Recent quotes:

Machine learning results: pay attention to what you don't see - STAT

Beyond examining multiple overall metrics of performance for machine learning, we should also assess how tools perform in subgroups as a step toward avoiding bias and discrimination. For example, artificial intelligence-based facial recognition software performed poorly when analyzing darker-skinned women. Many measures of algorithmic fairness center on performance in subgroups. Bias in algorithms has largely not been a focus in health care research. That needs to change. A new study found substantial racial bias against black patients in a commercial algorithm used by many hospitals and other health care systems. Other work developed algorithms to improve fairness for subgroups in health care spending formulas.

Living a longer, healthier life: A systems approach to medicine at WesternU's Pumerantz Lecture | Benzinga

"We think that clinical trials in the future ought to be done as N-of-1 (single subject) experiments," Hood said. "In a cancer trial we can use the individual data clouds to actually identify biomarkers that distinguish the responders from the non-responders. Then what we will do is a second trial of 50 patients with all responders. And if you get a 98 percent response rate, the FDA will approve your drug in the blink of an eye. You go from spending $1.5 billion on a clinical trial to spending hundreds of thousands of dollars on a clinical trial." Hood and his collaborators completed a study that used dynamic data clouds to improve wellness. "A wellness study of 108 individuals using personal, dense, dynamic data clouds," published in Nature Biotechnology in 2017, involved collecting the complete genome sequences of 108 volunteers, including Hood. The subjects each gave blood draws every three months to measure 1,200 analytes of three classes: clinical chemistries, metabolites and proteins. They measured gut microbiome every three months and used Fitbit and other devices for digital health measurements. "What these gave us for each individual was longitudinal data clouds that, when analyzed, led to actionable possibilities that if executed by the individual could either improve their wellness and/or let them ameliorate or avoid disease," Hood said. "The big question in all of this is: How do you change social behavior? What we used were wellness coaches, trained in psychology and nutrition and nursing, who were magnificent. They would elicit from the individual exactly what they wanted in their health objectives, and this is not easy to do."

Potential diagnostic tool, treatment for Parkinson's disease -- ScienceDaily

In the new study, Wang's team obtained skin samples from 83 Parkinson's patients, five asymptomatic close relatives considered to be at heightened risk, 22 patients diagnosed with other movement disorders and 52 healthy control subjects. They extracted fibroblasts -- cells that are common in skin tissue -- from the samples, cultured them in petri dishes and subjected them to a stressful process that messes up mitochondria. This should result in their clearance, necessarily preceded by removal of Miro molecules tethering them to the grid. Yet the researchers found the Miro-removal defect in 78 of the 83 Parkinson's fibroblasts (94%) and in all 5 of the "high-risk" samples, but not in fibroblasts from the control group or other or from patients with other movement-disorders.

Semiconductor Engineering .:. Spreading Intelligence From The Cloud To The Edge

To handle all of these bits, at least some processing has to be done at the edge. It takes far too much time, energy and money to move it all—and the bulk of it is useless. But so far there is no agreement on how or where this will be done, or by whom. Cloud providers still believe hyperscale data centers are the most efficient tool to grind down the mountains of operational data produced by IoT devices every day. Device makers, in contrast, believe they can pre-process much of that data at or close to the source if they can put a smart enough, purpose-built machine learning inference accelerator in the device.

Next for Apple Watch: a clinical trial with J&J to track heart health - STAT

Typically, clinical trials use their own research infrastructure to track how patients do, but the HEARTLINE trial will plug into insurance claims databases to track patients. This approach, called a pragmatic clinical trial, could be cheaper and more efficient than the way studies are conducted now, without sacrificing the clarity and certainty that comes from having a control group in a study. But this is also risky: Using insurance claims data this way is new and untested. “It’s certainly a vast and gross departure form the bricks and mortar model,” said Gibson. “This is one of the most exciting things, the idea that you’re going to find participants through the media, the news, potentially through insurers and even health care providers.”

Reference genome is threatening dream of personalized medicine - STAT

On March 23, 1997, the then-nascent Human Genome Project placed an ad in a newspaper in Buffalo, N.Y. (site of a project scientist’s lab), seeking volunteers to donate blood from which they would sequence DNA. Through a quirk of whose DNA got processed when, about 70 percent of the reference genome comes from an anonymous man designated RP11, said UW genome scientist Evan Eichler, with the rest from a few score other volunteers. The reference genome is therefore a mashup of the sequences of these everyday people. As a result, it isn’t a perfectly healthy genome: It has at least 3,556 variants that increase the risk of diseases, including type 1 diabetes and hypertension. Its most serious shortcoming, however, reflects the fact that 1990s Buffalo was not exactly the United Nations. Its ethnic populations are almost all European — German, Irish, Polish, and others. The reference genome, therefore, is as well. That had long been known, but was largely swept under the rug.

The neuroactive potential of the human gut microbiota in quality of life and depression | Nature Microbiology

Butyrate-producing Faecalibacterium and Coprococcus bacteria were consistently associated with higher quality of life indicators. Together with Dialister, Coprococcus spp. were also depleted in depression, even after correcting for the confounding effects of antidepressants. Using a module-based analytical framework, we assembled a catalogue of neuroactive potential of sequenced gut prokaryotes. Gut–brain module analysis of faecal metagenomes identified the microbial synthesis potential of the dopamine metabolite 3,4-dihydroxyphenylacetic acid as correlating positively with mental quality of life and indicated a potential role of microbial γ-aminobutyric acid production in depression.

The shifting model in clinical diagnostics: how next-generation sequencing and families are altering the way rare diseases are discovered, studied, and treated | Genetics in Medicine

Until very recently, the fragmented distribution of patients across institutions hindered the discovery of new rare diseases. Clinicians working with a single, isolated patient could steadily eliminate known disorders but do little more. Families would seek clinicians with the longest history and largest clinic volume to increase their chances of finding a second case, but what does a physician do when N = 1 or if the phenotype is inconsistent across patients? These challenges are driving an increase in the use of NGS. Yet this technological advance presents new challenges of its own. Perhaps the most daunting, in our opinion, is the inability to share sequencing data quickly and universally. Standards and bioinformatic tools are needed that allow for a national repository where families or scientists can bring clinical results and NGS data for comparison. This challenge can be circumvented by tools already created for and by the Internet and social media.

When the insurance company monitors your driving in real time does it help? New research finds that it helps on a number of levels, from safety to consumer cost -- ScienceDaily

"We found that UBI users tend to improve the safety of their driving in general, and in once specific area by decreasing their daily average number of hard-brakes by an average of 21 percent after six months," said Miremad Soleymanian. "Our research found that the number miles driven tend to stay the same and that both younger drivers and females tend to improve their UBI scores more than older drivers and males."

Side-effects not fully reported in more than 30 percent of healthcare reviews -- ScienceDaily

The new study looked at the reporting of adverse events in 187 systematic reviews published between 2017 and 2018. Systematic reviews in health research aim to summarise the results of controlled healthcare interventions and provide evidence of the effectiveness of a healthcare intervention. Research showed that 35 per cent of reviewers did not fully report the side-effects of the medical intervention under review. Dr Su Golder, from the University of York's Department of Health Sciences, said: "Despite reviewers stating in their own protocols that adverse events should be included in the review, 65 per cent fully reported the event as intended by the protocol, eight per cent entirely excluded them, and the remaining 27 per cent either partially reported or changed the adverse event outcomes." "Just over 60 per cent, however, didn't even include adverse events in their protocols, which suggests that a more proactive approach is needed to prompt reviewers to report on potential harmful side-effects in their reporting of healthcare interventions."

Do differences in gait predict the risk of developing depression in later life? -- ScienceDaily

Gait parameters and mental health both have significant impacts on functional status in later life. The study's findings suggest that gait problems may represent a potentially modifiable risk factor for depression.

Gut Bacteria Linked to Depression Identified – Neuroscience News

Mireia Valles-Colomer (VIB-KU Leuven): ‘Many neuroactive compounds are produced in the human gut. We wanted to see which gut microbes could participate in producing, degrading, or modifying these molecules. Our toolbox not only allows to identify the different bacteria that could play a role in mental health conditions, but also the mechanisms potentially involved in this interaction with the host. For example, we found that the ability of microorganisms to produce DOPAC, a metabolite of the human neurotransmitter dopamine, was associated with better mental quality of life.’

‘Omnigenic’ Model Suggests That All Genes Affect Every Complex Trait | Quanta Magazine

Starting about 15 years ago, geneticists began to collect DNA from thousands of people who shared traits, to look for clues to each trait’s cause in commonalities between their genomes, a kind of analysis called a genome-wide association study (GWAS). What they found, first, was that you need an enormous number of people to get statistically significant results — one recent GWAS seeking correlations between genetics and insomnia, for instance, included more than a million people. Second, in study after study, even the most significant genetic connections turned out to have surprisingly small effects. The conclusion, sometimes called the polygenic hypothesis, was that multiple loci, or positions in the genome, were likely to be involved in every trait, with each contributing just a small part. (A single large gene can contain several loci, each representing a distinct part of the DNA where mutations make a detectable difference.)

will.i.am on personal data ownership

Personal data needs to be regarded as a human right, just as access to water is a human right. The ability for people to own and control their data should be considered a central human value. The data itself should be treated like property and people should be fairly compensated for it.

WillIAm-on-consumer-health-data-services

Today, my gadgets may count my steps, but they aren’t seeing the big picture: what I ate, how I felt, what my blood pressure is. New services, built from the point of view of the consumer, will benefit me by sharing and interconnecting my own data, rather than selling it on. When more trust is established, my personal “agent” or “assistant” should merge relevant things together that are currently just disconnected data points.

BARBARIANS AT THE GATE: CONSUMER-DRIVEN HEALTH DATA COMMONS AND THE TRANSFORMATION OF CITIZEN SCIENCE

A few state court cases have found patients own their medical records under specific circumstances.118 Unfortunately, the pertinent body of state medical records law generally applies in traditional healthcare settings and seemingly does not govern commercial providers of PHD devices and services, such as purveyors of medical and fitness devices. Courts do not recognize an individual property right in personal information such as one’s name, address, and social security number.119 Commercial databases that hold such information are generally treated as the property of the companies that compiled them.120 In a famous case121 where plaintiffs sought to block a company from disclosing their personal information by selling its mailing lists, Vera Bergelson notes an implicit judicial bias “that, to the extent personal information may be viewed as property, that property belongs to the one who collects it.”122 This bias— if it exists—is reminiscent of the ancient res nullius doctrine from natural resource law, which treated assets such as subsurface mineral deposits and wild animals as unowned until somebody discovers and captures (takes possession of) them.123 “Rarely used today, it let private owners stake claims as in the Klondike gold rush.”124

BARBARIANS AT THE GATE: CONSUMER-DRIVEN HEALTH DATA COMMONS AND THE TRANSFORMATION OF CITIZEN SCIENCE

This article explores how these mechanisms, imbedded in major federal research and privacy regulations, enshrine institutional data holders—entities such as hospitals, research institutions, and insurers that store people’s health data—as the prime movers in assembling large-scale data resources for research and public health. They rely on approaches—such as de-identification of data and waivers of informed consent—that are increasingly unworkable going forward. They shower individuals with unwanted, paternalistic protections—such as barriers to access to their own research results—while denying them a voice in what will be done with their data.

Privacy in the age of medical big data | Nature Medicine

Big data is often defined by ‘three Vs’: volume (large amounts of data), velocity (high speed of access and analysis), and variety (substantial data heterogeneity across individuals and data types), all of which appear in medical data2.

All of Us enrollees can now share health data from their Fitbit accounts with researchers | MobiHealthNews

“Collecting real-world, real-time data through digital technologies will become a fundamental part of the program,” Eric Dishman, director of the All of Us Research Program, said in a statement. “This information in combination with many other data types will give us an unprecedented ability to better understand the impact of lifestyle and environment on health outcomes and, ultimately, develop better strategies for keeping people healthy in a very precise, individualized way.”

Everything big data claims to know about you could be wrong: To understand human health and behavior, researchers would do better to study individuals, not groups -- ScienceDaily

"If you want to know what individuals feel or how they become sick, you have to conduct research on individuals, not on groups," said study lead author Aaron Fisher, an assistant professor of psychology at UC Berkeley. "Diseases, mental disorders, emotions, and behaviors are expressed within individual people, over time. A snapshot of many people at one moment in time can't capture these phenomena." Moreover, the consequences of continuing to rely on group data in the medical, social and behavioral sciences include misdiagnoses, prescribing the wrong treatments and generally perpetuating scientific theory and experimentation that is not properly calibrated to the differences between individuals, Fisher said.

Can Big Data Help Psychiatry Unravel the Complexity of Mental Illness? - Scientific American

Psychiatrist Charles DeBattista of Stanford University and colleagues, compared electroencephalograms (EEGs) collected from depressed patients, with a database of EEGs from over 1,800 patients that included information about response to specific treatments. Using EEG measures to guide decisions about treatment alternatives led to significantly better outcomes than clinical treatment selection.

Is soda bad for your brain? (And is diet soda worse?): Both sugary, diet drinks correlated with accelerated brain aging -- ScienceDaily

Now, new research suggests that excess sugar -- especially the fructose in sugary drinks -- might damage your brain. Researchers using data from the Framingham Heart Study (FHS) found that people who drink sugary beverages frequently are more likely to have poorer memory, smaller overall brain volume, and a significantly smaller hippocampus -- an area of the brain important for learning and memory. But before you chuck your sweet tea and reach for a diet soda, there's more: a follow-up study found that people who drank diet soda daily were almost three times as likely to develop stroke and dementia when compared to those who did not.

Fitbit creates research library with Fitabase, publishes results of corporate wellness study | MobiHealthNews

The library currently has 163 different published studies that mention using a Fitbit (or a few of them) as part of their study design. The pace of research using the wearables has been accelerating every year, Ramirez said, posing what Fitabase believed was a need for a comprehensive library. “So we wanted to make it a public resource where anyone who wants to explore Fitbit research can have a one-stop shop. It’s meant to be a library down the street, and it will continue to grow as people do more research.”

Instagram photos reveal predictive markers of depression

Using Instagram data from 166 individuals, we applied machine learning tools to successfully identify markers of depression. Statistical features were computationally extracted from 43,950 participant Instagram photos, using color analysis, metadata components, and algorithmic face detection. Resulting models outperformed general practitioners' average diagnostic success rate for depression. These results held even when the analysis was restricted to posts made before depressed individuals were first diagnosed. Photos posted by depressed individuals were more likely to be bluer, grayer, and darker. Human ratings of photo attributes (happy, sad, etc.) were weaker predictors of depression, and were uncorrelated with computationally-generated features. These findings suggest new avenues for early screening and detection of mental illness.

How Vector Space Mathematics Reveals the Hidden Sexism in Language

The team does this by searching the vector space for word pairs that produce a similar vector to “she: he.” This reveals a huge list of gender analogies. For example, she;he::midwife:doctor; sewing:carpentry; registered_nurse:physician; whore:coward; hairdresser:barber; nude:shirtless; boobs:ass; giggling:grinning; nanny:chauffeur, and so on. The question they want to answer is whether these analogies are appropriate or inappropriate. So they use Amazon’s Mechanical Turk to ask. They showed each analogy to 10 turkers and asked them whether the analogy was biased or not. They consider the analogy biased if more than half of the turkers thought it was biased.

New Depression Model Outperforms Psychiatrists

Data mined from clinical trials may soon help doctors tailor antidepressant therapy to their patients, the authors say. Currently, only about 30% of patients get relief from the first drug they are prescribed, and it can often take a year or more before doctors find the right medication to alleviate symptoms of depression. The Yale team analyzed data from a large clinical trial on depression and pinpointed 25 questions that best predicted the patients’ response to a particular antidepressant. Using these questions, they developed a mathematical model to predict whether a patient will respond to Celexa after three months of treatment. “These are questions any patient can fill out in 5 or 10 minutes, on any laptop or smartphone, and get a prediction immediately,” explained Adam Chekroud, Ph.D. candidate in the Human Neuroscience Lab and lead author of the paper.