This study analyzes the data coming from a web conjoint survey administered to a sample of HCWs in Europe. The web survey was implemented through the online software platform Qualtrics Experience Management. The survey was launched in April 2022 and ended in November 2023. To get access to the survey, the respondent required only an anonymous survey link, which could be easily shared via email, social networks, QR codes, or private messages, ensuring a user-friendly and widespread dissemination. Within the DigiCare4You Project, local partners from the 4 implementation countries were in charge of distributing the survey link to HWs within their network, and those who received the link were asked to further circulate it among their peers. By actively encouraging participants to share the survey link within their working networks, we seek to foster a widespread reach, initiating a dynamic snowball sampling mechanism [45]. While acknowledging that the resultant sample is nonprobabilistic, this strategic approach plays a pivotal role in attracting a diverse range of respondents and enriching the data with a multitude of perspectives.

The WCS is structured across three main sections: (1) the first section includes demographic details of respondents, encompassing age, employment, and years of education; (2) the second section evaluates the baseline individual attitude toward technology. In particular, the goal is to understand the propensity of the respondents to use technological devices in their working life and in the managing of their lifestyle habits, by looking at their level of agreement with some statements; (3) the third section provides the DCE experiment to elicit the HCWs preferences on the novel platform to monitor the health status of their patients. The survey was implemented in 5 languages, which include the 4 official languages in the 4 implementation countries of the DigiCare4You project plus the English language. The survey was completely anonymous and does not collect personally identifiable information. The complete survey (in English) and the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) checklist are available as Multimedia Appendix 1 and Checklist 1, respectively.

Considering previous evidence pointing out that the usage of technologies in working activities and in daily life may influence the adoption of digital tools by the health workers [20], the second section of the survey explores the HCWs’ baseline attitudes toward technology. Participants were asked to indicate, on a scale ranging from 1-10, their level of agreement with 10 statements. These statements were designed to capture attitudes and beliefs related to the use of technological devices, with 5 statements focusing on working activities and the remaining 5 statements on the integration of technology into lifestyle management. In this scale, a rating of 1 signifies “totally disagree,” while a rating of 10 indicates “totally agree.” The list of statements is presented in Table 1.

Using their responses to such statements, we intend to quantify the HCWs’ baseline attitude toward technology. In particular, we intend to estimate their latent attitude in using technological tools for working activities and lifestyle habits using IRT models. These models provide a robust framework for estimating an individual’s latent trait, based on their responses to a set of items [44]. This methodology offers valuable insights into abilities or latent characteristics by considering individual responses across a range of assessment items. IRTs are widely used in educational [46,47], psychological [48], and clinical [49,50] measurement. Using an IRT model, we can estimate, through the maximum likelihood approach, the values assumed by the latent trait of each respondent.

In this study, our primary goal was to assess the attitudes of HCWs toward technology usage, distinguishing between their perspectives on professional tasks and leisure pursuits. To achieve this, we analyzed their responses to distinct sets of items tailored to each aspect, with each item rated on a scale from 1 to 10. To accurately capture and interpret these ordinal responses, we used a specialized form of IRT known as the Graded Response Model (GRM) [51,52]. This model is well suited for assessing attitudes expressed through ordered categories, providing a nuanced understanding of the varying degrees of agreement or disagreement with the given statements related to technology use. In this setting, we estimated 2 separate GRMs to predict the individual attitude of health workers to use technological tools in their working activities and in the management of their lifestyle habits. We denoted these predicted traits as θ^W and θ^H, respectively. In addition, to enhance the robustness of our analysis, we also estimated the overall latent HCWs’ attitude toward technology θ^WH by applying GRM to all items pertaining to statements on technology usage in both professional tasks and daily routine. Before fitting GRMs, we noted that there were a few missing entries in the responses to the items. Since, upon exploring the data, we did not observe relevant patterns that rule the missingness in the items, we can impute them through Predictive Mean Matching (PMM) [53,54]. All the descriptives related to the missing data in the items and the performance of the PMM algorithm in imputing missing entries are available in the Multimedia Appendix 2. Finally, after estimating the latent traits, we identified the determinants influencing these traits by means of linear regression models with the latent traits θ^W , θ^H and θ^WH as the outcome variables and individual health worker’s characteristics as regressors. The independent variables in these models encompassed the health worker’s gender, the country of residence, the binary variable less than 30 (indicating whether the health worker is less than 30 years old), the binary variable high education (equal to 1 if the health worker has more than 14 years of education), the variable electronic health record (indicating the health worker’s experience with electronic health records), and the binary variable doctor (taking the value 1 for general practitioners and specialized doctors and 0 for the other health workers’ employment positions).

The third section of the survey provided a DCE [42] to assess the HCW’s preferences on the digital health platform they might use to monitor the health status of their patients. DCEs are widely used in several fields of social sciences [55,56] and health care applications [42] as they are flexible and easy to interpret. They enable researchers to elicit the user preferences on a specific issue without explicitly asking them, as respondents face some pairwise comparisons and choose their preferred option between 2 competing scenarios, whose characteristics (attributes) provide different options (levels).

In this setting, HCWs indicated the preferred platform for patient data monitoring between 2 competing platforms that differed in the characteristics of some of their aspects. These aspects have been identified based on a literature review of studies using DCE in similar contexts. In addition, insights were gathered through focused group discussions involving health workers and health care policymakers across the 4 implementation countries (details about the participants and the output of the focus groups are available in the Multimedia Appendix 3).

The key aspects considered for the web platform are as follows:

Table 2 describes the attributes and levels that have been used in the DCE. In the proposed DCE, respondents are asked to pick their preferred platform for patient data monitoring between two competing platforms, whose main characteristics –summarized by the attributes- provide different options –represented by the levels. The DCE data were analyzed using alternative-specific conditional logit regression models, in which the outcome was HCW’s preference (ie, platform A or platform B).

The study was carried out in compliance and accordance with the General Data Protection Regulation (2016/679) and the Italian Legislative Decree No. 196/2003 (“Personal Data Protection Code ").

The data used in the study were collected from a web survey, which is completely anonymous and does not collect personally identifiable information. In particular, a link to a survey was sent to individual email addresses or through a QR code. Responses were submitted directly to a survey software package (not returned via email). The survey included a disclosure letter describing the project and the purpose of the survey, the section of the questionnaire, the length of time of the survey, and that participation and consent were totally voluntary. The questionnaire did not gather any personally identifiable information for any purpose, or a combination of identifiers that might make it more likely to identify an individual. The individual respondents’ responses and data could not be linked to their email. No incentives were offered.

Since the data do not interfere with the rights and freedoms of the respondents, an evaluation by the Ethics Committee is not necessary. According to the Declaration of Helsinki on “ethical principles for medical research involving human participants” (Preamble, Art. 1), such evaluation is required only when the medical research involves humanly identifiable material or data. In addition, the DigiCare4You project received the ethical approval number 34-19/11/2021 from the Harokopio University Research Ethics Committee.

In this section, we present some preliminary descriptive evidence based on our dataset, (additional descriptive results are available in the Multimedia Appendix 4). The final dataset comprises responses from 260 HCWs situated in the 4 implementation countries of the DigiCare4You project; 102 respondents from Albania, 62 from Bulgaria, 68 from Greece, and 28 respondents from Spain. As more extensively discussed in the Multimedia Appendix 2, data notably exhibit some missing entries, as respondents had the option to skip questions they chose not to answer. Specifically, 3 respondents abstained from answering over half of the available questions; hence, we made the decision to exclude them from the analysis.

The distribution of sociodemographic characteristics across the 4 implementation countries is depicted in Table 3.

The reports include both the overall distribution of the variables and the stratified distribution by countries. Reported data include both absolute and relative frequencies, expressed in percentages. Upon examination of the table, it becomes evident that the majority of respondents were female (204/260, 78%), relatively young (171/260, 66% aged between 18 and 44 years old), and highly educated, with 200/260 respondents reporting having more than 16 years of education. Notably, 116/260 (45%) respondents were employed as medical specialists, followed by 62/260 (24%) respondents employed as general practitioners. Differences emerge when the 4 countries are compared. For instance, Albanian HCWs tend to be younger compared to their counterparts, while respondents from Spain exhibit the highest percentage of higher-educated respondents (26/28, 93% of the respondents declared having more than 15 years of education). Greece exhibits the highest percentage of respondents employed as dietitians or nutritionists (19/68, 28%) and health visitors (13/68, 18%)—qualified nurses who have additional training in public health and who offer support to families with young children. Albania has the highest percentage of respondents employed as general practitioners (40/102, 39%), followed by Spain and Bulgaria (7/28, 25% and 15/62, 24%, respectively).

As we have previously hinted throughout the paper, we used the information about the level of assessment of the 10 statements related to the HCWs’ attitude toward technology to predict their latent attitude in using such technologies. Specifically, we separately used the items related to the usage of digital technologies in working activities and to the usage of technological devices in the management of daily habits. The IRT models fit the data well, as all parameters resulted in being statistically significant, and all items contributed to explaining the latent traits. The complete estimation and fitting graphs are available in the Multimedia Appendix 5.

Once these traits were estimated, we explored whether socioeconomic factors have a role in determining their variability. Figure 1 graphically reports the main results of the 3 linear regression results in which the outcome variables were θw^θH^ and θ^WH the latent attitudes of HCW toward using the technology during working activities, daily habits, and both, respectively. Complete tables are available in the Multimedia Appendix 5. In particular, Figure 1 provides an overview of the estimated coefficients in the 3 models, together with their corresponding 95% CIs. Results suggest that (1) males exhibit a less pronounced attitude in using technologies compared to women, and this effect is particularly significant when estimating the latent trait targeted at working activities θ^W and the comprehensive latent trait θ^WH; (2) HCWs in Spain and Bulgaria demonstrate a lower propensity to use technologies compared to their Albanian counterparts; (3) being highly educated leads to a higher inclination to use technologies; and (4) doctors exhibit a lower propensity to use technologies compared to HCWs referred to other employment categories.

In this section, we reported the DCE results regarding the features that represent potential barriers or facilitators that may encourage or discourage HCWs in integrating a given digital health platform for patients’ data monitoring in their working routine. Moreover, we investigate whether their attitude toward the usage of technologies in working activities and in the management of their habits shapes their preferences about the ideal mHealth platform for patients’ data monitoring.

In the DCE used in this setting, HCWs faced 6 pairwise comparisons (named Choice Sets), each presenting 2 alternative characterizations of the platform, which differed in the levels of their attributes. Figure 2 reports the main results of the DCE analysis, obtained by applying conditional logit models [57], where the response variable is the choice variable, and the regressors are the 4 attributes of the alternatives.

Specifically, to analyze data from the DCE, we run three conditional logit models. Figure 2 reports the estimated odds ratios, together with their corresponding 95% CIs of each model: (1) the first one (blue lines) provides the inclusion as independent variables of the 4 attributes only, (2) the second one (orange lines) also includes pairwise interactions between the 4 attributes and the 2 estimated latent traits θ^W and θ^H, and (3) the third one (green lines) still includes interactions but with respect to the comprehensive latent trait θ^WH. Results here are presented in terms of odds ratios for ease of interpretation (odds ratios greater than 1 signal that the presence of a given feature encourages the respondents to choose that alternative). The coefficients related to the attributes that affect the HCWs’ choice of the ideal platform for patient data monitoring are highly significant.

By looking at the figure, we can definitely state that HCWs (1) prefer to analyze data through graphs rather than through tables, (2) find it very useful to have comparative statistics that relate data of their patients to a standard, rather than simply observing data of their patients, (3) prefer to have information on the patients’ clinical history rather than on their emotional status, and (4) prefer to receive an active group training on the platform rather than completing self-tutorials.

By looking at the interactions, we can also state that (1) having a higher attitude in using technologies in working activities (θ^W) and having a high comprehensive attitude toward technology (θ^WH) significantly increases the preference of having statistical data about the comparison of patients against a standard, instead of having data related to the given patient only; and (2) an increased attitude toward the usage of technological devices in managing lifestyle habits (θ^H) and toward technologies as a whole (θ^WH) decreases the preference of having a self-training program to learn the platform, instead of a group training.

As a consequence, the ideal digital health platform for patient data monitoring should provide intuitive graphs, statistics that compare patients’ data against a standard, information about the patients’ family clinical history, and health workers should be trained for the usage of that platform through a group training led by an instructor. The characteristics of the ideal platform chosen by HCW for patient data monitoring are represented in Figure 3.

This contribution proposes an integrated approach based on IRT and DCE (1) to evaluate the HCWs’ attitude toward the usage of technological devices in their working activities and in managing their lifestyle habits, (2) to explore the role of sociodemographic characteristics in shaping these attitudes, and (3) to appraise whether their attitude toward technology shapes their preferences about a novel digital health platform for patient data monitoring. First, results suggest that health workers show an overall positive attitude toward the usage of technological devices both in their working activities and in their lifestyle habits. This propensity is particularly strong for women and highly educated professionals, while it appears to be weaker for doctors compared to other health care professionals. Similarly to the study by Muiruri et al [17], the employment position plays a role in determining the attitude toward technology.

The different country of the respondents also affects the attitude toward technology, as Albanian HCWs have a stronger propensity to integrate technological tools within their daily routines compared to the other countries. These findings are in line with the review done by Borges et al [26], who identified that country-specific technologies can affect the adoption of digital health devices. Thus, statistically different results may be hypothesized as the respondents being a proxy of their countries’ readiness to adopt these technologies.

In some studies [17], sociodemographic variables, such as gender and education, do not play a role in determining the HCWs’ willingness to adopt technologies, while in our contribution, they result in significantly impacting the HCWs’ attitude toward technology, as already pointed out in Borges do Nascimento et al [26]. Second, the DCE results suggest that the type of training, the way data are presented, and the information provided by the platform play a pivotal role in shaping the HCWs’ willingness to adopt a digital platform for patients’ data monitoring. The inclination for personal training instead of self-guided tutorials was already pointed out before [4,19,20], besides the findings of [21,26], indicating a concern with the quality of the training received by the staff.

The preferences regarding data looking in the format of graphs and statistics compared to a standard hint to preferences aligned with more intuitive and user-friendly interfaces, which are also previously mentioned in other studies [4,14,20,21] as facilitators of the adoption of digital health tools by the HCW.

To the best of our knowledge, there are no existing studies that jointly explore the HCWs’ attitude toward technology and their preferences about such digital technologies. Our findings suggest that there are relevant interactions between these 2 aforementioned topics, and this issue should be further explored, as it provides a global understanding of the approach of HCWs toward digital health systems.

This work has some limitations. The main limitation is that the sample of individuals asked to complete the WCS might be affected by a selection bias due to the snowball sampling nonprobabilistic design. The sampling technique was chosen based on the availability of resources in the 4 countries under study. An additional limitation of the study is that the sample size in each of the 4 countries, even if it is relatively large, does not allow us to conduct a stratified analysis, where we separately implement the integrated analytical framework in each country and then we inspect differences among countries. The relatively low response rates and participation, especially in some countries like Spain, can be due to people’s fatigue with invitations and participation in various surveys (oversurveying), especially after the COVID-19 period of dominance of e-communication. However, this work also offers valuable contributions. First, it proposes an integrated analytical approach based on IRT and DCE to simultaneously investigate the HCWs’ attitude toward the usage of technologies, the HCWs’ preferences about potential digital tools to be integrated within their working routine, and potential interactions between the aforementioned topics. This approach empowers us to delve into the intricate dynamics between HCWs’ fundamental attitudes toward technology and their specific inclinations toward the novel platform. This exploration adds a layer of depth to our understanding, allowing for a more comprehensive assessment of the factors influencing their preferences. Second, this study covers a wide international cohort of HCWs enrolled in multiple European countries. Therefore, it does not focus on specific experiences that refer to empirical contexts that are likely to be narrow both from the spatial and the temporal perspective but attempts to obtain findings that can be potentially generalized.

The analytical framework presented in this work, merging IRT and DCE, elevates our comprehension of HCWs’ propensities regarding the integration of digital health tools. The breadth of the analysis, encompassing a wide international cohort of HCWs, places the results as not just informative but also highly applicable across a spectrum of contexts. This potential generalizability underscores the significance of the findings, making them a relevant resource for shaping strategies and targeting interventions in various health care settings.

The findings from this study have significant policy implications for the integration of digital systems in health care systems. First, the study reveals that HCWs generally hold a positive attitude toward the use of technological devices, emphasizing the potential benefits of these tools in enhancing health care efficiency. It follows that political authorities should push for the progressive integration of digital health systems, integrating patient-generated data into the daily workflows of health care workers to enhance care delivery by improving the available infrastructures through which digital tools are implemented and promoting laws and programs to encourage the gradual development of a technologically empowered health care system. Second, the research highlights the influence of socioeconomic factors in shaping health workers’ attitudes toward digital health tools. Policymakers should take into account these factors when designing strategies for the adoption of such technologies, targeting specific support to those categories that have a weaker propensity for adopting these technologies. Furthermore, the study underscores the importance of considering specific characteristics of platforms for patients’ data monitoring, as these features significantly impact HCWs’ willingness to integrate them into their working routines. Despite the relevance of our results, the attitude of HCWs toward digital tools and their preferences toward the ideal characteristics of these tools should be further explored in future studies to enhance the global understanding of this topic and to provide even stronger evidence for policymakers. Future works may consider wide international samples. In addition, researchers should deeply understand which are the organizational and infrastructural barriers that still limit the development of digital health systems and which are the possible approaches to enhance the HCWs’ attitude toward these technologies.