Feasibility and adherence to ecological momentary assessment among community-dwelling adults with suicide risk

For nursing students

Study summary

Suicidal thoughts do not stay constant. They can rise and fall within hours, which makes them hard to capture with a single clinic visit or a one-time questionnaire. Researchers in South Korea tested a way to track these fluctuations in real time using a method called ecological momentary assessment, or EMA. This study, published in BMC Nursing in 2025, asked whether community-dwelling adults with a history of suicidal ideation could and would keep up with an intensive, month-long EMA protocol.

Twenty-two adults with suicidal ideation were recruited through a suicide-prevention center. Participants were adults aged 19 and older who owned a smartphone and were able to wear a wrist device. People with active psychosis needing hospitalization or significant cognitive impairment were excluded. Over 28 days, participants received three text-message prompts per day, timed individually around each person's wake time, bedtime, and reported peak times for suicidal thoughts. At each prompt, they rated their depression, anxiety, and stress on short 1-to-5 scales. At the same time, for the first 14 days, participants wore an Actiwatch Spectrum PRO, a research-grade actigraphic wristband that records movement and sleep patterns. Whenever they felt a strong urge to act on suicidal thoughts, they were asked to press an event-marker button on the device, creating a timestamped record of suicidal impulses as they happened, not as recalled later.

Of the 22 enrolled, 20 completed the study, a 90.9% retention rate. The average EMA survey response rate across the month was 82.05%, but it was not steady: response rates were higher in the first two weeks (86.96%) and dropped in the second two weeks (76.31%), suggesting that participants grew tired of the frequent check-ins over time. In contrast, adherence to wearing the Actiwatch stayed high throughout, averaging 98.1%. These two adherence measures were moderately linked (r = .53), meaning people who wore the device consistently also tended to keep answering the surveys.

The researchers also looked at who struggled most with the protocol. Participants with higher depression and anxiety scores at baseline were more likely to have lower Actiwatch adherence, while participants with higher perceived stress were more likely to have lower EMA response rates. In other words, the people whose mental health symptoms were most severe were also the ones most likely to fall behind on the monitoring tasks meant to help understand their risk. This is an important paradox for anyone designing a monitoring program: the tool works, but it may work least well for the people who need it most unless extra support is built in.

Looking at when suicidal impulses actually happened, button presses on the Actiwatch clustered around 9 to 10 p.m., with smaller peaks around midday and early evening. Activations were lowest in the very early morning, between about 4 and 6 a.m. This pattern fits with other research showing that people with depression often have delayed or disrupted daily biological rhythms, and it hints that evening hours may be a higher-risk window worth watching for this population.

Overall, the study's authors concluded that EMA combined with a wearable actigraphic device is feasible for monitoring suicidal ideation and impulses in the community over about a month, and that it could be a useful complementary tool alongside standard clinical suicide-risk assessment. But feasibility came with a catch: the burden of frequent self-report surveys caused fatigue, and that fatigue landed hardest on participants who were already more symptomatic.

For nursing and other mental health professionals, this suggests that any EMA-based remote monitoring program should include a plan to watch for declining response rates as a signal in itself, and to build in flexibility, reminders, or check-ins for participants whose engagement drops. It also reinforces that a wearable device that gives participants direct feedback (in this case, about their sleep) may support better adherence than pure self-report alone. The study was small, drawn from one country and one recruitment site, and observational, so its numbers should be read as encouraging early evidence rather than a settled protocol. Larger, more diverse trials will be needed before EMA becomes a standard part of community suicide-risk monitoring, but this study is a useful, concrete step toward understanding what such monitoring will require.

For educators

Teaching and appraisal notes

This secondary analysis, published in BMC Nursing (2025) by Kim and colleagues, examines the feasibility and adherence patterns of a 28-day ecological momentary assessment (EMA) protocol combined with actigraphic monitoring in 22 community-dwelling South Korean adults with a history of suicidal ideation, recruited through a suicide-prevention center between May and October 2021. Twenty participants (90.9%) completed the study. The design pairs three daily individualized text-message EMA prompts (rating depression, anxiety, and stress on 5-point scales, timed around each participant's wake time, bedtime, and self-reported peak ideation periods) with 14 days of Actiwatch Spectrum PRO wear, on which participants pressed an event-marker button whenever they experienced a strong suicidal impulse. Baseline mental health characteristics were captured with the Beck Scale for Suicidal Ideation, PHQ-9, GAD-7, PSS, and AUDIT-K.

For appraisal purposes, this is a feasibility/adherence study, not an efficacy or intervention trial — it does not test whether EMA changes suicide outcomes, only whether the monitoring protocol itself can be sustained by a vulnerable population and what factors predict engagement. The core findings worth foregrounding with students: overall EMA response was 82.05%, but it declined from 86.96% in the first two weeks to 76.31% in the second two weeks, an adherence-fatigue pattern the authors attribute directly to participant burden from the intensity of the protocol. Actiwatch adherence, by contrast, remained high (98.1%) and was moderately correlated with EMA response (r = .53, p = .016). Critically, higher baseline depression and anxiety were associated with lower Actiwatch adherence, and higher perceived stress was associated with lower EMA response — meaning the participants with the greatest symptom burden were also the least able to sustain the monitoring protocol designed to track them. A secondary, exploratory finding of clinical interest is the temporal clustering of suicidal-impulse button presses around 9–10 p.m., with the fewest activations between 4 and 6 a.m., a pattern the authors relate to literature on evening chronotype and circadian misalignment in depression.

In seminar, this article supports several appraisal threads. First, sample size (n = 20 analyzed) is very small for correlational claims; the r = .53–.67 range reported should be treated as hypothesis-generating, not confirmatory, and students should practice distinguishing "statistically associated" from "clinically actionable." Second, the study is single-site, single-country, and recruited through one suicide-prevention center, limiting generalizability across health systems, including the Canadian context. Third, the differential-adherence finding — that the most symptomatic participants adhered least — is a methodological caution with direct clinical translation: any digital monitoring protocol deployed with high-risk populations needs built-in adherence-support mechanisms (reminders, simplified prompts, staff check-ins triggered by declining response) rather than assuming uniform engagement. Fourth, students should note the study offers no data on whether real-time monitoring changed any participant's clinical trajectory or care; it establishes feasibility of data collection only, not of intervention.

Practice-relevant discussion points include: how declining EMA response rate could itself function as a clinical warning sign prompting outreach; how wearable-generated sleep feedback may have supported adherence better than repeated self-report alone; and how identified evening-hour risk clustering might inform timing of check-ins or safety planning conversations. Faculty may pair this article with the study's cited protocol paper and with broader digital-phenotyping literature on suicide risk to contrast passive sensing approaches against the active, effortful EMA approach used here.