=Paper=
{{Paper
|id=Vol-1388/latebreaking_paper8
|storemode=property
|title=Adaptive Recommendations for Patients with Diabetes
|pdfUrl=https://ceur-ws.org/Vol-1388/latebreaking_paper8.pdf
|volume=Vol-1388
|dblpUrl=https://dblp.org/rec/conf/um/WeibelzahlHHMS15
}}
==Adaptive Recommendations for Patients with Diabetes==
https://ceur-ws.org/Vol-1388/latebreaking_paper8.pdf
Adaptive Recommendations for Patients with
Diabetes
Stephan Weibelzahl1 , Dominikus Heckmann2 , Eelco Herder3 , Karsten
Müssig4,5,6 , Janko Schildt7
1
PFH Private University of Applied Sciences Göttingen, Germany
2
Ostbayerische Technische Hochschule Amberg-Weiden, Germany
3
L3S Research Center, Leibniz University Hannover, Germany
4
Institute for Clinical Diabetology, German Diabetes Center at Heinrich Heine
University, Leibniz Center for Diabetes Research, Düsseldorf, Germany
5
German Center for Diabetes Research, Partner Düsseldorf, Germany
6
Department of Endocrinology and Diabetology, Medical Faculty, Heinrich Heine
University, Düsseldorf, Germany
7
Emperra E-Health Technologies GmbH, Germany
weibelzahl@pfh.de, d.heckmann@oth-aw.de, herder@l3s.de,
Karsten.Muessig@DDZ.uni-duesseldorf.de, j.schildt@emperra.com
Abstract. Diabetes mellitus is a major epidemic with about 8.3% of the
world population being affected. Proper treatment minimizes the risk of
secondary diseases. The GlycoRec system aims to support patients in
making decision that are related to the treatment by modeling their
behavior and their physiology. Here we describe the aims and first steps
towards the development of GlycoRec.
1 Diabetes Mellitus
Diabetes mellitus is a major epidemic and a threat to public health with about
8.3% of the world adult population being affected (Shi & Hu, 2014). The high-
est increase in diagnoses in recent years is observed in patients aged 60 and
over. While there is no known cure for diabetes, it can be managed through
a combination of diet, exercise and appropriate medication. However, when not
managed in an appropriate way, patients are at high risk of developing secondary
conditions comprising in particular as cardiovascular disease resulting in a sig-
nificantly increased morbidity and mortality. Therefore, it is of high importance
that patients are able to manage their diabetes treatment on their own aiming
at near normal blood glucose levels (American Diabetes Association, 2014).
Most patients who treat their diabetes with insulin go through the same rou-
tine several times each day: they monitor their current blood glucose level using a
glucometer; they estimate their carbohydrate intake; they calculate the required
insulin doses and inject an appropriate amount. Different types of insulin that
vary in onset and duration of action may be used.
One of the challenges in diabetes management is the patients’ need to learn
how their body reacts to food intake, activity and insulin application. Mobile
apps currently available for calculating insulin dose without reference to indi-
vidual needs show systemic issues such as missing validity checks of input data
affecting the safety of patients (Huckvale, Adomaviciute, Prieto, Leow, & Car,
2015).
2 Aims and Objectives
The GlycoRec system aims to support diabetes patients in managing their dis-
ease. It supports decisions and gives individualized recommendations based on
the patient’s behavior, physiology and treatment history. Individualized advise
may include
– estimation of nutritional characteristics such as carbohydrate content and
glycemic index of meals
– recommendations on insulin application based on glucose level, activity and
food intake
– warnings if blood glucose levels are at risk of leaving the target range
3 Requirements Engineering
Complex adaptive interactive systems such as GlycoRec require systematic elic-
itation and documentation of requirements (Gena & Weibelzahl, 2007).
3.1 Requirements Elicitation
Based on an extensive review of the literature, we designed a survey for patients
to explore both the patients’ situation as well as the main barriers they en-
counter. Questions on the patients’ current situation referred to their strategies
for managing their disease as well as the technologies available to them. The ex-
ploration of barriers encountered will help to tailor functionality to patients and
prioritize features. Moreover, semi-structured interviews with diabetes nurses
will be conducted in order to validate the survey results and to elicit expert
knowledge on diabetes management strategies (Dix, Finlay, Abowd, & Beale,
1998; Weibelzahl, Jedlitschka, & Ayari, 2006).
3.2 Preliminary Personas
In order to support the modeling process, we developed a set of personas (Cooper,
1999) that represent the main target groups of the system in regard to their needs
and preferences. Figure 1 shows a condensed version of two of the personas de-
veloped based on the survey data.
Andreas Beate
– male, age 59 – female, age 30
– accountant – shop assistant
– type 2 diabetes – type 1 diabetes
– diagnosed 12 months ago – diagnosed at age 6
– owns smart phone – uses smart phone and tablet
– likes to prepare his own PC on regular basis
meals – long-standing experience
– has lunch in company’s cafe- with diabetes management
teria – focus on healthy life-style,
– feels insecure when taking exercises three times a week
treatment decisions – wants flexibility, e.g., eat out,
clubbing
Fig. 1. Condensed version of two of the GlycoRec personas
4 System Architecture
The GlycoRec architecture follows the high level pattern of interactive adaptive
systems in accordance with Jameson (2008) comprising inference, modeling and
adaptation decision. Figure 2 depicts an outline of the high-level system archi-
tecture. A variety of sensor data are collected including actual glucose level as
measured by a glucometer, level of activity and insulin application. Data are
gathered through smart phone, smart watch and networked glucometer and in-
sulin pen, stored in a central database and analyzed in order to model current
glucose level.
Patients interact with the system via smart TV, tablet or smart watch. While
the smart watch interface is designed for interaction during the day where both
the patient and the system can initiate interaction, the smart TV interface sup-
ports review and reflection on historical data and facilitates identification of
patterns over time. Patients can also share their records with their physician or
their diabetes nurse for discussion of their diabetes management.
5 User Modeling and Adaptation
From a user modeling perspective, GlycoRec tackles a number of challenges,
including but not limited to:
Firstly, physical reactions to insulin, food intake and activity in diabetes are
idiosyncratic. While the general patterns are known, individual patients seem
to respond differently in similar situations, depending on factors such as age,
Fig. 2. Overview of the architecture of the GlycoRec system
weight, general heath, medication, comorbidities, to name but a few. Individual
response patterns need to be observed and learned.
Secondly, this will involve combining a variety of sensor data. We have se-
lected a number of candidates, but it will be necessary to narrow down the list
for both modeling and practical reasons.
Thirdly, the available data vary greatly in granularity and quality. While for
instance activity level can be assessed on a continuous basis, most patients mea-
sure their glucose level three to seven times a day, with some patients measuring
only once a day. So while validation and readjustment of glucose level measures
are sparse, the (predicted) glucose level need to be assessed at any time in or-
der to be able to issue warnings. Accordingly, models will differ in certainty at
different points in time.
Fourthly, the development process is subject to a number of regulations, as
any device involved in the treatment of patients is considered a medical device
that needs to be compliant with ISO 13485 (International Standards Organiza-
tion, 2003). User testing and iterative development is less flexible under these
conditions.
Lastly, designing the adaptive user experience for patients is challenging as
the disease has huge impact on the patients’ lives anyway. Any additional effort
and new processes in managing their disease will only be accepted if the benefits
are obvious and the required input is minimal, i.e., data collection and modeling
need to happen with minimal or no user interaction in the background, but if
and only if intervention is required the system needs to take initiative and make
reliable recommendations.
6 Future Perspectives
This three-year project commenced in January 2015 and is in its early stages.
Requirements have been gathered. Significant involvement of patients in the de-
velopment process and the application of further user centered design methods
(Norman, 1988) is planned for the next phase. A user evaluation including vali-
dation against physiological parameters of treatment quality such as the HbA1c
value (Larsen, Hørder, & Mogensen, 1990) will demonstrate the effects of the
system.
Acknowledgment
The GlycoRec project is funded by the Federal Ministry of Education and Re-
search (BMBF) under the funding scheme Adaptive, Learning Systems (Adap-
tive, lernende Systeme).
References
American Diabetes Association. (2014). Standards of medical care in diabetes
2014. Diabetes Care, 37 (suppl 1), S14–S80. doi: 10.2337/dc14-S014
Cooper, A. (1999). The inmates are running the asylum: Why high-tech products
drive us crazy and how to restore the sanity. Indianapolis, IN: SAMS.
Dix, A., Finlay, J. E., Abowd, G. D., & Beale, R. (1998). Human computer
interaction (2nd ed.). Harlow, UK: Prentice Hall.
Gena, C., & Weibelzahl, S. (2007). Usability engineering for the adaptive web. In
P. Brusilovsky, A. Kobsa, & W. Nejdl (Eds.), The Adaptive Web: Methods
and Strategies of Web Personalization (pp. 720–762). Berlin: Springer.
doi: 10.1007/978-3-540-72079-9 24
Huckvale, K., Adomaviciute, S., Prieto, J. T., Leow, M. K.-S., & Car, J. (2015).
Smartphone apps for calculating insulin dose: a systematic assessment.
BMC Medicine, 13 , 106. doi: 10.1186/s12916-015-0314-7
International Standards Organization. (2003). ISO 13485:2003 – Medical devices
– Quality management systems – Requirements for regulatory purposes
(2nd ed.). Geneva, Switzerland: ISO.
Jameson, A. (2008). Adaptive user interfaces and agents. In A. Sears & J. Jacko
(Eds.), The human-computer interaction handbook: Fundamentals, evolv-
ing technologies and emerging applications (2nd ed., pp. 433–458). Boca
Raton, FL: CRC Press.
Larsen, M. L., Hørder, M., & Mogensen, E. F. (1990). Effect of long-
term monitoring of glycosylated haemoglobin levels in insulin-dependent
diabetes mellitus. N. Engl. J. Med., 323 (15), 1021–1025. doi:
10.1056/NEJM199010113231503
Norman, D. (1988). The design of everyday things. New York: Basic Books.
Shi, Y., & Hu, F. B. (2014). The global implications of diabetes and cancer. The
Lancet, 383 (9933), 1947—1948. doi: 10.1016/S0140-6736(14)60886-2
Weibelzahl, S., Jedlitschka, A., & Ayari, B. (2006). Eliciting requirements for an
adaptive decision support system through structured user interviews. In
Proceedings of the Fifth Workshop on User-Centred Design and Evaluation
of Adaptive Systems, held in conjunction with the 4th International Con-
ference on Adaptive Hypermedia & Adaptive Web-based Systems (AH’06),
Dublin, Ireland, 20 June 2006 (pp. 770–778). Dublin: National College of
Ireland.