2018
Sascha Korzec
AMC Klinische Informatiekunde
Furore Informatica B.V.
8/29/2018
Local outbreak analysis: Fighting antibiotic resistance
with Statistical Process Control
1 Local outbreak analysis: Fighting antibiotic resistance with statistical process control
Student Sascha Korzec Student number: 6060463 E-mail: Saschakorzec@gmail.com Tutor Dr. M.C. Schut Faculty of Medicine
Department of Medical Informatics, Amsterdam UMC Email: m.c.schut@amc.uva.nl
Mentor
Remco Piening, Msc.
Email: r.piening@furore.com
Implementatieconsultant, Furore Informatica B.V.
Mentor Dr. D. Budding
Department of Medical Microbiology and Infection control Amsterdam UMC Email: D.budding@VUmc.nl SRP address Furore Informatica B.V. Bos en Lommerplein 280 1055 RW Amsterdam Email: info@furore.com Period November 2017 – August 2018
2 Table of Contents
Preface & Acknowledgements ... 4
Abstract ... 5 Dutch Abstract ... 6 1. Introduction ... 7 1.1. Research Aim ... 8 1.2. Objectives ... 8 1.3. Research Questions ... 9 1.4. Approach ... 9 1.5. Expected results ... 9 1.6. Chapter organization ... 10
2. Background: Medical Microbiology and infection control ... 11
2.1. Bacteria and Fungi ... 11
2.2. Antibiotics ... 11
2.3. Antibiotic resistance ... 11
2.3.1. Minimum inhibitory concentration ... 11
2.3.2. Causal origin of antibiotic resistance ... 12
2.4. MDRO ... 12
2.4.1. Extended Spectrum Beta-Lactamase ... 12
2.4.2. MR/MR+ ... 13
2.4.3. Vancomycin resistant Enterococcus ... 13
2.4.4. Methicillin Resistant Staphylococcus Aureus ... 13
2.4.5. Carbapenemase producing organisms ... 13
2.4.6. Multi-resistant Acinetobacter ... 13
2.5. Outbreak management ... 14
3. Data, Software and Conversion ... 15
3.1. Research Data Platform ... 15
3.2. RDP data ... 16
3.3. TIBCO Spotfire: Data Visualization & Analytics Software ... 18
3.4. Data Conversion in Spotfire ... 18
3.5. Data in R ... 19
3.6. Annual reports ... 19
4. Background: Statistical Proces Control ...20
4.1. Outbreak detection: Method ...20
4.2. Statistical Process control ...20
4.3. Statistical Process Control with MDROs ... 21
4.4. Charts ... 21 4.5. Control chart ... 21 4.6. CUSUM chart ... 22 4.7. EWMA chart ... 22 5. Results ... 24 5.1. SPC analysis ... 24 5.2. Annual Reports ... 24
3
5.3. Comparison of annual reports and SPC analysis ... 0
6. Discussion ... 1
7. Conclusion ... 3
8. References ... 4
9. List of Abbreviations ... 8
10. Appendix ... 9
10.1. Appendix A: Geographical dashboard ... 10
10.2. Appendix B: MDRO classification ... 11
10.3. Appendix C: Spotfire data conversion ... 12
10.4. Appendix D: Outbreak dates... 13
4 PREFACE & ACKNOWLEDGEMENTS
In this scientific research project I combined theoretical and practical work to finish my Master thesis for Medical Informatics. I felt thrown in the deep, with a project uncertain of an outcome. While swimming in data I learned that if you just try you can achieve practical things, which is not a guarantee while studying. I hope the geographical dashboard that is developed will have an added value in clinical practice. It was altogether a very instructive journey.
I would like to thank the people that made it all possible. I would like to especially thank my tutor Martijn Schut and my mentors Dries Budding and Remco Piening. Every time I did not see the next step they could always point me into the right direction, even if they were very busy. You can always learn something, but it is most valuable that people instruct you in such a way you learn the most of it. Also I want to thank Annie Kaiser, Joffrey van Prehn, and Alex Wagemakers for a fruitful
collaboration in challenging data questions and helping me along the road in understanding medical microbiology and infection control.
Ofcourse a special thanks to my parents, who during my whole study always believed and supported me and special thanks to my girlfriend Eline for infinite support.
5 ABSTRACT
Introduction
The World Health Organisation alarms for increasing antibiotic resistance. One of the consequences of antibiotic resistance is the rise of multi drug resistant organisms. While there do not seem any good alternatives being developed for antibiotic resistance in the long run an aim is set to improve prevention of bacterial infection. There are a lot of different outbreak detection tools, with different scopes and algorithms, but no golden rule exist.
Methods
Statistical process control is used in present research to analyse the multi drug resistant organism outbreaks in a hospital. If an incidence of a multi drug resistant organism per week is outside control limits this is seen as an outbreak. To validate the found outbreaks annual reports of the medical microbiology and infection control department are used.
Results
Statistical process control identified all outbreak reported in the annual reports from the hospital; also 30% other outbreaks, which were not identified in the annual reports. Another finding is that one MRSA outbreak in 2017 was not found by SPC, but reported in the annual reports. Remaining, four out of seven type of multi drug resistant organisms were not usable for analysis, due to lack of incidence.
Discussion
Statistical process control seems a good tool to find extra outbreaks of multi drug resistant
organisms. However, it cannot replace regular infection control methods and requires high incidence of the MDRO that is investigated. Statistical process control combining with regular methods and including employee data in the future can help further improve outbreak detection of multi drug resistant organisms.
Keywords: Multidrug resistant organisms, Statistical Process control, Infection control, Research data platform
6 DUTCH ABSTRACT
Introductie
De wereld gezondheid organisatie stuurt geregeld alarmerende berichten over toenemende antibiotica resistentie. Een van de consequenties van toenemende antibiotica resistentie is de toename van bijzonder resistente micro organismen. Zolang er geen goede alternatieven worden ontwikkeld voor de toenemende resistentie richt dit onderzoek zich op preventie van bacteriële infectie. Er zijn veel verschillende uitbraak detectie systemen, van verschillende omvang en verschillende algoritmen.
Methode
Statistical process control wordt in onderhavig onderzoek gebruikt om de uitbraken van bijzonder resistente micro organismen te analyseren. Wanneer de incidentie van een bijzonder resistent micro organisme per week buiten de controle limiet valt is er sprake van een uitbraak. De gevonden uitbraken van statistical process control worden gevalideerd door deze te vergelijken met de gerapporteerde uitbraken uit de jaarverslagen van de infectiepreventie.
Resultaten
Statistical process control signaleerde alle uitbraken genoemd in de jaarverslagen van het ziekenhuis en 30% extra. Een ander resultaat van dit onderzoek is dat een MRSA uitbraak uit 2017, die wel in het jaarverslag wordt genoemd, niet door SPC is gevonden. Verder waren vier van de zeven bijzonder resistente micro organismen niet bruikbaar voor analyse door een te lage incidentie.
Discussie
Statistical process control lijkt een bruikbare statistiek om extra uitbraken bij bijzonder resistente micro organismen te vinden. Echter, Statistical process control kan reguliere infectiepreventie niet vervangen. Reguliere methoden zouden met Statistical process control gecombineerd kunnen worden samen met data van verplegers en artsen zodat er meer inzicht verschaft kan worden in bijzonder resistente micro organisme uitbraken.
Steekwoorden: Bijzonder resistente micro organismen, Statistical process control, Infectiepreventie, Research data platform
7 1. INTRODUCTION
Antimicrobial resistance is growing in an alarming rate. The World Health Organization (WHO) increases alarm for the growing antimicrobial resistance [1 - 3]. The WHO states that antibiotic resistance is one of the biggest threats to public health. A review of alternative therapies for antibiotics compared alternative antibiotic therapies to antibiotics that serve as preventive or complementary treatment at best [4]. The review on alternative therapies on antibiotics suggests that on the short run antibiotics are still needed in treatment with bacterial threats [4]. On the long run current antibiotic treatments seem not sufficient and alternatives have to be investigated more.
A way to find bacterial threats is by using outbreak detection methods. One of the first outbreak detection methods on bacterial infections was developed in London in September 1854. The British scientist John Snow marked the homes of the 500 people that died in ten days from a cholera outbreak. Snow discovered thus the origin of the cholera outbreak. It originated from a water pump [5].
Nowadays water pumps in the western world are not a dangerous source of bacteria anymore. People do not think anymore their illness originates from the air. Bacterial infections can be hazardous to health and the increase in antibiotic resistance is a danger for public health [1 -3]. The antibiotic resistance struggle is a problem all healthcare institutions face [6, 7].
In healthcare institutions hospitalized patients are generally more vulnerable to bacterial infections than healthy people. Due to higher vulnerability of hospitalized patients, more antibiotics are used in healthcare institutions for treatment and prophylaxis. With more use of antibiotics, a healthcare institution has a high risk of increasing resistance in antibiotics [8].
To prevent transference of antibiotic resistant microorganisms, quick localisation and
identification of a microorganism has high clinical relevance. If a microorganism is resistant to one or multiple first choice clinical antibiotic treatment, it is called a multidrug resistant organism (MDRO) [9]. Hospitalized patients with an MDRO infection have a higher mortality risk [10]. The localisation is of an MDRO infection is difficult. A MDRO detected more than 48 hours after patient hospitalisation is not necessarily acquired in the hospital [11]. Searching for the origin of a bacterial infection is difficult, which makes preventing transference important. Detecting a transference of a MDRO can help prevent people from getting infected and achieve quicker adherence to infection control and isolation protocols within hospitals to prevent an MDRO outbreak.
To detect MDRO outbreaks different researchers explored how to define an outbreak or how to manage different surveillance scopes with different algorithms and other statistical challenges [12 - 17]. A golden rule for outbreak detection analysis does not exist.
8 A systematic review [17] of 29 studies showed difference in surveillance scope and
algorithms per outbreak detection method. The difference in scope and algorithms makes comparing of the studies difficult. A research in 2003 [18] tried to increase speed in the recognition of
outbreaks, by analysing an automated national laboratory data based system in the Netherlands. The national laboratory data analysis produced positive result. The global outbreaks were faster
recognized, but still had its limitations. A conclusion was that local outbreaks will still be faster detected by conventional lab methods [18]. Another study in the Netherlands tried to achieve faster and more accurate detection with an automated detection tool on three clusters of infectious diseases [20]. Eight outbreak alerts in two years were generated, but only two out of eight alerts were true positives. The other six alerts were false positives. Considering outbreak detection methods, there is a broad spectrum of outbreak detection studies that try to handle bacterial threats, however local analysis is still difficult to improve.
1.1. Research Aim
The aim of the present investigation is an analysis of outbreaks in a large academic hospital. The aim is specifically to improve outbreak detection. The study will aim to compare currents infection control outcomes to the outcomes of a new method. In example this research is named a data-driven data collection comparison research. The data-driven part consists of an analysis with Statistical process control (SPC), where MDRO incidence per week per MDRO will be analysed. SPC will help detect outbreaks from the data. SPC is used as a retrospective tool to analyse the outbreaks per MDRO occurrence in the large academic hospital. The data collection part consists of annual reports from the infection control of the hospital. The annual reports will function as validation for the SPC analysis. The outbreaks found with SPC will be compared to the outbreaks reported in the annual reports.
In addition to the SPC analysis a geographical dashboard will be developed with the map chart functionality of TIBCO Spotfire [19]. In order to get more insight in the spread of the outbreak among hospitalized patients, the map chart provides an overview which patient with which MDRO lies in which bed.
1.2. Objectives
Present study analyses outbreaks of multidrug resistant organisms using SPC. The first objective is to provide more accurate outbreak detection and evaluate SPC as a tool for quality improvement in healthcare regarding prevention of outbreaks. The annual reports of the large academic hospital will function as validation for the SPC analysis. The second objective is to develop a geographical
9 geographical dashboard made with Spotfire [19] will present an overview of all bed departments in the hospital. With this dashboard a spread of microorganisms can be quicker analysed. Present research focusses on the first objective.
1.3. Research Questions Main question:
● Can statistical process control help identify multidrug resistant organism outbreaks? Sub questions:
● Is statistical process control an applicable method to analyse the available data? o Which charts from statistical process control are applicable to our case? ● Which multidrug resistant organisms are representative enough to analyse the data?
● Which multidrug resistant organism outbreaks are found in the annual reports from infection control?
o Do the outbreaks from the annual reports from the Infection control match the found outbreaks from statistical process control at each multidrug resistant organism?
● How are the performance measures (control limits] of the used statistical process control charts set?
1.4. Approach
Background information on antibiotic resistance and multidrug resistant organisms is collected, followed by the development of the graphical tool to be used as a dashboard. After the graphical tool is developed, a data driven data collection comparison study will be conducted. The SPC outbreaks will be compared with the outbreaks from the annual reports. The performance measure of the data driven data collection study will be the control limits calculated by different SPC charts. The found outbreaks will be compared to the annuals reports of the infection control department.
1.5. Expected results
SPC as an outbreak detection method can help find MDRO outbreaks in a large academic hospital and the geographical Spotfire component can provide faster analysis of bacterial spread in a hospital.
10 1.6. Chapter organization
The next chapter will consist of background information on antibiotics and antibiotic resistance. Afterwards, the data and software that is used, will be discussed. Next, the methods of the data driven approach with SPC as method will be discussed. The results will present the outcomes of the outbreaks found with SPC and the annual reports. Lastly, a discussion and a conclusion will mention strengths and weaknesses and a direction for future research with SPC and MDROs.
11 2. BACKGROUND: MEDICAL MICROBIOLOGY AND INFECTION CONTROL
Relevant background information will be provided on bacteria and antibiotics.
2.1. Bacteria and Fungi
Microorganisms is a broad concept of all small living things that have an organic mechanism that interact with each other. Microorganisms are too small to detect with the human eye. In this study the focus lies on the interaction between bacteria and fungi. Whereas the dangerous MDROs are bacteria and the antibiotics are the fungi that restrict the growth of bacteria. Every human carries about one and a half kilograms of bacteria with them for mainly good purposes. If the concentration of bacteria in a substance is too high we can speak of an infection, which can be cured by the fungi.
2.2. Antibiotics
The accidental discovery of antibiotics by Sir Alexander Fleming in 1928 has led to a major amelioration in public health. Alexander Fleming discovered that a fungus growing in a bacteria culture inhibits the growth of the bacteria [21]. The mold would later be known as penicillin, one of the first antibiotics. The discovery of Alexander Fleming changed medicine. Simple infections, like pneumonia, that first were difficult to cure were suddenly easily curable. However, the use of antibiotics nowadays is also a risk for public health. Everybody does not indicate pneumonia as a dangerous disease anymore, whereas a bacteria that infects a human or animal can have a resistance for multiple antibiotics. Additionally, the resistance of bacteria for all antibiotics is increasing [22 - 24].
2.3. Antibiotic resistance
2.3.1. Minimum inhibitory concentration
Antibiotic resistance is the rate in which a bacteria is resistant to an antimicrobial agent, which is mostly stated as a minimum inhibitory concentration (MIC) value [37]. MIC is a value that indicates the lowest value of antibiotic concentration that is needed to stop the bacteria from growing. The MIC is measured through agar or broth dilution methods. Both techniques use different kinds of concentrations of the antimicrobial agents to measure a MIC value of a bacteria. The MIC value also represent the clinical definition of antibiotic resistance [37]. The clinical definition focuses on the therapeutic applicability of antibiotics for patient treatment. In practice the clinical definition entails the dose of antibiotics to treat the patient.
12 2.3.2. Causal origin of antibiotic resistance
One of the main causes for the prevailing resistance is overuse [22]. One of the reasons for overuse is the widespread availability. High availability increases to usage of antibiotics and therefor increases the resistance [23]. The multi-resistant version of the microorganism called Staphylococcus Aureus (MRSA) spread worldwide in 2010 [25]. The increasing prevalence of MRSA is start of increasing antibiotic resistance next to other bacteria, such as VRE (Vancomycin-resistant Enterococcus) and drug resistant streptococcus pneumoniae.
2.4. MDRO
An overview is created of the different terminology used in the paradigm that entails antimicrobial resistant bacteria and the multidrug resistance [26]. Different terms and frameworks are mentioned concerning increasing antibiotic resistance and multidrug resistant organisms. To avoid confusion we will only use MDRO here to address antimicrobial resistance regarding multi drug resistant
organisms.
The causal origin of a MDRO differs. Some MDROs are multi-resistant by nature and other can become multi-resistant by overuse of antibiotics [27]. Both type MDROs are increasingly reported worldwide [28]. Additionally, antibiotic resistance from MDRO can originate locally or remotely. Local origin implies an antibiotic resistance increase that originates from an increase of antibiotic prescriptions in a regional hospital or other local origin. Remote antibiotic increase implies the antibiotic increase originates from another country [22, 23]. In the Netherlands there is a yearly report from the national institute for public health and environment, named Nethmap. Nethmap mainly gives an overview of the use of antimicrobials, and surveillance of resistance in medically important bacteria [30]. Nethmap mentions in the synopsis that the overall resistance of antimicrobials does not increases, but more outbreaks were reported. The increase of reported outbreaks can indicate an improvement of outbreak detection. In the next paragraphs the different kind of MDROs that appeared are shortly explained. In the large academic hospital of our research seven different MDROs are encountered. The seven different MDRO will be shortly explained, MR and MR+ are combined. See Appendix B for the hospital classification.
2.4.1. Extended Spectrum Beta-Lactamase
Extended Spectrum Beta-Lactamase (ESBL) is an enzyme that breaks down beta-lactam antibiotics, like penicillin’s and celafosporines [30, 41]. Microorganisms producing the ESBL enzyme are common bacteria, like Klebsiella Pneumoniae and Escherichia Coli. In the Netherlands, 5% of the population is carrier of a bacteria producing the ESBL enzyme [31]. The standard antibiotic treatment will become difficult, if a patient in a hospital gets infected with a bacteria producing ESBL.
13 2.4.2. MR/MR+
MR or MR+ is a collective label for bacteria that are multi-resistant (MR) or have extra resistance to other antibiotics (MR+). MR or MR+ refers to a collection of bacteria that is resistant to multiple clinical first choice antibiotics. Common examples of MR or MR+ are Pseudomonas Aeruginosa or Klebsiella Pneumoniae. VRE and MRSA have their own label, due to their higher prevalence.
2.4.3. Vancomycin resistant Enterococcus
Vancomycin resistant Enterococcus (VRE) is a bacteria that is resistant to antibiotics for the standard treatment for an infection. VRE spread worldwide in 2002 [44]. Within the species of VRE
(Enterococcus faecium) two subspecies are clinically relevant [44] species, namely E. faecium and E. faecalis. E. faecium is more pathogenic, but E. faecalis generates more antibiotic resistance.
Enterococci are a persistent inhabitant of hospitals and have developed multiple antibiotic resistant mechanisms, which makes treatment difficult [44].
2.4.4. Methicillin Resistant Staphylococcus Aureus
MRSA stands for Methicillin Resistant Staphylococcus Aureus. Also known as a hospital bacteria, because its prevalence is much higher in hospitals then in the regular community [42, 43]. Staphylococci are bacteria that have a high prevalence in healthy people. However some
staphylococcus have a strain that is resistant to standard antibiotic treatment. When staphylococci have a specific strain, they are named MRSA.
2.4.5. Carbapenemase producing organisms
Carbapenemase-producing organisms (CPO) refers to bacteria in the family of the
Enterobacteriaceae that are resistant to the carbapenem antibiotic [39]. These Enterococci are resistant to carbapenems. Carbapenem is an antibiotic, which is part of the lactams. Other beta-lactams include penicillin, cephalosporins and monobactams. Carbapenems is a broad spectrum antibiotic that is mentioned as one of the most reliable last-resort treatment for bacterial infections [41]. The resistance for carbapenems is very dangerous for hospitalized patients.
2.4.6. Multi-resistant Acinetobacter
MR Acinetobacter stands for the multi-resistant Acinetobacter bacteria. The Acinetobacter baumanii species within the Acinetobacter is the most clinically relevant [40]. It is difficult to determine what the natural environment of the Acinetobacter baumannii is. It can stick to a wall, but also appear more than often in hospitals in a multidrug resistant form. Furthermore, the baumannii species colonizes effectively in immunosuppressed patients [45]. The Acinetobacter baumannii is a major cause of hospital-acquired infection and increases the risk of death [44]. The prevalence of
14 Acinetobacter baumannii infections increases, as are the resistant strains [40]. Additionally the Acinetobacter baumannii is sometimes even pan-resistant, which means no antibiotic treatment can help.
2.5. Outbreak management
If a healthcare professional suspects an outbreak or spread of a microorganism, contact investigation will take place among employees and patients. Patients who were in the same room as the infected patient will be checked for the same microorganism. With contact investigation the same
microorganisms can be found, but it is only called an outbreak if the microorganism grown from the contact investigation patient has the same strain as the grown bacteria1. However, a different strain
will not definitely mean there was no transference of a bacteria, because a strain can mutate.
15 3. DATA, SOFTWARE AND CONVERSION
In this chapter we will describe the data and software used for the data driven data collection comparison research. In the next chapter (the results) a comparison will be made to see if SPC analysis with MDRO has added value to standard outbreak detection.
3.1. Research Data Platform
The Research Data Platform (RDP) is an aggregated database, built by the hospital business intelligence department. The RDP presents data from different hospital information systems (like GLIMS and EPIC) in Detailed Clinical Models (DCMs) for use by researchers. A DCM is a
standardization of data in healthcare, which usage is increased in modern healthcare [47]. An example of the RDP catalogue and some DCMs is shown in Figure 1. The data in our research was obtained from the RDP with Spotfire software. In the RDP catalogus on the left the DCM is shown and on the right the hospital information system and the time since when the data was obtained.
16 3.2. RDP data
From the RDP a specific DCM is selected to extract to Spotfire. The used DCM contains
microorganism isolates from the MMI laboratory. Example values from the DCM are seen in Table 1. A flow diagram of the complete process is shown in Figure 2.
Table 1. Columns, description and example value of the DCM on lab isolates.
Kolom Omschrijving Voorbeeld
Monsternummer Unieke identifier van het monster 1.45e+11 isolaatnummer Unieke identifier van isolatt van monster 3
patientnummer Patientnummer 1111111
MicroOrganismeCode Code micro-organisme anaeme
AfnameDatum Afname datum monster 10/19/2016
ArtsCode Arts code 97063065
AfdelingCodeAanvrager Code afdeling aanvrager PVVO
AfdelingNaamAanvragen Naam afdeling aanvrager Polikliniek voorplantingsgeneeskunde AfdelingKliniekPoliAanvragen
Is afdeling aanvrager
poli/kliniek/extern/laboratorium poli/kliniek/extern/laboratorium OrganisatieCodeAanvrager Code organisate aanvrager ESZA
OrganisatieNaamAanvragen Naam organisatie aanvragen Ziekenhuis Amstelland
StudieNummer Uniek intern studie id 1111
MicroOrganismeOuder Micro-organisme ouder StrepBHS
MicroOrganismeOuderOuder Micro-organisme ouder-ouder Streptococcus MicroBiologieProcedureCode Code microbiologische procedure M_banaal_UR
MicroOrganismeName Naam micro-organisme Streptococcus pyogenes (Groep A)
MicroOrganismeType Type micro-organisme Germ
MicroOrganismeParentCode Code micro-organisme ouder Streptococcus MateriaalCode Unieke code gebruikt materiaal M_BL_SG_P MateriaalDescription Beschrijving gebruikt materiaal Serum MateriaalShortName Beschrijving gebruikt materiaal kort Serum ExternCommentaar Extern commentaar
morfologische verschillend van andere E.coli
17 Figure 2. Flow diagram of the data.
From the DCM five variables were used. The first is a code for microorganism, the second is the date of the sample which from the patient is collected, the third is the parent of the microorganism species, the fourth is the actual name of the microorganism, and the last used variable is extra commentary on the observed microorganism isolate. The extra comment is important to determine if the microorganism is a MDRO. With the extra comment a translation table is made by healthcare professionals from the MMI department. An example of the translation table is shown in Table 2.
Research Data Platform TIBCO Spotfire Visualisation Software Import in Rstudio for analysis Load via SQL server Export data to txt/csv file ‘Extern Commentaar’ Translation table Local data drive Tool development Spotfire Geographical Dashboard
18 Table 2. Examples of the translation table in Microsoft Excel. The example rows represent the extra comment in the first row and the MDRO in the second column.
Extern Commentaar MDRO
multi-resistent, Carbapenemase positief; voor gevoeligheid zie andere kweek van 6-8-2015 Carbapenemase multi-resistent. {<Carba_feno} {<Carba_pos} Carbapenemase Multi-resistent, Carbapenemase PCR; positief, ESBL PCR; negatief. Carbapenemase
multi-resistent, Carbapenemase positief Carbapenemase
multi-resistent, Carbapenemase positief Carbapenemase
multi-resistent, Carbapenemase positief. Uitslag Carbapenemase PCR positief Carbapenemase multi-resistent, Carbapenemase positief. Uitslag Carbapenemase PCR: Positief Carbapenemase multi-resistent Uitslag Carbapenemase PCR positief. Carbapenemase voor gevoeligheid zie keelkweek van 16-11-2013 {<Carba_pos} {<ESBL_neg} Carbapenemase multi-resistent en ESBL en Cabapenemase positief Carbapenemase
{<multi_res_ESBL_pos} ESBL
multi-resistent en ESBL positief ESBL
multi-resistent en ESBL positief. Carbapenemase PCR negatief ESBL
{<aantal_kolonies}1{<multi_res_ESBL_pos} ESBL
{<aantal_kolonies}1;{<multi_res_ESBL_pos} ESBL
{<aantal_kolonies}2{<multi_res_ESBL_pos} ESBL
{ESBL_pos} ESBL
{ESBL_pos} morfologisch verschillend van andere Enterobacter ESBL {ESBL_pos} voor gevoeligheid zie andere anuskweek van 05-06-2013 ESBL {ESBL_pos} voor gevoeligheid zie drainvocht van 16-06-2013 ESBL
3.3. TIBCO Spotfire: Data Visualization & Analytics Software
Spotfire is software that can create easy understandable visualizations. Additionally, Spotfire has a broad variety of data analysis tools [48]. Spotfire enables MMI department to create a quick and simple overview of all the microorganisms obtained by the lab in the hospital. For the present research Spotfire uses data from the RDP by connecting via a SQL server. In Spotfire the data was also loaded into the geographical dashboard of the hospital beds. For a screenshot of the dashboard see Appendix A. The data in Spotfire has a delay of 36 hours when a healthcare professional fills in electronic data to when the data is loaded into Spotfire.
3.4. Data Conversion in Spotfire
After the required data is loaded from the RDP in Spotfire, the data needs some adjustment to make it ready to use with SPC in R. The ‘AfnameDatum’ variable is transferred from date per day to week number per year. Furthermore, the incidence for each separate MDRO per week is calculated with Spotfire. A conversion of the data in Spotfire can be seen in appendix C. After these calculations are done the data can be exported from Spotfire into a CSV file, which can be imported into R. To overview the data flow see Figure 2.
19 3.5. Data in R
Further processing of the data was conducted in R. R is a free software environment for statistical programming and graphics [32]. For further pre-processing, analysing and reporting of the data, RStudio version 1.1.383 and R version 3.4.2 are used. The preprocessing packages qicharts2 [33], qcc [34], dplyr [35] and data.table [36] were used.
3.6. Annual reports
After all the data is processed, outbreaks are obtained from the annual reports to validate the outbreaks from the SPC analysis. The Medical Microbiology and Infection control (MMI) department presents an annual report every year. These annual reports contain the events that happened at the department, mainly containing surveillance, infection control policy and outbreak management. The information on MDRO outbreaks was extracted from the outbreak management section. All sections of the reports are read to be sure no information outside the outbreak management section was missed.
20 4. BACKGROUND: STATISTICAL PROCES CONTROL
4.1. Outbreak detection: Method
To improve clinical care and patient treatment quality, automated outbreak detection systems are used and new ones are being developed in hospitals. The literature provides what surveillance scope and what kind of method best is used with present research. The publication rate of SPC in health care research keeps growing since 1990 [50]. A systematic review [17] on automated outbreak detection methods see that SPC is used in most studies. It is hard to pinpoint the best method to analyse MDRO outbreaks. Nonetheless, SPC has an advantage over other methods, because it has a wide variability and applicability on different problems [49].
4.2. Statistical Process control
Statistical Process control (SPC) is a statistical tool that was developed by Dr. Walter Shewart around 1920s [51]. SPC consists of four basic principles [51]:
- Individual measurement from any process will exhibit variation.
- If the data comes from a stable common cause process, its variability is predictable within a knowable range that can be computed from a statistical model such as the Gaussian, binomial or Poisson distribution.
- If processes produce data with special causes, measured values will deviate in some observable way from these random distribution models.
- Assuming the data are in control, we can establish statistical limits and test for data that deviate from predictions, providing statistical evidence of a change.
The same research provided a good explanation of what data control is: Processes that exhibit only common cause variation are said to be stable, predictable, and in “statistical control”, hence the major tool of SPC is called the “statistical control chart” [51].
It is argued [52] that in the health sector worldwide there is a lack of use in SPC, whilst it is a useful tool for quality improvement. Furthermore, the strength of SPC research can have on quality improvement of healthcare is not emphasized enough [52].
There are two advantages of using SPC in healthcare. The first advantage consists of the easy understanding and interpretation of SPC charts. Everyone can interpret SPC charts. Only the
knowledge of what a mean and standard deviation is can suffice. Second, the charts can be updated with information that prompt decision making. For example with controlling infectious outbreaks in hospitals [52].
21 4.3. Statistical Process Control with MDROs
In present research the incidence of a MDRO per week is seen as the process. The incidence per MDRO differs a lot. For example, VRE is very rare, while ESBL is pretty common. There is not a straight cut-off point. Leavenworth and Grant [53] state that: "No general rules may be laid down for frequency of subgroups. Each case must be decided on its own merits, considering both the cost of taking and analysing measurements, and the benefits to be derived from the action based on the control charts." (p. 42). Therefore the frequency of the subgroups from each case can be deducted from analysing the outcomes.
To analyse the incidence per week a couple of SPC methods can be applied. Because each MDRO has a different definition different SPC charts are applied. This means that each MDRO can have a different distribution. On MDRO can have a high incidence with values around the mean, but if the max incidence is only three patients per week we have to look for an analysis which detects smaller shifts (CUSUM). For an analysis of incidence per week three charts are common to use. Control charts are used for exploration of the data, CUSUM is used for detecting smaller shifts in the data, whilst EWMA can put extra weight in more recent observations. Additionally if the process has to control small variations and the sample has an individual unit (which is our case) a EWMA or CUSUM is advised to use [50].
4.4. Charts
For the analysis of MDROs in a large academic hospital three charts of SPC are used. The control chart the CUSUM chart and the EWMA chart will now be discussed.
4.5. Control chart
In a control chart the incidents per MDRO are plotted per week. In a control chart the middle line represent the mean of the MDRO and the upper and lower line represent the control limit of three standard deviations (SDs) above or under the mean. If the number of MDROs exceeds the upper or lower control limit this is defined as an outbreak, because the observed values are beyond limits. In control charts also a trend can be visible. For example when a measure is above the mean for seven times in a row, this can be attributed to special variation. In this research we call this a violating run. In Figure 3 an example of a control chart is shown. In this chart the number of injuries per month are shown. The red dashed line is the upper control limit and the green line is the mean. In Figure 3 there are no violating runs or measures out of the control limits.
22 Figure 3. Example control chart, number of injuries per month.
4.6. CUSUM chart
A cumulative sum (CUSUM) chart in SPC is used to detect smaller shifts from the mean of the data. With the qcc package in Rstudio a standardized chart can be made to detect the outbreaks per MDRO. With MDRO data sometimes it is more common to focus on smaller shifts in data then on larger shifts (with 3 SDs). CUSUM charts use the mean of the sample per group (in our case 1 value per sample per week) to detect outbreaks from 2 SDs above or under the mean. In Figure 4 an example of a CUSUM chart is shown. In the example the mean per sample is the blue line and the control limits are the red lines.
Figure 4. Example of a CUSUM chart.
4.7. EWMA chart
An Estimated Weighted Moving Average (EWMA) chart is a chart that averages the data that more recent data is given more weight. EWMA charts are like CUSUM charts especially useful in detecting small shifts in the mean or variance of a process.
EWMA charts are made available in R to the qcc package, the same package as the other used charts. A Lambda (smoothing parameter) is set between 0 and 1 can be set to set weights to the
23 data. A lambda of ‘1’ suggest that only the most recent data is most valid, whereas a lambda of ‘0’ suggests there is no different weight of data across time. In present research a lambda of 0.8 is chosen. For expected change over time 0.8 is an often chosen value, as is 0.2 when little change is expected. The lambda values of 0.2 and 0.8 can be seen as a heuristic, because it is difficult beforehand to calculate a precise number. Antibiotic resistance increases over time and the large academic hospital keeps improving their detection methods, which indicates recent data is more valid than older data. In EWMA charts rule of thumb of 0.2 and 0.8 are commonly used. In Figure 5 an example of a EWMA chart is shown.
24 5. RESULTS
5.1. SPC analysis
In table 3 the outbreaks that resulted from the SPC analysis are reported. Per type of chart the number of weeks analyzed are reported. The data starts at 31 October 2011 (week 1 of the data gathered from the RDP) and ends at April 16th 2018 (week 338). Table 3 also reports the mean
incidence of unique patients per week and the control limits calculated by SPC. The number of weeks with points outside limits represent the number of weeks where the incidence of particular MDRO exceeded the control limits that were set by SPC.
5.2. Annual Reports
With exception of the of year report of 2011 all reports contained a specific outbreak management part. The annual report of 2011 therefore is difficult to interpret, because no specific outbreaks were mentioned. However the annual report of 2011 contained information on raised ESBL incidence, which we interpreted as a signal for an ESBL outbreak. The reports of 2012 and 2013 were combined, but it was not mentioned why these two years were combined. The other reports all had the same layout.
In table 4 the outbreaks of the MDRO presented by the year reports are gathered. Table 4 also includes additional information (if reported), namely, location (if a specific department), number of contact investigations, how many people with infected by the spread, and a comment to provide extra info regarding that specific reported outbreak. The year reports found an outbreak of ESBL in 2011, an outbreak of MRSA in 2012, 2013, 2015 and 2016, an outbreak of VRE in 2015 and 2017, an outbreak of a multi-resistant Pseudomonas Aeruginosa in 2012,2013,2014,2015, and an outbreak of a muti resistant Klebsiella in 2012 and 2013.
Table 3. Overview of the outbreaks found with control charts, CUSUM charts and EWMA charts. MDRO Type of chart Number of weeks Mean (sd) Control Limits Number of weeks outside
limits*
Number of violating runs
Mean sample size per
week Year of outbreak
ESBL Control chart 338 8.16 (2.86) 0 ; 16.73 2 32 1 2011
CUSUM 338 8.16 (2.91) -5.0 ; 5.0 39** Na 1 2011,2012,2013,2014
EWMA 338 8.16 (2.54) 1.93 ; 14.39 3 Na 1 2011,2013
VRE Control chart 338 0.18 (0.42) 0 ; 1.45 15 223 1 2013
CUSUM 338 0.18 (0.62) -5.0 ; 5.0 76 Na 1 2013,2015,2017.2018****
EWMA 338 0.18 (0.19) -0.28 ; 0.64 39 Na 1 2013,2014,2015,2016,2017,2018****
Carbapenemase Control chart 338 0.15 (0.39) 0 ; 1.33 6 156 1 2014,2015,2016,2017
CUSUM 338 0.15 (0.40) -5.0 ; 5.0 46 Na 1 2015,2016,2017
EWMA 338 0.15 (0.20) -0.36 ; 0.64 46 Na 1 2014,2015,2016,2017
MRSA Control chart 338 1.70 (1.30) 0 ; 5.61 3 19 1 2012
CUSUM 338 1.70 (1.38) -5.0 ; 5.0 36 Na 1 2012,2013 EWMA 338 1.70 (1.22) -1.29 ; 4.69 7 Na 1 2012,2013,2015,2016 MR Control chart 338 4.21 (2.05) 0 ; 10.37 3 29 1 2012,2014,2015 CUSUM 338 4.21 (2.27) -5.0 ; 5.0 36*** Na 1 2013,2014 EWMA 338 4.21 (1.94) -0.55 ; 8.97 5 Na 1 2012,2014,2015,2017 MR+ Control chart 129 0.31 (0.56) 0 ; 1.98 5 18 1 2016,2017 CUSUM 129 0.31 (0.57) -5.0 ; 5.0 10 Na 1 2016 EWMA 129 0.31 (0.33) -0.48 ; 1.11 6 Na 1 2016,2017 MR
Acinetobacter Control chart 338 0.086 (0.29) 0 ; 0.96 16 304 1 2013,2015,2016,2018**** CUSUM 338 0.086 (0.46) -5.0 ; 5.0 9 Na 1 2018 EWMA 338 0.086 (0.039) -0.01 ; 0.18 17 Na 1 2013, 2015, 2016,2018****
* Date of the week of the outbreaks is found in appendix D
* includes 9 points below lower control limit (represents a low incidence, therefor no outbreak) *** includes 10 points below lower control limit (represents a low incidence, therefor no outbreak) **** no year report available yet
1 Table 4. Overview of the reported outbreak management of the Infection control 2011-2017.
Annual reports Outbreak MDRO* Location
extra time info # Contact investigations conducted Found MDRO spread Comment
2011 ESBL Whole hospital none Na (not available) Na
The report only mentions increase, but not when were an outbreak was found
MRSA INEO**
March
2013 Na Na
2012 & 2013 MR Klebsiella VHEM*** none Na Na different patients appeared to have links with each other MR pseudomonas VHEM*** none Na Na different patients appeared to have links with each other 2014
MR Pseudomonas
aeruginosa VHEM*** none Na Na relation of outbreak between patients
MR Pseudomonas
aeruginosa IC/MICV**** none Na Na
MRSA IC none 1 1 employee part of IC was closed for short time
2015 VRE Dialysis***** none 1 3 patients Unsuspected finding but unrelated VRE strain.
MR Pseudomonas
aerigunosa (M)IC none Na 16 patients
same AFLP type, maybe originated from IC surrounding, research from 2012
2016 MRSA Whole hospital none 24
3 employees, 2 patients
2017****** MRSA Whole hospital none 30
4 employees 2 patients
In 4 of the 30 CIs a spread of MRSA was found. Three of the four were proven transmission, one was accidence.
VRE Whole hospital none 4
10 employees,
2 patients In 2 of the 4 CIs a spread of VRE was found * More outbreaks were mentioned in the reports but we only took the MDROs into account.
** INEO is the department of Neonatology & intensive care for children *** VHEM is the department for hematology nuclear medicine & PET. **** IC/MICV is the Intensive care/Medium care for adults
***** No data from the dialysis unit was available at the RDP ****** This report was not final yet, but a first version was used.
5.3. Comparison of annual reports and SPC analysis
Table 5 presents a summary of findings to compare the results of the SPC analysis with the reported outbreaks from the annual reports. However, we removed the found outbreaks of VRE,
Carbapenemase, MR+ and MR Acinetobacter in table 4. These four MDROs had not enough incidence to render reliable data in the SPC analysis.
Table 5. Comparison found MDROs in SPC analysis and annual report
Year Annual report
Control
chart CUSUM chart EWMA chart
2011 ESBL ESBL ESBL ESBL
2012 MRSA, MR MRSA, MR ESBL*, MRSA MRSA, MR
2013 MRSA, MR -- ESBL*, MRSA, MR ESBL*, MRSA
2014 MR MR ESBL*, MR MR
2015 MRSA, (VRE), MR MR -- MRSA, MR
2016 MRSA -- -- MRSA
2017 MRSA**, (VRE) -- -- MR*
* Extra outbreak found with SPC analysis **Outbreak not found in SPC analysis
Table 4 shows that all found MDRO outbreaks in year reports, were also found in at least one of the used charts, with the exception of MRSA in 2017. Also table 4 shows that the CUSUM chart found an extra outbreak of ESBL in 2012, 2013, and 2014. Furthermore the EWMA chart found an extra outbreak of ESBL in 2013 and an extra outbreak of MR in 2017. The outbreaks found by SPC analysis implicate 30% missed outbreaks, mainly ESBL. For all generated SPC charts, see appendix E.
1 6. DISCUSSION
Principal findings statement
The aim of this study was to find undetected outbreaks in a large academic hospital. In order to improve the finding of outbreaks in the annual reports from the infection control were compared to the outbreaks found with SPC in the RDP data. The main question was if SPC can detect extra MDRO outbreaks at a large academic hospital. The data driven data collection comparison study found 30% more outbreaks in SPC than in the annual reports. The 30% more found outbreaks indicate that regular outbreak detection protocols differ from the findings of SPC. Shortly stated, SPC can help identify MDRO outbreaks. Almost all outbreaks (except from one) that were reported in the annual reports were also found with SPC analysis.
Strengths and weaknesses of study
As seen in the systematic review of outbreak detection methods [17], SPC seems to be a helpful tool to analyze MDRO outbreak detection methods. SPC is easy to understand, easy to interpret and applicable to different healthcare data. SPC is an easy method that should be leading in order to tackle a quality improvement problem [52]. As Pimentel and Barrueto [52] said: “control charts make large data sets intuitively coherent by integrating statistical and visual descriptions”. However, SPC does not explain causal relations, but helps understand the data.
Another strength of present research is the large data set used. The seven years of used data strengthens the MDRO incidence as a healthcare process. In the discussed literature [17] MDRO incidence is not used as a process. The large dataset shows the usability of MDRO as a stable process. The coherent results over a long time period indicate the expansion of SPC to analyze more
healthcare processes. However, not all MDROs qualify as a stable process due to a lack of incidence. Despite a large dataset used, analyzing MDROs has some problems. The first problem is not all MDROs have enough data to analyze. If a MDRO does not have enough data it cannot be handled as a stable process and it is not available for SPC analysis [53]. Four out of seven MDROs did not have enough incidence to be a stable process. Another disadvantage of MDROs is that the specific MDRO Multi-resistant (MR) does not make a difference between species. With the MDRO MR there is no difference between different species, whilst the other MDROs refer to specific species. The MR MDRO can for instance be Klesbiella Pneumoniae or a Pseudomonas Aeruginosa. This makes MR less reliable in SPC analysis.
One more interesting result is the MRSA outbreak in 2017 that was not found by SPC. A reason for this finding could be that MRSA had a low incidence, but the infection control already had
2 a suspicion on an outbreak, due to nonconformities to hand cleaning protocols. Another reason could be that the spread was mainly on employees, which were not a part of our dataset.
SPC cannot replace regular detection methods. The first reason is that the MRSA outbreak in 2017 was not found in the SPC analysis. It seems best to combine methods, because the regular methods does not exclude the SPC analysis. Therefor it seems better to combine methods. As explained in the background, an outbreak is formulated by a finding of the same strain of the same bacteria. SPC analysis focusses on higher incidence. However, if one patient is infected and transfer it to only one other, this will not come out in a SPC analysis. Therefore SPC is best used at big data analysis and not if only one transference took place.
Present research shows that SPC is a useful tool and that preventing MDRO outbreaks is a useful addition to SPC research. Present research also shows that MDRO is a stable process and the results show missed outbreak from infection control. Our results show a lower false alarm rate than research of Groeneveld et al. [20]. Present research also showed that local outbreaks are ready to improve by combining big data analysis (for example with SPC) with regular methods.
3 7. CONCLUSION
The results of this study provide an extra tool to healthcare professionals to help detect outbreaks in a hospital. In addition to the regular methods to detect an outbreak, a big data analysis with (in our case) SPC shows that some outbreaks can be missed and can help less spread of harmful bacteria. Adding employee data in future research will complicate the analysis, but validate the outcome more.
Altogether present research learns us more about how local outbreak detection can be improved. The clinical relevance can be improved. The unanswered question remains how to include the employees in future research. Nonetheless a data warehouse for a hospital is a very useful tool to improve clinical care and broadens the spectrum of information research in healthcare.
4 8. REFERENCES
1. World Health Organization. Antimicrobial resistance: global report on surveillance. 2014; Available at: http://www.who.int/drugresistance/documents/surveillancereport/en/. Accessed 8 November 2017.
2. World Health Organization. Global action plan on antimicrobial resistance. 2015; Available at: http://apps.who.int/iris/handle/10665/193736. Accessed 08 November 2017.
3. World Health Organization. Guidelines on core components of infection prevention and control programmes at the national and acute health care facility level. 2016. Available
at: http://apps.who.int/iris/handle/10665/251730. Accessed 08 november 2017.
4. Czaplewski, L., Bax, R., Clokie, M., Dawson, M., Fairhead, H., Fischetti, V. A., ... & Henderson, I. R. (2016). Alternatives to antibiotics—a pipeline portfolio review. The Lancet Infectious Diseases, 16(2), 239-251.
5. Shiode, N., Shiode, S., Rod-Thatcher, E., Rana, S., & Vinten-Johansen, P. (2015). The mortality rates and the space-time patterns of John Snow’s cholera epidemic map. International journal of health geographics, 14(1), 21.
6. Ventola, C. L. (2015). The antibiotic resistance crisis: part 1: causes and threats. Pharmacy and Therapeutics, 40(4), 277.
7. Ventola, C. L. (2015). The antibiotic resistance crisis: part 2: management strategies and new agents. Pharmacy and Therapeutics, 40(5), 344.
8. Wang, A., Daneman, N., Tan, C., Brownstein, J. S., & MacFadden, D. R. (2017). Evaluating the relationship between hospital antibiotic use and antibiotic resistance in common nosocomial pathogens. Infection Control & Hospital Epidemiology, 38(12), 1457-1463.
9. Khan, A., Miller, W. R., & Arias, C. A. (2018). Mechanisms of antimicrobial resistance among hospital-associated pathogens. Expert review of anti-infective therapy, 16(4), 269-287. 10. Barrasa-Villar, J. I., Aibar-Remón, C., Prieto-Andrés, P., Mareca-Doñate, R., & Moliner-Lahoz,
J. (2017). Impact on Morbidity, Mortality, and Length of Stay of Hospital-Acquired Infections by Resistant Microorganisms. Clinical Infectious Diseases, 65(4), 644-652.
11. Erb, S., Frei, R., Dangel, M., & Widmer, A. F. (2017). Multidrug-resistant organisms detected more than 48 hours after hospital admission are not necessarily hospital-acquired. infection control & hospital epidemiology, 38(1), 18-23.
12. Sonesson, C., & Bock, D. (2003). A review and discussion of prospective statistical surveillance in public health. Journal of the Royal Statistical Society: Series A (Statistics in Society), 166(1), 5-21.
5 13. Farrington, P., & Andrews, N. (2004). Application to Infectious. Monitoring the Health of
Populations: Statistical Principles and Methods for Public Health Surveillance, 203. 14. Buckeridge, D. L., Burkom, H., Campbell, M., Hogan, W. R., & Moore, A. W. (2005).
Algorithms for rapid outbreak detection: a research synthesis. Journal of biomedical informatics, 38(2), 99-113.
15. Shmueli, G., & Burkom, H. (2010). Statistical challenges facing early outbreak detection in biosurveillance. Technometrics, 52(1), 39-51.
16. Unkel, S., Farrington, C., Garthwaite, P. H., Robertson, C., & Andrews, N. (2012). Statistical methods for the prospective detection of infectious disease outbreaks: a review. Journal of the Royal Statistical Society: Series A (Statistics in Society), 175(1), 49-82.
17. Leclère, B., Buckeridge, D. L., Boëlle, P. Y., Astagneau, P., & Lepelletier, D. (2017). Automated detection of hospital outbreaks: A systematic review of methods. PloS one, 12(4), e0176438. 18. Widdowson, M. A., Bosman, A., van Straten, E., Tinga, M., Chaves, S., van Eerden, L., & van
Pelt, W. (2003). Automated, laboratory-based system using the Internet for disease outbreak detection, the Netherlands. Emerging Infectious Diseases, 9(9), 1046.
19. Phillips, M. (2015). TIBCO Spotfire–A Comprehensive Primer. Packt Publishing Ltd.
20. Groeneveld, G. H., Dalhuijsen, A., Kara-Zaïtri, C., Hamilton, B., de Waal, M. W., van Dissel, J. T., & van Steenbergen, J. E. (2017). ICARES: a real-time automated detection tool for clusters of infectious diseases in the Netherlands. BMC infectious diseases, 17(1), 201.
21. Rogers, J. P. (2017). A History of Penicillin: The Miracle of Medicine.
22. Ventola, C. L. (2015). The antibiotic resistance crisis: part 1: causes and threats. Pharmacy and Therapeutics, 40(4), 277.
23. Ventola, C. L. (2015). The antibiotic resistance crisis: part 2: management strategies and new agents. Pharmacy and Therapeutics, 40(5), 344.
24. Gaynes, R. (2017). The Discovery of Penicillin—New Insights After More Than 75 Years of Clinical Use. Emerging Infectious Diseases, 23(5), 849.
25. Spellberg, B., & Gilbert, D. N. (2014). The future of antibiotics and resistance: a tribute to a career of leadership by John Bartlett. Clinical infectious diseases, 59(suppl_2), S71-S75. 26. Magiorakos, A. P., Srinivasan, A., Carey, R. B., Carmeli, Y., Falagas, M. E., Giske, C. G., ... &
Paterson, D. L. (2012). Multidrug‐resistant, extensively drug‐resistant and pandrug‐resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clinical microbiology and infection, 18(3), 268-281.
27. Sautter, R. L., & Halstead, D. C. (2018). Need of the Hour: Addressing the Challenges of Multi-Drug-Resistant Health Care-Associated Infections and the Role of the Laboratory in Lowering Infection Rates. Clinical Microbiology Newsletter, 40(2), 11-16.
6 28. Peleg, A. Y., & Hooper, D. C. (2010). Hospital-acquired infections due to gram-negative
bacteria. New England Journal of Medicine, 362(19), 1804-1813.
29. Martínez, J. L., Coque, T. M., & Baquero, F. (2015). What is a resistance gene? Ranking risk in resistomes. Nature reviews microbiology, 13(2), 116.
30. De Greeff, S. C., & Mouton, J. W. (2017). NethMap 2017: consumption of antimicrobial agents and antimicrobial resistance among medically important bacteria in the
Netherlands/MARAN 2017: monitoring of antimicrobial resistance and antibiotic usage in animals in the Netherlands in 2016.
31. Evers, E. G., & van Duijkeren, E. (2017). Vergelijkende humane blootstellingsschatting van ESBL-producerende Escherichia coli via consumptie van vlees.
32. Matloff, N. (2011). The art of R programming: A tour of statistical software design. No Starch Press.
33. Anhoej, J., Roeder, T., & Anhoej, M. J. (2017). Package ‘qicharts’.
34. Scrucca, L. (2004). qcc: an R package for quality control charting and statistical process control. dim(pistonrings), 1(200), 3.
35. Wickham, H. (2007). Reshaping data with the reshape package. Journal of Statistical Software, 21(12), 1-20.
36. Dowle, M., Short, T., Lianoglou, S., Saporta, R., Srinivasan, A., & Antonyan, E. (2014). data. table: Extension of data. frame.
37. Wiegand, I., Hilpert, K., & Hancock, R. E. (2008). Agar and broth dilution methods to determine the minimal inhibitory concentration (MIC) of antimicrobial substances. Nature protocols, 3(2), 163.
38. Miller, W. R., Murray, B. E., Rice, L. B., & Arias, C. A. (2016). Vancomycin-Resistant Enterococci. Infectious Disease Clinics, 30(2), 415-439.
39. Quale, J., Spelman, D., Hooper, D. C., & Bloom, A. (2016). Overview of carbapenemase producing gram-negative bacilli. UpToDate: Waltham, MA, USA, Accessed on, 21. 40. Subramaniyan, A., Nair, S., Joseph, N. M., & Kanungo, R. (2017). Profile of Multidrug
Resistant Acinetobacter Baumannii Infections among Hospitalized Patients. J Med Sci Clin Res, 5, 23111-5.
41. Meletis, G. (2016). Carbapenem resistance: overview of the problem and future perspectives. Therapeutic advances in infectious disease, 3(1), 15-21
42. Witte, W., Kresken, M., Braulke, C., & Cuny, C. (1997). Increasing incidence and widespread dissemination of methicillin‐resistant Staphylococcus aureus (MRSA) in hospitals in central Europe, with special reference to German hospitals. Clinical Microbiology and Infection, 3(4), 414-422.
7 43. Dulon, M., Haamann, F., Peters, C., Schablon, A., & Nienhaus, A. (2011). MRSA prevalence in
European healthcare settings: a review. BMC infectious diseases, 11(1), 138. 44. O’Driscoll, T., & Crank, C. W. (2015). Vancomycin-resistant enterococcal infections:
epidemiology, clinical manifestations, and optimal management. Infection and drug resistance, 8, 217.
45. Dijkshoorn, L., Nemec, A., & Seifert, H. (2007). An increasing threat in hospitals: multidrug-resistant Acinetobacter baumannii. Nature reviews microbiology, 5(12), 939.
46. Siegel, J. D., Rhinehart, E., Jackson, M., & Chiarello, L. (2007). Management of multidrug-resistant organisms in health care settings, 2006. American journal of infection
control, 35(10), S165-S193.
47. Goossen, W., Goossen-Baremans, A., & Van Der Zel, M. (2010). Detailed clinical models: a review. Healthcare informatics research, 16(4), 201-214.
48. Phillips, M. (2015). TIBCO Spotfire–A Comprehensive Primer. Packt Publishing Ltd.
49. Gomes IC, Mingoti SA, Di Lorenzo Oliveira C. A novel experience in the use of control charts for the detection of nosocomial infection outbreaks. Clinics. 2011;66: 1681–1689.
pmid:22012038
50. Karakus, Z. (2013). Performance comparison of statistical process control methods to monitor grouped data and rare events in healthcare applications (Doctoral dissertation, Northeastern University).
51. Benneyan, J. C., Lloyd, R. C., & Plsek, P. E. (2003). Statistical process control as a tool for research and healthcare improvement. BMJ Quality & Safety, 12(6), 458-464.
52. Pimentel, L., & Barrueto, F. (2015). Statistical process control: separating signal from noise in emergency department operations. Journal of Emergency Medicine, 48(5), 628-638.
53. Leavenworth, R. S., & Grant, E. L. (2000). Statistical quality control. Tata McGraw-Hill Education.
8 9. LIST OF ABBREVIATIONS
WHO – World Health Organisation MDRO – Multidrug Resistant Organism SPC – Statistical Process Control
MIC - Minimum Inhibitory Concentration MRSA – Methicillin Resistant Staphylococcus VRE – Vancomycin Resistant Enterococcus ESBL – Extended Spectrum Beta-Lactamase MR – Multi-resistant
MR+ - Multi-resistant Plus
CPO – Carbapenemase Producing Organisms RDP – Research Data Platform
DCM – Detailed Clinical Model
MMI – Medical Microbiology and Infection control CUSUM – Cumulative Sum
9 10. APPENDIX
A Geographical dashboard B MDRO classification C Spotfire Data conversion
D Date of outbreaks per MDRO (SPC analysis) E MDRO SPC charts
10 10.1. Appendix A: Geographical dashboard
Each of the nine pictures represent a floor. A blue marker indicates a room with a low occurence, a red marker represents a room with high occurence. In Spotfire filters for each patient data can be set manually. In our case the color of the marker indicates the microorganism of the investigated MDRO.
11 10.2. Appendix B: MDRO classification
Lijst van Resistente Micro Organismen en Bijzonder Resistente Micro Organismen (BRMO) die gemeld moeten worden aan de infectiepreventie.
Resistent tegen één van de genoemde antibiotica of combinatie van antibiotica (aangegeven met een “+”); zie onderstaande tabel
LET OP: intermediaire gevoeligheid (I) wordt beschouwd als resistent (R); zie opmerkingen pagina 2 opmerking: voor de te nemen isolatie maatregelen raadpleegt men het handboek van de infectiepreventie
Enterobacteriaceae ESBL positief carbapenems
aminoglycosiden + quinolonen
Pseudomonas aeruginosa carbapenem resistent o.b.v. een carbapenemase (bv. MBL) ESBL positief
combinatie van resistentie tegen 3 uit de volgende groepen antibiotica / middelen (indien carbapenem resistent, maar CIM negatief, vermeld dan MR+, in andere gevallen MR) - Carbapenems
- Aminoglycosiden - Quinolonen
- Piperacilline/tazobactam - Ceftazidime
Acinetobacter spp ESBL positief (alleen met PCR aantoonbaar)
carbapenem resistent o.b.v. een carbapenemase (bv. MBL)
carbapenems
aminoglycosiden + quinolonen Stenotrophomonas maltophilia cotrimoxazol
Staphylococcus aureus meticilline/oxacilline Streptococcus pneumoniae penicilline
vancomycine
Enterococcus faecium vancomycine + amoxicilline - intermediaire gevoeligheid (I) wordt beschouwd als resistent (R) - aminoglycosiden: gentamicine of tobramycine of amikacine
- quinolonen: ciprofloxacin of levofloxacin of ofloxacin of moxifloxacin - carbapenem: imipenem of meropenem of ertapenem
12 10.3. Appendix C: Spotfire data conversion
Table 1. Relevant Variables from Lab_MMI_Isolaten (source: GLIMS 2012-2018).
Lab_Isolaten Kolom Omschrijving Voorbeeld
Pseudo_id
Geanonimiseerde
patientnummer 111111
AfnameDatum Afname datum monster 10/19/2016 MicroOrganismeName Naam micro-organisme Streptococcus
ExternCommentaar Extern commentaar
Morfologisch verschillende van E.Coli
Table 2. Three examples of computed variables needed for analysis
Created Columns
Used column(s) to create
column Example or Translation Function
Created Example
BRMO*
ExternCommentaar &
translation table {<MR} {<Carba_pos} Carbapenemase
Year.Week Afnamedatum YearAndWeek([AfnameDatum]) 201701
Aantal.BRMO.per.week BRMO*
Count (DISTINCT [Pseudo_id]) Over
([Year-Week],[BRMO]) 12
13 10.4. Appendix D: Outbreak dates
MDRO
Type of
chart date of outbreak (number of beyond limits) ESBL Control chart 14/11/2011, 12/12/2011 CUSUM 12/12/2011 t/m 07-05-2012, 26/08/2013 t/m 21/10/2013, 03/11/2014 t/m 10/11/2014 EWMA 14/11/2011, 12/12/2011, 19/08/2013 VRE Control chart 20/05/2013 t/m 03/06/2013 CUSUM 20/05/2013 t/m 16/09/2013, 16/11/2015 t/m 23/11/2015, 27/02/2017 t/m 19/03/2018 EWMA 08/04/2013, 22/04/2013 t/m 03/06/2013, 26/06/2013 t/m 01/07/2013, 02/10/2014, 02/03/2015, 06/07/2015, 26/10/2015, 02/11/2015, 09/11/2015 t/m 16/11/2015, 05/30/2016, 11/14/2016 t/m 21/11/2016, 09/01/2017, 06/02/2017, 20/02/2017 t/m 13/03/2017, 27/03/2017 t/m 08/05/2017, 29/05/2017, 12/06/2017, 26/06/2017, 04/12/2017, 01/01/2018, 22/01/2018 Carbapenemase Control chart 08/09/2014, 20/04/2015, 31/08/2015, 26/10/2015, 30/05/2016, 13/11/2017 CUSUM 20/04/2015 t/m 27/04/2015, 24/08/2015 t/m 04/07/2016, 13/11/2017 EWMA 10/12/2012, 04/02/2013, 18/03/2013, 15/04/2013, 09/09/2013, 18/811/2013 t/m 25/11/2013, 06/01/2014 t/m 13/01/2013, 08/09/2014, 09/22/2014, 09/03/2015 t/m 16/03/2015, 06/04/2015, 20/04/2015, 07/06/2015, 20/07/2015, 10/08/2015 t/m 07/09/2017, 19/10/2015 t/m 26/10/2015, 30/11/2015, 21/12/2015 t/m 28/12/2015, 18/01/2016, 15/02/2016 tm 29/02/2016, 30/05/2016 t/m 06/06/2016, 22/08/2016, 14/11/2016, 30/01/2017, 13/02/2017 t/m 20/02/2017, 27/03/2017, 03/07/2017, 31/07/2017, 14/08/2017, 28/08/2017, 06/11/2017 t/m 13/11/2017, 22/01/2018 MRSA Control chart 30/01/2012, 19/03/2012, 09/07/2012 CUSUM 30/01/2012 t/m 16/07/2012, 01/10/2012 t/m 12/11/2012, 18/03/2017 t/m 01/04/2017, 15/04/2013, 06/05/2013 EWMA 16/01/2012, 30/01/2012, 19/03/2012, 17/09/2012, 18/03/2013, 28/12/2015, 24/10/2016 MR Control chart 28/05/2012, 21/07/2014, 12/10/2015 CUSUM 24/06/2013 t/m 06/01/2014 EWMA 28/05/2012, 08/10/2012, 21/07/2014, 12/10/2015, 07/08/2017 MR+ Control chart 18/04/2016 t/m 25/04/2016, 26/09/2016, 19/06/2017 t/m 26/06/2017, 23/10/2017 CUSUM 25/04/2016 t/m 09/05/2016, 30/05/2016 t/m 04/07/2016, 26/09/2016 EWMA 18/04/2016 t/m 25/04/2016, 26/09/2016, 19/06/2017 t/m 26/06/2017, 23/10/2017 MR Acinetobacter Control chart 01/04/2013, 17/08/2015 t/m 24/08/2015, 12/09/2016, 29/01/2018 t/m 16/04/2018 CUSUM 19/02/2018 t/m 16/04/2018, EWMA 01/04/2013, 17/08/2015 t/m 31/08/2015, 12/09/2016, 29/01/2018 t/m 16/04/2018
14 10.5. Appendix E: MDRO SPC charts
15 VRE
16 MRSA
17 Multi-resistant
18 Multi-resistant plus
19 Carbapenemase
20 MR acinetobacter
