Diffusion of subsidized ACTs in accredited drug shops in Tanzania: determinants of stocking and characteristics of early and late adopters

Despite the successful development of effective medications to treat health conditions such as malaria, diarrheal disease, HIV and tuberculosis, access to these medications remains highly constrained in most developing countries [1]. Crowded public health facilities, drug stock outs and lengthy travel times have all been noted as barriers to improving access to effective medications in public sector health facilities [2‐5]. It is no wonder, then, that the private sector is typically the first choice in treating common health conditions such as malaria and diarrhea [6]. A major problem, though, is that the latest and most effective medications and technology to conduct effective diagnoses are often unavailable at private drug shops. Many factors likely contribute to low availability of new drugs in retail shops. Retail prices have been shown to impact shop level decisions to stock particular anti-malarial medications in Kenya [7]. Research has indicated that shops sometimes stock specific medications in response to customer demands, rather than policy recommendations [8]. Sellers’ knowledge of pharmaceutical products and proper dosages has been shown to be low in some contexts and could compromise efforts to introduce new products into retail markets [9]. Expanding access to safe and effective medications, along with basic diagnostics for common conditions such as rapid diagnostic tests (RDTs) for malaria, in privately owned drug shops might allow households to receive proper treatment for life threatening diseases earlier, reducing the chance of severe disease and mortality.

Larger scale development of new technologies for global health in the last decade implies that many new health technologies will be introduced across developing countries in the next few years [10]. Some of these technologies, such as new anti-diarrheal drugs, will sell in the same retail markets as those used for malaria medicines. It therefore becomes important to assess how these technologies will diffuse in the market, which drug shops are most likely to stock them first, and the impact of different variables (such as number of employees, number of customers, sourcing pattern and competition) on the rate of diffusion. Understanding diffusion has important strategic value in order to ensure the rapid and widespread adoption of these health technologies. Diffusion has been studied extensively in developed country markets [11]. In his theory of the diffusion of innovations, Rogers introduces the concept of new “adopters.” Early adopters make use of technologies and subsequently encourage the “late majority” and “laggards” to adopt until the new technology or product spreads through the entire market [12]. Rogers’ model assumed that first adopters of new products differed from people who adopt at later times. Bass, however, expanded up Rogers work by hypothesizing that “innovators” lead the introduction of new products, while “imitators” are inspired to purchase new goods following the example of those around them [13]. He created a mathematical model of adoption and applied it to a range of durable consumer goods. The interpretation of Bass’ “innovators vs. imitators” has been since generalized to address marketing mechanisms of broad advertising vs. social contagion. Since that time, the Bass model has been applied extensively to determine how internal and external factors, such as within group market competition vs. broad media messages, influence product diffusion [14‐16]. Bass’ basic ideas work well with current social network theory, where high profile and well-connected individuals influence proximal contacts, who in turn imitate one another [17, 18]. See Figure 1 for a graphical representation of both the Rogers and Bass models of product diffusion. Where the Rogers’ model divides adopters on the basis of timing alone, the Bass model assumes that adopters are intrinsically different and may adopt new technologies at any time, but in distinct ways. “Innovators” will choose to adopt new technologies independently of other individuals within the same group, while “imitators” will accept the new products or technologies in response to others around them.

Although much has been done to understand the adoption and diffusion of health technologies in developed countries, little attention has been paid to this in Sub-Saharan African countries. Even less is known for health products which diffuse through retail markets. Market phenomenon are driven by social, economic, and cultural factors which vary significantly between developed countries and low income countries in Sub-Saharan Africa. One study, though, indicated that developing countries are known to adopt new technologies more slowly than developed countries once available, but that diffusion of products within countries occurs quickly [19]. Other studies found, however, that average penetration potential in developing countries is one third as that of developed countries, that peak sales in developing countries take longer than in developed countries, and that delayed product introduction tends to not have positive effects on the overall adoption rate [20]. High prices have been associated with a reduced rate of diffusion of prescription medications in developing countries [16]. Studies of pharmaceutical product diffusion in developed country settings indicated that large health facilities tend to lead the market in adopting new drugs [21]. Some work on supply chains for malaria drugs in developing countries has suggested that the market for drugs in developing country markets is highly complex [22], making studies of diffusion in developing economies difficult. Of secondary interest to the authors of this paper is how products diffuse through networks of shops, connected either through perceived mutual competition, or through common suppliers. A network based approach similar to techniques used in social network analyses was used to analyze competition between shops in the study areas. Network analysis has recently been used to study problems of social networks, such as friend networks for information sharing or large and intricate assemblages of partnerships as applied to the study of sexually transmitted diseases [23‐26]. Some work has been done applying network methodologies to economic questions [27‐29]. We are unaware of any work that has been done that analyzes network based competition to understand stocking behavior of certain goods and/or pricing.

Knowing how and whether existing models of product diffusion can be applied to drug shops and private markets in sub-Saharan Africa could help public health groups more effectively expand access to essential medications. By identifying specific characteristics of early adopters, new drugs may be introduced more efficiently into the market, thus maximizing availability and minimizing time to universal adoption. We hypothesize that stocking of new products will be the result of a combination of factors which include both local demand for effective products, and a desire for shops to remain competitive in the market by emulating the behavior of peer shops.

In this paper, we leveraged data from a study which longitudinally tracked the adoption of artimisinin combination therapies (ACTs), an effective anti-malarial drug, in Tanzanian Accredited Drug Dispensing Outlets (ADDOs). ADDOs are the lowest level of drug shops that currently have approval from the Tanzania Food and Drug Authority (TFDA) to sell formal anti-malarial medications such as ACTs. The ADDO system was created to improve access to affordable medicines and services in retail pharmaceutical outlets in underserved rural and peri-urban areas [30]. The study was conducted as a part of an operational research project to assess the Affordable Medicines Facility – Malaria (AMFm), an ACT subsidy program hosted by the Global Fund to Fight Tuberculosis, HIV and Malaria. The AMFm provided a supply side subsidy to reduce prices and to increase availability of ACTs in private drug shops in eight pilot countries. Data were collected close to when AMFm subsidized ACTs first became available to ADDOs in Tanzania. Prior to our study, ACTs were prohibitively expensive and generally unavailable in private drug shops [31]. We emphasize that studies of product diffusion usually model the pattern of adoption of new technologies or products into a particular market. At the time of our study, however, ACTs had already been available in the public clinics and hospitals and thus not new to Tanzania. High prices of ACTs, however, heavily constrained availability in private sector outlets, confirmed by the initial round of our survey (see Results). Given our focus on a specific class of private sector drug outlets (ADDOs), the near unavailability of ACTs in these shops and the presence of competing anti-malarial medications, we feel that models of product diffusion as presented in this paper are relevant and applicable. Though the patterns of diffusion and characteristics of early and late adoption might be different for an entirely new product to Tanzania, we feel that this data set offers a unique opportunity to follow a specific product (low cost ACTs) from time of introduction in a specific and important pharmaceutical sector (ADDOs).

This paper will be structured as follows. First, we will describe patterns of ACT stocking over time. Second, we will test associations of competition and demand variables with ACT stocking over the study period. Third, we will characterize shops as early and late adopters and test associations of possible determinants with adopter status. Finally, we will measure and test the contributions of internal and external influence using a mixed influence model. A broad goal of this paper will be to discover whether new product adoption in ADDOs is the result of a shop level motivation to stock ACTs due to actual or yet to be realized customer demand (innovator), or whether shops stock new products due to influence by market competitors.

An operational research study of the AMFm provided us with an opportunity to track the diffusion of a health product in the private market in a developing country from its introduction to widespread availability. We know of no other study which has focused on the question of how pharmaceutical products diffuse in developing world markets using a comprehensive longitudinal series of surveys. Understanding the specific characteristics of shops which stock new products and shops which adopt new products earlier than others could help inform targeting and design of future campaigns to increase availability.

Through our study, we found that busier, more prominent shops in direct competition with others in densely populated areas were more likely to stock ACTs than isolated shops which served fewer customers. Another study also found that shops proximal to outlets which also sold ACTs were more likely to stock them, confirming that shop level competition is a factor in the stocking of new medicinal products [38]. We also found evidence that might suggest that shops which have a higher fraction of their customers presenting for malaria concerns are more likely to stock anti-malarial medications overall and are more likely to adopt new products targeting malaria more quickly. We noted that a specific wholesaler location (Mtwara) was associated with early product availability in small shops suggesting that wholesalers also may play an important role in expanding product availability. Though the limited number of time points prevented more rigorous analysis using a mixed influence model, the parameters estimates from the Bass model used suggest that shops predominantly imitate one another when stocking ACTs, rather than respond exclusively to customer demands. This result is to be expected if shops are responding to suggestions from wholesalers to stock new products. An optimal multivariate model included a combination of increased customer demand, both overall and for malaria, and increased numbers of proximal sources of pharmaceutical goods. This could imply that decisions to stock new products are the result of a complex mix of factors related to both a need to imitate and thus remain competitive in the market and to simultaneously satisfy the potential demands of customers. More work needs to be done to parse out the relative impacts of competition and demand factors in influencing shop level decisions to stock new products.

Our study suffered from many limitations. First, the ability to generalize our results to other regions of Tanzania and to other Sub Saharan African countries may be limited. Rukwa and Mtwara respectively represent inland and coastal areas of Tanzania in terms of access to public and private health care. The results might differ in the area surrounding heavily urbanized Dar es Salaam. Follow up research might include urbanized Dar es Salaam to determine whether patterns of diffusion differ from those in other, more remote regions. Further, presence of the shop accreditation system in Tanzania might limit the ability to generalize our findings to other African countries. Though private, the ADDOs are licensed and accountable to the central regulatory body which, in cooperation with local health authorities and development partners, attempts to ensure standards of shop keeper knowledge and product quality. The private pharmaceutical sector in other Sub Saharan African countries might not be as formalized. Future research might attempt to monitor patterns of diffusion of new pharmaceutical products in private sector health markets in a number of countries simultaneously.

Second, though we had access to detailed data such as shop locations, names of competing shops and customer numbers, we lacked more detailed knowledge of the attitudes and experiences of the shop proprietors themselves. Informal interviews while study teams were in Tanzania indicated that business knowledge and attitudes were heterogeneous among shops. Future studies might seek to collect more detailed data on how shop owners strategize their businesses, assuming they do at all. Knowing more information about what criteria shop owners use to determine which new products will they start stocking might help create efficient strategies for the launch and faster adoption of new health technologies in private markets. There is evidence that educational strategies that seek to improve the knowledge and thus efficiency of small drug providers has been shown to be effective in some contexts [39].

Third, the unique nature of the AMFm might have produced a case of product diffusion that may not always be applicable to other contexts. New products that lack such public support and deep price subsidies might not diffuse as quickly. The AMFm was widely publicized and ACTs were likely known to customers and shopkeepers. Shops may be reluctant to stock a product that is unfamiliar to them or has yet to be accepted by or is unknown to current customers. Also, deeper information on the demographics and experiences of the catchment populations, such as socio-economic profiles or the incidence of fever in the catchment, might have better informed the demand variables. Further country level research might also examine the role of spatial heterogeneities of malaria risk in influencing shop level stocking behavior.

Decisions to stock new pharmaceutical products in Tanzanian ADDOs are influenced by a combination of factors related to both market competition and customer demand, but are particularly influenced by the behavior of competing shops. Efforts to expand access to new pharmaceutical products in developing country markets could benefit from initial targeting of high profile shops in competitive markets and wholesale suppliers to encourage faster product diffusion across all drug retailers. Our results might be specific to AMFm ACTs, given their subsidized nature and the previous availability of ACTs in the public sector. Thus, future research might follow an entirely new product in a market like Tanzania. Studies of product diffusion in developing country pharmaceutical markets will help inform policy and market based strategies to maximize uptake while minimizing time to adoption, improving access to life saving medications.

^aRCH clinics were used as a proxy in place of all public health facilities as we did not have GPS coordinates of all public health facilities. A large fraction of public health facilities offer RCH services so this was a reasonable approximation.