Developmental motor problems and health‐related quality of life in 5‐year‐old children born extremely preterm: A European cohort study

To measure the association between cerebral palsy (CP) and non‐CP‐related movement difficulties and health‐related quality of life (HRQoL) among 5‐year‐old children born extremely preterm (<28 weeks gestational age).

With advances in obstetric and neonatal care, survival after being born extremely preterm (<28 weeks gestational age) has increased in recent decades, 1 but children born extremely preterm continue to face substantially higher risks of developmental problems and disorders than their peers born at term. 1,2Developmental motor problems, which include cerebral palsy (CP), developmental coordination disorder (DCD), and other movement difficulties, are common consequences of being born extremely preterm.5][6][7] While CP is clearly defined as a motor disorder with established guidelines for health care provision, [8][9][10] there is less agreement about whether all children with movement difficulties should be classified as having a disorder.The low use of motor function-related health care services by children born extremely preterm with movement difficulties raises questions about whether this reflects the absence of a health disorder or an unmet need. 11mpaired motor function can limit daily activities, schooling, and social relationships, but the consequences of CP and non-CP-related movement difficulties on children's lives and well-being are not well described. 12One approach to quantifying this impact is to measure health-related quality of life (HRQoL).4][15] In a systematic review, Zwicker and Harris 15 concluded that preschool children born preterm or with a very low birthweight (<1500 g) tend to have lower parent-reported HRQoL, whereas evidence in older children is scarce.However, these studies did not explore the extent to which lower HRQoL is associated with specific preterm birth-related developmental comorbidities, including developmental motor problems.Among children with CP or DCD in the general population, studies generally reported worse physical, psychosocial, and social functioning compared with typically developing peers. 12,16he existing literature suggests that children's HRQoL is likely to be impacted by movement difficulties associated with being born extremely preterm, but these effects have not been studied and may be heterogeneous.Our principal aim was to measure the associations of CP and non-CP-related movement difficulties with HRQoL among children born extremely preterm.We further sought to assess whether these associations existed in the absence of other neurodevelopmental impairments (NDIs) and among children not receiving motor function-related health care services.

Study design and participants
This study used data from the Screening to Improve Health in Very Preterm Infants in Europe (SHIPS) populationbased, prospective cohort of children born very preterm (<32 weeks gestational age) in 2011 to 2012 in 19 regions in 11 European countries.As described in more detail in the cohort profile, 17 inclusions took place over a 12month period (except for regions in France where they occurred over a 6-month period only), with all births (including stillbirths and terminations of pregnancy) between 22 weeks + 0 days and 31 weeks + 6 days occurring in all maternity units from the selected regions included in the study.Data were collected from obstetric and neonatal records during neonatal hospitalization as well as parental questionnaires at 2 and 5 years of age.At 5 years of age, clinical assessments of neurocognitive and motor functioning were also carried out for the subgroup of children born extremely preterm.Our study population was limited to this subgroup born extremely preterm, including 1021 children born extremely preterm (out of 1654 eligible children) followed at 5 years of age.

Ethics statement
All study regions obtained ethical approval according to national legislation.The study was also approved by the French Advisory Committee on the Use of Health Data in Medical Research and the French National Commission for Data to −6.1], and − 26.1 [−31.0 to −21.2]).Quantile regression analyses showed similar decreases in HRQoL for all children with CP, whereas for children with non-CP-related movement difficulties, reductions in HRQoL were more pronounced at lower centiles.Interpretation: CP and non-CP-related movement difficulties were associated with lower HRQoL, even for children with less severe difficulties.Heterogeneous associations for non-CP-related movement difficulties raise questions for research about mitigating and protective factors.

What this paper adds
• Children born extremely preterm with motor function problems had lower health-related quality of life (HRQoL).• Clinically significant lower HRQoL was observed in children with less severe difficulties.• Lower HRQoL was also found in children without motor function-related health care use.• Reductions in HRQoL were less marked in children with higher reported HRQoL.
Protection and Liberties.Parents gave their written informed consent to participate in the study before any data collection.

Developmental motor difficulties
Movement difficulties were assessed and classified using the Movement Assessment Battery for Children, Second Edition (MABC-2), 18 a validated test previously used in high-risk populations such as children born extremely preterm. 5It considers performance on eight motor tasks in three domains: manual dexterity; aiming and catching; and balance.All tasks are summed and converted to an ageadjusted centile score and classified into: no movement difficulties (>15th centile); at risk of movement difficulties (6th-15th centile); and significant movement difficulties (≤5th centile).Because national norms do not exist in all countries, we applied UK norms, which were originally developed for the test and are most commonly used to derive centile scores for all children. 18,19The MABC-2 was administered by trained psychologists or physiotherapists in local routine follow-up programmes where available (Belgium, the Netherlands, Sweden), or in the SHIPS research teams (other countries).Common data collection guidelines and a core data collection form were developed to standardize procedures and reporting.Information on CP diagnosis was reported by parents in the 5-year questionnaire, except in France where CP diagnosis was ascertained during a medical visit.
We classified developmental motor problems into four groups.Children with CP were considered a distinct group because CP is a well-defined neurodevelopmental disorder with a specific aetiology; 8 children without CP were classified according to their MABC-2 centiles.The MABC-2 is not designed to assess movement abilities in children with severe NDIs who have difficulty accomplishing the tasks because of their sensory or cognitive deficits. 18We therefore excluded children with an IQ < 55 (<-3SD) or severe hearing or visual impairment (Figure S1).

Health-related quality of life
The child's HRQoL was measured using the parent-reported Pediatric Quality of Life Inventory (PedsQL) generic core scales 20 for the ages 5 to 7 years.This instrument has demonstrated reliability and validity in typically developing school populations and in pediatric populations with acute or chronic health conditions. 21,22It consists of 23 items on a 5-point Likert scale in four domains: physical (eight items); emotional (five items); social (five items); and school functioning (five items).Items were reverse-scored and linearly transformed to a 0 to 100 scale, with higher scores indicating a better HRQoL. 20hree summary scores were computed: the physical health summary score (eight items on physical functioning); the psychosocial health summary score (15 items on emotional, social, and school functioning); and the total score (23 items).

Covariates
The following covariates were considered as potential confounders: child age and sex, and family sociodemographic characteristics, including maternal age and parity at the child's birth, maternal country of birth, parental cohabiting status, maternal educational level, and household unemployment status.For subgroup analyses, we used information on other NDIs, including low full-scale IQ and parent-reported hearing or visual impairments.

Analytical approach
We first described the characteristics of children included in the study sample and those excluded because of missing MABC-2 or PedsQL scores using means and SDs, as well as frequencies and proportions for continuous and categorical variables respectively.We compared the characteristics of children according to motor category (no movement difficulties, at risk of movement difficulties, significant movement difficulties, and CP).Linear regression analyses were used to measure associations between motor categories and HRQoL, adjusting for child age, sex, and family sociodemographic characteristics.
We carried out subgroup analyses first by excluding children with moderate NDIs (i.e.IQ between 55 and 69 [−2SD to −3SD], implants or hearing aids, or difficulty seeing even with glasses).Because multiple developmental impairments are frequent after extremely preterm birth and the association between these impairments is not clear, we aimed to measure whether developmental motor problems have an impact on HRQoL in the absence of these other difficulties.For the group with CP, children with severe NDIs were also excluded from these analyses that compared them with children without CP, where severe NDI is an exclusion criterion.In a second analysis, we additionally excluded children who were not receiving motor function-related health care services, defined as having at least one visit during the past year with: (1) a physiotherapist, motor development or psychomotor therapist, or kinesiologist; (2) an occupational therapist; or (3) an early intervention specialist. 11 final analysis explored possible differential effects over the HRQoL distribution using quantile regression models. 23,24While linear regression analysis measures the general relationship, on average, between developmental motor problems and HRQoL, quantile regression seeks to determine whether this relationship differs at multiple points of the outcome's distribution, 24 that is, whether this might be stronger or weaker for children experiencing different levels of HRQoL.This analysis was carried out on the population of children without moderate NDIs.Lastly, to facilitate the interpretation of differences in scores, given that HRQoL measurements have no intrinsic clinical thresholds, we calculated Cohen's δ as a measure of effect size, and used thresholds to interpret small (0.2), medium (0.5), and large (0.8) effect sizes. 25

Loss to follow-up and missing data
We used inverse probability weighting (IPW) to take into account loss to follow-up. 26As described previously, 7 the characteristics of responders and non-responders were compared (Table S1) and used to estimate the probability of response and create the weight (IPW) (Table S2).Missing data for covariates were imputed using multiple imputation by chained equations for the weights (20 imputed data sets) and the final models (20 imputed data sets). 27Missing data were less than 2% for covariates in the final models.
We did not impute data for children with missing MABC-2 or PedsQL scores because the 'missing at random' assumptions probably did not hold.However, for eligible children with missing MABC-2 data, neurodevelopmental specialists and an epidemiologist (RC, UA, SJ, and JZ) reviewed each case.If a child was unable to complete a task or component because of severe motor impairment, the lowest score on that particular scale was assigned.If data were missing for a task in the absence of other developmental problems, the average of the other tasks was assigned.In Belgium, some children had centile scores from the Movement Assessment Battery for Children, First Edition that were used. 28In all other cases, scores were left as missing.

Sensitivity analyses
Linear regression models were used for our main analysis, in line with the HRQoL literature. 29,30However, these scores are not normally distributed because of right censoring due to bounded data at the maximum (Figure S2).Ignoring this ceiling effect might result in biased and inconsistent estimates. 31,32In addition, there is no clear criterion for how large the sample size has to be before the central limit theorem applies. 31Censored regression models make it possible to produce consistent estimates in this situation.2][33] Because these censored regression models gave very similar results to the linear regression models, and the latter are commonly used in the literature, facilitating the comparison with other studies, we decided to present the results from the linear regression models in the main analysis and the results from the censored regression models as the sensitivity analysis.Lastly, we reran the final models using unweighted and complete case samples, and after excluding children with severe NDIs in the CP group (n = 15), as this was an exclusion criterion for children without CP.
All analyses were carried out with Stata version 15.0 (StataCorp, College Station, TX, USA).

R E SU LTS
Of 1671 infants born extremely preterm and alive at discharge, 1654 survived to 5 years and 1021 (61.7%) were followed (Figure S1).One hundred children had a CP diagnosis (unweighted prevalence = 9.8%, weighted prevalence = 11.1%).Among children without CP, 29 with severe NDIs were excluded.After excluding missing MABC-2 and PedsQL measures, the final sample consisted of 810 children.
Children were assessed at mean 5 years 7 months (SD 5 months) (Table 1); 42.0% were born at 27 weeks gestational age, 51.6% were males, and 72.5% were singleton births.Their mothers were aged 35 years or older for 26.7%, 58.9% were primiparous, 37.3% had at least a bachelor's degree, and 18.6% were born outside Europe.Because loss to follow-up was mainly related to social disadvantage (Table S1), IPW affected the distribution of most sociodemographic characteristics.
Younger maternal age and not breastfeeding were associated with missing MABC-2 (n = 116) and PedsQL scores (n = 63) (Tables S3 and S4).Missing PedsQL was also more frequent among multiparous and non-European mothers.
The proportion of males increased with the severity of movement difficulties and was highest among children with CP (Table 2).Children without movement difficulties and those at risk of movement difficulties had mothers with similarly distributed educational levels, while those with significant movement difficulties or CP were more likely to have mothers with a lower education.A larger proportion of household unemployment was observed for all children with movement difficulties or CP than children without movement difficulties.Among children with CP, 25.3% and 16.2% had severe and moderate NDIs respectively.For analyses of movement difficulties, severe NDIs were excluded, but moderate NDIs increased with the severity of movement difficulties.In the analyses without IPW, the results were similar, although the proportion of children with significant movement difficulties or CP was reduced while there was a higher proportion of children without movement difficulties or at risk of movement difficulties (Table S5).
Compared to children without movement difficulties, those at risk of movement difficulties, and with significant movement difficulties and CP had lower PedsQL total scores (β and 95% confidence intervals: −5.3]).Decreases were greater for physical than psychosocial scores.Models adjusted on sociodemographic characteristics provided similar results.After exclusion of children with moderate NDIs (n = 123), associations with the PedsQL total score were slightly attenuated, particularly for the group with CP.Finally, among children without moderate NDIs, those not receiving motor function-related health care services had higher PedsQL scores than those receiving services.However, they still had lower PedsQL scores than children without movement difficulties.Children with CP were not included in this analysis because almost all received motor function-related health care services.When these estimates were converted to Cohen's δ (Table S6), they were in the ranges of 0.31 to 0.41, 0.38 to 0.60, and 1.16 to 1.61 for children at risk of movement difficulties, children with significant movement difficulties, and children with CP respectively.
In quantile analyses, the estimated decline in PedsQL scores for children at risk or with significant movement difficulties in comparison to children without movement difficulties widened with decreasing quantiles, with non-significant T A B L E 1 (Continued) decreases at the 90th centile but up to 15.4 points for the 10th centile (Figure 1).In contrast, children with CP had lower scores for all quantiles.Similar patterns were observed for physical and psychosocial scores (Figures S3 and S4).
Analysis using censored regression models (i.e.Tobit and censored least absolute deviation models) did not show appreciable differences with linear models (Table S7).For instance, total score reductions for significant movement Values are weighted frequencies (rounded to a whole number), percentages (excluding missing values and rounded to one decimal point), and mean (SD) for continuous variables, all using inverse probability weighting to correct for loss to follow-up.Unweighted results are presented in Table S5.b Integrated cognitive, hearing, and visual impairment.Mild impairment was defined as an IQ between 84 and 70 (−1SD to −2SD), difficulties hearing but not requiring hearing aids or implants, or needing glasses but seeing well when wearing them.Moderate impairment was defined as an IQ between 69 and 55 (−2SD to −3SD), using hearing aids or implants but hearing well when wearing them, or having difficulties seeing even when wearing glasses.Severe impairment was defined as an IQ < 55 (<-3SD), deafness or difficulties hearing even when using hearing aids or implants, or blindness or seeing light only.c Other situations included student, parental leave, home parent, and other.
Abbreviations: CP, cerebral palsy; ISCED, International Standard Classification of Education.2) for the linear, Tobit, and censored least absolute deviation models respectively.One difference, however, was a smaller reduction in physical score for the atrisk category in the censored least absolute deviation model compared to the linear and Tobit models.Linear regressions using the unweighted and complete case samples, and excluding children with a severe NDI (n = 15) in the group with CP, yielded similar conclusions, although in the latter analysis, scores were slightly higher for children with CP.

Summary of findings
Developmental motor problems were associated with poorer HRQoL among children born extremely preterm aged 5 years, with the lowest level for children with CP, followed by those with significant movement difficulties and at risk of movement difficulties compared with children without movement difficulties.Decreases were greater for physical than psychosocial scores.Decreases were also more pronounced at lower HRQoL centiles for children at risk of movement difficulties or with significant movement difficulties, whereas decreases were observed at all centiles for children with CP.Differences persisted after exclusion of children with moderate NDIs and those not receiving any motor function-related health care services over the past year.

Comparison with the published literature
The results from this European multi-regional cohort, the largest with information on HRQoL and developmental motor problems in children born extremely Weighted frequencies (rounded to a whole number).b PedsQL score reductions are expressed as β and 95% CI using inverse probability weighting to correct for loss to follow-up.
c Maternal age and parity at child's birth, maternal country of birth, parental cohabiting status, maternal educational level, and household unemployment status.
d Integrated cognitive, hearing, and visual impairment.Moderate NDI was defined as an IQ between 69 and 55 (−2SD to −3SD), using hearing aids or implants but hearing well when wearing them, or difficulties seeing even when wearing glasses.For the group with CP, children with severe NDI were excluded from these analyses to be comparable to children without CP, where severe NDI is an exclusion criterion (i.e.IQ < 55 [<-3SD], deafness or difficulties hearing even when using hearing aids or implants, or blindness or seeing light only).e Receiving motor function-related health care services was defined as having at least one visit during the last 12 months to any of the following: (1) a physiotherapist, motor development or psychomotor therapist, or kinesiologist; (2) an occupational therapist; or (3) an early intervention specialist. 11bbreviations: CI, confidence interval; CP, cerebral palsy; MD, movement difficulty; NDI, neurodevelopmental impairment; PedsQL, Pediatric Quality of Life Inventory; REF, reference category.
preterm, 12,15,34 are consistent with the literature on DCDs where most studies reported worse physical, psychological, and social functioning in children with DCDs than their peers. 12,35Specifically among children born very preterm, 11-year-old children with DCDs (defined as an MABC-2 score ≤ 5th centile) self-reported lower HRQoL than children born very preterm without motor impairments; 36 children aged 6 to 10 years with global developmental disability and CP had reduced parent-reported HRQoL. 37Studies that used the 15th or lower centile of the MABC-2 as a cut-off score to indicate DCD suggested that children in the 'borderline' range for motor function experience significant reductions in HRQoL, in line with our findings for the at-risk group. 12Among children with CP, Makris et al. 16 reported worse physical functioning compared with typically developing peers, but the results were less clear concerning psychosocial functioning, with variations depending on the HRQoL instrument used, whereas we observed decreases in all domains.

Interpretation
There were marked differences in HRQoL scores between children with CP and those with non-CP-related movement difficulties, although they were less striking once children with moderate NDIs were excluded.We did not have information on the type and severity of CP, which are needed to further understand the results in the group with CP. 38 The fact that HRQoL scores followed a gradient reflecting the severity of reduced motor function provides support for a direct association; however, interpreting the scores is complex.0][41] Other studies on HRQoL regarded differences of 5 to 10 points on a 100-point scale as clinically significant. 42Hilliard et al. 39 established that the amount of change that children and young people with diabetes and their parents perceive as meaningful in everyday functioning or activities was between 4.5 and 6.5 points of the PedsQL score. 39This result is of the same order of magnitude  as the difference between children with and without health problems found in the initial PedsQL validation study and corresponds to the effect found for children at risk of movement difficulties and without NDIs in our study. 20Our effect sizes were also consistent with this interpretation using Cohen's δ.
Another question relates to the validity of parent-reported HRQoL with differences existing between self-reported and parent-reported HRQoL. 15,43Compared to other quality of life measures, the PedsQL instrument focuses on the child's functional status, 44 with better agreement between parents and children on observable items than measures focusing on feelings and lived experiences. 34,43Given the young age of the children and the high prevalence of NDIs in this study, self-report was not appropriate 34 but future research should assess whether parents' characteristics or beliefs contribute to the variation in HRQoL and confirm these findings using self-reported HRQoL.

Implications for clinical practice, policy, and research
8][49] These results showing the consequences of developmental motor problems on the HRQoL of children born extremely preterm at 5 years of age therefore highlight the importance of early screening and support.These findings also reinforce calls for continued follow-up of children born extremely preterm beyond 2 years and for better diagnosis and health service provision for children at risk of movement difficulties or with significant movement difficulties, 50 in particular for children without other developmental problems who may be less likely to receive motor functionrelated health care services. 11nalyses using quantile regression illustrated a heterogeneous association of movement difficulties with HRQoL and suggested that there may be mitigating or protective factors that preserve HRQoL for some children. 23,24Future studies should focus on identifying protective individual, familial, or broader environmental factors because these may lead to more effective interventions and better identification of children requiring more support. 51Using integrated measures of day-to-day functioning, such as the PedsQL, for research alongside more traditional neurodevelopmental measures opens up new opportunities for discovery and responds to families' concerns about the well-being of their child born extremely preterm. 52

Strengths and limitations
The main strengths of this study are its large populationbased sample of more than 800 children born extremely preterm with standardized collection of data on motor function and HRQoL using validated instruments.Although test and examiner reliabilities were not assessed across sites, the MABC-2 has good-to-excellent interrater reliability and test−retest reliability. 53CP was based on a parental report of a clinical diagnosis, which may lead to some misclassification, although almost all children with CP will have received a diagnosis by the age of 5 years in cohorts with high health service use 54,55 such as ours. 56This is supported by a CP prevalence of 11%, which is in the range of recent cohort studies (9−12%). 3,57,58However, analysis of the effect of type and severity of CP on HRQoL was not possible.We carried out subgroup analyses excluding children with moderate NDIs to assess the association between motor problems and HRQoL in the absence of these impairments.However, children born extremely preterm experience a range of developmental comorbidities that may affect their HRQoL.How other health and developmental conditions interact with CP and non-CP-related movement difficulties to affect HRQoL is an important area for future research.Finally, there may have been bias from loss to follow-up, although information was available on factors affecting attrition, which were primarily related to social disadvantage and IPW corrected for those characteristics that were under-represented in the follow-up sample.

Conclusion
In children born extremely preterm, developmental motor problems were associated with lower HRQoL at 5 years of age, even among children with less severe difficulties.These results illustrate the importance of follow-up and health service provision for children born extremely preterm across the full spectrum of motor problems to reduce their impact on HRQoL.

59
Abbreviations: CP, cerebral palsy; ISCED, International Standard Classification of Education.59 AND HEALTH-RELATED QUALITY OF LIFE IN 5-YEAR-OLD CHILDREN BORN EXTREMELY PRETERM: A EUROPEAN COHORT STUDY difficulties were − 9.1 (−12.0 to −6.1), −8.1 (−10.8 to −5.4), and − 9.2 (−13.6 to −6. Funding was received from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no.259882 and Horizon 2020 Research and Innovation Program under grant agreement nos.633724 and 733280.Raquel Costa was supported by the FSE and FCT under a postdoctoral grant no.SFRH/BPD/117597/2016. EPIUnit (UIDB/04750/2020) and ITR (LA/P/0064/2020) are financed by the Fundação para a Ciência e a Tecnologia.Additional funding is acknowledged from the following regions: France (French Institute of Public Health Research and Institute of Public Health and their partners the French Health Ministry, the National Institute of Health and Medical Research, the National Institute of Cancer, and the National Solidarity Fund for Autonomy, and grant no.ANR-11-EQPX-0038 from the National Research Agency through the French Equipex Program of Investments in the Future and the PremUp Foundation); UK (funding from The Neonatal Survey from Neonatal Networks for the East Midlands and Yorkshire and Humber regions).
Characteristics of children included in the study with and without IPW (n = 810).
T A B L E 1 Sample characteristics according to motor category (n = 810).a T A B L E 2 Unadjusted and adjusted association of PedsQL scores with motor category (n = 810).Subgroup analysis after exclusion of children with moderate NDI and stratified on receipt of motor-function related health care services (n = 638) e T A B L E 3 a Adjusted associations of Pediatric Quality of Life Inventory (PedsQL) quantiles (total score) with motor category for children without moderate neurodevelopmental impairment (NDI).PedsQL total score (REF) and reductions are expressed as β values and their 95% confidence intervals (CIs) rounded to one decimal point, all using inverse probability weighting to correct for loss to follow-up and adjusted for child age and sex and family sociodemographic characteristics, including maternal age and parity at child's birth, maternal country of birth, parental cohabiting status, maternal educational level, and household unemployment status.Children with moderate NDI defined as having an IQ between 69 and 55 (−2SD to −3SD), using hearing aids or implants but hearing well when wearing them, or having difficulties seeing even wearing glasses were excluded.For the group with cerebral palsy (CP), children with severe NDI were excluded from these analyses to be comparable with children without CP, where severe NDI is an exclusion criterion (i.e.IQ < 55 [<−3SD], deafness or difficulties hearing even when wearing hearing aids or implants, or blindness or seeing light only).
E 1