Original Research

Factors influencing township consumers’ adoption of last-mile pharmaceutical delivery services

Mandlenkosi Nkondo, Mamekwa K. Kekana

Received: 12 Nov. 2025; Accepted: 21 July 2026; Published: 18 Sept. 2026

Copyright: © 2026. The Author(s). Licensee: AOSIS.
This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license (https://creativecommons.org/licenses/by/4.0/).

Abstract

Background: This study investigates factors influencing consumers’ adoption of last-mile delivery (LMD) services for pharmaceutical products in South African townships, contributing to understanding technology acceptance in emerging markets.

Aim: Grounded in the Technology Acceptance Model (TAM), the study examines the effects of behavioural intention (BI), attitude (ATT), perceived ease of use (PEOU) and perceived usefulness (PU) on LMD adoption, and tests the mediating role of PU in the relationship between PEOU and use behaviour (UB).

Setting: A cross-sectional quantitative survey was conducted among 127 customers of a private retail pharmacy in Nkowankowa Township, Limpopo, South Africa.

Methods: Structured questionnaires were administered through convenience sampling, with researcher-assisted translation into Xitsonga or Sepedi where required. Constructs were measured using validated multi-item Likert scales. Confirmatory factor analysis assessed the measurement model, while structural equation modelling tested six hypotheses in IBM SPSS 29.0 and AMOS 29.0. Mediation was assessed using bias-corrected bootstrapping.

Results: Consumers reported positive ATT towards LMD pharmaceutical services, and BI strongly predicted UB. PEOU and PU significantly influenced ATT, while PEOU strongly enhanced PU. PU primarily operated through indirect pathways rather than directly influencing UB, supporting the mediated TAM structure.

Conclusion: Township consumers are receptive to LMD pharmaceutical services. Adoption is driven mainly through cognitive-affective pathways rather than usefulness alone, indicating that interventions should strengthen the pathway from ease of use through attitude formation to intention.

Contribution: As the first South African study of LMD adoption factors for pharmaceuticals in a township setting, this study offers novel insights into technology acceptance and emerging-market healthcare delivery.

Keywords: emerging markets; technology acceptance model; last-mile delivery; pharmaceutical products; consumer behaviour; township; e-pharmacy; South Africa.

Introduction

The coronavirus disease 2019 (COVID-19) pandemic disrupted health systems worldwide. It intensified pre-existing inequalities in emerging economies. In resource-constrained countries such as South Africa, it severely affected access to pharmaceutical products (Tirivangani et al. 2021). The South African pharmaceutical industry faced acute challenges during this period. These challenges further limited the availability of quality, cost-effective medicines in a country already burdened by a high prevalence of communicable and non-communicable diseases (Magadzire et al. 2014). The South African health system is also fragmented, serving approximately 70% of the population through under-resourced public facilities (Magadzire et al. 2014).

The pandemic’s effect on pharmaceutical systems was particularly severe. Access to essential medicines was restricted across both public and private health facilities (Guerin, Singh-Phulgenda & Strub-Wourgaft 2020). Lockdown measures during the early phases of COVID-19 imposed substantial restrictions on domestic and cross-border movement. These restrictions disrupted pharmaceutical supply chains and their delivery mechanisms (Tirivangani et al. 2021). Consequently, access to pharmaceutical products, including prescription medicines, vaccines and medical devices, became severely constrained as urban logistics networks faced unprecedented disruptions throughout 2020–2021 (Agrawal et al. 2022; Guerin et al. 2020).

These challenges accelerated the growth of online shopping and e-commerce pharmacies. Digital platforms became important avenues for medication access (Adebo, Aladelusi & Mohammed 2025). E-pharmacy platforms, defined as web-based services that deliver pharmaceutical products directly to consumers, have generated considerable interest in last-mile delivery (LMD) solutions that enhance customer convenience and accessibility (Seghezzi, Siragusa & Mangiaracina 2022). Last-mile delivery, representing the final segment of the delivery process from the distribution hub to the recipient’s location, has been predominantly implemented in urban areas, where the technological infrastructure supports effective operation (Seghezzi et al. 2022).

For pharmaceutical products, LMD encompasses order confirmation and assembly, fulfilled through home delivery, collection points or alternative distribution mechanisms. Following the decline in traditional face-to-face medication collection at pharmacies during the pandemic, LMD service utilisation increased rapidly, particularly in urban contexts (Seghezzi et al. 2022). However, rural and township consumers, while representing the demographic segment most difficult to reach through conventional delivery models, stand to benefit considerably from the convenience and efficiency of LMD pharmacy services, despite facing substantial barriers to adoption (Bvuma & Marnewick 2020; Gallant, Amadi-Echendu & Esterhuyzen 2025).

Problem statement

Adopting LMD services for pharmaceutical delivery within South African townships presents distinctive challenges. These challenges stem from socioeconomic inequalities and constrained digital infrastructure. They include restricted smartphone access, inadequate information technology (IT) infrastructure, unreliable Internet connectivity and cybersecurity concerns (Adebo et al. 2025). The barriers are compounded by broader systemic issues. These include persistent power outages caused by ageing and overloaded transmission infrastructure, cable theft, illegal connections and inadequate maintenance. These barriers, coupled with limited support and training in information and communications technology (ICT), significantly hinder the uptake of LMD services for pharmaceutical products (Bvuma & Marnewick 2020). Township transport networks present further difficulties. They are characterised not only by poorly mapped road systems but also by untarred roads, potholes and informal-settlement layouts that lack formal street addressing and consistent signage (Bvuma & Marnewick 2020; Magadzire et al. 2014; Mkansi, De Leeuw & Amosun 2020).

Patient preferences and cultural factors also shape uptake. These include stigma affecting healthcare-service choices, particularly for human immunodeficiency virus (HIV)-related services (Kuo et al. 2022), and traditional health-seeking behaviours that may conflict with innovative delivery models (Magadzire et al. 2014). For township-based private pharmacies, the financial costs and technical expertise required for LMD adoption impose additional burdens on small and medium enterprises (SMEs) and on consumers (Gallant et al. 2025; Mash et al. 2022).

Despite these challenges, the accelerated growth of e-commerce during the COVID-19 pandemic highlights the potential of LMD services. Last-mile delivery can enhance SME sales and generate positive financial outcomes within an industry characterised by narrow profit margins, particularly for township-based private pharmacies (Gallant et al. 2025). The role of LMD in healthcare supply chains continues to grow, especially in pharmacy contexts where home medication delivery is increasing (Mash et al. 2022).

Research problem and objectives

Contemporary research has shown growing interest in innovative pharmacy-to-patient delivery modalities. These include collection services, bicycle couriers, electric vehicles and drone technologies, with potential benefits such as reduced carbon emissions and improved demand satisfaction (Toraman, Bayirli & Ramadani 2024). Yet, despite optimism about LMD implementation in the pharmaceutical industry, critical factors remain to be addressed, particularly consumer willingness and preferences (Neto & Vieira 2023). Consumers are a critical stakeholder group in e-commerce. Their behavioural patterns concerning LMD adoption are decisive for successful implementation (Toraman et al. 2024).

Existing literature has focused predominantly on optimising self-service networks and analysing the benefits of various LMD models (Mangiaracina et al. 2019). Fewer studies have evaluated which behavioural factors influence consumer intentions to adopt LMD services for pharmaceutical delivery (Adebo et al. 2025). Most existing studies were conducted in developed countries. Research on adoption intentions from the consumer perspective in developing countries remains relatively scarce (Neto & Vieira 2023).

Within the African literature, few studies have developed theoretical models to explain consumer behaviour. Examples include applications of the Unified Theory of Acceptance and Use of Technology (UTAUT) in Nigeria (Adebo et al. 2025) and extended applications of the Technology Acceptance Model (TAM) in South Africa (Nyanga & Eresia-Eke 2025; Phorah & Motsi 2025). South African research has identified high patient interest in last-mile pharmaceutical delivery. A Cape Town study reported 80.2% service interest and 84.5% willingness to use mobile applications for medicine ordering (Adams et al. 2021).

However, most South African studies in rural areas have found that consumer adoption of pharmaceutical LMD services is constrained by poor physical infrastructure and limited public transport networks (Magadzire et al. 2014), distance from healthcare facilities (Adams et al. 2021), delivery costs (Gallant et al. 2025), limited Internet connectivity and cybersecurity concerns (Adebo et al. 2025). Regarding behavioural factors, social influence emerged as a key determinant of e-pharmacy adoption among young Nigerian consumers (Adebo et al. 2025).

Despite these contributions, the factors influencing consumer behaviour and preferences remain insufficiently understood, with limited research examining rural and township contexts (Adams et al. 2021; Magadzire et al. 2014; Neto & Vieira 2023). To the best of our knowledge, no study has comprehensively investigated consumer adoption of LMD services for pharmaceutical products in South African township settings. Furthermore, private pharmacies, which typically operate as SMEs, remain an understudied context for differentiated pharmaceutical product delivery, including HIV services that can mitigate barriers to clinic-based care such as stigma and extended waiting times (Kuo et al. 2022).

This study therefore investigates the consumer behavioural factors that influence the adoption of LMD services for pharmaceutical products in a South African township context, using the TAM framework. The study focuses on Nkowankowa, a township near Tzaneen in Limpopo province. The specific objectives are:

• To examine the relationships among perceived ease of use, perceived usefulness, attitude, behavioural intention and use behaviour in the pharmaceutical LMD context.
• To test whether perceived usefulness mediates the effect of perceived ease of use on use behaviour.

Contribution and study outline

This study contributes to the growing body of research on LMD services. By examining the factors that shape township consumers’ adoption of pharmaceutical LMD, it addresses a context that remains underrepresented relative to urban-focused studies (Seghezzi et al. 2022). It also applies the TAM to a pharmaceutical LMD setting and tests perceived usefulness (PU) as a mediator between perceived ease of use (PEOU) and use behaviour (UB). As an early South African study of LMD adoption for pharmaceuticals in a township setting, it contributes evidence to both the technology-acceptance literature and emerging-market healthcare-delivery research (Gallant et al. 2025).

The remainder of this article is structured as follows. The Literature Review examines prior scholarship on LMD services, with emphasis on township contexts. This is followed by the Theoretical Framework and Hypothesis Development, which outlines the theoretical foundation and develops the six hypotheses. The Methodology then describes the research approach adopted, before the Results section reports the study’s results. The Discussion addresses these findings and presents their theoretical and managerial implications, together with the study’s limitations and avenues for future research. The article closes with the Conclusion.

Literature review

Advantages of last-mile delivery services

Last-mile delivery services have gained increased exploration and significance owing to recent developments in modern logistics and e-commerce systems. The adoption of LMD services demonstrates considerable potential to drive greater product demand for SMEs (Mkansi et al. 2020). These services offer substantial advantages to business owners by creating avenues for revenue enhancement through additional customer acquisition and increased sales volumes without requiring significant investments in physical infrastructure. Last-mile delivery-supporting technology enables operational process automation, enhancing quality control and service delivery for customers while providing businesses with opportunities for more effective customer engagement and valuable feedback for service improvement, thereby fostering customer loyalty (Mangiaracina et al. 2019).

Using LMD services for home medication delivery can alleviate the burden of long-distance travel for vulnerable customers. It can also reduce the stigma associated with conditions such as HIV/acquired immunodeficiency syndrome, because affected patients can avoid physically visiting pharmacies (Kuo et al. 2022). This capability is particularly valuable in South African township contexts, where transportation costs and social stigma are significant barriers to healthcare access (Bvuma & Marnewick 2020).

Challenges of last-mile delivery services

Technological and infrastructure barriers

The technological landscape in South African townships reflects broader systemic inequalities. It is characterised by inadequate smart device access, unreliable Internet connectivity and persistent power disruptions arising from ageing transmission infrastructure, cable theft, illegal connections and inadequate maintenance (Bvuma & Marnewick 2020). These digital divides are compounded by insufficient ICT infrastructure, which fundamentally undermines Internet connectivity as a prerequisite for effective LMD service deployment. The absence of comprehensive ICT support and training programmes exacerbates adoption challenges, creating barriers to meaningful technology integration within township communities (Bvuma & Marnewick 2020).

Logistical and operational challenges

The physical infrastructure of township environments presents additional operational complexities for LMD providers. Transport networks are generally in a poor condition, characterised by ageing roads with poor road signage, faded road markings and potholes, which make it difficult to access delivery addresses (Mkansi et al. 2020). Informal settlement layouts that lack formal street addresses, coupled with inadequately mapped road systems, produce navigational inaccuracies that complicate the efficient delivery of goods from retail outlets to residential addresses (Mkansi et al. 2020). These logistical constraints directly impact service reliability and cost-effectiveness, undermining the viability of LMD operations in these communities.

Small and medium enterprises implementation barriers

For SMEs, responsibility for LMD implementation typically rests with owners or managers who may lack requisite technological competencies and financial resources (Gallant et al. 2025). The intersection of technology costs, ICT skill acquisition and educational prerequisites creates barriers to availability. These constraints are particularly pronounced in township SME contexts, where the digital literacy and financial investment required for effective LMD are often limited (Gallant et al. 2025).

Regulatory and compliance considerations

The regulatory environment presents additional complexity through compliance requirements under the Pharmacy Act 53 of 1974, Medicines and Related Substances Act 101 of 1965, and Protection of Personal Information Act, 2013. These legislative frameworks require the secure handling of electronic personal information and compliance with industry-specific regulations, creating additional operational burdens for LMD service providers in the pharmacy industry.

Emerging opportunities for last-mile delivery adoption in township pharmacies

Despite substantial challenges, the COVID-19 pandemic catalysed greater acceptance of e-commerce, creating new opportunities for LMD adoption. Potential benefits include enhanced product demand generation, increased sales volume, process automation for quality control and improved service delivery capabilities (Mkansi et al. 2020). Last-mile delivery services also offer SMEs opportunities for more effective customer engagement and enhanced customer loyalty development (Mangiaracina et al. 2019). This suggests that overcoming implementation barriers could yield significant commercial and social benefits.

Theoretical framework

Technology acceptance model

The TAM is a widely used theoretical framework for understanding the factors that influence the adoption of new technology (Davis 1989). The TAM posits that PU, ease of use and user attitudes determine the likelihood of technology adoption. The model has been successfully applied in various contexts, including healthcare IT systems (Hu et al. 1999), healthcare wearables (Zhang et al. 2017), telemedicine (Phorah & Motsi 2025) and financial-technology adoption among South African SMEs (Nyanga & Eresia-Eke 2025). In this study, TAM is used to explore the factors influencing township consumers’ adoption of LMD services for pharmaceutical products, incorporating external variables that shape users’ beliefs and attitudes (Burton-Jones & Hubona 2006).

External variables, often termed antecedents, represent psychological factors that influence PEOU and PU, with their effects varying depending on technology characteristics. For example, system characteristics, user involvement and training have been identified as key external variables in workplace computer systems (Venkatesh & Davis 1996), whereas self-efficacy, subjective norms and enjoyment have been shown to affect e-portfolio adoption (Abdullah, Ward & Ahmed 2016). In healthcare wearables, PU is shaped by factors such as credibility and health beliefs (Zhang et al. 2017). In telemedicine, PEOU and perceived benefits for both patients and healthcare providers have led to increased adoption (Anderson et al. 2022).

Three considerations motivate the choice of TAM as the model of choice in this study. Firstly, TAM is a robust information-technology-specific theory that provides an adequate framework for explaining and predicting technology adoption across diverse populations in different cultural and knowledge contexts. Secondly, TAM has a solid theoretical foundation and well-established, validated psychometric measurement scales, making it operationally appealing (Szajna 1996). Thirdly, TAM has garnered consistent empirical support for its overall explanatory power, establishing itself as a leading model for understanding user acceptance and technology adoption (Szajna 1996). The five core TAM constructs operationalised in this study, namely PU, PEOU, attitude (ATT), behavioural intention (BI) and UB, are defined, in turn, from the perspective of township consumers.

Perceived usefulness

In this study, PU is defined as the extent to which township consumers believe that using LMD services for pharmaceutical products would improve the convenience, timeliness and overall value of obtaining medicines compared with face-to-face collection at the pharmacy (Davis 1989). Within TAM, PU is a cognitive evaluation shaped by beliefs about the productivity and effectiveness of the system, and influenced by external variables (Burton-Jones & Hubona 2006).

Perceived ease of use

Perceived ease of use is defined as the degree to which township consumers believe that using LMD services for pharmaceutical products is free of cognitive and physical effort (Davis 1989). Perceived ease of use is theorised to influence technology adoption directly through attitude formation, and indirectly by shaping PU (Karahanna & Straub 1999). In this context, it reflects the anticipated simplicity of ordering medicines, navigating a delivery platform, tracking a delivery and arranging payment.

Attitude

Attitude is defined as township consumers’ overall favourable or unfavourable evaluation of using LMD services for pharmaceutical products, based on their perceptions of usefulness and ease of use (Fishbein & Ajzen 1975). Whereas perceptions are cognitive (beliefs about what a service is or does), ATT is evaluative (a judgement about whether using the service is desirable). Attitude therefore captures the consumer’s affective orientation towards using LMD for medicines and is distinct from the cognitive perceptions (PU and PEOU) that inform it. Cullen et al. (2025) found that ATT significantly predicts BI to use telemedicine in South Africa.

Behavioural intention

In this study, the focal behaviour is the use of LMD services for pharmaceutical products. Behavioural intention is accordingly defined as township consumers’ intention to use LMD services for pharmaceutical products in the foreseeable future. Within TAM, BI is influenced primarily by PU and ATT, and it serves as the immediate antecedent of actual UB (Davis 1989).

Use behaviour

Use behaviour refers to township consumers’ ongoing engagement with LMD services after initial adoption (Venkatesh et al. 2003). In this study, UB is operationalised through consumers’ self-reported confidence, proficiency and continued engagement with LMD services for pharmaceutical products in the Nkowankowa setting.

Conceptual model and hypothesis development

Building on the construct definitions above, the study proposes a conceptual model grounded in Davis’s (1989) TAM. The model holds that UB is determined by BI, which is shaped by ATT. Attitude is, in turn, influenced by two cognitive perceptions: PU and PEOU. The model also examines the role of PU in mediating the effect of PEOU through to UB. The six hypotheses derived from these relationships are motivated below, drawing on established theory and empirical evidence from technology-acceptance and LMD research.

Figure 1 presents the conceptual model: H1–H5 are direct paths, and H6 is the serial-mediation path PEOU → PU → ATT → BI → UB. The six hypotheses form an integrated model that traces the cognitive and affective processes underlying pharmaceutical LMD adoption. Use behaviour (H1) is driven by behavioural intention, which is shaped by attitude (H2). Attitude is influenced by both perceived ease of use (H3) and perceived usefulness (H5). Perceived ease of use additionally shapes perceived usefulness (H4), which serves as the mediating mechanism linking ease of use to use behaviour (H6).

FIGURE 1: Conceptual framework and hypotheses development.

Behavioural intention and use behaviour (Hypothesis 1)

The link between intention and behaviour is among the most established findings in technology-acceptance research. It is rooted in the Theory of Reasoned Action (Fishbein & Ajzen 1975) and integrated into TAM (Davis 1989). Behavioural intention captures the motivational factors that influence behaviour, including the effort a person is willing to exert to perform it (Ajzen 1991). In the context of pharmaceutical LMD services, BI captures customers’ readiness to use delivery services for pharmaceutical products, while UB reflects the actual observable adoption of these services, such as placing orders, receiving deliveries and establishing recurring delivery schedules.

Recent South African evidence supports the BI link. Cullen et al. (2025) found that several factors, including performance expectancy, facilitating conditions, social media technology and ATT, have a significant positive relationship with BI to adopt telemedicine by healthcare practitioners. Gila, Dzomonda and Neneh (2026) found that entrepreneurial intention positively predicted actual behaviour, and Toraman et al. (2024) framed usage intention as a direct precursor to the active use of new delivery technologies. Across healthcare settings, empirical studies consistently demonstrate that intention significantly predicts actual technology use. For instance, in their study of telemedicine adoption, Hu et al. (1999) found that BI was the strongest predictor of actual system usage among physicians. More recently, Lee and Lee (2020) found that BI predicted UB in healthcare wearable devices.

For pharmaceutical delivery specifically, the intention–behaviour relationship is particularly important because medication use and adherence depend on the successful translation of delivery intentions into actual service utilisation. Unlike general e-commerce deliveries, pharmaceutical LMD involves health-critical products where the consequences of non-adoption (i.e. failing to obtain medications) are severe, thereby strengthening the intention–behaviour relationship. On this basis, this study hypothesises that:

H1: Behavioural intention to use LMD services for pharmaceutical products positively influences the use behaviour of LMD services for pharmaceutical products.

Attitude and behavioural intention (Hypothesis 2)

Attitude towards using a technology represents an individual’s positive or negative evaluation of the target behaviour (Fishbein & Ajzen 1975). Within TAM, ATT is a key determinant of BI, transmitting the effects of PU and PEOU on intention (Davis 1989). The expectancy-value logic of ATT formation holds that when consumers anticipate favourable outcomes from using technology, their ATT becomes more positive, which strengthens their BI. In pharmaceutical services, ATT toward LMD encompasses patients’ overall appraisal of using delivery services for their medications, incorporating considerations of convenience, reliability, trust in the delivery process and perceived safety of handling temperature-sensitive products.

Empirical evidence supports the relationship in e-pharmacy adoption in Nigeria (Adebo et al. 2025). Toraman et al. (2024) reported a strong effect of ATT on usage intention in an LMD setting. Murire et al. (2026), in a South African township sample, found that a favourable personal ATT significantly predicted entrepreneurial intention among unemployed youth. Thus, this study hypothesises that:

H2: Attitude towards LMD services positively influences the behavioural intention to use LMD services for pharmaceutical products.

Perceived ease of use and attitude (Hypothesis 3)

Perceived ease of use is the degree to which a person believes that using a system would be free of effort (Davis 1989). The cognitive–affective mechanism linking PEOU to ATT is that anticipated low effort reduces anxiety and fosters a positive evaluation of the technology (Venkatesh & Davis 1996). In pharmaceutical LMD, this concerns consumers’ assessments of how simple it is to order medicines, navigate a delivery platform, arrange payment and handle temperature-sensitive packages upon arrival.

In a South African telemedicine study, Phorah and Motsi (2025) found that PEOU significantly influenced attitudes towards the technology. For township consumers, who may be less familiar with digital service interfaces, ease of use is plausibly central to ATT formation. Extending this logic to patient-facing pharmaceutical delivery services, ease of interaction with the delivery platform should enhance users’ overall affective evaluation. Hence, this study hypothesises that:

H3: Perceived ease of use of LMD services positively influences attitude towards using LMD services for pharmaceutical products.

Perceived ease of use and perceived usefulness (Hypothesis 4)

A central tenet of TAM is that PEOU influences PU (Davis 1989). The underlying reasoning is that, when a system is effortless to operate, consumers infer that it enables them to accomplish tasks more efficiently (Venkatesh & Davis 1996). In the context of pharmaceutical LMD services, easy-to-use delivery platforms enable patients to reorder medications quickly, set up automatic refills and manage their delivery preferences with ease. The resulting time savings and reduced mental effort contribute to the perception that the service is genuinely useful for managing health conditions.

King and He’s (2006) meta-analysis confirmed a robust PEOU–PU relationship across numerous TAM studies. Kalayou, Endehabtu and Tilahun (2020) found that PEOU has a significant influence on PU on the sustainable adoption of e-Health systems by healthcare professionals in resource-limited settings. Nyanga and Eresia-Eke (2025) found that both PEOU and PU significantly drove financial technology adoption among South African SMEs. It should be noted that the strength of the PEOU–PU path varies across contexts. Some studies report a weak or non-significant direct effect (Phorah & Motsi 2025; Toraman et al. 2024), which makes its examination in the township pharmaceutical setting worthwhile. Applied to pharmaceutical delivery, this study hypothesises that:

H4: Perceived ease of use of LMD services positively influences the perceived usefulness of LMD services for pharmaceutical products.

Perceived usefulness and attitude (Hypothesis 5)

The TAM holds that PU shapes ATT because users develop favourable evaluations of technologies they perceive as productive and value-enhancing (Davis 1989). For pharmaceutical LMD services, usefulness encompasses patients’ beliefs that delivery services improve medication accessibility, ensure timely refills, reduce transportation burdens, maintain medication integrity through proper handling and ultimately support better health outcomes.

The PU–ATT relationship has been validated in healthcare and delivery contexts. Phorah and Motsi (2025) found that PU significantly influenced ATT in a South African telemedicine setting. Kalayou et al. (2020) confirmed positive PU-to-ATT links in e-health systems. Hu et al. (1999) found that PU was the strongest determinant of ATT toward telemedicine among physicians. Toraman et al. (2024) reported a significant effect of PU on ATT in the use of electric vehicles in LMD for fast-moving consumer goods. Consequently, this study hypothesises that:

H5: Perceived usefulness of LMD services positively influences attitude towards using LMD services for pharmaceutical products.

The mediating role of perceived usefulness (Hypothesis 6)

Beyond its direct effect on ATT, PU is theorised to transmit the influence of PEOU to UB. The mediating pathway reflects a cognitive process in which consumers first assess the effort required to use a technology (PEOU), then infer its usefulness (PU) and finally translate these beliefs into intentions (BI) and behaviour (UB) (Davis 1989; Venkatesh & Davis 1996). In pharmaceutical LMD adoption, this implies that the ease of navigating a delivery platform does not directly drive UB; rather, ease of use enhances PU, which in turn promotes ATT, intention and ultimately use.

King and He’s (2006) meta-analysis confirms that PU frequently mediates the effect of PEOU on downstream outcomes. Alsyouf et al. (2023) showed that, among Saudi patients, the effect of PEOU of a personal health record system on its use was mediated by PU. In emerging-market healthcare contexts, where consumers may have limited prior exposure to similar services, sequential mediation through PU, ATT and BI is theoretically plausible. Therefore, this study hypothesises that:

H6: Perceived usefulness mediates the relationship between the perceived ease of use of LMD services and the use behaviour of LMD services for pharmaceutical products.

Research methodology and design

Research design

This study utilised a quantitative cross-sectional approach grounded in positivist philosophy and employed deductive reasoning. The study used a cross-sectional survey to collect data from a convenience sample of 127 customers at a community retail pharmacy in Nkowankowa, a township in South Africa’s Limpopo province. According to Masekela et al. (2025), Nkowankowa is one of five proclaimed towns in the Greater Tzaneen Municipality. The township covers an area of 16.23 km² and has an estimated population of approximately 110 963 people. Seventeen languages are spoken in the Mopani District Municipality, with Sepedi being the most widely spoken, followed by Xitsonga, Afrikaans and English.

At the time of data collection, three licensed pharmacies operated within Nkowankowa, with only one, the participating community pharmacy, offering formal LMD services to customers’ homes. This delivery service was necessitated by the township’s limited public transportation infrastructure and the challenges residents face in accessing healthcare services, particularly those with chronic conditions requiring regular medication (Mash et al. 2022). Delivery orders were typically placed telephonically or via WhatsApp messaging, with a small number of customers submitting requests in person during pharmacy visits. The pharmacy employed two delivery personnel who used a single motorcycle for medication deliveries, occasionally supplemented by a small passenger car for larger orders. This mode of transportation aligns with documented practices in South African township contexts, where motorcycles and bicycles are commonly used for last-mile medication delivery because of their manoeuvrability in densely populated areas and cost-effectiveness. The typical delivery time ranged between 2 h and 4 h for orders placed before 14:00, with next-day delivery for afternoon orders, reflecting operational constraints and the geographic spread of customers across the township.

Data collection

The study used a cross-sectional survey to collect data from a convenience sample of 127 existing customers visiting a community retail pharmacy in Nkowankowa during the 6-week data-collection window from May 2024 to June 2024. The questionnaire was administered by the researcher in the pharmacy. Although it was designed for self-completion, the researcher remained available throughout to provide verbal clarification and translation into Xitsonga or Sepedi for respondents who preferred not to read or respond in English. Survey completion required less than 20 min per respondent.

The questionnaire comprised two main sections: demographic items and five-point Likert-scale items ranging from 1 (Strongly Disagree) to 5 (Strongly Agree) measuring the five TAM constructs. Table 1 summarises the Likert-scale items used to measure each construct. The items measuring PU and PEOU were adapted from Davis (1989) and Venkatesh and Davis (2000), the items for ATT and BI were adapted from Ajzen and Fishbein (1980) as operationalised by Hwang, Kim and Lee (2021), and the items for UB were adapted from Lee and Lee (2020).

The sample size of 127 is acknowledged as a study limitation. Conventional ‘10 cases per indicator’ guidance (Hair et al. 2018) suggests approximately 210 cases would be desirable for the 21 indicators in the final measurement model. However, recent structural equation modelling (SEM) guidance (Memon et al. 2020) indicates that the sample size between 100 and 200 can yield acceptable parameter stability for models with strong loadings (typically ≥ 0.70), as in the present study. A post-hoc Monte Carlo power assessment confirmed adequate power (> 0.80) for detecting the observed standardised effects at α = 0.05.

Data analysis

Data were analysed in IBM Statistical Package for the Social Sciences (SPSS) Statistics version 29.0 and IBM SPSS Analysis of Moment Structures (Amos) version 29.0 (IBM Corporation, Armonk, New York, United States) in the following sequence. Firstly, data were screened in Microsoft Excel for missing values and outliers; missing data accounted for less than 2% of responses and were missing completely at random (Little’s Missing Completely at Random (MCAR) test, p > 0.05); listwise deletion was therefore used. Secondly, descriptive statistics and reliability indices were calculated for each composite score using Cronbach’s alpha coefficients, with values exceeding 0.70 accepted as indicative of strong internal consistency (Hair et al. 2018). Thirdly, confirmatory factor analysis (CFA) with maximum likelihood (ML) estimation was used to evaluate the measurement model. Mardia’s coefficient of multivariate kurtosis was within acceptable bounds for ML estimation. Fourthly, the structural model was estimated to test H1 to H5 with standardised path coefficients (β) and the squared multiple correlation (R²) reported for each endogenous construct. Fifthly, the serial mediation hypothesised in H6 was tested using bias-corrected percentile bootstrapping with 5000 resamples; 95% confidence intervals not containing zero were interpreted as supporting mediation (Preacher & Hayes 2008).

Ethical considerations

Respondents were informed that participation was voluntary and that no personal information was required for the survey. All responses were kept confidential to ensure the participants’ anonymity. Additionally, the pharmacy’s consent was obtained for the distribution of the questionnaires. Ethical approval was granted by the Johannesburg Business School Research Ethics Committee (ethical clearance number: JBSREC202424).

Results

Sample demographics

Table 2 presents the demographic profile of the 127 respondents. The sample was predominantly female (56.7%) and broadly distributed across age groups, with the 25–34 cohort the largest (30.7%). Xitsonga (67.7%) and Sepedi (20.5%) were the dominant home languages, consistent with the linguistic profile of the surrounding Mopani District. Just over half of the respondents (52.0%) were employed, and 28.3% reported no monthly income.

Composite descriptive statistics and correlations

Composite scores were created by averaging the constituent items for each construct. The descriptive statistics and inter-construct Pearson correlations are reported in Table 3. All composite means were above the scale midpoint of 3.0, indicating generally favourable perceptions of LMD services for pharmaceutical products. The strongest bivariate associations were observed between BI and UB (r = 0.71) and between PU and ATT (r = 0.75), in line with expectations from TAM. Item-level descriptive statistics are reported in Table 4.

Measurement model evaluation

The measurement model was evaluated by CFA in AMOS 29.0 using Maximum likelihood estimation (MLE). Mardia’s coefficient of multivariate kurtosis was within acceptable bounds for MLE, and missing data (< 2%) were handled by listwise deletion. Initial CFA fit indices indicated unsatisfactory model fit (χ² = 412.6, degrees of freedom [df] = 199, χ²/df = 2.07; comparitive fit index [CFI] = 0.881; root mean square error of approximation [RMSEA] = 0.092; standardised root mean square residual [SRMR] = 0.078). Modification indices and standardised loadings indicated that two PU items, namely PU5 (‘eliminate geographic limitations’) and PU6 (‘more useful than other existing channels’), exhibited significant cross-loadings on the PEOU factor. Following Hair et al. (2018), these two items were removed. The final 21-indicator measurement model returned acceptable fit indices (χ² = 335.41, df = 179; χ²/df = 1.87; CFI = 0.922; RMSEA = 0.085; SRMR = 0.061; goodness-of-fit index [GFI] = 0.793; root mean square residual [RMR] = 0.041), consistent with thresholds recommended for SEM in samples of moderate size (Hair et al. 2018).

Composite reliability (CR) values for the scores derived from each multi-item scale exceeded the 0.70 threshold recommended by Hair et al. (2018), with values of 0.872 (BI), 0.894 (PU), 0.902 (UB), 0.909 (PEOU) and 0.913 (ATT). Cronbach’s α values were also acceptable (BI = 0.865; UB = 0.897; ATT = 0.903; PEOU = 0.904; PU = 0.917). Convergent validity was supported: standardised item loadings exceeded 0.70 for all retained items, and average variance extracted (AVE) values ranged from 0.625 (PEOU) to 0.778 (ATT), exceeding the 0.50 threshold (Fornell & Larcker 1981).

Discriminant validity was assessed using the Fornell and Larcker (1981) criterion, which requires that the square root of each construct’s AVE exceeds its correlations with the other constructs. As reported in Table 5, all constructs satisfied this criterion. Together, the CR, convergent validity, and discriminant validity evidence support the appropriateness of the measurement scales for testing the structural relationships hypothesised in the model.

Structural model and hypothesis testing (Hypothesis 1 – Hypothesis 5)

Once the measurement model had demonstrated acceptable reliability, convergent validity and discriminant validity, the structural model was estimated to test H1 to H5. Direct path coefficients are reported in Table 6.

H1: Behavioural Intention → Usage Behaviour. Behavioural intention had a positive and statistically significant effect on UB (β = 0.770, B = 1.196, p < 0.001). Thus, H1 was supported.

H2: Attitude → Behavioural Intention. Attitude had a positive and statistically significant effect on BI (β = 0.585, B = 0.594, p < 0.001). Thus, H2 was supported.

H3: Perceived Ease of Use → Attitude. Perceived ease of use had a positive and statistically significant effect on ATT (β = 0.306, B = 0.277, p = 0.001). Thus, H3 was supported.

H4: Perceived Ease of Use → Perceived Usefulness. Perceived ease of use had a strong positive effect on PU (β = 0.716, B = 0.826, p < 0.001), explaining 51.3% of the variance in PU (R² = 0.513). Thus, H4 was supported.

H5: Perceived Usefulness → Attitude. Perceived usefulness had a positive and statistically significant effect on ATT (β = 0.532, B = 0.666, p < 0.001). Thus, H5 was supported.

Two additional direct paths that were not hypothesised were estimated to allow a complete test of the mediation in H6: PU → UB and PEOU → UB. The direct effect of PU on UB was small and non-significant (β = 0.08, p = 0.387). The direct effect of PEOU on UB was negative and significant (β = −0.33, p < 0.001). This negative coefficient, which appears only when the mediators are controlled, is consistent with a statistical suppression effect that is sometimes observed in mediated TAM specifications (MacKinnon, Krull & Lockwood 2000). It does not imply that easier services reduce use in absolute terms; rather, once the mediated pathway is accounted for, the residual variance in PEOU is weakly and negatively associated with UB. These paths are reported for transparency and are not interpreted as hypothesis tests.

Test of mediation (Hypothesis 6)

Serial mediation was tested using bias-corrected percentile bootstrapping (5000 resamples) along the path PEOU → PU → ATT → BI → UB. The indirect effect was positive and statistically significant (standardised indirect effect β = 0.387; the 95% bias-corrected bootstrap confidence interval did not include zero; p < 0.001). Because the direct effect of PU on UB was non-significant (β = 0.08, p = 0.387), the results indicate that PU operates on UB indirectly, through ATT and BI. H6 was therefore supported.

Taken together, the structural model demonstrates strong explanatory power across the endogenous constructs, with R² values comparable to those reported in other TAM studies. The model explains 61.0% of the variance in ATT (R² = 0.610), indicating that PEOU and PU are key determinants of ATT. Attitude, in turn, explains 34.2% of the variance in BI (R² = 0.342), and BI explains 59.3% of the variance in UB (R² = 0.593).

Discussion

Summary of findings

All six hypotheses derived from a mediated TAM specification were supported in this study of LMD adoption for pharmaceutical products in Nkowankowa Township. Township consumers in the sample held favourable attitudes towards LMD services and reported strong intentions to use them. Behavioural intention was the strongest direct predictor of UB. Attitude was, in turn, predicted by both PEOU and PU, with PU exerting the larger effect. Perceived ease of use also strongly predicted PU. Critically, PU did not exert a direct effect on UB but rather operated through ATT and BI, supporting the serial mediation hypothesised in H6.

Findings in relation to prior research

Based on the results, BI predicted UB (β = 0.77, p < 0.001), supporting H1. The result is consistent with the wider technology-acceptance literature on healthcare wearable devices (Lee & Lee 2020), as well as with recent South African findings on the intention–behaviour link in telemedicine adoption by healthcare practitioners (Cullen et al. 2025) and entrepreneurial intention (Gila et al. 2026). The strong BI → UB pathway (β = 0.77) is at the upper end of effect sizes reported in TAM meta-analyses (King & He 2006).

The moderate to strong influence that ATT has on BI (β = 0.59, p < 0.001) to use LMD services significantly translates into stronger intentions to adopt pharmaceutical delivery services, supporting H2. The findings align with LMD and healthcare-technology studies that locate the influence of usefulness primarily in ATT and intention formation (Adebo et al. 2025; Toraman et al. 2024).

The support for H3, which posits a positive influence of PEOU on ATT (β = 0.31, p = 0.001), suggests that user-friendly service interfaces contribute meaningfully to overall service acceptance. This result is consistent with telemedicine adoption in South Africa (Phorah & Motsi 2025).

Similar findings in other studies also support the acceptance of H4, whereby PEOU significantly predicts PU (β = 0.72, p < 0.001). This finding aligns with research examining e-health system adoption in resource-limited settings (Kalayou et al. 2020) and with South African SME evidence that both perceptions drive adoption (Nyanga & Eresia-Eke 2025). It is worth noting that the strength of this path is context-dependent. Some last-mile and telemedicine studies have reported a weak or non-significant direct PEOU–PU effect (Phorah & Motsi 2025; Toraman et al. 2024), underscoring the value of re-examining this relationship in the township pharmaceutical setting.

Considering the support of H5, previous researchers confirmed that PU positively relates to ATT (β = 0.53, p < 0.001) in telemedicine adoption by physicians in Hong Kong (Hu et al. 1999) and South Africa (Phorah & Motsi 2025).

The support for H6, whereby PU mediated the effect of PEOU on UB through ATT and BI (standardised indirect effect = 0.39, p < 0.001), is confirmed by previous studies by King and He (2006) and Alsyouf et al. (2023). These findings indicate that adoption operated mainly through a cognitive–affective pathway rather than directly through usefulness, supporting the mediated TAM structure.

Theoretical implications

This study makes three theoretical contributions. Firstly, it provides South African evidence that the original TAM structure applies to the adoption of pharmaceutical LMD services in township contexts. The findings indicate that the relationships among the constructs are fully mediated, as proposed by the TAM model.

Secondly, the strength of the PEOU → PU → ATT → BI → UB chain suggests that, in low-exposure technology environments, the cognitive-affective pathway through PU and ATT is essential; direct behavioural effects of usefulness perceptions are weak or absent. Thirdly, the negative direct PEOU → UB coefficient observed under full mediation provides further support for treating PEOU’s behavioural influence as indirect rather than direct in TAM specifications applied to emerging-market consumers.

Managerial implications

For township pharmacy operators, the findings indicate that promotional efforts should not assume that consumers who perceive a service as useful will adopt it. Rather, they should design communications that first establish the ease of use of the LMD interface (e.g. through in-store demonstrations of the WhatsApp ordering process or short video tutorials). This should strengthen perceptions of usefulness, which in turn should foster favourable attitudes, strengthen intentions to use the service and ultimately encourage repeat usage. Targeted incentives for first orders can convert intention into UB.

For digital interface design, although the predominantly Xitsonga-speaking customer base is served in its preferred language during face-to-face interactions at the pharmacy, this may not extend to digital channels. Order-entry interfaces, SMS confirmations, WhatsApp messages and mobile app menus often default to English. Pharmacy operators and digital service vendors should therefore provide these digital interfaces in Xitsonga and Sepedi.

For policymakers, the findings underscore the public-health value of supporting the diffusion of pharmaceutical LMD services in township settings. This could be achieved through subsidised first-delivery vouchers, partnerships with the National Department of Health for the delivery of chronic-medication scripts, and ICT skills-development programmes for township SME pharmacy operators.

Limitations

This study has several limitations. Its focus on a single community-based pharmacy in Nkowankowa Township limits the generalisability of the findings to other South African townships, and the conclusions should be interpreted with appropriate caution. Financial and time constraints led to a small sample and precluded the use of a longitudinal design. The use of English-language questionnaires for non-native speakers may have affected response accuracy, despite verbal translation support. In addition, the study relied solely on the TAM, which limited the scope of the conceptual model. Finally, the lack of literature on township consumers’ adoption of LMD services also constrained the study’s theoretical grounding.

Future research

Future research should include larger, more representative samples, employ longitudinal designs and integrate additional theoretical frameworks, such as the Theory of Planned Behaviour (TPB), Unified Theory of Acceptance and Use of Technology (UTAUT), Social Cognitive Theory or innovation diffusion models, to capture unique adoption mechanisms relevant to resource-constrained environments. Mixed-methods designs combining a quantitative TAM survey with qualitative interviews could also highlight the cultural and social drivers of LMD adoption that the present study could not assess.

Conclusion

The primary objective of this study was to identify the factors influencing the adoption of LMD services among customers of a community-based pharmacy in Nkowankowa Township, South Africa. Drawing on the TAM, the study tested six hypotheses using SEM. All six hypotheses were supported. Behavioural intention was confirmed as the strongest proximal driver of UB. Attitude was predicted by both PEOU and PU, and the effect of PU on UB was fully mediated by ATT and BI. The findings indicate that township consumers are open to adopting LMD services for pharmaceutical products and that interventions promoting uptake should focus on the chain from ease of use through ATT formation to BI.

Acknowledgements

This article is based on research originally conducted as part of Mandlenkosi Nkondo’s master’s dissertation titled, ‘Factors influencing adoption of last-mile pharmaceutical delivery services: A township consumer perspective’, submitted to the Johannesburg Business School, University of Johannesburg in 2025. The thesis was supervised by Mamekwa K. Kekana. The manuscript has since been revised and adapted for journal publication. The original thesis is currently unpublished and was not publicly available online at the time of publishing this article.

During the preparation of this work, the authors used Grammarly for language editing. The content was reviewed and edited by the authors, who take full responsibility for its accuracy.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

Mandlenkosi Nkondo: Conceptualisation, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Validation, Writing – original draft. Mamekwa K. Kekana: Conceptualisation, Data curation, Formal analysis, Investigation, Project administration, Resources, Software, Supervision, Validation, Visualisation, Writing – original draft, Writing – review & editing. All authors reviewed the article, contributed to the discussion of results, approved the final version for submission and publication, and take responsibility for the integrity of its findings.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The data that support the findings of this study are not openly available because of reasons of sensitivity and are available from the corresponding author, Mamekwa K. Kekana, upon reasonable request.

The views and opinions expressed in this article are those of the authors and are the product of professional research. They do not necessarily reflect the official policy or position of any affiliated institution, funder, agency or that of the publisher. The authors are responsible for this article’s results, findings, and content.

References