Many healthcare professionals view the advent of AI with considerable concern and a fair measure of trepidation. For virtually everyone, their worries are misplaced because AI is a tool that will make their jobs easier and more personally and professionally rewarding. Here are some prime examples.
Artificial intelligence has been incorporating itself into healthcare for years. Equipment gets smarter and more capable. Systems become easier to use and more intuitive. Routine tasks that used to be done by people are now routinely done by EMR and other integrated systems. Only recently has the spectra of AI become worrisome as its role has accelerated enough to become noticeable. Nonetheless, AI is still a tool that will make our jobs easier and more personally and professionally rewarding. While time will tell where things go, at present, AI is here to augment, not replace, humans and these are exciting times.
While the threat of our computer overlords remains stuff of science fiction, here is a fact based look at how AI is impacting the healthcare workforce in some critical ways:
The largest immediate impact by far is at the primary care level where most chronic and rare disease is identified and treated. Early diagnosis and focused treatment is the key to improving health status and mitigating long term impact of chronic conditions. There is a long list of routine administrative tasks that are being assumed by AI at the point of care allowing primary care providers to focus their attention on their patients instead of electronic medical record systems. Chronic conditions represent over three quarters of the cost of care under the age of 65 and over 95 percent of the cost over the age of 65. Rare diseases, of which there are about 7,000, are often less rare than rarely diagnosed before it is too late to treat them effectively.
The first operating entity in this area is Navina, which uses a novel AI to sift through EMR data including handwritten notes and disparate data such as specialist, laboratory and imaging reports to create an organized, one-page clinical summary for clinicians prior to each visit. This summary saves up to 25 percent of the time spent on EMR focus for each visit. Further, the company is exploring a partnership with a non-profit rare disease support group to identify potential patients using its platform that can save patients a lifetime of misery.
Once identified, the system can direct physicians on proper diagnosis, treatment, and support of this and prospectively many other rare diseases.
AI is already on its way to revolutionizing primary care by relieving it of the following administrative burdens, further freeing clinicians to focus on patient care and to practice at the top of their licenses:
1. Data Entry and Documentation: Natural Language Processing (NLP) systems can transcribe spoken words into text or extract information from medical records.
2. Appointment Scheduling: Chatbots and other automated scheduling systems can handle appointment bookings and reminders. Patients can schedule appointments online or through automated phone systems, reducing the need for staff to manage appointment logistics.
3. Medical Coding: Automatic assignment and recommendation of diagnostic and procedural codes allows medical coders to focus on high level issues instead of low-level clerical duties while reducing human error.
4. Billing and Claims Processing: Automating the billing process by generating invoices, verifying insurance information, and processing claims ensures timely and accurate billing, reducing administrative overhead, human error and enhancing cash flow.
5. Patient Communication: Communication tools can send automated follow-up messages, appointment reminders, and educational materials to patients, improving patient engagement and reducing workload.
6. Transcription Services: Voice to text transcription allows medical records to be updated and shared with the patient at the point of care.
Patient privacy and security regulations must be updated to ensure these enhanced capabilities protect patient rights to confidentiality.
AI is on the verge of revolutionizing clinical decision support by providing healthcare professionals with data-driven recommendations and best practices insights to augment clinical decision-making in ten critical areas. Overall, these advances will allow physicians and other clinicians to focus their practices at the top of their licenses and to do more with less support staff. The net effect will be a more streamlined, efficient system of higher qualified and trained people:
1. Diagnosis and Risk Assessment: New AI systems integrated into EMR platforms promise to analyze patient data that includes medical history, symptoms, and test results, to assist healthcare providers in diagnosing diseases and relative risk management using population health database driven comparisons. These comparisons can identify patterns and correlations indicating early onset of disease including rare diseases. Early diagnosis is key to the most effective and successful treatment.
2. Medical Imaging Analysis: AI-powered image recognition and analysis can help radiologists and pathologists interpret medical images more accurately by detecting anomalies in X-rays, MRIs, CT scans, and pathology slides that are not visible to the human eye.
3. Pharmacogenomics, or Precision Medicine: Using AI to compare a patient’s genetic profile to identify genetic variations that influence a patient’s ability to metabolize, absorb, and respond to a medication to tailor drug therapy to their unique genetic profile optimizes treatment outcomes and to reduces the risk of side effects. Combined with the patient’s medical history and current health status, healthcare providers can make more informed decisions about the most appropriate treatments and therapies.
4. Clinical Pathways and Guidelines: Consistency in adhering to clinical guidelines and best practices is a challenge at the practice and healthcare system level. By providing real-time guidance to clinicians. AI can suggest appropriate tests and interventions based on a patient's condition consistent with the most current evidence-based medicine.
5. Patient Monitoring: Just because we can now collect volumes of clinical data does not mean that healthcare providers can consume it. AI-driven predictive analytics can continuously monitor patients and identify subtle changes in vital signs or health parameters eliminating the need to absorb oftentimes overwhelming amounts of clinical data. This enables early intervention when deterioration is detected, improving patient outcomes.
6. Genomic Medicine: When genomic data is available AI can compare them with a large database to identify genetic markers associated with diseases to, combined with health and diagnostic status, guide personalized treatment decisions.
7. Population Health Management: Population health analytics without AI are more art than science. AI can analyze population health data empirically, comparing it to larger datasets to identify trends and risk factors, identifying health risks and trends, allocating resources efficiently and effectively. This will create new job opportunities in a wide range of new healthcare planning and management areas.
8. New Drug Discovery: By analyzing vast datasets to identify potential new drug candidates and predict their prospective effectiveness. This can lead to the development of new treatments and therapies creating a boon in research and development jobs.
9. Resource Allocation: Predictive models can forecast inventory needs, automatically reorder taking advantage of preferred pricing, predict patient admissions based on historical data and population health status and estimate resource requirements from staff to disease specific equipment and supplies, helping hospitals to effectively meet demand.
10. Empirical End of Life and Organ Allocation Support: Complex ethical dilemmas in healthcare such as end-of-life care or organ allocation are often subjective leading to high-stress situations and emotional distress. Fact-based, empirical direction can often ease these tensions by providing an impartial third-party empirical rationale that all parties can rely on.
An article in the JAMA Open Network entitled “Use of GPT-4 to Analyze Medical Records of Patients with Extensive Investigations and Delayed Diagnosis” explains in depth how GPT-4 can “improve the diagnostic accuracy of clinicians by supplying the most probable diagnosis or suggesting differential diagnoses in complex cases.”
The thrust of the article is how Open AI’s Chat GPT-4 can do more than augment the effectiveness of healthcare workers in parts of the world where more must be done with much less. Now, with a laptop and Internet connection, the quality of medical care can be elevated to be on par with the best care practices and science.
1. Reviewing Medical Literature: GPT-4 can quickly go through vast volumes of medical literature, research articles, and clinical guidelines to provide clinicians with up-to-date information relevant to a patient's condition anywhere an Internet connection is available. Providers always have the latest best practices and treatment options at their fingertips and added services such as medication and therapeutic equipment acquisition can be added.
2. Case-Based Knowledge: By evaluating complex patient cases using available medical data and comparing it to a database of similar cases, GPT-4 can identify potential diagnoses and suggest inferred or differential diagnoses and treatment protocols and the reasoning behind the recommendations.
3. Risk Assessment: Using available medical data, GPT-4 can compare it to population health risk assessed databases to indicate a range of likely diagnoses helping clinicians narrow down their risk assessment and diagnosis. This risk assessment can guide further testing and treatment decisions.
4. Red Flag Alerts: GPT-4 can monitor medical records and identify potentially severe or urgent conditions requiring immediate attention and recommend diagnostic and treatment protocols.
5. Automatic Monitoring: Staffing in many areas is spread thin and generally multitasked. AI can fill a critical gap by continuously monitoring patient data and updating its diagnostic suggestions as new information becomes available. It can detect subtle changes in a patient's condition that might indicate an increasing or decreasing risk factor or even different diagnosis.
6. Patient Education: Patient-friendly diagnostic and treatment plan explanations and summaries can help patients and their families understand their medical condition and treatment options in their native language. Where literacy is an issue, text to voice options are available.
7. Multi-Location Collaboration: GPT-4 provides a common platform for sharing and discussing complex cases and potential diagnoses.
AI is already having a significant impact on the non-physician healthcare workforce by assuming more and more routine tasks, which are transforming roles, streamlining processes, and enhancing efficiency in all healthcare settings. The net effect is that almost every side of healthcare can do more with less people. The upside is that those that remain are better trained and better paid. This trend will continue except in some areas. Vitals will still need to be taken and bedpans changed by people. For now.
Here's how AI is affecting different non-physician healthcare roles:
Clinical Decision Support: Advanced Practice Registered Nurses and Physician Assistants practice using algorithms while physicians use training and experience-based medicine. AI-powered clinical decision support systems recommend diagnostic protocols and provide real-time evidence based, best practice diagnosis and treatment presently at a physician level.
Patient Monitoring and Communications: AI-enabled wearable devices and remote monitoring solutions provide up to data patient data on demand via smartphone or tablet and instantly when a patient calls for assistance. It also provides alarms and updates to allow nursing staff to be proactive instead of reactive to patients’ needs.
Over time, the role of licensed practical nursing staff will be reduced or replaced with less skilled medical assistants.
Medication Administration: Medication error remains a problem in hospitals and is a vexing problem in long term nursing facilities. AI systems help ensure accurate medication administration by verifying dosages, checking for drug interactions, and providing medication administration records including interfacing with patient data to suggest medication changes to prescribing physicians to allow them to be more proactive with inpatients.
Automated Dispensing Systems: Hospital and retain pharmacies may be staffed by a single pharmacist overseeing an automated system.
Automated Testing: Like pharmacies, laboratory equipment is becoming more and more automated and autonomous. While the role of phlebotomists is likely to be merged with laboratory technicians or eliminated outright, someone will still have to collect the samples and load the machines. Similarly, higher skilled technologists will still have to make more complex decisions and run maintenance and quality control programs on the equipment.
Automated Coding: AI is increasingly assuming the role of assigning diagnostic and procedural codes to patient records, reducing, or eliminating the clerical workload for coders and improving coding and billing accuracy.
Data Abstraction: AI can extract relevant information from unstructured clinical notes and documents including handwritten documents and data from disparate sources, aiding in accurate coding and data abstraction for research and billing purposes.
As AI systems come online and improve through machine learning, lower skilled clerical positions will be reduced or eliminated.
These positions will be largely unaffected because obtaining images for X-Ray through PET Scans requires trained human intervention.
From front desk staff in physician’s offices to back-office staff in health systems, overall staffing will be cut by as much as three quarters over the next ten years while other, more skilled positions will open. Training for existing employees for more skilled positions is an obvious win-win for both parties and a very likely outcome.
Roles in these areas are unlikely to change except for added training in the installation and maintenance of remote monitoring equipment.
Overall, virtually every position, professional and support, in healthcare will change as AI is implemented and incorporated into the system. How those changes manifest will be a matter of the type of AI used, the type of healthcare system and the payment, political and institutional focus. Resistance, as the famous science fiction saying goes, is futile.