Background
The large-scale introduction of malaria rapid diagnostic tests (RDTs) promises to improve management of fever patients and the rational use of valuable anti-malarials. However, evidence on the impact of RDT introduction on the overprescription of anti-malarials has been mixed. This study explored determinants of provider decision-making to prescribe anti-malarials following a negative RDT result.

Methods
A qualitative study was conducted in a rural district in mid-western Uganda in 2011, ten months after RDT introduction. Prescriptions for all patients with negative RDT results were first audited from outpatient registers for a two month period at all facilities using RDTs (n = 30). Facilities were then ranked according to overall prescribing performance, defined as the proportion of patients with a negative RDT result prescribed any anti-malarial. Positive and negative deviant facilities were sampled for qualitative investigation; positive deviants (n = 5) were defined ex post facto as <0.75 % and negative deviants (n = 7) as >5 %. All prescribing clinicians were targeted for qualitative observation and in-depth interview; 55 fever cases were observed and 22 providers interviewed. Thematic analysis followed the ‘framework’ approach.

Results
8344 RDT-negative patients were recorded at the 30 facilities (prescription audit); 339 (4.06 %) were prescribed an anti-malarial. Of the 55 observed patients, 38 tested negative; one of these was prescribed an anti-malarial. Treatment decision-making was influenced by providers’ clinical beliefs, capacity constraints, and perception of patient demands. Although providers generally trusted the accuracy of RDTs, anti-malarial prescription was driven by perceptions of treatment failure or undetectable malaria in patients who had already taken artemisinin-based combination therapy prior to facility arrival. Patient assessment and other diagnostic practices were minimal and providers demonstrated limited ability to identify alternative causes of fever. Provider perceptions of patient expectations sometimes appeared to influence treatment decisions.

Conclusions
The study found high provider adherence to RDT results, but that providers believed in certain clinical exceptions and felt they lacked alternative options. Guidance on how the RDT works and testing following partial treatment, better methods for assisting providers in diagnostic decision-making, and a context-appropriate provider behaviour change intervention package are needed.