Suicidal ideation arises from a complex interplay of multiple interacting risk factors over time. Recently, ecological momentary assessment (EMA) has increased our understanding of factors associated with real-time suicidal ideation, as well as those predicting ideation at the level of hours and days. Here we used statistical network methods to investigate which cognitive-affective risk and protective factors are associated with the temporal dynamics of suicidal ideation.

The SAFE study is a longitudinal cohort study of 82 participants with current suicidal ideation who completed 4×/day EMA over 21 days. We modeled contemporaneous (t) and temporal (t + 1) associations of three suicidal ideation components (passive ideation, active ideation, and acquired capability) and their predictors (positive and negative affect, anxiety, hopelessness, loneliness, burdensomeness, and optimism) using multilevel vector auto-regression models.

Contemporaneously, passive suicidal ideation was positively associated with sadness, hopelessness, loneliness, and burdensomeness, and negatively with happiness, calmness, and optimism; active suicidal ideation was positively associated with passive suicidal ideation, sadness, and shame; and acquired capability only with passive and active suicidal ideation. Acquired capability and hopelessness positively predicted passive ideation at t + 1, which in turn predicted active ideation; acquired capability was positively predicted at t + 1 by shame, and negatively by burdensomeness.

Our findings show that systematic real-time associations exist between suicidal ideation and its predictors, and that different factors may uniquely influence distinct components of ideation. These factors may represent important targets for safety planning and risk detection.

Depression is a highly recurrent disorder, with more than 50% of those affected experiencing a subsequent episode. Although there is relatively little stability in symptoms across episodes, some evidence indicates that suicidal ideation may be an exception. However, these findings warrant replication, especially over longer periods and across multiple episodes.

To assess the relative stability of suicidal ideation in comparison with other non-core depressive symptoms across episodes.

We examined 490 individuals with current major depressive disorder (MDD) at baseline and at least one subsequent episode during 9-year follow-up within the Netherlands Study of Depression and Anxiety (NESDA). The Inventory of Depressive Symptomatology (IDS) was used to assess DSM-5 non-core MDD symptoms (fatigue, appetite/weight change, sleep disturbance, psychomotor disturbance, concentration difficulties, worthlessness/guilt, suicidal ideation) at baseline and 2-, 4-, 6- and 9-year follow-up. We examined consistency in symptom presentation (i.e. whether the symptom met the diagnostic threshold, based on a binary categorisation of the IDS) using kappa (κ) and percentage agreement, and stability in symptom severity using Spearman correlation, based on the continuous IDS scores.

Out of all non-core depressive symptoms, insomnia appeared the most stable across episodes (r = 0.55–0.69, κ = 0.31–0.47) and weight decrease the least stable (r = 0.03–0.33, κ = 0.06–0.19). For suicidal ideation, correlations across episodes ranged from r = 0.36 to r = 0.55 and consistency ranged from κ = 0.28 to κ = 0.49.

Suicidal ideation is moderately stable in recurrent depression over 9 years. Contrary to prior reports, however, it does not exhibit substantially more stability than most other non-core symptoms of depression.

The DSM-5 features hundreds of diagnoses comprising a multitude of symptoms, and there is considerable repetition in the symptoms among diagnoses. This repetition undermines what we can learn from studying individual diagnostic constructs because it can obscure both disorder- and symptom-specific signals. However, these lost opportunities are currently veiled because symptom repetition in the DSM-5 has not been quantified.

This descriptive study mapped the repetition among the 1419 symptoms described in 202 diagnoses of adult psychopathology in section II of the DSM-5. Over a million possible symptom comparisons needed to be conducted, for which we used both qualitative content coding and natural language processing.

In total, we identified 628 distinct symptoms: 397 symptoms (63.2%) were unique to a single diagnosis, whereas 231 symptoms (36.8%) repeated across multiple diagnoses a total of 1022 times (median 3 times per symptom; range 2–22). Some chapters had more repetition than others: For example, every symptom of every diagnosis in the bipolar and related disorders chapter was repeated in other chapters, but there was no repetition for any symptoms of any diagnoses in the elimination disorders, gender dysphoria or paraphilic disorders. The most frequently repeated symptoms included insomnia, difficulty concentrating, and irritability – listed in 22, 17 and 16 diagnoses, respectively. Notably, the top 15 most frequently repeating diagnostic criteria were dominated by symptoms of major depressive disorder.

Overall, our findings lay the foundation for a better understanding of the extent and potential consequences of symptom overlap.

Few factor analyses and no network analyses have examined the structure of DSM phobic fears or tested the specificity of the relationship between panic disorder and agoraphobic fears.

Histories of 21 lifetime phobic fears, coded as four-level ordinal variables (no fear to fear with major interference) were assessed at personal interview in 7514 adults from the Virginia Twin Registry. We estimated Gaussian Graphical Models on individual phobic fears; compared network structures of women and men using the Network Comparison Test; used community detection to determine the number and nature of groups in which phobic fears hang together; and validated the anticipated specific relationship between panic disorder and agoraphobia.

All networks were densely and positively inter-connected; networks of women and men were structurally similar. Our most frequent and stable solution identified four phobic clusters: (i) blood-injection, (ii) social-agoraphobia, (iii) situational, and (iv) animal-disease. Fear of public restrooms and of diseases clustered with animal and not, respectively, social and blood-injury phobias. When added to the network, the three strongest connections with lifetime panic disorder were all agoraphobic fears: being in crowds, going out of the house alone, and being in open spaces

Using network analyses applied to a large epidemiologic twin sample, we broadly validated the DSM-IV typography but did not entirely support the distinction of agoraphobic and social phobic fears or the DSM placements for fears of public restrooms and diseases. We found strong support for the specificity of the relationship between panic disorder and agoraphobic fears.

While taxonomy segregates anxiety symptoms into diagnoses, patients typically present with multiple diagnoses; this poses major challenges, particularly for youth, where mixed presentation is particularly common. Anxiety comorbidity could reflect multivariate, cross-domain interactions insufficiently emphasized in current taxonomy. We utilize network analytic approaches that model these interactions by characterizing pediatric anxiety as involving distinct, inter-connected, symptom domains. Quantifying this network structure could inform views of pediatric anxiety that shape clinical practice and research.

Participants were 4964 youths (ages 5–17 years) from seven international sites. Participants completed standard symptom inventory assessing severity along distinct domains that follow pediatric anxiety diagnostic categories. We first applied network analytic tools to quantify the anxiety domain network structure. We then examined whether variation in the network structure related to age (3-year longitudinal assessments) and sex, key moderators of pediatric anxiety expression.

The anxiety network featured a highly inter-connected structure; all domains correlated positively but to varying degrees. Anxiety patients and healthy youth differed in severity but demonstrated a comparable network structure. We noted specific sex differences in the network structure; longitudinal data indicated additional structural changes during childhood. Generalized-anxiety and panic symptoms consistently emerged as central domains.

Pediatric anxiety manifests along multiple, inter-connected symptom domains. By quantifying cross-domain associations and related moderation effects, the current study might shape views on the diagnosis, treatment, and study of pediatric anxiety.

In clinical research, populations are often selected on the sum-score of diagnostic criteria such as symptoms. Estimating statistical models where a subset of the data is selected based on a function of the analyzed variables introduces Berkson's bias, which presents a potential threat to the validity of findings in the clinical literature. The aim of the present paper is to investigate the effect of Berkson's bias on the performance of the two most commonly used psychological network models: the Gaussian Graphical Model (GGM) for continuous and ordinal data, and the Ising Model for binary data.

In two simulation studies, we test how well the two models recover a true network structure when estimation is based on a subset of the data typically seen in clinical studies. The network is based on a dataset of 2807 patients diagnosed with major depression, and nodes in the network are items from the Hamilton Rating Scale for Depression (HRSD). The simulation studies test different scenarios by varying (1) sample size and (2) the cut-off value of the sum-score which governs the selection of participants.

The results of both studies indicate that higher cut-off values are associated with worse recovery of the network structure. As expected from the Berkson's bias literature, selection reduced recovery rates by inducing negative connections between the items.

Our findings provide evidence that Berkson's bias is a considerable and underappreciated problem in the clinical network literature. Furthermore, we discuss potential solutions to circumvent Berkson's bias and their pitfalls.