Content analysis of psychological first aid training manuals via topic modelling

ABSTRACT

Background: Psychological First Aid (PFA) is practiced worldwide. This practice in English is guided through a small collection of training manuals. Despite ubiquitous practice and formal training materials, little is known about what topics are covered and in what depth in these influential manuals. As such, we analyzed the topic structure of these training manuals.

Objective: To model the PFA manuals’ topics with the goal of identifying a set of topics with recurrent themes and evaluating the extent to which each manual demonstrated those themes.

Method: This machine learning study employed an unsupervised topic modelling design using Latent Dirichlet Allocation. The variables are (1) the distribution of a word across documents and (2) the distribution of a word across topics. The level of measurement for all variables is continuous. The unit of analysis is words. Preprocessing and data analysis were carried out using the Orange Data Mining Toolbox (Demšar et al., 2013). This programme is a Python GUI.

Results: Results indicated a ten-topic structure to the universe of the English PFA training manuals. These topics were: (1) Refugees, (2) Orientation Activities, (3) Community-Based Applications, (4) PTSD & Other Psychological Issues, (5) Training Materials, (6) Specific Helper Instructions, (7) PFA Scholarship, (8) MHPSS, (9) General Curriculum, and (10) Australian Specific Delivery. The depth of discourse on each topic varied widely between manuals.

Conclusions: The Academics of the PFA topic shows a strong representation of the corpus and suggests current training manuals have stayed true to its evidence-supported practice. The topic of Community-Based Applications strongly represents the corpus and suggests that training models incorporate community-based applications. The scientific foundation and practical implementation of the training guides are essential elements. Limitations and implications were also discussed.

HIGHLIGHTS

• Little is known about what topics are covered and in what depth in the influential PFA English manuals.

• We conducted a topic modelling study using Latent Dirichlet Allocation, aiming to discover a set of topics with recurring themes and analyze the degree to which each manual exhibited those topics.

• Results indicated a 10-topic structure to the universe of the English PFA training manuals.

• The training manuals’ scientific basis and practical application are key components, while notable gaps presented.

Antecedentes: Los Primeros Auxilios Psicológicos (PAP) son practicados en todo el mundo. Esta práctica en Inglés está guiada a través de una pequeña colección de manuales de formación. A pesar de la práctica omnipresente y los materiales de capacitación formales, se sabe poco sobre qué temas son tratados, y con qué profundidad, en estos manuales influyentes. De esta manera, analizamos la estructura temática de estos manuales de capacitación.

Objetivo: Modelar los temas de los manuales de PAP con el objetivo de identificar un conjunto de temáticas con temas recurrentes y evaluar en qué medida cada manual demostró tales temas.

Método: Este estudio de aprendizaje automático empleó un diseño de Modelado no supervisado de temas utilizando la asignación de Dirichlet latente. Las variables son (1) la distribución de una palabra en los documentos y (2) la distribución de una palabra en los temas. El nivel de medición para todas las variables es continuo. La unidad de análisis son las palabras. El pre-procesamiento y el análisis de datos se llevaron a cabo utilizando el software de minería de datos Orange (Demšar et al., 2013). Este programa es una interfaz gráfica de Python.

Resultados: Los resultados indicaron una estructura de diez temáticas para el universo de los manuales de capacitación de PAP en inglés. Estos temas fueron: (1) Refugiados, (2) Actividades de orientación, (3) Aplicaciones basadas en la comunidad, (4) TEPT y otros problemas psicológicos, (5) Materiales de capacitación, (6) Instrucciones específicas para ayudantes, (7) Becas PAP, (8) salud mental y el apoyo psicosocial (SMASS), (9) Plan de estudios general y (10) Entrega específica de Australia. La profundidad del discurso sobre cada tema varió ampliamente entre los manuales.

Conclusiones: La Academia de los PAP muestra una fuerte representación del corpus y sugiere que los manuales de capacitación actuales se han mantenido fieles a su práctica respaldada por evidencia. El tema de las aplicaciones basadas en la comunidad representa fuertemente el corpus y sugiere que los modelos de capacitación incorporan aplicaciones basadas en la comunidad. La fundamentación científica y la implementación práctica de las guías de formación son elementos esenciales. También se discutieron las limitaciones y las implicaciones.

背景: 心理急救 (PFA) 在全球廣泛實施。這種英語實行是靠一小部分培訓 手冊來指導。儘管實行和訓練教材無處不在，但這些有影響力的手冊涵蓋了哪些主題以及深度知之甚少。

目的: 使用 Latent Dirichlet Allocation 對 心理急救手冊進行主題建模研究，旨在發現一組具有重複主題的題材 ，並分析每本手冊展示這些主題的程度。

方法: 本機器學習研究採用了無監督主題建構設計。變量是 (1) 一個詞在文檔中的分佈和 (2) 一個詞在主題中的分佈。我們使用橘子數據挖掘工具箱 (Demšar et al., 2013) 進行預處理和數據分析。這個程序是一個 Python 圖形用戶界面。

結果: 結果表明英語心理急救 培訓手冊的整體結構有十個主題。這些主題是：(1) 難民，(2) 導引活動，(3)社區為主的應用，(4) 創傷創傷壓力症候群和其他心理問題，(5) 培訓材料，(6) 助人者的特定指導，(7) PFA 學術研究， (8) 心理健康和社會心理支持，(9) 一般課程，以及 (10) 澳大利亞特定教導。不同手冊對每個主題的討論深度差異很大。

結論: 心理急救學術主題顯示了語料庫的強大代表性，表明當前的培訓手冊一直忠實於其循證實踐。基於社區的應用主題強烈代表了語料庫，表明訓練模型包含社區的應用實施。培訓手冊的科學依據和實際應用是關鍵組成部分。此外還討論了局限性和影響。

KEYWORDS:

• Psychological First Aid
• PFA
• multiculturalism
• latent Dirichlet allocation
• LDA
• artificial intelligence
• topic modelling
• LDAvis
• Gensim
• Orange Data Mining

PALABRAS CLAVE:

• Primeros auxilios psicológicos; PAP
• multiculturalismo
• asignación de Dirichlet latente
• LDA
• inteligencia artificial
• modelado de temas
• LDAvis
• Gensim
• Orange Data Mining

关键词:

• 心理急救
• PFA
• 多元文化主义
• 潜在狄利克雷分配
• LDA
• 人工智能
• 自然语言处理
• 主题模型
• 语料库语言学
• LDAvis
• Gensim
• 橙色数据挖掘

Open science information

Given the study's exploratory nature, preregistration of the research questions was not done. Because of copyright issues with the PFA manuals, the corpus could not be shared. All other data will be placed on this study’s Open Science Foundation (OSF) research project webpage: https://osf.io/u7jr2/. All study materials (e.g. stop word list) will also be placed on this webpage.

Psychological first aid (PFA) is a widely used tool by trained providers or first responders in administering early interventions to those who have undergone disasters or traumatic experiences during calamities, pandemics, or war. Over the past decade, numerous PFA training manuals have been developed, emphasizing different components or populations. These manuals typically include modules with core components and actions to provide clear step-by-step instructions with easy implementation. However, it can be overwhelming when training organizations must decide which PFA manual to use among the large pool of possibilities. The purpose of this study is to bring some clarity and organization to this selection.

The goal of the present research effort was twofold. The first goal was to address a major gap in the PFA literature. Despite robust PFA training activities worldwide using a variety of manuals, no systematic knowledge of the universe of topics covered in these manuals exists. Without such knowledge, training organizations have no sense of the comprehensiveness of each manual. Moreover, assessing possible missing elements in the present PFA training manuals remains impossible. The second goal is to potentially expand current training practices by increasing trainers’ awareness of the content of the manuals from which they can choose. The hope is that trainers will then be able to make selections that best fit the needs of trainees in their localities.

Examining the research literature related to potential variables for this study revealed five areas for consideration. These areas were: (1) the definition of PFA; (2) the reason for PFA training, (3) current knowledge of topics covered in PFA training; (4) the reason for topic modelling PFA manuals, and (5) unsupervised machine learning approaches to topic modelling. After examining these areas, the research questions that guided the present study will be detailed.

Understanding the structure of the PFA research field allows for critical reflection and identifying blind spots for advancing the field.

PFA refers to evidence-supported interventions by non-mental health professionals and first responders to assist disaster victims in regaining stability (Ruzek et al., 2007). It includes providing information, comfort, emotional care, and instrumental support to those experiencing disaster-related stress (Forbes et al., 2011; Fox et al., 2012). Previous research has demonstrated the effectiveness of specific essential components of PFA (Hobfoll et al., 2021). In particular, a worldwide expert panel has identified five empirically supported interventions for individuals exposed to mass trauma in the early to mid-term stages, including promoting a sense of safety, calming, self and community efficacy, connectedness, and hope (Hobfoll et al., 2021).

PFA is a response to the need for laypeople to be trained to offer psychological aid to people in circumstances where the need for such assistance considerably outstrips the supply of mental health professionals. Thus, there is a dual need to train people with no prior knowledge of mental health and to ensure that these laypersons are competent enough to avoid causing further harm. Without such training, a PFA layperson could, with the best of intentions, pursue activities injurious to victims. For instance, elicit detailed trauma narratives that are best left to a mental health professional trained in exposure therapy (e.g. Foa, 2011). As such, sound and comprehensive PFA manuals are essential to effective PFA work.

The term ‘Psychological First Aid’ (PFA) first appeared during World War II as an intervention for merchant marine sailors (Blain et al., 1945). Since then, PFA has been a reliable tool for working with people stressed by disasters, armed conflict, or other traumatic events. It has supported individuals and communities in coping with fear and anxiety during public health crises (e.g. SARS, Ebola, COVID-19) (Cénat et al., 2020; Grubin et al., 2022a; WHO, 2014). The multiple settings of PFA include schools, workplaces, mass disaster venues, and circumscribed critical incidents such as armed conflict, floods, fires, accidents, and other traumatic events (Allen et al., 2010; Hechanova et al., 2015; Schafer et al., 2016). The application of psychological first aid in situations of armed conflict continues to the present day in places as diverse as Syria (El-Khani et al., 2016) and Ukraine (Prykhodko et al., 2021). Recently, a digital psychological first aid chatbot named ‘Друг. Перша допомога’ (‘Friend. First Aid’) was developed in Ukraine to support adults and children and cope with the stress of the war based on the users’ environment (Frankova et al., 2022).

Several models of PFA exist, but their effectiveness varies. A recent randomized controlled trial found that PFA reduced distress immediately after traumatic events or disasters (Figueroa et al., 2022). However, earlier systematic reviews (Dieltjens et al., 2014; Fox et al., 2012; Shultz & Forbes, 2014) highlighted the lack of evidence-based data for developing guidelines and training procedures for PFA. Fox et al. (2012) specifically examined the PFA evidence from 2000 to 2010 and concluded that while many experts supported PFA, there was insufficient evidence to validate it. Some important topics may be missing from PFA manuals, which could explain the inconsistency of evidence (Hermosilla et al., 2022; Schoultz et al., 2022).

The practice of content analysis frequently results in discussions about what the content should be. Blain et al.'s (1945) seven curriculum components and the National Child Traumatic Stress Network's (Brymer et al., 2006) eight fundamental actions impacted us. These 15 subjects are listed on the webpage for this research project: https://osf.io/u7jr2/. After carefully reading these documents, it became clear to us that we could expect discussions on the following subjects: (1) psychological foundations (developmental and abnormal), (2) communication skills, (3) PFA interventions, (4) training techniques, (5) adaptations to different settings, and (6) prevention methods. However, we chose to use an unsupervised topic modelling technique rather than build off a priori notions. In other words, preconceived shoulds did not guide this topical exploration.

An explosion of textual material available to professionals has marked the past decade. The field of PFA training is no exception. At present, PFA trainers have no less than 14 general manuals and an additional surfeit of special application manuals. What to do with such a situation where the amount of textual material far exceeds any professional’s ability to read, assess, and integrate it? Professionals in allied fields such as psychology (e.g. session content), public health (e.g. health attitudes), education (e.g. technology use), and psychiatry (e.g. patient information) have used topic modelling to penetrate vast walls of data to pull out meaningful information (Hwang et al., 2023; Kleine et al., 2023; Li et al., 2023; Lin et al., 2023). The present topic modelling study of PFA manuals aims to chunk the textual data into a form that trainers and future manual authors can use.

The preferred approach to understanding the topical structure of any field is the unsupervised machine learning approach to topic modelling called Latent Dirichlet allocation (LDA). Blei et al. (2003) were the first to use LDA to discover topics in a group of documents and automatically classify each document according to its relevance to each identified topic. Blei et al. (2003) described LDA as a generative probabilistic model for collections of discrete data, such as text corpora. LDA is a three-level hierarchical Bayesian model in which a collection item is modelled as a finite mixture over an underlying set of topics. Each topic is, in turn, modelled as an infinite mixture over an underlying set of topic probabilities. In text modelling, topic probabilities clearly show what a document is about (Blei et al., 2003).

Two research questions were created to direct this study considering the aforespecified need to condense the enormous amount of textual data present in PFA manuals to a level suitable for training management and design:

RQ1: What is the topic structure of the Psychological First Aid training manuals?

RQ2: How do the level of topics vary by Psychological First Aid training manuals?

1. Method

1.1. Design

This machine-learning study employed an unsupervised topic modelling design (Blei, 2012b). From an algorithmic perspective, topics are conceptualized as patterns of word co-occurrences in a set of documents (Rüdiger et al., 2022). Topic modelling of textual data is similar to the clustering of numeric data (Almuzaini & Azmi, 2022). The variables are (1) the distribution of a word across documents and (2) the distribution of a word across topics. The level of measurement for all variables is continuous. The unit of analysis is words. Given the published and public nature of the data, no human subjects review was needed.

1.2. Corpus

1.2.1. Register and inclusion criteria

The register was academic prose, and the subregister was psychological training manuals. The inclusion criteria for the PFA training manual corpus were as follows:

1. Published in English.

2. Appears in a Google search with the term ‘psychological first aid’ along with one or more of the following terms: ‘guide’, ‘manual’, ‘handbook’, and ‘frontline’.

3. In the case of multiple editions, the latest edition only.

4. Not a close derivative of another manual (e.g. Brymer et al., 2006 but not Brymer et al., 2012).

Using these criteria, we found 14 manuals for inclusion in the corpus. This corpus contained 392,633 words and 13,160 individual word types. These 14 manuals are marked with an asterisk in the reference section. Australian Red Cross and Australian Psychological Society (2013); Center for Public Health Preparedness, School of Public Health, University at Albany (2017); Dybdal et al. (2013); Everly and Lating (2017); Jacobs (2016); Mayer and Lurie (2022); Ministry of Labour and Social Welfare & UNICEF Lao People's Democratic Republic (2021); NSW Health & University of Western Sydney (2012); and Singaravelu (2013).

1.2.2. Preprocessing

Preprocessing included removing function words that add no meaning to the text and lemmatization. A lemma is the basic form of a word and consists of all forms of the word, such as the lemma ‘go’; which also includes goes, went, going, and gone (Kilgarriff et al., 2022). A complete listing of preprocessing can be reviewed on this study’s webpage (https://osf.io/u7jr2/).

1.3. Measures

1.3.1. Topic modeling

Topic modelling is a form of analysis involving a suite of algorithms to discover the hidden thematic structure in extensive collections of texts (Blei, 2012a). This analysis operated with two goals in mind: (1) discovering a set of topics defined as recurring themes discussed in the corpus and (2) the degree to which each document exhibits those topics (Blei, 2012b).

1.3.2. Heatmap

A heatmap shows values for the main variable of interest across two axis variables as a grid of coloured squares, where each cell’s colour represents the value of the main variable in the corresponding cell range (Yi, 2022). In the case of this study, the main variable is the topic, and the other variable is the manual.

1.3.3. Log perplexity

Log perplexity measures how probable some new unseen data is given the model learned earlier (Kapadia, 2019).

1.3.4. Topic coherence

Topic Coherence represents the degree of semantic similarity between high-scoring words in the topic to distinguish between semantically interpretable topics and statistical artifacts (Kapadia, 2019).

1.3.5. k

k is the symbol in LDA analyses for the number of topics. LDA permits the researcher to experiment with multiple numbers of topics. The goal is to find a k that minimizes perplexity and maximizes coherence (Dua, 2021).

1.3.6. Lift

Lift is the ratio of (1) a term’s probability within a topic to (2) its marginal probability across the corpus (Sievert & Shirley, 2014).

1.3.7. Topic Relevance (λ)

Relevance (λ) can be defined as the weight given to the probability of term w under topic k relative to its lift (Sievert & Shirley, 2014).

1.4. Apparatus

Preprocessing and data analysis were carried out using the Orange Data Mining Toolbox (Demšar et al., 2013). This programme is a Python GUI. For topic modelling, it is a GUI for the Python programme Gensim (Řehůřek & Sojka, 2019). The Orange topic modelling widget was used for topic modelling with the Latent Dirichlet Allocation option selected. The results will be visualized for topic interpretation using Orange’s LDAvis widget and MDS (Multidimensional Scaling) widgets. The LDAvis widget provides the right panel of Sievert and Shirley’s original LDA two-panel data visualization programme (Sievert, 2018). These two noted the following concerning this panel: (a) the horizontal bar charts with bars representing the individual terms are the most useful for interpreting a selected topic, and (b) A pair of overlaid bars represent both the corpus-wide frequency of a given term as well as the topic-specific frequency (Sievert & Shirley, 2015). The Multidimensional Scaling (MDS) widget is Orange’s operationalization of the left side of the original LDAvis programme. Concerning this left panel, the two noted that it produces a data visualization project that contains the following: (a) plot the topics as circles in the two-dimensional plane whose centres are determined by computing the distance between topics, and (b) encodes each topic’s overall prevalence using the areas of the circles (Sievert & Shirley, 2015). Basic word counts, bigram counts, and Key Word in Context (KWIC) lines were produced using Antconc (Anthony, 2022).

1.5. Data analysis

In RQ1, topic modelling will be accomplished with Latent Dirichlet Allocation. The k selected for analysis will be between three and 15, representing the best mix of low log perplexity and high topic coherence. We will use the LDAvis and MDS widgets to assist with topic interpretation and report their results. For the LDAvis, the relevance setting used will be the one found to be optimal by Sievert and Shirley (2015; λ = .6). Regarding RQ2, a heatmap of the topics and manuals will be produced.

2. Results

With the RQ1, a 10-topic solution (k = 10) returned the best balance between lower perplexity (229.06) and higher topic coherence (.52). A list of the top 10 keywords for each topic can be inspected in Table 1. A display of the relationship between the topics can be observed in Figure 1. Topic interpretation can also be found in Table 1. This interpretation draws upon the bar chart output for each topic from the LDAvis widget. Given journal space limitations, the complete LDAvis output can be examined on this study’s webpage (https://osf.io/u7jr2/). Concerning RQ2, a heatmap showing the level of each topic in each of the manuals can be viewed in Figure 2.

Figure 1. Relationships Between the Topics via Multidimensional Scaling.

Note. The size of the points of each topic represents that point’s marginal topic probability which indicates how strongly represented the topic is in the corpus (Orange Data Mining, 2022).

Figure 2. Heatmap of Topic Levels by Manual.

Note. The darker the red hue the stronger the topic’s presence in the manual.

Table 1. Topic Key Words.

Topic #	Topic Interpretation	Topic Key Words
1	Refugees	refugee, model, treatment, responder, australian, asylum, therapist, seeker, assessment, evidence
2	Orientation Activities	role, minute, helper, orientation, play, exercise, link, scenario, instruction, discussion
3	Community Based Applications	individual, cbpfa, culture, rural, old, community, experience, area, year, ask
4	PTSD & Other Psychological Issues	survivor, national, center, parent, pts, want, covid, operation, sleep, something
5	Training Materials	session, slide, save, hand, activity, parent, programme, giver, section, staff
6	Specific Helper Instructions	helper, individual, slide, role, national, experience, survivor, parent, minute, volunteer
7	Academics	journal, pts, study, report, year, depression, american, model, attack, psychiatric
8	MHPSS	trainer, helper, volunteer, mhpss, slide, service, national, link, module, group
9	General Curriculum	course, public, recommend, content, video, scenario, survivor, exercise, link, client
10	Austrialian Specific Delivery	australian, core, cross, deliver, national, society, organization, university, australia, centre

Note. Italicized words indicate LDAvis results of greater than .50 regarding term frequency in topic versus overall frequency in the corpus. Bolded words indicate the highest term frequency in topic versus the overall frequency in the corpus.

3. Discussion

This study aimed to identify the latent topic structure of English-language PFA training manuals. The results indicate a latent topic structure composed of 10 topics. Concerning RQ1, the obtained topic structure is consistent with what one might expect in a study in the training manual subregister. We found topics on training materials, curriculum, delivery, orientation, and research basis; as well as populations, programmes, and settings. All these topics were consistent with other topic modelling studies on training materials (e.g. Micheli et al., 2021). While these results point to the strengths of these manuals, the present study’s examination also encountered some problematic aspects of this training discourse. Our attention now turns to these aspects.

The results from RQ1 also point to four notable gaps in the topical structure of PFA manuals. The first of these gaps is the absence of a self-care topic. For example, of the 392,619 bigrams appearing in the PFA manual corpus, only 47 were the bigrams of ‘secondary stress’, ‘secondary trauma*’, or ‘vicarious trauma*’ (* denotes wild card for word ending). The term ‘self-care’ itself appears just 188 times. While it is not inevitable that those offering psychological first aid will develop secondary trauma (i.e. Greinacher et al., 2019a), it does occur. Happily, resiliency factors and processes exist to prevent deleterious outcomes in these first-line psychosocial responders (Alshahrani et al., 2022; Greinacher et al., 2019b; Greinacher et al., 2022). However, without substantive training on resiliency factors and processes, front-line workers supplying psychosocial support are at risk. As such, trainers should supplement their manual of choice with training and exercises on self-care. PFA manuals have been designed with the flexibility to be adaptable to the specific needs and contexts of different populations and communities (Brymer et al., 2006; WHO, 2013). It is possible that self-care could be addressed in other training, or directed to resources within their organization, or available elsewhere. Given that the main purpose of PFA is to help others, it could be considered a topic in the formation course to use the manuals, but not necessarily the manuals’ content.

The second gap is the absence of an armed conflict topic. The term ‘armed conflict’ appears rarely in the text. Only 11 times does the term appears in a corpus of 14 manuals. The KWIC lines for the appearance of this term are as follows:

Table

Manual	Left Context	Term	Right Context
Red Cross PFA	natural disasters, migration crises or	armed conflict	are provided with
Red Cross PFA	Cross Armed conflict. Situations of	Armed conflict	include acts of
Red Cross PFA	Living in a situation of	armed conflict	is dangerous and
Red Cross PFA	Mirva Helenius/Finnish Red Cross	armed conflict	Situations of armed
Red Cross PFA	in the future. PFA in	armed conflict	When the soldiers
STC PFA	threatened as a result of	armed conflict,	disaster or the
STC PFA	threatened as a result of	armed conflict,	disaster or the
STC PFA	against Children in Situations of	armed conflict	United Nations Rule
Red Cross PFA	or in a situation of	armed conflict.	In this phase
Red Cross PFA	or on-going situations of	armed conflict.	It is a
Red Cross PFA	events, such as natural disasters,	armed conflicts,	physical assault, abuse

Examining these KWIC lines reveals a limited and superficial discourse on armed conflict. This discourse's poverty directly contrasts with the founding reasons for PFA and its wide and deep application to armed conflict situations. The reason for this absence is unclear. It could be because PFA manual authors have little direct experience with armed conflict. Alternatively, it could be that agencies that fund PFA training manual development have less interest in this topic. Despite the overall lack of interest in the manuals, PFA practitioners have applied the extant generic manuals to conflict settings.

Schafer et al. (2016) did a qualitative study on how the WHO PFA manual was used in Gaza (World Health Organization, War Trauma Foundation, and World Vision International, 2013). Their focus group interviews indicate success in adapting this broad handbook to such a situation. The International Federation of Red Cross has developed training for applying their general manual to armed conflict situations (2018, 2022). While outcome research on conducting professional interventions in the midst of armed conflict has occurred (e.g. Schmitt et al., 2022), no PFA research has been conducted. Despite the complicated ethical and practical issues that come with this type of research, efforts in this direction are required.

A third notable gap in PFA manuals is the topic of cyberspace. Words related to this topic (web*, internet, online, text*) appear only 182 times in the corpus. The internet serves as a vehicle for many critical aspects of PFA, including access to (1) PFA interventions; (2) training; (3) training of trainers; (4) training materials; (5) PFA consultation; and (6) PFA practitioner and trainer communities. Kılıç Bayageldi (2021) laid out helpful guidelines for PFA practitioners delivering services online that every trainer should review. Tye et al. (2022) detailed the outcomes of a study on Zoom-delivered PFA training using the WHO manual (World Health Organization, War Trauma Foundation, and World Vision International, 2013). They found significant gains in trainees’ PFA self-efficacy and knowledge. Given the aforesaid, the topic structure findings of this study suggest that the present PFA training materials do not reflect the online nature of the current world. The reasons for this manual-reality disconnect are puzzling.

The fourth significant gap in PFA manuals is the lack of attention to cultural humility and multicultural awareness, particularly concerning gender and LGBTQIA2 + identities (lesbian, gay, bisexual, transgender, questioning, intersex, asexual, and two-spirit). Upon examining KWIC lines, it became apparent that there was minimal and tangentially related discourse on LGBTQI2S topics, while gender discourse varied significantly across manuals. A previous study found similar results, indicating that diversity and equity were not prioritized in existing PFA manuals (Ni et al., 2023). This trend suggests that more emphasis is needed on these critical topics. On a positive note, newer manuals, such as the UNICEF Indian Alaska Native Communities manual (Grubin et al., 2022b), are beginning to address the unique challenges communities might face in the aftermath of a crisis. This includes being mindful of the potential impact of discrimination, stigma, and marginalization on mental health and well-being. Example KWIC lines for the appearance of these terms are as follows:

Table

Manual	Left Context	Term	Right Context
UNICEF Indian Alaska Native	violence or intimate partner violence, including women, LGBTQI2S lesbian,	gay,	bisexual, transgender, questioning, intersex, and two-spirit community members,
Johns Hopkins Guide to PFA	stress over sexual orientation on self-esteem of	gay,	lesbian, and bisexual individuals. International Journal for the Advancement
UNICEF Indian Alaska Native	centres and resources near you. https://findtreatment.samhsa.gov SAMHSA	LQBTQ2S	A list of resources to support two-spirit
WHO PFA Guide	PFA can be offered to anyone, regardless of age,	gender,	social status, etc. But not everyone will want
KOICA UNICEF Lao PFA	principles are applied to everyone regardless of age,	gender,	social status, ethnic background, etc.
Singaravelu PFA	Language used to greet people and the language they speak _	gender &	Age – should affected women only be approached by women
STC PFA for Children	consider asking someone else of the same	gender	as the child to take over
STC PFA for Children	feel unsafe talking to someone of the opposite	gender.	Ask the child if s/he feels safe talking
STC PFA for Children	to ask about the concept of age and	gender.	At what age would they presume that a child
STC PFA for Children	the same number to group together. Either keep a good	gender	balance in the groups, or, depending on the context,
Red Cross PFA	person knows personally or if they are of a different	gender.	Be conscious of possible differences regarding this,
Singaravelu PFA	keep an appropriate distance depending on their age,	gender &	culture. Don't judge what they have or haven'
KOICA UNICEF Lao PFA	disaster can have an impact on anybody regardless of age,	gender,	culture, etc. There are often people of various cultural
Johns Hopkins Guide to PFA	Awareness of cultural and even	gender	differences is important when reflecting emotions.
Red Cross PFA	identified by characteristics such as	gender,	ethnicity, religious beliefs, race, sexual orientation or age.
STC PFA	in the groups, or, depending on the context, make same-	gender	groups. Contextualize the training.
Red Cross PFA	considering the person's culture, age and	gender	listen actively be aware of
Jacobs Community Based PFA	the safety of the workers, whether the teams are mixed	gender	or not. There are some problems
Jacobs Community Based PFA	send professionals out to work in the community in mixed-	gender	pairs. This is done partially to improve the team's
Singaravelu PFA	body language appropriate to their culture, age,	gender,	religion and custom. Always be considerate
Johns Hopkins Guide to PFA	Cournoyer, & Mahalik, (1995). Cross-sectional study of	gender	role conflict examining college-aged and middle-aged men.
NCTSN PFA Field Operations	reestablish customs, traditions, rituals, family structure,	gender	roles, and social bonds is important in helping survivors
STC PFA	when you have to talk to women and girls about	gender-	‘sensitive issues.’ ‘Sometimes you have to use relatives’

The preceding KWIC lines underscored the crucial role of gender in creating a safe and supportive environment sensitive to the community’s unique needs. Connecting individuals with appropriate resources and support networks is vital to helping them navigate their challenges.

Concerning RQ2, this investigation sought to examine how the topic structure is reflected in the different manuals in the sample. As noted, the analysis landed on a ten-topic model interpretation. This is shown in Table 1, and the heat map is shown in Figure 2, where each manual examined demonstrates the topic interpretations. Some topics, such as ‘PTSD & Other Psychological Issues’ (#4), are covered robustly across many manuals. Other topics, like ‘Specific Helper Instructions’ (#6), are not found distinctly in the manuals. Finally, the topic ‘Academics’ (#7) is associated with just one manual.

Three limitations of the present study should be considered. First, this was a topical examination of the universe of PFA training manuals published in English. The extent to which the resultant topical structure can be applied to non-English manuals is unknown. Second, this corpus was designed to minimize duplication. Thus, derivative manuals and manuals falling outside the search criteria were not assessed. As such, the extent to which the findings can be generalized to the whole universe of PFA training manuals remains unknown. The readers should consider this limitation when considering the following implication section. Third, LDA’s statistical nature does not allow for analysis over time (Naushan, 2020). Thus, the question of what, if any, changes have occurred over successive generations of the FPA training manuals remains open.

Regarding the implications for research emerging from the findings, two should be noted. First, this study can be used as an initial step toward investigating what the ‘active ingredients’ are in PFA training outcomes. For example, a research question like: ‘Does more of Topic X or less of Topic Y lead to better outcomes?’ At present, we simply don’t know. Also, in concert with the present study, greater knowledge of active ingredients could lead to the production of PFA training manuals that are more comprehensive and effective. A second research implication is the absence of a topical-level discourse on programme assessment in these manuals. Further research on best practices in this area may include this topic in future generations of these manuals. Based on randomized control trial outcomes, researchers can develop manuals that cater to learners at different competency levels, ranging from beginners to intermediate and advanced learners. These manuals are carefully designed to provide tailored training based on the individual's level of expertise and community needs and resources to enhance their learning experience. A prime example of a manual for beginners is the ABCDE Psychological First Aid Application Handbook (Cortés Montenegro & Figueroa Cabello, 2016), which offers basic yet essential information that beginners can easily relate to. It includes brief self-care instructions, which are crucial for beginners to understand as they embark on their learning journey. In addition to written content, each module is complemented with video demonstrations to enhance the learning experience.

As learners progress to intermediate and advanced levels, the manuals must provide more in-depth content to address the nuances of multicultural awareness and practical online training and learning in response to today's changing world. Intermediate and advanced responders require a deeper understanding of the complexities of various cultures and diverse perspectives to further develop their skills. The manuals for these levels should also focus on effective online training and learning strategies as the digital landscape continues evolving and shaping how we acquire knowledge and skills.

For PFA trainers, several practical implications can be drawn from the result of this study. One implication is that, at present, there is no comprehensive manual. As such, elements of PFA training will be missing, whatever manual one uses. Therefore, PFA trainers are advised to review multiple manuals so they can supplement elements missing from their manual of choice. Trainers must be well-equipped to prepare first responders from the general citizenry or agency resources to provide psychosocial support, psychological treatments, and foundational helping skills. The World Health Organization (WHO) offers a solution in the form of the Ensuring Quality in Psychological Support (EQUIP) platform (Watts et al., 2021). This online resource enables trainers to assess competencies, evaluate training, support trainee development, and modify curricula to ensure quality training. Another tool, the ENhancing Assessment of Common Therapeutic factors (ENACT) (Kohrt et al., 2015; Kohrt et al., 2015), measures a set of identified common factors necessary for effective psychological intervention, even when delivered by non-specialists.

Another practical implication concerns multicultural applications. Cultural considerations are sparse in these manuals. Fortunately, exceptions do exist. The PFA manual for American Indian and Alaska Native communities by Grubin et al. (2022b) is an excellent example of integrating culture and PFA training richly and effectively. The article on developing this manual is a must-read for trainers (Grubin et al., 2022b). Another practical implication concerns internet activities. It should be noted that because the PFA training manuals have not kept pace with the growing online nature of social interaction, PFA trainers must develop a list of internet best practices for their PFA trainees to learn. A final practical implication for trainers emerges from Greinacher et al.’s (2022) study on the psychological burden of PFA work. Given the paucity of discussion on self-care noted earlier, trainers should supplement the manual to address this topic, make trainees aware of the need to receive regular supervision, and modulate the frequency of emergency responses.

Author note

This study did not receive extramural funding. Given the published and public nature of the data, no human subjects review was required. The detailed data analysis results can be found at https://osf.io/u7jr2/

Disclosure statement

No potential conflict of interest was reported by the author(s).