【Objective】 Forest ecosystems have an obvious hierarchical structure; thus, the concept of health assessment results from the sample points or sample areas at the regional scale is currently an important issue in sustainable forest management. Therefore, the goal of this study was to achieve the efficient conversion of forest health assessment results among trees, stands and regional levels using statistical methods. It may provide a theoretical basis and technical support for forest health management in northeastern China. 【Method】 Based on the datasets of 51 sample plots and forest resource inventories in the Pangu Forest Farm in the Greater Khingan Mountains, a tree-level health assessment model was constructed using the entropy-AHP method. Then, five statistical indicators, namely, mean value (H_m), standard deviation (H_sd), coefficient of variation (H_cv), skewness (H_pd) and kurtosis (H_fd), were generated for each plot, which were treated as the results of a stand-level health assessment. The error-in-variable regression model was applied to develop a comprehensive evaluation model of stand-level health status. Finally, the spatial distributions of regional-level forest health scores were mapped using the estimated regression model and forest management inventory datasets. The basic characteristics and patterns of the five statistical indicators were analyzed. 【Result】 The prediction accuracy of the estimated forest health comprehensive evaluation model was relatively high, in which the values of determination coefficients (R²) of the H_m, H_sd, H_cv, H_pd and H_fd models were as high as 0.464 3, 0.305 6, 0.909 6, 0.298 1 and 0.448 5, respectively, meeting the potential demands of forest health assessment. The results of tree-, stand-, and regional level health assessments indicated that the forests within the study area mainly belonged to the category of sub-health, in which the stand (height of individual tree, height of dominant tree, number of tree species, number of trees per hectare, stand basal area and stand volume) and topographic factors (elevation, slope and slope position) affected the forest health scores and their distributions. The average values of forest health were 0.623 4, which belonged to the sub-health level. With respect to the five statistical indicators, they all exhibited obvious spatial patterns, in which the stands with larger H_m, H_cv, H_pd and H_fd values were mainly concentrated in the northern part of the forest farm, in particular being close to settlements and roads. However, the values were significantly lower in the southern portion where transportation was usually not convenient. Furthermore, the spatial distribution pattern of H_sd was completely opposite to that of the other indicators, revealing that the adaptive management is meaningful in the improvement of forest health levels. 【Conclusion】 The statistical methods used in this study could efficiently achieve scale conversions of forest health assessments at the trees, stands and regional levels.