Measuring Mental Health Quality of Life in R: A Psychometric Evaluation and Standardised MHQoL Toolbox

🧠 Why Standardised Scoring and Psychometric Evaluation Matter

• ✅ Validated health instruments (like MHQoL) are essential in health research
• ❌ Manual scoring is error-prone and inefficient
• Psychometric evaluation ensures:
  • Internal consistency
  • Validity (construct & convergent)
• ⚙️ R enables:
  • Automated scoring
  • Reproducible evaluation

📦 What this Project Delivers

• 📊 Psychometric analysis workflow in R
• Custom R package for:
  • MHQoL score recalculation
  • Consistent scoring logic
• 🌐 Interactive Shiny App for:
  • User-friendly input
  • Automated results & visualisation
• 🔁 Fully reproducible pipeline

⚙️ Recalculate Scores Using R Package or Shiny App

🧠 What is the MHQoL?

• A 7-item questionnaire for mental health-related quality of life

• Includes a Visual Analogue Scale (VAS) for overall mental well-being

💡 Why use it — especially next to the EQ-5D?

• The EQ-5D focuses primarily on physical health

• MHQoL captures additional aspects such as mood, relationships, and future outlook

• Offers a more complete picture of wellbeing in mental health populations

• Supports richer outcome measurement in evaluations and research

⚙️ Recalculate Scores Using the MHQoL Package

• 🗂️ Contains 8 function, highlight two:
  • mhqol()
  • mhqol_LSS()
• 📑 One dataframe:
  • mhqol_valueset()
• 🖥️ Shiny Tool is integrated in the package
  • shiny_mhqol()

⚙️

Which to use? 🤔

Both! 🥳

Example - Output in utilities?🤔

library(MHQoL)

# Prepare input: 7 MHQoL dimensions
dimensions <- example_data_scores[, c("SI", "IN", "MO", "RE", "DA", "PH", "FU")]

# Calculate total utility score (Netherlands value set)
mhqol_utility <- mhqol(
  dimensions = dimensions,
  country = "Netherlands",   # 🇳🇱 Default option
  metric = "total",          # "total" or "average"
  ignore_invalid = TRUE,     # Skip incomplete rows
  ignore_NA = FALSE          # Stop if NAs are present
)

Example - Output in scores?🤔

library(MHQoL)

# Prepare input: 7 MHQoL dimensions
dimensions <- example_data_text[, c("SI", "IN", "MO", "RE", "DA", "PH", "FU")] #Input from the text dataframe

# Calculate total utility score (Netherlands value set)
mhqol_LSS <- mhqol_LSS(
  dimensions = dimensions,
  metric = "total",          # "total" or "average"
  ignore_invalid = TRUE,     # Skip incomplete rows
  ignore_NA = FALSE          # Stop if NAs are present
)

Or if you want to use an Interface (from Shiny): the MHQoL cooker👩‍🍳

::: :::

Internal Consistency using the ltm package

library(ltm)

# Select MHQoL dimensions (7 items)
dimensions <- example_data_scores[, c("SI", "IN", "MO", "RE", "DA", "PH", "FU")]

# Compute Cronbach’s Alpha
cronbach.alpha(dimensions)

✅ Sow we Tested it in R (lavaan)

library(lavaan)

# Latent model: 1 factor with 7 observed MHQoL items
model <- '
  MHQoL =~ SI + IN + MO + RE + DA + PH + FU
'

# 1. Configural invariance
fit_config <- cfa(model, data = df, group = "group_var",
                  estimator = "WLSMV", ordered = TRUE)

# 2. Metric invariance (equal loadings)
fit_metric <- cfa(model, data = df, group = "group_var",
                  estimator = "WLSMV", ordered = TRUE,
                  group.equal = "loadings")

# 3. Scalar invariance (equal loadings + intercepts)
fit_scalar <- cfa(model, data = df, group = "group_var",
                  estimator = "WLSMV", ordered = TRUE,
                  group.equal = c("loadings", "intercepts"))

# Compare model fits
anova(fit_config, fit_metric, fit_scalar)

📊 Comparing Invariance Models

Partial scalar: free some intercepts, for example

fit_part_scalar <- cfa(model, data = df, group = "group_var",
                  estimator = "WLSMV", ordered = TRUE,
                  group.equal = c("loadings", C("SI|T1", "IN|T2")))

🔗 Spearman’s Rank Correlation Coefficients

# Pearson
cor.test(x = mhqol_score, y = eq5d_score, method = "pearson")

# Spearman
cor.test(x = mhqol_score, y = eq5d_score, method = "spearman")

📊 Example output correlations with MHQoL Score

How it was done in R

library(lme4)

# Random intercept model: MHQoL dimensions → MHQoL-VAS
model1 <- lmer(
  mhqol_vas ~ SI + IN + MO + RE + DA + PH + FU +
  age + gender + (1 | country),
  data = example_data_scores
)

# Random intercept + slope model (example: FU varies by country)
model2 <- lmer(
  mhqol_vas ~ SI + IN + MO + RE + DA + PH + FU +
  age + gender + (FU | country),
  data = example_data_scores
)

# Compare models
anova(model1, model2)

# Model performance
summary(model1)

✅ Key Takeaways

• 🧠 R provides a complete psychometric workflow, from internal consistency to measurement invariance and multilevel modeling

• 📦 We used specialized packages:

  • ltm for Cronbach’s Alpha
  • lavaan for Confirmatory Factor Analysis and Measurement Invariance
  • lme4 for Multilevel Regression

• 🧰 We developed the MHQoL R package and Shiny app:

  • Standardized scoring
  • Enhances reproducibility and ease of use

• 🎯 Together, these tools support both rigorous evaluation and practical implementation of mental health QoL measures

Any Questions?

The MHQoL R package and Shiny app streamline scoring, reduce manual errors, and make validated mental health quality of life measurement more accessible and reproducible.