What’s the most amazing thing about being in the intersection of healthcare and technologies? With the press of a button, we can comb and combine vast amounts of data to help a patient, find new medical treatments, and make new diagnoses.
We call the science of processing data informatics, or information science. And successful healthcare systems know that using and analyzing this wealth of data can lead to great things.
These things can include new tests for breast cancer based on the genes in one’s gut. Or making it easier and more accurate to diagnosis high blood pressure. There are even ways to combine Google search questions with city smoking rates and poverty levels to come up with lung cancer probabilities.
— Steven Chan, MD MBA (@StevenChanMD) November 13, 2016
This picture, for instance, shows a Venn diagram. This Venn diagram shows how combining data from billing, clinical notes, medication prescriptions, and lab and test results can help determine a person’s diagnosis.
Many people work in the informatics space for healthcare. The American Medical Informatics Association—a professional organization for this space—recently welcomed over 2,500 attendees to its annual conference. They came together in late 2016 to share ideas on information technology and data.
AMIA counts “translational scientists, clinicians, computer scientists, nurse informaticians, clinical researchers, public health officials, and those who specialize in telemedicine and consumer health informatics” amongst its attendees.
The conference itself is a glimpse into the future. Here are some of the highlights of AMIA conference and their implications for healthcare technologies in 2017:
New tests for diseases and treatments
The concept of patient-generated data is where devices and sensors track different parts of our lives—a heartbeat, a breath, the sugar levels in one’s blood. This patient-generated data can be used for wellness, health, and biomedical research.
This can be used for cancers, such as head & neck cancer. One such patient-generated data project, CYCORE—or Cyberinfrastructure for Comparative Effectiveness Research—brings together researchers from the University of California, San Diego (@UCSD) and the MD Anderson Cancer Center (@MDAndersonNews).
“CYCORE includes sensors, ecological momentary assessment, video, and tracking this data over time,” explained researcher Susan Peterson (@S_K_Peterson). The concept of ecological momentary assessment brings questionnaires to someone at particular times of the day (i.e., scheduled, fixed times, or if a person is away from their home).
CYCORE uses heart rate, blood pressure, and weight to monitor dehydration risk because, in head & neck cancers, the chemotherapy and radiation can cause swallowing issues and pain. This hardship discourages cancer patients from drinking enough fluids. When that happens, dehydration results and can cause falls and injuries. Watching out for this could help reduce hospital and emergency department visits.
— Steven Chan, MD MBA (@StevenChanMD) November 15, 2016
But informatics can go beyond patient-generated data, clinical data, and population-level statistics and incorporate the exposome, or individual environmental health risk factors.
According to Fernando Martin (@Fermarsan) Ph.D. FACMI, at New York-based Cornell University, there is a link between this exposome—environmental risk factors (including pollution and radiation)—with one’s anatomy and physiology (how the body works) and biomarkers (including DNA sequences). All this can be linked with data to lead to a more precise picture of one’s health and status.
Large-scale community testing
“Everyone has a computer in their pocket and access to next-generation DNA science,” said Richman.
Her company, uBiome, sequences genes found in fecal matter to learn about the bacteria within one’s gut. And they anticipate becoming the standard of care, allowing drugs to be discovered more readily.
uBiome itself has the largest research microbiome dataset in the world, storing over 100,000 microbiome samples from over 65 countries. This rich dataset attracts researchers from institutions like Oxford University (@UniofOxford) and the United States Centers for Disease Control (CDC, @CDCgov). For instance, Oxford researchers are studying correlations between personality and the microbiome in United States and United Kingdom populations.
This large-scale use of data can also help hospitals learn who is at higher risk for coming back to the hospital. At Washington Hospital Center, a recent study by Bayati and colleagues published in 2014 combed 20 years of hospital data to predict this so that they could avoid the cost of readmissions—US$13,679, to be exact.
Eric Horvitz M.D., Ph.D. (@EricHorvitz, @MSFTResearch, @Microsoft), who manages Microsoft Research’s machine learning and intelligence research, also discussed how techniques like this can apply to home life. Horvitz’s team combed through smoking rates, poverty levels, U.S. Environmental Protection Agency data, home age, and any searches that people made in their web browser for “hoarseness” or “blood” or “Nicorette gum”. With this data, they were able to predict who was at greater risk for lung cancer.
Assisting doctors to let them make better decisions for you
Horvitz also said that this data can help achieve complementarity so that the human mind is complemented by computers in areas in which humans are weaker—in the areas of memory, attention, and judgment.
For instance, Wang, Khosla, and colleagues studied pathologists—physicians who study tissue samples—and compared them with pathologists who are helped by computers. The pathologists were looking at metastatic breast cancer slides. The study found that the pathologists had an error rate of 3.4%, but when they are helped by computers, the error rate drops to 0.5%.
The United States National Library of Medicine (NLM, @NLM_news) also helps doctors make better decisions. The NLM goes beyond bookshelves as a central worldwide data bank for healthcare.
“We are a platform for discovery, an enabler of science, ensuring science serves society,” said NLM Director Patti Brennan (@NLMdirector). Brennan spoke on the NLM’s accomplishments, including automatic image analysis to detect cervical cancer and running innovation contests for medication image recognition. The NLM is even working on a way to automatically analyze chest X-rays.
By late 2017, direct deposit of data will allow NLM’s popular PubMed service to store researchers’ data, along with journal articles.
“It’s not your mother’s library anymore,” Brennan said.
Getting data is important, tough work
As easy as informatics may sound, it’s not as simple as plugging in a USB stick and double-clicking on files. Many systems have thousands of database tables and petabytes of data that make it a challenge to make any sense of data.
— Steven Chan, MD MBA (@StevenChanMD) November 15, 2016
This has real-world implications. Katherine Kim, Ph.D., professor at the University of California, Davis (UC Davis) Betty Irene Moore School of Nursing, has observed that patients with cancer must navigate a maze of different websites and health systems. Because of data coordination challenges, each website portal can have incomplete information, and that can be dangerous.
Even using just one electronic health record (EHR) system can be challenging. According to Robert Grundmeier (@RGrundmeier, @ChildrensPhila), Epic, one of the most popular EHR’s, requires at least 4 people to implement new modules.
One initiative, the Cisco-UCSF Connected Health Interoperability Platform, is actively developing new ways to take this grunt work and complexity out of EHR systems. The platform, led by Aaron Neinstein, MD (@AaronNeinstein), is part of the UCSF Center for Digital Health Innovation (@UCSF, @UCSFCDHI).
— Steven Chan, MD MBA (@StevenChanMD) November 14, 2016
According to Neinstein, however, it is difficult to make new changes in healthcare informatics. Innovation does not scale because of tight healthcare technology budgets, expensive app pilots, no sustainable business model for healthcare information exchanges, and standards problems. Having a standard platform can make the innovation scale much faster, Neinstein said.
“When Apple released the iPhone in 2007, who imagined we’d use it for Angry Birds, Snapchat, and Uber?” The next step is to build a platform for healthcare, Neinstein said, is with interchangeable apps. And consumers are now familiar with apps embedded in apps.
Compounding this problem is that a critical factor of one’s health—mental health—is cordoned off with even stricter, fragmented regulations. Ethical, medical, and regulatory hurdles put more restrictions on the sharing of psychiatric and addictions data—and these include “42 CFR Part 2,” which makes “sensitive notes” that only mental health and emergency medicine clinicians are able to read.
“Providers are nervous about having mental health and substance use diagnoses on problem lists because of stigma and insurance concerns,” David Bates MD (@DBatesSafety) said.
Despite all of these difficulties, there is enormous promise in healthcare informatics. The White House Office of Science & Technology Policy (@WhiteHouseOSTP), for instance, has invested US$200M for big data initiatives at the U.S. National Institutes of Health (NIH) and the Defense Advanced Research Projects Agency (DARPA).
Next year, 2017, will include further advances in large-scale community testing, population-level health, patient-generated data, doctors assisted by computers, and overcoming hurdles for data sharing and systems implementation.