A Self-Help Online Intervention Is Associated with Reduced Distress and Improved Mental Wellbeing in Australian Farmers: The Evaluation and Key Mechanisms of www.ifarmwell.com.au

ABSTRACT

Objectives: Farmers are faced with many stressors, along with numerous barriers to accessing traditional forms of mental health support. The ifarmwell online intervention was co-designed with farmers and is informed by Acceptance and Commitment Therapy to equip farmers with transferable coping strategies, particularly to help them cope with circumstances beyond their control. We aimed to evaluate the effect of ifarmwell on farmers’ short- and long-term distress and mental wellbeing. Methods: Australian farmers (21–73 years) who registered on www.ifarmwell.com.au completed measures at the commencement of module 1 (N = 228), the end of the intervention (N = 77) and 6-months post-intervention (N = 61). Primary outcomes included distress (Kessler Psychological Distress Scale) and mental wellbeing (Mental Health Continuum – Short Form). We also examined the relationship between distress, mental wellbeing, neuroticism (Quickscales-R) and ACT-based psychological mechanisms; psychological inflexibility (Acceptance and Action Questionnaire-II), cognitive fusion (Cognitive Fusion Questionnaire), believability of automatic thoughts (Automatic Thoughts Questionnaire-B), coping via acceptance (4 items from the situational COPE) and mindfulness (Five Facet Mindfulness Questionnaire-Short Form). Secondary outcomes of acceptability (Client Satisfaction Questionnaire-8) and usability (System Usability Scale) were also explored. Results: Pre- to post-intervention, farmers’ distress decreased and mental wellbeing increased. These effects were maintained at 6-month follow-up. Changes in these outcomes were greatest for participants who entered the intervention with high baseline distress ($\hat{\beta }$= −0.59, 95%CI =[−0.70, −0.47]) and low mental wellbeing ($\hat{\beta }$= −0.33, 95%CI = [−0.47, −0.19]). Decreases in distress and increases in mental wellbeing were associated with decreases in psychological inflexibility, cognitive fusion, and believability of thoughts, and increases in coping via acceptance and mindfulness. Levels of satisfaction (M = 26.92/32) and usability (M = 84.70/100) were high, and 94.6% of participants said they would recommend ifarmwell to a friend in need of similar advice and tools. Conclusions: ifarmwell is an effective and usable intervention that is likely to help farmers reduce their levels of distress and improve their mental wellbeing, by improving their psychological flexibility, ability to focus on the present and accept things beyond their control, as well as by reducing the extent to which they believe unhelpful thoughts.

KEYWORDS:

• Acceptance and commitment therapy
• agriculture
• coping
• drought
• farm
• mental health
• online intervention
• rural
• ; stress

Introduction

Farming is a psychologically challenging occupation,¹ with many of the factors that determine their occupational success (or otherwise), being factors such as the weather, which are beyond their control.² Despite widespread exposure to stressors,² farmers are known to be stoic, self-reliant, to minimize the problems they experience,³ and many consider mental health a lower priority than farm work.⁴ Even if they do consider seeking help, long wait lists can be off-putting, and many farmers consider that doing so via traditional face-to-face consultations is too time consuming and expensive due to travel distances, the need to employ labor while they are away, and/or the direct costs associated with accessing services.⁴ Farmers are also known to show high levels of skepticism towards non-rural people, and to perceive a lack of understanding of their way of life from “outsiders” and/or “experts”, and cultural inappropriateness of many services.^2,5–7 These attitudinal issues act as further barriers to their help-seeking in times of need. Farmers are half as likely to seek help than their non-farming rural employed counterparts.⁸ They already have higher rates of suicide than the general population,^9–12 and with climate change models predicting more frequent and extreme climate events (floods, fire, drought, wind) in the future,¹³ it is foreseeable that farmers’ levels of uncertainty and distress² will continue to increase.

There is a clear need to develop and implement evidence-based strategies to prevent and reduce distress, as well as improve mental wellbeing, in this culturally unique and difficult to reach population. An online intervention, ifarmwell¹⁴ is co-designed with farmers, specifically for farmers. As detailed elsewhere,¹⁵ the ifarmwell intervention is also informed by Acceptance and Commitment Therapy (ACT),¹⁶ behavior change techniques¹⁷ and persuasive system design elements¹⁸ to aid engagement. ACT is a transdiagnostic psychotherapeutic approach that focuses on helping people to accept what they cannot control, to distance themselves from challenging thoughts and feelings, and put energy into things that are consistent with their values. More specifically, it aims to help people become “psychologically flexible” (or accepting of difficult thoughts and feelings) via six core processes; (1) cognitive defusion, (2) acceptance, (3) being present, (4) values, (5) committed action, and (6) self-as-context.¹⁹ Definitions of these processes, how they are addressed in the ifarmwell intervention, as well as how they are measured in the following evaluation, are outlined below in Table 1.

Table 1. Ifarmwell modules and corresponding acceptance and commitment therapy (ACT) processes and measures.

Module Name	Acceptance and Commitment Therapy (ACT) Processes Targeted	ACT Process and Associated Measure
Module 1: Taking stock of your current wellbeing and some practical strategies to get you started	No ACT processes targeted	Acceptance = 4 items from the situational version of the COPE (Carver, Scheier, Weintraub, 1989), Acceptance and Action Questionnaire-II (AAQ-II; Bond et al., 2011) Cognitive defusion= Cognitive Fusion Questionnaire (CFQ; Gillanders et al., 2014), Automatic Thoughts Questionnaire-B (ATQ-B; Hollon & Kendall, 1980) Being present= Five Facet Mindfulness Questionnaire-Short Form (FFMQ-SF; Bohlmeijer, Klooster & Fledderus, 2011) Self as context= Five Facet Mindfulness Questionnaire-Short Form Values= Valuing Questionnaire (Smout et al., 2014) NB this questionnaire was optional to reduce participant burden. A lack of uptake of the optional questions/abundance of missing data meant this variable had to be excluded from the analyses. Committed action= Acceptance and Action Questionnaire-II
Module 2: Thoughts are like bullies – how to spend less time “in your head”	Acceptance (that the stressful event has occurred and is real, and acceptance rather than avoidance, of difficult thoughts and feelings) Cognitive defusion (ability to separate self from thoughts) Being present (focused on now) Self as context (being aware of your experiences without being attached to them)
Module 3: Doing what really matters – how to get the most out of life	Values (awareness of own values) Committed action (action in line with values)
Module 4: Training your “attention muscle” and focusing on the “here and now” – a more pleasant, less exhausting place to be	Acceptance Cognitive defusion Being present Self as context Values Committed action
Module 5: Putting it all together and moving forward	Acceptance Cognitive defusion Being present Self as context Values Committed action

• Carver CS, Scheier MF, Weintraub KJ. Assessing coping strategies: a theoretically based approach. Journal of Personality and Social Psychology. 1989;56(2):267–83.

• Bond FW, Hayes SC, Baer RA, Carpenter KM, Guenole N, Orcutt HK, et al. Preliminary psychometric properties of the Acceptance and Action Questionnaire-II: A revised measure of psychological flexibility and experiential avoidance. Behav Ther. 2011;42:676–88.

• Gillanders DT, Bolderston H, Bond FW, Dempster M, Flaxman PE, Campbell L, et al. The development and initial validation of The Cognitive Fusion Questionnaire. Behav Ther. 2014;45:83–101.

• Hollon SD, Kendall PC. Cognitive self-statements in depression: Development of an Automatic Thoughts Questionnaire. Cog Therap Res. 1980;4:383–95.

• Bohlmeijer E, ten Klooster PM, Fledderus M, Veehof M, Baer R. Psychometric properties of the five facet mindfulness questionnaire in depressed adults and development of a short form. Assessment. 2011;18:308–20.

• Smout M, Davies M, Burns N, Christie A. Development of the Valuing Questionnaire (VQ). J Context Behav Sci. 2014;3:164–72.

As detailed elsewhere,¹⁵ ACT was viewed as an appropriate therapeutic approach to test in the context of farmer distress, as “acceptance” has been shown to be a particularly adaptive coping strategy for farmers to adopt in times of drought, and behavioral disengagement (giving up or withdrawing effort from attempting to reach the goal with which the stressor is interfering, i.e. the opposite of committed action) is maladaptive.²⁰ By targeting these two key coping strategies, ACT is well-suited to assisting people in contexts where the stressor (e.g. illness, or in the case of farmers, the weather) cannot be controlled.²¹

The ifarmwell website has already been found to be highly acceptable to users.¹⁵ Based upon data collected from users between February 2018 and October 2018, the average participant rating of modules was high. Further, completion of the intervention components was not influenced by participant demographics, level of distress or satisfaction with the previous module. Instead “too busy/not got to it yet” was found to be the major reason for non-completion.

This paper builds upon that acceptability testing and reports on the effectiveness of the ifarmwell intervention and the possible ACT-based mechanisms that may be responsible for these changes. A secondary aim was to examine the role of neuroticism (being sensitive/nervous). In addition to developing farmers’ coping strategies, there is a need to better understand the individual-level factors that may influence farmers’ well-being. Our previous work identified that neuroticism (which is generally considered a non-modifiable personality trait) accounted for 37% of the variance in the distress experienced by farmers exposed to stressful events during drought.²⁰ Building on this, the current study sought to examine the role that neuroticism may play in determining the effectiveness of the ifarmwell intervention. A final aim was to examine ratings of usability and satisfaction with the website. The following hypotheses guided this evaluation:

H1.

Pre- to post-intervention, distress will decrease and levels of mental wellbeing will increase, and these effects will be maintained at a 6-month follow-up.

H2.

Improvements over time will be greater for users who enter the intervention with higher distress or lower mental wellbeing.

H3.

Improvements over time in distress and mental wellbeing will be associated with key ACT processes including lower psychological inflexibility, lower cognitive fusion, reduced believability of thoughts, greater use of acceptance as a coping strategy, and greater mindfulness.

H4.

Neuroticism, a dispositional trait, will be positively associated with distress and negatively associated with mental wellbeing at baseline, but it will not be associated with change in distress or mental wellbeing over the course of the intervention.

H5.

Satisfaction and usability ratings will be high.

Methods

Study design

This was a single-arm, pre- and post-study with users of the ifarmwell intervention between February 2018 and February 2020. The University of South Australia Human Research Ethics Committee approved this study (ID 0000035637). The protocol was registered with the Australian New Zealand Clinical Trials Registry (ACTRN: ACTRN12617000506392) on April 3, 2017 .

Sample size

The sample size was calculated for pre- to post-intervention change in the K10 distress score. Based on cross-sectional data from a previous study, we were interested in changes of at least 2.5 units (SD = 7, range 0–35) [20]. Being an exploratory study, we prioritized the reduction in Type II errors. Assuming a within individual correlation >0.5, then with N = 52 a one sample t-test (1-sided alpha = 0.10) has 90% power to detect a within individual change in distress of 2.5 units. Assuming an 80% retention rate, we planned to accrue at least 65 individuals.

Intervention details

The ifarmwell intervention is described in detail elsewhere.¹⁵ Five interactive modules address the six core ACT processes. A new module is made available to users every fortnight, meaning that the intervention is completed over a minimum of 10 weeks (but users may take longer if desired). The first module also includes distress screening and advice on how to seek help from a general practitioner (or other services) for mental health concerns, but users are also encouraged to continue progressing through the modules. Emails and SMS reminders and supportive messages are sent to users throughout the intervention.

Sample

Participants were adults who owned or played an active role in the operation of a farming or pastoral enterprise in Australia (or their spouse), were fluent in English, had access to the Internet, and had access to a mobile phone with reception at least once per week. Participants with cognitive impairment were not eligible to take part. Consent to participate in the evaluation was established when users registered on the website.

Measures

Measures asked at each timepoint (baseline, post-intervention, 6-month follow-up) are outlined below. All measures were compulsory, first asked immediately prior to module 1 (baseline), and repeated post-intervention (end of module 5) and 6 months following completion (follow-up), unless otherwise indicated. The Valuing Questionnaire²² and Five Facet Mindfulness Questionnaire-Short Form (FFMQ-SF)²³ were optional to reduce participant burden. Due to a substantial amount of missing data on the constructs measured by the Valuing Questionnaire²² (Baseline 17.5% Progress and Obstruction subscales; Post-intervention 25.9% Progress and Obstruction subscales), this scale was excluded from the analysis, so is not detailed below. All measures showed high internal consistency (see Supplementary Table A for Cronbach’s alphas at baseline).

Demographics

Participants gave their age, gender, marital status, highest education level, farm type and residential postcode at baseline only. Postcode was used to calculate remoteness of residence using the Accessibility and Remoteness Index of Australia (ARIA) from the Australian Bureau of Statistics.²⁴ Due to small numbers, several variable categories were collapsed for the purpose of regression analyses. Remoteness of residence was collapsed into two categories by combining Major Cities with Inner Regional, and Remote Australia with Very Remote Australia categories. Farm type was collapsed into: (1) grain, sheep and/or cattle, (2) sheep and/or cattle, (3) dairy, and (4) horticulture, cane, poultry, viticulture or other. Marital status was collapsed into (1) married or partnered and (2) not married or partnered, and education was collapsed into (1) finished primary or high school, (2) trade certificate, or (3) tertiary degree/postgraduate degree. Whether users signed up to the website before (Version 1) or after changes (Version 2) were made following acceptability and usability testing (11/4/2019) (as detailed elsewhere)¹⁵ was also recorded.

Primary outcomes

Distress. The 10-item Kessler Psychological Distress Scale (K10)²⁵ assessed distress over the past 30 days, with questions rated on Likert scales from 1 (None of the time) to 5 (All of the time) and higher scores indicating greater distress. A continuous score can be used in analyses (range 10–50) or participants can be classified as having low (10–19), mild (20–24), moderate (25–29), or severe (30–50) levels of distress.²⁶

Mental wellbeing. Mental wellbeing (including emotional, psychological and social wellbeing) over the last 30 days was assessed using the 14-item Mental Health Continuum – Short Form (MHC-SF).²⁷ Questions are measured on Likert scales from 0 (Never) to 5 (Every day), with higher scores indicating better wellbeing. The scale can be scored as a continuous measure or to categorically indicate “languishing”, “moderate” or “flourishing” wellbeing as described in detail elsewhere.²⁸

Acceptance and Commitment Therapy (ACT) variables

Psychological inflexibility/experiential avoidance. Psychological inflexibility/experiential avoidance (i.e. non-acceptance of difficult thoughts and feelings) was measured using the 7-item Acceptance and Action Questionnaire-II (AAQ-II).²⁹ Items are rated on Likert scales from 1 (Never true) to 7 (Always true), with higher scores indicating greater psychological inflexibility.

Cognitive fusion. The 7-item Cognitive Fusion Questionnaire (CFQ),³⁰ examined participants’ ability to separate themselves from their thoughts. Items were rated on Likert scales from 1 (Never true) to 7 (Always true), with higher scores reflecting greater fusion to one’s thoughts.

Believability of automatic thoughts. The tendency to believe negative automatic thoughts was assessed using the 30-item Automatic Thoughts Questionnaire-Believability (ATQ-B).³¹ Items were rated on Likert scales from 1 (Not at all) to 5 (Totally), with higher scores indicating more negative automatic thoughts.

Mindfulness. Mindfulness (observing, describing, acting with awareness, non-judging and non-reactivity) was measured using the Five Facet Mindfulness Questionnaire-Short Form (FFMQ-SF).²³ Items were rated on Likert scales from 1 (Never or very rarely) to 5 (Very often or always true), with higher scores indicating greater mindfulness.1 1 Item 13 from the FFMQ-SF was incorrectly worded in the participant questionnaire and was subsequently removed for data analysis. This resulted in a 23-item measure.

Coping via acceptance. Four items from the original COPE scale³² examined participants’ use of acceptance as a coping strategy (accepting the fact that the stressful event has occurred and is real). Items were scored on Likert scales from 1 (I usually don’t do this at all) to 4 (I usually do this a lot), with higher scores indicating greater use of acceptance coping.

Confounder variables

Neuroticism. The 6-item Neuroticism subscale of the Quickscales-R³³ examined the extent to which participants are sensitive/nervous on 7-point Likert scales from 1 (Not at all) to 7 (Extremely), with higher scores indicating greater neuroticism.

Secondary outcomes: satisfaction with and usability of the intervention

In the two post-intervention surveys, participants were asked to rate their satisfaction with and the usability of the intervention using the Client Satisfaction Questionnaire-8 (CSQ-8; items scored 1–4, with higher scores indicating greater satisfaction) and the System Usability Scale (SUS; items scored 0–4, with higher scores indicating greater usability and acceptability), respectively. A score on the SUS of 68 or more is considered “above average”.³⁴ Cronbach’s alphas were 0.89 and 0.78, respectively.

Procedure

Participants accessing the website between February 2018 and February 2020 were invited to participate in the website evaluation and were presented with an online information sheet and consent form. Measures were taken immediately prior to module 1 (baseline), at the end of the final module (post-intervention), and 6 months following intervention completion (follow-up).

Statistical methods

Analyses were conducted in Stata version 16.1. Means (standard deviations) and frequencies (percentages) were used to describe the sample at baseline. Logistic regressions were conducted to identify demographic and baseline predictors of drop-out between those starting Module 1 and those completing the post-intervention questionnaire.

Linear mixed models using maximum likelihood estimation to assess change over time in distress and mental wellbeing (hypotheses H1 and H2). The outcome variable for these models was the change score, calculated as post-intervention score minus baseline score. Timepoint (post-intervention, follow-up), mean-centered baseline score of the outcome variable, mean-centered age, gender and website version were included as fixed effect factors. Random intercept per individual were included to account for repeated measures with an identity covariance structure for the error distribution (i.e., compound symmetry covariance amongst observations from the same individual). To assess the influence of possible post-intervention change in improvements in distress and wellbeing (H3), and the relationship of neuroticism with change in distress and wellbeing over time (H4), change scores for each of the explanatory ACT variables were entered into linear regressions for the distress and mental wellbeing outcomes. Change between post-intervention and baseline was the specified outcome measure in these models. Mean-centered baseline score, age, gender and website version were entered as covariates. Missing data imputation for this analysis is outlined in detail in Supplementary Materials B.

Multiple linear regressions were conducted to examine associations between neuroticism and the key outcome variables at baseline (H4). These controlled for baseline age, gender and website version. Descriptive statistics were used to examine satisfaction and usability outcome measures (H5).

Results

Primary analyses include the 228 people who started the intervention. An additional 36 people registered and provided baseline scores on the measures taken at registration. However, these people did not start the intervention and are subsequently not included in the analyses below. Seventy-seven people (33.8%) completed the post-intervention questionnaire and 61 people (26.8%) completed the 6-month follow-up. Please note that module completion rates and drop out are examined in detail in a prior publication [15].

Participant characteristics

At baseline, participants were aged 21 to 73 (mean = 45.9, SD = 12.2), primarily female (63%), married or partnered (84.7%), living in outer regional (40.5%) or inner regional Australia (39.2%), on a grain, sheep and/or cattle farm (46.9%), and had completed a university degree or diploma (37.3%) (Table 2).

Table 2. Demographics for users at baseline and post-intervention.

	Participants who completed baseline measures (N=228)	Participants who completed pre- and post-intervention measures (N=77)	Participants who did not complete all five modules (N=151)
Age (years)
Mean (SD)	45.9 (12.2)	48.1 (12.3)	44.9 (12.0)
Range	21-73	21-73	22-71
Gender, n(%)
Male	84 (36.8)	28 (36.4)	56 (37.1)
Female	144 (63.2)	49 (63.6)	95 (62.9)
Marital status, n(%)
Married or living with partner	193 (84.7)	68 (88.3)	125 (82.8)
Never married	18 (7.9)	2 (2.6)	16 (10.6)
Separated/divorced	11 (4.8)	4 (5.2)	7 (4.6)
Widowed	6 (2.6)	3 (3.9)	3 (2.0)
Education level (highest qualification), n(%)
Finished Primary or High School	53 (23.3)	21 (27.3)	32 (21.2)
Trade Certificate	60 (26.3)	18 (23.4)	42 (27.8)
University degree or diploma	85 (37.3)	28 (36.4)	57 (37.8)
Postgraduate Degree	30 (13.2)	10 (13.0)	20 (13.3)
Remoteness of residence, n(%)
Major Cities of Australia	12 (5.3)	5 (6.6)	7 (4.6)
Inner Regional Australia	89 (39.2)	26 (34.2)	63 (41.7)
Outer Regional Australia	92 (40.5)	32 (42.1)	60 (39.7)
Remote Australia	27 (11.9)	13 (17.1)	14 (9.3)
Very Remote Australia	7 (3.1)	0	7 (4.6)
Farm type, n(%)
Dairy	23 (10.2)	9 (11.8)	14 (9.3)
Grain, Sheep and/or Cattle	106 (46.9)	35 (46.1)	71 (47.3)
Horticulture, Market Garden, Fruit	18 (8.0)	7 (9.2)	11 (7.3)
Sheep and/or Cattle	65 (28.8)	21 (27.6)	44 (29.3)
Poultry/cane/viticulture/other	14 (6.2)	4 (5.3)	10 (6.7)

Table 3 shows the mean values of all measures at each of the three timepoints. At baseline, the majority of the sample had scores consistent with “low” or “mild” levels of distress (70.6%), and only 13 participants (6.5%) had scores that indicated that their mental wellbeing was “languishing” according to the MHC-SF criteria. Supplementary Table A shows the correlations between key measures at baseline.

Table 3. Description of each measure pre- and post-intervention, and at the 6-month follow-up.

	Higher scores =	Baseline M(SD) N=228	Post M(SD) N=77	Follow-up M(SD) N=61
K10^a	More distress	21.1(7.2)	18.1(4.9)	17.2(4.8)
Mental Health Continuum – Short Form	Better wellbeing	45.4(13.0)	46.8(12.1)	48.2(13.3)
Acceptance and action questionnaire ^b	More psychological inflexibility	22.7(8.2)	19.0(6.9)	19.4(7.2)
Cognitive Fusion Questionnaire ^b	More fusion	24.6(9.1)	21.5(7.6)	20.9(8.4)
Automatic Thoughts Questionnaire^c	More negative automatic thoughts	62.5(22.0)	49.3(18.0)	50.7(20.8)
Coping via acceptance ^b	More use of acceptance coping	12.3(2.7)	13.1(2.6)	13.3(2.4)
Five Facet Mindfulness Questionnaire ^d	More mindful	74.5(12.4)	79(11.2)	80(11.3)
Neuroticism^b	More neurotic	21.8(7.2)	20.1(7.4)	18.8(7.2)
Client Satisfaction Questionnairee	More satisfied	-	26.9(4.1)	26.9(4.3)
System Usability Scale^e	More useable	-	84.7(13.1)	84.6(11.8)

^aThis questionnaire was asked in Module 1. Of the 264 people who registered for ifarmwell, only 228 started Module 1 and completed the K10.

^bPost-intervention n = 76, follow-up n = 60.

^cBaseline n = 137. This questionnaire was asked in Module 2. One-hundred and thirty-seven people started Module 2 and completed these questions.

^dOptional questionnaires. Baseline n = 168, post-intervention n = 57, follow-up n = 48.

^ePost-intervention n = 74, follow-up n = 60.

Note: Some users did not complete all modules and the post-intervention questionnaire could be taken at any time, without completing all five modules, leading to missing data on some variables. Data shown are pre-imputation of missing values.

Supplementary Table C summarizes each measure for the subset of participants who completed both the pre- and post-intervention measures (n = 77, 33.8%). Univariable and multivariable logistic regression models indicated no association between drop-out and age, gender, marital status, education level, farm type, remoteness of residence, or baseline scores on distress, psychological inflexibility or mental wellbeing (see Supplementary Table D).

Change in primary outcome measures

In support of H1, the predictive marginal means show that post-intervention distress scores were significantly lower than baseline, with a mean change of −2.76 units [95%CI: −3.61, −1.91] from baseline (Table 4). This continued at 6-month follow-up, with a mean change of −3.52 units [95%CI: −4.44, −2.60] from baseline, although the difference between post-intervention and follow-up scores was not significant ($\hat{\beta }$ = −0.76, 95%CI = [−1.66, 0.13]). Mental wellbeing increased, from baseline to post-intervention. Post-intervention scores did not significantly differ from 6-month follow-up scores on either outcome, indicating a maintenance of these effects over time.

Table 4. Linear mixed models predicting change from baseline for distress and mental wellbeing (N = 77).

	Predictive marginal means
	M(SE)	95% CI	$\hat{\beta }$ [95% CI]	P value
Change in distress from baseline
Time
Post vs. pre	−2.76(.43)	−3.61, −1.91	-
Follow-up vs. pre	−3.52(.47)	−4.44, −2.60	−0.76 [−1.66, 0.13]	.10
Baseline distress			−0.59 [−0.70, −0.47]	<.001
Age			−0.03 [−0.09, 0.03]	.35
Gender
Male v Female			0.19 [−1.48, 1.85]	.83
Website version
Version 2 v. Version 1			−0.93 [−3.24, 1.39]	.43
Intercept			−2.71 [−3.78, −1.65]	<.001
Change in mental wellbeing from baseline
Time
Post vs. pre	2.14(1.00)	0.18, 4.09	-
Follow-up vs. pre	2.94(1.08)	0.83, 5.04	0.80 [−1.11, 2.71]	.41
Baseline mental wellbeing			−0.33 [−0.47, −0.19]	<.001
Age			−0.06 [−0.21, 0.09]	.42
Gender
Male v. Female			−1.64 [−5.56, 2.28]	.41
Website version
Version 2 v. Version 1			1.24 [−4.18, 6.66]	.65
Intercept			2.60 [0.12, 5.08]	.04

^aUnstandardized regression coefficients.

In support of H2, Table 4 also indicates that a decrease in distress was greater in individuals reporting higher baseline distress ($\hat{\beta }$ = −0.59, 95%CI = [−0.70, −0.47]). Similarly, improvement in mental wellbeing was greater in individuals with lower baseline mental wellbeing ($\hat{\beta }$ = −0.33, 95%CI = [−0.47, −0.19]).

Relationship of these effects with ACT-based processes

As shown in the Supplementary Table E, linear regressions indicate that from pre- to post-intervention, psychological inflexibility, cognitive fusion and automatic thoughts significantly decreased, and mindfulness significantly increased. Coping via acceptance increased, but this was only borderline significant (P = .06).

As shown in Table 5, in support of H3, a decrease in psychological inflexibility from pre- to post-intervention was associated with a decrease in distress from pre- to post-intervention ($\hat{\beta }$ = 0.45, 95%CI = [0.33, 0.57]). Decreases in distress were also associated with decreases in cognitive fusion, believability of thoughts, and increases in coping via acceptance, and mindfulness, also in support of H3.

Table 5. Associations with change from baseline ACT variables for distress and mental wellbeing (n = 77).

	Univariable models
	$\hat{\beta }$ [95% CI]	P value	$\hat{\beta }$ [95% CI]	P value	$\hat{\beta }$ [95% CI]	P value	$\hat{\beta }$ [95% CI]			P value	$\hat{\beta }$ [95% CI]	P value	$\hat{\beta }$ [95% CI]	P value
Change in distress from baseline
Intercept	−0.64 [−1.66, 0.37]	.21	−1.49 [−2.53, −0.45]	.01	−2.32 [−3.48, −1.16]	<.001		−0.70 [−1.98, 0.58]		.28	−1.89 [−3.10, −0.67]	.003	−2.63 [−3.77, −1.48]	<.001
Age	−0.02 [−0.08, 0.04]	.48	0.00 [−0.06, 0.07]	.90	−0.00 [−0.08, 0.07]	.95		−0.01 [−0.08, 0.05]		.66	−0.01 [−0.08, 0.06]	.78	−0.01 [−0.08, 0.06]	.78
Gender
Male v Female	−0.14 [−1.60, 1.32]	.85	0.45 [−1.15, 2.05]	.58	0.29 [−1.62, 2.20]	.77		−0.59 [−2.24, 1.07]		.48	0.00 [−1.88, 1.90]	.99	0.39 [−1.53, 2.31]	.69
Website version
Version 2 v. Version 1	−1.78 [−3.79, 0.23]	.08	−1.25 [−3.49, 0.99]	.27	−2.44 [−5.10, 0.22]	.07		−0.50 [−2.80, 1.81]		.67	−1.68 [−4.26, 0.91]	.20	−2.17 [−4.83, 0.50]	.11
Baseline measures
Distress	−0.72 [−0.86, −0.57]	<.001	−0.69 [−0.86, −0.52]	<.001	−0.58 [−0.71, −0.44]	<.001		−0.76 [−0.91, −0.61]		<.001	−0.62 [−0.79, −0.46]	<.001	−0.71 [−0.92, −0.50]	<.001
Psych. inflex.	0.37 [0.23, 0.52]	<.001
Cognitive fusion			0.29 [0.14, 0.43]	<.001
Coping via acceptance					−0.18 [−0.64, 0.28]	.43
Automatic thoughts								0.15 [0.10, 0.21]		<.001
Mindfulness											−0.12 [−0.23, −0.02]	.02
Neuroticism													0.17 [−0.02, 0.37]	.08
ACT change scores
Psych. inflex.	0.45 [0.33, 0.57]	<.001
Cognitive fusion			0.36 [0.23, 0.49]	<.001
Coping via acceptance					−0.43 [−0.78, −0.07]	.02
Automatic thoughts								0.14 [0.08, 0.19]		<.001
Mindfulness											−0.16 [−0.27, −0.05]	.005
Intercept	1.04 [−1.72, 3.80]	.46	1.70 [−0.92, 4.31]	.20	2.21 [−0.22, 4.65]	.07		−1.24 [−4.24, 1.76]	.41		1.80 [−0.78, 4.37]	.17	2.91 [0.33, 5.48]	.03
Age	−0.02 [−0.17, 0.13]	.79	−0.04 [−0.19, 0.12]	.62	−0.01 [−0.16, 0.15]	.94		−0.03 [−0.17, 0.12]	.73		0.01 [−0.15, 0.16]	.93	−0.01 [−0.17, 0.16]	0.95
Gender
Male v Female	−2.50 [−6.52, 1.53]	.22	−2.51 [−6.58, 1.56]	.22	−1.70 [−5.74, 2.34]	.40		−0.77 [−4.64, 3.11]	.69		−2.72 [−6.88, 1.43]	.20	−2.13 [−6.43, 2.16]	.33
Website version
Version 2 v. Version 1	0.45 [−5.12, 6.03]	.87	−.17 [−5.84, 5.50]	.95	0.88 [−4.67, 6.43]	.75		−3.02 [−8.53, 2.49]	.28		−0.21 [−5.85, 5.42]	.94	0.31 [−5.59, 6.20]	.92
Baseline measures
Mental wellbeing	−0.33 [−0.53, −0.13]	.002	−0.28 [−0.48, −0.08]	.007	−0.35 [−0.49, −0.20]	<.001		−0.52 [−0.70, −0.33]	<.001		−0.34 [−0.53, −0.16]	.001	−0.34 [−0.55, −0.13]	.002
Psych. inflex.	−0.06 [−0.47, 0.35]	.76
Cognitive fusion			0.03 [−0.31, 0.36]	.88
Coping via acceptance					−0.55 [−1.51, 0.41]	.26
Automatic thoughts								−0.23 [−0.37, −0.10]	.001
Mindfulness											0.07 [−0.18, 0.31]	.59
Neuroticism													0.07 [−0.31, 0.45]	.71
ACT change scores
Psych. inflex.	−0.49 [−0.82, −0.15]	.005
Cognitive fusion			−0.40 [−0.72, −0.08]	.02
Coping via acceptance					0.83 [0.08, 1.58]	.03
Automatic thoughts								−0.31 [−0.44, −0.17]	<.001
Mindfulness											0.37 [0.12, 0.62]	.005

^aUnstandardized regression coefficient.

Additionally, Table 5 indicates that a decrease in believability of thoughts from pre- to post-intervention was associated with an increase in mental wellbeing from pre- to post-intervention ($\hat{\beta }$ = −0.31, 95%CI = [−0.44, −0.17]). Improvements in mental wellbeing were also associated with a decrease in psychological inflexibility, cognitive fusion, and an increase in coping via acceptance and mindfulness (Table 5).

The role of neuroticism

In support of H4, neuroticism was positively associated with distress ($\hat{\beta }$ = 0.71, 95%CI = [0.61, 0.80]) and negatively associated with mental wellbeing ($\hat{\beta }$ = −1.13, 95%CI = [−1.32, −0.95]) at baseline (Table 6). Additionally, in support of H4, Table 5 indicates that baseline neuroticism was not associated with the change in either distress ($\hat{\beta }$ = 0.17, 95%CI = [−0.02, 0.37]) or mental wellbeing ($\hat{\beta }$ = 0.07, 95%CI = [−0.31, 0.45]) over time.

Table 6. Adjusted linear multiple regressions for associations between baseline distress, wellbeing and neuroticism.

	$\hat{\beta }$ [95% CI]	P value
Baseline distress
Intercept	−15.63 [−17.84, −13.42]	<.001
Baseline neuroticism	0.71 [0.61, 0.80]	<.001
Age	−0.03 [−0.09, 0.02]	.24
Gender
Male v. Female	1.28 [−0.11, 2.67]	.07
Website version
Version 2 v. Version 1	−0.79 [−2.32, 0.74]	.31
Baseline well-being
Intercept	25.95 [21.53, 30.36]	<.001
Baseline neuroticism	−1.13 [−1.32, −0.95]	<.001
Age	−0.07 [−0.18, 0.04]	.23
Gender
Male v. Female	−3.85 [−6.62, −1.09]	.007
Website version
Version 2 v. Version 1	0.51 [−2.53, 3.56]	.74

^aUnstandardised regression coefficients.

Secondary outcomes: satisfaction and usability of the website

In support of H5, average satisfaction post-intervention was 26.92 (SD = 4.12) on a scale of 8 to 32. The majority of participants (94.6%) would recommend the website to a friend in need of similar advice and tools. Average usability post-intervention was 84.70 (SD = 13.14) on a scale of 0 to 100.

Discussion

These findings suggest that use of the ifarmwell intervention was associated with decreasing distress and improving mental wellbeing over time. Moreover, these effects were maintained at 6 months follow-up, showing that the development of skills within the intervention appeared to have positive, lasting effects on farmers’ wellbeing. Given the low cost and potential broad geographical reach of this intervention, this is a promising achievement, with significant social and economic implications for rural Australia and beyond, given the current burden of untreated distress in farming communities.

As expected, greatest reductions in distress and improvements in mental wellbeing were achieved for those farmers who started the intervention with very high levels of distress and low levels of mental wellbeing. These findings provide reassurance that ifarmwell is a safe and appropriate online intervention to recommend for farmers who are currently experiencing poor mental health. The ifarmwell intervention is designed to teach transferable coping strategies to all farmers, and these findings also suggest that benefits were not limited to farmers with only mild distress. The results also indicate that participants scoring highly on neuroticism were more likely to be distressed and have poorer mental wellbeing at the start of the intervention, consistent with our past research.²⁰ However, neuroticism did not influence the effectiveness of the intervention, further suggesting that ifarmwell is appropriate to recommend broadly, across the farming population.

A key strength of this work is the fact that we not only examined if the ifarmwell intervention was associated with reductions in distress and improvement in mental health (our primary outcome measures), but we also examined how this occurred by examining potential mechanisms. These analyses demonstrated that shifts in psychological inflexibility, cognitive fusion, the believability of thoughts, mindfulness and coping via acceptance (all of the ACT constructs targeted in the modules, apart from values targeted in module 3 which we were unable to analyze due to missing data), were associated with these improvements. This is largely consistent with a recent systematic review of mechanisms of change for ACT-based interventions, which identified psychological inflexibility and acceptance as key mediators on a number of outcomes.³⁵ Interestingly, the results of that review indicated that cognitive fusion is generally not a significant mediator, in contrast to the current findings. It is possible that because farmers tend to spend long periods of time on their own, they are more likely than the general population to fuse with their thoughts, and therefore benefit from defusion to a greater degree. Given this finding, further exploration of the types of thoughts farmers tend to fuse with, and the extent of their fusion compared to other populations, is worth exploring. It was unfortunate that the role of values could not be examined due to missing data. Our previous examination of ifarmwell indicated that the values module (module 3) was rated lower on user satisfaction than the other components of the intervention.¹⁵ Given this, future exploration of farmers’ values and how they influence distress and well-being is warranted.

Together, these results confirm the relevance of an ACT-based therapeutic approach for farmers and support its application to an online format. Additionally, previous research has linked psychological inflexibility to suicide risk in various contexts.^36,37 Therefore, these results indicate the potential for the ifarmwell intervention to help to prevent suicide in the at-risk farming population, although further exploration of the direct impact of the intervention on suicidal ideation and behavior would be useful.

It is acknowledged that the present study was potentially limited by the lack of a control group, so we cannot rule out the possibility that participants’ distress and wellbeing would have improved over time regardless of their participation in ifarmwell. Given farmers’ involvement in the co-design of this resource and the strong push for the intervention to be made available as soon as possible, due to the great unmet need in this group, it was deemed impractical and unethical to implement a control group in this instance. Fortunately, our findings on the mechanisms at work within the intervention strengthen confidence in our pre-post findings, by suggesting that it was the successful development of these targeted skills that led to improvements, rather than natural improvement over time. Additionally, in the absence of targeted interventions specifically addressing distress, we do not expect farmers’ distress to decrease over time. For instance, analysis of a recent online intervention addressing suicide stigma in Australian farmers found no change in the Isolation-Depression measure pre- to post-intervention.³⁸ Similarly, a Scottish online Cognitive Behavior Therapy-based life skills course for farmers (with weekly, personalized email support from a counsellor) found no significant change in depression or daily functioning, over the course of their intervention.³⁹

It should be noted, since all users of the website were required to participate in the research evaluating ifarmwell, the adherence rate reported here (33.8%) does not distinguish between adherence to the website versus adherence to the research. Nonetheless, the adherence rate reported here is higher than that reported in a web-based CBT life skills course for farmers (adherence rate 14%)³⁹ and an internet-based intervention targeting depression in “green professions” such as farmers (adherence rate of 22%).⁴⁰ Importantly, the current study showed that satisfaction with ifarmwell and overall usability was high, and an overwhelming majority of the participants (94.6%) who completed the post-intervention questionnaire would recommend this website to a friend. While these ratings are similar to other studies of web-based interventions, for example to aid the management of depressive symptoms in adults with diabetes,⁴¹ given the aforementioned challenges in engaging the farming population in mental health and wellbeing focused interventions, this is a pleasing result.

Despite these affirming results for ifarmwell, barriers to extending new approaches – even if their value has been scientifically proven – are well known in the agricultural sector.⁴² If an ambitious dissemination goal is to be achieved, we will need to continue to build upon lessons from agricultural extension,⁴³ educational theory,⁴⁴ implementation science,⁴⁵ our own research on what farmers want from online mental health and wellbeing interventions^46,47 and how ifarmwell can assist key stakeholders such as rural financial counsellors,⁴⁸ lessons from other parts of the world,^49,50 and, importantly, the existing strengths of farming communities.

Conclusions

This study provides the first published evidence that a fully automated online intervention, co-designed with farmers, for farmers, can reduce farmers’ distress and improve their wellbeing. The next challenge is to increase uptake of ifarmwell in the broader Australian farming community, so that less farmers experience preventable and/or treatable distress. Based upon the demographic profile of the farmers who took part in this study, targeted promotion of the resource at men and younger farmers may need to be a particular focus. Further, given predictions that Australian farmers are likely to be exposed to more extreme climate events in the future,¹³ this is a timely future-proofing, resilience-building and suicide-prevention tool, that is likely to be of interest to governments and industry groups.

There are also implications from this work for farmers in other parts of the world, who also face many stressors beyond their control, and are likely to benefit from the ACT-based strategies contained within the ifarmwell intervention. Findings may also inform the development of additional ACT-based online interventions for other at-risk populations, who face similar or different (e.g. COVID19-related) challenges with accessing face-to-face based mental health education and care.

Authors contributions

KMG: Conceptualization, Methodology, Funding Acquisition, Supervision, Project Administration, Data Collection, Writing- original draft; GS: Data Analysis (quantitative), Writing- original draft; JD: Conceptualization, Methodology, Funding Acquisition, Writing – Review & Editing; ADV: Conceptualization, Methodology, Data Analysis (quantitative), Writing – Review & Editing; SB: Conceptualization, Methodology, Writing – Review & Editing; CES: Conceptualization, Methodology, Writing – Review & Editing; DT: Conceptualization, Methodology, Writing – Review & Editing.

Acknowledgment

We gratefully acknowledge the farmers who took part in this research and assisted with the development and testing of the intervention and Mrs Margaret Mc Gee for her assistance with supporting this project.

Disclosure statement

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

Additional information

Funding

This project was funded by the NAB Foundation, University of South Australia, Freemasons Centre for Male Health and Wellbeing and the Freemasons Foundation.