1 Introduction

The consequences of Canada’s aging population are numerous, not the least of which is the aging of Canada’s workforce. Though the amount of workers aged 55 and older decreased steadily from 1976 to 1997, the trend has reversed sharply with their employment rate climbing from 22% in 1997 to 41.9% in 2017 [1, 2]. The reasons for this are varied, ranging from an increase in life expectancy and the amount of good-health years later in life, to the economic pressures of rising consumer and housing debt, as well as higher costs of living.

As companies seek to upgrade work processes with new technologies, the demographic composition of their staff will become an increasingly important issue. The difficulties of large-scale organizational change are well documented and the failure rates for major IS systems implementation averages around 70% [3]. The successful introduction of a new technology within an organization relies heavily on its acceptance among the workers and users themselves [4,5,6], and with such a significant amount of older workers in the workplace, special attention must be paid to how new technologies are best introduced to this segment of the workforce.

The use of new technologies is increasing rapidly in many sectors, though there has been a notable lag within Canada’s healthcare industry, particularly in the area of electronic health record adoption. Currently, 78% of general practitioners use electronic health records and this rate drops to a disappointing 21% when considering only EHR systems with basic meaningful use functionalities, such as allowing for the exchange of health information with other providers, generating patient information, and providing routine decision support [7]. Clearly, there is still much work to be done in order to implement EHRs in facilities that do not yet use it and upgrade existing systems in those organizations that do not have the system functionalities to fully realize the benefits EHRs can provide. Both of these tasks required well-planned implementation techniques, tailored to the needs of the daily users, to be successful.

2 Literature Review

2.1 Measuring Implementation Success

Research has sought to better understand the factors driving technology implementation success through analyzing user satisfaction, professional identity, and user acceptance (see Table 1). User satisfaction is seen as an important aspect of EHR implementation and according to this theory, satisfied users are more likely to accept and use the technology as intended. When reviewing studies on user satisfaction, the positive or negative effects on satisfaction were most often tied to system functionalities. Lee et al. [8] designed a survey to assess user satisfaction of a newly implemented physician order entry system among physicians and nurses at a women’s hospital. Efficiency was found to be the most influential factor, taking precedence over improvements in patient care [8]. Although, physicians reported greater satisfaction than nurses, revealing that there are significant differences between the two groups of users that must be addressed [8]. Lee et al. [8] also acknowledged the fact that the manner in which the technology was introduced also affected satisfaction. Clinical leadership, staff involvement, and customization to user needs were found to have a positive impact on user satisfaction [8].

Table 1. Healthcare technology implementation research from the individual perspective
Full size table

Goorman and Berg [9] used observation and in-depth interviews to discover that the compatibility of a new health information system with current work processes was also an important aspect of user satisfaction. If the functionality of the system did not match with current process or if workflows were not adequately modified to suit the new system during implementation, satisfaction was very difficult to achieve [9]. Lee et al. [10] built on the theory of functionality being closely linked with user satisfaction by studying a group of intensive care nurses in Taiwan. Through staff interviews, Lee et al. [10] confirmed that efficiencies in information entry and paper consumption lead to user satisfaction, while slow information retrieval and lack of workload alleviation were found to lead to user dissatisfaction [10]. Likourezos et al. [11] also identified efficiency as a main driver in satisfaction. The improved entering and accessing of data EHR systems facilitated proved to be significant in the emergency setting and so resulted in high levels of user satisfaction [11].

While system functionality is undeniably important in the acceptance of health care technology, the above user satisfaction research overlooks the effect social constructions, such as professional identity, can have on acceptance. Prasad [12] explored the relationship between new technology in the healthcare sector and professional identity through the lens of symbolic interaction. According to Prasad, technology could take on meanings and become symbols of concepts and perspectives through the social interaction of those who used it [12]. The computerization of work within an American HMO studied by Prasad took on the meaning of professionalism for both physicians and nurses [12]. These healthcare providers felt their professionalism was validated through their work with computers and felt pride as well as a sense of increased organizational status as a result [12]. Prasad demonstrated that when new technologies became a symbol for social constructions that were important to the user, such as professionalism and status, acceptance became easier to achieve. Apesoa-Varano also explored the idea of professional identity among nurses, noting that historically, upgrades in skills and technical knowledge were often sought by the nursing profession in order to legitimize their status within the health care sector [13]. Since education and technological skill were seen as symbols of professionalism, nurses used them as tools to bring legitimacy to their occupation and increase their status within the organization. Quantitative measurement of exactly how much this process impacted nurse acceptance of new technologies, however, remained to be studied.

2.2 Targeting Further Understanding

Past research surrounding electronic health record implementation is abundant though incomplete. Earlier studies regarding the introduction of EHRs focused first on its benefits and how to best increase adoption rates [18,19,20,21,22,23,24]. Research addressing the processes and factors affecting implementation itself, while present, has had a narrow scope with little focus on the effect user age would have on acceptance. Furthermore, the majority of research on this subject addresses primarily physician acceptance and takes place within acute care settings or private clinics [9,10,11, 14, 25,26,27,28]. With the large role senior health care will play in the near future, studies that can establish EHR implementation best practices for long-term care facilities are needed.

Furthermore, the important user population of nurses has yet to be sufficiently addressed. Nurses are vital in all healthcare settings, often outnumbering other healthcare professionals. This is especially true in the field of long-term care, where patients and residents interact most with nursing staff. It is also important to note that 42% of all regulated nurses in Canada are over the age of 50, with 28% being 55 years of age or older [29]. In order to effectively introduce EHR systems, the factors that drive electronic health record acceptance among older nurses must be identified.

This research seeks to identify the factors necessary for the successful implementation of electronic health records among older nurses in health care facilities. Though electronic health record systems were introduced more than a decade ago, many facilities still face considerable challenges implementing them. Evidence of these challenges can be seen in the slow and incomplete adoption of EHR systems in North America and Canada especially. As the nursing workforce ages, these challenges become more varied, differing across age groups. By identifying predictive factors in EHR adoption and analyzing any changes in their significance among these groups, this research can help health care organizations tailor their implementation efforts effectively to their older nursing staff.

3 Method

This research used a hypothetico-deductive approach to investigate the user acceptance of an EHR system implemented in a long-term care facility. This study employed a modified version of the Technology Acceptance Model (TAM) questionnaire survey to collect empirical observations and the partial least squares method for its analysis.

3.1 Research Model

TAM has proven to be an effective model for examining technology acceptance within the workplace and the extended TAM2 model provided the framework necessary to better understand the more specific influences on this acceptance. Therefore, based on the goal of this study, a slightly modified TAM2 was chosen as the research model and can be seen in the Fig. 1 below. Within this framework, the variable of perceived usefulness is the measure of user acceptance. The determinants to be tested included experience, subjective norm, image, job relevance, output quality, result demonstrability, perceived usefulness, and perceived ease of use. For the purpose of this study, three factors present within the original TAM2 framework were excluded in the final research model: voluntariness, intention to use, and usage behaviour. Voluntariness was excluded, as the EHR system to be evaluated was not voluntary. For similar reasons, intention to use was also removed as all nurses must use and intend to use the EHR system in the course of their patient care. As voluntariness and intention to use were inputs required to measure usage behaviour, this factor was excluded as well.

Fig. 1.
figure 1

Research model

Full size image

3.2 Hypotheses

Based on the research model above, the following hypotheses were formulated: Subjective norm is a factor that was originally present in the Theory of Reasoned Action (TRA) and is defined as a “person’s perception that most people who are important to him think he should or should not perform the behaviour in question” [30]. Within the workplace setting, this subjective norm can be viewed as the opinions of both management/superiors and peers towards the technology. The effect of subjective norm on technology acceptance has varied; with Mathieson [31] finding that subjective norm had no significant effect on usage, while others such as Taylor and Todd [32], Venkatesh and Davis [33], and Wu et al. [34] found that subjective norm did have significant effects. Though excluded from the original TAM model, Davis [35] stated that subjective norm could be more influential in settings where a system’s use is mandatory. Users would feel compelled to use a system as result of managerial expectations as opposed to individual beliefs towards a system [35]. Electronic health records are often an organization-wide update in documentation processes and as a result subjective norm becomes an important variable to examine.

  • H1. Subjective norm will be positively associated with perceived usefulness.

Experience is defined as user history and interaction with the technology. Fuerst and Cheney [36] identified computer experience as a variable in the usage of information systems. Hartwick and Barki [37] continued with this concept stating that testing the effects of experience on IS use is an important subject of future study. Hartwick and Barki [37] theorized that the less experience a user had with a system, the more that user would rely on the opinions of others. Venkatesh and Davis [33] built on this view, concluding that experience would have a significant effect on the influence of subjective norm. The more experience one had with a system, the less subjective norm could influence one’s perceptions [33] and the less experience one had, the more important subjective norm would become.

  • H2. Experience will moderate the influence of subjective norm on perceived usefulness.

While studying the factors affecting the diffusion of innovations, Rogers [38] stated that, “undoubtedly one of the most important motivations for almost any individual to adopt an innovation is the desire to gain social status.” This motivation can readily be applied to the acceptance of new technology within the workplace. In the context of this research, image is defined as, “the degree to which use of an innovation is perceived to enhance one’s … status in one’s social system” [39]. The importance of image in a healthcare setting was presented by Prasad [12] in her observations of HMO employees. New technology knowledge came to represent an increase in professional status, leading workers to more readily accept and learn new technologies in order to gain status within the organization. Therefore, the image of an EHR system should have a significant effect on how useful it is perceived to be. Additionally, the social nature of image suggests there is a close relationship with subjective norm and so the effect of subject norm on image is also the subject of further investigation.

  • H3. Subjective norm will be positively associated with image.

  • H4. Image will be positively associated with perceived usefulness.

The next three variables of job relevance, output quality, and result demonstrability focus on the instrumental rather than social characteristics of a new system. Bailey and Pearson [40] established that relevancy was one of the most important factors in determining user satisfaction. The more relevant the system to the task, the more positive the reaction to its use [40]. Even the most impressive functionalities are at best meaningless and at worst burdensome if it is not appropriate for a given task. Therefore, the more relevant a system is to the job, the more likely it will be viewed as useful and ultimately accepted by the user. The effects of output quality on technology acceptance in the workplace have been the subject of several research studies [4, 41,42,43]. Davis [4] examined output quality’s effect on perceived usefulness and found it to be significant, though moderated by its task importance. Therefore, the higher the quality of the system output, the more useful the system is perceived to be. Lastly, result demonstrability is defined as, “tangibility of the results of using the innovation” [39]. New systems are most often introduced as a means to improve job/task performance. Should these improvements not present themselves in the course of the system’s use, the perception of usefulness will most likely decrease. Agarwal and Prasad [44] and Mun et al. [45] both found evidence that result demonstrability was a significant determinant of perceived ease of use, concluding that a system’s benefits must not just be talked about but readily identifiable to the user.

  • H5. Job relevance will be positively associated with perceived usefulness.

  • H6. Output quality will be positively associated with perceived usefulness.

  • H7. Result demonstrability will be positively associated with perceived usefulness.

Evidence of the effect of perceived ease of use on perceived usefulness has been well documented in a variety of studies [4, 6, 35, 46,47,48,49]. Not surprisingly, the easier a system is to use, the more useful it is perceived to be.

  • H8. Perceived ease of use will be positively associated with perceived usefulness.

3.3 Subject Selection

As the target population of this research is long-term care nursing staff, sample subjects chosen for this study were employees of a Canadian long-term care facility. At the time of study, the electronic health record system implementation had occurred three years prior. To test the proposed model hypotheses outlined above, data was collected through a program-wide questionnaire. Nursing staff members present during the implementation of the electronic health care system were eligible to participate. As the goal of this research is to highlight the needs of older nurses, nurses of all ages were invited to complete the survey in order to facilitate age comparisons.

3.4 Questionnaire Design

The survey created was composed of two distinct sections. The first consisted of questions relating directly to the constructs outlined in the research model. The validated Venkatesh and Davis [33] TAM2 survey questions were used as the basis of this questionnaire. The responses were measured on a 7-point Likert scale. The second section of the questionnaire reflected important demographic information relating to job status, job title, age, and experience that could then be used to compare survey results of various sub-groups within the sample. A copy of the questionnaire as well as the theoretical mapping of the questions can be found in the appendix.

3.5 Data Collection and Analysis

In order to collect the data necessary through questionnaires, voluntary sampling was used. The questionnaire was made available to all nursing staff both electronically through email and as hard copies from June 1st to July 3rd, 2014. A total of 126 completed questionnaires were submitted, resulting in a 28% response rate. The following chart details the descriptive statistics of the sample (Table 2).

Table 2. Descriptive statistics of sample
Full size table

Analysis of the sample questionnaire data was done in four steps outlined below.

  • Step 1: Preliminary Review & Confirmatory Factor Analysis

After preliminary evaluation of the data collected, the overview of the sample and construct statistics were the following (Table 3):

Table 3. Sample mean and t-value statistics. *p < 0.01
Full size table

After initial review of the sample statistics, formative construct validity could be assessed. Constructs result demonstrability as well as experience presented high standard error levels and low t-statistic values. The t-values indicate the significance of the difference between means, with statistical significance increasing the higher the t-value becomes. Those t-values over 2.58 are considered significant with a 99% confidence level. The low t-values of result demonstrability and experience indicated that these items were not measuring as significant, especially when compared to the other constructs of perceived usefulness, perceived ease of use, subjective norm, image, job relevance, and output quality.

To further examine the sample data, confirmatory factor analysis was conducted. The resulting cross-loadings of the factors can be seen below (Table 4). When assessing these factors, values above 0.6 are considered high and those below 0.4 considered low [50]. The high loadings of the factors within their own construct suggested the measurements fit the theoretical model well.

Table 4. Construct cross-loadings
Full size table

After this initial analysis, the partial least squares (PLS) method was used to further analyze the data collected.

  • Step 2: Reliability Analysis

In this step, the reliability of the proposed construct measurements was determined. All the constructs within this study were tested for its Cronbach alpha value. Cronbach’s alpha is continually used in statistical research as an index of reliability, with higher values being viewed as more reliable [51]. Acceptable values for the alpha threshold can vary from 0.7 to 0.95 [51]. For the purposes of this study, the constructs that resulted in values greater than 0.7 were considered a reliable measurement. Those constructs that resulted in Cronbach α values of less than 0.7 were omitted from further consideration. As can be seen below, experience failed to meet the criteria of reliability and was therefore excluded from further analysis (Table 5).

Table 5. Latent variable reliability
Full size table

  • Step 3: Construct Validation

This step was used to validate the constructs of this study using convergent and discriminant validity. To establish convergent validity both composite reliability and average variance extracted (AVE) were analyzed. Those constructs with an AVE value greater than 0.5 and with a composite reliability reading greater than 0.7 demonstrated acceptable convergent validity and were included in further analysis. As result demonstrability did not meet the convergent validity threshold, it was omitted from further consideration. The results of this process can be seen below (Table 6).

Table 6. Latent variable parameters
Full size table

Discriminant validity was determined by evaluating whether or not an item loaded most heavily on its own construct. The items that met this condition were used in further analysis. The items that did not were omitted from further consideration. The results of this analysis are shown in Table 7 below. As all constructs that remained loaded most heavily within its own category, the validity of all remaining latent variables was confirmed.

Table 7. Latent variable correlations
Full size table

4 Results

4.1 Sample Group Results

The PLS analysis of the five remaining latent variables (perceived ease of use, subjective norm, image, job relevance, output quality) resulted in an R2 of 0.487. This indicated that 49% of the variance present in perceived usefulness can be explained by the five predictive variables within the model. The results of this analysis can be seen in the Fig. 2 below.

Fig. 2.
figure 2

Results of PLS model

Full size image

Though nearly half of the variance in perceived usefulness can be explained by these variables, some were significant in their influence while others were not. The results of this analysis revealed that for all employees sampled only Hypotheses 3, 4, and 8 were adequately supported. Subjective norm was highly significant in determining image at a 99% confidence level. A positive subjective norm had a positive influence on image, supporting H3. Image itself was also highly significant in determining perceived usefulness at a confidence level of 99%. A positive image was shown to have a positive effect on perceived usefulness, supporting H4. Finally, perceived ease of use also proved to be a predictor of perceived usefulness at a 95% confidence level. Positive perceived ease of use had a positive effect on perceived usefulness, supporting H8. However, there was no evidence of job relevance (H5) or output quality (H6) having a significant effect of perceived usefulness. H5 and H6 did not reach an acceptable p value of less than 0.05, and therefore could not be classified as having any significant effect, positive or negative, on perceived usefulness.

4.2 Multi-group Comparison Results

Though certain conclusions may be drawn from the analysis of the entire sample as a whole, the goal of this study is to determine the success factors for mature, experienced workers. As a result, emphasis was placed on comparisons between older and younger nurses as well as those with less experience compared to those with more. In order to do these comparisons, the Stats Tool Package created by James Gaskin was used [52]. The path coefficients of the various groups were calculated by entering individual group data into SmartPLS and creating a bootstrap report. The sample mean and standard error were then entered into the following formula to determine the t-statistic:

$$ t = \frac{{Path_{sample\_1} - Path_{sample\_2} }}{{\left[ {\sqrt {\frac{{\left( {m - 1} \right)^{2} }}{{\left( {m + n - 2} \right)}} *S.E._{sample1}^{2} + \frac{{\left( {n - 1} \right)^{2} }}{{\left( {m + n - 2} \right)}} *S.E._{sample2}^{2} } } \right]*\left[ {\sqrt {\frac{1}{m} + \frac{1}{n}} } \right]}} $$

That t-statistic was then converted into its 2-tailed p-value using the Excel Stat Tools Package. If the resulting p-value was less than 0.05, the difference between the groups compared was considered statistically significant considered statistically significant.

Table 8. Path coefficient comparison 1
Full size table

The first comparison was that of the age ranges within the sample group (Table 8). The three sub groups within this comparison were ages 25 to 34 years old, 35 to 54 years old, and 55 years or older. These ages have been grouped to closely align with generational divisions. Those aged 25 to 34 fall within the Millennial generation, those 35 to 54 within Generation X, and those over 55 within the Baby Boomer generation [53]. There was no evidence of significant path coefficient differences between 25 to 34 year olds and 35 to 54 year olds. However, there was evidence of significant differences between nurses 25 to 34 years old and those 55 and older regarding the effect of perceived ease of use (H8) and image (H4) on perceived usefulness. Perceived ease of use was found to have significantly more influence with nurses 25 to 34 years old compared to nurses 55 years old and over, while image had a significantly stronger effect on nurses 55 and older compared to those aged 25 to 34. There was also evidence of significant path coefficient differences regarding image between 35 to 54 year olds and those over 55. Once again, image had a significantly stronger effect on nurses 55 and older compared to those age 35 to 54 years old.

Table 9. Path coefficient comparison 2, part 1
Full size table
Table 10. Path coefficient comparison 2, part 2
Full size table

The second sub group comparison conducted measured significant differences among workers with varying years of nursing experience. As nurses within the sample entered the profession at varying ages, comparing sub groups based on experience could lead to differences not present among age groups. The results of this comparison can be found above (Tables 9 and 10). There was no evidence of any significant path coefficient differences between those with 3 to 5 years experience and those with 5 to 10 years experience. However, there was evidence of significant differences between nurses with 3 to 5 years experience and those with 10 to 20 years experience concerning Hypothesis 3, 4, 5, and 8. The influence of image (H4) and perceived ease of use (H8) on perceived usefulness were significantly stronger in those nurses with 3 to 5 years experience compared to those with 10 to 20 years. The influence of job relevance (H5) and output quality (H6) on perceived usefulness, however, was significantly lower in those nurses with 3 to 5 years experience compared to those with 10 to 20 years.

There was also evidence of many significant path coefficient differences between those with 10 to 20 years experience compared to those with over 20 years experience. The path coefficients from subjective norm to image (H3), image to perceived usefulness (H4), and perceived ease of use to perceived usefulness (H8) were all significantly more influential for those with more than 20 years of experience compared to those with 10 to 20 years. The path coefficients from job relevance (H5) and output quality (H5) to perceived usefulness were significantly more influential for those with 10 to 20 years experience compared to those with over 20 years experience.

When comparing the 3 to 5 years experience sub group to the over 20 years experience sub group, there was evidence of two statistically significant path coefficient differences. The positive influence of subjective norm on image (H3) was significantly higher with those having over 20 years experience compared with nurses who had 3 to 5 years experience, while the positive effect of output quality (H6) was significantly stronger for those with 3 to 5 years experience compared to nurses with over 20.

There was evidence of only two significant differences between nurses with 5 to 10 years experience and nurses with 10 to 20 years. The path of output quality to perceived usefulness (H6) was significantly more influential for nurses with 10 to 20 years experience compared to those with 5 to 10. However, the positive effect of perceived ease of use on perceived usefulness (H8) was significantly stronger for nurses with 5 to 10 years experience compared to 10 to 20 years.

Nurses with 5 to 10 years experience compared to those with over 20 showed evidence of significant differences regarding the path coefficients of subjective norm to image (H3). The effect of subject norm on image was significantly stronger among nurses with over 20 years experience compared to those with only 5 to 10 years experience.

The differences between the various age groups and levels of experience revealed that not only is it important to examine nurses separately from the larger group of clinicians, but also that there are significant differences among the nurse population itself that must be considered in order to effectively introduce new systems to older employees.

5 Discussion and Conclusions

The goal of this research study was to identify the factors that best predicted the successful adoption of an electronic health record system among older nurses within a long-term care facility. After analyzing the group sampled as a whole and then comparing various age groups, this study did reveal significant differences. The influence of image is significantly higher in the 55 and over age group compared to the other age groupings. This suggests that as nurses age, their professional identity and status begin to matter more. Older nurses hold their professional status within the organization in high regard and are, as a result, willing to more readily accept the introduction of new technology in order to secure and elevate this status.

The comparison of years of experience also resulted in significant differences. Image was significantly influential to those with the least experience as well as to those with the most, suggesting professional status was more important to nurses toward the beginning and end of their careers. While nurses in the beginning of their careers sought to establish their professional identities through acquiring technical knowledge, those towards the end of their careers sought to increase and secure the status they had already earned. Organizations can therefore call particular attention to the status benefits of the EHR system in question among older workers to facilitate their acceptance and support. The importance of subjective norm was also significantly higher in nurses with over 20 years experience compared to those with less, indicating that highly experienced workers also require strong leadership and support of their superiors to encourage the successful adoption of a new system.

The effect of perceived ease of use was also significantly different among various experience levels. Once again, nurses with the most experience as well as those with the least experience were influenced the most by ease of use. Older nurses reported having less experience with computer programs and this lack of experience can often lead to resistance in using newly implemented systems. These survey responses suggest, for nurses 55 and over, it is the lack of experience with computer programs that makes perceived ease of use so influential. The easier the program is to use, the more comfortable older nurses will feel using it, and the more likely they are to accept it in their well-established nursing routine.

Many health care facilities have recognized the benefits of electronic health record systems but the transition to such systems remains slow to such systems remains slow for some. Implementing new technologies in organizations as large and complex as long-term care facilities can be challenging, however this study has revealed that using subjective norms, image, and perceived ease of use as tools during implementation can greatly increase the likelihood of successful adoption among the increasing number of mature and experienced nurses in the health care system.

6 Limitations

Though it is believed the results of this study can make a significant contribution to the senior health care field, the conclusions may not be applicable to all long-term care facilities as a result of this study’s inductive structure. The primary research is limited to one long-term care facility and there is a possibility that it is not representative of all the long-term care facilities across the province and country. Ideally, there will be an opportunity to conduct similar research in other facilities and verify these findings.

In terms of data collection, the risk of response bias – such as having a certain type of nurse be more likely to respond – is always present. There is also room for error in the interpretation of the survey as the responses available were limited, there was little opportunity to expand on answers, and respondents may have misunderstood the questions posed.