What AI Can Do In Healthcare In The Coming Years: 8 Examples

We previously looked at what artificial intelligence (AI) can do already in healthcare and we continue this series on the technology’s potential in this article. This time, our focus is on what we can expect the technology to do in the near future. We share 8 examples of what AI is likely to do in the healthcare field in order to better anticipate the future.

1. Predicting disease progression

Prior to the worsening of a condition, there are often telltale signals from various health metrics that indicate a downward trend. However, such signals are not easily picked up, which leads to millions of deaths that could be prevented with earlier detection.

AI models have been developed to analyse electronic health records (EHRs) to accurately predict long-term outcomes for various conditions. Google DeepMind has developed an algorithm that can accurately predict acute kidney injury in patients up to 48 hours earlier than it is currently diagnosed. Researchers in Belgium have trained an AI model to reliably predict the probability of disability progression in the next two years among people with multiple sclerosis.

While such models are in the research and testing stages, we can expect them to be implemented in practice in the coming years as they scale up. However, this is subject to clinicians having the adequate tools to be alerted of disease progression and acting at the right time.

2. Real-time, real surgical assistance

Unequal distribution of specialist surgical workforce disproportionately affects rural and resource-poor areas. Some models even estimate that such regions experience a shortage of one million specialist surgical, anesthetic, and obstetric providers. AI could bridge that gap by enabling remote collaboration in surgical rooms.

AI-powered systems like Proximie connect surgeons virtually to any operating room in real-time. This enables collaboration and broadens surgical expertise access while the AI provides performance metrics and improves workflow. Similar solutions are also on the way. For example, in early 2024 Johnson & Johnson MedTech partnered with Nvidia to develop AI tools aimed at delivering real-time analyses of surgical data.

3. Coping with alarm fatigue

In healthcare, alarm fatigue refers to a phenomenon where caregivers become desensitized to alarm signs from numerous beeping devices. In fact, between 72% and 99% of all alarms are false which adds to alarm fatigue. Unfortunately, this means that some alarms that actually necessitate clinical attention are overlooked, leading to medical mistakes. 

AI could tune out the false alarms and make clinical staff aware of the alarms that require their attention. Researchers have developed an algorithm to reduce notifications received by the caregivers by up to 99.3%. This automated AI reasoning mechanism analyses patient monitoring data and vital signs to decide whether to group notifications rather than send individual ones so as to prevent alarm fatigue. 

Medtronic has partnered with other institutions to develop a similar tool. Their ‘Beyond the Noise’ project aims to develop an AI filtration tool to ensure only critical alerts reach medical staff. 

4. Remote patient monitoring

Remote care has been on the rise in recent years and such modalities can be enhanced with wearables for at-home monitoring. We have become accustomed with the likes of smartwatches and blood pressure monitors; but the next wave of wearables will increasingly be combined with AI.

One example is that of Biofourmis which provides care-at-home solutions for the continuous monitoring of both acute and chronic patients. Their platform combines clinical-grade wearable devices and AI algorithms for remote monitoring and detection of deterioration. The company has noted a 49% decrease in readmission rates for patients with congestive heart failure and the ability to detect deterioration 21 hours sooner.

The makers of the fitness tracker Whoop are also considering the integration of AI. Will Ahmed, the company’s founder, highlights AI’s unique ability to notice health trends or crunch biometric numbers for personalised health insights.

5. Genomics and precision medicine

One’s genetic makeup can be responsible for 30% of individual health outcomes. As such, genetic and genomic analyses can provide valuable insights into the risk of developing certain conditions and help in mitigating them. Such tests have become more affordable over the years but the ability to draw meaningful interpretations remains a barrier.

This hurdle can be addressed with AI tools like Google’s DeepVariant and Helix. These models can analyse genetic data and improve the identification of disease-causing variants. Subsequently, such findings can help in the development of more personalised treatments. DeepVariant can reduce the error rate of identifying variant locations by more than 50%. Helix’s AI can mine genomic data for more targeted drug discovery.

6. Automated insurance approvals and billing 

The manual handling of health insurance claims has been a reason for significant turndowns. In some cases, insurers have denied 49% of claims. AI systems can make this process more efficient. They can streamline prior authorizations and claim processes, reducing administrative delays. The patient experience is also made smoother as wait times are reduced.

Despite such promises, some companies are facing financial and ethical challenges. Healthcare AI startup Olive, which focused on revenue cycle automation tools, closed down due to strained resources after a period of fast growth. Other tools like NaviHealth appear to have targeted beneficiaries of Medicare Advantage to deny them care. This highlights the need for adequate moderation of such tools to ensure ethical and equitable automation.

7. Early detection of rare diseases

By virtue of their uncommon occurrence, rare diseases are challenging to identify and treat. They can sometimes present with specific physical features that can help in their identification. Clinicians can expect to receive the aid of AI to help them detect such conditions. 

Researchers in Germany developed such a tool, trained on multitude photographs, to improve the accuracy of detecting rare conditions such as mucopolysaccharidosis, Mabry syndrome and Kabuki syndrome, where those affected have characteristic facial features. 

Another similar tool comes from OM1. Its AI analyses patterns from patient data to form a “digital phenotype” that can be used to detect early warning signals of rare conditions such as generalized pustular psoriasis. 

Using such techniques could fast-track the identification and treatment of those affected from early on based on patient-reported symptoms and clinical data.

8. Patient-specific virtual health coaches

Health coaching can train people to adopt healthier behaviours and reduce the risk of preventable diseases. This approach has enticed companies to integrate AI for patient-specific virtual health coaching. 

One major player is OpenAI which has partnered with Thrive AI Health to develop an AI-powered health coach. Other apps like Healthify track diet intake and provide personalised health coaching with the assistance of AI. We can expect such types of virtual health coaches to become more commonplace in the near future.

This concludes our collection of what AI is likely to do in healthcare. We would encourage you to also take a look at our first entry in this series of articles to learn more about what AI can already do in this field. We will be back with a final entry focusing on what the technology might bring to healthcare.