**Background:** The WIC Infant and Toddler Feeding Practices Study-2 (WIC ITFPS-2) is a longitudinal cohort study funded by USDA to assess feeding practices and behaviors of women and young children participating in WIC. The study required a representative sample of infants enrolled in WIC for the first time, either prenatally or within 2.5 months after birth. Because no national list of WIC enrollees was available at the time of sampling (2012), the researchers employed time-location sampling (TLS), a method typically used for rare or stigmatized populations, to construct a probability sample of new WIC enrollees.
**Methods:** The sample design used a stratified, multistage approach. First, a sampling frame was constructed from the WIC Participant and Program Characteristics 2010 (PC 2010) data, which contained 4,979 reporting units (a mix of individual sites and local agencies). Reporting units in American Samoa, Guam, Northern Mariana Islands, and US Virgin Islands were excluded for geographic reasons, and those with fewer than 30 estimated eligible new WIC enrollees per month were excluded due to low enrollment. After exclusions, 1,834 eligible units remained. In phase 1 of the first stage, 160 reporting units (4 per stratum across 40 strata) were selected with probabilities proportional to size (PPS). Eligibility was verified using updated January 2012 enrollment data, and in phase 2, 80 units (2 per stratum) were subsampled with equal probabilities. For units that were local agencies, one site per agency was selected in the second stage with PPS, yielding a final sample of 80 sites associated with 27 state agencies. Six sites (7.5%) were replaced by predesignated replacements due to non-cooperation.
Recruitment windows were assigned to each site based on estimated daily enrollment of eligible participants (prenatal enrollees plus 20% of postnatal enrollees) using July 2012 data. The target was to approach 98 women per site. Window lengths ranged from 4 to 87 recruitment days (2 weeks to nearly 18 weeks), with 54% of sites requiring 3–8 weeks. Start dates were randomly assigned across nine weekly release groups to stagger recruitment. Two recruitment models were used: on-site recruitment (75 sites) and referral-form-only recruitment by phone (5 sites). Bilingual recruiters were assigned to 46 sites where more than 15% of enrollees spoke only Spanish. The recruitment period spanned July through November 2013.
To meet precision requirements for subgroup analyses, a core sample and a supplemental sample were designated. Initially, half of new enrollees were assigned to each, with supplemental sample members subject to further subsampling based on mother's race, ethnicity, trimester at WIC enrollment, and pre-pregnancy BMI. Due to lower-than-expected WIC enrollment, recruitment windows were extended by 3% before recruitment and by an additional 10% during recruitment, and subsampling of the supplemental sample was eventually eliminated.
**Key Results:** The study approached 6,775 women for recruitment (target: 7,840) and enrolled 4,367 women/infants (target: 4,435), with 4,046 live births (target: 3,990). Of enrolled live births, 3,235 (80%) were assigned to the core sample and 811 to the supplemental sample. The overall response rate through study enrollment was 82%. However, subgroup enrollment deviated from targets: Black mothers (0.8 ratio of actual to expected), White mothers (1.7), Other race (0.4); Hispanic (1.2), Non-Hispanic (0.9); first trimester enrollees (0.8), second/third trimester (1.6), postnatal (0.5); overweight mothers (1.1), obese (1.0), other BMI (0.9). There was considerable site-to-site variation in the number of women approached, with one outlier site approaching 264 women during 65 recruitment days.
**Clinical Implications:** This study demonstrates that TLS can be successfully adapted to sample participants of public health nutrition programs when no comprehensive sampling frame exists, producing a probability sample with known selection probabilities. The approach allowed for oversampling of key subgroups and real-time adjustments to recruitment procedures. The findings highlight the importance of building flexibility into TLS designs to accommodate unexpected fluctuations in enrollment flow, and the need for careful monitoring and adjustment of recruitment windows and subsampling procedures. The WIC ITFPS-2 will provide valuable longitudinal data on feeding practices and health outcomes for WIC participants through age 9.