Tuberculosis is an infectious bacterial disease caused predominantly by the microorganism Mycobacterium tuberculosis. Although the number of deaths from tuberculosis has been falling in recent years, the disease still kills more than 1 million people every year, mainly in developing countries. Tuberculosis can be treated with antibiotics, but the emergence of bacteria that are resistant to existing drugs is threatening efforts to eradicate the disease.

Preventing the spread of tuberculosis is heavily dependent on accurate diagnosis of individuals with the disease. This is challenging because the initial symptoms are often mild, usually just a cough, which means that someone can spread the disease to many others over a period of several months before the symptoms become worse—fever, night sweats, and weight loss—and they realize that they are sick. Multiple diagnostic strategies are available, from the relatively low-tech—examining sputum samples under a microscope to detect tuberculosis bacteria—to more sophisticated tests that can detect bacterial DNA and determine whether the bacteria are drug-resistant in less than 2 hr.

Choosing which diagnostic strategy to adopt can be challenging because the optimal solution in a region will depend on the specific local conditions. To overcome this problem, Dowdy et al. have developed a computer program that enables decision-makers to input four key parameters that describe the tuberculosis situation in their region, and to obtain 5-year projections of the rate of new infections, mortality, and total costs likely to result from adopting any of nine different diagnostic strategies.

The four parameters are the number of new cases of tuberculosis each year (incidence), the proportion of new cases that are multi-drug resistant, the proportion of the adult population that has HIV, and the local costs of various diagnostic techniques and treatments. Since the entire computer program is written in a freely available open-source programming language (Python), any user can tweak these parameters to provide a more precise fit to their own region. Alternatively, the standard version of the program can be run directly from a website without any need to interact with computer code.

This model is the first to enable local decision-makers to evaluate the impact of different diagnostic strategies for tuberculosis under the conditions specific to their region. The model predicts, for example, that in areas where there is little money available for up-front investment, same-day microscopy analysis of sputum samples and starting patients on treatment is the most cost-effective strategy for reducing the rate of new infections. Given the wide variation in conditions within even small geographical areas, this more flexible approach should lead to the more efficient use of resources and may, ultimately, help to reduce the spread of tuberculosis.

Infectious disease transmission models are important tools for translating the best current knowledge of the natural history and epidemiology of infectious diseases into projections of epidemiological impact (e.g., incidence, mortality) and costs under alternative strategies for disease control (Garnett et al., 2011). Currently, most published transmission models are either loosely calibrated to reflect global/regional outcomes or more tightly fit to specific epidemiological settings; in either case, model results may be difficult for local decision-makers in the majority of public health settings to utilize. Simplified models designed for in-country use by decision-makers, most notably the Spectrum suite of models supported by the Futures Institute (Stover et al., 2010), have been used to inform decision-making in the fields of reproductive health and human immunodeficiency virus (HIV) for over a decade (Stover, 2004). Estimates from the Spectrum models are now routinely incorporated into official global and country-level estimates of HIV disease burden (Brown et al., 2010) and intervention impact (Farnham et al., 2013). Other simplified models are readily available for impact projections related to non-infectious diseases, where transmission assumptions are less important (Betz Brown et al., 2000; Walker et al., 2013). However, to date, simple, user-friendly transmission models have not been widely used for decision-making related to many infectious diseases other than HIV. Diagnosis of active tuberculosis (TB) is an example of a public health intervention for which transmission models may provide guidance on both global (Dowdy et al., 2006; Abu-Raddad et al., 2009) and country-specific levels (Menzies et al., 2012). Specifically, an unprecedented number of new diagnostic strategies for active TB are now recommended by the World Health Organization (WHO), including same-day microscopy (World Health Organization, 2011), microcolony-based culture techniques (Leung et al., 2012) such as the microscopic-observation drug-susceptibility (MODS) assay (Moore et al., 2006), line-probe assays for drug susceptibility testing (Bwanga et al., 2009), and Xpert MTB/RIF (‘Xpert’), a molecular assay capable of providing results (including rifampin resistance) in 90 min with minimal human resource requirements (Boehme et al., 2010, 2011). TB program decision-makers must repeatedly determine when to invest in scaling up a novel diagnostic test, which test(s) to promote, and whether the implementation strategy should differ by epidemiological situation (Cobelens et al., 2012). Without transmission models to provide locally relevant estimates of cost and impact under alternative implementation strategies, such decisions will be made without systematically considering the implications of available scientific evidence.

To aid in this decision-making process, we created a flexible, simple modeling tool that allows non-expert users to define their local situation according to three key epidemiological parameters (TB incidence, proportion of new TB cases that are multidrug-resistant [MDR], and adult human immunodeficiency virus [HIV] prevalence) and local unit costs of TB diagnosis and treatment. This tool then incorporates those estimates into a combined decision analysis-transmission framework to generate 5-year projections of TB incidence, mortality, and control costs for nine diagnostic strategies (Figures 1 and 2). These strategies are:

1. ‘Baseline’: Sputum smear microscopy for each diagnostic attempt, with liquid-media TB culture only to evaluate smear-positive cases with a history of previous TB treatment for drug resistance. (Cultures in all scenarios trigger drug-susceptibility testing if positive.)

2. ‘TB culture if previously treated’: Sputum smear microscopy used for patients without a history of TB treatment; smear plus liquid-media culture used to diagnose TB in any previously treated individual with symptoms (regardless of smear status).

3. ‘Xpert if HIV-positive’: Xpert MTB/RIF for HIV-infected patients only, with a positive test for rifampin resistance triggering treatment for MDR-TB. Xpert is assumed to be deployed at the district level, such that results cannot generally be provided during the same clinical encounter (Lawn et al., 2012). This strategy is conceived as a ‘best-case’ scenario for HIV-targeted TB testing: if individuals unaware of their HIV status are not tested with Xpert, this strategy will overestimate effectiveness, and if those unaware of their status are tested, it will underestimate costs.

4. ‘Xpert if Smear-Positive’: Xpert MTB/RIF for smear-positive patients only (i.e., for rapid DST), with a positive test for rifampin resistance triggering treatment for MDR-TB.

5. ‘Xpert for All’: Xpert MTB/RIF for all patients.

6. ‘Xpert with Culture DST Confirmation’: Xpert MTB/RIF for all patients, but treatment for MDR-TB only initiated if rifampin resistance is confirmed by culture.

7. ‘MODS/TLA’: Sputum smear, plus microcolony-based TB culture (e.g., MODS or thin-layer agar, TLA) for all patients.

8. ‘Same-Day Microscopy’: Double the per-test cost of sputum smear microscopy, in exchange for the ability to provide results to patients in the same clinical encounter (e.g., with peripheral, unbatched reading of sputum smears).

9. ‘Same-Day Xpert’: Double the per-test cost of Xpert MTB/RIF, in exchange for the ability to provide results to patients in the same clinical encounter (e.g., peripheral deployment, with greater costs reflecting lower volume per machine [Vassall et al., 2011]).

Figure 1

Download asset Open asset

Figure 2

Download asset Open asset

For purposes of illustration, we evaluated each of these diagnostic strategies in four emblematic epidemiological settings, defined by TB incidence, MDR-TB prevalence among new cases, and adult HIV prevalence:

1. ‘Reference/High-Incidence Setting’ (e.g., Southeast Asia): TB incidence 250 per 100,000/year (twice the global incidence [World Health Organization, 2012]), MDR-TB prevalence of 3.7% in new TB cases (the estimated global prevalence [World Health Organization, 2012]), adult HIV prevalence of 0·83% (the estimated global prevalence [UNAIDS, 2012]);

2. ‘Low-Incidence Setting’ (e.g., United States, Western Europe): TB incidence at entry of 8.9 per 100,000/year and declining, with similar MDR-TB (as a proportion of new cases) and HIV prevalence as above;

3. ‘High MDR Setting’ (e.g., former Soviet Union): TB incidence 100 per 100,000/year, MDR-TB prevalence of 10% in new TB cases, adult HIV prevalence of 0.83%; and

4. ‘High HIV Setting’ (e.g., sub-Saharan Africa): Adult HIV prevalence of 20%, TB incidence 500 per 100,000, and MDR-TB prevalence among new cases of 3.7%.

In each setting, we used a uniform set of costs for purposes of comparison (Table 1). Although these four settings form the basis of the results presented here, decision-makers can use the open-source model program (included as Supplementary file 1, written in the open-source programming language Python, Version 2.7, www.python.org) to re-define any parameter in Table 1 according to their best local knowledge; a manual for doing so is also included as Supplementary file 2. Capacity to create non-equilibrium settings (e.g., declining TB incidence, increasing MDR-TB) is included. We also provide a web-based interface (flexdx.modeltb.org) that allows users to input local TB incidence, MDR-TB prevalence, HIV prevalence, and unit costs; this interface provides customized results without the requirement to manipulate programming code. The program corresponding to the web version is also available on a public repository (https://github.com/JJPennington/FlexDx-TB-Web-Django).

To date, most transmission models of infectious disease control interventions present results that are not directly usable by decision makers because they are not customizable to local conditions. We present a flexible, user-friendly model of TB diagnosis and transmission that allows users without modeling expertise to define an epidemiologic setting (according to TB incidence, MDR-TB prevalence, and HIV prevalence) and unit costs, evaluating various diagnostic strategies in that setting in terms of population-level costs and impact. While this model cannot precisely replicate the epidemiological situation in any given location, it applies a standardized methodology across a wide range of settings, thereby illustrating important interactions between epidemiological parameters and projected impact. We provide both a web interface for rapid calculations and full model code whereby users can change any model parameter for ‘personalized’ sensitivity analysis. Our model results suggest that the rank-ordering of diagnostic strategies may be relatively stable across epidemiological settings, but that the actual population-level costs and impact differ dramatically. This model also serves as an example of how epidemiologists can provide decision-makers across a wide variety of local settings with rapid access to customizable ‘first-pass’ projections of cost and impact from transmission models without the need to construct tightly fitted models to represent all epidemiological settings.

Our results provide important guidance to TB decision-makers. Specifically, in settings where little additional up-front investment is possible (<25% increase in TB control budget), same-day microscopy has the greatest impact on TB incidence with no increase in overall cost at 5 years, provided that same-day microscopy can be feasibly delivered for less than $10 per patient. For settings in which containing MDR-TB is the most important consideration, Xpert for smear-positives has the greatest effectiveness for this outcome, but at substantial price and very little impact on overall TB incidence and mortality. Xpert for HIV-positives and culture for retreatment cases both offer meaningful, albeit small, gains; the cost and impact of these strategies are the highest in HIV-endemic settings. In settings where more initial investment is possible (about 50% increase in TB control budget), broader scale-up of either Xpert or MODS/TLA for all TB diagnosis can offer substantially greater benefits, both in terms of reduced incidence (often 10 times more TB cases averted than with the more narrowly-targeted strategies above) and long-term costs (in that the incremental cost of these strategies declines greatly by Year 5). In general, culture confirmation of positive Xpert results before committing to a course of second-line therapy is preferred. Finally, where the greatest impact is sought, combination of Xpert for all plus infrastructure for same-day diagnosis achieves this aim in all settings, but at the highest cost.

Although this model represents a highly simplified framework, it compares well to other global estimates to which it was not fit (e.g., WHO estimates of TB mortality, previously treated TB, HIV-associated TB, and MDR-TB prevalence among previously treated cases). Its results are also similar to those of other mathematical models of TB diagnosis that are fit to specific locations. For example, Menzies et al. (2012) modeled scale-up of Xpert for people living with HIV in southern Africa, estimating a 6% reduction in incidence, 21% reduction in TB mortality, 25% reduction in MDR-TB incidence that declines to <10% over a longer time frame, and 40% increase in costs (not including costs of antiretroviral therapy). Our corresponding estimates for the Xpert for HIV-positive strategy in the high-HIV setting are 5% reduction in incidence, 25% reduction in TB mortality, 7% reduction in MDR-TB incidence, and increase in costs from 41% (year 1) to 28% (year 5). For purposes of deciding between alternative strategies of scaling up TB diagnostics, the estimates from these two models are likely to provide similar guidance.

For any such simplified transmission model to have an impact on decision-making, it is important to consider who the eventual users of such a model might be, and to develop the model in consultation with those groups. In this case, we identified a number of potential end-user organizations including technical assistance agencies (e.g., KNCV Tuberculosis Foundation), funding bodies (e.g., Global Fund for AIDS, Tuberculosis, and Malaria), non-governmental organizations (e.g., Medecins Sans Frontiers), National Tuberculosis Programs, and regional offices of the World Health Organization. We then invited representatives of these organizations to attend a 1-day workshop (April 2014, The Hague, The Netherlands) describing methods and challenges related to modeling of TB diagnostics in general, including this transmission model in particular. We developed ‘hands-on’ exercises for participants to better learn the model and solicited specific feedback, which is being incorporated into the model structure and web interface in ongoing fashion. We next intend to develop a series of informal ‘case studies’ whereby the use of this model for in-country decision-making can be demonstrated and disseminated.

As with any modeling analysis, this research has important limitations. In order to provide sufficient flexibility and generalizability, we make a number of strong assumptions that include a constant population, homogeneous mixing, no change in parameter values over time, and simplistic incorporation of HIV and drug resistance. Without making such simplifying assumptions, it is impossible to deliver a flexible modeling framework that can generate transparent, customizable, rapid results (i.e., without complex statistical fitting to each individual epidemiological scenario). This model can replicate user-defined TB incidence, MDR-TB prevalence, and HIV incidence but does not describe the breakdown of these values in key subpopulations (e.g., congregate settings or geographical ‘hotspots’), nor does it incorporate operational aspects of the TB diagnostic system in any one setting. Thus, this model does not ameliorate the need for more detailed models in settings where precise estimates are needed; rather it provides access to ‘first-pass’ estimates in settings (i.e., the vast majority of local decision-makers) where such tightly calibrated model projections are not available. This model focuses on transmission dynamics and thus does not include pediatric TB and extrapulmonary TB; these largely non-infectious disease manifestations remain very important components of the TB epidemic that are not captured here. We validate our model against global estimates and other models; ideally, further validation would be performed over time using field data across a wide variety of settings. Our model allows users to define three key epidemiological parameters (as well as other model assumptions within the program), but data to inform even these three parameters—as well as unit costs—are unlikely to be available on sub-national levels. As a result, users wishing to adapt the model to a smaller geographic scale will need to perform additional data-gathering exercises to inform these estimates if they wish to maximize the utility of the model.

Nevertheless, even if high-quality data are not available at the local level, this model allows decision-makers to estimate epidemiological and economic values according to reasonable assumptions (e.g., comparison of TB notification rates or budget line-items to those in the published literature or at the national level), vary those assumptions in real-time, and obtain corresponding projections of comparative impact that incorporate the best available current data on epidemiological or natural history parameters (e.g., TB progression and reactivation). Future efforts might also provide flexibility to specify operational characteristics (e.g., health system capacity) as well.

A final concern is that, by providing users the ability to specify TB incidence, MDR-TB prevalence, and adult HIV prevalence, a number of scenarios can be created (e.g., low TB incidence and very high adult HIV prevalence) that are not epidemiologically realistic. Although there is danger in allowing uninformed users to make projections for such scenarios (and the model will reject or alert users to highly implausible values), we believe that this risk is outweighed by the benefit of providing full flexibility to model epidemiological scenarios (e.g., sub-district level data) that will never be captured by a limited number of closely-calibrated TB transmission models.

In summary, we have created a flexible modeling framework that allows users without modeling expertise to generate simulated populations with locally relevant values for TB incidence, MDR-TB prevalence, adult HIV prevalence, and TB treatment costs. By comparing an array of diagnostic options across emblematic epidemiological scenarios, we provide guidance to decision-makers who seek to ascertain the optimal diagnostic strategy to achieve their selected disease control targets, and to do so using a standardized methodology. Success in the fight against infectious disease generally, and TB specifically, depends on our ability to place global knowledge in the hands of local decision-makers, enabling them to choose those interventions that are likely to have the greatest impact, given existing resources and local epidemiological realities. This flexible modeling framework of diagnostic interventions takes an important step in that direction.

1. 1. Abu-Raddad LJ
   2. Sabatelli L
   3. Achterberg JT
   4. Sugimoto JD
   5. Longini IM Jnr
   6. Dye C
   7. Halloran ME

   (2009) Epidemiological benefits of more-effective tuberculosis vaccines, drugs, and diagnostics

   Proceedings of the National Academy of Sciences of the United States of America 106:13980–13985.

   https://doi.org/10.1073/pnas.0901720106

   • Google Scholar
2. 1. Andrews JR
   2. Noubary F
   3. Walensky RP
   4. Cerda R
   5. Losina E
   6. Horsburgh CR

   (2012) Risk of progression to active tuberculosis following reinfection with Mycobacterium tuberculosis

   Clinical Infectious Diseases 54:784–791.

   https://doi.org/10.1093/cid/cir951

   • Google Scholar
3. 1. Antonucci G
   2. Girardi E
   3. Raviglione MC
   4. Ippolito G

   (1995) Risk factors for tuberculosis in HIV-infected persons. A prospective cohort study. The Gruppo Italiano di Studio Tubercolosi e AIDS (GISTA)

   The Journal of the American Medical Association 274:143–148.

   https://doi.org/10.1001/jama.1995.03530020061033

   • Google Scholar
4. 1. Behr MA
   2. Warren SA
   3. Salamon H
   4. Hopewell PC
   5. Ponce de Leon A
   6. Daley CL
   7. Small PM

   (1999) Transmission of Mycobacterium tuberculosis from patients smear-negative for acid-fast bacilli

   Lancet 353:444–449.

   https://doi.org/10.1016/S0140-6736(98)03406-0

   • Google Scholar
5. 1. Betz Brown J
   2. Russell A
   3. Chan W
   4. Pedula K
   5. Aickin M

   (2000) The global diabetes model: user friendly version 3.0

   Diabetes Research and Clinical Practice 50:S15–S46.

   https://doi.org/10.1016/S0168-8227(00)00215-1

   • Google Scholar
6. 1. Boehme CC
   2. Nabeta P
   3. Hillemann D
   4. Nicol MP
   5. Shenai S
   6. Krapp F
   7. Allen J
   8. Tahirli R
   9. Blakemore R
   10. Rustomjee R
   11. Milovic A
   12. Jones M
   13. O'Brien SM
   14. Persing DH
   15. Ruesch-Gerdes S
   16. Gotuzzo E
   17. Rodrigues C
   18. Alland D
   19. Perkins MD

   (2010) Rapid molecular detection of tuberculosis and rifampin resistance

   The New England Journal of Medicine 363:1005–1015.

   https://doi.org/10.1056/NEJMoa0907847

   • Google Scholar
7. 1. Boehme CC
   2. Nicol MP
   3. Nabeta P
   4. Michael JS
   5. Gotuzzo E
   6. Tahirli R
   7. Gler MT
   8. Blakemore R
   9. Worodria W
   10. Gray C
   11. Huang L
   12. Caceres T
   13. Mehdiyev R
   14. Raymond L
   15. Whitelaw A
   16. Sagadevan K
   17. Alexander H
   18. Albert H
   19. Cobelens F
   20. Cox H
   21. Alland D
   22. Perkins MD

   (2011) Feasibility, diagnostic accuracy, and effectiveness of decentralised use of the Xpert MTB/RIF test for diagnosis of tuberculosis and multidrug resistance: a multicentre implementation study

   Lancet 377:1495–1505.

   https://doi.org/10.1016/S0140-6736(11)60438-8

   • Google Scholar
8. 1. Brown T
   2. Bao L
   3. Raftery AE
   4. Salomon JA
   5. Baggaley RF
   6. Stover J
   7. Gerland P

   (2010) Modelling HIV epidemics in the antiretroviral era: the UNAIDS Estimation and Projection package 2009

   Sexually Transmitted Infections 86:ii3–ii10.

   https://doi.org/10.1136/sti.2010.044784

   • Google Scholar
9. 1. Brownell R
   2. Metcalfe J
   3. Millman AJ
   4. Miller C
   5. Cattamanchi AP

   (2012)

   Performance of Xpert MTB/RIF for diagnosis of pulmonary and extrapulmonary tuberculosis - a systematic review and meta-anaylsis

   American Journal of Respiratory and Critical Care Medicine 185:A4715.

   • Google Scholar
10. 1. Bwanga F
    2. Hoffner S
    3. Haile M
    4. Joloba ML

    (2009) Direct susceptibility testing for multi drug resistant tuberculosis: a meta-analysis

    BMC Infectious Diseases 9:67.

    https://doi.org/10.1186/1471-2334-9-67

    • Google Scholar
11. 1. Cobelens F
    2. van den Hof S
    3. Pai M
    4. Squire SB
    5. Ramsay A
    6. Kimerling ME

    (2012) Which new diagnostics for tuberculosis, and when?

    The Journal of Infectious Diseases 205:S191–S198.

    https://doi.org/10.1093/infdis/jis188

    • Google Scholar
12. 1. Corbett EL
    2. Bandason T
    3. Cheung YB
    4. Munyati S
    5. Godfrey-Faussett P
    6. Hayes R
    7. Churchyard G
    8. Butterworth A
    9. Mason P

    (2007) Epidemiology of tuberculosis in a high HIV prevalence population provided with enhanced diagnosis of symptomatic disease

    PLOS Medicine 4:e22.

    https://doi.org/10.1371/journal.pmed.0040022

    • Google Scholar
13. 1. Corbett EL
    2. Charalambous S
    3. Moloi VM
    4. Fielding K
    5. Grant AD
    6. Dye C
    7. De Cock KM
    8. Hayes RJ
    9. Williams BG
    10. Churchyard GJ

    (2004) Human immunodeficiency virus and the prevalence of undiagnosed tuberculosis in African gold miners

    American Journal of Respiratory and Critical Care Medicine 170:673–679.

    https://doi.org/10.1164/rccm.200405-590OC

    • Google Scholar
14. 1. Corbett EL
    2. Watt CJ
    3. Walker N
    4. Maher D
    5. Williams BG
    6. Raviglione MC
    7. Dye C

    (2003) The growing burden of tuberculosis: global trends and interactions with the HIV epidemic

    Archives of Internal Medicine 163:1009–1021.

    https://doi.org/10.1001/archinte.163.9.1009

    • Google Scholar
15. 1. Cruciani M
    2. Scarparo C
    3. Malena M
    4. Bosco O
    5. Serpelloni G
    6. Mengoli C

    (2004) Meta-analysis of BACTEC MGIT 960 and BACTEC 460 TB, with or without solid media, for detection of mycobacteria

    Journal of Clinical Microbiology 42:2321–2325.

    https://doi.org/10.1128/JCM.42.5.2321-2325.2004

    • Google Scholar
16. 1. Daley CL
    2. Small PM
    3. Schecter GF
    4. Schoolnik GK
    5. McAdam RA
    6. Jacobs WR Jnr
    7. Hopewell PC

    (1992) An outbreak of tuberculosis with accelerated progression among persons infected with the human immunodeficiency virus. An analysis using restriction-fragment-length polymorphisms

    The New England Journal of Medicine 326:231–235.

    https://doi.org/10.1056/NEJM199201233260404

    • Google Scholar
17. 1. Dowdy DW
    2. Basu S
    3. Andrews JR

    (2013) Is passive diagnosis enough? the impact of subclinical disease on diagnostic strategies for tuberculosis

    American Journal of Respiratory and Critical Care Medicine 187:543–551.

    https://doi.org/10.1164/rccm.201207-1217OC

    • Google Scholar
18. 1. Dowdy DW
    2. Chaisson RE
    3. Moulton LH
    4. Dorman SE

    (2006) The potential impact of enhanced diagnostic techniques for tuberculosis driven by HIV: a mathematical model

    AIDS 20:751–762.

    https://doi.org/10.1097/01.aids.0000216376.07185.cc

    • Google Scholar
19. 1. Dowdy DW
    2. Lourenço MC
    3. Cavalcante SC
    4. Saraceni V
    5. King B
    6. Golub JE
    7. Bishai D
    8. Durovni B
    9. Chaisson RE
    10. Dorman SE

    (2008) Impact and cost-effectiveness of culture for diagnosis of tuberculosis in HIV-infected Brazilian adults

    PLOS ONE 3:e4057.

    https://doi.org/10.1371/journal.pone.0004057

    • Google Scholar
20. 1. Dye C
    2. Garnett GP
    3. Sleeman K
    4. Williams BG

    (1998) Prospects for worldwide tuberculosis control under the WHO DOTS strategy

    Lancet 352:1886–1891.

    https://doi.org/10.1016/S0140-6736(98)03199-7

    • Google Scholar
21. 1. Espinal MA
    2. Kim SJ
    3. Suarez PG
    4. Kam KM
    5. Khomenko AG
    6. Migliori GB
    7. Baéz J
    8. Kochi A
    9. Dye C
    10. Raviglione MC

    (2000) Standard short-course chemotherapy for drug-resistant tuberculosis: treatment outcomes in 6 countries

    The Journal of the American Medical Association 283:2537–2545.

    https://doi.org/10.1001/jama.283.19.2537

    • Google Scholar
22. 1. Farnham PG
    2. Holtgrave DR
    3. Gopalappa C
    4. Hutchinson AB
    5. Sansom SL

    (2013) Lifetime costs and quality-adjusted life years saved from HIV prevention in the test and treat era

    Journal of Acquired Immune Deficiency Syndromes 64:e15–e18.

    https://doi.org/10.1097/QAI.0b013e3182a5c8d4

    • Google Scholar
23. 1. Garnett GP
    2. Cousens S
    3. Hallett TB
    4. Steketee R
    5. Walker N

    (2011) Mathematical models in the evaluation of health programmes

    Lancet 378:515–525.

    https://doi.org/10.1016/S0140-6736(10)61505-X

    • Google Scholar
24. 1. Getahun H
    2. Harrington M
    3. O'Brien R
    4. Nunn P

    (2007) Diagnosis of smear-negative pulmonary tuberculosis in people with HIV infection or AIDS in resource-constrained settings: informing urgent policy changes

    Lancet 369:2042–2049.

    https://doi.org/10.1016/S0140-6736(07)60284-0

    • Google Scholar
25. 1. Horsburgh CR Jnr
    2. O'Donnell M
    3. Chamblee S
    4. Moreland JL
    5. Johnson J
    6. Marsh BJ
    7. Narita M
    8. Johnson LS
    9. von Reyn CF

    (2010) Revisiting rates of reactivation tuberculosis: a population-based approach

    American Journal of Respiratory and Critical Care Medicine 182:420–425.

    https://doi.org/10.1164/rccm.200909-1355OC

    • Google Scholar
26. 1. Kendall MG

    (1942) Partial rank correlation

    Biometrika 32:277–283.

    https://doi.org/10.1093/biomet/32.3-4.277

    • Google Scholar
27. 1. Lawn SD
    2. Kerkhoff AD
    3. Wood R

    (2012) Location of Xpert(R) MTB/RIF in centralised laboratories in South Africa undermines potential impact

    The International Journal of Tuberculosis and Lung Disease 16:701.

    https://doi.org/10.5588/ijtld.12.0131

    • Google Scholar
28. 1. Leung E
    2. Minion J
    3. Benedetti A
    4. Pai M
    5. Menzies D

    (2012) Microcolony culture techniques for tuberculosis diagnosis: a systematic review

    The International Journal of Tuberculosis and Lung Disease 16:16–23.

    https://doi.org/10.5588/ijtld.10.0065

    • Google Scholar
29. 1. Lew W
    2. Pai M
    3. Oxlade O
    4. Martin D
    5. Menzies D

    (2008) Initial drug resistance and tuberculosis treatment outcomes: systematic review and meta-analysis

    Annals of Internal Medicine 149:123–134.

    https://doi.org/10.7326/0003-4819-149-2-200807150-00008

    • Google Scholar
30. 1. MacPherson P
    2. Houben RM
    3. Glynn JR
    4. Corbett EL
    5. Kranzer K

    (2014) Pre-treatment loss to follow-up in low- and lower-middle-income countries and high-burden countries: a systematic review and meta-analysis

    Bulletin of the World Health Organization 92:126–138.

    https://doi.org/10.2471/BLT.13.124800

    • Google Scholar
31. 1. McKay MD
    2. Beckman RJ
    3. Conover WJ

    (1979)

    Comparison of three methods for selecting values of input variables in the analysis of output from a computer code

    Technometrics 21:239–245.

    • Google Scholar
32. 1. Menzies D
    2. Benedetti A
    3. Paydar A
    4. Martin I
    5. Royce S
    6. Pai M
    7. Vernon A
    8. Lienhardt C
    9. Burman W

    (2009a) Effect of duration and intermittency of rifampin on tuberculosis treatment outcomes: a systematic review and meta-analysis

    PLOS Medicine 6:e1000146.

    https://doi.org/10.1371/journal.pmed.1000146

    • Google Scholar
33. 1. Menzies D
    2. Benedetti A
    3. Paydar A
    4. Royce S
    5. Madhukar P
    6. Burman W
    7. Vernon A
    8. Lienhardt C

    (2009b) Standardized treatment of active tuberculosis in patients with previous treatment and/or with mono-resistance to isoniazid: a systematic review and meta-analysis

    PLOS Medicine 6:e1000150.

    https://doi.org/10.1371/journal.pmed.1000150

    • Google Scholar
34. 1. Menzies NA
    2. Cohen T
    3. Lin HH
    4. Murray M
    5. Salomon JA

    (2012) Population health impact and cost-effectiveness of tuberculosis diagnosis with xpert MTB/RIF: a dynamic simulation and economic evaluation

    PLOS Medicine 9:e1001347.

    https://doi.org/10.1371/journal.pmed.1001347

    • Google Scholar
35. 1. Minion J
    2. Leung E
    3. Menzies D
    4. Pai M

    (2010) Microscopic-observation drug susceptibility and thin layer agar assays for the detection of drug resistant tuberculosis: a systematic review and meta-analysis

    The Lancet Infectious Diseases 10:688–698.

    https://doi.org/10.1016/S1473-3099(10)70165-1

    • Google Scholar
36. 1. Moore DAJ
    2. Evans CAW
    3. Gilman RH
    4. Caviedes L
    5. Coronel J
    6. Vivar A
    7. Sanchez E
    8. Piñedo Y
    9. Saravia JC
    10. Salazar C
    11. Oberhelman R
    12. Hollm-Delgado MG
    13. LaChira D
    14. Escombe AR
    15. Friedland JS

    (2006) Microscopic-observation drug-susceptibility assay for the diagnosis of TB

    The New England Journal of Medicine 355:1539–1550.

    https://doi.org/10.1056/NEJMoa055524

    • Google Scholar
37. 1. Solari L
    2. Gutierrez A
    3. Suarez C
    4. Jave O
    5. Castillo E
    6. Yale G
    7. Ascencios L
    8. Quispe N
    9. Valencia E
    10. Suárez V

    (2011) Cost analysis of rapid methods for diagnosis of multidrug resistant tuberculosis in different epidemiologic groups in Peru

    Revista Peruana de Medicina Experimental y Salud Publica 28:426–431.

    https://doi.org/10.1590/S1726-46342011000300004

    • Google Scholar
38. 1. Sreeramareddy CT
    2. Panduru KV
    3. Menten J
    4. Van den Ende J

    (2009) Time delays in diagnosis of pulmonary tuberculosis: a systematic review of literature

    BMC Infectious Diseases 9:91.

    https://doi.org/10.1186/1471-2334-9-91

    • Google Scholar
39. 1. Steingart KR
    2. Henry M
    3. Ng V
    4. Hopewell PC
    5. Ramsay A
    6. Cunningham J
    7. Urbanczik R
    8. Perkins M
    9. Aziz MA
    10. Pai M

    (2006a) Fluorescence versus conventional sputum smear microscopy for tuberculosis: a systematic review

    The Lancet Infectious Diseases 6:570–581.

    https://doi.org/10.1016/S1473-3099(06)70578-3

    • Google Scholar
40. 1. Steingart KR
    2. Ng V
    3. Henry M
    4. Hopewell PC
    5. Ramsay A
    6. Cunningham J
    7. Urbanczik R
    8. Perkins MD
    9. Aziz MA
    10. Pai M

    (2006b) Sputum processing methods to improve the sensitivity of smear microscopy for tuberculosis: a systematic review

    The Lancet Infectious Diseases 6:664–674.

    https://doi.org/10.1016/S1473-3099(06)70602-8

    • Google Scholar
41. 1. Storla DG
    2. Yimer S
    3. Bjune GA

    (2008) A systematic review of delay in the diagnosis and treatment of tuberculosis

    BMC Public Health 8:15.

    https://doi.org/10.1186/1471-2458-8-15

    • Google Scholar
42. 1. Stover J

    (2004) Projecting the demographic consequences of adult HIV prevalence trends: the Spectrum Projection Package

    Sexually Transmitted Infections 80:i14–i18.

    https://doi.org/10.1136/sti.2004.010157

    • Google Scholar
43. 1. Stover J
    2. MacKinnon R
    3. Winfrey B

    (2010) Spectrum: a model platform for linking maternal and child survival interventions with AIDS, family planning and demographic projections

    International Journal of Epidemiology 39:i7–i10.

    https://doi.org/10.1093/ije/dyq016

    • Google Scholar
44. 1. Tiemersma EW
    2. van der Werf MJ
    3. Borgdorff MW
    4. Williams BG
    5. Nagelkerke NJ

    (2011) Natural history of tuberculosis: duration and fatality of untreated pulmonary tuberculosis in HIV negative patients: a systematic review

    PLOS ONE 6:e17601.

    https://doi.org/10.1371/journal.pone.0017601

    • Google Scholar
45. Book

    1. UNAIDS

    (2012)

    UNAIDS report on the global AIDS epidemic 2012

    Geneva: UNAIDS.

    • Google Scholar
46. 1. Vassall A
    2. van Kampen S
    3. Sohn H
    4. Michael JS
    5. John KR
    6. den Boon S
    7. Davis JL
    8. Whitelaw A
    9. Nicol MP
    10. Gler MT
    11. Khaliqov A
    12. Zamudio C
    13. Perkins MD
    14. Boehme CC
    15. Cobelens F

    (2011) Rapid diagnosis of tuberculosis with the xpert MTB/RIF assay in high burden countries: a cost-effectiveness analysis

    PLOS Medicine 8:e1001120.

    https://doi.org/10.1371/journal.pmed.1001120

    • Google Scholar
47. 1. Vynnycky E
    2. Fine PEM

    (1997) The natural history of tuberculosis: the implications of age-dependent risks of disease and the role of reinfection

    Epidemiology and Infection 119:183–201.

    https://doi.org/10.1017/S0950268897007917

    • Google Scholar
48. 1. Walker N
    2. Tam Y
    3. Friberg IK

    (2013) Overview of the Lives Saved Tool (LiST)

    BMC Public Health 13:S1.

    https://doi.org/10.1186/1471-2458-13-S3-S1

    • Google Scholar
49. 1. Wilkinson D
    2. Newman W
    3. Reid A
    4. Squire SB
    5. Sturm AW
    6. Gilks CF

    (2000)

    Trial-of-antibiotic algorithm for the diagnosis of tuberculosis in a district hospital in a developing country with high HIV prevalence

    The International Journal of Tuberculosis and Lung Disease 4:513–518.

    • Google Scholar
50. 1. Williams BG
    2. Granich R
    3. De Cock KM
    4. Glaziou P
    5. Sharma A
    6. Dye C

    (2010) Antiretroviral therapy for tuberculosis control in nine African countries

    Proceedings of the National Academy of Sciences of the United States of America 107:19485–19489.

    https://doi.org/10.1073/pnas.1005660107

    • Google Scholar
51. 1. Wood R
    2. Middelkoop K
    3. Myer L
    4. Grant AD
    5. Whitelaw A
    6. Lawn SD
    7. Kaplan G
    8. Huebner R
    9. McIntyre J
    10. Bekker LG

    (2007) Undiagnosed tuberculosis in a community with high HIV prevalence: implications for tuberculosis control

    American Journal of Respiratory and Critical Care Medicine 175:87–93.

    https://doi.org/10.1164/rccm.200606-759OC

    • Google Scholar
52. 1. World Bank

    (2012) Life expectancy

    http://www.google.com/publicdata/explore?ds=d5bncppjof8f9_&ctype=l&strail=false&nselm=h&hl=en&dl=en#!ctype=l&strail=false&bcs=d&nselm=h&met_y=sp_dyn_le00_in&scale_y=lin&ind_y=false&rdim=region&ifdim=region&tdim=true&hl=en_US&dl=en&ind=false, Accessed July 11.

    http://www.google.com/publicdata/explore?ds=d5bncppjof8f9_%26ctype=l%26strail=false%26nselm=h%26hl=en%26dl=en#!ctype=l%26strail=false%26bcs=d%26nselm=h%26met_y=sp_dyn_le00_in%26scale_y=lin%26ind_y=false%26rdim=region%26ifdim=region%26tdim=true%26hl=en_US%26dl=en%26ind=false

    • Google Scholar
53. Book

    1. World Health Organization

    (2010)

    Multidrug and extensively drug-resistant TB (M/XDR-TB): 2010 global report on surveillance and response

    Geneva: WHO.

    • Google Scholar
54. Book

    1. World Health Organization

    (2011)

    Same-day diagnosis of tuberculosis by microscopy: policy statement

    Geneva: WHO.

    • Google Scholar
55. Book

    1. World Health Organization

    (2012)

    Global tuberculosis report 2012

    Geneva: WHO.

    • Google Scholar

Author details

1. David W Dowdy

   1. Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States
   2. Center for Tuberculosis Research, Johns Hopkins University, Baltimore, United States

   Contribution

   DWD, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   ddowdy1@jhmi.edu

   Competing interests

   The authors declare that no competing interests exist.

2. Jason R Andrews

   Division of Infectious Diseases, Department of Medicine, Massachusetts General Hospital, Boston, United States

   Present address

   Division of Infectious Diseases, Department of Medicine, Stanford University School of Medicine, Palo Alto, United States

   Contribution

   JRA, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

3. Peter J Dodd

   School of Health and Related Research, University of Sheffield, Sheffield, United Kingdom

   Contribution

   PJD, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

4. Robert H Gilman

   Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, United States

   Contribution

   RHG, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

• 1,897

  views

• 160

  downloads

• 13

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.