Background: In many countries, disease registries are the main source of information that is necessary for planning and monitoring health-improving programs. In Iran, there is no unified registry system for deaths due to traffic accidents, so organizations that are involved in traffic accidents report different statistics. This study aimed at estimation of traffic accident mortality rate and completeness of death registries in Northern provinces of Iran from March 21, 2010 to March 20, 2011 using capture-recapture method. Methods: Attributes of all traffic accident victims including their first name, last name, age, sex, and date of death enrolled by traffic police department, legal medicine organization, and hospitals were used to count the number of cases that were captured by one, two or all the three aforementioned organizations. Then, different log-linear models proposed in the context of capture-recapture method were fitted to predict the number of cases missed by the three sources. Akaike information criterion (AIC) and Bayesian information criterion (BIC) were used as goodness of fit indices to identify preferred fitted models. Results: According to the prediction of the most fitted model, the total number of traffic accident deaths was 3857 of which, 727 cases were not captured by any of the responsible organizations. Mortality rate due to traffic accidents was estimated as 53.73 with 95% confidence interval (52.41 -55.15) per 100000 population of the three provinces. Coverage rates of legal medicine organization, hospitals, and traffic police department were 65.95%, 36.06%, and 16.85%, respectively. Conclusions: Incidence rate of fatal traffic accidents in Northern provinces of Iran is higher in comparison with other provinces. Besides, it seems that statistics reported by responsible organizations like legal medicine organization, traffic police department, and hospitals are underestimated, so it highlights the need for application of methods like capture-recapture to correct for this underestimation.