NIHR Public Health Review Team

First Episode Psychosis (FEP) is a term used to describe the first time a person experiences symptoms like hallucinations or delusions severe enough to affect how they think, see reality, and cope with everyday life.

Although FEP is most common in late adolescence and early adulthood, it can occur at any age. Despite advances in detection and early intervention, the evidence base for prevention remains fragmented. As it stands, most reviews have focused on clinical high-risk samples only, with interventions aimed at youth who already show early warning signs prior to the onset of FEP.

Longitudinal research (studies that follow participants over time) has identified multiple modifiable factors, including trauma, substance use, discrimination, and social disadvantage. However, these have not been brought together systematically.

We will address these evidence gaps by examining preventive interventions and modifiable risk factors (things that can be changed via interventions such as substance misuse) for FEP.

Review questions

The protocol for this work is available at: https://www.crd.york.ac.uk/PROSPERO/view/CRD420261309356

Background

Modern slavery is an umbrella term that refers to human trafficking and exploitation. It is a serious issue, that is often under-reported. Up to 1.8 per 1,000 people may be impacted in the UK each year. There are many types of exploitation that people might experience, including labour exploitation, forced servitude, criminal exploitation and sexual exploitation.

In recent years policymakers, researchers and practitioners have started to adopt a public health approach to tackling modern slavery. This involves investment in programmes to prevent exploitation and trafficking, along with re-exploitation and re-trafficking.

Examples of programmes might include community awareness raising, improving reporting processes, and working with organisations that work with people at risk of exploitation.

However, it is not clear what public health policies and services exist to prevent exploitation and trafficking, and what research has been undertaken to find out if they work. We conducted a scoping review to systematically map public health interventions targeting modern slavery to find out:

1. What types of interventions target modern slavery?
2. How are these interventions meant to bring about change?
3. Who are the intended participants for  these interventions?
4. What outcomes are these interventions hoping to change?
5. What types of evidence has been generated to evaluate these interventions?

 Incorporating lived experience

To help us understand more about the priorities of people affected by modern slavery, and contextualise the research evidence we found, we worked with the charity Unseen and a group of survivor consultants with lived experience of modern slavery throughout the review. We held a series of workshops, creating opportunities for survivor consultants to shape the research at different stages.

The first workshop introduced the review and worked to build a shared understanding of what a public health approach to modern slavery means to survivor consultants. In the second, we shared initial findings from the review through a series of vignettes, which survivor consultants responded to, contextualised, and expanded on using their own knowledge and experiences.

In the third workshop, survivor consultants worked together to develop recommendations for national government, local government, charities and researchers, drawing on both the review findings and their lived experience.

Finally, we worked with survivor consultants to develop creative outputs for disseminating the research. Using excerpts from studies included in the review, we arranged words and phrases into lines of poetry, which were then collectively developed into a full poem.

What we found

After searching a range of evidence sources we included a total of 83 programmes (total 90 evaluations) in our review.

Most evaluations were conducted in England (28). The majority of programmes targeted modern slavery among children and young people aged under 18 years. Sexual exploitation was the most frequently targeted type of modern slavery.

Most programmes worked by trying to improve literacy (51). For example, raising knowledge and awareness of sexual exploitation among shopkeepers who work at night and might be able to spot young people at risk and report them. Other ways of helping people were:

Improving working between organisations (35)

Improving access to housing, legal advice and education (35)

Improving survivors feeling of power and control (34)

Deterring perpetrators and disrupting opportunities (26)

Most (32) programmes were delivered at the service level, providing support to those at risk as opposed to, for example, national policies.  

There were not many evaluations of interventions. Only three assessed if programmes actually worked to prevent modern slavery. Most looked at whether the people involved liked it or found it easy to deliver.

Recommendations

Practice

• Ensure delivery of interventions that target different  populations and types of modern slavery, such as labour exploitation, forced marriage and forced servitude.
• Make use of evidence-based interventions where available and help to produce evidence for current practice where feasible.

Policy

• Work to deliver a public health approach to modern slavery, with a particular focus on developing national-level approaches.
• Invest in interventions that target different populations and types of modern slavery.
• Invest in building an evidence-base for interventions across public health domains.

Research

• Develop, refine and adapt interventions that have a clear rationale for why they should work and have measurable outcomes. 
• Conduct robust outcome evaluations of interventions to establish if they are effective in changing outcomes.

Sharing our findings

The protocol can be found here.

Background

Public health aims to protect and improve the health of communities and populations at local, regional, national, and global level, but typically suffers from limited resources. Artificial intelligence (AI) can help automate tasks that may otherwise have needed human input. It has the potential to support many aspects of public health including intervention delivery. But, as with any new technology, the risks need to be considered before it is implemented more widely.

What we did

We conducted a total of five reviews. Our aim was to map the current landscape, visualise future uses, and balance the potential of AI in public health against the potential risks:

1. Three systematic reviews to find evidence for how effective AI-assisted interventions have been for reducing alcohol drinking, smoking, and physical inactivity

For these reviews, we looked specifically for randomised controlled trials (RCTs) to see how effective AI-assisted interventions could be at helping people to reduce drinking alcohol or smoking, or to increase physical activity.

2. A scoping review to see how AI is currently being used by public health professionals

For this review, we looked for published articles where the authors talked about having used AI in their work as public health professionals. The authors didn’t need to specify how effective their use of AI ultimately was, just that AI was used.

3. A scoping review to bring together all suggested potential uses and risks of AI in public health

For this review, we looked at published articles where the authors talked about how AI could be used in public health, and the associated risks of doing so.

What we found

For past use of AI, we found no strong evidence of a beneficial effect of AI-assistance in public health interventions for reducing alcohol consumption (four trials), smoking (12 trials), or physical inactivity (17 trials) in high-income countries. This held true both for AI that was added to existing interventions, or when comparing AI-assisted interventions to no intervention. This is not necessarily to say that such interventions may not be effective: mostly, this represents a lack of evidence, particularly high-quality evidence from which we could draw strong conclusions. We found no evidence at all for interventions that used large language models.

For the present use of AI, we found a total of eight reports of public health professionals using AI in practice. While assessing the effectiveness of the AI tools was beyond the scope of this review, the evidence from included reports was mixed. One memorable quote for a chatbot used on the WHO website was: “S.A.R.A.H. [the chatbot] had all the ingredients to serve not only as a useless tool in the quest to improve the information ecosystem and reduce the dangers infodemics pose to public health but also as a potential threat to erode trust in the WHO itself".

For future uses of AI, we found 67 reports that talked about potential uses of AI and their associated risks. In total, we identified 44 potential uses of AI across 8 public health contexts, including in communication (18 uses), education (3 uses), emergencies (10 uses), evidence reviews (8 uses), forecasting (12 uses), interventions (19 uses), public health organisations (12 uses), and surveillance (2 uses). We also identified a substantial number of risks, including bias and generalisability, which could lead to health inequalities, discrimination, reduction in public trust, inaccurate or unreliable predictions, unintended consequences, and ineffective interventions. These potential risks add weight to the view that understanding the potential risks of an intervention before adoption is necessary even for non-clinical interventions.

Overall, there was little evidence to support the use of AI in public health, either as behaviour change interventions, or in public health practice. Ultimately, there needs to be strategic and careful thought and planning, including workforce training and support, to prevent the adoption of AI tools that are ineffective, embed inequity, or reduce trust in a public health organisation.

Sharing our findings

Protocols for these reviews are hosted on Prospero (CRD42025642317, CRD42025642334, and CRD42025642339) and Zenodo (15971109 and 15971614)

The role of Artificial Intelligence in interventions to reduce alcohol consumption and smoking: A systematic review

The role of Artificial Intelligence in interventions to reduce physical inactivity: A systematic review

The current roles of Artificial Intelligence in public health: A scoping review

The potential roles and risks of Artificial Intelligence in public health: A scoping review

Background

Young people face increased health risks, including substance misuse, mental ill-health, and sexual health concerns. These can and often do co-occur. Having depression or anxiety may affect sexual health decision making and lead to sexually transmitted infections (STIs) or make substance misuse more likely. In turn, factors such as peer pressure, stress and lack of support are also influential.

Such challenges tend to be more pronounced for sexual and ethnic minoritised, cared for and care experienced, disabled and disadvantaged young people.

Digital public health interventions are seen as a promising solution to address these challenges by improving accessibility, efficiency, and personalisation of care. However, robust evidence on their effectiveness for underserved populations remains limited.

We set out to systematically review the evidence on digital public health interventions tailored to young people, particularly those from underserved groups, focusing on mental health, sexual health, and substance use outcomes. We looked specifically at interventions delivered through digital technologies such as text messages, mobile phone apps and websites.

What we did

We wanted to find out:

• What is the quantity, quality, and strength of evidence for digital public health interventions tailored to underserved groups?
• How are digital public health interventions tailored to underserved groups? And what is the evidence of effectiveness for them?
• Is there evidence for differential effectiveness of digital public health interventions by underserved groups?

We wrote three corresponding papers:

• Digital public health interventions targeting young people's mental health, substance use and sexual health: systematic review of reviews and equity analysis.
• Inclusion of underserved young people in digital public health interventions for mental health, sexual health and substance use. A systematic review and intervention component analysis.
• Inclusion of underserved young people in digital public health interventions for mental health, sexual health and substance use: Systematic review and equity synthesis of moderation analyses.

What we found

We found mixed evidence for the effectiveness of digital interventions, and inconsistent consideration of equity impacts across diverse groups of young people. As a result, the promise of digital approaches is constrained by uncertainty about their ability to support underserved populations and address needs equitably. Further research is needed to understand their impact more fully and to ensure they do not inadvertently widen existing health inequalities.

Further details about each review:

• The first review included 75 reviews. It found that digital public health interventions were effective for mental health outcomes, most substance use outcomes, but evidence was mixed for sexual health outcomes. Findings were limited by review quality, variation across the reviews (heterogeneity) that limited direct comparison, and inconsistent reporting of equity characteristics.
• In the second review, we identified twelve intervention components that were consistent to tailored mental health, substance use and sexual health interventions. Four related to how young people were involved in the design of interventions. Six related to how interventions worked to engage young people and to turn knowledge into action. Evidence of effectiveness was strongest for substance use interventions, was mixed for sexual heath and limited for mental health.
• In our third review analysis, we found that digital public health interventions generally produced inconsistent and non-significant moderation effects, and that evidence is lacking for many important equity-relevant characteristics (such as socioeconomic status, education, sexual orientation, and disability). Therefore, it remains uncertain how well these interventions are able to equitably address mental health, substance use, and sexual health challenges in young people.

Sharing our findings

• Protocols for the reviews are hosted on PROSPERO: CRD42025641364
• Digital public health interventions targeting young people's mental health, substance use and sexual health: systematic review of reviews and equity analysis. (forthcoming)
• Inclusion of underserved young people in digital public health interventions for mental health, sexual health and substance use. A systematic review and intervention component analysis. (forthcoming)
• Inclusion of underserved young people in digital public health interventions for mental health, sexual health and substance use: Systematic review and equity synthesis of moderation analyses. (forthcoming)

Background

The UK has low breastfeeding rates with substantial variation both regionally and socio-demographically. Whilst some women make an informed decision not to breastfeed, many stop before they intended. 

Pain and latching problems are key reasons, but there are other factors. These include mothers’ psychological adjustment and systemic barriers.

Breastfeeding support offered in the community can potentially help women navigate these. This includes support delivered by women with similar social or cultural backgrounds (peer support) as well as non-hospital-based healthcare professionals, for example health visitors.

However, existing evidence suggests that some women may benefit more from this type of support than others.

For this project we looked at research evidence to help understand how peer and community support services could be delivered to make sure that more women have a positive breastfeeding experience and are encouraged and supported to breastfeed if they want to.

What we did

We conducted two reviews:

1. Peer support and community interventions targeting breastfeeding in the UK: systematic review and equity synthesis of qualitative evidence

This drew together qualitative evidence from the UK to understand how the characteristics of mothers shaped their experiences of peer and community support.

2. Equity-focused peer support and community interventions for breastfeeding in high-income countries: Systematic review and intervention component analysis

This looked at studies of peer and community support in higher-income countries to understand what activities (components) are delivered during breastfeeding peer support, and if they prevent inequities in breastfeeding rates.

What we found

Peer support and community interventions targeting breastfeeding in the UK: systematic review and equity synthesis of qualitative evidence

Fifty-five studies were included in this review. Mothers from lower socio-economic backgrounds and black and minority ethnic communities tend to have poorer experiences of support.

There were different experiences across the different phases of receiving support, which may lead to inequities in breastfeeding rates, these include:

• Inadequate consideration of mothers’ social, economic and cultural background, and how this might shape their needs (e.g. not comfortable breastfeeding in public spaces).
• Lack of consideration of different physical characteristics (e.g. body type, disability).
• Structural barriers (e.g. lack of community acceptance, unsupportive workplace policies) that might discourage breastfeeding

Equity-focused peer support and community interventions for breastfeeding in high-income countries: Systematic review and intervention component analysis

Thirty-one studies were included in this review. It found that peer and community support tend to use four different types of principles when working with underserved communities to prevent inequities:

• Crosscutting principles: Support underpinned by principle of continued availability throughout key parenting transitions, while keeping a holistic focus on mothers’ health and wellbeing.
• Contextual fit: Support providers with similar cultural, social and economic background to mothers; support to be delivered in a suitable place (e.g. home), through appropriate mode (e.g. face-to-face or video call); and include people from the target population in intervention design.
• Delivery: Information and materials accessible to varying cultural and language needs; support providers build relationships with mothers; and offer practical breastfeeding resources (e.g. practical devices such as nipple shields, breast pump, nipple cream, nursing bras).
• Wayfinding: Support for mothers throughout the breastfeeding journey by strengthening support in existing social network (e.g. family involvement and support); creating new networks of breastfeeding support through a community asset approach; offering activities and resources to support breastfeeding in conjunction with returning to work or education (e.g. addressing concerns about breastfeeding in public).

Whilst mothers who received interventions were around 10% less likely to stop breastfeeding up to one year, there was no conclusive evidence that any intervention components were more effective than others in targeting breastfeeding rates among underserved communities.

Sharing our findings

• Protocols for the reviews are hosted on PROSPERO: CRD42024537108 and  CRD42024589555

Publications:

• Peer support and community interventions targeting breastfeeding in the UK: systematic review and equity synthesis of qualitative evidence
• Equity-focused peer support and community interventions for breastfeeding in high-income countries: Systematic review and intervention component analysis (forthcoming)
• Blog posts discussing this work is available here and here
• A briefing paper can be found here
• A policy briefing paper can be found here