Development & validation of a scalable mobile platform for screening of developmental neuropsychiatric disorders in low-resource settings

Worldwide, 1 in 161 children are diagnosed or diagnosable with Autism Spectrum Disorders (ASD). In India, our own study found a prevalence just over 1% in children aged 2 to 9 years. Although parents may notice ASD symptoms as early as 24 months, social and economic barriers to access to qualified health personnel mean that most of these children do not receive early diagnosis or intervention - or indeed any diagnosis or intervention at all. Moreover, many parents are unaware of age-appropriate developmental milestones and therefore clinical opinion is sought only when symptoms become obvious and impossible to ignore. This avoidable delay is an unfolding tragedy in light of evidence showing that front-line worker-delivered, parent-mediated interventions (e.g. PASS) lead to better behavioural and social outcomes. Scalable methods to flag up probable ASD would promote early referral to scarce specialist diagnostic facilities, connecting families with affordable, scalable, community-based interventions. We will realise this goal by developing a low-cost mobile screening platform ('Screening Tools for Autism Risk using Technology' (START)) usable in the home or in a routine health facility by the same non-specialist health workers who deliver PASS or other front-line interventions.

START will collect data from multiple sources. First, parents will be asked simple questions about their child's everyday behaviour, based on established questionnaires that have been validated in the Indian cultural context. To help parents identify and report subtle abnormalities, these questions will be supplemented by videos illustrating typical versus ASD behaviours. This video-enabled questionnaire will be delivered on a tablet PC. Second, low-cost eye-tracking technology on the same tablet PC will be used to monitor the child's eye movements in simple tasks, such as those assessing preference for social versus non-social images, and measuring how quickly attention shifts to new objects appearing on the screen. Finally, a segment of parent and child interaction will be recorded using the inbuilt camera, and used to code for signs of atypical behaviour. This combination of multiple measures will provide independent channels of data collected on a single platform, potentially improving on current screening methods that rely on one technique solely. Combining these techniques becomes especially important in a setting where challenges such as poor awareness about ASD may impact on parent reporting of abnormal behaviour (for example, parents may not have noticed that their child avoids direct gaze).

START's central innovation is its integration with existing front-line workers in LMICs, by adapting methods successfully employed in high income countries (HICs) to LMIC environments and resources. By virtue of its design to require minimal training, it will promote task-shifting - a concept the World Health Organization defines as "a process of delegation whereby tasks are moved, where appropriate, to less specialized health workers" and thus will reach wider populations. This task-shifting approach reduces the burden on the small number of highly-skilled mental-health professionals in LMICs, who then are free to focus on confirming diagnoses and prescribing appropriate interventions which, in turn, can also be delivered by non-specialist workers. The resulting enhancement to detection of ASD may also help develop community awareness and, in the longer term, address the barrier of low demand for services in these areas.

The development and application of the START platform involves collaborations across the breadth of basic and applied sciences - including public health research, technology, clinical psychology and neuroscience, and will be led by world-class institutions in the UK and India. Such global partnerships will build research capacity in India to bridge these wide detection gaps for neuropsychiatric disorders in low resource settings.

Autism Spectrum Disorders (ASD) emerge early and persist throughout life, contributing significantly to global years lived with disability. Typically, ASD diagnosis depends on clinical diagnostic assessments by highly trained professionals. This high resource demand poses a challenge in resource-limited areas where skilled personnel are scarce and public literacy and awareness of neurodevelopmental symptoms are low. The resulting delay or outright absence of diagnosis obstructs the community-based, parent-mediated interventions that promise to improve outcomes. Closing this detection gap will depend on task-shifting from scarce experts to more plentiful community health workers, via scalable screening tools usable in the community and with only minimal training. Specifically, this project aims to:
1. Develop a mobile platform incorporating parent report and child performance based assessment methods for detection of ASD, delivered by non-specialist health workers in community-based settings; and
2. Create a data transfer, storage, and analysis pipeline, which will enable subsequent combination of data streams from large-scale studies, to build and optimise probabilistic classifiers for estimating ASD risk and guiding referral to specialised diagnostic clinics and appropriate care pathways during translation to scale; and
3. Evaluate the usability, utility, and validity of the platform by pilot-testing in ~120 children and parents in and around Delhi (equal numbers of typically developing, DSM-5-diagnosed ASD, and non-ASD children with Intellectual disability (ID)).

These objectives will be achieved by an interdisciplinary team of public health researchers, clinicians, computer scientists, software designers and developers, and neuroscientists from UK and India. This robust international partnership will form the programmatic basis of a long-term research effort translating neuroscience into community health in low-resource settings.

South Asia has the largest number of ASD children in the world, with India alone accounting for over 5 million children with ASD. In India, as in other LMICs, the treatment gap is close to 100% due to a) the paucity of mental health professionals, b) their concentration in urban private healthcare facilities, c) the use of expensive proprietary tools for diagnosis, and d) the lack of awareness of neuropsychiatric symptoms and consequent failures to seek treatment. This project will develop a mobile platform that can be administered by non-specialist health workers to identify probable cases of ASD. It will thus take a definite step to address the imbalance between where most autistic individuals live (LMICs) and where most of the autism research is conducted (HICs). This platform can potentially reduce this detection gap for ASD in India, and can be extended in future to other neuropsychiatric disorders.
The design phase of the project will incorporate iterative feedback from parents of children with ASD. This process will help maximise the impact on the target population once the full-scale study is launched.
Early intervention improves behavioural outcomes in ASD. A recent trial by project investigators demonstrated that a non-specialist-delivered ASD intervention (PASS) in community settings in India and Pakistan is acceptable, feasible and effective. The availability of a scalable tool that detects probable ASD cases, and transfers them to a referral pathway for clinical diagnosis and interventions such as PASS, would give more families access to these evidence-based interventions.

Three groups of stakeholders will benefit directly from this research: academics, healthcare systems as well as families of children with ASD. The impact for academics will be the availability of data from a scalable tool for ASD detection implemented in an LMIC. Project data and code will be shared freely through open access mechanisms, thus reducing dependence on proprietary tools. The combination of measures recorded by the platform may help develop future diagnostic tools for ASD. Since the platform will be designed to be extensible to other disorders, a prototype will be available for academics interested to develop similar projects in other other low-resource settings. The facilitators and barriers to achieving our targets will be widely shared with all stakeholders, thus ensuring the implementation of increasingly feasible and efficient approaches in future projects. Beyond the development of the platform, one of the deliverables of this project will be the cementing of an interdisciplinary autism research partnership across UK and India. Such a partnership will serve as the foundation for future academic collaborations.

A well-defined referral pathway that will be set up as a next-step of developing the detection tool will have a twofold benefit for healthcare systems : a) reducing the burden on the limited number of specialist clinicians by suggesting probable conditions, enabling them to focus on formal diagnosis; and b) reducing health-care related costs by promoting earlier diagnosis and onset of interventions. Capacity in the form of local expertise will be built within cadres of community health workers. Families of children with ASD will benefit by becoming more aware of the signs of probable ASD, and developmental disorders in general. Increased awareness about these disorders is expected not only to improve early treatment seeking behaviour, but also to reduce anxiety over the outcome of an affected child by connecting them to a defined referral pathway. A widely applied screening mechanism for childhood developmental disorders would also help increase general awareness on this topic, which over time, would lead to reduced stigma associated with these disorders. Perhaps most significantly for the long term, these changes will be codified into public policy via links with governmental organisations and advisory bodies.