The effectiveness of health recommendations and treatment plans depends on theextent to which individuals follow them. For the individual, medication adherence in-volves an inter-temporal trade-off between expected future health benefits and immedi-ate effort costs. Therefore examining time preferences may help us to understand whysome people fail to follow health recommendations and treatment plans. In this paper,we use a simple, real-effort task implemented via text message to elicit the time prefer-ences of pregnant women in South Africa. We find evidence that high discounters aresignificantly less likely to report to adhere to the recommendation of taking daily ironsupplements daily during pregnancy. There is some indication that time-inconsistencyalso negatively affects adherence. Together our results suggest that measuring timepreferences could help predict medication adherence and thus be used to improve pre-ventive health care measures.