INT J TUBERC LUNG DIS 19(3):288–294 Q 2015 The Union http://dx.doi.org/10.5588/ijtld.14.0317 An approach to estimating tuberculosis incidence and case detection rate from routine notification data K. K. Avilov,* † A. A. Romanyukha,* ‡§ S. E. Borisov, ¶ E. M. Belilovsky, ¶ O. B. Nechaeva, † A. S. Karkach* *Institute for Numerical Mathematics, Russian Academy of Sciences, Moscow, † Federal Research Institute for Health Organization and Informatics, Ministry of Health of the Russian Federation, Moscow, ‡ Lomonosov Moscow State University, Moscow, § Moscow Institute of Physics and Technology (State University), Moscow, ¶ Moscow Scientific and Practical Center for Tuberculosis Control, Moscow Health Department, Moscow, Russia SUMMARY OBJECTIVE: To estimate tuberculosis (TB) incidence and case detection rate (CDR) using routine TB surveillance data only. METHODS: A mathematical model of the case detection process, representing competition between disease pro- gression and case finding, is proposed. The model describes disease progression as a two-stage process (bacillary and non-bacillary TB), and so relates the proportion of bacillary TB cases on detection to the effectiveness of detection. Thus, given the annual numbers of newly detected TB cases stratified by bacillary status, the model estimates detection rates, incidence and CDR. Routine notification data from eight provinces in Russia, 2000–2011, were used for the study. RESULTS: Subnational level estimates of incidence and CDR were obtained. Incidence estimates varied by two- fold among the provinces; corrected CDR estimates varied by 1.5 times. The trend in the incidence estimates was similar to that in the World Health Organization estimates for the whole of Russia. The change in the trend in WHO CDR estimates in 2008–2009 was not supported by our estimates. CONCLUSION: The general approach that uses multi- stage models of disease progression and accordingly stratified notification data can be applied in various settings for the routine estimation of incidence and CDR. KEY WORDS: mathematical modelling; CDR; active and passive case detection; mathematical epidemiology; multistage model of disease progression TUBERCULOSIS (TB) incidence is a key indicator both for the assessment of the epidemiological situation as well as for the calculation of the TB case detection rate (CDR), the latter being one of the target indicators of the United Nations Millennium Development Goals and the World Health Organi- zation (WHO) Stop TB Strategy. 1,2 In many settings, case notification cannot be deemed complete; the number of TB cases notified to a national tubercu- losis programme (NTP) is therefore lower than the number of incident cases in the population. Sites with an inadequate case-finding system may show significant differences between notification and incidence. 3 The direct measurement of TB incidence by a prospective cohort study is possible, but is prohibitively complicated both logistically and financially. Indirect estimations of TB incidence are therefore used. There are a number of approaches to estimating TB incidence. First, using ‘Styblo’s rule, an estimated TB incidence may be calculated from the annual risk of tuberculous infection (ARI). 4 This method, which involves the derivation of ARI from the prevalence of tuberculous infection from a tuberculin survey and the mean age of the sampled patients (usually children), has been in use for some time; however, in its 2009 impact measurement document the WHO considered it unreliable. 5,6 Second, TB incidence can be derived from active TB prevalence and the average duration of TB disease. This method requires one or, preferably, several consecutive prevalence surveys, which may be costly. Third, TB incidence may be estimated from the number of TB-associated deaths among non-notified TB cases and an estimate of the death rate of untreated TB patients. This approach may be inaccurate because it depends upon the quality of ascertaining cause of death that is poor in some regions. Fourth, capture-recapture analysis of multiregister data may yield an estimate of TB incidence; however, this requires two or three semi- Correspondence to: Konstantin K Avilov, Institute for Numerical Mathematics, Russian Academy of Sciences, 8 Gubkina str, Moscow, Russia, 119333. Tel: ( þ 7) 495 984 81 20. Fax: ( þ 7) 495 989 80 23. e-mail: kkavilov@gmail.com Article submitted 21 April 2014. Final version accepted 29 October 2014. //titan/production/j/jtld/live_jobs/jtld-19-03/jtld-19-03-16/layouts/jtld-19-03-16.3d  22 January 2015  11:20 pm  Allen Press, Inc. Page 288