Intensity, Not Just Instability: What Moment-to-Moment Affect Reveals About Suicidal Risk
- Across 74 depressed psychiatric inpatients sampled ten times a day for six days, the average level of affect – not its turbulence – carried the strongest and most consistent link to suicidal ideation.
- Within a person, a rise in positive affect predicted lower passive and active suicidal ideation at the same moment, while a rise in negative affect predicted higher ideation, even after adjusting for ideation at the previous prompt.
- Counterintuitively, greater variability and higher inertia of positive affect across persons were tied to lower overall ideation, suggesting that a fuller, more responsive positive-affect system may be protective rather than a sign of fragility.
- Dynamic features of negative affect – its instability and abrupt swings – showed no reliable within-person association with ideation once intensity was accounted for.
The clinical folklore around mood instability runs in one direction: the more an affect swings, the more dangerous the state. This high-frequency study from Leipzig, Duisburg-Essen and Luxembourg interrogates that assumption directly, and the answer is more disciplined than the folklore. In a high-risk sample of inpatients with major depressive disorder, the team measured positive affect, negative affect and both passive and active suicidal ideation sixty times over six days, then decomposed the affect signal into intensity (mean level) and several dynamic indices – variability, instability, acute change and inertia.
The headline is a correction of emphasis. Affective intensity, not affective dynamics, showed the most robust associations with suicidal ideation at both the moment-to-moment and the between-person level. A within-person uptick in positive affect tracked a concurrent drop in ideation; an uptick in negative affect tracked a rise. These links survived adjustment for ideation at the prior prompt, which matters: it argues that affect is not merely echoing a stable suicidal state but moving alongside it in real time.
Where the study becomes genuinely instructive is in what failed to predict. Moment-to-moment instability of negative affect, and abrupt acute changes in affect, did not hold a within-person relationship with ideation once intensity was in the model. The intuition that a jagged negative-affect trace flags imminent risk is not supported here. Even more striking, at the between-person level, people whose positive affect was more variable and showed higher inertia reported lower overall ideation. A positive-affect system that still moves – that can be lifted and can carry that lift forward – reads as a marker of preserved capacity, not instability.
For the practitioner, the operational lesson is to resist over-reading the shape of the curve. When a patient's negative affect is volatile, the volatility itself is weaker signal than the simple question of how high the negative affect sits and how flat the positive affect has become. Anhedonic flattening of positive affect – low level, low responsiveness – is the configuration this dataset associates with elevated risk, and it is precisely the configuration that a clinician scanning for dramatic swings may overlook. The study is correlational and confined to a single diagnostic group, so it cannot license a risk algorithm. But it sharpens a transdiagnostic question that runs well beyond depression: when we monitor affect in daily life, are we measuring the right property of it?
Why dynamics seduced the field
Ecological momentary assessment made affective variability cheap to compute, and a decade of work treated higher variability as inherently pathological. This study is a reminder that a dynamic index is only as meaningful as its competition: once mean intensity is held constant, several celebrated dynamic markers lose their predictive grip.
What a clinician should watch instead
The configuration worth flagging is a depressed positive-affect system – low, unreactive, slow to recover – paired with elevated mean negative affect. That static portrait outperformed the turbulence metrics here, and it is one a structured daily check-in can capture without sophisticated modelling.
In this high-risk inpatient sample, how high or low affect sat predicted suicidal ideation better than how violently it swung.
The sample was 74 inpatients with major depressive disorder, so generalisation to outpatients, other diagnoses and milder presentations is untested. The design is observational and within-person associations are concurrent, which cannot establish that affect change causes ideation change. Suicidal ideation was self-reported during a hospital stay, a context that may itself compress the range of affect and risk.