Dr Reid Ponte: Could you tell me a bit about your background?
Dr Cummins: I call myself an "accidental" nurse. I originally entered into an electrical engineering major at the University of Illinois but didn't connect with the discipline. I had several friends at the time who were starting the nursing program on the Chicago campus. I decided to do the same. I was fortunate to be offered a job as a student nurse assistant in the emergency department (ED) at the University of Illinois Hospital, which ended up transforming me. I learned firsthand the importance of effective interprofessional teamwork and the importance of timely access to accurate information in caring for patients and families.
I practiced as an ED nurse in Chicago and later a faculty member at Northern Kentucky University (NKU) where I graduated from the Family Nurse Practitioner track. I worked on a chart abstraction project to examine whether primary care provider practice was consistent with guidelines for treating major depressive disorder. In abstracting assessments and decisions from paper charts, I realized how poorly paper records supported our ability to accomplish what we now call "learning health," the capacity to observe our own practice patterns and their links to outcomes as the basis for improvement. This sparked an interest in electronic health records (EHRs) and the power of electronic data, what we now call big data.
Prof Denise Robinson and other faculty members at NKU encouraged me to pursue doctoral education in nursing. This led me to Indiana University's PhD program and an exceptional scientific mentor, Dr Anna McDaniel. Dr McDaniel helped me to develop a dissertation involving artificial neural networks, a method for discovering patterns in big data, and the method upon which current deep learning approaches are based. I applied these methods in national survey data in order to develop a predictive model of smoking cessation status. It was one of the earliest dissertations of its nature in the discipline of nursing. The use of big data methods was criticized as a fishing expedition by some faculty. I was later recruited to the University of Utah where I began my work in teaching and mentoring PhD students and conducting my program of research. I benefitted greatly from the mentoring and support of faculty members in the Nursing Informatics and the Department of Biomedical Informatics. At the University of Utah, I began to pursue a program of applied informatics research in collaboration with the Utah Poison Control Center. I was later appointed associate dean for Research and PhD Program.
Dr Reid Ponte: How and why do big data, population health, and interoperability of electronic health data/records matter to you and your program of research?
Dr Cummins: These concepts are particularly compelling when you consider the case of poison control centers. These are 24/7 call centers that serve both the public and healthcare providers with specialty consultation when someone is exposed to a poisonous substance. Their ability to be effective depends upon rapid, effective, communication, and information sharing. However, they have some of the least developed information technology (IT) systems in healthcare. Since they don't bill for services, there aren't strong financial incentives to invest in IT in these settings. For example, the recent meaningful use incentives and disincentives implemented by the Centers for Medicare & Medicaid Services simply do not apply to poison control centers. Moreover, these centers tend to be underfunded and lack necessary resources to advance their IT infrastructure. Poison control centers depend almost entirely upon the telephone to share information and consult with providers.
I obtained funding from the Agency for Healthcare Research and Quality (AHRQ) to study this telephone-based process and determine the requirements for a new process that would be strengthened with health information exchange. We identified a number of inefficiencies and safety vulnerabilities in the verbal, telephone-based process. AHRQ awarded further funding to develop and evaluate a bidirectional health information exchange process between EDs. We built software called SNOWHITE that enables poison control centers to participate in health information exchange and approximately 2 years ago accomplished the 1st participation of a US poison control center in standards-based health information exchange.
I believe that data should be private and secure, but fluid. It should follow patients from care setting to care setting, minimizing patient and provider burden of data entry, and I believe that relevant information should be distilled and presented to care providers at the point of decision making. There is a lot we can do in this area, and I'm hopeful that new regulations associated with information blocking provisions of the 21st Century Cures Act (https://www.congress.gov/114/bills/hr34/BILLS-114hr34enr.pdf; https://www.congress.gov/bill/114th-congress/house-bill/34) will help us realize the potential of interoperability and health information exchange.
Dr Reid Ponte: You mentioned the term Machine Learning Methods when describing the logistics of your research program. What does that mean?
Dr Cummins: Machine learning methods are algorithmic approaches to identifying patterns in data. These methods largely originated from the field of artificial intelligence (AI) and are called learning methods because they typically involve incremental adjustments to a model based on exposure to new data, new examples. These methods can be used to induce algorithms that predict a certain outcome based on other variables. Those algorithms can then be implemented in information systems to provide support for decision making or the operation of devices. These methods are widely used in e-commerce to recommend books, movies, or clothing. Insurance and mortgage companies use these methods to predict outcomes of interest in their sector. In healthcare, they are found in many medical devices and used to predict your risk of certain health conditions.
Dr Reid Ponte: What would you like JONA readers to know about the power of informatics in care delivery and the health of the nation that they may not already?
Dr Cummins: I realize that JONA readers are a diverse group of nurse leaders who work across the healthcare sector. The priorities differ among settings. In settings without a strong nursing informatics presence, nurses become the victim when they have little input into the way systems are designed or configured, they experience severe workflow disruption from using the EHR, and the systems do not support their work through decision support tools or other features that add value. In these situations, a focus on nursing engagement and representation in IT decision making is essential. I hope JONA readers will work to educate and empower nurses to positively contribute to better use of IT in their settings.
For those working in settings where they have an IT-engaged nursing community and good access to nurse scientists who are interested in innovating and advancing healthcare through the power of informatics, so much becomes possible. Pursuing innovations that make the data work for patients, families, and healthcare providers in a learning health cycle is possible. Think about partnering with patients and families to incorporate data from their personal data and sensors (activity, blood pressure, blood glucose, symptom logs, etc), for more effective decision making, especially in the management of chronic disease. Push for usable dashboards and displays that support and simplify nursing processes.