ARTIFICIAL INTELLIGENCE IN HEALTH CARE

Artificial intelligence in healthcare is introduced with the aim to simplify the lives of stakeholders in the healthcare industry (patients, doctors, hospital administrators, pharmaceuticals, insurance companies, etc.). AI is reinventing modern healthcare by machines that can predict, comprehend, learn, and act. New companies are emerging in this sector every day.

Currently, AI is being used in healthcare in the following ways to tackle cancer, cardiovascular & neurological issues:

1. Diagnosis — make an accurate diagnosis, predict the possibility of a disease in a patient, make an early diagnosis in comparison to its human counterpart, listens to symptoms to predict the diagnosis.
2. Pharma industry — to find new applications for existing drugs, find new patients for drugs or clinical trials, mapping of diseases to improve breakthrough medicines, predict bioactivity.
3. Patient experience — to tackle real-time patient flow, generate insights on patient journeys, prioritize patients according to conditions, improve clinical workflows.
4. Finance in hospitals — predict the healthcare costs for patients during admission.
5. As a resource to doctors — helps in diagnosing by combing massive dataset for comparable symptoms, reduce workload, more efficient use of doctor’s time.
6. In AI robot-assisted surgeries.
7. To avoid deaths by misdiagnosis or errors.

AI uses machine learning to analyze structured health data such as imaging, genetic and electrodiagnostic data & clusters patients’ traits such as age, gender, disease history, etc., and infer the probability of disease outcomes. Deep learning then extracts higher-level insights that identify may be specific cancer cells. Natural language processing (NLP) methods extract information from unstructured data such as medical notes to enrich structured data. All this data is generated for electronic health reports of different patients which are then structured and fed into the system.

Data:

Health data is scattered across public and private hospitals. There is no central system that maintains all the data of patients at one point. The Indian government has proposed a national stack of health data in 2018. Again, centralized data would also not guarantee correct results in predictions. As the data is collected from patients across different demographics, a correct prediction for a patient in one place would not mean the same algorithm can be used to predict for a patient in another place. It is also important to have diversity in data to ensure bias and discrimination is avoided. There also arises the point of ethics in data. Where is this data coming from and who owns this data? Does patient decide if his/ her data can be used? Do hospitals own patient data because they treated them? Do patients or hospitals make money out of selling this data to AI tech companies?

AI healthcare — influence over stakeholders:

Patients are eligible for effective care from early diagnosis to predictive treatment to early discharge and high patient satisfaction which seems great on the front end. But depending on how it is deployed, AI could either improve inequalities and bias in the system or make them worse. According to WHO, half of the world’s population lacks access to essential healthcare and nearly 100 million people are pushed into extreme poverty by health care expenses. This raises the question that who would pay for the AI? If it is the patient, then does this service only cater to patients who can already afford healthcare. That would make early detection, the genetic analysis only accessible to rich.

Another perspective to look at this scenario would be through the pharma industry, with all the available data, better medicines could be manufactured. The section of society which cannot afford the early detections could contribute to the creation of new breakthrough medicine by contributing their data which might be using them in the future.

Insurance companies can make the most out of AI. With all data skimmed through for potential diseases in a patient, insurance companies can have tailored premiums for every individual.

The hospital industry could use AI to study patient flow and satisfaction to improve its efficiency. It could also prepare itself in planned investments based on the insights generated on available data. For example, the winter season might bring X flu among Y age groups in the Z area, which will bring an influx of M number of patients. This would imply N number of beds be prepared for admission and a P number of medicines should be procured for treatment. The AI algorithms predict the future, but it is important to know that the result varies in different places based on demographics.

Doctors, nurses, and other medical staff would be free of data clerk jobs if AI takes over that sector. While AI reviews histories, compares records, keeps notes — doctors can spend their time seeing more patients. Currently, deep learning in radiology has been trained to identify cancer cells so well that they might replace radiologists. But similar AI systems tested in other fields were able to make judgments for one demography but failed in other. It is hence better to create an environment where AI systems simply expand access to current medical standards rather than replace human experts.

Accenture (consulting firm) predicted that AI applications could save the US economy $150 billion per year by 2026. It is unclear if patients would benefit or if tech companies, healthcare providers & insurers end up making more money. Even if AI makes cost-saving recommendations, doctors are offered incentives based on the number of investigations prescribed — hence, they might hesitate to take AI advise if that would result in less profit. With the influence of AI in healthcare on doctors, hospitals, pharma and insurance companies, patients might gain better health care on the big picture but the healthcare industry might become a costly industry, because of more players in the picture.

Ethics & Policy:

When AI starts to make decisions instead of humans, issues of accountability, transparency, permission, and privacy rise. It would be difficult to explain why an AI system suggested a diagnosis as a deep learning outcome is impossible to interpret. Mistakes will happen in AI systems and it is a question of who should be held accountable patients might receive a diagnosis from an AI system in place of an empathetic doctor.

There is also a question of ethics with regard to patients — whether the patient needs to know that AI is guiding his/her care. AI offers a brilliant future in healthcare — it would develop precision medicine and promote preventive health care. It is also important for the rest of the world to catch up with technology in terms of policy and regulations. It is important to calculate what negative effects it would generate on patients even though it promises great healthcare.

Further Future:

AI with its ability to make sense of data, would help other streams of technology in medicine to achieve great results. Its predictions can be used to build precision nanostructures, regenerative medicine, and customized organs through 3D printing.

The advent of AI also creates new business models that would shift healthcare from caregiving when sick to well-being. Data would be pushed to become radically interoperable and healthcare products would move from wearable devices to biosensing objects around humans. This would decentralize healthcare and hospitals might become hubs for specialized care. While patients would be able to solve simple health issues at home.

This would change the dynamics of the healthcare industry — bringing major players in the healthcare industry to create partnerships with industry disruptors like tech companies. New business models would be created to deliver these services — hubs, health products, virtual health space, consumer-centric health systems.

Conclusion:

AI promises a great future, but it is also constructed by humans. The system might adopt the same bias and framework of knowledge that humans possess. With no backtracking of a result produced by a deep learning algorithm, how do we ensure the quality of the outcome in terms of equality? There are a lot of unanswered questions with regard to data ownership, costs, accountability, and policy. Until we develop a system that could answer these questions and develop policies that keep the system in check, the promised value-based scenario (where patients gain most out of the system) might not be delivered.

There would be unchecked power dynamics established an average human might still be burdened by costs and unfair practices.