The current stage of malaria elimination in China requires experienced local health workers with sufficient knowledge of malaria who help to keep the public health system vigilant about a possible resurgence. However, the influencing factors of local health workers’ knowledge level are not fully comprehended. This study aims to explore the factors with heavy impact on local health worker’s knowledge of malaria and propose corresponding suggestions. Underpinned by stratified sampling method, a cross-sectional survey was carried out between November 2014 and April 2016. Chi square test was performed to identify the factors with potential influence on health workers’ knowledge level of malaria. Bivariate logistic regression was employed to explore the relationship between the predictors and local health workers’ knowledge level of malaria. Layered Chi square test was used to calculate the homogeneity of the interaction between training approaches and the percentage of participants with high-level knowledge. The endemic type of county and type of organization played the most significant role in influencing local health workers’ knowledge level regarding malaria in the sample population. The participants from Type 1 and Type 2 counties were 4.3 times (4.336 and 4.328, respectively) more likely to have high-level knowledge of malaria than those who work in Type 3 counties. The probability of having high-level knowledge amongst the participants from county-level facilities (county hospitals and CDCs) were more than 2.2 times higher than those who work in villages. Other socio-demographic factors, such as education and work experience, also affected one’s knowledge regarding malaria. Amongst the six most-used training approaches, electronic material (OR = 2.356, 95% CI 1.112–4.989), thematic series (OR = 1.784, 95% CI 0.907–3.508) and supervision (OR = 2.788, 95% CI 1.018–7.632) were proven with significant positive impact on local health workers’ knowledge of malaria. Village doctors and who served in Type 3 counties were identified as the ones in urgent need of effective training. Three types of training approaches, including electronic material, thematic series and supervision, were proven to be effective in improving local health workers’ knowledge. Nevertheless, the coverage of these training approaches was still limited. This study suggests expanding the coverage of training, especially the three particular types of training, to local health workers, particularly to the target populations (village doctors and who served in Type 3 counties). Online training, small group discussion and targeted skill development may be the directions for the future development of training programmes.