Femke De Backere

BACKGROUND Hospitalization is a stressful experience for children, yet standard well-being assessments often rely on resource-intensive proxy reports, completed by parents, or repetitive burdensome questionnaires, which can lead to reduced data quality over time. OBJECTIVE This study aimed to develop and evaluate MediPals, a serious game incorporating Large Language Model (LLM)-driven non-player characters (NPCs) to interactively assess the well-being of hospitalized children through gameplay. METHODS MediPals was designed using a privacy-preserving hybrid architecture in which an external LLM dynamically generates dialogue trees, while a self-hosted LLM summarizes clinically relevant insights for nurses. A pilot feasibility study was conducted at the University Hospital Ghent with hospitalized children and nurses to assess usability, acceptability, and clinical potential. RESULTS Five patients and four nurses participated in the pilot study, with participation durations ranging from 2 to 10 days. Nurses reported a satisfaction rate of 71% while children rated their experience at 62%. On average, patients engaged in 2.7 daily conversations with NPCs. These findings suggest moderate acceptance and engagement from both children and healthcare providers. CONCLUSIONS While the MediPals pilot study demonstrates that hybrid, safety-validated LLM architectures can securely lower communication barriers for hospitalized children, consistent well-being assessment remains heavily dependent on long-term player engagement. By pairing structural privacy safeguards with a multi-genre gamified ecosystem, future interventions have the potential to transform longitudinal pediatric well-being monitoring in hospitals.

Background: Building on the assertion that nearly every clinician will eventually use artificial intelligence (AI), this study provides a triangulated qualitative analysis of the requirements, challenges, and prospects for integrating AI into routine health care practice. This skills gap contributes to cautious and uneven adoption across clinical settings. Despite advancements, many health care professionals report a self-perceived lack of proficiency in comprehending, critically evaluating, and ethically deploying AI tools, which contributes to cautious adoption in clinical settings. Objective: While addressing key research questions, the study investigates the necessary prerequisites, barriers, and opportunities for AI adoption and specific training priorities that medical staff require. The study is uniquely focused on the health care workforce, moving beyond the predominant emphasis in the literature on medical students. Methods: Situated in Flanders, Belgium, a recognized innovation leader but with moderate lifelong learning participation, this research combines 15 semistructured expert interviews, a regional survey of 134 health care professionals, and 3 co-interpretive focus groups with 39 stakeholders, all conducted in 2024. Results: The results expose small generational and mainly occupational divides. For instance, 85.07% (114/134) of survey respondents expressed interest in introductory AI courses tailored to health care, while 80% (134) of them sought practical, job-relevant AI skills. However, only 13.8% (134) of clinicians felt that their training adequately prepared them for AI integration. Notably, younger professionals (50 years of age) prioritized reducing administrative burden. Physicians and dentists reported higher self-assessed AI knowledge, whereas nurses and physiotherapists showed the lowest familiarity. The survey also revealed differences in preferred learning formats, with doctors favoring flexible, asynchronous learning and nurses emphasizing the need for accredited, employer-supported training during work hours. Ethics, though emphasized in academic literature, ranked low in training interest among most practitioners, except for younger and palliative care professionals. Focus group participants confirmed the need for clear regulatory guidance and access to accredited, practically oriented training. A significant insight was that nurses often lacked institutional support and funding for training, despite their pivotal role in AI-enabled workflows. Conclusions: Taken together, these findings indicate that a one-size-fits-all approach to AI education in health care is unlikely to be effective. By triangulating insights across research stages, this study highlights the need for occupation-specific, accessible, and accredited AI training programs that bridge gaps in digital literacy and align with practical clinical priorities. The qualitative insights obtained can inform policy and training priorities in light of the European Union (EU) AI literacy mandates, while highlighting persistent gaps in workforce preparation.

Background: The optimal moment to deliver digital prompts for reducing sedentary behavior (SB) is when individuals are susceptible to prolonged SB (vulnerability) and open to behavior change (opportunity). This study aims to examine both vulnerability and opportunity. Methods: A 14-day ecological momentary assessment study was conducted with 105 older adults. Participants wore a Fitbit activity tracker and an ActivPAL accelerometer and used a smartphone-based ecological momentary assessment app. Sensor-triggered questionnaires were delivered via the app after 30 minutes of SB to capture participants’ activity, physical, social, and temporal contexts; feelings of pain and fatigue; and willingness to interrupt SB. Descriptive statistics identified states of vulnerability, whereas linear mixed models examined when participants are willing to interrupt prolonged SB, shedding light on states of opportunity. Results: Ecological momentary assessment data (n = 2580) showed that older adults were most vulnerable to prolonged SB while watching TV (36.3%) and using digital devices (14.7%). Prolonged SB predominantly occurred at home (91.2%), in the evening (46.4%), when feeling fatigue (52.2%), and when experiencing mild pain (21.1%). The willingness to interrupt SB varied by activity, physical and social context, perceived pain, and level of fatigue, but not by time of day. Specifically, participants were more willing to interrupt their SB when engaging in mentally active SB and when indoors, alone, or with close relatives. Conclusions: Contextual factors significantly influence older adults’ vulnerability to prolonged SB and their willingness to interrupt it. This empirical information can guide the design of just-in-time adaptive interventions that deliver strategically timed prompts to reduce prolonged SB in older adults.

BACKGROUND The integration of Artificial Intelligence (AI) in healthcare is advancing rapidly, promising to enhance clinical decision-making, streamline administrative tasks, and personalize patient care. However, many healthcare professionals report a lack of confidence in understanding, critically evaluating, and ethically applying AI technologies. In regions like Flanders, Belgium—recognized for innovation yet facing moderate lifelong learning participation—these challenges are pronounced, especially amid an aging healthcare workforce and resource disparities between professions. OBJECTIVE This study aimed to explore the requirements, obstacles, and prospects of AI adoption among healthcare professionals, and to identify the specific training priorities needed to bridge the AI-literacy gap in clinical practice in the Flanders region. METHODS A multi-stage qualitative methodology was employed. First, 15 semi-structured interviews with key informants were conducted to inform the survey design. Then, a survey was distributed to healthcare professionals across Flanders, gathering 134 valid responses. Finally, three focus groups involving 39 participants were conducted to co-interpret the survey findings. Thematic analysis and descriptive statistics were used to synthesize insights across stages. RESULTS Healthcare professionals recognized AI’s potential to reduce administrative burdens and enhance clinical care but reported low self-perceived AI literacy, especially among older and non-physician staff. Interest in AI training was high, particularly for practical applications and basic AI knowledge, rather than technical coding or standalone ethics courses. Differences emerged based on occupation, age, and perceived job security. Nurses and younger professionals were especially concerned about the risks and opportunities of AI adoption. A lack of legally approved AI tools and practical hands-on training were identified as major barriers. Focus group discussions highlighted disparities in access to AI training between doctors and nurses, skepticism about private-sector-led courses, and the need for hospital management support in facilitating AI education. CONCLUSIONS A one-size-fits-all approach to AI training in healthcare is inadequate. Training programs must be stratified by occupation, age, and resource availability, emphasizing immediate practical applications while embedding ethical considerations within broader curricula. Addressing barriers to training accessibility and clarifying regulatory frameworks will be crucial to scaling AI integration in healthcare systems, starting in Flanders and potentially informing broader European initiatives under frameworks like the EU AI Act.

Background. Mobile health (mHealth) interventions hold promise in assisting older adults to meet physical activity (PA) guidelines. Yet, little is known about how older adults perceive using smartphones to enhance their PA. This study explored older adults’ experiences with the “My Health Plan” mHealth intervention and examined which person‐level factors were associated with adherence. Methods. Forty older adults (52.5% female, mean age: 72.6 years) were instructed to use the My Health Plan application, which provided up to six stretch‐ and strengthening exercises per day and to wear a Fitbit Charge activity monitor for seven consecutive days. Person‐level factors (e.g., gender and age) and psychosocial factors (e.g., intention and motivation) were assessed using a questionnaire. Afterwards, semistructured interviews were conducted to gather participants’ experiences with the application and Fitbit. Deductive thematic analysis (qualitative data) and regression analyses (quantitative data) were conducted. Results. Key themes emerging from the interviews were (1) smartphone notifications (including timing and number) and carrying the smartphone throughout the day, (2) suggested stretch‐ and strengthening exercises, (3) providing feedback, (4) experiences with Fitbit, and (5) overall suggestions for improving the application and mHealth interventions. Overall, participants reported positive experiences with the application and Fitbit. Being male, having higher baseline PA, lower anxiety, and greater ability to participate in social roles and activities were related to increased engagement with the proposed exercises. Conclusions. This study provides valuable insights to optimize future mHealth interventions tailored to older adults’ specific needs, aligning with their perceptions of the digital transformation in health promotion.

Background: Despite the availability of physical activity (PA) interventions, many older adults are still not active enough. This might be partially explained by the often-limited effects of PA interventions. In general, health behavior change interventions often do not focus on contextual and time-varying determinants, which may limit their effectiveness. However, before the dynamic tailoring of interventions can be developed, one should know which time-dependent determinants are associated with PA and how strong these associations are. Objective: The aim of this study was to examine within-person associations between multiple determinants of the capability, opportunity, motivation, and behavior framework assessed using Ecological Momentary Assessment (EMA) and accelerometer-assessed light PA, moderate to vigorous PA, and total PA performed at 15, 30, 60, and 120 minutes after the EMA trigger. Methods: Observational data were collected from 64 healthy older adults (36/64, 56% men; mean age 72.1, SD 5.6 y). Participants were asked to answer a time-based EMA questionnaire 6 times per day that assessed emotions (ie, relaxation, satisfaction, irritation, and feeling down), the physical complaint fatigue, intention, intention, and self-efficacy. An Axivity AX3 was wrist worn to capture the participants’ PA. Multilevel regression analyses in R were performed to examine these within-person associations. Results: Irritation, feeling down, intention, and self-efficacy were positively associated with subsequent light PA or moderate to vigorous PA at 15, 30, 60, or 120 minutes after the trigger, whereas relaxation, satisfaction, and fatigue were negatively associated. Conclusions: Multiple associations were observed in this study. This knowledge in combination with the time dependency of the determinants is valuable information for future interventions so that suggestions to be active can be provided when the older adult is most receptive.

Background: The use of serious games in health care is on the rise, as these games motivate treatment adherence, reduce treatment costs, and educate patients and families. However, current serious games fail to offer personalized interventions, ignoring the need to abandon the one-size-fits-all approach. Moreover, these games, with a primary objective other than pure entertainment, are costly and complex to develop and require the constant involvement of a multidisciplinary team. No standardized approach exists on how serious games can be personalized, as existing literature focuses on specific use cases and scenarios. The serious game development domain fails to consider any transfer of domain knowledge, which means this labor-intensive process must be repeated for each serious game. Objective: We proposed a software engineering framework that aims to streamline the multidisciplinary design process of personalized serious games in health care and facilitates the reuse of domain knowledge and personalization algorithms. By focusing on the transfer of knowledge to new serious games by reusing components and personalization algorithms, the comparison and evaluation of different personalization strategies can be simplified and expedited. In doing so, the first steps are taken in advancing the state of the art of knowledge regarding personalized serious games in health care. Methods: The proposed framework aimed to answer 3 questions that need to be asked when designing personalized serious games: Why is the game personalized? What parameters can be used for personalization? and How is the personalization achieved? The 3 involved stakeholders, namely, the domain expert, the (game) developer, and the software engineer, were each assigned a question and then assigned responsibilities regarding the design of the personalized serious game. The (game) developer was responsible for all the game-related components; the domain expert was in charge of the modeling of the domain knowledge using simple or complex concepts (eg, ontologies); and the software engineer managed the personalization algorithms or models integrated into the system. The framework acted as an intermediate step between game conceptualization and implementation; it was illustrated by developing and evaluating a proof of concept. Results: The proof of concept, a serious game for shoulder rehabilitation, was evaluated using simulations of heart rate and game scores to assess how personalization was achieved and whether the framework responded as expected. The simulations indicated the value of both real-time and offline personalization. The proof of concept illustrated how the interaction between different components worked and how the framework was used to simplify the design process. Conclusions: The proposed framework for personalized serious games in health care identifies the responsibilities of the involved stakeholders in the design process, using 3 key questions for personalization. The framework focuses on the transferability of knowledge and reusability of personalization algorithms to simplify the design process of personalized serious games.

BACKGROUND Adopting a physically active lifestyle results in substantial health benefits. However, many adults and older adults are insufficiently physically active and engage in high levels of (SB). Just-in-time adaptive interventions (JITAIs) are promising in targeting these health behaviors by providing real-time supportive prompts during opportunistic moments. Developing effective JITAIs requires better insight into the dynamic determinants of PA and SB, which can be obtained through event-based ecological momentary assessment (EMA). In both JITAIs and event-based EMA research, real-time monitoring of participants’ behavior is needed to enable the detection of the events of interest (e.g., 5 minutes of sustained walking or 30 minutes of prolonged sitting). OBJECTIVE To ensure Fitbits are suitable devices for use in event-based EMA studies and JITAIs, we explored the concurrent validity, accuracy and inter-device agreement of two Fitbit models (Inspire 2 and Ionic) in measuring short bouts of stepping and SB among free-living adults and older adults. METHODS Thirty-seven adults (18-65y) and 32 older adults (65+) from two countries (Belgium and the Czech Republic) wore four devices simultaneously for three consecutive days. In addition to two Fitbit models at the non-dominant wrist, they wore an ActiGraph GT3X+ at the hip and an ActivPAL at the thigh as reference devices for step count and SB, respectively. To investigate the concurrent validity for measuring step count during sustained walking (≥ 60 steps/min), the mean error and mean absolute error were calculated for 1-, 5-, 10-, and 20-minute epochs. In addition, concordance correlation coefficients were calculated and Bland-Altman plots were constructed. To explore Fitbit’s accuracy in detecting short bouts of stepping, the sensitivities and specificities were calculated for various thresholds (60 to 120 steps/minute) and bout durations (1 to 20 min). Finally, to define Fitbit’s optimal thresholds to capture prolonged bouts of sitting (1 to 60 min), receiver operating characteristics curves were constructed. RESULTS The Fitbit appeared to be valid for measuring steps on a short time scale (i.e., 1, 20, 30 and 60 minutes) compared to the ActiGraph. Except for stepping above 120 steps/min in older adults, both Fitbits detected stepping bouts of various durations in both adults and older adults with sensitivities and specificities greater than 87% and 97%, respectively. In addition, optimal cut-off values to detect prolonged bouts of sitting were identified, obtaining sensitivities and specificities greater than 93% and 89%, respectively. CONCLUSIONS This paper provides practical considerations for using Fitbit devices in JITAIs and event-based EMA studies in adults and older adults. Considering Fitbits' reasonable accuracy in detecting short bouts of stepping and SB, they can be used to trigger JITAI prompts or EMA questionnaires following an event of interest (e.g., 5 minutes of sustained walking).

BACKGROUND Adopting a physically active lifestyle results in substantial health benefits. However, many adults and older adults are insufficiently physically active and engage in high levels of (SB). Just-in-time adaptive interventions (JITAIs) are promising in targeting these health behaviors by providing real-time supportive prompts during opportunistic moments. Developing effective JITAIs requires better insight into the dynamic determinants of PA and SB, which can be obtained through event-based ecological momentary assessment (EMA). In both JITAIs and event-based EMA research, real-time monitoring of participants’ behavior is needed to enable the detection of the events of interest (e.g., 5 minutes of sustained walking or 30 minutes of prolonged sitting). OBJECTIVE To ensure Fitbits are suitable devices for use in event-based EMA studies and JITAIs, we explored the concurrent validity, accuracy and inter-device agreement of two Fitbit models (Inspire 2 and Ionic) in measuring short bouts of stepping and SB among free-living adults and older adults. METHODS Thirty-seven adults (18-65y) and 32 older adults (65+) from two countries (Belgium and the Czech Republic) wore four devices simultaneously for three consecutive days. In addition to two Fitbit models at the non-dominant wrist, they wore an ActiGraph GT3X+ at the hip and an ActivPAL at the thigh as reference devices for step count and SB, respectively. To investigate the concurrent validity for measuring step count during sustained walking (≥ 60 steps/min), the mean error and mean absolute error were calculated for 1-, 5-, 10-, and 20-minute epochs. In addition, concordance correlation coefficients were calculated and Bland-Altman plots were constructed. To explore Fitbit’s accuracy in detecting short bouts of stepping, the sensitivities and specificities were calculated for various thresholds (60 to 120 steps/minute) and bout durations (1 to 20 min). Finally, to define Fitbit’s optimal thresholds to capture prolonged bouts of sitting (1 to 60 min), receiver operating characteristics curves were constructed. RESULTS The Fitbit appeared to be valid for measuring steps on a short time scale (i.e., 1, 20, 30 and 60 minutes) compared to the ActiGraph. Except for stepping above 120 steps/min in older adults, both Fitbits detected stepping bouts of various durations in both adults and older adults with sensitivities and specificities greater than 87% and 97%, respectively. In addition, optimal cut-off values to detect prolonged bouts of sitting were identified, obtaining sensitivities and specificities greater than 93% and 89%, respectively. CONCLUSIONS This paper provides practical considerations for using Fitbit devices in JITAIs and event-based EMA studies in adults and older adults. Considering Fitbits' reasonable accuracy in detecting short bouts of stepping and SB, they can be used to trigger JITAI prompts or EMA questionnaires following an event of interest (e.g., 5 minutes of sustained walking).

BACKGROUND Despite the availability of physical activity (PA) interventions, many older adults are still not active enough. This might be partially explained by the often-limited effects of PA interventions. In general, health behavior change interventions often do not focus on contextual and time-varying determinants, which may limit their effectiveness. However, before the dynamic tailoring of interventions can be developed, one should know which time-dependent determinants are associated with PA and how strong these associations are. OBJECTIVE The aim of this study was to examine within-person associations between multiple determinants of the capability, opportunity, motivation, and behavior framework assessed using Ecological Momentary Assessment (EMA) and accelerometer-assessed light PA, moderate to vigorous PA, and total PA performed at 15, 30, 60, and 120 minutes after the EMA trigger. METHODS Observational data were collected from 64 healthy older adults (36/64, 56% men; mean age 72.1, SD 5.6 y). Participants were asked to answer a time-based EMA questionnaire 6 times per day that assessed emotions (ie, relaxation, satisfaction, irritation, and feeling down), the physical complaint fatigue, intention, intention, and self-efficacy. An Axivity AX3 was wrist worn to capture the participants’ PA. Multilevel regression analyses in R were performed to examine these within-person associations. RESULTS Irritation, feeling down, intention, and self-efficacy were positively associated with subsequent light PA or moderate to vigorous PA at 15, 30, 60, or 120 minutes after the trigger, whereas relaxation, satisfaction, and fatigue were negatively associated. CONCLUSIONS Multiple associations were observed in this study. This knowledge in combination with the time dependency of the determinants is valuable information for future interventions so that suggestions to be active can be provided when the older adult is most receptive.