The “Preamble to the Convention on the Rights of Persons with Disabilities and Optional Protocol United Nations” considers disability as a consequence of the interaction between the impairments and factors such as environmental and personal. A conventional disability evaluation is in line mainly with the medical model of disability. It states the percentage of permanent physical impairment (PPI) and does not accurately depict the individual’s functioning. This apparent flaw lies in the fact that a PPI calculation does not consider the relevant environmental and personal factors, as elaborately mentioned in the International Classification of Functioning Disability and Health (ICF). Recent research on disability highlights how barriers, such as environmental and personal, have a significant impact on disability. In instances where the matter is under judicial consideration, such as compensation due to an accident, the awarded sum is based on functional disability (FD).
ICF by the World Health Organization (WHO) can be a useful tool that can combine the physical impairments with environmental and functional factors to close the gap between the medical and social models by using core sets and evaluating FD. However, its use will need augmentation by technology, given certain shortcomings. The use of ICF entails a lot of options where subjectivity can creep in. Besides, the measurement of environmental and personal factors is resource-consuming. The need of the hour is to embrace technology, such as telemedicine and artificial intelligence (AI), to make life easier for all the stakeholders, such as the medical professionals, the judiciary, and persons with disability. This article intends to be informative regarding ICF’s use augmented by telemedicine and AI in the evaluation of FD and its anticipated benefits in addressing the existing workflow problems.
Existing workflow and the need for reforms
The person requiring a disability assessment presents to a designated medical board. After evaluation and examination, the committee awards a PPI based on the prevailing guidelines. This evaluation does not take into account the personal and environmental factors affecting the applicant’s daily life. Instances of disability assessment by doctors to benefit the claimant are rampant. On the other hand, mistakes due to human error are also commonplace. The calculation of FD is restricted to the domain where compensation is involved. The judicial tribunal invariably asks for evidence to arrive at FD’s level in cases where compensation is involved. Other benefits are still accorded based on the percentage of impairment arrived at by medical boards.
AI refers to the use of computer technology to simulate intelligent behavior and critical thinking comparable to humans. AI programs have conquered chess; however, fully automated disability evaluation may require some more time. It will require the requisite hardware of various dimensions to accommodate the diversity that exists amongst human beings.
The concept of a functioning profile
Filling up the ICF creates a profile similar as depicted in [Figure 1]. This profile is a hypothetical one with only one parameter chosen to simplify the concept. It explains how ICF helps to arrive at an FD level. It covers most domains in routine life; however, it leaves the rating to the person evaluating the disability. There needs to be a consensus amongst various members amongst the authorities evaluating disability regarding how we would grade different ICF domains. Training data for AI requires agreement on how grading of parameters will happen, irrespective of the location. AI can objectively replicate the same.
The suggested workflow
The suggested workflow is that the person is supposed to apply for the certificate on the website, which will first ascertain whether a benchmark disability exists or not. A smart assistant along with a video analysis will help in making this decision. A webcam will be set-up with pre-specified questions and a pre-specified protocol with instructions. Video guides and instruction manuals provided to evaluate disability will help to conclude with adequate confidence whether the patient has a benchmark disability or not. An appellate authority will take care of any appeals. In case it does qualify as a benchmark disability, AI fills WHO’s ICF core set to create a functioning profile; uses telemedicine to measure capacity and performance, which may depend on the environment and the societal norms existing; assign weights to various parameters to arrive at a percentage. Supercomputers will collate the data and issue an accurate depiction of FD after giving appropriate weights to the functioning profile generated.
This workflow can be adopted in compensations, and other spheres such as disability pensions, reservations in educational institutions, reservations in government jobs, etcetera [Figures 2 and 3].
There can be three lower limb amputees with the same degree of PPI but with different FD as per the existing workflow. The reason could be as simple as age and vocation. The proposed system will consider many environmental and personal factors with weights ascribed to each based on the AI algorithm to arrive at an accurate functional level.
Existing tools, software and hardware
AI and machine learning (ML) are encompassing every sphere of life. The requirement of complicated hardware at the point of contact (edge of the system) is minimized while the cloud is bearing a significant workload. After being released in 2010, the Microsoft Kinect™ has become a research interest as an affordable, miniaturized, and easy to use tool for markerless motion tracking. It has been used in several musculoskeletal and neurological conditions such as Parkinson’s disease, spinal cord injury, stroke, limited shoulder range of motion to evaluate physical impairment. Tools such as Lokomat™ and ReWalk™ are exoskeletons that have been used for motor evaluation for lower limb with success. Other categories of impairments like speech and language, vision, hearing, cognitive and mental retardation are relatively easy to be processed by AI systems. High-fidelity microphones, headsets, and online questionnaires are the only required hardware at the edge, and these are readily available. Multiple studies have shown encouraging findings. Aira™ is a technology used for the assessment of vision as well as rehabilitation. ML, deep learning (DL), and natural language processing (NLP) are being used for cognitive and mental impairments. Audiology is using advanced technology already for assessment and rehabilitation.
Level 2 AI workflow
Examination is done by the doctors in Level 1 AI, even though it may be through telemedicine. However, the objective is to automate disability evaluation in Level 2 AI entirely. It will need exoskeletons, robots, motion trackers, and the latest hardware and the technology used in Level 1 AI. The benefits of a disability analyzer are numerous, starting from a straightforward application system, accessibility to the closest disability analyzer, reproducibility of results throughout the country as the machines would be the same, and linkage of all disability analyzers [Figure 4].
Any AI system will require an adequate representation of the various types and degrees of disability. The process will require training until it starts matching and eventually doing better than medical boards requiring human beings. Three phases are involved in the actual process of AI teaching itself: training, validating, and checking. By feeding data into the computer system, it is trained to generate a specific prediction with each loop.
What is essential is that the training data be free of bias, and this can happen when people training these models; keep this in mind as the model will work in a biased manner if not trained without a partial data set. The model will reproduce that behavior trained on data that has a bias in it. AI-based decision-making approaches have started being used in situations where experts often disagree. Disability evaluation is undoubtedly one category where experts often disagree. Regulatory issues, economics, and liability also need to keep in mind.
Primary care physicians
Primary care physicians will play a major role if this shift in assessment is to be made successfully. These physicians are usually the first point of contact for persons with disabilities. The proposed automated system will enhance the accessibility to certification of disability, and aid physicians in guiding the patient through the process of assessment. Constraints such as difficulty in traveling to the nearest assessment board, financial implications of such a journey, and the absence of availability of experts across the country and especially in rural areas pose a unique challenge in the assessment of disability in India.
The new technology will come at a price, but this should not deter us as constituting medical boards have their challenges and economic issues, and it would be wrong to compare the two. Any new technology will require widespread acceptance. It is essential to mention the problems that AI can reinforce.
A potential source of bias arises if the training data lacks an adequate representation. It was commonly seen in face analytics systems with much higher error rates for black women than whites.
The benefits to the patient, the medical fraternity, the judiciary, and the society are immense and summarized [Table 1].
Persons with disabilities form a sizable population as per WHO estimates. This article discusses how AI can help benefit the medical and the legal system, persons with disabilities, and society. If we are not consistent with the degree of disability, we cannot deliver justice in the real sense.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
1. Convention on the Rights of Persons with Disabilities and Optional Protocol, United Nations. Preamble 1. Available from: https://www.un.org/disabilities/documents/convention/convoptprot-e.pdf
Last accessed on 2021 Feb 22
2. Charlton J. Nothing about us without us:Disability, oppression and empowerment. Berkeley, University of California Press 1998. Nothing About Us without Us by James I Charlton-Paperback-University of California Press Available from: ucpress.edu
Last accessed on 2021 Feb 22
3. McConachie H, Colver AF, Forsyth RJ, Jarvis SN, Parkinson KN. Participation of disabled children:How should it be characterized and measured? Disabil Rehabil. 2006;28:1157–64
4. Oliver M. The Politics of Disablement:A Sociological Approach St Martin's Press USA 1990;1:4–5
5. Purshotam Dass vs New India Asso. Co. Ltd. and Ors. on 8 April, 2011 Supreme Court of India. Purshotam Dass vs New India Asso. Co. Ltd. and Ors. on 8 April, 2011 Available from: indiankanoon.org
Last accessed on 2021 Feb 22
6. Silver D, Hubert T, Schrittwieser J, Antonoglou I, Lai M, Guez A, et al A general reinforcement learning algorithm that master's chess, shogi, and Go through self-play. Science. 2018;362:1140–4
7. Ruda D, Einarsson G, Andersen ASS, Matthiassen JB, Correll CU, Winge K, et al Exploring movement impairments in patients with Parkinson's disease using the Microsoft Kinect sensor:A feasibility study. Front Neurol. 2021;11:610614
8. Tran H, Nguyen KD, Pathirana PN, Horne MK, Power L, Szmulewicz DJ. A comprehensive scheme for the objective upper body assessments of subjects with cerebellar ataxia. J Neuroeng Rehabil. 2020;17:162
9. iÁlvarez I, Latorre J, Aguilar M, Pastor P, Llorens R. Validity and sensitivity of instrumented postural and gait assessment using low-cost devices in Parkinson's disease J Neuroeng Rehabil. 2020;17:149
10. Lee YM, Lee S, Uhm KE, Kurillo G, Han JJ, Lee J. Upper limb three-dimensional reachable workspace analysis using the Kinect sensor in hemiplegic stroke patients:A cross-sectional observational study. Am J Phys Med Rehabil. 2020;99:397–403
11. Scano A, Molteni F, Molinari Tosatti L. Low-cost tracking systems allow fine biomechanical evaluation of upper-limb daily-life gestures in healthy people and post-stroke patients. Sensors (Basel). 2019;19:1224
12. Scano A, Chiavenna A, Malosio M, Molinari Tosatti L, Molteni F. Kinect V2 implementation and testing of the reaching performance scale for motor evaluation of patients with neurological impairment. Med Eng Phys. 2018;56:54–8
13. Lee S, Lee YS, Kim J. Automated evaluation of upper-limb motor function impairment using Fugl-Meyer assessment. IEEE Trans Neural Syst Rehabil Eng. 2018;26:125–34
14. Napoli A, Glass SM, Tucker C, Obeid I. The automated assessment of postural stability:Balance detection algorithm. Ann Biomed Eng. 2017;45:2784–93
15. Gritsenko V, Dailey E, Kyle N, Taylor M, Whittacre S, Swisher AK. Feasibility of using low-cost motion capture for automated screening of shoulder motion limitation after breast cancer surgery. PLoS One. 2015;10:e0128809
16. Maggioni S, Melendez-Calderon A, van Asseldonk E, Klamroth-Marganska V, Lünenburger L, Riener R, et al Robot-aided assessment of lower extremity functions:A review. J Neuroeng Rehabil. 2016;13:72
17. Fuente GS de la, Ritchie CW, Luz S. Artificial intelligence speech and language processing approaches to monitoring Alzheimer's disease:A systematic review. J Alzheimers Dis. 2020;78:1547–74
18. Nguyen BJ, Chen WS, Chen AJ, Utt A, Hill E, Apgar R, et al Large-scale assessment of needs in low vision individuals using the Aira assistive technology. Clin Ophthalmol. 2019;13:1853–68
19. Graham SA, Lee EE, Jeste DV, Van Patten R, Twamley EW, Nebeker C, et al Artificial intelligence approaches to predicting and detecting cognitive decline in older adults:A conceptual review. Psychiatry Res. 2020;284:112732
20. Nkyekyer J, Meyer D, Blamey PJ, Pipingas A, Bhar S. Investigating the impact of hearing aid use and auditory training on cognition, depressive symptoms, and social interaction in adults with hearing loss:Protocol for a crossover trial. JMIR Res Protoc. 2018;7:e85
21. Sanchez LR, Fereczkowski M, Neher T, Santurette S, Dau T. Robust data-driven auditory profiling towards precision audiology. Trends Hear. 2020;24:2331216520973539
22. Lakhani P, Sundaram B. Deep learning at chest radiography:Automated classification of pulmonary tuberculosis by using convolutional neural networks. Radiology. 2017;284:574–82
23. Altman R. Artificial intelligence (AI) systems for interpreting complex medical datasets. Clin Pharmacol Ther. 2017;101:585–6
24. Angwin J, Larson J, Mattu S, Kirchner L. Machine Bias:There's software used across the country to predict future criminals. And it's biased against blacks 2016 Available from: https://www.propublica.org/article/machine-bias-risk-assessmentsin-criminal-sentencing
Last accessed on 2021 Feb 22