|FULL TITLE||Chief Investigator||Sponsor||Research Location||Approval Date|
|Changes in the incidence and importance of endocarditis as a diagnosis code in Oxfordshire hospitals See lay summary below||Prof Sarah Walker||NIHR OxBRC||OUH NHS Trust||21-Jul-16|
|What is driving changes in the incidence of E. coli bloodstream infections in Oxfordshire? See lay summary below||Prof Sarah Walker||NIHR OxBRC||OUH NHS Trust||21-Jul-16|
|Antibiotic use- if you use less antibiotic in general medicine are there adverse consequences? See lay summary below
See also blog on Modernising Medical Microbiology
|Profs Sarah Walker & Tim Peto||NIHR OxBRC||OUH NHS Trust||19-Mar-15|
|Do birth characteristics influence susceptibility to childhood infections? See lay summary below||Prof Stephen Kennedy||NIHR OxBRC||OUH NHS Trust||25-Feb-15|
|What is driving increases in paediatric A&E attendances? See lay summary below||Prof Sarah Walker||NIHR OxBRC||Oxford University||25-Feb-15|
|Can we identify surgical site infections using routine electronic health record data? See lay summary below||Prof Sarah Walker||NIHR OxBRC||Oxford University||25-Feb-15|
|Trends in bacterial pathogens, antibiotic resistance, antibiotic usage and patient outcomes in critical care in the Oxford University NHS Trust hospitals, 1999-2014 See lay summary below||Dr Nicole Stoesser||NIHR OxBRC||Oxford University||12-Jan-15|
|Determining patterns of Gram Negative resistance and their genetic basis See lay summary below||Dr David Clifton||NIHR OxBRC||Oxford University||17-Sep-14|
|Antimicrobial susceptibility profiles of urinary isolates from samples collected in the community See lay summary below||Dr Kyle Knox||NIHR OxBRC||OUH NHS Trust||04-Jul-14|
|Outcomes after elective surgery: prognostic information in baseline blood tests See lay summary below||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||04-Jul-14|
|To what extent do urine cultures provide prognostic information for poor outcomes? See lay summary below||Phuong Quan||NIHR OxBRC||OUH NHS Trust||03-May-13|
|Incidence and outcomes following C. difficile infections in Oxfordshire, and predictors of poor outcomes See lay summary below||Dr Sarah Walker||NIHR OxBRC||OUH NHS Trust||14-Sep-12|
|C. difficile recurrence: incidence, predictors and risk scores See lay summary below||Dr Sarah Walker||NIHR OxBRC||OUH NHS Trust||14-Sep-12|
|Using secondary data sources
to investigate incidence and severity of respiratory infections
See also blog on Modernising Medical Microbiology
|Phuong Quan||NIHR OxBRC||OUH NHS Trust||07-Jun-12|
|A search for temporospatial clusters of blood stream infection in intensive care and high dependency units||Dr Angela Minassian||NIHR OxBRC||OUH NHS Trust||14-Nov-11|
|Understanding the causes of changes in severity biomarkers||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||21-Oct-11|
|Statistical Network Modelling of Ward Transfer Patterns and theirRelationship to Hospital Acquired Infections||John Finney||NIHR OxBRC||OUH NHS Trust, OU Dept of Statistics||22-Jul-11|
|Asymptomatic carriage candidate selection||John Finney||NIHR OxBRC||OUH NHS Trust||22-Jul-11|
|Mortality following invasive pneumococcal disease in Oxfordshire 1995-2010||Dr Kyle Knox||NIHR OxBRC||OUH NHS Trust||05-Nov-10|
|C. difficile transmission in Oxfordshire hospitals 2006-2009 using Markov Chain Monte Carlo estimation and information from patients not tested for C. difficile||Dr Madeleine Cule||NIHR OxBRC||OU Dept of Statistics||05-Nov-10|
|Panoramic view of the John Radcliffe Hospital infection networks||Mr John Finney||NIHR OxBRC||OUH NHS Trust||05-Nov-10|
|S. aureus transmission in ITU, Trauma, Geratology and Vascular specialities in the John Radcliffe Hospital, Oxford 2009-||Prof Derrick Crook||NIHR OxBRC||OUH NHS Trust/OU Dept of Statistics and WTCHG||05-Nov-10|
|Microbiology of Bronchiectasis and Cystic fibrosis||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||05-Nov-10|
|Analysis of 2009/10 Hospital Norovirus outbreak||Dr Nick Wong||NIHR OxBRC||OUH NHS Trust||05-Nov-10|
|C. difficile testing specificity in Oxfordshire 1997-2009||Dr A. Sarah Walker||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|C. difficile transmission in Oxfordshire hospitals 2006-2009||Prof Tim Peto||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Incidence of and competition between different pathogens isolated from blood and other sterile sites in Oxfordshire 1997-2009||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Competition between MRSA and MSSA at an individual level in Oxfordshire 1997-2009||Dr A. Sarah Walker||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|MRSA incidence 1998-2009||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Klebsiella in Oxfordshire 1998-2009||Dr A. Sarah Walker||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Detecting and adjusting for severe illness in medical patients using passively collected data||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Improving threat detection and quality surveillance: tools for infection management||Dr David Wyllie||NIHR OxBRC||OUH NHS Trust||4 Mar 2010, amended 7 Jun 2012|
|Infection and mortality in patients with bronchoalveolar lavage/pleural fluid samples||Dr Susan Williamson||NIHR OxBRC||OUH NHS Trust||04-Mar-10|
|Nosocomial MRSA endemicity: addition or replacement (multi-centre, multi-cohort study)||Prof Marc Bonten||Utrecht University||University Med Centre Utrecht||04-Mar-10|
NIHR: National Institutes of Health Research. OxBRC: Oxford Biomedical Research Centre. OUH: Oxford University Hospitals Trust.
Changes in the incidence and importance of endocarditis as a diagnosis code in Oxfordshire hospitals
In 2008 guidance recommended no longer using antibiotics to prevent people getting infections of the heart after having surgery at their dentists. This was because there was no strong evidence showing this was necessary, and to reduce overall use of antibiotics to avoid increasing antibiotic resistance in future. However, in 2015, a group looked at all admissions to NHS hospitals that had been recorded as being for heart infections. They found that these happened more frequently than would have been expected after the guidance was changed. The problem with these kind of analyses is that it is difficult to be confident about exactly what has caused a specific code to be recorded in a patient’s records. From other studies we know that there have been changes over time in how different diagnostic codes are used, reflecting changes in the way hospitals are managed and paid. We also know that extra codes are used much more frequently now than in the past. We plan to use the extra information about germs grown from patient samples and results of blood tests available in IORD to look in more detail at heart infections in Oxfordshire, and how these may have changed over time.
What is driving changes in the incidence of E. coli bloodstream infections in Oxfordshire?
Across England, the number of bloodstream infections caused by a bacteria called ‘Esherichia coli’ (E. coli) have been rising strongly for the last few years. However, the reasons behind these increases are not well understood at the moment. We plan to use the extra information in the IORD database to look at some different reasons why these increases might be occurring. We will look at whether the increases are happening more in hospital, outside of hospital or in patients who have been recently discharged. We will see whether the infections being reported are making patients less severely ill, to find out whether changes could just be down to how doctors are making decisions about who to test. We will also explore whether the aging population explains most of the increase, or whether it could be due to changes in the kinds of bacteria causing the bloodstream infections, particularly how resistant they are to different antibiotics. We will also particularly look at whether there is any evidence that infections of the bladder could be leading to more germs passing over into the blood.
Antibiotic use – if you use less antibiotic in general medicine are there adverse consequences?
Reducing antibiotic use to combat antimicrobial resistance is a priority. We want to know if we can do this safely and without putting patients at undue risk. Antibiotic use was measured over 1 week in an intensive audit of practice in the Acute/General Medicine service at the John Radcliffe. This showed that one Consultant team (Consultant A) used substantially less antibiotics compared to other teams. There was no difference in poor outcomes (death or admission to hospital) over this 1 week audit, but this is a short period of time. Every time a patient is admitted, a large amount of routine electronic data is recorded, including when they were admitted, what they were treated for, if the patient died, and which Consultant team they were treated by. We want to use this information over a longer 3 year period to see if more patients died or were admitted to hospital when they were managed under Consultant A compared to other consultant teams. If there is no difference between Consultant A and other teams, it suggests that their strategy of using substantially fewer antibiotics is safe, and can be used to reduce antibiotic use in General Medicine without risk to patients. See Manuscript on BMJ
Panoramic view of the John Radcliffe Hospital infection networks
Within a hospital, infectious disease can potentially pass from person-to-person through close contact. Without comprehensive data on actual contacts between different patients (contact ‘networks’), it is reasonable to assume that there is an increased chance of transmission of an infection between patients on the same ward, and that this chance would be even higher if patients are in beds next to each other. The probability of person-to-person transmission may also be influenced by multiple factors such as length of contact time, health of individual, age, comorbidities and other factors. The aim of this analysis is to look at the history of which wards and beds (where available) patients have been in within the OUH Trust in order to generate a longitudinal network model. These network models are most famous from applications to Facebook and other social networks, but can also be used with other types of data like this. Adding this time and space data to information about what infections patients had and their clinical records will let us investigate how infections move around hospitals. They also provide a basis for computer simulations that can help explore how different types of prevention measures might work.
Do birth characteristics influence susceptibility to childhood infections?
Newborns differ a great deal in terms of their size at birth. In particular, their weight, length and head circumference are determined by how well they have grown in the womb and their age at birth, i.e. premature babies are smaller than babies born at term. We seek to understand whether these differences at birth influence the chances of developing infections during childhood. The reason we are asking the question is that our immune system, which helps us fight infections, starts to develop as we are growing inside our mother’s womb. When we are born, we get exposed to different microbes, which train our immune system to fight the infections they cause. The vaccines that babies receive also do this. If a baby’s immune system is poorly developed at birth, for example because of premature delivery, it may take longer for the infant to acquire immunity or their immune system may simply never be as strong. As these infections are quite rare, we need access to large databases of health records, such as IORD, to answer these questions.
What is driving increases in paediatric A&E attendances?
The numbers of children coming to A&E have steadily increased over the last 5 years, and are continuing to increase. This puts a large burden on NHS staff. However, the reasons behind this increase are not entirely clear. Parents may be bringing children who are less sick because they struggle to get emergency appointments with their GP, or because they cannot access other out-of-hours services. Alternatively, more children who really are sick may being sent to A&E by services like the NHS-111 telephone line. To manage limited NHS resources better, we want to understand more about which groups of children are coming to A&E more and more, and particularly how infections may be contributing. We plan to investigate how numbers of A&E attendances are changing over time by various factors including (i) registered GP surgery (ii) time of day, day of week, (iii) reasons for attending, (iv) underlying illness (assess using blood test requests and results, and whether the child was admitted to hospital from A&E), (v) antibiotics prescribed. If we can identify key subgroups of children, we may be able to introduce ways to manage paediatric A&E better, including better information for parents or different systems in A&E.
Can we identify surgical site infections using routine electronic health record data?
Millions of operations are carried out in the NHS every year. Infection is a rare but important complication that can happen after surgery. Specific operations are routinely monitored every year in every NHS hospital to see how often these post-surgery infections occur. This is very time-consuming as it is done in person by an infection control nurse for 3 months every year in every hospital, and continuously in cardiac surgery. Even then, the nurses are only able to follow-up patients whilst they are in hospital, and cannot find out if patients get infections after they have been discharged. Every time a person comes to hospital or has an operation, a large amount of routine electronic data is recorded, including about the operation (what type it was, how long it went on for), what blood tests were done and what was found, how long people stayed in different wards for etc. In this project we want to try to use this data to see if we can accurately and reliably predict who developed post-surgery infections, using the data collected by hand by the nurse as the “gold standard”. If we can do this, we could monitor these infections throughout the year.
Trends in bacterial pathogens, antibiotic resistance, antibiotic usage and patient outcomes in critical care in the Oxford University NHS Trust hospitals, 1999-2014
Antibiotics are medicines that are hugely important in treating many infections, and antibiotic resistance is a major clinical problem. Research has shown that increased use of antibiotics, whether appropriate or inappropriate, seems to be associated with higher numbers of antibiotic-resistant infections. To develop the best treatment strategies, it is important to understand which bacteria (bugs) are causing most infections and which antibiotics they are resistant to. This may vary in different hospital settings. Patients in critical care (intensive/high dependency care units) are particularly vulnerable to infections. There are however limited data on infections, antibiotic resistance and antibiotic usage in UK critical care units, and it is unclear: (i) what bugs are causing most infections in these patients; (ii) which antibiotics would be the best ones to use; and (iii) whether there is any change over time. This study would use the IORD database to answer these questions for critical care units in the OUH NHS Trust hospitals. We will look for links between antibiotic use, types of infection and common bugs, and whether this is different from other hospital wards. This information can be used to determine which antibiotic(s) is/are the best one(s) to use in critical care.
Determining patterns of Gram Negative resistance and their genetic basis
Our knowledge of resistance of bacteria, such as E. coli, to antibiotic drugs is incomplete. An understanding of how such resistance is passed on from generation to generation of bacteria is important in helping us to determine which drugs should be given to patients, and in helping us formulate new drugs. This project aims to be use “big data” methods to combine information from many infection-related databases, with the goal of improving our understanding of how resistance to different types of drug are inherited by bacteria such as E. coli. The “gold standard” means of testing resistance to drugs is a series of laboratory tests, however such tests vary from hospital to hospital (and, over time, even within the same hospital), and where resistance to different sets of drugs are typically tested. Also, the results of these tests are imperfect, and occasional mislabelling occurs (where a test might, for example, indicate that a bacteria sample is resistant to a drug, when in fact it is not resistant). With data being collected across many hospitals, there is great potential for combining these datasets using modern methods of machine learning. Such methods are able to cope with occasionally-mislabelled data, and can be used to combine datasets of different kinds: for example, this project will consider the results of lab tests from blood and urine samples, data concerning the severity of infection of the patient from which each sample came, and, for some samples, their genomic sequences.
Outcomes after elective surgery: prognostic information in baseline blood tests
Although relatively rare (under 5%), surgical site infections are the most common complication of planned operations (hospital admissions for what is termed “elective surgery”). Other complications (such as needing to be re-operated on, or even dying) are also rare, but may also be related to underlying infections, even when these cannot be directly identified by growing micro-organisms from patient samples. These complications cause ill-health and are extremely distressing for patients. Some predictors of complications are known – for example, making sure patients glucose levels remain in normal ranges, and that they do not get too hot or too cold reduces the risk of complications. However, recent studies in other infections including tuberculosis and malaria suggest that the relative numbers of some different types of blood cells might also be related to poor outcomes. We want to find out whether these blood cell values could also be a predictor of complications following planned operations, and whether this could be a useful way to identify patients at high risk, who could then be managed more intensively to reduce complications.
Antimicrobial susceptibility profiles of urinary isolates from samples collected in the community
Urinary tract infections comprise around 1% of the 300 million general practice consultations annually in the UK; most of these are acute uncomplicated UTIs (AUUTIs) in women of childbearing age. AUUTIs affect around 50% of women during their lifetime and, by 24 years of age, one-third of women have had a UTI. Although AUUTIs generally resolve without complications, they result in significant morbidity and the goal of treatment is to reduce the severity and duration of symptoms. For most non-pregnant women with AUUTIs, the decision to start treatment is based on a combination of characteristic symptoms, without the need for urine testing. Current national guidelines advise empiric treatment with 3 days of either nitrofurantoin or trimethoprim as resistance to these oral antibiotics is reported to remain low. This low rate of resistance is in contrast to other oral antibiotics. The aim of this proposal is to examine whether there has been a change in (i) the bacterial species and (ii) the antimicrobial sensitivities of bacteria isolated from urine samples submitted from Primary Care over the past 2 decades in Oxfordshire. This data would guide the choice of antimicrobial in a proposal to improve the management of AUUTIs in women.
To what extent do urine cultures provide prognostic information for poor outcome?
Urine cultures are the most commonly performed microbiology test at the Oxford University Hospitals but clinical impression is that many of these tests may be unnecessary. IORD provides an opportunity to see if we can answer clinical questions about the value of tests such as urine cultures for clinical management using routinely collected data. We will investigate whether the results of a urine test (negative, positive for a plausible infecting microbe, positive but only with a probable contaminant) have any impact on subsequent mortality, adjusting for co-morbidities and factors influencing whether or not tests are requested in the first place. We will see whether the main diagnosis (reason the patient came to hospital) is an independent predictor of poor outcome. We will explore whether it is possible to identify high- and low- risk subgroups based on urine test results which could enable better targeting of different management approaches in future.
C. difficile recurrence: incidence, predictors and risk scores
This project focusses on patients who do not recover straightaway from C. difficile, but in whom either initial treatment for the infection does not work, or in whom the infection comes back (termed “recurrence” of infection). We want to work how often C. difficile recurrence happens, and how long does this typically take. We will try to find factors that predict patients having a C. difficile recurrence, and will develop a risk score to predict patients at high risk of this outcome which could be used with electronic patient records. We will consider important demographic (age, sex) and hospital-associated factors (speciality, admitted as emergency/elective, admission speciality, duration and type of previous hospital exposure including days since current admission, number of previous admissions, total prior hospital stay, time since last discharged, previous isolation of S. aureus or C. difficile). Separate models will also include results of laboratory tests at the time a diagnosis was made, that may not be available in all individuals, and strain type. The final regression model will be used to develop an integerised risk score.
Summary of findings: 22% of patients surviving 14 days after their first C. difficile infection will go onto to have a recurrence of C. difficile, mostly in the next 4 months. Risk factors, including increasing age, initial disease severity, and hospital exposure, predict CDI recurrence and identify patients likely to benefit from enhanced initial CDI treatment. See Manuscript on Oxford Journals
Incidence and outcomes following C. difficile infections in Oxfordshire, and predictors of poor outcomes
This project aims to estimate the incidence of C. difficile infection in Oxfordshire, inside and outside hospitals (inpatient, outpatient/day case/A&E, community), over calendar time, analysing the trends to identify potential new strains (as measured by multi-locus sequence type, MLST) threatening the health of Oxfordshire residents. We will also explore whether and how outcomes following C. difficile infection, might be changing, particularly in terms of mortality and changes in laboratory test results. We will then compare the impact of different C. difficile strains (by MLST) on different outcomes, including mortality, admission to ITU, and laboratory tests such as C-reactive protein, and numbers of cells which fight infections (white cells, lymphocytes, neutrophils), in order to identify whether there are specific strains that are more likely to be associated with poor outcomes that we should be particularly looking out for. We will adjust for important demographic (age, sex) and hospital-associated factors (speciality, admitted as emergency/elective, admission speciality, duration and type of previous hospital exposure including days since current admission, number of previous admissions, total prior hospital stay, time since last discharged, previous isolation of S. aureus or C. difficile).
Summary of findings: We found that Clostridium difficile genotype predicts 14-day mortality in 1893 EIA-positive-culture-positive adults. Excess mortality correlates with genotype-specific changes in biomarkers, strongly implicating inflammatory pathways as a major influence on poor outcome. PCR-ribotype-078/ST-11(clade 5) is associated with high mortality; ongoing surveillance remains essential. See Manuscript on Oxford Journals
C. difficile transmission in Oxfordshire hospitals 2006-2009
Summary of findings: In an endemic setting, with well-implemented infection control measures, ward-based contact with other patients with C. difficile diarrhoea cannot account for the majority of new cases – no more than 25% of new cases in Oxfordshire over 2 years could be linked to a previous source based on strain-typing and ward movements. See Manuscript on PLOS
MRSA incidence 1998-2009
Summary of findings: Rates of MRSA infection in blood and other sites were falling before intensification of infection-control measures in 2006. This, together with strain-specific changes in MRSA isolation, strongly suggests that incompletely understood biological factors are responsible for the much recent variation in MRSA isolation. A major, mainly meticillin-sensitive, S aureus burden remains. See Manuscript on BMJ
Using secondary data sources to investigate incidence and severity of respiratory infections
Summary of findings: In this project, we looked at the hospital database of all people admitted to hospital, and found that admissions for pneumonia had almost tripled over the space of 15 years, an alarming increase. By looking at different types of data available in IORD, it became clear that there wasn’t a simple explanation for the increase in pneumonia. It’s not just changing diagnostic codes, it’s not just risk-averse GPs, and it’s not just an ageing population. It appears there is a genuine increase in pneumonia in our region. This has been reported before, but never with this level of detailed investigation. See Manuscript on BMJ
Improving threat detection and quality surveillance: tools for infection management
Summary of findings: We found that increases in severity of C. difficile diarrhoea, likely to due to the arrival of the hypervirulent 027 strain in the UK, could have been detected three years early using routine monitoring of neutrophil counts at diagnosis, compared with looking at patient survival. This study shows how automated electronic systems providing early warning of the changing severity of infectious conditions could be established more generally using routinely collected laboratory hospital data. See Manuscript on PLOS
Nosocomial MRSA endemicity: addition or replacement (multi-centre, multi-cohort study)
Summary of findings: Trends in the rates of serious bloodstream infections caused by antibiotic-sensitive microbes were similar between 1998-2007 across 14 hospitals in Europe, but rates of serious bloodstream infections caused by antibiotic-resistant microbes increased much more in hospitals in countries were antibiotic resistance is in endemic. So antibiotic-resistant infections increase the total disease burden, rather than simply replacing infections with susceptible strains. See Manuscript on Oxford Academic