Bacillus stearothermophilus large fragment (BSTLF) DNA polymerase is reported, isolated on silica via a fused R5 silica-affinity peptide and used in nucleic acid diagnostics. mCherry (mCh), included in the fusion construct, was shown as an efficient fluorescent label to follow the workflow from gene to diagnostic. The R5 immobilisation on silica from cell lysate was consistent with cooperative R5-specific binding of R52-mCh-FL-BSTLF or R52-mCh-H10-BSTLF fusion proteins followed by non-specific protein binding (including E. coli native proteins). Higher R5-binding could be achieved in the presence of phosphate, but phosphate residue reduced loop-mediated isothermal amplification (LAMP) performance, possibly blocking sites on the BSTLF for binding of β- and γ-phosphates of the dNTPs. Quantitative assessment showed that cations (Mg2+ and Mn2+) that complex the PPi product optimised enzyme activity. In malaria testing, the limit of detection depended on Plasmodium species and primer set. For example, 1000 copies of P. knowlesi 18S rRNA could be detected with the P.KNO-LAU primer set with Si-R52-mCh-FL-BSTLF , but 10 copies of P. ovale 18S rRNA could be detected with the P.OVA-HAN primer set using the same enzyme. The Si-immobilised BSTLF outperformed the commercial enzyme for four of the nine Plasmodium LAMP primer sets tested. Si-R52-mCh-FL-BSTLF production was transferred from Cambridge to Accra and set up de novo for a trial with clinical samples. Different detection limits were found, targeting the mitochondrial DNA or the 18S rRNA gene for P. falciparum. The results are discussed in comparison with qPCR and sampling protocol and show that the Si-BSTLF polymerase can be optimised to meet the WHO recommended guidelines.