Smartphones have a central role in digital health by enabling (1) data collection through built-in functions (eg, GPS for geolocation) and interfaces for patient-reported outcome measures, (2) connection to the IoT via a Bluetooth connection, and (3) patient feedback. In the best-equipped countries, smartphone ownership is limited to 25%, 50%, and 75% of children aged 8 years to 9 years, 10 years to 11 years, and 12 years to 13 years, respectively, complicating the use of digital health interventions for the remote management of children’s diseases [11-14]. Furthermore, it is not intended to promote better smartphone equipment among CYP. In addition to the negative effects of screen time on children's cognitive and socioemotional development, smartphone addiction affects 1 in 4 CYP [15,16]. Thus, even if the DTS interfaces presented on smartphones do not pose an addiction problem, since they will be limited to supporting children in the daily management of their disease, providing a smartphone to children on this occasion could have negative effects on their development due to the other uses they would make of it (eg, games, social media). It is therefore preferable to use specific standalone devices to link the different connected objects as has been done for artificial pancreas systems. In the case of childhood asthma, a smartwatch that would both collect data relevant to monitoring the child's asthma control status (heart rate, oxygen saturation, activity) and provide appropriate and timely recommendations on its screen without allowing other types of internet access would be an interesting solution.

All parents have experienced the recurrent changes in size of clothing and shoes as children grow. A child's height doubles in the first 4 years, from an average of 50 cm to 100 cm, and then increases again by at least 1.5 times over the next 15 years. Similarly, a child's abilities develop impressively, from the infant who does not yet walk and talk to the adolescent capable of the most extreme sports and complex reasoning. Finally, in the medical field, physiological values vary constantly with the age of the child (eg, a heart rate of 50 bpm is normal for a teenager but abnormally low for an infant), as do the expected results of additional tests. This complicates the task for manufacturers of connected devices who need to provide different model sizes, develop appropriate interfaces for children and adolescents of different ages, and adapt the standards for physiological parameters according to the age of the child. These adaptations are not without risk: In the field of cardiac medical devices, an attempt to miniaturize an implantable cardioverter defibrillator resulted in a higher risk of failure in young patients [17]. In the case of an asthma DTS providing recommendations for children, we shall ensure that the recommendations given are age-appropriate, using oral instructions for a 6- or 7-year-old rather than text messages, which would be appropriate for teenagers.

The use of connected objects to collect data from children poses particular risks. Similar to a toy with a defective design, children may choke on, or simply ingest, small parts that may come off the object; they can also be exposed to chemicals that are carcinogenic or mutagenic or to endocrine-disrupting substances. Thus, devices intended to collect data from children’s bodies must be manufactured taking into account the additional risks they may pose to children, which is another challenge compared with manufacturing devices for adults.

In the other direction, children pose specific risks to the connected objects. In the same way that children regularly come home having lost or broken their glasses, connected objects are more likely to be lost or broken when used with children than with adults. This is the consequence of their age-related activities but also of a less cautious attitude toward their personal belongings. For example, connected inhalers that automatically record children's use of asthma medication were reported lost or damaged by up to 50% of families [18,19].

CYP pose particular problems when it comes to creating models with traditional supervised approaches from real-world data. The amount of data available is lower, due to the young age of patients (less historical data), logistical and legal difficulties in obtaining the data, and the lower prevalence of diseases in children than in adults in high-income countries [20]. Furthermore, the heterogeneity of the pediatric population is such that it is unsound to learn a single model and necessary to split data in several subgroups. This leads to smaller data sets, which makes it more difficult to obtain performant models for CYP. Fine-tuning models developed for adults may be an option to overcome these limitations, but to our knowledge, it has not yet been applied to create pediatric models.

Autonomy means being governed by oneself, in thought and action [24]. As the principle of autonomy cannot be applied as such to children, who are inherently dependent on adults to meet their needs, it was proposed instead to protect the development of the child's autonomy [25]. Pediatricians encourage the development of autonomy in children with diseases by encouraging them to describe their symptoms themselves and by supporting them in taking responsibility for their own care.

Beneficence refers to the responsibility of professionals to promote the well-being of their patients [31]. It is clear that, before being used in children, DTS must be evaluated for both efficacy and safety through appropriate clinical studies. However, DTS may also pose particular threats to young people and their development.

The principle of justice refers to the delivery of equal treatment and care according to the particular patient’s needs, as well as just allocation of available resources [36]. Providing appropriate care for each child according to their needs is all the more important as poor child health limits the potential and development of children, leading to reduced health and life chances in adulthood [37,38]. Yet, the main risk of DTS is that it exacerbates inequalities in health care for children. First, disadvantaged families may not be able to afford DTS, thereby excluding their children from the most effective (if proven) treatment strategies. Second, families with low levels of digital literacy may find it more difficult to use DTS, resulting in less effective treatment adaptations. The third risk is that of self-exclusion of disadvantaged families. Several studies found that lower levels of education, lower levels of employment, and lower household income are associated with negative views and reluctance to participate in research programs involving AI [39-42], raising the risk that algorithms will be trained on data from advantaged families and optimized for these populations. However, DTS can also contribute to reducing inequalities in children's care; for example, by standardizing the care of children and bypassing the doctor for certain decisions, children from disadvantaged families, living in remote areas or with out-of-date doctors, can receive the same care with DTS as children living in privileged areas. For example, the use of connected inhalers that automatically send alerts to health care professionals in case of an asthma attack has been shown to particularly improve the care of the most disadvantaged children [19].

In conclusion, ethical dilemmas can arise in many ways when developing a DTS for CYP, given the need to preserve the health of the child; preserve their physical, mental, and social development; and be anticipated.

Children need special protection from DTS risks for 2 reasons: They have special characteristics that expose them to greater or potentially different risks in comparison with adults, and they are developing beings who may face dramatic consequences for the rest of their lives if their health is adversely affected. Because of these specificities, the approval of a drug for children is subject to a specific process in the United States and European Union. Sponsors are required to provide a pediatric study plan to the Food and Drug Administration (FDA) in the United States or its equivalent pediatric investigation plan to the European Medicines Agency (EMA), which are reviewed by specific committees taking in account pediatric considerations: the FDA's Pediatric Review Committee in the United States and the EMA's Paediatric Committee in the European Union [44].

DTS are not subject to this regulation. Under the current regimes, they are considered medical devices and not medicines. The regulation of medical devices, although recently strengthened in the European Union by the regulation 2017/745 [45], is not as strict as that of medicines from a pediatric perspective [46]. In the United States, the FDA has issued specific guidance for the development of pediatric medical devices [47], but there is no dedicated pediatric committee to assess how these devices are evaluated or adapted for children as there is for medicines and as advocated by the American Academy of Pediatrics [48]. In the European Union, the regulation is limited to stating that the presence of carcinogenic, mutagenic, reprotoxic, or endocrine-disrupting substances must be justified for devices intended for children [45]. Because medical devices are no less risky than medicines, it is important that a similar pathway with pediatric investigation and study plans, reviewed by specialized pediatric committees, is established by the legislators for the approval of pediatric DTS.

If legislators must protect children from the risks that a DTS could pose, they must also protect children from the risk of not having access to DTS that could improve their health. Industry has consistently been less interested in developing drugs for children because of the small number of children affected, increased regulatory constraints, and difficulty of conducting clinical trials in this population, and this trend continues with medical devices including DTS [49-52]. As an example, if multiple algorithms have been used by manufacturers of implantable cardioverter-defibrillators to correctly diagnose atrial and ventricular arrhythmias, these have not been tested in children [17]. For drugs, the situation improved with the introduction of regulations in the European Union and the United States with 2 objectives: (1) to compel pharmaceutical companies to include studies with pediatric populations if the disease or condition for which the drug is indicated occurs in children and (2) to encourage these same companies to carry out these studies by providing a financial incentive in the form of an extension of intellectual property rights [49]. However, such schemes do not exist for medical devices. In the United States, the FDA has launched a series of initiatives to encourage manufacturers to develop pediatric medical devices, with some results [53]. Such initiatives are awaited in Europe.

In conclusion, the risk is that, in the absence of legislative constraints or incentives, market forces will continue to widen the gap between the quantity and quality of DTS developed for adults and children.

If most medical decisions are made autonomously and remotely by DTS, this would result in fewer in-person visits to pediatricians, in line with a more general trend of reduced face-to-face interactions in society [54]. The risk of dehumanization associated with DTS may be a greater problem for pediatricians, who made the choice of a particularly “people-oriented” specialty, in contrast to “technology-oriented” specialties (eg, anesthesiology, radiology) [55]. The risk is that an increasingly technological approach to child health care, combined with a reduction in face-to-face interactions, will lead many to a loss of meaning in practice and to burnout. This may result in a gradual replacement of these doctors by professionals proficient with information technology but less skilled in human relations.

The use of DTS in pediatrics may contribute to a growing distrust of pediatricians. The digitalization of the world, whether through access to medical information on the Internet or conspiracy groups on social networks, has already increased parents' distrust of their pediatricians [56,57]. Families may have less confidence in their pediatrician if they feel that the latter is clearly not mastering the DTS or if the recommendations made by the DTS and the pediatrician differ. In the latter case, explainable models would help the pediatrician to justify disagreement with the DTS and maintain trust. An additional difficulty is that most DTS will aim to prevent a complication or exacerbation from occurring and therefore act while the patient has few or no symptoms. The positive effect of DTS will therefore be less perceptible (absence of an uncertain event) than an exceptional side effect highlighted in social networks. For example, a DTS for childhood asthma may prevent thousands of asthma attacks by providing appropriate recommendations to families when the risk of an asthma attack is predicted for the coming week but be blamed for an undetected severe asthma attack. Finally, pediatricians are perceived to be acting in the best interests of the child, whereas this may not be the case for the companies behind DTS, increasing mistrust [58].

The use of DTS requires the collection of a large amount of data, which are increasingly being collected continuously from the child's home or even body. This constant monitoring of children to ensure their safety is a general trend in our societies, in the same way that the proliferation of surveillance cameras was intended to ensure the safety of the population. It can lead to a surveillance society where everyone watches each other, with children being a main target because of their vulnerability. Parents could demand to monitor their children at all times for their safety, since the data are already available, being continuously collected for health purposes. Pediatricians may also ask to use indirect information about parents' behavior collected from the home devices of the child to monitor the risk of child abuse. If such a surveillance society was to emerge, children would grow up in an optimized state of physical health but would probably be more anxious, dependent, and conventional in adulthood, unable to make decisions on their own values [59,60]. Again, legislation implemented by each society will be crucial in ensuring that the best interest of the child is taken into account, at both the individual and societal levels.

Climate change is one of the major challenges of our century. Young people understand this and are actively campaigning for policies to reduce carbon emissions [61]. CYP are indeed the first to be affected by the consequences of global warming: not only will children born today live in a world with a temperature 4 degrees higher than the pre-industrial era but the effects of global warming on health are much more significant for children than for adults [62]. For example, with regard to the development of diseases favored by global warming, 93% fall upon children [63].

Among the sectors responsible for greenhouse gas emissions, the digital sector is growing in importance every year, increasing in the contribution to global emissions from 2.5% to 3.7% between 2013 and 2019 [64]. Medical devices are the first part of the problem. They usually require raw materials such as rare metals, and most of them are single-use devices with limited recyclability [65]. The flow and storage of the data generated by these devices in data centers are another part of the problem. Currently, data centers account for 1% of total global electricity demand, with about one-half of this energy being used to cool servers [66,67]. Finally, it was recently shown that training a single AI model could generate CO₂ emissions equivalent to those of a passenger making 300 flights between New York and San Francisco [68]. Thus, even with efforts to reduce electricity consumption in data centers and to move toward “green AI” with efficient models, the general use of DTS would generate a significant amount of greenhouse gas emissions [69]. Conversely, if DTS are effective and improve children's health, they would reduce the use of health care (eg, travel to hospitals, hospital admissions) and in turn would reduce the associated amount of greenhouse gas emissions. Thus, it is currently essential to carry out studies of the environmental impact of digital health interventions, in addition to efficacy trials and medicoeconomic studies, as envisaged by the National Institute for Health and Care Excellence in the United Kingdom [70]. This will participate in safeguarding the future of children.