Early interventions for primary care patients with stress-related mental ill health: a randomised controlled non-inferiority trial of a nurse-led collaborative care intervention versus cognitive behavioural therapy

Abstract

Background

Stress-related mental ill health is highly prevalent, and most individuals first seek help in primary care where access to psychological support is often limited.

Objective

To test whether nurse-led collaborative care (CC) is non-inferior to cognitive behavioural therapy delivered as psychologist-guided self-help (GSH-CBT) in reducing stress for adult primary care patients with stress-related mental ill health.

Methods

A randomised controlled non-inferiority trial where 172 adult primary care patients with mild to moderate stress-related mental ill health were allocated to nurse-led CC or GSH-CBT (1:1). The primary outcome was the Perceived Stress Scale-10 administered every 3 weeks during the 12-week treatment phase. The a priori non-inferiority margin was 0.3 d, requiring the one-sided 95% CI of the effect size to fall within this bound. Secondary outcomes were symptoms of exhaustion, sleep problems, depressive symptoms, general anxiety and quality of life.

Findings

In both conditions, participants showed large, significant within-group improvements on the primary outcome and all secondary outcomes. The primary outcome showed a small between-group difference (d=0.22), favouring GSH-CBT, with the CI upper bound (0.52) exceeding the non-inferiority margin. For secondary outcomes, GSH-CBT led to a greater reduction in exhaustion (d=0.28; 95% CI 0.64 to 0.002). No other between-group differences were observed.

Conclusions

Nurse-led CC was not non-inferior to GSH-CBT. Effect differences were small, but improvements were slightly larger in GSH-CBT.

Clinical implications

Given that nurse-led CC is safe, easy to administer, and that significant within-group improvements were observed across all measures, this intervention could be a viable option in primary care settings where psychologist-delivered treatment is limited.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Cognitive behavioural therapy (CBT) is arguably the most effective treatment for stress-related mental ill health, but limited access to psychologists in primary care creates barriers to early intervention.

• Nurse-led collaborative care (CC) has shown promise for mental health problems, but has not been compared against CBT for patients with stress-related mental ill health.

WHAT THIS STUDY ADDS

• This randomised controlled trial found that nurse-led CC was not non-inferior to guided self-help CBT in reducing perceived stress for primary care patients with mild to moderate stress-related mental ill health, though both interventions led to significant improvements across all symptom measures that were sustained at 1 year follow-up.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Given that nurse-led CC is safe, requires minimal clinician time (three to four visits), and produces significant symptom reduction, it may serve as a viable resource-efficient alternative in primary care settings where access to psychological specialists is limited.

Background

Stress is defined as the physiological or psychological response to internal or external stressors.¹ From an evolutionary perspective, stress serves an important function by activating behavioural systems that aid in handling demands and threats, thereby increasing the chances of survival. However, prolonged or repeated stress without sufficient recovery can contribute to medical and psychiatric disorders. Examples of symptoms that commonly develop in response to prolonged stress are disturbed sleep, depressed mood, cognitive impairments, somatoform complaints and decreased quality of life.^{2 3} Hereafter, we conceptualise this symptom profile as stress-related mental ill health, encompassing both subclinical symptoms that do not meet the criteria for a formal clinical diagnosis and more severe symptom profiles that meet diagnostic criteria for established conditions, such as adjustment disorder. Adjustment disorders are further characterised by a preoccupation with the stressor(s) or their implications and a perceived reduced ability to cope.⁴

In recent decades, stress-related mental ill health has become one of the most considerable health challenges in the Western world.⁵ Adjustment disorder is one of the most frequently used psychiatric diagnoses globally,⁶ and 30% of primary care patients report symptoms of stress and exhaustion.⁷ Stress-related disorders are one of the most common causes of sick leave and account for long sick leave periods.⁸ The costs for employers and society are large due to the high incidence of sickness absence, reduced productivity at work, and loss of labour supply.⁹ For afflicted individuals, who are predominantly women,¹⁰ stress-related mental ill health can lead to increased medical and insurance expenses and decreased income.⁵

Cognitive behavioural therapy (CBT) is the most promising treatment to reduce stress in the working population.¹¹ Unfortunately, a minority receive evidence-based treatments as the availability of psychological treatment is limited due to barriers such as a lack of resources and qualified therapists.¹²

Because early intervention can prevent long-term disability, limited treatment accessibility constitutes a major clinical challenge, underscoring the need for resource-efficient treatment models. One viable approach is delivering interventions as guided self-help (GSH), where patients receive the treatment via a book (ie, bibliotherapy) or the internet while receiving limited support from the therapist. There is solid scientific literature showing that these treatment formats are feasible and resource-efficient for delivering CBT with effect sizes on par with face-to-face treatment.^{13 14} For stress-related mental ill health specifically, several randomised controlled trials have shown that internet-based CBT is effective in reducing symptoms and increasing quality of life.¹¹

Although delivering CBT as GSH is a step towards making treatment more accessible, it remains difficult for psychologists to meet the large treatment demands, especially with the current shortage of psychologists.¹² This has sparked a call for task-shifting in mental healthcare, as indicated by the development of treatment models that can be delivered by other healthcare professionals than mental health specialists.

Collaborative care (CC) has been shown to be an effective working model for primary care patients with mental health problems.¹⁵ A recent systematic review reported that nurse-led task-shifting interventions led to significant improvements in mental health outcomes among adults with psychiatric disorders.¹⁶ Despite these promising findings, the impact of nurse-led interventions in primary care remains underexplored, underscoring the need for further research in this field. While existing studies indicate that nurse-led interventions can successfully reduce symptoms of depression and anxiety,¹⁵ there is a notable lack of research on their effects on stress-related mental ill health. CC is particularly well suited to primary care nurses’ existing skills and training, as it draws on structured problem-solving, psychoeducation and validation rather than requiring the specialist competence needed to deliver CBT.

Essential features of the CC model are that a care manager, a nurse in the current trial, provides the patient with self-management support and, when needed, guides the patient through the healthcare system. An important advantage of this treatment model is that the key clinician is a nurse, which is typically a cornerstone profession in primary healthcare contexts, which could enable higher accessibility to an early intervention. This type of care intervention has, however, not been compared against CBT, which is arguably the most empirically supported treatment for patients with stress-related mental ill health.¹¹

As the provision of treatment in the early phase of symptom development may reduce the risk of long-term disability,¹⁷ both CC and GSH-CBT are potentially important therapeutic tools for primary care to achieve its objective of offering access to empirically supported treatments for the large group of patients with stress-related mental ill health.

Objective

In this study, we aimed to investigate if implementation of nurse-led CC and GSH-CBT can be effective as early interventions for adult primary care patients with mild to moderate symptoms of stress-related mental ill health. Specifically, we hypothesised that a nurse-led CC intervention would be non-inferior to GSH-CBT delivered by a psychologist in reducing perceived stress. We also investigated the effects of the two treatments on psychiatric symptoms and quality of life. Lastly, we assessed treatment feasibility in terms of patient satisfaction, completion rates and negative effects.

Methods

Study design

This was a randomised controlled non-inferiority trial in which 172 consecutively recruited adult primary care patients with mild to moderate stress-related mental ill health were randomised in a 1:1 ratio to nurse-led CC (n=85) or GSH-CBT (n=87). The primary outcome was assessed at baseline, every 3 weeks, during the 12-week treatment phase, post-treatment (12-week follow-up) and 1 year follow-up. Randomisation was done by an external person using a true random number generator (www.random.org) and conducted after inclusion to prevent knowledge of forthcoming allocations. All participants provided informed consent. The trial was preregistered at Clinicaltrials.gov (Identifier NCT04820283) on 24 March 2021, a data analysis plan was preregistered at Open Science Framework on 23 September 2024 (https://doi.org/10.17605/OSF.IO/F872T) and the trial was reported according to the Consolidated Standards of Reporting Trials CONSORT 2025 checklist (online supplemental file 1).

Participants, study setting and recruitment

Participants were consecutively recruited from routine care at Gustavsberg Primary Care Center in Stockholm, Sweden, between 29 April 2021 and 11 June 2024. All 1 year follow-up data were available in July 2025. As the trial was intended to reflect the regular primary care context, there were no media advertisements and all clinicians in the study were primary care staff. Inclusion criteria required participants to (1) Be at least 16 years of age, (2) Score 20 points or higher on the 10-item version of the Perceived Stress Scale (PSS-10), (3) Report stress-related mental ill health as the principal problem with (4) The duration of problems for at least 1 month, and (5) Experience symptoms in the mild to moderate range as assessed by a psychologist in a structured clinical assessment preceding inclusion (see below). Participants also had to (6) Be able to read and write in Swedish and (7) Have access to a computer with an internet connection. Participants were excluded if (8) They had been on sickness absence for more than 2 months, (9) Had ongoing psychotic symptoms or substance abuse and/or (9) Had other ongoing concurrent psychological treatment for stress-related problems. Table 1 presents baseline sociodemographic variables of included participants and online supplemental table S1 presents common stressors.

Table 1. Baseline characteristics of included participants.

	Total (n=172)	GSH-CBT (n=87)	CC (n=85)
Gender, n (%)
Female	129 (75.0)	63 (72.4)	66 (77.7)
Age
Mean age (SD)	42.3 (11.6)	40.9 (11.3)	43.7 (11.8)
Range	16-77	16-64	20-77
Relationship and parental status, n (%)
Married/in a relationship	138 (80.2)	68 (78.2)	70 (82.4)
Parent, yes	137 (79.7)	69 (79.3)	68 (80.0)
Highest education, n (%)
College/University >3 years	75 (43.6)	39 (44.8)	36 (42.4)
College/University ≤3 years	38 (22.1)	20 (23.0)	18 (21.2)
Secondary school=3 years	38 (22.1)	18 (20.7)	20 (23.5)
Secondary school ≤2 years	14 (8.1)	5 (5.8)	9 (10.6)
Primary school	7 (4.1)	5 (5.8)	2 (2.4)
Employment, n (%)
Full-time	140 (81.4)	69 (79.3)	71 (83.5)
Part-time	15 (8.7)	10 (11.5)	5 (5.9)
Student	10 (5.8)	5 (5.8)	5 (5.9)
Pension	3 (1.7)	0	3 (3.5)
Parental leave	2 (1.2)	2 (2.3)	0
Unemployed	2 (1.2)	1 (1.2)	1 (1.2)
Sickness absence, n (%)
Full-time	38 (22.1)	21 (24.1)	17 (20.0)
Part-time	19 (11.0)	8 (9.2)	11 (12.9)
None	115 (66.9)	58 (66.7)	57 (67.1)
Symptom duration current episode (months)
Mean (SD)	14.7 (23.1)	13.5 (16.8)	16.0 (28.2)
Median	8	8	8
Range	1-240	1-120	1-240
Medication, n (%)
Antidepressants	20 (11.6)	7 (8.1)	13 (15.3)
Anxiolytics	7 (4.1)	3 (3.5)	4 (4.7)
Hypnotics	7 (4.1)	1 (1.1)	6 (7.1)
Other sleep medication	18 (10.5)	7 (8.0)	11 (12.9)
Psychiatric comorbidity*
Depression	16 (9.3)	6 (6.9)	10 (11.8)
Anxiety disorder	8 (4.7)	4 (4.6)	4 (4.7)
Insomnia	1 (0.6)	0	1 (1.2)
Other†	1 (0.6)	1 (1.2)	0

Note. All participants (100%) provided baseline data, as presented in the table above.

^*Current psychiatric comorbidity, as assessed in the clinical interview.

^†Obsessive–compulsive disorder.

CC, collaborative care; GSH-CBT, guided self-help cognitive behavioural therapy.

Patients seeking help for stress-related mental ill health were asked by healthcare professionals working at the clinic if they would be interested in participation. Those who showed interest received both verbal and written information about the study prior to signing a digital informed consent form. This form, along with several screening questionnaires, was completed via BASS4, a secure online platform provided by the eHealth Core Facility at Karolinska Institutet in Stockholm. Once this was completed, the patient underwent a structured assessment interview with a psychologist where all inclusion and exclusion criteria were assessed. The interview took approximately 60 min and followed a structured template that drew on the participant’s responses from the online screening process. Whether stress was the principal problem was judged with reference to the participant’s responses on the Adjustment Disorder – New Module 8 (ADNM-8). The assessor also evaluated whether criteria for exhaustion disorder according to the International Classification of Diseases, 11th Revision (ICD-11) were met, as this was an exclusion criterion, and whether ICD-11 criteria for adjustment disorder were met; the latter was recorded to characterise the sample diagnostically but was not required for inclusion. Beyond confirming the inclusion and exclusion criteria and establishing that stress was the participant’s principal problem, participants were asked about psychiatric comorbidity, with diagnoses assigned by clinical judgement against ICD-11 criteria. A structured diagnostic interview was available to the assessor to support differential diagnosis when judged necessary, but was not administered routinely. Participants received no compensation for their participation, but healthcare visits within the study (i.e., assessments and intervention sessions) were free of charge.

The inclusion criteria were slightly revised after 20 out of 172 participants had been included to better operationalise the patient population and enhance participant recruitment for the study. Initially, participants were required to fulfil criteria for adjustment disorder. This criterion was changed to requiring a score of 20 or higher on the PSS-10 and having stress-related mental ill health as their primary issue, with the duration of at least 1 month. Additionally, the criteria for sick leave were modified, allowing participants to have been on full-time sick leave for a maximum of 2 months instead of allowing 50% sick leave for a maximum of 1 month. These revisions were approved by the Ethical Review Authority.

Measures

Baseline variables

References for all measures other than the primary outcome are provided in the online supplemental file 1. Sociodemographic data were gathered during the screening process via the online study platform. Participants were also asked to fill in the ADNM-8 to provide details on their stressors and how they had been affected by them. Information about symptom duration, psychiatric morbidity and medication use was obtained during the clinical assessment prior to inclusion.

Primary outcome

The primary outcome was PSS-10,¹⁸ which measures perceived stress and is generally regarded as the gold standard measure for measuring stress perception.¹⁹ The instrument measures the degree to which life events are perceived as stressful in terms of being unpredictable, uncontrollable and overloading in the past month. The total scale range is 0–40 where a higher score indicates more stress. As in previous research on digital interventions for stress-related mental ill health,²⁰ we used an inclusion cut-off representing the general population mean +1 SD, which in the current trial was ≥20 based on Swedish norms.²¹ The Swedish version of the PSS-10, used in the present study, has shown good internal consistency, construct validity and sensitivity to change.²¹

Secondary outcome measures

The Shirom-Melamed Burnout Questionnaire (SMBQ), consisting of 22 items scored from 1 to 7, was used to evaluate the severity of burn-out, with higher scores indicating more pronounced symptoms.

Sleep problems were assessed using the Insomnia Severity Index (ISI), which consists of seven items (scale range 0–28), with higher scores indicating more severe sleep difficulties.

To assess anxiety and worry, the 7-item Generalised Anxiety Disorder (GAD-7) Scale was used, comprising seven items (scale range 0–21), where higher scores represent more pronounced symptoms.

Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9), which includes nine items (scale range 0–28), with higher scores indicating more severe depression.

Lastly, the Brunnsviken Brief Quality of Life (BBQ) Scale was used to assess perceived quality of life. The BBQ consists of 12 items (scale range 0 to 96), with higher scores indicating higher quality of life.

In addition to the measures described above, we administered instruments relating to putative mediational processes and health economic assessment, which will be reported elsewhere.

Feasibility measures

Treatment credibility was evaluated using the Treatment Credibility Scale (C-scale), and patient satisfaction with the treatment was assessed post-treatment with the Client Satisfaction Questionnaire (CSQ-8). Potential negative effects of treatment were assessed post-treatment using the Negative Effects Questionnaire (NEQ), a validated self-report measure on which respondents report negative events experienced during the treatment period and indicate whether they attribute them to the treatment. All questionnaires were self-rated by participants. We also measured the number of sessions attended and number of completed modules (internet-based CBT), which were recorded by the treating clinician in the respective treatment arm.

Assessment points

The PSS-10, SMBQ, PHQ-9 and ISI were assessed at baseline, every 3 weeks during treatment, post-treatment (12-week follow-up) and 1 year follow-up from baseline. The BBQ and GAD-7 were assessed at baseline, post-treatment and 1 year follow-up. CSQ and NEQ were assessed at post-treatment only and the C-scale was administered in week 3. All assessments were conducted online, a format proven to be reliable for administration of questionnaires.²² As the PSS-10 refers to the preceding month, the recall windows of adjacent treatment-phase assessments overlap to a limited extent; this interval was chosen to characterise the trajectory of change in perceived stress during treatment while limiting participant burden.

Treatments

Nurse-led CC

The CC intervention was a brief, structured, nurse-led intervention delivered in two to six visits over a maximum of 12 weeks, typically lasting 30–45 min depending on patient needs. A central aim was to avoid medicalising stress reactions and instead guide patients towards a problem-solving approach. Rather than prescribing solutions, the nurse encouraged patients to draw on their own resources to address the causes of their stress. In doing so, the patient and nurse collaboratively did an inventory of the patient’s problems, and the patient was asked to generate possible solutions to the most relevant problems. After discussing the pros and cons of possible solutions, the patient was encouraged to select and implement at least one, which was followed up during a subsequent visit. The nurse also provided psychoeducation on stress and recovery and acted as a care coordinator, facilitating connections with other services at the primary care centre, typically the general practitioner, when necessary. Two experienced licensed nurses with basic training in managing mental health for primary care patients, but who were not psychiatry specialists, delivered this intervention. This approach reflects a nurse-led, primary-care-integrated model of CC, in which psychological support is delivered within the patient’s existing care relationship rather than through specialist referral, consistent with the broader literature on task-shifted interventions in primary care.^{16 23 24}

Guided self-help cognitive behavioural therapy

In this trial, GSH-CBT was considered the empirically supported reference treatment. We used a CBT protocol specifically developed to address stress-related mental ill health, that has shown to be effective in reducing symptoms of stress in several previous randomised controlled clinical trials.^{25 26} Key components of this 12-week treatment included recovery training, interventions to improve sleep, behavioural activation, and exposure techniques to help participants break cognitive and behavioural patterns postulated to maintain and exacerbate stress-related symptoms.

The treatment was delivered by nine psychologists, either via the internet or as bibliotherapy, depending on the patient’s preference. All psychologists were trained in CBT and the specific protocol. This form of therapist-guided self-help treatment has been shown to be a valid and effective treatment format in numerous randomised controlled trials and meta-analyses.^{13 14} The treatment includes psychoeducation, worksheets and exercises organised into 12 modules (internet-based CBT) or chapters (bibliotherapy CBT) and participants were expected to complete approximately one module or chapter per week. The internet-based CBT was delivered through a secure web platform and participants received written feedback on their progress from a psychologist via the digital platform, typically once per week.

Participants in bibliotherapy CBT received the same treatment content in the form of a booklet and met with a psychologist who provided support and feedback on treatment progress for two to five face-to-face sessions, with each session lasting approximately 40 min. Although the two formats differ in delivery mode, they share the same treatment content, self-administered structure and therapeutic purpose: in both, the patient works largely independently through the treatment material with the psychologist providing guidance and support rather than delivering content flexibly. The option to choose either internet-based or bibliotherapy as the format for GSH-CBT was implemented after 20 participants had been included and before any data were analysed. This amendment was made because some participants were hesitant to be included in the study because they did not want internet-delivered treatment (ie, they preferred face-to-face sessions). Hence, to increase the recruitment rate, the option to receive bibliotherapy CBT was added. This amendment was approved by the Swedish Ethical Review Authority.

Formal treatment fidelity ratings were not conducted. In the GSH-CBT condition, treatment content was standardised and identical across participants, delivered via fixed internet modules or a structured self-help book, with the psychologist providing guidance and support rather than delivering content flexibly. In the CC condition, the nurse recorded at the post-treatment assessment which intervention components had been used with each patient (see online supplemental table S2). All therapists in both treatment arms had substantial prior experience with the respective treatments, and therefore, only brief study-specific training was provided. Therapists in both conditions received weekly supervision, in which project-related and treatment-specific matters were discussed.

Statistical analysis

Statistical analyses were conducted using STATA, V.18 (StataCorp). The a priori set non-inferiority margin on the primary outcome PSS-10 was 0.3 Cohen’s d. This meant that CC would only be regarded as non-inferior to GSH-CBT if the upper limit of the one-sided 95% CI of the between-group effect size on the primary outcome did not include 0.3 d. This difference was regarded as a small clinical effect, and as a comparison, the median non-inferiority margin in the literature is typically around 0.42–0.50 d and only a small proportion of studies have used a non-inferiority margin as narrow as 0.3.²⁷ This margin is conservative in the context of psychotherapy research, where a prominent meta-analysis adopted an a priori margin of 0.25 (g) as a particularly conservative benchmark.²⁸

Data analysis of continuous outcomes was conducted using linear mixed models regression where the most central estimate was the interaction effect of group (nurse-led CC vs GSH-CBT) and time on the primary outcome PSS-10 and its corresponding one-sided 95% CI. Time was coded from 0 (pretreatment) to 1 (post-treatment), so that the group-by-time interaction represented the difference between groups in change on the PSS-10 from pretreatment to post-treatment. These models included random effects for intercept and slope (unstructured covariance). No baseline covariates were included, since randomisation was expected to balance the groups at baseline. Missing data were not imputed prior to fitting the linear mixed-effects models described below; instead all available observations were included in the analyses under the missing-at-random assumption.

As stated in the preregistered data analytical plan (osf.io/f872t), the absence of significant differences between treatments was not regarded as sufficient evidence of non-inferiority. To align with the power analysis, d was calculated based on the interaction effect estimate and standardised in accordance with Westfall et al.²⁹ Baseline to primary end point (post-treatment) analyses were conducted using masked coding of allocation status.

Analyses followed the intention-to-treat principle, including all randomised participants in their allocated groups. The primary analysis concerned the baseline to post-treatment (12-week follow-up) period, but we also analysed differences in improvement slopes from baseline to 1 year follow-up. As the trial’s power calculation was based on the primary outcome PSS-10, we regarded all other outcome measures as secondary and therefore analysed these measures exploratively using linear mixed models analysis with two-sided 95% CIs.

As described in the protocol at clinicaltrials.gov, which was completed before enrolment started, our primary analyses concerned change from baseline to post-treatment, that is, an analytical model using data from repeated assessments. Before initial recruitment, we conducted a crude power calculation for a post-treatment between-group t-test, which indicated a target of approximately 300 participants. However, during the treatment phase, but before any data had been analysed, an independent statistician conducted a power analysis for a longitudinal linear mixed effects model (with the time*group interaction as the main estimate) using simulations based on PSS-10 data from a previously conducted trial of CBT for stress-related mental ill health. This power analysis was thus more appropriate considering our preregistered primary outcome. With an estimated attrition of 10%, this power calculation showed that approximately 170 patients (85 in each treatment condition) needed to be included to achieve 80% power to detect non-inferiority on the primary outcome PSS-10 with a non-inferiority margin of d=0.3, given no true effect difference.

Findings

Attrition and adherence

Figure 1 illustrates the participant flow throughout the trial and data completion rates on the primary outcome. Post-treatment assessment completion rates were 87.4% (GSH-CBT) and 85.9% (CC). One-year follow-up completion rates were 73.6% (GSH-CBT) and 75.3% (CC).

Figure 1. Participant flow chart. Note. Data on completed assessments refer to the primary outcome. GSH-CBT, guided self-help cognitive behavioural therapy; PSS-10, 10-item version of the Perceived Stress Scale.

Participants assigned to the CC group attended an average of 3.68 sessions (SD=2.01, median=3) with a nurse. Sessions were conducted either face to face or via video, depending on patient preference.

Participants who received bibliotherapy CBT (n=36) attended an average of 4.06 sessions (SD=1.66, median=4) with a psychologist, likewise conducted either face to face or via video, depending on patient preference.

Participants who received internet-based CBT (n=51) instead had asynchronous, text-based contact, receiving written feedback from a psychologist through the secure platform, typically once per week, in response to their submitted exercises, with little or no real-time contact. Participants who received internet-based CBT completed an average of 8.61 modules (SD=3.85) out of 12 over the 12-week treatment period. Among these, 39 participants (76.47%) completed at least six modules, which is equivalent to being a treatment completer since all central elements of the treatment are presented in the first six modules. Psychologists delivering internet-based CBT spent an average of 113.63 min (SD=65.99) per patient, equating to approximately 9.47 min per week.

Participants in the GSH-CBT group reported spending 2.78 (SD=1.60) hours per week working with the treatment compared with 2.31 (SD=1.96) hours for participants in the CC group (see online supplemental table S3).

Primary outcome

Table 2 presents means, SDs and effect sizes (d values of approximately 0.2, 0.5 and 0.8 are considered small, moderate and large, respectively) and online supplemental table S4 presents unstandardised linear mixed model estimates of the outcomes relating to psychiatric symptoms and quality of life. In both treatment conditions, participants made large and significant within-group improvements on the primary outcome (table 2). The analysis of the primary outcome PSS-10 showed that the standardised between-group difference in baseline to post-treatment change was small (d=0.22), favouring GSH-CBT with an upper bound for the CI of 0.52, which was outside of the non-inferiority margin. Improvements were sustained at the 1 year follow-up in both treatment conditions (see figure 2 for PSS-10 trajectories during the study period).

Table 2. Means and effect sizes (d) on primary and secondary outcome measures.

Measure (scale range)	Mean (SD)			Effect size d (95% CI)
	Pretreatment	Post-treatment	1YFU	Within-group pretreatment to post-treatment	Within-group pre to 1-year follow-up	Between group pretreatment to post-treatment	Between group pre to 1-year follow-up
PSS-10 (0–40)
GSH-CBT	25.8 (4.5)	18.4 (7.6)	15.2 (6.6)	−0.98 (-0.77 to −1.18)	−1.12 (−0.94 to −1.30)	−0.22 (−0.52)	−0.26 (−0.88)
CC	24.9 (4.9)	19.4 (5.9)	16.3 (7.1)	−0.76 (−0.55 to −0.96)	−0.90 (−0.72 to −1.09)
SMBQ (1–7)
GSH-CBT	5.4 (0.8)	4.1 (1.3)	3.6 (1.4)	−0.92 (−0.73 to −1.11)	−1.09 (−0.89 to −1.28)	−0.28 (−0.64 to −0.002)	−0.27 (−0.89 to 0.02)
CC	5.2 (0.9)	4.3 (1.1)	3.8 (1.3)	−0.64 (−0.44 to −0.84)	−0.84 (−0.65 to −1.03)
PHQ-9 (0–28)
GSH-CBT	13.0 (5.4)	7.5 (5.4)	6.8 (5.9)	−0.72 (−0.54 to −0.90)	−0.72 (−0.57 to −0.88)	−0.21 (−0.54 to 0.04)	−0.14 (−0.76 to 0.09)
CC	11.6 (5.1)	7.9 (4.7)	6.7 (5.5)	−0.51 (−0.33 to −0.69)	−0.61 (−0.45 to −0.76)
ISI (0–28)
GSH-CBT	14.8 (6.7)	9.5 (6.7)	8.6 (6.3)	−0.62 (−0.46 to −0.79)	−0.65 (−0.48 to −0.81)	−0.23 (−0.54 to 0.002)	−0.20 (−0.67 to 0.05)
CC	14.8 (6.1)	10.9 (6.2)	10.1 (6.9)	−0.39 (−0.23 to −0.56)	−0.47 (−0.30 to −0.63)
GAD-7 (0–21)
GSH-CBT	12.7 (4.4)	7.7 (5.7)	6.6 (5.2)	−0.90 (−0.66 to −1.15)	−1.14 (−0.88 to −1.40)	−0.30 (−0.84 to 0.04)	−0.09 (−0.55 to 0.23)
CC	11.9 (4.4)	8.6 (5.2)	6.3 (5.2)	−0.61 (−0.36 to −0.85)	−1.05 (−0.79 to −1.32)
BBQ (0–96)
GSH-CBT	41.7 (19.5)	48.8 (21.2)	54.3 (22.2)	0.25 (0.08 to 0.41)	0.51 (0.32 to 0.71)	−0.07 (−0.30 to 0.17)	0.02 (−0.31 to 0.24)
CC	40.6 (18.0)	46.3 (18.3)	51.6 (22.1)	0.25 (0.09 to 0.42)	0.50 (0.30 to 0.70)

Significant estimates in bold font. For BBQ, there was one specific attrition; hence, the results are based on 75 answers in the GSH-CBT group.

BBQ, Brunnsviken Brief Quality of Life ; CC, collaborative care; GAD-7, 7-item Generalised Anxiety Disorder Scale; GSH-CBT, guided self-help cognitive behavioural therapy; ISI, Insomnia Severity Index; PHQ-9, 9-item Patient Health Questionnaire; PSS-10, 10-item Perceived Stress Scale; SMBQ, Shirom-Melamed Burnout Questionnaire; 1YFU, 1 year follow-up.

Figure 2. Symptom development on the primary outcome measure PSS-10 during the study period. 1YFU, 1 year follow-up; CBT, cognitive behavioural therapy; CC, collaborative care; Post, after the 12-week treatment period; Pre, before treatment start; PSS-10, Perceived Stress Scale-10. Note: vertical bars represent 95% CIs and CBT refers to GSH-CBT (guided self-help cognitive behavioural therapy).

Additional analyses are presented in the online supplemental file 1. The sensitivity analysis excluding the first 20 participants yielded similar results (d=0.22, one-sided 95% CI upper bound=0.52), supporting the robustness of the primary finding (see online supplemental file 1). An exploratory comparison of the two GSH-CBT formats showed near-identical primary outcome trajectories (slope difference=−0.05, p=0.972), supporting the treatment of GSH-CBT as a unified comparator (online supplemental figure S1). Scatterplots relating the number of completed modules or visits to PSS-10 change scores are presented in online supplemental figures S2–S4, and primary outcome trajectories by gender, baseline severity and individual trajectories are presented in online supplemental figures S5–S7.

Secondary outcomes

As shown in table 2 and online supplemental table S4, there were significant within-group improvements on all secondary outcomes in both treatment conditions. Analysis of the secondary outcomes showed that GSH-CBT led to a significantly larger baseline to post-treatment reduction of exhaustion (SMBQ) compared with CC (d=0.28; 95% CI 0.64 to 0.002). There were no significant differences between the treatment groups on the other secondary outcomes and the upper bounds of the 95% CIs for effect sizes were typically in the moderate to large range (table 2).

Treatment credibility, treatment satisfaction and negative effects

The mean rating of treatment credibility (C-scale) for participants in the GSH-CBT group was 34.17 out of a possible total score of 50 (SD=9.21) compared with participants in the CC group with a mean rating of 31.96 (SD=9.74), which was non-significant, (t(161) = −1.49, p=0.139). Similarly, participants in the GSH-CBT group reported a mean treatment satisfaction (CSQ-8) rating of 24.41 out of a possible total score of 32 (SD=5.41) compared with 23.14 in the CC group (SD=5.20), a difference that was also non-significant (t(145) = −1.46, p=0.148). See online supplemental tables S5 and S6 in the online supplemental file for further details.

At post-treatment, participants reported any negative effects they experienced during the treatment period (assessed by the NEQ). For those who received GSH-CBT, 58 participants (77.3%) reported at least one negative event, compared with 39 participants (54.2%) who received CC. The most frequently reported negative event in both groups was an increased feeling of stress. No serious negative effects, such as those posing acute health risks requiring hospitalisation, were reported. Negative effects were summarised descriptively and were not compared between groups; full details are provided in online supplemental table S7. The online supplemental file also reports the number of participants referred to further healthcare services during and after treatment (online supplemental table S8), the number with a prolonged treatment period (online supplemental table S9), self-reported concurrent treatments received between post-treatment and 1 year follow-up (online supplemental table S10) and changes in stress-related medication over the same period (online supplemental table S11).

Discussion

This study evaluated nurse-led CC and GSH-CBT as early interventions for adult primary care patients with stress-related mental ill health, hypothesising that nurse-led CC would be non-inferior to GSH-CBT in reducing perceived stress. Contrary to our hypothesis, non-inferiority was not demonstrated. Both groups showed significant within-group improvements on the primary and all secondary outcomes, which were sustained at 1 year follow-up. Analysis of the primary outcome (PSS-10) showed a small standardised between-group difference in baseline to post-treatment change (d=0.22), favouring GSH-CBT, with the upper CI bound (0.52) exceeding the non-inferiority margin. GSH-CBT also produced significantly larger reductions in exhaustion symptoms. Both treatments were rated as credible and satisfactory, with no serious negative effects reported.

To our knowledge, this is the first non-inferiority trial comparing two active psychosocial interventions for consecutively recruited primary care patients with stress-related mental ill health. The sample primarily consisted of middle-aged women reporting work-related stress, aligning with previous research.^{3 10 25} CBT is generally regarded as the most empirically supported treatment,¹¹ and consistent with previous studies,^{25 26} participants who received GSH-CBT showed substantial reductions in perceived stress and adhered to the treatment to the same extent as in previous clinical trials, indicating that this reference intervention was implemented as intended.

Two conclusions should be distinguished. First, the trial did not demonstrate non-inferiority of nurse-led CC to GSH-CBT: the upper limit of the CI for the between-group difference exceeded the prespecified margin, and nurse-led CC can therefore not be regarded as non-inferior to GSH-CBT on the primary outcome. Second, this absence of evidence for non-inferiority does not indicate that CC lacks clinical value. This study adds to the body of knowledge on nurse-led CC for mental health problems, suggesting that this approach is safe, feasible and associated with reductions in mental health symptoms when delivered in the primary care context. While non-inferiority was not demonstrated, the findings support that GSH-CBT may not be the only viable treatment option for individuals with stress-related mental ill health. In the present study no significant differences emerged between the two groups on any secondary outcomes, except for exhaustion symptoms (SMBQ), which showed greater reductions in the GSH-CBT group. One tentative explanation for this result is that complex and multifaceted symptoms like exhaustion may require more specialised therapeutic techniques such as systematic exposure to conditioned stressors, a central component in GSH-CBT but not in CC.

In both treatments, the time spent by clinicians delivering the interventions equated to approximately three to four visits, with slightly less observed time spent in nurse-led CC, which highlights their clinical utility in healthcare contexts where therapeutic resources are limited. In primary care settings, characterised by a large influx of patients and constant organisational concern regarding time needed to treat,³⁰ this finding suggests that both GSH-CBT and nurse-led CC could be viable options for addressing stress-related mental health issues without overwhelming healthcare professionals.

Given limited access to psychological treatment,¹² alternative interventions warrant consideration. While GSH-CBT was developed partly to increase access to evidence-based treatment by reducing therapist contact time, its delivery still requires psychologist involvement, which remains a limiting factor where psychologists are scarce. The shortage of psychologists in primary care coupled with long waiting times remain significant issues. Early identification and treatment of stress-related mental ill health could have positive implications, potentially reducing prolonged sick leave.¹⁷ Both GSH-CBT and CC represent task-shifting approaches that reduce reliance on therapist-delivered treatment; they differ primarily in that GSH-CBT shifts the therapeutic work to the patient through structured self-help guided by a psychologist, whereas CC shifts the clinical role to a nurse delivering brief structured support. The choice of CC over, for example, nurse-delivered CBT reflects the pragmatic constraints of primary care, where brief training in structured problem-solving and psychoeducation was regarded more feasible than the extended training required to deliver CBT with fidelity. While CBT is considered the most empirically supported treatment for stress-related mental ill health,¹¹ nurse-led CC presents a promising alternative in settings where access to psychologist-delivered treatment is limited. Although this study did not support its non-inferiority to GSH-CBT, the significant within-group improvements observed suggest that nurse-led CC may be a clinically meaningful option for individuals and healthcare organisations seeking to expand the range of available primary care interventions.

Strengths and limitations

This study has several strengths. The randomised design ensures that both known and unknown variables are evenly distributed between groups, reducing the risk of bias. Additionally, the study was conducted in a routine primary care setting, with no media advertisements and all clinicians being primary care staff, which supports the ecological validity of the findings and their potential relevance to real-world clinical practice. Despite being implemented within regular healthcare, data completion rates were high, further strengthening the reliability of the findings.

Moreover, we used validated outcome measures that covered a broad range of relevant domains, including not only perceived stress but also anxiety, depression, sleep and quality of life, and the non-inferiority margin on the primary outcome was clearly defined and prespecified. A further strength was that the study used a reference treatment, GSH-CBT, that has been shown to be effective in a series of randomised trials and that was delivered by primary care clinicians trained in CBT and with previous experience of delivering this treatment, which speaks for the validity of the comparison. Finally, the inclusion of a controlled long-term follow-up allows for a more comprehensive understanding of the intervention’s sustained effects.

Several limitations should be considered when interpreting the results of this study. First, the lack of clinician-rated assessments following treatment completion is a potential limitation. Although we believe that the measures used were suitable for this population, clinician assessments could have provided additional insights, for example, regarding experiences of working with the treatments.

Second, the absence of an untreated control group may be seen as a limitation, making it difficult to determine whether symptom reductions resulted from treatment effects or natural symptom course. While comparing both treatments against a psychological placebo could have been informative, this was not practically feasible in the current trial.

A third limitation concerns the generalisability of the findings. The sample was predominantly middle-aged, well-educated women, which is broadly consistent with the demographic profile reported in comparable Swedish trials,³ but may limit the applicability of the results to other groups, such as men or individuals with lower educational backgrounds. It is possible that both interventions may be more effective for individuals accustomed to engaging with written material or problem-solving tasks. Additionally, participants were recruited from a single primary care centre in Stockholm; although the centre operated under standard primary care conditions, single-site recruitment limits generalisability to other settings, regions and healthcare systems, and replication in multisite and international contexts would strengthen the evidence base. A notable feature of the sample was the high proportion of participants in full-time employment (over 80%). This is considerably higher than in comparable Swedish trials of psychological treatment for stress-related mental ill health; for example, a trial with internet-based CBT for adults with diagnosed stress-related disorders reported 53.3% in full-time employment.²⁵ A likely explanation is that the present trial deliberately targeted a milder, mild-to-moderate symptom severity population and did not require a formal psychiatric diagnosis for inclusion, whereas most existing trials in this field have enrolled participants meeting criteria for a diagnosis such as adjustment or exhaustion disorder, who tend to be more severely affected and more frequently sick-listed. To our knowledge, few trials internationally have examined this milder, non-diagnostic, help-seeking primary care stress population, which represents a large group in routine care that is at risk of symptom exacerbation but not yet meeting diagnostic thresholds. With that said, employment status and sickness-absence status were recorded as distinct variables (table 1), which meant that a participant could be in full-time employment while simultaneously being on full or partial sick leave from work. The high proportion in full-time employment should therefore not be taken to indicate an absence of work impairment.

A fourth limitation concerns the heterogeneity of the GSH-CBT comparator. The GSH-CBT condition comprised two delivery formats, internet-based CBT and bibliotherapy CBT, which differed in contact mode (asynchronous written feedback vs face-to-face or video sessions) and contact intensity. Allocation to format was determined by patient preference rather than randomisation, so the two subgroups are not directly exchangeable. We treated GSH-CBT as a single comparator on the a priori basis that the two formats share the same treatment content, self-administered structure and therapeutic purpose, and this was supported by the exploratory post hoc analysis, which showed near-identical primary outcome trajectories across formats (online supplemental figure S1).

Lastly, a related limitation concerns the choice of analysis population. The primary analysis followed the intention-to-treat principle, as prespecified, which preserves the advantages of randomisation but is known to be biased towards the non-inferiority conclusion in non-inferiority trials, since dropouts and noise tend to attenuate between-group differences. A per-protocol analysis would provide a complementary perspective; however, defining a treatment completer in a clinically meaningful way proved difficult across all three formats, as the degree of engagement required for benefit varied considerably between patients.

Conclusions

This study did not demonstrate that nurse-led CC is non-inferior to GSH-CBT in reducing stress among adult primary care patients with stress-related mental ill health. The effect size differences were small, with the overall pattern showing slightly larger improvements in the GSH-CBT group. However, nurse-led CC required limited clinician resources and was associated with significant reductions in perceived stress and other psychiatric symptoms when delivered to primary care patients. Based on the results of this trial, we cannot conclude that nurse-led CC is non-inferior to GSH-CBT.

Clinical implications

As nurse-led CC is safe, relatively easy to administer and yielded significant within-group improvements across all measures, it may still be a useful intervention for this patient group in primary care settings. Both GSH-CBT and nurse-led CC require only about three to four clinician visits, suggesting they are viable, resource-efficient options for managing stress-related health issues in time-pressured primary care settings.

Data availability statement

Data may be obtained from a third party and are not publicly available.