A method is presented for evaluating the impact of a modification in a surveillance system for an epidemic disease among reporting sources susceptible to the modification. The approach requires a control series with reports from sources that are not affected by the modification. The measure of effect is the reporting odds ratio (ROR) which estimates the ratio of reporting proportions (completeness) before and after the change. The proportions themselves remain unknown. Various potential biases are identified and an investigation of one potential bias is proposed. The method may prove useful in supplementing techniques that estimate absolute values of completeness of registry systems. The application to AIDS surveillance data in Switzerland shows that the transition from voluntary to mandatory reporting had an impact that was enhanced by the 1987 revision of the Centers for Disease Control case definition for AIDS. Reporting sources were 1.9 (95% confidence interval [CI]: 1.4-2.6) times more likely to report their patients after introduction of mandatory reporting. It is estimated that 15% of all reports were attributable to mandatory reporting.