We found two organizational networks of nearly identical size providing services for HIV care and family planning in two sub-cities of Addis Ababa. However, there were few additional similarities. In one network, a government hospital and associated clinics figured prominently. The other, lacking a hospital and with fewer government clinics, had more NGOs and FBOs and was less centralized.
We are unaware of any other publications reporting on HIV- or family planning-related organizational networks in developing countries. The only networks for comparison are in the US. A network of 30 HIV/AIDS services in Baltimore in 1997 had a referral density of 0.43 [
28]. Thomas and colleagues measured the densities of information exchanges among HIV prevention organizations in ten counties of North Carolina [
11]. The six counties with a high rate of syphilis (greater than 4.5 cases/100,000 population) had a median density of 0.156. The four counties with a low rate of syphilis (less than 3.5 cases/100,000) had a median density of 0.384. The higher densities in low rate counties suggest that better coordination may contribute to better disease control.
Evaluating networks
Compared to the Baltimore network, the referral densities in the two Addis Ababa sub-cities are relatively low. The Baltimore connections may have existed in part from a donor program requiring a collaborative approach [
28]. There were organizations in both sub-cities that provided care for people with HIV but did not attend to their family planning needs and were not referring their clients to other organizations that could. The opposite was also evident, though less clearly so since the family planning organizations also provided some HIV services. When they provided only a few HIV services and did not refer clients to other HIV organizations, one can infer that some HIV services were not being received, or the organization in question was not facilitating access to them.
Although the referral densities of the two sub-city networks were similar to each other on an absolute scale, they were composed of different organization mixes. Each network arose from a particular context in response to local needs and opportunities; each resulting structure has strengths and weaknesses. Hypothetically, a more centralized network, as in Kolfe-Keranyo, can afford quicker action and more coordination when the central organization dictates changes or procedures. However, if the central organization encounters a crisis, such as a loss of funding, negative consequences can ripple out to the rest of the network quickly. Centralized networks can thus be more efficient, but they are also more dependent on the success of the central organization.
A more decentralized network, as in Kirkos, has several organizations that are well connected but none that dominates. Should one of these encounter a crisis, the others could keep the network functioning. Decentralized networks, then, might not be as efficient as centralized ones but they could exhibit more resilience to shocks.
Our data offer little basis for determining whether either network is right; that it offers the optimal mix of services or the organizations coordinate in ways that best serve the clients’ needs. There is no gold standard for the density of referrals. More is not necessarily better. In some instances, referrals can be unnecessary and inefficient. Moreover, even though an outsider can point to organizations that could refer to each other based on the services offered, there may be reasons not captured by a questionnaire that would argue against those referrals. We learned in our presentation of results for example, that one organization to which others were referring was overwhelmed with received referrals and needed to decrease them.
In cultural anthropology, an emic perspective is that of cultural insiders. It stands in contrast to the etic perspective of outsiders. What a network should be depends largely on the emic perspective of providers and their clients. When providers of HIV care and reproductive health in Addis Ababa see objective documentation of how they are interacting, they need to address a number of questions.
Among them are the following: Is this the mix of services needed? Are clients’ needs being met? Are they connecting as you, the providers, would like them to? Are they connecting as the clients would like them to? What are the barriers to connection? Can they be addressed? With the insiders’ answers to questions such as these, insiders and outsiders alike will be better able to move the network towards the insiders’ ideal.
The etic and systematic perspective provided by the network sociogram and the gap analysis is a necessary complement to the insiders’ views. It enables the service providers to see themselves as a network when they are more naturally inclined to think about themselves as discrete organizations. And when they see the network, the data can guide them through questions about their connections and inform their desires for change.
Network analysis limits
There are limits, of course, to the network perspective we describe. A single organization can be part of several networks at the same time. A medical clinic, for example, would likely address health concerns beyond HIV care or family planning, such as mental health or diabetes. There will be other organizations that it refers clients to for services it doesn’t provide itself; organizations that were not part of the network we described for HIV care and family planning. The HIV care and family planning network information provides limited guidance for those other connections.
We also did not ask respondents about referrals to or from organizations outside of their sub-city. For example, we did not ask organizations in Kirkos about referrals to the government hospital in Kolfe- Keranyo. The organizations to include in a network analysis depend on the question at hand. In our case, it was how organizations with similar or complementary missions located within the same sub-city interact with each other. Nor do we have information from community members about how they view the organizational network or how their behavior affects the network.
Our analysis presents a picture of the network at one particular point in time. Networks are inherently dynamic, constantly adapting to new challenges and opportunities. The network we described could be significantly different a few months later. The relevance of our analysis, however, was strengthened by the stakeholders’ view that the results reflected their experience. The results were thus helpful for discussing desired changes. In a forthcoming publication we will report the results of an intervention that implemented their recommendations.