Isn't statistical analysis a legitimate tool for helping diagnosis since forever? It's not exactly surprising that a pattern matcher does reasonably well at matching symptoms to diseases.