Epidemiology and Association Rules Analysis for Pulmonary Tuberculosis Cases with Extrapulmonary Tuberculosis from Age and Gender Perspective: A Large-Scale Retrospective Multicenter Observational Study in China

Wanli Kang, Jiajia Yu, Liang Chen, Quanhong Wang, L. Li, Jian Du, Hongyan Chen, Jianxiong Liu, Jinshan Ma, Mingwu Li, Jingmin Qin, Wei Shu, Peilan Zong, Yi Zhang, Xiaofeng Yan, Zhiyi Yang, Zaoxian Mei, Qunyi Deng, Pu Wang, Wenge Han, Meiying Wu, Ling Chen, Xinguo Zhao, Lei Tan, Fujian Li, Chao Zheng, Hongwei Liu, Xinjie Li, A Ertai, Yingrong Du, Fenglin Liu, Wenyu Cui, Song‐I Yang, Xiaohong Chen, Junfeng Han, Qingyao Xie, Yanmei Feng, Wenyu Liu, Peijun Tang, Jianyong Zhang, Jian Zheng, Dawei Chen, Xiangyang Yao, Tong Ren, Yan Li, Yuanyuan Li, Lei Wu, Qiang Song, Mei Yang, Yuanyuan Liu, Shuliang Guo, Kun Yan, Xinghua Shen, Dan Lei, Yanli Zhang, Youcai Li, Yongkang Dong, Shenjie TangInternational Journal of Clinical Practice 2023 · DOI

TL;DR

This study leverages association rule mining on a massive dataset of over 438,000 tuberculosis (TB) patients in China to uncover complex relationships between pulmonary TB (PTB) and various forms of extrapulmonary TB (EPTB). The findings reveal that concurrent PTB-EPTB is common and its patterns vary significantly by age and gender, suggesting a need for tailored screening and treatment strategies.

What Problem Does This Paper Address?

Tuberculosis (TB) is a major global health challenge, and while it most commonly affects the lungs (PTB), it can also manifest in other parts of the body as extrapulmonary TB (EPTB). Clinicians often encounter patients with both PTB and EPTB simultaneously, but understanding the epidemiology and specific associations between these co-occurring conditions, especially across different demographics, has been limited. This lack of detailed knowledge can complicate diagnosis and treatment. This paper aims to fill this gap by systematically analyzing a large cohort to identify common concurrent PTB-EPTB presentations and the rules governing their co-occurrence, particularly considering age and gender.

Methods

The researchers employed a retrospective, multicenter observational study design, collecting data from 21 hospitals across 15 Chinese provinces. The core analytical technique used was association rule mining, specifically the Apriori algorithm. This algorithm is adept at discovering relationships between items in large datasets. In this context, "items" are different types of TB diagnoses (PTB and various EPTB forms).

The study defined key metrics for association rules:

  • Support: The frequency of a particular TB combination (e.g., PTB and tuberculous pleurisy) within the entire dataset.
  • Confidence: The probability that if a patient has one type of TB (e.g., PTB), they also have another (e.g., tuberculous pleurisy).
  • Lift: A measure of how much more likely two TB types are to occur together than would be expected by chance. A lift value greater than 1 indicates a positive association.

In addition to association rule mining, multivariable logistic regression was used to examine the associations of gender and age group with concurrent PTB-EPTB.

Datasets

The study utilized a large dataset of 438,979 TB inpatients diagnosed between January 2011 and December 2017 from 21 hospitals in 15 provinces in China. This dataset represents confirmed TB diagnoses, encompassing both bacteriologically confirmed and clinically diagnosed cases. No new benchmarks were introduced; the focus was on analyzing existing clinical data.

Key Insights and Ideas

  • Concurrent PTB-EPTB is prevalent: Nearly 30% of TB patients in this cohort had concurrent PTB and EPTB, highlighting that this is not an uncommon scenario.
  • Tuberculous pleurisy is the most frequent EPTB co-occurrence: This finding is consistent across the study, suggesting it's a primary concern when PTB is present.
  • Age and Gender Matter Significantly: The likelihood of having concurrent PTB-EPTB is not uniform. Females and younger age groups (<15 to 35-44 years) showed a higher propensity for concurrent PTB-EPTB compared to older age groups.
  • Specific EPTB types have strong associations: The study identified distinct EPTB forms that are strongly linked to PTB, and these associations differ depending on whether PTB is considered the antecedent or consequent.
  • Tailored screening is crucial: The varying associations by age and gender underscore the need for clinicians to consider specific EPTB screenings based on patient demographics and initial PTB diagnosis.

Key Results

  • Overall PTB-EPTB prevalence: 129,422 out of 438,979 patients (29.48%) had concurrent PTB-EPTB.
  • Most common concurrent EPTB type: Tuberculous pleurisy was the most common concurrent EPTB observed.
  • Strongest association rule (PTB as antecedent): PTB with concurrent bronchial tuberculosis showed a lift of 1.09, indicating a positive association.
  • Strongest association rule (EPTB as antecedent): Pharyngeal/laryngeal tuberculosis with concurrent PTB had a lift of 1.11.
  • Gender differences: Females were more likely to have concurrent PTB-EPTB (OR = 1.119). Specific EPTB associations also differed: tuberculous empyema was strongest in males (lift = 1.20), while bronchial tuberculosis was strongest in females (lift = 1.64).
  • Age group variations: Tuberculous meningitis had the strongest association with PTB in children (<15 years, lift = 3.89). Other age groups showed varying strongest associations, such as splenic tuberculosis in 15-24 year olds (lift = 2.23) and oviduct tuberculosis in 25-34 year olds (lift = 3.09).

Limitations and Future Work

The authors acknowledge several limitations:

  • Berkson's Bias: The study population consisted of hospitalized TB patients, which may overestimate the prevalence of concurrent PTB-EPTB as these cases might be more likely to be hospitalized.
  • Hospital Specialization: Most participating hospitals were TB-specialized, potentially limiting the generalizability of findings to the broader TB patient population.
  • Pediatric TB: Some specialized hospitals may not admit pediatric TB patients, affecting the representation of this age group.
  • Diagnostic Information: The study did not collect detailed information on the number of bacteriologically versus clinically diagnosed TB cases.
  • Unaccounted Factors: The analysis did not consider low-frequency disease complications, comorbidities like HIV, or factors such as income and smoking status.

Future work could involve population-based studies to mitigate bias, broader inclusion of different hospital types, and more detailed diagnostic stratification. The authors emphasize the need for customized treatment regimens for concurrent PTB-EPTB patients.

Summary updated

Abstract

Numerous concurrent PTB-EPTB case types were observed, with confidence and lift values varying with gender and age. Clinicians should screen for concurrent PTB-EPTB in order to improve treatment outcomes.
BibTeX
@article{kang2023epidemiology,
  author = {Wanli Kang and Jiajia Yu and Liang Chen and Quanhong Wang and L. Li and Jian Du and Hongyan Chen and Jianxiong Liu and Jinshan Ma and Mingwu Li and Jingmin Qin and Wei Shu and Peilan Zong and Yi Zhang and Xiaofeng Yan and Zhiyi Yang and Zaoxian Mei and Qunyi Deng and Pu Wang and Wenge Han and Meiying Wu and Ling Chen and Xinguo Zhao and Lei Tan and Fujian Li and Chao Zheng and Hongwei Liu and Xinjie Li and A Ertai and Yingrong Du and Fenglin Liu and Wenyu Cui and Song‐I Yang and Xiaohong Chen and Junfeng Han and Qingyao Xie and Yanmei Feng and Wenyu Liu and Peijun Tang and Jianyong Zhang and Jian Zheng and Dawei Chen and Xiangyang Yao and Tong Ren and Yan Li and Yuanyuan Li and Lei Wu and Qiang Song and Mei Yang and Jian Zhang and Yuanyuan Liu and Shuliang Guo and Kun Yan and Xinghua Shen and Dan Lei and Yanli Zhang and Youcai Li and Yongkang Dong and Shenjie Tang and Youcai Li and Yongkang Dong and Shenjie Tang},
  title = {{Epidemiology and Association Rules Analysis for Pulmonary Tuberculosis Cases with Extrapulmonary Tuberculosis from Age and Gender Perspective: A Large-Scale Retrospective Multicenter Observational Study in China}},
  journal = {International Journal of Clinical Practice},
  year = {2023},
  doi = {10.1155/2023/5562495},
}
Figure