Sharing aggregated electronic health records (EHRs) for integrated health care and public health studies is increasingly demanded. Patient privacy demands that anonymisation procedures are in place for data sharing. However traditional methods such as k-anonymity and its derivations are often overgeneralizing resulting in lower data accuracy. To tackle this issue, we present the Semantic Linkage K-Anonymity (SLKA) approach supporting ongoing record linkages. We show how SLKA balances privacy and utility preservation through detecting risky combinations hidden in data releases.