Background: The medical community fails dramatically in the understanding of chronic diseases and development of causal therapies. Statistical associations between vaccinations or infections and autoimmune diseases or a multitude of chronic diseases remains without any scientific rational. The Gulf War illness, with an unexpected high health risk (34%) for military personnel after a manifold of vaccinations, points to the need of a scientific approach different from insufficient linear clonal selection concepts in immunology. Methods: Non-linear dynamics analyzed in time series, phase portraits, and mathematical models. Results: Three different types of chronic diseases, different from lasting acute diseases, are discriminated by Cerebrospinal fluid analysis and nonlinear dynamics. A nonlinear model for a virus-driven chronic disease explains the chronification of the infection, as one of the stable optional states between complete immunity and death. The model also helps create causal therapies and facilitating phase transitions to complete immunity based on individual connectivity in the immune network. Chronic diseases with a sudden change from stable health to a pathological but stable state are phase transitions based on metabolic fluctuations spontaneous or facilitated by external influences (via the immune, endocrine or nervous system). A reduced complexity in diseases is shown by the numerical analysis of time series (e.g., heart rate) or attractor phase portraits. In chronic diseases, with immune system associated pathology, vaccination or infection has to be regarded as a facilitator for a phase transition, like a catalyst, but not as the cause. With this causation we understand why no traces, like causative antigens, are found in Gulf War illness, chronic fatigue syndrome, post lyme syndrome, or the mild encephalitis in schizophrenia.Conclusions: The complexity approach shows why antiviral or antibiotic medications fail in chronic diseases. Disease as an emergent property should be investigated on the phenotype level. Nonlinear analysis of time series of the individual patient gives information not available from group statistics of molecular "multiscale" systems or systems biology. New research perspectives could base on the extended time of registry after vaccinations and should focus on causal therapies, which need financial support, independent from the medical-industrial complex, with its divergent interests.