The role of digital technologies in health and long-term care sectors

HEALTH AND LONG-TERM CARE WORKFORCE 87 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS 4. THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS Emilio Gómez-González, Emilia Gómez, Songül Tolan, Fernando Martínez-Plumed, Annarosa Pesole, Enrique Fernández-Macías, José Hernández-Orallo, Emilia Gómez, Alessandra Conte, Sara Grubanov Boskovic The role of digital technologies in health and long-term care sectors has been manifold. Digital technologies have, amongst others, transformed the delivery of health and care services and generated new type of professional roles and skillsets, therefore affecting both the demand for and the supply of healthcare workers. However, the full implementation of digital technologies – and its potential effect on health and care workforce – remain closely related to numerous ethical, social and labour market aspects. Given the broad scope of the topic, this chapter focuses on three specific thematic areas. The first contribution aims to provide an overview of the main applications brought to the healthcare sector by Artificial Intelligence (AI), together with their benefits and challenges. KEY MESSAGES Currently, the spectrum of AI applications in healthcare is extremely broad: it ranges from applications with high technology availability value, such as algorithms for computer-aided diagnosis or imaging tools, to applications that are still immature, such as mind reading or whole-brain simulation. Three groups of ethical and social aspects related to AI in the healthcare sector can be distinguished: 1) issues such as data privacy, fairness or human oversight which have been broadly discussed in the context of general AI; 2) issues of particular relevance in medicine and healthcare, but also common in other domains, e.g. transparency, the required updates in evaluation, benchmarking and legislation; 3) controversial aspects not yet considered in other fields, e.g. ethical guidelines related to self-experimentation medicine. HEALTH AND LONG-TERM CARE WORKFORCE 88 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS COVID-19 has emphasised both the opportunities and ethical challenges of the use of AI in medicine, bringing an increased interest for the public. Given the strong implications of health data-related AI systems and the overlap with public health policies, an analysis of opportunities and risks has to be carried out before systems are fully deployed. The second contribution aims to assess the impact of AI progress on four specific health-related occupations: medical doctors; nursing and midwifery professionals; paramedical practitioners; and medical and pharmaceutical technicians. KEY MESSAGES Overall, the findings on AI research intensity suggest high activity in AI areas that contribute to abilities dealing with things and ideas and low activity for abilities dealing with people. In particular, medical doctors are the category most exposed to AI. The majority of AI exposure is driven by its impact on tasks that require abilities dealing with ideas (e.g. comprehension, attention and search as well as conceptualisation). On other hand, little AI exposure can be expected through basic processing abilities (e.g. visual processing or auditory processing) or through abilities that deal with people (e.g. modelling and social interaction or communication). AI could also play a novel role in the context of technology-driven labour market polarisation, depending on whether AI exposure is labour-replacing or labourenhancing. In the labour-replacement scenario, it could lead to unpolarising effects and a reduction in income inequality; whereas in the labour-enhancing scenario, it could imply an expansion of productivity for high-skilled occupations, potentially leading to occupational upgrading effects and an expansion of income inequality. The third thematic contribution aims to analyse how the divide in internet access and digital skills within elderly Europeans poses a barrier to the implementation of telemedicine. KEY MESSAGES Digital technologies, such as telemedicine, have great potential to improve the population’s access to health and LTC. However, the digital divide within certain socio-demographic groups remains considerable in the EU to the point of becoming a barrier to the implementation of telemedicine. The findings show that strengthening the potential of telemedicine among elderly people requires additional efforts in promoting digital inclusion, especially for elderly people living alone in their homes, the elderly with a low level of education and those living in rural and remote areas. Moreover, particular attention should HEALTH AND LONG-TERM CARE WORKFORCE 89 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS also be paid to bridging the digital divide between elderly men and women, the latter group having lower percentages of internet use and digital skills than men. Internet access is also a social and economic affordability issue. For example, only one third of the EU’s population aged 80+ who lives in rural and remote areas owns a computer. At the same time, almost all households in the EU own a telephone, which opens up the possibility of strengthening the mobile healthcare practices among the EU’s elderly. 4.1 ARTIFICIAL INTELLIGENCE IN MEDICINE AND HEALTHCARE: SOCIAL IMPACTS AND CHALLENGES Artificial intelligence (AI) has become a technological domain of strategic importance and a key driver of economic development in all sectors. In the domain of medicine and healthcare, AI is giving rise to new applications, paradigms – even defying the traditional roles of doctors and patients – and risks. Here we present a brief overview of the main applications brought to medicine and healthcare by AI, together with their benefits and challenges. We then introduce these issues in the context of the COVID-19 health emergency. Lastly, we illustrate some specific, very recent examples of AI systems in medicine and healthcare related to demography and migration, and, from the questions that arise: the risk of data bias and the dual potential of communication and conversational platforms and of systems for border control. AI IN MEDICINE AND HEALTHCARE: BENEFITS AND CHALLENGES Gómez-González & Gómez (2020) provide an updated review of the current and future applications of AI in the area of medicine and healthcare based on an analysis of state-of-the-art research and technology, including software, personal monitoring devices, genetic tests and editing tools, personalised digital models, online platforms, augmented reality devices and surgical and companion robotics. Figure 24 presents a ‘visual overview’ of the Gómez-González & Gómez (2020) review, including well-established applications such as the use of algorithms to support medical diagnosis, robots in surgery or conversational platforms (‘chatbots’) for patient assistance. In Figure 24, the different applications are assigned a Technology Availability Level (TAL) scale, presented in Table 5. The TAL provides a qualitative description of the degree of availability of a technology in a numerical scale in 10 steps (levels), ranging from 0 (unknown status, not considered feasible) to 9 (available for the general public). We observe in Figure 24 that applications such as algorithms for computer-aided diagnosis or imaging tools have a high TAL value, while others such as mind reading HEALTH AND LONG-TERM CARE WORKFORCE 90 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS or whole-brain simulation are still immature according to this scale. The TAL scale is similar in format (and related) to the standard ‘Technology Readiness Level’ (TRL) scale commonly used to assess research and development figures, but it is based on published references (in scientific and academic literature, industrial or corporate reports, and in general media citing sources considered to be reliable according to standards). These kinds of scales are useful for conveying practical information about the proximity to the market of any given technology. The technological realm expands to the social and ethical aspects associated with the use of medical AI systems. Beyond their technology availability level, AI systems offer extraordinary opportunities – e.g. derived from greater efficiency – in medical and clinical areas of deep social interest such as oncology, genetics and neurosciences, but also present possible risks and ethical questions raised by their implementation. This balance between benefits and risk is represented by the ‘controversy level’ in Figure 24. This level ranges from (commonly assumed) ‘positive’ or beneficial applications of AI (e.g. software for decision support to improve diagnostic efficiency) to (commonly considered) ‘negative’ or harmful areas (e.g. new tools for bioterrorism or the possibility of engineering biologic weapons targeted against certain populations), crossing through many intermediate domains where a balance between the potential benefits and associated risks needs to be carefully sought. There are different types of controversial issues. Some of them show a clear potential duality, such as the possibility of preventing diseases through genetic editing and the use of neural interfaces and neurostimulation for controlling advanced prosthesis, or for unwanted registration of brain signals (‘thought reading’) and interference with neural signals to impose limits on human free will. Other questionable topics refer to autonomous systems which may make vital decisions for people, and the use of AI-mediated genetic research to challenge fundamental boundaries and the very basic definitions of life (e.g. ‘engineered, enhanced humans’, human-animal hybrids) (Gómez-González & Gómez, 2020). TAL Score Status of viability of the technology TAL 0 Unknown status. Not considered feasible according to references. TAL 1 Unknown status. Considered feasible according to related, indirect references. TAL 2 General/basic idea publicly proposed. TAL 3 Calls for public funding of research and development (R&D) open. TAL 4 Results of academic/partial projects disclosed. TAL 5 Early design of product disclosed. TAL 6 Operational prototype/‘first case’ disclosed. TAL 7 Products disclosed but not available. TAL 8 Available products for restricted (e.g. professional) users. TAL 9 Available for the public. TABLE 5. The Technology Availability Level (TAL) scale defined in Gómez-González and Gómez (2020) HEALTH AND LONG-TERM CARE WORKFORCE 91 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS FIGURE 24. - A visual overview of the classification of AI and AI-mediated technologies in medicine and healthcare according to their ethical and social impact. SW: software, AR: augmented reality, VR: virtual reality, IoT: internet of things, TAL: Technology Availability Level Note: reproduced with permission from Gómez-González & Gómez (2020) HEALTH AND LONG-TERM CARE WORKFORCE 92 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS These controversial aspects are being dealt with at different levels. We can distinguish three groups of ethical and social aspects related to AI in medicine, according to how they are considered in comparison to other application domains (Gómez- González & Gómez, 2020). The first group includes issues which are common to other areas of the application of AI systems, namely social networks, electronic commerce, automation of manufacturing processes and autonomous vehicles. These are topics such as data privacy, fairness or human oversight, and have been broadly discussed in the context of general AI ethical frameworks (EU, 2019b). The second group comprises topics also common in other application domains but of particular relevance in medicine and healthcare such as transparency, the trust in the relationship between doctors and patients or the required updates in evaluation, benchmarking and legislation. Lastly, a third group refers to controversial aspects not yet considered in other fields. Among them, ethical guidelines related to selfexperimentation medicine (including gene editing) or the generation of artificial life forms. From the previous analysis, we can emphasise three novel emerging paradigms. Firstly, we observe a division of medicine into three main streams, all of them having AI as a supporting tool: (1) ‘fake-based medicine’, based on (unfounded, unconfirmed) rumours and presenting ‘ancient, natural knowledge’ as opposed to scientific, evidence-based medicine, supposedly under malicious control by academia, institutions and governments; (2) ‘patient-generated medicine’, derived from the growing online availability of the many (correct and unsupervised, unreliable) sources of medical information; and (3) ‘scientifically tailored medicine’, evolved from the most advanced scientific research into extended personalised and precision medicine (Gómez-González & Gómez, 2020). The second emerging effect is the increase in social differences and inequalities in the access to AI systems in medicine and healthcare, due to the technical complexities and high costs associated with these systems, e.g. personalised drug design or (genetically) tailored treatments (Gómez-González & Gómez, 2020). Lastly, this analysis also warns about new forms of ‘digital health scammers’, bio-hacking and bioterrorism, arguing how a disorderly development of technology – without analysis and debate about ethical and societal impact – may bring strong conflicts with fundamental rights and principles of our free, democratic, particularly European, societies. A SOCIAL DEBATE A public debate has already started around some of the issues presented above. Most of them relate to the human perception of AI-based diagnostics, the (un) trust generated by increasingly autonomous systems and the well-known concerns about the privacy and security of personal data. In addition, there is a growing number of voices (including highly qualified scientists, physicians and entrepreneurs) who demand for open and truthful information on the actual results of AI-based medicine, particularly in areas of high social interest (e.g. cancer and neuroscience). They also ask for preventive regulations, especially on HEALTH AND LONG-TERM CARE WORKFORCE 93 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS the most dangerous and controversial topics – before it’s ‘too late’ – and advocate a clear focus on human-centred AI development in medicine and healthcare. Most of these concerns are also explicitly included in the ‘urgent priorities for the next decade’ defined by the World Health Organization in early 2020 (WHO, 2020). However, to date, there are still no European or international references to a coordinated overview and analysis of the ethical aspects and social impact of AI in the medical field and related areas. Nor are there any specific regulations on many of the most conflictive issues mentioned below. THE COVID-19 PANDEMIC: CONSEQUENCES AND AN ENHANCED ROLE FOR AI Since early 2020, the unexpected Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV-2) outbreak and the corresponding COVID-19 disease have had strong consequences for individuals and societies all over the world. In many areas, including the most industrialised, the effectiveness of the mechanisms of transmission and contagion of the virus has caused an overflow of patients and the immediate scarcity of health resources – from supplies to professionals and hospital facilities – leading to severe societal effects and very high death tolls. The many unknowns about the disease and the lack of vaccine or effective treatment have prompted many harsh measurements of compulsory population confinement, prohibition of travel and economic lockdown at a very fast pace and with many uncertainties about their duration and future evolution. In this extraordinarily difficult context, there has been an explosion of research in all related scientific fields in which AI-mediated technologies have proven to be essential tools for the common goals of controlling the spread, preventing the contagion and curing the many sick people. Machine learning techniques are being exploited to support COVID-19 diagnosis, to develop potential vaccines and drugs and to build epidemiological models of transmission and spread. AI is also exploited in online information platforms – including the fight against fake information – robotics and telemedicine, and datadriven models are exploited for individual contact tracing and social distancing, as well as quarantine and population confinement control (Nigris et al., 2020). In these extraordinary circumstances, some of the ethical questions mentioned in the previous section become more relevant, bringing a sudden interest for the public. They include conceptual reflections on how to prioritise health attention (Ahuja, 2020) and assign reduced resources (e.g. who should be attended to? How should an automated system assist the triage of incoming patients? (Walsh, 2020)); concerns of massive (including genetic) data collection (Wee, 2020) and population monitoring (Schechner et al., 2020); the opposition of the industry to additional regulations about data control and the training of systems which – in their view – would slow down the ability to respond to crises such as COVID-19 (Chee, 2020); and the possibility of deleting personal information after the pandemic can be considered as ‘controlled’ – as in Norway (Klesty & Macfie, 2020). HEALTH AND LONG-TERM CARE WORKFORCE 94 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS In addition, AI-mediated technologies are being used for different levels of social monitoring, from aerial drones to enforce the confinement of the population (Linder, 2020a) to the control of interpersonal distancing in public spaces (Linder, 2020b). However, devices with an original ‘healthcare orientation’ (e.g. hand washing (Kelly et al., 2020)) can be easily employed to monitor the individual behaviour – even in private environments – and send the information to third parties. DEMOGRAPHY AND DATA BIAS IN AI SYSTEMS One important factor influencing the performance of AI systems in general (and machine learning models in particular) is the ‘quantity’ and ‘quality’ of the data used to train them. This has been a subject of extensive analysis since the early days of AI that becomes a particularly complex question for health-related data since they are commonly fragmented and potentially biased with respect to the many demographic features (e.g. age, gender, ethnicity, body mass and others) accounting for the variability of the population in which they will be exploited (Panch et al., 2019). Systems trained with non-representative, biased datasets will not only produce ‘operational’ (strictly technical) errors if applied to individuals for which they have not been suited, but the resulting outputs may be completely wrong, with very serious consequences and the corresponding – yet unsolved – questions about liability and responsibility. In another area – security monitoring – an initial example of this new type of errors was recently brought into the headlines as the incorrect identification by a facial – possibly racially biased – recognition system led to the arrest of an innocent person (Hill, 2020). AI algorithms require representative numbers of cases to ‘learn’ and it may not be easy to provide them with enough, well-balanced sets to achieve an appropriate representation of the different human groups for generalised medical applications. This is a particularly high risk for under-represented populations. PERSONAL CROSS-LANGUAGE COMMUNICATION AND ‘PSYCHOLOGICAL HACKING’ Of particular interest is the development of AI-mediated devices to enhance interpersonal communication. In the context of the COVID-19 pandemic, they are potentially useful for overcoming the limitations in physical contact required to fight the disease. New technological designs include conversational assistants (e.g. ‘chatbots’) that can also work in a cross-language setting, incorporating, for instance, devices or ‘connected’ face masks capable of performing real-time translations (Kelly & Tomoshige, 2020). These types of systems evolve from an extensive research literature on natural language processing and automatic translation. With an obvious HEALTH AND LONG-TERM CARE WORKFORCE 95 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS potential ‘for good’, they are exploited to facilitate medical assistance, interactive translations, cross-language communication (e.g. the integration of migrants), cross-language telemedicine and to fight loneliness in quarantine periods or in certain remote, isolated environments. Nevertheless, automated (autonomous) conversational platforms may also result in the development of psychological, emotional links between people and machines. This has been a topic of extensive analysis since the early days of AI which has come to the interest of the general public in the context of the current COVID-19 pandemic. During prolonged periods of social isolation – with available online platforms – some users declare that they ‘feel very connected’ to the AI systems they use (Metz, 2020). Questions arise as to whether such platforms are used by vulnerable populations – children, the elderly and people with mental ailments – which may develop trust in and affection for them. There may be beneficial applications of affective computing, for example to support certain therapies linked to neurological or psychiatric disorders (e.g. dementia), but the risk of manipulation is high. A recent report68 shows that malicious interference (‘hacking’) of a wearable can (relatively easily) generate fake signals for the user to take medicines or other actions, with the obvious risks of inducing severely damaging consequences (e.g. overdose). As pointed out in Gómez-González & Gómez (2020), health data present a worrying vulnerability to illicit ‘manipulation’, since alterations of data would be extremely difficult to track and identify. It seems that the (evil design of) conversational chatbots may open a window to new forms of interference in people (particularly for those more vulnerable). This can be considered as ‘psychological hacking’ (‘psycho-hacking’), and calls for its further analysis. HEALTH AND AI SYSTEMS FOR BORDER CONTROL One of the areas in which AI systems are beginning to be exploited is that of border control. This is a very sensitive application with many different aspects to consider, from the strictly logistic ones (many hundreds of millions of people enter the European Union each year) to governmental requirements to fight crime and terrorism, the need for userfriendly procedures and the requirement for the robust protection of the human rights of citizens, migrants and asylum-seekers. Automated systems for screening at borders, with conversational capabilities in particular, are of great interest (Accenture, 2017), even the subject of EU-funded projects,69 and some countries are in different stages of testing (Kendrick, 2019). In 2018, a detailed analysis of the use of automated systems at the Canadian border warned how they may have a strong, negative impact from the point of view of human rights and exacerbate disparities with the more vulnerable, under-resourced communities (Molnar & Gill, 2018). In the current international situation established by the COVID-19 pandemic, health information related to the disease is an additional requirement for entering Europe HEALTH AND LONG-TERM CARE WORKFORCE 102 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS Figure 29 depicts the computed AI exposure score differentiated by cognitive abilities. We obtain the AI exposure score by mapping AI progress to work contents through the layer of cognitive abilities for the four medical occupations selected. Firstly, the figure shows that out of this group of occupations, medical doctors are most exposed to AI. Secondly, Figure 29 clearly shows that most AI exposure is driven by its impact on tasks that require abilities that deal with ideas, such as comprehension (CE), attention and search (AS) as well as conceptualisation (CL). This is not because we assign more cognitive abilities (six) to the ideas category than to the other categories (four each), since the smallest exposure score from the ideas abilities (in most cases quantitative reasoning (QL)) is still always greater than the highest exposure score from the people category. Compared to this, the exposure scores in the things category are negligibly small. That is, little AI exposure can be expected through basic processing abilities, such as visual processing (VP) or auditory processing (AP), nor through abilities that deal with people, such as mind modelling and social interaction (MS) or communication (CO). However, our findings based on the tasks and occupation data indicate a relatively high need for people abilities in most occupations and a relatively low need for abilities dealing with things. Equivalently, the findings on AI research intensity suggest high activity in AI areas that contribute to abilities dealing with things and ideas and low activity for abilities dealing with people. Lastly, we compute a single AI exposure and plot the score against average wage percentiles for all occupations in our dataset, illustrated in Figure 30.We clearly observe a positive relationship between wages and AI exposure. That is, high-income occupations seem to be more likely to be affected by AI progress than low-income occupations. FIGURE 28. AI research activity per cognitive ability weighted by average intensity per period Source: JRC CAS – HUMAINT. 0.00000 0.00025 0.00050 0.00075 0.00100 MP SI VP AP AS PA CE CO EC NV CL QL MS MC Ability intensity 2008−2010 2011−2013 2014−2016 2017−2018 HEALTH AND LONG-TERM CARE WORKFORCE 103 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS FIGURE 29. Ability-specific AI exposure scores for selected occupations Note: patterns reflect ability categories, where stripes represent people abilities, checked patterns represent ideas abilities and no pattern represents things abilities. Source: JRC CAS – HUMAINT. FIGURE 30. Scatterplot and best fit line, AI exposure score against wage percentiles Source: JRC CAS – HUMAINT elaborations of Structure of Earnings Survey, 2014. HEALTH AND LONG-TERM CARE WORKFORCE 104 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS CONCLUSIONS: THE IMPACT OF AI ON LABOUR MARKETS It is clear that there are many effects to consider when analysing the impact of AI on work (Brynjolfsson & Mitchell, 2017). This analysis is limited to the technical potential of AI (i.e. the things that AI could potentially do at work). We can use this approach to highlight occupations and abilities involved where AI could play a role. However, this framework remains silent about the complementary conditions necessary to enable the integration of AI in the workplace and the processes that occur after the integration of AI. Consequently, our results have to be interpreted in light of this limitation. Nevertheless, this study sheds light on some aspects of the relationship between AI progress and labour markets. Firstly, these findings show that AI – as an emerging technology – could potentially play a novel role in the context of technology-driven labour market polarisation. According to some studies (Autor et al., 2003; Goos et al., 2014), previous waves of technological progress led to polarisation on the labour market where the automation of medium-skilled occupations pushed medium-skilled workers to either low- or high-skilled occupations. In contrast, our findings (according to Figure 30) suggest relatively high AI exposure for high-skilled medical doctors but lower AI exposure for medium-skilled occupations such as nursing and paramedicals. However, AI exposure can also be relatively high for medium-skilled occupations such as medical technicians. In the end, it depends on the cognitive abilities required. Overall, this can have different implications for labour market polarisation (and consequently inequality) depending on whether AI exposure is labour-replacing or labour-enhancing. If this effect is in fact a labour-replacing one, it could potentially lead to unpolarising effects and a reduction in income inequality (Webb, 2019). If this effect is a labour-enhancing, it could imply a significant expansion in productivity for high-skilled occupations, potentially leading to occupational upgrading effects and an expansion of income inequality (very much like the traditional hypothesis of skills-biased technological change; see Acemoglu (2002)). Furthermore, our findings show that AI progress could affect how specific skills are rewarded (e.g. in terms of wages and working conditions) on the labour market. The finding of low exposure through people abilities versus high exposure through ideas abilities is parallel to Deming (2017) who explores the relationship in the labour market returns to social skills and, what he calls, cognitive skills which we refer to here as analytical skills. More specifically, we use social skills to interact with people and we use abilities that deal with ideas in areas that require analytic skills. Deming (2017) finds that social and analytical skills are complements rather than substitutes. That is, an increased labour demand for analytical skills, which increases wages for people with analytical skills, leads to an increased labour demand for people that, in addition to analytical skills, also have strong social skills. In addition, we find that many labour market tasks require high levels of people as well as ideas abilities, but AI exposure occurs mostly through ideas abilities only. Consequently, we can expect an increase in the wages for workers that combine their strong ideas abilities with strong people abilities. HEALTH AND LONG-TERM CARE WORKFORCE 105 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS 4.3 AGE AND TERRITORIAL DIGITAL DIVIDE FOR TELEMEDICINE The adoption of digital technology in the healthcare industry is generating major changes in the way healthcare services are delivered and in patients’ interactions with medical workers. Many EU Member States are removing some of the regulatory and financial barriers to remote healthcare in order to strengthen their capacities and reduce the increasing cost pressures arising from healthcare expenditure related to the ageing population (EU, 2019a, 2018b; Alotaibi & Federico, 2017; Stroetmann et al., 2015; Nouhi et al., 2012). Additionally, the acceptance of digital products for healthcare is increasingly widespread among consumers (Safi et al., 2019), and people are generally supportive of using their data to create new knowledge and improve care (OECD, 2020d). The ongoing COVID-19 pandemic is further changing users’ perspectives in favour of remote healthcare even in the new normality, thus acting as an additional driver for the implementation and use of remote consultation.74 The main technological solutions in healthcare include mhealth, ehealth, telehealth and telemedicine, and primarily target issues related to mobility, communication, interactivity, remote monitoring and the timely provision of patient-specific information. In short, the terms ‘eHealth’ and ‘mHealth’ are used to describe the provision of health services using the internet and wearable devices respectively. The term ‘telehealth’ is used to describe various electronic procedures related to health, while ‘telemedicine’ specifically refers to the remote treatment of patients (see Box 6 for definitions). Telemedicine is categorised into three types of services using different ICT solutions, which are real-time communication, store-and-forward approach, and patient tele monitoring. Real-time communication makes use of standard communication technologies for patient contact and data exchange, including video visits, live chat, and email. Video/audio quality is therefore essential for physician and patient appointments. Official communication methods and platforms are typically used to ensure data privacy and security. Telemedicine also uses the store-and-forward approach through which clinical data - typically demographic data or lab reports - are filled in and transmitted. The healthcare provider can either use a mobile device or desktop computer to collect and send the information via email or upload it to a secure platform. Finally, with the remote patient monitoring, it is possible to track patients’ vital statistics remotely through the use of electronic devices that transmit patient statistics to a healthcare provider’s analytic interface. In particular, wearable devices (e.g. cardiac and activity sensors) are becoming increasingly common in people’s daily lives, especially in areas with better internet coverage. Overall, these technologies and applications demand users to have adequate training on how to use them. While systems for healthcare providers may consist of a variety of analytical interfaces, ICT applications for older users are getting more user-friendly, although they require them to acquire the knowledge and skills to use electronic devices. BOX 6 Definition of key concepts: eHealth, mHealth, Telemedicine and Telehealth HEALTH AND LONG-TERM CARE WORKFORCE 106 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS The adoption of telemedicine has so far been rather fragmented and limited in the EU. Challenges related to the diffusion of telemedicine are manifold and include regulatory, cultural and commercial barriers; substantial investment in infrastructure or human resources; territorial differences in broadband internet services; 75 and a significant digital and e-skills divide among the elderly patients that prevents their full involvement via mhealth. The age-based digital divide in Europe in particular is very deep, both between and within countries, and internet access is still a luxury for the inhabitants of some remote and rural areas. All these factors challenge the potential of the new digital health systems that can be used by a large part of the population. The WHO defines eHealth as the ‘costeffective and secure use of information and communications technologies in support of health and health-related fields, including health-care services, health surveillance, health literature, and health education, knowledge and research’ 76 . eHealth therefore includes a wide range of solutions including electronic health record systems, patient and laboratory administration systems, telemedicine and mhealth. mHealth (or mobile health) is defined by the WHO Global Observatory for eHealth as ‘medical and public health practice supported by mobile devices, such as mobile phones, patient monitoring devices, personal digital assistants (PDAs), and other wireless devices’. Patients can store and monitor their health data, consult electronic medical records on their mobile devices, communicate directly with doctors and therapists through text messages or video visits, and use reminders and medical applications to follow appointments or pursue a healthy lifestyle. Telemedicine refers exclusively to the provision of remote clinical services to patients. It is defined as ‘The delivery of health care services, where distance is a critical factor, by all health care professionals using information and communication technologies for the exchange of valid information for diagnosis, treatment and prevention of disease and injuries, research and evaluation, and for the continuing education of health care providers, all in the interests of advancing the health of individuals and their communities’ (WHO, 2010b). Telehealth refers to both remote clinical and non-clinical services. It is associated with telemedicine but includes a wider application of technologies, such as distance medical training, consumer awareness, nursing call centres and other digital applications designed to support health services. The terms telehealth and telemedicine are often used interchangeably as there are no universal definitions of these concepts. HEALTH AND LONG-TERM CARE WORKFORCE 107 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS ACCESS TO THE INTERNET The presence of a significant age-based digital divide in the access to the internet 77 and the lack of ICT skills for productive purposes by a large part of the population most in need of healthcare are key obstacles to the effective implementation of telemedicine. A high percentage of the older European population, particularly those with a low level of formal education and residing in rural and remote areas, indeed do not use the internet and are not familiar with ICT. According to the EU-SILC 2018 microdata, elderly citizens have less access to the internet in their homes than the total European population. On average, 54% of the adult population aged 65 and over had an internet connection for personal use in 2018, compared to an average of 81% of the total population. The EUSILC survey identifies the internet access that can take place via smartphones and other devices such as tablets, laptops, desktop computers, TVs, etc. Internet activities for personal use include creating social networks, sending/receiving emails, creating web pages, internet banking, reading or downloading videos or news, searching for information, making phone/video calls, participating in online consultations or voting on civic or political issues. There are considerable disparities between EU Member States in the proportion of adult individuals who habitually use the internet and an uneven distribution of such an access among different age groups in the population. In 2018, the percentage of individuals in the EU-27 aged 65 and over with internet access for private use was 54%. The percentages of regular users are much higher among younger individuals, with these percentages above 94% for individuals aged 16-29, between 90% and 93% for users aged 30-54 and around 55% for those aged 55-59. At national level, the percentage of the adult population aged 65 years and over with internet access varies considerably from a minimum of 20% in Romania to a maximum of 89% in Denmark. In general, northern and western Member States demonstrate the highest levels of internet use compared to Eastern and Southern European countries. The Member States with the highest percentages of adults with internet access are indeed Denmark (89%), the Netherlands (88%) and Sweden (87%), while those with the lowest percentages – beyond Romania – are Bulgaria (22%), Croatia (24%) and Greece (23%) (Figure 31). In order to understand the size and heterogeneity of the population that could be more involved in digital applications for telemedicine, it is necessary to determine the extent to which these people are connected to the internet and some of their specific needs. We look at a series of questions from the 2018 EU-SILC survey on household material deprivation related to internet access – i.e. the possession of a PC, telephone, the internet – for more than 270,000 people over the age of 50 years. In particular, we highlight both the age and the geographical aspects of the internet access divide, HEALTH AND LONG-TERM CARE WORKFORCE 108 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS FIGURE 31. Proportion of population using the internet, 2018 Note: Malta excluded due to missing values on the age of internet users Source: KCMD elaboration based on EU-SILC microdata, 2018. presenting figures for four age groups (50-59, 60-69, 70-79, 80+) in the population and for the degree of urbanisation of the place of residence.78 Table 7 shows the presence of a divide in overall internet access across the different age groups of the adult population and by the degree of urbanisation. Overall, internet access is greater among the adult population living in large cities (87%), while the lowest percentages are observed among those living in less densely populated areas (80%). Among the adult population, the youngest group (50-59 years old) uses the internet the most, with percentages above 90% in the three geographical areas by the degree of urbanisation. In contrast, the older group (80+) has the lowest percentage of internet users, with the lowest scores observed in remote areas (41%) and the highest scores reported among the residents of densely populated areas (57%). These data also show differences by age groups in the availability of a computer at home; the 50-59 age group has higher rates of availability of a computer at home than older groups, reporting a percentage of over 90% in the three areas of residence. Among the over-80s, only 34% of those living in remote areas have a PC, compared to 41% of older people living in intermediate areas and 46% of the group living in large cities. Most EU households have a telephone, regardless of age or place of residence. We now turn our attention to demography and some aspects of material deprivation of the adult off-line population. The EU-SILC microdata contain information on the HEALTH AND LONG-TERM CARE WORKFORCE 109 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS population, sufficiently detailed in their economic and socio-demographic breakdown to identify where the main challenges of internet access may operate. In Table 8, we use a range of information on health conditions, socio-economic status, social and family networks and the area of residence, as follows. The general health status assessment is expressed with a categorical variable ranging from 1 to 5, with the lowest values indicating the lowest health condition; the presence of chronic illness or condition is a dummy variable equal to 1 (and equal to 0 in the absence of disease); activity restriction indicates the percentage of individuals who are limited in their usual activities due to a health problem; female indicates the sex of the respondent; the variable single person household indicates whether the household is composed of one person, while the household members variable indicates the total number of household components; higher education corresponds to post-secondary non-tertiary education and tertiary education (ISCED levels 4 and above). The material deprivation is measured by the indicator of poverty and social exclusion risk, expressed on a scale from 0 to 7, with 7 indicating the highest material deprivation; the variable financial ability indicates the ability to cope with unexpected expenses; financial satisfaction indicates a positive perception of one’s finances measured on a scale from 0 to 10, with 10 corresponding to the highest levels of satisfaction; lastly, being together is a binary indicator indicating whether meetings with friends and relatives take place regularly. Access to the internet is a social and economic issue and it remains particularly limited for the population that would need it the most (Table TABLE 7. Internet access and material deprivation items by age groups and degree of urbanisation Source: KCMD elaboration based on EU-SILC microdata, 2018. Age groups Degree of urbanisation Internet access Computer Telephone (including mobile phone) 50-59 Thinly populated area 93% 90% 100% 60-69 86% 85% 99% 70-79 68% 65% 99% 80+ 41% 34% 100% Total (50+) 80% 78% 100% 50-59 Intermediate area 94% 92% 100% 60-69 92% 90% 100% 70-79 80% 75% 100% 80+ 51% 41% 100% Total (50+) 85% 82% 100% 50-59 Densely populated area 93% 90% 100% 60-69 92% 87% 99% 70-79 83% 74% 100% 80+ 57% 46% 100% Total (50+) 87% 82% 100% HEALTH AND LONG-TERM CARE WORKFORCE 110 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS 8). Around half of the respondents without an internet connection stated that they are in a medium state of health, presenting a ‘health’ index of approximately three points in each age group and level of urbanisation. However, more than half of the adult offline population report having a chronic illness. The older group of people over 80 certainly require more attention and healthcare, presenting the highest rate of chronic diseases, equal to 71% in remote areas, 62% in intermediate areas and 71% in densely populated areas. 71% of the population in the age group 50-59 years old, and resident in large cities, also report suffering from chronic diseases. These health conditions represent a limit in the performance of daily activities for around 60% of the total adult population (50 years and over) and for the different levels of urbanisation. Among those lacking access to the internet, women have the highest percentage across all age groups and levels of urbanisation (Table 8). We therefore highlight an internet gender divide with differences in the access and effective use of ICT within and between countries. The highest percentages of women without access to the internet are observed among residents of intermediate areas and among the older groups of the adult population. Approximately 52% of the adult offline population lives alone in intermediate and densely populated areas, while this percentage is around 35% of the sample interviewed in remote areas. It is in large cities where older people aged over 80 have the highest percentage of single person households (73%), followed by 65% in intermediate areas and 53% in remote areas. Families are indeed more numerous among the residents of small cities or remote areas where, on average, Thinly populated area Intermediate area Densely populated area Age groups 50- 59 60- 69 70- 79 80+ Total (50+) 50- 59 60- 69 70- 79 80+ Total (50+) 50- 59 60- 69 70- 79 80+ Total (50+) Health status (1 Very bad –5 Very good) 3.311 3.403 3.306 3.055 3.254 3.136 3.166 3.12 3.099 3.123 2.796 3.09 3.204 3.038 3.044 Chronic illness 54% 54% 58% 71% 60% 59% 61% 59% 62% 60% 71% 59% 55% 71% 64% Activity restriction 51% 52% 54% 73% 59% 55% 54% 64% 70% 63% 65% 53% 57% 64% 60% Female 52% 53% 54% 66% 57% 55% 54% 60% 71% 62% 56% 66% 60% 65% 62% Single person household 27% 16% 34% 53% 35% 46% 47% 42% 65% 52% 42% 46% 44% 73% 53% Household members 2.123 2.416 1.779 1.673 1.935 1.732 1.725 1.686 1.403 1.596 1.86 1.83 1.63 1.272 1.603 High education 9% 5% 8% 9% 8% 11% 8% 10% 9% 9% 21% 13% 14% 16% 16% Risk of poverty and social exclusion 1.135 0.808 1.023 1.293 1.069 2.059 1.088 0.905 0.904 1.106 3.455 1.808 0.69 0.584 1.447 Financial ability 81% 79% 85% 81% 82% 51% 66% 80% 87% 76% 27% 48% 71% 78% 60% Financial satisfaction (0 Low – 10 High) 6.631 6.867 7.299 7.595 7.206 5.79 6.437 7.013 7.2 6.8 4.258 5.922 7.078 7.482 6.398 Get-together 75% 81% 78% 74% 77% 59% 63% 70% 68% 66% 59% 62% 70% 76% 68% TABLE 8. Demographics and socio-economic breakdown of the adult offline population Source: KCMD elaboration based on EU-SILC microdata, 2018. HEALTH AND LONG-TERM CARE WORKFORCE 111 THE ROLE OF DIGITAL TECHNOLOGY IN HEALTH AND LONG-TERM CARE SECTORS the family unit is made up of two components. It is therefore essential to improve the accessibility to digital services, as digital activity increasingly contributes to social and cultural inclusion and may help in preventing social isolation. Overall, individuals with a high level of education (post-secondary and tertiary) are almost all regular internet users (Table 8). The percentage of the adult population without internet access but with a high level of education is indeed below 10% in rural and intermediate areas and for the different population age groups. Among those with a high level of secondary education, the highest percentage of offline users is in the population between 50 and 59 years old living in large cities (21%). On average, adults aged 50 years and over are exposed to a similar risk of poverty and social exclusion through the different levels of urbanisation (Table 8). However, respondents aged between 50 and 59 have slightly higher rates of material deprivation than older groups in large cities and intermediate areas, and economic affordability remains an important barrier for ICT use within the offline population. Similarly, the younger group living in large cities and intermediate areas report lower financial satisfaction, less financial capacity to sustain unexpected expenses and reduced social encounters than older groups living in the same areas. DIGITAL SKILLS OF ELDERLY PEOPLE In addition to having a device that can be connected to the internet (phone, computer, etc.) and an access to the internet, the implementation of telemedicine also requires a sufficient level of digital skills in the patient. Basic digital skills or above basic digital skills represent the two highest levels of the general e-skills indicator, which is a composite Eurostat indicator based on the activities undertaken by individuals aged between 16 and 74 on the internet predominantly in the areas of information, communication, problem solving and content creation. Table 9 summarises Eurostat’s information for the year 2019 on the proportion of the total population, adult population aged 65 to 74, adult population aged 55 to 74 (by educational level and gender) and the population living in rural, medium and large cities with basic or higher digital skills. The highest percentages of individuals self-reporting to have basic or higher digital skills are in Denmark and Germany (70%), the Netherlands (79%) and Sweden (72%), while the lowest are in Bulgaria (29%) and Romania (31%). With regard to e-skills within the population aged 65-74, Finland, Sweden, Denmark and the Netherlands have the highest percentages for the adult population with basic or higher digital skills (over 40%); on the other hand, in Bulgaria, Romania, Latvia, Poland, Greece, less than 10% of the population aged 65-74 reports to have good e-skills. Lastly, in Bulgaria, Cyprus, Latvia, Lithuania, Slovakia and Poland we find the lowest percentages of women between 55 and 74 years old and the population in the same age group with a low level of formal education claiming to have good digital skills.