Keywords

1 Autism Spectrum Disorders and the Challenges Associated with Diagnosis in Young Children

According to the National Autistic Society [1] it is thought around 700,000 people in the UK are living with autism, with many remaining undiagnosed. Autism and autism spectrum disorders (ASDs) are lifelong developmental disabilities [2]. Autism is considered a spectrum condition, meaning that those at the lower-functioning end may require lifelong specialist support, whilst those at the high-functioning end may lead relatively independent lives [8].

A diagnosis of autism has been traditionally reached in the UK on the basis of the so-called triad of impairments: social impairments, communication difficulties and rigid and repetitive interests and activities [3]. More recently, however, the Diagnostic and Statistical Manual of Mental Disorders–Fifth Edition (DSM-5) of the American Psychiatric Association (APA) has redefined ASD based on a dyad of impairments in social-communication and inflexible behavioural traits [4]. It should be noted, however, that, although much of the research literature refers to the US DSM-4 and DSM-5 manuals, the UK’s National Health Service use the World Health Organisation’s International Classification of Diseases (ICD) [5]. In spite of this, it could be argued that the publication of the DSM-5 has had some cultural impact in the UK and within the National Health Service (NHS) more specifically, given the controversy over the need for a separate diagnosis for Asperger Syndrome.

Despite progression in understanding of the nature of ASD and possible etiological factors, there is no single biological marker for ASD at this time and so diagnosis is based on a portfolio of evidence collected from clinical observations and parental reports, which tend to reveal different information [6–8]. The accuracy and stability of a diagnosis is dependent, then, on reliable and comprehensive information being obtained from multiple sources [8]. In the absence of definitive tests, there remains the potential for bias in reporting and Taylor et al. [6] assert that the MCHAT-type (Modified Checklist for Autism in Toddlers) questionnaire, typically completed by caregivers, has the potential to result in over and under identification of ASD in young children in the absence of supporting evidence.

Whilst the average age of diagnosis of autism has reduced [9], early diagnosis remains key to improving the prognosis. Studies have shown that the impairments associated with ASDs can be ameliorated when interventions and support are accessed early on [10].

Within the NHS, cases of suspected ASD in young children are typically discussed at cross-disciplinary meetings, during which, a number of health care professionals such as speech and language therapists (SALTs), pediatricians and psychologists will present their findings based on their interactions with and observations of the child. These discussions may also be supported by parental reports or evidence supplied by educational professionals. This process, however, relies on snap-shot observations and accurate reporting in order to reach an accurate diagnosis of autism and this can result in the process taking several months or even years. In turn, this can result in families and schools experiencing delays in accessing support services and funding which are unlocked only once a formal diagnosis has been reached.

With estimates of prevalence being between 1 in 68 [11] and 1 in 88 children, accurate and effective identification of ASD in young children remains a pressing public health issue [6]. With continued financial pressures on the National Health Service there is perhaps a need to find cost-effective, efficient and reliable ways to of collecting evidence of behaviours indicative of the presence of an autism times in order to support and expedite the diagnosis process.

In this paper, we consider the potential for user-behaviour analysis software on tablet computers or smart phones, along with other m-health solutions, to provide a cost-effective opportunity for the NHS to support the diagnostic process and to assist in the ongoing monitoring and development of children with ASD.

2 M-Health Solutions to Support the Diagnosis Process

Digital Health is the umbrella term referring to the interventional and diagnostic technologies that depend on the use of ICT (Information Communication Technology) [12]. According to Thummler [12] the growth of the digital health industry has been rapid in recent years, facilitated by a move towards distributed patient-centered care (Fig. 1), where the need for physical contact between patients and professionals is reduced. The EU Horizon 2020 Research and Innovation programme calls for research into improving ability to “monitor health and to prevent, detect, treat and manage disease” (European Commission, 2015) and spending on ICT in healthcare is expected to grow to €18 billion globally by 2017 [12].

Fig. 1.
figure 1

(Source: Thummler, 2015) [12]

Distributed Patient Centred Care in the 21st Century

Full size image

M-health (Mobile Health) refers more specifically to mobile computing, medical sensor and communication technologies for health [13]. According to Istepanian (2010) [14], m-health was first introduced in 2003 and has since been of increasing importance in key areas of healthcare, including wellbeing, disease management and diagnostics.

Mobile technology, particularly over the past ten years, has become commonplace in the everyday lives of the general populous with 93 % of UK adults owning a mobile phone, around 7 out of 10 of which are smartphones [15, 16]. Similarly, the average price of tablet computing device has dropped from £400 in 2011 to £179 in 2014 resulting in an increase in the UK households owning a tablet from 22 % in 2012 to 55 % in July 2014 [16].

Given the popular nature of mobile technology, there is growing interest in the role of mobile devices in healthcare initiatives [17, 18]. Tablet computers and mobile phones present a potentially affordable technological solution to epidemiological and health-based challenges, particularly in the context of a global move towards distributed patient-centred healthcare models [12, 19]. As part of a broader initiative to improve digital health provision, m-health can play a role in bridging a health workforce gap where resources are stretched [20] such as the United Kingdom’s NHS and, with the proliferation of mobile technology in developing countries, improve access to healthcare globally [20, 21].

Whilst many children enjoy interacting with technology, children on the autism spectrum often feel a strong affinity for computing technology, as it is seen as a safe environment to learn and practice skills that may be difficult in everyday life [22]. Children with ASD may experience difficulty with cooperative play with other children and often prefer their own repetitive activities [23], e.g. playing with specific games or apps for extended periods or in a repetitive manner. User-behaviour analysis software on mobile computing devices, therefore, represent an interesting opportunity to collect rich data that can be analysed in order to look for signs of possible ASD.

Whilst laptop or desktop computers have typically relied on peripherals such as traditional keyboards, mice or click-pads for user interaction, tablet computers and smart phones typically rely on touch-screen technology. With the advent of touch-screen technology, young children are now able to independently interact with computing devices from a younger age than might have previously been possible.

Each user’s interaction with a touchscreen device is unique and touch-usage based profiling can reveal significant information about a user, including gender and age [24]. Consequently, there has been considerable interest in the role that mobile technology can play in improving understanding and supporting the educational and communication needs of those on the autism spectrum [25–28]. There is also an emerging area of research and a number of recent initiatives in using mobile technology to collect evidence to support the diagnostic process and the ongoing observation of autism in children, with a view to identifying support strategies and managing their effectiveness [19, 29, 30]. Sensors contained within mobile devices can be a more affordable alternative to expensive eye-tracking software often used in assessing communication skills of those with or who are suspected of having autism [19, 31].

Tablet computers and mobile phones not only serve as communication devices, but also recreation devices and are packed with embedded sensors such as gyroscopes, GPS, accelerometers and touch-screens [32]. In combination with access to user’s calendars, contacts and other personal information, mobile computing devices can infer where their users are and what they are doing, potentially providing rich contextual information for the purposes of user-behaviour analysis.

Such m-health solutions provide the advantage of being able to collect evidence of behaviours suggesting the presence of an autism spectrum disorder at home and within the child’s natural environment [33]. Tablet computers are also increasingly used within primary school education settings, widening the scope to collect data outside of the clinician’s office. Data could then be uploaded to the clinical teams via a secure, bidirectional wireless communication link (Fig. 2). The ability to access such data in advance of appointment times could facilitate the clinician in planning the agenda for an appointment and planned observations.

Fig. 2.
figure 2

Example of child to healthcare worker software

Full size image

User-behaviour analysis software can be acquired as off-the-shelf or bespoke software packages or could be created using open-source software such as Funf [34].

The Funf framework consists of a number of basic data collection objects that collect data on user interaction with a smartphone or tablet computer, including GPS, location, accelerometer, browser history, running apps and screen on/off sate. The modular architecture of Funf also enables the addition of new ‘probes’ by 3rd party app developers [34]. Such software could be used as a platform on which to develop user-behaviour analysis software relatively quickly. It would require, however, some level of technical expertise, which may be lacking or expensive to obtain for health care providers.

One notable example of off-the-shelf user-behaviour analysis software is the Play.Care Apple iPad application developed by Harimata [35] in conjunction with researchers at various academic institutions across Europe. According to Harimata [35] the child begins the assessment by playing a set of two educational games which have been designed to “encourage motor, social, and cognitive behaviours”. During the session, the software gathers data in relation to the user’s behaviour using the device’s touch-screen and motion sensors, including the accelerometer and gyroscope. Through the application of machine learning algorithms, a risk assessment of autism is conducted based on comparisons with children in which the condition has already been identified. Harimata claim that the algorithms underpinning their tablet-based software, which principally aims to identify the possible risk of autism through movement pattern analysis, achieved 90 % accuracy in differentiating behavioural patterns related to autism from those in typically developing children [35]. Potential risks associated with this approach, however, might include a lack of engagement with the specific games the user is required to play, particularly given the restrictive and repetitive nature of play behaviours that children on the autism spectrum might display. Nevertheless, the Play.Care application represents an interesting and novel application of machine learning techniques in supporting the diagnosis process.

M-health solutions can also guide the user through the process of collecting evidence of behaviours using a smartphone video camera. NODA smartCapture [33] is a mobile phone-based application that enables parents to record clinically relevant prescribed video evidence of their child’s behavior. This approach guides the parent through the process based on the clinical needs of the healthcare professional. According to Nazneed et al. [33], it supports the recording and uploading of four, up to 10-min long naturalistic observation diagnostic assessment (NODA) scenarios, that were chosen based on pilot research on video-based diagnosis of autism. These scenarios include:

  1. 1.

    the child playing alone,

  2. 2.

    the child playing with a sibling or peer,

  3. 3.

    a family mealtime, and

  4. 4.

    any behavior that is of concern to the parent.

Working alongside NODA smartCapture, the NODA Connect web portal has been developed to allow healthcare professionals to access the child’s developmental history. This functionality facilitates remote diagnostic assessments by liking evidence of behaviours tagged in the videos to DSM criteria. This method still calls on the clinical judgment of the healthcare professionals. As already noted in this paper, however, the NHS in the UK relies on the World Health Organisation’s International Classification of Diseases (ICD) [5].

3 Conclusions

Identifying signs of autism spectrum disorders in young children is key to securing an early diagnosis and the accessing the subsequent support and interventions necessary to ensuring the best possible prognosis. In the absence of a single biomarker for autism, the diagnosis of ASD in a young child is still heavily reliant on a portfolio of evidence based on clinical observations made by healthcare professionals, parental reporting and the feedback of educational professionals. This process can be difficult where there is a risk of parental bias in terms of reporting and observations and where there are lengthy gaps between appointment times with health care professionals, particularly within a stretched health service. Within the NHS, the various teams involved in the care and monitoring of a child suspected of having an ASD may be working to different agendas and experiencing unique pressures making it difficult to harmonize the process across different departments.

Whilst there are m-health solutions emerging to assist in the diagnosis and ongoing monitoring of autism in young children, there are also limitations associated with these approaches. In order for these software products to support the NHS, it is vital that the user-requirements elicitation and modelling processes effectively capture the unique and evolving needs of the various professionals working within a dynamic organization such as the NHS. This will also ensure that software can evolve to reflect changes in our understanding of ASDs. M-health solutions, however, do present an interesting opportunity for health care professionals to make observations of children between appointment times and within their home environment or familiar education setting, thus potentially speeding up the diagnosis process.