Humanwide pilot collects data with mobile monitoring devices then pulls it into an EHR so the care team can help patients manage conditions.
A clinical trial program initiated by Stanford Medicine has deployed a data-driven, integrated team approach to predict and prevent disease and better detect overlooked health conditions and risks.
WHY IT MATTERS
The Humanwide pilot project uses science and technology to understand each patient, from lifestyle to DNA, and apply that knowledge to transform their health.
The organization’s model combines tools of biomedicine with a data-driven, team-based approach to focus on predicting and preventing disease before it strikes.
As part of the pilot, Humanwide patients used mobile monitoring devices, including a glucometer, pedometer, scale and blood pressure cuff, to regularly measure key health metrics.
The data automatically uploaded to their electronic health records for remote monitoring by their health care team, and the care team then helps the patient manage current health conditions and address future risks through a plan aligned with his or her personal goals.
ON THE RECORD
“With Humanwide, we’re able to focus on the whole human: who they are when they’re working, who they are when they’re playing, who they are when they’re at home,” Mahoney said in a statement. “This program demonstrates how we can zero in on what matters to a patient, to craft the entire care plan around their goals.”
In a paper in the Annals of Family Medicine, she and co-author Steven Asch outlined the early lessons of the year-long project, which involved 50 patients.
The paper noted encouraging the use of wearable devices in a healthy population helped identify multiple patients with early diabetes or hypertension, prompting early intervention and self-management.
The pilot participants underwent genetic assessments that gauged their risk for cancer and other diseases, and a pharmacogenomic evaluation that determined which types of drugs are most effective for their individual biology and cause the fewest side effects.
The patients also tracked key health metrics, such as blood glucose levels and blood pressure, using portable digital devices that beamed their readings back to their electronic health records for remote monitoring by care teams.
The teams, which included a primary care physician, nutritionist, behavioral health specialist and clinical pharmacist, used this data to inform each patient's care.
They also considered other factors, such as social and environmental determinants, and succeeded in identifying several previously overlooked health conditions and risks for different participants, from hypertension to heightened risk for breast cancer.
“Looking at genomic data and other factors that actually predict patient health allows us to be proactive instead of waiting for something to happen and having to react to that,” David Entwistle, president and CEO of Stanford Health Care, said in a statement. “Humanwide is an opportunity to build a deep understanding of each patient in a unique way.”
Computer algorithms were used to analyze 29 clinical variables in UPMC’s electronic health record systems, and were able to recognize patients with sepsis within six hours of arrival.
But it took a lot of learning to reach this stage and be able to spot the signs of sepsis and the hidden subtypes of sepsis, say researchers at Pitt Health Sciences, part of University of Pittsburgh Medical Center.
“For over a decade there have been no major breakthroughs in the treatment of sepsis; the largest improvements we’ve seen involve the enforcing of ‘one size fits all’ protocols,” says study lead author Christopher Seymour, MD, an associate professor in Pitt’s department of critical care medicine. But these protocols ignore that sepsis patients are not all the same.”
In fact, use of algorithms have found four distinct sepsis types:
- Alpha: the most common type (33 percent), comprising patients with the fewest abnormal laboratory test results, least organ dysfunction and lowest in-hospital death rate at 2 percent.
- Beta: older patients comprising 27 percent with the most chronic illnesses and kidney dysfunction.
- Gamma: similar frequency to beta but with elevated levels of inflammation and primarily pulmonary dysfunction.
- Delta: the least common at 13 percent, but the most deadly type, often with liver dysfunction and shock, and the highest in-hospital death rate at 32 percent.
Having analyzed the clinical variables of 20,000 patients, researchers then studied the electronic health records of 43,000 other UPMC sepsis patients and the four findings held. The findings held again when the team studied rich clinical data and immune response biomarkers from about 500 pneumonia patients enrolled at 28 hospitals across the nation.
The next step was to apply their findings to recently completed international clinical trials that tested promising therapies, but results were unremarkable.
Sepsis recognition can be tricky, says Derek Angus, MD, senior author of the study and an associate professor in Pitts’ department of critical care medicine. Most doctors are not confused about a classic case of sepsis, but those are only a very small portion of all cases, meaning that in most other cases the recognition of sepsis is known only when it has become obvious and is too late to make the first correct treatment moves, Angus notes.
In an “early goal-directed therapy (EGDT),” an aggressive resuscitation protocol that includes placing a catheter to monitor blood pressure and oxygen levels, delivery of drugs fluids and blood transfusions was found to have no benefit following a five-year $8.4 million study. But when Seymour’s team-reexamined the results, they found that EGDT was beneficial for patients with the Alpha type of sepsis, but EGDT resulted in worse outcomes for those with the Delta subtype.
“Intuitively, this makes sense as you would not give all breast cancer patients the same treatment,” Angus explains. “Some breast cancers are more invasive and must be treated aggressively. Some are positive or negative for different biomarkers and respond to different medications. The next step is to do the same for sepsis that we have for cancer—to find therapies that apply to the specific types of sepsis and then new clinical trials to test them.”
That’s why it is imperative that patients have their vitals and labs captured upon arrival at the hospital, Seymour says. Sepsis requires the presence of organ disruption and six organs can be effected by the disease. Consequently, early treatment intervention should be done within 6 hours of suspected sepsis as the time window for capturing data at hospital presentation is 6 hours.
Capturing the vitals and labs early, with additional information available in the electronic health record, quickly helps physicians at the bedside to wrap their minds around the patient’s physiology. But now, physicians have another powerful tool at their disposal—machine learning technology.
Machine learning can find patterns that doctors cannot—much more than the three to four variables that doctors usually use. Data in the EHR can help doctors select variables to consider and then run machine learning models in collaboration with biostatisticians and computer scientists, says Seymour.
“We rely on doctors to find sepsis and quickly get patients on antibiotics, and we have machine learning and the EHR to parse out the type of sepsis,” he adds.
Samyukta Mullangi, John P. Pollak, Said Ibrahim
Health systems do not systematically collect information on social determinants of health (SDH) — the conditions in which people are born, live, grow, and age — despite knowing that they have a big impact on individual and population health. But the shift from reimbursing providers for the volume of services they deliver (fee for service) to the quality of patient outcomes relative to cost (value) is causing them to focus more on maintaining patients health and not just curing disease. This shift is causing providers to start investing in population health management strategies, which require them to better understand the local population and identify unmet needs.
The challenge is that the SDH information that physicians collect from patients and enter into their electronic medical records (EMRs) is pretty limited. Even though 83% of family physicians agree that the Institute of Medicine’s 2014 recommendation that they collect sociodemographic, psychological, and behavioral information from patients and put it into their EMRs, only 20% say they have the time to do so. But alternative means of collecting such information are emerging: smartphones, credit card transactions, and social media.
Smartphones. The Pew Research Center estimates that more than three-fourths of Americans now own smartphones. One example of how these devices could be used to collect SDH information involves the mobile applications that health systems offer to allow patients to easily book appointments or contact medical providers. These apps can also access information on patients’ location, which can be cross-referenced with rich databases like Foursquare’s book of local businesses or city-level heat maps on crime/domestic violence to understand a patient’s experience of his or her neighborhood — e.g., the availability of fresh food via local grocers or bodegas and the ability to exercise outside in relative safety. In a research setting, this type of location sharing has yielded startling insights.
In one interesting study on smoking cessation and relapses, patients’ location data, along with their self-reporting on their craving levels and smoking status, was overlaid on a point-of-sale tobacco outlet geodatabase to demonstrate that an individuals’ daily exposure to these retail outlets was significantly associated with lapses even when cravings were low. This real-time quantification about an individual’s interactions with her local environment unearthed novel influences on health behaviors that were likely invisible to the patient herself. This type of geolocation data is currently still being developed and tested in the research setting, but one day it may be used to make patients more aware of these triggers and resist unhealthy temptations.
Credit-card transactions. These are another goldmine of information that can help round out the medical record. For instance, a Gates Foundation- and United Nations Foundation-funded investigation into the economic, social, and health status of women in developing countries combined credit card records with records on their phone calls to identify patterns in people’s socioeconomic behaviors. The analysis resulted in six distinct lifestyle clusters in terms of expenditure patterns, age, mobility, and social networks. One can imagine that this type of aggregation can be useful as health systems increasingly work to tailor community and outreach programs to patients.
Credit-card statements do not contain the details necessary to generate insights ( i.e., what actual items make up a bill from the grocery store). That level of granular detail would go a long way into understanding whether patients fill their prescriptions, purchase cigarettes, or order salads. Some digital grocers (e.g., Instacart, Peapod), drug retailers (e.g., CVS, Walgreens), and payment kiosks (e.g., Square) are now emailing itemized receipts to consumers (with their consent). One group at Cornell Tech has created software tools that scrape these receipts and analyze purchases against a patient’s personal nutritional goals, a research effort with commercial potential. Such approaches not only collect information on SDH but also raise the patients’ level of awareness of the relationship between healthy behaviors and health itself.
Social media. Leveraging the willingness of people to divulge personal details on social media is yet another emerging frontierin the effort to collect SDH data. It is being used to successfully access populations that have historically been considered hard to reach: younger people, females, and low-income individuals. New features on popular sites like Facebook that allow individuals to mark themselves safe during natural disasters represent an initial foray to using this medium for gathering more SDH data. Health systems that engage patients via social media can elicit answers to questions around food insecurity, employment status, physical activity, and so on. In fact, new research suggests that many adult Facebook and Twitter users are willing to share their social media and medical data and link it with EMR data for research purposes.
Certainly, several pragmatic issues might create barriers to applying these approaches. An obvious one is privacy. More research will need to be done to ascertain patients’ comfort with novel ideas such as giving physicians access to their purchase histories or locations. It is also critical that the information gathered through these novel mechanisms not be used in a punitive manner but rather to inform clinician counseling and to support patients in their efforts to pursue healthy behaviors. Patients are not likely to share credit card or social media data, for example, if they perceive there to be a link between the information gathered and punitive responses such as the denial of insurance coverage or increased co-pays.
Another obstacle lies in the very act of obtaining consent from a large number of patients to participate in such information-gathering programs. One notable effort at Parkland Hospital in Dallas, which linked data about patients’ usage of food pantries, homeless shelters, and other social services with their medical records, found that patients were more willing to be enrolled into a digital database when asked to do so by community partners that had earned their trust rather than in the emergency room. Discouragingly, privacy concerns over the Trump administration’s policies tying social services usage with legal status has caused many undocumented immigrants to ask to be erased from social services’ IT systems.
Finally, it may be difficult to obtain buy-in from physicians who are already suffering from information overload. To overcome it, data will need to be turned into intelligent summaries with clear visuals and actionable takeaways. Additionally, clinics need to invest in support staff and ancillary services that help at-risk patients. For example, clinics can be outfitted with connections to community-based resources (housing programs, job training centers, and nutritional supplement programs). These investments will go a long way to ensuring that physicians are receptive to the work of monitoring additional data about SDH.
With these elements in place, health care systems will be able to harness digital technologies to identify the needs and interventions required to create healthier communities.
The authors wish to acknowledge Jessica Ancker for her critical review of this manuscript.
An artificial intelligence tool can help diagnose post-traumatic stress disorder in veterans by analyzing their voices, a new study found.
Medical researchers and engineers designed an AI tool that can distinguish, with 89% accuracy, between the voices of those with or without PTSD, according to their study published Monday in Depression and Anxiety. The findings open up the possibility of using the AI-based voice analysis tool to diagnose PTSD more rapidly or through telemedicine.
“Our findings suggest that speech-based characteristics can be used to diagnose this disease, and with further refinement and validation, may be employed in the clinic in the near future,” senior study author Charles Marmar, M.D., from the department of psychiatry at NYU School of Medicine, said in a statement. A division of the U.S. Army supported the study.
The U.S. Department of Veterans Affairs reports that between 11% and 20% of veterans who served in operations in Iraq and Afghanistan have PTSD, while about 12% of Gulf War veterans have PTSD. Additionally, it is estimated that 30% of Vietnam veterans have had PTSD in their lifetimes.
The ability to improve PTSD diagnosis has wider implications, as more than 70% of adults worldwide experience a traumatic event at some point in their lives, with up to 12% of people in some struggling countries suffering from PTSD, according to the Sidran Institute.
According to researchers, the ability to accurately screen for and diagnose PTSD remains challenging. The diagnosis is usually based on clinical interviews or self-report measures. The gold standard for diagnosing the condition is the clinician-administered PTSD scale, a structured clinical interview to assess the frequency and severity of PTSD symptoms and related functional impairments. However, even that assessment is subject to biases. The interviews also require a lengthy visit to a clinician’s office, which some patients may be unwilling or unable to do.
An objective test is lacking, according to the researchers, who developed a classifier of PTSD based on objective speech-marker features that discriminate PTSD cases from controls. The research team included psychiatrists from New York University School of Medicine, Steven and Alexandra Cohen Veterans Center for the Study of Post-Traumatic Stress and Traumatic Brain Injury and engineers from SRI International, the institute that also invented Apple’s Siri feature.
For the study, researchers used speech samples from war zone-exposed veterans, 53 cases with PTSD and 78 controls, assessed with the clinician-administered PTSD Scale. Audio recordings of clinical interviews were used to obtain 40,526 speech features, which the team’s AI program sifted through for patterns.
The program linked patterns of specific voice features with PTSD, including less clear speech and a lifeless, metallic tone, both of which had long been reported anecdotally as helpful in diagnosis.
The theory is that traumatic events change brain circuits that process emotion and muscle tone, which affects a person’s voice, according to researchers.
“We believe that our panel of voice markers represents a rich, multidimensional set of features which with further validation holds promise for developing an objective, low cost, noninvasive, and, given the ubiquity of smartphones, widely accessible tool for assessing PTSD in veteran, military, and civilian contexts,” the researchers said.
Other healthcare researchers are also exploring the use of voice analysis to detect and diagnose disease. A team at Mayo Clinic is exploring how to use AI-supported voice analysis as a noninvasive diagnostic tool to identify changes in tone or cadence that could potentially be predictive of an outcome, such as high blood pressure, stroke or heart attack.
The research team behind this latest study plans to train the AI voice tool with more data, further validate it on an independent sample and apply for government approval to use the tool clinically.