Figure 1

Probability of being allocated to each incentive structure in each of the four strata over the enrollment period. A, B, C, and D indicate protocol changes in the randomization procedure. These changes were made to address observed disparities in acceptance rates. On 11 April 2012, we implemented a 60%/40% split between [usual care + individual reward + collaborative reward arms] and [individual deposit + competitive deposit arms] (A). On 25 April 2012, we changed to an 80%/20% split between the same two groups (B). On 14 September 2012, we changed to a 50%/50% split between the same two groups (C). On 26 September 2012, we retained this 50%/50% split, continued the adaptive randomization for the deposit arms, but among the 50% of participants randomized to the group containing the other three arms, we fixed the allocation probabilities for the usual care (15%), individual reward (15%) and collaborative reward (70%) arms to bolster assignment to the latter (D).