Longterm Temporal Dynamics of Suicidal Ideation: A Dynamic Time Warping Analysis of Depression, Anxiety, Worry, and Mastery

Using Dynamic Time Warping analysis on a 13-year longitudinal dataset, this study reveals that suicidal ideation is a dynamic phenomenon that most strongly co-fluctuates with specific affective depressive symptoms, as well as acute fear and persistent worry, suggesting that personalized prevention strategies should target these specific temporal symptom patterns.

The Gist

The Big Picture: Why Static Maps Fail

Imagine you are trying to predict a storm. Traditional research often looks at a static map: "This person lives in a low-lying area (depression), so they are at risk." But storms don't just happen because of where you live; they happen because of how the wind, pressure, and temperature change right before the rain starts.

This study argues that Suicidal Ideation (SI)—the thoughts of wanting to die—is like a storm. It's not a permanent state for most people; it fluctuates up and down over time. Traditional studies take a single snapshot and miss the movement. This study wanted to watch the "weather patterns" of the mind over 13 years to see which symptoms rise and fall together with suicidal thoughts.

The Tool: "Dynamic Time Warping" (The Elastic Ruler)

The researchers used a fancy math tool called Dynamic Time Warping (DTW).

The Analogy: Imagine two runners on a track.

• Runner A (Suicidal Thoughts) speeds up, slows down, and stops.
• Runner B (Sadness) does the same, but maybe they speed up a little later than Runner A.

A standard ruler (traditional statistics) would say, "They aren't running together because they aren't side-by-side at the exact same second."
DTW is like an elastic ruler. It stretches and squishes the timeline to match the shape of their running. Even if Runner B is a few minutes late, the elastic ruler stretches the time to show, "Hey, they actually ran the same pattern!"

This allowed the researchers to see which symptoms "danced" in sync with suicidal thoughts over many years, even if the timing wasn't perfect.

The Participants: A Long-Term Journey

The study followed 305 people from the Netherlands who had depression or anxiety. They checked in on them six times over 13 years.

• They looked at: Sadness, Anxiety, Worry, and a sense of control over life (called "Mastery").
• They only included people whose suicidal thoughts actually changed (went up and down) over time, because you can't study a dance if one partner never moves.

The Findings: Who Dances with the Storm?

When the researchers stretched their elastic rulers to see what moved in sync with suicidal thoughts, they found some clear partners and some that didn't dance at all.

1. The Closest Dancers (Strong Alignment)

These symptoms rose and fell right alongside suicidal thoughts:

• The "Heavy Heart" Symptoms: Sad mood, feeling like nothing is fun (anhedonia), low energy, and pessimism.
  • Analogy: Think of these as the heavy clouds that always appear right before the lightning.
• The "Scary" Symptoms: Feeling terrified or afraid.
• The "Overwhelmed" Symptoms: Feeling like your worries are too big to handle.
  • Analogy: These are like the sudden gusts of wind that push the storm over the edge.

2. The Distant Partners (Weak Alignment)

• Mastery (Feeling in Control): This was the surprise. Feeling like you can't control your life did not rise and fall in sync with suicidal thoughts.
  • Analogy: Mastery is like the foundation of a house. It's stable. If the foundation cracks, the house is in trouble. But the foundation doesn't shake every time a storm passes. It's a long-term risk factor, not a short-term trigger. The study suggests that low mastery might make you vulnerable to the storm, but it doesn't necessarily "dance" with the storm as it happens.

The Network: A Web of Connections

The researchers drew a map (a network graph) of how these symptoms connect.

• The Core: Suicidal thoughts were tightly glued to the "depression cluster" (sadness, no joy).
• The Bridges: However, suicidal thoughts also had "bridges" connecting them to fear and worry.
• Takeaway: Suicidal thoughts aren't just about being sad. They are a mix of deep sadness plus a sudden spike of fear or a feeling of being overwhelmed by worry.

Did Age or Gender Matter?

The researchers checked if this dance was different for young men, older women, etc.

• The Result: The dance was surprisingly the same for everyone. Whether you were a young woman or an older man, the symptoms that moved with suicidal thoughts were the same: sadness, lack of joy, fear, and overwhelming worry.

What Does This Mean for Helping People?

This study changes how we might think about preventing suicide:

1. Stop Waiting for the "Bad Day": Since suicidal thoughts fluctuate, we need to catch them when they are rising, not just when someone is at their lowest point.
2. Target the Specific Triggers: If a patient is struggling, don't just treat "depression" generally. Specifically target:
   • The inability to feel pleasure (anhedonia).
   • The sudden spikes of fear.
   • The feeling of being overwhelmed by worry.
3. Real-Time Monitoring: Because these thoughts change quickly (like weather), checking in on people frequently (perhaps via apps or daily check-ins) is better than waiting for a yearly appointment.

Summary

Think of suicidal thoughts not as a permanent wall, but as a wave. This study found that the wave is most closely linked to the feeling of sadness, the loss of joy, and the sudden spike of fear or worry. While feeling out of control (low mastery) is a risk, it's the emotional storm happening right now that is most dangerous. To stop the wave, we need to calm the specific winds of fear and the heavy clouds of sadness as they happen.

Technical Summary

1. Problem Statement

Traditional suicide risk assessment relies heavily on static risk factors (e.g., history of attempts, demographic variables, or cross-sectional symptom severity). However, these factors are poor predictors of the dynamic emergence and fluctuation of suicidal ideation (SI), which often varies significantly over short-to-intermediate timescales.

• The Gap: Existing longitudinal research often uses cross-sectional designs or limited two-wave assessments, failing to capture the temporal interplay between SI and co-occurring symptoms (depression, anxiety, worry, mastery) over extended periods.
• The Challenge: Standard longitudinal models assume synchronized timing or fixed functional forms, making it difficult to compare trajectories when individuals change at different rates or times. Furthermore, the "ergodicity problem" suggests that group-level averages often fail to reflect individual within-person processes.
• Objective: To identify specific symptom dynamics that co-fluctuate with SI over a 13-year period using a method capable of aligning non-synchronous time-series data.

2. Methodology

Study Design and Sample

• Data Source: The Netherlands Study of Depression and Anxiety (NESDA), a naturalistic multisite cohort study.
• Participants: $n = 305$ adults (mean age 40.8, 62% female) with a history of depressive/anxiety disorders who exhibited at least one fluctuation in suicidal ideation across the study waves.
• Timeframe: Six assessment waves: Baseline (W1) and follow-ups at 2, 4, 6, 9, and 13 years.
• Stratification: Analyses were stratified into four subgroups: Younger Women (<40), Older Women (≥40), Younger Men (<40), and Older Men (≥40).

Measures

• Suicidal Ideation (SI): Composite score derived from two items:
  1. Item 2 of the Scale for Suicide Ideation (SSI): "Wish to die."
  2. Item 18 of the Inventory of Depressive Symptomatology (IDS): "Suicidal thoughts."
• Depression: 29 items from the IDS (self-report).
• Anxiety: 21 items from the Beck Anxiety Inventory (BAI).
• Worry: 11 items from the Penn State Worry Questionnaire (PSWQ-11).
• Mastery: 7 items from the Pearlin Mastery Scale (belief in control over life circumstances).

Analytical Approach: Dynamic Time Warping (DTW)

The study employed Dynamic Time Warping (DTW), a non-linear time-series alignment algorithm, to quantify the temporal similarity between trajectories.

• Preprocessing:
  • Variables were standardized (z-scores) and winsorized (truncated at $\pm 2.5$) to handle outliers.
  • Interpolation: Five equally spaced values were added between observed time points to create smoother trajectories and reduce endpoint distortion.
  • Directionality: Scores were recoded so higher values always indicated greater distress (e.g., lower mastery scores were inverted).
• DTW Application:
  • DTW calculated the "warped distance" between the composite SI trajectory and each individual symptom trajectory for every participant.
  • Constraint: A Sakoe-Chiba band of 1 time point was applied to prevent overfitting and ensure alignment occurred within a reasonable temporal window.
  • Metric: A smaller DTW distance indicates stronger temporal alignment (i.e., symptoms rise and fall together over time).
• Statistical Modeling:
  • Linear mixed-effects models estimated relative alignment, adjusting for age, sex, education, and baseline antidepressant use.
  • Random intercepts accounted for clustering within individuals.
  • Network Analysis: An undirected network visualized co-fluctuations, where edge thickness represented the inverse of the mean DTW distance.

3. Key Contributions

1. Methodological Innovation: Successfully applied DTW to long-term, multi-year longitudinal data (13 years, 6 waves) rather than just high-frequency Ecological Momentary Assessment (EMA). This demonstrates DTW's utility in capturing slower, clinically meaningful shifts in symptomatology.
2. Within-Person Focus: Addressed the ergodicity problem by analyzing individual temporal alignment before aggregating, revealing dynamics that group averages obscure.
3. Granular Symptom Mapping: Moved beyond broad diagnostic categories to identify specific items (e.g., "sad mood" vs. general depression) that drive SI fluctuations.

4. Results

Overall Symptom Network

• Depressive Symptoms: SI showed the strongest temporal alignment with affective and anhedonic depressive symptoms. Key aligned items included:
  • Sad mood (IDS5)
  • Low capacity for pleasure/Anhedonia (IDS21)
  • Low general interest (IDS19)
  • Pessimism (IDS17)
  • Quality of mood (IDS10)
  • Responsiveness of mood (IDS8)
  • Decreased appetite (IDS11)
• Anxiety and Worry:
  • Anxiety: Only specific cognitive-affective items aligned significantly, primarily "Terrified or afraid" (BAI9).
  • Worry: "Overwhelming worries" (PSWQ item) showed significant alignment with SI.
• Mastery: No mastery items showed significant temporal alignment with SI.
• Network Structure: SI acted as a bridge within a cluster of depressogenic symptoms but also connected to acute fear and persistent worry.

Subgroup Analysis

• Consistency: The pattern of alignment (Depression > Anxiety/Worry > Mastery) was broadly consistent across all four subgroups (Young/Older Men and Women).
• Nuance: Younger women showed slightly stronger proximity to SI regarding "Responsiveness of mood" and "Low capacity for pleasure" compared to other subgroups, but the overall dynamic structure remained stable.

5. Significance and Implications

Clinical Implications

• Targeted Interventions: Suicide prevention strategies should move beyond general depression treatment to target specific dynamic drivers: mood instability, anhedonia (inability to feel pleasure), and uncontrollable worry.
• Real-Time Monitoring: The findings support the use of real-time monitoring to detect spikes in these specific symptoms (sadness, anhedonia, overwhelming worry) as immediate precursors to SI escalation.
• Mastery as a Trait: The lack of temporal alignment for mastery suggests it is a stable trait that influences vulnerability indirectly (perhaps by buffering anxiety) rather than a proximal, fluctuating driver of SI. Interventions should not expect mastery to fluctuate in sync with SI.

Theoretical Contribution

• Dynamic vs. Static: The study validates that SI is a dynamic phenomenon closely linked to specific symptom clusters rather than a static outcome of overall severity.
• Transdiagnostic Pathways: The alignment of SI with both "terrified/afraid" (anxiety) and "overwhelming worries" (cognitive) supports the Integrated Motivational-Volitional (IMV) model, highlighting fear and entrapment/worry as key bridging mechanisms.

Limitations

• Correlational Nature: DTW identifies alignment but does not prove causality.
• Data Sparsity: While DTW handles uneven spacing, the 6-wave design (spanning 13 years) may miss rapid, short-term fluctuations that EMA studies capture.
• Sample Bias: The cohort oversampled individuals with diagnosed disorders, potentially limiting generalizability to the general population.

Conclusion

This study demonstrates that suicidal ideation fluctuates in tandem with specific affective depressive symptoms (sadness, anhedonia) and cognitive worry, rather than general distress or stable traits like mastery. By utilizing Dynamic Time Warping on long-term longitudinal data, the authors provide a robust framework for identifying temporally proximal symptom patterns, offering a pathway toward more personalized and dynamic suicide prevention strategies.