Electronic　medical　records　are　an　essential　basis　for　doctors　in　clinical　diagnosis　and　treatment.　Accurate　and　effective　retrieval　of　similar　medical　records　can　not　only　offer　great　help　to　clinical　decision-making　but　also　bring　benefits　and　convenience　to　case-based　patient　research　and　the　unearthing　of　similar　patient　groups.　However,　the　existing　electronic　medical　record　retrieval　model　cannot　accurately　and　efficiently　retrieve　similar　medical　records.　In　this　paper,　we　propose　a　Chinese　electronic　medical　record　retrieval　method　using　fine-tuned　Roberta　and　hybrid　features.　Firstly,　RoBERTa　pre-trained　language　model　was　adopted　and　finetuned　by　using　real　clinical　Chinese　electronic　medical　records　to　make　character-level　embedding　more　suitable　for　Chinese　electronic　medical　records.　Then,　the　character　vectors　generated　from　Fine-tuned　RoBERTa　model　are　input　to　BiLSTM　and　CNN　respectively　and　the　features　of　electronic　medical　records　respectively　output　from　BiLSTM　and　CNN　are　combined.　Finally,　the　similarities　between　combined　features　of　Chinese　electronic　medical　records　are　calculated　to　obtain　the　most　similar　medical　records.　A　comparative　experiment　conducted　on　a　real　dataset　shows　that　our　method　can　make　a　progress　in　the　accuracy　of　Chinese　electronic　medical　record　retrieval.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.