E-learner’s continuance usage behavior of online learning: integration of ECM and TAM

The novelty and advancement of technology have explored new avenues in the education sectors. Now e-learning has blended with face-to-face learning to increase its effectiveness. An integrated model is being proposed in this study to measure the continuance usage intention of students adopting e-learning services. With the constructs of ECM (confirmation, perceived usefulness), perceived ease of use and perceived self-efficacy proposed to measure their effect on satisfaction. Besides, to explore the impact of quality features on satisfaction and continuance intention, three major determinants of D&M’s ISS model such as system quality, service quality and information quality included in this study. 410 responses collected to do quantitative analysis. The PLS-SEM analysis showed that perceived ease of use has the strongest effect on perceived usefulness, followed by perceived self-efficacy on perceived ease of use and satisfaction with continuance intention. The study reveals that challenges in developing countries are unique and drives the learners differently than developed countries. The growing e-learning industry requires widespread research from national and institutional perspectives for holistic development in the Bangladesh education market.


Introduction
The emergence and advancement of information technology creates the scope of digital learning, which replaces traditional learning.The web-based innovative learning platform through increasing engagement, accessibility to resources, getting immediate responses, greater flexibility and remote learning has provided digital learning.Though online learning already is an established, active and significant learning system in many learning institutions and organizations (Lee et al. 2020), it has become highly popular and the only source of learning during the pandemic.During COVID-19, most of the countries concentrated on online learning as a crisis reaction, which was supposed to reach all understudies.Severe impact of pandemic is being observed on higher education in developing countries than in developed countries.
Many researchers have observed and shown this severity in their studies such as India (Sengupta, 2022), Bangladesh (Al-Amin, et al. 2023;Maisah & Shetu, 2023), Sri Lanka (Lakmal et al. 2021), Nigeria (Egielewa, et al. 2022) and Iran (Taghizadeh et al. 2021).Among the students, this web-based learning is highly accepted and still exceptionally popular among them.Online learning leads students toward self-directed and self-arranged learning, interactivity, personalized learning, flexibility of educational activity (Bidin & Ziden, 2013;Docimini & Palumbo, 2013).Because of these unique characteristics, most higher education institutions started online learning and decided to continue blended education in the post-pandemic situation (Rosli and Saleh, 2022;Taghizadeh et al. 2021).Therefore, the demand for e-learning is still growing and there is a gap for research to explore the potential factors affecting the technology-based learning system.COVID-19 pandemic has changed the acceptance behavior of students toward e-learning and remains after the lockdown situation withdrawn.
Revolution of online learning not only changed the offline education practices in schools and universities but also changed the tuition culture in Bangladesh among students.Now they are experiencing easily accessible interactive online tuition and performing well in assessment and tests.Many institutions are also offering blended approaches where distant learning is offered by traditional face-toface learning practices.In a study, Islam et al. (2022) found that the success of blended learning highly related to the subject of competence of the course teachers/instructors to blend face-to-face and online learning to achieve ideal student participation and education.Many researchers have suggested that blended learning was the best medium of education and is a good topic of research and experimentation (Kituyi & Tusubira, 2013).Though many researchers have explored some barriers and challenges online learning has faced (Anwar et al. 2020;Abuhammad, 2020).Among them some major barriers mentioned by Hoque et al. (2021) and Li & Lalani, (2020) are lack of resources, infrastructure deficiency, poor internet connections, learning alone at home and access to alternative e-learning platforms.Whereas, in another study, a student's mental status, technical resources, the plan and design and courses and the interaction between educators and students enhances the effectiveness of online e-learning in Bangladesh (Islam & Habib, 2021;Alam, 2020).After the pandemic majority of the studies concentrated on the adoption and satisfaction of the online learning platform and argued that services of e-learning and its sustainable development will be the next focus point of post-pandemic education (García-Morales et al. 2021;Chattaraj & Vijayaraghavan, 2021).
As a result, the Technology Continuance Theory (TCT) is considered as the basis for this study and tries to identify the constructs that influence higher studies students' decisions to continue using e-learning in the post-COVID era.Besides, the continuation of new technology also highly depends on the success of its implementation.

Model Conceptualization Technology Continuance Theory (TCT)
To define continuance usage intention of ISs, previous studies incorporated three models, such as expectation confirmation model (ECM), technology acceptance model (TAM), and cognitive model (COGM).According to Liao et al. (2009), TCT shows a significant advancement over ECM, TAM and COGM both qualitatively and quantitatively, to determine the user's adoption behavior at different stages.Expectancy confirmation theory (ECT), introduced by Oliver (1980), has been widely used in IS to investigate the continuous usage behavior of IT.The aim of ECT is to understand the continuance behavior and loyalty to system use and claims that satisfaction of the user is the most important criteria to determine the user's continuance intention.The ECT model explains that consumer satisfaction results from a five-step process.First, consumers develop initial expectations for a particular product or service before making a purchase.Then accept and decide to use the product or service.After an initial period of use, an idea about the performance is formed based on its salient characteristics.Third, these performance perceptions are compared to prior expectations to determine the extent to which expectations are confirmed.Expectations can be positively invalidated (perceived product performance over expectations), confirmed (perceived product performance meets expectations), or negatively overridden (perceived product performance exceeds expectations) may fall short of expectations.Fourth, they feel satisfied or dissatisfied depending on their level of disconfirmation.Moderate satisfaction is maintained by confirmation, increased by pleasure because of positive disconfirmation, and reduced by disappointment because of negative disconfirmation.And finally, satisfied consumers intend to reuse the service or product in the future, whereas dissatisfied users later discontinue using the product or service.In this model, confirmation has an instant influence on satisfaction.As confirmation is the extent to which performance meets, exceeds, or falls short of an individual's expectations, subsequent to zero, positive and negative negation, accordingly (Oliver & Swan 1989).
While TAM introduced by Davis et al. (1989) originated an on theory of reasoned action (TRA) (Fishbein & Ajzen, 1975).Researchers (Davis, 1986;Davis et al. 1989) claimed that perceived ease of use (PE) and perceived usefulness (PU) are two important constructs which influence the attitude and usage intention of IS.PU is one of the significant elements of the TAM model and defines the level of adoption of noble technologies to enhance efficiency and performance (Davis et al. 1989).PU represents the conception that consumers usually adopt a service if it improves by technology use (Ryu et al. 2017) and positively impacts usage intentions (Hong & Zhu, 2006;Ng & Kwok, 2017).
Another component of TAM, PE, describes the level of effort related to using new technologies (Davis, 1986).The TAM describes that PE has a significant impact on the adoption behavior of noble technologies (Kumar & Chand, 2019).Many researchers have tested empirically and found significant relationships between PU and PE with behavioral intention (Davis et al. 1989;Venkatesh & Davis, 2000;Venkatesh, 2000).Karahanna & Straub (1999) and Taylor & Todd (1995) have applied TAM to study post-adoption and continuance behavior.As TAM focuses on initial acceptance of information technology, many researchers have added more external factors in combination with TAM to enhance explanatory power (Teo, 2019;Roca & Gagné, 2008).For individual satisfaction and continuance behavior, student confidence enhances efficiency and reduced effort.Self-efficacy (SE) is an important construct developed by Bandura (1977) in Social-Cognitive theory.SE denotes the domain and task specific views that people have about their capacity to organize resources and to carry out the actions necessary to perform the tasks successfully (Bandura, 1977).Strong SE can lead learners to behave in ways that are likely to improve their learning and academic performance.Thus, self-efficacy is considered having a high impact on PE (Ching-Ter et al. 2017;Rosli & Saleh 2022;Chien, 2012).Findings explored that student who have high SE are highly eager to use online learning and computer-assisted education (Hanif et al. 2018;Downey & Kher, 2015).

DeLone & McLean's ISS Model
The updated D&M's ISS model developed by DeLone & McLean (2003) to understand the field of e-commerce, but it applied in different sectors to measure the quality factors of information systems.Many authors have extended this model in e-commerce (Hsu et al. (2014)), acceptance of internet banking (Jagannathan et al. 2018) by adding new constructs.Whereas, other researchers have validated the model by studying Knowledge Repository Systems (Qian & Bock, 2005), e-commerce (DeLone & McLean, 2004), e-Government systems (Wang & Liao, 2008) and information systems (Iivari, 2005).In recent times, many researchers have focused on online education using the D&M's ISS model (Mohammadi, 2015;Seta et al. 2018;Alotaibi et al. 2022) as the use of the system is a significant criterion of the system's success.DeLone and McLean's ISS model contains six variables as system quality, service quality, information quality, use/user intention, satisfaction, net benefits (DeLone & McLean, 2003).System quality (SY) defines performance of the system, user-friendliness, and technical features.Continuous connection and access, high speed, ease of use and navigation are the common features of SY (Cidral et al. 2018;Gao et al. 2015).E-learning creates an opportunity for learning from any place at any time.Information quality (IN) refers to the accuracy, validity and availability of the content provided through the system.Almaiah et al. (2016) mentioned that IN is a significant variable that requires to be considered and plays an important role in the m-learning system's success.IN can measure as a cluster of dimensions of timeliness, relevance, completeness, correctness, and effective learning content (Almaiah et al. 2016).Service quality (SQ) explains the responsiveness of the technical staff and competences of them (DeLone & McLean, 2003).Whereas, use/ user intention is the first step in the system's success.User's satisfaction in the level of acceptance and satisfaction of the system.Net benefit refers to efficiency and job performance that arises from the perceived organizational and individual influences (Cidral et al. 2018).Mohammadi (2015) and Hassanzadeh et al. (2012) introduced an integrated ISS and TAM model to determine the achievement of the e-learning system in Iran.Very limited works have been conducted to understand the impact of e-learning usage drivers using ISS and ECM models in developing countries.This study tries to discover the research gap by applying the quality features (system quality, service quality & information quality) accompanied with confirmation, perceived usefulness, perceived ease of use and perceived self-efficacy on students' satisfaction and continuation of usage.

Research Framework & Hypotheses Development
According to Oliver's COG (Oliver, 1980) and Bhattacherjee's ECM (Bhattacherjee, 2001), satisfaction is referred to as the overall post-consumption judgement of an individual that arises after a particular transaction.The research framework of this study is shown in the fig-1.User satisfaction is a prime determinant of continued usage behavior.Individual satisfaction is defined as a linear function relational with confirmation/disconfirmation (Churchill & Suprenant, 1982;Oliver, 1980).Confirmation is positively involved with satisfaction, whereas disconfirmation indicates failure to reach expectation (Bhattacherjee, 2001).The ECM suggests that continued usage behavior of the user is impacted by the confirmation of expectation and PU.On the other hand, the TAM focuses that PE and PU are the key predictors of the user's behavioral intentions towards actual use.The PE is also often observed to influence the PU of technology adoption.In this study, researchers adopt technology continuance theory (TCT) by integrating ECM and TAM to predict satisfaction and continued usage behavior.Therefore, the following hypotheses are proposed: H1: Confirmation (C) has significant positive influence on PU.H2: Confirmation (C) has a significant positive influence on usage satisfaction.H3: Perceived usefulness (PU) has a significant positive influence on MFS usage satisfaction.
H4: Perceived ease of use (PE) has a significant positive effect on PU.
H5: Perceived ease of use (PE) has a significant positive effect on MFS usage satisfaction.PE shows how much an individual accepts that using an innovation will be low in effort (Davis, 1989).Many researchers have tried to explore the influence of external predictors on PE to better understand the technology acceptance behavior.Thus, in this study, perceived self-efficacy (PS) used as an external predictor of PE.Liaw (2008) explains that perception of self-efficacy is the capability to interconnect with technical tools.In e-learning, self-efficacy has been identified as an important motivating factor on the acceptance of technology-based learning (Wang et al. 2019).In a study, Saeed et al.User's satisfaction defines a general evaluation of an IS and reflects a sensation-based reaction aligned with the aim of IS (Lam et al., 2004).Many researchers have been studying satisfaction as an important predictor of the perception of users of the effectiveness of IS (DeLone & McLean, 2003;DeLone & McLean, 1992).Therefore, the hypothesis stands as: H11: Satisfaction (S) has a significant positive influence on continuance usage intention (CI).

Research and Methodology
Based on the theoretical framework, a model is proposed to investigate the continuance behavioral intention of e-learning.For the data collection purpose, the researcher conducted an online survey, as it is the swiftest and effective method to collect replies (Alraimi et al., 2015).A questionnaire has developed to collect demographic information and added items against each construct based on the literature.During the pandemic, most of the educational institutions converted to online learning, Bangladeshi is not exceptional.Now though most of the educational institutions started face-to-face learning, many are continuing e-learning approaches yet.Therefore, to continue the e-learning process and achieve sustainability, it is necessary to measure the satisfaction of students and the quality features of the IS.In this study, researchers added nine constructs to measure continuance usage behavior, integrating two popular models, such as ECM and TAM.Items related with perceived usefulness (PU), perceived ease of use (PE), perceived selfefficacy (PS), confirmation (C), system quality (SY), information quality (IN), service quality (SQ), satisfaction (S) and continuance usage intention (CI) from different literature sources shown in the table-2.Initially, a questionnaire was developed, including 35 items to measure the CI of the respondents toward online courses and classes.Researcher followed a two-step research process to confirm the content validity of the items in the questionnaire.In the first step, fifteen industry experts (teachers & students) were selected who are directly related to the development of the e-learning process.Based on their suggestions, 32 items included in the final questionnaire and sent to the selected respondents for the data collection purpose.For the better understanding of the respondents, the questionnaire was arranged in two languages, Bengali and English.

Data Collection Process
Convenience sampling method based on the subjective judgment of the researchers have chosen this study purpose (Saunders et al. 2019).Students recruited from different public and private universities who are continuing online learning practices after the pandemic period, regardless of gender, age and year of study.The questionnaire directed to the respondents (students/learners) by their instructors and teachers, and requested to fill-up and back again their responses within the due time.This online survey was continued for the period of 15 th October-07 th November, 2023.Initially, 500 questionnaires were sent to the respondents, 435 responses were returned.Among these, 25 questionnaires were incomplete and 410 responses were accepted for the final analysis purpose.The demographic information of the respondents presented in the table-1.(Hair et al. 2017).Perhaps, PLS-SEM forecast the level of changes in dependent variables because of independent variables (Wang et al. 2019).In this study, the confirmatory factor analysis (CFA) and structured relationships among the variables measured using SMART PLS-3 software (Hair et al. 2017).

Reliability & Validity
By following the suggestion of Hair et al. (2017), the proposed research model confirmed by testing the external measurement model.We have tested construct reliability of this model by applying Cronbach's Alpha, composite reliability (CR) and average variance extracted (AVE).The recommended composite reliability (CR) is more than >0.7, explaining 70% of the measurement model's variation (Hair et al. 2017;Bagozzi and Yi, 1988).The mentioned cut-off value of Cronbach's Alpha is >0.7 and AVE is >0.5 recommended by Hair et al. (2017).The tested value of Cronbach's Alpha, CR and AVE of this researcher has confirmed the required threshold level.To test discriminant validity of this model, we tested Fornell and Lacker criterion and showed the values in the table-3.According to Hair et al. (2017), discriminant validity is "the degree to which two conceptually similar constructs are distinct".When the coefficient of correlation between the pair of variables is smaller than 0.85 and the most values of the square roots of the AVE estimates are higher than the consistent coefficient of correlations, discriminant validity is met (Chen et al. 2020;Kline, 2005).Table -3 shows the discriminant validity tested by following Fornell and Larcker criterion.The bold values in the crosswise scales are the square root of AVE.CI: continuance usage intention, C: confirmation, IN: information quality, PE: perceived ease of use, PS: perceived self-efficacy, PU: perceived usefulness, SQ: service quality, SY: system quality, S: satisfaction

Hypotheses Testing & Structural Equation Modeling:
To analyze the collected data and to estimate the underlying relationship amongst the variables, Structural equation modeling (SEM) was applied in this study.The results of the model fit statistics were χ 2 /df = 2.498, GFI=.926,NFI= .918,CFI= .907,RMSEA= .033,and met the acceptance level of goodness-of-fit indices.In the first stage, this model starts with the dimensions of ECM, TAM and ISS model (confirmation, perceived usefulness, perceived ease of use, perceived self-efficacy, system quality, service quality and information quality), consisting of seven independent variables, to measure the satisfaction and continuance behavioral intention of the students.

Table 4: Standardized Regression Weight for Path Coefficients
The results of the structured model displayed in the table-5 comprising standardized path coefficients (β), t-values and p-value for the hypothesized relationships.The higher the R2 value explained, the interpretation, indicates better the model.The R2 value of this model for continuance intention is R2= 0.675, satisfaction is R2=0.778,PU is R2=0.656 and PE is R2=0.766 which were acceptable.Therefore, the research model shows good explanatory power.The significant path coefficient shows that H2, H3, H4, H5, H7, H8, and H11 were supported based on the t-value and p-value.Whereas, H1, H6, H9 and H10 not supported.

Discussion
The results explore the effects of the constructs of the TCT and ISS model on the e-learning continuance intention.Most of the hypotheses of the TCT and ISS models are supported, with some exceptions.In the study, PE found to have the strongest significant effect on PU, which supports many findings of the researchers (Mohammadi, 2015;Iqbal & Bhatti, 2015) in IS.The results shows that students found the usefulness of accepting online learning when it is easy to use.PU is an important predictor of student satisfaction and found significant effect which is like to the other findings (Al Amin et al. 2022;Gilani et al. 2016;Liao et al. 2009).
Whereas, in this study, PE also found a straight influence on satisfaction, where most of the researchers found an indirect effect on satisfaction through PU (Baki et al. 2018;Cheng-Hsun, 2010).Besides, as an external variable, PS has found a significant effect on PE in this study, which supports other research findings ( Al-Adwan et al. 2023;Luo & Du, 2022;Baki, et al. 2018).It showed that students relied on the ability of online learning to increase their confidence and control over their learning through technology/devices.But no direct PS on satisfaction is not found in the analysis which is different (Hassanzadeh et al. 2012;Wang and Chiu, 2011).Another important construct of ECM is confirmation is found no significant effect on PU which supports the findings of Li & Phongsatha (2022); Liao et al. (2009) but contradicts the findings of previous research (Al Amin et al. 2022), indicates that student's prior expectation is confirmed then satisfaction is formed.Whereas, confirmation has significant positive relationship with satisfaction and supports some prior research (Al Amin et al. 2022;Taghizadeh et al. 2021).It might be claimed that students usually first try to find that the system meets their expectations.Then they realize that e-learning is the better option than conventional learning, creating a sense of confirmation.
The results show that system quality, one of the important constructs of the ISS model, helps to measure the effectiveness of the technology and found a positive effect on user's satisfaction in this study which corresponds with the studies of Mohammadi (2015), Hassanzadeh et al. (2012).Information quality showed no significant impact on satisfaction, contracting the findings of Mohammadi (2015), Salam & Farooq (2020), Hassanzadeh et al. (2012).Here, contrary to previous studies (Poulova and Simonova, 2014;Xu et al. 2014), service quality does not positively affect user's satisfaction.The findings suggested that information quality and service quality are not significant predictors in this country context, as the service providers are not much conscious about these issues yet.This observes that future research may find out the causes to confirm the assumptions.Finally, satisfaction was explored as a significant predictor affecting continuance intention in the e-learning context parallels the studies of Taghizadeh et al. (2021); Hassanzadeh et al. (2012).As the interconnection between satisfaction and continuance intention is the most significant relationship in ECM (Bhattacherjee, 2001).Consequently, the degree of satisfaction results in continuance intention among the students.
Universities and institutions should enhance satisfaction which will in turn increase continued e-learning.

Conclusion
Technology based online learning is a noble innovation in education.It opens an opportunity for learners to gain knowledge from the best sources from anywhere in this world.As technology ensures reliability and accessibility, many researches have focused on the attitude and behavior of e-learners and instructors, playing a significant role in the achievement of e-learning (Cheng, 2011;Liaw et al. 2007).
Technology continuance theory is an effective theory designed for modeling the continuous usage behavior of users.In this study, we also examined a combined model of DeLone and McLean's ISS model and expectation confirmation model to investigate user's continuance behavior toward e-learning systems and the impact of quality features affecting users' satisfaction.
This study shows that perceived ease of use has the strongest influence on perceived usefulness.It can be clarified that student's feeling of perceived easiness and effectiveness of e-learning direct them toward technology assisted online learning.In addition, students' confidence and ability to use and interact through the internet enhance their performance and efficiency.Therefore, perceived self-efficacy has a positive influence on learner's perceived easiness, which is proved in this study as well.A strong significant impact of satisfaction is also observed in the study that satisfies the established relationship mentioned in the ECM model.To attain satisfaction, the results of confirmation of expectations from prior use of IS are very important.
A significant impact of system quality on satisfaction is observed in this study.Whereas, service quality and information quality are found to have no impact on satisfaction which indicates people are still unaware about content quality and support services offered in e-learning.In the context of emerging economies, e-learning is still in the development stage and requires intensive research and proper guidelines to increase acceptance among the users.Thus, e-learning faces unique challenges from developed countries.Despite the challenges, proper attention of teachers, administrators and instructors in the content development and incessant upgradation may ensure more engagement and satisfaction among e-learners (Sumi & Raju, 2021).This study is conducted on one culture context, more studies on cross-cultural contexts and settings can explore different possibilities and findings.A comprehensive contribution of researchers, national policy-makers, industry experts and stakeholders can ensure a substantial development and growth in the online industry.
Therefore, this study aims to design a research model based on Technology Continuance Theory (TCT) which integrated the Technology Acceptance Model (TAM) and Expectation Confirmation Model (ECM) to predict the individual continuance usage intention.To identify the success of the system and predict an individual's satisfaction and continuance usage behavior, researchers added DeLone & McLean's ISS Model with the ECM in this study.To analyze the relationship between the constructs researcher used Structural Equation Modelling (SEM).
(2020)  identifies that student's PS, PE and PU are the most dominant factors influencing continuous intention to adopt technology.Therefore, the hypotheses stand as: H6: Perceived self-efficacy (PS) has a significant positive impact on PE.H7: Perceived self-efficacy (PS) has a positive impact on satisfaction.In this study, researchers also added DeLone & McLean's ISS Model(DeLone & McLean, 2003) to discover the effects of quality features on satisfaction and continued acceptance behavior.The success of the ISS and user's satisfaction, system quality, quality of information and services quality are important predictors of ISS model.Thus, the hypotheses are: 385 H8: System quality (SY) has significant positive effect on MFS usage satisfaction.H9: The quality of information (IN) positively affects MFS usage satisfaction.H10: The service quality (SQ) has a positive effect on MFS usage satisfaction.

Figure 1 :
Figure 1: The Conceptual Model

Figure 2 :
Figure 2: The Output of The Coefficients (P-Values) Are Reported in The Following Fig SEM is used to measuring the parameters of ECM and the TAM on the student's satisfaction level and continuance usage intention.Figure-2 shows the results of the structural model of online learning satisfaction and continuance behavior (p<.001).In the first stage, this model starts with the dimensions of ECM, TAM and ISS model (confirmation, perceived usefulness, perceived ease of use, perceived self-efficacy, system quality, service quality and information quality), consisting of seven independent variables, to measure the satisfaction and continuance behavioral intention of the students.

Table 1 :
Respondents' Demographic Profile To analyze the models and equations, Structural Equation and Modeling (SEM) is applied in this research work.SEM analyzes a multifaceted model with several variables(Cohen et al. 2018) and underlies the relationship amongst the variables.Two categories of SEM are available namely, covariance-based SEM (CB-SEM) and Partial least square (PLS-SEM).CB-SEM analyzes the fit statistics among the exogenous and endogenous variables to exploit the explained variances based on forecasting and estimation Table-2 represents factor loadings, value of Cronbach's Alpha, CR and AVE of the constructs and all items.

Table 2 :
Standardized Factor Loadings, Cronbach's Alpha, CR and AVE Value

Table 3 :
The Discriminate Validity Was Accepted Following Fornell and Larcker Criterion Note: