Gamification for Hand Hygiene Compliance

Joao Guilherme Silva da Cunha

Thesis to obtain the Master of Science Degree in

Information Systems and Computer Engineering

Supervisors: Prof. Miguel Leitao Bignolas Mira da SilvaProf. Luıs Velez Lapao

Examination Committee

Chairperson: Prof. Paolo RomanoSupervisor: Prof. Miguel Leitao Bignolas Mira da Silva

Member of the Committee: Prof. Daniel Jorge Viegas Goncalves

November 2018

Acknowledgments

Muitas pessoas contribuıram para todo o percurso que culminou na escrita deste documento. A

todas elas e a muitas mais que nao tenho espaco para referir dirijo um enorme agradecimento.

Aos meus orientadores, Miguel Mira da Silva e Luıs Velez Lapao, que sempre me guiaram para o

caminho certo, me deram apoio e ajudaram a resolver todo e qualquer problema que aparecesse.

A Rita, que me acompanhou de perto na tese e nunca me deixou desmotivar. Estiveste sempre la

quando precisava de tirar uma duvida e tiveste a paciencia para rever tudo com olhos de lince. Nao

podia ter pedido melhor pessoa para a tarefa e isto tambem e teu.

Aos outros mestrandos que se encontravam na mesma situacao que eu. Aquelas reunioes em que

todos ajudavam todos a nao desmotivar jamais serao esquecidas.

A Susana por me ouvir desabafar e me guiar naquele que foi o meu perıodo mais difıcil.

Ao pessoal do curso que comecou por ser um bando de desconhecidos que se sentou numa mesa

do bar amarelo e agora esta ca para a vida! Sem voces nao tinha tido a piada que teve fazer aque-

las semanas de projeto a bater com a cabeca na parede por causa de um ”;” e a descomprimir nos

matraquilhos. Levo de voces muitos momentos bons e muito apoio, sempre.

Aos amigos de sempre, voces sabem quem sao, sabem que sao famılia e que se a minha hibernacao

para terminar este percurso nao nos afastou, porque estiveram la para me puxar para descansar um

pouco e cheirar as flores quando precisava, nada mais nos afasta. Nao me esquecerei das vossas

palavras nos momentos difıceis assim como tambem nao esqueco os brindes nos melhores momentos.

A toda a minha famılia, aos meus primos, tios e avos, especialmente aos que ja partiram. Sem as

vossas licoes de humildade, etica, esforco e dedicacao nunca teria conseguido fazer tudo para que o

trabalho ficasse feito.

Aos meus pais, Dina e Joao Paulo, porque sem eles, sem a insistencia para que nunca desistisse

quer no percurso academico quer na vida nunca teria chegado onde cheguei. Sem o seu amor, carinho

e compreensao para aturar as frustracoes e incertezas o caminho teria sido bem mais difıcil. Sem os

sacrifıcios que fizeram nunca teria tido as oportunidades que tive que me trouxeram ate aqui.

A todos estes e aqueles que passaram ainda que brevemente na minha vida o meu mais sincero

Bem Haja!

Abstract

Hospital acquired infections are one of the major problems of healthcare in the world, resulting in in-

creased morbidity and mortality. This problem could be minimized by simply performing good hand

hygiene. However, due to factors like forgetfulness, healthcare workers still have low levels of hand hy-

giene compliance. The main method to monitor compliance levels, direct monitoring, is very expensive

both in time and money. We propose a gamification solution capable of providing feedback about com-

pliance levels and at the same time motivate people to improve their behaviour. This solution requires

many different components working together each posing different challenges, which will be discussed

in this document. In this thesis we propose a model of steps to minimize such challenges impact. The

bottom layer of the model corresponds to the indoor location technology, for which we reviewed and

compared some alternatives. Above that, an AMS is used to monitor hand hygiene, whose data is pre-

sented using gamification. For the gamification layer, we used the solution proposed in the OSYRISH

project. This solution’s interface (the upper layer of the model we propose) was evaluated using user

testing, and although it still needs developing the results were considered acceptable.

Keywords

Hand hygiene compliance; Hospital acquired infections; Indoor location; Gamification; Automated moni-

toring systems;

iii

Resumo

As infecoes hospitalares sao um dos maiores problemas de saude no mundo levando a taxas ele-

vadas de morbidade e mortalidade. Estas infecoes podem ser reduzidas se os profissionais de saude

cumprirem o simples ato de higienizar as maos nos momentos corretos. Infelizmente, devido a fatores

como o esquecimento, estes mesmos profissionais continuam a ter baixos nıveis de cumprimento das

diretrizes para a higienizacao das maos. O metodo mais usado para medir estes nıveis, observacao

direta, e dispendioso tanto em termos de tempo como de dinheiro. Propomos, por isso, uma solucao

com base em gamificacao capaz de dar feedback sobre o cumprimento dos momentos de higienizacao

ao mesmo tempo que motiva os profissionais a melhorar o seu comportamento. Uma solucao deste

genero requer muitos componentes diferentes a trabalhar em conjunto o que levanta inumeros desafios.

Para minimizar o impacto destes desafios esta tese propoe um modelo de passos a seguir. A camada

base do modelo corresponde a tecnologia de localizacao indoor para a qual analisamos e comparamos

varias alternativas. Por cima dessa camada, temos um Sistema de monitorizacao automatica que mon-

itoriza os nıveis de higienizacao das maos, nıveis esses que sao apresentados usando gamificacao.

Para a camada de gamificacao, usamos a solucao proposta pelo projeto OSYRISH. A interface desta

solucao (camada superior do modelo proposto) foi avaliada usando testes com utilizadores, e embora

precise de ser mais desenvolvida os resultados dos testes foram aceitaveis.

Palavras Chave

Higienizacao das maos; Infecoes Hospitalares; Localizacao Indoor; Gamificacao; Sistema de monitorizacao

automatica;

v

Contents

1 Introduction 1

1.1 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Organization of the Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Problem Identification 7

3 Related Work 11

3.1 Monitoring Hand Hygiene Compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.1.1 The Five Moments for Hand Hygiene . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.1.2 Methods for Hand Hygiene Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.1.2.A Direct Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.1.2.B Indirect monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 Indoor Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.2.1 Signal properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.2.2 Location Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.2.3 Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3.3 Gamification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3.1 Gamification in Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4 Proposal 27

4.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.2 Model for a Gamification Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.2.1 User Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5 Design and Development 33

5.1 Indoor Location Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5.1.1 Commercial Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5.2 Indoor Location Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.3 Automated Monitoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.4 Gamification Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

vii

5.5 Gamification System Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

6 Evaluation 45

6.1 Indoor location Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.1.1 Pozyx . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.1.1.A Test Set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.1.1.B Testing and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.1.2 Saninudge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6.1.3 Estimote . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6.1.4 Comparing Commercial Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6.2 Indoor Location Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6.3 Automated Monitoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.4 Gamification System Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.4.1 User Tests Set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

6.4.2 User Tests Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

7 Conclusion 59

7.1 Lessons Learned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

7.4 Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

7.5 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

A User tests materials 69

A.1 Consent Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

A.2 Test Script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

viii

List of Figures

1.1 DSRM Process in this thesis context (adapted from [1]) . . . . . . . . . . . . . . . . . . . 4

2.1 Observed prevalence of HAIs with 95% confidence intervals and predicted prevalence of

HAI in acute care hospitals based on patient case-mix and hospital characteristics, by

country, ECDC PPS 2011-2012 [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1 The 5 moments for Hand Hygiene [3] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2 Game elements Hierarchy (adapted from [4]) . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.1 Model artefact proposed to solve the problem of developing a gamification based solution

for improving hand hygiene compliance in healthcare . . . . . . . . . . . . . . . . . . . . . 30

5.1 Pozyx tag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

5.2 Estimote Proximity Beacon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.3 Flowchart of the program flow if the dispenser tag was used . . . . . . . . . . . . . . . . . 40

5.4 Indoor App Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.5 Automated Monitoring System Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

6.1 Deployed Anchors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.2 Predefined Route . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6.3 Test results with default settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.4 Test results with default values plus filtering . . . . . . . . . . . . . . . . . . . . . . . . . . 50

6.5 Pozyx tag with 9V battery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6.6 Test results with preamble length 256 symbols plus filtering . . . . . . . . . . . . . . . . . 52

6.7 Test Dispenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6.8 European Style Dispenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

ix

x

List of Tables

1.1 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

6.1 Commercial Solutions Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6.2 User tests number of clicks by task (confidence level of 95%) . . . . . . . . . . . . . . . . 56

xi

xii

Acronyms

AMS Automated Monitoring System

AOA Angle Of Arrival

API Application Program Interface

BLE Bluetooth Low Energy

DSRM Design Science Research Methodology

DToA Difference Time of Arrival

GPS Global Position System

HAI Hospital Acquired Infection

HW Healthcare Worker

ICU Intensive Care Unit

IR Infrared

OSYRISH Organizational and Informational System to Improve the Management of Healthcare

Associated Infections in Hospitals

RF Radio Frequency

RSS Received Signal Strength

SDK Software Development Kit

SUS System Usability Scale

TDoA Time Difference of Arrival

ToA Time of Arrival

xiii

ToF Time of Flight

UWB Ultra Wide Band

WHO World Health Organization

WLAN Wireless Local Area Network

xiv

1Introduction

Contents

1.1 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Organization of the Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1

2

Hospital Acquired Infections (HAIs) are one of the biggest problems in healthcare around the world

[5]. HAIs cause extra-days of hospital stay, increase the cost for both hospitals and patients as well and

contribute to higher morbidity and mortality rates [6].

Preventing this particular type of infections can be done in a simple and inexpensive manner: per-

forming hand hygiene at the right moments. Nevertheless, the compliance by Healthcare Workers (HWs)

is still far from desired levels.

There are several factors leading to this low level of compliance with hand hygiene guidelines, like

Lack of time, forgetfulness and skin damage caused by products used in hand hygiene [7].

Despite the existence of several methods to measure the hand hygiene compliance rates, direct

observation is still considered the golden standard. This method is very expensive and time-consuming.

For this reason, there is a need for new and innovative ways to monitor and improve HWs’ compliance

with hand hygiene guidelines.

A earlier attempt at solving this problem was part of the Organizational and Informational System

to Improve the Management of Healthcare Associated Infections in Hospitals (OSYRISH)1 program [8].

The solution aimed at measuring HWs’ hand hygiene compliance levels in a fun and innovative way.

However, this solution had several obstacles before it could be properly deployed and tested. Several

indoor location technologies were tested to track the position of HWs in order to accurately monitor their

hand hygiene compliance, none produced good enough results for pilot testing.

This thesis builds on that solution’s work, and proposes a way for minimizing the problems in the de-

velopment of a gamification based solution of an system capable of monitoring hand hygiene compliance

and presents the development of said solution.

1.1 Research Methodology

We chose Design Science Research Methodology (DSRM) as the methodology to follow during this

research as it is a problem-solving process based on the understanding and formulation of a design

problem.

The main goal of this methodology is the creation and evaluation of artefacts. These artefacts which

can be constructs, models, representations, methods or instantiations must be useful. They should be

able solve unsolved problems in innovative ways, or solve already solved problems in a more efficient or

effective way [9].

DSRM consists of six phases [1]:

Problem Identification: Define the specific research problem and justify the value of a solution. In

order for the artefact to provide a solution to the defined problem said definition should atomize the

1OSYRISH: http://osyrish.org accessed January 2018

3

problem conceptually. This allows the solution to capture the problem’s complexity.

Define the solution’s objective: Based on the problem definition and on the knowledge of what is

possible and feasible infer the objectives of a solution.

Design & Development: After determining the artefact’s desired functionality and architecture

create it.

Demonstration: Use the artefact to solve one or more instances of the problem. This can be done

in a variety of ways: experimentation, simulation, case study or any other appropriate activity.

Evaluation: Observe and measure how well the artefact supports a solution to the problem, by

comparing the results obtained when using the artefact in the demonstration with the objectives of

the solution. The evaluation could include any appropriate empirical evidence or logical proof, and

the knowledge of relevant metrics and analysis techniques.

Communication: Communicate the problem and its importance, the artefact and its utility, novelty,

rigour design, and effectiveness to researchers and other relevant audiences.

This is an iterative process where the output of one phase is the input of another. Although this se-

quence is structured in a nominal order, there is no specified beginning. Because of this, and depending

on the approach, the first phase will vary. One example of this variation is, while on a problem-centred

approach there is an observation of the problem and we start in the identification phase. On a design and

development-centred approach where an artefact could have been developed for a different research or

domain, we would start with the Design & Development phase [1].

In Figure 1.1 we can see the DSRM process adapted to our research, following a problem-centred

approach. The Demonstration step was omitted because we simply tested the artefacts and measured

the results in terms of the objectives as described in the Evaluation activity and were.

Identify Problem Define solution'sobjective

EvaluationDesign &Development

Communication

The challenges ofimplementing,deploying andevaluating agamification solutionfor improving handhygiene compliance.

Show and follow thesteps needed to develop and deploya gamificationsolution for handhygiene compliance:

Accurately monitorhand hygienecompliance;

Crate a cheap, highaccuracy, low batteryconsumption ,automated and noninvasive monitoringsystem;

Integrate the AMSwith the gamificationplatform;

Validate the systemprior to deployment.

Controlledenvironment testing;

User Testing.

Dissertation;

Papers.

Model used toreduce challengesimpact ondeveloping anddeploying agamification basedsolution for handhygiene compliance;

System comprisedof each layer of themodel resulting in agamification basedsolution for handhygiene compliance.

Figure 1.1: DSRM Process in this thesis context (adapted from [1])

4

1.2 Organization of the Document

The chosen methodology, DSRM influences the structure of the document. This structure is presented

in Table 1.1.

Table 1.1: Thesis Outline

2. Problem Identification Statement of the problem and motivation for this work3. Related Work Critical analysis of the main concepts related to this thesis context4. Proposal Detailed description of the research proposal5. Design and Development Design of the architecture and development of the solution6. Evaluation Evaluation of the alternatives for each choice made during the design and development phase7. Conclusion Final conclusions, lessons learned, limitations, communications and future work

5

6

2Problem Identification

7

8

HAIs (also known as nosocomial or hospital infections) are infections acquired by a patient while

receiving treatment for medical or surgical conditions, in a hospital or other healthcare facility, which

were not present or in incubation at the time of admission [10]. HAIs can manifest after discharge and

can also affect the hospital staff.

These infections are responsible for 37000 deaths and 16 million extra days of hospital stay in Europe

alone [6]. Latest show the HAIs prevalence percentage, corresponding to patients that contracted at

least one HAI, is 5.70% in Europe and 10.87% in Portugal [11](Figure 2.1). If we consider only patients

staying more than two days Intensive Care Units (ICUs) the number is 8.4% [12].

Figure 2.1: Observed prevalence of HAIs with 95% confidence intervals and predicted prevalence of HAI in acutecare hospitals based on patient case-mix and hospital characteristics, by country, ECDC PPS 2011-2012 [2]

HAIs are preventable and can occur due to numerous factors, being HWs’ poor hand hygiene one

of main causes [3]. The World Health Organization (WHO) proposed a technique to carry out hand

hygiene, and the five main moments where it should be performed [3].

Several studies show, that although most HWs are aware of these guidelines, their hand hygiene

compliance levels continue to be below desired [13,14].

The most common method to observe the levels of hand hygiene compliance is direct observation [3].

In this method, a trained observer will accompany the HW and register the worker’s compliance, or not,

when presented with an opportunity for hand hygiene.

9

Despite its accuracy, direct observation has some disadvantages, the high cost in time and money

for training and validating the observers, and the possibility of the presence of Hawthorne effect [15].

Following this, hospitals need to address this problem by using approaches which are less costly,

non invasive and can motivate HWs to keep a high level of compliance even when they are not being

monitored.

Gamification has been used in health for several different reasons and can be used as a motivating

tool [16]. Due to the technological advancements in recent years,Automated Monitoring Systems (AMSs)

have been attempted as an alternative to direct monitoring.

Implementing a gamification solution however, is not easy and AMSs still have several problems

when reporting hand hygiene compliance accurately.

Succinctly, this thesis addresses the challenges of implementing, deploying and evaluating a

gami fication solution for improving hand hygiene compliance.

10

3Related Work

Contents

3.1 Monitoring Hand Hygiene Compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3.2 Indoor Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.3 Gamification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

11

12

Developing systems or solutions to address the identified problem requires us to monitor the actual

hand hygiene compliance. For that, we need to understand the guidelines for hand hygiene, and then

the methods used to measure the compliance levels.

Furthermore, to measure hand hygiene AMSs are often used. Because of this we need to un-

derstand how AMSs this means understanding the technologies and methods used to perform indoor

location. Additionally, we studied gamification as it has successfully been used as a tool to help promote

behavioural change [16].

3.1 Monitoring Hand Hygiene Compliance

Monitoring hand hygiene compliance is essential to measure HWs’ performance so they can improve

their rates. For this purpose we need to understand the WHO 5 moments for hand hygiene.

3.1.1 The Five Moments for Hand Hygiene

Hand hygiene levels are still very low [17]. In order to improve this the WHO released a framework listing

guidelines, known as the five moments, where hand hygiene should be performed.

Theses five moments, presented in Figure 3.1, are as follows:

1. Before touching a patient: the period between the last contact with a surface or object outside

of the immediate surroundings of the patient and the first contact with the patient’s zone, hand

hygiene at this moment will mainly prevent cross-colonisation of the patient.

2. Before a clean/aseptic procedure: hand hygiene should be performed after the HW touches a

surface, or an object, in the patient’s immediate surroundings and before performing a clean/asep-

tic procedure, for example performing wound care. This prevents colonization and HAIs.

3. After body fluid exposure risk: after performing a task associated with risk of exposure to body

fluids and before touching anything else, even in the patient’s vicinity ,the HW should perform hand

hygiene. This reduces the risk of colonisation or infection of HWs with infectious agents and the

risk of transmission of micro-organisms from a colonised to a clean body site within the same

patient.

4. After touching a patient: after a care sequence and before touching an object hand hygiene

substantially reduces contamination of HWs’ hands with the flora from the patient. This behaviour

minimizes the risk of dissemination to the healthcare environment, and protects the HW.

5. After touching patient surroundings: if the HW performs hand hygiene after touching the pa-

tient’s surroundings, it reduces hand contamination.

13

Figure 3.1: The 5 moments for Hand Hygiene [3]

3.1.2 Methods for Hand Hygiene Monitoring

After understanding the moments for hand hygiene we need to measure the HWs’ compliance levels

against those guidelines. The more known methods, both for direct and indirect monitoring, are pre-

sented below, alongside their advantages and disadvantages.

3.1.2.A Direct Monitoring

Direct monitoring requires a human component to observe the hand hygiene actions being performed.

The two most common methods of direct monitoring are:

• Direct Observation: This method requires trained observers to accompany HWs. The observer

registers the opportunities for hand hygiene according to a previously established set of rules,

along with the compliance, or the lack of it by the HW. This method is the only one that allows

the evaluation of the hand hygiene technique. Another advantage is the ability to find the hand

hygiene compliance rate by type of HWs (nurse, doctor, ...) [15].

There are several disadvantages associated with the use of direct observation, the very high cost

and time of the observers’ training are two of them. Different hospitals have different heuristics

to evaluate the data. This fact makes comparing the obtained hand hygiene compliance rates

between facilities difficult.

Finally, direct observation is subject to observer bias, meaning different observers may gather

different information within the same setting, and observation bias also known as Hawthorne effect

14

[15], which refers to people changing their behaviour when they know they are being observed,

leading to deceitful hand hygiene compliance rates [18].

• Self-assessment: HWs report their own hand hygiene compliance rates, for example by filling

cards or questionnaires at the end of the shift. Compared to all the other methods, this is by far

the least expensive, both in cost and in time consumption. However, people tend to overestimate

their compliance rates making this method a poor choice [19].

3.1.2.B Indirect monitoring

As opposed to the direct observations, indirect monitoring methods, minimize the human component in

the calculation of the hand hygiene compliance rates, either using devices or estimating the number of

hand hygiene actions performed.

• Monitoring Hand Hygiene Product Usage: In this type of monitoring, the hand hygiene compli-

ance is calculated based on the observation of the amount of product consumed when performing

hand hygiene.

This can be done by counting the number of bottles that needed replacement and then, based

on the average quantity of product required to perform hygiene, calculate the number of hygiene

operations performed. The next step is to divide the previously calculated amount by patient-

days [15].

The observation and observer bias, which were disadvantages in the previously presented meth-

ods, do not apply to this one. Also, the labour required to perform this method is significantly

lower. Nonetheless, there are some disadvantages, such as the inability to confirm that the hy-

giene actions were performed according to the five moments presented before and if the employed

technique was the correct one.

• AMS: In recent years there has been an emergence in intelligent systems to help monitor hygiene

compliance rates. These systems consist of some type of wearable, such as a tag or a small

device, the HW carries during the shift. Based on information captured by these devices, HWs are

prompted with reminders when they should perform hygiene, and positive reinforcement whenever

they comply with the five moments for hygiene.

AMSs, as with the method above, are unable to evaluate the correctness of the applied hygiene

technique. Another limitation is the high cost to implement such a system, and the elevated main-

tenance one may require [20].

Despite the limitations, AMSs show some promise [21, 22] hence the need to understand indoor

location.

15

3.2 Indoor Location

Indoor location consists of the tracking of individuals inside enclosed spaces.

Indoor location is challenging because the devices cannot obtain enough Global Position System

(GPS) signals. Also, the environment in which the localization is performed, a healthcare facility, in this

case, has many different objects, materials, and people causing what is commonly referred to as noise.

In this section, we will describe some of the most mentioned technologies to perform indoor location,

as well as the products resulting from the application of those technologies. No technology is perfect,

sometimes there is a need for a combination of several, in order to achieve the best outcome for a

specific situation.

3.2.1 Signal properties

Different indoor location methods use different signal propagation properties when calculating the po-

sition. In order to better understand how the different methods perform the calculations we need to

comprehend these properties.

• Angle Of Arrival (AOA): By having several antennae in an array we can gather the angle of

the arriving signal based on its arrival time at each antenna. This provides information about the

direction of a node. After determining angular separation between a node and a certain number of

anchors and using angulation methods, we can determine a node position [23].

• Time of Arrival (ToA): Time the signal takes to arrive at a determined number of sensors. We can

estimate the distance to the anchor because the signal has known propagation properties. One of

the problems with this approach is the necessity to know when the signal was transmitted. Another

is the time synchronization required between the sender of the signal and the receiver. Finally, it’s

very difficult to precisely record the arrival time of radio signals which travel close to the speed of

light [24].

• Time Difference of Arrival (TDoA): An improvement on TDoA since it eliminates the need to

know when the signal was transmitted. It does this by using several synchronized nodes which

receive the signal, and at a specified time instant check the difference in arriving times. Since the

signal travels at constant speed, the position can be determined if we have enough nodes [24].

• Difference Time of Arrival (DToA): The receiving nodes are precisely synchronized so that they

can all receive the same mobile node’s beacon, and calculate the difference in distance based on

this precision clock. One advantage is the mobile node itself doesn’t need to be synchronized with

the network.

16

3.2.2 Location Methods

Calculating the indoor position of a subject can be done using several methods, including:

• Triangulation: is a family of methods including both lateration and angulation. Lateration is the

calculation of position the subject based on the relative distance to several previously known fixed

points. This calculation is made indirectly by obtaining and measuring certain parameters which

are proportional to the distance. These parameters include Time of Flight (ToF) and power atten-

uation of a radio signal among others [25]. Angulation uses the AOA of signals, emitted form fixed

points, to calculate the position of the subject [23].

• Trilateration: as opposed to triangulation, this method does not include the measurement of

angles. It uses three different points (anchors), it then creates three circles (2D) or spheres (3D),

centred on these anchors and with a radius equal to the distance to each of them. The distance

is calculated using ToF. The point where these circles or spheres overlap is the position of the

subject.

• Proximity: confirmation of the subject’s presence in the immediacy of a sensor either by physical

contact or presentation of a device. This method has limited range and analysis capabilities [25].

• Scene Analysis: monitoring of a wide area from a vantage point using sensors with ample cover-

age area and range. Ceiling mounted video cameras or passive infrared sensors, are some of the

examples [25].

• Dead Reckoning: using sensors which provide location updates. These updates are calculated

based on a previously estimated location. Estimation of the location is commonly based on ac-

celerometry and gyroscopy [25].

3.2.3 Technologies

There is a variety of technologies used to perform indoor location, each one with its own features. In this

subsection we will study the most popular.

Radio Frequency (RF) is the generation of electromagnetic waves. Location techniques in this

category use one or more parameters of electromagnetic waves. The values of these parameters

typically depend both on the distance travelled by the signal and the characteristics of the surrounding

environment.

There is an attenuation of RF signals caused by wood and concrete, there are also reflections, and

scattering caused by metal and water. Both of these factors, in conjunction with the diffraction of

radio waves, lead to multi-path radio wave propagation, which, in turn contributes to an inaccurate

calculation of distance between the transmitter and the receiver.

17

Many of these technologies use fingerprinting to calculate the location of the subject. The fingerprint-

ing technique consists of two steps. On the first step, metrics (fingerprints), from the radio signal are

measured at specific points of the environment. These values are then stored, along with the spatial

coordinates of the measurement, in a map.

For the second step, the mobile receiver extracts the fingerprint of the radio signal at his location

and then one of two things may occur. Either the receiver finds the closest match, between the

obtained fingerprint and the ones in the map, [26] or it follows a more triangulation like approach,

where several candidate locations similar to the received signal, are merged to estimate the location

of the receiver in space [27].

In both approaches, the discovery of the most similar match(es) is done through algorithms such as

nearest neighbours and maximum likelihood. Due to the unstable nature of the environment in terms

of presence of people and objects, the signal properties are always changing, because of this, the

map is easily outdated.

Some technologies which use the fingerprint method are presented below:

• Bluetooth Low Energy (BLE): is a successor of classical Bluetooth, designed to ease com-

munication within a short range, for devices that require little quantity of data transfer. A benefit

of this technology is most portable devices already possess a BLE interface, also due to its low

power consumption, it is expected to have an autonomy of several years on 2xAA batteries.

Due to its small dimensions, it’s very easy to install, and can be placed in strategic locations to

improve localization. The advantage of this technology when compared to the regular Bluetooth

is the fact that it does not require the long handshake period. And compared to wi-fi it’s much

cheaper [28].

• Wi-Fi: one of the most used approaches for indoor positioning. Wi-fi is a Wireless Local Area

Network (WLAN) technology defined by IEEE 802.11 standards that allow communication be-

tween electronic devices over a wireless signal. This approach takes advantage of both, the

available WLAN infrastructure deployed in most modern buildings and the fact that most of the

electronic devices already possess a WLAN interface. Because wi-fi operates in the 2.4 GHz

band, it’s highly affected by humans and objects present in the room, this makes it less re-

liable [28, 29]. Another problem is the fluctuation of Received Signal Strength (RSS) of wi-fi

signals.

Examples of technologies that don’t usually use fingerprinting are:

• Ultra Wide Band (UWB): consists of short duration pulses to mitigate propagation problems.

This technology is characterized by possessing a very good accuracy. Obstructions and reflec-

tions of the radio signal cause problems when using triangulation by creating non-line-of-sight

18

scenarios. In these, the signal travels to the receiver through an indirect path leading to a wrong

distance calculation. However, the short pulses of UWB mitigate this problem [30].

• RFID tags: is a system composed of one or more reading devices capable of wirelessly ob-

taining the IDs of tags present in the environment. The reader transmits a RF signal, and the

tags reflect and modulate it by adding an identification code. The RFID tags can be of two

types, the active tags are powered by a battery, while the passive ones draw energy from the

incoming radio signals. The last ones have the problem of having a limited range due to power

constraints [31].

Photonic Energy is the energy carried by electromagnetic radiation. Visible light refers to these

waves with wavelengths between 380 and 750 nanometres. Ultraviolet and infrared are located in

the lower and upper vicinity of this spectrum respectively. This category comprises solutions which

estimate the position of the subject based on the photonic energy received from infrared or visible

light emissions or reflections. The most relevant technology of this type is:

• Infrared (IR): refers to the energy carried by electromagnetic radiation in the upper vicinity of

the visible spectrum. This technology has several problems such as the decay of IR power in a

non-linear way related to distance, and the disturbance of IR signals caused by light and thermal

radiation [32]. Also, occlusions caused by objects or humans contribute to the inefficacy of the

system.

Sonic Waves are mechanical vibrations transmitted over a solid, liquid or gaseous medium. Dis-

tance calculation made by sonic waves technologies use the quasi-constant speed of the waves in

the air. The most common technique used is the ultrasound combined with RF.

• Ultrasound: is sound waves produced by vibration above the threshold of human hearing.

The main method used for the calculation of a subject position by this type of technologies is

triangulation based on the ToF of the sonic waves on the air.

This technology is usually combined with RF waves because these waves travel several orders

of magnitude faster than a sonic one. So, when a combination of these two types of signals is

emitted in unison, the difference between the ToA of both types of waves at the receiver side is

a good approximation of the ToF of the sonic wave.

19

There are also ultrasound only approaches that use broadband ultrasound waves to solve the

synchronization problem. This is done by allowing several emitters to transmit signals sharing

the same frequency band concurrently, causing minimal interference to the other signals being

transmitted. Then, direct sequence code division multiplexing techniques are used to combine

multiple signals simultaneously over the same frequency spectrum. Finally, the DToA of the

signal is used for trilateration in order to estimate the position.

Common problems of this type of technology are the difficulty in detecting the signal in the midst

of the ambient noise, and the complication in isolating a single source when there are multiple

emitters interfering with each other. Also, the echoes caused in the indoor environment, and the

temperature affect the propagations properties leading to wrong distance calculation [33].

Summing up, there are many possible combinations between location methods and technologies.

Considering our problem, using indoor location to create an AMS, the best approach is to consider either

triangulation or proximity. This is due to the fact that scene analysis may invade the privacy, triangulation

is an improvement of triangulation and finally dead reckoning location is based in estimations.

Technology based in RF is less expensive than the alternatives. This means a good approach to our

problem is to find a RF based technology that uses triangulation or proximity methods to perform indoor

location.

3.3 Gamification

In this section, we will describe the concept of gamification, what it is and what it isn’t, and attempt

to clarify some misconceptions about this subject. We will show how to approach a problem using

gamification and some good and bad examples of this concept’s use.

Companies have always tried to stimulate certain behaviours from their employees and customers

using competition, leaderboards and participation badges. Adding that to the development and in-

creased impact of technology gamifiction has been on the rise [34].

It is important to differentiate gamification from serious games. While serious games are games that

are designed to be played in order to learn something, gamification is widely accepted as the use of

game design elements in non-game contexts [35].

The key difference is gamification is not a game itself. The idea behind gamification is to take con-

cepts from game design and apply it to processes or tasks in order to modify or reinforce a certain

behaviour, not develop a playable game.

Gamification is a relatively recent concept that has gained some popularity. However, gamification is

sometimes not taken seriously. It is seen as simply playing games at work or a glorified point system [4].

Although gamification’s adoption is rising, gamification is not a guaranteed recipe for success. In

20

fact, many apps using gamification fail [4].

This happens due to poor understanding of what gamification is, how it works and how to design the

experiences that inspire behaviour changes and result on desired outcomes [36].

Gamification can be used in a more personal context, for example to improve the performance of

housework and company use. In this last context it can target both people outside of the company,

customers for example, and inside, employees. It can also focus specific processes for example the ac-

quisition of customers or outcomes, number of sales for example [36]. In all of these cases gamification

is about ”finding the fun in the things we have to do” [4].

In order to understand if the process is suitable for gamification we can start by answering these four

questions [4]:

1. Motivation: Where would you derive value from encouraging behaviour?

2. Meaningful Choices: Are your target activities sufficiently interesting?

3. Structure: Can the desired behaviours be modelled through a set of algorithms?

4. Potential Conflicts: Can the game avoid conflicts with existing motivational structures?

The best processes to be targeted by gamification are those in which motivating the target population

will have a positive effect on the problem, where we can offer participants some degree of choice and

challenges, that in turn can be turned into code and monitored, and where gamification can increment

the benefits of already in place reward systems [4].

Not all processes will answer positively to all these questions. Nonetheless, the more positive an-

swers we get to each one of them, the better candidate the process is to be the subject of gamification.

Gamification relies heavily on motivation. Motivation can be extrinsic or intrinsic. Extrinsic motivation

factors are those external to the person, money and prizes. On the other hand, intrinsic motivation

factors are things like fun or enjoyment [36].

If we can create extrinsic motivators that create a sense of autonomy, competence, and relatedness

we will obtain better results, as people will feel more motivated to perform the desired behaviour.

However, we should not thoughtlessly attach extrinsic motivators to activities which can be intrinsi-

cally motivated. The user may start to take the reward for granted, performing only the needed work to

achieve it. On the other hand boring and repetitive activities actually benefit from extrinsic rewards [36].

An important aspect of game design to keep in mind is the importance of how the player feels about

the experience and not the formulation of the reward [4].

One very important tool is feedback as players will feel motivated if they receive it unexpectedly, or

about the progress they are making. The metrics we provide feedback about, have the power to steer

the focus of the player. If we only show feedback relating to a particular subject, the player will be more

interested in that subject than the others [4].

21

After understanding this we must consider game elements. Game elements are the tools used to

implement gamification. Werbach and Hunter propose three categories of game elements relevant to

gamification: dynamics, mechanics, and components. Their relation can be described as a pyramid

where dynamics are the more abstract concept and the components the more concrete ones, as illus-

trated in Figure 3.2.

Figure 3.2: Game elements Hierarchy (adapted from [4])

Dynamics are big-picture aspects of the game that we must be aware but can never directly imple-

ment into the game. Some examples of dynamics are emotions, constraints, storyline, progression, and

relationships.

Mechanics are the next level of the pyramid, each one is a way of achieving one or more dynamics.

They are what prompts the action forward and generates player engagement. Challenges, chance,

competition, cooperation, feedback, resource acquisition, rewards, transactions, turns and win state.

Components are the most specific forms of dynamics. Some of them are: achievements, avatars,

badges, collections, content unlocking, gifting, leaderboards, levels, points, teams, virtual goods.

Not all the elements listed above will be used in a project. It is the responsibility of the designer

to choose the right ones. We should start with a large assortment of them and then start to narrow

it down to those suitable for the project. The most commonly used elements are points, badges, and

leaderboards.

Gamification is much more than just gluing these elements together and expect the gamified system

to work. It involves merging emotional concepts such as fun, play and users’ experiences with the

22

creation of systems that serve concrete business objectives. Therefore, we should follow a process like

the one proposed by Werbach and Hunter [4]. This process consists of six steps for the creation of a

gamified system:

1. Define business objectives: define the specific goals of the gamified system. There are many

outcomes of a gamified system, however, we need to verify which outcomes can be considered a

success when measured against what intend to achieve.

2. Delineate target behaviours: define the behaviours in which we want the players to engage and

how these will be measured. These behaviours should align with the previously defined objectives.

3. Describe your players: describe the target players of the gamified system. We need to un-

derstand their relationship with us and what motivates or demotivate them. Not all players are

the same, they may be motivated by different things, segmenting the target population into small

groups can help in making the system fitting for all of them. While some may thrive on competition,

others may find the idea of sharing an experience in a camaraderie environment more stimulating.

4. Devise activity cycles: devise activity cycles, both engagement loops, and progression stairs.

Engagement loops describe what the players do, why they do it and how the system responds.

Progression stairs give a perspective on the player’s journey. It is what creates a sense of progres-

sion, these cycles can be, for example, the actions that need to be taken by the player in order to

receive a reward. These should be of increasing difficulty or complexity between reward tiers.

5. Don’t forget the fun! fun is important in gamification, if the system is not fun to play with players

will just leave.

6. Deploy the appropriate tools: tie in the tasks performed in all the other steps and finally build

and deploy the system.

Sometimes following these steps can be difficult and that leads to the failure of the gamification

initiative. One example of this is the the Marriott Hotel chain failed attempt at attracting potential employ-

ees [37]. This happened because Marriot Hotel may not have understood the motivations of potential

employees when it designed its gamified ‘My Marriott Hotels’ in 2011.

In order to attract new employees, Marriott developed a game in which players simulated work in

an actual Marriott hotel kitchen. Players imitated activities including decorating the hotel dining room,

ordering food inventory, and adhering to a budget. The mechanics were structured in a way that players

would earn points for making customers happy, and would lose points when poor customer service was

delivered. However this points served no purpose whatsoever. The only thing tied to the objective of the

game was the link to the career page.

23

Ultimately this lead to the failure of the game and serves as an example of why we need to carefully

plan when we intend to use gamification.

3.3.1 Gamification in Healthcare

Before we wrote that gamification can be applied in many contexts, healthcare is no exception. In

2016 [38] at the University of Connecticut an experiment was conducted to determine if gamification

could help medical residents improve their score in American Board of Surgery In-Service Training

Examination.

For this purpose they used a gamified microblogging project using Twitter as the platform and modi-

fied questions from a question bank. Everyday a question was posted and the participants were awarded

points for speed, accuracy, and contribution to discussion topics.

The moderator challenged respondents by asking additional questions and prompted them to find

evidence for their claims to fuel further discussion. This solution had very good results as participants

of the program increased their grades when compared to non participants.

Another project [16] used gamification on physical activity, health care utilization, medication overuse,

empowerment, and disease knowledge of the patients affected by rheumatoid arthritis. This study con-

cluded that gamification alone or with social support increased physical activity and empowerment and

decreased health care utilization.

The OSYRISH project proposed the use of gamification as a mean to improve HWs’ hand hygiene

compliance levels. This project used indoor location technologies to monitor and measure the hand

hygiene compliance levels. The information was then gamified in a platform as badges, levels, rankings,

etc. to motivate the HWs to improve their behaviour. In this platform the HWs could share documents

about the hand hygiene topic, review their levels across a period of time and find out the collective

objectives.

3.4 Summary

In this chapter we listed the 5 moments for hand hygiene as proposed by the WHO. We listed the

most know methods for hand hygiene monitoring and the difficulties of implementing an AMS capable of

minimizing the Hawthorne effect at the same time it provides accurate measurements.

In order to implement an AMS system we needed to better understand indoor technologies and its

challenges. For this purpose we conducted an overview of both the methods used to perform indoor

location as well as the several available technologies. Following this we arrived at the conclusion that

the technology used in our project should be triangulation or proximity based, and due to the usually

cheaper price use RF.

24

We also gave an overview on gamification as we needed to understand how to stimulate behaviour

change on HWs and showed several examples of gamification in healthcare.

25

26

4Proposal

Contents

4.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4.2 Model for a Gamification Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

27

28

In this chapter we outline the objectives of the solution and present an overview of how we developed

the solution.

4.1 Objectives

The main objective of this thesis is to show the necessary steps to take when implementing a

gamification system for hand hygiene compliance in an healthcare environment, mainly:

• Accurately monitor hand hygiene compliance levels.

• Create a cheap, high accuracy, low battery consumption, automated and non invasive monitoring

system for hand hygiene compliance.

• Integrate the AMS with the gamification platform

• Validate the system prior to deployment.

4.2 Model for a Gamification Solution

Implementing a gamification based solution to improve hand hygiene compliance in HWs can be a

challenging task [39]. The OSYRISH project had an indoor location problem that hindered the ability

to collect accurate data about the HWs’ hand hygiene compliance. During the development of that

particular platform 3 technologies were tested and several problems arose: the ranging of beacons was

inconsistent, network coverage was poor, hardware falling and disturbing HWs’ work [8].

To minimize the difficulties of developing this type of solution, and achieve the objectives above,

we propose the ArchiMate model in Figure 4.1. We modelled this proposal in ArchiMate because we

needed a non ambiguous notation capable of being recognized, and where we could easily model each

part of the proposed solution and the relations between the parts.

The model comprises five layers:

• Indoor Location Hardware: Indoor location technology equipment capable of transmitting raw

position data (coordinates, signals for proximity detection,...).

• Indoor Location App: Software algorithms capable of transforming the raw positioning data into

data capable of being used in the automated monitoring system.

• Automated Monitoring System: Set of algorithms to transform the position data into hand hy-

giene compliance metrics.

• Gamification Platform: Platform responsible for gamifying the hand hygiene compliance metrics.

29

Figure 4.1: Model artefact proposed to solve the problem of developing a gamification based solution for improvinghand hygiene compliance in healthcare

• Gamification Platform Interface: Part of the gamification platform the users interact with.

We chose to build the solution in an incremental manner because, as we have seen in other attempts

at solving this problem in a similar manner, the OSYRISH project, [8], a gamification platform can not be

properly deployed or tested without the data to gamify.

The data serving as input for the gamification platform is the HWs’ compliance metrics. These

compliance metrics are calculated by an AMS based on indoor location data. The most simple ones are

the number of opportunities where the HW should have performed hand hygiene, and the number of

compliances with these opportunities. This allows the measurement of the compliance level.

The indoor location data is the result of filtering and extracting information from various sensors

capable of monitoring HWs’ actions.

Another important element to take into consideration is the way the gamified information is presented

to the users, that is why the gamification platform’s interface is represented on the model.

30

As we can see each layer requires the one below to be implemented. Only when all of these layers

work together can the solution be properly tested.

On the following sections we detail the choices and developments we made for each layer of the

model adapted to our solution. We evaluate the bottom four layers by performing practical tests in a

controlled environment and the gamification platform interface is evaluated using user testing.

4.2.1 User Testing

The main purpose of these tests is to evaluate the interface of the gamification platform. Such evaluation

must be performed in order to find problems within the user/system interaction.

These problems can hinder the performance of the system, as well as its adoption, when we deploy

it to be tested in the real world. For this reason we must identify and try to solve these problems before

the deployment.

Performing these tests will allow us to identify the problems with the interface and see if everything

is clear enough to be used by any user without getting confused or frustrated.

Despite being more expensive, user testing is one of the testing methods achieving better results [40].

Admittedly, this particular type of testing requires meticulous planning and the creation of a tests’ script.

Some of the questions we needed to answer on the script include:

Objective - What do we want to achieve with the test?

Duration - How long will each test session last?

Equipment - What is the required equipment?

Software - What is the needed software to perform the tests?

State - What is the state of the system at the start of the test?

Coordinator and Observer - Who will coordinate and observe the test?

Users - Who are the users performing the tests, how do we contact them and how many will be

necessary?

Tasks - How many tasks will the users be performing, what tasks are these and in what order will

they be done?

Correct ending - When will we consider that a task was accomplished successfully?

Data - What data will we be collecting and analysing?

Success - What criteria will determine if the interface is a success?

Furthermore, we should create the consent forms, the information gathering forms, satisfaction ques-

tionnaires and the needed material to explain the systems to be evaluated. We also must guarantee all

the users perform the tests in the same conditions.

During the tests we should take into consideration how important the users are. Without their collab-

31

oration the tests would not be possible. For this reason we should not waste their time by performing

tasks that are in no way related to the test.

Also, they should feel comfortable and not pressured. The goal is to test the system, if the users

fail or have a hard time accomplishing a task, they should be reassured that the problem lies within the

system and the fact they were struggling is actually helpful.

32

5Design and Development

Contents

5.1 Indoor Location Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5.2 Indoor Location Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5.3 Automated Monitoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5.4 Gamification Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.5 Gamification System Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

33

34

This section we described the process behind our decisions to implement each layer of the model

and build our proposed solution.

5.1 Indoor Location Hardware

Sensors are the bottom layer of the model. These sensors use the technologies described in Subsection

3.2.3 to emit the signals used in the indoor location calculations. However, not all the technologies are

ideal for the unique challenges a healthcare facility presents.

Implementation of RFID based systems is very costly [41] and has several accuracy problems [42].

Based on this all solutions using RFID were not considered.

IR technology requires an unobstructed line of sight from the emitter to the receiver. Therefore IR

technology was not suitable for our solution.

Patients’ privacy is one of the concerns when performing monitoring actions in a healthcare envi-

ronment. Video monitoring technologies can’t distinguish between HWs and patients, because we can’t

protect patients’ privacy video based technologies were not taken into consideration.

As stated in Subsection 3.2.3 ultrasound technology is unreliable in environments were there is

ambient noise and several emitters. Healthcare facilities have many equipments with different materials

and people, in need of monitoring, moving around. This creates an environment rich in ambient noise

and emitters, as a result ultrasound was not considered for our solution.

Most of the solutions available in the market are based in RF technologies also, these solutions are

usually cheaper than other types of solutions.

Despite the advantages of Wi-fi: pre-existent infrastructure and multitude of devices Wi-fi ready,

the required accuracy for a hand hygiene monitoring system is a big obstacle for Wi-fi. This happens

because the RSSI fluctuates greatly and similarly to ultrasound is affected by the objects and people

present in the environment. Due to this inaccuracy solutions based on Wi-fi were not considered.

The remaining RF based technologies are BLE and UWB.

5.1.1 Commercial Solutions

After narrowing down the technologies the next step was to find indoor location solutions based on

the remaining technologies, BLE and BLE. In this subsection we present the various alternatives and

choose one to use on the next layer.

Pozyx

The first solution we looked into was Pozyx1. This product uses UWB and has two main components:

the anchors and the tags. The anchors are modules with a fixed position, which are used to build a1Pozyx: https://www.pozyx.io/ accessed January 2018

35

coordinate system, they act like satellites in a GPS.

It has the ability to perform three different types of positioning. The 2D positioning where the anchors

must be placed in the same horizontal plane. The semi-3D which requires the height of the tag to be

supplied and fixed and finally the 3D positioning, where the height is variable but requires a minimum

of four anchors to be accurate.

The trackable components of this system are the tags represented in Figure 5.1. In environments

where there are several tags to be tracked, one tag takes on the role of master receiving the infor-

mation provided by all the others, referred to as slaves.

Figure 5.1: Pozyx tag

Every slave-tag has the ability to send its coordinates to the master-tag. These coordinates are the

position of the tag in the coordinate referential created by the position of the anchors. The master-tag

then processes the information.

It is possible for the master tag to use either the Arduino or the Python libraries. As the name

suggests, using the Arduino library requires a connection to the computer via Arduino, this is more

complex but gives us greater control over the settings. The Python library, however, only requires a

µ-usb cable.

Pozyx devices have a few settings we can configure:

• Channel: The channel where the UWB is broadcast, the pozyx device can use 6 independent

UWB channels. Devices on different UWB channels cannot communicate and do not interfere

with each other. In general, lower frequencies (i.e., lower channel numbers) also result in an

increased communication range.

• Data bitrate: This sets the UWB bitrate. Three settings are possible: 110kbit/sec, 850kbit/sec

and 6.81Mbit/sec. A higher bitrate will result is shorter messages and thus faster communica-

36

tion. However, this comes at the expense of a reduced operating range.

• Pulse Repetition Frequency: This sets the UWB pulse repetition frequency. There are three

possible settings: 16MHz or 64MHz. This settings has little effect on the communication rate.

However, on the same UWB channel, these two settings can live next to each other without

interfering.

• Preamble length: This sets the UWB preamble length. This setting has 8 different options:

4096, 2048, 1536, 1024, 512 , 256 , 128 , or 64 symbols. A shorter preamble length results in

shorter messages and thus faster communication. However, this again comes at the expense

of a reduced operating range.

• Tx Gain: This setting configures the UWB transmit gain. A larger gain will result in more

transmission power and a larger range. Altering this setting can make the device fall out off

regulation.

However, it is important to understand that settings which increase range also drain more battery.

Saninudge

The second solution we investigated was provided by Saninudge2. Saninudge is a Danish company

focused on improving hand hygiene compliance based on measuring the current compliance of a

HW and nudging them when they forget to sanitize their hands.

Saninudge’s solution consists of two types of sensors, one placed above the patient’s bed and an-

other placed on the dispensers. The first type of sensor creates a bubble around the bed, called the

patient zone, the second one registers when sanitation occurs.

When this information is combined with the SaniId, a small coin sized tag that attaches to the HW’s

identification tag we can collect the compliance by HW.

Sensors used in Saninudge’s solution were already hospital tested, and the data provided was al-

ready being processed to allow easy calculation of HWs’ hand hygiene compliance.

Estimote Beacons

The last solution we researched was Estimote3. Estimote offers a variety of beacons for several

purposes using either UWB or BLE technology. The BLE ones are much cheaper than the UWB and

can be divided into two categories, location beacons, and proximity beacons.

Location beacons are proximity beacons with more capabilities. While proximity beacons, shown in

Figure 5.2, have a range of 70m, operate in the 2.4GHz to 2.48GHz frequency range, have both

2Saninudge: https://www.saninudge.com/ accessed February 20183Estimote: https://estimote.com/ accessed June 2018

37

motion and temperature sensors, the location beacons have an increased range of 200m and in

addition to the sensors present in the first have both ambient light and pressure sensors and a

magnetometer.

Additionally, only the location beacons can estimate the position of a subject in the room. Proximity

beacons can only sense if a subject is present in the vicinity or not.

Figure 5.2: Estimote Proximity Beacon

Based on the performance of each technology during our tests, detailed in Section 6.1 we decided

to use Estimote beacons as our indoor location hardware.

5.2 Indoor Location Software

We needed to develop an app capable of detecting the signals from the beacons and transform them to

the type of data we wanted to collect: entrance in the patient’s zone and use of hand solution dispensers.

Each beacon has a tag associated making it easier to create groups of beacons with the same

functions, and the events are specific for each tag. There are also different settings we can alter in each

beacon, the strength of the signal the advertising interval and the type of packet transmitted: Eddystone,

iBeacon or Estimote Monitoring.

Eddystone4 is an open beacon format developed by Google and designed with transparency and

robustness in mind, it can be detected by both Android and iOS devices. iBeacon5 created by Apple

provides location awareness, and interactivity between iOS devices and iBeacon hardware. Finally

4Eddystone: https://developers.google.com/beacons/eddystone accessed in June 20185iBeacon: https://developer.apple.com/ibeacon/ accessed in June 2018

38

Estimote Monitoring6 is a proximity monitoring algorithm,developed specificaly for Estimote beacons,

that registers enter and exit events to a defined area. It relies on the set of Estimote propriety packets

optimized for reliable monitoring.

We chose to use Estimote Monitoring for the development of our system as it was specifically devel-

oped to work with the Estimote beacons and had an Application Program Interface (API) ready to detect

certain types of events.

Estimote provides a Software Development Kit (SDK) both in Android and iOS to communicate with

the beacons. Independently of the system used in the device the tasks to performed remained the same

as well as the events triggered. There are 3 types of events a device can detect:

• onEnter: when the device enters beacon range.

• onExit: when the device exits beacon range.

• onContextChange: when the number of detected beacons sharing the same tag changes.

The implementation of the onContextChange event means it is possible to detect that we are in

range of more than one beacon. This results in the onEnter event being triggered when the first beacon

is detected and while we don’t leave all the beacons’ range all the changes on the number of beacons

in range trigger an onContextChange event.

For our software we used the tag ”bed” associated with the beacons position at the top of the bed

on the wall, responsible for creating the patient’s bubble. We also established the range at which the

events are triggered, this range is an approximation because there are multiple factors which alter the

propagation properties of the signal.

Although we could have a tag associated with dispensers, we chose not to. Because, when the

device enters the onEnter event code block for a tag, it stays there until the onExit event is triggered

and is not able to scan for different tags. This behaviour would pose a problem if a dispenser was too

close to a bed. The program would be waiting for the onExit event and would not be able to detect the

dispenser’s use. This is illustrated in Figure 5.3, as soon as a bed is detected, the flow is directed to

the right side of the diagram and waits there until the device leaves the patient’s bubble. If we use a

dispenser there is no scan being performed and therefore this action is not detected.

Addressing the challenge of detection of the dispenser’s use while in bed range, required us to take

advantage of two other beacon properties. The first was the Estimote Telemetry packets7. These pack-

ets broadcast several informations about the beacon, the ones we needed were motionState, describes

if the beacon is moving or not, and currentMotionStateDurationInSeconds gives us the seconds the

beacon was moving for.6Estimote Monitoring: https://community.estimote.com/hc/en-us/articles/360003252832-What-is-Estimote-Monitoring-

accessed in June 20187Estimote Telemetry: https://developer.estimote.com/sensors/estimote-telemetry/ accessed in June 2018

39

Run bed'sonEnter code

block

bed dispenser Which tag wasdetected?

yes onExit triggered?

no

Run bed's onExitcode block

Wait for tagdetection

Run dispenser'sonEnter code

block

yesonExit triggered?

no

Run dispenser'sonExit code block

Figure 5.3: Flowchart of the program flow if the dispenser tag was used

The second was the broadcast only when in motion mode, meaning the beacon only broadcasts

when it’s in motion. Having this broadcasting mode active allowed us to ensure the the Telemetry packet

was received by the device and the corresponding actions were executed as an interruption without

interfering on the tag events. Otherwise, the device would be continuously receiving this packet stating

the device was either in motion or stopped, blocking the scan of the bed areas.

Finally the last task was gathering and storing the data. Storing was done on a database. Gathering

the data was done as illustrated in Figure 5.4.

The first step was to start the app with 0 beds. Then we waited for the first bed to be detected and

trigger the onEnter event. Following that, we store the time of the entrance on the patient’s bubble.

The app then enters a loop on the onContextChange event. In the simplest of cases the context

changes for 0 beds in range, leading to the trigger of the onExit event and the storing of the exit time of

that particular bed.

On more complex cases up to 2 beds can be in range. Either because one signal was wrongly picked

up by the app or because the healthcare worker is actually in range of 2 beds.

If the app detects a second bed it registers the entering time stamp of that second bed. Then, on the

next trigger of the onContextChange event, where only one bed stayed in range, it compares the ids of

the beds to register the exit time of the corresponding one.

This process is repeated until finally there are no more beds in range and the onExit event is triggered

storing the exit time of the last remaining bed.

As we’ve previously stated the use of the dispenser is registered automatically and therefore was not

40

Trigger onEnterevent

Register entrance on first bed area

Register last bed exit

Trigger onExitevent

TriggeronContextChange

event

2

2

1

How many beds incontext?

0

1 2How many beds in range?

2

Register 2nd bed entry

Update context to0 beds

YesFirst bed detected?

No

App started

No

YesApp terminated? App terminated

How many beds incontext?

Update context to1 bed

Register exit oncorresponding bed

Update context to 2 beds

1

Figure 5.4: Indoor App Flowchart

represented. Also, the filtering of the relevant visits to the beds will be done by the AMS.

5.3 Automated Monitoring System

Our approach was modeled as shown in Figure 5.5 and was based on the work of Marques, illustrated

in Figure 11 of ”Using Gamification for Reducing Infections in Hospitals” [8] that proposed a general

approach to the measuring of hand hygiene compliance. The resulting algorithm is adapted to the use

of Estimote beacons, there is no longer the necessity to detect the entrance in a room and the proximity

to the sink.

41

Yes

No Difference between entrance and exit

> 2s

Shift started

No

YesShift ended?

No

Yes

Patient's bubbleapproached?

Register opportunity

ABHR/Sink usedBed approachedWhat was the previous

event detected?

Complied opportunity

No YesMore than 30s passed?

Non-complied opportunityNon-complied

opportunity

Shift ended

Figure 5.5: Automated Monitoring System Flowchart

All the data provided by the indoor location software was stored in the database. The AMS would go

trough this data at a set interval of time.

The AMS ignored each visit to a patient’s bubble with a duration under 2s. We considered this interval

as a HW just passing by or a wrong signal detection.

If the duration of the visit was considered valid the AMS registered the opportunity. Then, the AMS

checked what was the previous event detected. If it was a bed, same or other the HW had not complied

with the guidelines for hand hygiene. On the other hand, if the previous event was the use of a dispenser

the HW complied with the hand hygiene guidelines.

Since the dispenser for hand hygiene can be in the hallways far from the beds, we considered the

interval of 30s as the maximum time interval between the dispenser use and the patient’s bubble en-

42

trance. For any time interval greater than those 30s we considered the HW didn’t go directly from the

dispenser to the patient’s bubble.

All the time intervals should be adjusted on a case by case basis and further real world testing is

required.

5.4 Gamification Platform

After the AMS layer we needed to focus on the gamification of the metrics resultant of AMS calculations.

The OSYRISH was a project aimed at tackling the hand hygiene lack of compliance by HWs. This

gamified system had several gamification elements and functionalities. The platform had badges and

rankings based on performance as well as a levelling system, to reward the HWs and motivate them.

It also allowed users to customize their appearance and to interact with each other creating a social

component. There was also a common objective to the unit to encourage teamwork and inter unit

competition. Another interesting feature was the possibility for HWs to visualize their hand hygiene

performance across time.

Unfortunately, this solution was not tested in the real world. This happened due to the inability of

having an functional AMS, mainly due to difficulties related with the on the indoor location systems [43].

The platform had the advantage of being developed with the collaboration of HWs. Understanding

the target users is key in motivating the desired behaviours. Also, the fact that there were several

attempts at testing indoor location solutions in an healthcare facility provided us with valuable insights to

the problems of developing this type of solution.

5.5 Gamification System Interface

The last layer of our model was the interface of the gamification platform.

The interface is a very important part of the solution as it is the way the information is displayed to

the users.

In the OSYRISH project however, the interface was never available. To understand if information is

displayed in an intuitive way we needed to test the interface, the results are presented in Section 4.2.1.

43

44

6Evaluation

Contents

6.1 Indoor location Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.2 Indoor Location Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6.3 Automated Monitoring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

6.4 Gamification System Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

45

46

This chapter corresponds to the evaluation step of the DSRM. We present the detailed testing

process of each layer and the respective results. By evaluating each layer we evaluate the solution’s

model itself. Since the gamification platform was already evaluated [8] we did not conduct tests for that

layer.

6.1 Indoor location Hardware

In this section we evaluate each of the different choices we considered for the sensors component.

6.1.1 Pozyx

Pozyx has several settings that should be configured to better suit our needs, in this subsection we test

these settings to understand the potential of this technology in our solution.

6.1.1.A Test Set-up

Firstly we positioned 4 anchors in a room, as shown in Figure 6.1, according to the rules described in

the Pozyx website:

1. Place the anchors high and in line-of-sight of the user.

2. Spread the anchors around the user. Never on a straight line!

3. Place the anchors vertically with the antenna at the top.

4. For 3D positioning: place the anchors at different heights.

Figure 6.1: Deployed Anchors

We then connected the master-tag to a computer running the Pozyx platform. This master-tag is the

link between the anchors, the slave-tags and the platform. After this we measured each anchor’s height

47

and registered this data in the platform. Next, we used the auto calibration function to both ensure all

the anchors could successfully detect at least 2 neighbours, and they could find the distance between

them.

The software also allowed us to upload a schematic of the room and overlay the location of the

anchors in order to facilitate information extraction.

The final step of the set-up was to have a predefined route, shown in Figure 6.2 to compare against

the obtained route in order to determine how accurate the Pozyx system was.

Figure 6.2: Predefined Route

48

6.1.1.B Testing and Results

We conducted several tests with different parameters to understand how the different settings influence

the accuracy and battery consumption.

For the first test, Figure 6.3, we kept the default settings:

• Channel: 5

• Data bitrate: 110kbps

• Pulse Repetition Frequency: 64MHz

• Preamble length: 1024 symbols

• Tx Gain: 11.5dB

Figure 6.3: Test results with default settings

As it can be seen from Figure 6.3 the resulting route was full of noise. It was very difficult to extract

valid information about the real position of the slave-tag with these settings.

For this reason we needed to apply a filter, filters are algorithms capable of reducing the noise of a

route. There were three different filters:

• Low Pass: reduces high frequency jittering.

• Moving Average: allows for a smoother trajectory.

• Moving Median: filters out most outliers.

We can also change the strength of the filter between 0 and 10. The higher the value the less jittery

the route is however, the time required for processing information increases.

After running a test for each filter with the strength setting at 5 and the default values we obtained

the results in Figure 6.4.

49

(a) Low pass filter (b) Moving average filter

(c) Moving median filter

Figure 6.4: Test results with default values plus filtering

By comparing these results with the predefined route, Figure 6.2, the positioning seemed quite ac-

curate, we then tested the autonomy.

Testing for autonomy consisted of having the slave-tag in the default settings and connected to a

9V battery via an adaptor as seen in Figure 6.5. We then proceeded to measure how long the battery

lasted.

The battery lasted for approximately 2h 30min. These results were considered non satisfactory for

our objective, as the process of changing batteries every 2h 30min would be invasive and costly.

To improve this we tested the accuracy with a lower range capacity. This was accomplished by

changing the preamble length to 256 symbols. These tests build on those of the previous ones, as we

applied all the three filters with the same strength.

The results of testing with this new preamble length are displayed in Figure 6.6. By comparing these

results with the ones in Figure 6.4 and the predefined route, Figure 6.2 no significant variance seemed

to be observed.

We then tested the autonomy of the slave-tag in an analogue way. It seemed no significant improve-

50

Figure 6.5: Pozyx tag with 9V battery

ment was observed.

After contacting Pozyx, we were informed that the best solution was to use a power bank to power

the slave-tag. This posed a problem due to the dimensions and weight of the power bank that the HWs

would have to carry in their pockets.

6.1.2 Saninudge

After the first contacts ad meetings Saninudge technology seemed to be a good fit for our problem, as

it had already been tested in healthcare environments. After we described the project of implementing

a gamification based solution to measure HWs’ hand hygiene compliance using the AMS developed by

them Saninudge was interested but there was a need to sort the bureaucratic aspect.

After the signed protocol, and wile we waited for the sensors, Saninudge provided us with the nec-

essary documentation to integrate their technology with our system. The process behind the integration

of the Saninudge’s AMS data and our system cannot be disclosed due to confidentiality policies.

With the integration ready, we inquired Saninudge about the status of the sensors’ arrival. Saninudge

assured us the sensors would be arriving a few weeks later. Unfortunately, not much time had passed

when we received an email informing that Saninudge was no longer available to aid in our project.

6.1.3 Estimote

Estimote provides an out of the box app to test its beacons. We labelled 3 beacons as desk1, desk2

and desk3 and gave each of them a owner attribute: owner1, owner2 and owner3 respectively. Then

we spread the beacons 2m apart in a small room with no great interferences.

Finally we used an Android device running the Estimote proximity app and as we approached each

beacon the app would display the message welcome to ownerx desk or welcome to ownerx, ownery

desk.

This allowed us to understand how the beacons worked and how these proximity zones and beacons

51

(a) Low pass filter (b) Moving average filter

(c) Moving median filter

Figure 6.6: Test results with preamble length 256 symbols plus filtering

could be used as part of our solution. Although we did not alter the range on the out of the box app 1m,

the proximity zones would detect us when we walked near that range and the message would change

promptly.

6.1.4 Comparing Commercial Solutions

Pozyx was the first technology we tested. The idea behind a solution based on the Pozyx technology was

as follows: the HW would carry around a slave-tag, this tag would transmit its position to the master-tag

which in turn would store the data for further processing.

Estimote technology had already been tested, middle 2016, for similar solution [8] with poor results.

However, since then both the hardware and the SDK suffered major changes.

This solution would be based on proximity. The HW would carry a device with BLE capabilities and

this device would be able to detect if it was near a bed or a dispenser being used and store the data for

further processing.

Estimote also provides an SDK so programmers can develop applications to communicate with the

52

beacons. These applications can be developed either in iOS or Android and can obtain the information

broadcast by the beacons, process it and then carry out determined actions based on the received

information. The results of comparing both solutions is summarized in Table 6.1.

Table 6.1: Commercial Solutions Comparison

Pozyx EstimoteLocation Technology UWB BLELocation Method Triangulation Proximity

Pricing

4 anchors + 5 tags: 999C

extra tag/anchor: 135C

extra anchor:

4 beacons: 99$(device to run the software purchased separately)

Advantages Very accurateBeacons have own power source lasting over a year

App running in a cellphone (small and with internal power source)

DisadvantagesTags are power bank dependent

Anchors need to be connected to a power sourceSometimes inconsistent/delayed detection

Proximity can only tell if the subject is in the vicinity while triangulation can calculate the exact position

of the subject. This means that, while the Poxyz tags only allowed us to check if a HW was near a

dispenser, Estimote’s Telemetry packet allows us to know a dispenser is actually being used.

Proximity based technology also makes the developing of the indoor location software easier. If a

bed is moved and the beacon goes with it the patient zone stays the same. With Pozyx we would need

to set the new coordinates for the bed each time it moved.

Another factor we considered was the autonomy. The only way to power a Pozyx tag through an

entire 8h shift was to use a power bank. Using this solution would increase the weight and dimension of

the objects in the HW’s pockets. This was considered too invasive as the HWs had already expressed

concerns about using the smaller version.

The mobile phones used in an Estimote based solution already have a battery source capable of

lasting for more than one shift and the devices are more resistant than a Pozyx tag.

Another problem with Pozyx was the need for the anchors to be connected to an outlet, this would

require cables to be run from the power outlet to the anchor making this solution even more invasive.

Estimote beacons on the other hand, can last for over a year depending on the power of transmission

and advertising interval.

6.2 Indoor Location Software

Testing the indoor location software to extract accurate information about the positioning of the moni-

toring device (mobile phone) required us to position the beacons as if they were being deployed in the

healthcare facility.

We placed two beacons on the wall 3m apart. This was the distance between the center of the beds

53

on the healthcare facility were the solution would be deployed. We also delimited the bed area as 1m

wide and 2m length as it was the dimensions of the patient’s bubble at said healthcare facility.

To simulate the dispensers we used a dispenser with a beacon attached on top as seen in Figure 6.7.

The dispensers at an healthcare facility, shown in Figure 6.8, are European style. Despite the differences

between the dispensers at the healthcare facility and the test dispensers, what were interested in was

the pumping motion the HW performs when extracting the solution for hand hygiene.

Figure 6.7: Test Dispenser

Figure 6.8: European Style Dispenser

We then developed an Android app with onEnter and onExit events implemented just to display an

entering bed and exiting bed message respectively. We set the custom range to trigger these events at

2.0m (length of the bed) and began to approach the first beacon.

Unfortunately at 4m the onEnter event was already being triggered. We repeated the test with the

custom range set to 1m. The triggering of the events was inconsistent. The onEnter event would trigger

at 1m and if we moved to the side of the beacon, still under 1m, the onExit event would trigger.

We then decided to try iOS, so we repeated the test with the custom range at 2m. The results were

54

satisfactory as the device was successfully detecting the entering and exit of the patient’s bubble.

We then moved to testing when the HW was between two beds. We used an improved version of the

app with the onContext event implemented. The device showed correctly the beds in range. However,

sometimes it took a few seconds to recognize those changes.

Moving on to the next phase we performed the exact same test, except this time we used a neck

pouch. There seemed to be no difference in the results. This meant the HWs could carry the cellphone

with little interference with their work.

Finally, we moved into testing the dispenser broadcasting its telemetry data and set on transmitting

only when in motion. Firstly, we used the dispenser out of the range of the beds but in range of the

device. The device picked up the use of the dispenser.

Then, we used the dispenser while we were in range of the beds and the device successfully picked

up on the use of the dispenser while maintaining the correct bed area.

In conclusion the technology seemed to perform according to what was expected, detecting enter

and exit events at appropriate times, although sometimes with a small delay and correctly identified the

use of the dispensers. However, further testing in the real world should be performed .

6.3 Automated Monitoring System

The AMS is a script capable of extracting metrics from data collected by the indoor location software.

The tests were run based on data collected using the indoor location software. We ran the AMS over

the data collected and compared the compliance cases observed with the ones obtained by the AMS.

The comparison from between the two gave an efficacy of 100%. However, the tests were performed

in a controlled environment with little to no interference and further real world testing is required.

6.4 Gamification System Interface

In this section we describe how we performed the user tests and analyse the results.

Before carrying out the user tests it’s recommended to perform pilot tests [40]. These tests serve the

purpose of detecting there is no confusion about what to do, if all the necessary information is provided

in a clear way and if the time is appropriate.

The pilot tests were conducted with several students, different from the target of the actual testing.

After everything was approved we carried out the user tests.

55

6.4.1 User Tests Set-up

We performed all the tests in the same meeting room over two different days, according to the partici-

pants availability. Each test took about 15 minutes, 5 minutes to explain the project and an additional 10

minutes of actually testing. Only one user carried out the test at a time.

The test was performed in a laptop running the gamification platform with a previously populated

database containing fictitious data of healthcare professionals, levels and badges.

All the users started in the home screen, after login. The test profile was the same for every user

and was reset between tests.

The tests were supervised by one researcher and the users were students with ages between 22

and 25.

The tasks to be carried out were part of the testing script displayed in Appendix A.2 and they were:

1. View the game objectives;

2. Check the requirements to obtain the badge: ”Vedeta de Turno” nıvel 2;

3. Change the profile description to ”Hello!”;

4. View the percentage of hand hygiene compliance on 2015/09/29;

5. Edit the profile to not display the number of badges.

These tasks considered successful only when either the user pointed out the task was finished or

read out the required information.

After each test we also asked every participant to answer a small questionnaire regarding the plat-

form usability. This form was fully based on the System Usability Scale (SUS) as it the standard when

mesuring a system’s usability [44].

The SUS consists of 10 statements where the user uses a scale from 1 to 5 ( 1 meaning that the

user strongly disagrees and 5 meaning the user strongly agrees with the statement).

6.4.2 User Tests Results

In Table 6.2 we can see the interval of number of clicks a user takes to complete each task with a

confidence of 95%.

Table 6.2: User tests number of clicks by task (confidence level of 95%)

Tasks1 2 3 4 5

Confidence interval [3 - 5] [3 - 8] [11 - 21] [11 - 16] [4 - 6]Optimal number of clicks 2 2 6 3 4

56

Comparing these intervals to the optimal number of clicks needed to accomplish each task we can

see a big difference for tasks 3 and 4. This was consistent with the values for the times, as these were

the tasks with the highest values. During the tests both tasks 3 and 4 were also the tasks were the users

gave the highest number of suggestions for improvement.

The main problem detected in both of these tasks was described, by the users, as not intuitive

interface design.

Some suggestions were made to improve these tasks. Regarding task 3 the main suggestions were

to allow the profile definitions to be accessed by clicking the icon on the top right corner. Another, was

to allow the edit of description on the description box itself.

The last most given suggestion was to have a check box to manage the display of profile information.

This was to avoid confusion between the alteration of the content of the profile information and the

display preferences of said information.

Regarding task 4, users suggested to have a date selector plug-in instead of having to write the date,

or remove the icon to the plug-in selector. Another one was having a different option if we wanted to

view information for just one single day.

Regarding the SUS we observed an average score of 70.38, this would result in a C score which

meant that the interface was just acceptable when considering the work of Bangor et al [45].

Having an interface with a just acceptable score ias a problem for a gamification based solution as

the main way of communicating the metrics, engaging the HWs and change their behaviour would be

this interface.

57

58

7Conclusion

Contents

7.1 Lessons Learned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

7.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

7.4 Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

7.5 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

59

60

This chapter provides a summary of the worked performed during this thesis. We reflect on the

lessons learned and the possible limitations. We list the contributions and finish by providing future work

suggestions.

This thesis built on the work of the OSYRISH project, we started by addressing the problem of HAIs

and their negative impact on both patients and healthcare facilities. HAIs’ impact could be minimized if

the HWs complied with the guidelines for hand hygiene by the WHO.

Monitoring this compliance can be an expensive process as direct observation is still the golden

standard. This compels healthcare facilities to invent new and innovative ways of performing hand

hygiene monitoring.

Learning from the difficulties of implementing a gamification solution to monitor hand hygiene com-

pliance and provide feedback to the HWs, faced in the OSYRISH project, we developed a model to

structure the development of this solution and tackle these challenges.

By following the steps of the model, we developed a solution however, we were not able to deploy

the system in the real world, because it seems there is still no indoor location technology suitable for

solving this problem.

7.1 Lessons Learned

From the first steps of the solution’s development we understood that it’s very difficult to perform indoor

location. Finding a indoor location technology that is both accurate cheap and has good autonomy is a

challenge.

By compromising and finding a balance between the parameters, and making sure it still fitted our

needs we were able to develop a pilot that gave us very good results in the tests.

Developing a solution in an incremental manner allowed us to have a better grasp on the specificities

of the bottom layers and make sure the choices we made for the top layers fitted the solution in the best

way possible.

7.2 Contributions

During this thesis work there were several contributions. The main one is the model we applied during

the development of the solution.

Another contribution was the way we used the latest Estimote sensors and API to create an iOS

app to successfully monitor the HW’s approach of the patient’s bubble and dispenser use as well as the

development of the AMS to monitor the HWs hand hygiene compliance based on the collected data.

61

Likewise the integration of this AMS system with the OSYRISH and all the testing done in terms of

the choices and use of different technologies used are contributions in the process of development of a

system capable of monitoring HWs’ compliance with hand hygiene guidelines.

Finally, the user testing of the gamification platform’s interface contributed to the understanding of

possible problems and the development of solutions to be implemented in the future.

7.3 Limitations

The main limitation the AMS suffers from is the fact that it is only capable of monitoring moments 1, 4

and 5 of the WHO’s framework for hand hygiene. However, these moments correspond to 80% of the

moments for hand hygiene [46].

Unfortunately, we were not able to test in a real world environment. This means we had no way of

accessing the accuracy of the system after deployment despite the promising lab results.

Saninudge pulling out of the project after a few months work had been done and after promising the

sensors was unfortunate. The technology had great chances of working as it had already been tested for

the measuring of hand hygiene compliance levels in an healthcare environment with good results. With

this technology integrated with the platform the real world testing would have been a realistic possibility.

7.4 Communications

Although there was no opportunity to submit papers during the course of this thesis, a paper is cur-

rently being prepared for submission to the International Journal of Distributed Sensor Networks, which

addresses the main points of the thesis: the model to develop a gamification based solution for hand

hygigene compliance and the use of indoor location to measure this compliance.

Also, the dissertation discussion serves as a mean of communicating the results of our work.

7.5 Future Work

We were able to successfully monitor the and hygiene compliance in a laboratory setting. Future work

might include testing of the developed AMS in an healthcare facility.

Finally, if the AMS proves successful, in the healthcare facility environment a full deploy of the gam-

ification solution should provide us with enough data to understand if this is a viable path towards im-

proving HWs’s compliance rate and a step forward in reducing HAI in healthcare.

62

—————————————————————————–

63

Bibliography

[1] K. Peffers, T. Tuunanen, M. A. Rothenberger, and S. Chatterjee, “A Design Science Research

Methodology for Information Systems Research,” Journal of Management Information Systems,

vol. 24, no. 3, pp. 45–77, 2007.

[2] European Centre for Disease Prevention and Control, “Observed and

predicted prevalence of HAIs and antimicrobial use.” [Online]. Avail-

able: https://ecdc.europa.eu/en/healthcare-associated-infections-acute-care-hospitals/database/

prevalence-hais-and-antimicrobial-use/observed

[3] D. Pittet, B. Allegranzi, and J. Boyce, “The World Health Organization Guidelines on

Hand Hygiene in Health Care and Their Consensus Recommendations,” Infection Control

& Hospital Epidemiology, vol. 30, no. 07, pp. 611–622, 2009. [Online]. Available: https:

//www.cambridge.org/core/product/identifier/S019594170003722X/type/journal article

[4] K. Werbach and D. Hunter, For the win: How game thinking can revolutionize your business. Whar-

ton Digital Press, 2012.

[5] N. Safdar, D. J. Anderson, B. I. Braun, P. Carling, S. Cohen, C. Donskey, M. Drees,

A. Harris, D. K. Henderson, S. S. Huang, M. Juthani-Mehta, E. Lautenbach, D. R.

Linkin, J. Meddings, L. G. Miller, A. Milstone, D. Morgan, S. Sengupta, M. Varman,

D. Yokoe, and D. M. Zerr, “The Evolving Landscape of Healthcare-Associated Infections:

Recent Advances in Prevention and a Road Map for Research,” Infection Control &

Hospital Epidemiology, vol. 35, no. 05, pp. 480–493, may 2014. [Online]. Available:

https://www.cambridge.org/core/product/identifier/S0899823X00191676/type/journal article

[6] W. World Health Organization, “Report on the Burden of Endemic Health Care-Associated Infection

Worldwide.” WHO Library Cataloguing-in-Publication Data, p. 40, 2011.

[7] M. R. Chassin, C. Mayer, and K. Nether, “Improving Hand Hygiene at Eight Hospitals in the United

States by Targeting Specific Causes of Noncompliance,” Journal on Quality and Patient Safety,

vol. 41, no. 1, pp. 4–12, 2015.

64

[8] R. Marques, “Using Gamification for Reducing Infections in Hospitals,” no. April, 2016.

[9] A. R. Hevner, S. T. March, J. Park, and S. Ram, “Design Science in Information

Systems Research,” MIS Quarterly, vol. 28, no. 1, pp. 75–105, 2004. [Online]. Available:

http://dl.acm.org/citation.cfm?id=2017217

[10] B. Allegranzi, S. B. Nejad, C. Combescure, W. Graafmans, H. Attar, L. Donaldson, and

D. Pittet, “Burden of endemic health-care-associated infection in developing countries: systematic

review and meta-analysis,” The Lancet, vol. 377, no. 9761, pp. 228–241, jan 2011. [Online].

Available: http://dx.doi.org/10.1016/S0140-6736(10)61458-4http://linkinghub.elsevier.com/retrieve/

pii/S0140673610614584

[11] “European Centre for Disease Prevention and Control. Point prevalence survey of healthcare- as-

sociated infections and antimicrobial use in European acute care hospitals.” Stockholm:, 2013.

[12] European Centre for Disease Prevention and Control, “Healthcare-associated infections

acquired in intensive care units - Annual Epidemiological Report for 2016,” ECDC,

Stockholm, Tech. Rep., 2018. [Online]. Available: https://ecdc.europa.eu/en/publications-data/

healthcare-associated-infections-acquired-intensive-care-units-annual-0%0Ahttps://ecdc.

europa.eu/sites/portal/files/documents/AER for 2015-healthcare-associated-infections.pdf

[13] D. Pittet, B. Allegranzi, H. Sax, S. Dharan, C. L. Pessoa-Silva, L. Donaldson, and J. M. Boyce,

“Evidence-based model for hand transmission during patient care and the role of improved prac-

tices,” Lancet Infectious Diseases, vol. 6, no. 10, pp. 641–652, 2006.

[14] B. Allegranzi and D. Pittet, “Role of hand hygiene in healthcare-associated infection prevention,”

Journal of Hospital Infection, vol. 73, no. 4, pp. 305–315, dec 2009. [Online]. Available: http://dx.

doi.org/10.1016/j.jhin.2009.04.019http://linkinghub.elsevier.com/retrieve/pii/S0195670109001868

[15] J. M. Boyce, “Hand hygiene compliance monitoring: current perspectives from the USA,”

Journal of Hospital Infection, vol. 70, no. SUPPL. 1, pp. 2–7, 2008. [Online]. Available:

http://dx.doi.org/10.1016/S0195-6701(08)60003-1

[16] A. Allam, Z. Kostova, K. Nakamoto, and P. J. Schulz, “The effect of social support features and

gamification on a web-based intervention for rheumatoid arthritis patients: Randomized controlled

trial,” Journal of Medical Internet Research, vol. 17, no. 1, p. e14, 2015.

[17] W. Zingg, A. Holmes, M. Dettenkofer, T. Goetting, F. Secci, L. Clack, B. Allegranzi, A. P. Magiorakos,

D. Pittet, Y. Carmeli, A. Dittrich, W. Ebner, R. Edwards, E. Ferlie, P. Gastmeier, W. Hryniewicz,

S. Kalenic, C. Kilpatrick, N. Sorknes, E. Szilagyi, R. Vatcheva-Dobrevska, and C. Vincent, “Hospital

organisation, management, and structure for prevention of health-care-associated infection: A

65

systematic review and expert consensus,” The Lancet Infectious Diseases, vol. 15, no. 2, pp.

212–224, 2015. [Online]. Available: http://dx.doi.org/10.1016/S1473-3099(14)70854-0

[18] S. Hagel, J. Reischke, M. Kesselmeier, J. Winning, P. Gastmeier, F. M. Brunkhorst,

A. Scherag, and M. W. Pletz, “Quantifying the Hawthorne Effect in Hand Hygiene Compliance

Through Comparing Direct Observation With Automated Hand Hygiene Monitoring,” Infection

Control & Hospital Epidemiology, vol. 36, no. 08, pp. 957–962, 2015. [Online]. Available:

https://www.cambridge.org/core/product/identifier/S0899823X15000938/type/journal article

[19] N. Contzen, S. De Pasquale, and H. J. Mosler, “Over-reporting in handwashing self-reports: Po-

tential explanatory factors and alternative measurements,” PLoS ONE, vol. 10, no. 8, pp. 1–22,

2015.

[20] M. A. Ward, M. L. Schweizer, P. M. Polgreen, K. Gupta, H. S. Reisinger, and E. N. Perencevich,

“Automated and electronically assisted hand hygiene monitoring systems: A systematic review,”

American Journal of Infection Control, vol. 42, no. 5, pp. 472–478, 2014. [Online]. Available:

http://dx.doi.org/10.1016/j.ajic.2014.01.002

[21] J. M. Al Salman, S. Hani, N. de Marcellis-Warin, and S. Fatima Isa, “Effectiveness of an

electronic hand hygiene monitoring system on healthcare workers’ compliance to guidelines,”

Journal of Infection and Public Health, vol. 8, no. 2, pp. 117–126, 2015. [Online]. Available:

http://dx.doi.org/10.1016/j.jiph.2014.07.019

[22] S. J. Storey, G. FitzGerald, G. Moore, E. Knights, S. Atkinson, S. Smith, O. Freeman, P. Cryer,

and A. P. Wilson, “Effect of a contact monitoring system with immediate visual feedback on hand

hygiene compliance,” Journal of Hospital Infection, vol. 88, no. 2, pp. 84–88, 2014. [Online].

Available: http://dx.doi.org/10.1016/j.jhin.2014.06.014

[23] H. Soganci, S. Gezici, and H. V. Poor, “Accurate positioning in ultra-wideband systems,” IEEE

Wireless Communications, vol. 18, no. 2, pp. 19–27, 2011.

[24] I. Amundson and X. Koutsoukos, “A survey on localization for mobile wireless sensor networks,”

Mobile entity localization and tracking in GPS-less environnments, pp. 235–254, 2009.

[25] G. B. Jeffrey Hightower, “A Survey and Taxonomy of Location Systems for Ubiquitous Computing,”

IEEE Computer, vol. 34, pp. 57–66, 2001. [Online]. Available: http://citeseerx.ist.psu.edu/viewdoc/

summary?doi=10.1.1.12.1169

[26] U. Ahmad, A. V. Gavrilov, S. Lee, and Y.-K. Lee, “A modular classification model for received signal

strength based location systems,” Neurocomputing, vol. 71, no. 13-15, pp. 2657–2669, 2008.

66

[27] A. M. Ladd, K. E. Bekris, A. Rudys, L. E. Kavraki, and D. S. Wallach, “On the feasibility of Using

Wireless Ethernet for Indoor Localization,” IEEE Transactions on Robotics and Automation (TRA),

vol. 20, no. 3, pp. 555–559, 2004.

[28] X. Zhao, Z. Xiao, A. Markham, N. Trigoni, and Y. Ren, “Does BTLE measure up against WiFi? A

comparison of indoor location performance,” 20th European Wireless Conference, pp. 1–6, 2014.

[29] Y. Chen, D. Lymberopoulos, J. Liu, and B. Priyantha, “FM-based indoor localization,” Proceedings of

the 10th international conference on Mobile systems, applications, and services - MobiSys ’12, no.

September, p. 169, 2012. [Online]. Available: http://dl.acm.org/citation.cfm?doid=2307636.2307653

[30] S. Venkatesh and R. Buehrer, “Non-line-of-sight identification in ultra-wideband systems based on

received signal statistics,” European Space Agency, (Special Publication) ESA SP, vol. 626 SP,

no. 6, pp. 1120–1130, 2007.

[31] R. Tesoriero, J. Gallud, M. Lozano, and V. Penichet, “Using active and passive RFID technology

to support indoor location-aware systems,” IEEE Transactions on Consumer Electronics, vol. 54,

no. 2, pp. 578–583, 2008.

[32] N. Petrellis, N. Konofaos, and G. P. Alexiou, “Target localization utilizing the success rate in infrared

pattern recognition,” IEEE Sensors Journal, vol. 6, no. 5, pp. 1355–1364, 2006.

[33] R. Casas, D. Cuartielles, A. Marco, H. J. Gracia, and J. L. Falco, “Hidden issues in deploying an

indoor location system,” IEEE Pervasive Computing, vol. 6, no. 2, pp. 62–69, 2007.

[34] J. Hamari, J. Koivisto, and H. Sarsa, “Does gamification work?–a literature review of empirical

studies on gamification,” System Sciences (HICSS), 2014 47th Hawaii International Conference on,

pp. 3025–3034, 2014. [Online]. Available: http://ieeexplore.ieee.org/xpls/abs all.jsp?arnumber=

6758978

[35] S. Deterding, D. Dixon, R. Khaled, and L. Nacke, “From game design elements to

gamefulness,” Proceedings of the 15th International Academic MindTrek Conference on

Envisioning Future Media Environments - MindTrek ’11, p. 9, 2011. [Online]. Available:

http://dl.acm.org/citation.cfm?doid=2181037.2181040

[36] K. Robson, K. Plangger, J. H. Kietzmann, I. McCarthy, and L. Pitt, “Is it all a game? Understanding

the principles of gamification,” Business Horizons, vol. 58, no. 4, pp. 411–420, 2015. [Online].

Available: http://dx.doi.org/10.1016/j.bushor.2015.03.006

[37] ——, “Game on: Engaging customers and employees through gamification,” Business Horizons,

vol. 59, no. 1, pp. 29–36, 2016.

67

[38] L. C. Lamb, M. M. DiFiori, V. Jayaraman, B. D. Shames, and J. M. Feeney, “Gamified Twitter

Microblogging to Support Resident Preparation for the American Board of Surgery In-Service

Training Examination,” Journal of Surgical Education, vol. 74, no. 6, pp. 986–991, 2017. [Online].

Available: http://dx.doi.org/10.1016/j.jsurg.2017.05.010

[39] R. Marques, J. Gregorio, F. Pinheiro, P. Povoa, M. M. da Silva, and L. V. Lapao, “How can

information systems provide support to nurses’ hand hygiene performance? Using gamification

and indoor location to improve hand hygiene awareness and reduce hospital infections,” BMC

Medical Informatics and Decision Making, vol. 17, no. 1, p. 15, 2017. [Online]. Available:

http://bmcmedinformdecismak.biomedcentral.com/articles/10.1186/s12911-017-0410-z

[40] M. J. Fonseca, P. Campos, and D. Goncalves, Introducao ao Design de Interfaces, 2012.

[41] H. J. Yazici, “An exploratory analysis of hospital perspectives on real time information

requirements and perceived benefits of RFID technology for future adoption,” International

Journal of Information Management, vol. 34, no. 5, pp. 603–621, 2014. [Online]. Available:

http://dx.doi.org/10.1016/j.ijinfomgt.2014.04.010

[42] L. L. Pineles, D. J. Morgan, H. M. Limper, S. G. Weber, K. A. Thom, E. N. Perencevich,

A. D. Harris, and E. Landon, “Accuracy of a radiofrequency identification (RFID) badge

system to monitor hand hygiene behavior during routine clinical activities,” American

Journal of Infection Control, vol. 42, no. 2, pp. 144–147, feb 2014. [Online]. Available:

http://linkinghub.elsevier.com/retrieve/pii/S0196655313011504

[43] R. Marques, J. Gregorio, M. M. Da Silva, and L. V. Lapao, “The promise of the internet of things in

healthcare: How hard is it to keep?” Studies in Health Technology and Informatics, vol. 228, pp.

665–669, 2017.

[44] P. Kortum and S. C. Peres, “The Relationship Between System Effectiveness and Subjective Usabil-

ity Scores Using the System Usability Scale,” International Journal of Human-Computer Interaction,

vol. 30, no. 7, pp. 575–584, 2014.

[45] A. Bangor, P. T. Kortum, J. T. Miller, A. Bangor, P. T. Kortum, J. T. Miller, A. Empirical, A. Bangor, P. T.

Kortum, and J. T. Miller, “An Empirical Evaluation of the System Usability Scale Usability Scale,” vol.

7318, 2008.

[46] J. M. Boyce, “Measuring Healthcare Worker Hand Hygiene Activity: Current Practices and

Emerging Technologies,” Infection Control & Hospital Epidemiology, vol. 32, no. 10, pp. 1016–1028,

2011. [Online]. Available: https://www.cambridge.org/core/product/identifier/S0195941700040765/

type/journal article

68

AUser tests materials

69

A.1 Consent Form

Consentimento Informado para a realização de

testes com os utilizadores

Eu abaixo assinado, declaro que tomei conhecimento e autorizo a que no decorrer

dos testes com utilizadores no âmbito da tese de mestrado: “Gamification for Hand

Hygiene Compliance” as minhas interações com o sistema sejam gravadas. Consinto

ainda a tomada de notas sobre essas mesmas interações. Declaro também que

autorizo o uso desses dados, de forma anónima, para fins educativos e de

investigação, sem mais nenhum fim alternativo fora dos mencionados anteriormente.

(assinatura)

___________________________________________

70

A.2 Test Script

Teste de Interface:

Este teste tem como objectivo avaliar a usabilidade da interface do sistema

desenvolvido para melhorar a higienização das mãos por parte dos profissionais de

saúde usando gamification. O teste tem a duração aproximada de 10 min.

As tarefas devem ser realizadas pela ordem apresentada.

Por favor quando a tarefa estiver concluída indique ao responsável.

Não se esqueça que o que está a ser testado é o sistema, não o participante! Se

em algum momento optar por não realizar o teste ou alguma tarefa em particular tem

sempre essa opção.

As tarefas a realizar encontram-se descritas abaixo:

1. Consultar os objectivos do jogo;

2. Consultar os requisitos para ter o distintivo: “Vedeta do Turno” nível 2;

3. Alterar a descrição do perfil para: “Hello!”;

4. Consultar a percentagem de higienização no dia 2015/09/29

5. Editar o perfil para não mostrar o número de distintivos.

Obrigado pela sua participação!

71

72