Diagnostic Accuracy

Diagnostic Accuracy Analysis in R

Overview

Comprehensive diagnostic test accuracy analysis covering ROC curve analysis, optimal cutpoint determination, sensitivity and specificity estimation, likelihood ratios, decision curve analysis, inter-rater reliability measures, and diagnostic meta-analysis.

Basic Diagnostic Measures

2x2 Table Analysis

library(epiR)

# Create 2x2 table
# Format: [TP, FN; FP, TN]
diag_table  0.9)
pauc_spec  0.9)
pauc_sens = 0.9
cp_constrained 
  plot(type = "clinical_impact")

# Net benefit with confidence intervals (bootstrap)
dca_boot  Post-test probability:", round(post_prob_pos * 100, 1), "%\n")
  cat("LR-:", lr_neg, "-> Post-test probability:", round(post_prob_neg * 100, 1), "%\n")

  return(list(
    pre_test = pre_test_prob,
    post_test_positive = post_prob_pos,
    post_test_negative = post_prob_neg
  ))
}

# Example usage
fagan_nomogram(pre_test_prob = 0.20, lr_pos = 8, lr_neg = 0.15)

Reporting Diagnostic Study Results

# Create comprehensive diagnostic report
create_diagnostic_report = cutpoint, 1, 0)

  # Confusion matrix
  cm <- table(Predicted = pred_class, Actual = df$disease)
  TP <- cm[2, 2]; FN <- cm[1, 2]
  FP <- cm[2, 1]; TN <- cm[1, 1]

  # Calculate metrics
  metrics <- data.frame(
    Metric = c("AUC", "Cutpoint", "Sensitivity", "Specificity",
               "PPV", "NPV", "LR+", "LR-", "Accuracy", "Youden Index"),
    Value = c(
      round(auc(roc_obj), 3),
      round(cutpoint, 2),
      round(TP / (TP + FN), 3),
      round(TN / (TN + FP), 3),
      round(TP / (TP + FP), 3),
      round(TN / (TN + FN), 3),
      round((TP / (TP + FN)) / (FP / (FP + TN)), 2),
      round((FN / (TP + FN)) / (TN / (FP + TN)), 2),
      round((TP + TN) / (TP + TN + FP + FN), 3),
      round(TP / (TP + FN) + TN / (TN + FP) - 1, 3)
    )
  )

  return(metrics)
}

Key Packages Summary

| Package | Purpose | |---------|---------| | pROC | ROC curve analysis and AUC | | cutpointr | Optimal cutpoint selection | | OptimalCutpoints | Multiple cutpoint methods | | dcurves | Decision curve analysis | | irr | Inter-rater reliability (kappa, ICC) | | mada | Diagnostic meta-analysis | | BlandAltmanLeh | Method agreement plots | | epiR | Diagnostic test evaluation | | DiagTest3Grp | Three-group ROC analysis | | caret | Confusion matrix utilities |

Best Practices

1. Report multiple metrics: Sensitivity, specificity, PPV, NPV, and likelihood ratios
2. Account for prevalence: PPV/NPV depend heavily on disease prevalence
3. Use appropriate cutpoint method: Consider clinical consequences (cost of FN vs FP)
4. Provide confidence intervals: Especially for AUC and diagnostic measures
5. Check calibration: Predicted probabilities should match observed frequencies
6. Decision curve analysis: Evaluates clinical utility across threshold range
7. Consider spectrum bias: Ensure representative disease severity range
8. Report according to STARD: Standards for Reporting of Diagnostic Accuracy Studies

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: choxos
• Source: choxos/BiostatAgent
• License: MIT

Install and usage instructions live in the source repository linked above.