This is the 34th edition of Things & Thinks and covers quite a bit of ground on the research brief and digital health news updates. Of course, the ending is always with its own flavor, with Longread, Tweet and Chart o the Month. Happy to hear your thoughts and feedback…
Unexpected & Potentially harmful algorithmic behavior of AI systems
This study, published in The Lancet Digital Health, did not stop at validating their of a deep learning system for the detection of proximal femoral fractures but also performed an algorithmic audit. The authors noted that
- Although the model shows high performance, and does not appear to deviate from human performance in prespecified subgroups, it does still make the occasional inhuman error (eg, misdiagnosing a highly displaced fracture).
- Although the model reproduces some recognisable aspects of human practice (eg, it appears to pay attention to Shenton’s line), the visualisations nonetheless raise concerns about the regions that are not highlighted in the heatmaps.
- Saliency maps should be interpreted with caution due to known failings of these methods, but current findings together raise the concern that, despite the model performing extremely well at the task of proximal femoral fracture detection when assessed with summary statistics, the model appears to be prone to making unexpected mistakes and can behave unpredictably on cases that humans would consider simple to interpret.
Long way to interoperability
Researchers, using data from the 2019 US National Electronic Health Record Survey, have come up with interesting insights in this article, as summarized here.
- About 42% of physicians used an EHR from one of the five market-leading developers, (Epic, Cerner, athenahealth, eClincalWorks and Allscripts), while 21% of physicians used an EHR from the sixth to 10th market-leading developers, meaning NextGen, Practice Fusion, Greenway, GE and Modernizing Medicine, and 27% used a different vendor.
- The researchers found “substantial variation” in data exchange capabilities based on practice size. Nearly half, or 44%, of solo practitioners were able to send, receive or query for information, compared to 71% of physicians in practices of four to 10 physicians and 76% of physicians in practices of more than 50 physicians. Integrating data into EHR varied by size of practice. Some 17% of solo practitioners said they were able to do so, compared to 30% of physicians in practices of four to 10 physicians and 51% of physicians in practices of greater than 50 physicians.
- According to the authors, whereas patients treated by large, integrated practices may receive all care within the organization, patients of small practice physicians are likely to receive care from physicians or healthcare organizations outside of the small practice. Physicians in small practices may, therefore, need to rely on interoperable technology to a greater extent to receive important clinical information in a timely and useable manner.
- Despite this potential value, small practice and independent physicians have been observed to adopt EHRs and engage in interoperable exchange at lower rates than other physicians
Digital Healthcare News
Tech in Digital Health
Fitbit received clearance from the USFDA for a new PPG (photoplethysmography) algorithm to identify atrial fibrillation (AFib).
Global analytics provider SAS Health is teaming up with Microsoft to leverage Microsoft’s Azure Health Data Services platform to make SAS Health’s AI tools more efficient and secure to power data exchange and analytics, according to the companies.
The European Commission presented its ‘health data space’ plans, aiming to dramatically reshape access to medical data and its use in research and policy. It is poised to change how patients, doctors, researchers and policymakers access and use health data; it’s also hoped the plan will lead saving up to €11 billion over the next 10 years, with half coming from improved data exchanges in health care itself, and the other half from the use of health data in research and policy.
The USFDA granted PEAR Therapeutics its Safer Technologies Program (STeP) designation to a virtual reality-based treatment intended to treat adults with acute and chronic pain after surgery.
The European Commission’s medtech advisory group has added guidance documents on legacy devices and Eudamed to the In Vitro Diagnostic Medical Devices Regulation (IVDR) transition plan.
Israeli startup AIdoc received its ninth FDA approval, this time for an AI tool designed to automatically identify potential cases of brain aneurysm in head scans.
Alio receives FDA 510(k) for its remote patient monitoring system that collects data on skin temperature, auscultation, or internal body sounds, and heart rate.
Gleamer- a Paris-based company- received US FDA approval for its BoneView platform, which automatically highlights bone fractures and traumatic injuries and had previously received a CE mark in Europe.
The US FDA released a guidance to remind hospital professionals that artificial intelligence programs for detecting conditions such as strokes are currently not a replacement for attentive, hands-on care and that clinicians should keep themselves in the loop.
AI-enabled pathology company PathAI has entered into a multiyear drug discovery partnership with pharma giant GlaxoSmithKline focusing on advancing clinical trials and drug development in oncology and for nonalcoholic steatohepatitis (NASH).
Digital Medicine Society partnered with Anthem and major pharmaceutical companies including Biogen, Eli Lilly and Pfizer to develop a toolkit to support reimbursement for new drugs developed using digital endpoints.
Bayer is setting up 27 telemedicine centres in 12 districts across eight states in India over the next two years through Bayer Foundation India. Bayer partnered with multispecialty clinic RxDx Healthcare and Piramal Swasthya, the public health arm of the Piramal Foundation, which are both experienced in running telemedicine in primary health centres located in some targeted districts.
Funding, Deals, Mergers & acquisitions
Teladoc Health recorded a $6.6 billion impairment charge in the first quarter for its Livongo acquisition (purchased for $18.5 billion in 2020). Impairment occurs when the acquired asset no longer generates the financial results that were expected of them at the time of purchase.
UnitedHealth is planning to sell ClaimsXten, Change Healthcare’s claims payment and editing business to private equity firm TPG Capital for $2.2 billion, in a bid to ease antitrust concerns dogging the UnitedHealth-Change merger.
GE Healthcare and Medtronic partnered on outpatient care in which both will provide medical device products and software to ambulatory surgery centers (ASCs) and office-based labs (OBLs).
3 kidney care providers — Fresenius Health Partners, Cricket Health and InterWell Health — plan to merge in deal valued at $2.4B
Three US national credit reporting agencies (NCRAs) — Equifax, Experian, and TransUnion announced significant changes to medical collection debt reporting to support consumers faced with unexpected medical bills. These joint measures will remove nearly 70% of medical collection debt tradelines from consumer credit reports
AI-backed remote monitoring and digital therapeutics company Biofourmis raised $300 million in Series D funding, boosting the company to unicorn status with a valuation of $1.3 billion.
Global digital health funding reached a six-quarter low in the first quarter of 2022, ringing in at $10.4 billion, according to a new report from CB Insights. That’s a 36% decrease from the fourth quarter of 2021.
Palo Alto Networks used crowdsourced data from scans of more than 200,000 smart infusion pumps and discovered 3 in 4 had known security gaps, making them vulnerable to digital attacks.
Indian health and fitness app HealthifyMe has launched HealthifyPro, that comes with a biosensor system of continuous glucose monitor (BIOS) and other sensors and body monitors; a connected Smart Scale metabolic panel; and support from in-house coaches and AI assistant, Ria.
Longread of the Month
Quite an interesting overview of the digital health stack here by Christina Farr and and Marissa Schlueter Moore
Tweet of the Month
This tweet sums up (in a nice tongue-in-cheek fashion) what’s going on in the ‘Machine Learning’ field!
Chart of the Month
This chart, from FiercePharma’s review of Top 20 Pharma by revenue is quite telling!