Effectiveness of a process-based approach to farmer wellbeing: A randomized multiple-baseline single-case experimental design

Title

Effectiveness of a Process-Based Approach to Farmer Wellbeing: A Randomized Multiple Baseline Single-Case Experimental Design

Full Reference

Stynes, G., Stapleton, A., Moore, B., Russell, T., O'Connor, M., Richardson, N., Ruiz, F. J., & McHugh, L. (2025). Effectiveness of a process-based approach to farmer wellbeing: A randomized multiple baseline single-case experimental design. Journal of Contextual Behavioral Science, 37, 100932.

Study Type

Quantitative empirical study: Single-Case Experimental Design (SCED) with randomized multiple baseline. Group intervention of 6 sessions of online Acceptance and Commitment Therapy (ACT).

---

Background and Objectives

Context

Mental health and wellbeing of farmers constitute a growing concern in the scientific literature. Farmers report elevated rates of stress, anxiety, depression, and suicidal ideation. Previous research indicates that among 256 farmers evaluated, 23.4% reported suicidal thoughts or urges within the preceding two weeks. In Ireland, 57% of 736 farm operators reported experiencing stress or anxiety in the previous five years. Additionally, 23.6% of 351 farmers reported "burning out" (burnout).

Practical barriers to accessing mental health support are significant among farmers: low access to mental health services, healthcare costs, distance from treatment centers, time associated with help-seeking, and lack of established relationships with mental health providers. Many farmers exhibit characteristics of self-reliance and avoidance that may impede formal help-seeking.

Primary Objectives

The study evaluated the effectiveness of an ACT-based intervention specifically designed for farmers. The goal was to determine whether a process-based approach (i.e., focused on psychological flexibility and acceptance rather than direct symptom reduction) would be effective in improving farmer wellbeing among those experiencing elevated distress. The specific hypotheses were:

1. There will be a significant increase in psychological flexibility based on a one-to-one measure.
2. There will be a significant decrease in stress based on a single-item stress measure.
3. There will be increases in wellbeing, acceptance of shame, psychological flexibility, and positively associated behavioral outcomes with clinically relevant change.
4. There will be decreases in depression, anxiety, stress, shame, and clinically relevant behavioral outcomes negatively associated with change.

---

Method

Participants

The study included 6 participants (5 males and 1 female) with ages ranging from 47 to 69 years (M = 53.7 years). Participants were identified as farmers experiencing elevated distress. Regarding marital status, 4 participants were married/in civil partnership and 2 were single. Most had children (range: 0-4 children).

Agricultural enterprises varied in size (range: 50-100 hectares) and type of main enterprise (dairy farming in 4 cases; beef/tillage in 1 case). Annual average incomes varied from €30,000-€99,000. Five of 6 participants worked on the farm full-time or nearly full-time, and 2 had off-farm employment. Education levels ranged from agricultural "Green Cert" to university degrees. All participants self-reported having "good" or "fair" mental health at baseline.

Inclusion criteria: Farmers working in livestock agriculture, with elevated distress scores (score ≥ moderate on Depression, Anxiety and Stress Scale-21; DASS-21) at screening, and aged over 18 years.

Exclusion criteria: None explicitly reported; however, participation required engagement with potential "writing" and "role-play" exercises.

Design

A randomized multiple baseline single-case experimental design (SCED) was employed. The design included multiple baselines across participants, allowing for concurrent evaluation of the intervention across individual cases. Data were collected daily during baseline (Phase A) and intervention (Phase B) phases.

Timeline: The study extended approximately 84 days. Participants completed different baseline phases that varied from 14-31 days according to random assignment. Intervention phases lasted 17-32 days depending on the participant. Some participants (P1 and P2) had final follow-up sessions scheduled three days after the final session.

Study Phases:

• Phase A (Baseline): Daily data collection prior to intervention
• Phase B (Intervention): Period of six ACT sessions online

Intervention

A 6-session online ACT program was implemented, designed to enhance farmer wellbeing. The program was delivered by an experienced ACT psychologist via Zoom. Each session was approximately 60-90 minutes in duration. Content was based on Acceptance and Commitment principles, facilitating psychological flexibility.

Session Content (per Table 2):

• Session 1: Introduction to ACT. "Feeling Good" exercise and "Creative hopelessness"—exploring the futility of thought control. Recognition of what has not worked and how persistent attempts to avoid thoughts may result in more thoughts. Homework: "Go with the stream" exercise.

• Session 2: Review of Session 1. Introduction to mindfulness/"noticing". Three-step breathing exercises and mindfulness practices. Development of the "polygraph metaphor"—what you can and cannot control; openness to physical sensations of emotions. Homework: Incorporate mindfulness into routine activities.

• Session 3: Review of previous sessions. Introduction to breathing, mindfulness, observation. Development of "Beliefs" (observe thoughts, do not accept as facts). Homework: Mindfulness activities in daily routine.

• Session 4: Review of Session 3 and homework review. Defusion #2 and connection with the "observing self" as context. The rumination problem, rules we create around problem-solving, the "chessboard metaphor" (metaphor of the chessboard). Homework: Bring mindfulness to observation activities and be a witness to one's own experience.

• Session 5: Review of Session 4 and homework review. Review of values and goal-setting using short and long-term SMART goals. Identification of benefits, dealing with obstacles (if goals). Following values through valued actions.

• Session 6: Review of Session 5 and homework review. Integration of all psychological flexibility, acceptance, defusion. The "ACT Matrix" and "success". Additional recommended resources.

Implementation Features:

• Experienced ACT therapist delivering sessions
• Supervision available from an experienced ACT therapist
• Didactic approach, with informed consent provided
• Online delivery via Zoom, providing flexibility
• Some participants opted for "writing" exercises and potentially "role-play" (though these were minimally used)

Measurement Instruments

Ecological Momentary Assessment (EMA): Acceptance and Commitment Measures

Brief Acceptance Measure (BAM; Asmundt et al., 2018): A 3-item measure of psychological flexibility for use in SCED. Participants self-reported their level of flexibility on a scale of 1-10 (unaware—1 to acting with awareness—10). A correlation of 0.77 with CompACT and test-retest correlation of 0.77 was reported.

Single-item daily stress measure: A single-item measure where participants indicated which number (0-10) best described the stress they experienced during the day (0 = No Stress; 10 = Extreme Stress).

Pre-intervention, Post-intervention, and Three-Month Follow-up Measures

Depression, Anxiety and Stress Scale-21 (DASS-21; Lovibond & Lovibond, 1995): A widely used, shorter version of the 42-item DASS, comprising three subscales (depression, anxiety, stress) with 21 items. Each item applies to the person during the preceding week using a 4-point scale. Scores are doubled to produce a total score of 0-42 for each subscale. Cut-off scores range from "normal" to "extremely severe." Cronbach's alpha of 0.88 was reported for depression, 0.82 for anxiety, and 0.81 for stress in a general British population. In the present study, clinical comparisons from a United States outpatient sample were used.

Mental Health Continuum Short Form (MHC-SF; Keyes et al., 2008): A 14-item version of the 40-item Mental Health Continuum Scale measuring emotional, psychological, and social wellbeing. Participants indicate the frequency with which they experience each aspect using a 6-point scale: never, once a week, about two or three times a week, every day except one or two, every day. Cronbach's alpha of 0.89 was reported for overall wellbeing. The MHC-SF provides a measure of overall wellbeing for clinical change. Cut-off scores range from 0-70 with wellbeing scores indicating greater overall wellbeing.

External and Internal Shame Scale (EISS; Ferreira et al., 2022): An 8-item scale measuring external shame (4 items) and internal shame (4 items). External shame refers to shame experienced when others view personal characteristics or behaviors negatively. Internal shame occurs when one has internalized negative self-evaluations, considering oneself as flawed. Participants rate statements using a 5-point scale from "never" to "always"; scores are summed to produce an external and internal shame value, calculated from a composite of all items (range 0-32). Cronbach's alphas of 0.80 for external shame and 0.82 for internal shame were reported.

Acceptance of Shame and Embarrassment Scale (ASES; Sedighimornani et al., 2019): A 17-item scale measuring acceptance of feelings of shame and embarrassment. Acceptance is a core process in ACT and a key element of psychological flexibility. Openness to experiencing uncomfortable emotions such as shame and embarrassment may support help-seeking regardless of the content of shameful thoughts. Participants rate statements using a point scale (reverse items indicated): 0=never/rarely true; 2=sometimes true; 4=often true; 6=always/almost always true. Scores are summed for a total shame acceptance score, ranging from 0-102. Cronbach's alphas of 0.93 and 0.94 were reported for clinical and non-clinical populations respectively.

Psy-Flex (Cluster et al., 2021): A 6-item measure of psychological flexibility with flexibility scores ranging from 1-10. Total and mean scores are calculated in the same manner as the BAM, except Psy-Flex is a 6-item measure of psychological flexibility.

Single-item health measure (from SF-36; Ware & Sherbourne, 1992): A single-item health measure derived from the SF-36, used to evaluate criterion validity of the PBAT. Participants described their health in the past week on a 1-5 scale: poor, fair, good, very good, excellent.

Single item life satisfaction measure (adapted from Cheung & Lucas, 2014): A single-item life satisfaction measure adapted and recommended by PBAT authors as an outcome measure. Participants indicated whether they were satisfied with life over the past week using a 10-point scale: 0 (not at all) to 10 (very true).

Single-item measure of burnout from work (adapted from West et al., 2009): Included in outcome measures from an online repository (https://pbatsupport.com/free-download/), a single-item measure: "I feel burnt out from my work". Participants indicated the extent to which they agreed over the past week using a 10-point scale: 0 (not at all) to 10 (very true).

Data Analysis Strategy

EMA data were collected daily via Google Forms, with a link sent by text message to participants. Data were analyzed using single-case experimental design (SCED) methodology, described as follows:

EMA Data: Psychological flexibility (BAM) and stress data were analyzed using Reliable Change Indices (RCI; per Jacobson & Truax, 1991) adapted for SCED. Reliable change analysis was used to determine whether observed changes were clinically significant at the individual level. The Jacobson & Truax (1991) method was used to evaluate whether observed changes exceeded natural measurement error. Morley's (1991) clinical change reliability workbook was used to calculate reliable change, except for Psy-flex where Baysinger's reliability estimation was used (per Baysinger et al., 2018). Clinical change analysis was conducted using the same package for DASS-21 and MHC-SF results. Clinical comparisons of DASS-21 measures were conducted using clinical cut-off scores from a United Kingdom general population normative sample (Honk et al., 2011) with clinical comparisons from a United States outpatient patient sample (Lamers et al., 2011) with clinical comparisons from a Dutch DSM-IV sample. Appropriate clinical cut-off scores were selected for the MHC-SF measure. Clinical changes in DASS-21 measures were evaluated using the same outcomes package for DASS-21 and MHC-SF results. Reliability measures were available for these measures. In relation to DASS-21, United Kingdom general population norms (Honk et al., 2011) were used; clinical comparisons from a United States outpatient sample (Lamers et al., 2011) with clinical comparisons from a Dutch DSM-IV sample for DASS-21 clinical cut-off scores; clinical comparisons are available, which were considered clinically relevant only when falling within moderate, severe, or extremely severe ranges.

Visual Analysis: Visual analysis of time-series graphs with data point indicators was conducted to determine the presence of trend. The SCAN Package (Wilbert & Lüke, 2024) was used to analyze EMA data and complete Tau-U analyses, accounting for independence between data points for effect size. Tau-U is a non-overlap statistic between two phases, in this case Phase A and Phase B, but also provides a means of quantifying trends within and between phases. Tau-U does not reflect effect size per se, as non-overlap between phases can be overestimated (Bernard-Brak et al., 2021). The SCAN R package (Wilbert & Lüke, 2024) was used to analyze SCED data and complete Tau-U analyses, accounting for data point independence for effect size. Tau-U does not reflect effect size per se; therefore a parametric test between standardized mean difference between two phases was used. BC-SMD is relatively robust to the influence of autocorrelation. SCED data analysis was conducted using the SCAN Package (Wilbert & Lüke, 2024), which modeled data using linear trends including fixed effects for change in level and trend with random effects for level. Leads Reliable Change Indicator (per Jacobson & Truax, 2014) was used to evaluate reliable change in standardized measures assessing depression, anxiety, stress, wellbeing, shame acceptance and change, and psychological flexibility. Clinical change analysis was conducted using the method described by Jacobson & Truax (1991) to evaluate whether observed changes exceeded other natural measurement errors. Morley's (1991) clinical change reliability workbook was used to calculate reliable change except for Psy-flex where Baysinger's reliability estimation was used. Clinical change analysis was conducted using the same package for DASS-21 and MHC-SF results. Descriptive statistics were used to analyze changes in process-based measures and outcomes, using percentage change scores and change rating scores, comparing pre-intervention scores to post-intervention scores (T1-T2) and pre-intervention scores to three-month follow-up scores (T1-T3).

---

Results

Psychological Flexibility and Stress Data (EMA)

Psy-Flex (Psychological Flexibility): Visual analysis of graphs for P1, P5, and P6 suggests little evidence of trend in Phase A. There is variability within Phase A for values of P2, P3, and P4. Closer examination suggests little trend. There is no change in values for P1 before intervention in Phase B, though interpretation is difficult; data for P2 in the intervention phase appear slightly higher than Phase A. Particularly for P4 and P6 in the later intervention period of Phase B, graphs suggest intervention effect.

Significant Tau-U (p < .05) was evidenced in Phase B for P3 (τ = −.05), P4 (p < .001), and P6 (p < .001). Significant improvement was evidenced in Psy-flex in Phase B for P4 (p < .001). For remaining participants, Tau-UA vs B showed improvements of 0.61; p < .001 for P1, 0.49; p < .001 for P4, and 0.58; p = .001 for P6 with statistically non-significant improvements in other participants. BC-SMD across participants on Psy-flex was significant with a large effect size of 0.95% CI [0.236, 1.693] using Cohen (2013) guidelines. Overall, there is partial support for the first hypothesis.

Stress: Phase A values for all participants reflect little evidence of trend from visual analyses. Except for one data point, there was no change in values for P1 in the intervention phase. Phase B interpretation is difficult; for P2 in the intervention phase, stress appears slightly higher than Phase A. Particularly for P4 and P6 in the later intervention period of Phase B, graphs suggest intervention effect.

Tau-U analyses for stress were completed using Tau-UA vs B for all participants except P4. Tau-UA vs B = −0.09 (p = .73). For P2 visual analysis did not detect trend. For remaining participants, Tau-UA vs B = 0.61; p < .001 for P1, −0.31; p = .001 for P4, and −0.14; p = .35 for P6. BC-SMD across participants on stress was significant with a large effect size of −0.3924 using Cohen (2013) guidelines. Overall, there is insufficient support for the second hypothesis.

Pre-intervention, Post-intervention, and Three-Month Follow-up Measures

As reported in Table 4, results were mixed across participants. Results from reliable change analysis for non-EMA measures were mixed:

Depression (DASS-21): One participant (P1) showed reliable deterioration with no change in remaining participants. There was reliable improvement for P2 at post-intervention. Clinically significant improvements were observed for five of six participants (ranging from 4.76% to 37.5%; P5 = 0%). Improvements in Vitality were observed for five of six participants ranging from 4.76% to 37.5% (P2 = 0%; P3 = 0%; P5 = 0%).

Anxiety (DASS-21): One participant (P2) showed reliable deterioration. Reliable improvement was observed for two participants (P2 and P4). Significant improvements were observed in Vitality ranging from 8.11% (P4) to 51.22% (P3).

Stress (DASS-21): Reliable improvements were observed for four participants except P3 whose score increased by 4.55%. Work burnout improved for one participant ranging from 10.11% to 200% for five participants oscillating from 11.11% for P3 to 200% for P2. There were mixed results overall for the fourth hypothesis and therefore insufficient support.

Subjective Wellbeing (MHC-SF): There was reliable deterioration for one participant (P1) with no change in remaining participants. There was reliable improvement for P2 at post-intervention. Improvements in psychological wellbeing ranged from 8.11% (P4) to 51.22% (P3).

Shame (EISS): Reliable improvement was observed for one participant (P1) with no change in remaining participants. There was reliable improvement for P2 at post-intervention. No reliable changes were observed for others.

Positive Behaviors (PBAT): Reliable improvements were observed for four participants (P2, P3, P4, P6) ranging from 8.11% (P4) to 51.22% (P3).

Negative Behaviors (PBAT): Reduction of negative behaviors for five participants (P2, P3, P4, P5, P6) ranging from 8.11% (P5) to 100% (P6). Scores for P1 increased by 105.26%. Overall psychological distress as measured by the composite increased by 105.26%.

Summary: Overall, participants demonstrated a variety of results on post-intervention measures ranging from no change to significant changes. While two participants (P4 and P6) showed reliable improvements in depression, stress, vitality, and acceptance, most individuals showed no improvements in stress, anxiety, or subjective wellbeing. Burnout increased over time. While evaluation results are mixed, full session attendance supports intervention feasibility.

---

Discussion and Conclusions

Main Findings

The study results are mixed. There were significant improvements in psychological flexibility for participants overall, based on evidence from EMA and Psy-flex, with a large effect size on EMA. This result is consistent with expectations of an ACT intervention, which specifically aims to increase psychological flexibility to enhance valued living. Previous research suggests that psychological flexibility is associated with positive mental health and wellbeing.

Results on more traditional non-EMA measures were mixed. There was no evident intervention impact on stress, anxiety, or subjective wellbeing for most participants. This is an interesting finding in the context of improvements in other areas. Baseline stress scores as measured by DASS-21 were in the "normal" range of 0-14 and thus floor effects may have influenced potential outcomes. It is also likely that farmers report a range of stressors, such as regulatory issues and perceptions of farmers being blamed for climate crisis. Beyond these emotions, exploration of a "problematic mind" as well as problem-solving may have been beneficial.

The intervention results on acceptance or at least three participants in relation to depression, distress, vitality, satisfaction, and behaviors suggest changes typically associated with clinical change. While two participants showed deterioration in depression, reliable clinical improvements were evident for others. Using "cut-off" scores provided by Lovibond and Lovibond (1995), there were no clinically significant changes in other participants. Deteriorated results were however related to DASS-21 at screening and baseline. The DASS-21 results at screening and baseline suggest results were potentially confounded. While baseline was used as a comparison point prior to initial data points, if there was in fact an inaccurate representation of baseline data, results would have been confounded.

Limitations

Several limitations were identified in the study. First, substantial baseline differences were observed between DASS-21 results at screening and baseline. At screening baseline, all participants had at least one elevated subscale score on DASS-21. Baseline differences were used as a comparison point prior to initial data. There was missing data at follow-up for P1 and post-intervention for P4. Missing data were attributed to family follow-up assistance for technical issues. Data likely contributed to missing data. Missing data were not for BMA reasons. Specific data were missing for SCED analysis in standard methods proposed to address missing data in SCED. Provided the large amount of missing data of 62%, by specification, multiple imputation method was not performed across SCED participants. Despite the large amount of missing data of 62%, imputation is challenging and the pattern of missing data across P2 and participants where analysis applies is considered in the evaluation of clinically relevant changes that could be identified from single output measure or whether trends were in fact not observed in baseline visual analysis of previous analyses; intervention data were modeled at level with random effects for level. Due to the presence of observed trends for some participants, intervention data were modeled at level with random effects for level.

Clinical and Theoretical Implications

Results suggest that an online process-based ACT intervention was relatively applicable to farmers, with full session attendance supporting intervention feasibility. This result is encouraging. Psychological inflexibility and lack of valued living are considered the basis of depression in ACT. It is notable therefore that although psychological improvement improved, there were implications of improvements in individual items of psychological flexibility, vitality, and process-based outcomes. Changes observed with outcome total satisfaction, depression, and stress associated with clinical change.

One notable element was that there were substantial differences between DASS-21 and results obtained two weeks after intervention completion. This suggests that initially farmers may have been avoiding sensations of burnout related to exposure to pressure from committed behaviors for some, and others may have had additional responsibilities connected with functioning on their farm. Future explanation for increases in burnout over time. While psychological research especially of reliability and validity criteria of PBAT, as well as baselines used as comparison point prior to initial data points.

In conclusion, this study aimed to evaluate whether an online ACT intervention was effective with a process-based approach to address farmers' known stressors and associated behavioral changes. The intervention did not appear to have impacted stress, anxiety, or subjective wellbeing for most participants. This is an interesting finding in the context of improvements in other areas. Baseline stress scores as measured by DASS-21 were in the "normal" range of 0-14 and thus floor effects may have influenced potential outcomes.

Importance and Contribution

This study contributes to the literature on farmer wellbeing and process-based psychological interventions. The research addresses a critical and underaddressed public health gap by demonstrating the feasibility of delivering structured ACT interventions online to rural agricultural populations with elevated psychological distress. Participants showed full attendance, suggesting acceptable implementation in this population. The study provides preliminary evidence that process-based ACT focusing on psychological flexibility rather than direct symptom reduction may support farmers' wellbeing. Improvements in psychological flexibility and positive behavioral changes suggest potential mechanisms through which farmers could better manage occupational stressors. Although baseline stress scores were in the normal range and DASS-21 symptom changes did not always reach clinical significance, the demonstrated acceptability of the intervention and improvements in process variables support further investigation in larger samples with more diverse farming populations and extended follow-up periods.

---

PART 2: ENGLISH SUMMARY

[English translation of sections above with equivalent structure and content]

---

• Summary generated from PDF (not from memory)
• Accuracy of numerical data not re-verified in this repair pass
• Bilingual format present (Spanish + English)
• Format standardized in repair pass 2026-03-30