Spatially-targeted digital chest radiography-based screening to reduce tuberculosis in high-burden settings : simulation of an adaptive decision-making approach and cost-effectiveness analysis

Date
2022-04
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH SUMMARY: Background. Interventions to actively detect individuals with tuberculosis (TB) and to enable timely treatment could accelerate TB control. Spatially-targeted interventions that concentrate screening within geographic “hotspots” of tuberculosis incidence, could lead to more effective allocation of already limited resources. However, identifying and prioritizing areas with the highest burdens of undetected TB is challenging. The aim of this thesis was to investigate the effectiveness and cost-effectiveness of an adaptive decision-making approach for spatially-targeted, community-based TB screening in high burden settings. Towards this principal aim, there were three objectives: (1) to develop a Monte-Carlo simulation model of a hypothetical digital chest radiography (dCXR)-based TB screening intervention and then to use the simulation to (2) project, in 24 high TB burden communities in South Africa and Zambia, the case-finding yield under the adaptive approach compared to untargeted (random) allocation of screening resources and to (3) investigate the cost-effectiveness of the adaptive approach relative to random and notification based allocation in 12 high burden communities in metropolitan Cape Town, South Africa. Methods. A probabilistic simulation model to simulate a TB screening intervention with TB prevalence estimates derived from a large community-randomized trial was developed. A hypothetical scenario of TB screening was assumed under which mobile screening units were allocated among communities during a 52-week period. A Thompson sampling algorithm was implemented to adaptively allocate screening units based on Bayesian probabilities of local TB prevalence that are continuously updated during weekly screening rounds. The simulation was used to estimate and compare yields of bacteriologically-confirmed TB patients detected per 1,000 screenings performed. Thereafter, the simulation was extended to estimate costs and disability-adjusted life years averted (DALYs). Results. Random allocation of four screening units among the 24 communities would result in an expected 665 (95% uncertainty interval 523-819) TB cases detected over one year, equivalent to 8.9 (7.5-10.4) per 1,000 screened. Spatially-targeted allocation informed by the adaptive decision-making approach would increase this yield. Balanced, adaptive allocation resulted in an expected 1,234 (983-1,487) TB cases detected, 16.5 (14.5-19.0) per 1,000 screened. Numbers of dCXR-based screenings to find one additional TB case declined during the first 12-14 weeks as a result of Bayesian learning. Random-allocation of three screening units among the 12 communities was estimated to avert 1,523(980 – 2,181) DALYs at a cost of $216 ($149 – $321) per DALY averted. Balanced, adaptive allocation could yield an additional 19% of DALYs at an incremental cost of $61 ($24 – $177) per additional DALY averted relative to random allocation. However, this incremental cost per DALY averted increased with increasing costs incurred to adaptively relocate screening units among communities. Conclusion. An approach for spatially-targeted TB screening is proposed that could reduce the number of dCXR screenings necessary to detect additional TB cases in high-burden settings. Furthermore, the approach could result in additional health benefits at relatively low additional cost. However, the extent to which this approach is cost-effective and feasible depends on the additional logistics and costs incurred to relocate the screening units. Empirical research is needed to determine whether this approach could be successfully implemented.
AFRIKAANSE OPSOMMING: Agtergrond. Intervensies om individue met aktiewe tuberkulose (TB) op te spoor om tydige behandeling moontlik te maak, kan TB-beheer versnel. Ruimtelik-geteikende intervensies wat sifting binne geografiese "brandpunte" van tuberkulose insidensiekonsentreer, kan lei tot meer effektiewe aanwysing van reeds beperkte hulpbronne. Die identifisering en prioritisering van gebiede met die hoogste prevalensie van onopgemerkte TB is egter uitdagend. Die doel van hierdie tesis was om die doeltreffendheid en koste-effektiwiteit van 'n aanpasbare besluitnemingsbenadering vir ruimtelik-geteikende, gemeenskapsgebaseerde TB-sifting in hoe-prevalensie -omgewings te ondersoek. Met betrekking tot hierdie hoofdoel was daar drie doelwitte: (1) om 'n Monte-Carlo simulasiemodel van 'n hipotetiese digitale borskasradiografie (dCXR)-gebaseerde TB-siftingsintervensie te ontwikkel en dan die simulasie te gebruik om (2) in 24 hoe TB prevalensie gemeenskappe in Suid-Afrika en Zambie die gevalle-opbrengs te projekteer onder die aanpasbare benadering in vergelyking met ongeteikende (ewekansige) toewysing van siftingshulpbronne en om (3) in 12 hoe prevalensie-gemeenskappe in die Kaapstadse metropool, Suid-Afrika, die koste-effektiwiteit van die aanpasbare benadering relatief tot ewekansige en diagnose gebaseerde toewysing te ondersoek. Metodes. 'n Waarskynlikheidssimulasiemodel is ontwikkel om 'n TB-siftingsintervensie te simuleer met TB-prevalensieskattings afgelei van 'n groot gemeenskap-verewekansigde proef. 'n Hipotetiese scenario van TB-sifting is aangeneem waarvolgens mobiele siftingseenhede gedurende 'n tydperk van 52 weke onder gemeenskappe verdeel is. 'n Thompson-steekproefalgoritme is geïmplementeer om siftingseenhede aanpasbaar toe te ken gebaseer op Bayesiese waarskynlikhede van plaaslike TB-prevalensie wat deurlopend tydens weeklikse siftingsrondtes bygewerk word. Die simulasie is gebruik om opbrengste van bakteriologies-bevestigde TB-pasiente opgespoor per 1 000 siftingstoetse wat uitgevoer is, te skat en te vergelyk. Daarna is die simulasie uitgebrei om koste en ongeskiktheids-aangepaste lewensjare afgeweer (DALY's) te skat. Resultate Die ewekansige toewysing van vier siftingseenhede tussen die 24 gemeenskappe sal 'n verwagte 665 (95% onsekerheidsinterval 523-819) TB-gevalle oor een jaar opspoor, gelykstaande aan 8,9 (7,5-10,4) per 1 000 siftingstoetse. Ruimtelik-geteikende toekenning ingelig deur die aanpasbare besluitnemingsbenadering sal hierdie opbrengs verhoog. Gebalanseerde, aanpasbare toekenning het gelei tot 'n verwagte 1 234 (983-1 487) TB-gevalle wat opgespoor is, 16,5 (14,5-19,0) per 1 000 siftingstoetse. Die aantal dCXR-gebaseerde siftings om een bykomende TB-geval te vind, het gedurende die eerste 12-14 weke afgeneem as gevolg van Bayesiaanse leer. Die ewekansige toekenning van die drie siftingseenhede tussen die 12 gemeenskappe was geskat om 1,523 (980 – 2,181) DALY’s vermy teen ‘n koste van $216 ($149 - $321) per bykomende DALY. Gebalanseerde, aanpasbare toekenning kan 'n bykomende 19% van DALY's oplewer teen 'n inkrementele koste van $61 ($24 - $177) per bykomende DALY wat afgeweer word relatief tot ewekansige toekenning. Hierdie inkrementele koste per DALY wat afgeweer is, het egter toegeneem met toenemende koste wat aangegaan is om siftingseenhede aanpasbaar tussen gemeenskappe te hervestig. Afsluiting. 'n Benadering vir ruimtelik-geteikende TB-sifting wat kan lei tot die vermindering van die aantal dCXR-siftingtoetse wat nodig is om bykomende TB-gevalle in hoe-prevalensie gebiede op te spoor, is voorgestel. Verder kan die benadering tot bykomende gesondheidsvoordele teen relatief lae bykomende koste lei. Die mate waarin hierdie benadering kostedoeltreffend en uitvoerbaar is, hang egter af van die bykomende logistiek en kostes wat aangegaan word om die siftingseenhede te hervestig. Empiriese navorsing is nodig om te bepaal of hierdie benadering suksesvol geimplementeer kan word.
Description
Thesis (MSc)--Stellenbosch University, 2022.
Keywords
Citation