AI and Technology in Healthcare

AI and technology have changed how medicine is practiced and documented around the world. Learn how technology upgrades have been historically implemented and now how are being quickly utilized in the healthcare landscape.

Clinical care has evolved considerably for patients including lab testing, medications, and surgical procedures.
 
Other technologies have replaced how medicine is practiced and documented. For example, the Electronic Health Record (EHR) has largely replaced paper charts. Healthcare technological upgrades are typically applied on a relatively incremental and fragmented scale. However, that changed in 2020 with the novel SARS-CoV-2 (COVID-19) pandemic. The role of technology in healthcare has recently shown to be critical for patient care.

Indeed, the COVID-19 pandemic caused worldwide disruption on a massive scale. Experts estimate the healthcare system made 10 years of technological upgrades in a few short months after the pandemic was declared by the World Health Organization (WHO).

This is largely due to the response to local and national shutdowns that were implemented to slow the spread of the virus. Many healthcare facilities switched to telemedicine (phone or video call consultations) for non-emergent care to comply with new social distancing mandates.

While telehealth appointments continue to be an integral part of these technological adaptations, there are many other technologies that have been implemented to improve patient care, increase access, and mitigate the spread of COVID-19 infections.

It’s estimated the healthcare system made 10 years of technological upgrades in a few short months.

• Artificial Intelligence (AI) is an umbrella term that describes a computer’s ability to copy human cognition, such as problem solving and learning. There are different types of AI processes.
• Machine Learning (ML) is a method of using data, computer algorithms, and mathematical modeling so a computer can learn without human input. It can improve independently, based on experience. Cardiac arrhythmia detection algorithms that are used in Automated Electronic Defibrillators (AEDs) is an example of ML technology.
• An Expert System (ES), which is different from ML, uses a rules-based system to replicate a human subject matter expert’s ability to interpret data and make decisions. It makes a recommendation based on a series of “if/then” statements. ES and ML systems are not necessarily mutually exclusive.
• Remote Patient Monitoring (RPM) utilizes a series of AI cascade decision trees which may incorporate both ML and ES technologies for optimal outcomes. RPM uses digital technologies to capture real-time patient health information in one location and transmit it to a healthcare provider or system in a different location for assessment and management. It can be used for acute or chronic health problems and may cut down on in-person medical visits. It is estimated over 39 million people in the United States used one form of RPM in 2021, and that number is expected to grow to over 70 million in 2025.
  
  • RPM devices are available for asthma, diabetes, high blood pressure, and small intestinal bacterial growth (SIBO), among other health conditions.
  • Benchmarks and thresholds can be preset in RPM systems to recommend an intervention based on static medical guidelines, though the physician still needs to use their clinical expertise for management.

Technology offers a variety of healthcare solutions as patient loads increase and the global pandemic continues. ES, ML, and RPM tools have the capacity to increase efficiencies and patient access if adopted on a large scale. The second article will take a deeper dive regarding specific applications where technology is already helping patients and providers.

Clinical utility of health IT software and hardware

For example, AI-based chatbots have the potential to screen patients for hereditary cancer risk using a rules-based ES. The National Comprehensive Cancer Network (NCCN) publishes guidelines for patients who may meet the criteria for genetic screening for hereditary cancer risk.

In this case, a chatbot interface (via mobile health app or Internet) can ask a patient personal and family history questions about cancer. The interface may include a human avatar and uses patient-friendly language, thus making the interaction seem more like a conversation with a person. Depending on the patient’s answers to these questions, the chatbot can determine who should be offered hereditary cancer genetic screening.

Chatbots may be updated as new guidelines are published to reflect the most current medical recommendations. Studies have shown chatbots engage patients, provide education, and facilitate a referral that may include the option of genetic testing.

The global pandemic allowed AI software and hardware to be more commonplace in healthcare. However, AI technologies in healthcare are not a simple “plug and play” situation. These technologies come with some ethical and logistical challenges:

• Privacy: How will the patient be protected from unwanted intrusions into their personal domain?
• Data Management: How will the data be encrypted and protected so only healthcare professionals who are part of the patient’s care will have access to Personal Health Information (PHI)? How will data management comply with the Health Insurance Portability and Accountability Act (HIPAA)?
• Bias: Will the dataset used to train the AI system(s) be inclusive of different patient populations to prevent bias?
• Consent: How will smart healthcare devices be limited to perform only the health measurements consented by the patient, and not produce unnecessary extraneous data?
• Practice Implications: Will the implementation of AI change the way the physician views their patient? Will using AI increase the liability of clinical staff members who make a decision based on their clinical judgment that conflicts with AI recommendations?

Transparency is absolutely critical

The solutions to these issues are multifaceted and will likely change over time. Transparency is absolutely critical among healthcare providers, companies developing these technologies, and the federal government.

HIPAA covers 18 identifiers that must remain private and are considered PHI. End to end encryption must be standardized to prevent data breaches, where hackers illegally obtain patients’ PHI and other data that may be used to steal someone’s identity. A consent process should be standard, so patient autonomy is respected.

While the aim is to have AI systems work independently, they will still need human oversight. Involving patients as soon as possible while these technologies are being developed is important. There is a higher success rate for AI processes if patients are consulted earlier in the development of a new technology, compared with technologies that do not consider the patient’s experience in the design phase.

The data used to train AI systems must be representative of a diverse patient population incorporating a variety of sexes, genders, genetic ancestries, socioeconomic backgrounds, and geographies.

Just as new laws in the United States have been passed to regulate healthcare, such as HIPAA (1996); GINA (2008); HITECH (2009); No Surprises Act (2022); new laws regulating AI should be implemented, preferably before they are released for clinical use.

Clinical healthcare providers must be trained in the social implications of incorporating AI technologies into routine patient care. AI has the potential to cause a massive paradigm shift in healthcare, and providers must be prepared for how this will change their clinical practice.

The role of technology in healthcare will continue to expand as patient loads increase.

AI and ML tools have a multitude of applications, and the literature demonstrates the benefits of different technologies. However, multiple stakeholders must work cooperatively with one another to overcome the ethical and legal challenges that come with the exciting promises of different technological advances.

What began with prior authorization processing has evolved into a personalized platform that includes automations for benefits eligibility and prior auth requirements in real time.

1. Lab receives test request form. Lab sends request to careviso.
2. seeQer runs 5 step process. Lab receives seeQer report.
   
   1. Identify correct insurance
   2. Run benefits and eligibility algorithm
   3. Confirm patient requirements and reflex to PA if needed
3. If PA is not required, Lab submits claim to billing. If PA is required, seeQer reflexes to PA workflow.

We use multiple eligibility vendors, build datasets of standard medical codes, leverage pricing, and configure prior authorization requirements all in a single platform: seeQer.

seeQer accurately identifies coverage and benefits, and standardizes the configuration of pricing, coverage, and prior authorization requirements. It also streamlines the prior auth process while maintaining or increasing current PA volume, and calculates patient cost in real-time.

By leveraging seeQer, clients can streamline their workflows, optimize productivity, and minimize delays to maximize profits.