Over 100,000 people are living with human immunodeficiency virus (HIV) in the UK, a virus that attacks the immune system. If HIV-positive people are treated early and well then they have a near normal life expectancy and are less likely to infect others with their HIV accidently. However, a quarter of people infected with HIV do not know that they carry the virus. Although you could test (screen) everyone for HIV, this would not be cost effective. We want to better target testing to people who are most likely to have HIV. Patients with HIV often see their GP with other illnesses which can be due to HIV before their HIV is detected as the cause. In this study, we aim to develop a computer program to alert the practice that an individual should be offered an HIV test by their GP when the patients visits the surgery. To do this we need to know what symptoms people with HIV go to their GP with. We will find all the HIV cases and see which symptoms are best predictors of being HIV-positive and use this to build a computer program that will flag people with these symptoms in GP records.
We aim to detect undiagnosed HIV infection (25% of all people living with HIV in the UK) using a risk algorithm that can be incorporated into GP IT systems (e.g EMIS). These individuals are often seen in Primary Care with risk factors for HIV acquisition or clinical symptoms associated with HIV disease but the need for an HIV test may not be apparent to the GP, particularly in lower prevalence areas. We will undertake a feasibility study using retrospective data:
i) Using CPRD data we will determine HIV testing rates in the GP setting by calendar year, sex and age.
ii) We will investigate which risk factors are predictive for HIV and the clinical symptoms they experienced prior to or at diagnosis of HIV. We will fit candidate predictive models based on these factors to discriminate between HIV-positive and negative individuals.
iii) We will evaluate the specificity, sensitivity and positive predictive value of candidate models based on different configurations of risk factor information.
iv) The discrimination of the model with optimum performance characteristics will be further assessed in a retrospective analysis of GP data in two local practices and refined if sensitivity and specificity suggests this is needed.
The outcome for determining testing rates in primary care is whether a person has undergone an HIV test.The outcome for research on the HIV risk algorithm is HIV positive, denoted either by a positive HIV test result or recorded that HIV positive/AIDS recorded in CPRD, or died with HIV mentioned on the death certificate as recorded by ONS death registry.
John Macleod - Chief Investigator - University of Bristol
John Macleod - Corresponding Applicant - University of Bristol
Jonathan Rougier - Collaborator - University of Bristol
Margaret May - Collaborator - University of Bristol
Mark Gompels - Collaborator - North Bristol NHS Trust
Skevi Michael - Collaborator - University of Bristol
Tim Jones - Collaborator - University of Bristol
Charlotte Davies - Collaborator - University of Bristol
HES Admitted Patient Care;HES Outpatient;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation