SUMMARYChildren with an autistic spectrum disorder (ASD) are at significantly increased risk of overweight/obesity. Exercise interventions have been attempted, but health outcomes are challenging to measure in this population. This research proposes to develop a platform for non-medical-device-based access to heart rate and heart rate variability (HRV), robust measures of cardiovascular health. The platform is designed to record seamless, zero-participation, and laboratory-accurate measurements, with negligible setup time.TEXTOverweight/obesity and autistic spectrum disorder (ASD) are both pediatric health concerns throughout the developed world. ASDs describe a family of developmental disorders, diagnosed during childhood, which encompass a variety of difficulties with social interaction/communication, motor impairment, attentional deficits, aggression, self-injury, and characteristic repetitive or restrictive behaviours; they occur more frequently in boys (1 in 42) than girls (1 in 189) (CDC, 2017).There is a substantial cross-over between the symptomatology of ASD and increased bodyfat/mass. ASD children are more likely to demonstrate food selectivity, show a strong preference for nutrient-dense foods, are less physically active, and often suffer from behavioural problems such that typical pediatric exercise interventions are not tenable. In addition, the prescription of anti-psychotic medication, reasonably common in the ASD population, is a strong predictor of weight gain \cite{Srinivasan_2014}.Consequently, the highest recent estimate describes a substantial increase in risk for both overweight (autism, OR = 2.24, Asperger Syndrome = 1.49) and obesity (autism, OR = 4.83, Asperger Syndrome, OR = 5.69) over control children \cite{Broder_Fingert_2014}. This represents a burgeoning public health problem, as obesity is associated with both short- and long-term health problems such as cardiovascular disease, hypertension, osteoarthritis, various forms of cancer, gallbladder disease, sleep apnoea, or a combination of the above (i.e. metabolic syndrome) \cite{Bhaskaran_2014,Poirier_2006}. Researchers and clinicians have thus been strongly urged to develop and implement effective exercise interventions for this population \cite{Srinivasan_2014}.Some interventions have previously taken place within this population i.e. they have have administered an exercise intervention to a pediatric/young adult ASD population numbers \cite{Pitetti_2006,Lochbaum_2003,Pan_2011,Hinckson_2013,Hayakawa_2011}. This is a small and heterogenous body of literature; each study addresses a different exercise modalities and includes different outcomes. Overall, the work is of poor quality (most did not conduct follow-up, small sample sizes/no control group, no indication of the details of diagnosis, etc.) While all report improvements in various instantiations of physical health; cardiovascular health, musculoskeletal strength, coordination, gait pattern, walking endurance, gross motor skills, etc. the area is entirely without standardization.This proposal is to develop a tool which changes the answer to the question "how should exercise interventions in ASD populations be assessed?" All measurements outlined in the study above occur before and after study participation, within-subjects. A wearable, smart-device or ubiquitous computing model replaces these endpoint measurements with chronic instrumentation, and offers some significant advantages for measurement fidelity:single measurements are subject to motivation and performance bias, repeated measurements are substantially more robust to individual measurement anomaliessingle measurements have rigidly defined start and endpoints which require significant coordination for participants and experimenters, repeated measurements do notsingle measurements mean small-sample studies are invariably underpowered, as they provide two timepoint scores and one change score, repeated measurements are robustly powered even within individual participantsThe above are of particular significance when we consider a small, hard-to-recruit clinical population. Consider \citet{Hinckson_2013}, where n=17 autism/intellectual disability participants underwent 10 weeks of bi-weekly exercise training, whose effectiveness was assessed by a common exercise tolerance test (6 Minute Walk Test; 6MWT). This gives n=34 assessment points from a test whose reliability due to motivation and learning effects has repeatedly been questioned \citep[see][]{Hamilton_2000}. A measurement of cardiovascular health or capacity antecedent to each exercise session would return n=340 (17*10*2) comparison points, a similar daily measurement would return n=1190 (17*70). Even if these individual measurements have a reduced observational capacity, are occasionally erroneous, or are not recorded, their overall effect is to provide a massive within-subjects comparison, and are obviously more powerful.In addition, exercise interventions are generally considered to be independent variables - overwhelmingly, they are set and parameterised at the beginning of the study into pre-defined intervention delivered in pre-defined blocks over time. This is easily administered in a sports science context, however, a non-cooperative population with ASD or similar intellectual disability (ID) may not find such a regime stimulating or tolerable. Previous research into ASD/ID has recommended mixed-method or mixed-mode interventions \cite{Bandini_2015}; he best-case practice administration of an exercise and lifestyle modification program drawn from traditional pediatric obesity interventions to meet individual cognitive and learning needs \cite{Fleming_2008}. Monitoring from a ubiquitous computing model allows any adherence to an intervention, or any modality of that intervention (be it exercise, dietary or lifestyle change), to be a dependent variable - we can instead calculate dose-response within set parameters, rather than treat non-cooperation as incomplete data.Wearable and ambulatory approaches have shown a great deal of promise in making physiological observations with an ASD population outside of the context of health improvement \cite{Prince_2016,Levine_2014,Goodwin_2016,Kientz_2013}, but have their own drawbacks: they frequently require adherence to the skin or body and can be poorly tolerated by sensitive populations, they require compact power sources/housing and algorithms, and are not considered to be as reliable as analytical devices. However, a related ubiquitous computing approach to physiological measurement (where a common household item, space or similar is instrumented with a measurement device) solves these problems - generally, devices can be of analytical quality, are not subject to space restrictions, and take measurements requiring only the most minimal engagement from experimental participants.Specifically, this proposal seeks to validate the design of a chair with large contact electrodes on the armrests, adjustable to hand position, used to record heart rate. These contact electrodes are 12" 316 stainless globes/orbs which are connected to (a) the two lead wires of an electrocardiogram, and (b) a capacitive-type circuit which essentially couples 'stray noise' through the body with a signal to the ECG (Figure 1). Essentially, if a participant sits in the chair and any skin surface is applied to both globes, an ECG records automatically to a stored device. Similar instrumented devices are reasonably common, and are relatively simple to deploy - this device has switching and recording functions similar to previous designs which required bilateral contact from the human body over multiple electrodes and measure physiological signals, including a toilet seat \cite{Huang_2012}, steering wheel \cite{Matsuno_2016} and office chair \cite{Yong_Gyu_Lim_2006}.