While decentralization is currently a fashionable idea, the concept has been around for centuries. Antony Jay in Management and Machiavelli notes that colonization by Athens, Rome, Spain and England are examples of decentralization. The founding country said in effect to its colonists, "Off you go. Here's a map, and this is where you'll be founding your colony. From now on it's up to you. Of course we'll help you if we can, but don't depend on us." The author concludes that this sort of independence has a powerful liberating effect on latent or suppressed creative and leadership qualities.

A more recent, and more cynical, view of decentralization comes from Jared F. Harrison in Management by Obstruction or How to Save Your Organization From Needless Efficiency. In this tongue-in-cheek look at management, the author devotes an entire chapter to concentrated decentralization which he defines as an obstructive organizational principle in which decision power appears to be located in decentralized units, but is in fact highly concentrated within a centralized headquarters. Harrison gives several examples of what can be done when a decentralized unit becomes overly efficient:

• increase paperwork by requiring management by objectives and performance evaluation;
• provide career advancement to move talented people out of the decentralized unit; and
• standardize planning procedures, information systems, and integration of services.

These techniques if properly applied should be sufficient to stifle the creativity and to smother the independence of a decentralized unit. The author also proposes the concept of concentrated centralized decentralized centralization, which I think I've seen successfully applied in several countries.

On another track, current literature usually describes four types of decentralization:

• Deconcentration: shifting of powers from center to peripheral units;
• Delegation: transfer of responsibility to parastatal organizations;
• Devolution: transfer of powers from central to local government; and
• Privatization: transfer of governmental functions to NGOs or private sector.

It is important to note, however, that these types of decentralization are never clear cut, but are a kind of shorthand for various forms of structural arrangements to distribute power. In fact, Patrick Vaughan cautions planners not to place too much emphasis on this classification system:

Classifying decentralization into four major types is less helpful for analyzing specific country examples, because of the wide variety of mechanisms adopted. It is probably even less useful in formulating decentralization policies, as a strict use of the four types can severely limit the possible options that a country might adopt. It would be more appropriate for a country to work out its own proposals without the constraints of this classification.

Many countries now have decentralization policies, usually with a focus on creating geographically defined health districts. A health district is a well defined geographical area of 100,000-300,000 inhabitants comprising a referral hospital, satellite health centers and communities. In an article entitled "Building the Infrastructure for Primary Health Care: an Overview of Vertical and Integrated Approaches," authors Smith and Bryant concluded that:

district health systems based on primary health care provide an excellent practical model for health development, including an appropriate health system infrastructure. Within this model the concerns with accelerating the application of known and effective technologies and the concerns with strengthening of community involvement and intersectoral action for health are both accommodated. The district health system provides a realistic setting for dialogue and planning involving both professionals and non-professionals concerned with health and social development.

The health district concept has been actively promoted by WHO for many years, yet only recently seems to be receiving the renewed attention that it deserves. In my alphabet of primary health care, "D" is for decentralization and districts. Both USAID and the World Bank appear to be taking a more systems strengthening approach to decentralization through health districts.

USAID's Child Survival projects have emphasized oral rehydration therapy (ORT) and immunizations as the "twin engines" to develop PHC systems. However, a recent proposal to revise the Child Survival strategy for Africa noted that:

There is growing consensus that while immunization and ORT efforts were very successful during the 1980's, coverage levels have now plateaued or are even decreasing. Child survival interventions cannot be sustained without a functioning health system. More attention must be given to reinforcing the support components of integrated health systems. Each level (national, regional, and district) of an integrated health system must include subsystems to support the health interventions.

Similarly, a recent World Bank study which made an extensive analysis of strategies for implementation of health programs in Africa has concluded that:

The experience of several African countries suggests that a cost-effective "package" of services - comprising curative, preventive and promotive care- can be extended to rural, peri-urban and even urban residents in the form of district-based health care, featuring a district hospital, a network of health centers, and prominent community involvement in multi-sectoral interventions such as water and sanitation. African health systems will have their greatest impact - in terms of efficiency, equity and sustainability- if they are community-based and are organized to provide a basic package of health care services in a limited number of levels, often in health districts.

Despite all this interest, few countries have really succeeded in establishing functional decentralized health districts. There are numerous reasons why effective decentralization has not happened, most of which are closely linked to the general stagnation or decline in health development, especially in Africa. A special issue of Social Science and Medicine examined why health and development has plateaued, or is declining, in many African countries. Concerns and obstacles raised in the introduction entitled "Health Revolution in Africa?" were that:

• Policy decisions are often suggested from outside the health system;
• Improvements are piecemeal without coherent integration;
• Vested interests in status quo protects vertical programs;
• Conflict exists between top-down and bottom-up approaches;
• Decentralization and community involvement have been too slow;
• Women and communities have not been empowered to make decisions;
• There is a lack of monitoring and adequate information system;
• Conflict between high vs. appropriate technology;
• Training is seldom used as a development tool;
• Incentives to work are often inadequate;
• Few successes exist in high level intersectoral coordination;
• A preoccupation exists with recurrent emergencies and not new approaches;
• High population growth rates are accepted with complacency; and
• Universities seldom play an innovative role in primary health care.

These problems are common to almost all decentralization efforts. The "art" of successful implementation of decentralization is finding some way to resolve or work around these obstacles. Five factors which I have found to be associated with the successful decentralization of health districts are discussed below.

1) Poor communication and transportation systems can facilitate decentralization:

Zaire successfully decentralized to health districts despite having only six short (and badly maintained) paved roads in the whole country, and a telephone system that worked part-time. However, this weakness became an advantage for decentralization. In fact, the MOH decentralization strategy corresponds to that mentioned earlier, i.e., "Off you go. Here's a map, and this is where you'll be founding your [health zone]. From now on it's up to you. Of course we'll help you if we can, but don't depend on us."

During my many years in Zaire, I struggled with the logic of this seeming contradiction, and was therefore pleased to find the explanation in Management and Machiavelli:

It is arguable that one reason why the Roman Empire grew so large and survived so long - a prodigious feat of management - is that there was no railway, automobile, airplane, radio, paper, or telephone. Above all, no telephone. And therefore you could not maintain any illusion of direct control...

2) Objectives should be defined in terms of the decentralized unit:

One of the reasons that the Promoting Health Interventions for Child Survival (PHICS) project of Malawi made little progress in decentralization to health districts was that most of the project objectives and indicators were defined in terms supporting activities at the central level. This simply reinforced the existing centralized MOH management capacity. The mid-term evaluation of PHICS in 1993 found that the project had had almost no impact on service delivery.

In order to help PHICS refocus efforts on decentralization and strengthening service delivery, the project logical framework was rewritten with objectives and indicators defined in terms of health districts and its system support components (see table below). This strategy, which I have also used for project design in Zaire, Cameroon and Niger, helps to target the majority of activities and resources at the health district, rather than waiting for a "trickle down" effect from the national or regional level.

3) NGOs can establish management precedents for decentralization:

In Zaire around 50% of the 306 Health Zones are managed by or in collaboration with NGOs, primarily those of the indigenous catholic and protestant churches. These NGOs take on a much larger responsibility for managing primary health care than is found in most countries. In fact of the first 85 health zones to become functional during the 1982-1984 period, 88% were managed by NGOs or received important NGO management assistance. Health zones were "built" on the existing infrastructure of functional hospitals and primary health care activities. This not only accelerated development of PHC and health zones, but also established important precedents in the areas of cost-recovery, local budgeting and financial management, and the control of vehicles.

4) Geographic decentralization should not be limited to administrative units:

While there are advantages to limiting decentralization to administrative units, e.g. departments or arrondissements, there are also real advantages in allowing flexibility in defining the boundaries of health districts. The existing health infrastructure, population size, ease of supervision, and health seeking behaviors of the population should be determinants of health district boundaries.

In Cameroon the MOH first attempted to decentralize to the arrondissement level. This proved unworkable because many arrondissements had populations of less than 10,000 people. It was necessary, therefore, to allow a more flexible definition of limits such that in some regions an entire department, comprising several arrondissements, made up a health district, while in the densely populated Far North, one arrondissement usually became a health district.

In Malawi the region and district levels are to be strengthened to enable increase supervision and local decision-making. However, Malawi has not yet decided what it means by "district." While there are 24 administrative districts, there are also numerous church hospitals throughout the country. The MOH has decided to create 40 Health Delivery Areas, however the relationship between the Health Delivery Area and the existing district health officers is not yet clear. Malawi needs to establish some working models of Health Delivery Area before nationalizing this policy.

Zaire's 306 health zones were carved from some 145 administrative zones, and are a well known example of geographic decentralization. While the concept of health zones was approved in 1975 it was not until there were enough self-defined health zones that the MOH pursued their official delimitation. This was done during regional workshops by representatives from all potential reference hospitals, and was a real stimulus to promoting local ownership and decentralized management of the developing health zones.

5) A project-of-projects environment can encourage decentralized ownership:

Too much planning for decentralization at the national level can lead to centralized decentralization. On the other hand, a "project of projects" environment can facilitate decentralization without dictating all the implementation details. Each decentralized unit, e.g., health district, should pull resources from the project, rather having resources pushed on them.

A project-of-projects strategy is a more efficient use of resources because it by provides assistance to those who are ready to roll, rather than trying to bring everyone up to the same standard before providing a certain type of assistance. It also defines the role of the project staff as facilitators whose main responsibility is to get resources to sub-projects, rather than telling sub-projects what to do. The "fast-track" development of just several health districts serve as a catalyst and "classroom" for other districts. This strategy also creates healthy competition between health districts.

The SANRU Basic Rural Health project in Zaire evolved into a project-of-projects and established its reputation as "a doer" by providing a broad spectrum of assistance, on demand, to health zones. This assistance included national and local training, building rehabilitation, communication systems, medical equipment and medicines, and operations research (see table below).

In conclusion, while the above strategies are certainly not the only ones which contribute to successful decentralization of health districts, their implementation could accelerate the decentralization process, promote increased ownership of programs and local decision-making, and improve the provision of primary health care services.