Multilingual abstracts

Please see Additional file 1 for translations of the abstract into the five official working languages of the United Nations.

Background

Population aging is a critical public health concern in modern China. The proportion of elderly people aged 65 and above was 10.8%, and people aged 60 and above 16.7% of a population of 1382 million people in 2016 [1]. The risk of tuberculosis (TB) increases with age, as shown by studies internationally [2,3,4,5] and in China [6]. Almost half (48.8%) of the diagnosed TB cases in China’s Fifth National Prevalence Survey of TB were elderly people, among whom 39.8% were asymptomatic, and 53.2% didn’t seek any medical care [7]. Bele et al. found that population aging was one of the bottlenecks in TB control in rural China, and more TB control efforts are needed to reach the most vulnerable populations at high priority [8].

Rapid case identification and early treatment are the most important interventions to prevent TB transmission and reduce its incidence [9, 10]. Systematic screening of high-risk groups to diagnose TB early has proven effective to help end the global TB epidemic [11, 12]. However, it’s not feasible to screen all elderly people due to the huge population numbers in China. On the other hand, combining age with one or more TB risk factors could detect a higher TB prevalence in a more concentrated population. In this study, we aim to determine the prevalence and identify the risk factors of TB among elderly people to shape the TB screening algorithm for this high-risk population in China.

Methods

Study setting and sampling

This was a cross-sectional study. Sample size was estimated using a method appropriate to estimate a single population proportion. The 369/100000 prevalence of bacteriologically positive pulmonary TB (PTB) among elderly people (≥ 65 years) from the most recent national TB prevalence survey was used as the estimated prevalence of elderly population (unpublished data). We assumed 95% confidence interval and 0.2 allowable error and used the formula \( n=\frac{pq}{{\left(\frac{d}{{\mathrm{Z}}_{\upalpha}}\right)}^2} \) (p = 369/100000, q = 1 – p, d = 0.25p, α = 0.05, Zα = 1.96). A design effect of 1.8 from the simple random sampling was considered. So, a total of 29 873 elderly participants were needed and a sample size of 33 192 was determined to allow 10% of non-response.

We applied the multistage sampling in 27 study sites (10 townships in rural areas and 17 communities in urban areas) from ten counties of ten provinces in China. In the first stage, ten out of 31 provinces were selected, of which three were from eastern China, three from central China, three from western China, and one from four municipalities directly under the central government (Beijing, Shanghai, Tianjin and Chongqing). Selection at this stage was also based on willingness to participate and the human resources of each province. In the end, Sichuan, Guangxi, Yunnan of western China, Heilongjiang, Henan, Hubei of central China, Zhejiang, Jiangsu, Guangdong of eastern China, and Shanghai were chosen. In the second stage, one county per district that had more than 500 000 people was randomly selected in each province. In the third stage, random cluster sampling was used to select communities or townships. If the total number of general population in selected community or township was less than 30 000, then the nearest community or township would also be included in the study site, to reach 30 000. Finally, a total of 38 888 elderly people were eligible participants in the study.

Definitions

Suspected TB symptoms were defined by this study, as any of the following conditions: (1) cough for more than 2 weeks; (2) hemoptysis; (3) cough for more than 1 week yet less than 2 weeks, and accompanied with any of the following symptoms: fever, chest pain, night sweating, loss of appetite, fatigue, and/or weight loss (> 3 kg). The definitions of previous TB cases, human immunodeficiency virus or acquired immune deficiency syndrome (HIV/AIDS), known diabetes, close contacts, underweight, tobacco use, drinking history, chronic bronchitis, average family annual income per capita and average family living area, are shown in Table 1. It is necessary to note that China has its own criteria of body mass index (BMI) for adults: underweight BMI < 18.5, normal 18.5 ≤ BMI < 23.9, overweight BMI ≥ 24 [13].

Table 1 Definitions of terms used in this study
Full size table

Data collection

In our study, each province organized a research team of 50–100 staff for data collection, including researchers, health-care workers, enumerators, and local government staff. From June to September 2013, participants were interviewed for any suspected TB symptoms. Meanwhile, information of participants’ sex, age, marital status, education, medical history, smoking and drinking habit, and socioeconomic status were collected. Their height and weight were measured to calculate the BMI as an indicator of nutritional status. All participants were offered chest X-ray (CXR) examination. The interviews took approximately 15–20 min each CXR was completed in less than 1 h including waiting time for one participant.

Participants with any one of the suspected TB symptoms or CXR abnormalities consistent with TB were asked to submit three sputum samples (morning, night and spot sputum) for sputum smear and culture. Patients with smear-positive and/or culture-positive sputum were diagnosed as bacteriologically positive TB. Patients with active PTB included those with bacteriologically positive sputum, and those diagnosed only by changes on their chest radiographs—known as clinically diagnosed PTB [6].

Each county in China, according to the national guideline [14], has a TB diagnosis group, composed of at least three health staff, including a clinical doctor, a laboratory technician and a radiologist. They are trained to diagnose active PTB cases based on patient symptoms and clinical history, radiographic findings, bacteriological results, and response to antibiotics. A national expert group reviewed the data from each patient involved in this study to confirm the diagnosis. Quality checks were done according to the national guidelines [14]. The survey process is shown in Fig. 1.

Fig. 1
figure 1

Flowchart of the survey process in China in 2013. CXR: Chest X-ray; TB: Tuberculosis; PTB: Pulmonary tuberculosis

Full size image

The data collected was reviewed by the enumerators carefully on the same day, and any mistakes were corrected within 24 h. During the investigation period, 5 % of questionnaires were randomly selected for review by supervisors. Meanwhile, the study invited the Westat Company (Rockville, Maryland, USA) as a third party to supervise the implementation process.

Statistical analysis

All data were double entered using an online input system developed by a local software company. TB prevalence of subgroups was respectively calculated, including bacteriologically positive PTB and active PTB. Chi-square (χ2) test or Fisher’s exact test was used to compare the difference of subgroups. Variables identified as potentially relevant for active PTB included: sex, age, residence, marital status, education level, family average annual income per capita, family living area per capita, tobacco use, drinking history, diabetes, close contacts, previous TB cases, and chronic bronchitis. Univariate logistic regression analysis was used to identify potential risk factors associated with active PTB. To control potential confounders, a multiple logistic regression model with stepwise selection was performed. All statistical tests were two-tailed, and the significance level was set at P = 0.05 or less. Odds ratio (OR) with 95%confidence interval (CI) was calculated for categorical variables in the study to assess the strength of association between risk factors and TB prevalence. All tests were performed using SAS 9.3 (SAS Institute Inc., Cary, North Carolina, USA).

The population attributable fraction (PAF) is most commonly defined as the proportional reduction in average disease risk over a specified time interval that would be achieved by eliminating the exposure(s) of interest from the population while distributions of other risk factors in the population remain unchanged [15, 16]. Formula as below was used to calculate PAF of each adjusted risk factor.

$$ PAF=\frac{p_e\left( RR-1\right)}{p_e\left( RR-1\right)+1} $$

pe = proportion of source population exposed to the factor of interest. RR (risk ratio) may be the ratio of two cumulative incidence proportions, two (average) incidence rates (rate ratio), or an approximation of one of these ratios. In this study, pe means proportion of population exposed to adjusted risk factors, and OR was used to replace RR.

Ethical considerations

The study was reviewed and approved by the Institutional Review Board of Chinese Center for Disease Control and Prevention. Written informed consent was signed by each participant before enrollment. All patients identified were referred to the local designated TB clinic or hospital for treatment according to national guidelines [14].

Results

Demographic characteristics

There were 38 888 eligible elderly people (≥ 65), of which 34 269 (88.12%) participants were included in this study and finished both the questionnaire and physical examination. Of these 18 212 were male (53.14%) and 16 057 were female (46.86%) (Table 2). The median age was 72 (interquartile range: 68–77).

Table 2 Demographic characteristics of the elderly participants aged 65 or over in China in 2013
Full size table

Prevalence of bacteriologically positive TB and active PTB

Among the 34 269 participants, 817 (2.38%) reported to have one or more of the TB symptoms identified, 2542 (7.42%) had CXR abnormalities, and 168 (0.49%) had both.

A total of 62 bacteriologically positive TB cases and 193 active PTB cases were identified. The prevalence of bacteriologically positive PTB and active PTB were 180.92/100000, and 563.19/100000 respectively. The prevalence of bacteriologically positive TB and active PTB were higher in males than in females, increased with age and were higher in rural over urban areas (Table 3).

Table 3 TB Prevalence and comparison of subgroups of the elderly participants aged 65 or over in China in 2013
Full size table

Multivariable logistic regression analysis for independent determinants of active PTB

For active PTB, after adjustment in multivariable analysis, sex (M:F) (OR = 3.26, 95% CI: 2.34–4.55); using age group 65–74 as reference, age group 75–84 (OR = 1.59, 95% CI: 1.17–2.17), and age group ≥85 (OR = 2.05, 95% CI: 1.25–3.36); living in rural area (OR = 2.65, 95% CI: 1.81–3.88); lower family income (OR: 1.64, 95% CI: 1.17–2.31); using BMI group 18.5–23.9 as reference, BMI group < 18.5 (OR = 1.55, 95% CI: 1.09–2.22), and BMI group ≥24 (OR: 0.33, 95% CI: 0.19–0.55); diabetes (OR = 1.83, 95% CI: 1.08–3.10); close contacts (OR = 7.30, 95% CI: 2.15–24.82); and previous TB cases (OR = 9.23, 95% CI: 6.16–13.83), remained strongly associated (Table 4).

Table 4 Multivariable logistic regression analysis of active PTB and PAF of high-risk factors among the elderly participants aged 65 or over in China in 2013
Full size table

PAFs were considered for eight risk factors, as shown in Table 4. Male sex had the highest PAF (51.43%), followed by living in a rural area (50.67%), lower than average family annul income per capita (30.42%), previous TB cases (12.70%), underweight (5.52%), diabetes (5.50%), and close contacts (1.70%). PAFs for age were 15.51% for 75–84 years, and 5.57% for ≥85 years.

Discussion

This study found that the TB prevalence of elderly people is very high in China, and confirmed that male sex, older age, living in rural areas, underweight, diabetes, close contact of PTB and previously TB history are high risk factors. The risk of TB increased with increasing age and decreasing BMI after adjusting for other factors, and there is a positive dose–response relationship.

Male sex is a strong risk factor for TB disease, supported by both this study and studies from high-income and low-income countries [2, 17,18,19]. A study from the United States found that, among older adults aged 65 and above, TB rates increased with age [18], which was in accordance with our study. Residence in rural areas and with lower family incomes were risk factors in this study. However, this is not supported by studies from other low-income countries [3, 4, 20, 21]. This may be because in China, the difference between living in urban versus rural areas are more significant than other Southeast Asian and African countries.

Our study showed TB risk decreased with increasing BMI, which is also found in two other studies [22, 23]. TB and diabetes mellitus comorbidity is universal globally, and older age is a risk factor for this comorbidity [24]. Jeon and Murray found the relative risk of diabetes was 3.11 (95% CI: 2.27–4.26) [25], higher than this study. A systematic review by Morrison et al. of 41 studies showed that 4.5% of TB patient’s household contacts were diagnosed with active TB [26], and another review by Shah et al. of 25 studies showed the pooled yield was 7.8% (95% CI: 5.6–10.0%) for active TB in household contacts of drug-resistant TB [27]. In our study, 3.2% of TB patient’s household contacts were diagnosed with active TB, and the OR was 7.30 (95% CI: 2.15–24.82). In total, 5.4% (33/193) of previous TB patients had a recurrence, with a high OR of 9.23 (95% CI: 6.16–13.83) in this study. A retrospective cohort study conducted in Shanghai, China, showed 5.3% (710/13417) of successfully treated cases had a recurrence, a rate of 7.55 (95% CI: 7.01–8.13) episodes per 1000 person-years, more than 18 times the rate of TB in the general population [28]. More attention should be paid to patients with a history of previous TB diagnosis.

Heavy alcohol use and alcohol use disorders have a strong association with TB [29, 30], and this study showed the pooled relative risk was 2.94 (95% CI: 1.89–4.59), which used an exposure cut-off level set at 40 g of alcohol per day or above, or defined exposure as a clinical diagnosis of an alcohol use disorder [29]. In our study, there was no association between alcohol use and TB, no matter if the alcohol use was prior to or during the study. We defined alcohol use as using 20 g alcohol per week or above, which was much lower than the above-mentioned study. This may have led to the difference. Systematic reviews also showed tobacco smoking and chronic respiratory disease were associated with an increased risk of TB [31, 32]. Our study found the TB prevalence was higher in tobacco smokers and patients with chronic bronchitis, but with no significant statistical difference.

Lönnroth et al. estimated PAFs of selected TB risk factors for 22 high-burden countries, and the PAFs of underweight and diabetes for the total Chinese population were 16.5 and 4.4% respectively [10]. The PAF of underweight in our study was different from that estimation, because of a lower exposure rate and different target population. A study in Zimbabwe had a similar PAF of male sex (40%) and household contacts (10%), but with a high PAF of HIV (33%) [33]. Our study came to the same conclusion as a study in Tanzania [34], that a factor strongly associated with an outcome on an individual level, such as close contact and TB (OR = 7.30, PAF = 1.70%), may not have much population impact, if the exposure is not common in the population at large.

Dye and Williams thought that control programs had been less effective than expected in cutting transmission mainly because patients were not diagnosed and cured quickly enough [35]. To find more undiagnosed TB cases, ACF is an effective method. However, ACF can be a costly undertaking, depending on the target population and the diagnostic strategy used [36]. So, the first step to implement ACF is to confirm the target population with high enough prevalence. This study indicated that ACF could be implemented among elderly people aged 65 and above with low BMI (< 18.5), diabetes, close contact history and previous TB history as a priority, because of the higher risks and minor population number. The proportions of low BMI (< 18.5), diabetes, close contact history and previous TB history among elderly people are 10.62, 7.01, 0.27 and 1.77% respectively. Conducting ACF among these high-risk groups of elderly people will be more efficient and cost-effective. For other high-risk factors with a large population number, like male sex or residence in rural areas, it will be not as easy to screen them all. However, high-risk factors combination will be a better choice, which can increase the risk of development of recent active TB and narrow the target population [37].

The strengths of this study were that it was a carefully designed and implemented survey, which used the current TB diagnostic protocols and tests in China for diagnosis. And the results may represent the prevalence and case distribution characteristics of TB among elderly people in China. The study also showed the yield of ACF. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [38] and sound ethics principles for the conduct and reporting of this study [39].

The study had a few limitations. Some TB risk factors were collected by self-reporting (such as tobacco use and drinking history) or based on self-reporting and local health documents (such as diabetes), which might not be sufficient to estimate the real distribution among the elderly people. The prevalence of diabetes in this study was 7%, much lower than the 10.9% of the national diabetes prevalence survey in China in 2013 [40]. Meanwhile, our study was unable to detect any association between HIV and TB, as there was only one person who was living with HIV.

Conclusions

The TB prevalence of elderly people is very high in China, and male sex, older age, living in rural areas, underweight, diabetes, close contact of PTB and previous TB history are high risk factors. The risk of TB increased with increasing age and decreasing BMI after adjusting for other factors, and there is a positive dose–response relationship. ACF could be implemented among elderly people aged 65 and above with underweight, diabetes, close contact history and previous TB history as a priority, which will get significant yields and be cost-effective.