Wales Census Research Articles

Data resource profile: Exploring freely accessible data describing wider determinants of health in England

IntroductionIn England, life expectancy has stalled and significant decreases observed in certain geographical areas and populations. The cause of this involves complex dynamics between an individual's health, characteristics, lifestyle, and their wider environment known as the wider determinants of health which are key to good life expectancy, healthy life expectancy, and prevention of long-term medical conditions. Knowing the availability, breadth, features, and linkage potential of datasets relevant to wider determinants of health is important for exploring trends and associations for policy and public health planning. MethodsA systematic mapping of internet content identified accessible datasets relevant to wider determinants of health in England with town level geographical granularity or lower. Search terms were used in search engines and chatbots to identify weblinks subsequently examined for eligible datasets. Results105 potential weblinks to datasets were identified. Of these, twenty-one weblinks were explored further after exclusion of those: not accessible or currently live (n = 13); duplicated across search engines (n = 17); providing information only (i.e. no raw data, n = 14); did not provide freely accessible data (n 3); were not relevant to wider determinants of health (n = 17); lacked geographical granularity (n = 26). Eighty-nine datasets of interest were compiled with sub-town level data aggregation. Approximately half (n = 47, 52%) were from the England and Wales census 2021, with the remaining sources including government bodies, public services, and research datasets. Datasets covered many valuable categories of wider determinants of health. Key data gaps included food consumption, social care data and community/voluntary services. ConclusionIn England, access to data relevant to wider determinants of health is good and available at relatively small geographical resolution. Accessible datasets were identified and compiled within multiple categories of wider determinants of health as a useful data resource to explore wider determinants of health at place if linked to relevant health data or population studies. Key Features This data resource profile describes a systematic mapping of freely accessible population data on wider determinants of health in England. To the authors knowledge this is the first comprehensive compilation of freely accessible data resources of this kind. This data resource profile was created to support research into the mechanisms and impact of wider determinants on the health of populations in England but is applicable to research and populations studies wider than this. Eighty-nine datasets were identified that may be of use to researchers in health and other population data fields. Datasets are held separately but many have the potential to be linked through common geographical area codes to other relevant datasets such as health data. The datasets originate from multiple sources including government bodies, public services, and research datasets. They cover many key categories of wider determinants of health including socioeconomics, employment, housing, support services, and the environment. Key data gaps included food consumption, social care data and community/voluntary services. The data resource profile in this article links directly to freely accessible datasets at the time of publication in the tables provided. The authors welcome contact from researchers for collaboration on exploring wider determinants of health through data use. Please contact Dr Melanie Rees-Roberts m.rees-roberts@kent.ac.uk

Using linkage to assess coverage of population estimates

ObjectivesThe Demographic Index (DI) comprises of five linked administrative datasets, used for population estimation. Current linkage methods are not ideal to utilise the power of this asset. Using the 2021 England and Wales Census, we developed an innovative composite linkage method to fully utilise the power of the DI. ApproachUsing non-greedy deterministic and probabilistic linkage methods, we linked the DI to the Census at a composite level where we believe links exist – i.e., linking a Census record with a DI cluster (consisting of linked records from the data sources used to make the DI). Next, pairwise linkage of records within these clusters was performed to ensure that every DI record within the cluster was a link. Clerical review of a sample of the full linkage was performed using clerical resolution and search to resolve uncertain and conflicting links and to inform the quality of our linkage. Due to data quality errors in the administrative data, clerical techniques were not always able to determine true match status, so clerical identified ‘uncertain’ links to estimate precision and recall ranges rather than a set estimate. Results A quality analysis to estimate linkage accuracy was conducted and suggests that the linkage has a precision of between 99.3% and 99.7%, and recall between 99.1% and 99.7%. ConclusionWe have linked the ONS’ composite population-level dataset to the 2021 England and Wales Census. The presentation will showcase the methods developed and how we ensured the highest quality possible. *** Please note that this was subbmited in 2022 but the presenters were unable to present due to the UK Queen's death.

Linkage and Demographic Indexing for Nursing Dynamics in the Integrated Data Service

Objective and ApproachThis project links UK Nursing and Midwifery Council (NMC) Registrant data to the 2021 England and Wales Census (“Census 21”). The Office for National Statistics (ONS) has composed a framework of index tables to facilitate linkage between prominent datasets in the UK. In linking person-level data to Census 21, ONS’ Demographic Index enables onward linkage to other administrative datasets that are indexed alike. NMC Register and Census 21 records have been initially linked using bespoke deterministic and probabilistic methods, with resulting matches being joined to the Demographic Index. ONS’ Demographic Index Matching Service (DIMS) will also be used to link NMC data to the Demographic Index to enable comparisons between methods. ResultsAnalysis of both approaches will compare quality metrics such as match rate, recall and precision and tools will be used to assess bias in matches made through each method. These metrics will provide an insight into how the index matching services have been designed and whether improvements can be made to account for any shortcomings in their performance. ConclusionsHybrid approaches using both bespoke and indexing methods are proving valuable where sole use of the matching services is not suitable. This research will help ONS to refine which methods will yield the highest quality linkage and when. ImplicationsAs ONS pursues an ‘indexing first’ approach to linkage, the outcomes of this and similar projects will influence the future of utilising indexes in varied capacities when creating or maximising the potential of linked data.

LGBT+ representation higher education in England and Wales

This paper investigates the level of LGBT+ representation among staff and students in higher education in England and Wales. We compare data from the 2021 England and Wales Census to Higher Education Statistics Agency (HESA) data. We find that LGBT+ people are more highly represented in higher education among all staff groups and students than in relevant comparator groups according to age and educational level in the general population. LGBT+ representation among students and STEM academics is modestly higher than the general population comparator group, while representation among non-STEM academics is substantially higher than one would expect from the general population comparator group. We found no statistically significant under-representation in any particular higher education institution.

Using linkage to assess coverage of population estimates.

ObjectivesThe Demographic Index (DI) comprises of five linked administrative datasets, used for population estimation. Current linkage methods are not ideal to utilise the power of this asset. Using the 2021 England and Wales Census, we are developing an innovative composite linkage method to fully utilise the power of the DI.
 ApproachUsing non-greedy deterministic and probabilistic linkage methods, we will link the DI to the Census at a composite level where we believe links exist – i.e., linking a Census cluster (consisting of linked Census and Census Coverage Survey (CCS) records) with a DI cluster (consisting of linked records from the data sources used to make the DI). We will then conduct a pairwise linkage of records from these linked clusters to link individual source records to the Census. We will utilise clerical review to resolve uncertain and conflicting links and to inform the quality of our linkage.
 ResultsWe anticipate producing a high-quality linkage that will inform how the coverage of the DI compares to Census (through the composite-level linkage) and the quality of the DI itself (through the pairwise-level linkage). We have developed a clerical matching system that can display composite-level linkage, i.e., candidate cluster-pairs. We will tailor our clerical review and quality assessment to records that fall within carefully chosen postcode areas, to ensure all hard-to-count groups and geographical areas are sampled. Working with large datasets is a challenge we are overcoming by using distributed computing and search space reduction.
 The 2021 Census has been previously linked to the CCS with high accuracy; these records are considered intrinsically linked.
 ConclusionTo assess national population estimates’ quality and the policy decisions based upon them, we are linking a key composite population-level dataset to the 2021 England and Wales Census. The presentation will showcase the methods we are developing and how we are ensuring the highest quality possible.

2021 Census to Census Coverage Survey Matching Results.

The 2021 England and Wales Census was matched to the Census Coverage Survey (CCS). This was an essential requisite for estimating undercount in the Census. To ensure outputs could be produced within a year of census day, matching had to be completed in eight weeks.
 We used the Data Access Platform to write efficient and accurate automatic matching algorithms. Probabilistic and associative matching methods were used to present candidate pairs that could not be automatically matched but likely consisted of many true matches. These were resolved by matchers in the clerical matching system. Active learning was used to iteratively improve probabilistic parameters as matches were made.
 We also developed a pre-search algorithm that simplifies clerical matching by replacing clerical searching (here’s a record, can you find a match?) with clerical resolution (here are two records, do they match?).
 As a result of our improvements, we increased our automatic matching rates from 70% (2011 Census) to 93% (2021 Census) for person matching and from 60% to 95% for household matching, without loss of accuracy. Precision and recall were estimated to both be 99.96% for person matching and 100% and 99.78% respectively for household matching. For both persons and households, the same trends seen in 2011 are seen here too; however, most of the biases have decreased and most match rates have increased.
 Dual system estimation using the matched and unmatched records enabled us to estimate that coverage in the 2021 England and Wales Census was 97%.
 The matching methodology implemented for the 2021 Census to CCS matching improved both person and household matching greatly when compared with the 2011 matching. The clerical matching system enabled clerical matching at scale and at pace. Thus, we successfully completed all clerical matching within our eight-week deadline.

Using an Alternative Method to Estimate Overcount for Census 2021.

ObjectivesThe project aimed to test an alternative method to estimate the number of duplicate responses in the 2021 England and Wales Census. The method utilises information from all census records instead of relying on samples. It requires less clerical review than the original inverse sampling method used for overcount estimation. ApproachWe used the Splink implementation of Fellegi-Sunter to match the 2021 Census to itself. The resulting linked dataset was filtered to retain only the top scoring record pair for each unique census record, giving a final dataset of around 71.5 million record pairs. These pairs were divided by score into 13 homogeneous buckets. Random samples of 1000 pairs per bucket were clerically reviewed to determine whether each pair was a true duplicate or not. The clerical results were used to assign an estimated probability of being a duplicate to each bucket and hence to every census record within the bucket. ResultsA dashboard was created which contained percentages of duplicates by region and ‘overcount group’ for the original and alternative methods. This enabled us to view the data side by side and create visualisations to aid analysis. The alternative method had higher average duplicate percentages in overcount groups for communal establishments and was also higher for 9 out of 10 regions in the armed forces overcount group. We found that both the original and alternative methods of overcount estimation followed the same pattern in terms of minimum and maximum duplicate percentages except for those in communal establishments where minimum and maximums were both higher than the original method. Additionally, the minimum was marginally higher for the student and armed forces overcount groups for the alternative method. ConclusionOn initial comparison, estimated rates of duplication resulting from the different methods are comparable. We plan to conduct further analysis on the similarities and differences of the two methods, and research whether the new method could be applied to estimate the duplication rate in other large datasets including administrative data.

Exploring Goldstein et al.’s Scalelink method of data linkage.

ObjectivesScalelink is an innovative probabilistic data linkage method based on correspondence analysis. Unlike the popular and widely-used Fellegi-Sunter algorithm, it does not assume linkage variable independence. It also claims to be more intuitive and computationally efficient. We aim to test this method for the first time on real-world big data.
 ApproachScalelink uses agreement states for each linkage variable and candidate pair. These are compared to determine how frequently, for all candidate pairs, any given agreement state is held at the same time as any other agreement state (this accounts for variable dependence). The results of this comparison are inputted into a loss function and the minimisation of this function is derived within constraints to produce weights. Currently, the method is accessible via Goldstein et al.’s paper and R package. We are translating it into PySpark to enable testing on datasets that are too large to link without using distributed computing.
 ResultsInitial testing of Goldstein et al.’s Scalelink method on small samples of real-world datasets shows that it performs as expected for a probabilistic linkage method, although cannot currently deal with missingness. To test the quality of the method on real-world big data, a high-quality linked dataset of the 2021 England and Wales Census and follow-up Census Coverage Survey will be used as a Gold Standard (GS). After developing a method that enables Scalelink to deal with missingness, we will apply Scalelink and automatic Fellegi-Sunter probabilistic linkage to this GS. We can thus establish and compare the precision and recall of both methods. We will also investigate linkage bias for particular demographics, test computational efficiency and estimate the clerical review burden for each method.
 ConclusionGoldstein et al.’s Scalelink algorithm shows promise as a high quality, scalable, dependence-free linkage algorithm for use in any matching project. Here, for the first time, we research the method’s quality and feasibility with real-world big data. From this we will produce recommendations regarding its utility.

Bespoke automated linkage to enable analysis of covid deaths by ethnicity.

In early 2020 there was intense media speculation that ethnicity and Covid-19 deaths were correlated. However, the existing method of adding ethnicity to death records resulted in low linkage rates for very recent deaths. We designed and implemented a bespoke linkage in three days enabling accurate reporting to the nation.
 We linked the 2011 England and Wales Census to death records using a range of personal identifiers. Due to time pressure, we focused on executing a single linkage method well. Deterministic linkage was chosen, using a variety of matchkeys which were tested via clerical review. To overcome the issue of addresses changing since 2011, we also linked 2020 death record residuals to the 2019 Patient Register (PR) and then made use of the 2011 PR address where it existed. This additionally provided an indication of whether unmatched death records might be attributable to migration into England and Wales post-2011.
 The prior linking method used NHS Number only. Although the overall linkage rate was approximately 90%, the rate for recent deaths (2nd March 2020 to 10th April 2020 in the first iteration of the linkage) was closer to 30% due to an administrative lag in adding NHS Numbers to death records. Our novel bespoke linkage method linked over 39,000 extra death records. Whilst this had minimal impact on the overall linkage rate, it improved the linkage rate for recent deaths to approximately 90%. This was without an impact on accuracy: clerical review demonstrated that the false positive rate was approximately 0.2%. A report was published using this data showing that the risk of death involving Covid-19 among some ethnic groups was significantly higher than others.
 Determining whether Covid-19 disproportionally affected certain ethnicities was of crucial importance in the early phase of the pandemic to enable appropriate government strategies to be developed. We delivered a bespoke linkage under an exceptional time-limit without compromising on accuracy, enabling this impactful analysis with nation-wide interest and impact.

Do Conceptualisations of ‘Mixed Race’, ‘Interracial Unions’, and Race’s ‘Centrality to Understandings of Racism’ Challenge the UK’s Official Categorisation by Ethnic Group?

A focus on ‘mixed race’ and mixedness in Britain has revived a debate around the central question of whether the decennial census and other official data collections should be capturing ‘race’ rather than ethnic group and producing ‘racial’ outputs. The British practice may seem out of step by some commentators, given that ‘mixed race’ is the term of choice amongst those it describes, and given scholarly interest in interracial unions. Moreover, the resurgence of interest in ‘race’ and racisms in the context of the ‘Black Lives Matter’ movement and concern over the down-playing in a UK Government-commissioned report of the role of structural racism has enlivened the debate. However, this paper argues against a shift to ‘race’ in official data collection and for continued use of the conceptually preferable ‘ethnic group’ in the census question title, the section label ‘mixed/multiple ethnic groups’, and the ongoing provision of data on unions at the pan-ethnic and granular levels. A measure of socially constructed ‘race’ is already available in all but name in the pan-ethnic section labels (White, Asian, Black, Mixed, and Other) and the tick boxes under the ‘mixed/multiple’ heading. Ethnic group has been the conceptual basis of the question since the field trials for the 1991 Census, and its position has been strengthened by the increasing granularity of the categorisation (19 categories in the 2021 England and Wales Census) and by substantial distributed free-text provision that underpins the question’s context of self-identification. The wider understanding of ‘race’ identifications invokes ascription, imposition, and social categorisation rather than self-identification and subscription. There is also evidence of the unacceptability of ‘race’ in the context of the census amongst the wider society.

Factors associated with round-trip carsharing frequency and driving-mileage impacts in London

This study draws on respondents from Greater London within the 2016/17 wave (n = 2640) of Britain’s Annual Survey of carsharing users, which we enrich with external data from the 2011 England and Wales Census and small-area income estimates. Focusing on round-trip carsharing users, we present multivariate analyses of frequency-of-usage of carsharing vehicles and impacts on annual vehicle miles traveled (VMT). Published attribute effects from other geographic and contextual circumstances are compiled and compared (where direct comparison is possible) with the specific attribute effects that we report in this paper. We demonstrate a statistically significant link between customer satisfaction with proximity of carsharing vehicles and VMT impacts. Car ownership (both current, and changes upon joining a carsharing service) is shown to have intuitive structural impacts. We find that frequent usage is associated ceteris paribus with increasing VMT after joining a carsharing service, and that subscribing to multiple types of carsharing is associated with frequent carsharing usage and a reduction in VMT. Interestingly, we did not find any significant effect of household income on either frequency-of-usage or VMT impacts.

35.2 PREVENTING PSYCHOSIS: WHAT, (IF ANYTHING) CAN WE LEARN FROM THE EU-GEI INCIDENCE STUDY?

BackgroundThe incidence of psychotic disorders varies across replicable social and environmental gradients at both an individual and a population level, such as higher rates of disorder in urban and migrant populations. However, the factors underpinning this are unclear. The EU-GEI study was established to investigate the incidence as well as the genetic and environmental determinants of first episode psychosis in a multi-national setting. The aim of the present study was to investigate the variance found in the incidence across the 17 catchment areas in the 6 countries (England, France, Italy, the Netherlands, Spain and Brazil) included in this study at both individual and population level, and identify putative predictors of psychosis risk.MethodsWe conducted a population-based study of the incidence of non-organic adult ICD-10 psychotic disorders (F20-F33). Demographic data (age, sex, ethnicity) and OPCRIT diagnoses were collected, and denominator data was estimated from government sources. Crude incidence rates were directly standardised to the 2011 England and Wales Census population to account for population differences in age, sex and ethnicity. Multilevel Poisson regression was carried out to investigate variance in incidence between catchment areas by latitude, population density, and percentage of unemployment, owner-occupied houses and single-person households as markers of catchment-area level social fragmentation, using official government statistics and data from the 2011 European Population and Housing Census.ResultsWe identified a total of 2,774 cases over 12.94 million person-years at risk, leading to a crude incidence of 21.4 per 100,000 person-years (95%CI: 19.4–23.4). The age pattern of incidence differed between men and women: crude incidence peaked in men aged 18–24 (61.0 per 100,000 person-years, 95%CI: 59.0–63.1) and declined sharply thereafter, for women rates also peaked in the youngest age group (27.0 per 100,000 person-years, 95%CI: 24.9–29.1) but decline was more gradual and there was a small but robust secondary peak after age 45. By age 35, 68% of male cases had presented to services, compared to 51% of female cases. Age-sex-ethnicity standardised incidence of all psychotic disorders varied 8-fold across settings. Poisson regression revealed higher rates in minority groups (IRR: 1.6, 95%CI: 1.5–1.7), and an association between greater catchment area-level owner-occupancy and lower incidence (IRR for a 10% increase: 0.8, 95%CI: 0.7–0.8). No relationship was found for other putative environmental risk factors, including latitude and population density. Results were similar for non-affective and affective disorders.DiscussionVariance in treated incidence was substantial and was only partially explained by standardisation for age, sex and ethnicity, and Poisson regression including catchment-area level risk factors. For the prevention of psychosis two main lessons can be learned: services focused on early intervention should not have an upper age limit as half of all female (and 32% of male) cases present after age 35, and future examinations of variance should focus on socioenvironmental and not geographical determinants.

Generational change and how we conceptualize and measure multiracial people and “mixture”

ABSTRACTUntil relatively recently, in countries such as the U.S.A. and U.K., individuals could only opt for “single race” categories with which they identified. However, in the 2000 decennial census, respondents in the U.S. were able to choose more than one racial category, while in 2001, a “Mixed” box (with further subcategories) was provided in the England and Wales census for the first time. But the very success of this racial project in these countries has spawned a number of questions for policy-makers and academics who theorize, enumerate and study the experiences of multiracial people. With demographic changes such as generational change, who counts as multiracial or mixed race? This question has yet to receive significant attention. Although mixing is becoming more commonplace, the question of who counts as multiracial is far from straightforward, especially as we look down the generational pipeline.

Explaining geographic patterns of small group internal migration

AbstractInternal migration plays a key role in shaping the demographic characteristics of areas. In this paper, data from the 2011 England and Wales census are used to assess the geographic patterns of migration for 4 small cultural groups that each constitute about 0.5% of the population—Arabs, Chinese, Jews, and Sikhs—with a White British “benchmark” group. It examines the sensitivity of the scale of intercommunity moves to distance, having controlled for other migrant characteristics, through the development of spatial interaction models. The analysis finds that, where a choice exists, Jews are more averse to making a longer move than other small groups, all of whom favour shorter moves than the White British. The paper also investigates the influence of origin location and socioeconomic characteristics on the choice of migration destination using multinomial logistic regression. It finds that the influence of student status, age, qualifications, and home tenure vary by group though a number of patterns are shared between groups. Finally, it probes the presence in these smaller groups of patterns found historically in the wider population, such as counter‐urbanisation. Overall, this paper broadens the understanding of minority group migration patterns by examining, for the first time, Arabs (identified separately only in the 2011 census) and 2 groups based on religion (Jews and Sikhs) and by revisiting, with new questions, the White British and Chinese groups using the latest census data.

The Garden Palace: Building an Early Sydney Icon

The Garden Palace: Building an Early Sydney Icon

Bringing the Prisoner into View: English and Welsh Census Data and the Victorian Prison Population

The 1881 England and Wales census is now available as a fully-searchable electronic file. What can this source of big data tell us about the shape of the late-Victorian prison population? This article explores how the 1881 census compares to other sources of prison population data, and examines how the census can be analysed to reveal the demographic structure of the prison population in England and Wales. We argue that the census enables us to bring breadth of information about incarcerated individuals, identifying the ‘typical’ and ‘atypical’ prisoner, to a depth of information already available through small-scale qualitative research. In doing so, census data provides both quantitative and qualitative data essential for bringing context and scale to social and personal histories of crime.

Understanding recent internal migration of small cultural groups in England and Wales

ABSTRACTTaking advantage of the availability of 2011 England and Wales census microdata, and recognising the importance of internal migration in shaping the size and nature of communities, this paper seeks to identify and quantify the underlying determinants of internal migration of small cultural groups. The Jewish group is one of the longest present minority groups in Britain. Three other groups (Arab, Chinese, and Sikh), which have been present in significant numbers for a much shorter period, are also examined. Multivariate binary logistic regression has been applied to data extracted from the 2011 safeguarded microdata files, to understand whether, having controlled for the variables identified, there remain residual unexplained differences between Jewish, other smaller group, and general migration levels. The study shows that the initial wide variation in migration propensity between these cultural groups is partly explained by compositional differences between groups, but that even after controlling for individual-level socio-demographic characteristics, regional location and distance of migration, cultural differences in migration behaviour remain. Overall, the study shows that there are fewer differences between Jewish and white British migration levels than for the other three groups, for whom a small but significant ‘cultural group penalty’, inhibiting migration propensity, remains.

Employers and the 1881 Population Census of England and Wales

This article presents the method and first results of using the 1881 England and Wales Census Enumerators' Books (CEBs) to identify and extract employer records using occupational information. Over 230,000 employers are identified, of which about four fifths employ others. Important sub-groups are also identified of the own account selfemployed, company proprietors, directors and partnerships. The article demonstrates the feasibility of the method and uses the example of the building industry to illustrate firm-size distribution at parish level across England and Wales. The paper indicates the applicability of the extraction method to other censuses, which is now possible using the recently released I-CeM database. The paper also demonstrates some difficulties in the database for 1881, including data keying and coding errors, ranging from 0.5 to 5.5 per cent of entries for larger businesses. Gender miscoding appears to be a systematic error of about 0.7 per 1,000 people. The analysis suggests that where small or atypical sample groups are involved, users of the census database should make detailed checks with manuscript CEBs.

What are the parenting practices of multiracial people in Britain?

ABSTRACTSince ‘mixed’ was first offered as an option in the ethnicity question in the 2001 England and Wales Census, Britain's recognition of, and interest in, mixed (or ‘multiracial’) people and families has not abated. Recent studies have focused primarily upon how mixed people identify themselves, or how parents racially identify their multiracial children. But Britain now has a population of multiracial individuals who are themselves parents, about whom we know very little. What are the particular concerns for multiracial individuals who are parents? Do multiracial people (who are parents) want to steer their children toward a particular kind of upbringing, and if so, toward what (and why)? This paper is an in-depth exploration of the various ways in which different types of multiracial people in Britain raise their children. Our main finding is that a significant majority of multiracial parents engage in parenting practices that emphasize a cosmopolitan ethos.

Management of adult spontaneous spondylodiscitis and its rising incidence.

Spondylodiscitis refers to the infection of the intervertebral disc and osteomyelitis of the adjacent endplates, and it is uncommon in the developed world. Broad consensus indicates its incidence is on the rise. The aim of this retrospective study was twofold. First, we sought to give an up-to-date incidence estimate by comparing case presentation over two time periods (1995-1999 and 2008-2011). Data from the England and Wales census in 2001 and 2011 were used for incidence estimation. The second part of this study aimed to generate management guidance from data from medical and radiographic records of the 2008-2011 patient cohort. The incidence of adult spontaneous spondylodiscitis in the local region between 2008 and 2011 was 3.67/100,000 per year, representing a 150% increase from the incidence in 1995-1999. Our data demonstrate that methicillin sensitive Staphylococcus aureus remains the most common offending pathogen of spontaneous spondylodiscitis. The mean C-reactive protein (CRP) level remained at >30mg/l after a month of starting antibiotic treatment in both medically and surgically managed groups. Evidence suggests that the incidence of spondylodiscitis is on the rise. A review of our case series has demonstrated the effectiveness of intravenous antibiotic therapy. While no official guidance exists for when to switch from intravenous to oral antibiotics, our study shows that CRP at 1 month is >30mg/l and we recommend 6 weeks of intravenous therapy, followed by 6 further weeks of oral therapy.