Impact of Multiple Policy Interventions on the Screening and Diagnosis of Drug-Resistant Tuberculosis Patients: Cascade Analysis on 6 Cities from Eastern China

Background The detection of drug resistant tuberculosis (DR-TB) is a major concern in China. China has implemented multiple DR-TB interventions in partnership with development partners and via domestic policy reforms. We aim to summarize the interventions related to the screening and detection of DR-TB in Jiangsu Province, analyse their impact, and highlight policy implications for the prevention and control of DR-TB. Methods We selected six prefectures across Jiangsu Province, a relatively developed area in China with a long history of collaboration with development partners and provincial policy reforms. We collected policy documents between 2008 and 2019, and extracted routine TB patient registration data from the TB Information Management System (TBIMS) between 2013 and 2019. We used the High-quality Health System Framework to guide the analysis. We performed statistical analysis on the percentage of pulmonary TB cases conrmed by bacteriology, percentage of bacteriologically conrmed TB patients tested for drug susceptibility by prefecture and by year. We used logistic regression models on TB cases conrmed by bacteriology and bacteriologically conrmed TB patients tested for drug susceptibility to further assess the impact of different policy interventions. Results Our study found that three prefectures in Jiangsu introduced DR-TB related interventions between 2008 and 2010 in partnership with the Global Fund and the Bill and Melinda Gates Foundation. By 2017, all prefectures in Jiangsu had implemented provincial level DR-TB policies, such as increases in human resources for health, use of rapid molecular tests (RMT), expanded drug susceptibility testing (DST) for populations at risk of DR-TB, and allocation of funds specically targeting DR-TB. With the use of RMT, the diagnostic process of DR-TB was optimized and the testing time shortened. The percentage of pulmonary TB cases conrmed by bacteriology increased from around 30% to over 50% between 2013 and 2019, indicating that the implementation of new diagnostics, including RMT, have provided more sensitive testing results than the traditional smear microscopy. At the same time, the proportion of bacteriologically conrmed cases tested for drug resistance has increased substantially, indicating that the intervention of expanding the coverage of DST has reached more of the population at risk of DR-TB. Prefectures that implemented interventions with the Global Fund and the Bill and Melinda Gates Foundation had better performance in detecting DR-TB patients. The disparities in detecting DR-TB among prefectures narrowed after of provincial samples county only progress in the implementation of policies on DR-TB the DR-TB detection across all prefectures after 2017. However, most indicators assessed did not have a linear trend of improvement, indicating the complexity of policy interventions affected by other factors such as domestic migration and economic developments. This is an observational study that used routine administrative data to reect the effects of policy interventions across different regions at different time points between the window of 2013 and 2019. We attempted to capture the impact of a group of interventions that have been implemented in different combinations and sequences. Our study brought real-world evidence but also has some limitations. First, random control design was not possible since this study relied on routine, retrospective data. Second, we were only able to analyse indicators already collected by the TBIMS. Other factors impacting the results (i.e., the insurance status of TB patients) were not fully recognized or reected in the analysis. Third, the data from TBIMS is designed for routine administration. It does not necessarily have robust quality assurance, and errors in data entry cannot be ruled out. We found international projects promoted local development in the detection of DR-TB, but the achievenements has only been observed in project sites. To fully scale up DR-TB detection across the country, long-term sustainable policies at the provincial, or even the national level, are needed. The application of new diagnostic devices and expanded screening strategy have improved the detection of DR-TB patients. Along with these improvements in DR-TB detection, resources and commitment from local government have played an important role in improving the implementation of DR-TB detection.


Background
Tuberculosis (TB) is a communicable disease that remains a key global health concern. In 2017, it led to the highest number of deaths and the second highest Disability-Adjusted Life Years (DALYs) among all communicable diseases globally [1]. The End TB Strategy led by the World Health Organization endeavours to reduce TB deaths by 95% and eliminate catastrophic health expenditures caused by TB by 2035 [2]. China has made great achievements in TB care and control: China's incidence of TB has been decreasing from 170/100 000 in 1990 to 61/100 000 in 2018 [3]. However, drug-resistant TB (DR-TB) continues to remain a public health threat in China. A 2007 national survey reported that 5.7% of new TB patients and 25.6% of previously treated TB patients were infected with multidrug-resistant TB (MDR-TB) [4]. Comparing to regular TB, MDR/RR-TB is more resource intensive with a longer treatment course, but the treatment success rate is less than 60% globally [5,6]. In 2018, less than one fourth of MDR/RR-TB patients were estimated to be detected, and of those that were detected, only 60% have initiated treatment [6].
Multiple strategies have been implemented in China to increase the detection of and treatment rates for DR-TB patients. In 2006, China started a partnership with the Global Fund to Fight AIDS, Tuberculosis and Malaria (the Global Fund) in two provinces. The partnership aimed to improve DR-TB diagnosis, treatment, and management [7]. Through another partnership in 2009 with the Bill & Melinda Gates Foundation (BMGF), rapid molecular tests (RMT) for DR-TB diagnosis and other treatment and nancing interventions were piloted in four prefectures of four provinces in China [8,9]. Given its rapid economic development, the Global Fund ended its partnership with China in 2014 and as a response to the end in external support, the Chinese government put more funding and policy support into DR-TB control and care. Since 2012, a national action plan has been announced to set goals on establishing a new nancing mechanism for DR-TB control, regulating the use of second line TB drugs, establishing DR-TB treatment and management sites at the prefecture level, coordinating the health services for DR-TB, enhancing capacity building and implementing DR-TB case nding strategy [7,10]. Provinces across China have implemented different TB strategies, and those that were supported by donor programs also introduced additional TB strategies. However, most of existing evidence provided assessment on single policy intervention on DR-TB, none has evaluated the impact of multiple interventions funded by domestic government and development partners. Wang et al summarized the implementation of projects funded by the Global Fund across China, but there was no assessment on the impact of different interventions [7]. Li et al evaluated the effect of a comprehensive DR-TB project with focus on access to diagnosis, quality treatment, and affordable treatment for MDR-TB [8]. Huang et al evaluated the bene t of expanding the diagnostic algorithm for diagnosis and treating bacteriologically con rmed pulmonary TB and RR-TB [11]. Considering the signi cant variance in DR-TB policies across the country and the various external factors beyond the health sector that may also in uence DR-TB control and care, it is essential to understand the complexity of DR-TB interventions to further improve the DR-TB policies at both the national and provincial level. The purpose of this study is to summarize the interventions and policy changes in the screening and detection of DR-TB patients in Jiangsu Province, analyse their impact, and provide policy implications for the prevention and control of DR-TB. Jiangsu has a long history of DR-TB control through partnership with international organizations and domestic efforts, and ranked top-tier among all provinces in China for the detection of DR-TB. By examining DR-TB policies and their impact using real-world evidence from Jiangsu, useful insights on DR-TB control could be extended to other regions of China.

Settings
We selected Jiangsu province as our study site. Jiangsu lies on the east coast of China. It is the fourth wealthiest province in China, with a GDP per capita $17,450 as of 2018. There are thirteen prefectures in Jiangsu Province with populations ranging from 3.2 million to 10.7 million. There is regional variance in economic development across the province, with prefectures in southern Jiangsu being the most developed followed by those located in central and northern Jiangsu. We selected Jiangsu province for several reasons. First, Jiangsu hosted one of the rst project sites for international collaboration on DR-TB control and care, including collaborations with the Global Fund and BMGF. Second, the provincial and prefecture-level governments of Jiangsu were very supportive of the prevention and treatment for DR-TB. Finally, Jiangsu province has had digitalized Hospital Information System (HIS) and TB Information Management System (TBIMS) since 2005 [12]. The accessibility and quality of historical data made our analysis feasible.
We selected two prefectures from each region of Jiangsu Province to ensure representation of the different levels of development and re ect variation among local policies: Zhenjiang and Changzhou (South), Nantong and Yangzhou (Central), Huai'an and Lianyungang (North). Socioeconomic information of selected prefectures can be found in Appendix 1a.

Conceptual framework
We used the High-quality Health System Framework to guide the analysis [13]. We rst looked at the Foundations that have been invested and used in DR-TB prevention and treatment, including policy, workforce, diagnostic devices, etc. Then, we described the Processes of care that indicated the activities for DR-TB testing, such as turnaround time for testing results. Finally, we presented the percentage of pulmonary TB cases con rmed by bacteriology and the percentage of bacteriologically con rmed TB patients tested for drug susceptibility as the Impact indicators of the DR-TB interventions.

Data collection
We collected government policy documents related to prevention, screening, diagnosis, and management of DR-TB. We also collected the protocols of DR-TB related projects from 2008 to 2019 at provincial, prefecture, and county levels to understand the interventions by project funders as well as the policies issued by local government. We collected noti ed TB cases and presumptive DR-TB patients from TBIMS [12]. Data were restructured into three case-based databases: one for all active TB cases, one for patients with presumptive DR-TB, and one for con rmed DR-TB cases.
We extracted noti ed TB case records and presumptive DR-TB case records from TBIMS from the six selected sites. Records with a diagnosis of "pulmonary tuberculosis" between Jan 1, 2013 and Dec 31, 2019 were included for analysis. We collected the following indicators for noti ed TB patients (i.e., those with pulmonary TB): registration number, age, sex, residency (within the county, within the prefecture, etc.), and TB related information including bacteriological results, previous TB treatment, whether diagnosed at county or prefecture agency, type of diagnosis agency, and date of entry (into TBIMS). Similarly, we collected the following indicators for presumptive DR-TB: diagnosis region, registration number, age, sex, source of presumptive DR-TB (linked with TB database, directly registered and linked with con rmed DR-TB database), the registration date of basic management unit (BMU) [14,15], smear results, culture results, strain identi cation results, types of drug susceptibility testing (DST) (phenotypic or genotypic), results of DST, con rmed date of DST, and date of entry. No personally identi able information was extracted from the database.

Data analysis
Descriptive analysis on key indicators was disaggregated by year and by prefecture. The de nitions of key indicators of interest are listed below.
1) Bacteriologically con rmed TB case: A patient from whom a biological specimen is positive by smear microscopy, culture, or rapid diagnostic test.
2) Percentage of pulmonary TB cases con rmed by bacteriology: The number of bacteriologically con rmed TB cases divided by the number of noti ed active pulmonary TB cases.
3) Percentage of bacteriologically con rmed TB patients tested for drug susceptibility: The number of bacteriologically con rmed TB cases tested for drug resistance (including tested only for rifampicin resistance) divided by the number of bacteriologically con rmed TB cases. [7]: TB patients with at least one of the following conditions: a) failure to respond to the initial or retreatment efforts, chronic cases of TB, b) close contact with a known MDR-TB patient, c) smear-positive at the end of 2 to 3 months of initial treatment.

4) Suspected high risk (M)DR-TB patients
We used logistic regression models on noti ed TB records to further assess the impact of different policy interventions. In model 1, the dependent variable was the bacteriological results. The key explanatory variable was use of the new diagnostic devices. The control variables were: year of registry (data must be entered within 24 hours of diagnosis), prefecture, sex, age group (under 45, 45 to 60, and over 60), treatment history (new patient, previously treated patient), cavitary disease (no, yes), administrative level of diagnosis agency (prefecture, county), type of diagnosis agency (general hospital, specialized hospital, CDC clinic or TB dispensary), current residency (within county, within prefecture, within Jiangsu, outside Jiangsu). For model 2, the dependent variable was con rmed TB patients tested for drug susceptibility. We used a group of dummy variables as a key explanatory variable to indicate the interventions on the DST coverage, while the control variables were the same as model 1.

Policies and major interventions
Based on our review of the policy documents and intervention protocols, we mapped out different TB interventions that were implemented in each prefecture between 2013 and 2019. Considering the impact of health interventions may last longer after its implementation, we summarized the DR-TB related interventions between 2008-2019. Figure 1 shows the timeline, scope of each intervention, and implementation sites of major interventions related to the screening and diagnosis of DR-TB. Zhenjiang was the rst prefecture to implement DR-TB interventions on treatment and patient management (Global Fund, 2008-2014), followed by Lianyungang and Nantong. These three prefectures started to follow provincial level DR-TB interventions after the end of international projects. Conversely, Changzhou, Huai'an, and Yangzhou, implemented DR-TB interventions relatively late, beginning only in 2012. In contrast to the other three prefectures of focus, these three prefectures did not host any TB-related international projects, and therefore only implemented provincial level DR-TB interventions. comprehensive TB control model for effective diagnosis, treatment, and management of TB patients using the molecular diagnosis technology and expanding DST coverage [16]. At the beginning of the provincial project, it focused on diagnosis, treatment, management, nancing of MDR-TB patients, and DST with suspected high risk (M)DR-TB patients. To accelerate the pace of TB control, RMT was introduced to all TB labs by 2017 and DST with bacteriologically con rmed TB cases.
2 Workforce and funding for DR-TB We reviewed and compared the changes in TB related staff, funding for DR-TB and laboratory equipment between 2013 and 2018 as the input of DR-TB detection. Between 2013 and 2018, the number of designated TB hospitals remained the same for most prefectures except for Zhenjiang, while the number of RMT devices have been increased substantially, indicating more TB designated TB hospitals were facilitated with RMT devices (Appendix 1b). In 2013, there were only three TB laboratories equipped with RMT devices: by 2018 all TB laboratories were equipped with RMT devices, funded either by international programs or provincial government. Between 2013 and 2018, the number of full-time staff at local CDCs dropped slightly while that of the designated TB hospitals increased. Full-time staff of local CDCs declined in ve prefectures, with the greatest declines found in Zhenjiang and Nantong. Between 2013 and 2018, full-time staff of designated TB hospitals increased in most prefectures, among which Yangzhou saw the most growth (from 32 to 71). Funds allocated speci cally for DR-TB control and prevention from the provincial government increased substantially in ve prefectures and the disparity of special funds dedicated to DR-TB narrowed in 2018: Lianyungang was the only prefecture that saw a decline in special funds and Nantong had the most rapid growth, followed by Zhenjiang and Huai'an (Appendix 1b).

Screening and testing process
Screening and testing are regarded as inter-mediate changes of DR-TB detection practice. As mentioned in previous sections, more devices have been put into use in Jiangsu while the detection population has been expanded according to some DR-TB detection polices. We compared the screening and testing process for DR-TB between 2013 and 2019 to observe changes.
Conventional diagnosis methods (including sputum smear microscopy, culture and phenotypic DST) were mainly used before 2017 (see Appendix 2). Speci cally, sputum smear microscopy, culture testing, and risk assessment for drug-resistance were performed at the county level. Clinical isolates or sputum samples with positive results had to be transferred to the prefecture-level laboratory for DST before 2017. The rapid molecular DST was accessed in prefectures implementing international projects such as Zhenjiang, Lianyungang, and Nantong.

Testing time
Testing time is jointly in uenced by the type of test used and the testing process. Rapid turnaround time for testing can bring timely treatment and reduce potential risk of spreading infection. We calculated the time required for the diagnosis of DR-TB (de ned as days from the registration date in BMU to the con rmed date of DST). RMT has a signi cantly shorter time compared with solid culture and DST. From 2013 to 2019, the average time for RMT shortened by two thirds (from 18.86 to 6.36) while the average time for solid culture and DST dropped slightly, but still remains over 60 days (Fig. 2).

Percentage of pulmonary TB cases con rmed by bacteriology
The percentage of pulmonary TB cases con rmed by bacteriology has increased from around 30% in 2013 to 55% in 2019. The application of various detection methods, such as culture and rapid TB testing resulted in the increase of the percentage of pulmonary TB cases con rmed by bacteriology (Fig. 3A-F). Prefectures that implemented international projects generally had better performance between 2013 and 2016. For instance, Zhenjiang, one site for Gates project between 2014 and 2016, had the highest percentage of pulmonary TB cases con rmed by bacteriology among the six selected prefectures. Most prefectures experienced signi cant improvement between 2017-2018 given the implementation of provincial-level policies in all prefectures (including the application of culture and RMT for suspected TB patients). The percentage of pulmonary TB cases con rmed by bacteriology in our six focus prefectures all reached 50% by 2019 while the disparities among prefectures gradually narrowed over the seven year time period reviewed ( Table 2, Fig. 3A-F). 6 Percentage of bacteriologically con rmed TB patients tested for drug susceptibility In 2013, the percentage of bacteriologically con rmed TB patients tested for drug susceptibility varied from below 10% to over 90% across the prefectures. For prefectures that either performed DST on wider suspected DR-TB patients or perform DST before 2017, the proportion of bacteriologically con rmed TB patients tested for drug susceptibility was higher. The proportion of bacteriologically con rmed TB patients tested for drug susceptibility has risen sharply, particularly among prefectures with low rates at the beginning since the implementation of provincial-level policies in 2017. In 2019, the percentage of bacteriologically con rmed TB patients tested for drug susceptibility was around 95% for all the prefectures ( Table 2, Fig. 4A-F). Factors associated with pulmonary TB cases con rmed by bacteriology were analyzed. Results showed that the application of RMT (Xpert) is 1.11 times (p < 0.01) more likely to con rm TB cases than the smear test. The overall likelihood of pulmonary TB cases con rmed by bacteriology improved between 2013 and 2019. Females, the elderly (over 60), previously treated TB patients (compared with new patients), those with a cavity in a lung, and those residing outside the prefecture (compared with within county) are more likely to be bacteriologically con rmed. Patients are more likely to be bacteriologically con rmed by a county level diagnosis agency (compared with prefecture-level) and at general hospitals (compared with specialized hospitals, CDC, or TB dispensary).
We analyzed factors associated with bacteriologically con rmed cases tested for drug resistance. Results showed that the following types of patients were signi cantly more likely to be tested for drug resistance than patients with coverage of suspected high-risk DR-TB patients: patients with DST coverage of biologically con rmed TB cases (3.84 times), smear-positive TB patients (2.54 times), and smear-positive TB patients and smear-negative but culture-positive TB patients (2.30 times). Females, the elderly (over 60), previously patients treated (compared with new patients), and those residing within the county (compared with outside the county), were more likely to be tested for drug resistance. Bacteriologically con rmed cases were more likely to be tested by county level diagnosis agency (compared with prefecture-level) and at general hospitals (compared with specialized hospitals, CDC, or TB dispensary) for drug resistance. Table 3 In uential factors for pulmonary TB cases con rmed by bacteriology and bacteriologically con rmed cases tested for drug resistance

Discussion
Our study found that Jiangsu province has implemented a series of interventions to improve DR-TB control and prevention, such as workforce increases, the use of RMT devices, and allocation of special funds focused on DR-TB. With the application of RMT, the diagnostic process of DR-TB has become more convenient to patients with reduced testing time. The percentage of pulmonary TB cases con rmed by bacteriology has increased from around 30% in 2013 to over 50% in 2019, indicating that the implementation of new diagnostics, including RMT, have provided testing results more than the traditional smear microscopy. At the same time, the proportion of bacteriologically con rmed cases tested for drug resistance has increased substantially, indicating that the intervention of expanding the coverage of DST has reached more populations at risk of DR-TB.
Although there is no similar research analysing the impact of multiple DR-TB interventions on diagnosis, our ndings are consistent with results from other evaluation studies on single DR-TB interventions, such as the implementation of RMT. For example, in Armenia the introduction of RMT increased the detection rate of RR-TB [17]; O. Oxlade et al found that the implementation of rapid DST at the moment of diagnosis was the most cost-effective strategy [18]; and a study in South Africa found that the implementation of molecular DST reduced the laboratory turnaround time from 55 days to 27 days [19]. These studies also found the same results as we did: the application of rapid DST has improved the diagnosis of DR-TB, and improved diagnosis is critical for achieving desired treatment results [20].
We observed that the three prefectures that participated in a Global Fund project and/or China-Gates project (Zhenjiang, Lianyungang, and Nantong) generally performed better than other prefectures, even after the partnerships concluded. This nding aligns with other studies. For example, in Nicaragua Plamondon et al also found that the presence of a Global Fund project improved TB control, built human resource capacity, and strengthened community involvement [21]. Our ndings show that support from the Global Fund or BMGF had a strong emphasis on capacity building. Whether their support was for diagnostic devices or other best practices, such as DST to smear positive TB patients, the capacity building from international projects has had a long-lasting role in promoting the diagnosis for DR-TB [8].
Another signi cant nding is that the disparity in diagnosis for DR-TB among the six prefectures disappeared after 2017. This improvement can be attributed to the strong political commitment of the Jiangsu government in TB control and care. The provincial DR-TB intervention policies that were implemented were accompanied by su cient resources including staff, funding, and diagnostic devices, which ensured the sound implementation of interventions. In addition to the infrastructure acquired from international projects, Jiangsu province made a signi cant investment to strengthen the infrastructure and capacity of laboratories of the prefectures that did not have international partnerships. For instance, both conventional equipment and RMT devices were introduced to all prefectures in Jiangsu (GeneChip to prefectures and LAMP to counties). In comparison, throughout the country only 44.2% of labs at the prefecture level and 27.5% at the county level have the same capacity as Jiangsu [22].
The diagnostic algorithm substantially improved after the application of RMT. The new process is more effective for the diagnosis of TB and DR-TB and reduces the testing time, enabling patients to receive more timely treatment. While the overall testing time was reduced, we noticed that in 2016, testing time for RMT (Hain or GeneChip) was slightly longer than previous year after its initial introduction to prefectures. This increase may re ect the learning curve of the lab staff to understand and execute the operating procedures. The average time for solid culture and DST continuously decreased after 2016, but it was still two times longer than the rapid molecular DST. Therefore, the application of the rapid molecular DST can reduce diagnosis delay and enable prompt treatment.
We observed signi cant changes across various indicators after the implementation of DR-TB detection interventions. The proportion of bacteriologically con rmed cases tested for drug resistance varied greatly across prefectures in 2013. Some prefectures, particularly those with the least progress in DR-TB detection, experienced dramatic increases between 2016 and 2019 (Yangzhou and Huai'an). Not as international projects that only made progress in project sites, the implementation of provincial policies on DR-TB improved the DR-TB detection across all prefectures after 2017. However, most indicators assessed did not have a linear trend of improvement, indicating the complexity of policy interventions affected by other factors such as domestic migration and economic developments.
This is an observational study that used routine administrative data to re ect the effects of policy interventions across different regions at different time points between the window of 2013 and 2019. We attempted to capture the impact of a group of interventions that have been implemented in different combinations and sequences. Our study brought real-world evidence but also has some limitations. First, random control design was not possible since this study relied on routine, retrospective data. Second, we were only able to analyse indicators already collected by the TBIMS. Other factors impacting the results (i.e., the insurance status of TB patients) were not fully recognized or re ected in the analysis. Third, the data from TBIMS is designed for routine administration. It does not necessarily have robust quality assurance, and errors in data entry cannot be ruled out.

Conclusion
We found international projects promoted local development in the detection of DR-TB, but the achievenements has only been observed in project sites. To fully scale up DR-TB detection across the country, long-term sustainable policies at the provincial, or even the national level, are needed. The application of new diagnostic devices and expanded screening strategy have improved the detection of DR-TB patients. Along with these improvements in DR-TB detection, resources and commitment from local government have played an important role in improving the implementation of DR-TB detection. Table   Table 1 not available with this version.