Health promoting schools and health promotion in schools: two systematic reviews.

The review examined two broad categories of school-based health promotion interventions:

**Health promoting schools (HPS):** Whole-school approaches that integrate health into the school's ethos, curriculum, environment, and community links. These are multi-component programmes addressing multiple health behaviours (e.g., nutrition, physical activity, mental health, substance use) simultaneously, often with staff training, policy changes, and parental involvement.

**Health promotion in schools (non-HPS):** Single-component or curriculum-only programmes focused on one health behaviour (e.g., a smoking prevention lesson series, a physical activity break programme, or a nutrition education module). These are typically delivered by teachers or external health educators within existing school structures.

**Comparators:** Most studies compared the intervention group to a control group receiving usual school health education (no special programme) or a delayed-intervention control. A few used active controls (e.g., a different health topic).

**Outcome measures:** The review synthesised outcomes across multiple domains:

Smoking initiation and prevalence (self-reported, sometimes validated with cotinine)

Alcohol and drug use (self-reported frequency and quantity)

Physical activity levels (self-reported minutes per week, sometimes pedometer or accelerometer)

Dietary intake (self-reported fruit/vegetable servings, fat intake)

Mental health and well-being (self-reported scales like the Strengths and Difficulties Questionnaire or the General Health Questionnaire)

Body mass index (BMI, measured or self-reported height/weight)

Academic attendance and attainment (school records)

The review included 29 studies (21 randomised controlled trials, 8 quasi-experimental designs) published between 1985 and 1998. The total sample across all studies was approximately 45,000 students, aged 5–18 years, from schools in the UK, USA, Australia, Canada, and several European countries. Most studies focused on secondary school students (ages 11–16), with a smaller number in primary schools (ages 5–11). Settings ranged from urban inner-city schools to suburban and rural schools. Socioeconomic status varied, but many studies targeted disadvantaged communities. No studies exclusively recruited clinical populations (e.g., students with diagnosed health conditions).

The review did not collect original data; it synthesised findings from individual studies. The primary measurement tools used across the included studies were:

**Self-report questionnaires:** Most common. Examples include the Youth Risk Behavior Surveillance System (YRBSS) for smoking, alcohol, and physical activity; the Food Frequency Questionnaire for diet; and the Strengths and Difficulties Questionnaire (SDQ) for mental health. Reliability and validity varied by instrument and age group.

**Biomarkers:** A minority of studies used objective measures. For smoking, some used salivary cotinine (a nicotine metabolite) to validate self-reports. For physical activity, a few used pedometers or accelerometers. For BMI, some studies measured height and weight directly; others relied on self-report.

**School records:** Attendance and academic attainment were extracted from school databases.

**Process measures:** Implementation fidelity was assessed via teacher logs, classroom observations, and student attendance records.

**Study design:** This is a systematic review of 29 studies, including 21 randomised controlled trials (RCTs) and 8 quasi-experimental designs (non-randomised comparison groups). The review followed standard systematic review methodology: a comprehensive search of 12 electronic databases (e.g., MEDLINE, PsycINFO, ERIC, Cochrane Library), hand-searching of 15 key journals, and reference list checking. Two reviewers independently screened titles/abstracts and full texts, extracted data, and assessed study quality using a validated checklist (the Jadad scale for RCTs and a modified version for quasi-experimental studies).

**Randomisation and blinding:** Among the 21 RCTs, randomisation was at the school level (cluster randomisation) in 18 studies, and at the student level in 3 studies. Only 5 RCTs reported adequate allocation concealment (e.g., central randomisation or sealed envelopes). Blinding was rare: 2 studies blinded outcome assessors (e.g., researchers measuring BMI were unaware of group allocation), but no studies blinded participants or teachers due to the nature of the intervention (you cannot hide a school-wide programme). This is a major limitation because participants and implementers knew they were in the intervention group, which can bias self-reported outcomes (e.g., students may over-report healthy behaviours due to social desirability).

**Duration:** Intervention duration ranged from 6 weeks to 3 school years (approximately 30 weeks per year). Follow-up periods ranged from immediate post-test to 2 years post-intervention. Only 4 studies had follow-up beyond 12 months. This is critical because health behaviour change often requires sustained reinforcement; short follow-ups cannot distinguish temporary effects from lasting habit formation.

**Statistical approach:** The review used narrative synthesis (no meta-analysis) because of heterogeneity in outcomes, measurement tools, and study designs. Effect sizes were reported as standardised mean differences (Cohen's d) where possible, or as raw percentage changes. The authors assessed publication bias using funnel plots (asymmetry suggested possible publication bias favouring positive results).

**What this design can and cannot prove:**

**Can prove:** The review can identify patterns of effectiveness across multiple studies, providing stronger evidence than any single study. It can highlight which intervention components (e.g., whole-school vs. curriculum-only) are associated with larger effects, and which populations or settings show the most promise.

**Cannot prove:** Because the review is based on studies with weak blinding, short follow-ups, and self-reported outcomes, it cannot prove that school-based health promotion causes lasting behaviour change. The lack of meta-analysis means we cannot quantify an overall effect size with precision. The review also cannot rule out that observed effects are due to social desirability bias (students reporting what they think the researchers want to hear) rather than actual behaviour change.

**Major methodological weaknesses:**

Only 5 of 29 studies had adequate allocation concealment, increasing risk of selection bias.

Only 2 studies blinded outcome assessors; none blinded participants or teachers.

Only 4 studies had follow-up beyond 12 months; most measured outcomes immediately post-intervention.

Self-report outcomes were used in 27 of 29 studies, with no validation against objective measures in most cases.

Attrition was high (20–40% in many studies), and only 8 studies reported intention-to-treat analyses.

Cluster randomisation was not always accounted for in statistical analyses (only 12 of 18 cluster-RCTs adjusted for clustering), which can inflate Type I error rates.

**Primary outcomes (health behaviours):**

**Smoking initiation:** Among 12 studies measuring smoking, 7 reported statistically significant reductions in smoking initiation in the intervention group compared to controls. Effect sizes ranged from a 5% to 15% absolute reduction in smoking prevalence at follow-up (e.g., from 25% to 20% in one study, p < 0.05). However, 5 studies found no significant difference. The largest effects were seen in whole-school HPS programmes (10–15% reduction) compared to curriculum-only programmes (5–8% reduction).

**Alcohol and drug use:** Among 8 studies, 4 reported significant reductions in alcohol use (e.g., 8–12% fewer students reporting binge drinking in the past month, p < 0.05). For illicit drugs, 3 of 5 studies found significant reductions (e.g., 6–10% lower cannabis use, p < 0.05). Effects were inconsistent across studies, and no clear pattern emerged by programme type.

**Physical activity:** Among 6 studies, 4 reported significant increases in physical activity (e.g., 3–8% more students meeting recommended 60 minutes/day, p < 0.05). Effect sizes were small (Cohen's d = 0.15–0.30). The largest effects were in programmes that included structured physical activity breaks during the school day (not just curriculum lessons).

**Dietary intake:** Among 7 studies, 5 reported significant improvements in fruit and vegetable consumption (e.g., 0.3–0.8 additional servings per day, p < 0.05). Effects on fat intake were mixed (2 of 4 studies found reductions). Whole-school approaches with cafeteria changes and parental involvement showed larger effects than curriculum-only programmes.

**Secondary outcomes (mental health, BMI, attendance):**

**Mental health:** Among 5 studies, 3 reported small but significant improvements in self-reported well-being (e.g., 2–4 point reduction on the SDQ total difficulties score, p < 0.05). Effects were larger in programmes that included social-emotional learning components.

**BMI:** Among 4 studies measuring BMI, only 1 found a significant reduction (mean difference of –0.4 kg/m², p < 0.05). The other 3 found no significant effect.

**School attendance:** Among 3 studies, 2 reported small improvements in attendance (e.g., 2–5% fewer absences, p < 0.05). No studies found significant effects on academic attainment.

**Moderators of effectiveness:**

Programme duration: Programmes lasting ≥1 school year showed larger effects than shorter programmes (e.g., 10–15% vs. 3–8% reduction in smoking initiation).

Implementation fidelity: Studies with high fidelity (≥80% of planned sessions delivered) showed larger effects than those with low fidelity.

Age: Programmes starting in primary school (ages 5–11) showed larger effects on smoking and alcohol use than those starting in secondary school (ages 11–16).

Socioeconomic status: Effects were similar across socioeconomic groups, but disadvantaged schools showed higher attrition and lower fidelity.

The effects found in this review are best described as **small to moderate** in size. For context:

A 5–15% reduction in smoking initiation means that if 25% of students in a control group start smoking, the intervention group would see 10–20% starting — a difference of 5–15 percentage points. This is roughly equivalent to the effect of a 50% increase in cigarette taxes (based on other research at the time).

A 0.3–0.8 serving increase in fruit/vegetable intake per day is about one extra apple or half a cup of broccoli per day. This is a modest change — equivalent to adding one piece of fruit to a lunchbox.

A 3–8% increase in physical activity means that if 40% of students meet recommended activity levels at baseline, the intervention group would see 43–48% meeting them — a small shift.

The mental health improvements (2–4 points on the SDQ) are roughly equivalent to the difference between a child with no mental health concerns and one with mild difficulties — noticeable but not clinically significant for most.

These effects are **not large enough** to single-handedly solve population-level health problems, but they are **meaningful at scale** — if implemented across thousands of schools, even a 5% reduction in smoking initiation could prevent tens of thousands of young people from starting to smoke.

**What the authors acknowledge:**

High heterogeneity across studies (different interventions, outcomes, populations) prevented meta-analysis.

Most studies had short follow-up (≤12 months), so long-term effects are unknown.

Self-report bias is a major concern, especially for socially undesirable behaviours like smoking and drug use.

Only a minority of studies used objective measures (e.g., cotinine, accelerometers).

Publication bias was suggested by funnel plot asymmetry — studies with null or negative results may be missing.

Many studies had high attrition (20–40%), and few used intention-to-treat analysis, which can overestimate effects.

Cluster randomisation was not always properly accounted for, inflating Type I error rates.

**Additional limitations a critical reader would note:**

The review is from 1999, so it does not include modern programmes (e.g., digital interventions, mindfulness-based approaches). The evidence base is now 25+ years old.

The quality of included studies was generally low: only 5 of 29 had adequate allocation concealment, and only 2 blinded outcome assessors.

The review did not formally assess risk of bias using modern tools (e.g., Cochrane Risk of Bias tool), which may have revealed more serious flaws.

The focus on school-based programmes means results may not generalise to out-of-school settings (e.g., community centres, family-based interventions).

The review did not examine cost-effectiveness, which is critical for real-world implementation.

No studies examined potential harms (e.g., increased anxiety from health messaging, stigmatisation of overweight students).

For someone running their own n=1 experiment (e.g., a parent, teacher, or school administrator testing a health promotion programme in their own school or family):

### What to test

**Specific intervention:** A whole-school approach (HPS) that combines curriculum lessons (e.g., 10–15 sessions on nutrition and physical activity), environmental changes (e.g., healthier cafeteria options, active breaks during class), and parental involvement (e.g., newsletters, family activity challenges). This is the most evidence-based approach from the review.

**Dose:** Run the programme for at least one full school year (30–40 weeks). Shorter programmes (6–12 weeks) showed smaller and less durable effects.

**Comparator:** Compare to a control group receiving usual health education (e.g., standard curriculum without extra components). If you cannot randomise, use a matched comparison school or a delayed-intervention design.

### Minimum meaningful duration

**Intervention:** At least 1 school year (30–40 weeks). The review found that programmes lasting ≥1 year had 2–3 times larger effects than shorter programmes.

**Follow-up:** Measure outcomes immediately post-intervention and again 6–12 months later. The review found that effects often decayed after 12 months, so long-term follow-up is critical to assess lasting change.

### What to measure

**Primary outcomes:** Smoking initiation (self-reported, ideally validated with cotinine if possible), physical activity (minutes per week, measured by pedometer or accelerometer if available), fruit/vegetable intake (servings per day, using a validated food frequency questionnaire).

**Secondary outcomes:** Mental well-being (SDQ or General Health Questionnaire), BMI (measured height/weight), school attendance (from school records).

**Process measures:** Implementation fidelity (e.g., percentage of planned sessions delivered, teacher logs), student engagement (e.g., attendance at programme sessions), and parental involvement (e.g., number of newsletters sent, family activity logs returned).

### Key confounds to control for

**Social desirability bias:** Students may over-report healthy behaviours. Use objective measures where possible (e.g., pedometers for activity, cotinine for smoking). If using self-report, include a social desirability scale (e.g., the Marlowe-Crowne scale) to adjust for bias.

**Seasonal effects:** Physical activity and diet vary by season (e.g., more outdoor activity in summer, more holiday treats in winter). Run the programme across full school years to average out seasonal variation.

**School-level factors:** School size, socioeconomic status, and existing health policies can confound results. If you cannot randomise, match schools on these variables or use statistical adjustment.

**Teacher effects:** Teacher enthusiasm and fidelity vary widely. Train all teachers uniformly, and monitor fidelity with logs and observations.

**Maturation:** Students naturally change their behaviours as they age (e.g., smoking initiation increases in adolescence). Use a control group to separate programme effects from maturation.

**Attrition:** Students may drop out of the programme or the study. Track reasons for dropout, and use intention-to-treat analysis (analyse all participants as randomised, regardless of completion).

### What a positive result would look like

**Smoking:** A 5–15% absolute reduction in smoking initiation (e.g., from 25% to 15–20% of students reporting any smoking in the past