Lessons learned through piloting a community-based SMS referral system for common mental health disorders used by female community health volunteers in rural Nepal

Objective The Community Informant Detection Tool (CIDT) is a paper-based proactive case detection strategy with evidence for improving help-seeking behavior for mental healthcare. Key implementation barriers for the paper-based CIDT include delayed reporting of cases and lack of active follow up. We used mobile phones and structured text messages to improve timeliness of case reporting, encouraging follow up, and case record keeping. 36 female community health volunteers piloted this mobile phone CIDT (mCIDT) for three months in 2017 in rural Nepal. Results Only 8 cases were identified by health volunteers using mCIDT, and only two of these cases engaged with health services post-referral. Accuracy with the mCIDT was considerably lower than paper-based CIDT, especially among older health volunteers, those with lower education, and those having difficulties sending text messages. Qualitative findings revealed implementation challenges including cases not following through on referrals due to perceived lack of staff at health facilities, assumptions among health volunteers that all earthquake-related mental health needs had been met, and lack of financial incentives for use of mCIDT. Based on study findings, we provide 5 recommendations—in particular attitudinal and system preparedness changes—to effectively introduce new mental healthcare technology in low resource health systems.


Introduction
Mobile health (mHealth) has the potential to improve mental health in low-and middle-income countries (LMICs) by increasing awareness of and access to treatment [1][2][3][4]. The Community Informant Detection Tool (CIDT) was developed by Transcultural Psychosocial Organization Nepal (TPO-Nepal), under the Programme for Improving Mental Health Care (PRIME) [5,6] to facilitate detection of and help-seeking for mental illness at the community level. The validated tool is a paper-based form consisting of vignettes of common symptoms using local idioms and illustrations [7][8][9]. The CIDT works by training trusted people in the community to detect people who match the vignettes and encourage them to seek care from mental health trained facilitybased health workers (HW). In a pragmatic trial, use of paper-based CIDT was associated with 47% greater mental health treatment initiation compared to referral as usual [10].
Challenges in CIDT implementation have been identified, such as communication gaps between FCHVs and HW along with poorly maintained outpatient logs of CIDT referrals [8]. To address this, we designed a structure Short Messaging Service (SMS), i.e., text messaging, referral system to complement the paper-based CIDT process. This aimed to (i) reduce the communication gap between FCHVs and HWs at health posts, (ii) digitize the referral process and (iii) maintain online documentation. Our goals were to increase the rate of help-seeking, the number of people initiating care after being referred, and facilitating active follow-up from FCHVs.

Setting
This study was conducted in Nepal in 2017, 2 years after a 7.8 magnitude earthquake that saw an increase in mental illnesses in affected regions. The study site, 4 village development committees (VDCs) in Sindhuli district, was an earthquake affected region that had received immediate post-earthquake mental health services sponsored by an international organization and was now transitioning to government-funded operations. The introduction of the mCIDT was to foster sustaining and scaling of mental health referrals for this government transition.

Development of mCIDT
Due to the widespread use of standard feature phones (i.e., not smartphones), an SMS approach was selected for digitizing CIDT. An iterative process was used to develop and finalize the workflow for the SMS system. Digitization of the referral component of the CIDT was selected as the high value target to achieve this and a workflow was drafted (Fig. 1a). Because we only digitized the referral mechanism, the FCHVs continued to use the paper-based CIDT for detection of people with potential mental illness. Based on our previous CIDT experience, we estimated that one week would be an appropriate time window to visit the health facility once referred by the FCHV. If the person visited the health facility, s/he was registered in the system as a "complete case", otherwise a reminder SMS was sent to the FCHV to follow up.

Project implementation
A 3-day training for the FCHVs and HWs was conducted in July 2017. The training included use of basic mobile phone functions, role-playing using mCIDT, and a review of conditions included in CIDT. A codebook describing basic phone functions and the steps to refer using mCIDT was given to each FCHV for reference. After the training, we conducted 4 focus group discussions (FGD) with 36 FCHVs, 8 key informant interviews (KIIs) with HW, 2 KIIs with mental health experts, and 1 KII with an mHealth expert.
After the training, the mCIDT platform was implemented from August-October 2017. Supervision visits were completed twice at the health post at each VDC by the community clinical supervisor. In response to the lower than expected referrals seen near the end of the third month, a simulation workshop was held to retrain the FCHVs. A written vignette describing a mock case was read out loud to the FCHVs, who were asked to identify the case using CIDT. FCHVs were asked to use the mCIDT platform to refer the mock case (See Additional file 1: Figure S1). These steps were completed for a total of five vignettes addressing depression, postpartum depression, alcohol use disorder (AUD), psychosis, and epilepsy.

Results
Demographics of the 36 FCHVs trained on mCIDT are included in Additional file 2: Table S1. Of these 36 FCHVs, only 8 FCHVs successfully implemented mCIDT, defined as referring someone. Several error messages including typos, missing spaces, wrong disorder codes, and incorrect sequencing were recorded in the system (Additional file 3: Table S2). Over 3 months of implementation, 8 FCHVs registered and referred 8 cases through mCIDT: 4 depression, 2 psychosis, 1 epilepsy, and 1 antenatal depression. No FCHV registered more than 1 case in the study period. Of those 8 cases who were referred, 2 cases visited the health facility, 2 could not be contacted in the follow-up. Four cases stated they did not seek treatment because the government health facility was not staffed with health workers.
After piloting the technology for 3 months, a simulation exercise was held with the FCHVs (n = 34) to determine their accuracy of using mCIDT (Fig. 1b). Level of education was significantly higher and age was significantly lower for the FCHVs who were able to correctly use mCIDT in comparison to those who were not able (Additional file 4: Table S3). Those who selfreported the ability to send an SMS and use the mCIDT Codebook were significantly more likely to correctly use mCIDT across all disorders.
Qualitative analysis of the KIIs, FGDs, field notes and observations by the research team elucidated the benefits of mCIDT, challenges, and recommendations to improve the program (See sample quotes from KIIs and FGDs in Table 1).

Acceptability, feasibility and benefits
The FCHVs, HWs and mental health experts, thought the greatest benefit of mCIDT could be reducing the burden of work on FCHVs. Secondary benefits included the potential for better communication between HWs and FCHVs. Mental health and mHealth experts were wary of the FCHVs ability to use the mobile phones and stated that maintaining patient confidentiality was not feasible.

Challenges
The simulation results showed that most FCHVs were unable to utilize mCIDT, this challenge was evident in the interviews as well. We summarized these challenges in 5 domains: Community, Participant, Facility, Program, and Technological. The most prominent challenges were mentioned at the community level. FCHVs repeatedly said no mental health cases were present in the community, which is inconsistent with assessments finding high rates of mental health and psychosocial problems in the area [11]. This pointed to a lack of community awareness of the burden of mental illnesses. FCHVs mentioned that previously Home-Based Community Workers (HBCW) in the area were responsible for identification of mental health cases, and they had been paid by an international organization to implement CIDT after the earthquake.
FCHVs acknowledged that stigma towards mental health is persistent in the community. If an FCHV identified someone with a potential mental illness, it was difficult to gain support from the family to get the patient to care. FCHVs were aware that AUD cases resided in the communities, but they were uncomfortable interacting with the patient fearing he was violent or thinking the patient cannot get better. FCHVs particularly felt discomfort dealing with male patients citing their gender roles. The fact that these FCHVs lived in the same community and they did not want to have potential conflict also contributed. FCHVs said that community members were worried about breaches of confidentiality due to the use of a mobile phone.
FCHVs struggled to use the mobile phone for reasons ranging from poor eyesight among older FCHVs to lack of confidence using technology. Lack of technological literacy was the most frequent issue observed during training sessions. It was also noted by trainers that the need to focus on how to use a mobile phone was unanticipated. Lower education became a barrier when trying to type and send the structured SMS. A major challenge was transferring the visual information on paper-based CIDTs into appropriate syntax for the structured SMS. Lower education also became an issue when receiving error messages and the inability to read and respond with the correction. Interviewees, who were not FCHVs, discussed government challenges in the context of implementing a policy that would set an educational threshold for FCHVs.
FCHVs are also overburdened through engagement in many parts of the health sector. Absenteeism of HW at the health post discouraged one FCHV whose referred case had to return without services. Financial incentives were brought up by most FCHVs as a way to increase motivation for them to engage in the mental health sector. Network instability was one technological challenge.

Suggested recommendations from participants
The main recommendations centered around more supervision for the FCHVs and increasing the level of awareness about mental health in the community.

Discussion
mCIDT was designed as a tool to strengthen the government's mental health services by increasing the number of cases referred from the community for care-seeking. Piloting of the mCIDT platform for 3 months in 4 VDCs resulted in only 8 referrals. This equates to 0.67 mCIDT referrals per VDC per month, with a 25% treatment seeking rate. This is considerably lower than the 615 paper CIDT referrals with 364 seeking treatment in 24 VDCs during the 2-year post-earthquake period, which equates to 1 CIDT per VDC per month, with 60% treatment seeking rate [11]. Our prior studies in non-earthquake settings in Nepal also identified a 67% treatment seeking rate [9]. Qualitative and quantitative data from the study revealed numerous barriers and challenges with implementation of mCIDT. After this pilot period, mCIDT was not continued or scaled-up due to these challenges. Although prior studies have highlighted technological barriers (e.g., lack of electricity and poor networks) as prominent barriers to mHealth in LMIC settings [12], we found that these were not among the most important barriers when introducing mCIDT as a novel mental health mHealth initiative in rural Nepal. In our study, lower educational levels, older age, and inability to send text messages were associated with incorrect use of mCIDT. This is consistent with systematic review of mHealth tools used by frontline health workers in LMICs that identified age, level of education, and years of experience of the health workers as primary barriers to adoption of mobile technologies [13]. Similarly, a study in rural Nigeria also showed that younger and more educated midwives demonstrated higher scores on knowledge assessments for mHealth technology use [14].
A striking lesson from our study was the lack of motivation to engage in mental health work. This is consistent with overall lack of importance placed upon mental health among health workers and stigmatization of providing mental health care [15]. Another challenge to motivation among FCHVs was the shift from CIDT work being compensated through an international organization in the immediate post-earthquake period to FCHVs expected to conduct their work, including CIDT, through the government system without pay after the earthquake relief phase.

Limitations
With the mCIDT tool only piloted for 3-months, the study duration was relatively short and it is difficult to identify patterns with only 8 potential mental illness cases identified and only 2 seeking treatment after the referral. When the paper version was used in other areas, this was in the context of a broader district mental health plan to improve primary care services and community awareness [16]. We did not evaluate the community's perceptions of the feasibility and acceptability of the tool prior to its implementation. In our other work with introduction of mHealth technologies in Nepal, we first conducted qualitative studies to identify acceptability and potential barriers to implementation [17].
Based on the experiences and challenges encountered in this study, we have provided recommendations for future studies introducing new technology in low resource settings to address mental health needs (see Table 2 for recommendations). Of particular importance, we recommend that any technological introduction for mental health services should be accompanied by approaches to also assure motivation to work with mental health patients [15]. This is crucial not only among community health volunteers using new technology but also throughout the health system to assure that there are health workers are motivated to treat newly identified mental health patients.
Additional file 1: Figure S1 Steps for using mCIDT to refer patient.

Additional file 2: Table S1
Demographics of Female Community Health Volunteers (n=36) trained in mCIDT platform.

Additional file 3: Table S2
Types of messages stored in server. Table S3 a Factors associated with correct and incorrect use of mCIDT for each diagnosis, categorical variables (n = 34); b Factors In settings with high stigma against mental illness, consider how technology could be used to address anonymity and reduce risk of breaching confidentiality 3. Assure that adequate services are in place for mental health care When piloting new technology, assure availability of basic mental health services to reduce risk of low adoption because of perceived lack of care/ supportive human resources or infrastructure 4. Address perceived need for both care and technological solutions Prior to implementation, explore community and health worker perceptions of need, and address these needs alongside technical skills in trainings 5. Explore multiple stakeholder groups for adopting technology Given the potential for differential uptake of technology, pilot with multiple stakeholder groups (e.g., community health workers, teachers, community groups, religious leaders)