Birth incidence, deaths and hospitalisations of children and young people with Down syndrome, 1990-2015: birth cohort study.

OBJECTIVE
To investigate current Down syndrome live birth and death rates, and childhood hospitalisations, compared with peers.


SETTING
General community.


PARTICIPANTS
All live births with Down syndrome, 1990-2015, identified via Scottish regional cytogenetic laboratories, each age-sex-neighbourhood deprivation matched with five non-Down syndrome controls. Record linkage to Scotland's hospital admissions and death data.


PRIMARY OUTCOME
HRs comparing risk of first hospitalisation (any and emergency), readmission for children with Down syndrome and matched controls were calculated using stratified Cox proportional hazards (PH) model, and length of hospital stay was calculated using a conditional log-linear regression model.


RESULTS
689/1479 (46.6%) female and 769/1479 (51.9%) male children/young people with Down syndrome were identified (1.0/1000 births, with no reduction over time); 1235 were matched. 92/1235 (7.4%) died during the period, 18.5 times more than controls. More of the Down syndrome group had at least one admission (incidence rate ratio(IRR) 72.89 (68.72-77.32) vs 40.51 (39.15-41.92); adjusted HR=1.84 (1.68, 2.01)) and readmissions (IRR 54.85 (51.46-58.46) vs 15.06 (14.36-15.80); adjusted HR=2.56 (2.08, 3.14)). More of their admissions were emergencies (IRR 56.78 (53.13-60.72) vs 28.88 (27.73-30.07); first emergency admission adjusted HR=2.87 (2.61, 3.15)). Children with Down syndrome had 28% longer first admission after birth. Admission rate increased from 1990-2003 to 2004-2014 for the Down syndrome group (from 90.7% to 92.2%) and decreased for controls (from 63.3% to 44.8%).


CONCLUSIONS
We provide contemporaneous statistics on the live birth rate of babies with Down syndrome, and their childhood death rate. They require more hospital admissions, readmissions emergency admissions and longer lengths of stays than their peers, which has received scant research attention in the past. This demonstrates the importance of statutory planning as well as informal support to families to avoid added problems in child development and family bonding over and above that brought by the intellectual disabilities associated with Down syndrome.


Strengths and limitations of this study
► A whole country population-based study on Down syndrome, with comprehensive identification via all NHS Scottish Regional Genetics Centres. ► 25-Year period of linked data. ► Account taken of the potential confounders of age, sex and extent of neighbourhood deprivation; underlying conditions were not investigated. ► Some of the Down syndrome individuals could not be matched, mostly due to missing data on postcode, and so were omitted from analyses. ► The use of routinely collected hospital and death statistics has many advantages including largescale coverage, but there is likely to be a degree of coding inaccuracy which is not quantified.

AbStrACt
Objective Conclusions We provide contemporaneous statistics on the live birth rate of babies with Down syndrome, and their childhood death rate. They require more hospital admissions, readmissions emergency admissions and longer lengths of stays than their peers, which has received scant research attention in the past. This demonstrates the importance of statutory planning as well as informal support to families to avoid added problems in child development and family bonding over and above that brought by the intellectual disabilities associated with Down syndrome.

IntrOduCtIOn
Improvements in health and social care and attitudes to disability have increased the survival of people with Down syndrome in recent decades. [1][2][3][4][5][6] In England and Wales, it was estimated that there were 37 090 people with Down syndrome in 2011, 7  Open access not available for Scotland. Given the changing population demographic for people with Down syndrome, and limited evidence base, we require a better understanding of their live birth rate and future health service needs. 4 Down syndrome is associated with a range of congenital anomalies, particularly cardiac, and a higher risk for respiratory, immunological, endocrine and gastrointestinal conditions. 4 11 This phenotype puts people with Down syndrome at risk of hospitalisation. This may impact on their development and family relationships, and so is an added disadvantage on top of their intellectual disabilities. Despite this, few previous studies have investigated hospitalisation of children and young people with Down syndrome, compared with the general population. Fitzgerald et al 12 1977 and 2008. The people with Down syndrome had more than twice the rate of hospital admissions, and nearly three times as many bed days. Both studies found that length of admissions of the people with Down syndrome had reduced over the periods studied, and Zhu et al 13 reported more bed days for males, children under 5 years, and those with congenital heart anomalies. Other studies have lacked general population comparison data or been limited to single conditions. We are not aware of any large-scale studies specifically of children and young people with Down syndrome investigating these issues compared with their peers. There has been very little other investigation of the factors that influence length of hospital stay of children and young people with Down syndrome, and we located no studies on risk of readmission or on emergency admissions of children and young people.
Without accurate information on live birth and death rates, and hospital admission needs, it is not possible to plan for the healthcare support that children with Down syndrome and their families need.
This study's aims were to investigate the (1) incidence of live births of people with Down syndrome over a 25-year period, (2) frequency of deaths of children and young people with Down syndrome compared with matched controls and (3) hospital admission frequency and duration, emergency admissions and readmissions of children and young people with Down syndrome compared with matched controls.

MethOdS
Written informed consent was not obtained from participants for their clinical records to be used in this study, but patient records were anonymised and de-identified prior to analysis. Safehaven approval was granted to approved researchers to analyse the data.

Patient and public involvement
This study was undertaken by the Scottish Learning Disabilities Observatory, which has a specific remit for people with intellectual disabilities, including people with Down syndrome; its steering group includes partners from third sector organisations, including Down Syndrome Scotland. Results from this study will be disseminated for people with Down syndrome and their families/carers in easy-read version via the Scottish Learning Disabilities Observatory website and newsletters.
Study sample, setting and process All NHS Scottish Regional Genetics Centres (the east, north, south-east and west Scotland centres) identified all live birth infants screened positive for Down syndrome postnatally (live births with trisomy 21, mosaic trisomy 21 or unbalanced translocation resulting in trisomy 21) between 1 January 1990 and 1 December 2015. In Scotland, everyone is given a unique Community Health Index (CHI) at birth which is used in their health records. The CHI database is held centrally by National Services Scotland (NSS), and was used to identify five controls without Down syndrome for each person with Down syndrome, matched on sex, age (by month and year of birth) and neighbourhood deprivation (using the Scottish Index of Multiple Deprivation (SIMD)). 14 15 Down syndrome infants who could not be matched due to missing data or had less than 1 year of potential follow-up time (ie, births in 2015) were excluded from statisitcal analyses. The CHI also provides a means to record link each person identified with Down syndrome and their matched controls to routinely collected hospital admissions data (Scottish Morbidity Records 01: SMR01), 16 and National Records of Scotland death certificate data. 17 The end of the period of follow-up was 1 December 2015. However, children born in 2015 were excluded from the statistical analyses to allow a minimum potential follow-up time of 1 year for all children. Hospital admissions SMR01 contains episode-based records for all nonpsychiatric, non-obstetric acute hospital admissions in

Open access
Scotland. The information collected includes the date of admission to hospital and discharge from it and whether the admission was routine or emergency. In SMR01, continuous periods of care are accounted for with a continuous inpatient stay (CIS) marker. This CIS marker ensures that a series of individual episodes over an unbroken period of care (eg, transfers between wards or hospitals) can be identified and treated as one admission rather than several admissions. Transfer of a baby to a neonatal intensive care unit after birth is counted as a first admission. Data quality assurance assessments for SMR01 are performed periodically. 19 20 Admission type Admission type is coded as emergency or routine. 'Emergency' admissions are those which were unplanned in advance; planned admissions are 'routine' admissions.
Scottish Index of Multiple Deprivation SIMD matching with controls and reporting is by quintiles, where SIMD 1 is the most deprived neighbourhoods and SIMD 5 is the most affluent neighbourhoods. SIMD is calculated at datazone level, identified from postcodes.

Discharge type
Discharge type is coded as a regular discharge or irregular discharge. Irregular discharges include, for example, a patient discharging himself/herself against medical advice, or death.
All follow-up/censoring Children were followed up from birth, and all models were censored on death or 1 December 2015 (whichever came first) unless stated otherwise.

Analyses
Incidence of Down syndrome Incidence of Down syndrome births was calculated for each calendar year from 1990 to 2015 inclusive.

Death
Incidence rates for mortality per person time (per 1000) were described, split by age group (for all ages, 0-1 month of life, from 1 to 12 months, from 13 to 60 months, from 61 to 120 months, from 121 to 180 months and from 181 to 240 months), as were underlying causes of death, for people with Down syndrome and their matched controls.

First hospital admissions
Incidence rates for first admissions per person time (per 1000) for people with Down syndrome compared with controls were described, split by age group (for all ages, 0-1 month, from 1 to 12 months and from 13+ months). Descriptive data for first admissions were presented including by sex, SIMD, year of birth and duration of first admissions in days, and this was compared using dependent t-tests for continuous variables and χ 2 tests for categorical variables or Mann-Whitney U tests for length of hospital stay. The relative risk of first hospital admission was compared for people with Down syndrome and controls using stratified (by sex, birth year group, SIMD) Cox regression models (log-log plots confirmed the proportional hazards assumption was met). The follow-up period was defined as from date of birth until date of first admission, and admission type was entered as a potential confounder.

Risk of emergency hospital admission
The relative risk of emergency first hospital admission was compared for people with Down syndrome and controls, analysed using stratified (by sex, birth year group, SIMD) Cox regression models. The follow-up period was defined as from date of birth until date of first admission.

Duration of hospital admission
Conditional (stratified by sex, birth year group, SIMD) linear regression was used to model the duration of first hospital admission (log transformed to help negate potential skewing of results) as the outcome, comparing Down syndrome and control groups. Admission type was entered as a potential confounder.

Risk of readmission
The relative risk of readmission (the next admission after the first admission) was compared for people with Down syndrome and controls, analysed using stratified (by sex, birth year group, SIMD) Cox regression models. The follow-up period was defined as date of discharge from the first admission until date of the next admission. First admission type and discharge type from first admission were entered as potential confounders, and the interaction of group (Down syndrome vs controls)*admission type (emergency vs routine) was included.
Reference groups for all regression analyses were: not having Down syndrome, females, most affluent neighbourhoods, routine admissions and regular type of hospital discharges. All analyses were conducted with IBM SPSS V.22.
reSultS Between 1 January 1990 and 1 December 2015, the Scottish Regional Genetics Centres identified 1479 live births with Down syndrome. One thousand two hundred and thirty-five people were matched with 6175 controls. One hundred eighty-seven (12.6%) infants with Down syndrome were excluded from statistical analyses, as they could not be matched due to missing data on SIMD quintile (n=160) in view of missing postcode information, sex (n=21), month and year of birth (n=<5) and CHI for example, due to migration out of Scotland (n=<5). A further 57 children were excluded as they were born in 2015 (and had less than 1 year of potential follow-up time). Of the 1479 infants born with Down syndrome, 689 Open access    Table 3

duration of hospitalisation
The duration of first hospital admission was significantly longer for the Down syndrome group at 5.03 days (median=2 days) versus 1.78 days (median=1 day) by 3.25 days. After adjustment (with a log transformed time variable, to help account for any data skew) for sex, birth year group, SIMD and type of admission, the Coefficient=0.25 (95% CI 0.22 to 0.28) (table 5). This log transformed coefficient can be interpreted as a ratio of difference for the Down syndrome population compared with the control group (exponentiate of the log transformed coefficient, exp(0.248)=1.28145993219-1×100=28, with a baseline length of stay (intercept) exp(0.12)=1.12749685158). This means the average length of stay was 28% higher in the Down syndrome population, compared with the controls. All exponentiated values are presented as percentages in table 5.

risk of readmission
Of the Down syndrome group, 945/1105 (85.5%) had at least one readmission, compared with 1685/3305 (51.0%) of the control group. Readmission incidence rates per person time (per 1000) for the Down syndrome population versus controls were 54.85 (95% CI 51.46 to 58.46) (n=945) vs 15.06 (95% CI 14.36 to 15.80) (n=1695). The baseline age at readmission was lower for the Down syndrome group with a median age of 6 months (IQR 1-17) compared with the controls with a median age of 11 months (IQR 5-20; p<0.001). The unadjusted HR of Down syndrome for time to readmission following discharge from first admission was 2.44 (95% CI 2.13 to 2.79). The HR when adjusted for type of admission, type of first discharge and group*emergency interaction was 2.56 (95% CI 2.08 to 3.14). Results are displayed in table 6.

dISCuSSIOn Principal findings and interpretation
We found the Scottish incidence of Down syndrome live births to be 1.0/1000 births over the last 25 years, with it being possibly higher now than in the early 1990s. The Scottish birth rate has fallen overall, but incidence of Down syndrome has not. Death rate was 18.5 times higher for the children/young people with Down syndrome than for their controls, and their causes of death differed, including causes that might be expected to have been amenable to good healthcare such as infections.
Over a 25-year period, we found that children/young people with Down syndrome compared with matched controls were considerably more likely to have had at least one hospital admission (IRR of 72.89 compared with 40.51; adjusted HR=1.84) and readmissions (IRR of 54.85 compared with 15.06; adjusted HR=2.56). Their admissions were more likely to have been emergency admissions (IRR of 56.78 compared with 28.88; first emergency admission HR=2.87) and of longer duration 28% longer first admission after birth. Given the higher rate of first emergency admissions in the Down syndrome group, we Open access  included the interaction of group×type of first admission (emergency or routine) in the analysis of readmissions, but they still remained more likely to be readmitted. We are not aware of any previous study of readmissions and emergency admissions in this population. The rate of admissions slightly increased over time (from 1990-2003 to 2004-2014) for the children/young people with Down syndrome (from 90.7% to 92.2%), while it decreased for the control group (from 63.3% to 44.8%), perhaps reflecting changes in medical practice over time, to keep children out of hospital as far as possible, but admitting for serious problems which the Down syndrome children may be more likely to experience. These findings are important, as they demonstrate the impact of additional health needs to both children and young people with Down syndrome, and to their families. This demonstrates the importance of statutory as well as informal support to families to avoid added problems in child development over and above the intellectual disabilities that the children with Down syndrome experience, and the impact caring can have on families. Additionally, the high rates of emergency admissions and readmissions may indicate some poorer quality management of healthcare at the primary care level (to avoid admissions), and have not to our knowledge been reported before, though other factors, including underlying health conditions may contribute. The Down syndrome group were younger at the time of first hospital admission, had a more equal sex distribution for first admission than the control group and an equal distribution across extent of neighbourhood deprivation, while the control group had a gradient with those in the most deprived areas more likely to have had a hospital admission. This suggests that at present, the children with Down syndrome appear to receive comparable care in terms of first admission, regardless of their neighbourhood, and highlights that findings in the general population cannot be relied on when planning services Open access for people with Down syndrome/intellectual disabilities, such as focussing services in areas of greatest deprivation. Families of children and young people with Down syndrome need focused support regardless of the areas they reside in.
Comparison with previous literature As well as being 18.5 times more likely to die than their controls, we found that the children/young people with Down syndrome were more likely to be hospitalised, emergency hospitalised, readmitted and for longer durations. We have not identified any previous studies on readmission with which to compare our findings, nor on emergency admissions. For the earlier period of 1977-2008 in Denmark, Zhu et al 13 reported that children and adults with Down syndrome had more than twice the rate of hospital admissions (all admissions, not separately reported for first admissions or repeat admissions), similar to our adjusted HRs for first and repeat admissions, but we specifically investigated children and young people unlike that study. The smaller Australian study for the earlier period of 1983-2004 suggested that the hospitalisation rate was five times higher, but relied on previously reported general population data from 1995. 12 As our study shows, and as previously reported in a study of children with Down syndrome identified from population-based sources in 1997 (n=210) and 2004 (n=208) in Western Australia, 21 it is highly likely that admission rates change over time, so these two studies are not strictly comparable to ours, and we have found no studies contemporaneous with ours with which to draw comparisons. Previously, small-scale studies reported length of admissions of children with Down syndrome to be 4.7 days in a Swedish study of 211 children followed up for at least 17 years, 22 and 7 days specifically for bronchiolitis in a small retrospective case-control study in Singapore, 23 compared with the 5.0 days we found. A further small study of 213 children with Down syndrome aged 0-3 years in America reported high admission rates. 24 The live birth incidence for Down syndrome we report is similar to that recently reported in England, USA and Canada. [8][9][10] Antenatal screening changes may underlie the variability in the birth rates found over time. Previous research in Scotland showed that decisions to terminate pregnancy following antenatal diagnosis of aneuploidy had fallen over time, 1 which our results are also possibly suggestive of from the early 1990s to more recently. In contrast, in some other countries (eg, Denmark), following the introduction of antenatal combined screening for Down syndrome in 2004, there was a concurrent increase in prenatal diagnoses and the number of Down syndrome live births decreased suddenly and significantly before then stabilising. 25 A higher proportion of the Scottish population have catholic faith compared with Denmark which might contribute in part to differences between countries: a range of complex factors influence women's decision making regarding prenatal screening and diagnosis of Down syndrome, and it is possible that more societal acceptance of diversity and people with disabilities, or better availability of family support may influence decisions in Scotland over time, 1 26 although this requires further investigation.

Strengths and limitations
This is the first population-based study examining incidence of Down syndrome and healthcare utilisation over time in Scotland compared with the general population, and the first worldwide to investigate emergency hospital admissions and readmissions. It controlled for the potential confounders of age, sex and SIMD. It was not subject to selection bias, as the population identification was based on routine data sources from all genetic service laboratories in Scotland (ie, the whole population). The Cox regression analysis for risk of admissions and readmissions provides advantages over logistic or linear techniques as this includes consideration of time to event (person time at risk). However, there were some missing control data, mostly due to missing postcodes, and this is a limitation of the study. There were no data on ethnicity which is a limitation; the Scottish population is 96% white. The use of routinely collected hospital and death statistics has many advantages including large-scale coverage, but there is likely to be a degree of coding inaccuracy which is not quantified. Children with Down syndrome have higher risk of comorbidities, so may be more likely to be transferred to neonatal intensive care units which might explain the increased risk of first hospital admission rates, but we could not distinguish such admissions in our dataset. We did not aim to investigate underlying conditions which are associated with admissions and death in the two groups.

COnCluSIOnS
As people with Down syndrome are living longer than in the past, we have provided contemporaneous statistics for the live birth rate of babies with Down syndrome and their childhood death rate compared with peers. We have found that their increasing survival is not without health consequences for the child, their family and health services-we report the majority require hospital admissions, readmissions and longer lengths of stays than their peers, which has received scant research attention in the past. Some factors may even suggest poorer healthcare than their peers, such as higher rates of emergency admissions, although we did not aim to investigate the conditions associated with admission. Together this brings important new findings necessary to address the challenge of planning support that children with Down syndrome and their families need. They require adequate healthcare, and also an awareness of the impact that hospitalisation can have on child development and family bonding. There may be a need to develop tailored packages of support for families with children/young people with Down syndrome to cope with the higher healthcare needs of their offspring.