A central pillar of universal health coverage (UHC) is to achieve financial protection from catastrophic health expenditure. There are concerns, however, that national health insurance programmes with premiums may not benefit impoverished groups. In 2003, Ghana became the first sub-Saharan African country to introduce a National Health Insurance Scheme (NHIS) with progressively structured premium charges. In this study, we test the impact of being insured on utilization and financial risk protection compared with no enrolment, using the 2012-13 Ghana Living Standards Survey (n = 72 372). Consistent with previous studies, we observed that participating in health insurance significantly decreased the probability of unmet medical needs by 15 percentage points (p.p.) and that of incurring catastrophic out-of-pocket (OOP) health payments by 7 p.p. relative to no enrolment in the NHIS. Households living outside a 1-h radius to the nearest hospital had lower reductions in financial risk from excess OOP medical spending relative to households living closer (-5 p.p. vs -9 p.p.). We also find evidence that in Ghana, the scheme was highly pro-poor. Once insured, the poorest 40% of households experienced significantly larger improvements in medical utilization (18 p.p. vs. 8 p.p.) and substantively larger reductions in catastrophic OOP health expenditure (-10 p.p. vs. -6 p.p.) compared with that of the richest households. However, health insurance did not benefit vulnerable persons equally from financial risk. Once insured, poor, low-educated and self-employed households living far from hospitals had significantly lower reductions in catastrophic OOP medical spending compared with their counterparts living closer. Taken together, we show that enrolment in the NHIS is associated with improved financial protection but less so among geographically remote vulnerable groups. Efforts to boost not just insurance uptake but also health service delivery may be needed as a supplement for insurance schemes to accelerate progress towards UHC.