This is an excerpt from a story delivered exclusively to Business Insider Intelligence Digital Health Pro subscribers. To receive the full story plus other insights each morning, click here.
The health unit of Google’s London-based AI powerhouse DeepMind published research that claims its machine learning (ML) AI can accurately predict 90% of acute kidney injuries (AKI) that would require dialysis and spot potentially fatal kidney injuries 48 hours before symptoms are recognized by doctors, per CNBC.
AKI refers to sudden impairment or failure of normal kidney functions, and it claims the lives of approximately 2 million people globally each year, according to research from the University of Pittsburgh School of Medicine.
Here’s what it means: DeepMind’s tech is a groundbreaking application of ML for clinical diagnostics, but false positives and a lack of diversity in the study’s participants means this is just the start of its efforts.
- This study is a breakthrough for the early detection of AKI — a notoriously difficult disease for physicians to identify. The disease is hard to detect given how little is known about its causes and the speed at which the kidney breakdown occurs. But by assessing insights gleaned from electronic health records (EHRs), DeepMind’s ML algorithm outperformed current clinical models used for AKI risk assessment by 20%, per STAT.
- Still, shortcomings in DeepMind’s study suggest its algorithm is just the first step — not a final solution — for the detection of AKI. Because DeepMind trained its AI detection tool on anonymized data from 700,000 patients in US Veteran’s Affairs (VA) hospitals across the country, the study suffered from a lack of patient diversity due to the demographics of the VA: Only about 6% of all patients used in the data set were female, for instance. When asked about how DeepMind Health addressed concerns around gender and racial diversity in its VA study a spokesperson for DeepMind Health informed Business Insider Intelligence that its model predicted 56% of AKI early in men, but only 44.8% overall for patients whose gender was known, indicating worse outcomes for female patients. Furthermore, they shared that the model actually performed 6% better among African American patients than all other ethnicities in aggregate, which can be critical given that recent studies have shown African Americans are 30–50% more likely to experience AKI.
The bigger picture: DeepMind’s VA study marks the first US partnership for the London-based Google subsidiary — and it’s likely the first of many as one of DeepMind Health’s most prominent teams is soon to be absorbed into Google Health.
DeepMind is coming stateside with long-term plans to bring AI-informed decision making to "nurses and doctors everywhere." New information about the performance of DeepMind’s clinical alert and patient management app, Streams, in UK hospitals was also announced alongside the VA study results.
A peer review revealed that Streams reduced cost of care by more than £2,000 ($2,400) per patient, according to Digital Health. Notably, Streams was designed in part to more efficiently notify caregivers of preventable, but difficult-to-detect conditions like sepsis and AKI via assessment tools already used by the UK’s National Health System. And the new algorithm trained on data from the VA study will eventually be incorporated into the app.
This all comes less than a year after Google Health announced it would absorb the DeepMind Health team behind Streams. I (Zach) think it’s highly likely we’ll see an updated version of Streams released in the US very soon — which would be a major US healthcare play for Google Health.
Interested in getting the full story? Here are three ways to get access:
- Sign up for Digital Health Pro, Business Insider Intelligence’s expert product suite keeping you up-to-date on the people, technologies, trends, and companies shaping the future of healthcare, delivered to your inbox 6x a week. >> Get Started
- Subscribe to a Premium pass to Business Insider Intelligence and gain immediate access to Digital Health Pro, plus more than 250 other expertly researched reports. As an added bonus, you’ll also gain access to all future reports and daily newsletters to ensure you stay ahead of the curve and benefit personally and professionally. >> Learn More Now
- Current subscribers can read the full briefing here.
- Google Nest is providing 100,000 Google Home Minis to help people living with paralysis retain their independence
- Livongo’s and Health Catalyst’s market debuts ended the three-year digital health IPO drought
- Genomic testing firm Helix is offering free genetic screenings in Florida