Dataset Information

Assessing machine learning for fair prediction of ADHD in school pupils using a retrospective cohort study of linked education and healthcare data.

ABSTRACT:

Objectives

Attention deficit hyperactivity disorder (ADHD) is a prevalent childhood disorder, but often goes unrecognised and untreated. To improve access to services, accurate predictions of populations at high risk of ADHD are needed for effective resource allocation. Using a unique linked health and education data resource, we examined how machine learning (ML) approaches can predict risk of ADHD.

Design

Retrospective population cohort study.

Setting

South London (2007-2013).

Participants

n=56 258 pupils with linked education and health data.

Primary outcome measures

Using area under the curve (AUC), we compared the predictive accuracy of four ML models and one neural network for ADHD diagnosis. Ethnic group and language biases were weighted using a fair pre-processing algorithm.

Results

Random forest and logistic regression prediction models provided the highest predictive accuracy for ADHD in population samples (AUC 0.86 and 0.86, respectively) and clinical samples (AUC 0.72 and 0.70). Precision-recall curve analyses were less favourable. Sociodemographic biases were effectively reduced by a fair pre-processing algorithm without loss of accuracy.

Conclusions

ML approaches using linked routinely collected education and health data offer accurate, low-cost and scalable prediction models of ADHD. These approaches could help identify areas of need and inform resource allocation. Introducing 'fairness weighting' attenuates some sociodemographic biases which would otherwise underestimate ADHD risk within minority groups.

SUBMITTER: Ter-Minassian L

PROVIDER: S-EPMC9723859 | biostudies-literature | 2022 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Assessing machine learning for fair prediction of ADHD in school pupils using a retrospective cohort study of linked education and healthcare data.

Ter-Minassian Lucile L Viani Natalia N Wickersham Alice A Cross Lauren L Stewart Robert R Velupillai Sumithra S Downs Johnny J

BMJ open 20221205 12

<h4>Objectives</h4>Attention deficit hyperactivity disorder (ADHD) is a prevalent childhood disorder, but often goes unrecognised and untreated. To improve access to services, accurate predictions of populations at high risk of ADHD are needed for effective resource allocation. Using a unique linked health and education data resource, we examined how machine learning (ML) approaches can predict risk of ADHD.<h4>Design</h4>Retrospective population cohort study.<h4>Setting</h4>South London (2007-2 ...[more]

PMID: 36576182

Similar Datasets

Project description:Gross motor coordination (GMC) is essential for the development of specific motor skills and long-term participation in physical activities and sports. Group analysis reveals that, on average, children develop these skills gradually; however, how individuals develop GMC is less clear. The main aim of this study is to increase the understanding of developmental patterns within one school year, and whether children's grade, gender, or baseline GMC proficiency are associated with these developmental patterns. In total, 2594 Dutch children aged 6-12 years performed the modified Körper Koordinations Test für Kinder (KTK3+) twice in one school year (autumn and spring). The KTK3+ includes four subtests: walking backwards, moving sideways, jumping sideways, and eye-hand coordination (EHC) test. On average, children developed significantly on all subtests (p < 0.001). At baseline, children in higher grades scored significantly higher than children in lower grades, and children in grades 5 and 6 (age 9 and 10 years) showed most development (raw scores on average, p < 0.001). Boys outperformed girls on EHC across all grades, whereas girls outperformed boys on walking backwards. Nevertheless, both boys and girls developed similarly. Children with lower scores at baseline developed more quickly across all grades. Noteworthy is that 12.1-24% (depending on the test item) of the children scored lower in the spring than in the autumn tests. On average, children develop their GMC; however, we report large differences in their individual trajectories and note that a substantial number did not show a positive GMC development. Further research should examine GMC development with more possible influencing factors as well as over a longer time span to better understand differences in children's GMC development. This may result in more individualized programs in PE lessons, enabling children to optimally develop their GMC, and better use of GMC assessment tools to monitor children's development.

Project description:BackgroundData suggest that Ghana has made significant improvements in the educational system, resulting in some accomplishments. Nonetheless, pupils' academic performance at the junior high school level and in the Basic Education Certificate Examination (BECE) remains poor nationally. Some factors have been identified to influence poor academic performance in some parts of Ghana, but none in the Krachi West district. Thus, it has become imperative to determine the local factors accountable for this trend and identify the most appropriate localised solutions.Materials and methodsAn embedded research design was employed to ascertain the influence of pupils' characteristics and school environment factors on academic performance. In all, 361 participants (325 BECE candidates, 12 teachers, 12 head teachers and 12 PTA/SMC Chairpersons) were selected using cluster sampling, simple random and purposive sampling across 12 schools in the district. The academic performance of pupils was assessed using scores from six (6) subjects in the standardised district-level mock examination.ResultsBased on the examination scores, half of the pupils performed poorly in the six subjects. Even though from the quantitative study, pupils' characteristics had no significant influence on academic performance, school location (β = -3.29, p < 0.01), school type (β = 1.15, p < 0.01), and school environmental factors (β = 0.807, p = 0.024) were significant predictors of academic performance in the district. Pupils in private schools were thrice more likely to achieve average academic performance than pupils from public schools (OR = 3.2, CI = 1.06-9.47). Also, schools with good environmental factors were twice as likely to have average academic performance than schools with poor environmental factors (OR = 2.2 CI = 1.11-4.52).ConclusionWhile school environment factors and pupil characteristics have a relationship with academic performance as suggested by the ecological theory, it was established in this study that only school location, school type and school environment factors were the significant predictors of academic performance in the Krachi West District. Therefore, education stakeholders need to consider these predictors when coming up with integrated but local strategies to improve pupils' academic performance in that district.

Dataset Information

Assessing machine learning for fair prediction of ADHD in school pupils using a retrospective cohort study of linked education and healthcare data.

Objectives

Design

Setting

Participants

Primary outcome measures

Results

Conclusions

Publications

Assessing machine learning for fair prediction of ADHD in school pupils using a retrospective cohort study of linked education and healthcare data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets