RefCheck Maintenance Notice

On Monday, December 3, 2018, from 16:00-18:00 EST, RefCheck will be undergoing maintenance. RefCheck is the process where, during copyediting, all references are extracted from the manuscript file, parsed, matched against various databases (eg, PubMed and CrossRef), and automatically corrected. For more information on RefCheck, please visit our Knowledge Base.

Who will be affected?


Citing this Article

Right click to copy or hit: ctrl+c (cmd+c on mac)

Published on 30.11.18 in Vol 6, No 4 (2018): Oct-Dec

Preprints (earlier versions) of this paper are available at, first published Jun 08, 2018.

This paper is in the following e-collection/theme issue:

    Original Paper

    Most Influential Qualities in Creating Satisfaction Among the Users of Health Information Systems: Study in Seven European Union Countries

    1Department of Health Science, Blekinge Institute of Technology, Karlskrona, Sweden

    2Department of Mechanical Engineering, Blekinge Institute of Technology, Karlskrona, Sweden

    3Department of Technology and Aesthetics, Blekinge Institute of Technology, Karlskrona, Sweden

    Corresponding Author:

    Shahryar Eivazzadeh, MSc, Lic

    Department of Health Science

    Blekinge Institute of Technology

    Blekinge Tekniska Högskola

    Karlskrona, 371 79


    Phone: 46 455 38 50 00



    Background: Several models suggest how the qualities of a product or service influence user satisfaction. Models such as the Customer Satisfaction Index (CSI), Technology Acceptance Model (TAM), and Delone and McLean Information Systems Success demonstrate those relations and have been used in the context of health information systems.

    Objective: This study aimed to investigate which qualities foster greater satisfaction among patient and professional users. In addition, we are interested in knowing to what extent improvement in those qualities can explain user satisfaction and whether this makes user satisfaction a proxy indicator of those qualities.

    Methods: The Unified eValuation using ONtology (UVON) method was used to construct an ontology of the required qualities for 7 electronic health (eHealth) apps being developed in the Future Internet Social and Technological Alignment Research (FI-STAR) project, a European Union (EU) project in electronic health (eHealth). The eHealth apps were deployed across 7 EU countries. The ontology included and unified the required qualities of those systems together with the aspects suggested by the Model for ASsessment of Telemedicine apps (MAST) evaluation framework. Moreover, 2 similar questionnaires for 87 patient users and 31 health professional users were elicited from the ontology. In the questionnaires, the user was asked if the system has improved the specified qualities and if the user was satisfied with the system. The results were analyzed using Kendall correlation coefficients matrices, incorporating the quality and satisfaction aspects. For the next step, 2 partial least squares structural equation modeling (PLS-SEM) path models were developed using the quality and satisfaction measure variables and the latent construct variables that were suggested by the UVON method.

    Results: Most of the quality aspects grouped by the UVON method are highly correlated. Strong correlations in each group suggest that the grouped qualities can be measures that reflect a latent quality construct. The PLS-SEM path analysis for the patients reveals that the effectiveness, safety, and efficiency of treatment provided by the system are the most influential qualities in achieving and predicting user satisfaction. For the professional users, effectiveness and affordability are the most influential. The parameters of the PLS-SEM that are calculated allow for the measurement of a user satisfaction index similar to CSI for similar health information systems.

    Conclusions: For both patients and professionals, the effectiveness of systems highly contributes to their satisfaction. Patients care about improvements in safety and efficiency, whereas professionals care about improvements in the affordability of treatments with health information systems. User satisfaction is reflected more in the users’ evaluation of system output and fulfillment of expectations but slightly less in how far the system is from ideal. Investigating satisfaction scores can be a simple and fast way to infer if the system has improved the abovementioned qualities in treatment and care.

    JMIR Med Inform 2018;6(4):e11252





    The normative evaluation of health information systems is articulated through a frequently used set of keywords such as acceptance or adoption [1,2], success [3], and satisfaction [4,5]. Each of these keywords reminds us how a health information system inherits traits from its conceptual ancestors, that is, the information system, technology, and product. For an overall evaluation of these systems, one might measure how well these information systems succeed [6,7], how these technologies are accepted by users [8,9], or how the customers of these systems, patients, or professionals are satisfied with these products [10]. Below this layer of top indicators, there exist sets of constructs and relationships that cause success, acceptance, or satisfaction. Researchers have tried to capture and demonstrate through the models how success [6,10], acceptance [9], or satisfaction [11-13] are created by constructs such as perceived quality, perceived expectation, ease of use, and other variables. Some of these models have largely been employed in diverse contexts [14]. There are also models, whether novel or customized from the mainstream, that are specific to a smaller context such as health information systems [15-17].

    The Customer Satisfaction Index (CSI) model family places the satisfaction construct at the core of their path structures. There, the satisfaction construct is affected by leading indicators such as perceived quality. At the same time, it impacts lagging indicators such as user loyalty. There are at least three versions of CSI widely being used. The original CSI model was introduced in Sweden [11]. The American Customer Satisfaction Index improved the Swedish version [12], and then the European Customer Satisfaction Index (ECSI) enhanced the American version [18]. The ECSI model consists of 9 latent construct variables [18], which are measured by a series of measure or manifest variables. Historically, CSI models have been used at macro levels where the satisfaction of customers at the national level or the level of an enterprise was the matter of concern. The wording of CSI models, such as the customer term, and the inclusion of some constructs, such as loyalty, suggest that measuring user satisfaction at the micro level, that is, the product level, was not their main target. However, the manifestation of the satisfaction concept in the CSI models through its 3 manifest variables [18] is versatile enough to measure satisfaction at both micro or macro levels with the same wording. CSI models introduce a way of measuring satisfaction scores through adding the weighted scores of those 3 variables, which inspires similar approaches in various disciplines. In comparison with using the CSI models for health information systems, one might consider the patient satisfaction models [19] that share a set of common constructs and relations with the CSI models but do not necessarily embed the same structures or components.

    The Delone and McLean Information Systems Success (D&M IS) model, a prevalent model for analyzing the success of information systems, sets out the relationship between user satisfaction and quality dimensions [6,20]. In this model, 3 categories of qualities, information, system, and service contribute to user satisfaction. There are 2 other constructs, net benefits and intention to use, that are in a bidirectional impact relationship with user satisfaction [20]. Several studies have validated the causal relationships between the constructs in the D&M IS model [21]. In addition, there exists a long list of validated measures for each of the constructs [22]. The D&M IS model has broadly been used in the health information system context [23]. Furthermore, it has also been extended and customized to be more specific for this context, such as in the Human, Organization, and Technology Fit model [24], but the extension has been directed more toward a wider perspective of organization and technology. Although some studies have incorporated [25] or prioritized [26] more specific qualities, further investigation is needed to be more specific about the impacting qualities and their degree of importance.

    Technology acceptance models, such as Technology Acceptance Model (TAM) as well as the Unified Theory of Acceptance and Use of Technology, which are supported by a great many of studies [8,9], placed the acceptance of a system or technology at the core of relationships. These models have been applied in health information system studies, although they have reported the significance of some of relationships differently [8]. The user acceptance of a health information system can be a prelude to or reflection of their satisfaction in using that system, but acceptance is not the same as satisfaction. Some researchers have considered acceptance, that is, the behavioral intention to use in TAM, an equivalent for satisfaction [27], but the intention to use is a different construct from satisfaction in a well-studied model such as D&M IS.

    Contextualizing TAM by adding variables has been a common practice [8,28,29]. Indeed, contextualizing TAM for health information technology has led to the introduction of some frequently employed variables, such as fit [30]. However, there is a shortage of studies applying a systematic approach, such as belief elicitation [31], when introducing a new variable [8].


    For the CSI, D&M IS, and TAM, the set of relationships between their proposed constructs have been already examined in various contexts [8,14,13]. Nevertheless, for a specific context such as health information systems, one might seek to develop new models, probably inspired by those that are well established, to expand a construct into more detailed constructs or find manifest variables more relevant to a case. For example, the constructs that represent the qualities of a system are generalized in those models or their variations as perceived quality [13]; system, service, or information quality [20]; and output quality [8]. However, none of those models represents a specific quality, such as safety, as a stand-alone construct.

    In addition, finding a systematic approach to define manifests for construct variables, as mentioned before, is another direction for extending a model [8]. Many of the evaluation frameworks for health information systems suggest the qualities to evaluate [32], arranged as categories or domains. These frameworks implicitly suggest constructs and the manifests to each construct. Simultaneously, the end users of health information systems are another source for eliciting the qualities and their groupings [33].

    In the forthcoming sections, we put forth a list of qualities that create and predict user satisfaction with health information systems. The qualities are embedded within a path model that demonstrates their relationships with user satisfaction. This study’s methods and materials are discussed in the Methods section. The qualities elicited from the Future Internet Social and Technological Alignment Research (FI-STAR) project by applying the Unified eValuation using ONtology (UVON) method are reported in the Results section, an exploratory result is demonstrated in the Correlation Patterns section, and an estimated model is presented in the Partial Least Squares Structural Equation Modeling Path Models section. The European Union (EU)–wide empirical data collected through the FI-STAR project, detailed in the Methods section and Multimedia Appendix 1, is used to calculate and validate the model in the Partial Least Squares Structural Equation Modeling Path Models section. The results of exploration and model calculations are discussed in the Correlations section and the Partial Least Squares Structural Equation Modeling Path Models section. Subsequently, based on the model, the relative importance of qualities in creating and predicting user satisfaction is discussed in the Most Influential Qualities section. In the section Satisfaction Index, weightings are suggested for the calculation of a satisfaction index for health information systems. Finally, we examine the limits and extensions to our approach and suggested model in the Extensions and Limitations section.


    Data Collection

    The empirical data for this study have been collected from the FI-STAR project, an EU electronic health (eHealth) project with 7 subprojects across the EU [34]. A convenience sampling approach was used for the recruitment of the participants. Each eHealth app was deployed in a hospital or health facility site and the users on the site—patients or health professionals—were asked about their assessment of the impact of that specific solution on treatment. Participation in the trials, and therefore the survey, was voluntary with no mentioned preconditions. There was no constraint on the type of eHealth project being developed provided that they follow the FI-STAR requirements, especially using the FIWARE infrastructure. However, most of the subprojects could be categorized as telehealth apps. A summary of all subprojects can be found in Multimedia Appendix 1.

    Data Analysis

    We applied the UVON method [33] to the FI-STAR requirement documents, together with the evaluation aspects from the Model for ASsessment of Telemedicine framework [35]. The quality aspects appearing in the result of the UVON method are supposed to be provided by the eHealth apps developed in the FI-STAR project. For each quality appearing in the UVON’s output, a question was formulated according to that specific quality in the treatment. The questions were categorized according to the resulting domains in the UVON’s output. The answer alternatives to the questions were formed as a 5-point Likert rating scale with unweighted scores. There were 20 qualities alongside the 3 user satisfaction questions from ECSI [18] that were converted to 2 questionnaires. One questionnaire was customized for the patients and the other one for the health professionals. The content of the questionnaires can be found in Multimedia Appendices 2 and 3. Responses to the questionnaires by 87 patients and 31 health professionals, physicians or nurses, were used for the models suggested in this study.

    In 2 steps, we arrived at a model based on the empirical data from the answers to the questionnaires. The calculations were done using the R language, version 3.4.0 [36]. The bootstrapped significance calculation was performed in SmartPLS software version 3.2.7 (SmartPLS GmbH) [37].

    In the first step, a matrix of Kendall correlation coefficients tau (τ) for each of the patient and professional questionnaires was formed. The results are presented in Figures 1 and 2. Moreover, we used Cronbach alpha (α) test to measure the consistency of the results in the UVON-suggested families of qualities as hints for finding constructs in the later steps.

    In the second step, we created a PLS-SEM path model. For each set of the qualities that have already been grouped by the UVON method and show a high degree of correlation, a latent construct variable was considered. These latent variables are not directly measurable but manifest themselves through quality and satisfaction variables. If a quality corresponds only with one question in the questionnaire, one latent proxy variable was considered. Consequently, it would be possible to add more measure variables to the same latent variable in future studies. The PLS-SEM analysis was performed using the matrixpls library in R, version 1.0 [38]. The sample size adequacy calculations were performed using G*Power version 3.1 (Heinrich Heine University Düsseldorf), a program for statistical power analysis for a variety of statistical tests [39].

    The result of PLS-SEM should be interpreted in the context of the questionnaire. Accordingly, as discussed in the Partial Least Squares Structural Equation Modeling Path Models section, negative coefficients were considered noninformative and were excluded from the final results. The validity of the result was demonstrated through a toolbox of significance, discrimination analysis, internal consistency reliability, and convergence validity. The calculation of significance indicators was performed applying the bootstrapping approach using SmartPLS software [37]. Whenever applicable, the noninformative nature of negative coefficients was considered during the validity and fitness calculations [40].

    Figure 1. Correlation matrix for the patient questionnaire results across all cases. For the details of each variable, refer to the corresponding question in Multimedia Appendix 2. Insignificant (P>.05) results are left blank. Negative results are marked with leftward slanting lines. Note that the qualities grouped by the Unified eValuation using ONtology (UVON) method usually show higher correlations together.
    View this figure
    Figure 2. Correlation matrix for the professional questionnaire results across all cases. For the details of each variable, refer to the corresponding question in Multimedia Appendix 3. Insignificant (P>.05) results are left blank. Note that the qualities grouped by the Unified eValuation using ONtology (UVON) method usually show higher correlations together.
    View this figure


    Unified Evaluation Using Ontology Method Outcome

    The result of applying the UVON method on the FI-STAR project was a tree-style ontology of qualities [33], of which the top-level qualities are listed in Table 1. The questionnaires articulate those qualities and their more specific subqualities. Table 1 is just an overview of the qualities; details of the questions that were created for each quality can be found in Multimedia Appendices 2 and 3.

    Descriptive statistics of the variables in the patient and professional questionnaires, including mean, SD, and median, are shown in Tables 2 and 3. Details of the corresponding question for each quality in Tables 2 and 3 can be found in Multimedia Appendices 2 and 3.

    Correlation Pattern

    As can be seen in Figures 1 and 2, a spectrum of weak to strong correlation coefficients appeared in the Corrgram diagrams [41] for the patient and professional questionnaire. The blank cells are the results that were not statistically important (P>.05). The results of the Cronbach alpha (α) test can be found in Table 4.

    Table 1. The quality attributes resulting from applying the Unified eValuation using ONtology (UVON) method to Future Internet Social and Technological Alignment Research (FI-STAR) requirement documents.
    View this table
    Table 2. Descriptive statistics of the variables in the patient questionnaire.
    View this table
    Table 3. Descriptive statistics of the variables in the professional questionnaire.
    View this table
    Table 4. Cronbach alpha (α) test results for the quality groups.
    View this table

    Partial Least Squares Structural Equation Modeling Path Models

    The 2 PLS-SEM models and their loadings and coefficient values are depicted in Figures 3 and 4. As is common with path models, latent variables are depicted as ovals, whereas manifests are shown as boxes. We considered all the measure variables as reflective, that is, they do not construct their respected latent variables, but they measure or manifest them. The number label on the edge between a manifest and a latent variable is the loading, and the number label on the edge between two latent variables is the coefficient.

    The most contributing and predictive qualities regarding satisfaction are reported in Table 5 by specifying the coefficients of relationships between qualities and the satisfaction construct. For the patients, the coefficients of effectiveness, safety, and efficiency qualities were higher than the average (.13) of all coefficients. Similar to the professional, coefficients of the affordability and effectiveness qualities were higher than the average (.51) of all coefficients.

    The relationship of each measure variable in the path model with its construct is associated with weights. The standardized weights for the satisfaction construct measure variables are required for calculating the user satisfaction index and can be found in Table 6.

    Figure 3. Partial least squares path model for the patient questionnaire. The constructs are shown as ovals, and the number between constructs is the coefficient value. Manifests are shown as rectangles, and the number between a manifest and a construct is the loading value of that manifest.
    View this figure
    Figure 4. Partial least squares path model for the professional questionnaire. The constructs are shown as ovals, and the number between constructs is the coefficient value. Manifests are shown as rectangles, and the number between a manifest and a construct is the loading value of that manifest.
    View this figure
    Table 5. The coefficients of the qualities to satisfaction relationships in the partial least squares structural equation modeling path model show which qualities contribute more to satisfaction.
    View this table
    Table 6. Standard weights for calculating the satisfaction index, based on the manifest variable loadings for the Satisfaction constructs in the patient and professional path models.
    View this table
    Table 7. Significance of the quality to satisfaction relationships by calculating the P values of the relationships between the qualities and the Satisfaction construct. If a relation does not exist in model, the corresponding cell in the table is left blank.
    View this table

    The P value results of calculating the significance of quality to success relationships using the bootstrapping approach are shown in Table 7. The P values below .05 are marked with footnotes. Regarding discriminant analysis, the results of the Heterotrait-Monotrait (HTMT) ratio are demonstrated in Table 8. HTMT being below 1.0, preferably 0.9, satisfies the discriminatory criterion [42]. In addition, Multimedia Appendix 4 depicts cross-loadings in the path models. The effect sizes of the samples were enough to show significant results for the highest loading constructs, as shown in Table 5. Details of the effect sizes and their associated power, by recalculating the PLS-SEM and focusing on significant relations, are shown in Multimedia Appendix 5. The list of acronyms is provided in Multimedia Appendix 6.

    Table 8. The discriminant validity analysis shows if the manifests of a construct in the patient or professional Partial Least Squares Structural Equation Modeling models have the strongest relationship with that construct compared with another construct. If a relation does not exist in model, the corresponding cell in the table is left blank.
    View this table
    Table 9. The result of internal consistency reliability of the manifest variables by calculating composite reliability and their convergence by measuring average variance extracted, grouped by constructs.
    View this table

    For internal consistency reliability, composite reliability (CR) values, and for convergence validity, average variance extracted (AVE) values were calculated for each construct, as depicted in Table 9. The minimum CR should preferably be above 0.7 [43]. The minimum AVE should preferably be above 0.5 [44].



    This study advances and prioritizes the qualities in health information systems that determine and predict user satisfaction, both for patients and health professionals. As a secondary outcome, it also suggests weightings for calculating the satisfaction index. The outcomes of the study exhibit the effectiveness of the UVON method in proposing quality constructs that can be applied to a path analysis. Conclusions from the results are achieved in 2 steps. First, the correlations give better insight about the groupings of the qualities as manifest variables of the latent constructs. Second, the path model justifies and quantifies the relationship between those grouped qualities, their latent construct variables, and the satisfaction construct.


    In the exploratory step, as shown in Figures 1 and 2, strong correlations appear between the qualities that have been already grouped into a family by the UVON method. These high correlations result from the semantic unification of qualities across branches of an ontological tree by the UVON method [33]. For example, a set of above-medium correlations exist within the efficiency, effectiveness, and safety family of qualities in the patient questionnaire results, as well as within the empowerment, effectiveness, safety, and efficiency families in the professional questionnaire results. The same is true for the satisfaction questions in both groups of questionnaires.

    The above correlations suggest that the members of a quality or satisfaction group can be combined. Alternatively, in other words, they manifest a common latent variable. It is worth mentioning that the Kendall correlation tau (τ) is less generous than Spearman rank correlation rho (ρ) in confirming the correlations [45]. Hence, there would be more confidence in positively interpreting the correlation results and suggesting a common latent origin. The Cronbach alpha results in Table 4 confirm the same explanation in the quality groups.

    Besides the possible existence of latent variables, there are 2 other corollaries to the correlations. First, the high degree of correlation between a family of qualities results in the multicollinearity problem. Multicollinearity makes ordinary regression techniques inefficient and the interpretation of the regression coefficients challenging [46]. Overall, 2 solutions can be taken here: choose one of the variables that show high correlation with each other by using variable selection methods, such as Least Absolute Shrinkage and Selection Operator, or apply a method that is tolerant to the problem. The PLS-SEM approach used in this study is tolerant to multicollinearity; meanwhile, it can investigate the causality relations between some correlated quality groups.

    Second, the correlation between qualities and satisfaction aspects suggests a causality relationship between quality and user satisfaction. Similarly, there are models, such as the CSI family of models [11-13] as well as the D&M IS model [20], that demonstrate a causality relationship between qualities and satisfaction in parts of their structure. We can draw on corroborations from the extensive amount of literature about those models, both to enrich our model and verify the results.

    A summary of the above discussion is that we can group the qualities within a family as manifests of a latent variable, consider a causal relationship from those quality latent variables to a satisfaction latent variable, and present these groupings and relationships through a PLS-SEM model.

    Partial Least Squares Structural Equation Modeling Path Models

    The PLS-SEM path model has traditionally been used to represent causalities for the CSI series [11,12,47], the models related to D&M IS [48,49], and similar intentions [50]. This prevalence of usage gives the opportunity to reuse some of those models’ parts, compare their structures, and collate their results. Other advantages of the PLS-SEM approach are the need for small sample size and the ability of handling non-normal data [50].

    The 2 PLS-SEM path models in this study comply with the general pattern in CSI, D&M IS, and TAM series models in which a central construct—be it called system success, user satisfaction, customer satisfaction, or user acceptance—is influenced by system qualities. Besides the use of different constructs, each model captures a distinct level of detail for the same or similar concepts. The CSI and TAM models are more concerned about the general perception of quality, whereas D&M IS examines further details about the qualities by considering 3 separate constructs: system quality, information quality, and service quality. The model presented in this study is inclined to be more domain-specific by focusing on the health information system domain. The model is also more concerned with the qualities improved in the whole treatment setting by using the health information system rather than the qualities of the system. Finally, in comparison with the previously mentioned mainstream models, the model in this study is more specific about the type of qualities and how they compare in determining and predicting user satisfaction.

    Before discussing the qualities with the most influence on satisfaction, model validity and the right way of interpreting the results need to be investigated.

    Regarding internal consistency reliability, CR indicators need to be higher than 0.7 [43], which is well satisfied (see Table 9). The CR shows if the manifest variables of each construct measure the same thing. The convergent indicator AVE needs to be more than 0.5 to indicate that more than half of the variance in the measures is because of the variance in the construct [44]. All the constructs in our PLS-SEM model satisfy this criterion (refer to Table 9).

    From the discriminatory validation perspective, both patient and professional models show indications of correctly assigning the measure to construct variables. The HTMT ratio demonstrates if the assignment of measuring variables to a specific construct is better, that is, more relevant, than other alternatives. All the HTMT results in both models, except one as shown in Table 8, satisfy the specified criterion of being less than 1. Furthermore, most of the HTMT values are less than 0.9, which confirms discriminant validity [42]. The only pair of constructs that have a ratio value above 1 is the effectiveness and satisfaction pair for professionals. However, this unfulfilled criterion might be justified considering that, in the domain, one should segregate satisfaction and effectiveness, while effectiveness highly contributes to satisfaction. The relatively high HTMT ratio for the pair of empowerment and effectiveness can indicate that the users’ empowerment to reach effectiveness is not very distinctive from the improved effectiveness.

    The interpretation of the negative coefficients that appeared in the models must be discussed. The wording in the FI-STAR questionnaires captures only the user perspective on positive relationships but not the negative ones. The information gathered from the questions is unidirectional. Therefore, one cannot interpret the negative coefficients as an indication that some qualities are inversely related to satisfaction.

    For example, if it is asked whether the app has increased the effectiveness of a treatment by decreasing the number of mistakes, a responder might answer “disagree.” This answer can mean whether the user does not believe that the app has decreased the number of mistakes in the treatment or the user might think that the app has decreased the number of mistakes but does not contribute to the system’s effectiveness. On the other hand, it is still a possible interpretation that the app has caused more mistakes; however, we cannot separate this interpretation from the other previously mentioned valid interpretations. Therefore, we can only confirm the positive part of relationships, where quality contributes to user satisfaction.

    Most Influential Qualities

    As shown in Table 7, there are constructs whose relationships to the satisfaction construct are statistically significant. Those constructs also have a considerable impact on satisfaction, as depicted in Table 5. Regarding satisfaction as the major contributor to success [51] and the indicator of voluntary acceptance [14], we extend our discussion to cover similar studies that report on these 2 indicators.

    Within the list of qualities in Table 7 with a significant relationship to the satisfaction construct, the degree of effectiveness is considerably predictive in creating satisfaction. Both patients and professionals care considerably whether an app has increased the effectiveness of treatment and care. This result highlights the nonintuitive contrast between the effectiveness of other qualities, such as the efficiency for patients, in affecting their satisfaction. Nevertheless, there can be alternative interpretations. For example, if the apps in the FI-STAR project could significantly improve efficiency for patients, efficiency might have shown an impact as the effectiveness on satisfaction.

    This result confirms the studies that consider effectiveness the major contributor or even equivalent to user satisfaction, generally, in information systems [14,51,52]. More specifically, this study parallels the studies that reported effectiveness (sometimes expressed as usefulness) as the most, or one of the most, influential qualities for the satisfaction of patients [53] or health professional users [54-56] in a variety of health information systems. Nevertheless, there exist studies that reached a different conclusion in prioritizing the most influential qualities [57].

    It should be noted that improvements in the effectiveness of the treatment are not articulated identically in all those studies. For comparison, one needs to consider this discrepancy [8]. Some studies have used performance [54], a term presumably borrowed from the D&M IS and TAM families. In addition, some studies reported similar manifest variables to effectiveness in our study—see Multimedia Appendices 2 and 3 —such as making fewer mistakes [58].

    According to our results, affordability has a high degree of impact on satisfaction for professionals, similar to effectiveness. The affordability to satisfaction relationship for the patients was not statistically significant in this study, whereas its magnitude was also negligible compared with other relations. An explanation might be that, in the FI-STAR setting, patients did not have to be concerned about the costs and affordability of solutions, whereas professionals might have a more holistic perspective. In a different setting, where patients are more concerned about treatment costs, their satisfaction might be more influenced by the improvements in affordability, showing higher significance and magnitude in the coefficient that relates affordability to satisfaction constructs.

    Although studies report increased affordability and cost reduction in treatment can improve the satisfaction and acceptance of health information systems [2,59], only some of these studies quantify or compare the degree that affordability and cost reduction affect satisfaction. In addition, many studies such as ours could not report definitively how patients perceived cost-reduction quality considering patients usually do not pay for treatment in the context of a study [5,60]. Some studies that rely on TAM models have considered affordability, alongside other factors, as a manifest to perceived usefulness [8,61,62]. These studies show a relatively high or above-average impact on professional users’ satisfaction, acceptance, or intention to use [61,62]. Our results in Table 5 comply with these studies. It is worth noting that some studies declared the same idea in a negative form, where being costly is considered a barrier to acceptance [2,60,63] or success [7].

    Patients showed some degree of improved satisfaction when there was an improvement in efficiency or safety. It is important to note that health information systems, as a side effect, can degrade the status of efficiency or safety in a health care setting [64]. Hence, their contribution to overall satisfaction can even be negative. However, as highlighted before, our questionnaires were not designed to differentiate between the states of negative impact and no impact.

    Considering most of the apps in the FI-STAR project could be categorized as telemedicine apps (see Multimedia Appendix 1), it could be predicted that efficiency, achieved by eliminating the hassle of distant travels, contributes to user satisfaction to some extent. Despite our initial expectations, the degree of impact on satisfaction, although existing, was less than the previously mentioned factors. Our expectation was based on similar studies that investigated the impact of efficiency improvements on satisfaction and acceptance: the degree of impact was recorded relatively more than our study [1]. It is also important to note that we considered a separate category for affordability and cost saving quality, whereas some studies considered cost saving a form of improving efficiency [65] or a manifest to perceived usefulness [62]; therefore, other results should be compared with more attention to this detail. From the other side, some studies considered other forms of efficiency, rather than affordability, as manifests to perceived usefulness [55,66,67]. Regarding the coefficients and loadings in models of the aforementioned studies, in comparison with our results, they have recorded a higher impact from efficiency on the satisfaction or acceptance of users.

    Looking at the safety questions for the patients in Multimedia Appendix 2, it seems that being informed about the situation and capable of keeping the situation in check is the source of the safety to satisfaction causality. Similar to other qualities, safety has been categorized in various constructs in the studies, whether as a manifest to provider performance [68], perceived usefulness [62], outcome [62], or information satisfaction [62].

    To the best of our knowledge, there are few studies that investigate the impact of safety brought by health information systems on the satisfaction or acceptance of the users of those systems. However, one should pay attention to this caveat that the safety concept might have divergent embodiments in various studies. Two of the manifest variables in our models, and, refer to providing correct information, which is also mentioned in distinct studies, mostly as a manifest variable for the information quality construct [26,58,69]. In these studies, providing the correct information influences satisfaction or acceptance relatively high or above average. Moreover, there are some studies on the systems in which their primary function is to improve safety aspects. As can be anticipated, they report a high impact from safety on the intention to use [70,71].

    Satisfaction Index

    Each of the satisfaction constructs in the PLS-SEM models is operationalized by 3 measure variables, embodied as 3 questions. The relationship of each of these measure variables with the latent satisfaction construct is characterized by loading and weighting values. The weighting values make it possible to calculate a weighted satisfaction index, both for patients and professionals. Using this index can be a makeshift way of assessing qualities inside health information systems.

    The standardized weightings for the scores of the 3 measure questions of satisfaction are determined by the overall balance in the model as it is implemented by the PLS-SEM algorithm. The weightings, along with the scores of those 3 questions, make it possible to calculate the user satisfaction that is engendered by the improvement of qualities. Without the weightings, the satisfaction score represents the evaluation of a kind of unidentified trait that is relevant but not necessarily the same as a quality engendered satisfaction. A larger and more diverse sample population of respondents and apps might be needed to stabilize the weightings for a larger scope.

    Collecting satisfaction scores is a common practice in health-related studies. However, to the best of our knowledge, the studies on health information systems that suggest a kind of satisfaction index with adjusted weights regarding other qualities are limited. Conversely, the studies that have calculated the parameters of TAM or D&M IS models in their contexts or have used a path model that includes user satisfaction alongside other impacted qualities, such as studies by Jo et al, Schaper et al, and Pai et al [53,56,69], implicitly suggest a kind of indexing for adoption and satisfaction aspects. Nevertheless, there are barriers to utilizing the measures suggested for the satisfaction construct. First, the extent and diversity of subjects in these studies are important factors to reuse their suggested measures and their associated weights. Second, using arbitrary measure questions or the number of items for measurement can create a burden beyond a study’s resources. A sample of this case is the End User Computing Satisfaction measures that range from 12 to 39 items [52,72]. Although some studies have the required resources to apply them [25], using those instruments is not feasible in many other cases. Our focus on the 3 standard satisfaction measures from CSI makes it easier to implement the study and simultaneously facilitates the running of interdisciplinary comparisons and knowledge about the satisfaction of users (customers) of health information systems and other services and products.

    The qualities investigated in this study can explain different amounts of satisfaction variations with R2=.43 for patient satisfaction and R2=.88 for professional satisfaction (as depicted in Figures 3 and 4). The satisfaction index can facilitate an informed guess about the qualities when the user perspectives on those qualities are missing or hard to elicit. This approach is a makeshift way to evaluate the qualities improved by a system. A practical application might be to use the satisfaction index when comparing a pair of similar systems in similar contexts. Another application is to compare the past and present state of the same system that has undergone quality improvement, but no other system or environmental aspect has been changed. Generally, if there are similarities in context and functionalities, and there has been no drastic change or difference in qualities improved by systems, the satisfaction index can serve as a good indicator for an informed guess about those qualities.

    Extensions and Limitations

    The list of most influential qualities should be read with the precaution of how the similar or even the same qualities have been articulated differently in studies [73]. Studies that recruit highly cited frameworks also tend to recruit similar wordings for the qualities. However, other studies practice their freedom to use the wording that best matches their case, resulting in divergent wordings for similar concepts. In our case, notably, the improvement of a treatment’s effectiveness is largely similar to performance or performance expectancy in the studies that are based on the TAM family, and consider the performance expectancy definition as “the degree to which an individual believes that using the system will help him or her to attain gains in job performance” [74]. It is similar to what we asked about effectiveness (refer to Multimedia Appendices 2 and 3). However, one can still find research based on TAM studies that used performance expectancy as a form of efficiency [29]. In addition, efficiency in our study is more or less similar to effort expectancy or perceived usefulness in the TAM series. Some other studies have used productivity or even performance instead of what we called efficiency [68], and others considered efficiency as a manifest to perceived usefulness [8,55,66,67].

    Another caveat when comparing the results is that the qualities specified in our study are the qualities of the treatment or care that using FI-STAR systems might have improved, which are different from the intrinsic qualities of those systems. Accordingly, for example, a higher speed system might increase the affordability of treatment and care in some way. What we have focused on was the affordability of care but not the speed of the system. Therefore, reports on the intrinsic system qualities that have increased user satisfaction or acceptance cannot be compared directly with our study, unless those qualities get translated to their final impact on treatment and care.

    All the eHealth apps, being developed in the FI-STAR project, were supposed to support the required general technical specifications, which includes using the FIWARE infrastructure and being based on software to data paradigm [75]. These requirements were not constraints on the diversity of apps as indicated in Multimedia Appendix 1. Nevertheless, the diversity of the apps should be considered when generalizing the results of this study. For example, if the main outcome of an app is improving the safety of treatment, users might consider both safety and effectiveness almost the same; however, in other apps, they can consider them as distinctive qualities. As another example, the effect of some apps on affordability of treatment can be varied in a different context, when the patients pay for the treatment. This study relies on the output of the UVON method, which extracts common qualities between a set of apps; however, as it is shown in abovementioned examples, user perspectives on those common qualities could be diverse. Therefore, generalization of the results of this study should be done bearing this caveat in mind.

    The predictive power of qualities in projecting user satisfaction can support design decision making for health information systems. When trade-offs are necessary, designers can prioritize features if they can compare their user satisfaction yield. Knowing the quality profile of each feature, one can combine that with the table of quality to satisfaction magnitudes, such as in Table 5, to arrive at more informed feature selection decisions [76].

    Another extension to the model of this study is to consider the qualities or system usage ramifications that impact satisfaction negatively. This needs articulation of questions to capture negative attitudes—not noninformative ones—about the impacts of systems. That kind of wording permits elicitation of constructs that are negatively related to other constructs or new manifest variables for the current constructs that reflect the construct negatively.

    A possible future extension to our PLS-SEM model is to consider relationships between quality latent variables. In the model presented in this study, no relationship has been proposed between the quality constructs, but one might try, for example, to investigate if a system that improves adherence also changes user attitudes about its contribution to effectiveness. However, considering that only recursive relationships can be used in the PLS-SEM models, we cannot investigate the circular impact between qualities and satisfaction with this technique [77,78].


    The satisfaction of health information system users is highly influenced by certain qualities that are improved by those systems. Both patient and professional users consider improvements to the effectiveness of health care a highly important quality that makes them satisfied with the system. For patient users, safety and efficiency qualities come after effectiveness in creating satisfaction. For health professionals, better health care affordability brought by health information systems is important, much like effectiveness, in creating their satisfaction.

    The PLS-SEM model presented in this study can demonstrate the above ranking of qualities in the creation of user satisfaction. Furthermore, the model suggests weightings to calculate the satisfaction index for health information systems. The satisfaction index can be used to compare and monitor health information systems from user satisfaction and quality improvement perspectives.


    The authors would like to acknowledge the contribution of project partners, especially Dr Samuel Fricker from the FI-STAR project, for providing the context of this study. The FI-STAR project was funded by the European Commission under the Seventh Framework Programme under grant agreement number 604691.

    Authors' Contributions

    SE drafted the manuscript and incorporated the contributions from other authors, contributed to the design of the study, developed the proposed models and methods, and analyzed results. JSB contributed to the design of the study, supervised the research process, and critically reviewed the manuscript. TCL contributed to the design of the study and critically reviewed the manuscript. MF facilitated the collection of data and critically reviewed the manuscript. PA contributed to the design of the study, contributed to the method development, supervised the research process, and critically reviewed the manuscript.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Summary of the FI-STAR trial cases.

    PDF File, 14KB

    Multimedia Appendix 2

    Questionnaire for the patients.

    DOCX File, 15KB

    Multimedia Appendix 3

    Questionnaire for the health professionals.

    PDF File (Adobe PDF File), 31KB

    Multimedia Appendix 4

    Crossloadings in the path models.

    PDF File (Adobe PDF File), 61KB

    Multimedia Appendix 5

    Effect size and power analysis.

    PDF File (Adobe PDF File), 38KB

    Multimedia Appendix 6


    PDF File (Adobe PDF File), 32KB


    1. Fitterer R, Rohner P, Mettler T, Winter R. A Taxonomy for Multi-Perspective Ex Ante Evaluation of the Value of Complementary Health Information Systems - Applying the Unified Theory of Acceptance and Use of Technology. 2010 Presented at: 43rd Hawaii International Conference on System Sciences; January 5-8, 2010; Hawaii p. 1-10. [CrossRef]
    2. Li J, Talaei-Khoei A, Seale H, Ray P, Macintyre CR. Health care provider adoption of ehealth: systematic literature review. Interact J Med Res 2013;2(1):e7 [FREE Full text] [CrossRef] [Medline]
    3. Fritz F, Tilahun B, Dugas M. Success criteria for electronic medical record implementations in low-resource settings: a systematic review. J Am Med Inform Assoc 2015 Mar;22(2):479-488. [CrossRef] [Medline]
    4. Ammenwerth E, Gräber S, Herrmann G, Bürkle T, König J. Evaluation of health information systems-problems and challenges. Int J Med Inform 2003 Sep;71(2-3):125-135. [Medline]
    5. Mair F, Whitten P. Systematic review of studies of patient satisfaction with telemedicine. Br Med J 2000 Jun 3;320(7248):1517-1520 [FREE Full text] [Medline]
    6. DeLone WH, McLean ER. Information systems success: the quest for the dependent variable. Inf Syst Res 1992 Mar;3(1):60-95. [CrossRef]
    7. Heeks R. Health information systems: failure, success and improvisation. Int J Med Inform 2006 Feb;75(2):125-137. [CrossRef] [Medline]
    8. Holden RJ, Karsh BT. The technology acceptance model: its past and its future in health care. J Biomed Inform 2010 Feb;43(1):159-172 [FREE Full text] [CrossRef] [Medline]
    9. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS Q 2003;27(3):425-478. [CrossRef]
    10. Cronin JJ, Brady MK, Hult GM. Assessing the effects of quality, value, and customer satisfaction on consumer behavioral intentions in service environments. J Retailing 2000 Jun;76(2):193-218. [CrossRef]
    11. Fornell C. A national customer satisfaction barometer: the Swedish experience. J Mark 1992 Jan;56(1):6-21. [CrossRef]
    12. Fornell C, Johnson MD, Anderson EW, Cha J, Bryant BE. The American Customer Satisfaction Index: nature, purpose, and findings. J Mark 1996 Oct;60(4):7-18. [CrossRef]
    13. Eklöf JA, Westlund AH. The Pan-European Customer Satisfaction Index Programme—current work and the way ahead. Total Qual Manag 2006 Jul 28;13(8):1099-1106. [CrossRef]
    14. Urbach N, Smolnik S, Riempp G. The state of research on information systems success. Bus Inf Syst Eng 2009 Jul 18;1(4):315-325. [CrossRef]
    15. Aragon SJ, Gesell SB. A patient satisfaction theory and its robustness across gender in emergency departments: a multigroup structural equation modeling investigation. Am J Med Qual 2003;18(6):229-241. [CrossRef] [Medline]
    16. Karimi F, Poo DC, Tan YM. Clinical information systems end user satisfaction: the expectations and needs congruencies effects. J Biomed Inform 2015 Feb;53:342-354 [FREE Full text] [CrossRef] [Medline]
    17. Hadji B, Degoulet P. Information system end-user satisfaction and continuance intention: a unified modeling approach. J Biomed Inform 2016 Jun;61:185-193 [FREE Full text] [CrossRef] [Medline]
    18. Bayol MP, de la Foye A, Tellier C, Tenenhaus M. Use of PLS path modeling to estimate the European Consumer Satisfaction Index (ECSI) model. Statistica Applicata 2000;12(3):361-375 [FREE Full text]
    19. Jackson JL, Chamberlin J, Kroenke K. Predictors of patient satisfaction. Soc Sci Med 2001 Feb;52(4):609-620. [Medline]
    20. DeLone W, McLean E. The DeLone and McLean Model of Information Systems Success: a ten-year update. J Manag Inf Syst 2014 Dec 23;19(4):9-30. [CrossRef]
    21. Urbach N, Müller B. The Updated DeLone and McLean Model of Information Systems Success. In: Information Systems Theory: Explaining and Predicting Our Digital Society. New York: Springer; 2012:1-18.
    22. Petter S, DeLone W, McLean E. Measuring information systems success: models, dimensions, measures, and interrelationships. Eur J Inf Syst 2008 Jun;17(3):236-263. [CrossRef]
    23. Yu P, Li H, Gagnon M. Health IT acceptance factors in long-term care facilities: a cross-sectional survey. Int J Med Inform 2009 Apr;78(4):219-229. [CrossRef] [Medline]
    24. Yusof MM, Paul RJ, Stergioulas LK. Towards a Framework for Health Information Systems Evaluation. 2006 Jan Presented at: 39th Annual Hawaii International Conference on System Sciences; January 4-7, 2006; Hawaii p. 95a-95a. [CrossRef]
    25. Aggelidis VP, Chatzoglou PD. Hospital information systems: measuring end user computing satisfaction (EUCS). J Biomed Inform 2012 Jun;45(3):566-579 [FREE Full text] [CrossRef] [Medline]
    26. Kimiafar K, Sadoughi F, Sheikhtaheri A, Sarbaz M. Prioritizing factors influencing nurses' satisfaction with hospital information systems: a fuzzy analytic hierarchy process approach. Comput Inform Nurs 2014 Apr;32(4):174-181. [CrossRef] [Medline]
    27. Ives B, Olson MH, Baroudi JJ. The measurement of user information satisfaction. Commun ACM 1983 Oct 10;26(10):785-793. [CrossRef]
    28. Venkatesh V. Determinants of perceived ease of use: integrating control, intrinsic motivation, and emotion into the technology acceptance model. Inf Syst Res 2000 Dec 1;11(4):342-365. [CrossRef]
    29. Phichitchaisopa N, Naenna T. Factors affecting the adoption of healthcare information technology. EXCLI J 2013;12:413-436 [FREE Full text] [Medline]
    30. Ammenwerth E, Iller C, Mahler C. IT-adoption and the interaction of task, technology and individuals: a fit framework and a case study. BMC Med Inform Decis Mak 2006;6:3 [FREE Full text] [CrossRef] [Medline]
    31. Francis J, Johnston M, Eccles M, Walker A, Grimshaw JM, Foy R, et al. Constructing questionnaires based on the theory of planned behaviour: A manual for Health Services Researchers. In: Quality of Life and Management of Living Resources. Newcastle: Centre for Health Services Research; 2004.
    32. Eslami Andargoli VP, Scheepers H, Rajendran D, Sohal A. Health information systems evaluation frameworks: a systematic review. Int J Med Inform 2017 Dec;97:195-209. [CrossRef] [Medline]
    33. Eivazzadeh S, Anderberg P, Larsson TC, Fricker SA, Berglund J. Evaluating health information systems using ontologies. JMIR Med Inform 2016 Jun 16;4(2):e20 [FREE Full text] [CrossRef] [Medline]
    34. European Commision: CORDIS; 2016. Future Internet Social and Technological Alignment Research   URL: [accessed 2018-06-05] [WebCite Cache]
    35. Kidholm K, Ekeland AG, Jensen LK, Rasmussen J, Pedersen CD, Bowes A, et al. A model for assessment of telemedicine applications: MAST. Int J Technol Assess Health Care 2012 Jan;28(1):44-51. [CrossRef] [Medline]
    36. R-project. Vienna, Austria: R Foundation for Statistical Computing; 2017. R: A Language and Environment for Statistical Computing   URL: [accessed 2018-10-17] [WebCite Cache]
    37. Ringle CM, Wende S, Becker JM. SmartPLS.: SmartPLS GmbH; 2015. SmartPLS 3   URL: [accessed 2018-10-17] [WebCite Cache]
    38. Rönkkö M. Github. 2017. Matrixpls: Matrix-Based Partial Least Squares Estimation   URL: [accessed 2018-10-17] [WebCite Cache]
    39. Faul F, Erdfelder E, Buchner A, Lang AG. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav Res Methods 2009 Nov;41(4):1149-1160. [CrossRef] [Medline]
    40. Kock N. One-Tailed or Two-Tailed P Values in PLS-SEM? Int J e-Collab 2015 Apr;11(2):1-7. [CrossRef]
    41. Friendly M. Corrgrams. Exploratory displays for correlation matrices. Am Stat 2002 Nov;56(4):316-324. [CrossRef]
    42. Henseler J, Ringle CM, Sarstedt M. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J Acad Mark Sci 2014 Aug 22;43(1):115-135. [CrossRef]
    43. Henseler J, Ringle CM, Sarstedt M. Using partial least squares path modeling in advertising research: basic concepts and recent issues. In: Okazaki S, editor. Handbook of Research on International Advertising. Massachusetts, USA: Edward Elgar Publishing; 2012:252-276.
    44. Chin WW. The partial least squares approach to structural equation modeling. In: Marcoulides GA, editor. Methodology for Business and Management. Modern Methods for Business Research. Mahwah, New Jersey: Lawrence Erlbaum Associates Publishers; 1998:295-336.
    45. Gilpin AR. Table for conversion of Kendall's Tau to Spearman's Rho within the context of measures of magnitude of effect for meta-analysis. Educ Psychol Meas 2016 Sep 7;53(1):87-92. [CrossRef]
    46. Farrar DE, Glauber RR. Multicollinearity in regression analysis: the problem revisited. Rev Econ Stat 1967 Feb;49(1):92-107. [CrossRef]
    47. Henseler J, Hubona G, Ray PA. Using PLS path modeling in new technology research: updated guidelines. Ind Manage Data Syst 2016 Feb;116(1):2-20. [CrossRef]
    48. Hsu SH. Developing an index for online customer satisfaction: adaptation of American Customer Satisfaction Index. Expert Syst Appl 2008 May;34(4):3033-3042. [CrossRef]
    49. Iivari J. An empirical test of the DeLone-McLean Model of Information System Success. ACM SIGMIS Database 2005 Jun 7;36(2):8-27. [CrossRef]
    50. Hair Jr JF, Sarstedt M, Hopkins L, Kuppelwieser V. Partial least squares structural equation modeling (PLS-SEM): an emerging tool in business research. Eur Bus Rev 2014 Mar 4;26(2):106-121. [CrossRef]
    51. Seddon P, Kiew MY. A partial test and development of Delone and Mclean's model of IS success. Australas J Inf Syst 1996 Nov 1;4(1). [CrossRef]
    52. Bailey JE, Pearson SW. Development of a tool for measuring and analyzing computer user satisfaction. Manag Sci 1983 May;29(5):530-545. [CrossRef]
    53. Jo HS, Song TM, Kim BG. Analysis of the factors affecting consumer acceptance of accredited online health information. J Korean Med Sci 2017 Nov;32(11):1757-1763 [FREE Full text] [CrossRef] [Medline]
    54. Liu L, Miguel Cruz A, Rios Rincon A, Buttar V, Ranson Q, Goertzen D. What factors determine therapists' acceptance of new technologies for rehabilitation – a study using the Unified Theory of Acceptance and Use of Technology (UTAUT). Disabil Rehabil 2015;37(5):447-455. [CrossRef] [Medline]
    55. Hu PJ, Chau PY, Sheng ORL, Tam KY. Examining the technology acceptance model using physician acceptance of telemedicine technology. J Manag Inf Syst 2015 Dec 2;16(2):91-112. [CrossRef]
    56. Schaper LK, Pervan GP. ICT and OTs: a model of information and communication technology acceptance and utilisation by occupational therapists. Int J Med Inform 2007 Jun;76 Suppl 1:S212-S221. [CrossRef] [Medline]
    57. Wu JH, Wang SC, Lin LM. Mobile computing acceptance factors in the healthcare industry: a structural equation model. Int J Med Inform 2007 Jan;76(1):66-77. [CrossRef] [Medline]
    58. Kim J, Chae YM, Kim S, Ho SH, Kim HH, Park CB. A study on user satisfaction regarding the clinical decision support system (CDSS) for medication. Healthc Inform Res 2012 Mar;18(1):35-43 [FREE Full text] [CrossRef] [Medline]
    59. Akter S, D'Ambra J, Ray P. Trustworthiness in mHealth information services: an assessment of a hierarchical model with mediating and moderating effects using partial least squares (PLS). J Am Soc Inf Sci 2010 Oct 25;62(1):100-116. [CrossRef]
    60. Handayani PW, Hidayanto AN, Budi I. User acceptance factors of hospital information systems and related technologies: systematic review. Inform Health Soc Care 2017 Aug 22:1-26. [CrossRef] [Medline]
    61. Hsiao JL, Wu WC, Chen RF. Factors of accepting pain management decision support systems by nurse anesthetists. BMC Med Inform Decis Mak 2013 Jan 29;13:16 [FREE Full text] [CrossRef] [Medline]
    62. Steininger K, Stiglbauer B, Baumgartner B, Engleder B. Factors Explaining Physicians? Acceptance of Electronic Health Records. 2014 Jan Presented at: 47th Hawaii International Conference on System Sciences; 2014; Hawaii p. 2768-2777. [CrossRef]
    63. Berner ES, Detmer DE, Simborg D. Will the wave finally break? A brief view of the adoption of electronic medical records in the United States. J Am Med Inform Assoc 2005;12(1):3-7 [FREE Full text] [CrossRef] [Medline]
    64. Boonstra A, Broekhuis M. Barriers to the acceptance of electronic medical records by physicians from systematic review to taxonomy and interventions. BMC Health Serv Res 2010;10:231 [FREE Full text] [CrossRef] [Medline]
    65. Schweitzer J, Synowiec C. The economics of eHealth and mHealth. J Health Commun 2012;17 Suppl 1:73-81 [FREE Full text] [CrossRef] [Medline]
    66. Tung FC, Chang SC, Chou CM. An extension of trust and TAM model with IDT in the adoption of the electronic logistics information system in HIS in the medical industry. Int J Med Inform 2008 May;77(5):324-335. [CrossRef] [Medline]
    67. Zailani S, Gilani MS, Nikbin D, Iranmanesh M. Determinants of telemedicine acceptance in selected public hospitals in Malaysia: clinical perspective. J Med Syst 2014 Sep;38(9):111. [CrossRef] [Medline]
    68. Mohd Salleh MI, Zakaria N, Abdullah R. The influence of system quality characteristics on health care providers' performance: empirical evidence from Malaysia. J Infect Public Health 2016;9(6):698-707 [FREE Full text] [CrossRef] [Medline]
    69. Pai FY, Huang KI. Applying the Technology Acceptance model to the introduction of healthcare information systems. Technol Forecast Soc Change 2011 May;78(4):650-660. [CrossRef]
    70. Wong AM, Chang WH, Ke PC, Huang CK, Tsai TH, Chang HT, et al. Technology acceptance for an Intelligent Comprehensive Interactive Care (ICIC) system for care of the elderly: a survey-questionnaire study. PLoS One 2012;7(8):e40591 [FREE Full text] [CrossRef] [Medline]
    71. Peeters JM, de Veer AJ, van der Hoek L, Francke AL. Factors influencing the adoption of home telecare by elderly or chronically ill people: a national survey. J Clin Nurs 2012 Nov;21(21-22):3183-3193. [CrossRef] [Medline]
    72. Doll WJ, Deng X, Raghunathan T, Torkzadeh G, Xia W. The meaning and measurement of user satisfaction: a multigroup invariance analysis of the end-user computing satisfaction instrument. J Manage Inf Syst 2014 Dec 8;21(1):227-262. [CrossRef]
    73. DeFranco J, Kassab M, Laplante P, Laplante N. The nonfunctional requirement focus in medical device software: a systematic mapping study and taxonomy. Innovations Syst Softw Eng 2017 Aug 12;13(2-3):81-100. [CrossRef]
    74. Davis FD. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Q 1989 Sep;13(3):319. [CrossRef]
    75. 2014. About FI-STAR   URL: [accessed 2015-09-29] [WebCite Cache]
    76. Eivazzadeh S, Anderberg P, Berglund J, Larsson T. Designing with Priorities and Thresholds for Health Care Heterogeneity: The Approach of Constructing Parametric Ontology. In: Proceedings of the 20th International Conference on Engineering Design (iced 15) Vol 2: Design Theory and Research Methodology Design Processes. Glasgow: Design Soc; 2015 Presented at: 20th International Conference on Engineering Design; 2015; Milan.
    77. Monecke A, Leisch F. semPLS: structural equation modeling using partial least squares. J Stat Softw 2012;48(3). [CrossRef]
    78. Fricker S, Thuemmler C, Paulin A. 2015. General, social and ethical-legal requirements   URL:,%20social%20and%20ethical-legal%20requireme [accessed 2017-02-03] [WebCite Cache]


    AVE: average variance extracted
    CR: composite reliability
    CSI: Customer Satisfaction Index
    D&M IS: Delone and McLean Information Systems Success
    ECSI: European Customer Satisfaction Index
    eHealth: electronic health
    EU: European Union
    FI-STAR: Future Internet Social and Technological Alignment Research
    HTMT: Heterotrait-Monotrait
    PLS-SEM: partial least squares structural equation modeling
    TAM: technology acceptance model
    UVON: Unified eValuation using ONtology

    Edited by G Eysenbach; submitted 08.06.18; peer-reviewed by A Sheikhtaheri, S Koch-Weser; comments to author 12.07.18; revised version received 05.09.18; accepted 08.09.18; published 30.11.18

    ©Shahryar Eivazzadeh, Johan S Berglund, Tobias C Larsson, Markus Fiedler, Peter Anderberg. Originally published in JMIR Medical Informatics (, 30.11.2018.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Informatics, is properly cited. The complete bibliographic information, a link to the original publication on, as well as this copyright and license information must be included.