Negative Affect Linked to Suicidal Ideation in EMA Study

Key Takeaways

• Among high-risk psychiatric outpatients in Madrid, greater negative affect was significantly associated with greater suicidal ideation during smartphone monitoring.
• Interpersonal difficulties were also significantly associated with suicidal ideation, including when modeled alongside negative affect.
• Suicidal ideation fluctuated considerably within individuals without a significant overall time trend.
• Sleep problems and appetite were not significantly associated with mean suicidal ideation in the corresponding smartphone models, although separate analyses linked their variability with variability in ideation.

Suicidal thoughts and emotional distress can change between psychiatric outpatient visits, making their relationship difficult to characterize through retrospective assessment alone. Patients’ daily smartphone answers can capture experiences closer to when they occur and show how those experiences vary during outpatient care. Adults recently treated after suicidal behavior or intent provided repeated reports that allowed investigators to examine these changes.

Investigators analyzed prospective, observational ecological momentary assessment (EMA) data from both arms of SmartCrisis 2.0, as described in Porras-Segovia and colleagues’ six-month ecological momentary assessment study in the Journal of Medical Internet Research. The analysis was not a randomized test of the app or safety plan. Hospitals in Madrid, Spain, recruited 115 adults who had experienced a suicide attempt or emergency intervention for suicidal intention in the preceding month; 99 met the minimum response threshold for analysis. Over six months, smartphone prompts appeared daily at random times. Each included a suicidal-ideation item, while other questions rotated among negative affect, interpersonal experiences, sleep, eating, and nonsuicidal self-injury. Investigators also collected baseline clinical assessments and used mixed-effects models to examine associations and time trends.

Suicidal ideation varied considerably within individuals during monitoring: its intraclass correlation coefficient (ICC) was 0.17. Participants submitted 13,900 EMA responses, for an approximately 29.3% overall response rate. Suicidal ideation, negative affect, and interpersonal difficulties showed no statistically significant overall time trends. Sleep problems and appetite had no significant association with mean suicidal ideation in their respective EMA models, although separate participant-level analyses linked variability in each domain with variability in ideation.

Greater negative affect (β=0.193) and greater interpersonal difficulties (β=0.17) were each associated with greater concurrent suicidal ideation; both associations had P<.001 in the full-text results. Both domains remained significantly associated with ideation when included in the same model. Their interactions with time were not significant. These models do not establish whether emotional or interpersonal distress preceded subsequent suicidal thoughts.

Rotating questions limited continuity for individual items and prevented lagged analyses, leaving the temporal order of these associations unresolved. The concurrent findings cannot establish causation, individual-level forecasting, or a reduction in suicide attempts. Overall completion was limited, and rotation precluded item-specific response rates. Question time frames differed, and participants could respond until midnight, tempering a strictly momentary reading of their answers. Baseline psychological-pain and depression measures were available only for a smaller subset. Monitoring data were pooled across randomized arms; a nonsignificant safety-plan covariate analysis does not demonstrate intervention effectiveness.

The authors observed that suicidal ideation fluctuated within individuals during repeated monitoring of high-risk Madrid outpatients and was associated with emotional and interpersonal distress.

Clinician Questions

Which emotions made up the negative-affect measure in smartphone monitoring of suicidal ideation?

Investigators combined questions about stress, restlessness, hopelessness, anxiety, sadness, psychological pain, and anger toward oneself or others, with happiness reverse-coded. This measure operationalized negative affect rather than identifying a separate clinical diagnosis.

Did baseline psychological pain or depression relate to suicidal ideation during six-month smartphone monitoring?

Among 37 high-risk Madrid psychiatric outpatients with baseline symptom data, psychological pain interacted with follow-up duration in analyses of average suicidal ideation: lower initial pain was linked to less ideation with longer follow-up, but baseline psychological pain alone did not significantly predict average ideation. Baseline depression analyses were not significantly associated with suicidal ideation; a depression-by-follow-up interaction was associated instead with negative affect. These were secondary subset associations, not evidence that changing pain or depression prevents suicidal thoughts.

Were passive wishes to die included in the suicidal-ideation definition used for Madrid psychiatric outpatients?

Yes. Suicidal ideation in the Madrid outpatient smartphone investigation included passive wishes to be dead and active thoughts of ending one’s life when no preparatory behavior was involved. That definition differs from suicide attempts and nonsuicidal self-injury; it does not establish either as a predicted outcome.