Type 1 diabetes is a disease in which the frequent use of data, by patients and care providers alike, is critical to care. Yet current software tools for this purpose are highly inadequate.
Type 1 diabetes is a complex autoimmune disease impacting up to 3 million Americans and an estimated 30 million patients worldwide. Patients must continuously monitor and control blood glucose levels; essentially reproducing the complicated functions of a no longer functional biological pancreas. Successful management is a demanding, 24-7-365 commitment—so much so that learning sufficient managment strategies has been compared to learning to “fly an aircraft.”
With the advent of intensive care guidelines supporting frequent data use by patients and care providers in the 1980s and 1990s, it was believed that the combination of simple rules for data-driven care decisions and new insulin delivery devices would result in successful management in the majority of patients. As of the early 2000s, it has become clear that this is not the case. Over two-thirds of patients in top clinics with access to sophisticated technologies and knowledgeable providers are wildly uncontrolled, leading to devastating health effects.’
Initially, patient noncompliance was held to be the cause for this lack of successful outcomes. However, today, the care community comprised of highly motivated patients, providers, diabetes educators, family members, and other specialists have realized that their needs are not well understood. There is now strong demand for greater insight to be extracted from more holistic data sources. The status quo of software, sensor and device design at present gives rise to subpar diabetes management performance, avoidable life-threatening adverse events like hypoglycemia or diabetic ketoacidosis, and increased risk of future cardiovascular outcomes in the majority of type 1 diabetic patients. Moreover, it leads to severe reductions in quality of life via imposed limits to lifestyle freedom, as well as emotional distress caused by weakened perceptions of self-efficacy.
There is widespread recognition, within the type 1 diabetes community, of a need for a variety of software products to support type 1 diabetics by enabling meaningful use of data.
There is a call for the development of data software solutions to meaningfully provide better type 1 diabetes care. The major challenge, in our view, is not the recognition of this need, but enabling innovators to develop the understanding of diabetes patient end-user needs. In addition innovator must be cognizant of design and engineering possibilities. For instance, improved human factors considerations and understanding of the variations in type 1 diabetes manifestation and care strategies in the diverse patient population it affects would enable software innovators to develop design possibilities that can better leverage device and user-generated data. Accordingly, many in the community are spearheading efforts to develop improved software solutions to do so. Yet, while this idea has been in play for years and existing technology and knowledge dispersed within the community of would-be innovators would seem ripe for the development of meaningful products, the current trends in Type 1 diabetes management software develop are on suboptimal trajectories with respect to timing, meaningful impact on care outcomes and/or scalability. These include, for instance:
Efforts to develop data software by small independent developers or teams, often of patients or close participants in care (e.g., relatives) who are focused on developing, first and foremost, apps to support basic feedback for optimizing one’s care regimen, or to make more targeted goals, such as adjusting insulin administration strategies to allow for greater freedom in activity or food intake, or understand other detailed interactions between non-basic variables and blood glucose levels to fine tune their care.
R&D groups developing improved patient simulation models to better understand nuances and variability in type 1 diabetes management in different patients, for instance, to make automated or “closed-loop” artificial pancreas systems work optimally in as large a patient population as possible.
Physicians’ practices developing solutions to better capture, integrate, and share data and annotations for enhanced troubleshooting and doctor-patient collaboration in the limited timeframe of quarterly office visits and limited phone and email consults.
Researchers looking to aggregate information including both device and “human” data from larger populations of patients for data mining and inference making between and within patient groups.
Why, with so many brilliant teams and organizations working to develop diabetes software, is there still great need for innovations in this space?
We believe that the root cause for a lack of effective software for type 1 diabetes care is the following: Recent data indicates that type 1 diabetes management is a more complex context than previously understood. There is an urgent need to make better use of data as well as human intelligence to improve care outcomes.
While there are many brilliant teams and organizations working to develop solutions, it is difficult, if not impossible, for many to possess holistic understanding of user requirements and design possibilities that will lead to meaningful improvement. As a result, we are seeing minimal or niche viability products emerge at present, but a lack of widely scalable, high impact solutions.
Through wide networking and deep conversations with a variety of stakeholders in type 1 diabetes software innovation, and personal involvement in many projects, we have come to an important realization that drives this proposal. Simply put, this state of affairs is the result of a mismatch between the complexity of the situation (specifically, data use in type 1 diabetes care) they are trying to improve, and the level of understanding it is possible for innovating groups to glean separately, using merely their own methods and resources, time, and prior knowledge to develop intelligence to guide their efforts.
The need we have identified is for a site for combining, refining, and clearly communicating high quality intelligence from the dispersed knowledge of these diverse innovators. It is our firm belief that this will facilitate and accelerate the innovation of effective, scalable diabetes data software.
A web-based portal for trans-stakeholder “sensemaking” to generate sufficient shared intelligence to enable the innovation of successful data software products to improve type 1 diabetes care. We would like to support diabetes data software innovators in their projects with a website built to reduce the costs of “sensemaking” of the complex challenge of developing effective software for type 1 diabetes. Currently, there is a wealth of information and understanding dispersed in the community of innovators in this space. Our interface is designed to decrease the immense time and attention that innovators need to devote to understanding this context.
Proposed Portal Content and User Interface
To help diabetes software innovators develop effective products, the portal integrates information from diverse sources to generate applicable intelligence of five key areas in which we have observed knowledge gaps. In addition, users of the portal will have a personal “shoebox” page in which they may save, annotate, and create collections of information sources relevant to their own software innovation projects. The portal will be structured as follows:
1. User needs in diabetes data software. The specific needs of type 1 diabetic patients and their care teams within the general goal of using data to make improved care decisions. In this section, First, we will showcase the challenges faced by the diverse target users of the software products in an information map designed as a 2D matrix. Over time, will collect robust information to clarify the number of users as well as the experiences they desire and a collection of data about the users that may be used to identify or segment target groups, and incorporate “scale” or “number of prospective patients affected” into the matrix visualization. When a portal user clicks through to view more in depth information about any of these needs, he or she will find a page of curated, crosschecked, and pithy pieces of information excerpted from diverse sources along with succinct summaries of insights derived from these. Diverse information items will be navigable by insights, source type, format, or other metadata, based on the preferences of users of the web portal. Moreover, they will be displayed in the most distilled form for rapid scanning by portal users of only the most relevant and meaningful content, along with links to the full original content and citations. The types of structure and unstructured data sources displayed will include:
2. Management models and constructs. The concepts and schemas related to the software needs. This area of our portal will be centered around a clickable information mapping of various “constructs” useful for troubleshooting type 1 diabetes care using data. These will be featured on an entry page in intuitive categorical groupings. When a user clicks through a title and brief description on the map page, he or she will be directed to a page dedicated to that construct. Each construct-dedicated page will feature the following:
3. Guidance and tools for designing data systems. This portion of the site will contain information necessary for dealing with data and contextual information when building a software product, including tools available to collect this data, and in development:
Secondly, users will find in this section instructions for integrating data from diverse collection tools as inputs into their products.
Thirdly, this section will offer access to guidance about analytic approaches to processing the information.
4. Design of reports and visualizations. This section addresses the many questions about how innovators can output information within a specific softare product in ways that effectively augment human managment decisions. Designing the output from data systems to support human decision making, especially in a cognitively taxing situation such as type 1 diabetes management, is no easy feat. That said, there is considerable work that has been put forth, publicly, by a variety of innovators in the type 1 diabetes software space. We have collected numerous examples, begun to classify them according to various “types” of visualizations and conventions that have emerged, as well as other criteria such as the tools with which they were made, and their availability for open use in projects (e.g. links to code repositories). In this section, software innovators will be able to browse visualizations by relevant search terms, filters, and tags. We want to enable innovators to navigate through examples of visualizations based on their visualization skill level, user needs and data models relevant to their projects, users access to tools and funding, and other considerations, to make it easy for them to find what they are looking for.
5. User “Shoebox” Pages. Individuals or teams using the site will have their own space to save relevant information items for their projects. On this page, users will be able to create collections of items they find relevant, and add their own annotations and inferences. Ultimately, these shoeboxes will not only help the individual innovators using our site to conduct sensemaking relevant to their own projects, but it will allow us as the creators of the site to collect a richer base of content to filter and draw in to the open sections of the site, as well as site analytics to better understand what content users are finding to be valuable, and what meaning they derive from them. As a secondary phase, we’d like to allow users to upload content they perceive as relevant to their own Shoebox pages, which can then be triangulated, annotated, referenced and, where helpful, brought in to the section of the site open to all users with relevant placement and formatting.
Improvements over Existing Interfaces for Trans-stakeholder Sensemaking
It would be one thing to suggest a website that allows sharing of information by diverse stakeholders, in the style of today’s Web 2.0 sites on type 1 diabetes and other diseases. (We note that wonderful examples in that category exist for type 1 diabetes.) It is quite another to propose a web portal for trans-stakeholder sensemaking that truly allows visitors to come up with the requisite “sense” of how to make their innovations successful in such a complex arena as type 1 diabetes management. Intelligent portals for any subject area provide (a) subject matter insight, (b) user tools to help with sensemaking tasks, (c) machine learning cognitive support, and (d) the benefits of collaboration. They reduce the time and attention demands on users for developing insight into complex innovation contexts by pre-vetting, curating, and organizing information according to site user needs. Looking underneath the hood of our portal quickly reveals some of the major complementary cognitive support advances: machine learning recommendation tools and crowd-sourcing for intuition inspiring content.
Existing Resources and Feedback for Future Improvements
The information displayed within each section, and its architecture, will be built on the basis of an already assembled collection of knowledge of existing projects and identified needs based on our interactions in the community as well as our primary research. In addition, a survey designed by our team members and already in field will be used to obtain further information about the needs of future software users to display on the site.
With anticipated grant funding, we propose to use two types of feedback loops—human and machine—in order to continue to improve the usefulness of the site by understanding the needs of site users, their reactions to design prototypes, as site analytics to see how they actually interact with our site during the process of sensemaking for their own projects. In particular, we propose:
1. In-depth interviews and co-design sessions with diverse innovators working on diabetes data software.
In these sessions, we will engage a wide network of innovators whom we have pre-recruited or with whom we have pre-established relationships.
We will conduct user interviews with participants who test the site in beta versions to understand their perceptions of usability on various dimensions and in different scenarios. In this way, we will seek to understand site performance with regard to the several subtasks of intelligence sensemaking, as well as help to better define these tasks, which are often relatively vaguely described in the sensemaking literature.
Additionally, we will conduct co-design sessions with lead users to understand ways to improve the site with regard to content, organization, and functionality.
2. Machine learning algorithm feedback developed on the back end and Netflix-like recommend to users of the site on the front-end, such as metrics related to data source relevance, information usage, suggested information foraging paths, and other design to facilitate learning.
Web portals for collaborative intelligence in adjacent domains to leverage the power of site analytics and machine learning algorithm to generate feedback. For instance, offering statistics regarding the usage, perceived relevance, and linkages of particular sources. We are aware of products for sensemaking in large scale intelligence domains such as national security and government policy. Typically, these products also collect analytics on entire documents or media files as well, rather than on relevant, pre-filtered excerpts and triangulation clusters, as we are proposing here.
We believe the human-mind should never be divorced from intuition building, but we develop cognitive support to make digestible, torrent and data dumps of unwieldy data, that can provide gems of useful patient information such as blogs and recordings. Our platform design features (i) both curated and uncrated ingestion of relevant medical information from the web- information from the fat-tail distribution of patient experiences; (ii) suggested exploration of content using underlying sematic network structures facilitated by machine learning tools.
VALUE PROPOSITION AND USER SCENARIOS
Use Case Scenario 1 :: Physician Software Developer MaverickSensemaking Value Proposition:
Scenario: Endocrinologists are principally responsible for administering type 1 diabetic care and have large communities’ bases of patients. A common issue faced by practices is the lack of adequate access to holistic patient data, during interim periods between quarterly visits, in an easily accessible format. Endocrinology practices are actively seeking to develop or sponsor software which would maximize provider time & inform clinical decision making. Useful device feature could include: the ability to quickly and easily pull in data from multiple patient devices; capacity to utilize applications which would allow patients to capture relevant contextual information such as food intake, activity, emotional status, menstruation, and so forth, as applicable to individual cases; and integrate relevant pattern recognition and micro-narratives as a way to better facilitate the human intelligence faculty of pattern recognition.
Knowledge Gaps: To develop this software, physician practices groups require access to a variety of “pieces of knowledge” missing from current awareness. These can include: human factors considerations (what is useful or desirable for patient annotation); APIs and protocols for pulling data from devices into open source (non-device-maker) software; data design conventions and considerations as well as opinions for improvements; and access to example datasets to understand how to “handle and process” streams generated by patients.
Use Case Scenario #2 :: iPhone App Developer w/ Behavioral Insight and Data SciencesSensemaking Value Proposition:
Knowledge Gaps: The father is knowledgeable in programming for mobile platforms and in the design of phone user interfaces. However, understanding how to handle clinical data—glucose, insulin, hBA1C – or user-derived behavioral data is difficult. Packaging data stream information such that it is clinically meaningful useful inform health decision require another layer of complexity. To develop this software application concept, the father needs to understand the following: • Data extraction guidance document and API for type 1 diabetes devices. • Under device peripherals and wireless transmitter options to extract in real time • Understand obfuscated file structures based and file processing • Understand statistical visualization and modeling lagged processes to derive meaning from blood glucose meters and insulin administration pumps
Use Case Scenario 3 :: Sensors, Peripheral and Fitness BioMed EngineerSensemaking Value Proposition:
Knowledge Gaps: The goal of the biomed engineer is to simplify the complicated and tedious ordeal of logging physical activity and glucose intake to regulate blood glucose. The current process is inefficient and burdensome. To design an integrated solution, the biomed engineering will require: • Relationships between information which can be “sensed” (e.g. physical activity wattage, blood glucose, some composite measure of “intensity,” and adequate capture of patient variability) • APIs & code linking protocols for sensors of external devices, along with diabetes devices • Conceptual and causal relationship of confounding variables that need to be considered & modeled • Usability considerations and suggestion for his wife and other diabetic athletes. • Study designs strategies to test his approach Use Case Scenario #4 :: Medical Researcher Interested in Industry PartnershipsSensemaking Value Proposition:
Scenario: A behavioral research scientists at a top tier university has been working to evaluate a novel visual “life-logging” camera device and algorithm that visually matches pictures of food with an online catalogue of foods that can estimate nutritional value. The researchers are not specifically involved in type 1 diabetes research, but think their product may provide value to patients. The combination of devices and algorithm shows high validity in accurately capturing caloric carb ingestion based randomized clinical studies in non-diabetic patients. The research team thinks that have a solid product-software combination. They do have some concerns that some users complained about comfort and that their devices has not been piloted in a type 1 diabetic population.
Knowledge Gaps: The academic researcher is skilled in establishing statistical validity, but lack training or skill sets that would let them evaluate the human factor components of the system to ensure reliable usability. The researchers require insight and feedback from human factors specialist and user interface designers to complete the following aspects of bring a product to market: • Conduct front-end analysis to understand major functions of the product, the environment where it is used, and the preferences, requirements, and expectations of the user. • Task analysis of the camera system to understand any user complications • Collaboration with type 1 diabetes specialist who may understand how to best integrate food data with alternate biometrics sources for meaningful glucose control. • Integrate camera sensor data into a mobile smart device platform with interoperability with iOS, android, and Facebook • Conduct a viability study in a small population of type 1 diabetes patients.
Use Case Scenario 5 :: Community, Patient Advocacy & Innovation Information SeekersSensemaking Value Proposition:
Scenario: A medical technology analyst for CNET wants to complete a news feature on the state of sensor innovation in the type 1 diabetes community. The reporter is not familiar with the research space but is eager to reach out. He learns that many key researches, patient groups, and start-ups frequent a type 1 diabetes sensemaking portal for collaboration and design initiatives.
Knowledge Gaps: The reporter is an intelligent, resourceful individual that is used to learning new subjects and writing about them as part of his work profile. His needs are to quickly forage for important information and make sense of active areas of research that show promise. He needs to understand: • The extent of groups working within a design space for specific niche devices and applications • He needs to suggestion on research domains that semantic linkage as he is learning • He needs to understand where and what information resources are being frequently accessed by top developers. • He needs overviews of the conceptual frameworks, human factors, clinical and statistics language utilized by the active type 1 diabetes community.
A VALUE PROPOSITION AT MULTIPLE LEVELS
At present, despite advances in technology, two-thirds of type 1 diabetic patients are controlled well below care guidelines even with sophisticated technology and access to care at top clinics.
A major short term innovation on the horizon is the Artificial Pancreas. A dual-hormone version of this device (essentially an insulin pump, sensor, and algorithmic model for automated insulin delivery to control blood glucose ranges) is projected to reach the U.S. market in 2017. We support this innovation but believe there is still huge opportunity to improve care through sophisticated software and systems for data use in the short term, and even beyond 2017. This is because, in diabetes care, time is of the essence. Every year prior to the artificial pancreas is a year suffering the consequences of a devastating disease. Moreover, researchers acknowledge that the Artificial Pancreas will scale to many patients in its early years, many patients will still require or prefer data software to enhance their ability to make intelligent care decisions and collaborate with their care providers.
Furthermore, we aim to evaluate our model of development to track clinical success of software tools generated from usage of this platform design. It’s not sufficient to churn out more ideas, faster. Rather we want to positively impact the clinical outcomes and quality of lives of type 1 diabetes patients. Finally, we recognize a unique opportunity to develop, in the context of type 1 diabetes, which is a leading edge healthcare space for data use (as it is critical for survival and long-term health in this disease), a model for a trans-stakeholder portal that enables collaborative intelligence pooling and understanding to guide innovations that will allow for developing future solutions in other diseases as sensors and data models advance, and the cost and usability of related tools improves.
SUMMARY >Describe your project in one sentence. We are proposing to build a web-based intelligence portal to support the innovation of software products that improve care outcomes in type 1 diabetes by making effective use of diverse data sources; specifically, a site for trans-stakeholder “sensemaking” among diverse groups of innovators working to improve data use in type 1 diabetes care by developing scalable, high impact software products.
Who is the audience for this project? How does it meet their needs? Our audience includes a wide variety of innovators, represented by different disciplines and stakeholders in the type 1 diabetes space, with diverse needs when it comes to using data to improve outcomes in type 1 diabetes. Our audience consists of participants in software innovation efforts ranging from small patient-led projects, to efforts of physicians to enhance information sharing with their patients; device manufacturers with rigid development cycles in need of greater user and design insight to make meaningful improvements to their proprietary data software, and researchers working to better understand patient scenarios for simulation models.
What does success look like? Our web-based portal aims to maximize the potential for effective innovation of software products to make better use of data in type 1 diabetes, and vastly care improve outcomes in the patient population. Specifically, we aim to support both short term innovations (such as those facilitating management with existing technology) and long term innovations (such as maximizing the scalability of the artificial pancreas) in type 1 diabetes, AS WELL AS develop a model for similar web portals which can be used to support collaborative innovation in the use of data in other diseases.
TEAM AND CORE COLLABORATORS
Elsa Kaminsky Elsa is a type 1 diabetic, former consumer researcher for Colgate-Palmolive and Fortune 500 clients, and an MFA candidate in Transdisciplinary Design at Parsons The New School for Design. She now focuses on developing information architecture, multi-methodology research protocols, and information interfaces for collaborative sensemaking in large scale intelligence contexts.
Jorge Luna Jorge is an MPH, Epidemiology PhD and Operations Research MS candidate at Columbia University, with broad ranging experience with the use of data modeling and simulations for healthcare-related applications.
Patrick Schlafer Patrick is a designer and strategist with a broad skillset. Adept at solving complex situations with advanced web technologies, and demonstrated information visualization talent, he has the ability to intuitively convey complex information at a glance, as well as build user-friendly websites for both visitors and administrators, organize intricate systems of information into easily-understandable taxonomies, and think beyond current definitions of a problem. Some clients include: Inc Design, Country Music Association, United Methodist Communications. Patrick holds an MFA in Transdisciplinary Design from Parsons The New School.
Leon Gold Leon Gold, PhD, is a Human Factors consultant who has taught Human Factors and Ergonomics at Columbia University’s, Department of Industrial Engineering and Operations for over 2O years. Previously he was Director of Usability at the New York Stock Exchange and has worked for numerous clients including Bell Laboratories, AT&T, IBM, and the American Stock Exchange. Leon’s interests include the design and evaluation of user interfaces, consumer products, and human/system interactions along with the development of personal selection and evaluation methodologies.
Sara Krugman Sara Krugman’s practice as a designer focuses on the importance of the relationships people have with health care data and devices. She has a passion for creating personalized and simple health care solutions, from mobile apps to human services. Studied at The Copenhagen Institute for Interaction Design. She works through her own company, Line (linehq.com) with various partners to create apps and devices for T1diabetes.
Evan Grauer Database Director at Harrison Scott Publications, a leading intelligence provider for the structured finance, commercial real estate and hedge fund markets. Evan has led a team tasked with designing databases and data retrieval systems needed to illuminate the darkest corners of multi-billion dollar markets. His team’s research has been cited in dozens of national and trade publications, including the New York Times, Wall Street Journal and Bloomberg News. Evan holds a Bachelor of Arts in Computer Science from Dartmouth College.
For exhibition and sponsorship opportunities, contact Susan Stewart at email@example.com
For information on trade opportunities with O'Reilly conferences email mediapartners
For media-related inquiries, contact Maureen Jennings at firstname.lastname@example.org
View a complete list of Strata + Hadoop World 2013 contacts