Using Latent Class Analysis to Identify Hidden Clinical Phenotypes

INTRODUCTION

This JAMA Guide to Statistics and Methods summarizes latent class analysis, a statistical technique that estimates the probability of patients belonging to a discrete group that shares specific combinations of observed variables, and explains how the technique is used and can be interpreted in observational research.

In precision medicine, a common question for researchers is whether patients can be classified with others who have similar risks and treatment responses. Such groupings can assist in predicting risk and matching patients with appropriate treatment strategies. The challenge is that it is often not easy to identify meaningful clusters of people with the observable data.

Latent class analysis (LCA) is a common explanatory modeling technique that allows researchers to identify groups of people who have similar characteristics that can include demographics, clinical characteristics, treatments, comorbidities, and outcomes.¹ The term latent derives from the fact that the classes are not directly observable. Latent class analysis estimates the probability of each participant being a member of each latent class.²

In a cohort study published in JAMA Cardiology, Patel et al³ used group-based trajectory modeling (GBTM), a type of LCA, and identified 5 distinct patterns of change in participant urine albumin-creatinine ratio (UACR) observed over 20 years. These 5 classes were independently associated with adverse changes in cardiac structure and ventricular function.³ Notably, participants belonging to the identified trajectory classes could not be distinguished by the baseline UACR alone, highlighting the value of this technique. This Guide to Statistics and Methods article describes LCA, its potential application, and limitations.

USE OF THE METHOD

What Is LCA?

Latent class analysis is a statistical technique that identifies groups defined by specific combinations of observed variables.² Latent class analysis assigns each participant a probability of being in each subgroup based on maximum likelihood estimation. Then, each participant is assigned to the group to which they have the highest probability of belonging. In GBTM, the trajectories’ shape can be a straight or curvilinear form; shapes are based on the maximum likelihood estimation. Selecting the number of groups requires manual reconciliation of the trajectories’ shape, the minimum number of participants assigned to a trajectory, and measures indicating how well the model fits, such as the Akaike information criterion or Bayes information criterion.⁴ Although there is no single criterion to select the number and shape of classes or trajectories, the number of classes that yield the best fit to the observed data, the highest average probability of group membership, and the fewest poor-fitting participants (ie, those with a highest probability of group membership <0.7) is chosen.⁵ Therefore, reporting the decisions and rationales behind this process of manual reconciliation is crucial.

Why Is LCA Used?

Latent class analysis is useful when the patterns that constitute distinct ...