International Journal of Health Care Quality Assurance Perceived service quality’s effect on patient satisfaction and behavioural compliance Bahari Mohamed Noor Azlinna Azizan
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Article information: To cite this document: Bahari Mohamed Noor Azlinna Azizan , (2015),"Perceived service quality’s effect on patient satisfaction and behavioural compliance", International Journal of Health Care Quality Assurance, Vol. 28 Iss 3 pp. 300 - 314 Permanent link to this document: http://dx.doi.org/10.1108/IJHCQA-06-2014-0074 Downloaded on: 30 January 2016, At: 13:22 (PT) References: this document contains references to 54 other documents. To copy this document: [email protected] The fulltext of this document has been downloaded 413 times since 2015*
Users who downloaded this article also downloaded: Rama Koteswara Rao Kondasani, Rajeev Kumar Panda, (2015),"Customer perceived service quality, satisfaction and loyalty in Indian private healthcare", International Journal of Health Care Quality Assurance, Vol. 28 Iss 5 pp. 452-467 http://dx.doi.org/10.1108/IJHCQA-01-2015-0008 Muslim Amin, Siti Zahora Nasharuddin, (2013),"Hospital service quality and its effects on patient satisfaction and behavioural intention", Clinical Governance: An International Journal, Vol. 18 Iss 3 pp. 238-254 http://dx.doi.org/10.1108/CGIJ-05-2012-0016 Antonio Carrizo Moreira, Pedro Miguel Silva, (2015),"The trust-commitment challenge in service quality-loyalty relationships", International Journal of Health Care Quality Assurance, Vol. 28 Iss 3 pp. 253-266 http://dx.doi.org/10.1108/IJHCQA-02-2014-0017
Access to this document was granted through an Emerald subscription provided by emeraldsrm:195450 []
For Authors If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online products and additional customer resources and services. Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
*Related content and download information correct at time of download.
The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/0952-6862.htm
IJHCQA 28,3
Perceived service quality’s effect on patient satisfaction and behavioural compliance
300 Received 30 June 2014 Revised 23 October 2014 Accepted 12 November 2014
Bahari Mohamed Faculty of Business and Management, University College Shahputra, Kuantan, Malaysia, and
Noor Azlinna Azizan
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Faculty of Industrial Management, University Malaysia Pahang, Kuantan, Malaysia Abstract Purpose – The purpose of this paper is to advance healthcare service quality research using hierarchical component models. Design/methodology/approach – This study used a quantitative approach with cross-sectional design as a survey method, combining cluster and convenience sampling and partial least square structural equation modelling (PLS-SEM) to validate the research model and test the hypotheses. Findings – The study extends health service quality literature by showing that: patient satisfaction (PS) is dominant, significant and indirect determinant of behavioural compliance (BC); perceived service quality has the strongest effect on BC via PS. Research limitations/implications – Only one hospital was evaluated. Practical implications – The study provides managers with a service quality model for conducting integrated service delivery systems analysis and design. Originality/value – Overall, the study makes a significant contribution to healthcare organizations, better health outcomes for patients and better quality of life for the community. Keywords Healthcare, Hierarchical perceived service quality, Population-based sampling Paper type Research paper
International Journal of Health Care Quality Assurance Vol. 28 No. 3, 2015 pp. 300-314 © Emerald Group Publishing Limited 0952-6862 DOI 10.1108/IJHCQA-06-2014-0074
Introduction Service quality measurement interests many researchers and professionals. Grönroos (1984) define perceived service quality (PSQ) as an evaluation process, which the consumer compares his/her service expectations with perceptions. Thus, healthcare service quality measurement is becoming an important factor and should be addressed from the patient’s perspective because patients provide valid and unique information about the care they received (Cho et al., 2004). Grönroos (1984) postulated two service quality aspects: technical, which refers to core service delivery or service outcome, including the provider competence as staff go about performing their routines. In healthcare contexts, these include doctors and nurses’ skills and clinical outcomes; and functional care, which refers to service delivery processes or the way in which the customer receives the service. In healthcare, these represent interpersonal care and the socio-psychological relationships between patient and providers. This involves interaction quality, courtesy and the affection that providers conveyed to the patients. There is few studies that develop and test comprehensive models for capturing causality between various constructs (Badri et al., 2008). The study’s main aim is to expand the growing healthcare quality research and its outcomes in a developing country – a Malaysian public hospital. The Malaysian
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
population relies on public hospitals, which provide relatively free services. Hence, the need to focus on hospital service quality has become more relevant. Measuring service quality as perceived by patients is one step to improve public hospital service quality. The study’s specific objectives were to: measure healthcare service quality using patient perceptions and a hierarchical model; test PSQ impact on patient satisfaction (PS) and compliance; and suggest service quality improvements for public hospitals. We developed an alternative method for measuring healthcare service quality in a larger framework where PSQ, PS and behavioural compliance (BC) are the central constructs in a hierarchical partial least squares (PLS) model (Figure 1). The model was based on Akter et al. (2011). The framework was developed from marketing, psychology and healthcare literature. Based on these, we identify provider characteristics and link them to PS and BC.
PSQ’s effect on patient satisfaction 301
PSQ components Parasuraman et al. (1988) through the SERVQUAL scale offered significant advances to PSQ understanding and measurement. Therefore, based on SERVQUAL, we conceived the PSQ as second-order hierarchical construct. Studies to assess PSQ have been undertaken in a healthcare context in different countries. Aagja and Garg (2010) and Chahal and Kumari (2010) in their effort to measure healthcare PSQ in India identified admission, medical and overall service, discharge, social responsibility as dimensions, physical environment, interaction and outcome as dimensions underlying the construct. Rose et al. (2004) recommends that interpersonal aspects, patient education, cost, technical aspects, outcomes, access times, facilities and social support as PSQ dimension in a Malaysia context. Others propose various healthcare PSQ dimensions, such as infrastructure, interaction, qualified staff, clinical and administrative processes, safety indicators, medical and nursing care (NC) and the service providers’ responsibilities (e.g. Chahal and Kumari, 2010; Duggirala et al., 2008). The literature shows that PSQ has different dimensions in different studies. Thus, further testing and validation is required before we accept any construct as dimensions underlying PSQ. Moreover, service quality is context specific; studies done in other contexts, cannot be generalized (Dagger et al., 2007). In this study, based on
INF PS INT
AD
MC
NC
H1 H2 PSQ
H3
BC
Figure 1. Research model and hypotheses
IJHCQA 28,3
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
302
literature, the infrastructure, interaction, administrative procedure (ADM), medical and NC constructs were proposed as second-order hierarchical PSQ components. Infrastructure The healthcare service relies on physical evidence to improve customer experience. Healthcare services are high in credence qualities as such physical evidence, which provides a cue for patients’ service quality perceptions (Ramsaran-Fowdar, 2008). Generally, in hospital, infrastructure such as physical facilities, equipment, personnel and written materials must be appear good to create positive impressions and to influence favourable patient perceptions (Andaleeb et al., 2007). Studies show that the relationship between infrastructure and PSQ is significant and positive (Dagger et al., 2007; Chahal and Kumari, 2010). Interaction The interactions that take place between customers and service personnel during service delivery often affect service quality (Brady and Cronin, 2001). Healthcare services are intangible and often require patient involvement in the treatment process. In the process, patients also need information about their health status and outcomes, because absent information may affect the healing process. This situation contributes to intimate interactions between patient and medical staff. Thus, healthcare services emphasize the interactions between patients and care providers (Hausman, 2004; Zineldin, 2006). We proposed communication, attention to patient problems, interest in solving patient problems, professionalism and understanding patient problems are the main items in establishing this construct. ADM The ADM assists core services and simultaneously adds value for a customer (Baalbaki et al., 2008). In hospital, ADM includes admission, stay and discharge processes, clinical appointments and waiting time for consultation. The ADM process is important for ensuring favourable perceptions (Atinga et al., 2011). Efficient ADM makes patients appreciate hospital services. Thus, properly organized ADM is required to make the patients feel safe and have a pleasant experience in hospital. Many studies show that the relationship between ADM and PSQ is significant and positive (Duggirala et al., 2008; Dagger et al., 2007). Medical care (MC) MC is healthcare’s technical quality, which describes the doctors’ performance, actions, activities and conduct in relation to patient care. The construct evaluates patient experience regarding medical quality (Duggirala et al., 2008). Its definition is based on diagnostic accuracy and procedures or conformance to professional specifications (Lam, 1997). It reflects the doctor’s expertise, professionalism and competency. Doctors also need soft skills such as interpersonal, communication, courtesy and caring skills to elicit patient cooperation with treatment and thus improve patient perception towards MC (Andaleeb, 1998; Hasin et al., 2001). Although patients place high priority on MC, how to evaluate the construct is generally not well understood. This service aspect is taken for granted by patients because they may not possess sufficient background knowledge to evaluate the service quality offered by a doctor (Rohini and Mahadevappa, 2006). Studies show that the relationship between MC and PSQ is significant and positive (Rose et al., 2004; Dagger et al., 2007).
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
NC Nurses are the majority employees in healthcare, the primary service provider and spend more time with patients compared to other staff (Tafreshi et al., 2007). Thus, NC is experienced and measured by patients (Wagner and Bear, 2009). Tafreshi et al. (2007, p. 320) defined the NC as “delivery of safety care based on nursing standards, which eventuates in patient satisfaction”. Researchers confirm that NC is strongly and significantly related to PSQ (Dagger et al., 2007). NC is regarded as the most important factor in patient perceptions. If the nurse is unable to fulfil this role then healthcare service quality will not be achieved. Patients satisfied with NC develop positive reactions to nurses and other staff. PS PSQ and PS are two related but distinct constructs; i.e., PSQ is a cognitive construct, while PS is an affective construct (Choi et al., 2004). According to Zineldin (2006), PS is an emotional response; it includes a collective indicator supporting satisfaction with various hospital quality aspects such as technical, functional, infrastructure, interaction and atmosphere. It reflects care quality and is frequently considered an important healthcare quality construct (Naidu, 2009). Researchers use numerous factors to represent PS, such as communication skills, staff competence, demeanour and accessibility (Eiriz and Figueiredo, 2005). BC In this study, BC refers to a specific patient behaviour during and after receiving healthcare. According to Hausman (2004), BC refers to the extent to which patients follow instructions and advice. Therefore, healthcare relies heavily on patient cooperation during service delivery to ensure effective service. Compliance with treatment is among the most important healthcare issues for ensuring effective treatment, costs control, patient safety and PS (Cárdenas-Valladolid et al., 2010). Accordingly, evidence shows that patients satisfied with healthcare comply with the treatment regime and have better outcomes (Murphy and Coster, 1997). PSQ, PS and BC In the marketing literature, several studies show that PSQ and satisfaction with services have a mixed impact on behavioural intentions. Studies confirm that service quality is an antecedent to customer satisfaction (e.g. Dabholkar et al., 2000; Dagger and Sweeney, 2007). In healthcare contexts, PSQ has a positive relationship with PS (Tucker and Adams, 2001). Studies show that PSQ is indirectly related to behavioural intentions with service satisfaction as a mediating variable (Dabholkar et al., 2000; Olorunniwo and Hsu, 2006). However, studies also found that service quality has a direct impact on behavioural intentions (Olorunniwo et al., 2006). In healthcare, researchers found that the relationship between PSQ and BC is mediated by PS; they also found PSQ directly effects BC (Papanikolaou and Ntani, 2008; Cárdenas-Valladolid et al., 2010). Thus, there is a strong link between PSQ, PS and BC in healthcare service. Therefore, we are interested in whether PSQ has a direct impact on PS and BC. We are also interested exploring PS’s mediated effect on the relationship between PSQ and BC. Developing the conceptual framework Based on the literature, we propose that PSQ has five constructs: infrastructure, interaction, ADM, medical and NC, which parsimoniously summarizes PSQ.
PSQ’s effect on patient satisfaction 303
IJHCQA 28,3
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
304
We reconceptualised PSQ as a second-order hierarchical component model (HCM) construct in accordance with Ko and Pastore (2005). The HCM has two construct layers: higher-order (PSQ) and lower-order components (five constructs). In this study, HCM was appropriate, it reduces total relationships in the structural model, thus making the model parsimonious (Hair et al., 2014). It was modelled as reflective-reflective, which means the relationships between the lower-order and higher order component and the relationships between lower order components and their associated indicators, were all reflective. We propose BC as an outcome construct because it determined overall service quality effectiveness. Many studies suggested that patient BC is influenced by PS with service quality (Drennan et al., 2011; Regnault et al., 2012). Thus, excellent service quality leads to PS and enhanced patient BC with provider instructions and advice. Therefore, successful healthcare services often depend on patient compliance with services. Hypotheses The healthcare quality literature confirmed that PSQ leads to PS and increased PS leads to BC (Papanikolaou and Ntani, 2008; Cárdenas-Valladolid et al., 2010). We associate the hierarchical PSQ construct with PS and BC. We predict that hierarchical PSQ has a significant positive impact on PS and BC. In this relationship, PS, as a mediator, influences PSQ and BC relationships. According to Baron and Kenny (1986) mediation is a situation when the predictor: has significant influence on the mediator; has a significant influence on the criterion variable; and the predictor has a significant influence on the criterion variable when mediator is missing. Thus, we posit the following hypotheses: H1. PSQ has a significant positive impact on PS. H2. PS has a significant positive impact on BC. H3. PSQ has a significant positive impact on BC. H4. PS mediates the relationship between PSQ and BC. Methodology Survey questionnaire To operationalize the constructs in our research model, previously validated scales were adapted and modified (constructs and their sources are presented in the Table I). All items were measured using a five-point Likert-type scale, ranging from 1 (strongly disagree) to 5 (strongly agree). Items were endorsed by experts in the hospital and the academic sector. Based on their suggestions, minor revisions were made to the questionnaire. We also
Table I. Constructs and their sources
Construct
Source
Perceived service quality (PSQ) components 1. Infrastructure (INF) 2. Interaction (INT) 3. Administrative procedure (ADM) 4. Medical care (MC) 5. Nursing care (NC) Patient satisfaction (PS) Behavioural compliance (BC)
Duggirala et al. (2008) and Arasli et al. (2008) Dagger and Sweeney (2007) Dagger et al. (2007) and Duggirala et al. (2008) Andaleeb and Millet (2010) Andaleeb and Millet (2010) and Dagger et al. (2007) Dagger et al. (2007) Lin and Hsieh (2011)
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
conducted a pre-test with 30 newly discharge patients to ensure that the question content, wording, sequence, layout, question difficulty, instructions and scale range were appropriate. Pre-test feedback was used to refine the questionnaire before it was ready for data collection. In total, 35 items were included in the model. Sampling and data collection We measure causal relationships between the hierarchical PSQ, PS and BC in healthcare. To obtain the most reliable healthcare quality estimates from hospital users and without population lists from which a random sample could not be drawn, we used the methods adopted by previous researchers; i.e., population-based survey, especially stage-wise area sampling (Akter et al., 2011; Andaleeb, 2008). The respondents answered the questionnaire away from the hospital; i.e., they are not in contact with service providers (doctors and nurses), so less affected by courtesy or gratitude bias (Glick, 2009; Baalbaki et al., 2008). Moreover, the population-based survey is exempted from ethics committee review (Ong et al., 2013). The population was defined as working public servants from government establishments in Kuantan, Malaysia. The sample is considered a good theoretical population because respondents come from different government establishments and comprise various age groups, different experience and different working cultural (Leong et al., 2013). The respondents were patients admitted to hospital for at least two days to accumulate enough hospital experience (Papanikolaou and Ntani, 2008). They were discharged from the hospital no more than 12 months from the data collection process and were at least 18 years old. Government establishments were clustered according to three categories; schools, government departments and statutory bodies. Each category representatives were chosen using random sampling so that each establishment had an equal chance of being selected. All eligible individuals were invited to participate in the survey and permission to conduct the survey was sought from each manager. The survey questionnaires were distributed to respondents between 24th February and 10th March, 2014. Finally, 235 completed surveys were analyzed (a 59 per cent response rate). Based on the 235 sample size (n), we estimate 0.99 post-hoc power for the research model at a 0.05 significance level (α) and a 0.3 effect size (medium effect size) by using G*Power 3.1.9.2 for windows (Faul et al., 2009). Thus, we are confident in the research model’s hypothesized relationships. Table II shows respondent demographics. Analysis and results We used SmartPLS (Ringle et al., 2005) to estimate research model parameters. We applied non parametric bootstrapping on 235 cases with 1,000 replications to obtain the standard error estimates (Navarro et al., 2011). We used the repeated indicators approach to estimate the HCM and similar items for each construct (i.e. five items) in the lower-order components to ensure the model’s operationalization (Hair et al., 2014). Common methods variance Harman’s one factor test was carried out to examine possible common method bias (Podsakoff et al., 2003). An exploratory factor analysis, using principal axis factoring, with factors extracted based on eigenvalues W 1, was conducted on all items measuring first order latent constructs. Results showed that no single factor emerged and no factor accounted for more than 50 per cent of the variance, suggesting that common-method variance was not an issue in the current study.
PSQ’s effect on patient satisfaction 305
IJHCQA 28,3
Measure
Item
Gender
Male Female Malay Chinese Indian Primary PMR/LCE SPM/MCE STPM/HSC Diploma Bachelor Post graduate 18-25 26-35 36-45 46-55 56 and above Single Married
Ethnic
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
306
Education
Age
Table II. Sample demographics
Marital status
Frequency 63 172 226 4 5 3 11 39 8 29 119 26 30 90 75 32 8 38 197
% 26.8 73.2 96.2 1.7 2.1 1.3 4.7 16.6 3.4 12.3 50.6 11.1 12.8 38.3 31.9 13.6 3.4 16.2 83.8
PLS-SEM analysis The PLS-SEM technique has two components when examining latent constructs: measurement model, which is related to indicator loadings; and structural model, which relates to path coefficient measures. Measurement model Assessing the measurement model determines how well the indicators load on the theoretically defined constructs. Thus, it can be employed for testing hypotheses and verifying the research model. Table III shows that all composite reliability (CR) values were W 0.7, above the accepted value (Hair et al., 2010) and all Cronbach’s α values exceeded the 0.7 cut-off value (Nunnally and Bernstein, 1994); ensuring adequate reliability. Table III also established model convergent validity by showing that: all item loadings were exceeded than 0.5 (Hair et al., 2010) and significant at p o 0.01; and all average variance extracted (AVEs) were above 0.5 the accepted value (Fornell and Larcker, 1981). Table IV shows the discriminant validity results based on Fornell and Larcker’s (1981) criterion. The italics diagonal AVE element for each construct was significantly larger than any correlations involving the construct. All constructs share greater variance with their own measures than with other constructs in the model, thus establishing adequate discriminant validity. Therefore, the measurement model has adequate reliability, convergent validity and discriminant validity (Hair et al., 2014). Hierarchical PSQ model We specified PSQ as a second-order hierarchical reflective construct, which comprises five first-order reflective constructs (infrastructure, interaction, ADM, MC and NC), representing 25 (5 × 5) items. Table V shows that the hierarchical construct explained variance, which was reflected in its components: infrastructure (55.5 per cent); interaction (72.6 per cent); ADM (69.8 per cent); and medical (66.0 per cent) and NC
Constructs
Items
Loadings
α
CR
AVE
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
INF
INF1 0.760 0.823 0.876 0.586 INF2 0.795 INF3 0.765 INF4 0.740 INF5 0.765 INT INT1 0.825 0.909 0.932 0.734 INT2 0.863 INT3 0.861 INT4 0.884 INT5 0.848 ADM ADM1 0.754 0.890 0.921 0.701 ADM2 0.700 ADM3 0.907 ADM4 0.889 ADM5 0.912 MC MC1 0.777 0.868 0.905 0.655 MC2 0.827 MC3 0.839 MC4 0.795 MC5 0.807 NC NC1 0.823 0.908 0.932 0.732 NC2 0.814 NC3 0.894 NC4 0.874 NC5 0.871 PS PS1 0.878 0.925 0.945 0.769 PS2 0.840 PS3 0.904 PS4 0.875 PS5 0.884 BC BC1 0.788 0.909 0.932 0.733 BC2 0.850 BC3 0.876 BC4 0.881 BC5 0.882 Notes: α, Cronbach’s α; CR, composite reliability; AVE, average variance extracted; INF, infrastructure; INT, interaction; ADM, administration procedure; MC, medical care; NC, nursing care; PS, patient satisfaction; BC, behavioural compliance
(75.1 per cent). All the path coefficients from PSQ to its components were significant at p o 0.01. Additionally, we calculated the PSQ’s CR and AVE and the results were 0.913 and 0.678, respectively, which were well-above the cut-off values. Structural model After analyzing the measurement model, the next step was establishing a structural model. The structural model test is included the path coefficients estimates for hypothesis testing. We ran PLS-SEM on the research model twice for hypothesis testing by determining path coefficients and the path significant levels. First, hypothesis testing was carried out by examining the path between PSQ to PS (H1) and PS to BC (H2). Second, PLS-SEM was run to determine the direct effect
PSQ’s effect on patient satisfaction 307
Table III. First-order constructs – psychometric properties
IJHCQA 28,3
INF
INT
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
INF 0.765 INT 0.554 0.857 ADM 0.538 0.593 MC 0.497 0.683 NC 0.593 0.643 308 PS 0.529 0.642 BC 0.364 0.530 Notes: Square root of the AVE on the istration procedure; MC, medical care; Table IV. Discriminant validity compliance
Table V. HCM – perceived service quality
2
R β p
ADM
MC
NC
PS
BC
0.837 0.583 0.809 0.704 0.585 0.856 0.686 0.620 0.706 0.877 0.432 0.590 0.547 0.582 0.856 diagonal; INF, infrastructure; INT, interaction; ADM, adminNC, nursing care; PS, patient satisfaction; BC, behavioural
Infrastructure
Interaction
Administrative procedure
Medical care
Nursing care
0.555 0.745 W0.01
0.726 0.852 W 0.01
0.698 0.836 W0.01
0.660 0.812 W 0.051
0.751 0.869 W 0.01
(without the mediating construct) between PSQ to BC (H3). We ran the bootstrapping technique with a 1,000 re-sampling to determine the path coefficient’s significance level (Navarro et al., 2011). Table VI presents the results and hypothesis testing. The findings support hypotheses H1, H2 and H3 (t-values range from 14.046 to 36.458; all povalues 0.01). Mediating effects and H4 H4, the PS mediating effect on the relationship between PSQ and BC were tested using Hair et al. (2014) procedure. The mediation test criteria were already established: first, H1, the predictor (PSQ), was significantly correlated with the mediator (PS); second, H2, the mediator (PS), significantly influenced the criterion (BC) and third, H3, the direct relation (without the mediator) between PSQ and BC was significant (Hair et al., 2014). All three constructs and all three paths (Figure 2) were tested simultaneously, which ensures superior results to other methods (Hair et al., 2014). To establish the mediating effect, the ab indirect effect (Figure 2) has to be significant. We used the non parametric bootstrapping procedure to test the indirect effect’s significance by the PLS-SEM path modelling approach (Hair et al., 2014) (Figure 2). From the analysis, path a (β ¼ 0.778, t ¼ 34.370) and path b (β ¼ 0.283, t ¼ 3.228) were significant. Introducing mediation reduces the coefficient value between PSQ and
Hypothesis
Table VI. Path coefficient and hypothesis testing
Relationship
β
t-value
Result
H1 PSQ→PS 0.778 36.458, p o0.01 Supported H2 PS→BC 0.582 14.046, p o0.01 Supported H3 PSQ→BC 0.604 14.620, p o0.01 Supported Notes: PSQ, perceived service quality; PS, patient satisfaction; BC, behavioural compliance. Significant at one tail; p W0.05(t o1.645); p o 0.05(t⩽2.33); p W0.01(t⩾2.33)
PSQ’s effect on patient satisfaction
PS 2 R = 0.602 a
b
PSQ
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Path
BC 2 R = 0.396
c
309
Path coefficients
t - statistic
Significant level
a
0.778
34.370
b
0.283
3.228
p < 0.01
c
0.383
4.402
p < 0.01
a×b
0.220
ab STDEV (ab)
Hypothesis
Relationship
β a×b
H4
Perceived Service quality
0.220
→ Patient Satisfaction → Behavioural Compliance VAF =
ab ab + c
=
= 3.218
t -value 3.218
p < 0.01
p < 0.01
Supported Yes
p < 0.01
0.220 (0.220 + 0.383)
= 0.365
Notes: One-tail significance: p > 0.05(t
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Article information: To cite this document: Bahari Mohamed Noor Azlinna Azizan , (2015),"Perceived service quality’s effect on patient satisfaction and behavioural compliance", International Journal of Health Care Quality Assurance, Vol. 28 Iss 3 pp. 300 - 314 Permanent link to this document: http://dx.doi.org/10.1108/IJHCQA-06-2014-0074 Downloaded on: 30 January 2016, At: 13:22 (PT) References: this document contains references to 54 other documents. To copy this document: [email protected] The fulltext of this document has been downloaded 413 times since 2015*
Users who downloaded this article also downloaded: Rama Koteswara Rao Kondasani, Rajeev Kumar Panda, (2015),"Customer perceived service quality, satisfaction and loyalty in Indian private healthcare", International Journal of Health Care Quality Assurance, Vol. 28 Iss 5 pp. 452-467 http://dx.doi.org/10.1108/IJHCQA-01-2015-0008 Muslim Amin, Siti Zahora Nasharuddin, (2013),"Hospital service quality and its effects on patient satisfaction and behavioural intention", Clinical Governance: An International Journal, Vol. 18 Iss 3 pp. 238-254 http://dx.doi.org/10.1108/CGIJ-05-2012-0016 Antonio Carrizo Moreira, Pedro Miguel Silva, (2015),"The trust-commitment challenge in service quality-loyalty relationships", International Journal of Health Care Quality Assurance, Vol. 28 Iss 3 pp. 253-266 http://dx.doi.org/10.1108/IJHCQA-02-2014-0017
Access to this document was granted through an Emerald subscription provided by emeraldsrm:195450 []
For Authors If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online products and additional customer resources and services. Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
*Related content and download information correct at time of download.
The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/0952-6862.htm
IJHCQA 28,3
Perceived service quality’s effect on patient satisfaction and behavioural compliance
300 Received 30 June 2014 Revised 23 October 2014 Accepted 12 November 2014
Bahari Mohamed Faculty of Business and Management, University College Shahputra, Kuantan, Malaysia, and
Noor Azlinna Azizan
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Faculty of Industrial Management, University Malaysia Pahang, Kuantan, Malaysia Abstract Purpose – The purpose of this paper is to advance healthcare service quality research using hierarchical component models. Design/methodology/approach – This study used a quantitative approach with cross-sectional design as a survey method, combining cluster and convenience sampling and partial least square structural equation modelling (PLS-SEM) to validate the research model and test the hypotheses. Findings – The study extends health service quality literature by showing that: patient satisfaction (PS) is dominant, significant and indirect determinant of behavioural compliance (BC); perceived service quality has the strongest effect on BC via PS. Research limitations/implications – Only one hospital was evaluated. Practical implications – The study provides managers with a service quality model for conducting integrated service delivery systems analysis and design. Originality/value – Overall, the study makes a significant contribution to healthcare organizations, better health outcomes for patients and better quality of life for the community. Keywords Healthcare, Hierarchical perceived service quality, Population-based sampling Paper type Research paper
International Journal of Health Care Quality Assurance Vol. 28 No. 3, 2015 pp. 300-314 © Emerald Group Publishing Limited 0952-6862 DOI 10.1108/IJHCQA-06-2014-0074
Introduction Service quality measurement interests many researchers and professionals. Grönroos (1984) define perceived service quality (PSQ) as an evaluation process, which the consumer compares his/her service expectations with perceptions. Thus, healthcare service quality measurement is becoming an important factor and should be addressed from the patient’s perspective because patients provide valid and unique information about the care they received (Cho et al., 2004). Grönroos (1984) postulated two service quality aspects: technical, which refers to core service delivery or service outcome, including the provider competence as staff go about performing their routines. In healthcare contexts, these include doctors and nurses’ skills and clinical outcomes; and functional care, which refers to service delivery processes or the way in which the customer receives the service. In healthcare, these represent interpersonal care and the socio-psychological relationships between patient and providers. This involves interaction quality, courtesy and the affection that providers conveyed to the patients. There is few studies that develop and test comprehensive models for capturing causality between various constructs (Badri et al., 2008). The study’s main aim is to expand the growing healthcare quality research and its outcomes in a developing country – a Malaysian public hospital. The Malaysian
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
population relies on public hospitals, which provide relatively free services. Hence, the need to focus on hospital service quality has become more relevant. Measuring service quality as perceived by patients is one step to improve public hospital service quality. The study’s specific objectives were to: measure healthcare service quality using patient perceptions and a hierarchical model; test PSQ impact on patient satisfaction (PS) and compliance; and suggest service quality improvements for public hospitals. We developed an alternative method for measuring healthcare service quality in a larger framework where PSQ, PS and behavioural compliance (BC) are the central constructs in a hierarchical partial least squares (PLS) model (Figure 1). The model was based on Akter et al. (2011). The framework was developed from marketing, psychology and healthcare literature. Based on these, we identify provider characteristics and link them to PS and BC.
PSQ’s effect on patient satisfaction 301
PSQ components Parasuraman et al. (1988) through the SERVQUAL scale offered significant advances to PSQ understanding and measurement. Therefore, based on SERVQUAL, we conceived the PSQ as second-order hierarchical construct. Studies to assess PSQ have been undertaken in a healthcare context in different countries. Aagja and Garg (2010) and Chahal and Kumari (2010) in their effort to measure healthcare PSQ in India identified admission, medical and overall service, discharge, social responsibility as dimensions, physical environment, interaction and outcome as dimensions underlying the construct. Rose et al. (2004) recommends that interpersonal aspects, patient education, cost, technical aspects, outcomes, access times, facilities and social support as PSQ dimension in a Malaysia context. Others propose various healthcare PSQ dimensions, such as infrastructure, interaction, qualified staff, clinical and administrative processes, safety indicators, medical and nursing care (NC) and the service providers’ responsibilities (e.g. Chahal and Kumari, 2010; Duggirala et al., 2008). The literature shows that PSQ has different dimensions in different studies. Thus, further testing and validation is required before we accept any construct as dimensions underlying PSQ. Moreover, service quality is context specific; studies done in other contexts, cannot be generalized (Dagger et al., 2007). In this study, based on
INF PS INT
AD
MC
NC
H1 H2 PSQ
H3
BC
Figure 1. Research model and hypotheses
IJHCQA 28,3
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
302
literature, the infrastructure, interaction, administrative procedure (ADM), medical and NC constructs were proposed as second-order hierarchical PSQ components. Infrastructure The healthcare service relies on physical evidence to improve customer experience. Healthcare services are high in credence qualities as such physical evidence, which provides a cue for patients’ service quality perceptions (Ramsaran-Fowdar, 2008). Generally, in hospital, infrastructure such as physical facilities, equipment, personnel and written materials must be appear good to create positive impressions and to influence favourable patient perceptions (Andaleeb et al., 2007). Studies show that the relationship between infrastructure and PSQ is significant and positive (Dagger et al., 2007; Chahal and Kumari, 2010). Interaction The interactions that take place between customers and service personnel during service delivery often affect service quality (Brady and Cronin, 2001). Healthcare services are intangible and often require patient involvement in the treatment process. In the process, patients also need information about their health status and outcomes, because absent information may affect the healing process. This situation contributes to intimate interactions between patient and medical staff. Thus, healthcare services emphasize the interactions between patients and care providers (Hausman, 2004; Zineldin, 2006). We proposed communication, attention to patient problems, interest in solving patient problems, professionalism and understanding patient problems are the main items in establishing this construct. ADM The ADM assists core services and simultaneously adds value for a customer (Baalbaki et al., 2008). In hospital, ADM includes admission, stay and discharge processes, clinical appointments and waiting time for consultation. The ADM process is important for ensuring favourable perceptions (Atinga et al., 2011). Efficient ADM makes patients appreciate hospital services. Thus, properly organized ADM is required to make the patients feel safe and have a pleasant experience in hospital. Many studies show that the relationship between ADM and PSQ is significant and positive (Duggirala et al., 2008; Dagger et al., 2007). Medical care (MC) MC is healthcare’s technical quality, which describes the doctors’ performance, actions, activities and conduct in relation to patient care. The construct evaluates patient experience regarding medical quality (Duggirala et al., 2008). Its definition is based on diagnostic accuracy and procedures or conformance to professional specifications (Lam, 1997). It reflects the doctor’s expertise, professionalism and competency. Doctors also need soft skills such as interpersonal, communication, courtesy and caring skills to elicit patient cooperation with treatment and thus improve patient perception towards MC (Andaleeb, 1998; Hasin et al., 2001). Although patients place high priority on MC, how to evaluate the construct is generally not well understood. This service aspect is taken for granted by patients because they may not possess sufficient background knowledge to evaluate the service quality offered by a doctor (Rohini and Mahadevappa, 2006). Studies show that the relationship between MC and PSQ is significant and positive (Rose et al., 2004; Dagger et al., 2007).
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
NC Nurses are the majority employees in healthcare, the primary service provider and spend more time with patients compared to other staff (Tafreshi et al., 2007). Thus, NC is experienced and measured by patients (Wagner and Bear, 2009). Tafreshi et al. (2007, p. 320) defined the NC as “delivery of safety care based on nursing standards, which eventuates in patient satisfaction”. Researchers confirm that NC is strongly and significantly related to PSQ (Dagger et al., 2007). NC is regarded as the most important factor in patient perceptions. If the nurse is unable to fulfil this role then healthcare service quality will not be achieved. Patients satisfied with NC develop positive reactions to nurses and other staff. PS PSQ and PS are two related but distinct constructs; i.e., PSQ is a cognitive construct, while PS is an affective construct (Choi et al., 2004). According to Zineldin (2006), PS is an emotional response; it includes a collective indicator supporting satisfaction with various hospital quality aspects such as technical, functional, infrastructure, interaction and atmosphere. It reflects care quality and is frequently considered an important healthcare quality construct (Naidu, 2009). Researchers use numerous factors to represent PS, such as communication skills, staff competence, demeanour and accessibility (Eiriz and Figueiredo, 2005). BC In this study, BC refers to a specific patient behaviour during and after receiving healthcare. According to Hausman (2004), BC refers to the extent to which patients follow instructions and advice. Therefore, healthcare relies heavily on patient cooperation during service delivery to ensure effective service. Compliance with treatment is among the most important healthcare issues for ensuring effective treatment, costs control, patient safety and PS (Cárdenas-Valladolid et al., 2010). Accordingly, evidence shows that patients satisfied with healthcare comply with the treatment regime and have better outcomes (Murphy and Coster, 1997). PSQ, PS and BC In the marketing literature, several studies show that PSQ and satisfaction with services have a mixed impact on behavioural intentions. Studies confirm that service quality is an antecedent to customer satisfaction (e.g. Dabholkar et al., 2000; Dagger and Sweeney, 2007). In healthcare contexts, PSQ has a positive relationship with PS (Tucker and Adams, 2001). Studies show that PSQ is indirectly related to behavioural intentions with service satisfaction as a mediating variable (Dabholkar et al., 2000; Olorunniwo and Hsu, 2006). However, studies also found that service quality has a direct impact on behavioural intentions (Olorunniwo et al., 2006). In healthcare, researchers found that the relationship between PSQ and BC is mediated by PS; they also found PSQ directly effects BC (Papanikolaou and Ntani, 2008; Cárdenas-Valladolid et al., 2010). Thus, there is a strong link between PSQ, PS and BC in healthcare service. Therefore, we are interested in whether PSQ has a direct impact on PS and BC. We are also interested exploring PS’s mediated effect on the relationship between PSQ and BC. Developing the conceptual framework Based on the literature, we propose that PSQ has five constructs: infrastructure, interaction, ADM, medical and NC, which parsimoniously summarizes PSQ.
PSQ’s effect on patient satisfaction 303
IJHCQA 28,3
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
304
We reconceptualised PSQ as a second-order hierarchical component model (HCM) construct in accordance with Ko and Pastore (2005). The HCM has two construct layers: higher-order (PSQ) and lower-order components (five constructs). In this study, HCM was appropriate, it reduces total relationships in the structural model, thus making the model parsimonious (Hair et al., 2014). It was modelled as reflective-reflective, which means the relationships between the lower-order and higher order component and the relationships between lower order components and their associated indicators, were all reflective. We propose BC as an outcome construct because it determined overall service quality effectiveness. Many studies suggested that patient BC is influenced by PS with service quality (Drennan et al., 2011; Regnault et al., 2012). Thus, excellent service quality leads to PS and enhanced patient BC with provider instructions and advice. Therefore, successful healthcare services often depend on patient compliance with services. Hypotheses The healthcare quality literature confirmed that PSQ leads to PS and increased PS leads to BC (Papanikolaou and Ntani, 2008; Cárdenas-Valladolid et al., 2010). We associate the hierarchical PSQ construct with PS and BC. We predict that hierarchical PSQ has a significant positive impact on PS and BC. In this relationship, PS, as a mediator, influences PSQ and BC relationships. According to Baron and Kenny (1986) mediation is a situation when the predictor: has significant influence on the mediator; has a significant influence on the criterion variable; and the predictor has a significant influence on the criterion variable when mediator is missing. Thus, we posit the following hypotheses: H1. PSQ has a significant positive impact on PS. H2. PS has a significant positive impact on BC. H3. PSQ has a significant positive impact on BC. H4. PS mediates the relationship between PSQ and BC. Methodology Survey questionnaire To operationalize the constructs in our research model, previously validated scales were adapted and modified (constructs and their sources are presented in the Table I). All items were measured using a five-point Likert-type scale, ranging from 1 (strongly disagree) to 5 (strongly agree). Items were endorsed by experts in the hospital and the academic sector. Based on their suggestions, minor revisions were made to the questionnaire. We also
Table I. Constructs and their sources
Construct
Source
Perceived service quality (PSQ) components 1. Infrastructure (INF) 2. Interaction (INT) 3. Administrative procedure (ADM) 4. Medical care (MC) 5. Nursing care (NC) Patient satisfaction (PS) Behavioural compliance (BC)
Duggirala et al. (2008) and Arasli et al. (2008) Dagger and Sweeney (2007) Dagger et al. (2007) and Duggirala et al. (2008) Andaleeb and Millet (2010) Andaleeb and Millet (2010) and Dagger et al. (2007) Dagger et al. (2007) Lin and Hsieh (2011)
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
conducted a pre-test with 30 newly discharge patients to ensure that the question content, wording, sequence, layout, question difficulty, instructions and scale range were appropriate. Pre-test feedback was used to refine the questionnaire before it was ready for data collection. In total, 35 items were included in the model. Sampling and data collection We measure causal relationships between the hierarchical PSQ, PS and BC in healthcare. To obtain the most reliable healthcare quality estimates from hospital users and without population lists from which a random sample could not be drawn, we used the methods adopted by previous researchers; i.e., population-based survey, especially stage-wise area sampling (Akter et al., 2011; Andaleeb, 2008). The respondents answered the questionnaire away from the hospital; i.e., they are not in contact with service providers (doctors and nurses), so less affected by courtesy or gratitude bias (Glick, 2009; Baalbaki et al., 2008). Moreover, the population-based survey is exempted from ethics committee review (Ong et al., 2013). The population was defined as working public servants from government establishments in Kuantan, Malaysia. The sample is considered a good theoretical population because respondents come from different government establishments and comprise various age groups, different experience and different working cultural (Leong et al., 2013). The respondents were patients admitted to hospital for at least two days to accumulate enough hospital experience (Papanikolaou and Ntani, 2008). They were discharged from the hospital no more than 12 months from the data collection process and were at least 18 years old. Government establishments were clustered according to three categories; schools, government departments and statutory bodies. Each category representatives were chosen using random sampling so that each establishment had an equal chance of being selected. All eligible individuals were invited to participate in the survey and permission to conduct the survey was sought from each manager. The survey questionnaires were distributed to respondents between 24th February and 10th March, 2014. Finally, 235 completed surveys were analyzed (a 59 per cent response rate). Based on the 235 sample size (n), we estimate 0.99 post-hoc power for the research model at a 0.05 significance level (α) and a 0.3 effect size (medium effect size) by using G*Power 3.1.9.2 for windows (Faul et al., 2009). Thus, we are confident in the research model’s hypothesized relationships. Table II shows respondent demographics. Analysis and results We used SmartPLS (Ringle et al., 2005) to estimate research model parameters. We applied non parametric bootstrapping on 235 cases with 1,000 replications to obtain the standard error estimates (Navarro et al., 2011). We used the repeated indicators approach to estimate the HCM and similar items for each construct (i.e. five items) in the lower-order components to ensure the model’s operationalization (Hair et al., 2014). Common methods variance Harman’s one factor test was carried out to examine possible common method bias (Podsakoff et al., 2003). An exploratory factor analysis, using principal axis factoring, with factors extracted based on eigenvalues W 1, was conducted on all items measuring first order latent constructs. Results showed that no single factor emerged and no factor accounted for more than 50 per cent of the variance, suggesting that common-method variance was not an issue in the current study.
PSQ’s effect on patient satisfaction 305
IJHCQA 28,3
Measure
Item
Gender
Male Female Malay Chinese Indian Primary PMR/LCE SPM/MCE STPM/HSC Diploma Bachelor Post graduate 18-25 26-35 36-45 46-55 56 and above Single Married
Ethnic
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
306
Education
Age
Table II. Sample demographics
Marital status
Frequency 63 172 226 4 5 3 11 39 8 29 119 26 30 90 75 32 8 38 197
% 26.8 73.2 96.2 1.7 2.1 1.3 4.7 16.6 3.4 12.3 50.6 11.1 12.8 38.3 31.9 13.6 3.4 16.2 83.8
PLS-SEM analysis The PLS-SEM technique has two components when examining latent constructs: measurement model, which is related to indicator loadings; and structural model, which relates to path coefficient measures. Measurement model Assessing the measurement model determines how well the indicators load on the theoretically defined constructs. Thus, it can be employed for testing hypotheses and verifying the research model. Table III shows that all composite reliability (CR) values were W 0.7, above the accepted value (Hair et al., 2010) and all Cronbach’s α values exceeded the 0.7 cut-off value (Nunnally and Bernstein, 1994); ensuring adequate reliability. Table III also established model convergent validity by showing that: all item loadings were exceeded than 0.5 (Hair et al., 2010) and significant at p o 0.01; and all average variance extracted (AVEs) were above 0.5 the accepted value (Fornell and Larcker, 1981). Table IV shows the discriminant validity results based on Fornell and Larcker’s (1981) criterion. The italics diagonal AVE element for each construct was significantly larger than any correlations involving the construct. All constructs share greater variance with their own measures than with other constructs in the model, thus establishing adequate discriminant validity. Therefore, the measurement model has adequate reliability, convergent validity and discriminant validity (Hair et al., 2014). Hierarchical PSQ model We specified PSQ as a second-order hierarchical reflective construct, which comprises five first-order reflective constructs (infrastructure, interaction, ADM, MC and NC), representing 25 (5 × 5) items. Table V shows that the hierarchical construct explained variance, which was reflected in its components: infrastructure (55.5 per cent); interaction (72.6 per cent); ADM (69.8 per cent); and medical (66.0 per cent) and NC
Constructs
Items
Loadings
α
CR
AVE
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
INF
INF1 0.760 0.823 0.876 0.586 INF2 0.795 INF3 0.765 INF4 0.740 INF5 0.765 INT INT1 0.825 0.909 0.932 0.734 INT2 0.863 INT3 0.861 INT4 0.884 INT5 0.848 ADM ADM1 0.754 0.890 0.921 0.701 ADM2 0.700 ADM3 0.907 ADM4 0.889 ADM5 0.912 MC MC1 0.777 0.868 0.905 0.655 MC2 0.827 MC3 0.839 MC4 0.795 MC5 0.807 NC NC1 0.823 0.908 0.932 0.732 NC2 0.814 NC3 0.894 NC4 0.874 NC5 0.871 PS PS1 0.878 0.925 0.945 0.769 PS2 0.840 PS3 0.904 PS4 0.875 PS5 0.884 BC BC1 0.788 0.909 0.932 0.733 BC2 0.850 BC3 0.876 BC4 0.881 BC5 0.882 Notes: α, Cronbach’s α; CR, composite reliability; AVE, average variance extracted; INF, infrastructure; INT, interaction; ADM, administration procedure; MC, medical care; NC, nursing care; PS, patient satisfaction; BC, behavioural compliance
(75.1 per cent). All the path coefficients from PSQ to its components were significant at p o 0.01. Additionally, we calculated the PSQ’s CR and AVE and the results were 0.913 and 0.678, respectively, which were well-above the cut-off values. Structural model After analyzing the measurement model, the next step was establishing a structural model. The structural model test is included the path coefficients estimates for hypothesis testing. We ran PLS-SEM on the research model twice for hypothesis testing by determining path coefficients and the path significant levels. First, hypothesis testing was carried out by examining the path between PSQ to PS (H1) and PS to BC (H2). Second, PLS-SEM was run to determine the direct effect
PSQ’s effect on patient satisfaction 307
Table III. First-order constructs – psychometric properties
IJHCQA 28,3
INF
INT
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
INF 0.765 INT 0.554 0.857 ADM 0.538 0.593 MC 0.497 0.683 NC 0.593 0.643 308 PS 0.529 0.642 BC 0.364 0.530 Notes: Square root of the AVE on the istration procedure; MC, medical care; Table IV. Discriminant validity compliance
Table V. HCM – perceived service quality
2
R β p
ADM
MC
NC
PS
BC
0.837 0.583 0.809 0.704 0.585 0.856 0.686 0.620 0.706 0.877 0.432 0.590 0.547 0.582 0.856 diagonal; INF, infrastructure; INT, interaction; ADM, adminNC, nursing care; PS, patient satisfaction; BC, behavioural
Infrastructure
Interaction
Administrative procedure
Medical care
Nursing care
0.555 0.745 W0.01
0.726 0.852 W 0.01
0.698 0.836 W0.01
0.660 0.812 W 0.051
0.751 0.869 W 0.01
(without the mediating construct) between PSQ to BC (H3). We ran the bootstrapping technique with a 1,000 re-sampling to determine the path coefficient’s significance level (Navarro et al., 2011). Table VI presents the results and hypothesis testing. The findings support hypotheses H1, H2 and H3 (t-values range from 14.046 to 36.458; all povalues 0.01). Mediating effects and H4 H4, the PS mediating effect on the relationship between PSQ and BC were tested using Hair et al. (2014) procedure. The mediation test criteria were already established: first, H1, the predictor (PSQ), was significantly correlated with the mediator (PS); second, H2, the mediator (PS), significantly influenced the criterion (BC) and third, H3, the direct relation (without the mediator) between PSQ and BC was significant (Hair et al., 2014). All three constructs and all three paths (Figure 2) were tested simultaneously, which ensures superior results to other methods (Hair et al., 2014). To establish the mediating effect, the ab indirect effect (Figure 2) has to be significant. We used the non parametric bootstrapping procedure to test the indirect effect’s significance by the PLS-SEM path modelling approach (Hair et al., 2014) (Figure 2). From the analysis, path a (β ¼ 0.778, t ¼ 34.370) and path b (β ¼ 0.283, t ¼ 3.228) were significant. Introducing mediation reduces the coefficient value between PSQ and
Hypothesis
Table VI. Path coefficient and hypothesis testing
Relationship
β
t-value
Result
H1 PSQ→PS 0.778 36.458, p o0.01 Supported H2 PS→BC 0.582 14.046, p o0.01 Supported H3 PSQ→BC 0.604 14.620, p o0.01 Supported Notes: PSQ, perceived service quality; PS, patient satisfaction; BC, behavioural compliance. Significant at one tail; p W0.05(t o1.645); p o 0.05(t⩽2.33); p W0.01(t⩾2.33)
PSQ’s effect on patient satisfaction
PS 2 R = 0.602 a
b
PSQ
Downloaded by Middle East Technical University At 13:22 30 January 2016 (PT)
Path
BC 2 R = 0.396
c
309
Path coefficients
t - statistic
Significant level
a
0.778
34.370
b
0.283
3.228
p < 0.01
c
0.383
4.402
p < 0.01
a×b
0.220
ab STDEV (ab)
Hypothesis
Relationship
β a×b
H4
Perceived Service quality
0.220
→ Patient Satisfaction → Behavioural Compliance VAF =
ab ab + c
=
= 3.218
t -value 3.218
p < 0.01
p < 0.01
Supported Yes
p < 0.01
0.220 (0.220 + 0.383)
= 0.365
Notes: One-tail significance: p > 0.05(t
