We approached 3240 mothers (), and 2331 mothers were invited to participate. Of these, 1833 mothers in 4 health regions consented to participate and were randomly assigned to permuted blocks of 2 or 4. Because our aim was to test the materials and not the delivery of the materials, we excluded participants who did not receive a visit from a nurse or who were lost to follow-up. The final intention-to-treat analysis included 649 mothers who received the PURPLE materials and 630 mothers who received the control materials who completed the interview, diary, or both. The baseline characteristics of mothers in both groups were similar, which suggests that there was no selection bias ().
Figure 1: Selection of participants for inclusion in the randomized controlled trial of the effect of educational materials on maternal knowledge and behaviour about shaken baby syndrome. Note: R = randomization.
Of the 1279 participants, data about age were missing for 3%, education for 3%, family income for 9% and marital status for 3%. Data for all 5 knowledge and behaviour scales were missing for 4% of women. Compared with women who responded to at least 1 primary outcome scale, women missing data for all 5 primary outcome scales had less education (completed high school, 24% v. 14%), lower incomes (< $40 000, 74% v. 23%) and were less often in the intervention arm (40% v 51%). Age and marital status were similar among women with data for at least 1 outcome and those missing data for 5 outcomes. All 4 of the diary measures were missing for 17% of the participants.
Scores on the scale that measured crying knowledge were higher among mothers who received the PURPLE materials (63.8 points) than among mothers who received the control materials (58.4 points, difference 5.4 points, 95% CI 4.1 to 6.5 points, p < 0.001). The difference in the scores for shaking knowledge were not statistically significant between the 2 groups (84.0 points v. 83.2 points, difference 0.8 points, 95% CI –0.4 to 1.9 points, p = 0.20).
Compared with the scores among mothers who received control materials, the scores were higher among mothers who received the PURPLE materials for response to crying (48.3 points v. 48.2 points, difference 0.1 points, 95% CI –1.5 to 1.7 points, p = 0.90), response to inconsolable crying (27.7 points v. 26.0 points, difference 1.7 points, 95% CI –0.6 to 4.1 points, p = 0.15), and self-talk response (36.0 points v. 32.9 points, difference 3.1 points, 95% CI –0.3 to 6.4, p = 0.08). These differences were not statistically significant.
More mothers in the PURPLE group than in the control group shared information with other caregivers about walking away if frustrated by inconsolable crying (51.5% v. 38.5%, difference 13.0%, 95% CI 6.9% to 19.2%, p < 0.001), the danger of shaking (49.3% v. 36.4%, difference 12.9%, 95% CI 6.8% to 19.0%, p < 0.001), and crying (67.6% v. 60.0%, difference 7.6%, 95% CI 1.7% to 13.5%, p = 0.01).
When we performed the analysis using only known data, mothers who received the PURPLE materials were more likely than mothers who received the control materials to walk away when their infant cried inconsolably (0.067 times per day v. 0.039 times per day, RR 1.7, 95% CI 1.1 to 2.6, p = 0.01). When we performed the analysis using multiply imputed data, walking away occurred 1.5 times more frequently among mothers in the PURPLE group than among mothers in the control group (0.066 times per day v. 0.044 times per day, RR 1.5, 95% CI 1.0 to 2.2, p = 0.05).
There was no significant difference in the frequency of picking up the infants between mothers in the PURPLE group (3.57 times per day) and mothers in the control group (3.61 times per day, RR 0.99, 95% CI –0.9 to 1.1, p = 0.83). There was no difference in the length of time of contact when the infant was distressed among mothers in the PURPLE group (109.8 minutes per day) and those in the control group (106.1 minutes per day; difference 3.7 minutes per day, 95% CI –4.2 to 11.6, p = 0.36). The level of frustration was not statistically different between mothers in the PURPLE group (1.19) and those in the control group (1.25) (difference –0.06, 95% CI –0.18 to 0.05, p = 0.26). The duration and frequency of distress, fussing, crying and inconsolable crying were not statistically different between the groups (all p > 0.1).
The effect of receiving the PURPLE materials on maternal knowledge about crying was greater among women who watched the DVD, read the materials, or both (). The effect of PURPLE materials on knowledge about shaking was greater among those with more education. There was little evidence that the effect of the PURPLE materials varied on any other scale by education level, reading or viewing the materials, being the mother's first baby, or inconsolable crying.