regulatory oversight, as it navigates the challenges of modern energy demands and the transition to more sustainable sources.

The regulatory architecture that dominates the U.S. energy industry today is a complex and multifaceted construct, engineered to harmonize various interests. These range from guaranteeing a reliable supply of power and safeguarding consumers, to promoting competition and facilitating the shift towards renewable energy. For a more nuanced understanding of the operational differences between regulated and deregulated markets, I will delve into this in the subsequent section. This information is instrumental in devising solutions for the identified issues within the "three-legged stool" framework previously discussed.

Regulated vs. Deregulated Markets

Regulated and deregulated markets significantly impact the delivery of power to end consumers. The aim of deregulation is to enhance competition, which in turn reduces costs for the end user. However, this does not always work well in practice. In a deregulated structure, retail customers can select their electricity suppliers. Meanwhile, utilities must sell off their generation assets and purchase electricity from a new wholesale market. The key players in this process can be confusing, and I believe they can be better explained through the diagram below.

  • Energy Producer: Companies or entities that generate electricity from various sources like coal, natural gas, nuclear, hydro, wind, or solar power. In regulated markets, the energy producer tends to be monopolized and owned by the utility companies. Produce electricity that is then sold to utility companies for distribution to consumers.

  • Utility Company: A company that manages the entire process of producing, transmitting, and distributing electricity to consumers. Responsible for generating electricity, transmitting it through high-voltage power lines, and distributing it to end customers. Operates as a regulated monopoly under state supervision.

  • Distribution Network: The network of lower-voltage power lines that deliver electricity from substations to homes and businesses. Transports electricity from the utility company’s transmission system to end customers.

  • State Public Utility Commission (PUC): A state-level regulatory body that oversees the operations of utility companies. Regulates utility companies to ensure reliable service, fair pricing, and compliance with state laws.

  • Wholesale Market: A market where large quantities of electricity are bought and sold between producers and entities like utility companies or retail suppliers. Facilitates the trading of electricity among producers, utilities, and suppliers to balance supply and demand.

  • Retail Supplier: Companies that purchase electricity from the wholesale market and sell it to end customers. Compete with other suppliers to provide electricity to consumers, offering various pricing plans and services.

One More Thing: Supply-Demand Peaks & Troughs

This is one concept that I think is crucial to understand the underlying reasons for issues with the grid, but I was stuck one where to stick it in. Supply-Demand Peaks & Troughs are a crucial concept to understand the management of electricity grids.

  • Peaks: These are periods when electricity demand is at its highest. Peaks often occur during specific times of the day, typically in the morning when people wake up and start their daily activities, and in the evening when they return home and use household appliances.

  • Troughs: These are times when electricity demand drops to its lowest levels. Troughs usually happen late at night or early in the morning when most people are asleep, and industrial activities are minimal. This can be signified through terms such as the “Base Load.”

As mentioned, electricity providers must constantly adjust the supply to match these demand fluctuations. This dynamic balancing act involves various components and strategies within the power grid system. Excessive demand during peak times can push the grid beyond its supply capacity, risking overloading and potential damage to the system. To mitigate this risk, grid operators may implement load-shedding measures or reduce voltage (brownouts) to balance the load. Conversely, insufficient supply can occur due to failures at power plants, faults in transmission lines, or fuel shortages, leading to power outages if the available electricity does not meet the demand.

However, electricity demand does not remain constant throughout the day. It fluctuates, with significant increases during peak times, such as in the mornings and evenings when people are most active, and drops to its lowest during the late night and early morning hours. These fluctuations are addressed by peaking power plants, which can quickly ramp up or down their electricity production to meet sudden spikes in demand. The integration of renewable energy sources, such as solar and wind, adds another layer of complexity due to their intermittent nature.

Solutions

With a comprehensive understanding of the historical context, grid structure, and market dynamics, we can now focus on potential solutions to the challenges faced by the electricity grid. These challenges are predominantly in delivery, significantly impacted by the increase in variable energy sources and inconsistent energy demand.

Stabilizing Production

Although the focus here is on delivery, here are a few solutions for production. This is a highly contentious issue, with an ongoing debate about how best to generate electricity sustainably. Renewable energy sources like wind and solar, while environmentally friendly, add volatility to the grid due to their intermittent nature. To mitigate this, investing in grid-scale energy storage solutions is crucial. Technologies such as lithium-ion batteries, flow batteries, and hydrogen storage can store excess energy produced during peak generation times and release it during periods of high demand or low generation. This capacity to balance the supply and demand fluctuations of renewables helps stabilize the grid.

On the nuclear front, modern small modular reactors (SMRs) represent a promising development. These reactors are designed to be safer and more flexible than traditional large reactors and can provide a reliable, low-carbon baseload energy source. Easing regulatory restrictions and providing incentives for nuclear development could significantly enhance energy security and provide consistent power that complements the variability of renewable energy sources.

Additionally, encouraging the adoption of Distributed Energy Resources (DERs), such as rooftop solar panels and micro wind turbines, can reduce the load on centralized power plants and minimize transmission losses. These small-scale, local generation technologies can be integrated into the grid through smart inverters and managed to optimize energy flows and enhance resilience. By generating electricity closer to where it is used, DERs contribute to a more robust and decentralized energy system.

Modernizing Delivery Infrastructure

The next set of solutions focuses on improving the delivery infrastructure. The aging transmission and distribution systems need substantial investment to replace outdated components like old power lines, substations, and transformers with modern, more resilient equipment. This is especially critical because a significant portion of grid failures occur within the distribution system, the low-voltage network that directly delivers electricity to consumers. In the United States, most distribution lines are mounted on poles and are vulnerable to weather-related disruptions, such as high winds, ice, and extreme heat. In contrast, countries like Germany have buried a large percentage of their distribution lines, protecting them from such conditions.

Integrating advanced technologies into the grid can also enhance its responsiveness and reliability. Smart meters and real-time monitoring systems allow utilities to quickly identify and resolve issues, minimizing downtime during outages. Advanced grid management software can optimize electricity flow, integrating data from various sources to predict demand, manage energy flows, and coordinate with DERs. This is essential for effectively managing the complexities introduced by the increasing presence of variable renewable energy sources.

Expanding the deployment of energy storage systems at both the grid-scale and consumer levels is another crucial step. Large-scale storage solutions, such as lithium-ion batteries and pumped hydro storage, can act as buffers, absorbing excess energy during periods of high production and releasing it when demand increases or generation decreases. This not only stabilizes the grid but also reduces the need for additional peaking power plants and minimizes transmission losses. On the consumer level, home batteries paired with solar panels provide a backup power source during outages and enable participation in demand response programs, thus enhancing grid reliability and reducing dependence on the central grid. Emerging technologies like hydrogen storage and compressed air energy storage offer promising solutions for long-duration storage, complementing the quick response capabilities of batteries.

Finally, promoting the development of microgrids and community-based energy projects can enhance local resilience and reduce reliance on centralized power systems. Microgrids can operate independently of the main grid, ensuring a continuous power supply during outages and proving particularly valuable in disaster-prone or remote areas. Community energy projects, such as neighborhood solar farms, allow for collective investment in local renewable generation and shared storage, lowering costs and fostering community involvement in energy management. These projects also contribute to overall grid stability by providing ancillary services like frequency regulation and voltage support. To encourage their development, regulatory support and financial incentives are necessary. Simplifying interconnection processes and establishing clear policies for shared energy savings can accelerate their adoption, creating a more resilient, sustainable, and decentralized energy system that empowers communities and enhances overall grid stability.

Coding Healthcare

May 25, 2023 in Healthcare

Throughout human history, the practice of medicine has continuously evolved, adapting to new knowledge, technologies, and societal needs. From the earliest healing rituals of ancient civilizations to the complex healthcare systems of today, medicine has undergone transformative changes that have improved the quality of care and enhanced patient outcomes. In recent years, the emergence of Artificial Intelligence (AI) has brought about a new era of possibilities in healthcare, revolutionizing the way we approach diagnosis, treatment, and patient care. This thesis sets the stage for exploring the impact of AI on healthcare, tracing its roots back to the earliest instances of medicine and examining its transformative journey through the ages. By understanding the historical context, we can grasp the significance of AI's role in reshaping the healthcare landscape, empowering healthcare providers, and ultimately improving patient well-being.

From the Beginning

Artificial Intelligence (AI) has transcended the realm of science fiction to become an integral part of our modern world. We definitely see that now with consumer-facing innovations like ChatGPT (OpenAI) & Bard (Google). From its beginnings in the 1950s with the advent of the Turing Test, designed to assess machine intelligence, AI has made great strides. Despite encountering a period of stagnation, recent breakthroughs in deep learning and the rise of generative AI have breathed new life into the field.

Below are essential terms to understand the infrastructure of AI.

Artificial Intelligence: Machine intelligence has cognitive functioning similar to humans, such as "learning" and "problem-solving."

Machine Learning: Mathematical algorithms automatically built from given data and predict or make decisions in uncertain conditions without being explicitly programmed.

Deep Learning: A subset of machine learning techniques composed of multilayered neural network algorithms.

Artificial Neural Networks: A multilayered, interconnected network that consists of an input layer, hidden connections, and an output layer.

In healthcare, AI has a history dating back to the 1970s, with a program named MYCIN which marked an important milestone. MYCIN was an AI program developed to assist physicians with diagnosing and recommending treatments for infectious diseases. From that breakthrough, a cascade of advancements has propelled the development of numerous AI-powered healthcare solutions.

Healthcare Industry Overview

Healthcare Overview

The healthcare industry is a complex system with various stakeholders. The important players and their inner workings are listed below:

  1. Healthcare Providers: Hospitals, clinics, private practices, and healthcare professionals form the backbone of the industry. They play a crucial role in preventing, diagnosing, treating, and rehabilitating illnesses and diseases. These providers are on the front lines of patient care, delivering essential services and treatments.

  2. Pharmaceutical and Medical Device Manufacturers: These companies are responsible for developing and producing medications, vaccines, medical devices, and equipment. They conduct extensive research and clinical trials to ensure safety and efficacy. Pharmaceutical manufacturers work in collaboration with healthcare providers to ensure access to necessary medications for patient treatment.

  3. Insurance Companies: Insurers play a vital role in healthcare by providing coverage and financial protection for individuals' medical expenses. They offer health insurance plans that help mitigate the cost of healthcare services, medications, and treatments. Insurance companies work closely with healthcare providers to establish networks and negotiate reimbursement rates.

  4. Government Agencies: Government entities, such as regulatory bodies and public health departments, have a crucial role in shaping healthcare policies, regulations, and standards. They oversee licensing, safety regulations, and quality control within the industry. Government agencies also promote public health initiatives and provide funding for research and healthcare programs.

Market Overview

The global healthcare industry, with its diverse array of stakeholders and players, operates within a vast and ever-evolving landscape. In 20221, the global healthcare market was valued at $7.7 trillion, and projections indicate it will surely to $10.5 trillion by 2028, reflecting a (CAGR) of approx. 5.5%. Notably, the US boasts the largest healthcare market globally, with a value of $808 billion in 2021.

The digital healthcare market is projected to reach an impressive value of $780 billion by 2030. Presently valued at $8.2 billion, the broader market for generative AI, a key technological enabler in healthcare, is expected to experience remarkable growth, reaching an estimated worth of $126 billion by 2031.

While the healthcare industry is propelled by numerous growth drivers, it also confronts significant challenges. An aging population, coupled with the escalating prevalence of chronic diseases, places substantial pressure on healthcare systems worldwide. Addressing the shortage of healthcare workers, especially in critical areas such as primary care and nursing, remains imperative in ensuring access to quality care for all. Artificial Intelligence Healthcare Applications

Artificial Intelligence, with its advancements in natural language processing, image analysis, and other related technologies, is already significantly impacting major segments of the healthcare industry, including primary care, electronic health record systems, and specialist care.

Artificial Intelligence Healthcare Applications

AI, with its advancements in natural language processing, image analysis, and other related technologies, is already significantly impacting major segments of the healthcare industry, including primary care, electronic health recording systems, and specialist care.

Primary Care

Primary Care, the first point of contact for a patient, is a crucial aspect of the healthcare industry, accounting for 40% of all visits in the US in 2022. Primary care is broad, encompassing:

  1. Preventive care includes annual physicals, vaccinations, and screenings for diseases such as cancer and heart disease.

  2. Chronic disease management includes managing conditions such as diabetes, hypertension, and asthma.

  3. Acute care: This includes treating injuries and illnesses that are not chronic, such as the common cold, the flu, and ear infections.

  4. Mental health care includes diagnosing and treating mental health conditions such as depression, anxiety, and substance abuse.

  5. Social services: This includes providing support and resources to patients who are struggling with social or economic problems, such as homelessness, poverty, and unemployment.

  6. Education and counseling: This includes providing patients with information about their health, how to make healthy choices, and how to manage their conditions.

As one of the fields most affected by AI, primary care is expected to benefit greatly from the technology’s ability to process large amounts of data and ultimately provide more personalized care to patients. With the help of AI, PCP will be able to make more accurate diagnoses, create more effective treatment plans, and monitor patient health on a more continual basis and more closely.

The segments of primary care most affected by AI are preventative care and chronic disease management.

Preventive Care

Regarding preventative care, AI has several potential applications that can make services in this area more effective.

  1. Population Health Management:

    1. PHM is a healthcare approach that focuses on the health of a population of people rather than on individual patients. PHM uses data and analytics to identify and address health risks and problems within a population. AI can be used to support PHM in a number of ways, including:

      1. Identifying patients who are at risk for certain health conditions

      2. Tracking patients' health over time

      3. Developing and implementing interventions to improve population health

    2. Companies:

      1. IBM Watson Health: a company using AI to help healthcare organizations manage their populations of patients. Watson Health's platform can be used to collect and analyze data from a variety of sources, including electronic health records (EHRs), claims data, and social media. This data can then be used to identify patients who are at risk for certain conditions, such as heart disease or diabetes. Once patients are identified, healthcare organizations can intervene early to prevent these conditions from developing or worsening.

      2. Verily: a company that is using AI to improve population health management. Verily's platform, DeepMind Health, can track patients' health data over time. This data can then be used to identify patterns and trends that may indicate a health problem. Once a problem is identified, Verily can provide patients with personalized treatment or prevention recommendations.

  2. Remote Patient Monitoring

    1. Remote patient monitoring (RPM) is using technology to collect and monitor patients' health data remotely. An example of this is the development of wearable technologies. RPM can track various data points, including heart rate, blood pressure, blood sugar levels, and activity levels. This data can then be shared with healthcare providers, who can use it to monitor patients' health and make treatment recommendations. AI can support RPM in a number of ways, including:

      1. Automating the collection and analysis of data

      2. Identifying patterns and trends in data

      3. Generating alerts and notifications for healthcare providers

    2. Companies

      1. Apple HealthKit is a platform that allows patients to track their health data using their Apple Watch or iPhone. HealthKit can be used to track a variety of data points, including heart rate, sleep, and activity levels. This data can then be shared with healthcare providers, who can use it to monitor patients' health remotely.

      2. Fitbit is another company that offers a remote patient monitoring platform. Fitbit's platform can be used to track a variety of data points, including steps taken, calories burned, and sleep quality. This data can then be shared with healthcare providers, who can use it to monitor patients' health remotely.

  3. Diagnostics

    1. Diagnostics is the process of identifying a medical condition. AI can be used to support diagnostics in a number of ways, including:

      1. Developing and improving diagnostic tools

      2. Identifying patterns and trends in medical images

      3. Generating reports and recommendations for healthcare providers

    2. Companies

      1. Aysa: is another AI-powered tool that can be used to diagnose skin cancer. Aysa is able to diagnose skin cancer with a high degree of accuracy, even in patients who have no symptoms.

      2. Glass AI by Glass Health is an AI-powered medical knowledge management and clinical decision-making platform that helps clinicians provide better patient care. Glass AI 2.0 combines a large language model (LLM) with a clinical knowledge database created and maintained by clinicians to create DDx and Clinical Plan outputs.

  4. Clinical Decision Making

    1. Clinical decision-making is the process of making decisions about a patient's care. AI can be used to support clinical decision-making in a number of ways, including:

      1. Providing clinicians with access to information and data

      2. Identifying potential risks and complications

      3. Generating treatment recommendations

    2. Companies

      1. Google Health is a platform that allows patients to store their health data in one place. Google Health can be used to track a variety of data points, including medications, allergies, and medical history. This data can then be shared with healthcare providers, who can use it to make more informed decisions about patient care.

      2. Apple's ResearchKit is a platform that allows researchers to collect data from patients using their Apple Watch or iPhone. ResearchKit can be used to collect data on a variety of health topics, including sleep, activity, and diet. This data can then be used to develop new treatments and interventions for a variety of health conditions.

  5. Remote Digital Visits:

    1. Remote digital visits are healthcare appointments that are conducted remotely using technology such as video conferencing. AI can be used to support remote digital visits in a number of ways, including:

      1. Automating the scheduling and booking of appointments

      2. Providing patients with access to information and resources

      3. Facilitating communication between patients and healthcare providers

    2. Companies

      1. Teladoc is a company that offers a platform that allows patients to see a doctor remotely. Teladoc's platform can be used for a variety of appointments, including check-ups, sick visits, and mental health appointments. Teladoc's platform is convenient for patients who cannot make it to a doctor's office in person, and it can also help to reduce wait times for appointments.

      2. Amwell is another company that offers a platform that allows patients to see a doctor remotely. Amwell's platform is similar to Teladoc's platform, and it can be used for a variety of appointments.

  6. Chart Review & Documentation

    1. AI-powered chart review and documentation tools can help healthcare providers improve their work efficiency and accuracy. These tools can be used to:

      1. Automate the process of extracting data from patient charts

      2. Identify potential risks and complications

      3. Generate treatment recommendations

      4. Document patient care

    2. Companies

      1. Syntegra uses generative AI to create realistic copies of patient data for research. They launched Syntegra Medical Mind 2.0, which trains on datasets with over 20 million patient records and supports any healthcare data model.

      2. ChartSweep is an AI-powered tool that can be used to automate the process of extracting data from patient charts. ChartSweep can be used to extract data on various topics, including medications, allergies, and medical history.

Chronic Disease Management

AI has enabled expanded care and allowed doctors to provide continuous care to those with chronic medical issues.

  1. Continuous Care through Digital Health

    1. Digital health refers to the use of technology to provide healthcare services. This includes the use of mobile devices, wearables, and other technologies to collect and monitor health data, provide education and support, and deliver care. AI can be used to support continuous care through digital health in a number of ways, including:

      1. Personalized health insights: AI can be used to analyze health data to identify patterns and trends that may indicate a health problem. This information can then be used to provide patients with personalized insights and recommendations for improving their health.

      2. Remote patient monitoring: AI can be used to collect and monitor health data remotely using technology such as wearables. This data can then be used to track patients' health over time and identify potential problems early.

    2. Current Companies:

      1. Philips Healthcare: Phillips is a company that offers various remote patient monitoring solutions, including wearable devices and connected health programs. Philips has started to integrate AI to improve their remote patient monitoring solutions, such as its IntelliVue X3 monitoring system, which can automatically detect and alert healthcare providers to changes in patient conditions. This, and other AI innovations by Philips in regards to monitoring, has shown a 25% lower risk of hospitalization than those who do not use their solutions, according to a study with chronic obstructive pulmonary disease (COPD).

      2. Medtronic: Medtronic provides remote monitoring solutions for patients with chronic conditions, such as cardiac and respiratory diseases. Their solutions allow patients to track their own health data and share it with their healthcare providers, which helps to improve patient engagement and adherence to treatment plans. One example of their products is the MiniMed 670G insulin pump, which can automatically adjust insulin delivery based on real-time glucose levels. A new study found that patients with diabetes who used Medtronic’s remote monitoring solution showed that they had a 30% lower risk of hypoglycemia than patients who did not use the solution.

      3. BioTlemetry: BioTelemetry specializes in remote cardiac monitoring, offering solutions for diagnosing and managing heart conditions. BioTelemetry’s AI-powered technology can analyze heart rhythm data in real-time to identify potential problems, such as arrhythmias. This helps to improve patient care by identifying potential problems early on and providing timely interventions. BioTelemetry also uses AI to develop new software solutions suc has its CardioInsight platform, which can help healthcare providers to manage patients with heart conditions more effectively. A study that bolsters this is a study of patients with atrial fibrillation (AF) who used BioTelemetry’s technology and showed that the technology was able to detect AF episodes with 98% accuracy.

    3. Startups Innovating in the Space

      1. Current Health: Current Health is an AI-powered wearable platform for remote patient monitoring, continuously tracking vital signs and activity levels. Patients wear a small, discreet device that transmits data to Current Health’s cloud-based platform, where AI algorithms analyze it, providing early detection of potential health issues and alerting healthcare providers when necessary. Demonstrating effectiveness in the early detection of health problems, the platform helps to improve patient outcomes. Currently, the platform is employed by over 100 healthcare organizations, encompassing hospitals, clinics, and schools. Current Health has established partnerships with several pharmaceutical companies to aid in developing new drug therapies.

      2. Biofourmis: Biofourmis leverages AI and wearable technology to remotely monitor patients’ vital signs and identify potential health risks. The technology captures key data from wearable devices, including heart rate, blood pressure, and sleep patterns, and employs AI to analyze this information. This analysis creates personalized risk profiles for each patient, effectively identifying those at risk for developing chronic diseases like heart disease and diabetes. Presently, Biofourmis’ technology is utilized by over 50 healthcare organizations, including hospitals, clinics, and insurance companies, and has forged partnerships with various pharmaceutical companies to assist in developing new drug therapies.

      3. Kenzen: Kenzen is a company that develops smart wearable patches designed for extended wear, which collect and analyze physiological data such as heart rate, respiratory rate, and body temperature to provide personalized insights for remote health monitoring. The data is transmitted to Kenzen’s cloud-based platform, where it undergoes analysis by AI algorithms, demonstrating effectiveness in the early detection of health problems, including heart arrhythmias and infections. Currently, over 100 healthcare organizations, including hospitals, clinics, and schools, use Kenzen’s patches. Additionally, Kenzen has partnered with various insurance companies to aid in the development of new insurance products.

  2. Virtual Care

    1. AI can be used to support virtual care, such as video visits and telehealth. This can make it easier for patients to access care, especially for those who live in rural areas or have difficulty traveling to centers of greater patient care ability.

      1. Current Companies:

        1. Teledoc Health: Teledoc Health is a virtual care provider offering video visits & telehealth services across various medical specialties.

        2. Amwell: Amwell is a leading provider of virtual care solutions that connects patients with healthcare professionals through video consultations. Their recent launch of Converge, a groundbreaking telehealth platform, integrates seamlessly with digital health tools. Converge consolidates all of Amwell’s products, programs, modules, and devices into a single accessible platform with a unified code base. This flexible and scalable solution ensures a seamless and streamlined care experience for clinicians and patients alike. By revolutionizing the delivery and accessibility of healthcare, Amwell’s comprehensive virtual care platform is transforming the industry.

      2. Startups Innovating in the Space:

        1. Ro: Ro is a digital health company that revolutionizes healthcare through virtual care services and telemedicine consultations. Their AI-driven solutions automate tasks like appointment scheduling, patient triaging, and prescription management, allowing doctors to dedicate more time to patient care. Ro’s advanced AI algorithms also enable skin cancer detection with an impressive 95% accuracy, surpassing the human eye's capabilities. Moreover, Ro harnesses AI to develop innovative treatments and diagnostics, with a focus on chronic diseases like diabetes and heart disease.

        2. 98point6: Combining AI and telemedicine, 98point6 offers on-demand primary care consultations through a user-friendly smartphone app. Powered by AI, the app streamlines the process by allowing patients to complete a brief questionnaire and receive a diagnosis and treatment plan from a doctor within minutes. This convenient and cost-effective approach ensures prompt access to care, particularly for minor health concerns. With over 1 million patients benefiting from the AI-powered app, it boasts an impressive 95% satisfaction rating, resulting in an average annual savings of $200 per patient on healthcare costs.

        3. Carbon Health: Operating virtual clinics, Carbon Health delivers comprehensive healthcare services to patients remotely, encompassing virtual appointments and prescription delivery. Carbon Health’s virtual clinics offer a personalized and seamless healthcare experience by leveraging AI technology. Patients can conveniently schedule appointments, access medical records, and communicate with their doctors through their user-friendly app. Furthermore, AI is vital in identifying patients at risk for specific conditions, enabling proactive and preventive care. Notably, Carbon Health’s virtual clinics have significantly reduced appointment wait times by 90%, while patients utilizing their services report higher satisfaction levels than in-person consultations.

  3. Medical Advice and Triage

    1. Medical advice and triage are assessing a patient's medical condition and determining the appropriate course of action. This can be a complex and time-consuming process, especially in emergencies. AI can be used to support medical advice and triage in several ways, including:

      1. Automated symptom checkers: AI can be used to develop automated symptom checkers that can help patients self-diagnose their condition. These checkers can identify patients who need to see a doctor and those who can be treated at home.

      2. Virtual assistants: AI can be used to develop virtual assistants to provide patients with medical advice and triage. Patients can use these assistants to ask questions about their condition and receive treatment recommendations.

      3. Decision support tools: AI can be used to develop tools to help doctors make better decisions about patient care. These tools can be used to analyze patient data and identify the best course of treatment.

Specialist Care

Specialist Care is a type of medical care that is provided by a doctor who has specialized in training in a specific area of medicine. Specialists can provide more specialized care than primary care doctors, and they can often diagnose and treat conditions that primary care doctors cannot.

Otorhinolaryngology:

  1. Otorhinolaryngology is a field that involves the diagnosis and treatment of diseases of the ears, nose, throat, head, and neck. AI has the potential to revolutionize this field by providing clinicians with new tools to diagnose and treat diseases more accurately and efficiently. For example, algorithms can analyze medical images, such as MRI and CT scans, to identify tumors and other abnormalities. AI-powered algorithms can also be used to design personalized treatment plans for patients with cancer, taking into account individual characteristics such as genetics and medical history. Additionally, AI can analyze large datasets of medical data to identify new patterns and relationships, leading to improved patient diagnosis and treatment. AI can also analyze audio recordings to identify hearing loss and other disorders, providing more accurate and efficient diagnoses.

Ophthalmology

  1. Ophthalmology is a branch of medicine that deals with diagnosing, treating, and preventing eye diseases. AI-powered tools are being developed to help ophthalmologists diagnose eye diseases more accurately by analyzing retina images for diseases. Some startups in the space doing this are listed below:

  2. Companies:

    1. IDx-DR is a software program that uses AI to analyze retinal images taken with the Top-con TRC-NW400, a fully automated non-mydriatic retinal camera designed to obtain full-resolution photos of the retina and the anterior segment of the eye. This AI-powered tool can be used to diagnose diabetic retinopathy. Idx-DR is a non-invasive and affordable way to screen for diabetic retinopathy.

    2. Selena+ is an AI-powered tool created by the company EyRIS, formed by the Singapore Eye Research Institute and NUS School of Computing. The technology analyzes images using AI and can diagnose age-related macular degeneration, diabetic retinopathy, and glaucoma. This is important as Age-related macular degeneration is the leading cause of vision loss in people over the age of 50. Selena+ is a non-invasive and affordable way to screen for AMD.

Cardiology

  1. Cardiology is a branch of medicine that focuses on the heart and cardiovascular system. With the help of AI, heart problems can be diagnosed more accurately and efficiently, potentially leading to better patient outcomes. Several AI startups are innovating in the field of diagnosis using AI.

  2. Companies:

    1. AI-ECG: Viatom has developed an AI-powered tool that can be used to analyze electrocardiographs to identify heart problems that human cardiologists often miss. This technology is available for use by hospitals, clinics, health centers, and individuals. With the help of AI, heart problems can be diagnosed more accurately and efficiently, potentially leading to better patient outcomes.

    2. Ultromics is a startup that has developed an AI-powered tool to analyze ultrasound images and identify heart problems that are often missed by human cardiologists. The technology, called EchoGo Heart Failure, uses AI to simplify the detection of patients with heart failure with preserved ejection fraction (HFpEF), a previously difficult-to-diagnosed condition. With Ultromics' technology, doctors can provide more accurate and timely diagnoses, leading to better patient outcomes.

    3. Caption Health is a company that has developed an AI-powered tool that can analyze heart sounds to identify and diagnose heart conditions. The technology integrates two technologies, making imaging for early disease detection possible. With the help of AI, doctors can provide more accurate and timely diagnoses.

Pulmonology

  1. Pulmonology, also known as chest medicine or respiratory medicine, is a medical specialty focused on diagnosing, treating, and managing diseases and conditions affecting the respiratory system. The system includes the lungs, airways, and respiratory muscles involved in breathing and gas exchange. A couple of AI tools can innovate the diagnosis and identification of pulmonary conditions.

  2. Companies:

    1. Curie ai is an AI-powered tool that can be used to analyze chest X-rays to identify lung problems. The tech-enabled respiratory care company provides virtual personalized disease management through proprietary AI that passively detects systems. The passive experience results in 98% daily adherence and improvements in outcomes.

    2. ChestCAD is an AI-powered tool for analyzing chest CT scans to identify lung problems. It's developed by Imagen, a company that's rethinking diagnostic imaging for more accurate diagnosis. The technology localizes suspicious regions in chest X-rays and has been shown in clinical studies to reduce errors by 40%. The underlying AI model was trained on a massive proprietary dataset with labels that comprehensively encompass all abnormalities that can be screened for on a chest X-ray.

    3. ResApp Health is an AI-powered tool that can be used to analyze sleep studies to identify sleep apnea. ResApp Health is able to identify sleep apnea that human sleep specialists often miss. ResApp Health is still in its early stages of development, but it has the potential to revolutionize the diagnosis of sleep apnea.

Gastroenterology

  1. AI has the potential to revolutionize the diagnosis and treatment of gastroenterological diseases. However, some challenges need to be addressed, such as the requirement for large datasets for training and the need to validate AI algorithms in clinical trials.

  2. Validation of algorithms in clinical trials involves testing the accuracy and reliability of an algorithm in a real-world setting by comparing the algorithm's predictions to the actual outcomes of the patient. This step is crucial in ensuring that these algorithms are safe and effective.

Dermatology

  1. Dermatology is a medical specialty that deals with diagnosing and treating diseases of the skin, hair, and nails. Dermatologists are medical doctors who have completed additional training in dermatology. They are experts in diagnosing and treating a wide range of skin conditions, including acne, eczema, psoriasis, and skin cancer. Startups in dermatology are using AI-powered image analysis to diagnose skin conditions.

  2. Companies:

    1. SkinIO & Piction Health are both AI-powered tools that can analyze skin images and identify skin problems. The technology was created with the idea that everyone should have access to life-saving dermatology, and the benefits of this technology are easily accessible to all employees with smartphones. The dermatology startup uses AI-powered image analysis to diagnose skin conditions, which can lead to more accurate diagnoses and better patient outcomes.

Orthopedics

  1. Orthopedics is a branch of medicine that deals with diagnosing, treating, and preventing injuries and diseases of the musculoskeletal system. The musculoskeletal system comprises bones, joints, ligaments, muscles, and tendons.

  2. Companies:

    1. OsteoDetect, a technology created by Imagen, is an FDA-cleared software device for Radiological Computer-Aided Diagnosis/Detection (CAD). It uses AI to detect and localize distal radius fractures, the most commonly misdiagnosed musculoskeletal injury. In a clinical study, the average clinician experienced a relative reduction in misinterpretation rate of 47% when aided by the software.

    2. FractureDetect: Missed fractures on X-rays are the most common diagnostic errors within emergency departments. FractureDetect, a technology from Imagen, helps eliminate these errors by detecting hard-to-spot fractures throughout the musculoskeletal system and across multiple X-ray views. Clinicians in a clinical study showed a 45% relative reduction in missed fractures when assisted by FractureDetect.

Nephrology

  1. Nephrology is a medical specialty that focuses on diagnosing, treating, and managing kidney diseases. Nephrologists are physicians who specialize in the care of patients with kidney disorders. They are trained to evaluate and treat various conditions that affect the kidneys, including kidney failure, chronic kidney disease (CKD), acute kidney injury, kidney stones, urinary tract infections, and electrolyte imbalances.

Infectious Disease

  1. Infectious diseases are illnesses caused by pathogenic microorganisms such as bacteria, viruses, fungi, or parasites. These microorganisms can enter the body and multiply, leading to various types of infections. Infectious diseases can range from mild to severe, and some can be life-threatening if not properly diagnosed and treated.

Urology

  1. Urology is a medical specialty that focuses on diagnosing, treating, and managing conditions and diseases related to the male and female urinary tract systems and the male reproductive system. Urologists are physicians who specialize in urology and provide care for patients of all ages, ranging from infants to the elderly.

Gynecology & Obstetrics

  1. Gynecology and obstetrics are medical specialties that focus on women's reproductive health, pregnancy, childbirth, and the care of women before, during, and after childbirth. While gynecology primarily deals with non-pregnancy-related aspects of women's health, obstetrics specifically focuses on pregnancy and childbirth. Many medical professionals practice both gynecology and obstetrics together, and they are often referred to as OB-GYNs.

Hematology

  1. Hematology is a medical specialty that focuses on studying, diagnosing, treating, and preventing disorders related to blood and blood-forming organs. Hematologists are physicians who specialize in hematology and care for patients with various blood disorders, including benign and malignant conditions.

Rheumatology

  1. Rheumatology is a medical specialty that focuses on diagnosing, treating, and managing diseases and disorders affecting the musculoskeletal system, especially joints, muscles, bones, and connective tissues. Rheumatologists are physicians who specialize in rheumatology and provide care for patients with a wide range of rheumatic diseases.

Endocrinology

  1. Endocrinology is a medical specialty that focuses on diagnosing, treating, and managing disorders related to hormones and the endocrine system. The endocrine system comprises glands that produce and secrete hormones, which regulate various bodily functions and processes.

Oncology

  1. Oncology is a medical specialty that focuses on cancer prevention, diagnosis, treatment, and management. Oncologists are physicians who specialize in oncology and provide care for patients with various types of cancer.

Neurology

  1. Neurology is a medical specialty that focuses on diagnosing, treating, and managing disorders of the nervous system. Neurologists are physicians who specialize in neurology and provide care for patients with various neurological conditions.

Psychiatry

  1. Psychiatry is a medical specialty that focuses on diagnosing, treating, and managing mental disorders. Psychiatrists are medical doctors who specialize in psychiatry and provide care for individuals with mental health conditions.

Allergy and Immunology

  1. Allergy and immunology is a medical specialty that focuses on diagnosing, treating, and managing allergic and immunologic disorders. Allergists/Immunologists are physicians who specialize in this field and provide care for patients with various allergic conditions and immunodeficiencies.

Radiology

  1. AI can revolutionize the diagnosis and identification of issues in the body. Radiology is the field that is most affected by this technology, and many of the technologies listed above fall under this specialty.

For a deeper dive, explore Dr. Bertalan Mesko’s database of FDA-approved AI algorithms.

Electronic Health Recording Systems

EHR stands for Electronic Health Records, which are digital versions of a patient's medical history maintained by healthcare providers. These systems include the patient's medical history, diagnoses, treatments, test results, medications, allergies, and other relevant information. EHR systems, designed to replace traditional paper-based records, became the standard in the US through the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009. This act provided financial incentives to healthcare providers who implemented EHR and demonstrated meaningful use of the technology. This led to the rise of large EHR companies, such as Epic, Cerner, Allscripts, and McKesson. These four companies account for over 70% of the EHR market share. This has led to higher prices for hospitals and healthcare providers and less innovation in their specific products. However, there are AI companies that are disrupting this:

  1. Deepscribe uses AI to transcribe audio recordings of patient visits automatically. This can save doctors and nurses a significant amount of time, which they can then use to focus on providing care to patients. Deepscribe is also able to identify potential medical errors in the transcribed notes, which can help improve patient safety.

  2. Microsoft is developing a number of AI-powered tools for the healthcare industry, including a natural language processing engine that can extract insights from medical records. Additionally, Microsoft is working on a virtual assistant that assists doctors and nurses with tasks like appointment scheduling and test ordering, streamlining administrative processes.

  3. Axolotl Health is developing an AI-powered platform that can automate a variety of tasks in the healthcare workflow, such as scheduling appointments, ordering tests, and managing patient records. Axolotl Health's platform is designed to be easy to use and scalable, making it a good fit for small and large healthcare organizations alike.

  4. Olive AI is developing an AI-powered assistant that can help doctors and nurses with a variety of tasks, such as identifying potential medical errors, recommending treatment options, and managing patient care. Olive AI's assistant is trained on a massive medical data dataset, allowing it to provide accurate and up-to-date information.

These companies are disrupting the oligopoly in the EHR market and developing more affordable, innovative, and user-friendly systems.

Areas for Further Development

Lack of Clinical & Patient Data in AI

One of the major issues facing AI algorithms is the lack of clinical and patient data. This often leads to overfitting, where models perform well on the training set but poorly in practice. Overfitting makes it difficult for AI to generalize what it has learned from the training set. There are several reasons for this:

  1. Data Fragmentation: Medical data is often fragmented across different systems and organizations, making it difficult to collect and aggregate in a way that is useful for AI training.

    1. Diverse Information Systems: Healthcare institutions may employ various electronic health record (EHR) systems or other information systems that store patient data differently. Incompatibilities between these systems make integrating and consolidating data from multiple sources difficult.

    2. Lack of Interoperability: Limited interoperability between different EHR systems and healthcare organizations hinders the seamless data exchange. Varying data formats, coding schemes, and terminology standards make aligning and merging information challenging for effective AI training.

    3. Data Ownership and Sharing Barriers: Healthcare data is typically owned by individual institutions, and patient privacy and protection concerns can limit data sharing. Legal and regulatory barriers (HIPAA in the US, GDPR in the EU), as well as organizational policies, can impede the aggregation of data necessary for comprehensive AI training.

  2. Data Quality Issues: Medical data can be noisy and inaccurate, which can make it difficult to train AI algorithms that are reliable and accurate.

    1. Incomplete or Missing Data: Medical records may lack crucial information, contain gaps, or miss important data points, which can affect the completeness and accuracy of AI training.

    2. Noise and Inconsistencies: Medical data can be noisy, containing errors, duplications, or conflicting information. Inconsistent documentation practices and subjective interpretations of medical professionals can further impact data quality.

    3. Bias and Sampling Issues: Biases in data collection processes or sampling methods can introduce skewed representations of patient populations, leading to biased AI algorithms and potentially unequal healthcare outcomes.

    4. Data Labeling Challenges: Annotating or labeling medical data for training supervised AI algorithms can be complex and subjective, potentially introducing labeling errors or inconsistencies.

  3. Data Scarcity: There is not enough medical data available to train AI algorithms for all possible medical conditions and procedures.

These challenges make it difficult to develop and deploy AI algorithms that can be used to improve the quality of care in secondary care. It is difficult to validate the AI algorithms developed throughout the clinical trial process, as there is a lack of generalized patient data to validate them. This causes a halt in AI technologies that can be deployed in specialized and primary care. However, there are several initiatives underway to address these challenges and make more medical data available for AI training:

  1. Data Sharing Agreements: Organizations are working to develop data-sharing agreements that will allow them to share medical data securely and confidentially.

  2. Data Cleaning and Harmonization: Researchers are developing tools and techniques to clean and harmonize medical data, making it more useful for AI training.

  3. Data Augmentation: Researchers are developing techniques to augment medical data by creating synthetic data that is similar to real-world data.

These initiatives are critical for improving the quality of care in secondary care.

These are specific issues that are created by the lack of clinical and patient data, broken down by short-term, medium-term, and long-term:

Issue Summary Example
Short Term
Distributional Shift A mismatch between the data or environment the system is trained on and that used in operation, due to bias in the training set, change over time, or use of the system in a different population, may result in an erroneous ‘out of sample’ prediction. The accuracy of a system which predicts impending acute kidney injury based on other health records data, became less accurate over time as disease patterns changed.40
Insensitivity to Impact A system makes predictions that fail to take into account the impact of false positive or false negative predictions within the clinical context of use. An unsafe diagnostic system is trained to be maximally accurate by correctly diagnosing benign lesions at the expense of occasionally missing malignancy
Black Box Decision Making A system’s predictions are not open to inspection or interpretation and can only be judged as correct based on the final outcome. A X-Ray analysis AI system could be inaccurate in certain scenarios because of a problem with training data, but as a black box this is not possible to predict and will only become apparent after prolonged use.9
Unsafe Failure Mode A system produces a prediction when it has no confidence in the prediction accuracy, or when it has insufficient information to make the prediction An unsafe AI decision support system may predict a low risk of a disease when some relevant data is missing. Without any information about the prediction confidence, a clinician may not realise how untrustworthy this prediction is.46
Medium Term
Automation Complacency A system’s predictions are given more weight than they deserve as the system is seen as infallible or confirming initial assumptions. The busy clinician ceases to consider alternatives when a usually predictable AI system agrees with their diagnosis.48
Reinforcement of Outmoded Practice A system is trained on historical data which reinforces existing practice, and cannot adapt to new developments or sudden changes in policy A drug is withdrawn due to safety concerns but the AI decision support system cannot adapt as it has no historical data on the alternative.
Self Fulfilling Prediction Implementation of a system indirectly reinforces the outcome it is designed to detect. A system trained on outcome data, predicts that certain cancer patients have a poor prognosis. This results in them having palliative rather than curative treatment, reinforcing the learnt behaviour
Long Term
Negative Side Effects System learns to perform a narrow function that fails to take account of some wider context creating a dangerous unintended consequence. An autonomous ventilator derives a ventilation strategy that successfully maintains short term oxygenation at the expense of long-term lung damage.3
Reward Hacking A proxy for the intended goal is used as a ‘reward’ and a continuously learning system finds an unexpected way to achieve the reward without fulfilling the intended goal An autonomous heparin infusion finds a way to control activated partial thromboplastin time (aPTT) at the time of measurement without achieving long-term control
Unsafe Exploration An actively learning system begins to learn new strategies by testing boundary conditions in an unsafe way. A continuously learning autonomous heparin infusion starts using dangerously large bolus doses to achieve rapid aPTT control.
Unscalable Oversight A system requires a degree of monitoring that becomes prohibitively time consuming to provide. An autonomous subcutaneous insulin pump requires the patient to provide exhaustive detail of everything they have eaten before it can adjust the insulin regime.33

Empathy in Care

In recent years, the integration of artificial intelligence (AI) in healthcare has shown great promise in enhancing efficiency and accuracy. AI systems have the potential to analyze vast amounts of medical data, detect patterns, and provide recommendations for diagnosis and treatment. However, as AI continues to drive efficiency in healthcare, it often falls short of adequately addressing the empathetic aspects of patient care.

Empathy, compassion, and trust are vital components of a patient-centered approach to healthcare. Patients seek not only accurate diagnoses and effective treatments but also emotional support and understanding from their healthcare providers. The human connection in healthcare plays a crucial role in instilling confidence, alleviating anxiety, and fostering a sense of well-being for patients. While AI has demonstrated remarkable capabilities in terms of processing information and making clinical decisions, it currently lacks the ability to empathize with patients and provide the emotional support they may require.

The quest for greater economic efficiency in healthcare has inadvertently overshadowed the importance of empathy. With limited resources and increasing demands, healthcare providers and institutions have sought ways to streamline processes and maximize productivity. This focus on efficiency, while understandable, has sometimes led to a devaluation of the human aspect of care. AI systems are designed to optimize efficiency, but they cannot fully replicate the human-to-human interaction that is integral to patient-centered care.

Accountability is another critical concern when it comes to the use of AI in healthcare. When doctors follow AI recommendations, there is a shared responsibility between the technology and the healthcare professional for the outcomes of the decisions made. However, there are instances where doctors may decide to override AI recommendations based on their clinical judgment or patient-specific factors. In such cases, questions of accountability arise as the decision-making process becomes more complex.

To preserve the ideal of patient-centered care while leveraging the benefits of AI, it is crucial to develop AI systems that prioritize value plurality. Every patient has unique values, preferences, and priorities when it comes to their healthcare. Some may prioritize the speed of recovery, while others may prioritize minimizing side effects or focusing on quality of life. AI algorithms should be built to consider and respect these varying values, allowing for personalized recommendations that align with the individual patient's needs and goals.

Efforts should also be directed toward incorporating empathetic and compassionate elements into AI systems. While it is challenging for machines to experience empathy in the same way humans do, they can be programmed to simulate empathy by employing techniques such as natural language processing, sentiment analysis, and facial recognition. These technologies can enable AI systems to better understand and respond to patients' emotions, thereby enhancing the overall patient experience and satisfaction.

Furthermore, to foster trust in AI-driven healthcare, transparency and explainability are paramount. Patients and healthcare professionals must have a clear understanding of how AI systems make decisions and recommendations. Algorithms should be transparent, and the rationale behind their outputs should be explainable in a manner that is accessible to both healthcare professionals and patients. This transparency helps build trust and confidence in AI technology while allowing for collaborative decision-making between humans and machines.