ACE2 interaction network caused by virus infection can affect the expression levels of miRNAs and influence severity of COVID19. This study aimed to analyse the diagnostic/predictive utility of ACE2-related miRNAs, identified by in silico analysis: miR10b, miR124, miR200b-3p, miR26b, miR302c-5p in patients with COVID-19, and aimed to unravel the functions of miRNAs, by using machine learning-SHAP analysis for clinical data and bioinformatic tools. Blood samples and clinical data of 79 patients and 32 healthy were collected at; day of admission, 7days and 21days after admission. Endpoint was hospitalisation length of stay (>21days) and/or death in follow-up. Delta low miR200b-3p expression (7days-admission) presents predictive utility in assessment of the hospital length of stay and/or death (AUC:0.73,p=0.002). Logistic regression analysis showed that delta low miR-200b-3p, diabetes are independent predictors of increased hospital length of stay and/or death (OR=5.8;CI=0.57-21.21;p= 0.008, OR= 4.9;CI=1-23.9;p=0.04). MiR-26b-5p and miR10b in patients were found lower at the baseline, 7 and 21days after admission compared to healthy controls (p<0.0001 for all time points). SHAP analysis showed miR200b-3p (day7), miR302c-5p (day7), CRP (day7), neutrophils (day0), and DDimer (day0) as most promising predictors of long hospitalisation. Pathway-enrichment showed the following top pathways of the interleukin-2 signaling pathway, and cancer pathways. MiR200b-3p showed regulation of COVID19-related targets associated with Tcell protein tyrosine phosphatase and HIF-1 transcriptional activity in hypoxia, key pathways in COVID19. Bioinformatics analysis showed miRNAs roles in multiple CVDs phenotypes associated with COVID19. We validated/characterized miRNAs which can serve as novel, predictive biomarkers of the long term COVID19 hospitalisation and can be used for early stratification of patients and prediction of severity of infection.