Cross-cultural generalizability of suicide first aid actions: an analysis of agreement across expert consensus studies from a range of countries and cultures

Abstract

Background

A number of Delphi expert consensus studies have been carried out with different countries and cultural groups to develop guidelines on how a member of the public should provide assistance to a person who is suicidal. The present study aimed to determine whether cross-culturally generalizable suicide first aid actions are possible by comparing agreement across these Delphi studies.

Methods

Data on endorsement rates for items were compared across six Delphi studies. These studies involved panels of professionals and consumer advocates from English-speaking countries, professionals from Sri Lanka, professionals from Japan, professionals from India, professionals from the Philippines, and professionals and consumer advocates in refugee and immigrant mental health. Correlations were calculated between item endorsement rates across panels.

Results

There were 18 items that were highly endorsed across all eight of the Delphi panels and an additional 15 items highly endorsed across the panels from the three lower middle-income countries (India, Philippines and Sri Lanka). Correlations across panels in item endorsement rates were all 0.60 or above, but were higher between panels from countries that are socioeconomically similar.

Conclusions

There is broad agreement across the diverse expert panels about what are appropriate suicide first aid actions for members of the public, indicating that cross-cultural generalizability is possible. However, there is also some cultural specificity, indicating the need for local tailoring.

Background

Individuals in a suicidal person’s social network have the potential to take action that may reduce the risk of suicide. Because it is not feasible to carry out trials in which family and friends are randomly allocated to carry out specific actions to assist a suicidal person, expert consensus has been used to determine what are likely to be helpful actions in this area. This approach involves systematically gathering ‘practice-based evidence’ using the Delphi method, based on experts’ professional and/or personal experience [1]. It accords with ‘wisdom of crowds’ research showing that, under certain conditions, groups of individuals who each have imperfect expertise can make good decisions when their expertise is aggregated [1].

In the initial Delphi study on this topic, Kelly and colleagues [2] did a systematic search of advice on how to assist a suicidal person and used this to construct a questionnaire of 114 statements. This questionnaire was presented to a panel of 22 professionals and 16 people with lived experience (they had been suicidal or had cared for someone who had been suicidal in the past) who were from developed English-speaking countries. There were 30 items which were endorsed at a high level by both panels and these were used to produce guidelines on mental health first aid for suicidal ideation and behavior. These guidelines were subsequently used to inform the content of a Mental Health First Aid training course which has been widely disseminated [3]. The guidelines were also made available for free download online, and a survey of people who downloaded them found that some users made positive use of the guidelines to assist a suicidal person [4].

A new Delphi study was later carried out in order to update these guidelines [5]. A systematic search led to development of a questionnaire with 436 items about knowledge or actions to assist a suicidal person. These items were rated by panels of 42 suicide prevention professionals and 35 consumer advocates from developed English-speaking countries, and 164 items were endorsed at a high level. The items cover identification of suicide risk, assessing the seriousness of risk, initial assistance, talking with the suicidal person, no-suicide contracts, what the first aider should know, confidentiality and adolescent-specific considerations. The resulting updated mental health first aid guidelines for suicidal thoughts and behaviours were used to revise the content of Mental Health First Aid training [6,7,8,9]. More recently, the guidelines have been used as the basis of a special training course on Mental Health First Aid for the Suicidal Person [10]. These guidelines for the public differ from clinical practice guidelines for general practitioners and psychiatrists in that they do not involve formal assessment of clinical state, diagnostic labels or treatment plans [11, 12].

Because there may be cultural factors that affect provision of assistance by family, friends and other helpers [13], the Delphi studies on suicide first aid [2, 5] only used experts from developed English-speaking countries. However, the authors cautioned that: “The application of the guidelines to non-western cultures and ethnic minorities is an area requiring further investigation and consultation with suicide prevention experts from these cultural and ethnic backgrounds” [5].

Because of the potential cultural differences in how to best assist a suicidal person, a series of Delphi studies has also been carried out with professional experts in a number of Asian countries, using the questionnaires of either of the English-speaking country Delphi studies as a starting point [2, 5], but supplemented with additional items relevant to the specific culture. These studies have been carried out for India [14], Japan [15], Philippines [16] and Sri Lanka [17]. Cultural differences may also be important for ethnic minority groups within countries [18, 19]. For this reason, a Delphi study has also been carried out on assisting suicidal persons from refugee and immigrant backgrounds [20].

While a culturally-specific approach to developing suicide first aid guidelines might be the ideal, it is resource intensive and would be a major task to implement for every potential cultural group in the world. The aim of the present study is to explore the feasibility of having cross-culturally generalizable suicide first aid guidelines. The study does this by comparing the endorsement rates of expert panels from the various existing Delphi studies to see whether there are common actions recommended across all the studies, and also across the three lower-middle income countries specifically. The study also quantifies the degree of agreement across expert panels from the various cultural groups.

Methods

Data compilation

Data from Delphi expert consensus studies to develop suicide first aid guidelines were available for English-speaking countries [5], Philippines [16], India [14], Japan [15], Sri Lanka [17] and Refugees and Immigrants [20]. These studies were conducted by research teams that included one or more of the authors. The studies are summarized in Table 1. All of these Delphi studies involved a panel of professional experts. In addition, the English-speaking Delphi study included a panel of consumers and the Refugee and Immigrant study included a panel of consumer advocates (who partially overlapped with its professional panel). In all of these studies, items were presented and rated in English for how important they were to include in the guidelines. The instructions were: “Please rate how important (from essential to should not be included) you think it is that each statement be included in the final guidelines”. The rating scale for each item had the following options: Essential, Important, Don’t know/depends, Unimportant, Should not be included. An item was regarded as endorsed by a panelist if it was rated as ‘essential’ or ‘important’ to include. Copies of the helping-action statements (items) included in each cultural Delphi study, as well as their endorsement ratings, were obtained by downloading relevant additional files from the publishing journal’s website (English-speaking, Philippines, India, Japan) or by contacting authors directly (Sri Lanka, Refugee and Immigrant). Some of the Delphi studies included items on suicide warning signs appropriate to the particular culture. These items were not included in the current analyses, which focus on suicide first aid actions.

Table 1 Characteristics of the Delphi studies

Because the Delphi studies involved multiple rounds of rating, we used the ratings for each item on the round where its outcome was decided, e.g. if an item was re-rated in Round 2, we used the data from that round. Identical items across the different sets of Delphi items were matched, with corresponding endorsement ratings compiled into one data file. Items were matched using the search function in Microsoft Excel to locate novel words in item text. Researchers’ prior knowledge of guidelines content was also used in matching items with slightly different wording but similar meaning.

All items deemed similar by one of the authors (AMR) were discussed with another author (AFJ) to determine if they were similar enough for endorsement ratings to be comparable, or whether they are too dissimilar in meaning and need to be treated as separate items.

Some items that comprised two sentences sometimes appeared in some of the cultural Delphi studies as two separate statements, split at the full-stop. For items that were split and received similar endorsement ratings between them, the two ratings for the split item were averaged and included as an endorsement rating for the non-split item. For split items with dissimilar endorsement ratings, the two ratings for the split item were treated as different items to the original two-sentence non-split statement.

Data analysis

To quantify agreement across a pair of expert panels, the percentage endorsement for each item by a panel was calculated and a Pearson correlation of item endorsement rates computed between the panels. This method has been previously used to compare item endorsement rates of multiple expert panels (e.g. professional, consumer advocate, carer advocate) within Delphi studies [1]. A correlation matrix was developed using data from all available Delphi studies. Because the Delphi studies had varying number of items in common, the number of items that each correlation was based on varied between pairs of studies from 85 (English-speaking country professionals with Japan professionals) to 542 (Refugee and Immigrant professionals with Refugee and Immigrant consumer advocates).

A missing values analysis showed that data on only 74 items were available across all Delphi panels. In order to include a greater number of items that were common across most of the Delphi panels, multiple imputation was carried out. Values were imputed for items that were missing ratings from up to 2 of the 8 Delphi panels using the Fully Conditional Specification algorithm (utilising the multiple chain method), with imputed values constrained between 0 to 100. This allowed a correlation matrix to be computed for 109 items across all panels. Data were analysed using SPSS v23.

To explore areas of difference between panels, an analysis was carried out of items that were endorsed by at least 80% of one panel, but another panel was 30% or more lower in endorsement. A difference of 30% has been described as a ‘large’ effect size [21]. This analysis was carried out for the items that were common across all Delphi studies. Because there are many comparisons involved, interpretation has been focused on areas of consistency where one panel had large differences from three or more other panels.

Results

Table 2 shows the 18 items that were endorsed across all the 8 Delphi panels. The percentage panel agreement required for an item to be endorsed was 80% rating it as ‘essential’ or ‘important’ for all panels, except for the Japanese one, which had an endorsement cutoff of 70% because of the greater use of ‘don’t know/depends’ by that panel. Table 3 shows additional items to those in Table 2 that were endorsed by the panels from the three lower-middle income countries.

Table 2 Items endorsed across all Delphi studies

Table 3 Additional items endorsed across the Delphi panels from the three lower-middle-income countries

Table 4 shows Pearson correlations across items for the various panels based on complete data (upper diagonal) and based on imputed data (lower diagonal). Correlations were all .60 or above, indicating broad agreement in priorities for inclusion in guidelines. The correlations were consistently highest (all .8 or above) between the English-speaking panels and the Refugee and Immigrant panels. The Asian country panels tended to show somewhat lower agreement with each other, with correlations in the range .67 to .87. The lowest agreement was between the English-speaking panels on the one hand and the Asian country panels on the other, with correlations in the range .60 to .79.

Table 4 Correlations of endorsement rates across items for the Delphi panels

Table 5 shows the items that were consistently more likely or less likely to be endorsed by one panel compared to others. It is difficult to distinguish consistent themes from these differences without knowing the panel members’ reasons for the ratings they made. However, a few themes are apparent. English-speaking panelists were less likely to endorse attempting to dissuade the person from suicide (e.g. by discussing the morality of suicide, telling them how much they would be missed and giving hope of the benefits of treatment). The Japan panel was also less likely to endorse dissuading actions. English-speaking panelists and the Japan panel were also less likely to endorse working with the person to deal with specific problems in their life than were panelists from the lower-middle-income countries or the immigrant and refugee panels. Something else that stands out in examining these differences is that the Japan panel was less likely to endorse a number of items, but there were no items it was more likely to endorse, whereas Philippines, India and Sri Lanka showed the opposite pattern.

Table 5 Items showing large and consistent differences between panels^a

Discussion

Comparison of items endorsed across all Delphi studies shows a number of actions that are recommended across a broad range of countries and cultures. These cover assessing the risk of suicide, listening to the person, showing care and respect, and ensuring the person’s safety. A notable omission from these items is recommendations to refer the person to professional help. This omission may reflect the fact that items about professional help-seeking were not consistently included across the various Delphi questionnaires. It might also reflect limitations in the availability or feasibility of professional help in some countries, where services like suicide hotlines are not always available and mental health services can be scarce, particularly in rural areas. Thus, the list of items that were endorsed across studies should be regarded as a minimal cross-cultural consensus rather than exhaustive.

We also specifically examined actions that were endorsed across the three lower-middle-income countries (India, Philippines and Sri Lanka), which resulted in a larger pool of common items, indicating the feasibility of developing guidelines that might be generalizable to a broader group of low- and middle-income countries. Although additional items were endorsed in the three lower-middle-income countries, these may not necessarily reflect cultural differences between these countries and the rest. It may instead be due to the fact that it is easier to get consistent endorsement across three expert panels than across eight panels.

Despite these commonalities, there were also a number of differences between panels For example, the English-speaking and Japan panels were less likely to endorse attempting to dissuade the person from suicide or attempt to solve their problems than were panelists from the lower-middle-income countries or the immigrant and refugee panels. This could partially be explained by the fact that the lower-middle income countries in these studies are generally considered collectivistic societies, i.e. people are expected to look after each other (e.g. helping to solve one’s problems). Furthermore, in English-speaking countries and more ‘westernized’ societies such as Japan the bio-medical model of suicide, which attributes suicidal behavior to mental illness, is widespread and as a consequence an individual might see him/herself as unable to ‘solve the problem’, as this requires the intervention of a mental health professional or other health/medical professional [13]. There were also differences between panels in endorsement rates that may reflect variation in willingness of panel member to state firm views on the items. Cultural differences were also reported in the original Delphi studies. For example, in the Filipino study it was noted that religious and spiritual concepts appear to be more important in suicide prevention than in India or Japan [16]. By contrast, in the Japanese study it was noted that the experts were less likely to endorse actions that involved a closer personal distance and that some first aid strategies highly endorsed in other countries might be seen as socially inappropriate in Japan [15]. In the Refugee and Immigrant study, it was noted that refugees and immigrants sometimes have a fear and distrust about emergency services and that females from some cultural backgrounds may not be permitted to make decisions regarding their own health alone [20]. Availability of services is another culturally varying factor. For example, in the Sri Lankan study the lack of services in rural areas was noted [17].

Developing separate suicide first aid guidelines for each cultural group is the ideal. As well as providing specific cultural tailoring, it provides buy-in from the target cultural group by involving relevant experts at an early stage. However, to do this for every country, including for cultural minority groups within a country, is very resource intensive. It can also be difficult to recruit lived-experience experts from cultural groups that lack consumer advocacy organizations, which was the case with the Asian Delphi studies included here.

The alternative is to develop guidelines based on the consensus of experts from a broader range of cultural groups, including lived-experience experts where they are available. These broader guidelines may then be supplemented by expert consensus guidelines on additional cultural considerations that are specific to the target group. For example, this approach has been used to develop guidelines on additional considerations when giving mental health first aid (which includes suicide first aid) to Aboriginal and Torres Strait Islander adolescents [22], Iraqi refugees [23] and LGBTIQ people [24]. These additional cultural considerations have then been incorporated in Mental Health First Aid training [7, 9].

Conclusions

While this study has involved only suicide first aid guidelines, the findings may have implications for the broader development of mental health first aid guidelines for assisting a person developing a mental health problem or in a mental health crisis [25,26,27,28,29,30,31,32,33]. While previous research has sought the expertise of professionals, consumer advocates and carer advocates from developed English-speaking countries, it may be feasible to develop mental health first aid guidelines that have broader applicability, for example to high-income countries or to low- and middle-income countries, which can then be supplemented by consensus on local cultural and health system considerations.