Evaluating the Agreement between k-median and Latent Class Analysis for Clustering of Psychological Distress Prevalence

Maryam Salari; Zahra Rahimi; Reza Kalantari; Jamshid Jamali

doi:10.18502/jbe.v8i4.13353

Maryam Salari Department of Biostatistics, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
Zahra Rahimi Department of Biostatistics, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
Reza Kalantari Department of Ergonomics, School of Health, Shiraz University of Medical Sciences, Shiraz, Iran.
Jamshid Jamali Department of Biostatistics, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.

DOI: https://doi.org/10.18502/jbe.v8i4.13353

Keywords: Cross-sectional study; Psychological distress; K-6 questionnaire; Latent class analysis; K- media; Mashhad;

Abstract

Introduction: Psychological distress (PD) is one of the most common mental disorders in the general population. Psychological distress is considered a public health priority due to its adverse effects on quality of life, health, performance, and productivity. It can also predict several serious mental illnesses, such as depressive disorder and anxiety. In this study, we intend to identify the behavioral pattern of PD in the population of 18 to 65 years old in Mashhad using two methods, K-median and Latent Class Analysis (LCA), and evaluate the agreement between the two methods.

Methods: This cross-sectional study was performed on 38058 individuals referred to community health care centers in Mashhad of Iran in 2019. The information used in this study was extracted from Sina Electronic Health Record System (SinaEHR) database. A demographic information checklist and a 6-item Kessler psychological distress scale (K-6) were used for data collection. K-median and LCA were used for data analysis.

Results: Out of 38058 participants, 49.3% were women, 86.1% were married, and 63.6% had a diploma and under diploma education. The LCA identified three patterns of PD in answering the items of the K-6 questionnaire, including severe PD (19.7%), low PD (36.7%), and no PD (43.5%). Three clusters were identified by the K-Median method: 1) severe PD (22.0%), 2) low PD (31.1%), and 3) and no PD (46.9%). The agreement between K-Median and LCA was kappa = 0.862.

Conclusion: About 20% of people were classified as having severe PD. Both LCA and k-median methods can reasonably identify the latent pattern of PD with significant entropy, and there was almost complete agreement between the two methods in data clustering. Considering the advantages of the LCA, this method is recommended to identify the latent pattern of PD based on the k-6 questionnaire.