Problem Statement and objectives. Since the 1990s there is a world wide trend of reforming health sectors in order to make them more efficient, equitable and effective; one important element is the decentralisation from a central administration, generally Ministries of Health, towards district health systems with administrative and financial autonomy. This process is also strong in Latin American countries where this research has been conducted. It implies a new role for district health managers. In the literature there are a number of publications about what district health managers should do, i.e. about their key functions, but little is known about what they are actually doing in reality and about how their performance can be improved through training. The objectives of this research are: 1. To describe and compare health managers' knowledge and performance in nine different Latin American countries. 2. To describe factors related to good and poor health management performance in these countries. 3. To measure the direct results of a new two-year training programme on participants' knowledge and performance in selected countries. In order to explore health managers' performance in a considerable number of Latin American countries, all adjacent countries from Mexico to Ecuador (including the Dominican Republic but excluding Nicaragua) were taken into account in this research. These countries were: Mexico, Guatemala, Honduras, El Salvador, Costa Rica, Panama, Dominican Republic, Colombia and Ecuador. Pilot work was carried out in Chile. Methodology. The selection of health managers, whose performance was to be studied, could not be carried out randomly as the permission for this study was linked to a new management training programme to be conducted by the author and her team. Instead, large geographical areas with varying numbers of health districts (ranging from 15 to 34) were selected by Ministries of Health, in some countries together with international aid agencies, in order to provide a management training programme in these regions. All district health managers in these areas (with few exceptions) were included in the sample. Only in El Salvador and Colombia a national training programme was envisaged so that a national sample was taken. The total number of health managers included in the research was 218 (the minimum sample size for estimating the proportion of specific management characteristics with a tolerated 95% Confidence Interval of +/- 10% was 192 managers). In relation to objectives one and two, health managers were interviewed formally using two different questionnaires but with some similar or identical questions for internal validity testing. Additionally, a large proportion of managers were visited in their workplace in order to carry out observations, conduct staff interviews and informal re-interviews of the managers. This triangulation permitted the check of the validity of interview answers and the collection of complementary information, which was useful for the interpretation of the findings in the quantitative study. The variable "good management performance at district level" has been operationalised by using 12 indicators, which were identified in the literature review. These include the categories: needs assessment (use of indicators, prioritisation of problems for taking actions, risk assessment); planning (design and implementation of plans, use of maps and other visual aids); resource management (co-ordination meetings for human resource management; adequate structure of meetings, communication); systems management (community involvement and inter-sectoral co-ordination); monitoring and evaluation (use of monitoring and evaluation techniques). These indicators were used for establishing the level of management performance in the nine study groups. In relation to the third objective, training programmes were conducted in all countries but only three countries were included in the impact analysis of training (Mexico, Colombia and El Salvador). This was because only in these countries was it possible to identify appropriate control groups, and to include a sufficient number of health managers completing the two years training programme large enough for the statistical analysis of the training impact. The impact analysis was done by comparing the 12 (outcome) indicators mentioned above between intervention and control groups before and after the training programme. Results: In relation to the validity of the information obtained, interview answers showed a high consistency when comparing the two questionnaires applied. Furthermore, there was a good overall agreement between interview answers of the quantitative study and results of the qualitative studies carried out in the district health offices. It was found that management weaknesses were widespread. Almost half (44% to 49%) of the managers did not involve communities, use any monitoring technique, produce minutes for meetings or co-ordinate with other sectors. Approximately one third of them (34% to 36%) did not prioritise health problems, use indicators for monitoring and evaluation or identify risk groups. About one fifth (17% to 21%) did not design or implement their plan, send information to peripheral units or use maps. Overall, there were countries with good performers (Mexico and Honduras), poor performers (Guatemala and Dominican Republic) and other countries in an intermediate position. When analysing the variables associated with relatively good or relatively poor performance (second research objective), a statistical association (using Relative Risk calculations) with good management performance was found for organisational factors such as "existence of a written job description" (RR=2.0; 95% CI 1.4-2.7) and for individual and team factors "power to propose staff for appointment" (RR=1.7; 95%CI 1.3-2.2) and "knowledge of the local situation" (RR=3.4; 95%CI 1.7-7.4). Other possible factors associated with good management were identified by the qualitative study: "organisational discipline", "good working environment and resources" and "good co-ordination with higher authorities". The impact of the two years training programme (third research objective) in three countries was measured firstly by the knowledge gain comparing the results of pre-tests and post-tests (i.e. examinations with short answer questions) in the intervention groups. In Mexico, the improvement was from 24.7% correct answers before the training to 80.3% after the training; in Colombia the improvement was from 21.0% to 83.8% and in El Salvador from 20.5% to 77.3%. Secondly, the improvement of management performance was measured by calculating the Relative Risk (risk ratio) of being a good manager with or without the training programme. The Mexican trainees showed an 8.3 (95%CI 4.7-14.6) times better management performance than the control group; the Colombian trainees were 3.6 (95%CI 2.5- 5.1) times better than their controls and the Salvadorian trainees were 2.4 (95%CI 1.7-3.4) times better than their controls (p< 0.001). Discussion and conclusions. The limitations of this study were that quantitative indicators were mainly used rather than indicators of quality and that equal weight was given to all management functions without prioritising some. Nevertheless, the study tools were precise enough to show severe district management weaknesses, to identify factors associated with good management and to measure the impact of a training programme. On the basis of this research recommendations for short and long term actions were developed focusing at the development of district management systems (which include a detailed job-description of district managers with defined performance standards, defined line managers and appraisal system, organised information exchange), monitoring of management performance (using locally developed set of indicators) and in service training programmes based on previously identified needs (including the information provided by this study and establishing partnerships between MoH and Universities). Long term policy recommendations are related to the need for comprehensive health sector reforms (which address the issue of district health managers being "political tokens" who will be easily changed) and a closer cooperation between MoH and Universities (in order to adapt pre- and post-graduate training to the needs of the services). Further research focussing on management issues, the development of indicators of quality and of cost-effective interventions can help to identify major problem areas and indicate options for interventions.