Healthcare and its integration with AI

Understaffed and Exhausted Healthcare workers

Amidst the challenges in global healthcare, AI solutions can be utilized to transform and enhance the efficiency of healthcare systems.

“Artificial intelligence is as revolutionary as mobile phones and the Internet”.

These words by Bill Gates aptly underline the prominence that AI is going to have in our lives in the coming years if not months. The healthcare industry isn’t immune to the reform that is being brought about by the rise of AI.

A survey of 500 medical professionals by Tebra (digital healthcare platform) in the US revealed the understated figures:

• California, Texas, and Georgia reported the most critical staffing shortages in the past year.
• 35% of health care workers have seen a co-worker fall asleep during a shift
• 73% of health care workers feel underpaid, and 59% feel unappreciated at work
• 77% of health care workers believe a health care crisis will occur within the next year due to understaffing and employee burnout.
• One in three health care workers plan to leave their job within the next year, and 14% plan to leave the industry entirely.

The survey presents a frightening picture for the healthcare system of a country where the median annual wage for healthcare practitioners and technical occupations (such as dental hygienists, physicians and surgeons, and registered nurses) was $77,760 in May 2022 as per the U.S. Bureau of Labor Statistics. The monetary compensation in the US is significantly higher than many of the developing and under-developed countries which along with lower monetary compensation, suffer from further shortage of staff and lack of proper infrastructure to cater to patients. This state of understaffed hospitals leads to exhaustion among the personnels which affects the quality of treatment that a patient receives. As per an article published in National Institute of health (U.S.) “Approximately, one in three physicians is experiencing burnout at any given time. This may not only interfere with own wellbeing but also with the quality of delivered care.”

The 2020 Medscape National Physician Burnout and Suicide Report reported a burnout rate of about 43%, which remains quite similar to the 46% reported in 2015 and 39.8% in 2013. The situation of healthcare workers in the U.S. can be extrapolated to assume a worse situation in countries with lower spending on healthcare and a lack of access to medical infrastructure.

Application of Gen-AI in healthcare

It is quite evident that the current detrimental state of healthcare around the globe has less to do with innovation in treatment methods but more of a lack of personnel in the field. As the population increases across regions – which affects and impacts the public health to deteriorate due to multiple factors – calls for a robust healthcare infrastructure now more than ever. There cannot be a better demonstration of the lack of capacity around the world than the covid-19 pandemic and how the world witnessed the crippling of healthcare systems of European countries which are supposed to be the best there is.

The rise in capability that Generative Artificial intelligence possesses now is the ‘messiah’ that the health care system needs.

The areas of application of AI are widespread from apps that allow patients to better manage their care; online symptom checkers and AI-powered diagnosis tools provide e-triage support. Virtual agents carry out tasks in hospitals. A bionic pancreas helps those with diabetes. Other technologies optimize scheduling and bed management in healthcare operations. Certain solutions predict hospital admission risks or help detect cancers early to improve population health outcomes. Some applications aid healthcare research and development – as well as – pharmacovigilance efforts. Together–  these advances have the potential to lead to better survival rates through earlier intervention and more efficient operations.

All of the above-mentioned solutions will help improve the efficiency of treatment and provide better care for patients. But the primary concern for now is to address the workforce crisis that the world currently faces.

AI-enabled solution for Healthcare Workforce

The low-hanging fruit in the healthcare domain is the administrative work which takes up a huge chunk of doctors’ and nurses’ time. Optimizing the routine and daily operations that take place in a hospital. Tasks like Patient scheduling, Bed management, medical record management, Supply chain management, Billing and insurance verification, Prior authorization and Discharge planning. AI/machine learning algorithms can be developed and utilized to optimize the above tasks to minimize the human inputs required and in some cases AI assistants can be developed to eliminate human input. These tasks take up a major chunk of time for hospital workers and eventually, these tasks would be automated.

However, it is to be noted that since many of these processes involve human interactions and are often sensitive in nature, substituting health personnel is not feasible. The goal of these AI-based applications is solely to lessen the burden on the personnel and allow them to interact with patients in a more meaningful and effective manner. A study conducted by McKinsey Global institute (MGI) showcases how automation and AI are likely to affect work hours in various sectors. In their analysis they demonstrate the share of hours currently worked that could be freed up by automation by 2030 for a wide range of healthcare occupations in selected European countries.

The demand for healthcare workforce is bound to increase in the future but with the intense nature of the work involved, the number of people getting into the field might not increase proportionally. Thus it becomes imperative to make use of AI solutions to combat the inevitable shortage of healthcare workers so that we can cater to patients in an equitable manner.

Challenges in implementing the newfound capabilities

It is certain now that AI has the potential to revolutionize clinical practice, but a major hurdle that we first need to address is the availability of quality medical data. First, it is important to understand that the AI/ML models are trained using structured data. Structured data is organized and formatted in a predefined manner that enables computer systems to easily interpret and process it. Examples include relational databases and spreadsheets containing fields for names, dates, currencies, text (with defined character limits), and numeric/alphanumeric codes. People inputting structured data are given clear guidelines on what information to provide and the format it should follow. For instance, insurance companies mandate healthcare providers strictly adhere to rules for submitting reimbursement claims. This consistency in structure and formatting allows the data to be readily searched, sorted, and analysed by machines. In contrast to unstructured data like images, audio, or free-form text, structured information has a uniform, well-defined schema that facilitates automated processing and extraction of insights.

But the availability of such structured data is rare and exists mostly in developed nations where the health infrastructure has already been digitalized. In the majority of the world today the healthcare system is fragmented, and the input of structured data is limited to advanced hospitals in the region which are not accessible to the majority of the population. But this challenge is a gold mine waiting to be explored. The reason is that the sheer enormity of unstructured data will give rise to leapfrog advancement in the sector.

The kind of data that can be used as an input for Electronic Health records (EHR) – which is dependent on structured data input – are as follows:

• Handwritten Notes

Natural language processing can be used to interpret handwritten notes recorded by physicians, nurses and other care providers. It can recognize characters as well as understand abbreviations, misspellings and intent by drawing conclusions from the text.

• Radiological Images

Medical images from X-rays, CT scans, MRIs etc can be analyzed using machine learning algorithms trained on archived images. They can recognize patterns to correctly interpret scans and distinguish healthy tissue from abnormalities.

• Image Metadata

Technicians and physicians can describe images using structured codes and keywords which can then be entered into the EHR system. This adds searchable information and context to images without requiring the images themselves to be structured data.

• Searches

Metadata and machine learning can enable searching within specific types of unstructured content like images and audio, allowing retrieval of relevant data.

• Streaming Data

Real-time data from IoT devices and sensors can be structured at the edge using AI/ML to enable alerts and actions based on analyzed trends in unstructured streaming data.

• Existing Records

Past unstructured healthcare content like old medical records still hold valuable insights and need to be accessible through appropriate structuring and integration with current EHRs.

Pre-Requisites for AI-enabled capacity building

The first step in the formulation of AI-enabled healthcare system in any country is collecting wide array of quality data which can be used as an input to train AI models. For this there needs to be an EMR(Electronic Medical Records)/EHR(Electronic Health Records) in place which will act as the primary source of data. Unfortunately, as discussed earlier, many of the developing nations lack an effective EMR.

EMR systems require investments in hardware, software, retraining, and IT support. Explaining the advantages of the EMR system to providers who have used paper charts for decades may be the greatest obstacle to implementation. Why is this system needed? Whom will it benefit? How does this system ensure safety? Continuous technology support must follow, addressing procedural questions about computer operation, information entry and access, corrections, and applications. EMR systems offer automated alerts, remote access, and data analysis.

EMR systems must balance patient and provider needs within the limits of business and political imperatives. While the primary goal of EMR systems is to improve quality of care, the incentives for building such a system include reducing medical errors, managing revenue streams, providing an effective means of communication, sharing information between healthcare providers, and collecting health information for institutional, educational, and research purposes ( Ayatollahi, 2014). The World Health Organization encourages buy-in from healthcare practitioners due to EMRs’ ability to:

1. Improve the accuracy and quality of data recorded in a health record
2. Enhance healthcare practitioners’ access to a patient’s healthcare information, enabling it to be shared between parties for ongoing care
3. Improve the quality of care as a result of having health information immediately available at all times.
4. Improve the efficiency of the health record service
5. Reduce healthcare costs (WHO, 2006)

NLP algorithms can analyze unstructured clinical text within electronic health records to extract meaningful information. By identifying keywords and phrases, NLP enables machines to determine the context and understand the documentation. This allows NLP to simplify medical documentation and perform automated tasks like voice transcription voice transcription.

Healthcare providers are increasingly adopting NLP due to the growing volumes of data in EHR systems. NLP assists providers by allowing them to focus on patient interactions while automatically generating accurate notes. It also facilitates record retrieval and populating forms, reducing administrative burdens.

As NLP and other clinical AI tools advance, their potential to enhance the care continuum will increase. When applied to EHR data, NLP can help overcome resource constraints by automating routine operations. This frees up personnel for direct patient care. NLP can also integrate data from diverse sources like mobile apps and devices. By gleaning insights from aggregated EHR data, NLP stands to improve outcomes at both individual and population levels.

For the nations who have not yet digitised their healthcare systems and created national portals for them, a fantastic opportunity has presented itself. The system can now be developed in the framework of artificial intelligence – which will greatly assist healthcare professionals and lessen their workload – if only somewhat. Thus – alongside developing and implementing AI-enabled solutions – training healthcare personnel to make optimal use of these solution should not be neglected. The development needs to take place at multiple fronts to ensure the most optimal path to reforming the public health care system so that we can reap the benefits of AI in an equitable manner.