BACKGROUND: In less developed countries, rheumatic fever still occurs. We started a long-term educational programme in two French Caribbean islands that was directed at the public and at health-care workers to see whether we could reduce the incidence of rheumatic fever. METHODS: Our 10-year programme started in 1981 in Martinique and Guadeloupe, and was based in the community and in clinics and hospitals. The programme established a registry of all cases of primary and secondary rheumatic fever (diagnosed by Jones' modified criteria), with systematic hospital admission of children. We graded carditis as severe, mild, or subclinical, and acute glomerulonephritis was defined by oedema, proteinuria, and haematuria for less than 3 months. The educational part of the programme targeted the public and health-care workers, including doctors, with written information distributed in schools or via radio and television broadcasts or videotapes. For the public, the benign clinical presentation of the initial streptococcal infection was contrasted with the severity of later heart disease. FINDINGS: The first months of the programme led to a 10-20% increase in the number of rheumatic fever cases admitted to hospital, because of the renewed attention paid to the disease. Therefore we took 1982 as the baseline year. In 1982-83 the incidence of rheumatic fever was 19.6 per 100 000 inhabitants aged under 20 in Martinique, and 17.4 per 100 000 in Guadeloupe. In 100 Martinique children and 97 Guadeloupe children in 1982-83, 40 and 71% had carditis, respectively (severe in 10 and 32%). Rheumatic fever was preceded by symptomatic sore throat in 52 and 41% of cases, respectively. The disease was not seen in children with active streptococcal cutaneous infections. Disease frequency was highest in the poorest areas and families, a finding that persisted over time. The programme was associated with a progressive decline in the frequency of rheumatic fever: final reduction of 78% in Martinique and 74% in Guadeloupe. The frequency of carditis also fell. Apart from two outbreaks in one hospital, the frequency of acute glomerulonephritis also declined; 31% of cases had had sore throat, while 56% had skin infections. The cost of the programme during the 4 most intensive years was FFr 250 000 (US$ 44 500) in each island. The cost of childhood rheumatic fever, excluding late sequelae, was initially (in 1982) about FFr 7.8 million (US$ 1426 000). The cost fell to an average of Ffr 550 000 (US$ 100 000) per year in 1991-92. INTERPRETATION: A rapid decline in rheumatic fever incidence was achieved at modest cost. Such a programme needs to be continued because of the risk of disease resurgence.