Ethnic differences in the prevalence of diseases are known but the underlying causes and mechanisms remain unclear. The field of ethnic inequalities in health is relatively young in the UK, compromised by the lack of relevant data. It is, nevertheless, curious that while the UK has long provided some of the best empirical evidence and theoretical understanding of social inequalities in health, the exploration of ethnicity as an alternative dimension of social inequality is relatively recent. It was not until 1984 that first national study of ethnic differences in mortality was published. Between 1984 and 1996 the epidemiological knowledge base increased rapidly but weaknesses in these studies were evident. They were cross-sectional, small, based mainly on migrants of working ages from the Caribbean and the Indian subcontinent and tended to focus on CHD in South Asians and hypertension in Black Caribbeans. The use of ethnicity was linked to a biological or cultural concept, and the examination of the role of socio-economic environmental factors was cursory. The ten papers that I am submitting for this thesis were published between 1996 and 2004 and are located within this developing context of research in ethnic differences in health. At a time when there were no other national longitudinal studies of ethnic minorities, these papers were among the first to exploit the potential of the Office for National Statistics Longitudinal Study to examine ethnic differences in health. The findings of my submitted papers enhance the understanding of mortality and morbidity patterns in migrant groups and their children. Briefly, Papers I-III show that, in spite of the progressive improvement in socio-economic circumstances across three generations (grandparents, parents and children), Irish people living in England and Wales continued to show adverse mortality and cancer incidence patterns; Paper IV signals that addressing the risk of cancers is a public health priority for all minority groups, even though rates may be lower than the national average in some groups; Paper V shows the continuity of limiting long-term illness patterns across most migrant and UK-born minority groups; Papers VI and VII show that cumulative disadvantage was more common in South Asian and Caribbean migrants and that downward social mobility is associated with a disproportionate impact on limiting long-term illness in migrants; Papers VIII and IX show that the duration of residence in England and Wales and age at migration are important influences on the mortality of South Asian and Caribbean migrants in England and Wales; Paper X shows that there was little shift in mean birth weight between babies of migrant mothers and babies of UK-born minority mothers in the same ethnic group, which may have a continuing legacy on ethnic health inequalities, notably in cardiovascular disease. In summary, these published papers shed light on the social patterning of ethnic differences in health. This work stimulated the development of my current research programme. The papers included in the submission for the degree of PhD by published work are I. Harding S, Balarajan R. Patterns of mortality in second generation Irish living in England and Wales: longitudinal study. British Medical Journal 1996;312(7043):1389-1392. II. Harding S, Balarajan R. Mortality of third generation Irish people living in England and Wales: longitudinal study. British Medical Journal 2001;322(7284):466-467. III. Harding S. The incidence of cancers among second generation Irish living in England and Wales. British Journal of Cancer 1998;78(7):958-961. IV. Harding S, Rosato M. Cancer incidence among first generation Scottish, Irish, West Indian and South Asian migrants living in England and Wales. Ethnicity and Health 1999;4(1-2):83-92. V. Harding S, Balarajan R. Limiting long-term illness among Black Caribbeans, Black Africans, Indians, Pakistanis, Bangladeshis and Chinese born in the UK. Ethnicity and Health 2000;5(1):41-46. VI. Harding S, Balarajan R. Longitudinal Study of Socio-economic Differences in Mortality Among South Asian and West Indian Migrants. Ethnicity and Health 2001;6(2):121-128. VII. Harding S. Social mobility and self-reported limiting long-term illness among West Indian and South Asian migrants living in England and Wales. Social Science & Medicine 2003;56(2):355-361. VIII. Harding S. Mortality of migrants from the Indian subcontinent to England and Wales: Effect of duration of residence. Epidemiology 2003;14(3):287-292. IX. Harding S. Mortality of migrants from the Caribbean to England and Wales: effect of duration of residence. Int. J. Epidemiol. 2004;33(2):382-386. X. Harding S, Rosato M, Cruickshank JK. Lack of change in birthweights of infants by generational status among Indian, Pakistani, Bangladeshi, Black Caribbean, and Black African mothers in a British cohort study. Int. J. Epidemiol. 2004;33(6):1279-1285.

The aims of this PhD study were to assess the feasibility of accessing non-medical settings within which to offer chlamydia screening, to ascertain the knowledge of chlamydia and young men’s and women’s views towards non-medical screening, and to assess relative willingness to be screened for chlamydia by young men and women. Results: Eighty-four percent of age eligible users approached participated in education, health and fitness and workplace settings (n=126, n=133 and n=104, respectively). Of all sexually active people 113 (32%) were willing to be tested for chlamydia in non-medical settings. Uptake of testing was highest in the health and fitness setting (50% uptake for both women and men compared with 20% in education and 30% in workplace settings). In each setting young men were more willing than women to accept the offer of a chlamydia test. Overall, 40% of men approached provided a sample compared with 27% of all women. Disease prevalence was 4.4% (4.9% in men; 3.8% in women). Interview data suggests young men’s willingness to be tested for chlamydia in non-medical settings is due to convenience and raised awareness of the largely asymptomatic nature of chlamydia infection. Whilst 94% of men screened had never been tested for chlamydia before, one in three young women screened had previous screening experience. Women’s lower uptake of screening was due to concerns about the public nature of the settings leading to stigma. Conclusions: Increasing opportunities for the take-up of screening in non-medical settings could be an effective approach to reaching young men and have a significant impact on the incidence and prevalence of this easily treated STI, thereby reducing the future burden of unwanted reproductive health sequelae.

This thesis analyses the findings of empirical research carried out in three case study UK health authorities in 1996-97, using repeat interviewing of senior managers. It aimed to test three competing hypotheses: i. Markets are one possible system for allocating scarce resources. The process of contract specification in a complex quasi-market is likely to make rationing more explicit than it would be in a hierarchical system ii. In the complex context of the NHS the quasi-market may fail to produce clear contracts and unambiguous allocations, because of prohibitive transaction costs, political costs and ethical costs of greater explicitness iii. Other pressures in favour of explicitness (e.g. rising expenditure, effectiveness evidence and the Patient's Charter) may be irresistible, whatever structural form the NHS takes. The complex relationship between explicit rationing, the internal market and other factors is discussed. Results suggest the quasi-market has contributed to the growth in explicit rationing, notably by decoupling purchasers and providers from their previously shared responsibility to manage resources. In other respects the market has speeded up or magnified the effect of other factors which would or could have happened anyway. Concern to control rising expenditure has led to more explicit decisions but is now rekindling interest in the value of fixed budgets for providers and implicit clinical decision-making. Factors such as the Patient's Charter have also had an independent effect on greater explicitness. Implicit rationing remains significant. The implications for health care rationing of government proposals to abolish the internal market are examined. The results suggest that explicit rationing will probably continue to grow, but with a greater emphasis on explicit criteria to guide clinicians in determining who gets treatment, rather than the exclusion of whole services. The retention of some form of commissioner provider split may also exercise continuing pressure towards explicitness.

Introduction There exists a ‘Scottish effect’, a residue of excess mortality that remains for Scotland relative to England and Wales after standardising for age, sex and local area deprivation status. This residue is largest for the most deprived segments of the Scottish population. Most Scottish areas that can be classified as deprived are located in West Central Scotland and, in particular, the City of Glasgow. Therefore the central aim of this thesis is to establish the existence of a similar ‘Glasgow effect’ and identify if the relationship between deprivation and all cause mortality is different in Glasgow to what is in other, comparable cities in the UK. Methods A method to compare the deprivation status of several UK cities was devised using the deprivation score first calculated by Carstairs and Morris. The population of mainland UK was broken into deciles according to the Carstairs score of Scottish postcode sectors and English wards. Deprivation profiles for particular cities were drawn according to the percentage of the local population that lived in each Carstairs decile. Using data from the three censuses since 1981, longitudinal trends in relative deprivation status for each city could be observed. Analysis of death rates in cities was also undertaken. Two methods were used to compare death rates in cities. Indirect standardisation was used to compare death rates adjusting for the categorical variables of age group, sex and Carstairs decile of postcode sector or ward of residence. Negative binomial models of death counts in small areas using local population as the exposure variable were also created; such models allow the calculation of SMRs with adjustment for continuous variables. Covariates used in these models included city of residence, age group, sex, Carstairs z-score and also the z-scores for each of the four variables from which the Carstairs score is comprised (lack of car ownership, low social class, household overcrowding and unemployment). Results The deprivation profiles confirmed that all UK cities have a high proportion of deprived residents, although some cities have far higher proportions than others. Some cities appeared to show relative improvement in deprivation status over time whilst others seemed resistant to change. Glasgow was the most deprived city at all census time points and the Clydeside conurbation was also more deprived than all other conurbations. SMRs calculated by indirect standardisation indicated that many cities have excess mortality compared to the whole of the UK when adjusting for age group and sex only. Three cities, Glasgow, Liverpool and Manchester, had SMRs that were significantly higher than all other cities at every census time point. Adjusting SMRs for Carstairs deprivation decile diminished the magnitude of this excess mortality in most cities. However, adjusting for Carstairs decile did not diminish the excess mortality in Glasgow sufficiently and there remained a significant, unexplained residue of excess mortality in Glasgow. SMRs generated by regression models adjusting for continuous variables were able to reduce the size of the excess mortality in most cities, though the model producing the lowest SMR varied from place to place and from time to time. In Glasgow, a regression model including age group, sex and lack of car ownership as covariates explained most of the excess mortality at all three time points. Discussion and Conclusion The relationship between deprivation (as measured by the Carstairs index) and death rates in Glasgow did appear to be different to other cities, and there seems to be evidence of a Glasgow effect. There are several reasons why this might be the case, including; the Glasgow effect may be apparent rather than real – an artefact of the Carstairs measure of deprivation failing to capture the complex nature of multiple deprivation; The effect may be the result of migration patterns to and from the city; the effect may be the result of historical levels of deprivation; or the effect may result from different behavioural patterns among Glasgow residents compared to residents of other UK cities. In conclusion, the results show that continued efforts by public health professionals, politicians and residents have failed to produce a step change in the city’s relative health status and Glasgow continues to lag some way behind other cities in the UK. The ability of the Carstairs measure to describe multiple deprivation is called into question. Future research should focus on identifying specific causes of mortality that contribute to the Glasgow effect; on qualitative work to identify if there is a distinct set of social norms in deprived neighbourhoods of Glasgow that contribute to unhealthy patterns of behaviour; and on creating a deprivation index that can be used on equivalent units of geographical area in both Scotland and England.

Background Homelessness shares many similarities with other socio-economically deprived circumstances. It was not known whether the health of homeless people was similar to that of other deprived non-homeless populations. Aims To describe hospital admissions and deaths in a cohort of homeless Glasgow adults and to compare these to socio-economically deprived groups within a matched sample of the non-homeless local population. Methods A retrospective 5-year cohort study was conducted comparing an exposed (homeless) cohort of adults with an age and sex matched unexposed (non-homeless) cohort from the local general population. All participants’ linked hospitalisation and death records were identified. Survival was analysed using comparisons of rates, Kaplan-Meier plots and Cox proportional hazards models. Hospitalisation rate ratios were compared using an exact Poisson method. Additional proportional hazards models were produced to adjust for morbidity, which was identified in hospital records up to 5 years before death. Results 6323 homeless and 12 625 non-homeless adults were studied. The mean ages of men and women in both cohorts at entry were 33 and 30 years, respectively, and 65% were men. After 5 years 1.7% of the general population and 7.2% of the homeless population had died. Age and sex adjusted hazards of death, compared with residents of the most affluent areas, were 2.6 (95% CI 1.5 – 4.4) for residents of the most deprived areas and 8.7 (95% CI 5.2 – 14.5) for homeless individuals. Men were at twice the risk of death as women. Homelessness was associated with death on average 12 years younger than the matched general population (41 versus 53 years). A third of deaths in the homeless were caused by drugs and a further 16% by alcohol. In the homeless, adjusted hazards ratios for deaths by drugs were 20.4 (95% CI 12.0 – 34.7), for suicide were 8.4 (95% CI 3.9 – 18.2), for assault were 7.0 (95% CI 2.6 – 19.0) and for alcohol were 4.7 (95% CI 3.1 – 7.1) compared with the non-homeless population. Homelessness remained an independent risk factor for death after adjustment for morbidities, with a hazard ratio of 2.4 (95% CI 1.3 – 4.3) compared with living in the most affluent non-homeless circumstances. Hospitalisation for alcohol related conditions increased the risk of death from alcohol by 42-fold but homelessness added no further hazard. In contrast, hospitalisation for drug-related causes raised the risk of death from them by 4-fold and homelessness added a further 7-fold risk. The risk ratio for emergency hospitalisation in the homeless was 6.4 compared with the non-homeless. Admission rates were higher in the homeless for all conditions except cancers. Risk ratios in the homeless compared to the most affluent non-homeless cohorts were highest for cellulitis (risk ratio 112.9, 95% CI 20.2 – 4472.0), drug poisoning (risk ratio 90.0, 95% CI 16.0 – 3565.9) and convulsions (risk ratio 71.5, 95% CI 12.7 – 2834.1) In men, lengths of stay were longest in patients from the most affluent areas and shortest in the homeless. In women, lengths of stay increased with greater socio-economic deprivation but homeless women had stays that were typical of the general population. There was little difference in elective admission rates across different socio-economic strata. Homelessness was associated with a small reduction in risk of elective hospitalisation in men and a small increase in women compared with the general population. Admissions for treatment of infectious and parasitic disease were 9 times more common in the homeless. Admissions for injuries, poisonings, mental and behavioural disorders, and maternity related diagnoses were around 2-3 more common in the homeless. Homelessness was associated with almost 3-fold increases in elective admissions for abortions but an 80% lower risk of vasectomy. Lengths of stay for elective admissions increased with deprivation and were longest in the homeless. Conclusions The morbidity and mortality of homeless adults is significantly worse than that of the most deprived non-homeless populations of Glasgow. Hospital inpatients who are homeless are at greater risk of death for a number of conditions and may benefit from more intensive treatment and follow-up.

Background Road traffic crash (RTC) injuries affect 20 to 50 million people worldwide every year, causing premature death or disability as well as incurring large costs to individuals and society. In the UK, the number of RTC casualties is underestimated if based solely on police records, as many casualties are unreported to the police. “Safety” (speed and red light) cameras have shown to be an effective way of combating RTCs and in 2000 a national scheme was rolled out in the UK. Aim and objectives The overall aim of the study was: To investigate the epidemiology, cost and prevention of RTC injuries in the Strathclyde police region of Scotland. The specific objectives of the study were: 1. To establish the overall epidemiology and accuracy of reporting of RTC injuries in Strathclyde. 2. To determine the epidemiology of RTC injuries and the effectiveness of safety cameras at the camera sites in Strathclyde with special reference to different road users, RTC types and severity, before and after camera installation. 3. To estimate the economic burden of hospital admissions due to RTC injuries in Strathclyde and at the camera sites before and after installation. Methods Nine years (1997 to 2005) of police road casualty records (STATS19) and National Health Service hospitalisation records (SMR01) from the Strathclyde region were linked. The linkage resulted in nearly 11,000 police casualty records relating to approximately 30,000 hospital and death records. Unlinked RTC hospital and police casualties (nearly 9,000 and 70,000 respectively) were also utilised in the analysis. The study employed a range of epidemiological and economic methods. These included descriptive epidemiology (evaluating distributions of linked and unlinked records, length of stay and cost analysis), analytical epidemiology (examining associations using chi square and logistic regression models) and interventional epidemiology (before and after study). The economic evaluation utilised weighted mean costs. The focus of analysis was threefold: 1. Epidemiology of RTC, injuries and accuracy of police recording, 2. Epidemiological impact of safety cameras, 3. Cost of road traffic crashes a) in Strathclyde and b) at safety camera sites. Results Epidemiology of RTC injuries in Strathclyde: Older age and less protected road users (i.e. pedestrians and two-wheeled vehicle users) had a higher risk of a more severe outcome in RTCs. Head injuries were more common among pedestrians and pedal cyclists, while car occupants more often suffered injuries to the thorax and abdomen/lower back/lumbar spine. Accuracy of police reporting: 45% of RTC hospital admissions were not recorded by police. Casualty characteristics significantly associated with underreporting were: no third party involvement, older age, casualties from early in the study period, type of road user (especially pedal cyclist), hospitalisation as a day case and female gender. Seriously injured casualties recorded by police (STATS19) declined in frequency more than the RTC hospitalised injuries (SMR01) (38% and 21% respectively). Linked SMR01 casualties that were coded “slight” by the police increased by 5% over time, while linked SMR01 casualties coded “serious” declined by 27%. Safety camera impact: Compared to the rest of Strathclyde, there was a significantly greater downward linear time trend of RTC incidence at the camera sites. The impact of cameras on RTCs over time appeared stable. Cameras seemed to be effective in reducing the incidence of serious or fatal RTC injuries, as well as injuries associated with multiple-vehicle and non-junction RTCs. Cost of RTC casualties in Strathclyde: Total inpatient costs were conservatively estimated at £7.3 million yearly (linked records). Head and lower extremity injuries incurred the highest total costs (28% and 34% respectively). Pedestrian injuries, constituting 36% of the total, incurred 44% of total costs. Casualties from deprived areas, and pedestrians in particular, incurred higher hospital costs than other road user groups. Cost of RTCs at safety camera sites: 17% of all injured before safety camera installation were hospitalised, while 13% of casualties after installation were hospitalised. The mean costs of (surviving) casualties admitted to hospital declined by 24% after installation and the mean daily cost declined by 55%. Conclusions RTC injury incidence in Strathclyde declined over the study period, which is in line with expectations of developed countries. Young and elderly people as well as unprotected road users carry a disproportionately great RTC injury burden. Many hospitalised RTC casualties were not recorded by police and there appears to have been an increasing tendency over time for police officers to report injuries as slight rather than serious. National (UK) statistics of RTCs should be interpreted with caution in the light of these findings and routinely linking police and hospital data would enhance the quality of RTC casualty statistics. Linking police and hospital RTC records provide a more comprehensive source for road traffic analysis than any of the sources separately. Routine data linkage would also facilitate the evaluation of time trends in relation to national road casualty reduction targets. The study indicates that the most costly RTCs occur in areas with high levels of deprivation, a history of pedestrian RTCs, elderly and child casualties, roads with many non-junction RTCs and 30 mph speed limits. The evaluation of safety cameras strongly suggests that they are effective in reducing both road casualty incidence and severity and that the reduction in incidence is sustained over time. Additionally, safety cameras in Strathclyde may have contributed to a saving of over £5 million. Cameras thus fulfil an important public health, as well as law enforcement, function and should continue to play a central role in traffic calming. This study has demonstrated the value of utilising multiple data sources in the road traffic injury field.

Background: The HIV/AIDS epidemic has contributed to a drastic increase in the number of orphans and vulnerable children in sub-Saharan Africa. However, little is known about the mental health of these children in low prevalence areas such as Ghana. The thesis investigated the relationship between orphanhood, parental HIV/AIDS status and mental health. It further examined the mediating effects of identified risk and protective factors on the relationship between orphanhood/parental HIV/AIDS status and psychological difficulties. Finally, the thesis identified pathways through which HIV/AIDS impacts children by exploring the interactive and cumulative effects of the various risk and protective factors on psychological difficulties. Method: The thesis employed cross-sectional, quantitative interviews that involved 291 children aged 10-18 years and their caregivers that compared children who have lost their parents to AIDS, those who have lost their parents to other causes, those who are living with HIV/AIDS-infected caregivers and children from intact families in the Manya Krobo district in Ghana. ANOVAs, T-tests, General Linear Models, Log-linear Analyses, Chi-Squares and Bivariate Correlations were used to analyze the data that were obtained from both the children and their caregivers. Results After controlling for relevant socio-demographic factors, both children and informants’ reports showed that children orphaned by AIDS and those living with infected parents showed higher delinquency (p <.01), peer problems (p <.001), hyperactivity (p <.01) and lower self esteem (p <.001) than other orphans and children from intact families. AIDS orphans, other orphans and those living with HIV/AIDS-infected parents all reported significantly more depression (p <.001) and relationship problems (p <.001) than those for intact families. Conduct problems as indicated by informants’ reports were generally, significantly higher for orphans and vulnerable children compared to children from intact families. Over 70% of both AIDS orphans and children living with infected parents showed internalising symptoms that were above clinical cut-offs for abnormality. AIDS orphans and children living with infected parents reported more stigma, abuse, child labour and lower levels of SES and lower perceived social support. These factors independently, strongly mediated the relationship between orphanhood, parental HIV/AIDS status and mental health. The interactive and cumulative effect of engagement in child labour and being physically abused heightened the risks for depressive symptoms from 38% to 66%. Neglect and psychological abuse increased the risks for symptoms of Reactive Attachment Disorder from 26.6% to 67.3%. The cumulative effect of stigma and either child labour or physical abuse substantially increased the likelihood of delinquency symptoms to approximately 67%. Conclusion: The findings demonstrated that both AIDS orphans and children living with HIV/AIDS-infected parents showed heightened psychosocial symptoms. The present evidence also highlighted the interactive, cumulative, co-occurrence of contextual factors and HIV/AIDS unique exposures to create heightened vulnerabilities for psychological difficulties among children. The findings call for a comprehensive intervention programme that addresses factors specific to HIV/AIDS and contextual variables.

Background and Aims: Judicial antibiotic prescribing and appropriate use of healthcare resources are public health priorities. Preschool-aged children that attend day care frequently consult general practitioners (GPs) and receive antibiotics, despite experiencing mainly self-limiting and/or viral infections. North-American surveys indicate that day care providers unnecessarily exclude children with infections, and make exceptions to exclusion on the basis of antibiotic treatment. Commentators suggest that this may lead to unnecessary consultations and inappropriate antibiotic requests. This study’s main aim was to explore whether UK-based day care providers’ management of infections encourages parents to unnecessarily consult GPs, and inappropriately seek antibiotics. A secondary aim was to describe the content and nature of written day care sickness exclusion policies. Questionnaire Methods and Results: Questionnaires were distributed to 329 day care providers in three socio-demographically contrasting areas of South-East Wales, to gather descriptive data regarding sickness exclusion policies. 216 (66%) responses were received. Policies were mostly self-written, diverse in content and detail, and often non-evidence-based. Qualitative Methods and Results: Day care providers’ management of infections, and the influence this had on parents’ consulting and antibiotic-seeking behaviours, were explored through semi-structured interviews with 24 purposefully selected day care providers, and 28 opportunistically-selected parents that used their services. Interviews underwent inductive thematic analysis. All day care providers encouraged parents to consult GPs for self-limiting infections, and often inappropriately advised antibiotic treatment through written policies and verbal communication. Some parents felt that day care attendance increased their tendency to consult for symptoms they would usually manage themselves. The purpose of consultation was often to expedite return to day care, rather than alleviate concern. Parents understood that antibiotics were unlikely to be beneficial, but still sought and received treatment in order to appease day care providers’ requirements. Conclusion: Day care providers’ inappropriate advice to parents, together with non-evidence-based exclusion policies, contribute to unnecessary GP consultations and inappropriate antibiotic-seeking behaviour.

The high-resolution molecular characterisation of the Cryptosporidium species and subtypes biquitous in environmental samples can provide important information regarding their potential public health significance. The purpose of this study was to develop, evaluate and apply sensitive screening and DNA recovery methods to environmental Cryptosporidium for molecular characterisation. A systemic literature review as undertaken to identify methods of subtyping Cryptosporidium species recovered from environmental samples. Prior to molecular characterisation, the recovery and detection of oocysts from water is by immunomagnetic separation and immunofluorescence microscopy. However, this method is not currently suited for screening large numbers of faecal samples. A commercially available faecal parasite concentrator protocol was modified and evaluated for the enhanced detection of Cryptosporidium oocysts. Once recovered from samples, DNA must be released from the oocyst bound sporozoites before molecular methods can be applied. Commonly used oocyst disruption methods were identified and using samples containing high numbers of oocysts, evaluated by microscopy and a SYBR Green real-time polymerase chain reaction (PCR) developed from the internal stage of a previously published nested small subunit ribosomal DNA PCR. The eight best methods were then evaluated with low numbers of oocysts in the presence of immunomagnetic beads to replicate field samples, using the published nested PCR and SYBR Green real-time PCR. The enhanced screening and optimal DAN recovery combined with microsatellite multilocus fragment analysis was applied in a study investigating the species and subtypes of Cryptosporidium recovered from water, non-clinical farmed and wild animal faeces as well as clinical human and cattle samples from within a single water catchment. Several human pathogenic Cryptosporidium subtypes were identified in addition to the unprecedented finding of Cryptosporidium as the predominant species in the catchment surface water. This study demonstrated the potential application of the developed methodology in the public health investigation of environmental Cryptosporidium.

Despite structured enforcement of food safety requirements known to prevent foodborne disease outbreaks, catering businesses continue to be the most common setting for outbreaks in England and Wales. Limited published evidence suggests that the way catering businesses are managed and operated may contribute to food safety control failures which in turn can result in an outbreak. The purpose of this study, funded by the Food Standards Agency1,2 was to identify the underlying management factors that may contribute to or prevent outbreaks in the catering industry. A matched case control study compared the management and operational practices of 148 catering businesses associated with foodborne outbreaks with 148 control catering businesses. High response rates were achieved: case businesses 90%, and control businesses 93%. To minimise false inferences from chance associations analysis followed a predefined hypothesised causal pathway. Hazard analysis critical control point systems and formal food hygiene training were found not to be protective and food hygiene inspection scores were not useful in predicting which businesses were likely to be associated with outbreaks. Larger small medium sized (SME) businesses were more likely to be associated with outbreaks compared to micro SME businesses. Operational and management practices did not differ significantly between case and control businesses when adjusted for SME size. SME size was not explained by other staff employment and management variables. However, businesses associated with Salmonella outbreaks were significantly more likely to use regional egg suppliers, the only significantly independent operational practice associated with Salmonella outbreaks. Regional egg suppliers were also more likely to supply businesses associated with outbreaks that were attributed to food vehicles containing eggs. Businesses associated with egg outbreaks were less likely to use eggs produced under an approved quality assurance scheme suggesting that the underlying risk associated with using regional suppliers may relate to the use of contaminated eggs.