In 1985, the Pan American Health Organization adopted the goal of eradication of poliomyelitis from the Americas by 1990. Strategies to accomplish this included high vaccination coverage, aggressive outbreak control, and active surveillance for acute flaccid paralysis (AFP). Although the sensitivity of AFP surveillance for detecting paralytic poliomyelitis cases is high, studies have shown the specificity to be low. In 1990, 2497 notified cases of AFP were investigated in the Americas of which 2146 had stool specimens collected. However, only 18 were confirmed as poliomyelitis by isolation of wild poliovirus from stool specimens, 71 were classified as being compatible with poliomyelitis. Cases of AFP due to causes other than poliomyelitis result in extensive but unnecessary outbreak control measures. To predict, at initial clinical evaluation, the likelihood of future confirmation as a case of poliomyelitis, likelihood ratios (LR) were calculated for different combinations of clinical characteristics of AFP cases (249) from Mexico in 1989 and 1990. The best predictors in a child with AFP were proximal muscle involvement which progressed '4 days together with fever at onset of paralysis, and proximal and unilateral involvement with either fever at onset or paralysis which progressed '4 days. The odds would increase by 12 that the child would eventually be confirmed as poliomyelitis (19), based on a stool culture positive for wild poliovirus (95% confidence interval (CI) 2.6-55.9). A guide for use in the field is proposed whereby local health officials, often with little training in neurological evaluation, can predict at initial clinical examination the likelihood that an AFP case will subsequently be confirmed as poliomyelitis.(ABSTRACT TRUNCATED AT 250 WORDS)