Opportunities for improved indicator-based HIV testing in the hospital setting: a structural equation model analysis.

Indicator condition (IC)-guided HIV testing, i.e., testing when diagnosing a condition associated with HIV, is a feasible and cost-effective testing strategy to identify undiagnosed individuals. Assessing determinants for IC-guided testing may identify opportunities for improvement. A survey study based on the Theory of Planned Behaviour (TPB) was conducted among 163 hospital physicians from five specialties in Amsterdam, the Netherlands. Structural equation models were used to determine the association between the TPB domains (i.e., attitude, belief, norms, self-efficacy and behavioural control) and (1) the intention to test as a mediator for HIV testing behaviour (intentional model) and (2) actual HIV testing behaviour (direct model). Both models accounted for the effect of guideline recommendations. Behaviour scored lower than intention on a five-point scale (mean score of 2.8, SD = 1.6 versus 3.8, SD = 1.1; p<0.0001). The direct model had a better fit than the intentional model based on fit statistics. Discrepancies between the determinants most important for intention versus those for behaviour led to the following recommendations: interventions to improve IC-guided testing in hospitals should primarily focus on implementation of guideline recommendations, followed by improving physicians' attitude towards IC-guided HIV testing and self-efficacy, as these were the most important correlates of actual HIV testing behaviour.