FHIR FLI: An Open Source Platform for Storing, Sharing and Analysing

Lifestyle Data

Kiran Gopinathan

, Nikolaos Alexandros Kaloumenos

, Kinnari Ajmera

, Alexandru Matei

Ian Williams

and Andrew Davis

University College London, United Kingdom

Nufﬁeld Health, United Kingdom

Keywords:

Personal Health Record, Lifestyle Data, FHIR.

Abstract:

Consumers and healthcare organisations alike are increasingly interested in using digital health solutions to

reduce the risk of chronic conditions or to help manage these conditions outside hospitals. Equally, there is

a strong public health beneﬁt in helping individuals adopt and improve healthy lifestyle behaviours. The ﬁrst

step in this direction is the ability to record and analyse lifestyle data. Currently, lifestyle logging platforms

use proprietary data formats. Data is segregated among different platforms, impacting consumers, service

providers, research institutes and public health bodies. Our aim is to facilitate the transfer of information

between individuals and organisations that hold or require their lifestyle data. We demonstrate that an open

source platform based on a clinically recognised interoperability standard - Fast Healthcare Interoperability

Resources (FHIR) - can meet both consumers and industry needs. We use as an example the case of people

managing arthritis. Our contributions are: (i) an extension of the FHIR standard for lifestyle data, (ii) a

reference architecture for a Personal Lifestyle Record, (iii) integration with voice-enabled digital assistants

for lifestyle data capture and (iv) an open source implementation of this architecture that retrieves, saves and

analyses lifestyle data from wearable devices.

1 INTRODUCTION

The UK population is living longer, but at the same

time people spend more years in ill-health, managing

chronic conditions (Public Health England, 2015).

To cope with the increasing burden of disease,

more and more resources are dedicated to preven-

tion and management of chronic conditions within the

community. For example, a recent study in the UK

has shown a remission in Type 2 diabetes for par-

ticipants following a low-calorie, diet-based, weight

management programme. Throughout the process,

the trial participants had interaction with their GP and

could access a dietician (Lean et al., 2017)

As ﬁtness, well-being and lifestyle data becomes

more relevant in the provision of care, consumers

and healthcare professionals need the ability to aggre-

gate disparate information about one’s lifestyle into a

consolidated view spanning across the years, across

providers, and covering many data sources. Improved

access to lifestyle data can lead to better quality of

care, better health outcomes and improved customer

satisfaction.

We consider lifestyle data any measurement re-

lated to:

• lifestyle risk factors (e.g. physical activity)

• mental health (e.g quality of sleep, stress levels)

• monitoring of chronic conditions (e.g blood glu-

cose)

• maintaining or improving ability (e.g physiother-

apy, rehabilitation, mental resilience, elite sports).

The current landscape in the industry is very frag-

mented - each organisation that generates lifestyle

data uses proprietary formats. The challenge we aim

to address in this paper is the lack of a true platform

approach, in the well-being and lifestyle domain.

In a platform approach, the analysis of data (for

day-to-day monitoring, service personalisation or be-

haviour change) is separated from data collection. For

this to happen, there is a need for an encompassing

and open standard to describe lifestyle data in a con-

sistent manner, agnostic of the source systems gen-

erating the data. This will open the way for services

speciﬁcally designed to deliver value by interpreting

Gopinathan, K., Kaloumenos, N., Ajmera, K., Matei, A., Williams, I. and Davis, A.

FHIR FLI: An Open Source Platform for Storing, Sharing and Analysing Lifestyle Data.

DOI: 10.5220/0006791302270233

In Proceedings of the 4th International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2018), pages 227-233

ISBN: 978-989-758-299-8

227

the data in new, useful ways. This is the path fol-

lowed in the medical community, where much effort

has been put into deﬁning interoperability standards

to describe clinical data.

In this paper, we propose an architecture for a Per-

sonal Lifestyle Record that offers capabilities to re-

trieve, store and analyse lifestyle data in a provider-

agnostic environment, illustrating the potential of a

platform approach. To inform our proposal, we use as

an illustrative case study the management of arthritis.

The paper is organised as follows. Section 2

presents the current technology landscape for man-

aging and using lifestyle data. Section 3 discusses

an illustrative customer journey, in arthritis manage-

ment. Section 4 proposes a method for standardis-

ation of lifestyle data and offers two speciﬁc exam-

ples. Section 5 presents a reference architecture that

decouples data generation from data analysis. Sec-

tion 6 discusses emerging user interaction channels,

such as voice-enabled digital assistants for capturing

lifestyle data. Section 7 describes the implementation

work done to date and Section 8 summarises our con-

tributions and discusses future work.

2 BACKGROUND

Smart-phones have become a leading consumer

choice for managing lifestyle data. They are a nat-

ural hub to bring together data coming from built-in

sensors on the phone, data coming from wearable de-

vices that synchronise with the phone and data com-

ing from mobile apps installed on the phone. There

are a number of limitations when using a smart-phone

based lifestyle record:

• it is not an inclusive approach: Apple Healthkit

data can only be accessed from an iPhone.

• it does not cover the needs of service providers,

who need to access and aggregate data from a

population of clients

Solutions created by wearable companies have the

advantage of being accessible across platforms, mo-

bile and web. They focus on analysing and providing

insight on data collected through the wearable. They

have two main limitations:

• the data model of each platforms is controlled

by the vendor and it only contains items relevant

within the context of use for each wearable - third

parties cannot extend the data model.

• the data is still locked, as Fitbit users can only

compare themselves against other Fitbit users for

example.

Cloud-based platforms like Open MHealth or Mi-

crosoft Health Vault have cross-platform accessibility

and allow developers to extend the data model. Infor-

mation can be shared with service providers. How-

ever, these platforms also have some limitations.

• at the moment, they still use proprietary data for-

mats

• third-party developers cannot develop custom vi-

sualisations or predictive models

There are two conclusions we can draw from re-

viewing the current landscape. First, there is a clear

direction towards enabling data movement. Many

wearable companies allow customers to download

their own data for example. Conversely, other compa-

nies position themselves as data repositories, offering

the necessary integrations for consumers to bring data

into their platforms.

Second, there is no open, standard information

model for lifestyle data, with clear governance sup-

porting it, which is not tied to a speciﬁc commercial

organisation.

We argue that such an information model could

well be developed following an open and inclusive

approach, as opposed to being developed internally

by any one company. There is a large body of work

which could inform these efforts, stemming from the

clinical community pursuit to standardise the way

medical data is represented.

FHIR (Bender and Sartipi, 2013) is such a stan-

dard, supported by HL7 and adopted by a grow-

ing number of healthcare organisations, including the

NHS (Kavanagh, 2017).

The FHIR speciﬁcation deﬁnes a set of core re-

sources (Person, Observation, Diagnostic etc). and

an infrastructure for handling resources. FHIR Ob-

servations are already used to describe:

• Vital signs: e.g. temperature

• Laboratory Data: e.g. blood glucose

• Devices Measurements: e.g. pulse oximetry

• Clinical assessments: e.g. Glasgow Coma Score

• Personal characteristics: e.g. eye colour

The FHIR standard can be extended to suit par-

ticular contexts of use - including management of

lifestyle data. Extending FHIR to suit particular sce-

narios of use is achieved by creating FHIR Proﬁles. A

proﬁle extends or restricts core FHIR resources, based

on the domain-speciﬁc needs.

A FHIR extension for lifestyle data is being devel-

oped as part of the Finnish national Personal Health

Record infrastructure (Kanta, 2017). However, their

approach is to create a limited number of FHIR ob-

servations, describing ﬁtness information aggregated

ICT4AWE 2018 - 4th International Conference on Information and Communication Technologies for Ageing Well and e-Health

228

over a period of time - for example ’Distance reached

in 24h’. We favour an approach where any data point

created at the source system is mapped to a FHIR ob-

servation. In our approach, three distinct walking /

running episodes in a 24 hour window would be rep-

resented as three FHIR observations, not a single one.

3 ILLUSTRATIVE CASE STUDY

Lifestyle data can be used to improve the health and

well-being of people of all ages.

To discuss our proposal for managing lifestyle

data, we consider the case of arthritis, a group of

chronic conditions that affect the joints in the body.

As of 2017, 400,000 adults in the UK have rheuma-

toid arthritis, and prevalence increases with age. Fur-

ther, 8.75 million people in the UK have sought

treatment for osteoarthritis, where prevalence also in-

creases with age (Arthritis Research UK, 2017).

For the management of rheumatoid arthritis, the

National Institute for Health and Care Excellence

(NICE) in the UK recommends access to physiother-

apy services, to improve ﬁtness and encourage reg-

ular exercise (Deighton et al., 2009). Similarly, for

the management of osteoarthritis NICE recommends

physical exercise for local muscle strengthening and

general aerobic ﬁtness (Conaghan et al., 2008).

In term of assessing pain intensity - a key mea-

surement for arthritis management - there are differ-

ent scales available, with good degree of correlation

among them (Downie et al., 1978). Once could use a

descriptive scale (nil, mild, moderate, severe, very se-

vere), a numeric scale (0 - no pain, 10 - worst possible

pain) (Farrar et al., 2001) or a visual analogue scale

(100 mm in length anchored by the two extremes)

(Hawker et al., 2011).

Based on the clinical context, there is a demon-

strable need for data related to physical activity and

pain levels to be shared between users, GPs and phys-

iotherapists. The hypothetical customer journey we

use as a source of requirements is then: a patient di-

agnosed with arthritis has been referred by his GP to

a physiotherapist. The physiotherapist recommends a

regime of moderate physical activity, which includes

outside running sessions and pilates classes at a local

gym. The user would like to share a full record of his

physical activity with the physiotherapist, as well as

be able to track how his symptoms (i.e. pain levels)

change over time, as a result of his efforts. In turn,

the physiotherapist would like to share summary data

with the patients GP.

To support this journey, we need to enable three

types of information ﬂows.

The ﬁrst scenario we aim to support is a person

centric view: allowing one individual to aggregate

data about himself which is held by different plat-

forms and providers. For example, somebody manag-

ing osteoarthritis could attend physiotherapy sessions,

he could run in the park during the weekend and at-

tend a T’ai chi class in the gym once a week. He

should have access, combine and own all of this data

- information generated by a specialist, by a wearable,

by a ﬁtness service provider, or self reported.

The second scenario we aim to support is an or-

ganisation centric view: allowing service providers to

receive data shared with them by their clients. For ex-

ample, a physiotherapist should be able to access data

about physical activity levels and self reported pain

levels from each individual under treatment. Equally,

a research institution may want to accumulate large

datasets of lifestyle data for clinical research, from

patients willing to share their data.

Third, we are looking to support organisations that

hold lifestyle data about their clients and want to of-

fer access to this data to the individuals themselves.

For example, a gym chain looking to share individual

attendance data with the gym members, so that they

can further share this data with a physiotherapist.

4 FHIR PROFILES FOR

LIFESTYLE DATA

Our ﬁrst contribution is a demonstration of how the

FHIR standard may be used to describe lifestyle data

in a provider agnostic fashion. There are many poten-

tial data sources for lifestyle data:

• self-recorded observations

• observations captured by providers of a certain

service: health assessments, personal coach

• smartphones, wearables and body sensors

• smart home devices: smart meters, environment

sensors, smart speakers with voice-enabled digital

assistants

In order to capture this variety of data we needed

to identify a set of proﬁles capable of representing

it. Following FHIR best practices (Furore, 2017), we

chose to extend the Observation proﬁle, prioritizing

removing ﬁelds irrelevant for our context of use, and

only introducing new ﬁelds when no existing ﬁelds

could possibly represent the data. The constructed ob-

servation proﬁle is shown in Figure 1. This could be

used for example to create a recording of self reported

pain intensity, in reference to our arthritis case study.

FHIR FLI: An Open Source Platform for Storing, Sharing and Analysing Lifestyle Data

229

Figure 1: FHIR Observation Proﬁle.

The code ﬁeld is used to specify the type of obser-

vation (in our case, a pain intensity recording). The

subject ﬁeld is a reference to the person the observa-

tion concerns. The performer ﬁeld is a reference to

the source system (e.g. device, organisation or per-

son). To record the measurement result, we use a set

of related ﬁelds, grouped under Component. These

include the actual value, but also the admissible range

and the method of measurement. Finally, we record

both the time period described by this observation -

the effective time (in our case, when did the pain oc-

cur) and also the date the FHIR observation has been

created - the issued date.

The ﬁelds discussed so far are applicable to any

type of observation. There are some additional con-

siderations that are speciﬁc to the lifestyle domain.

First, we need to account for the varying degree

of reliability for the data sources: looking at a simple

measure such as weight measured in kilograms - this

could be self-reported, retrieved from a smart scale,

or measured during a health assessment. To capture

this ’trustworthiness’ of the data, we introduce a ﬁeld

for reliability, that takes numerical values. Each or-

ganisation using these FHIR proﬁle may specify rules

that assign reliability scores to source systems. Every

FHIR observation created based on data coming from

a given source would then carry the same reliability

value as that assigned to the source system.

Second, the same event (measurement) may be

captured in more than one system. The same physical

workout (e.g. a pilates class) could be captured by a

consumer wearable device and by the class booking

system of the gym. However, the event would be de-

scribed differently in the two platforms. Our design

approach is to avoid any reconciliation or aggrega-

tion at this stage. We recommend creating two dis-

tinct FHIR observations. The relation between these

observations (duplicate data, contradictory data, com-

plementary data) shall be established if and when nec-

essary to answer a speciﬁc query.

Third, depending on the source system, data may

be more or less sensitive, so different access rules

would apply to data. To enable differentiated access

rules for different types of observation, we add to the

proﬁle a ﬁeld to record how sensitive the data is.

In reference to our arthritis case study, the other

type of lifestyle information that should be saved re-

lates to physical workout sessions. To document an

exercise session, we propose using a Diagnostic Re-

port proﬁle, shown in Figure 2. This is another type

of FHIR resource that groups observations related to

the same episode. First, our proﬁle on Diagnostic Re-

port speciﬁes the same ﬁelds we have discussed for

individual observations: a code, subject, performer,

time. Second, the result ﬁeld contains reference to

additional observations. The list we provide is not ex-

haustive. Depending on what information is available

from the source system, implementers may add addi-

tional references to observations, so that all available

information is captured.

Many lifestyle devices also make the distinction

between aggregate and trace data - running watches

are able to record a user’s heart rate every few sec-

onds during a session (trace), and also calculate an

average heart rate at the end (aggregate). To reﬂect

this, where applicable, ﬁelds for both types of data

have been added to the proﬁle - heartRate and aver-

ageHeartRate.

Figure 2: FHIR FLI Proﬁle - Workout Session.

ICT4AWE 2018 - 4th International Conference on Information and Communication Technologies for Ageing Well and e-Health

230

5 REFERENCE ARCHITECTURE

A range of EHR platforms already use the FHIR

standard as a data exchange mechanism. However,

FHIR messages are mapped to an internal clinical

model (for example, OpenEHR) and then saved in

a structured database. Data can be interrogated ei-

ther through structured query languages on the inter-

nal clinical model.

For example, SMART ON FHIR (Mandel et al.,

2016) is an EHR platform built to support importing

and exporting healthcare data through FHIR compli-

ant messages - however, all data is stored internally as

a set of key value pairs in a relational database.

In comparison, we aim to investigate whether an

infrastructure based solely on FHIR messages can be

used for reporting and analytics. By designing and

implementing such a platform, we aim to evaluate the

role of the FHIR standard when it comes to storing

data, as opposed to transmitting data (as it has been

traditionally seen).

In this section, we discuss the components of a

reference architecture built around the FHIR proﬁles

for lifestyle data.

First, the platform should have a Connector com-

ponent to allow consumers to retrieve lifestyle data

from different sources. The connector is required be-

cause lifestyle data is usually spread across a multi-

tude of systems and platforms, each offering an in-

complete view of the individual. We aim to offer con-

sumers an easy way to retrieve their own data from

these various platforms. The connector comes with

built-in integrations to the main consumer-oriented

lifestyle tracking platforms. The connector can also

be extended by each organisation implementing the

reference architecture so that it also retrieves data

from company-speciﬁc internal databases or systems.

The connector is also responsible to retrieve and send

data in the FHIR format. This ensures interoperability

between different organisations that have committed

to exposing data in FHIR format.

Second, the architecture requires a Converter

component, which maps data from proprietary for-

mats onto FHIR proﬁles for lifestyle data. The con-

verter is also responsible to record the source of the

information for each FHIR observation created. It

also enacts the business rules, assigning values to the

reliability and sensitivity ﬁelds of the FHIR proﬁles.

The converter component relies on a library of se-

lected FHIR proﬁles suitable to capture common data

items in the lifestyle domain. Organisations imple-

menting the infrastructure for internal use could de-

velop additional FHIR proﬁles for data they hold in-

ternally. We propose a development process whereby

whenever a new FHIR proﬁle for lifestyle data is cre-

ated by organisations implementing this architecture,

the converter component is updated accordingly.

Third, the reference architecture requires a docu-

ment database, to store each FHIR observation in its

native format - a JSON message.

Fourth, the reference architecture includes a data

visualisation component, which comes with built-in

visualisations for common lifestyle data. We pro-

pose a process whereby whenever a new FHIR pro-

ﬁle is created, the data visualisation component is ex-

tended so that it can graphically represent the new

type of data. Organisations implementing the archi-

tecture may develop custom data visualisations to suit

their speciﬁc needs. One clear beneﬁt of building vi-

sualisations on top of a standard data model such as

FHIR is that visualisation over the data will continue

to function as expected, even if the systems generat-

ing the data change.

Fifth, the reference architecture allows for an ad-

vanced analytics component. This enables machine

learning experts to build predictive models against a

standard representation of data. These models can be

evaluated and used independently of the underlying

data source. The type of models particularly suitable

to lifestyle data is based on time series analysis of

behavioural traces, in order to predict trends in be-

haviours of interest or occurrence of speciﬁc events.

6 USER EXPERIENCE

In this section, we discuss new ways for customers

and organisations to interact with lifestyle records.

Whereas health records are accessed by clinicians in

controlled environments (a GP clinic, a hospital etc),

interaction with lifestyle records are more diverse, in

terms of users, context and purpose.

One such interaction model is that of voice-

enabled services, which, aside from increasing con-

venience, could remove a host of accessibility barriers

from users.

A personal lifestyle record could allow consumers

to create self-reported observations (for example, re-

lated to pain levels) using voice-enabled digital as-

sistants. Through this, elderly people facing accessi-

bility problems when a using a smartphone can more

easily manage a chronic condition using voice to track

their symptoms. Conversely, when developing these

interfaces, it is important to ensure that they respond

logically as otherwise could lead to frustration.

Similarly, lifestyle records can also by queried by

voice, asking for example for the next physiotherapy

session booked at the gym.

FHIR FLI: An Open Source Platform for Storing, Sharing and Analysing Lifestyle Data

231

Due to the nature of the interfaces provided by

voice enabled services, the types of data that can be

recorded must be predeﬁned. For example, the Ama-

zon Alexa interface does not allow users to submit

voice recordings as inputs, hence it would not be

possible to record qualitative data. However most

lifestyle data can be recorded quantitatively. For those

that can’t, quantitative mappings can be deﬁned (a nu-

merical scale in the case of pain) mitigating this issue.

7 FHIR FLI - OPEN SOURCE

IMPLEMENTATION

Based on the reference architecture, we have started

development of an open-source implementation -

FHIR FLI. This acts as a starting point for organisa-

tions wishing to deploy a software solution that con-

forms to the proposed reference architecture.

FHIR FLI contains a number of FHIR proﬁles

speciﬁcally developed for the lifestyle domain. For

example, a proﬁle to standardise sleep information,

or for describing a workout session. The full list is

available at https://simpliﬁer.net/FhirFli/ resources

FHIR FLI offers out of the box connectors that

allow consumers to authenticate and pull lifestyle data

from Fitbit, Google Fit and Apple Healthkit. Data is

then converted into the new FHIR proﬁles and saved

into a document database.

Finally, FHIR FLI contains a data visualisation

component and an Amazon Alexa skill under current

development.

The FHIR FLI data visualization component pro-

vides an interface for both corporate and individual

users to view and analyse FHIR data. Using the

dashboard, individual users can view their own per-

sonal data, whereas corporate users can view aggre-

gate anonymised data for all users registered with

their companies. The dashboard draws its data from a

document database, but also reroutes requests through

an Apache Drill connector to allow for larger scale

queries (in the case of corporate users).

The FHIR FLI Amazon Alexa skill provides

users a ﬂuid interface for making speciﬁc prede-

ﬁned queries on FHIR data stored within a docu-

ment database - queries such as ”What was my aver-

age heart rate yesterday?” will retrieve and return any

matching data through the Amazon Alexa device. The

skill also allows users to record data into a document

database through an Amazon Alexa, and once a user

has submitted the data, it is converted into the new

FHIR proﬁles and stored into the document database.

The source code is available at https://github.com/

fhirﬂi

8 CONCLUSIONS

FHIR FLI is an open source Personal Lifestyle

Record that helps both consumers and organisations

to combine and analyse lifestyle data, independently

of the source systems.

Using FHIR ensures interoperability, allowing

consumers to share data with service providers or

research institutes and empowering organisations to

share more data with customers.

The next step we are pursuing is assessing the fea-

sibility and performance of a FHIR based data reposi-

tory compared with data repositories based on clinical

models such as OpenEHR.

Future work should also concentrate on the imple-

mentation of access control policies and development

of the machine learning module. This would demon-

strate how large data sets of lifestyle data can be

assembled from individuals willing to anonymously

share their data with research institutes, to build new

predictive models linking lifestyle behaviours to clin-

ical risks and outcomes.

ACKNOWLEDGEMENTS

We are indebted to Dr. Dean Mohamedally, our

project supervisor, for guiding our research.

REFERENCES

Arthritis Research UK (2017). State of musculoskeletal

health 2017. https://www.arthritisresearchuk.org/

arthritis-information/data-and-statistics/state-of-

musculoskeletal-health.aspx.

Bender, D. and Sartipi, K. (2013). Hl7 fhir: An agile and

restful approach to healthcare information exchange.

In Computer-Based Medical Systems (CBMS), 2013

IEEE 26th International Symposium on, pages 326–

331. IEEE.

Conaghan, P. G., Dickson, J., and Grant, R. L. (2008). Care

and management of osteoarthritis in adults: summary

of nice guidance. Bmj, 336(7642):502–503.

Deighton, C., OMahony, R., Tosh, J., Turner, C., Rudolf,

M., Group, G. D., et al. (2009). Management of

rheumatoid arthritis: summary of nice guidance. BMJ,

338(mar16

1):b702–b702.

Downie, W., Leatham, P., Rhind, V., Wright, V., Branco,

J., and Anderson, J. (1978). Studies with pain rating

scales. Annals of the rheumatic diseases, 37(4):378–

381.

Farrar, J. T., Young, J. P., LaMoreaux, L., Werth, J. L., and

Poole, R. M. (2001). Clinical importance of changes

in chronic pain intensity measured on an 11-point nu-

merical pain rating scale. Pain, 94(2):149–158.

ICT4AWE 2018 - 4th International Conference on Information and Communication Technologies for Ageing Well and e-Health

232

Furore (2017). Fhir best practices. https://simpliﬁer.net/

guide/ProﬁlingAcademy/Best-practices.

Hawker, G. A., Mian, S., Kendzerska, T., and French,

M. (2011). Measures of adult pain: Visual analog

scale for pain (vas pain), numeric rating scale for pain

(nrs pain), mcgill pain questionnaire (mpq), short-

form mcgill pain questionnaire (sf-mpq), chronic pain

grade scale (cpgs), short form-36 bodily pain scale (sf-

36 bps), and measure of intermittent and constant os-

teoarthritis pain (icoap). Arthritis care & research,

63(S11).

Kanta (2017). Phr for developers. https://www.hl7.org/

documentcenter/public temp B8B93A79-1C23-BA

17-0CD130736D1F56A7/wg/mobile/Kanta%20PHR

%20for%20Developers.pdf.

Kavanagh, R. (2017). How is nhs digital using fhir.

http://www.hl7.org.uk/doc store/FHIR/NHSDigital

FHIR.pdf.

Lean, M. E., Leslie, W. S., Barnes, A. C., Brosnahan, N.,

Thom, G., McCombie, L., Peters, C., Zhyzhneuskaya,

S., Al-Mrabeh, A., Hollingsworth, K. G., Rodrigues,

A. M., Rehackova, L., Adamson, A. J., Sniehotta,

F. F., Mathers, J. C., Ross, H. M., McIlvenna, Y., Ste-

fanetti, R., Trenell, M., Welsh, P., Kean, S., Ford, I.,

McConnachie, A., Sattar, N., and Taylor, R. (2017).

Primary care-led weight management for remission

of type 2 diabetes (direct): an open-label, cluster-

randomised trial. The Lancet.

Mandel, J. C., Kreda, D. A., Mandl, K. D., Kohane,

I. S., and Ramoni, R. B. (2016). Smart on fhir: a

standards-based, interoperable apps platform for elec-

tronic health records. Journal of the American Medi-

cal Informatics Association, 23(5):899–908.

Public Health England (2015). England has the potential

to have the lowest disease burden in the world.

https://www.gov.uk/government/news/england-has-

the-potential-to-have-the-lowest-disease-burden-in-

the-world.

FHIR FLI: An Open Source Platform for Storing, Sharing and Analysing Lifestyle Data

233