Augmenting Evidence-Based Care with a Texting Mobile Interventionist: Pilot Randomized Controlled Trial

Abstract

Objective

To evaluate the feasibility and clinical utility of training intensive psychiatric community care team members to serve as mobile interventionists who engage patients in recovery-oriented texting exchanges.

Methods

A 3-month pilot randomized controlled trial was conducted to compare the mobile interventionist approach as an add-on to Assertive Community Treatment (ACT) versus ACT alone. Participants were 49 individuals with serious mental illness (62% with schizophrenia/schizoaffective disorder, 24% with bipolar disorder, and 14% with depression). Clinical outcomes were evaluated at baseline, post-treatment, and 6-month follow-up, and satisfaction post-treatment.

Results

The intervention appeared feasible (95% of participants initiated the intervention, texting on 69% of possible days, averaging four messages per day), acceptable (91% reported satisfaction), and safe (no adverse events reported). Exploratory post-treatment clinical effect estimations suggested greater reductions in the severity of paranoid thoughts (d = −0.61) and depression (d = −0.59), and improved illness management (d = 0.31) and recovery (d = 0.23) in the mobile interventionist group.

Conclusions

Augmentation of care with a texting mobile interventionist proved to be feasible, acceptable, safe, and clinically promising. The findings are encouraging given the relative ease of training practitioners to serve as mobile interventionists, the low burden placed on patients and practitioners, and the simplicity of the technology that was used. The technical resources are widely-accessible to patients and practitioners, boding well for potential intervention scalability. When pandemics such as COVID-19 block the possibility of in-person patient/provider contact, evidence-based texting interventions can serve a crucial role in supporting continuity of care.

Introduction

Technology is helping us re-imagine mental health services. Wearable sensors, natural language processing, machine learning, augmented reality, and other innovations may usher in an era when patient monitoring, illness detection, and interventions are deployed with reduced need for time-sensitive involvement of human providers (1). If such approaches prove to be feasible and clinically useful, these cutting-edge technologies would provide opportunities for dramatic innovation of mental health services. But more pressingly and pragmatically, innovative use of existent and widely-accessible technologies may already enable us to expand the reach and potency of evidence-based services in the present (2,3)..

Clinic-based services models often fail to reach those in need of care (4). When individuals with serious mental illness experience symptom exacerbations, they may find it challenging to go to a treatment center. Those who seek care must contend with the challenges that are intrinsic to clinic-based services including distance from their residence, limited hours of operation, stigma associated with going to a mental health center, apprehension about interacting with others in group settings, gaps between meetings, and varying quality of services (5–7). The recent COVID-19 global pandemic has brought into stark reality how in-person services may be discontinued entirely, abruptly. Mobile technologies may help overcome these barriers by providing opportunities to deliver mental health services to patients in their own environments (8).

There are approximately seven billion mobile phones in use worldwide (9). Access to mobile phones is one of the few areas where the resource gap between people with serious mental illness and the general population is almost non-existent; community-based studies consistently find that most people with severe psychopathology have mobile phones and are open to using their devices as recovery support tools (10–12). Text messaging is a simple, feasible, and technically robust digital communication modality that has emerged as one of the most widely-used health technology resources on the planet (13). People with serious mental illness use Short Message Service (SMS) text messages at similar rates to the general population (14, 15). A recent meta-analysis suggested a range of texting interventions are feasible among people with serious mental illness (16). The majority of these approaches involved automated or semi-automated messaging rather than targeted personalized interventions.

Our group developed a technology-assisted illness management strategy that is delivered by a texting mobile interventionist. In this model a trained community-based mental health worker engages people with serious mental illness in daily recovery-oriented texting exchanges tailored to their individual needs. Unlike digital health apps or therapeutic websites that can only be accessed on smartphones with computational capacities, practically all mobile phones in use today can technically support this type of texting intervention. Furthermore, the model incorporates the clinical expertise of a clinical workforce that can be trained to deliver targeted patient centered mHealth. Repurposing rather than replacing existing clinical personnel may be more appealing to agencies that are looking to modernize their models of care, and may provide a more seamless transition to remote care. A previous proof-of-concept study found the mobile interventionist approach was engaging to people with serious mental illness and that it facilitated a strong patient/provider therapeutic alliance that was rated as superior to in-person care (17). Here we report on the next phase in this program of research: a pilot randomized controlled trial designed to evaluate real-world feasibility and acceptability, and to estimate clinical effects of the mobile interventionist approach when added to an existing practice.

Method

We conducted an assessor-blind, two-arm, randomized controlled trial between December 2017 and October 2019 in partnership with Assertive Community Treatment (ACT) teams providing services to individuals with serious mental illness in the Midwest and Pacific Northwest regions of the United States. The study was approved by the Institutional Review Boards of the University of Washington and Dartmouth College. The research was monitored by an independent safety monitoring board at the University of Washington’s School of Medicine. Participants provided informed consent. Individuals were all receiving ACT services and were randomized into one of two treatment arms: mobile interventionist (ACT augmented with mobile phone texting) or treatment as usual (standard ACT). Interventions were deployed for a period of 3 months. We conducted assessments at baseline (0 months), post-trial (3 months), and follow-up (6 months). Participants were compensated for completing assessments ($30 per assessment, plus reimbursement for travel). The study was registered in ClinicalTrials.gov (https://clinicaltrials.gov/ct2/show/NCT03062267).

Participants

Participants were identified by ACT teams at the study site and screened by research staff. Inclusion criteria included: chart diagnosis of schizophrenia, schizoaffective disorder, bipolar disorder, or major depressive disorder; 18 years or older; currently receiving ACT services; a rating of “3” or higher on one of three items from the Recovery Assessment Scale indicating need for additional support (I don’t have the skills to stop my symptoms from getting worse; I want to make changes need to help my recovery; I am not able to handle my symptoms on my own most of the time); and ownership of a mobile phone with data plan. Exclusion criteria were: hearing, vision, or motor impairment affecting operation of a mobile device, determined using candidate’s device for screening; less than 4^th grade English reading ability, determined using the reading section of the WRAT-4 (18).

Randomization and Blinding

The study statistician created a computer-generated randomization list organized in blocks of four participants (3:1 ratio, mobile interventionist vs. treatment as usual, respectively). Study assessors were blinded to group.

Intervention description

All participants received ACT services, which have been described extensively (19, 20). Briefly, it is an intensive team-based treatment model designed to provide comprehensive psychiatric, rehabilitation, and support services to people with serious mental illness residing in the community (21–23). In the experimental arm, the mobile intervention was tested as an add-on to ACT. ACT team practitioners were trained to serve as team-embedded mobile interventionists in a half-day training seminar. They participated in weekly supervision calls with a clinical psychologist trained in the intervention model. Interventionists were given guidance on how to introduce the intervention to clients, establish rapport via text, and optimize engagement over time. Each interventionist was encouraged to infuse their approach with their own “personal touch” so that the texts do not seem bland or robotic. They were instructed to provide daily texting support to clients assigned to them over a twelve-week period. The study involved a total of 7 ACT teams. Each team had one mobile interventionist. Study participants allocated to the experimental arm received texting services from their assigned mobile interventionist but did not receive any other services from this individual for the duration of the study. Participants received services from only one ACT team and were not exposed to other team interventionists.

At a baseline visit, mobile interventionists met with each participant to build rapport, review the treatment model, identify target areas, and complete a training regarding basic phone functions (e.g., charging the battery, making calls) and texting. Following this visit, mobile interventionists would provide daily support via text messages during the ACT team’s hours of operation, typically Monday-Friday between 9 AM and 5 PM (after-hours ACT crisis management resources continued to be available). During this period, mobile interventionists continued as members of the ACT team and regularly consulted with other team members on client issues and provided weekly reports in person or via secure email. If clients indicated a crisis via text (despite being directed that the intervention was not meant as a time-sensitive tool to communicate urgent matters), the mobile interventionists would relay the information to their ACT team colleagues who would engage in their standard crisis management protocols.

The maximum intervention dose was three “exchanges,” wherein an exchange is defined as a cluster of thematically connected back-and-forth messages between mobile interventionist and participant (e.g. three outgoing messages and participant responses). Texting strategies included: reminders (e.g., appointments, prescription refills), information provision (e.g., psychoeducation, links to regional events and resources), cognitive challenges (e.g., restructuring dysfunctional beliefs about voices, questioning the validity of self-sabotaging automatic beliefs), self-monitoring/self-reflection (e.g., guidance on self-evaluation of affect, journaling of symptomatic experiences), relaxation techniques (e.g., diaphragmatic breathing, guided imagery), social skills training (e.g., initiating conversations, maintaining eye contact), supportive messages (e.g., affirmations, inspirational quotes), and in-vivo instruction (e.g., pre-scheduled real-time support as the patient attempted a new activity).

The intervention continued for twelve weeks for all participants. To avoid abrupt discontinuation, mobile interventionists led a tapered termination during the last two weeks of this period. At the end of treatment, the mobile interventionist and participant met to reviewe gains, summarize the helpful tools acquired, and discuss how the participant could use skills independently.

Measures

Feasibility

Feasibility was assessed based on (1) the number of participants in the experimental arm who initiated texting, (2) the number of days in which texting occurred, (3) the average number of texts sent and received per day on days in which texting occurred, and (4) the number of text messages sent and received per participant.

Acceptability

Treatment acceptability was assessed with a 5-item measure rated on a 7-point scale adapted from previous research examining mHealth for people with serious mental illness (17,24). All items appear in Figure 1.

Figure 1.

Acceptability measures.*Lower scores on this item indicate greater acceptability.

Clinical Effects

Depression and anxiety were assessed with the Beck Depression Inventory, 2^nd Edition (BDI-II) (25) and Beck Anxiety Inventory (BAI) (26), both of which include 21 items rated on a 4-point scale, summed for total severity scores. Psychosis was assessed with the Psychotic 17-item Symptom Rating Scales (PSYRATS) (27). The PSYRATS includes dimensions of auditory hallucinations and delusions that are rated on a 5-point scale. Persecutory ideation was assessed with the Green Paranoid Thoughts Scale (GPTS) – Persecution Subscale (28), a 16-item questionnaire rated on a 5-point scale that is summed to generate a total score. Illness management was measured with the Illness Management Recovery Scale (IMRS) (29), a 15-item self-report measure that captures different aspects of illness management, including coping with symptoms, medication use, utilization of social/familial support, and relapse prevention planning. IMRS items are scored on a 5-point scale and summed to represent a composite score. Recovery was assessed using the Recovery Assessment Scale (RAS) (30,31), a 24-item, 5-point Likert scaled measure assessing self-rated recovery.

Adverse events were monitored by site clinical staff who interacted with participants regularly and had access to agency-wide reports on the Electronic Health Record. Participant issues were discussed during weekly calls with the study clinical supervisor and reported to the study team during project management calls. Negative events (e.g., marked deterioration in physical health, hospitalization, death) were evaluated by the investigative team and study Safety Monitoring Board to determine whether they might be associated with participation in the study.

Because this study was focused on piloting the intervention and estimating preliminary effect sizes for a larger-scale trial rather than robustly testing effectiveness, our sample size was designed to examine real-world feasibility. We had 0.80 power to detect only very large clinical effect sizes of d = 0.94 or greater.

Analytic Approach

Feasibility was modeled descriptively based on frequency of texting. We used an intent-to-treat (ITT) approach in preliminary outcomes analyses. Measures were analyzed using linear regression with treatment condition and baseline values included as covariates. Participants with missing follow-up measures were included by using multiple imputation, which reduces bias in effect estimates (32). Attrition analyses evaluated rates and predictors of missingness.

Results

The study enrolled 51 individuals but 49 were included in analyses (see study CONSORT diagram in online supplement). Participants were on average 44.8 years old, 55% male, and 52% White, 26% Black/African American, 17% multiracial, and 5% other racial groups. The sample included 39% patients with schizophrenia, 22% with schizoaffective disorder, 24% with bipolar disorder, and 14% with major depressive disorder. Participants did not differ between groups on demographic variables. Of the 49 patients, 43 (88%) provided 3-month follow-up data and 42 (86%) provided 6-month follow-up data. Follow-up rates did not differ between the treatment conditions and were not significantly associated with baseline demographic or clinical measures.

Patients randomized to the mobile interventionist arm had significantly higher PSYRATS (difference = 14.4, t = 2.32, df = 47, p = 0.03) and GPTS scores (difference=12.7, t = 2.05, df = 47, p = 0.047). To control for this difference, baseline PSYRATS scores were included as covariates in analyses, which obviated the need to control for baseline GPTS due to the strong correlation between these measures r(47)=0.60, p < .001.

Feasibility

No adverse events were reported in either intervention arm. Nearly all participants assigned to the experimental arm (35/37, 95%) commenced treatment by sending at least one text message. Among those who engaged, there was a mean of 41.3±17.1 days with any text exchanges between the patient and interventionist. This represents approximately 69% of the days in which texting could have occurred. Patients sent a mean of 4.0±3.9 messages and received a mean of 3.6±1.5 messages from the interventionist. Over the entire study period, participants sent a mean of 174.7±164.5 text messages and received a mean of 157.9±111.6.

Acceptability

Participants reported the mobile interventionist approach to be acceptable. The majority reported that they were satisfied with the intervention (n = 29 of 32, 91%), that it made them feel better (n = 30 of 32, 94%), and that they would recommend it to a friend (n = 27 of 31, 87%). Post-treatment satisfaction ratings were similar across arms (mobile interventionist M = 30.1±6.1; TAU M = 28.2±6.2) with no significant difference (t = 0.55, df = 11.5, p = 0.59) between the two (see Figure 1).

Clinical effects

Post-intervention (3-month) outcomes controlling for baseline values reflect medium effect sizes in the severity of persecutory ideation (GPTS d = −0.61) and depression (BDI-2 d = −0.63), and small effects indicating greater illness management (IMRS d = 0.23) and recovery (RAS d = 0.24) in the mobile interventionist group. PSYRATS and BAI scores were more stable over time for both conditions and between-condition effects at three months were smaller for these measures (PSYRATS d= 0.06, BAI d= −0.17). Between group effects were, as expected due to sample size, non-significant. Clinical outcomes are graphed in Figure 2 and illustrate a pattern in which some outcomes worsened for the treatment as usual condition from baseline to 3-months, while the mobile interventionist group remained more stable. Follow-up (6-month) assessments revealed that these gains were not maintained after the intervention was disvontinued.

Figure 2.

Clinical outcome measures.

Discussion

This study reports on the feasibility, acceptability, and preliminary effectiveness of integrating clinical texting into ACT services for individuals with serious mental illness. Results demonstrate feasibility (95% of participants commenced the intervention, exchanging messages on 69% of possible days, averaging four texts per day), acceptability (91% reported satisfaction), and safety (no adverse events were reported). The mobile interventionist approach produced greater clinical stability or improvements relative to treatment as usual, including medium reductions in depression (d = −0.59) and persecutory ideation (d = −0.61) and small improvements in illness management (d = 0.31) and self-reported recovery (d = 0.23). The advantage for the texting condition diminished at 6-month follow up, suggesting access to the mobile interventionist must persist to maintain treatment effects.

Multiple studies have demonstrated the feasibility of texting among people with serious mental illness (33–35). The mobile interventionist approach demonstrates that after a brief community-based training, mental health workers can provide ongoing psychosocial interventions via messaging. This personalized, adaptive, and team-integrated approach is a novel use of texting in this population. The model blends key elements of person-delivered care and mobile health strategies to maximize treatment engagement, potency, and reach. But unlike in standard in-person care that involves seeing clients sequentially, a mobile interventionist can engage with multiple patients in parallel (e.g., several separate exchanges per hour) and asynchronously, potentially expanding reach and efficiency of care.

This study is not without limitations. Analyses were not powered to detect statistically significant differences between groups. Study conditions were imbalanced to obtain feasibility and study protocol acceptability data by allocating more participants to the experimental intervention. While results of the present study are promising, larger fully-powered trials are required to address questions related to intervention effectiveness. All participants in the trial owned a mobile device with a data plan. Results may have differed if samples included individuals who did not own a phone, a small sub-population (10–12). Finally, it is possible that additional daily contact with another person is what produced positive patient outcomes. Future research employing a comparison arm with simpler supportive messaging may help elucidate the potent elements of the intervention.

Conclusion

ACT is an established model of care (23), but illness management and other psychosocial functioning outcomes have not been fully achieved by ACT alone (36, 37). Augmenting ACT with a mobile interventionist is feasible and may amplify treatment effects, a high priority for ACT funders, administrators, and practitioners (38). The findings of this study are encouraging given the relative ease of training and supervising ACT staff to serve as interventionists, the low burden placed on both patients and practitioners over the intervention period, and the simplicity of the technology used. The technological resources employed in the intervention – any type of mobile phone, existent nationwide mobile-cellular infrastructure, standard data plans – are already widely accessible. This bodes well for the potential scalability of the approach. If future research replicates our findings in larger samples supporting the clinical utility of the intervention, it could be disseminated broadly and rapidly. Finally, this study was completed before the global 2019–2020 COVID-19 pandemic, which led to sweeping social distancing practices and, in many instances, systemic collapses in capacity to deliver standard mental health services. In an era where in-person patient/provider contact may be discontinued, evidence-based texting interventions can serve a crucial role in supporting continuity of care.

Table 1.

Baseline demographic and clinical descriptive statistics.

	Full Sample		Mobile interventionist		Treatment as usual
	N = 49		n = 37		n = 12
	n	%	n	%	n	%
Female	22	45%	15	41%	7	58%
White	24	52%	17	49%	7	64%
Black/African American	12	26%	10	29%	2	18%
More than one race	8	17%	6	17%	2	18%
Hispanic/Latinx	6	13%	3	8%	3	25%
Schizophrenia	19	39%	17	46%	2	17%
Schizoaffective disorder	11	22%	9	24%	2	17%
Major depressive disorder	7	14%	4	11%	3	25%
Bipolar disorder	12	24%	7	19%	5	42%
	M	SD	M	SD	M	SD
Age	44.8	11.2	45.4	11.1	43.3	12.0
Education (years)	12.9	3.0	12.8	2.4	13.1	4.5
Lifetime hospitalizations	2.8	3.0	2.8	3.4	2.7	1.7
Last year hospitalizations	0.4	0.6	0.4	0.6	0.3	0.5

There were no significant differences between treatment conditions on the demographic measures presented here.

Table 2.

Three and six-month clinical outcomes.

		Mobile interventionist		Treatment as usual
Measure	Time	M	SD	M	SD	t	df	p	Cohen’s d	95% CI
PSYRATS	Baseline	19.6	20.4	5.2	11.2	2.32	47	.03
	3-month	15.1	16.7	6.2	14.4	0.18	34.67	.86	0.06	−0.61–0.73
	6-month	17.9	19.1	0.6	1.9	1.63	39.68	.11	0.62	−0.15–1.40
GPTS - Persecutory Ideation Subscale	Baseline	38.4	20.6	25.7	10.6	2.05	47	.047
	3-month	36.6	19.6	41.5	20.7	−1.80	30.9	.08	−0.61	−1.31–0.08
	6-month	36.9	20.5	22.5	14.2	1.04	36.7	.30	0.40	−0.38–1.17
BDI	Baseline	18.5	13.7	18.5	11.4	0.00	47	1.00
	3-month	17.2	11.3	23.4	12.4	−2.05	30.7	.049	−0.63	−1.25–−0.002
	6-month	15.5	11.2	14.8	10.3	0.22	33.7	.83	0.07	−0.57–0.71
BAI	Baseline	21.9	14.0	18.9	13.3	0.66	47	.52
	3-month	19.7	12.9	22.0	15.5	−0.49	37.4	.63	−0.17	−0.87–0.53
	6-month	20.7	12.5	18.3	9.9	0.09	36.5	.93	0.03	−0.63–0.69
IMRS	Baseline	48.8	9.0	48.8	7.5	0.03	47	.98
	3-month	47.8	9.7	43.3	6.5	0.58	31.4	.57	0.23	−0.59–1.05
	6-month	51.6	9.6	51.6	9.8	0.11	32.5	.92	0.04	−0.68–0.76
RAS	Baseline	89.8	17.3	91.8	14.6	−0.35	47	.73
	3-month	92.9	15.5	90.0	13.8	0.85	34.6	.40	0.24	−0.33–0.80
	6-month	94.4	14.9	93.8	13.1	0.43	32.2	.67	0.15	−0.55–0.85

Note. Descriptive values are computed using complete cases only. Test statistics and effect sizes are computed using regression models with multiple imputation for missing data and adjustment for baseline PSYRATS and baseline values of the respective outcome. Negative coefficients indicate lower scores for mobile interventionist relative to treatment as usual.

Highlights.

• This is the first pilot randomized controlled trial examining the feasibility of augmenting Assertive Community Treatment (ACT) with a texting mobile interventionist.

• The texting intervention proved to be feasible, acceptable, and safe.

• Preliminary findings suggest better symptom management and recovery outcomes for individuals randomized to receive texting-augmented ACT than for those receiving ACT alone.

• When in-person patient/provider contact is difficult or impossible (e.g., distance from a clinic, social distancing restrictions imposed as a result of COVID-19 pandemic), evidence-based texting interventions may serve a crucial role in supporting continuity of mental health care.