The Minimum Semitotal Dominating Set (MSTDS) problem is a classical variant of the Minimum Dominating Set (MDS) problem. Given a graph G = ( V , E ) , the goal is to compute a minimum-cardinality vertex subset D ⊆ V such that (i) D is a dominating set, and (ii) for every vertex u in D has a partner v ∈ D within distance two (the semitotal property ), i.e, dist G ( u , v ) ≤ 2 . While the problem is NP-complete in general graphs, recent work has studied approximation algorithms in special graph classes, including unit disk graphs (UDGs). In this paper, we present a deterministic approximation algorithm for the MSTDS problem on UDGs in the graph-based input model, where the graph is provided explicitly without geometric coordinates. Our algorithm achieves a 5.75-approximation factor, improving upon the previous best 6-approximation due to Rout and Das (2024). The core of our algorithm is a multi-stage strategy that starts with a Maximal Independent Set (MIS) and iteratively satisfies the semitotal property through cost-effective vertex swaps and additions of new vertices to the MIS.