Thesis (MScMedSc)--Stellenbosch University, 2011. / ENGLISH ABSTRACT: Introduction: Tuberculosis (TB) treatment registers and laboratory records are essential recording and reporting tools in TB control programmes. Reliable data are essential for any TB control programme but under-registration of TB cases has been well documented internationally, due to under-reporting of patients on treatment or failure to initiate treatment. The accuracy and completeness of routinely collected data are seldom monitored. Aim: This study used record linking to assess the accuracy and completeness of TB treatment register data and the feasibility of estimating the completeness of bacteriological confirmed pulmonary TB registration in two high incident communities in South Africa with capturerecapture methods. Methods: All cases of bacteriologically confirmed TB defined as 2 smear-positive results and/or at least one culture-positive result were included. Record linking was performed between three data sources: (1) TB treatment registers; and (2) all smear and culture results from (a) the nearest central laboratory, and (b) the referral hospital laboratory. To estimate the completeness of TB treatment recording three-source log-linear capture-recapture models were used, with internal validity analysis. Results: The TB treatment registers had 435 TB cases recorded of which 204 (47%) were bacteriologically confirmed cases. An additional 39 cases that were recorded as nonbacteriological cases in the TB treatment register, were reclassified as bacteriologically confirmed. In addition, there were 63 bacteriologically confirmed cases identified from the laboratory databases which were not recorded in the TB treatment register. The final total number of bacteriologically confirmed TB cases across all 3 databases was 306, an increase of 50% over what had initially been recorded in the TB treatment register. The log-linear capture-recapture model estimated the number of bacteriologically confirmed TB cases not found in any of the data sources at 20, resulting in a total number of bacteriologically confirmed TB cases of 326 (95% CI 314-355). The completeness of registration of bacteriologically confirmed pulmonary TB cases was 79% after record linking and 75% after the capture-recapture estimate. Conclusions: The results presented in this thesis highlighted the concern regarding the accuracy and completeness of routinely collected TB recording and reporting data. A high percentage of bacteriologically confirmed cases from both laboratories were not recorded in the TB treatment registers. Capture-recapture can be useful, but not essential, for evaluation of TB control programmes, also in resource-limited settings, but methodology and results should be carefully assessed. The present study estimated the extent of the problem of underreporting of TB in South Africa and identified challenges in the process. Interventions to reduce underreporting of TB are urgently needed. / AFRIKAANSE OPSOMMING: Inleiding: Registers van tuberkulose (TB) behandeling en laboratoriumrekords is noodsaaklike instrumente in die dokumentering van en verslagdoening oor TB beheerprogramme. Betroubare data is onontbeerlik vir enige TB beheerprogram maar onderregistrasie van TB gevalle is internasionaal goed gedokumenteer. Die akkuraatheid en volledigheid van roetine data word selde gemoniteer. Doel: Hierdie studie het rekordkoppeling gebruik om die akkuraatheid en volledigheid van data in TB behandelingsregisters te ondersoek. Voorts is die uitvoerbaarheid van die vangshervangsmetodes vir die beoordeling van die volledigheid van bakteriologies bevestigde pulmonale TB registrasie in twee hoë-insidensie gemeenskappe ondersoek. Metodes: Alle gevalle van bakteriologies bevestigde TB, gedefinieer as 2 smeer-positiewe resultate en/of ten minste een kultuur-positiewe resultaat, is in die studie ingesluit. Rekordkoppeling is onderneem tussen drie databronne: (1) TB behandelingsregisters; en (2) alle smeer- en kultuurpositiewe resultate van (a) die naaste sentrale laboratorium, en (b) die verwysende hospitaallaboratorium. Om die volledigheid van TB behandelingsrekords te ondersoek is drie-bron log-lineêre vangs-hervangs modelle gebruik met interne geldigheidsontleding. Resultate: Die TB registers het 435 aangetekende TB gevalle bevat waarvan 204 (47%) bakteriologies bevestigde gevalle was. 'n Bykomende 39 gevalle wat as nie-bakteriologies bevestigde gevalle aangeteken was in die TB register is hergeklassifiseer as bakteriologies bevestig. Daar is ook 63 bakteriologies bevestigde gevalle geïdentifiseer vanuit die laboratorium databasisse wat nie in die TB register aangeteken was nie. Die finale totale aantal bakteriologies bevestigede TB gevalle oor al drie databasisse heen was 306, 'n toename van 50% in vergelyking met wat aanvanklik in die TB register aangeteken was. Die log-lineêre vangs-hervangs model het die aantal bakteriologies bevestigde gevalle wat nie in enige van die databronne gevind kon word nie as 20 gevalle geskat, wat gelei het tot 'n totaal van 326 (95% VI 314-355) bakteriologies bevestigde gevalle. Die volledigheid van registrasie van bakteriologies bevestigde TB gevalle was 79% na rekordkoppeling en 75% na die vangs-hervangs skatting. Gevolgtrekkings: Die resultate wat in hierdie tesis voorgelê is beklemtoon die besorgdheid oor die akkuraatheid en volledigheid van die aanmelding en optekening van roetine TB data. 'n Hoë persentasie van bakteriologies bevestigde gevalle van beide laboratoriums is nie in die TB register opgeteken nie. Vangs-hervangs kan nuttig wees, maar nie noodsaaklik nie, in die evaluasie van TB beheerprogramme, ook in hulpbron-arm omgewings, maar die metodologie moet omsigtig beoordeel word. Die huidige studie het die omvang van die probleem van onderrapportering van TB in Suid-Afrika beraam en uitdagings in die proses geïdentifiseer. Intervensies om onderrapportering te verminder word dringend benodig.