Evaluating the impact of quality antecedents on university students’ e-learning continuance intentions: A post COVID-19 perspective

Abstract The COVID−19 pandemic forced service providers to resort to online servicescapes across diverse industries globally. However, with the epidemic curve flattening in most parts of the world and the pandemic showing indications to phasing out, educational institutions are now operating on the continuum between blended e-learning and full-scale e-learning services. The purpose of this study was to examine the impact of e-learning quality antecedents on university students’ e-learning continuance intentions in Zimbabwe post the COVID−19 pandemic. The study adopted a causal research design within the positivism research paradigm. The study population were undergraduate students using e-learning at three universities in Zimbabwe. A random sampling procedure was employed to collect data. A quantitative approach validated 311 responses using covariance-based Structural Equation Modelling (SEM) in SPSS AMOS. The study revealed positive and statistically significant relationships between three quality factors (system quality, instructor quality and information quality) and students’ e-learning continuance intentions (p < 0.001). However, the impact of support service quality on e-learning continuance intentions was not statistically significant. The paper flags the key roles of instructor quality, information quality and system quality as key determinants of students’ willingness to continue using e-learning systems in a post COVID−19 era. Universities were urged to design e-learning systems that engulf state of the art communication technology attributes to ensure robustness, flexibility, speed, security, visual appeal, compatibility, availability and user-friendliness. The study also recommends periodical reviews to motivation, resourcing, training and development strategies for e-learning instructors to enhance their delivery during e-learning service encounters.


PUBLIC INTEREST STATEMENT
Digital transformation in the higher education industry has advanced in recent decades in the most affluent economies globally. The COVID-19 pandemic forced universities, other service providers alike, to resort to online services. However, technology infrastructure among other constraints continues to cripple sustainable e-learning success in most pre-emerging economies. This research investigates the influence of e-learning quality factors on university students' e-learning continuance intentions in Zimbabwe. The survey data was collected from undergraduate students at three local universities and data analysis was conducted using quantitative analyses. Our research reveals that e-learning continuance is positively influenced by system quality, information quality and instructor quality. This study informs the universities' customercare strategy in a world where customer voice has grown too pervasive to disregard. The research further enlightens the higher education industry that student evaluations of e-learning service are critical to university success given their correlation with future student behavioural intentions.

Introduction
The Corona Virus (SARS-COV−2 or COVID−19) pandemic resulted in closure of universities in more than 180 countries globally Shehzadi et al., 2020;UNESCO, 2020;Wang et al., 2021). In order to remain afloat, universities resorted to emergency electronic learning (e-learning) by employing various e-learning management systems which supported e-service delivery (Edelhauser & Lupu Dima, 2020;Li et al., 2021). E-learning systems such as Blackboard Learn, Canvas, Moodle and Schoology were popularised globally whilst web-based services such as Google Classroom, Google Meet and Zoom are complementing e-learning (Wang et al., 2021). To date, most universities are surviving on the continuum between blended e-learning and full-scale virtual operations (Favale et al., 2020;Islam et al., 2022). Due to school closures at the height COVID−19 pandemic, much urgency has been thrust on engaging students over digital learning servicescapes (Al Muhem & Wang, 2020;Li et al., 2021). The higher education industry in Zimbabwe was not an exception as e-learning adoption has been widely observed to meet the demands of the new normal (Dzvimbo, 2020;ZIMCHE, 2020).
Notwithstanding the initial COVID−19 induced e-learning developments observed in Zimbabwean universities, Information Systems (IS) literature confirms that IS continued usage represents its ultimate success. In the view of COVID−19 as the push factor behind the digital learning transformation, the World Health Organisation (WHO) report on the COVID−19 prevalence informs that most countries have managed to flatten the epidemic curve through vaccination, wearing face masks, maintenance of stipulated hygienic practices and recommended social distancing (WHO, 2023). This implies that most universities that were not fully e-ready to implement e-learning might be subjected to resort back to face-to-face learning. Although digital learning has long been envisaged in higher education given its effectiveness in service delivery and relatively lower operational costs (Al Muhem & Wang, 2020;Lwoga & Sife, 2018;Obeng & Coleman, 2020), research from most pre-emerging economies Edelhauser & Lupu Dima, 2020) suggests that most universities lack the required financial resources and infrastructural establishments to fully support online learning.
Despite e-learning being imperative, a host of digital constraints has long been suggested in previous studies. Most students face financial constraints to keep up with the high cost of internet services (Chitanana et al., 2008;Dube & Scott, 2017;Lwoga & Sife, 2018). Poor service quality has also been observed as most learners face online disappointments attributed to poorly designed systems, poor user interface and unreliable internet services (Edelhauser & Lupu Dima, 2020;Mukeredzi et al., 2020;Mwiya et al., 2019). More so, university students are confronted with a steep learning curve to fully appreciate the requirements of digital learning (Al Muhem & Wang, 2020;Marandu et al., 2019). Hardware limitations have also been observed in previous research as learners do not have the adequate computers, tablets and other mobile gadgets needed to engage their lecturers on e-learning encounters (Dube & Scott, 2017;Mukeredzi et al., 2020).
Whilst universities are operating on a continuum between blended e-learning and full-scale virtual operations, most universities have resorted to blended e-learning due to capacity limitations (Edelhauser & Lupu Dima, 2020;Shehzadi et al., 2020). The phenomenon connotes that e-learning future continuance remains unclear due to these challenges and this presents a critical question for researchers and university management. Whilst implementing e-learning represents a milestone of digital progress for the higher education industry in Zimbabwe, continued usage remains key for its ultimate success (Bhattacherjee, 2001;Dehghan et al., 2014;DeLone & McLean, 2003;Wang et al., 2021). To the best of the authors' knowledge, no studies in Zimbabwe have been operationalized to evaluate university students' e-learning continuance intentions prior and post the COVID−19 pandemic. The exploratory literature focuses on the challenges and opportunities for universities as they adopt digital learning (Chitanana et al., 2008;Dube & Scott, 2017;Mukeredzi et al., 2020). These studies provide a key reference point from extant literature that lends a critical research gap for the current study.
Several studies have investigated information systems continuance in various research contexts E.g., MOOCs in China (Shanshan & Wenfei, 2022), e-learning management systems in Brazil (Cidral et al., 2020), faculty e-resources in Tanzania (Lwoga & Sife, 2018), nurses' blended e-learning systems in Taiwan (Cheng, 2014) and students e-learning in Jeddah (Khamis et al., 2021). Diverse theoretical perspectives have been employed to measure e-resources usage continuance, e.g., expectations disconfirmation model (Bhattacherjee, 2001), flow theory (Csikszentmihalyi, 1975), technology-task fit theory (Goodhue & Thompson, 1995), Information Systems Success Model (ISSM) (DeLone & McLean, 2003). From his meta-analytic research, Jeyaraj (2020) notes that the ISSM has been adopted in most e-learning continuance studies because of its inclusion of IS quality dimensions. This research borrows theoretical foundations from the ISSM and the flow theory to evaluate e-learning continuance. Instructor quality has been vastly applied in most e-learning evaluation studies, e.g., Cheng (2014), Al-Adwan et al. (2021) and Pham et al. (2019); however, prior research based on the ISSM model did not integrate instructor quality. This study merges predictors from the ISSM model (system quality, information quality and service quality) with instructor quality to examine e-learning continuance from the perspective of a pre-emerging economy.
This research aims to fill a practical gap in the way Zimbabwean universities design and implement their student Customer Relationship Management (CRM) strategy. The popularity of the student-centered approach implies that universities are obliged to pay more attention to student perceptions of educational service quality (Islam et al., 2022;Kilburn et al., 2016). Today, more universities are changing their strategy of managing relationships with students by considering them as customers and universities as providers of educational services (Kilburn et al., 2016;Martínez-Argüelles & Batalla-Busquets, 2016). Universities are now trying to leverage their marketing orientation to enhance student satisfaction, lifelong learning, word-of-mouth support and customer loyalty Dehghan et al., 2014;Shehzadi et al., 2020). Thus, this research awards an opportunity for universities to understand the perceptions of e-learning success from a student perspective. This informs university policy and strategy as they attempt to deliver a customer centric service.
It is against this background that this study examined the influence of e-learning quality antecedents on university students' e-learning continuance intentions. The current section also provides the research problem, purpose, significance of the study and research questions. Subsequent sections dwell on literature review, materials and methods, data analysis and results as well as conclusions, recommendations, limitations and prospects for future research.

Statement of the problem
E-learning has recently captured much attention given the eruption of the COVID−19 pandemic. However, given the significant drop in new infections (WHO, 2023), most service providers are now standing on the continuance-discontinuance dilemma; thus, e-learning continuance has become a critical topic in both research and practice. Although previous research has examined e-learning continuance (Chang, 2013;Li et al., 2021;Lwoga & Komba, 2015;Mamoodi et al., 2019;Ramayah et al., 2010;Wang et al., 2021), these studies were conducted in the most affluent economies with sound technology infrastructure. Previous research in Zimbabwe has shown that more investment on the technology infrastructure is still imperative in order to fully operationalize e-learning. Studies by Dube and Scott (2017), Marandu et al. (2019) and Mukeredzi et al. (2020) observe that significant network deficiencies, poor service quality, hardware constraints, financial resource inequalities among learners, low skillset and negative learner and instructor attitude have stalled the success of digital learning systems in Zimbabwe. Thus, post COVID−19 e-learning continuance has emerged a key research topic for researchers and most universities in subsistence markets. As e-learning continuance remains critical for Zimbabwean higher education industry, this research investigates the differential effect of e-learning quality antecedents on university students' e-learning continuance intentions.

Purpose of the study
The success of e-learning implementation greatly hinges on its continued use by the learners. The higher education industry in Zimbabwe needs to establish the future viability of its virtual learning servicescapes. Thus, the purpose of this research is to examine the influence of e-learning quality attributes on students' continued use intentions. The results of the study provide a proxy for determining the success of e-learning implementation in a post COVID−19 context.

Significance of the study
Most university management systems have been digitalized globally; therefore, it remains imperative for Higher Education Institutions (HEIs) to advance their research efforts towards digital sustainability. The findings of this research assists universities in Zimbabwe to understand student (customer) perceptions of their e-learning service quality, hence e-learning success from a customer perspective. Given the popularity of the student-centered approach in university management globally, this research enlightens HEIs on the adequacy or lack of it on their digital learning systems, and that provides a basis for continuous review and improvement. More so, universities can use this research as an impetus to measure e-learning service quality in future. The measures used in this research are validated from a Zimbabwean context; hence, universities may adopt them for their self-evaluation surveys. Furthermore, the recommendations proffered in this paper also provide valuable insights on the relative importance of e-learning quality dimensions to learners; hence, they serve as a proxy for designing a customer education centric service.

Research questions
The research intended to provide answers to the following questions; (1) Does system quality influence significantly students' e-learning continuance intentions?
(2) Does instructor quality significantly affect students' e-learning continuance intentions?
(3) Does support service quality significantly determine students' e-learning continuance intentions?
(4) Does information quality significantly influence students' e-learning continuance intentions?

Theoretical perspectives
IS continuance is not a new concept in IS research. In his post-acceptance IS confirmation model, Bhattacherjee (2001) claims that continuance intention has been conceptualised in earlier studies as "implementation" (Zmud, 1982), "incorporation" (Kwon & Zmud, 1987) and "routinisation" (Cooper & Zmud, 1990). According to Bhattacherjee, these earlier studies acknowledged the existence of a post acceptance stage when IS use transcends conscious behavior and becomes part of normalised routine activity. More so, there is an acceptance-discontinuance anomaly that happens when users initially accept IS and decide to discontinue IS use (Bhattacherjee, 2001). Thus, in his post acceptance IS model, Bhattacherjee explains the importance of measuring user continued usage intentions among IS users.
IS continuance intention is also rooted in the innovation diffusion theory (Rogers, 1995). The five stages of knowledge, persuasion, decision, implementation and confirmation suggest that adopters re-evaluate their earlier acceptance decision during the final confirmation stage and make a decision on whether to continue or discontinue IS use (Bhattacherjee, 2001;Rogers, 1995). Thus, it suffices to infer that the innovation diffusion theory recognizes the presence of continuance evaluation during the confirmation stage. Whilst initial acceptance demonstrates a huge step in IS success, long-term viability of an IS depends on its continued usage than first time use (Bhattacherjee, 2001;Li et al., 2021;Wang et al., 2021).
The flow theory (Csikszentmihalyi, 1975) has also been one of the most cited theoretical frameworks in IS continuance research. Flow describes a state in which individuals engaged in a satisfying activity get completely lost, immersed, and lose self-consciousness. The theory describes a person who feels completely indulged. Flow is an intrinsic motivator and a key determinant of human behavior (Csikszentmihalyi, 1975). This theory also lends theoretical support for this study. University students who are satisfied with e-learning have more intrinsic motivation for future e-learning usage continuance. This theory has also been vastly applied in several studies that investigated future human behavioral intentions in various research settings.
The ISSM (DeLone & McLean, 2003) has also been extensively used to explain IS usage and continuance. In their updated model, DeLone and McLean reiterate that IS success brings organizational benefits, mediated by user satisfaction, intention to use and actual use. The key determinants of intention to use, usage and user satisfaction being system quality, information quality and service quality. In his meta-analytic study, Jeyaraj (2020) notes that over 8000 IS success studies have employed the DeLone and McLean model in different contexts to account for IS continuance. Thus, DeLone and McLean (2003) also opine that IS continuance represents a key dimension of IS success and hence it occupies critical space in IS research. Wang et al. (2021) defined e-learning continuance intentions as the willingness to keep or continue using virtual learning services. In his study on the digital libraries continuance usage intentions in China, he described continuance intentions as the will to voluntarily make a decision to continue using e-learning. It is the degree to which a person has formulated conscious plans to perform or not future behaviors (Bhattacherjee, 2001;Chang, 2013). Lwoga and Sife (2018) also defined continuance intentions as the will of users of an e-learning management system to use it in future. Thus, continuance intention to use e-learning systems can be identified as a proxy for e-learning management systems success in the post the COVID−19 era.

Quality antecedents of e-learning continuance intentions
Whilst a number of models have been employed to determine the antecedents of e-learning continuance intention (Wang et al., 2021), studies directly modelling e-learning quality factors have been sparse. The post-acceptance expectation-confirmation model (Bhattacherjee, 2001) has been the most cited IS continuance model. The technology task fit theory (Goodhue & Thompson, 1995) has also been integrated to explain IS continuance intentions. Researchers have also attempted to use decomposed models involving technology adoption (Davies, 1989) and any of the above mentioned IS continuance theories (Al-Fraihat et al., 2020;Shehzadi et al., 2020;Wang et al., 2021). The antecedents were drawn from ISSM (DeLone & McLean, 2003) and Cheng (2014). The ISSM model proposed intention to use an IS as an outcome of system quality, information quality and support service quality. Cheng (2014) integrated the Expectation Confirmation theory and IS success model to evaluate blended e-learning continuance. Thus, instructor quality was borrowed from Cheng (2014). From a wide range of IS continuance models, we adopted these based on their inclusion of IS quality factors as predictors of e-learning continuance. (2003) defined system quality as the system processing capabilities and the ability of an information system to perform its functions. The dimensions of system quality were defined in terms of availability, ease of use, reliability, access, security, log in, speed and navigation. System quality includes all features that are required for an information system to be able to perform its functions as directed by the end users (Li et al., 2021). System quality has been cited a key determinant of information systems acceptance and usage continuance (Chang, 2013;Cidral et al., 2020;Khamis et al., 2021;Mamoodi et al., 2019).

Instructor quality.
Instructor quality has been defined as the ability of the instructor to motivate, stimulate, give direction and purpose, encourage and convey learning sessions to students (Cheng, 2014). It has been identified as the competence of the instructor to facilitate successful e-learning sessions to learners in a virtual class. Instructor quality was defined in terms of instructor efficacy, care, technical competency, friendliness, warmth, ease and depth (Cheng, 2014;Pham et al., 2019). Instructor quality has also been conceptualised as facilitator quality in other studies (Wang et al., 2021).
2.2.1.3. Support service quality. According to DeLone and McLean (2003) service quality refers to the ability of the system to rectify user problems whilst they are using it. System quality also accounts for the ability of the IT and administrative staff to assist users to use the information system successfully. Service quality was added to the updated DeLone and McLean (2003) after a plethora of empirical concerns from different contexts (DeLone & McLean, 2003). Attributes of good support service quality include empathy, responsiveness, friendliness and efficiency. Clarity of instructions, steps of how to use the information system and the ability of the IT support staff to rectify student challenges were also included as key dimensions of good e-learning support service quality (Cidral et al., 2020;Shehzadi et al., 2020). (2003) defined information quality as the quality of content that users access from the information system. Dimensions such as format, compatibility, relevancy, understandability and freshness were used as good attributes of information quality. Li et al. (2021) also defined information quality as the ability of the content uploaded on an information system to accurately convey the intended meaning to the user. Information quality has been cited as a quality antecedent of user continuance usage behavior (Lwoga & Komba, 2015;Lwoga & Sife, 2018;Wang et al., 2021). The quality of information coming out of the system cannot be better than the quality of information that was put in the system, Garbage in Garbage out (GIGO) (DeLone & McLean, 2003).

System quality and students' e-learning continuance intentions
The relationship between system quality and continued usage intentions has been discussed in literature. From his IS continuance studies, Chang (2013) investigated the relationship between system quality and students' digital library continuance intentions. Their study found a positive and significant impact of system quality on users' continuance intention. As a sequel, Lwoga and Sife (2018) also examined the impact of quality determinants on faculty members continued usage of e-resources at three Namibian universities. Using Confirmatory Factor Analysis (CFA) and Structural Equation Modelling (SEM), their findings reported a significant relationship between system quality and continued usage intentions. Their findings supported those results by Alkhawaja et al. (2022), DeLone and McLean (2016), Kuadey et al. (2022) and Li et al. (2021). Their results corroborated the earlier findings of Ramayaha and Ramayah et al. (2010), who confirmed the positive effect of system quality and e-learning continuance intentions. As a result, this study proposed that; H 1 : System quality positively affects students' e-learning continuance intentions.

Instructor quality and students' e-learning continuance intentions
The importance of instructor quality in e-learning success has been observed in e-learning literature (Pham et al., 2019). Pham et al. (2019) established that the objectives of the instructor are to instill self-motivation, confidence, direction, purpose and to facilitate the e-learning process. Wang et al. (2021) notes that the paradigm shift from face-to-face to electronic learning implies a change process that requires adequate and consistent human skills to motivate learners' active engagement and commitment to e-learning. Students' e-learning continuance intentions hugely depends on the ability of the instructors to instill interest, acceptance and comfort in learners (Cheng, 2014;Wang et al., 2021). Thus, borrowing from Wang, good instructor quality positively correlates with e-learning continuance intentions.
Although a few studies have operationalised the direct influence of instructor quality on e-learning continuance, Pham et al. (2019) also examined the influence of instructor quality on overall e-learning service quality, claiming that better student perception of instructor quality would positively influence student participation, commitment and satisfaction with e-learning. Earlier findings by Cheng (2014) claimed the positive influence of instructor quality on continuance intention, mediated by student satisfaction. These findings gained the support of Al-Adwan et al.
and Islam et al. (2022). Given the theoretical and empirical support, we also proposed that; H 2 : Instructor quality positively influences students' e-learning continuance intentions.

Support service quality and students' e-learning continuance intentions
The significance of service quality in e-learning success has been extensively discussed in literature (Al Muhem & Wang, 2020;DeLone & McLean, 2016;Li et al., 2021;Mwiya et al., 2019). Earlier on, Ramayah et al. (2010) modelled service quality as a continuance intention antecedent in Malaysia. They alluded to efficiency in the university technical support as a key factor which motivates students to continue using e-learning. According to Lwoga and Sife (2018) service quality is the users' perception of the support they receive from the technical staff. Their results were also confirmatory of the positive influence of service quality and faculty members' e-resources continuance usage intentions. Adequate support staff is required to motivate IS use in emerging economies given the limited information systems literacy (Khamis et al., 2021;Lwoga & Sife, 2018). These claims are consistent with the findings of previous studies (Alkhawaja et al., 2022;DeLone & McLean, 2016;Favale et al., 2020;Kuadey et al., 2022). In light of these conclusions, the study also predicted that; H 3 : Support service quality positively influences students' e-learning continuance intentions.

Information quality and students' e-learning continuance intentions
Information quality has long been cited as a key IS feature (Chang, 2013;Cidral et al., 2020;DeLone & McLean, 2003Li et al., 2021). A positive user perception of information quality also enhances the user's continuance intention (Shehzadi et al., 2020;Wang et al., 2021). According to Wang et al. (2021), information richness, update regularity, format, understandability and relevance are strong attributes of good information quality. Learners establish e-learning as a useful means of knowledge acquisition (Daultani et al., 2021). Panigrahi et al. (2021) investigated the relationship between information quality and user continuance intention. The results reflected that information quality positively influences user's continuance intention with e-resources. Lwoga and Sife (2018) also modelled information quality as a predictor of faculty members continued usage intentions in Namibian Higher Education. Their findings also confirmed the positive influence of information quality and continuance intentions. This was supported also by the findings of Daultani et al. (2021), Al-Adwan et al. (2021)and Naveed et al. (2021). However, Mamoodi et al. (2019) argued that information quality was an insignificant e-learning quality determinant from their studies with nursing students in Alborz, Iran. Amid this foregoing discussion, we also hypothesized that; H 4 : Information quality positively influences students' e-learning continuance intentions.

Conceptual framework
The study proposed that e-learning quality factors (system quality, instructor quality, information quality and support service quality) were positively drivers to university students' e-learning continuance intentions (Figure 1). These were proposed to influence students' e-learning continuance intentions.

Design
The study employed an explanatory research design within the positivism research philosophy (Malhotra et al., 2017;Saunders et al., 2018). The study examined a hypothesized research model using Structural Equation Modelling to estimate model parameters. Causal relationships proposed by the research hypotheses were evaluated, and this was the basis for generating research findings.

Population and sampling
The research was conducted in a cross-sectional survey. The study population were 8330 undergraduate students at three universities in Zimbabwe who had prior e-learning use. Fieldwork was done during the February-June semester, 2023. Using simple random sampling, 350 questionnaires and 327 were returned, a response rate of 93.42%. Common with the quantitative research approach, data was obtained using a random sampling procedure. The sample size determination was influenced by data analysis methods, population size, resource considerations and sample sizes used in similar studies (Hair et al., 2020;Malhotra et al., 2017). According to Hair et al. (2019), at least 300 cases suffice for a study based on Confirmatory Factor Analysis (CFA). Tabachnick and Fidell (2013) also suggest that in quantitative studies, at least 10 cases per observed variable are adequate for estimation of model parameters. The research model had 24 observed variables hence at least 240 cases were adequate. More so, related e-learning studies also employed similar sampling guidelines, e.g., Al Muhem and Wang (2020)

Measures
Research instruments were adopted from Cheng (2014) Lwoga and Sife (2018). They were conceptualised as system quality, instructor quality, information quality, support service quality and e-learning continuance intentions as shown in Table 1. All constructs were measured on a 7-point Likert scale from Strongly Disagree (SD) to Strongly Agree (SA). Content validity was obtained by using measurement instruments validated from previous research. The study also used a pilot study to verify the suitability of the measurement scales.

Data collection procedures and ethical compliance
Compliant with research ethics, access was granted to university students (Malhotra et al., 2017). The purpose of the research was openly shared with participants and participation was voluntary. There researchers ensured that the survey was conducted with strict adherence to requirements of informed consent, privacy, confidentiality (Hair et al., 2020). The researchers hand delivered the questionnaires and collected them after 1 week. This allowed the respondents to fill them at their own convenience, thus also minimizing social desirability bias (Saunders et al., 2018). The participants were also informed that there was no right or wrong response.

Data analysis methods
The study employed Structural Equation Modelling (SEM) to examine the model through a stepstep CFA-SEM procedure in AMOS (Anderson & Gerbin, 1988). The study employed Average Variance Explained (AVE) and the Fornell and Larcker (1981) criteria to evaluate convergent and discriminant validity, respectively (Hair et al., 2019). The AVE is the mean of the squared extracted variances between a set of observed variables on a parent factor (Hair et al., 2019;Kline, 2023). According to Hair et al. (2019), the AVE should be greater than 0.5 to confirm convergent validity. The AVEs were calculated using the standardized factor loadings obtained in CFA.
Furthermore, Fornell and Larcker (1981) established the statistical criteria for evaluating discriminant validity using the comparison between the construct correlations and their corresponding square root of the AVE. The square root of the AVE should be greater than the inter-construct correlations to establish discriminant validity, thus the squared extracted variances should be greater than the shared variances. The model fit indices used in this study belong to the absolute and relative/incremental fit indices (Hair et al., 2019). Both CFA and SEM used the Maximum-Likelihood estimation method to generate model estimates in AMOS (Kline, 2023). Research hypotheses were tested based on path estimates, t-statistics (greater than + 1.96) and p values (p < 0.05). Thus, at 95% confidence interval, t-values greater + 1.96 were adjudged significantly different from zero (Hair et al., 2019).

Sample characterisation
The study obtained a response rate of 88.9%. Survey data was comprised of 61.7% female respondents and 31.3% male respondents. Young participants dominated the sample as age groups 17-21 years, 22-26 years and 27-31 years accounted for 27.7%, 32.4% and 23.1% of the total sample, respectively. Students who indicated that they "Always" and "Frequently" used e-learning were 53.3% and 40.5%, respectively. This confirmed vast adoption of e-learning systems by universities in Zimbabwe (Dzvimbo, 2020;ZIMCHE, 2020).

Assessment of the measurement model
Data was subjected to Confirmatory Factor Analysis (CFA) using AMOS 21 to assess model fit and unidimensionality of items (Anderson & Gerbin, 1988;Kline, 2023). CFA involves specification and estimation of a hypothesized model of factor structure being measured by latent constructs to account for covariances among a set of observed variables (Hair et al., 2019). Using maximum likelihood estimation, the observed variables had significant t-values and acceptable loadings ranging from 0.679 (SEVQ4) to 0.937 (COI3) (Hair et al., 2019). A poor model fit model was observed and the Modification Indices (MI) suggested covariances between system quality, information quality, service quality, instructor quality and continuance intentions to be made to reduce the Chi square (CMIN) value. These covariances improved the model fit as the normed chi-square (x 2 /df) = 3.55, RMR = 0.07, CFI = 930, IFI = 0.950, TLI = 0.942, RMSEA = 0.09.
Further inspection of the standardized residual covariance matrix resulted in deletion of INFQ5 was deleted for having absolute residual values greater than 1. The re-specified model resulted in a good fit as evidenced by an x 2 /df ratio of 2.29, RMR = 0.058, RMSEA = 0.065, CFI = 0.959, NFI = 0.930, IFI = 0.959, TLI = 0.950. Accordingly, the recommended values for a good fitting model were satisfied (Hair et al., 2019;Kline, 2023). Table 1 shows the final CFA model properties and Figure 2 illustrates the measurement model.
The composite reliability scores satisfied the minimum requirements of 0.7 (Fornell & Larcker, 1981;Hair et al., 2019), as they ranged between 0.861 and 0.950. Construct validity was examined by evaluating both convergent and discriminant validity. Average variance extracted (AVE) was used to examine the shared variance between measures of the same latent construct. The lowest AVE score was for 0.663 whilst the highest was 0.792. Therefore, there were no issues with convergent validity in this study (Hair et al., 2019;Kline, 2023).
Discriminant validity was assessed using the Fornell and Larcker (1981) method ( Table 2). The square root of the AVE of a latent construct should be greater than any of its correlations with other variables in the model. Thus, the squared extracted variances should be greater than the shared variances (Fornell & Larcker, 1981;Hair et al., 2019). The squared multiple correlations  (SMC) were all above the recommended values of 0.25 (Hair et al., 2019), as SERVQ4 had the least SMC score of 0.461.

Assessment of the structural model
The structural model was evaluated based on three criterion, namely, the model fit, causal path estimation and the predictive power of the model. Firstly, the discrepancy between the sample covariance matrix and the hypothesized model covariance matrix was examined before testing hypotheses (Anderson & Gerbin, 1988;Kline, 2023). In this study, absolute, incremental and parsimony adjusted fit indices were used to evaluate the models (Kline, 2023). The model fit for the SEM model had a good fit as x 2 /df was 2.29, RMR = 0.058, RMSEA = 0.065, NFI = 0.930, IFI = 0.959, TLI = 0.950, CFI = 0.959 (Hair et al., 2019;Kline, 2023). Figure 3 illustrates the structural model and Table 3 shows the results of hypotheses testing. Secondly, the model was examined based on the estimates, t-statistics and p-values, which provided the basis for testing research hypotheses (Table 3).

Results and discussion
To estimate parameters between the causal paths in the model, SEM was used (Hair et al., 2019;Kline, 2023). Hypotheses testing was based on a statistical criterion from the results of path estimates run in SPSS AMOS Version 21. The path estimates, t-statistics and p-values were used  to evaluate research hypotheses (Hair et al., 2019). The t-statistic is the ratio between an estimate and its standard error (Kline, 2023). At 95% confidence interval, t-values greater than +1.96 and p-values less than 0.05 were used to denote statistically significant causal relationships (Hair et al., 2019).
The first hypothesis (H 1 ) examined the relationship between system quality and e-learning continuance intentions. That association was confirmed with a positive significant effect (β = 0.426, t = 7.589, p < 0.001). This means that students develop an intention to continue learning online if the online learning system is of high quality. E-learning system was considered to be high if quality of the digital learning platform is easy to navigate, allows students to find information easily, well structured, easy to use and does not crush unnecessarily. The findings in this study are confirmed in previous empirical work. Kuadey et al. (2022) found that the quality of the system also predicts e-learning continuance using machine learning algorithms. Conversely, Alkhawaja et al. (2022) concluded that perceived high system quality influences students' willingness to continue using an e-learning system. McLean (2003, 2016) and  also confirmed that learners value processing speed, navigation capabilities, system availability, ease of use, reliability, personalization, security, log in, responsiveness of the system and user interface. Thus, the e-learning system's functional capabilities is a key factor in determining e-learning continued usage intentions.
The second hypothesis (H 2 ) supported a positive and significant relationship between instructor quality and e-learning continuance intentions (β = 0.368, t = 6.042, p < 0.001). These results mean that students enjoy learning online more if the instructor offers prompt responses to student feedback, has a positive attitude to e-learning systems, enthusiastic, communicates and interacts well with students on e-learning systems. Pham et al. (2019) also confirmed the influence of instructor quality on overall e-learning service quality, which affected student loyalty with e-learning. Students who perceive the quality of the instructor as high would like to continue using e-learning systems in their studies (Al-Adwan et al., 2021). A good instructor stimulates the learners' mind to participate actively in the e-learning sessions (Cheng, 2014;Islam et al., 2022). Students' e-learning continuance intention is driven by the motivation, competence, care, selfefficacy, empathy and inspiration that they get from their facilitators online (Islam et al., 2022;Pham et al., 2019).
The causal relationship between support service quality and e-learning continuance intentions was weak and statistically insignificant (β = 0.051, t = 0.784, p = 0.348), hence H 3 was not supported. This means that the extent to which students develop e-learning continuance intentions was not driven by the support services rendered. The findings diverge from previous findings, as previous scholars found that students enjoy e-learning when they have support services for their hardware and software services (Alkhawaja et al., 2022;Favale et al., 2020;Kuadey et al., 2022). From a Zimbabwean context, the results suggest a high propensity to have e-learning continuance post the COVID−19 pandemic is driven by system quality, information quality and instructor quality. Thus, this paper argues that if these quality factors are present, support service quality becomes a peripheral dimension, hence not pre-condition for e-learning continuance.
The fourth hypothesis (H 4 ) of the study examined the influence of information quality on students' e-learning continuance intentions. The resulting hypothesis gained empirical support, as the path estimate was positive and statistically significant (β = 0.228, t = 3.257, p < 0.001). The results are confirmatory of the key role played by organised, up to date, compatible, relevant, understandable and well-formatted information on influencing students' e-learning continuance intentions. Perceived high content quality positively influences learners' e-learning usage. The results confirmed earlier findings by Al-Adwan et al. (2021), Daultani et al. (2021) and Panigrahi et al. (2021). Poor information quality compromises the e-learning process because students would not receive the core service offering (Naveed et al., 2021), the reason why they have enrolled in their degree programmes. Thus, information quality positively affects students' e-learning continuance intentions. Lastly, the model was also assessed in terms of its explanatory power (Hair et al., 2019). The model R square was 0.728 (73% on path diagram). It therefore indicated that 72.8% of the variability in students' e-learning continuance intentions was explained by system quality, information quality, support service quality and instructor quality. The study validates the DeLone and McLean (2003) model from a Zimbabwean higher education context by accounting for a significant amount of variance in e-learning continuance.
The theoretical implications of the study inform the need to drive e-learning continuance through augmenting e-learning systems quality, improving e-learning instructor quality and revamping e-learning information quality. Among the ISSM quality dimensions (DeLone & McLean, 2003), the study disconfirmed the role of service quality as a significant antecedent of e-learning continuance. When the perceived system competence is high and e-learning systems are error-free, that promotes e-learning continuance behaviour among students. High perceived instructor and information quality are also key drivers for continued e-learning sessions. The supporting role of service quality was minimal among the targeted group; hence, service quality was an insignificant antecedent of e-learning continuance intention. More so, the integration of instructor quality to the model merits the research and invites future researchers to examine this integration in other e-learning contexts.

Conclusions on the research objectives
The main objective of the paper was to examine the impact of e-learning quality antecedents on university students' e-learning continuance intentions post the COVID−19 crisis in Zimbabwe.
Using the model which was tested, the study concludes that the e-learning quality antecedents (system quality, information quality, and instructor quality) have a direct positive impact on e-learning continuance; with a high predictive strength of 72.8%. However, the study concludes that not all e-learning quality antecedents have a direct significant association with e-learning continuance. The role of support service quality was insignificant, and thus we conclude that service quality is not a key determinant of e-learning continuance intentions among university students in Zimbabwe. The study further concludes that there is a positive influence of system quality, instructor quality and information quality on e-learning continuance intentions.

Importance of the findings for future studies
The findings provide a key impetus for future studies that will examine e-learning continuance in a Zimbabwean context. More researchers can use the study as a baseline for measuring the impact of e-learning quality on students' continuance intentions. Given that most researches focused on exploratory dimensions of e-learning in Zimbabwe, this study advances the empirical work on digital learning transformation in pre-emerging economies. Future research work may integrate the current research model with other theoretical frameworks to evaluate e-learning continuance from a variety of theoretical perspectives.

Theoretical implications
The study validated the DeLone and McLean (2003) model of information systems in a Zimbabwean context, thus the research extends the work of DeLone and McLean (2003). The study also integrated and validated instructor quality as a key antecedent of e-learning continuance intentions in Zimbabwean universities. This validates an additional construct to the DeLone and McLean (2003) model to explain information systems success in an e-learning continuance context in Zimbabwe. Instructor quality represents an important dimension that students highly regard in e-learning environments. The research provides evidence that e-learning success can be enhanced if the universities adequately support their instructors with IT resources, technical skills and job motivation. The research model explained 72.8% of the variability in e-learning continuance intentions in the higher education industry in Zimbabwe. The research advances literature based on the DeLone and McLean (2003) model and Cheng's (2014) framework that emphasised the role of instructor quality. Future research on e-learning continuance may also examine our research model to improve its applicability to other research contexts.

Practical implications
The findings of this study bring important implications to the higher education service industry in Zimbabwe. Firstly, the study enlightens university administration of the importance to evaluate student perceptions of the service of the university Martínez-Argüelles & Batalla-Busquets, 2016). Using the student-centred approach means that universities need to periodically review their quality of service (Kilburn et al., 2016;Pham et al., 2019) and this study brings a key impetus for universities to measure e-learning service quality using the framework adopted in this research. Without measuring customer perceptions of service quality, businesses fail (Makudza, 2021). This approach has been adopted by universities and much thrust has been directed to student centeredness. Hence, this study equips them with insights for ideal measurement of e-learning service quality.
More so, the study highlights the relative importance of the e-learning quality determinants which are key for student continuance with e-learning. The results confirmed that system quality, instructor quality and information quality are the key factors which determine e-learning success and hence continuance. This provides an important springboard for Zimbabwean universities when they are designing and reviewing their e-learning infrastructure. Much emphasis should be placed on system availability, security, speed, user interface, compatibility, readability, content freshness and update regularity (Islam et al., 2022;Kuadey et al., 2022;Naveed et al., 2021). More so, instructors should be motivating, confident, technical, problem solving, reliable and available. Although support service quality was insignificant, this does not confirm that universities should neglect e-learning support services such as trouble shooting, back up office assistance, albeit, taking less priority relative to system quality, information quality and instructor quality. The study hopes that universities use the findings of this research as an impetus for designing e-learning management systems that are underpinned by an objective to enhance student online learning experience.

Recommendations
The study brings important findings of use to university administration both in Zimbabwe and beyond. The results from the study inform e-learning management strategy as they highlight key core competences that universities should always review in order to provide a seamless e-learning service. In the light of these findings, the study recommended that HEIs should invest in robust e-learning management systems that consistently provide the core functionalities of an e-learning system. System quality had the highest impact on students' e-learning continuance intention (β = 0.426, p < 0.001).
Instructors should be regularly trained, motivated, resourced and supported to develop more efficacy and digital skills so that they motivate e-learning among students. Instructor quality had the second highest influence on continuance intention (β = 0.368, p < 0.001). The information uploaded on e-learning portals should regularly be checked for relevancy, freshness, density and richness among other key attributes. Results showed moderate influence of content quality on e-learning continuance intention (β = 0.228, p < 0.001). The study also recommends that HEIs should consider regularly evaluating student perceptions of educational service quality. This helps to enhance the design of a more customer centric service and be able to project and anticipate future students' behavioural intentions.

Limitations and future studies
Although the key objective of the study was achieved, the study was subject to limitations, which present avenues for future research. One of the key limitations of the study was the sampling constraint. Data was collected from three universities in Zimbabwe. To enhance the generalizability of research findings, future studies may consider sampling all universities in Zimbabwe. Another limitation is attributed to the self-reported data collection method. This method can result in skewed findings resulting from sampling and social desirability bias. Future studies may employ triangulation of data collection methods to enhance the validity of the research findings. Lastly, this research evaluated student intentions rather than actual behaviors; thus, to have a behavioral perspective of e-learning continuance post the COVID−19 crisis, future researchers may consider measuring the actual student e-learning behaviors.