Trust does not only form the basis of all human interactions, but it is also one of the most complex constructs that governs behaviour (Hawlitschek, Teubner & Weinhardt 2016). In defining the concept of trust, Rousseau et al. (1998:395) argue that trust is ‘a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another’.

The multidimensional nature of trust is illustrated by the fact that the concept consists of an affective (Punyatoya 2018), a cognitive (Terres, Dos Santos & Basso 2015) and a contractual component (Krishnan, Geyskens & Steenkamp 2016). Affective trust refers to the confidence one places in another on the basis of feelings generated by the level of care and concern the other party demonstrates (Johnson & Grayson 2005). As this type of trust is embedded in emotions, it tends to deepen over time as the parties make a mutual, emotional commitment to the relationship (Erdem & Ozen 2003). Cognitive-based trust implies that individuals seek rational reasons to trust another party (Erdem & Ozen 2003). High-consequence exchanges are generally complex situations that force customers to make an extra cognitive effort to evaluate their decisions and alternatives carefully (Terres et al. 2015). Contractual trust is found in a mutual understanding between partners to adhere to a specified agreement, which means that this type of trust is strongly related to moral standards of honesty and trustworthiness (Sako 2006).

Trust is not only viewed as a central construct in healthcare context relationships, but it is also a crucial element through which health outcomes are reached (Hall et al. 2001). In terms of vulnerability, Gidman, Ward and McGregor (2012) argue that vulnerability arises when users of health care services are ill and require care in an environment of specialist knowledge.

The downward spiral in global healthcare trust levels is portrayed in Figure 1.

Compared to the global average of 63%, trust in the South African healthcare industry has decreased from 59% to 53% from 2017 to 2018 (see Figure 1). Of the 28 countries analysed, South Africa is not only in the seventh worst position, but the 6% drop in the country from 2017 to 2018 is also significantly higher than the average drop of only 2%. More specifically, it was found that trust in the pharmaceutical industry amounts to only 55%.

Healthcare services are often characterised by uncertainty, risk and interdependence (Hosseini & Behaboudi 2017). Since one of the purposes of trust is to reduce uncertainty and risk (Chiu et al. 2018), it is inferred that trust is also a key building block of successful relationships in healthcare services.

Trust is achieved by means of a central route of information processing that coincides with a client’s high level of involvement when visiting a pharmacy (Perepelkin & Zhang 2011). Therefore, firms have a direct impact on customers’ health, with regard to building of trust as an important objective (Meijboom, Visak & Brom 2006). Berry and Bendapudi (2007) add that healthcare services are highly reliant on trust as the patient has to surrender completely to the healthcare provider (seller) so that the latter can solve their problem and address their needs. From a service provider’s perspective, including healthcare, trust results in favourable outcomes such as customer loyalty, repurchase intentions and customer satisfaction (Kantsperger & Kunz 2010).

A frontline employee refers to a service worker who personally interacts with a customer in a retail or a service environment (Sirianni et al. 2009). Frontline employees are a major source of communication, linking customers to firms and, consequently, influencing customers’ intention to maintain relations with the particular firm (Guenzi & Georges 2010). These employees not only make satisfactory exchanges possible, but they also have the ability to drive customer trust (Chen & Quester 2015). Furthermore, the building of trust between customers and frontline employees has been found to be one of the most cited actions in establishing long-term customer–firm relationships (Gremler, Gwinner & Brown 2001). However, despite the importance of trust, research on the employee-related antecedents of trust remains limited, especially from the perspective of ICPs. This article proposes that five key employee related antecedents contribute to trust in the pharmaceutical industry: expertise, familiarity, communication skills, likeability and customer orientation.

Employee expertise refers to the knowledge, experience and general competencies of an employee (Crosby, Evans & Cowles 1990). Competency, which is often used interchangeably with expertise, refers to employees’ performance and the way in which they manage their tasks as well as the tactics they use to fulfil customers’ needs (Tohidinia & Haghighi 2011). Customers’ feelings of uncertainty and the perception of risk during service delivery may be reduced through employees’ knowledge, technical competence and ability to answer specific questions (Guenzi & Georges 2010). Thus, employees whom customers perceive as possessing a high level of expertise provide customers with added value, which may lead to stronger and longer lasting relational ties between them (Liu & Leach 2001). Based on the literature review, the following hypotheses were stated:

Familiarity refers to a customer’s perception that an employee recognises the customer and knows what the customer’s specific services needs are (Gremler et al. 2001). Pressey and Mathews (2000) assert that the level of contact between the customer and employee is central to establishing mutually committed relationships. Gremler et al. (2001) contend that a customer’s trust is more likely to develop when an employee and customer are well known to each other because of repeated encounters. In addition, repeated interactions between customers and employees can assist customers in the process of assessing the employee’s credibility and benevolence (Kantsperger & Kunz 2010), which are key components of trust. The following three hypotheses were therefore formulated:

Communication is defined as the formal and informal information sharing of meaningful and timely information between relevant parties (Theron & Terblanche 2010). Employees act as the primary source of communication, especially in retail market situations (Sharma et al. 1999). Based on a systematic review of the literature, Phelps and Campbell (2012) confirmed that communication is a key driver of trust in any relationship. It was therefore hypothesised that:

Likeability is described as the extent to which an employee is perceived to be friendly, courteous and pleasant (Guenzi & Georges 2010). Perceived likeability deals with personality-related factors such as the degree to which an employee has an empathetic attitude, behaves politely and shares similar tastes, preferences, values and beliefs, status, appearance, lifestyle and personality characteristics with a customer (Coulter & Coulter 2003). In addition, it was found that employee characteristics and behaviour play an important role in the development of trust and contribute to how customers perceive a firm (Ivens & Schaarschmidt 2015). Based on this literature review, the following hypotheses were formulated:

According to Kemp, Jillapalli and Becerra (2014), it is crucial for employees to demonstrate customer-oriented behaviour as it contributes to building trust. Luo, Hsu and Liu (2008) believe that customer orientation is a set of beliefs that puts the customer’s interest first, while not excluding those of all other stakeholders. Customer orientation also suggests that employees alter and customise services to suit customers’ specific needs (Coulter & Coulter 2003). Newell et al. (2011) argue that customer-oriented behaviour plays a vital role in influencing buyer perceptions of trust during interactions. Since the central focus of customer-orientated employees is the well-being of customers, the latter will be more inclined to place trust in businesses which they feel care about their well-being (Van Esterik-Plasmeijer & Van Raaij 2017). The following hypotheses were consequently stated:

The South African private pharmacy industry is primarily divided into two groups: corporate chains, which are large retail pharmacy chains under corporate ownerships, and independent community pharmacies (ICPs), which are either individually owned and/or managed or which are part of a smaller group. This study specifically focused on ICPs. Independent community pharmacies play a unique role in offering services in geographical areas where people do not necessarily have access to larger corporate pharmacy chains. The uniqueness of ICPs lies in the fact that these pharmacies can offer more than a mere commercial transaction only, by focusing on a more personal experience with their clients (Perepelkin & Zhang 2014). These authors argue that clients’ trust in a pharmacy is largely derived from their perceptions of the pharmacy’s employees per se rather than the actual products the employees sell. This notion implies that pharmacies have to redefine the roles of their employees (Smith et al. 2018).

A 40-item questionnaire was used to measure the perceptions of ICP clients. The items were mainly sourced from Guenzi and Georges (2010), Johnson and Grayson (2005), Miyamoto and Rexha (2004), Terres et al. (2015) and Tohidinia and Haghighi (2011). Five items were used to measure the five independent and the three dependent variables. All questions were assessed using a 7-point Likert scale.

Clients of a preselected ICP group that manages eight pharmacies across the Western Cape Province participated in the study. This ICP group has between 8000 and 41 000 active customer profiles per pharmacy. In order to obtain a representative number of respondents, it was decided to obtain 50 questionnaires from each pharmacy, envisaging 400 completed responses.

Structured interviews were conducted with clients waiting to be served by a pharmacist. Those customers visiting the pharmacies to purchase non-medical supplies (such as haircare products), were therefore excluded. Data collection occurred over a 4-week period.

All analyses were done using Statistica, an advanced software package originally developed by StatSoft in the USA. The reliability of the instrument was assessed by determining the Cronbach alpha coefficients for each variable, and structural equation modelling (SEM) with partial least squares (PLS) was used to assess both the measurement and the structural models.

Ethical clearance for the study was obtained from both the institution (the Research Ethics Committee at Stellenbosch University, approval number SU-HSD-000435) to which the authors are affiliated and the managing director of the ICP group participating in the study. In addition, written consent was obtained from each participant.

Reliability and validity tests were executed by employing partial least squares (PLS) in order to assess the measurement model. Firstly, four outer models were assessed by evaluating the internal consistency. Thereafter, the convergent and discriminant validity of these frameworks were also evaluated. Internal consistency signals that measure behave as expected, which proves that they are related to the same construct. Table 1 indicates the calculated composite reliability, the average variance extracted (AVE) and outer loadings for each variable.

Table 1 demonstrates that all constructs provide evidence of construct reliability as their values are all above the 0.70 threshold. In terms of convergent validity, outer loadings were inspected. The AVE values and outer loadings for Models A to D changed only marginally after each model was assessed individually. Therefore, it was decided not to report each one of these values, but only the reliability statistics for the initial full model, Model A. Convergent validity was therefore confirmed. Discriminant validity was assessed through the Heterotrait-Monotrait (HTMT) ratio of correlations, and the results appear in Table 2.

The HTMT ratio confirmed the discriminant validity of the majority of the reflective constructs as most of these ratios were below 1 (see Table 2). However, in all four models the dependent variables (affective trust, cognitive trust and contractual trust) correlated too highly with each other’s scales, implying that these constructs were not completely distinct from each other by empirical standards. The same correlation occurred between communication skills and customer orientation for Model A as shown in Table 3, indicating that these two constructs captured some phenomena represented by both. By excluding either one of, or both communication skills and customer orientation, discriminant validity for all the independent variables were established with an HTMT ratio below one.

Partial least squares was used to assess the structural model of Models A to D, which shows the relationships (paths) between the latent constructs. Three assessments were conducted, namely collinearity concerns in terms of the structural model, the determination of R-square (R²), and the significance and relevance of the structural model relationships.

Table 3 provides the collinearity statistics of the four separate models because of collinearity issues that surfaced after assessing the initial full model (Model A). This threat of multicollinearity then justified the assessment of the three additional models in order to explore whether different combinations of independent variables produced lower VIF values to report valid results.

As indicated by Table 3, all VIF values pertaining to Model A exceeded the generally accepted level of 5, with the exception of familiarity. Therefore, it may be inferred that there is a correlation between the remaining variables. Familiarity yielded satisfactory results pertaining to reliability and discriminant validity. In comparison with all other variables, familiarity could easily be distinguished by the respondents as a distinct concept, whereas the remaining variables could have been perceived as more ambiguous by some respondents, resulting in high VIF values. As explained earlier, this concern was addressed by creating three separate models (labelled as Models B, C and D). The collinearity statistics for the three models appear in Table 4.

Model B was created by excluding the variables that proved to have the highest VIF values, namely communication skills and expertise. As a result, the VIF values in Model B decreased significantly. Although likeability still produced a VIF value greater than five, it was only marginally so. Therefore, as likeability was only moderately correlated with the remaining variables, there was little cause for concern. Consequently, likeability and customer orientation were excluded from Model C. However, communication skills was reintroduced into the model and, ultimately, the two remaining independent variables produced satisfactory VIF values, as can be seen in Table 4. In addition, Model D was also created to assess the collinearity of the model if expertise were also to be reintroduced, but replaced communication skills in Model C. The outcome also proved to be satisfactory since the two remaining independent variables produced the lowest VIF values in comparison with the other three models.