Some papers claim that you can do HLA typing from exome sequencing data. I have not validated this myself. I also guess it depends on capture. Nonetheless, I don't think you can type HLA with VCF generated with a standard pipeline, which is not specifically designed for HLA. Current variant callers won't work on HLA genes due to their high diversity. Also in general, there are thousands of HLA-B alleles. In most cases, you can't find the allele by looking at only one SNP.
No, you can't check if the patient has HLA-B27 only from VCF. There might be a hope if you get raw reads and run a proper HLA typing tool designed for such data.