The scope of health technology is vast – spanning so many different conditions, professionals and technologies, digital health is always being used in new and innovative ways.
In this feature, we’ll take a look at the some of the ideas, analyses and projects from different countries, showing the breadth of work taking place in health tech across the globe. We cover how facial recognition tech could be used in healthcare, analysis of electronic clinical decision support in Tanzania, national digital mental health support, a digital prevention in Sweden, and a use of artificial intelligence in Korea.
Evaluating facial recognition tech for opioid substitution in Australia
Authors Sandeep Reddy, Max Mito and Mark Feldschuh contributed an article to the International Journal of Digital Health entitled ‘A Qualitative Evaluation of the Use of Facial Recognition Technology for Opioid Substitution Treatment in Community Pharmacies’.
The authors highlight how opioid-related issues including addiction are “on the rise” in Australia, with opioid substitution therapy (OST) used in attempts to treat addiction and reduce opioid-related morbidity and mortality. Due to the associated risks that OST can bring, it is currently coordinated through a permit system in Victoria state, requiring appropriate identification of patients before doses are administrated. However, “it has been identified informally that some community pharmacies have issues in dispensing OST because of less than robust patient verification processes.”
Facial recognition technology (FRT) is “now increasingly being incorporated into healthcare systems and processes including check-ins, patient matching, and even in medical diagnosis,” the authors write. They explain how in most cases, FRT “creates a template of the recipient’s facial image and then compares it to pre-existing facial images. FRT uses artificial intelligence (AI) algorithms to identify distinct details of the person’s face. These details can be distance between eyes or shape of the chin. The process commences with detection of the recipient’s face followed by extraction of patterns from the image. These details are then converted into a mathematical representation and compared to the image database. In instances, where relevant facial images of the target are not available in the database, a probability match score between the unknown person and existing templates or a message conveying inability to recognise the target is presented.”
There is promise for its use in patient verification, the authors note, with reduction of the risks presented by incorrect identification. However, “there is growing concern about privacy and rights that may be compromised” which has “led to resistance from citizens”, and as such results in uncertainty about the value of FRT in healthcare.
The researchers employed the ‘Integrated Model of Evaluation’ (IMoE), with main components consisting of programme theory, context, intervention, change and outcomes; each component is assessed separately and brought together to form a thorough assessment.
Through this framework, results showed that participants had varying degrees of involvement and experience when it came to their knowledge of FRT, with some expressing a good understanding of how FRT was generally used such as in security or surveillance and the reason for its use in OST. Results also showed that the use of FRT “streamlined” the OST programme, improved accuracy of identification, supported adherence, reduced medication error and deceptiveness in patients; however, some pharmacies stopped using it as they felt it slowed down the process, and it was “found redundant” in instances where pharmacies had a very small number of OST clients, all of whom were familiar to the pharmacist.
The authors conclude that in spite of some negative perception and resistance, “this evaluation showcases the potential benefits of using FRT in healthcare.” Acknowledging the small size of the study, they highlight that it still “provides indicators of where FRT can be used effectively”, such as improving business processes and reducing medication errors.
Citation: Reddy S, Mito M, Feldschuh M. A Qualitative Evaluation of the Use of Facial Recognition Technology for Opioid Substitution Treatment in Community Pharmacies. International Journal of Digital Health. 2022;2(1):5. DOI: http://doi.org/10.29337/ijdh.47
Analysis of electronic clinical decision and support system to improve antenatal care in Tanzania
For an article in Frontiers, authors Sandra van Pelt, Karlijn Massar, Laura Shields-Zeeman, John B. F. de Wit, Lisette van der Eem, Athanas S. Lughata and Robert A. C. Ruiter submitted ‘The Developing of an Electronic Clinical Decision and Support System to Improve the Quality of Antenatal Care in Rural Tanzania: Lessons Learned Using Intervention Mapping’.
In 2019, research from the World Health Organisation (Trends in Maternal Mortality, 2000 – 2017) showed that Tanzania is one of the countries with a “very high” maternal mortality radio, with 524 maternal deaths per 100,000 live births. With consequences of maternal death far-reaching, impacting women, children, families and communities, the authors state that “antenatal care (ANC) is a crucial intervention to promote maternal and child health and identify issues early during pregnancy.”
An electronic clinical decision and support system called the Nurse Assistant App was developed and implemented in Tanzania in 2016 in order to provide digital assistance to healthcare workers during antenatal care consultations; the Frontiers study seeks to systematically evaluate its development and implementation process with the aim of informing future plans for developing digital interventions.
Desk research took place, examining digital project archives, project documents, and content from the NAA team (the Women Centred Care Project, focusing on research in close collaboration with universities in the Netherlands and Tanzania). This research was combined with semi-structured interviews to analyse the development process of the Nurse Assistant App using the six steps of Intervention Mapping. This aims to provide systematic guidance to programme developers by examining needs assessment to specify causes and contributing factors to the health problem; behaviour and underlying personal determinants needed to change to reduce the problem; design of an intervention; pre-testing of intervention materials, activities and protocols; development of an implementation plan; and programme evaluation including effect and process evaluation.
The researchers found that five of the six steps were completed during the development and implementation of the Nurse Assistant App, with tasks related to identifying theory-based behaviour change methods not accomplished.
Based on this, they conclude that “programme developers are recommended to (1) engage the community and listen to their insights, (2), focus on clear programme goals and the desired change, (3), consult or involve a behaviour change specialist, and (4), anticipate potential problems in unexpected circumstances.”
Citation: van Pelt S, Massar K, Shields-Zeeman L, de Wit JBF, van der Eem L, Lughata AS and Ruiter RAC (2021) The Development of an Electronic Clinical Decision and Support System to Improve the Quality of Antenatal Care in Rural Tanzania: Lessons Learned Using Intervention Mapping. Front. Public Health 9:645521. doi: 10.3389/fpubh.2021.645521
Self-screening and referral aid suggested for mental health support in India
Authors Aditya Agarwal, Pawan Kumar Gupta, Amit Singh and Sujita Kumar Kar contributed an article entitled ‘Conceptual model of Remotely Accessed Mobile Application for Psychiatric disorder Screening (RoaDMAPS): A self-screening and referral aid for mental health in India’ to The Lancet, published earlier this month.
The authors state: “With enhanced access to the internet, particularly on mobile devices, awareness and recognition of mental health as an essential part of an individual’s life is increasing. It also paved the way to overcome the stigma related to psychiatric disorders”.
They acknowledge that there are many health applications in existence that focus on mental health; however, “only a few provide a platform to screen oneself for a mental disorder” whilst maintaining good accessibility and reliable information. Some of the challenges shared for existing applications include user fees and lack of information about local mental health facilities.
“Thus,” the authors write, “we propose a simple conceptual model to counter the problems mentioned above by incorporating into the currently available and widely used mobile applications – standardised mental illness screening tools, comprehensive information about mental health facilities and available services, and channel to national online consultation platforms. This could be an avenue to provide a host of reliable, affordable, omnipresent, and stigma-free services. These apps will act as a single point to identify, screen, refer and facilitate care to the underserved population with mental illness. Also, the availability of information in one place will facilitate the endeavours towards the integration of services and continuity of care.”
They suggest that their conceptual app could be modified and used as a governmental health app in the public domain to increase recognition and trust. Alternatively, a previously-used government tool such as the COVID-tracking AaroguaSetu app could be upgraded to incorporate recommended services “as anyway its utility after the COVID-19 pandemic is going to be limited.”
Users could select a section of problems they wish to get screened for, such as mood, anxiety and childhood psychiatric disorders, and standardised tools for screening particular health conditions can be presented through simple questionnaires to provide the individual with the probability of having the disorder along with details of the nearest mental health facilities.
The authors conclude: “The strengths of the proposed model are its simplicity, utilisation of well-established resources, the minimum effort required to administer screening tools, cutting across stigma, and minimal cost to the users. However, research is needed, in a staged manner, to establish its validity and applicability.”
The article can be found in full here.
Observations on reaching the older population with a digital falls intervention in Sweden
‘Reaching Older People With a Digital Fall Prevention Intervention in a Swedish Municipality Context – an Observational Study’ was published to Frontiers by authors Sandra Bajraktari, Magnus Zingmark, Beatrice Pettersson, Erik Rosendahl, Lillemor Lundin-Olsson and Marlene Sandlund.
“In Sweden,” the authors write, using 2020 figures from The Public Health Agency of Sweden, “fall-related injuries cause the highest number of deaths, hospitalisations, and visits to emergency services among older people.”
In the study, researchers aimed to evaluate the reach of the Safe Step digital fall prevention exercise intervention in people aged 70 and above, living in their own homes in a Swedish municipality. Safe Step is a self-managed programme developed in co-creation with older people, supplying a year’s exercise programme plus monthly educational videos and optional supportive strategies such as introductory drop-in meetings, technical support and group exercise programmes.
Participants were recruited between October 2019 and April 2020 through a variety of digital strategies including Facebook, websites and TV screens, and non-digital strategies including bus advertisements, a postal brochure and oral presentations.
Despite the range of strategies used, the authors note that the recruited sample predominantly consisted of women who “used the internet or applications on smart technologies almost daily”, with many reporting high levels of education, living alone and “better functional performance”.
They state: “Consistent with the literature, participants with higher education are more likely to participate in research studies in general and those with health literacy proficiency are more likely to participate in digital interventions,” the authors state. This means that the recruited cohort are “not entirely a representative sample of the population of focus.”
According to previous studies shared by the researchers, “older people represent the fastest growing group of internet users in Sweden… however, this group lags behind all other groups and are at the highest risk of digital exclusion.”
Ultimately, they conclude, “A variety of digital and non-digital interventions complemented with a variety of recruitment strategies is likely needed to reach a larger and more diverse group of older people with different needs.”
Citation: Bajraktari S, Zingmark M, Pettersson B, Rosendahl E, Lundin-Olsson L and Sandlund M (2022) Reaching Older People With a Digital Fall Prevention Intervention in a Swedish Municipality Context—an Observational Study. Front. Public Health 10:857652. doi: 10.3389/fpubh.2022.857652
Use of artificial intelligence in small bowel capsule endoscopy in Korea
Authors Dong Jun Oh, Youngbae Hwang, Ji Hyung Nam and Yun Jeong Lim submitted a study entitled ‘Small bowel cleanliness in capsule endoscopy: a case-control study using validated artificial intelligence algorithm’ to Nature.
“Small bowel capsule endoscopy (SBCE) is currently the key modality for diagnosing various SB diseases, such as obscure gastrointestinal (GI) bleeding, known or suspected SB Crohn’s disease without stenosis, small bowel tumours or polyposis, and refractory celiac disease,” the authors write. Therefore, it is important that the bowel is properly prepared; however, “the consensus on the timing and method of bowel preparation for SBCE is still controversial.”
Recently, they continue, a study using an artificial intelligence (AI) algorithm trained by PillCam images was conducted, with the validated AI algorithm calculating an objective and automated SB cleanliness score for the full-length SB images.
In their own study, the researchers used a validated AI algorithm and compared SB cleanliness scores between SBSE immediately after colonoscopy, and SBCE alone. “So,” they state, “we decided to identify whether colonoscopy before SBCE affected the difference in SB cleanliness scores.”
The AI was trained using a convolutional neural network (CNN) algorithm, which “consists of several convolutional layers and pooling layers. In each convolutional layer, different features of the image are extracted to identify the image.” In this algorithm, each SBCE image was calculated according to a five-step scoring model, from a cleanliness score of one (meaning mucosal visualisation was less than 25 percent) to a score of five (mucosal visualisation is 90 percent or more).
The researchers found no significant difference in mean cleanliness calculated by the AI algorithm between the two groups, describing how this confirms “that small bowel cleanliness was adequately maintained in SBCE immediately after a colonoscopy, similar to that in SBCE alone… however, in poor colon preparation, SB cleanliness score was not as good as in fair, good and excellent colon preparation. So, if the colon preparation is good, SBCE can be performed immediately after colonoscopy, but, if colon preparation is not good, additional bowel preparation before SBCE may be necessary.”
Citation: Oh, D.J., Hwang, Y., Nam, J.H. et al. Small bowel cleanliness in capsule endoscopy: a case–control study using validated artificial intelligence algorithm. Sci Rep 12, 18265 (2022). https://doi.org/10.1038/s41598-022-23181-1
