Introduction

Tuberculosis (TB) leads to a huge illness and mortality globally, particularly in low- and middle-income countries1,2. Despite the recent decline in PTB notification rate due to the implementation of the End TB strategy, it remains a global public health challenge. The advent of the COVID-19 pandemic has further complicated matters, exerting a profound impact on the diagnosis and treatment of PTB and leading to adverse outcomes 3,4. The effective identification and prompt diagnosis of PTB cases have become critical nodes eliminating PTB. Reducing patient and healthcare system delays can not only improve prognosis but also decrease the risk of community transmission, thereby lowering the incidence rate in the general population5,6. Despite improvements in surveillance systems and PTB control strategies, delays persist in the overall process, even as the number of actively diagnosed cases continues to rise7.

Zhejiang Province, a coastal province of eastern China, grapples with similar challenges8. Notably, spatial–temporal analysis has emerged as a valuable tool for identifying clustering patterns of infectious diseases within specific populations9. However, the relationship between spatial distribution delay rates and key public health indicators, such as the disparities in PTB awareness, remains unexplored. Additionally, insufficient awareness has been identified as an important factor elevating the risk of PTB exposure. Particularly, a lack of knowledge regarding symptoms, transmission routes, and curable outcomes not only steers PTB cases toward traditional self-medication methods but also contributes to delays in case identification, emphasizing the need for conclusive evidence10.

This study aimed to elucidate the general epidemiological features of PTB and identify potential influencing factors contributing to two common types of delays: patient delay and diagnosis delay among PTB cases. Additionally, the correlation between these delays and public awareness was explored.

Methods

Location and data collection

In this study, Zhejiang province was chosen as the study site, and detailed information regarding this location has been previously outlined in the literature11. Zhejiang contains 11 cities as well as 90 counties and districts, the exact distribution is shown in Fig. S1-1. From 2017 to 2022, all variables related to notified PTB were extracted from the Tuberculosis Information Management System (TBIMS). Subsequently, information regarding various delays was calculated after excluding records with illogical errors (e.g., negative numbers). These variables included demographic information and specific PTB treatment details. Additionally, two rounds of PTB public awareness surveys were conducted in 2018 and 2022, spanning 10 counties or districts across 11 cities in Zhejiang Province. The surveys gauged awareness levels in three dimensions: suspicious symptoms, transmission routes, and curability, and more details were shown in our previous research12,13.

Definitions

Notified PTB cases included both clinically diagnosed and laboratory-confirmed cases, with diagnosis criteria referenced from the National Diagnostic Criteria for Pulmonary Tuberculosis (WS288-2017) and the Classification of Tuberculosis (WS196-2017) in China14,15. The diagnostic classification for PTB had been proposed in our previous study11. In this study, long-term patient delay (LPD) referred to the time interval between symptom onset and the first healthcare-seeking activity was equal to or exceeding its median time (≥ 15 days). Similarly, long-term diagnosis delay (LDD) was defined as the interval between the first healthcare-seeking and confirmed diagnosis equal to or exceeding the median time (≥ 2 days)16. Active case findings referred to those identification through mass screening and physical examination, while passive case findings included individuals identified through direct visits to designated healthcare facilities due to obvious PTB symptoms or those referred to designated hospitals by subordinate hospitals or non-designated facilities17.

Epidemiological characteristics of PTB delay

Patient delay and diagnostic delay in notified PTB cases were described in terms of sex, age group, occupation, cases finding, treatment history, pathogenetic results, and anti-TB treatment using median and Interquartile Range (IQR). A multivariate logistic regression was used to explore the potential risk factors for LPD and LDD. After detecting no collinearity (Table S2-1), we conducted the analysis, including all available potential factors in the model without prior variable selection.

Spatial analysis and spatial–temporal analysis for delay

We employed spatial autocorrelation analysis to verify the correlation of LPD and LDD rates at the spatial level. Additionally, spatial–temporal scan statistics were used to identify clusters in both dimensions. The former includes global spatial autocorrelation and local spatial autocorrelation, which represent the spatial distribution patterns for the entire region and its subregions18. Global spatial autocorrelation was assessed using Moran's I, which takes values ranging from −1 to 1. The closer the Moran's I value is to 1, the stronger the positive correlation; the closer to −1, the stronger the negative correlation; and 0 indicates no spatial correlation. P-values ≤ 0.05 are considered statistically significant11. Additionally, the analysis of aggregation characteristics in both time and space was conducted using Kulldorff spatial–temporal scan statistics19. The method defines a two-dimensional round foundation as a cylindrical window, where the radius represents the risk population scanned, and the height represents time. The null hypothesis is that the relative risk (RR) within the window equals the outside’s RR. When RR is greater than 1, those within the window are more likely to have LPD and LDD. The spatial–temporal scan window with the log likelihood ratio (LLR) is termed the most likely cluster. Moreover, we tested the significance of the identified clusters using Monte Carlo simulations at the 95% confidence level20.

Consistency analysis of PTB delay and public awareness

We utilized the Mantel test to evaluate the correlation between public awareness obtained from the surveys and PTB delay from the surveillance system. The Mantel correlation provides an overview of the relationship between two matrices21. The standardized equation of the Mantel’s test (\({Z}_{N}\)) is given by:

$${Z}_{N}=\frac{\sum_{i=1}^{n}\sum_{j=1}^{n}\left({a}_{ij}-\overline{A }\right)\times ({b}_{ij}\times \overline{B })}{{var(A)}^{1/2}\times {var(B)}^{1/2}}$$
(1)

The \({Z}_{N}\) represents the correlation value Mantel’s r between the standardized elements of matrices \(A\) and \(B\). \({a}_{ij}\) and \({b}_{ij}\) represent variables in row i and column j of matrices \(A\) and \(B\), respectively. \(\overline{A }\) and \(\overline{B }\) indicate the means of matrices \(A\) and \(B\), while \(var(A)\) and \(var(B)\) indicate the variances of matrices \(A\) and \(B\) 22. Based on the non-ecological distance matrix, the values of \({Z}_{N}\) range from −1 to 1. \({Z}_{N}\) > 0, < 0 and = 0 indicate the presence of positive, negative and no correlation, respectively. At a P-value of < 0.05, the greater the \({Z}_{N}\), the stronger the correlation 21.

Statistical analysis

A descriptive analysis of general epidemiological features, along with the association analysis of delay and public awareness, was performed using R software (version 4.3.0, R Core Team, Vienna, Austria). The results of spatial and spatial–temporal analysis were calculated using Geoda (version 1.20) and SaTScan (version 10.1.1, Boston, MA, USA), and maps were visualized with ArcGIS software (version 10.8, SERI Inc.; Redlands, CA, USA).

Ethics approval and consent to participate

This study consisted of two parts, both of which were approved by the ethics committee of the Zhejiang Provincial Center for Disease Control and Prevention (ZJCDC). The first part involved a delay analysis of elderly PTB based on surveillance system data. As it was solely surveillance data, informed consent was exempted by the Ethics Committee of ZJCDC. The second part comprised two rounds of public awareness surveys for PTB conducted in 2018 and 2022 obtaining ethical approval (No. 2018-035), and all participants provided signed informed consent before the research. This study statement confirms that all methods were performed in accordance with relevant guidelines and regulations in the declaration.

Results

General epidemiological feature and multi-factor analysis of patient delay and diagnosis delay

A total of 151,799 PTB cases were notified in Zhejiang Province between 2017 and 2022. The median and IQR for patient delay and diagnostic delay were 15 (4–35) and 2 (0–8) days, respectively. Notably, we observed a total of 76,407 LPDs with a patient delay of ≥ 15 days, and 77,077 LDDs with a diagnostic delay of ≥ 2 days. The dataset comprised 104,375 male and 47,424 female patients, with median patient delays of 14 (IQR 4–35) and 15 (IQR 4–35) days, respectively. The age distribution revealed median patient delays of 12 (IQR 3–35), 14 (IQR 3–35), and 17 (IQR 6–35) days for the age groups 0–14, 15–64, and 65 and above, respectively. Among different occupations, agricultural workers, retirees, and industrial workers had the highest LPD rates. In terms of methods of case finding, there was no significant difference in the median patient delay and diagnostic delay between passive and active case finding. Regarding treatment history, the median patient delays were 15 (IQR 4–35) and 16 (IQR 4–35) days for initial treatment and retreatment, and the median diagnostic delays were 2 (IQR 0–8) and 1 (IQR 0–8) days, respectively. Multi-factor logistic regression analysis showed that females, agricultural workers, retirees, those identified through passive case finding, retreatment patients, and those with no bacteriological results were related to LPD rate (P < 0.01). Conversely, commercial service workers, those identified through active case finding, individuals with initial treatment, and those with positive bacteriological results were associated with LDD rate (P < 0.01). More details are shown in Fig. 1.

Figure 1
figure 1

Epidemiological characteristics of patient delays and diagnosis delays in Zhejiang Province, 2017–2022. OR odds ratio, CI confidence interval, IQR interquartile range, LPD long-term patient delay, LDD long-term diagnostic delay. The boxes represent odds ratios, and the error bars represent 95% confidence intervals.

Full size image

Spatial–temporal analysis

The spatial autocorrelation analysis and spatial–temporal scan statistics were conducted to evaluate the rates of LPD and LDD in Zhejiang Province. The results of LPD rate indicated spatial correlation from 2019 to 2022, as evidenced by the global Moran’s I statistic ranging from 0.120 to 0.223 (P < 0.05; Table 1; Figure S1-2). Furthermore, one most likely cluster and three secondary clusters were identified at the spatial–temporal level (Table S2-2). The most likely cluster was mainly distributed in the southwestern region of Zhejiang, including some areas of Jinhua, Quzhou, Lishui, and Wenzhou cities (Fig. 2A). The spatial–temporal clustering of LPD mainly occurred before the onset of the COVID-19 pandemic, with the most likely clusters dating from January 2017 to January 2020.

Table 1 Results of spatial autocorrelation of long-term delay of PTB among Zhejiang Province from 2017 to 2022.
Full size table
Figure 2
figure 2

Spatial–temporal clusters of long-term delay of PTB among Zhejiang province from 2017 to 2022. “a” was long-term patient delay and “b” was long-term diagnostic delay.

Full size image

In terms of LDD rate, the global Moran's Index statistic ranged from 0.117 to 0.165, indicating spatial correlation from 2018 to 2022 (P < 0.05; Table 1; Figure S1-3). Furthermore, one most likely cluster and four secondary clusters were identified at the spatial–temporal level (Table S2-3). The most likely cluster was mainly located in the northeastern area of Zhejiang Province from September 2018 to August 2021, including regions in Ningbo and Shaoxing City (Fig. 2B). Additional details about other secondary clusters are shown in Fig. 2.

Correlation analysis of long-term delay and public awareness

The public awareness levels of PTB symptoms, transmission routes, and knowledge of curability demonstrated an increasing trend, and there was a strong positive correlation between the three items (Pearson correlation coefficients r > 0.6; Table 2; Fig. 3). Furthermore, results indicated a moderately positive association between LPD and the awareness level of PTB suspicious symptoms in 2018 (Mantel’s r = 0.4, P = 0.04). However, no significant association was observed in LDD. More details are shown in Table S2-4.

Table 2 Delays and public awareness rates in 11 counties of Zhejiang Province in 2018 and 2022.
Full size table
Figure 3
figure 3

Correlation analysis between long-term delay and public awareness among Zhejiang Province in 2018 and 2022. “a” was the years of 2018 and “b” was 2022.

Full size image

Discussion

This study aimed to investigate patient delay, diagnosis delay, and their spatial–temporal variations in Zhejiang Province, providing critical insights for effective PTB management. Patient delay and diagnosis delay pose significant epidemiological challenges for PTB prevention and control. Specifically, patient delay contributes to community transmission, exacerbation of the patient’s symptoms, damage to parenchymal tissue, and can lead to adverse outcomes, including death. Conversely, diagnosis delay underscores weaknesses and deficiencies in the health system's diagnostic processes for PTB. To our knowledge, this study is the first to examine variability at the spatial–temporal level concerning both PTB delay rates and explore the correlation between long-term delay and various aspects of public awareness. This finding provides new perspectives on TB delay research.

Patient delay accounted for the largest portion of total delay 17. The median patient delay in Zhejiang province was 15 days, shorter than the national average of 20 days 17 and that of neighboring countries such as South Korea, Thailand, and Cambodia 23,24. This could be attributed to a series of individual public health measures implemented in Zhejiang Province, including robust public health promotion and funding for PTB diagnosis. However, a review by Alvin Kuo Jing Teo et al. showed that the median patient delay in upper-middle-income countries was only 10 days 2. Despite having fewer female than male patients with PTB, women experienced longer patient delays and a high possibility with LPD. This discrepancy may be due to the differences, both physiological and psychological, resulting in delayed healthcare seeking among females who may have milder symptoms than their male counterparts6. Some delays may also be attributed to resource constraints, power imbalances and stigma faced by women2. Occupationally, patient delays are more severe among agricultural workers and retirees. As a low-income group, farmers bear high economic burden, daily living, and household expenses are prioritized, which may result in them being less likely to seek PTB treatment than other high-income groups2,7. Previous studies have also found a threefold increase in patient delay for patients living in rural areas compared to urban areas25. In addition, retired individuals, commonly older people, may have difficulty in seeking timely healthcare due to limited mobility, dependence on others, and social discrimination. PTB in older adults often presents with atypical symptoms such as anorexia, chronic fatigue, and unexplained low-grade fever rather than conventional symptoms such as cough, hemoptysis, fever, and night sweats25. Therefore, large-scale active screening should prioritize women and older retirees in rural areas when resources permit. Furthermore, we observed that patient delays were more severe in retreatment patients compared to initial treatment patients. This phenomenon can be attributed to poor adherence and multiple risk factors for PTB in this specific group, such as malnutrition, low body weight, smoking, and human immunodeficiency virus26,27. Therefore, efforts in reducing patient delays in retreatment patients should focus on improving adherence and personalized support such as nutrition for all populations during treatment. Additionally, enhanced patient follow-up and examinations after treatment interruptions and throughout the course of treatment should be conducted.

This study revealed that the median diagnosis delay in Zhejiang Province was 2 (IQR 0–8) days, slightly higher than the national delay of 1 (IQR 0–8) day, which indicates that the rational allocation of medical resources such as laboratories, testing technology and personnel within the health system of Zhejiang still should be improved during the study period. Previous studies have indicated that patients with PTB visit healthcare facilities an average of four times before initiating treatment28, and nearly half of these initial healthcare facilities lack the capacity for PTP diagnosis and treatment29. To address ongoing diagnosis delay caused by variations in the capabilities of medical institutions across Zhejiang Province, it is recommended to enhance the awareness and chest radiograph interpretation skills of respiratory doctors in non-designated hospitals for PTB. Additionally, timely referrals and recommendations for suspected patients should be prioritized, particularly for those whose initial healthcare visit is in a designated hospital. Recommendations include optimizing referral tracking within the hospital and actively incorporating new diagnostic technologies, such as replacing solid and liquid cultures with more efficient alternatives, to reduce LDD.

Spatial analysis results indicate a correlation in the rates of both LPD and LDD at the county level. Further spatial–temporal scanning statistics revealed that the most likely LPD risk cluster occurred in parts of southwestern Zhejiang between February 2017 and January 2020. Previous studies have identified western Zhejiang Province as a high-risk area for PTB, with higher morbidity risk and lower levels of economic development and education, potentially exacerbating PTB delays30. Tailored comprehensive interventions, such as health education and active detection, are recommended for this region28. The disappearance of this LPD cluster after the COVID-19 pandemic may be attributed to increased attention to respiratory diseases during that period, somewhat shortening delay times. Moreover, recent mass screening initiatives targeting the older population, patients with diabetes, and individuals with previous PTB in western Zhejiang have contributed to a reduction in detection time. In addition, Zhejiang Province has good economic and medical conditions nationwide; however, our spatial–temporal analysis revealed that LDD was more pronounced in the northeastern region, especially in Ningbo city, from 2018 to 2021. This may be attributed to the fact that Zhejiang Province has been implementing the China-Bill Melinda Gate Phase III project on a global scale since 2017 and has begun to progressively promote the rapid molecular diagnosis technology such as Xpert-MTB/RIF (except in Ningbo)31. Xpert-MTB/RIF can detect Mycobacterium tuberculosis complex and rifampicin resistance in less than 2 h, which greatly shortens the time compared to traditional solid and liquid culture methods for Mycobacteria, and has the advantages of being fast and efficient32. Due to financial and other constraints, Ningbo only achieved full implementation of this diagnostic method in PTB experiments at the county level until 202033, resulting in the clustering of LDD in Ningbo from 2018 to 2021. This scenario helps illustrate the role of new technologies in shortening the diagnosis time of PTB.

In this study, we also explored the ecological aspects of PTB delays and public awareness levels. Notably, the results revealed a moderately positive correlation between public awareness level of symptoms and LPD, which was inconsistent with our previous cognizations2, prompting a reconsideration of our health education for shortening LPD. To explore the possible reasons, further analysis demonstrated a higher positive correlation between people knowing that the PTB was curable and those knowing the PTB symptom information. Thus, we postulated that the recognition of the curability of PTB weakens their urgency to seek medical care. In future health education of PTB, it suggested that we need to strengthen the public health attributes of this specific disease, including its harm to the community and household, and actively promote the shift of knowledge known to attitudinal and behavioral change in the public health education as well.

This study has certain limitations. Given that we incorporated administrative regions that underwent changes during the study period, this approach may overlook spatial–temporal correlation within the combined regions. In addition, as with other national surveillance systems, older individuals who face difficulties in accessing medical care may go unreported, potentially leading to underestimates of morbidity and delays.

Conclusion

The spatial distribution pattern of patient delays and diagnosis delays provided a new insight to identify our existing short slab in both health education and the available healthcare system. Comprehensive interventions and implementations for PTB accordingly should be deployed based on evidence-based findings, such as mass active screening and allocation of optimal resources to health facilities. Besides, gaping patient delays should be offset by strengthening the shift from knowledge known to real attitudes and behaviors in the health education field.