{
  "abstract": "Background The German National Cohort (NAKO) offers a large population sample for estimating Generalized anxiety disorder (GAD) prevalence; however, the accuracy of the prevalence estimate based on the GAD-7 screening tool is constrained by its sensitivity and specificity. We applied a Bayesian framework to estimate the GAD prevalence accounting for the screening tool’s accuracy and compared the results to naive prevalence estimates based on a conventional GAD-7 cutoff.Methods We analyzed data from the NAKO cohort, a population-based study of 204,697 adults aged 19–74 years across 18 study centers from 16 regions throughout Germany recruited between 2014 and 2019. The participants’ anxiety symptoms were assessed using the GAD-7 screening tool. Bayesian modeling incorporated prior distributions for GAD-7 test accuracy, with sensitivity (0.74, 95% CI: 0.61–0.84) and specificity (0.83, 95% CI: 0.68–0.92) obtained from a previous validation study. These Bayesian adjusted estimates were then compared with naive prevalence estimates based on the conventional GAD-7 cutoff of ≥10. Design and calibration weights, specific to each of the 16 study regions, were applied to enhance regional representativeness.Results Of 204,697 NAKO participants, 15,789 (7.7%) were excluded due to missing GAD-7 responses. Bayesian modeling estimated a GAD prevalence of 1.2% (95% credible interval (CrI): 0.0–3.7%), significantly lower than the naive estimate of 5.2% (95% CI: 5.1–5.3%). Prevalence was higher among females (1.6% vs. 0.9% in males), non-native German speakers (2.2% vs. 1.1% in native speakers), and those with a history of anxiety (11.7% vs. 0.8% in those without). The prevalence of GAD decreases with age, with the highest estimates in young adults (19–29 years: 1.7%, 95% CrI: 0.1%–4.9%) and the lowest in older adults (70–79 years: 0.5%, 95% CrI: 0.0%–1.5%). Bayesian estimates were consistently lower than naive estimates across study centers (0.8–1.6% vs. 3.5–6.5%).Conclusion Estimating prevalence using screening scores without accounting for sensitivity and specificity may result in an overestimation, potentially leading to a misallocation of limited resources. Bayesian methods offer more robust estimates. Our findings underscore the need to account for diagnostic uncertainty when estimating prevalence and highlight the value of Bayesian methods in improving mental health surveillance and ultimately informing targeted interventions.",
  "authors": [
    {
      "affiliations": [
        "Department of Health Care Management, Technische Universitt Berlin, Berlin, Germany"
      ],
      "name": "Dinara Yessimova"
    },
    {
      "affiliations": [
        "Institute of Public Health, Charit- Universitts Medizin Berlin, Berlin, Germany"
      ],
      "name": "Chisato Ito"
    },
    {
      "affiliations": [
        "Institute of Epidemiology and Social Medicine, University of Mnster, Mnster, Germany"
      ],
      "name": "Klaus Berger"
    },
    {
      "affiliations": [
        "Institute of Medical Biometry, Informatics and Epidemiology, University of Bonn, Bonn, Germany"
      ],
      "name": "Dario Zocholl"
    },
    {
      "affiliations": [
        "Institute of Public Health, Charit- Universitts Medizin Berlin, Berlin, Germany"
      ],
      "name": "Tobias Kurth"
    }
  ],
  "title": "OP33 Regional prevalence of generalized anxiety disorder in Germany: insights from the NAKO study accounting for GAD-7 diagnostic inaccuracies",
  "uid": "d21113a1-d27f-5028-8494-f6ef365c3375"
}