Utilising identifier error variation in linkage of large administrative data sources

Background

Linkage of administrative data sources often relies on probabilistic methods using a set of common identifiers (e.g. sex, date of birth, postcode). Variation in data quality on an individual or organisational level (e.g. by hospital) can result in clustering of identifier errors, violating the assumption of independence between identifiers required for traditional probabilistic match weight estimation. This potentially introduces selection bias to the resulting linked dataset. We aimed to measure variation in identifier error rates in a large English administrative data source (Hospital Episode Statistics; HES) and to incorporate this information into match weight calculation.

Methods

We used 30,000 randomly selected HES hospital admissions records of patients aged 0–1, 5–6 and 18–19 years, for 2011/2012, linked via NHS number with data from the Personal Demographic Service (PDS; our gold-standard). We calculated identifier error rates for sex, date of birth and postcode and used multi-level logistic regression to investigate associations with individual-level attributes (age, ethnicity, and gender) and organisational variation. We then derived: i) weights incorporating dependence between identifiers; ii) attribute-specific weights (varying by age, ethnicity and gender); and iii) organisation-specific weights (by hospital). Results were compared with traditional match weights using a simulation study.

Results

Identifier errors (where values disagreed in linked HES-PDS records) or missing values were found in 0.11% of records for sex and date of birth and in 53% of records for postcode. Identifier error rates differed significantly by age, ethnicity and sex (p < 0.0005). Errors were less frequent in males, in 5–6 year olds and 18–19 year olds compared with infants, and were lowest for the Asian ethic group. A simulation study demonstrated that substantial bias was introduced into estimated readmission rates in the presence of identifier errors. Attribute- and organisational-specific weights reduced this bias compared with weights estimated using traditional probabilistic matching algorithms.

Conclusions

We provide empirical evidence on variation in rates of identifier error in a widely-used administrative data source and propose a new method for deriving match weights that incorporates additional data attributes. Our results demonstrate that incorporating information on variation by individual-level characteristics can help to reduce bias due to linkage error.

Estimating the burden of respiratory syncytial virus (RSV) on respiratory hospital admissions in children less than five years of age in England, 2007-2012

Background

Respiratory syncytial virus (RSV) is a leading cause of hospital admission in young children. With several RSV vaccines candidates undergoing clinical trials, recent estimates of RSV burden are required to provide a baseline for vaccine impact studies.

Objectives

To estimate the number of RSV‐associated hospital admissions in children aged <5 years in England over a 5‐year period from 2007 using ecological time series modelling of national hospital administrative data.

Patients/Methods

Multiple linear regression modelling of weekly time series of laboratory surveillance data and Hospital Episode Statistics (HES) data was used to estimate the number of hospital admissions due to major respiratory pathogens including RSV in children <5 years of age in England from mid‐2007 to mid‐2012, stratified by age group (<6 months, 6‐11 months, 1‐4 years) and primary diagnosis: bronchiolitis, pneumonia, unspecified lower respiratory tract infection (LRTI), bronchitis and upper respiratory tract infection (URTI).

Results

On average, 33 561 (95% confidence interval 30 429‐38 489) RSV‐associated hospital admissions in children <5 years of age occurred annually from 2007 to 2012. Average annual admission rates were 35.1 (95% CI: 32.9‐38.9) per 1000 children aged <1 year and 5.31 (95% CI: 4.5‐6.6) per 1000 children aged 1‐4 years. About 84% (95% CI: 81‐91%) of RSV‐associated admissions were for LRTI. The diagnosis‐specific burden of RSV‐associated admissions differed significantly by age group.

Conclusions

RSV remains a significant cause of hospital admissions in young children in England. Individual‐level analysis of RSV‐associated admissions is required to fully describe the burden by age and risk group and identify optimal prevention strategies.

Factors associated with re-entry to out-of-home care among children in England

Exiting and re-entering out-of-home care (OHC) is considered a disruption to permanence which may have long-lasting, negative consequences for children due to a lack of stability and continuity. Each year approximately one-third of children in OHC in England exit, but information is lacking on rates of re-entries and associated factors. Using national administrative data, we calculated rates of re-entry among children exiting OHC from 2007 to 2012, identified key child and care factors associated with re-entry using Cox proportional hazards modelling, and developed a simple probability calculator to estimate which groups of children are most likely to re-enter OHC within three months. Between 2007 and 2012 re-entries to OHC in England decreased (from 23.3% to 14.4% within one year of exit, p<0.001), possibly due to concurrent changes in the way children exited OHC. Overall, more than one-third of children exiting OHC in 2008 re-entered within five years (35.3%, N=4076), but rates of re-entry varied by child and care characteristics including age, ethnicity, mode of exit, and placement stability. Based on these associated factors, we developed a calculator that can estimate the likelihood of rapid re-entry to OHC for a group of children and could be used by social care practitioners or service planners. Our findings provide insight into which groups of children are most likely to re-enter OHC, who may benefit from additional support or ongoing monitoring.

Linking Data for Mothers and Babies in De-Identified Electronic Health Data

OBJECTIVE

Linkage of longitudinal administrative data for mothers and babies supports research and service evaluation in several populations around the world. We established a linked mother-baby cohort using pseudonymised, population-level data for England.

DESIGN AND SETTING

Retrospective linkage study using electronic hospital records of mothers and babies admitted to NHS hospitals in England, captured in Hospital Episode Statistics between April 2001 and March 2013.

RESULTS

Of 672,955 baby records in 2012/13, 280,470 (42%) linked deterministically to a maternal record using hospital, GP practice, maternal age, birthweight, gestation, birth order and sex. A further 380,164 (56%) records linked using probabilistic methods incorporating additional variables that could differ between mother/baby records (admission dates, ethnicity, 3/4-character postcode district) or that include missing values (delivery variables). The false-match rate was estimated at 0.15% using synthetic data. Data quality improved over time: for 2001/02, 91% of baby records were linked (holding the estimated false-match rate at 0.15%). The linked cohort was representative of national distributions of gender, gestation, birth weight and maternal age, and captured approximately 97% of births in England.

CONCLUSION

Probabilistic linkage of maternal and baby healthcare characteristics offers an efficient way to enrich maternity data, improve data quality, and create longitudinal cohorts for research and service evaluation. This approach could be extended to linkage of other datasets that have non-disclosive characteristics in common.

Integrating primary and secondary care for children and young people: sharing practice

OBJECTIVE

To share innovative practice with enough detail to be useful for paediatricians involved in planning services.

DESIGN

A review of practice, adopting a realist approach.

SETTING

We collected detailed information about five initiatives which were presented at two meetings in July and October 2014 and telephone interviews between July and November 2014 with key informants, updating information again in February 2015.

RESULTS

The five case studies involved three clinical commissioning groups (CCGs): Islington CCG and Southwark and Lambeth CCG in London and Taunton CCG in the Southwest. All five initiatives involved acute paediatric units. We heard about four distinct types of services designed to bring paediatric expertise into primary care and/or improve joint working between paediatricians and primary care professionals: telephone multidisciplinary team, hospital at home, general practitioner (GP) outreach clinics, and advice and guidance. We defined four common ways that initiatives might work: promoting shared responsibility; upskilling GPs; establishing relationships between paediatricians and primary healthcare professionals; and by taking specialist care to the patient.

CONCLUSIONS

We derived common aims and mechanisms and generated programme (mid-level) theory for each integrated care initiative about how they might work. These descriptions of what is being done can inform debate about which interventions should be prioritised for wider implementation. There is an urgent need for evaluation of these interventions and more indepth research into how mechanisms and their effectiveness could be assessed.

Entering out-of-home care during childhood: Cumulative incidence study in Canada and Australia

Cumulative incidence provides a more accurate indicator than annual incidence rates of the number of children who experience out-of-home care during childhood. The study utilises a cohort of all children born in Western Australia (WA) 1994-2005 and Manitoba 1998-2008 using de-identified linked data. Life tables were used to calculate the age-specific cumulative incidence over time and for at-risk groups. Cox regression was used to compare risk factors for entry to care. Manitoba had a larger proportion of children entering care compared to WA (9.4% vs 1.5% by age 12). Over time children entered care at a younger age in both WA (HR=1.5, CI:1.4-1.5) and Manitoba (HR=1.5, CI:1.5-1.6). Similar factors were associated with earlier age care entries in both countries including: socioeconomic disadvantage, young maternal age, maternal hospital admissions for mental health issues, substance misuse and assault. Supplementary analysis for WA showed a time trend with young children (<3years of age) who entered care spending an increasing proportion of their early years in care. Whilst Manitoba had a larger proportion of children entering care, over time in Western Australia children have been entering care at a younger age and spending more time in care. These latter factors contribute to an increased burden on the out-of-home care system. Manitoba had over five times greater cumulative incidence than WA, however risk factors for entry to out-of-home care were consistent in both countries. Knowledge of the risk factors for entry to out-of-home care can inform targeted support and prevention programs.

Who comes back with what: a retrospective database study on reasons for emergency readmission to hospital in children and young people in England

OBJECTIVE

To determine the proportion of children and young people (CYP) in England who are readmitted for the same condition.

DESIGN

Retrospective cohort study.

SETTING

National administrative hospital data (Hospital Episode Statistics).

PARTICIPANTS

CYP (0-year-olds to 24-year-olds) discharged after an emergency admission to the National Health Service in England in 2009/2010.

MAIN OUTCOME MEASURES

Coded primary diagnosis classified in six broad groups indicating reason for admission (infection, chronic condition, injury, perinatal related or pregnancy related, sign or symptom or other). We grouped readmissions as ≤30 days or between 31 days and 2 years after the index discharge. We used multivariable logistic regression to determine factors at the index admission that were predictive of readmission within 30 days.

RESULTS

9% of CYP were readmitted within 30 days. Half of the 30-day readmissions and 40% of the recurrent admissions between 30 days and 2 years had the same primary diagnosis group as the original admission. These proportions were consistent across age, sex and diagnostic groups, except for infants and young women with pregnancy-related problems (15-24 years) who were more likely to be readmitted for the same primary diagnostic group. CYP with underlying chronic conditions were readmitted within 30 days twice as often (OR: 1.93, 95% CI 1.89 to 1.99) compared with CYP without chronic conditions.

CONCLUSIONS

Financial penalties for readmission are expected to incentivise more effective care of the original problem, thereby avoiding readmission. Our findings, that half of children come back with different problems, do not support this presumption.

Factors associated with influenza vaccine uptake during a universal vaccination programme of preschool children in England and Wales: a cohort study

INTRODUCTION

Influenza vaccination through primary care has been recommended for all preschool children in the UK since 2013 as part of a universal immunisation programme. Vaccination is required annually and effectiveness varies by season. Factors associated with influenza vaccine receipt and those for other childhood vaccines may therefore differ.

METHODS

We used The Health Improvement Network, a large primary care database, to create a cohort of children in England and Wales aged 2-4 years eligible for vaccination in the 2014/2015 season. Mixed-effects Poisson regression models were used to determine sociodemographic and clinical factors associated with influenza vaccine receipt, allowing for practice-level variation.

RESULTS

Overall, 38.7% (95% CI 38.3% to 39.1%) of 57 545 children were vaccinated against influenza. Children in the poorest deprivation quintile were 19% less likely to receive influenza vaccine than those in the wealthiest quintile (adjusted risk ratio (ARR) 0.81, 95% CI 0.77 to 0.86). Children who received a timely first dose of measles-mumps-rubella vaccine were twice as likely to receive influenza vaccine (ARR 2.00 95% CI 1.87 to 2.13). Being 4 years old, not in a clinical risk group, or living with 2 or more other children were also significantly associated with a lower probability of vaccination.

DISCUSSION

Children living in areas of higher deprivation and in larger families are less likely to receive influenza vaccine. Further research is required into whether interventions, such as offering vaccinations in other settings, could increase uptake in children, particularly in deprived areas.

Generalisability and Cost-Impact of Antibiotic-Impregnated Central Venous Catheters for Reducing Risk of Bloodstream Infection in Paediatric Intensive Care Units in England

Background

We determined the generalisability and cost-impact of adopting antibiotic-impregnated CVCs in all paediatric intensive care units (PICUs) in England, based on results from a large randomised controlled trial (the CATCH trial; ISRCTN34884569).

Methods

BSI rates using standard CVCs were estimated through linkage of national PICU audit data (PICANet) with laboratory surveillance data. We estimated the number of BSI averted if PICUs switched from standard to antibiotic-impregnated CVCs by applying the CATCH trial rate-ratio (0.40; 95% CI 0.17,0.97) to the BSI rate using standard CVCs. The value of healthcare resources made available by averting one BSI as estimated from the trial economic analysis was £10,975; 95% CI -£2,801,£24,751.

Results

The BSI rate using standard CVCs was 4.58 (95% CI 4.42,4.74) per 1000 CVC-days in 2012. Applying the rate-ratio gave 232 BSI averted using antibiotic CVCs. The additional cost of purchasing antibiotic-impregnated compared with standard CVCs was £36 for each child, corresponding to additional costs of £317,916 for an estimated 8831 CVCs required in PICUs in 2012. Based on 2012 BSI rates, management of BSI in PICUs cost £2.5 million annually (95% uncertainty interval: -£160,986, £5,603,005). The additional cost of antibiotic CVCs would be less than the value of resources associated with managing BSI in PICUs with standard BSI rates >1.2 per 1000 CVC-days.

Conclusions

The cost of introducing antibiotic-impregnated CVCs is less than the cost associated with managing BSIs occurring with standard CVCs. The long-term benefits of preventing BSI could mean that antibiotic CVCs are cost-effective even in PICUs with extremely low BSI rates.

Prescribing of neuraminidase inhibitors for influenza in UK primary care since the 2009 pandemic

We determined prescribing rates of neuraminidase inhibitors (NIs) for influenza in UK primary care since 2009 in relation to national prescribing guidelines. All NI prescriptions issued during the influenza seasons between October 2010 and May 2013 were extracted from The Health Improvement Network (THIN), a large UK primary-care database. We calculated NI prescribing rates per 100,000 person-weeks (pw) by age group, sex, deprivation level, influenza season and presence of chronic conditions with 95% confidence intervals (CIs), and used negative binomial regression models to determine the independent association between these variables and NI prescribing. NI prescribing was rare. The prescribing rate was 1·7/100,000 pw (95% CI 1·7-1·8) during influenza-active periods, and 0·1/100,000 (95% CI 0·1-0·1) during non-active periods. Prescribing rates were highest in 25- to 44-year-olds in 2010/2011 and in persons aged ⩾85 years in 2011/2012 and 2012/2013. Individuals with chronic conditions had significantly higher prescribing rates than persons without (rate ratio 2·62, 95% CI 2·27-3·03). GPs are more likely to prescribe NIs to high-risk individuals and during influenza active periods, as per national guidelines. We could not assess the proportion of patients with influenza-like illness who were prescribed an NI.

Selective serotonin reuptake inhibitors and congenital heart anomalies: comparative cohort studies of women treated before and during pregnancy and their children

BACKGROUND

Large databases and population registers are increasingly used to examine adverse birth outcomes, congenital heart anomalies, in particular, following antidepressant exposures in pregnancy. Yet many studies have failed to account for other characteristics of the women who were prescribed antidepressants.

OBJECTIVE

To examine the characteristics of women who are prescribed selective serotonin reuptake inhibitors (SSRIs) in pregnancy and women who are not, associations between SSRIs prescribed in pregnancy and congenital heart anomalies, and the association between social and lifestyle characteristics of pregnant women and congenital heart anomalies.

METHOD

Using data from The Health Improvement Network primary care database in the United Kingdom between January 1, 1990, and January 31, 2011, we set up a comparative study including 4 cohorts of children of women with and without different antidepressant exposures before and during pregnancy. 5,154 women were receiving SSRIs before pregnancy, 2,776 were receiving SSRIs during pregnancy, 992 were receiving other antidepressants during pregnancy, and 200,213 were receiving no antidepressants before or during pregnancy. Our primary outcome was congenital heart anomalies.

RESULTS

Less than 1% of children had a record of congenital heart anomalies within 5 years of birth, and there were no significant differences related to antidepressant exposure in pregnancy (women not prescribed antidepressants versus women prescribed SSRIs in first trimester: odds ratio [OR] = 1.00; 95% CI, 0.65-1.52); however, independent of antidepressant prescribing, diabetes (OR = 2.23; 95% CI, 1.79-2.77), increasing age (OR = 1.01; 95% CI, 1.00-1.02), alcohol problem (OR = 2.58; 95% CI, 1.55-4.29, illicit drug problems (OR = 1.89; 95% CI, 1.09-3.25), and obesity (OR = 1.38; 95% CI, 1.13-1.69) were associated with an increased risk of having a child with congenital heart anomalies.

CONCLUSIONS

There was no difference in congenital heart anomalies in children born to women with different antidepressant prescribing exposure status. However, we confirmed an increased risk of congenital heart anomalies in children of older women and in children of women with diabetes, a body mass index above 30 kg/m(2), and a history of alcohol and illicit drug problems independent of the prescription of antidepressants. Future research in this field must account for these characteristics. On the basis of existing evidence, advising women to stop antidepressant treatment in pregnancy may be counterproductive.

Neonatal drug withdrawal syndrome: cross-country comparison using hospital administrative data in England, the USA, Western Australia and Ontario, Canada

OBJECTIVES

We determined trends over time in the prevalence of neonatal drug withdrawal syndrome (NWS) in England compared with that reported in the USA, Western (W) Australia and Ontario, Canada. We also examined variation in prevalence of NWS according to maternal age, birth weight and across the English NHS by hospital trusts.

DESIGN AND SETTING

Retrospective study using national hospital administrative data (Hospital Episode Statistics) for the NHS in England between 1997 and 2011. NWS was identified using international classification of disease codes in hospital admission records. We searched the research literature and contacted researchers to identify studies reporting trends in the prevalence of NWS.

MAIN OUTCOME MEASURES

Prevalence of NWS by calendar year per 1000 live births for each country/state. For births in England, prevalence by maternal age group and birth weight group. Prevalence by NHS trust and region at birth, and funnel plot to show outlying English NHS hospital trusts (>3 SD of mean prevalence).

MAIN RESULTS

Mean prevalence rates of recorded NWS increased in all four countries. Rates stabilised in England and W. Australia from the early 2000s and rose steeply in the USA and Ontario during the late 2000s. The most recent prevalence rates were 2.7/1000 live births in England (2011; 1544 cases); 2.7/1000 in W. Australia (2009); 3.6/1000 in the USA (2009) and 5.1/1000 in Ontario (2011). The highest prevalence in England was among babies born to mothers aged 25-34 years at delivery and among babies born with low birth weight (1500-2500 g). In England in 2011, 8.6% of hospital trusts had a recorded prevalence outside 3 SD of the overall average (7% above, 1% below). The North East region of England had the highest recorded prevalence of NWS.

CONCLUSIONS

Although recorded NWS is stable in England and W. Australia, rising rates in the USA and Ontario may reflect better recognition and/or increased use of prescribed opiate analgesics and highlight the need for surveillance. The extent to which different prevalence rates by hospital trust reflect variation in occurrence, recognition or recording requires further investigation.

Changes in first entry to out-of-home care from 1992 to 2012 among children in England

Placement in out-of-home care (OHC) indicates serious childhood adversity and is associated with multiple adverse outcomes. Each year 0.5% of children in England live in OHC but evidence is lacking on the cumulative proportion who enter during childhood and how this varies over time. We measured the proportion of children born between 1992 and 2011 who entered OHC, including variation in rates of entry over time, and explored the determinants of these changes using decomposition methods. We also described changes in placement type, duration and stability. By age 18, 3.3% of children born 1992-94 entered OHC. This proportion varied by ethnicity (1.6% of White vs. 4.5% of Black children born 2001-03 entered OHC by age 9, 95% CI [1.5-1.7] and [4.4-4.6], p<0.001) and increased over time (0.8% of children born 2009-11 entered OHC by age 1 vs. 0.5% born 1992-94, 95% CI [0.7-0.9] and [0.4-0.6], p<0.001). This overall increase was driven primarily by the increased rate of entry among White children and not by concurrent changes in the population's ethnic composition. The proportion of children entering OHC in England is increasing and characteristics of the care they receive are changing with earlier intervention and longer, more stable placements. Further research is required to understand the reasons for these changes in practice and whether they are cost-effective, sustainable, and improve outcomes for children and society.

Clinicopathological characteristics and cell cycle proteins as potential prognostic factors in myoepithelial carcinoma of salivary glands

Myoepithelial carcinoma (MCA) is a rare malignancy of salivary glands that was included in the WHO Classification of Head and Neck Tumors in 1991. MCA has shown a broad spectrum of clinical outcomes, but attempts to identify prognostic markers for this malignancy have not resulted in significant progress. Conventional histopathological characteristics such as tumour grade, nuclear atypia, mitotic index and cell proliferation have failed to predict the outcome of MCA. In this study, we reviewed the histopathology of 19 cases of MCA focusing on nuclear atypia, mitotic count, tumour necrosis, nerve and vascular invasion and occurrence of a pre-existing pleomorphic adenoma in connection to the MCA. Histopathological characteristics and clinical information were correlated with the immunohistochemical expression of cell cycle proteins including c-Myc, p21, Cdk4 and Cyclin D3. The proportion of tumour cells immunoreactive for these markers and their intensity of staining were correlated with clinical information using logistic regression, Kaplan-Meier and Cox regression. Using logistic regression analysis, cytoplasmic c-Myc expression was associated with the occurrence of metastases (P = 0.019), but limitations of semi-quantitation of immunostaining and the limited number of cases preclude definitive conclusions. Our data show that the occurrence of tumour necrosis predicts poor disease-free survival in MCA (P = 0.035).

How can health systems be strengthened to control and prevent an Ebola outbreak? A narrative review

The emergence and re-emergence of infectious diseases are now more than ever considered threats to public health systems. There have been over 20 outbreaks of Ebola in the past 40 years. Only recently, the World Health Organization has declared a public health emergency of international concern (PHEIC) in West Africa, with a projected estimate of 1.2 million deaths expected in the next 6 months. Ebola virus is a highly virulent pathogen, often fatal in humans and non-human primates. Ebola is now a great priority for global health security and often becomes fatal if left untreated. This study employed a narrative review. Three major databases - MEDLINE, EMBASE, and Global Health - were searched using both 'text-words' and 'thesaurus terms'. Evidence shows that low- and middle-income countries (LMICs) are not coping well with the current challenges of Ebola, not only because they have poor and fragile systems but also because there are poor infectious disease surveillance and response systems in place. The identification of potential cases is problematic, particularly in the aspects of contact tracing, infection control, and prevention, prior to the diagnosis of the case. This review therefore aims to examine whether LMICs' health systems would be able to control and manage Ebola in future and identifies two key elements of health systems strengthening that are needed to ensure the robustness of the health system to respond effectively.

Deferred Consent for Randomized Controlled Trials in Emergency Care Settings

BACKGROUND

There is limited experience in using deferred consent for studies involving children, which was legalized in the United Kingdom in 2008. We aimed to inform future studies by evaluating consent rates and reasons for nonconsent in a large randomized controlled trial in pediatric intensive care.

METHODS

In the CATCH trial, eligible children from 14 PICUs in England and Wales were randomly assigned to 3 types of central venous catheters. To avoid delay in treatment, children admitted on an emergency basis were first randomly assigned to a trial central venous catheter, and we deferred seeking consent to use already collected data and blood samples until after stabilization.

RESULTS

Consent was obtained for 984/1358 (72%) of children admitted on an emergency basis. Failure to obtain consent resulted mainly from a lack of opportunity (early discharge or transfer) for survivors and difficulties in seeking consent for children who died. For admissions where there was an opportunity to approach (n = 1298), inclusion rates differed according to survival status: 93/984 (9%) of consented patients died, compared with 58/314 (18%) of nonconsented patients. For children admitted on an emergency basis whose families were approached, 984/1178 (84%) provided deferred consent (n = 15 sites), compared with 441/641 (69%) of children admitted on an elective basis who were approached for prospective consent (n = 9 sites).

CONCLUSIONS

Design of emergency randomized controlled trials should balance the potential burden that seeking consent in difficult situations may cause against risk of bias by disproportionately excluding children who die or are transferred. Ethics committees could consider approving the use of already collected data when best efforts to obtain deferred consent are unsuccessful.

Recording of Influenza-Like Illness in UK Primary Care 1995-2013: Cohort Study

BACKGROUND

There is a lack of recent studies examining recording of influenza-like illness (ILI) in primary care in the UK over time and according to population characteristics. Our aim was to determine time trends and socio-demographic patterns of ILI recorded consultations in primary care.

METHODS

We used The Health Improvement Network (THIN) UK primary care database and extracted data on all ILI consultations between 1995 and 2013. We estimated ILI recorded consultation rates per 100,000 person-weeks (pw) by age, gender, deprivation and winter season. Negative binomial regression models were used to examine time trends and the effect of socio-demographic characteristics. Trends in ILI recorded consultations were compared to trends in consultations with less specific symptoms (cough or fever) recorded.

RESULTS

The study involved 7,682,908 individuals in 542 general practices. The ILI consultation rate decreased from 32.5/100,000 pw (95% confidence interval (CI) 32.1, 32.9) in 1995-98 to 15.5/100,000 pw (95% CI 15.4, 15.7) by 2010-13. The decrease occurred prior to 2002/3, and rates have remained largely stable since then. Declines were evident in all age groups. In comparison, cough or fever consultation rates increased from 169.4/100,000 pw (95% CI 168.6, 170.3) in 1995-98 to 237.7/100,000 pw (95% CI 237.2, 238.2) in 2010-13. ILI consultation rates were highest among individuals aged 15-44 years, higher in women than men, and in individuals from deprived areas.

CONCLUSION

There is substantial variation in ILI recorded consultations over time and by population socio-demographic characteristics, most likely reflecting changing recording behaviour by GPs. These results highlight the difficulties in using coded information from electronic primary care records to measure the severity of influenza epidemics across time and assess the relative burden of ILI in different population subgroups.

How parents and practitioners experience research without prior consent (deferred consent) for emergency research involving children with life threatening conditions: a mixed method study

OBJECTIVE

Alternatives to prospective informed consent to enable children with life-threatening conditions to be entered into trials of emergency treatments are needed. Across Europe, a process called deferred consent has been developed as an alternative. Little is known about the views and experiences of those with first-hand experience of this controversial consent process. To inform how consent is sought for future paediatric critical care trials, we explored the views and experiences of parents and practitioners involved in the CATheter infections in CHildren (CATCH) trial, which allowed for deferred consent in certain circumstances.

DESIGN

Mixed method survey, interview and focus group study.

PARTICIPANTS

275 parents completed a questionnaire; 20 families participated in an interview (18 mothers, 5 fathers). 17 CATCH practitioners participated in one of four focus groups (10 nurses, 3 doctors and 4 clinical trial unit staff).

SETTING

12 UK children's hospitals.

RESULTS

Some parents were momentarily shocked or angered to discover that their child had or could have been entered into CATCH without their prior consent. Although these feelings resolved after the reasons why consent needed to be deferred were explained and that the CATCH interventions were already used in clinical care. Prior to seeking deferred consent for the first few times, CATCH practitioners were apprehensive, although their feelings abated with experience of talking to parents about CATCH. Parents reported that their decisions about their child's participation in the trial had been voluntary. However, mistiming the deferred consent discussion had caused distress for some. Practitioners and parents supported the use of deferred consent in CATCH and in future trials of interventions already used in clinical care.

CONCLUSIONS

Our study provides evidence to support the use of deferred consent in paediatric emergency medicine; it also indicates the crucial importance of practitioner communication and appropriate timing of deferred consent discussions.

Contribution of recurrent admissions in children and young people to emergency hospital admissions: retrospective cohort analysis of hospital episode statistics

OBJECTIVE

To examine the contribution of recurrent admissions to the high rate of emergency admissions among children and young people (CYP) in England, and to what extent readmissions are accounted for by patients with chronic conditions.

DESIGN

All hospital admissions to the National Health Service (NHS) in England using hospital episode statistics (HES) from 2009 to 2011 for CYP aged 0-24 years. We followed CYP for 2 years from discharge of their first emergency admission in 2009. We determined the number of subsequent emergency admissions, time to next admission, length of stay and the proportion of injury and chronic condition admissions measured by diagnostic codes in all following admissions.

RESULTS

869 895 children had an index emergency admission in 2009, resulting in a further 939 710 admissions (of which 600 322, or 64%, were emergency admissions) over the next 2 years. After discharge from the index admission, 32% of 274,986 (32%) children were readmitted within 2 years, 26% of these readmissions occurring within 30 days of discharge. Recurrent emergency admission accounted for 41% of all emergency admissions in the 2-year cohort and 66% of inpatient days. 41% of index admissions, but 76% of the recurrent emergency admissions, were in children with a chronic condition.

CONCLUSIONS

Recurrent admissions contribute substantially to total emergency admissions. They often occur soon after discharge, and disproportionately affect CYP with chronic conditions. Policies aiming to discourage readmissions should consider whether they could undermine necessary inpatient care for children with chronic conditions.

Data linkage errors in hospital administrative data when applying a pseudonymisation algorithm to paediatric intensive care records

OBJECTIVES

Our aim was to estimate the rate of data linkage error in Hospital Episode Statistics (HES) by testing the HESID pseudoanonymisation algorithm against a reference standard, in a national registry of paediatric intensive care records.

SETTING

The Paediatric Intensive Care Audit Network (PICANet) database, covering 33 paediatric intensive care units in England, Scotland and Wales.

PARTICIPANTS

Data from infants and young people aged 0-19 years admitted between 1 January 2004 and 21 February 2014.

PRIMARY AND SECONDARY OUTCOME MEASURES

PICANet admission records were classified as matches (records belonging to the same patient who had been readmitted) or non-matches (records belonging to different patients) after applying the HESID algorithm to PICANet records. False-match and missed-match rates were calculated by comparing results of the HESID algorithm with the reference standard PICANet ID. The effect of linkage errors on readmission rate was evaluated.

RESULTS

Of 166,406 admissions, 88,596 were true matches (where the same patient had been readmitted). The HESID pseudonymisation algorithm produced few false matches (n=176/77,810; 0.2%) but a larger proportion of missed matches (n=3609/88,596; 4.1%). The true readmission rate was underestimated by 3.8% due to linkage errors. Patients who were younger, male, from Asian/Black/Other ethnic groups (vs White) were more likely to experience a false match. Missed matches were more common for younger patients, for Asian/Black/Other ethnic groups (vs White) and for patients whose records had missing data.

CONCLUSIONS

The deterministic algorithm used to link all episodes of hospital care for the same patient in England has a high missed match rate which underestimates the true readmission rate and will produce biased analyses. To reduce linkage error, pseudoanonymisation algorithms need to be validated against good quality reference standards. Pseudonymisation of data 'at source' does not itself address errors in patient identifiers and the impact these errors have on data linkage.

Identifying Possible False Matches in Anonymized Hospital Administrative Data without Patient Identifiers

OBJECTIVE

To identify data linkage errors in the form of possible false matches, where two patients appear to share the same unique identification number.

DATA SOURCE

Hospital Episode Statistics (HES) in England, United Kingdom.

STUDY DESIGN

Data on births and re-admissions for infants (April 1, 2011 to March 31, 2012; age 0-1 year) and adolescents (April 1, 2004 to March 31, 2011; age 10-19 years).

DATA COLLECTION/EXTRACTION METHODS

Hospital records pseudo-anonymized using an algorithm designed to link multiple records belonging to the same person. Six implausible clinical scenarios were considered possible false matches: multiple births sharing HESID, re-admission after death, two birth episodes sharing HESID, simultaneous admission at different hospitals, infant episodes coded as deliveries, and adolescent episodes coded as births.

PRINCIPAL FINDINGS

Among 507,778 infants, possible false matches were relatively rare (n = 433, 0.1 percent). The most common scenario (simultaneous admission at two hospitals, n = 324) was more likely for infants with missing data, those born preterm, and for Asian infants. Among adolescents, this scenario (n = 320) was more common for males, younger patients, the Mixed ethnic group, and those re-admitted more frequently.

CONCLUSIONS

Researchers can identify clinically implausible scenarios and patients affected, at the data cleaning stage, to mitigate the impact of possible linkage errors.

Cumulative incidence of entry into out-of-home care: changes over time in Denmark and England

Policies and thresholds vary for placing children into out-of-home care (OHC) at different ages. Evidence is lacking that quantifies the risk of entering OHC by age, and how this varies over time and between countries. We determined the age-specific cumulative incidence of ever entering OHC during childhood in Denmark and in eight local authorities in England. We used administrative data for any form of OHC (except respite care) provided by children's social services in Denmark and England from 1992 to 2008. Using life tables and national population estimates, we calculated the cumulative incidence of entry into OHC by year of age for cohorts born in 1992-1994 through to 2006-2008. The cumulative incidence of entry into OHC decreased over time in Denmark and increased in England at all ages. Cumulative incidence of OHC in the first year of life was similar in Denmark and England for infants born in 1992-1994 (Denmark 2.83/1,000, England 2.89/1,000), but infants born in 2007-2008 were nearly three times as likely to enter OHC before their first birthday in England (4.50/1,000) than in Denmark (1.61/1,000). Entry into OHC during adolescence was more common in Denmark than in England so that by 16 years old the cumulative incidence of ever entering OHC during childhood was twice as high in Denmark (33.83/1,000) as in England (15.62/1,000). Diverging trends over time in the use of OHC in Denmark and England are likely to reflect changing policies in the two countries.

Association between antibiotic prescribing in pregnancy and cerebral palsy or epilepsy in children born at term: a cohort study using the health improvement network

BACKGROUND

Between 19%-44% pregnant women are prescribed antibiotics during pregnancy. A single, large randomised-controlled-trial (ORACLE Childhood Study II) found an increased risk of childhood cerebral palsy and possibly epilepsy following prophylactic antibiotic use in pregnant women with spontaneous preterm labour. We ascertained whether this outcome could be reproduced across the population of babies delivered at term and prospectively followed in primary-care using data from The Health Improvement Network.

METHODS

We determined the risk of cerebral palsy or epilepsy in children whose mothers were prescribed antibiotics during pregnancy using a cohort of 195,909 women linked to their live, term-born, singleton children. We compared the effect of antibiotic class, number of courses and timing of prescribing in pregnancy. Analyses were adjusted for maternal risk factors (e.g. recorded infection, age, chronic conditions, social deprivation, smoking status). Children were followed until age seven years or cessation of registration with the primary-care practitioner.

RESULTS

In total, 64,623 (33.0%) women were prescribed antibiotics in pregnancy and 1,170 (0.60%) children had records indicating cerebral palsy or epilepsy. Adjusted analyses showed no association between prescribing of any antibiotic and cerebral palsy or epilepsy (adj.HR 1.04, 95%CI 0.91-1.19). However, compared with penicillins, macrolides were associated with an increased risk of cerebral palsy or epilepsy (adj.HR 1.78, 95%CI 1.18-2.69; number needed to harm 153, 95%CI 71-671).

CONCLUSIONS

We found no overall association between antibiotic prescribing in pregnancy and cerebral palsy and/or epilepsy in childhood. However, our finding of an increased risk of cerebral palsy or epilepsy associated with macrolide prescribing in pregnancy adds to evidence that macrolide use is associated with serious harm.

Notifications for child safeguarding from an acute hospital in response to presentations to healthcare by parents

BACKGROUND

Consideration of child safeguarding is routine within maternity services but less common in other health services for adults. We audited notifications for child safeguarding from an acute general hospital where the policy includes questioning adults presenting with violence, mental health problems or drug or alcohol misuse to any department within the hospital about children at home and notifying to the local authority children's social care services if there are safeguarding concerns.

METHODS

Cross-sectional audit of notifications for child safeguarding, including abuse, neglect or victimization, from all departments in one hospital to the local authority children's social care department during 12 months (2010/11).

RESULTS

Of 681 notifications (57 per month), 40% (270/681) were triggered by parents' presentation to acute hospital services. Of these, 37% (100/270; 12 teenage mothers) presented for maternity care and 60% (162/270; 8 teenage parents) presented to the emergency department (ED). Of the 60% (411/681) of notifications prompted by children presenting for healthcare, most originated from the ED (358/411; 87%): two-thirds of these presented with injury (250/358; 70%).

CONCLUSION

Given a policy to ask adults about children at home, a substantial proportion of children notified for child safeguarding were recognized through presentations to acute healthcare by their parents. Further research and development of this policy needs to ensure that questioning results in effective interventions for the children and their parents.

Violence, self-harm and drug or alcohol misuse in adolescents admitted to hospitals in England for injury: a retrospective cohort study

OBJECTIVES

Of adolescents in the general population in England, we aimed to determine (1) the proportion that has an emergency admission to hospital for injury related to adversity (violence, self-harm or drug or alcohol misuse) and (2) the risk of recurrent emergency admissions for injury in adolescents admitted with adversity-related injury compared with those admitted with accident-related injury only.

DESIGN

We used longitudinally linked administrative hospital data (Hospital Episode Statistics) to identify participants aged 10-19 years with emergency admissions for injury (including day cases lasting more than 4 h) in England in 1998-2011. We used the Office for National Statistics mid-year estimates for population denominators.

RESULTS

Approximately 4.3% (n=141 248) of adolescents in the general population (n=3 254 046) had one or more emergency admissions for adversity-related injury (girls 4.6%, boys 4.1%), accounting for 50% of all emergency admissions for injury in girls and 29.1% in boys. Admissions for self-harm or drug or alcohol misuse commonly occurred in the same girls and boys. Recurrent emergency admissions for injury were more common in adolescents with adversity-related injury (girls 17.3%, boys 16.5%) than in those with accident-related injury only (girls 4.7%, boys 7.4%), particularly for adolescents with adversity-related injury related to multiple types of adversity (girls 21.1%, boys 24.2%).

CONCLUSIONS

Hospital-based interventions should be developed to reduce the risk of future injury in adolescents admitted for adversity-related injury.

Contribution of respiratory tract infections to child deaths: a data linkage study

Background

Respiratory tract infections (RTIs) are an important cause of death in children, and often contribute to the terminal decline in children with chronic conditions. RTIs are often underrecorded as the underlying cause of death; therefore the overall contribution of RTIs to child deaths and the potential preventability of RTI-related deaths have not been adequately quantified.

Methods

We analysed deaths in children resident in England who died of non-injury causes aged 28 days to 18 years between 2001 and 2010 using death certificates linked to a longitudinal hospital admission database. We defined deaths as RTI-related if RTIs or other respiratory conditions were recorded on death certificates or linked hospital records up to 30 days before death. We examined trends in mortality by age group, year and season (winter or summer) and determined the winter excess of RTI-related deaths using rate differencing techniques. We estimated the proportion of RTI-related deaths in children with chronic conditions.

Results

22.4% (5039/22509) of child deaths were RTI-related. RTI-related deaths declined by 2.3% per year in infants aged 28 to 364 days between 2001 and 2010. No decline was observed for older children. On average there were 161 winter excess RTI-related deaths annually, accounting for 32% of all RTI-related deaths. 89.0% of children with RTI-related deaths had at least one chronic condition; neurological conditions were the most prevalent.

Conclusions

RTI-related deaths have not declined in the last decade except in infants. Targeted strategies to prevent the winter excess of RTIs and to treat RTIs in children, particularly children with chronic conditions, may reduce RTI-related deaths.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2458-14-1191) contains supplementary material, which is available to authorized users.

Discontinuation of antipsychotic medication in pregnancy: a cohort study

BACKGROUND

Women prescribed antipsychotics face the dilemma on whether to continue medication in pregnancy in terms of balancing risks and benefits. Previous research on other psychotropic medications suggests that many women discontinue treatment in early pregnancy. However, very limited evidence exists on discontinuation of antipsychotic medication.

METHODS

We identified 495,953 pregnant women from THIN primary care database. Kaplan-Meier plots were used to examine time to last antipsychotic prescription. Poisson regression was used to examine characteristics of those who stopped treatment during pregnancy.

RESULTS

There has been an overall increase in prevalence of antipsychotic prescribing since 2007. However, antipsychotics were more likely to be stopped in pregnant than non-pregnant women. Only 107/279 (38%) of women on atypical antipsychotics and 39/207 (19%) of women on typical antipsychotics before pregnancy still received treatment at the start of third trimester. Older women were more likely to continue typical antipsychotic treatment in pregnancy (35+ versus <25 years risk ratio: 3.09 [95% CI 1.76, 5.44]). Likewise, those who received typical antipsychotics for longer periods before were most likely to continue treatment in pregnancy (12+ versus <6 months: RR: 3.12 [95% CI 1.97, 4.95]). For atypical antipsychotics length and dose of prior prescribing were also associated with continuation in pregnancy.

CONCLUSIONS

Pregnancy was a major determinant of cessation of antipsychotics. Only 38% of women on atypical and 19% on typical antipsychotics were still prescribed the drug in the third trimester. Duration of prior treatment, maternal age as well as dose was significantly associated with continued treatment of antipsychotics in pregnancy.

Estimating the prevalence of chronic conditions in children who die in England, Scotland and Wales: a data linkage cohort study

OBJECTIVES

To estimate the proportion of children who die with chronic conditions and examine time trends in childhood deaths involving chronic conditions.

DESIGN

Retrospective population-based death cohort study using linked death certificates and hospital discharge records.

SETTING

England, Scotland and Wales.

PARTICIPANTS

All resident children who died aged 1-18 years between 2001 and 2010.

PRIMARY AND SECONDARY OUTCOME MEASURES

The primary outcome was the proportion of children who died with chronic conditions according to age group and type of chronic condition. The secondary outcome was trends over time in mortality rates involving chronic conditions per 100,000 children and trends in the proportion of children who died with chronic conditions.

RESULTS

65.4% of 23,438 children (95% CI 64.8%, 66.0%) died with chronic conditions, using information from death certificates. This increased to 70.7% (95% CI 70.1% to 71.3%) if hospital records up to 1 year before death were also included and was highest (74.8-79.9% depending on age group) among children aged less than 15 years. Using data from death certificates only led to underascertainment of all types of chronic conditions apart from cancer/blood conditions. Neurological/sensory conditions were most common (present in 38.5%). The rate of children dying with a chronic condition has declined since 2001, whereas the proportion of deaths affected by chronic conditions remained stable.

CONCLUSIONS

The majority of children who died had a chronic condition. Neurological/sensory conditions were the most prevalent. Linkage between death certificate and hospital discharge data avoids some of the under-recording of non-cancer conditions on death certificates, and provides a low-cost, population-based method for monitoring chronic conditions in children who die.

Maltreatment or violence-related injury in children and adolescents admitted to the NHS: comparison of trends in England and Scotland between 2005 and 2011

OBJECTIVE

Legislation to safeguard children from maltreatment by carers or violence by others was advanced in England and Scotland around 2004-2005 and resulted in different policies and services. We examined whether subsequent trends in injury admissions to hospital related to maltreatment or violence varied between the two countries.

SETTING AND PARTICIPANTS

We analysed rates of all unplanned injury admission to National Health Service (NHS) hospitals in England and Scotland between 2005 and 2011 for children and adolescents aged less than 19 years.

OUTCOMES

We compared incidence trends for maltreatment or violence-related (MVR) injury and adjusted rate differences between 2005 and 2011 using Poisson or negative binomial regression models to adjust for seasonal effects and secular trends in non-MVR injury. Infants, children 1-10 years and adolescents 11-18 years were analysed separately.

RESULTS

In 2005, MVR rates were similar in England and Scotland for infants and 1-10-year-olds, but almost twice as high in Scotland for 11-18-year-olds. MVR rates for infants increased by similar amounts in both countries, in line with rising non-MVR rates in England but contrary to declines in Scotland. Among 1-10-year-olds, MVR rates increased in England and declined in Scotland, in line with increasing non-MVR rates in England and declining rates in Scotland. Among 11-18-year-olds, MVR rates declined more steeply in Scotland than in England along with declines in non-MVR trends.

CONCLUSIONS

Diverging trends in England and Scotland may reflect true changes in the occurrence of MVR injury or differences in the way services recognise and respond to these children, record such injuries or a combination of these factors. Further linkage of data from surveys and services for child maltreatment and violence could help distinguish the impact of policies.

Evaluating bias due to data linkage error in electronic healthcare records

Background

Linkage of electronic healthcare records is becoming increasingly important for research purposes. However, linkage error due to mis-recorded or missing identifiers can lead to biased results. We evaluated the impact of linkage error on estimated infection rates using two different methods for classifying links: highest-weight (HW) classification using probabilistic match weights and prior-informed imputation (PII) using match probabilities.

Methods

A gold-standard dataset was created through deterministic linkage of unique identifiers in admission data from two hospitals and infection data recorded at the hospital laboratories (original data). Unique identifiers were then removed and data were re-linked by date of birth, sex and Soundex using two classification methods: i) HW classification - accepting the candidate record with the highest weight exceeding a threshold and ii) PII–imputing values from a match probability distribution. To evaluate methods for linking data with different error rates, non-random error and different match rates, we generated simulation data. Each set of simulated files was linked using both classification methods. Infection rates in the linked data were compared with those in the gold-standard data.

Results

In the original gold-standard data, 1496/20924 admissions linked to an infection. In the linked original data, PII provided least biased results: 1481 and 1457 infections (upper/lower thresholds) compared with 1316 and 1287 (HW upper/lower thresholds). In the simulated data, substantial bias (up to 112%) was introduced when linkage error varied by hospital. Bias was also greater when the match rate was low or the identifier error rate was high and in these cases, PII performed better than HW classification at reducing bias due to false-matches.

Conclusions

This study highlights the importance of evaluating the potential impact of linkage error on results. PII can help incorporate linkage uncertainty into analysis and reduce bias due to linkage error, without requiring identifiers.