Drug-resistant malaria is a substantial problem throughout Africa and most countries must regularly adapt their antimalarial drug policies to ensure a continued coverage of effective antimalarial treatment. The timing of drug policy change can be guided by several sources of data: molecular markers of resistance, in-vitro parasite sensitivity, parasitological and clinical failure rates, and community morbidity and mortality rates. Through mathematical simulations of the spread of parasite mutations through a population exposed to high-endemic malaria, we explore the causal and chronological relations between these indicators and show which of them are obscured or confounded by other factors. Taking into account the logistical and practical advantages and disadvantages of each type of data collection, we critically appraise the value of each indicator. A major problem is shown to be that drug efficacy as perceived by people at risk will remain high even after drugs have become almost completely ineffective, resulting in a lack of community pressure for drug policy change. We show that parasitological failure is the most sensitive and timely indicator, which allows around 2-3 years for drug policy change to be implemented, so as to prevent the most rapid rise in malaria-related mortality.