Parent Screen-Time Programs Work, But the Average Effect Is Small

There is a particular kind of late-evening searching that happens after a hard day with a four-year-old. The tablet went back on at dinner because dinner had to get cooked. Somewhere near 11pm a parent closes the laptop on the eleventh listicle of the night, all of which said much the same thing: set limits, model good habits, keep screens out of the bedroom. What almost none of them said is whether any of that has ever been tested against a group of families who were told nothing at all.

That is the gap this week's coverage sat in. A brief online program for parents, the reports said, lowered children's screen time and parental stress along with it. The claim is real and it comes from a real randomised trial. The number attached to it, though, is one of the largest ever reported for this kind of program, and the wider evidence base it belongs to is considerably more modest.

The study is Stop and Play, a single-blind, two-arm cluster randomised controlled trial run through government preschools in the Petaling district of Malaysia and published in the Journal of Medical Internet Research in 2023. It enrolled 360 mother-child dyads from low-income families and randomly assigned whole preschools, not individual families, to the program or to a waitlist control group. Recruitment and follow-up ran from March to December 2021.

The intervention itself was deliberately light: four weeks of parent health education delivered over WhatsApp, built from whiteboard animation videos, infographics and a problem-solving session. Mothers, not children, were the recipients. Screen time was collected by validated self-administered questionnaire at three points — baseline, immediately after the four weeks, and three months later — and analysed with generalised linear mixed models adjusted for maternal age and education, household income, the child's sex, baseline scores, and clustering at the school level.

At baseline these were heavy-screen households by any standard. Children in the intervention group averaged 290.29 minutes a day (SD 162.46); the control group averaged 273.69 (SD 139.04). That is between four and five hours daily, in preschoolers, before anything was tried.

Three months after the program ended, children in the intervention group were reported to be using screens 202.29 minutes a day less than controls (95% CI -224.48 to -180.10, p < 0.001). Attrition was very low: 352 of 360 dyads finished, a 2.2% loss. The intraclass correlation for children's screen time was 0.04, meaning most of the variation sat between families rather than between schools.

The secondary outcomes moved in the same direction. Mothers' own leisure screen time fell by roughly 70 minutes a day (95% CI -91.51 to -49.35). Maternal knowledge scores rose (6.88, 95% CI 6.11 to 7.65) and self-efficacy for reducing a child's screen time rose (1.59, 95% CI 1.48 to 1.70). One measure did not move: the presence of a screen device in the child's bedroom changed by -0.03 (95% CI -0.08 to 0), which did not reach significance at p = 0.08. The most concrete, least self-reportable thing the trial measured is the one thing it could not shift.

The useful context arrived in July 2026, when Obesity Reviews published a systematic review and meta-analysis of randomised trials of screen-time interventions in children aged 0 to 5. Sixty-five studies met inclusion criteria and 43 carried enough data to pool. Most targeted children aged 2 to 5 (77%) and worked with caregiver-child pairs (57%), delivered in preschools (29%), online (20%) or in healthcare settings (12%).

Judged against their own stated objectives, 62% of the interventions succeeded — a genuine majority. But the pooled effect size was small: g = 0.16, 95% CI 0.01 to 0.31, p = 0.036. The confidence interval very nearly touches zero. The characteristics that clustered among the programs that worked were delivery in a healthcare setting, delivery by the research team itself, and caregivers as the sole recipients. Interventions aimed at 2- to 5-year-olds produced larger effects than those aimed at infants and toddlers (B = 0.31, p = 0.038), even though the younger studies more often met their objectives.

The trial's own authors were candid about this. They note that a 202-minute daily reduction is far above the 25 to 37 minutes a day reported by the comparable studies they set their result against, and they offer two explanations from inside their own study period. The first is that Malaysia's pandemic movement restrictions had pushed baseline screen time unusually high, leaving more room to fall. The second is that schools reopened during the third data-collection window, so children in both arms were simply out of the house more when the final measurement was taken.

There is a third, which the authors also name. Screen time here was reported by mothers who knew which group they were in and had just spent four weeks being taught that screen time matters. The team cross-checked questionnaire answers against screen-media diaries for 10% of participants (18 of 180) and found 83% agreement (15 of 18) — reassuring, but a check on recall, not on the social desirability the authors flag directly: a temptation to under-report a child's hours and over-state one's own parenting. A measure that is vulnerable in exactly the direction the intervention pushes is the weakest place to put a trial's headline number.

None of this makes the program worthless. A well-run cluster trial with 2.2% attrition found a real, adjusted, statistically robust effect, and its cheap WhatsApp format is the kind of thing a health system could actually deliver. It does suggest the honest expectation is the pooled one — a small shift — rather than the three-and-a-half hours a day the coverage carried.

The strongest reading of this evidence is also the least dramatic one: parent-directed screen-time programs are real interventions with real, modest effects, and the field's own pooled estimate of g = 0.16 is a better guide to what a family might see than any single trial's headline. What makes Stop and Play worth reading is not its 202 minutes, which its authors themselves treat as inflated by a pandemic and by maternal self-report, but its design — four weeks of ordinary messaging, delivered to parents rather than children, at a cost a preschool or clinic could plausibly absorb. That the same trial could not move a screen out of a bedroom is the detail worth holding onto, because it is the one outcome a mother could not soften in a questionnaire. The evidence supports the claim that these programs do something. It does not yet support the claim that they do a lot, and the gap between those two sentences is where most of the advice on the internet currently lives.