EVIDENCE BASED RECOMMENDATIONS FOR NATIONAL HEALTHCARE- ASSOCIATED INFECTION SURVEILLANCE Philip L Russo BN, M.Clin.Epid. Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy School of Public Health and Social Work Faculty of Health Queensland University of Technology 2016
286
Embed
EVIDENCE BASED RECOMMENDATIONS FOR NATIONAL … › 100034 › 1 › Philip_Russo_Thesis.pdf · surveillance organisational websites and supporting resources, and through structured
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
EVIDENCE BASED RECOMMENDATIONSFOR NATIONAL HEALTHCARE-
ASSOCIATED INFECTION SURVEILLANCE
Philip L Russo BN, M.Clin.Epid.
Submitted in fulfilment of the requirements for the degree of
Doctor of Philosophy
School of Public Health and Social Work
Faculty of Health
Queensland University of Technology
2016
Evidence based recommendations for national healthcare-associated infection surveillance i
Keywords
Key Words Infection prevention, infection control, nosocomial infection, surveillance, healthcare-associated infection, epidemiology, safety and quality, discrete choice experiment, public reporting
ii Evidence based recommendations for national healthcare-associated infection surveillance
Abstract
Background
Healthcare-associated infections (HAIs) cause significant morbidity and
mortality, and are the most common complication affecting patients. Most are
believed to be preventable, which requires an understanding of how, why and where
they are occurring. A HAI surveillance program informs such knowledge.
Surveillance of HAIs is fundamental to any infection prevention program and
provides data on which to develop an infection prevention program.
Australia is one of the few developed countries that does not have a national
HAI surveillance program. Several state-wide HAI surveillance programs based on
the National Health and Safety Network (NHSN) in the United States of America,
have developed independently. However, there has been no attempt to coordinate
surveillance activities to generate national data. As such, the national burden of HAIs
in Australia is unknown. This is important as it severely limits attempts to develop
national infection prevention policy based on evidence and implement best practice
across Australia.
This thesis aimed to develop evidence based recommendations for a national
HAI surveillance program through answering four research questions:
1. What are the similarities and differences between existing HAI surveillance
processes in Australia?
2. What level of agreement exists in the identification of HAI between those
participating in HAI surveillance, and are there any factors that influence
agreement level?
3. What are the key attributes of successful centrally coordinated HAI
surveillance programs?
4. What are the preferences and priorities of key stakeholders when considering
a national HAI surveillance program?
Method
This research was a multipart study comprising a scoping review, a cross-
sectional survey, qualitative interviews and a discrete choice experiment.
Evidence based recommendations for national healthcare-associated infection surveillance iii
First, the scoping review of statewide surveillance programs in Australia and
international programs was undertaken. This was done by reviewing information on
surveillance organisational websites and supporting resources, and through structured
discussions with representatives from these programs. This established the current
surveillance activities across Australia in the context of international programs, and
was used to inform the design of the first study.
The first study was a cross sectional online survey of those who undertake HAI
surveillance activities across Australia. The aim of this study was twofold; to
improve our understanding of current surveillance practices by identifying in detail
how surveillance is currently performed, and to measure agreement when identifying
HAIs through a series of clinical vignettes. Participants were recruited using a
snowballing method starting with an email to over 500 subscribers of the
Australasian College for Infection Prevention and Control list server. All infection
prevention staff in hospitals with more than 50 acute inpatient beds were invited to
complete an online survey.
The second study was a discrete choice experiment (DCE) that aimed to
identify key stakeholder preferences for a national surveillance program, and was
conducted in two parts. First, a series of seven semi-structured interviews with
leaders from international and state-wide Australian HAI surveillance programs
informed by a comprehensive literature review, was conducted to identify factors
that are influential in surveillance program implementation and success. The findings
enabled the identification of key characteristics of national surveillance programs,
which were then used to inform and construct the DCE. The DCE provided
quantitative evidence on which elements of a national HAI surveillance program key
stakeholders consider most important. A total of 184 clinical and non-clinical leaders
in infection prevention across Australia were purposively selected to participate in
the DCE.
Results
The scoping review highlighted many differences between statewide programs
in Australia such as the type of infections under surveillance, definitions and support
resources. These findings informed the design of the first study and have been
published in Australian Health Review.
iv Evidence based recommendations for national healthcare-associated infection surveillance
There were a total of 104 completed responses to the cross sectional study.
Large variation in surveillance methodology, definitions, reporting, staff skill and
support was identified across Australia highlighting the many gaps and issues
required to be addressed for a national program. These findings have been published
in the American Journal of Infection Control, and were further supported by the
results from the clinical vignettes, which identified only moderate agreement in HAI
identification (range 53%-75%, excluding the control vignette). Findings from the
clinical vignettes were published in Antimicrobial Resistance and Infection Control.
The outcomes from the scoping review and the cross-sectional study have
provided answers to research questions 1 and 2.
Data from the semi-structured interviews from the DCE were analysed to
identify main characteristics of national surveillance programs. This data identified
five distinct but related characteristics of large HAI surveillance programs; triggers,
purpose, data measurements, implementation and maintenance, and processes. These
findings have been accepted for publication in the American Journal of Infection
Control. The interview data also informed the construction of the DCE.
A total of 122 key stakeholders completed the DCE (response rate 66%). The
results identified key stakeholders strongest five preferences for national HAI
surveillance: 1) a mandatory program with continuous targeted surveillance on
specified HAIs, 2) a surveillance protocol which allows for risk adjustment of HAI
rates, 3) annual competency assessments of data collectors, 4) very accurate data, and
5) hospital level data publicly reported on a website but not associated with financial
penalties. These findings have been published the BMJ Open.
The results from the DCE provided answers to research questions 3 and 4.
Findings from this original research have provided a rich source of evidence on
which to base recommendations for a national surveillance program in Australia. The
recommendations include: a mandatory HAI surveillance program, standardised
national surveillance protocol, risk adjusted and publicly released hospital HAI data,
and regular competency assessments of surveillance staff. Success of the program
requires a comprehensive implementation strategy and central coordination with
regular evaluation and expansion.
Evidence based recommendations for national healthcare-associated infection surveillance v
Conclusion
Currently the true burden of HAIs in Australia remains unknown. The
recommendations within this PhD address the current surveillance gaps in Australia
identified from this research, reflect the key stakeholder preferences for a
surveillance program, and importantly, are in alignment with best practice. These
elements will also positively influence the likelihood of implementation and
sustainability
vi Evidence based recommendations for national healthcare-associated infection surveillance
Evidence based recommendations for national healthcare-associated infection surveillance vii
Table of Contents
Keywords .................................................................................................................................................. iAbstract ................................................................................................................................................... iiTable of Contents .................................................................................................................................. viiList of Figures ........................................................................................................................................ xiList of Tables ......................................................................................................................................... xiiList of Abbreviations ............................................................................................................................ xiiiStatement of Original Authorship ......................................................................................................... xvDeclarations of Interest ......................................................................................................................... xvAcknowledgements .............................................................................................................................. xviCHAPTER 1: INTRODUCTION ....................................................................................................... 11.1 Background .................................................................................................................................... 11.2 Context of the study ....................................................................................................................... 21.3 Background of the author and contribution ................................................................................... 41.4 Aim ................................................................................................................................................ 61.5 Thesis outline and significance ...................................................................................................... 61.6 Summary ........................................................................................................................................ 8CHAPTER 2:LITERATURE REVIEW ........................................................................................... 92.1 Surveillance programs ................................................................................................................. 10
2.1.1 Purpose of surveillance ................................................................................................... 112.1.2 Establishing a surveillance program ............................................................................... 13
2.2 Attributes of surveillance programs ............................................................................................ 152.3 A brief history of healthcare –associated infection surveillance ................................................. 192.4 Healthcare-associated infection surveillance methods ................................................................ 21
2.4.1 Automated surveillance systems ..................................................................................... 232.5 National healthcare-associated infection surveillance systems ................................................... 24
2.5.1 United States of America ................................................................................................ 262.5.2 United Kingdom ............................................................................................................. 272.5.3 Germany ......................................................................................................................... 282.5.4 France ............................................................................................................................. 292.5.5 Netherlands ..................................................................................................................... 292.5.6 ECDC .............................................................................................................................. 30
2.6 Effectiveness of large surveillance programs .............................................................................. 302.7 Healthcare-associated infection surveillance in Australia ........................................................... 33
2.7.1 Surveillance activity by Australian infection prevention staff ....................................... 352.8 Benchmarking, public reporting and financial penalties ............................................................. 372.9 Data Quality ................................................................................................................................. 41
2.11 Implementation Science ............................................................................................................ 482.12 Conclusion ................................................................................................................................. 50CHAPTER 3:THE RESEARCH QUESTIONS AND STUDY DESIGN ..................................... 533.1 Research Question 1 .................................................................................................................... 533.2 Research Question 2 .................................................................................................................... 543.3 Study 1 – Cross sectional survey: Current Australian hospital practices in healthcare-associated infection surveillance ............................................................................................................................ 54
3.3.1 Study 1 design ................................................................................................................. 553.4 Research Question 3 .................................................................................................................... 553.5 Research Question 4 .................................................................................................................... 563.6 Study 2 – Preferences for a healthcare-associated infection surveillance program using a discrete choice experiment .................................................................................................................................. 56
3.6.1 Study 2 design ................................................................................................................. 563.7 Ethics and Limitations ................................................................................................................. 58CHAPTER 4:HEALTHCARE-ASSOCIATED INFECTION IN AUSTRALIA ........................ 594.1 Introduction ................................................................................................................................. 594.2 Paper one: “Healthcare-associated infections in Australia: time for national surveillance” ....... 62
CHAPTER 5:VARIATION IN HAI SURVEILLANCE PRACTICES ....................................... 795.1 Introduction ................................................................................................................................. 795.2 Paper two: “Variation in healthcare-associated infection surveillance practices in Australia” .. 82
CHAPTER 6:DIFFERENCES IN IDENTIFYING HEALTHCARE-ASSOCIATED INFECTIONS ................................................................................................................................ 916.1 Introduction ................................................................................................................................. 916.2 Paper three: “Differences in identifying healthcare-associated infections using clinical vignettes and the influence of respondent characteristics: a cross-sectional survey of Australian infection prevention staff” .................................................................................................................................... 94
CHAPTER 7:CHARACTERISTICS OF LARGE HEALTHCARE -ASSOCIATED INFECTION SURVEILLANCE PROGRAMS ............................................................................. 115
Evidence based recommendations for national healthcare-associated infection surveillance ix
7.1 Introduction ............................................................................................................................... 1157.2 Paper four: “Characteristics of national and statewide healthcare-associated infection surveillance programs: A qualitative study” ....................................................................................... 119
CHAPTER 8:STAKEHOLDER PREFERENCES FOR A NATIONAL HEALTHCARE-ASSOCIATED INFECTION SURVEILLANCE PROGRAM .................................................... 1398.1 Introduction ............................................................................................................................... 1398.2 Paper five: “Novel application of a discrete choice experiment to identify preferences for a national healthcare associated infection surveillance programme: a cross-sectional study” .............. 142
CHAPTER 9:DISCUSSION ........................................................................................................... 1659.1 Introduction ............................................................................................................................... 1659.2 Answers to the Research Questions ........................................................................................... 1669.3 Purpose of a surveillance program ............................................................................................ 1689.4 Support for a surveillance program ........................................................................................... 1689.5 System ....................................................................................................................................... 169
9.8 Investing in national healthcare-associated infection surveillance ............................................ 1809.9 Coordination, implementation and sustainability ...................................................................... 183
9.9.1 Coordinating role .......................................................................................................... 1839.9.2 Implementation and Sustainability ............................................................................... 1839.9.3 Summary ....................................................................................................................... 188
9.10 Limitations ............................................................................................................................... 1899.11 Recommendations for a national healthcare-associated infection surveillance program ........ 190CHAPTER 10: CONCLUSION ................................................................................................... 195REFERENCES .................................................................................................................................. 197
x Evidence based recommendations for national healthcare-associated infection surveillance
APPENDICES ................................................................................................................................... 217Appendix A: Key search terms and outputs for literature review ............................................ 217Appendix B: Ethics approval - Current Australian hospital practices in healthcare-
associated infection surveillance .................................................................................. 219Appendix C: Ethics approval - Key attributes of a healthcare-associated infection
surveillance program .................................................................................................... 221Appendix D: Ethics approval - Preferences for a healthcare-associated infection
surveillance program using a discrete choice experiment ............................................ 223Appendix E: Letter of Support from the Australasian College for Infection Prevention
and Control ................................................................................................................... 225Appendix F: Survey tool - Current Australian hospital practices in healthcare-associated
infection surveillance .................................................................................................... 226Appendix G: Current Australian hospital practices in healthcare-associated infection
surveillance: Frequency of access to other healthcare professionals – data not included in Chapter 5 .................................................................................................... 241
Appendix H: Current Australian hospital practices in healthcare-associated infection surveillance: Frequency of where HAI data is reported – data not included in Chapter 5 ....................................................................................................................... 242
Appendix I: Semi-structured interview guide for participants ................................................. 243Appendix J: Survey tool – Discrete choice experiment ........................................................... 245Appendix K Results of attitudinal questions in the discrete choice experiment not
included in the manuscript Chapter 8 ........................................................................... 265Appendix L: Normalisation process theory questions ............................................................. 266
As this is a thesis by publication, there are references in each publication that
are relevant for the individual publication and in the style of the journal to
which they were submitted. The references at the end of this thesis represent
the references for the unpublished sections of the document (Chapters 1, 2, 3,
9 and 10).
Evidence based recommendations for national healthcare-associated infection surveillance xi
List of Figures
Figure 1 - Funding sources of Australian hospitals ................................................................................. 3Figure 2 - The surveillance cycle .......................................................................................................... 11Figure 3 - Centrally coordinated (national) healthcare-associated infection surveillance
xii Evidence based recommendations for national healthcare-associated infection surveillance
List of Tables
Table 1 - Common objectives of healthcare-associated infection surveillance .................................... 13Table 2 - Attributes of a surveillance program listed in CDC Guidelines ............................................ 16Table 3 - Attributes of a healthcare-associated infection surveillance program listed by NHSN ......... 17Table 4 - Healthcare-associated infection surveillance strategies ......................................................... 22Table 5 - Reductions in infection due to surveillance ........................................................................... 32Table 6 - Process and outcome measurements used in four high income countries ............................. 38
Evidence based recommendations for national healthcare-associated infection surveillance xiii
List of Abbreviations
ACD Administrative coding data
ACIPC Australasian College for Infection Prevention and Control
ACSQHC Australian Commission on Safety and Quality in Health Care
AMR Antimicrobial resistance
AU Antimicrobial usage
AURA Antimicrobial Use and Resistance in Australia
BSI Bloodstream infection
CAUTI Catheter associated urinary tract infection
CDC Centers for Disease Control (USA)
CDI Clostridium difficile infection
CFIR Consolidated Framework for Implementation Research
CLABSI
DCE
ECDC
FTE
HAI
HAUTI
Central line associated bloodstream infection
Discrete choice experiment
European Centre for Disease Control and Prevention
Full time equivalent
Healthcare-associated infection
Healthcare-associated urinary tract infection
HELICS Hospitals in Europe Link for Infection Control through Surveillance
ICU Intensive care unit
IHBI Institute of Health and Biomedical Innovation
IP Infection prevention
IPSE Improving Patient Safety in Europe
KISS Krankenhaus-Infektions-Surveillance-System (Germany)
MRSA Methicillin resistant Staphylococcus aureus
NHHI National Hand Hygiene Initiative
NHS National Health Service (UK)
NHSN National Health and Safety Network (USA)
NINSS Nosocomial Infection National Surveillance Scheme (UK)
NNIS National Nosocomial Infection Surveillance (USA)
xiv Evidence based recommendations for national healthcare-associated infection surveillance
NPT Normalisation process theory
NSQHSS National Safety and Quality Health Service Standards
OR Odds ratio
PPV Positive predictive value
QUT Queensland University of Technology
RR Risk ratio
SAB Staphylococcus aureus bacteraemia
SSI Surgical site infection
UK United Kingdom
USA United States of America
UTI Urinary tract infection
VAP Ventilator associated pneumonia
VICNISS Victorian Healthcare-associated Infection Surveillance System
VIF Variance inflation factor
Evidence based recommendations for national healthcare-associated infection surveillance xv
Statement of Original Authorship
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the
best of my knowledge and belief, the thesis contains no material previously
published or written by another person except where due reference is made.
Signature:
Date:
Declarations of Interest
I am a current member of the Board of Directors of the Australasian College
for Infection Prevention and Control, and Chair of its Research Committee.
I am also a member of the Healthcare Associated Infection Advisory
Committee of the Australian Commission for Safety and Quality in Health Care, a
member of the National Health and Medical Research Council’s Infection Control
Guidelines Advisory Committee, and previously Operations Director at the VICNISS
Coordinating Centre.
8th October 2016 _________________________
QUT Verified Signature
xvi Evidence based recommendations for national healthcare-associated infection surveillance
Acknowledgements
The very nature of nursing has meant that I have worked with many wonderful
people who have influenced my career. I would like to first acknowledge A/Professor
Denis Spelman who mentored me from very early beginnings. I am grateful to Denis
for many things, in particular the culture of continuous learning he encouraged
by asking at the end of every day, “What have you learnt today?”
I have also had the privilege of working closely with Professor Mike Richards
and Professor Lindsay Grayson for extended periods in my career and thank them
both for their support and guidance, and the extraordinary opportunity they provided
me to lead exciting initiatives.
The decision to undertake a PhD was a long time coming, and I sought the
advice of many. Thanks to A/Professor Brett Mitchell and Professor Ramon Shaban
who patiently played the roles of decision support systems during this time. They
have not only been great supports throughout this doctorate, they continue to inspire
me in their work and life.
My associate supervisors, Professor Mike Richards, Professor Allen Cheng,
and Professor Nick Graves have all been available when I needed them to be,
responded on short turn around, and always encouraging of my work. I thank them
for their support. I am of course grateful to Nick and his team at the Centre for
Research Excellence in Reducing Healthcare Associated Infections at Queensland
University of Technology (QUT) who provided me the opportunity to join their
collection of PhD candidates. Gratitude also to the other PhD candidates in this
cohort who always welcomed me into their various sessions even though I was often
present via a grainy screen and ad hoc audio! Their support has been much
appreciated.
My principal supervisor Dr Lisa Hall no doubt will be glad to see me off!
Weekly Skype sessions, phone calls, emails and frequent two day visits, Lisa has
provided me solid support. Her extraordinary ability to think a little left, right, above
and below was often the nudge I needed when my brain hit the wall. Lisa’s mixed
Evidence based recommendations for national healthcare-associated infection surveillance xvii
methods skill is unique. I am extremely grateful for her mentoring, guidance,
availability, and interest in my work and life.
I have been fortunate to receive financial support that has enabled my full time
studies. I wish to acknowledge the Rosemary Norman Foundation and the Nurses
Memorial Centre who awarded me the “Babe” Norman scholarship, the Centre for
Research Excellence in Reducing Healthcare-Associated Infection, QUT, Covidien
(Medtronic), and the Australian College of Nursing.
Undertaking a PhD is a completely selfish act, and three years is a long time in
the life of a family of five! Two sons finished school, the other landed his first full
time job. We celebrated an eighteenth, two twenty-firsts, two fiftieths, a twenty-fifth
wedding anniversary and my mother’s 80th! Regardless of work and study, family
life hurtles along, and I am grateful for their support and understanding throughout as
my study door was often closed. Special thanks to my wife Kate who not only
managed to complete a MPH during this time, but has encouraged and supported me
unconditionally.
I undertook this PhD for myself, and I have enjoyed it from the very start. I am
genuinely grateful to all those mentioned above, and many others who have played
their part in enabling me to take this once in a lifetime opportunity, albeit late!
xviii Evidence based recommendations for national healthcare-associated infection surveillance
“Systems awareness and systems design are important for health professionals,
but they are not enough. They are enabling mechanisms only. It is the ethical
dimensions of individuals that are essential to a system’s success. Ultimately, the
secret of quality is love. You have to love your patient, you have to love your
profession. If you have love, you can then work backward to monitor and improve
the system.”
Avedis Donabedian 2000
Chapter 1: Introduction 1
Chapter 1: Introduction
1.1 BACKGROUND
“Premum non nocere” is a guiding principle for medical personnel which,
when translated into English means, “first, do no harm”.1 Although dating back to
the early 1800’s, and despite its deficiencies as an absolute principle,1 it still has
relevance in todays healthcare setting. Patients seek out healthcare practitioners
generally expecting to gain some health benefit. Unfortunately there are some
patients who are the subject of harm, such as acquiring an infection.
A healthcare-associated infection (HAI) is defined as an infection that occurs
as a result of a healthcare intervention and may occur within, or after leaving, a
healthcare facility.2 Historically called a “nosocomial” infection, meaning “hospital
acquired”, the term “healthcare-associated” is now preferred, acknowledging that
today much healthcare is administered beyond the hospital walls. Various types of
infection can result from a healthcare intervention, such as pneumonia, urinary tract
infection, a bloodstream infection (BSI) caused by an intravenous device, or an
infected wound following a surgical procedure.
HAIs are the most common complication affecting patients in healthcare
facilities, and many result in significant morbidity and mortality.3 It is estimated that
in Europe and North America between 12%-32% of HAI BSIs result in death.4 In
developing countries, the burden of HAIs is significantly higher when compared to
developed countries, with the density of catheter related BSI estimated to be up to 19
times higher.5
In the field of safety and quality in healthcare, HAIs are considered preventable
adverse events, that is, a medical error resulting in injury.6 This places it alongside
other preventable adverse events such as fractures resulting from a patient fall, and
the adverse side effects following administration of an incorrect medication.
Logically, to prevent HAIs, it is important to know how often they are occurring,
why, where, how and to whom. A HAI surveillance program will deliver this
information.
2 Chapter 1: Introduction
Surveillance is the “ongoing and systematic collection, analysis and
interpretation of outcome specific data essential to the planning, implementation, and
evaluation of public health practice, closely integrated with the timely dissemination
of these data to those who need to know”.7 Surveillance has been likened to a nerve
cell, where an afferent arm receives information, data are analysed by the cell, and
the efferent arm then takes action.7
Surveillance of HAIs is the cornerstone of healthcare epidemiology and
infection prevention programs,8 and has been described as the single most important
factor in the prevention of HAIs.9 Surveillance is held in such esteem because it
provides the information on which an infection prevention program is planned, and
in a landmark study, has been shown to reduce HAI rates through the influence of
data on practices.10
Whilst many countries have well established national HAI surveillance
programs, Australia does not. This severely limits our understanding of the
epidemiology of HAIs in Australia, which in turn restricts our ability to implement
evidence based policy, and measure the real impact of any infection prevention
interventions. It has been suggested that in Australia, 175,000 HAIs occur annually,11
however this estimate was based on data from only two hospitals. The lack of a
national surveillance program means that a more precise estimate is unable to be
made. This is a significant gap in our knowledge of HAIs in Australia.
1.2 CONTEXT OF THE STUDY
Australia consists of six states and two territories with an estimated population
of 24 million.12 There are 1,359 hospitals, of which 55% are public. The state and
territory governments are the largest funders of the public hospitals, whilst health
insurance funds contribute most to funding private hospitals (Figure 1).
Approximately 80% of the public hospitals have less than 100 beds, only 3% over
500 beds, whilst 21% are considered remote.13
Chapter 1: Introduction 3
Figure 1 - Funding sources of Australian hospitals
Adapted from the Australian Institute for Health and Welfare 14
In 2012, the Australian Commission for Safety and Quality in Health Care
(ACSQHC) released the National Safety and Quality Health Service Standards
(NSQHSS).15 Standard 3 is specific to infection prevention and control, and lists as
actions required to meet the standard; “Surveillance systems for healthcare-
associated infections are in place” and “Healthcare-associated infections surveillance
data are regularly monitored by the delegated workforce and/or committees”.15 There
are no recommendations regarding the type, method or intensity of surveillance.
Although there is no national surveillance program, it would appear substantial
resources exist at a hospital level that are devoted to surveillance. A cross sectional
study undertaken across 152 hospitals in 2014 estimated the mean full time
equivalent (FTE) infection prevention nurses per 100 beds to be 0.66, or 1 FTE per
152 beds, and remained relatively constant when stratified by hospital size.16 From
the same study, infection prevention nurses estimated they spent 36% of their time
undertaking surveillance activities, of which 56% was spent on collecting data.17
The focal point of this PhD is the Australian healthcare setting, however part of
the second study required data collection outside of Australia, as it was deemed
crucial that existing national surveillance programs were explored. data were
0% 10% 20% 30% 40% 50% 60% 70% 80% 90%100%
Privatehospitals(n=612)
Publichospitals(n=747)
Stateandterritorygovernments AustralianGovernment
Individuals DeptofVeteranAffairs
Other Healthinsurancefunds
4 Chapter 1: Introduction
collected from four countries with populations ranging from 5 million to over 300
million. Three countries were English speaking. Despite differences in culture, size,
governance and funding structures, these countries all have well established national
surveillance programs that were explored in detail in the second study.
During the undertaking of this PhD, a large piece of work titled Antimicrobial
Use and Resistance in Australia (AURA) Project was commissioned by ACSQHC to
explore options for antimicrobial resistance (AMR) and antimicrobial usage (AU)
surveillance in Australia. Whilst acknowledging obvious synergies between a
national HAI surveillance program and national AMR and AU surveillance, it differs
from the focus of this PhD in that AMR and AU surveillance utilises population data
rather than patient level data to identify trends and distribution patterns. The final
report from the AURA project is not due for completion until late 2016, therefore
findings and recommendations are unable to be included in this thesis.
The scope of this PhD is directed towards outcome surveillance in acute care
health public and private facilities of greater than 50 beds. Larger acute care facilities
generally have a patient population at higher risk of acquiring a HAI as these
facilities are more likely to have sicker patients, intensive care units (ICUs) and
undertake complex procedures. It is in these facilities where surveillance is generally
thought to have the greatest impact with respect to reducing HAIs. However many of
the findings are generalisable to other healthcare facilities either directly or indirectly
by providing structure and context on how to approach surveillance of HAIs.
A point worth clarifying is the use of the terms “surveillance program” and
“surveillance system”. Frequently the terms are used interchangeably, and essentially
they are the same. For the purposes of consistency, I will be using the term
“surveillance program” except where a program describes itself in the literature as a
“system”.
1.3 BACKGROUND OF THE AUTHOR AND CONTRIBUTION
My familiarity with HAI surveillance reaches back several decades to my first
position in infection prevention at a large Melbourne hospital. Since this time I have
held positions as the inaugural Operations Director of the Victorian healthcare-
associated infection surveillance program, VICNISS, and the national project
manager for the National Hand Hygiene Initiative (NHHI). I also completed a
Chapter 1: Introduction 5
Masters in Clinical Epidemiology during this time. Through my roles at VICNISS
and with the NHHI, I have been on a number of state and national infection
prevention committees, including ongoing membership of the HAI Advisory
Committee of the ACSQHC.
HAI surveillance has been a crucial element in all my roles. Initially at a
hospital level, I was involved in the development and implementation of a surgical
site infection (SSI) surveillance program based on the National Nosocomial Infection
Surveillance (NNIS) System (now the National Health and Safety Network
[NHSN]). For the first time, risk adjusted, procedure specific, surgeon specific, SSI
rates were generated, analysed and importantly fedback to the surgeons and an
infection control committee. This surveillance activity expanded to include ICU
central line associated bloodstream infection (CLABSI) surveillance and also ad hoc
and point prevalence surveillance activities over the years.
At VICNISS, my team established and implemented a statewide HAI
surveillance program for all public acute care facilities in Victoria. The program
continues today and is arguably the most robust statewide HAI surveillance program
in Australia. My national role with the NHHI enabled me to work with hospitals and
health departments across all states and territories in Australia. It provided me with
an extraordinary and privileged national perspective of infection prevention.
Through my experience with HAI surveillance, it always struck me that a
major limitation was the inability to generate national data and make comparisons
with hospitals across Australia. Different to the United States of America (USA), the
United Kingdom (UK) and many European countries, who have well established
national HAI programs, Australia is small with fewer hospitals, yet no national
surveillance program. This means that where only state programs exist, context is
limited as they can only be compared with like facilities in that state, statewide
denominators are smaller, rates more variable, and data less robust. Over the years I
have networked with many international colleagues and been envious of their access
to national HAI data.
I have a firm belief that Australia has much to benefit from national HAI
surveillance data. This has led me to question how far away we are from a national
program, and explore what sort of surveillance program would best suit Australia.
6 Chapter 1: Introduction
I confirm my following contributions to both studies in this thesis: study
design, administration, data collection, data analysis and manuscript writing (study
design, and data analysis and manuscript writing was assisted by supervisors and
other authors listed in publications).
1.4 AIM
The overall aim of this research is to establish evidence based
recommendations for an Australian national HAI surveillance system.
The specific research questions are:
1. What are the similarities and differences between existing HAI
surveillance processes in Australia?
2. What level of agreement exists in the identification of HAI between
those participating in HAI surveillance, and are there any factors that
influence agreement level?
3. What are the key attributes of successful centrally coordinated HAI
surveillance programs?
4. What are the preferences and priorities of key stakeholders when
considering a national HAI surveillance program?
The answers to these questions were addressed by undertaking two studies. The
first was a cross sectional survey of infection prevention and control staff who are
currently involved in HAI surveillance across Australia working at acute care
hospitals with more than 50 beds. The second study was a discrete choice experiment
(DCE) involving a broader range of key stakeholders. These studies, and their
findings generated five papers - four published, with the fifth recently having been
accepted for publication.
1.5 THESIS OUTLINE AND SIGNIFICANCE
This thesis by publication comprises three main sections, Literature Review,
five papers from two studies, and the Discussion that includes the recommendations
for a national HAI surveillance program.
In Chapter 2, the literature review explores major historical studies on HAI
surveillance and key papers on national HAI surveillance programs demonstrating
Chapter 1: Introduction 7
the benefits of surveillance, the current gaps in Australian surveillance and the many
issues that are relevant to HAI surveillance. The last section of the literature review
describes the unique DCE method used in the second study, and its application in
health sciences.
Chapter 3 presents the research questions, and discusses why these questions
are considered important, and outlines the methods used in the two studies
undertaken. The following five chapters contain the publications that have been
generated from this research.
Chapter 4 provides the results of a broad overview of international surveillance
programs and a review of existing Australian surveillance activities. The findings
identified several well established international surveillance programs, and major
differences between surveillance activities and coordination between Australian
states and territories, contributing to answers for research question 1. This paper was
published in Australian Health Review.
The publication in Chapter 5 also provides answers to question 1 and explores
in detail the differences between surveillance practices currently undertaken across
Australia. This study identified broad variation in current surveillance practices
across Australia, and was published in the American Journal of Infection Control.
Chapter 6 presents the paper published in Antimicrobial Resistance and
Infection Control journal, and was also generated from the first study. Seven clinical
vignettes included as part of the first study demonstrated only moderate agreement
when identifying HAIs. Data from this analysis answers research question 2.
Such detailed analysis of Australian surveillance practices and measurement of
agreement has never previously been described in Australia.
Through a literature review and a series of semi-structured interviews with
international experts, qualitative analysis identified five key characteristics of HAI
surveillance programs, and are presented in Chapter 7. This work comprises the first
part of the second study, the DCE, and answers question 3. This manuscript has been
accepted for publication in the American Journal of Infection Control (June 2016).
The full results of the DCE are presented in Chapter 8, and identify key
stakeholder preferences for a national surveillance program. The findings from the
DCE answer question 4. This manuscript has been published in BMJ Open. This is
8 Chapter 1: Introduction
the first time in an Australian setting that stakeholder holder preferences for a
surveillance program have been described.
Chapter 9 provides a detailed discussion of the findings from the studies in the
context of current knowledge, and synthesis of the data. There is also a discussion on
a pragmatic implementation framework required for a new surveillance program. The
chapter concludes by providing recommendations for the establishment of a national
surveillance program. The recommendations listed in Chapter 9 have been based on
findings from the studies undertaken in this PhD, using both local and international
data.
The Conclusion within Chapter 10 then summarises this thesis.
The Appendices contain several items of interest, including ethics approvals,
the surveys used, survey results not included in the publications, letters of support
and other relevant material.
This is significant research both for Australia and for the international infection
prevention field. This work builds on current knowledge of HAI surveillance
programs, and adds new knowledge applicable for the development, implementation
and maintenance of a national HAI surveillance program.
1.6 SUMMARY
This chapter has provided an overview of the origins of this thesis, an
introduction to HAIs and surveillance, and an outline of the structure of this thesis. A
background of the author has also been provided to highlight the long association of
working in this topic area that has provided a rounded understanding of the many
challenges for those at both a local and national level. The next chapter provides a
literature review covering several important issues relating to HAI surveillance.
Chapter 2: Literature Review 9
Chapter 2: Literature Review
This literature review provides a narrative of key articles relevant to the topics
of interest included in this thesis. These topics include fundamental issues relating to
surveillance programs in general and specific to HAI surveillance.
As the second study undertaken as part of this doctorate included the unique
DCE method, which has not previously been described in infection prevention
literature, it was important to include literature discussing this novel method to
demonstrate its suitability for the study. Similarly, the review also presents key
papers on implementation science, which is a crucial element when considering
pragmatic, evidence based recommendations for a national surveillance program.
The literature was accessed through Pubmed using a number of different search
terms in a structured, systematic fashion. The key search terms used were
“healthcare-associated infection” and “surveillance”. As the term “healthcare-
associated infection” is reasonably new, a search was also conducted on the terms
“nosocomial infection” and “surveillance”. This produced a total of 144 articles. To
focus the search on national surveillance programs, their development and
implementation, key terms of “national”, “development”, “establishment” and
“implementation” were introduced individually. This resulted in a zero return
additional for each of these terms. The search terms of “public reporting” and “data
quality” were also added identifying 9 and 65 articles respectively.
A general search was then conducted using the term “discrete choice
experiment” and “implementation science” which generated 67 and 279 articles
respectively.
The topics of public reporting, data quality, discrete choice experiments and
implementation science are rapidly evolving areas, and to make the review more
manageable, the searches of these terms were limited to English, involving humans,
had an abstract and for the period 2005 to 2015.
10 Chapter 2: Literature Review
All article titles were reviewed for relevance to national surveillance or large
surveillance networks. Abstracts were reviewed for their relevance to the specific
topic under discussion.
Given the limited volume of articles specific to this work, grey literature, such
as government reports, recommendations and surveillance protocols were sourced
from organisation websites. This also included grey literature on public health
surveillance of which much of HAI surveillance is based. Many of the articles that
have been included in this review were sourced from references within the grey
literature.
A detailed description of the search terms used and articles identified are listed
in Appendix A.
2.1 SURVEILLANCE PROGRAMS
The origin of the word surveillance comes from early 19th century French,
translated from sur “over” and veillar “watch”, and is derived from the Latin, vigilare
which means to “keep watch”.18 Commonly used in the observation of suspect
persons, in healthcare, the term has been applied to observing individuals and or
diseases.19
Scientifically based healthcare surveillance programs first came to light with
the monitoring and isolation of people with serious communicable diseases in the
late 1800’s. American epidemiologist Alexander Langmuir is credited with shifting
the emphasis of surveillance from monitoring those with or at risk of communicable
diseases to the diseases themselves.7
Surveillance programs are designed to provide basic epidemiological
descriptive data such as the time, place and person involved in the particular event
under observation. Such basic information enables the monitoring of the event over
time.20
Surveillance can be viewed as an information cycle, typically commencing
with recognition of an event, data collection, data analysis, interpretation and
importantly, dissemination of results to enable action (Figure 2).20 It is this action
which differentiates surveillance from simply monitoring events.21
Chapter 2: Literature Review 11
Figure 2 - The surveillance cycle
2.1.1 Purpose of surveillance
By its very existence, the science of “infection prevention” implies that HAIs
are preventable. Exactly what proportion of HAIs are preventable is unclear and
difficult to establish due to limitations with study designs. Most published literature
is derived from before and after studies which suffer from lack of randomisation and
control.22 It is also suggested that a possible publication bias exists in that studies
undertaken in this topic with negative outcomes may remain unpublished.22
In a recent systematic review looking at HAI reduction studies restricted to
USA and published in the previous 10 years, almost 5000 potentially relevant articles
were identified, however only 15 were included in the review to estimate the
proportion of preventable HAIs. Looking only at four types of HAI, seven “good
quality” studies demonstrated reductions in catheter associated blood stream
infections of between 18%-66%, two “good quality” studies and one “moderate
quality” study demonstrated reductions in ventilator associated pneumonia (VAP) of
between 46% and 55%, two “moderate studies” demonstrated reductions in catheter
associated urinary tract infection (CAUTI) of between 17%-69%, and three
“moderate quality” studies on SSI demonstrated reductions of 26%, 54% and 29%.23
DataCollection
DataAnalysis
Establishmentofratesanddissemination
Implementinterventions
12 Chapter 2: Literature Review
Uschmeid et al.23 cautiously report their findings due to the general low quality
of the studies. They do however conclude that the study population with the highest
risk of infection were often those that demonstrated the greatest reductions when
compared to study populations with low risk of infection.23 This supports the notion
that those groups who are at greatest risk of infection should be targeted in
surveillance programs.
In a more recent study, Lambert24 estimated that 52% of VAP and 69% of BSIs
are preventable. Limitations of this study include its use of routine HAI surveillance
data that had not been rigorously validated, and participation bias in that ICUs
submitting data are likely to have lower rates than those that do not participate.24
Although it is challenging to quantify the preventable proportion of HAIs, there
is agreement that a significant proportion, and probably the majority of HAIs are
preventable.22,23 This underpins the purpose of HAI surveillance.
The stated purpose of the surveillance program should indicate why the
program exists.25 When establishing a surveillance program, the purpose and
objectives must be clearly understood. Thacker proposes that two questions be
considered to help clarify purpose: “What will be done with the data and analysis?”
and “What action will be taken?”7 Answers to these questions and specific, action
oriented commitment, will determine data requirements and analysis and avoid any
unnecessary data collection.7
It is suggested there is one simple purpose of public health surveillance, and
that is to “provide a scientific basis for appropriate policy decisions in public health
practice and allocation of resources”.7 The purpose of HAI surveillance is to
provide quality data that can act as an effective monitoring and alert system
and reduce the incidence of preventable infections.26,27
Accompanying the purpose should be a set of objectives, or goals of the
program, and include how data from the program can be used.25 Common objectives
of HAI surveillance programs are listed in Table 1.
Chapter 2: Literature Review 13
Table 1 - Common objectives of healthcare-associated infection surveillance
From a slightly different perspective, in 2001 before the creation of NHSN,
NNIS researchers listed three requirements they believed essential for a successful
multi centred surveillance program.31 First, the surveillance program must have a
clear purpose. Second, it must have standardised definitions, data fields and
protocols, and third, there must be a coordinating centre to standardise definitions
and surveillance protocols, receive, review quality, analyse and disseminate data, and
standardise risk adjustment approaches.31
The authors went on to list what they considered the seven “NNIS elements”
critical for the successful reduction of HAIs:
“1) Voluntary participation and confidentiality;
2) Standard definitions and protocols;
3) Defined populations at high risk (e.g., intensive care, surgical patients);
4) Site-specific, risk-adjusted infection rates comparable across institutions;
5) Adequate numbers of trained infection control practitioners;
6) Dissemination of data to health-care providers; and
7) A link between monitored rates and prevention efforts, where patient-carepersonnel relied on the data to alter their behaviour in ways that may have reduced
the incidence of nosocomial infections”.31
There is only one reported use of the CDC guidelines to evaluate a HAI
surveillance program. In a ten year review of the Krankenhaus-Infektions-
Surveillance-System (KISS), the HAI surveillance program program in Germany,32
Gastmeier et al.32 used the CDC guidelines25 to assess the KISS program. In terms of
simplicity and flexibility, Gastmeier notes that since its commencement in 1997, data
collection had become increasingly more electronic based improving simplicity, and
the recent inclusion of a new surveillance component for Clostridium difficile
demonstrated its flexibility.32 Acceptability and representatives was reflected in the
participation of over 500 hospitals of all sizes from across the country.32 KISS is now
a web based reporting system, this means that reports can be generated by hospitals
at any time, and the system is always available, demonstrating its timeliness and
stability.32
Chapter 2: Literature Review 19
In conclusion, surveillance systems have many attributes, though it appears
from the literature these attributes aren’t always relevant or identifiable, and some
attributes may be more important than others. A successful HAI surveillance
program must be epidemiologically sound and balance attributes such as; accuracy,
timeliness, usefulness, consistency, and practicality.30
2.3 A BRIEF HISTORY OF HEALTHCARE –ASSOCIATED INFECTION SURVEILLANCE
The foundations of HAI surveillance can be traced back to Vienna in the mid
19th century to the work of physician Ignaz Semmelwies. Although not trained in
epidemiology, Semmelweis’ observations of mortality data and differences between
those who died and survived resulted in an infection prevention intervention with
dramatic effect. The Vienna Lying hospital was divided into two divisions, the first a
medical teaching service where women were delivered by physicians and students,
the second staffed by midwives. Semmelweis noted that the 1847 maternal death rate
in the first division was 10% compared to the second divisions of 3%.33 A thorough
epidemiological review of data and the coincidental death of a colleague from sepsis
following a needlestick injury during a post mortem, led Semmelweis to note a major
behavioural difference. Medical students undertook autopsies and often went to the
autopsy room to deliver women in the first division, in comparison to the midwives
who did not perform autopsies. Based on this observation Semmelweis hypothesised
that “cadaveric material was the cause of death”.33 Famously, Semmelwies
implemented a hand washing intervention on entry to the delivery suite, which
produced a dramatic and significant decrease in mortality rates.34 A failure to
effectively communicate his controversial findings led to Semmelweis fleeing to
Budapest, where his performance as a physician was criticised, and together with
reported episodes of psychosis, he was committed to an asylum where he eventually
died in 1865.35
Around the same time in the UK, William Farr and Florence Nightingale’s
shared interest in hospital mortality rates resulted in a collaboration that
demonstrated a relationship between hospital hygiene and infectious post operative
complications. Nightingale proposed that nursing staff collect and report data on
hospital mortality.33
20 Chapter 2: Literature Review
In 1860, Scottish physician James Simpson reviewed mortality data following
amputation in country areas compared to metropolitan hospitals and found much
higher mortality rates in the hospitals. He also went on to demonstrate that the
mortality rates increased with the size of the hospital.33 Simpson recognised the
importance of cleanliness and the prompt containment of the excretions from
diseased patients.36
The first well documented report describing active SSI with routine reporting
was conducted by Brewer who provided systematic feedback to surgeons resulting in
a 95% reduction of SSIs in the early 1900’s.33,36,37
The most significant and comprehensive research into infection prevention
programs was undertaken in the USA by Robert Haley in the 1980’s.38 Clear benefits
of HAI surveillance programs were first demonstrated by Haley’s pioneering Study
on the Efficacy of Nosocomial Infection Control (SENIC) commissioned by the
CDC in response to rising criticism from hospitals of the resources required to
comply with their recommendations for infection prevention and surveillance.38 In a
retrospective multi-centred study, Haley set out to determine whether infection
control programs reduced rates of SSI, BSI, urinary tract infections (UTI) and
VAP.10 Over three phases using screening questionnaires, interview surveys, and
medical record reviews, Haley developed a surveillance index to measure the extent
to which each hospital conducted active surveillance, and a control index to measure
the intensity of efforts to reduce infections. After evaluating hospital infection
control programs over a 10 year period, and reviewing HAI rates of SSI, BSI, UTI
and VAP, Haley identified four essential components of an effective infection
prevention program;
• a structured surveillance program,
• one infection prevention nurse per 250 beds,
• an infection prevention physician, and
• a system for reporting infection rates to surgeons.10
Different combinations of these four factors reduced the rates of all four
infections, however the only factor that was present for each was an effective
surveillance program.8
Chapter 2: Literature Review 21
This seminal work has been the foundation on which many HAI surveillance
programs have been established, and has also been used to estimate the impact of
infection prevention on rates for specific infections, to classify patients as either high
risk or low risk of infection, develop risk strata to predict patients probability to
develop indication, and to estimate the costs of infections.38
With the advancement of epidemiology, fundamental principles of good HAI
surveillance have been recognised. The methods used for HAI surveillance have
evolved over time, and this is explored in the next section.
Modified from Perl and Chaiwarth,8 Allen-Bridson, Morell and Horan,30 Perl,38 and Pottinger
et al.39
Targeted surveillance strategies are also supported by the work of Glenister40
from the UK in the early 1990’s. Acknowledging that hospital wide strategy was
resource intensive, Glenister40 conducted a single site, prospective continuous study
to determine the effectiveness of eight surveillance methods using sensitivity,
specificity and time for data collection as outcome measurements. Of the eight, a
combination of two methods, “laboratory based ward surveillance”, which involved
daily review of case records of those identified from positive microbiology reports,
and “ward liaison surveillance” comprising routine twice weekly ward visits,
discussions with nursing staff and review of records of patients reported to have an
infection, was found to have the highest sensitivity of all methods, and required one
third of the time of the reference method. This combination had the advantage of
identifying infections even when specimens were not taken or had negative results.
Although possibly not appropriate for all hospitals, this method was recommended to
increase efficiency and use of resources.40
Even though this method is efficient when compared to other methods, manual
medical record review of patients at risk of a HAI, which may also involve visiting
patents, reviewing microbiology results, discussions with ward staff and team
Chapter 2: Literature Review 23
meetings remain resource intensive.41 Further, the application of definitions is
subject to interpretation and identification of cases is often dependent on effort.41 In
2009, results of a national infection control program survey in the USA reported that
infection prevention staff spend up to 45% of their time collecting, analysing and
interpreting HAI data, the largest percentage of time of any of their infection
prevention activities.42 The same study also noted that 35% of infection prevention
staff had assistance with data management, and 13% had statistical help.42 A recent
cross sectional study with infection prevention staff in Australian, it was identified
that 36% of their time is spent on surveillance, 56% of this time was used for
collecting data. 43
2.4.1 Automated surveillance systems
Acknowledging the burden of manual data collection, the move towards the
use of automated technology and electronic data as an aid to traditional HAI
surveillance methods is gathering momentum.
The use of automated technology and electronic data in HAI surveillance is
well described.44 In a large systematic review, Freeman et al.44 report that electronic
surveillance systems often produced higher sensitivity and specificity when
compared to traditional methods, however they are limited in that they are dependent
on the facility’s electronic information technology systems, and on occasion a HAI
detected by an electronic surveillance systems still requires confirmation by a
healthcare worker. Generally, automated systems ensure consistent application of
surveillance definitions, significantly reduce the burden of data management
associated with traditional methods, provide improved sensitivity and specificity, and
could be used as a tool by staff to enhance their surveillance programs.44
To improve the efficiency of surveillance resources, Perl and Chaiwarth8
believe that integration of rapidly developing surveillance technologies is essential. It
is estimated that electronic HAI surveillance systems reduce time spent on
surveillance by up to 65%, whilst also improving sensitivity and specificity.8
If electronic HAI surveillance systems are to play a bigger role in routine HAI
surveillance, then they need to be as good as, and less prone to subjectivity, than
existing methods. This is particularly important if HAI rates between hospitals are to
be compared or publicly reported. Inconsistent application of definitions and case
24 Chapter 2: Literature Review
finding methods will influence the meaning of data and any comparisons made. A
recent study highlighted these advantages when using a computerised algorithm to
detect bloodstream infections compared to traditional methods.45 Researchers found
the differences in the outcome of these two methods significantly changed the order
of hospitals rankings when bloodstream infection rates were used as an indicator.45,46
Attempts at using administrative coding data (ACD) as a passive method to
detect HAIs is increasing, particularly in the USA where insurance claims are often
used by quality improvement programs and researchers.47 The use of ACD is
attractive because codes are often uniform across hospitals, they are stored
electronically and therefore convenient for applying algorithms.47 In a systematic
review of studies reporting the use of ACD for identifying HAIs, moderate
sensitivity and high specificity was found when detecting Clostridium difficile
infections (CDI) and orthopaedic SSIs. Evidence for other types of HAIs was limited
due to the small number of studies. The moderate level of sensitivity means that
using ACD as the only method to detect cases will result in some HAIs being missed,
and consequently HAIs rates reported using only ACD will be underestimated. Goto
et al.47 recommend that ACD may be useful as a factor within an algorithm, but
should not be used as the primary case finding method.
In summary, the increasing demands for more data in less time means that
current manual data collection methods are unsustainable. Although gradually more
hospitals are moving towards electronic records, Hebden48 reports that the uptake of
automated surveillance systems is low, and calls for more qualitative research to
explore the human factors associated with this poor uptake. Hebden48 implies that a
lack of implementation strategies could be partly to blame, as any automated process
requires an adjustment of workflow and roles, an understanding of how the data are
to be interpreted and then translated into knowledge to guide decision making.
2.5 NATIONAL HEALTHCARE-ASSOCIATED INFECTION SURVEILLANCE SYSTEMS
National HAI surveillance programs are characterised by two interrelated
cycles. At the micro level (hospital), surveillance is used to establish endemic rates
and to detect outbreaks, identify priorities and measure the effect of interventions. At
the macro level (state or national), data from participating sites is collated to provide
aggregated data that may be used for benchmarking and made available to
Chapter 2: Literature Review 25
participating hospitals, policy makers and sometimes the public. Common to all
national surveillance programs is a central data coordination process, often
undertaken by a central body.49 The central body may also be responsible for
developing uniform definitions and methods, and provide education and support for
those involved in surveillance (Figure 3).
Figure 3 - Centrally coordinated (national) healthcare-associated infection surveillance program. Hospital activity in blue circles, central activity in rectangles.49
There are several well established national HAI surveillance programs. The
USA,50 Germany,32 the United Kingdom,51 Belgium,52 Switzerland,53 Spain54 and the
Netherlands55 are all well documented. Many European countries have further
collaborated to establish the European Centres for Disease Control and Prevention
(ECDC) HAI Surveillance network which prescribes uniform definitions and
methodology for participating countries, and facilitates a greater understanding of the
epidemiology of HAIs across Europe.56 A description of the national programs in
USA, Germany, France and UK, and an outline of the ECDC will now be provided.
Adherencetocentral-lineinsertionpractices P§§ .. T ..
Compliancewithsurgicalantimicrobialprophylaxisorskindisinfection P§§ .. T ..
V=voluntary reporting. M=mandatory, confidential reporting. P=reported publicly. T=subject to a government target. HCAI=health-care-associated infection. BSI=bloodstream infection. MRSA=meticillin-resistant Staphylococcus aureus. SSI=surgical-site infection. UTI=urinary-tract infection. VAP=ventilator-associated pneumonia. *In at least one federal state. †Used in Pennsylvania until 2007.37 ‡California.85 §As part of a target related to the overall rate of isolation of MRSA from clinical specimens.77 ¶Nevada.86 ||For level 3 neonatal units only.81 **Orthopaedic SSI only.87 ††Missouri.43 ‡‡In England, NHS hospital trusts must show full adherence to a national code of practice,88 which includes an extensive framework of processes and practice. Compliance and inspection reports are made publicly available. §§New Hampshire.89
Modified from Haustein et al.59
Chapter 2: Literature Review 39
In Australia, one HAI outcome measure, and one process measure are now
routinely reported. Following a broad consultative process, annual hospital
identifiable SAB rates have been publicly reported since 2012113,114 Hand hygiene
compliance rates by hospitals are also reported publicly.102 This process indicator
data has now become embedded in the Australian healthcare setting since
commencing in 2009, and has been associated with a reduction of SAB.115
Public reporting of HAI data attracts contrasting opinions. Proponents argue it
promotes transparency, motivates organisations to implement best practice, and
ultimately improves patient outcomes.116,117 It is also suggested that publicly reported
HAI data can be used by consumers to make informed choices when deciding which
hospitals to attend.118 Although there is little evidence of a direct effect on improved
patient outcomes, public reporting has been associated with organisational change
and increased awareness of infection prevention.59 Humphries119 argues that public
reporting of national data as a benchmark not only drives down infection rates in
hospitals within a country, but benchmarking between countries can also serve to
drive improvements.
Opponents argue that mandatory public reporting, particularly of outcome data,
is flawed due to the variability in measurements between hospitals, and the
competition it creates between hospitals places undue pressure on infection control
teams.111,112,120
In Australia, although reporting of SAB and hand hygiene compliance data is
now considered routine, there has been criticism of the lack of validation and
appropriate risk stratification of SAB data.100 Further, the resources required to
sustain the mandated volume of hand hygiene auditing has also been criticised.121
Aware of early concerns relating to public reporting of HAI data, in 2005 the
Healthcare Infection Control Practices Advisory Committee at the CDC, developed a
series of recommendations for policy makers when considering statewide public
reporting of HAIs.109 They included sound epidemiological methods, risk
adjustment, and suggested using a combination of process and outcome data for the
“production of useful reports for stakeholders”.109 Process measures are considered
ideal for public reporting and hospital performance measurement as they do not
require any risk adjustment.109 Outcome measures require appropriate risk
adjustment for comparison, without which they are prone to misinterpretation.109
40 Chapter 2: Literature Review
In a review of public reporting across Europe, Martin et al.111 noted that debate
continues about the utility of public reporting, and doubt as to whether the public are
able to interpret HAI data appropriately. Kiernan122 notes that even if public
reporting is not particularly useful for the public, it captures the attention of
politicians and organisations, which can then translate into action.
Given the momentum of public reporting HAI data internationally, it is
reasonable to assume that it will also continue to expand in Australia. Therefore the
discussion now is not so much about whether or not HAI data should be publicly
reported, but rather how it should be reported, and which HAI data are suitable to
be used as a performance measurement.
In a survey of infection prevention leaders from 34 European member
countries, despite general support for public reporting of HAI data and it being
considered a major driver to strengthen infection prevention in hospitals, there was
strong disagreement about the benefits of public reporting, as well as the type of data
and format of the reports. The expert group conceded that benchmarking needed to
be accompanied by standardised methods and validation, and preferred reporting of
process indicators over outcome indicators.111
By comparison, a more recent review of healthcare performance measures
undertaken by Berenson, Pronovost and Krumholz110 recommended moving from
process measures to outcome measures. Whilst acknowledging the many challenges
of outcome measures as performance indicators, the authors state that process
measures do not always predict outcomes, and often require resource intensive,
manual data collection.110
Acknowledging the concerns when using HAI data as performance indicators,
namely lack of objectivity in applying infection definitions and insufficient risk
adjustment, the Healthcare Infection Control Practices Advisory Committee have
produced recommendations for public reporting of HAI data112 as an adjunct to their
2005 recommendations.109 Whilst not specifically addressing which infections are
suitable for public reporting, the recommendations highlight uniformity of
definitions, acknowledge the difference between surveillance and clinical definitions
may result in discordance, and that the final decision of determination must rest with
infection prevention teams. The recommendations then emphasise validation of
reported data, and recommend clear documentation of decision making in
Chapter 2: Literature Review 41
determining presence of infection, external audit, and a review of any claims
regarding potential under-reporting.112
In summary, the demand for public reporting of HAI data appears to be
increasing. Regardless of any perceived informed decision making benefits for
patients, their use as a hospital performance indicator will likely continue. This
further emphasises the epidemiological surveillance principles for uniformity and
standardisation and appropriate risk adjustment that are key to any good HAI
surveillance program.
2.9 DATA QUALITY
2.9.1 Accuracy
Crucial to any surveillance program is the accuracy of the data. Determining
the accuracy of HAI surveillance data is commonly done using three measures,
sensitivity, specificity and PPV.28 Sensitivity provides a measure of the proportion of
people with true infection who are reported as having an infection, specificity refers
to the proportion of people without an infection who are reported as not having an
infection, and the PPV measures the proportion of people reported as having an
infection who do have a true infection.123
In practical terms, if a surveillance program reports a high sensitivity and a low
specificity, this means that most patients with an infection are captured, but the low
specificity means that many patients who don’t have in infection will be reported that
they do have an infection. The PPV is influenced by the sensitivity and specificity,
and the prevalence of the HAI. A low PPV will result in non HAIs being
investigated, meaning that surveillance resources are being wasted, but may also lead
to the implementation of unnecessary interventions.25
It is recommended that independent, trained observers be engaged to measure
the sensitivity, specificity and PPV of a surveillance program.28 However such
validation studies are expensive to conduct, have inherent methodological difficulties
and often tend to focus on one aspect of data collection.124
Nevertheless, the advent of public reporting, benchmarking, and the potential
for HAI outcome data to be linked to hospital funding, the importance of validity and
reliability of HAI data has increased.120,125,126
42 Chapter 2: Literature Review
Emori127 first measured the accuracy of reporting ICU HAIs to the NNIS
program in 1998 and identified a sensitivity range of 30%-85% and a PPV range of
72%-87% for prospectively identified HAIs.127 A good specificity of over 98% was
reported for all HAIs. Their encouraging conclusion was that when an ICU HAI is
reported it is likely to be a true HAI, patients who do not acquire a HAI are identified
accurately, however because of the low sensitivity reported it is likely that some
HAIs were not being identified. To address this Emori127 recommended the need for
the training of data collectors to facilitate consistent application of infection criteria.
In a review of fourteen validation studies on HAI surveillance, Fabry et al.124
noted large variation in designs, studies were often limited to one or a small number
of facilities and were often focussed on one aspect of surveillance. In the studies
under review, many had similar findings to Emori127 with low to moderate sensitivity
and high specificity. When PPV was estimated they were generally high.124
Whilst large validation studies on national HAI programs are complex to
conduct, in the USA several states have undertaken their own validation studies.
Horan et al.65 from the NHSN reported that at May 2011, at least 15 states in the
USA had conducted validation studies, but again with variable results, supporting
findings from other similar validation studies.128-130
Comparable results have been found in Australian validation studies that have
been performed in two statewide HAI surveillance programs.94,131,132 In a review of
SSI reported from over 4,500 coronary artery bypass surgery procedures under
surveillance as part of the VICNISS, Friedman et al.131 found a PPV of 96%, but a
sensitivity of 55% and a specificity of 100%. When the review was limited to only
those infections that occurred in the sternum, (as opposed to sternum and graft site),
the PPV was 91%, sensitivity 62% and specificity 100%.131 These results implied
that not all SSIs, particularly those not at the sternal site, are identified.
Another study conducted by researchers from VICNISS on ICU CLABSI
surveillance estimated the sensitivity to be 35%, specificity 87% and PPV 59%.132
These findings revealed poor accuracy and consistency from those participating in
ICU CLABSI surveillance in the VICNISS program.132
In Western Australia, a review of all SAB events reported by public hospitals
identified 164 that were classified as healthcare-associated events during 2008.94 On
Chapter 2: Literature Review 43
review of the medical records of each notified case, researchers estimated that the
overall sensitivity was 77% and specificity100%, and in hospitals that did not have
an on-site microbiology service, the sensitivity was only 40%.94
2.9.2 Method Variation
There are other influences that can affect the quality of surveillance data, such
as variation in methodology. In a review of HAI national surveillance programs in
ten countries and one multinational program (HELICS) which all report using the
CDC/NHSN definitions and methods, variation was identified in the type of surgical
procedures under surveillance and length of time of follow up.133 Further differences
were also found in data collection methods, category of staff performing
surveillance, prospective and retrospective data collection methods, data sources, and
the inclusion of routine post discharge surveillance as a routine part of case
finding.133 It was also noted that validation of data did not occur on a regular basis.
This, together with the differences identified between the programs, contributes some
uncertainty about the quality of the data and also limits the ability to make
comparison of rates between different programs, despite being based on the same
methodology.133
In a cross sectional study of 126 hospitals designed to characterise variation in
surveillance methods and application of HAI definitions, Keller et al.134 used a series
of clinical vignettes to measure variation amongst infection prevention staff. Despite
all sites participating in and following NHSN methods, only 61% responses correctly
identified a HAI. Interestingly, 24% of those collecting HAI data did not have a
clinical background, which on multivariate analysis was an independent predictor of
an incorrect application of the HAIs definitions.134
As mentioned earlier in the review of the German KISS program, researchers
recently demonstrated a sensitivity of 85.7% by measuring the sensitivity and
specificity of 189 surveillance personal through a series of clinical vignettes
presented over a period of three years.74 Accuracy was positively associated with
surveillance experience and higher education levels.74
In a large ethnographic study across 17 ICU’s all participating in the same HAI
surveillance program, Dixon-Woods et al.135 illustrated broad variation in data
collection systems and the application of infection criteria. Dixon-Woods et al.135
44 Chapter 2: Literature Review
concluded that HAI data reported and used as hospital performance indicators clearly
misrepresented real infection rates. Rather than any deliberate attempt to game data,
the study identified that those involved in surveillance occasionally disagreed with
the standardised definitions and applied local interpretations largely because of
inequity aversion i.e. a dislike of unfair outcomes.135 This is an important finding
particularly if penalties are associated with infection rates.
In a cross sectional survey of 106 hospitals participating in mandatory
orthopaedic surgical site surveillance in the UK, Tanner et al.136 identified variation
in a number of areas including definitions applied, and the extent of post discharge
follow up. Not surprisingly it was noted that those who conducted inpatient
surveillance alone, and those who conducted inpatient surveillance and readmission
surveillance reported significantly lower SSI rates than those who also undertook
post discharge surveillance.136 Furthermore, the methods used to identify HAIs on
readmission and through post discharge surveillance varied enormously, also
affecting the reported rates.136
In an era of increased public reporting and performance measurement,
validation studies highlight the limitations in interpreting HAI data, and despite
participating in networked surveillance programs, variations in surveillance methods
between facilities continue. It is reasonable to expect similar findings amongst
Australian facilities, but studies are lacking that describe and measure this variation,
and if such variation has any impact on reported HAI rates.
Data accuracy and surveillance methods play a major role in the reliability of a
surveillance program. Misinterpretation of surveillance definitions and inconsistent
surveillance methods are the primary reasons for misclassification of infections.137
Also influencing the quality of the data include who collects the data, who applies
the definitions, the skill of those collecting the data, the data sources, the intensity of
case finding and the activities under surveillance.
It remains unclear what level of accuracy is acceptable, and what level of
resources and effort can be justified to provide high levels of accuracy.
2.10 DISCRETE CHOICE EXPERIMENTS
Data from HAI surveillance can serve several purposes, and will be used by
different stakeholders who possibly have different preferences on what they consider
Chapter 2: Literature Review 45
to be a good HAI surveillance program. Discrete choice experiments are an emerging
method being applied in a variety of situations to identify preferences.
Discrete choice experiments (DCE) have their origins in mathematical
psychology, and have been used in marketing, transport, environmental economics
and more recently have been commonly used in health economics.138,139 DCEs are a
quantitative attribute based survey method, and can be used to elicit preferences for
healthcare products, interventions, services, policies or programs.140-142 They are a
form of stated preference measurement where participants say what they would
prefer rather than being observed what they prefer, such as occurs in revealed
preference measurements.143,144
Typically, DCEs offer participants hypothetical scenarios that vary along
several characteristics or attributes. The participants are required to choose one
scenario in favour of the other.145 The technique has been used to value health
outcomes, investigate trade-offs between these outcomes and recently to estimate
utility weights of quality adjusted life years.141
DCEs have been described as the simplest of choice techniques.146 The low
cognitive complexity required to participate in a DCE is considered a big advantage
when compared with other choice techniques.146 DCEs are considered superior to
ranking and rating methods as they provide quantitative data on the strength of
preferences and trade off, and probability of take up.140
There is general agreement that there are distinct stages, or components of a
DCE. Lancsar and Louviere142 identify three components; an experimental design
used to generate choice data, a discrete choice analysis to estimate preferences, and
use of the resulting model to obtain welfare measures and policy analysis. Others
propose five stages which include; identification of attributes, identification of levels,
the experimental design, data collection and data analysis.140,146 Each of these stages
are addressed below.
2.10.1 Identification of attributes and levels
The DCE is characterised by a proposal of alternatives of the state of the good
described. These descriptions are called attributes of the alternative.146 Each attribute
consists of levels, which offer variation in the alternatives of the choice sets.146 A
simple example can be demonstrated when deciding between two job choices based
46 Chapter 2: Literature Review
on location, salary and opportunity (attributes). Job A is located 5 kms away, pays
$20/hr and offers good opportunity for promotion. Job B is located 20 kms away,
pays $27/hr and offers limited opportunity for promotion. So the levels for the
attribute of location are 5kms and 20kms, for salary the levels are $20 and $27, and
for opportunity the levels are good and limited.
Attributes and levels can be qualitative or quantitative and are usually
identified through literature reviews and qualitative research.142 Whilst there is no
one way to define an attribute, it is generally agreed attributes need to be relevant to
the requirements of the policy makers, plausible, meaningful and important to the
respondents.142,146
The number of attributes and levels are an important consideration when
designing a DCE. It is generally advised that the number of attributes be kept to a
minimum as the higher the number the more choices will be generated.140 For
example, if a DCE has 6 attributes each with 3 levels, the total number of choices
will be 36[3x3x3x3x3x3=729] resulting in 729 possible combinations. Although
there is no limit on the number of attributes, a 2012 review of 144 DCE studies
identified that 70% had 4-6 attributes.141 Even when attributes are kept to a
minimum, the number of choices can still be too many to present to respondents.
This issue is managed in the next stage.
2.10.2 Experimental design
Once the attributes and levels have been established, the choices with different
combinations of attributes and levels must be constructed. If all the combinations of
choices (also called the full factorial) are too great to present to respondents, a
statistical design theory is commonly used to draw independent samples of choices
from the full factorial.147 The resultant sample is called a fractional factorial. The aim
with a fractional factorial design is to ensure the properties of the full factorial are
maintained and the effects of interest can be estimated as efficiently as possible.146
Lanscar and Louviere142 argue that fractional factorial designs should be avoided and
the largest possible design should be implemented. They suggest putting
combinations into different blocks and randomly assigning respondents to different
blocks.142
Chapter 2: Literature Review 47
Whilst acknowledging that a full factorial design has attractive statistical
properties, Kjaer explains that in reality the full factorial can only be use in very
small experiments, and the practical solution is to use fractional factorial technique,
even though there will be loss of statistical information.140,146 When using a
minimal overlap and utility balance must be considered to optimise design
efficiency.140,146
In the 2012 review by De Bekker-Grob et al.,141 all 114 (100%) of the studies
reviewed used a fractional factorial design. This compared with 74% from an earlier
review of 34 studies in 2003.147
2.10.3 Data Collection
To administer a DCE, choice sets comprising two or more alternatives which
vary in attribute levels are presented to respondents who are required to select one
alternative.145 Standard pilot tests of the choice sets are required to test respondents
understanding of choices and levels, appropriateness, complexity and timing.142
When presenting the choice sets, careful consideration must be given to the
contextual introduction.145,146 A major consideration is if an “opt out” alternative is
provided. The omission of an “opt out’ alternative forces the respondent to select an
option that might not be a clear preference and so introduces bias.140 The inclusion of
an “opt out” alternative is appropriate in many situations but it is important that
respondents understand that selection of the “opt out” alternative indicates they are
happy with the current status.140 One clear issue with providing an “opt out”
alternative is that if the respondent feels that the choice task is too cognitively
demanding they may select the “opt out” to prevent making difficult choices,146
introducing another bias.
It is also recommended that “warm up” choices be offered to familiarise
respondents with the method. This also provides some internal consistency testing by
constructing choices where one alternative is clearly dominant over another.140
The survey can be delivered in a variety of ways; face to face interviews,
telephone interviews, mailed questionnaires, internet/email or a combination,146 each
with their own advantages. Hand delivered or mailed self completed questionnaires
have been used most commonly,147 however the use of computer based surveys is
48 Chapter 2: Literature Review
expected to become more widespread given its ability to collect large amounts of
data at little expense.146
2.10.4 Data analysis
Analysis of data derived from the DCE is based on the random utility model.
To estimate the strength of the preferences and levels selected by respondents,
several models can be used including random effects probit and logit, conditional
logic and mixed logic.140 Random effects probit has been found to be the most
commonly used model.141 Variation between preferences based on respondents’
characteristics (e.g. profession, qualifications, work location) can also be estimated.
The analysis can be used to determine which attributes are most important and
preferred by the participants, and the strength of this preference in comparison to
another, and how willing respondents are to trade between attributes.148
Whilst use of DCE for eliciting patient and clinician preferences for healthcare
prevention, testing and treatment options is well described,141 the use of DCE in the
infection prevention setting has not previously been described.
2.11 IMPLEMENTATION SCIENCE
It is not within the scope of this PhD to comprehensively review literature on
implementation science, but rather explore its potential application in infection
prevention, particularly considering a new surveillance program, by introducing
some major concepts.
Despite research evidence to support specific practices in health care, many fail
to translate into improved patient outcomes.149 Examples of this include that despite
clear evidence on reducing HAIs, universally hand hygiene compliance rates remain
relatively low, and the uptake of infection prevention ‘bundles’ to prevent central
line associated blood stream infections remains low.150 To address the gap between
research and practice, healthcare is turning towards implementation strategies.
Implementation science is defined as “the study of methods to promote the
integration of research findings and evidence into healthcare policy and practice”,151
and aims to explore health care worker behaviour with respect to adoption and
maintaining interventions. Health researchers are now being encouraged to not only
Chapter 2: Literature Review 49
examine the endpoint outcomes of their interventions, but to also consider the
implementation of any intervention.149
Specific to infection prevention, Pronovost et al.152 developed a model for
translating research into practice, which when applied successfully, resulted in a
large and sustained reduction in central line associated bloodstream infections.153 The
model focuses on systems, engagement and ownership, support, adaptation and
collaboration.152 It is acknowledged that successful application of the model requires
substantial resources and best suited to large scale projects.152 Pronovosts’ model
underpinned a landmark study in implementing an evidence based bundle which
resulted in a significant and sustained reduction of catheter related bloodstream
infections across 103 intensive care units.153
Various implementation theories have been published over the last decade,
however not all can be generalised and many have overlapping constructs.149 In order
to identify which frameworks may be suited to specific situations, Damschorder et
al.149 reviewed 19 different implementation theories described in the literature, and
from that established the Consolidated Framework for Implementation Research
(CFIR). The CFIR comprises five domains: intervention characteristics, outer setting,
inner setting, characteristics of the individuals involved, and the process of
implementation. Within each of these domains are a number of different constructs
that may or may not be applicable to certain interventions. It offers a pragmatic
foundation to assist in the understanding of the various influences that must be
considered in implementation.149
An emerging implementation theory being used in health is the Normalisation
Process Theory (NPT) developed by May et al.154 NPT comprises four constructs,
Coherence, Cognitive participation, Collective Action and Reflexive Monitoring.
The attraction of the NPT is its emphasis on complex healthcare interventions, and
its utility in the planning and development stages of the intervention, as well as
embedding and evaluating the intervention.155 It also has a strong emphasis on key
stakeholder engagement and relationships between stakeholders. 155,156
Clearly the development and implementation of a national HAI surveillance
program is a complex intervention. It involves many stakeholders including
consumers, healthcare workers, clinicians, executive staff, and government staff.
Issues regarding engagement, education, new or modified practices, and outcomes all
50 Chapter 2: Literature Review
warrant careful consideration. Many of these aspects are potential barriers and
enablers of a new HAI surveillance program. It is therefore crucial to understand
exactly what these barriers and enablers are, and therefore include in an
implementation strategy.
2.12 CONCLUSION
This review has explored the literature on HAI surveillance programs and key
issues associated with them. Particular attention has focussed on the utility of HAI
surveillance and the benefits of national HAI surveillance in improving our
understanding of the epidemiology of HAIs. This ultimately informs infection
prevention interventions and reduces the incidence of HAI. Although it is difficult to
precisely measure the effect of surveillance alone, reductions in HAIs following
implementation of surveillance programs is evident from a broad range of literature.
The review has revealed a number of gaps that need to be addressed when
considering an Australian national surveillance program. These include uncertainty
regarding:
• exactly what surveillance activities and methods are being undertaken,
• suitability of any existing programs for expanding as a national program
• the extent to which current practices reflect best practice
• how well the data is being used to implement infection prevention strategy
• the accuracy of the existing data
• how much surveillance training is delivered
• agreement levels in identifying and classifying HAIs
• the suitability of data for comparing facilities and benchmarking
The literature relating to public reporting and data quality highlighted the
momentum towards publicly reporting of HAI data. This means it is imperative that
data quality be constantly monitored to add credibility to the surveillance program
particularly in light of HAI data being used for hospitals performance measurements.
Further, although the CDC guidelines for evaluation public health surveillance
programs highlight several attributes which provide a framework for evaluation, little
is known about barriers and enablers when it comes to developing and implementing
Chapter 2: Literature Review 51
a surveillance program, and what the characteristics of well established programs are.
The fact that Australia currently doesn’t have a program can be used as an advantage
as it provides the opportunity to identify what stakeholders want from a national
program.
To assist in filling these gaps, literature about the proposed method for the
second study, a discrete choice experiment, was presented. Although DCEs have
been used in health settings previously, this will be the first time it will be applied in
an infection prevention setting.
To finish the review, the relevant literature regarding implementation science
and infection prevention was introduced as an important consideration in the
development of a national HAI surveillance program. An appropriate implementation
strategy is crucial to ensure appropriate translation of research to practice.
Chapter 3: The research questions and study design 53
Chapter 3: The research questions and study design
Four research questions and two studies form the basis of this work. The first
two questions were specific to the first study, and questions three and four were
specific to the second study.
Each of the questions are listed below with an explanation as to why they are
important. Following the questions for each study, I have provided a description of
the study design that was used to answer the questions.
The overall approach to answer the questions was a mixed methods design.
Mixed methods research involves both qualitative and quantitative approaches, either
in a single study or in multiple phases of a program of study.157 Mixed methods has
become increasingly popular as its strengths offsets the weaknesses of both
qualitative and quantitative methods,157 and provides a more rounded understanding
of the issue at hand than either qualitative and quantitative methods alone.
Whilst the first study in this PhD comprised a cross sectional survey generating
quantitative data, the discrete choice experiment, the method used in second study,
used a mixed methods approach where qualitative data were used to inform the
design of the experiment which was analysed using quantitative methods. This is
explained in more detail in Chapter 8.
A mixed methods approach was appropriate for this body of research as HAI
surveillance is a complex process. As well as collecting and analysing data, the
establishment of successful surveillance programs require an understanding of how
people collect the data, what resources are required, what are the enablers and
barriers, what level of support is required, how the data are used and how a new
or revised program is implemented.
3.1 RESEARCH QUESTION 1
What are the similarities and differences between existing HAI surveillance
processes in Australia?
Chapter 3: The research questions and study design 54
The origins of this question come from the knowledge that not all states and
territories have coordinated surveillance programs, so it is important to understand
how similar the existing programs are, and what is occurring where there is no
coordinated program. If there are no differences in the coordinated programs, or if
differences are only minor, then a national program may only require coordination of
current activities. However, if the differences are broad and major, then
consideration would need to be given to a more comprehensive review and
development of a new program.
This question is specific to processes within a surveillance program including
definitions used, data sources, collection, analysis and reporting. The answer to this
question will identify any gaps in current practices and provide an understanding of
the effort required to develop a national program.
3.2 RESEARCH QUESTION 2
What level of agreement exists in the identification of HAI between those
participating in HAI surveillance, and are there any factors that influence agreement
level?
Several validation studies have described poor to moderate agreement amongst
those involved in surveillance when it comes to identifying HAIs in Australia. 94,131,132,158 The answer to this question will build upon knowledge gained from
question 1 that looks at processes, and attempts to quantify differences by using
clinical vignettes. Outcome from the vignettes will provide information on the effect
of any differences identified in answering question 1 has on outcome data. It will
improve our understanding on the quality of data currently being reported, and
further contribute knowledge in identifying what work needs to be done to develop a
national program.
3.3 STUDY 1 – CROSS SECTIONAL SURVEY: CURRENT AUSTRALIAN HOSPITAL PRACTICES IN HEALTHCARE-ASSOCIATED INFECTION SURVEILLANCE
The overall aim of this study was to improve our understanding of the current
status of HAI surveillance practices in Australia. To do this, a cross sectional survey
was conducted with infection prevention staff who undertook surveillance. Previous
surveys of Australian infection prevention staff have not provided the level of detail
Chapter 3: The research questions and study design 55
this survey will collect, with the most recent survey being completed in 2008.103,104
Novel to this survey is the inclusion of clinical vignettes. The findings from this
study answered research questions 1 and 2.
3.3.1 Study 1 design
An online survey was constructed which sought data on the characteristics of
infection prevention staff who undertake surveillance, their surveillance practices
and the characteristics of the environment in which surveillance is undertaken.
Within the survey, a series of seven clinical vignettes describing potential HAIs were
included. The vignettes sought to explore agreement of HAI identification,
classification and calculation of rates. The vignettes were constructed in
collaboration with infection prevention experts from a jurisdictional surveillance
program. The final survey consisted a total of 88 items, however no respondents
were required to answer all as the logical design guided participants to questions that
were specific to their work environment and the type of surveillance they undertook.
Recruitment of participants used the snowballing method through the
Australasian College of Infection Prevention and Control (ACIPC) list server called
“Infexion Connexion”, which over 500 ACIPC members subscribe to. List server
subscribers received an email describing the study with a link to the survey.
Recipients were asked to forward the email on to all involved in surveillance.
Data were analysed using Stata, version 13 (Stata Corp, College Station,
Texas). The chi square test was performed to compare proportions between
groups, and Kruskall-Wallis to test for influence of State and Territory.
For the analysis of the vignettes, univariate logistic regression was used to
identify any characteristics that influenced agreement levels. A multivariable Poisson
model of the total number correct was developed from characteristics identified in
the Poisson univariate analysis. Further analysis was undertaken to assess any
multicollinearity.
3.4 RESEARCH QUESTION 3
What are the key components of successful centrally coordinated HAI
surveillance programs?
Chapter 3: The research questions and study design 56
As I have identified from the literature, there are many well established
international HAI surveillance programs. When considering a new national program,
it is crucial to explore existing large surveillance programs to gain knowledge on a
broad range of issues such as how they commenced, implementation characteristics
associated with successful programs, barriers and enablers for engagement and
implementation, resource requirements, data usage, and strategies for long term
sustainability. Knowledge gained was used to inform recommendations for a national
program in Australia. Importantly, answers to this question were used in the
construction of the discrete choice experiment.
3.5 RESEARCH QUESTION 4
What are the preferences and priorities of key stakeholders when considering a
national HAI surveillance program?
The final question is specific to the Australian environment. There are many
elements involved in a national surveillance program including type of infections
under surveillance, data collection method, surveillance staff skill and competency,
and how the data are used. Even though an ideal program may be devised in
theory, practical success will largely depend on stakeholders belief in the value
of the program. Therefore this question attempts to identify what stakeholders want
from a national surveillance program, which elements of the program they
consider most important and those they consider least important.
3.6 STUDY 2 – PREFERENCES FOR A HEALTHCARE-ASSOCIATED INFECTION SURVEILLANCE PROGRAM USING A DISCRETE CHOICE EXPERIMENT
The overall aim of this study was to identify what type of national HAI
surveillance program stakeholders in Australia want. To do this we undertook a DCE
to elicit stakeholder preferences for a national HAI surveillance program. DCEs have
been used in health settings previously, however its use in an infection prevention
setting was novel. The outcomes of this study answered research questions 3 and 4.
3.6.1 Study 2 design
A crucial step in the development of the DCE was the identification of key
characteristics of a surveillance program. Identifying attributes for a DCE commonly
requires a literature review and the application of qualitative methods such as
Chapter 3: The research questions and study design 57
interview or focus groups.142 For this part of the study, a review of the literature was
undertaken, and a series of semi-structured interviews were conducted with three
leaders from Australian statewide programs, and four from international HAI
surveillance programs. Participants were selected because of their leadership and
experience in developing, implementing and maintaining large surveillance
programs. Qualitative analysis of the data generated attributes and levels for
inclusion in the DCE.
To construct the DCE, advice was provided by Professor Julie Ratcliffe,
Professor in Health Economics, and Dr Gang Chen, Research Fellow, at Flinders
University, Adelaide. Professor Ratcliffe and Dr Chen have strong backgrounds in
DCEs and have conducted research using DCEs in a variety of settings.
The identification of five attributes and their corresponding levels resulted in a
total of over 23,000 possible choice questions. Clearly this was too many to include
in a survey. Therefore a D-efficient design was used to reduce the number of choice
scenarios into a more pragmatic number. A series of hypothetical surveillance
program scenarios were created where participants were required to choose one
surveillance program over another. The model consisted of two blocks of 12 pair
wise choice questions. To test for internal consistency, one choice question was
duplicated in each block resulting in 13 in each block. Participants were randomised
into one of the blocks.
To allow for subgroup analysis, demographic data such as gender, age,
qualifications and occupation was collected. A series of attitudinal questions relating
to HAI surveillance were also included.
A total of 184 participants were purposively selected to participate based on
their senior leadership role in infection prevention in Australia. Reminder emails
were sent out to encourage participants to complete the DCE over a five week period.
Data was analysed using Stata, version 13 (Stata Corp, College Station, Texas)
and a mixed logit models applied to identify and measure the strength of the
preferences by generating coefficients.
Although novel, the application of a DCE for constructing a HAI surveillance
system was favourable for a number of reasons:
Chapter 3: The research questions and study design 58
- HAI surveillance systems have several attributes that have been
identified in the literature, though others may exist generating new
knowledge
- The influence of different attributes on the outcome may vary
- It is unknown which attributes of a HAI surveillance system are
considered more important than others, or if those using the HAI
surveillance system may be willing to trade off between different
attributes depending on their priorities
- DCEs have been used for priority setting frameworks where decision
makers are required to manage competing demands with limited
resources.159-161 Given that surveillance of HAIs is just one of the many
activities that must be resourced from the infection prevention budget,
which itself competes with other hospital services, the DCE provides
new knowledge in this setting.
By offering choices of attributes that make up a HAI surveillance system to
stakeholders, attributes that are considered most important were identified, the
strength of these weightings and the willingness of stakeholders to trade off attributes
in favour of others was also identified.
This has provided crucial information in constructing a HAI surveillance
program, and was used to support the evidence based recommendations for a national
HAI surveillance program.
3.7 ETHICS AND LIMITATIONS
Both studies were considered negligible/low risk research. Three ethics
approvals were granted. The first for the cross sectional survey, the second for the
semi-structured interviews, and third for the DCE. Ethics approval was granted from
the QUT University Human Research Ethics Committee (see Appendices B, C and
D)
Chapter 4: Healthcare-associated infection in Australia 59
Chapter 4: Healthcare-associated infection in Australia
4.1 INTRODUCTION
To help inform the aims and design of the first study on the current
surveillance practices of infection prevention staff in Australia (presented in
Chapters 5 and 6), a scoping review was undertaken of existing statewide Australian
surveillance activities, specifically observing the type of infections under
surveillance and the level of coordination of activities that occurs, and well
established international HAI surveillance programs
This scoping review identified disparity across many aspects of surveillance in
Australia. Some statewide surveillance programs have been introduced over a period
of time and evolved at different rates. Although some common factors were
identified, those that do have statewide programs mandate surveillance on a different
range of HAIs.
This review highlights the benefits of a national surveillance program as
demonstrated in international programs, identifies disparity in existing Australian
HAI surveillance, and outlines work necessary to establish a framework for a
national HAI surveillance program in Australia.
The findings of this review were published in the Australian Health Review
journal.
Chapter 4: Healthcare-associated infection in Australia 60
Statement of Contribution of Co-Authors for Thesis by Published Paper
The authors listed below have certified* that:
• they meet the criteria for authorship in that they have participated in the
conception, execution, or interpretation, of at least that part of the publication
in their field of expertise;
• they take public responsibility for their part of the publication, except for the
responsible author who accepts overall responsibility for the publication;
• there are no other authors of the publication according to these criteria;
• potential conflicts of interest have been disclosed to (a) granting bodies, (b)
the editor or publisher of journals or other publications, and (c) the head of
the responsible academic unit, and
• they agree to the use of the publication in the student’s thesis and its
publication on the Australasian Research Online database consistent with
any limitations set by publisher requirements.
In the case of this chapter:
Publication title and date of publication or status:
Validation of statewide surveillance system data on central line-associated
bloodstream infection in intensive care units in Australia. Infect Control
Hosp Epidemiol. 2009 Nov;30(11):1045-9.
48. Australian Institute for Health and Welfare. MyHospitals [15/03/2013].
Available from: http://www.myhospitals.gov.au.
49. Cheng AC. How should we interpret hospital infection statistics? MJA. 2014
December 12;199(11):735-6.
50. Worth L, Thursky, K.A., Slavin, M.A. Public disclosure of health care-
associated infections in Australia: quality improvement or parody? MJA.
2012;197(1):29.
51. Worth LJ, Bull, Ann L., Richards, MJ. Public reporting of health care-
associated infection data in Australia: time to refine. MJA.
2013;198(5):252-3.
52. Shaban RZ CM, Christiansen K & the Antimicrobial Resistance Standing
Committee, . National Surveillance and Reporting of Antimicrobial
Resistance and Antibiotic Usage for Human Health in Australia. 2013.
53. Cardo D, Dennehy PH, Halverson P, Fishman N, Kohn M, Murphy CL, et
al. Moving toward elimination of healthcare-associated infections: a call to
action. Infect Control Hosp Epidemiol. 2010 Nov;31(11):1101-5.
54. Mitchell BG, Gardner A. A model for influences on reliable and valid health
care-associated infection data. Am J Infect Control. 2014 Feb;42(2):190-2.
Chapter 4: Healthcare-associated infection in Australia 77
Table 1. Comparison of mandatory healthcare associated infection surveillance components in acute care public facilities by state
All states and territories in Australia undertake surveillance for Staphylococcus aureus bloodstream (SAB) infection and Clostridium difficile infection (CDI).
✔, surveillance performed; Ó, surveillance not performed; 1, with modifications; 2, including neonatal intensive care unit (NICU); 3, NICU only; 4, if >50
procedures per year; 5, Royal Women’s hospitals and Women’s Mercy Hospital only; 6, infections only; ICU, intensive care unit; MRSA, methicillin-resistant
Staphylococcus aureus; MRAB, multi-resistant Acinetobacter; NHSN, National Health and Safety Network; BSI, bloodstream infection; NA, not applicable.
StatewideHAIsurveillanceprogram
Centrallineassociated
bloo
dstreamin
fectionsin
ICU
(includ
esperiphe
rallyin
serted
)
Acqu
isition
ofM
RSAinIC
UB
Acqu
isition
ofM
RABinIC
UB
Corona
ryArteryBy
passgraft
Hipprosthe
sis
Knee
prosthe
sis
LowerCaesarean
sectio
n
SSIP
ostD
ischargeSurveillan
ce
includ
ed–excep
tfor
read
mission
s
NHSN
definition
s
AllM
RSAinfections
Haemod
ialysisaccessassociated
bloo
dstreamin
infection
Hospitalw
ideBS
I
NSWHealthcareAssociatedInfectionsProgram
✔ ✔ ✔ ✔ ✔ ✔ ✖ ✖ ✔ (1) ✖ ✖ ✖
78 Chapter 4: Healthcare-associated infection in Australia
I have sighted email or other correspondence from all Co-authors confirming
their certifying authorship.
Name
Signature
Date
Dr Lisa Hall
11/7/2016
94 Chapter 6: Differences in identifying healthcare-associated infections
6.2 PAPER THREE: “DIFFERENCES IN IDENTIFYING HEALTHCARE-ASSOCIATED INFECTIONS USING CLINICAL VIGNETTES AND THE INFLUENCE OF RESPONDENT CHARACTERISTICS: A CROSS-SECTIONAL SURVEY OF AUSTRALIAN INFECTION PREVENTION STAFF”
Russo PL, Barnett AG, Cheng AC, Richards M, Graves N, Hall L. Differences
in identifying healthcare-associated infections using clinical vignettes and the
influence of respondent characteristics: a cross-sectional survey of Australian
infection prevention staff. Antimicrob Resist Infect Control 2015; 4(29): 1-7.
6.2.1 Abstract
Background
Australia has commenced public reporting and benchmarking of healthcare
associated infections (HAIs), despite not having a standardised national HAI
I am pleased to extend this letter of support for the Australasian College for InfectionPrevention andControl's (ACIPC) involvement in the above research. Youhave requestedaccesstoourDiscussionList,InfexionConnexion,withvoluntaryinvolvement;thedatawillbede-identified(meetingchangestothePrivacyAct2014)andensureconfidentiality.
Australasian College for Infection Prevention and Control Ltd GPO Box 3254 Brisbane Qld 4001 . ABN 61 154 341 036
P + 61 7 3211 4695 F + 61 7 3211 4900 E [email protected] W www.acipc.org.au
Kindregards
226 Appendices
Appendix F: Survey tool - Current Australian hospital practices in healthcare-associated infection surveillance
Hello and welcome to the HAI surveillance survey!
Description This project is being undertaken as part of a PhD for Philip Russo The purpose of this project is to to identify and describe the differences between the healthcare associated infection (HAI) surveillance programs in Australia, measure agreement between clinicians when identifying HAI's and identify factors that may influence agreement levels. You are invited to participate in this project because you are an infection prevention and control professional involved in HAI surveillance.
Participation Participation will involve completing an 88 item, anonymous questionnaire with likert scale answers (strongly agree – strongly disagree), multiple choice answers, and a series of clinical vignettes that will take approximately 20 minutes of your time. Questions will include “Do you undertake post discharge surgical site infection surveillance?”, “Which surgical site infection definitions do you use?”. Your participation in this project is entirely voluntary. If you agree to participate, please note some questions must be answered before you can progress to the next. You do not have to have to complete any question(s) or the survey if you are uncomfortable answering. Your decision to participate or not participate will in no way impact upon your current or future relationship with QUT. If you do agree to participate it will not be possible to withdraw, once you have submitted your responses.
Expected benefits It is expected that this project will directly benefit you. Data from this study will be used to inform the larger Research Project to identify evidence based practices for national HAI surveillance. A national HAI surveillance program will result in uniform methodology and reporting. This will close the current gap we have with current systems and ensure we are measuring infections the same way across Australia. This will improve our understanding of the epidemiology of HAIs in Australia and enable meaningful national comparisons of HAI rates by hospital size, type, specialty and potentially by specific patient risk factors. Detailed data will enable us to identify problem areas that require more infection prevention resources and target interventions. Successful interventions could act as role models and also inform State and national policy
Risks There are no risks beyond normal day-to-day living associated with your participation in this project.
Privacy and Confidentiality All comments and responses are anonymous and will be treated confidentially unless required by law. You are not asked to provide your name or any contact details. Any data collected as part of this project will be stored securely as per QUT’s Management of research data policy. Consent to Participate Submitting the completed online questionnaire is accepted as an indication of your consent to participate in this project.
Questions / further information about the project Ifyouhaveanyquestionsorrequirefurtherinformationpleasecontactoneoftheresearchteammembersbelow.
Concerns / complaints regarding the conduct of the project QUT is committed to research integrity and the ethical conduct of research projects. However, if you do have any concerns or complaints about the ethical conduct of the project you may contact the QUT Research Ethics Unit on [+61 7] 3138 5123 or email [email protected]. The QUT Research Ethics Unit is not connected with the research project and can facilitate a resolution to your concern in an impartial manner. This study has been approved by the QUT Human Research Ethics Committee (approval number 1400000339).
Section7-FUTURESURVEILLANCEPRIORITIESThinking about conducting surgical site infection surveillance, of the surgicalprocedureslistedbelow,selectuptoamaximumoffive(5)whichyoubelievearethemostimportantprocedurestoundertakesurveillanceforyourfacility?
Appendix G: Current Australian hospital practices in healthcare-associated infection surveillance: Frequency of access to other healthcare professionals –
data not included in Chapter 5
StaffCategoryn=104 Daily Weekly Lessthan
weekly Rarelyornever
InfectiousDiseasesPhysician
58% 15% 15% 13%
Microbiologist 63% 8% 12% 17%
IPstaffwithmoreexperience
44% 6% 8% 43%
Epidemiologist 11% 3% 4% 83%
Statistician 9% 5% 4% 83%
Administrativeassistance 36% 4% 3% 57%
242 Appendices
Appendix H: Current Australian hospital practices in healthcare-associated infection surveillance: Frequency of where HAI data are
reported – data not included in Chapter 5
Infectiontype N Clinicia
ns ICC S&Q HospitalExec
HospitalBoard
Consumers
Statebody
SSI 63 83% 100% 78% 84% 49% 25% 67%
CLABSI 55 78% 96% 80% 82% 47% 40% 69%
VAP 20 65% 40% 10% 15% 20% 5% 5%
CAUTI 20 70% 50% 30% 30% 25% 0% 0%
Appendices 243
Appendix I: Semi-structured interview guide for participants
Appendix L: Normalisation process theory questions
Normalisation process Theory constructs Questions to consider when applying NPT
Coherence - i.e meaning and sense making be participants • Is the intervention easy to describe? • Is it clearly distinct from other interventions? • Does it have a clear purpose for all relevant participants? • Do participants have a shared sense of its purpose? • What benefits will the interventions bring and to whom? • Are these benefits likely to be valued by potential participants? • Will it fit with overall goals and activity of the organisation?
Cognitive participation - i.e. commitment and engagement by participants
• Are target user groups likely to think the intervention is a good idea? • Will they see the point easily? • Will they be prepared to invest time, energy and work in it?
Collective action - i.e the work participants do to make the trial function
• How will the intervention affect the work of the user groups? • Will it promote or impede their work? • What effect will it have on consultations? • Will staff require extensive training before they can use it? • How compatible is it with existing work practices? • What impact will it have on division of labor, resources, power and
responsibility between different professional groups? • Will it fit with the overall goals and activity of the organisation?
Reflexive monitoring - i.e. participants reflect on or appraise the trial
• How are the users likely to perceive the intervention once it has been in use for a while?
• Is it likely to be perceived advantageous for patients and staff? • Will it be clear what effects the intervention has had? • Can users/staff contribute feedback about the intervention once it is in use? • Can the intervention be adapted/improved on the basis of experience?