The validity of clinic-based samples in approximating out-of-care HIV-infected populations