This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Digital health tools and technologies are transforming health care and making significant impacts on how health and care information are collected, used, and shared to achieve best outcomes. As most of the efforts are still focused on clinical settings, the wealth of health information generated outside of clinical settings is not being fully tapped. This is especially true for children with medical complexity (CMC) and their families, as they frequently spend significant hours providing hands-on medical care within the home setting and coordinating activities among multiple providers and other caregivers. In this paper, a multidisciplinary team of stakeholders discusses the value of health information generated at home, how technology can enhance care coordination, and challenges of technology adoption from a patient-centered perspective. Voice interactive technology has been identified to have the potential to transform care coordination for CMC. This paper shares opinions on the promises, limitations, recommended approaches, and challenges of adopting voice technology in health care, especially for the targeted patient population of CMC.
Immense efforts have been placed on capturing health information electronically, thereby modernizing health communications. The majority of these efforts are provider driven and center around traditional clinical settings. However, a lot of health and care activities happen outside of clinical settings and are not systematically documented and integrated into the clinical systems. Such a practice limits the information captured per patient, which may lead to adverse effects in clinical decision making. This is especially concerning for children with special health care needs (CSHCN). CSHCN is defined by the federal Maternal and Child Health Bureau (MCHB) as children who have or are at an increased risk for chronic physical, developmental, behavioral, or emotional conditions and who also require health and related services of a type or amount beyond that required by children generally [
Children with medical complexity (CMC), a subset of CSHCN, have significant health issues that occur outside of the clinic, and they require complex home care provided by parents and other caregivers in addition to nurses [
A team (coauthors), consisting of caregivers of CMC; a clinician who specializes in treating CMC; a care coordinator who assists CMC and their families; a user experience designer; an application developer; and scientists and researchers who are experienced in clinical informatics, participatory design, and digital health, was formed at Nationwide Children’s Hospital. The team identified gaps in care coordination for CMC and their families, how various technologies can fill these gaps, and how they could be implemented and adopted, all from a patient-centered perspective. The team discussed the value of health information generated at home and the challenges and barriers associated with capturing that information. The team developed recommendations to improve not only record keeping of patient care at home but also communication among patients, caregivers, and care providers through technological solutions. The purpose of this paper was to present our opinions on employing emergent voice interactive technology to capture real-time health information and to enhance care coordination, the associated challenges in adopting this technology, and desired future development.
For CMC, the role of care coordination is highly valuable because the responsibilities (time spent, effort, and financial burden) are higher and navigating services is more difficult [
Helping CMC and families provide the right home-administered treatments at the right time, promptly documenting clinical events (medication, therapy, oxygen treatment, etc), recording symptoms as they happen, and reaching out for timely assistance are critical to promote self-management and coordinated care skills and for successful care coordination [
It is commonly observed in clinical practice that patients and families do not have accurate recall of symptoms, clinical events, or usage of over-the-counter medications [
...I couldn’t remember how many times I have given my daughter Albuterol treatments in the last two weeks…
However, many of these apps fail to promote timely documentation of health-related information because of cumbersome user interfaces, lack of functionality, or not providing evidence-based and personalized content. In many cases, the perceived value does not overcome the burden of using the apps [
Families of CMC could be apprehensive about leaving the clinical setting because of the complexity of their medical care responsibilities. Care coordination services could be supportive and helpful for transitioning after discharge.
...when we were going home (after NICU discharge), I wasn’t ready to go home.
For care coordination services in the home setting, generic emailing and messaging apps and special-purpose nonmedical care apps (reminders for medications, diaper changes, feedings, etc) have been commonly used. However, care coordination apps are limited or nonexistent for complex services, lacking functionalities such as allowing multiple users to communicate and coordinate or providing on-demand coaching of home care skills [
It is common practice to use a verbal or informal note as a
Unfortunately, there is limited literature addressing care coordination technologies and their utilities for CSHCN or CMC and caregivers in home setting. A majority of the studies focus on the use of health communication technologies (Web-based and mobile tools) in self-care, and the results show limited evidence regarding care coordination outcomes [
The most commonly used and accessible health care technologies are mobile apps. In general, long-term sustained usage of health care apps is low [
The app takes too many taps to get to the right screen. I stopped using it after a month...
Many existing apps have the potential to deliver great value to end users but have failed to keep users engaged long enough to reach that potential [
On the provider side, EMR systems have become a common tool with digitized clinical records, and their use has been mandated. However, patient engagement and caregiver engagement depend on the perceived value and whether value outweighs burden of use. Today, many caregivers use paper or other analog, nonunified, unshared, nonsystematic methods to capture medical events and subsequently rely on one-to-one direct communication with providers to coordinate care. An alternate strategy to promote communication would be to make it easier for caregivers to report and capture medical events. Millions of homes have adopted voice interactive devices, such as Amazon Alexa, because they are easy and convenient to use. Voice-enabled technology can be leveraged to more accurately report medication compliance, event documentation, and care coaching. In line with that, our previous study demonstrated that voice interactive technologies are expected to promote adherence in health tracking and increase adoption of communication technologies for care management among caregivers and CSHCN [
To address the previously identified gaps in the current apps and tools landscape, it is essential to engage all stakeholders of CMC using human-centered design principles to create an accessible and interoperable solution. Our multistakeholder team did not focus on finding the
Voice interactive devices and apps are currently embraced and used in daily life by millions of people. The technology is not a passing eccentricity but rather has multiple embodiments from major technology companies, including Amazon Alexa, Google Assistant, Apple’s Siri, Samsung’s Bixby, and other Internet of Things– and mobile-based platforms. These apps have just started gaining attraction in health care [
Voice interaction is particularly successful when the nature of interaction does not require any visual or tactile feedback, thereby removing personal attention to the device. Users simply speak to the app naturally, and information will be captured or recorded. The app can allow caregivers to provide details of symptoms and health events in the most natural and narrative way, enabling hands-free voice interactivity, which might be critical for people who have physical limitations and are not able to type in information. Shifting to an audio diary with voice interaction could increase adherence in keeping a log, specifically when a diary is prescribed to record the frequency of seizures, follow-up with diabetic laboratory tests, or to track general medical symptoms.
The ease of leveraging natural voice documentation needs to be supported with strong natural language processing (NLP) for both voice transcription to text and information extraction from the unstructured text. NLP, together with advanced data science methodologies, has been developed and continues to be improved to take full advantage of the richness of contextual information presented in natural narratives [
The recommended solution framework is illustrated in
An ecosystem of voice interactive care coordination unifying homecare with health care institutions. EMR: electronic medical record; IoT: Internet of Things.
1. Timely capture of complete and accurate health information at home
Voice enabled for natural unstructured, real-time health information capture
Can record audio (such as coughing) and video to communicate with care provider team
User validates and edits text transcribed from voice to address transcription inaccuracy and privacy
Direct free-text documentation is also available to accommodate multiple input modalities
Register care needs using trigger words to notify the care coordination team
2. Facilitate coordination among caregivers at home
Allow multiple users with different level of access
Segment health information to reflect privacy preferences (public, shared, private, etc)
Voice-enabled retrieval of recent care history using predefined trigger keywords
Provide instructions or coaching on relevant home treatment procedures
3. Foster adoption of the app and long-term engagement of users
Integrate with electronic health record to pull clinical information and push home care information back
Enroll patients to use this app and help them to set up linkage at clinical visit
Reminders for medications, next scheduled visits, and updating of symptoms
Customized reports to patients periodically to provide value to them and to keep them engaged
Leverage Health Insurance Portability and Accountability Act–compliant Web servers and services for data storing and analysis
Raw captured data are distilled to represent succinct and relevant historical clinical information
4. Integrate the solution with the health care delivery system for care coordination
Receive feedback from the care coordination center
Adopt fast health care interoperability resources application program interfaces for interoperability
Demonstrate integration with a health care delivery system including care coordination
Who: Parent of a child who has asthma
What: Cannot recall how often he has been given Albuterol in the last 3 months and how many times the child has woken up because of night time coughing
Why: The above missing information is needed to assess asthma severity and recommend the right treatment plan
Solution:
Using the app, the parent documented the child’s asthma symptoms and treatment as they occurred.
The parent clicked the links to review and update Albuterol dosage and treatment times. Also, increased coughing events at night were noted.
A week later at the doctor’s appointment, the parent filled out questionnaire on symptoms referencing the records in the app.
Doctor: “Do you have any concerns over the last 3 months?”
The parent pulled out the app to review the list of concerns and the relevant symptom histories
Who: Parents and a child with multiple health problems including cerebral palsy, epilepsy, tracheostomy, and gastrostomy who uses a wheelchair
What: Have trouble coordinating complex care at home (tracheostomy tube changes and medication administration, etc)
Why: Not knowing whether the previous caregiver has given antiepileptic, at what time and dosage, if it could be dangerous and negatively impact health outcome
Solution:
Mom: “Alexa, Depakote 5mL given to Ben” (timestamp captured and recorded)
Amazon Alexa: “Depakote 5mL given to Ben. Got it”
Mom “Alexa. Ben Trach changed”
Amazon Alexa: “Trach changed for Ben. Got it”
Dad: “Alexa. When was the Trach changed last for Ben?”
Amazon Alexa: “Trach was last changed at 3:05 pm today for Ben.”
Who: A diabetic teenager who needs daily insulin shots and caring parents
What: The teenager gets agitated when parents check in daily to make sure medications are taken
Why: The teenager perceives parents’ medication monitoring and reinforcements as nagging
Solution:
Teenager: “Hey Siri. Insulin given” (timestamp captured and recorded)
Apple watch: “Got it. Insulin given at <current time>”.
Parents are also authorized to see the records and would need to “nag” the teenager a lot less, resulting in less stress and better teenager-parent relationship.
The teenager can also view her compliance of treatments and glucose levels over time. She begins to take responsibility for monitoring her own health but continues to have oversight by parents.
Outcome metrics for voice interaction could be slightly different from other technologies because of the nature of information processing and technology interaction. Therefore, it is important to consider the differences and adjust the metrics. In voice interaction, outcome metrics could be collected and assessed in 2 categories: technical and engagement. To validate the technical performance of the technology, accuracy testing of artificial intelligence (AI) and NLP methods would be employed. Annotated notes, number of user-validated transcriptions and notes, and number of retaken or corrected notes could be used to test the performance of AI. Precision and recall rates could be used to assess the accuracy of NLP methods in predicting and providing note highlights.
As shown in
Spectrum of applications of voice assistants.
Within this context, voice assistants in care coordination, as the envisioned solution in this paper, may have implemental, ethical, regulatory, and technical limitations. Although significant progress has been made in terms of compliance of services (eg, Health Insurance Portability and Accountability Act [HIPAA] compliance of Amazon Alexa), consumer-facing voice interactive device apps currently have limited abilities to be used in health care. Some of the emerging limitations for leveraging voice are as follows:
Mainstream vendors are not providing full access and control of the voice input (eg, transcript and raw data) to developers and researchers with user consent, which can be used for improving health services.
Not all health care services are HIPAA compliant and have limited security and privacy protocols related to audio-formatted health data transmission, processing, and storage.
There is relatively lower demand in the market compared with other communication technologies (eg, mobile apps).
Access to voice-enabled devices is affected by the social economic status and may create inequality in access to the solution (eg, requirement for compatible device and data plan).
There is a major progress in voice recognition in the English language but limited efforts on foreign-accent recognition and lack of availability and analytical capability in a large selection of other languages.
New methods are needed for designing voice services in health care. Translating mobile or Web services to voice may be limited in terms of functionality and navigation.
Conversely, integration of unstructured patient-reported data with the health care system could create a systematic burden and may be hard to control and use in decision making [
Considering the increasing investments in health care and voice technologies and the current trajectory of voice interactive device adoption [
In this paper, we have shared the challenges and recommendations regarding the use of technology to promote coordination of the care of CMC in a home setting. We argue that the use of voice interactive technologies in the home setting could enhance communication of health events and improve coordination. Although the current literature is limited in relation to voice assistant use in care, our report contributes to the literature suggesting potential health informatics solutions, which address information needs for coordination [
Voice interactive assistants in the market.
artificial intelligence
children with special health care needs
children with medical complexity
electronic medical record
Health Insurance Portability and Accountability Act
Maternal and Child Health Bureau
natural language processing
Care Coordination for CSHCN Grand Challenge, sponsored by MCHB of the Health Resources and Services Administration, motivated the formulation of this multistakeholder team and pursuit of this proposed solution. Our team has been awarded as one of the phase I and II winners. The authors would also like to acknowledge Carrie Robinson for constructively reading the manuscript.
All authors contributed to the ideation, solution development, and the writing of this paper. ES led the manuscript preparation, writing, and coordinating the authors. YH supervised and led the solution development. The authors contributed with their viewpoints as, physician (GN), nurse scientist (VS), caregiver (AE), care coordinator (KC), digital health scientist and clinical informaticians (ES, YH, SL, SR, and AC), developer (MB), and designer (RS).
None declared.