Inside Healthcare AI: A good-news-bad-news prognosis

tablet vertical primary healthcare ai insider pro
IDG / Getty Images

The number of hospitals and other types of healthcare organizations ramping up AI projects and measuring results has increased dramatically over the past few years. Robot-assisted surgery, virtual nursing assistants and administrative workflow assistance are projects attracting big investments but the spectrum of AI-related healthcare initiatives is wide. Telehealth, predictive logistics and AI-generated diagnoses, just to name a few, are also in the mix.

What’s more, with genome research increasingly uncovering disease biomarkers, humans will likely require AI help to take advantage of all the new, related data.

healhcare ai thumb IDG

Download our guide to healthcare and artificial intelligence. 

Lack of trust is an obstacle to healthcare AI

The biggest hurdle to AI implementations in healthcare is lack of trust, according to practitioners in the field. The essential problem is that AI systems can be “black boxes,” applications that supply answers and predictions with only a select few health practitioners or data scientists understanding how the results were achieved.

The black box issue is endemic to AI in general, but is especially worrisome in healthcare since lives may be at stake and errors can be fatal. One approach is to gather multidisciplinary teams that bring together representatives from technology and medical staffs with patient advocates to work out guidelines for data that can be applied to various types of algorithms.

Data integration is key

AI requires data, and the more data that can be fed into AI applications the more accurate the results will be. Healthcare information is notoriously untidy, however. Data may reside in multiple databases, in different structured and unstructured formats, and also include images and handwritten notes.

Given the ubiquity and lack of structure of health-related information in most organizations, the formation of coherent data management practices is key to the success of any healthcare AI strategy. Best practices for data cleansing and ETL (extract, transform, load) for data collection and integration need to be followed to help assure success.

Despite the difficulties, most healthcare organizations in the U.S. are planning or currently implementing AI-related applications, with average spending among the larger enterprises in the tens of millions of dollars, according to recent research. The prognosis for these initiatives is upbeat: depending on the type of project, ROI is generally predicted to occur within three or four years.

Insider Healthcare AI