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

Tuberculosis (TB) remains the most important infectious disease worldwide. Its etiological agent, Mycobacterium tuberculosis, infects one-third of the world population and it accounted for some 1.5 million deaths in 2013 according to WHO estimates [1]. China accounted for 11 % of global TB cases in 2013. It ranked fifth in terms of the epidemiologic burden of TB, with an incidence rate of 70 per 100,000, a prevalence rate of 94 per 100,000 and an estimated 3.0 deaths per 100,000 attributed to TB in 2013 [1]. Attaining WHO’s end TB strategy, zero deaths, disease, and suffering due to TB by 2035, will be a major challenge [2]. Non-compliance with treatment remains a central concern [3–6]. The direct observation of treatment (DOT) approach has proved effective, but, as in many countries, considerable proportion of non-compliant patients exist [6–12].

Studies indicated that education level, marital status, employment, and lack of TB and its treatment related knowledge were important reasons for non-compliance [8, 9, 13–15]. A number of studies also assessed related costs and found that a high economic burden is associated with poor treatment compliance [16–21].

To reduce the financial burden on and barriers to patients to facilitate seeking essential healthcare, a “free-TB service policy” has been implemented [9, 14, 16, 21, 22]. Under this policy, tuberculosis suspects are provided a free diagnosis and anti-tuberculosis treatment, including designated first-line TB drugs (6 months for new patients, 8 months if previously treated), chest X-ray examination for the first and last months, and sputum smear test [16, 21, 23–25]. Going with this policy, there are three service model: dispensary model, specialist model, integrated model [26]. The dispensary model is the most prevalent model. A TB dispensary is usually a department of Centre for Disease Control (CDC), with free treatment is only provided in this place. General hospitals are just responsible for referring TB suspects to dispensaries, and usually are supposed should not treat TB patients, unless severe or complicated cases. The specialist model is similar to dispensary model, with severe TB cases should only be treated in the specialist hospital, but treatment cost in specialist hospital is not covered by the free treatment policy. The integrated model is a new development, with the most popular general hospital in the district, usually called as TB designated hospital, provides TB treatment. TB dispensary continues to provide related service but just limited in training, mass education and case supervision, and should report and ensure referred TB suspects arrived in the designated hospital. Expenditure beyond the “free-TB service policy” was supposed to be covered by insurance. In china, health care system is bifurcated in nature between rural and urban areas [27], and TB patients mainly covered by government-led complementary insurance programs: the Urban Employee Basic Medical Insurance (UEBMI), designed exclusively for urban worker; the Urban Resident Basic Medical Insurance (URBMI), designed for urban residents without formal employment; the New Cooperative Medical Scheme (NCMS), designed for rural population.

Despite the free policy and the cover of insurance, the medical and non-medical costs incurred by TB patients and their households are still substantial [16, 28–32]. The phenomenon that requesting repeated investigation, including blood tests, X-rays and even CT test, and prescribing liver protection and ancillary drugs is not uncommon [10, 16, 21, 25, 33].

Several studies have also found that the anticipation of a high economic burden and concern with ‘sunk costs’ can affect patient decisions [9, 34–38]. Some patients may have poor compliance because they expect that the overall cost of treatment will be high even though their actual out-of-pocket payments have been relatively limited. Others who are facing serious financial difficulties may persevere with treatment because of the considerable investment in care already undertaken: the sunk cost effect. The observation of such complex relationships between economic burden and compliance emphasizes the need for a combination of qualitative and quantitative approaches. However, previous studies have generally adopted one approach. The primary objective of this paper is systematic research this problem through a mixed method.

Another gap is that previous studies usually focus on the direct medical cost. Few studies involve with direct non-medical and indirect costs. For certified low income and unemployed patients, just transportation cost can account for a considerable proportion of their annual household income. How about the relationship between economic burden incurred by direct non-medical and indirect costs and treatment compliance? The second objective of this paper is to fill this gap, and to examine whether and to what extent this burden affects patient treatment compliance.

In this study, we found that age, education level, household income and patients’ subjective assessment whether TB causes a heavy burden to their family are significant different between group compliance and non-compliance. In consistence with previous studies, we also got the similar result that patients who are elderly, low education level are more likely to non-compliance to TB treatment [8, 9, 14, 15]. The elderly and low education level patients are more likely to have a limited ability to get and understand related TB knowledge. This cue that except the chemotherapy, a humanist intervention or assistance may be needed.

Compared with direct medical costs, the non-medical costs, dietary supplement costs and indirect costs were not high. This finding is not consistent with other studies, which often indicate that the main burden arises from transport, diet supplements and loss of income [17, 46]. Explanation is that most of the patients were located in rural areas with incomes mainly coming from farming or self-employment (Table 2). The time required for diagnosis and treatment have little impact on the time available for work and the food for diet supplementation usually be home grown.

Though the free-TB service policy, patients have to pay a high expenditure on diagnosis and treatment. Even though covered by insurance, this expenditure doesn’t reduce too much. In the integrated model, TB patients can only be diagnosed and treated under the free treatment policy in the designated hospital. The free treatment policy just covered first-line anti-TB drugs, TB sputum smears and cultures, and X-ray examinations for the first and last months [16, 21, 23, 24, 51], and these items’ profit is low. Driven by the revenue, doctors may over-prescript additional drugs and test [16, 33]. As the qualitative results showed, overuse of liver protection drugs and CT test is not uncommon. And also previous studies revealed that providers have the motivation to use second-line anti-TB drugs [16, 21, 26]. All these overused items neither covered by the free policy, nor by the insurance, resulting in high patient OOP medical expenditure. A systematic review based on 85 research articles found no reliable evidence to support prescription of drugs protects the liver [52]. studies also reported that NCMS did not reduce patient out-of-pocket costs per episode [31, 53, 54]. All of these indicated that improving providers’ salary system, expanding insurance’s coverage is needed.

This study revealed a very high economic burden for TB patients in China, in line with prior studies [16, 28, 29, 55]. Almost 68 % of TB patients reported catastrophic health expenditure (Table 3), much higher than the average for China [43, 44]. Patients with low household incomes or who felt that the economic burden was high were more likely to occur non-compliance (Table 5). These findings lend support to the argument that there is an association between TB and poverty and suggests that TB patients still face economic barriers in seeking diagnosis and treatment [20, 28, 30].

Contrary to expectations, the quantitative analysis suggests that a high economic burden did not in itself result in poor compliance (Table 5). This finding has two possible explanations. First, a sunk cost effect probably exists in relation to seeking diagnosis and treatment. If a patient has spent a lot of money at early stage, he or she may continue to seek treatment, to avoid that expenditure becoming a sunk cost, even borrowing money or selling property to continue treatment. Second, in the process of calculating the economic burden, we only include the actual costs, not the anticipated costs. Patients with low household incomes may interrupt their treatment because they expect that further treatment will become unaffordable [9]. Such patients would certainly feel that the economic burden of treatment was very high. And these reasons can also explain why we can’t get a significant result from the test of direct and indirect costs between compliant patients and non-compliant patients.

Even the costs related to transport, accommodation and food could lead to catastrophic expenditure (Table 3). And the results from both from quantitative and qualitative studies demonstrate that the burden from costs on transport, lodging and food had a substantial effect on treatment compliance, in line with Tadesse’s study [3].This also lend supports to the hypothesis above. In China, under the free policy, providers are supposed to educate TB patients related knowledge before chemotherapy. Previous studies indicated several points: lack of knowledge about anti-TB treatment was associated with higher non-compliance rates [8, 13]; some patients still confused about treatment regimen after health education [8]; adverse reactions contribute to non-compliance [15, 21]. For patients with low household income, if adverse reactions occurred or symptoms relieved and can’t convenient keep in touch with doctors, travelling long distances to a health care facility at considerable expense will challenge their willingness to continue treatment [3, 16]. Another situation is that as patients have to bring back their medicine monthly, the high cost of transport, lodging and food really impede them from doing so.

What can we do to improve TB patients’ compliance? We suggest we can through two aspects to influence patients’ behavior: knowledge and economic burden. As we discussed before, for patients, transport, accommodation and food costs can be a barrier to keep in touch with doctors. So one of the useful ways is to develop mobile phone based intervention, as several studies already emphasize this intervention’s potential effectiveness [56–58]. Another way is to expand the current package to cover patients’ transport, accommodation and food costs or offer related economic assistance. Previous studies indicated that lack of knowledge about anti-TB treatment was associated with higher non-compliance rates [8, 13], and some patients still confused about treatment regimen after health education [8]; we found that patients who are elderly, low education level are more likely to non-compliance to TB treatment. This implied that a humanist intervention tailored by these factors may be helpful.

For economic burden, no matter actually happened burden or anticipated burden, an experience-based opinion is that due to the free-TB service policy, the economic burden caused by OOP2 just happened when patients already played a visit to doctor, otherwise they wouldn’t know or anticipate whether the burden is high. According to the above discussion, improving doctors’ salary system to cut off the revenue-oriented incentive, and expanding current insurance’s coverage is useful. Chinese insurance is a multilevel system. In our study, the actually happened economic burden is high. To enhance the cooperation between different level’s insurance, for instance between basic insurance and critical illness insurance or assistance, is necessary.

Our results lend weight to the argument that there is a link between TB and poverty and that TB patients continue to face economic barriers in seeking diagnosis and treatment. TB patients with low household incomes or who feel that the economic burden of treatment is high are more likely to have poor compliance. As indicated above, the number of TB patients in China is very large. The provision of free anti-TB drugs has been a valuable step forward but to improve the coverage and effectiveness of control programs for TB. We suggest that the useful way is to develop mobile phone based intervention or offer patients transport, accommodation and food economic assistance to break down the barrier of keeping in touch with doctors. Another way is to improve doctors’ salary system to cut off the revenue-oriented incentive, expand current insurance’s coverage, and enhance cooperation between different level’s insurance.