# Heterogeneity in longitudinal medication adherence patterns among patients with spinal tuberculosis: a latent class analysis using multi-timepoint follow-up data and associations with clinical outcomes

## TL;DR

This study identifies different medication adherence patterns in spinal tuberculosis patients after discharge and links these patterns to factors like age, disability, and supervision.

## Contribution

The study introduces a novel use of group-based trajectory modeling to identify distinct medication adherence patterns and their clinical associations in spinal tuberculosis patients.

## Key findings

- Three medication adherence trajectories were identified: persistently high adherence, fluctuating decline, and early low adherence.
- Older age, higher disability scores, and adverse drug reactions were associated with non-adherent medication patterns.
- Tailored interventions are recommended for different adherence subgroups to improve clinical outcomes.

## Abstract

This study aims to explore the dynamic trajectory of medication adherence behavior and its influencing factors among patients with spinal tuberculosis within 6 months after discharge, providing a basis for developing individualized intervention strategies.

A retrospective analysis was conducted using data from a longitudinal follow-up cohort, enrolling 117 spinal tuberculosis patients who underwent surgical treatment at the General Hospital of Ningxia Medical University between January 2024 and August 2025. Data were collected via telephone follow-up at 1 week, 1 month, 3 months, and 6 months after discharge, including general demographic information, scores from the Morisky Medication Adherence Questionnaire, Oswestry Disability Index (ODI), and laboratory indicators such as erythrocyte sedimentation rate and C-reactive protein levels. Group-based trajectory modeling (GBTM) was employed to identify subgroups of medication adherence trajectories, and multivariate logistic regression analysis was conducted to examine the influencing factors of these trajectory patterns.

Among 117 patients, group-based trajectory modeling identified three medication adherence trajectories: “persistently high adherence” (35.0%, n = 41), “fluctuating decline” (48.7%, n = 57), and “early low adherence” (16.2%, n = 19). The early low adherence group was older (64.21 ± 5.92 years) and had higher baseline ODI (62.11 ± 5.15) and CRP (39.58 ± 12.40 mg/L). Factors associated with non-adherence in the fluctuating vs. high adherence group included higher ODI z-score (OR = 4.658, 95% CI: 3.851–5.581), adverse drug reactions (OR = 3.249, 95% CI: 2.664–3.868), older age z-score (OR = 1.974, 95% CI: 1.524–2.516), and higher log(CRP+1) (OR = 1.532, 95% CI: 1.314–1.776). Protective factors were family supervision (OR = 0.779, 95% CI: 0.633–0.943), higher education (OR = 0.576, 95% CI: 0.498–0.660), and higher monthly income z-score (OR = 0.471, 95% CI: 0.373–0.583). All analyses were based on 1,000 bootstrap samples with L2 regularization.

Medication adherence behavior among spinal tuberculosis patients exhibits heterogeneous dynamic trajectories. Clinically, tailored interventions should be developed according to the characteristics and influencing factors of different trajectory subgroups, with particular attention to older patients, those with lower educational levels, those experiencing adverse drug reactions, and those lacking supervision, in order to improve medication adherence.