Background
User fees are a known common barrier to using health services, particularly among the poor. When fees are present, many facilities have waiver systems for poor patients to exempt them from paying. Targeting waivers to patients who need them most has been a challenge, especially in fragile states, where relevant data are limited and trust in institutions is low.

Methods
Community-based targeting of vulnerable households was piloted in Afghanistan and evaluated for its feasibility, accuracy and effect on care-seeking. Waiver cards were distributed to very poor and female-headed households in catchment areas of 26 facilities in 10 provinces of Afghanistan in 2005 as one component of a larger health financing study. Households were nominated by community leaders using general guidelines to support 15% of the poorest members. In most cases, waiver cards were pro-actively distributed to them. Targeting accuracy, perceptions, as well the cards' effects on utilization were evaluated in 2007 through household surveys, health facility data, and in-depth interviews and focus group discussions with facility staff and community leaders.

Results
The waiver system was implemented quickly at all but one facility charging fees. Facility staff and community leaders reported favorable perceptions of implementation and targeting accuracy.

However, an analysis of the asset index of beneficiaries indicated that although targeting was progressive, significant leakage and high levels of under-coverage occurred; 42% of cards were used by people in the wealthiest three quintiles, and only 19% of people in the poorest quintile received a card. Households with waiver cards reported higher rates of care-seeking for recent illnesses compared to those without cards (p = 0.02).

Conclusions
Community identification of beneficiaries is feasible in a fragile state. Several recommendations are discussed to improve targeting accuracy of a waiver card system in the future, in light of this research and other international experiences.