📄 social_science

Should Child Pain Anxiety Be Modeled as a Multidimensional Latent Construct or as a Network of Interrelated Symptoms? A Psychometric Comparison in Spanish Clinical and Community Populations

This study demonstrates that pediatric pain anxiety is better conceptualized as a network of mutually reinforcing symptoms rather than a single latent trait, revealing that a network model outperforms traditional factor models in fit and clinical utility while highlighting the necessity of contextualized interpretation across clinical and community populations.

Original authors: Beatriz Huidobro Labarga, Elisa López Dolado, Elisa Aguirre Sánchez, Melchor Álvarez de Mon Soto, Raúl Castañeda Vozmediano

Published 2026-07-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Beatriz Huidobro Labarga, Elisa López Dolado, Elisa Aguirre Sánchez, Melchor Álvarez de Mon Soto, Raúl Castañeda Vozmediano

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to understand why a child is afraid of pain. For a long time, psychologists have treated this fear like a hidden treasure chest. They believed that if you asked enough questions, you could find a single, invisible "anxiety monster" inside the child's mind that caused all the scary thoughts, the racing heart, and the desire to hide. This is called the "Latent Variable" model.

However, this new study suggests a different idea: maybe there is no single monster. Instead, the fear is more like a web of dominoes. If one domino (like a scary thought) falls, it knocks over the next one (like a racing heart), which knocks over the next (like wanting to avoid moving). In this view, the symptoms talk to each other directly, creating a self-sustaining cycle.

Here is what the researchers did and found, using simple terms:

The Experiment: Two Ways to Look at the Map

The researchers took a questionnaire called the CPASS (Child Pain Anxiety Symptoms Scale), which has 20 questions about how kids feel when they hurt. They asked 415 children and teenagers (some from hospitals, some from regular schools) to fill it out.

They then tried to draw a map of the answers using two different GPS systems:

The "Hidden Treasure" Map (Latent Model): This tries to group the answers into hidden categories (like "Fear" or "Avoidance").
The "Domino Web" Map (Network Model): This draws lines between the specific questions to see which ones trigger each other directly.

The Results: Which Map is Better?

When they compared the two maps, the "Domino Web" (Network Model) fit the data much better.

The Winner: The network model showed that the symptoms are indeed a web of interacting parts. It was the most accurate way to describe how the children's minds were working.
The Runner-Up: If you must use the "Hidden Treasure" approach, the researchers found that the old idea of four separate categories was too messy. Instead, the answers grouped best into just two main buckets:
1. Fear and Catastrophizing: Thoughts like "This pain means something terrible is happening" or "I'll never move again."
2. Anxiety and Avoidance: The physical feeling of panic and the urge to run away or hide.

The "Bridge" Symptoms

In the web model, the researchers found some specific questions that acted like bridges. These are the "super-dominoes."

One key question was: "I worry when I feel pain."
This question didn't just belong to one group; it connected the "scary thoughts" to the "panic feelings."
Why it matters: In the old "Hidden Treasure" model, some of these bridge questions were considered "bad" or "confusing" and were sometimes thrown away. But in the new "Web" model, these bridges are actually the most important parts because they hold the whole system together. If you fix the bridge, you might stop the whole chain reaction.

The "Context" Problem

The researchers also checked if the map looked the same for different groups of kids.

Boys vs. Girls: The maps looked mostly the same, though girls scored slightly higher on anxiety.
Young vs. Old: The maps were similar for kids under 12 and teens over 12.
Hospital vs. School: This is where the maps were totally different.
- The web of symptoms for kids in the hospital (who have serious, ongoing medical issues) looked very different from the web for kids in regular schools (who might have a headache or a scraped knee).
- The Lesson: You cannot simply compare the scores of a sick child to a healthy child as if they are on the same scale. The "rules" of how their anxiety works are different depending on their situation.

The Bottom Line

This study tells us that to understand a child's fear of pain, we shouldn't just look for a single invisible cause. We should look at how their specific fears, thoughts, and behaviors are connected to each other, like a web.

For the Test: The Spanish version of the CPASS is a good tool, but it works best if we look at the two main groups (Fear vs. Avoidance) rather than trying to force it into four old categories.
For the Doctors: When looking at the results, they need to be careful. A score from a child in a hospital might mean something different than the same score from a child at home, because the "web" of their anxiety is built differently.

In short: Pain anxiety isn't just one thing hiding inside a child; it's a complex, interconnected system of dominoes that needs to be understood as a whole, especially when comparing different groups of kids.

Technical Summary: Modeling Child Pain Anxiety as a Latent Construct vs. a Network

Problem Statement
Pain anxiety in children and adolescents is a critical psychological process predicting functional disability and poor adjustment, often exceeding the predictive power of pain intensity itself. While the Child Pain Anxiety Symptoms Scale (CPASS) is the primary instrument for assessing this construct, a fundamental psychometric question remains unresolved: do the symptoms captured by the scale reflect a single underlying latent trait (a common cause) or a network of directly interacting, mutually reinforcing symptoms? This distinction carries significant theoretical and clinical implications. Historically, psychometric research in pediatric pain has relied exclusively on latent variable models (e.g., Confirmatory Factor Analysis), potentially overlooking the dynamic interplay between symptoms. Furthermore, previous validations of the Spanish CPASS have been limited by homogeneous samples, small sample sizes for advanced modeling, and a lack of systematic measurement invariance testing across clinical and community populations.

Methodology
The study employed a rigorous, fully cross-validated psychometric pipeline involving 415 children and adolescents (aged 8–18) recruited from both clinical settings ( $n=195$ ) and community schools ( $n=220$ ) in Spain. The sample was randomly split into two halves to ensure independent exploratory and confirmatory phases.

Exploratory Phase: The first half of the data underwent Exploratory Factor Analysis (EFA) and Exploratory Graph Analysis (EGA) to identify dimensionality.
Confirmatory Phase: The second half was used to test competing models:
- Latent Variable Models: Three competing structures were evaluated: a two-factor model, a four-factor model (the original theoretical structure), and a second-order four-factor model.
- Network Model: A confirmatory psychometric network model was estimated using the psychonetrics package, treating symptoms as directly interacting nodes.
Model Comparison: Fit indices (CFI, TLI, RMSEA, SRMR, AIC) were compared across all models. Non-overlapping confidence intervals for RMSEA were used to determine superior fit.
Invariance Testing: Both measurement invariance (within the latent framework) and network structure invariance were tested across sex, age (under/over 12), and clinical status (clinical vs. community).
Additional Psychometrics: Essential unidimensionality (via bifactor analysis and PRMSE), internal consistency (Cronbach's $\alpha$ and McDonald's $\omega$ ), and item-level statistics were assessed.

Key Results

Model Fit Comparison: The confirmatory network model demonstrated superior fit compared to all latent variable specifications. It yielded the highest CFI and TLI values, the lowest RMSEA (with non-overlapping 95% confidence intervals), and the lowest AIC. This suggests that pain anxiety symptoms are better characterized as a system of direct interactions rather than passive reflections of a latent trait.
Factorial Structure: Within the latent framework, a parsimonious two-factor structure ("Fear and Catastrophizing about Pain" and "Anxiety and Avoidance Response to Pain") outperformed the original four-factor model. The two-factor solution showed superior factorial congruence across clinical and community samples and higher internal consistency. The original four-factor model, particularly the "Physiological Anxiety" subscale, demonstrated marginal reliability ( $\alpha = .67$ ).
Bridge Symptoms: Network analysis identified specific items (notably items 9, 14, and 19) as highly central "bridge symptoms" connecting different symptom domains. These items, which showed cross-loadings in EFA, were previously proposed for removal in some validations but are retained here as clinically meaningful connectors.
Invariance Findings:
- Gender and Age: Both measurement and network invariance were largely supported across sex and age groups, allowing for valid mean comparisons. Females scored significantly higher than males; no significant age differences were found in mean scores.
- Clinical vs. Community: Crucially, both measurement invariance and network structure invariance broke down between clinical and community samples. The connectivity patterns and factor loadings differed significantly, indicating that the symptom architecture of pain anxiety operates differently in clinical populations compared to community settings.

Key Contributions

First Direct Comparison: This is the first study to directly compare latent variable and psychometric network models within a pediatric pain anxiety instrument.
Methodological Advancement: The study demonstrates that relying solely on Confirmatory Factor Analysis (CFA) can miss critical structural differences, such as the breakdown of invariance between clinical and community groups, which was only fully revealed through the combined latent-network approach.
Refined Dimensionality: The study supports a shift from the original four-factor model to a more robust two-factor structure for the Spanish CPASS, improving reliability and interpretability.
Clinical Target Identification: By identifying bridge symptoms (e.g., worry about pain) that connect cognitive and behavioral domains, the study highlights specific targets for intervention that may disrupt the self-reinforcing cycles of pain anxiety.

Significance and Claims
The paper claims that pediatric pain anxiety is best understood as a network of mutually reinforcing symptoms rather than a single latent trait. This perspective aligns with the fear-avoidance model, suggesting that symptoms like catastrophic thoughts, fear, and avoidance directly influence one another.

The authors emphasize that while the Spanish CPASS is a psychometrically sound instrument, its scores require contextualized interpretation. The lack of strict invariance between clinical and community populations means that direct score comparisons between these groups are problematic and may necessitate population-specific norms.

Ultimately, the study argues that moving beyond default CFA approaches to include network modeling can reshape clinical conclusions, offering a more mechanistic understanding of how pain anxiety is maintained and suggesting that interventions targeting central and bridge symptoms may yield broader therapeutic effects. The findings support the use of the two-factor structure for general assessment while highlighting the utility of network analysis for identifying specific intervention targets.