An ever-growing amount of diagnostic data has led to information overload, often making accurate diagnoses more difficult, impacting patient health.

 By Erez Na’aman

Diagnostics have an intrinsic need for information about what is going on inside the body, yet for all but a fraction of human history, such information was neither plentiful nor reliable. This left even the best medical practitioners helpless in the face of most ailments. Moreover, treatments were not available for many medical conditions, prompting medical practitioners to focus on just alleviating symptoms. With limited ability to change patient outcomes, there was little value to knowing the underlying cause of a patient’s disease or condition.

In the last century, though, there has been an unprecedented leap in the ability to provide suitable medical treatment. Most conditions, if diagnosed accurately and early enough, are treatable or even curable. In recent years, technology has also lent itself to providing us with more data as well as the means to organize it. As a result of these two developments, diagnostics has earned its place in the spotlight.

Although information has become digitally accessible and searchable, this abundance has drawbacks. Information is often organized, but it is frequently spread out across numerous systems and not always arranged in a way that makes it useful. The situation is akin to having sorted an endless mess of papers into neat files, only to then find oneself overwhelmed by endless (and growing) cabinets full of files. A frequent result is that medical staff are inundated with lots of data that they are unable to put to use. This information overload is making timely and accurate diagnoses more elusive.

The Consequences of Information Overload

Being bombarded with information creates serious risks. It can lead to alert fatigue, where medical practitioners miss important signals buried under a mountain of noise or are simply too tired to notice. 

It can also lead to overdiagnosis. This is because when one has information overload, one is bound to find something out of the norm, at which point one may be legally obligated (and may also feel ethically inclined) to notify the patient. A common consequence of overdiagnosis is overtreatment, whereby patients go down costly, time-consuming pathways that do not make them healthier and may even increase their suffering.

AI Is Coming of Age

Artificial intelligence (AI) is often accused of exacerbating the risk of information overload described above. Such accusations are due to a limited number of suboptimal implementations and do not reflect an actual deficiency. Modern iterations of AI are actually capable of being powerful tools in mitigating the risk. 

AI’s greatest power is its aptitude for learning to find patterns. When trained properly and fed the right questions on sufficiently large data sets, it can filter for valuable information quickly and efficiently.

In the context of data, AI is a fantastic complementary asset to human creativity. Humans excel at learning and thinking outside the box, but our accessible memory is quite limited. This is why people sometimes suffer from burnout when they try to sift through loads of data (even just thinking about those endless file cabinets can be unsettling). In contrast, when relevant, actionable information is right in front of us, we thrive at connecting the dots. Where AI can extract relevant and actionable information, humans can pick up the baton and make optimal decisions.

This can mean taking a course of action in line with an algorithm’s recommendation, but crucially, it doesn’t have to. For instance, AI can detect a cancerous tumor, at which point the medical expert can take over. Even if an especially powerful algorithm further determines actionability and urgency, suggesting, for example, chemotherapy over surgery if the former is statistically more likely to succeed, the medical expert can account for this information but may decide on a different course of action based on variables to which the algorithm is not privy, such as a patient’s desires, fears, and state of mental well-being. When AI finds the relevant information, organizes that information, and makes it available, a qualified human professional can make the judgment call, interact with the patient, and employ creativity and empathy to reach an optimal management path.

Doubling Down on Diagnostics

Well-designed AI does not pose a threat of worsening the ongoing information overload in medicine. On the contrary, it presents an opportunity to usher in a new era in diagnostics. 

With the help of AI, medical professionals can gain actionable insights without sifting through huge data sets or trying to act on inapplicable portions of those data sets. Instead of struggling to find lots of information more quickly, they could find the right information in a timely manner. This would help reduce physical and emotional suffering for patients, lessen strain on caregivers, and lower costs for healthcare networks.

By better predicting underlying causes for symptoms and conditions, targeted diagnostics technology could find the ideal moment to provide timely, actionable alerts. This would allow experts to avoid lateness without resorting to being too early. The benefits would be providing relevant treatment on time and avoiding overtreatment, helping patients live longer, healthier lives.

Why Isn’t This Available Everywhere?

AI’s capacities for diagnostics are relatively new. Beyond the time required before their full effect becomes evident, a few things need to happen:

  1. Policy refinement: Regulatory policies need to be tailored to AI’s distinctive qualities—first and foremost, its dependence on data and short, fast learning cycles. With adapted policies in place, developers will be drawn to design products and introduce them to the market.
  2. Access to data: Currently, the data sets to which AI companies (diagnostics-focused and otherwise) have access are extremely siloed, hindering these companies’ capacities to optimize tools. Granting them wider access to information can help mitigate the information overload plaguing healthcare networks. 
  3. Adaptation and fine-tuning: Diagnostics workflows need to adjust to reflect new AI capabilities. Healthcare networks and practitioners will retain the ability to set diagnostic thresholds according to what is actionable and helpful. With AI at their disposal, these thresholds will become more precise, leading to increased trust and wider acceptance and use.

By overcoming the drawbacks of too much information, AI can take its proper place in diagnostics—helping clinical professionals achieve timely detection and providing actionable insights. The resulting reductions in suffering, stress, and costs may prove to be some of AI’s most significant contributions to health care.

Erez Na’aman, MSc, is a trained physicist and expert on digitization and AI in laboratory diagnostics. He is a co-founder of Scopio Labs, currently serving as the company’s CTO.