Cost–effectiveness of a comprehensive programme for drug-resistant tuberculosis in China

Abstract Objective To investigate the cost–effectiveness of a comprehensive programme for drug-resistant tuberculosis launched in four sites in China in 2011. Methods In 2011–2012, we reviewed the records of 172 patients with drug-resistant tuberculosis who enrolled in the comprehensive programme and we collected relevant administrative data from hospitals and China’s public health agency. For comparison, we examined a cohort of 81 patients who were treated for drug-resistant tuberculosis in 2006−2009. We performed a cost–effectiveness analysis, from a societal perspective, that included probabilistic uncertainty. We measured early treatment outcomes based on three-month culture results and modelled longer-term outcomes to facilitate estimation of the comprehensive programme’s cost per disability-adjusted life-year (DALY) averted. Findings The comprehensive programme cost 8837 United States dollars (US$) per patient treated. Low enrolment rates meant that some fixed costs were higher, per patient, than expected. Although the comprehensive programme appeared 30 times more costly than the previous one, it resulted in greater health benefits. The comprehensive programme, which cost US$ 639 (95% credible interval: 112 to 1322) per DALY averted, satisfied the World Health Organization’s criterion for a very cost–effective intervention. Conclusion The comprehensive programme, which included rapid screening, standardized care and financial protection, improved individual outcomes for MDR tuberculosis in a cost-effective manner. To support post-2015 global heath targets, the comprehensive programme should be expanded to non-residents and other areas of China.


Introduction
China has more than halved tuberculosis prevalence and mortality since 1990, from 215 cases and 19 deaths per 100 000 population to 108 cases and 4 deaths per 100 000 population in 2010. 1 However, in 2012, China had an estimated 59 000 notifications of multidrug-resistant (MDR) tuberculosisresistant to both isoniazid and rifampicin -but only 1906 patients enrolled in treatment for MDR tuberculosis. 2 If China's attempts to end tuberculosis by 2035 3 are going to succeed, it is necessary that patients with MDR tuberculosis receive adequate treatment.
In 2011−2014, the Chinese Center for Disease Control and Prevention (CCDC) piloted a comprehensive programme for drug-resistant tuberculosis in four areas of China: Chongqing municipality, Henan province, the Inner Mongolia autonomous region and Jiangsu province. 4 The comprehensive programme included the screening of all patients with sputum-smear-positive pulmonary tuberculosis, rapid diagnosis, standardized care and financial protection against catastrophic health expenditure. In the previous baseline programme in effect in the four study areas, conventional drug-susceptibility testing was recommended for all newlydiagnosed patients with smear-positive pulmonary tuberculosis and for all previously-treated patients. In practice, such testing was confined to previously-treated patients and new patients who could afford to cover the full costs of such testing. The non-standard treatment lasted only as long as patients were willing and able to pay for it.
The aims of the study were to determine whether the comprehensive programme appeared cost-effective when compared against the baseline programme and a hypothetical no treatment, zero-cost alternative that we called the null programme.

Study perspective
We assessed cost-effectiveness from a societal perspective and therefore included direct medical costs incurred by hospitals, insurance schemes, patients and the Chinese CDC. We excluded the costs incurred by patients while seeking diagnosis or treatment from other providers before diagnosis under the comprehensive programme. We also excluded direct nonmedical costs and any time costs incurred by patients and their caregivers. We had no reliable data on the excluded costs for the baseline programme, and the assessment and quantification of time costs are controversial. Using the World Health Organization's cost-effectiveness threshold, we considered any Objective To investigate the cost-effectiveness of a comprehensive programme for drug-resistant tuberculosis launched in four sites in China in 2011. Methods In 2011-2012, we reviewed the records of 172 patients with drug-resistant tuberculosis who enrolled in the comprehensive programme and we collected relevant administrative data from hospitals and China's public health agency. For comparison, we examined a cohort of 81 patients who were treated for drug-resistant tuberculosis in 2006−2009. We performed a cost-effectiveness analysis, from a societal perspective, that included probabilistic uncertainty. We measured early treatment outcomes based on three-month culture results and modelled longer-term outcomes to facilitate estimation of the comprehensive programme's cost per disability-adjusted lifeyear (DALY) averted. Findings The comprehensive programme cost 8837 United States dollars (US$) per patient treated. Low enrolment rates meant that some fixed costs were higher, per patient, than expected. Although the comprehensive programme appeared 30 times more costly than the previous one, it resulted in greater health benefits. The comprehensive programme, which cost US$ 639 (95% credible interval: 112 to 1322) per DALY averted, satisfied the World Health Organization's criterion for a very cost-effective intervention. Conclusion The comprehensive programme, which included rapid screening, standardized care and financial protection, improved individual outcomes for MDR tuberculosis in a cost-effective manner. To support post-2015 global heath targets, the comprehensive programme should be expanded to non-residents and other areas of China.
intervention that cost less than China's gross domestic product (GDP) per capita to avert a disability-adjusted life year (DALY) to be very cost-effective. 5

Comprehensive programme
In the comprehensive programme, sputum samples were collected from all patients who presented with sputumsmear-positive pulmonary tuberculosis either at a hospital or a dispensary maintained by the Chinese CDC. These samples were sent to a hospital laboratory and tested for Mycobacterium tuberculosis and resistance to rifampicin, using the rapid diagnostic Genechip test (CapitalBio, Beijing, China). All patients found positive for rifampicinresistant M. tuberculosis were considered to be eligible for enrolment in the study if they were resident in one of the four study areas and had no contraindications. Although sputum samples were also cultured and then subjected to conventional drug susceptibility testing, enrolled patients were started on treatment for MDR tuberculosis before the results of such testing became available.
All enrollees were treated with the same standardized regimen. This regimen consisted of a six-month intensive phase -based on pyrazinamide, either amikacin or capreomycin, either levofloxacin or moxifloxacin, p-aminosalicylic acid and prothionamide -followed by an 18-month continuation phase, based on the same drugs except amikacin or capreomycin. A standard package included examinations before hospitalization, hospitalization up to two months, follow-up after discharge, health education and psychosocial support (available from the corresponding author).
To encourage collaboration in the referral, hospitalization and follow-up of patients, incentives were paid to the Chinese CDC and hospital staff and village doctors who were involved in the testing of the comprehensive programme (Table 1). A small transportation subsidy was paid to patients: about 15 United States dollars (US$) per hospital visit, for up to 13 visits.
The comprehensive programme's financial protection has been described in detail. 4 In brief, the programme involved a move away from fee-for-service payment towards a standard package with controls on out-of-package fees and, by design, 90% of the fees within the standard package were reimbursed by insurance schemes and the programme. Under the baseline programme, reimbursement by the most common scheme, the New Rural Cooperative Medical Scheme, was 20−80% for inpatient services and 0−80% for outpatient services.

Data collection
The data for the baseline programme came from a review of medical records and a previous survey. The data consisted of all people with confirmed MDR tuberculosis who were hospitalized for treatment in Henan province, the Inner Mongolia autonomous region or Jiangsu province between January 2006 and June 2009. 4 Chongqing municipality was not included in the baseline study because, in the relevant period, it offered no culture confirmation of drug resistance.
Our investigation of the comprehensive programme was based on a detailed review of medical records and a survey of 73 enrollees -conducted after the enrollees had completed six months of treatment. We also included routinely collected data -mostly sputum-smear microscopy and culture results -for enrollees who gave their informed consent. The last routine data were collected in September 2012, after all the enrollees had completed at least six months of treatment. Costs for the full period of treatment were estimated according to the use of services and items defined in the standard package. Unit costs for outpatient clinic visits, sputum-smear microscopy, radiographs and hospital inpatient stays -i.e. bed days -were estimated, in each study area, using standard methods and templates. 6 Unit costs for Genechip screening and conventional drug susceptibility testing were previously determined for each study area. 7 We based our estimates of the unit costs of culture on our estimates of the unit costs of smear microscopy, multiplied by the ratio of the cost of culture to smear microscopy reported for six other sites in China (ratio range: 2.7-15). 8

Research
Cost-effectiveness of a tuberculosis programme, China Christopher Fitzpatrick et al.
The cost of a bed day was multiplied by the mean number of bed days recorded for enrollees who had been discharged from hospital by September 2012 and then by the total number of enrollees. The unit cost of a Genechip test was multiplied by the total number of patients screened. For conventional drug susceptibility testing, outpatient visits, smear microscopy, culture and chest X-rays, unit costs were multiplied by quantities indicated in the standard care package and by the number of enrollees. The total cost of the drugs used within the standard package was extracted from hospital fee schedules. For all other tests and drugs used within or outside the standard package, we extracted hospital charges from the relevant patient records. We reduced the cost of drugs by 15% to remove the hospitals' permitted mark-up.
Administrative data were collected from the facilities maintained by the Chinese CDC, at both city and county levels. Project and data management costs were estimated by multiplying the number of full-time equivalent staff times their gross wages and adding the cost of meetings between the Chinese CDC and the hospitals involved in the comprehensive programme. Health promotion costs were estimated from the cost of the relevant printed materials and visits of staff from city-level facilities to county-level facilities. Infection control costs at county level included the costs of renovating sputum collection rooms and purchasing surgical masks and N95 respirators. The cost of testing a sputum sample in a hospital was calculated from the cost of each sputum sample collected, registered and transported to the hospital for screening times, the number of such samples.
Administration costs involving the Chinese CDC were based on data collected over a year and therefore had to be extrapolated to cover the 24 months of the standard treatment regimen. The costs of project and data management, incentives, infection control, health promotion and training were divided into two categories: variable costs and fixed costs. Variable cost items were multiplied by the total number of treatment days and then divided by the number of days of treatment elapsed. Fixed costs were classified as recurring annual costs or non-recurring, one-time costs. Non-recurring fixed costs for assets with a useful life of more than one year were annualized -assuming that office and laboratory equipment would remain useful for three to five years and that buildings would remain useful for 20 years. We used a discount rate of 3% per year for all costs.
In a probabilistic sensitivity analysis, we allowed total cost to vary by an additional 20% above and below the levels of uncertainty embedded in the unit costs. Cost-effectiveness of a tuberculosis programme, China Christopher Fitzpatrick et al.
We inflated the baseline costs to 2011 prices using a price deflator based on Chinese GDP. 9 We reflected sampling uncertainty in a gamma distribution. Although all costs were initially recorded in Chinese yuan (¥), we report them in US$ for the year 2011, when US$ 1.00 had a mean value of ¥ 6.5.

Modelling and analysis
We constructed mathematical models using R software (R foundation, Vienna, Austria). 10,11 The mc2d package allowed for probabilistic sensitivity analysis of the model parameters along two dimensions. 12 To the y dimension we attached parameter uncertainty -i.e. uncertainty around the final two-year costs and outcomes -and to the x dimension we attached all other -i.e. model and parameter -uncertainty. This two-dimensionality allowed us to assess how uncertainty about final

Research
Cost-effectiveness of a tuberculosis programme, China Christopher Fitzpatrick et al.
costs and outcomes affected the results and to estimate the expected value of so-called partially perfect information. We simulated 500 values along each dimension. We tested for robustness of the results to 10 000 simulations along a single dimension.
We predicted two-year treatment outcomes from the results of culturing sputum samples collected three months after the initiation of treatment. Using a systematic review, we assumed that for negative month three test results, the probability of successful treatment (cure or completion of treatment) is 75%, death is 9%, treatment default is 13% and treatment failure is 3%. 13 We used β distributions for the outcome classes to generate multinomially distributed random number vectors representing the number of patients who were culture-negative at three months and still on treatment at six months. We also generated random number vectors for patients who were culture-positive at three months and still on treatment at six months. We summed these vectors with the known number of deaths and defaulters at six months to obtain the final distributions for the numbers of successes, deaths, defaults and failures at two years.
The decision trees for the baseline and null programmes are available from the corresponding author, as well as descriptions of the long-term mortality parameters -e.g. the probability of relapse and subsequent death of a patient who had been successfully treated -and transmission parameters. Modelling of transmission was based on the duration of the patients' infectious period and decision trees and parameters that were consistent with those used in earlier cost-effectiveness analyses of interven-tions against MDR tuberculosis. 14- 16 We assumed that secondary patients -patients generated by the primary patients -would have the same costs and effects as the primary -i.e. index -cases.

Ethical approval
This economic evaluation is based in part on secondary data published as a result of a pilot study that was reviewed and approved by the Tuberculosis Operational Research Ethics Review Committee of the Chinese Ministry of Health. The collection of the other data that we used satisfied the WHO Ethics Review Committee's conditions for exemption.

Detection and enrolment
Although all 2816 notified smear-positive patients in the four study areas in 2011−2014 should have been screened for rifampicin resistance, data on such screening were only available for 2244 (80%) of them. Overall, 243 (11%) patients tested positive for resistance to rifampicin and 172 of these were enrolled. Subsequent conventional drug susceptibility testing revealed that 112 (65%) of the enrolees carried M. tuberculosis that was resistant to both rifampicin and isoniazid. There were several reasons for non-enrolment, including the presence of clinical complications. 4 In the baseline programme, only 92 patients with MDR tuberculosis were identified over a period of 2.8 years and only 81 (88%) of those patients were started on treatment for MDR tuberculosis.
The characteristics of the enrollees are summarized in Table 2. The propor-tion of enrollees insured under the New Rural Cooperative Medical Scheme exceeded the proportion of MDR tuberculosis patients detected in the baseline programme who were insured under the same scheme. To adjust for this difference in insurance coverage, we weighted the costs so that the estimated costs for the baseline programme were reduced.

Costs
The total cost of the comprehensive programme was US$ 8837 (95% credible interval, CrI: 6649 to 10 962) per patient. Drugs and tests within the standard package made up 23% (US$ 1961) and 17% (US$ 1532) of the total, respectively (Fig. 1). In the baseline programme, medical fees totalled US$ 1235 (95% CrI: 1149 to 1329) per patient. Basic treatment costs were lower in the baseline programme than in the comprehensive programme -in part because so few patients in the baseline programme completed treatment.
In the comprehensive programme, US$ 1506 per patient -i.e. 17% of the total -was spent on incentives and enablers, including variable incentives to village doctors for community-based outpatient treatment, a mix of fixed and variable incentives to hospital workers, variable incentives to the Chinese CDC's county workers, transportation subsidies to patients and fixed incentives to the Chinese CDC's city workers.
In the comprehensive programme, project management costs were considerably higher in the Inner Mongolia autonomous region -where fixed costs were spread over only 17 patients -than in the other three study areas. The same region also faced relatively high transportation costs and these contributed to relatively high screening and health

Effectiveness
In the comprehensive programme, 4% (7) of the enrollees died and 15% (26) defaulted within the first six months. For the 139 enrolees who remained on treatment for six months, 83% (115) had negative sputum cultures by their third month of treatment (Fig. 2). There was no significant increase in negative cultures between the third and sixth months. We did not find a statistically significant difference in outcomes between study areas or patient categories.
In the baseline programme, two (2%) of the 81 patients who started on treatment were still on treatment after six months. The other 66 died or defaulted; six (9%) died within six months of starting treatment.
The predicted two-year success rate for the patients enrolled in the comprehensive programme was 0.563 ( Table 3). The corresponding value for the enrollees who showed resistance to both rifampicin and isoniazid was similar. The success rates predicted for the comprehensive programme were much higher than the corresponding rates for the baseline programme, but similar to those reported in some relatively highincome settings in China. 17,18 When considering only the primary patients that were enrolled, we estimated that the comprehensive programme would avert over their lifetimes an estimated 41 tuberculosis (TB) deaths compared with the baseline programme and 52 TB deaths compared with the null programme (Table 4). Suboptimal enrolment in the comprehensive programme presumably led to some avoidable deaths.
When we extended our predictions to cover primary patients plus any secondary patients infected by primary patients, we found that the comprehensive programme would avert over their lifetimes an estimated 717 TB deaths and 15 800 DALYs compared with the baseline programme and 546 TB deaths and 11 600 DALYs compared with the null programme. There was some indication that, by extending the life of patients but not achieving a cure, the baseline programme might ultimately avert fewer DALYs than the null programme.

Cost-effectiveness
When we included all enrollees in our calculations, we found that, per death averted, the comprehensive programme cost US$ 14 257 compared with the baseline programme and US$ 12 772 compared with the null programme ( Table 4). The corresponding costs per DALY were US$ 715 and US$ 639, respectively. Fig. 3 shows the probability that the comprehensive programme will be very cost-effective at different thresholds of cost-effectiveness. We considered a conservative threshold to be US$ 4615 per DALY averted. This value is roughly equivalent to the GDP per capita in Henan province in 2011 and lower than the GDP per capita for the other three study areas and for mainland China as a whole. 19 The probability that the comprehensive programme is very costeffective -compared with either the baseline programme or the null -was found to be more than 99% under all of the scenarios that we considered.
The uncertainty around the final outcomes had a negligible influence on our main findings (Fig. 3).

Discussion
Our results add to the data on the cost and cost-effectiveness of treatment for MDR tuberculosis. 20 Earlier studies looked at treatment provided free of charge to patients by public and non-profit providers. [14][15][16] The present study included the costs associated with rapid screening and financial protection and considered the effects of such screening and protection on access and adherence. The costs of incentives and non-tuberculosis-specific drugs and tests in our study areas of China were higher than those reported in related studies in other countries -reflecting the unique and complex Chinese system of publicly-owned but profitseeking hospitals.  The comprehensive programme appeared to offer a very cost-effective approach to the control of MDR tuberculosis in all four areas where it was piloted. Early outcomes for the enrollees were much improved and the financial protection from controls on out-of-package fees and higher rates of reimbursement were associated with much better treatment adherence. Aggregate outcomes for the comprehensive programme would have been better if more patients had been enrolled, but some patients may not be able to afford the standard package even with 90% reimbursement. Due to extensive internal migration, coverage of this programme should be expanded to migrants within the study areas and scaled up to cover China as a whole. Given China's large share of the world's drug-resistant tuberculosis burden, implementation of the comprehensive programme across China will help ensure that the world meets its tuberculosis-related targets for the post-2015 development agenda.

Conclusion
The comprehensive programme improved individual outcomes for MDR tuberculosis in a cost-effective manner. To support post-2015 global heath targets, the comprehensive programme should be expanded to non-residents and other areas of China. This programme could also be applicable to other countries with low rates of de-tection and poor treatment outcomes. However, China is a middle-income country with a low prevalence of human immunodeficiency virus (HIV). Further research will be required to assess the cost-effectiveness of this approach in low-income settings with a high prevalence of HIV. ■