Use of spatial information to predict multidrug resistance in tuberculosis patients, Peru

Predicting resistance to fluoroquinolones among patients with rifampicin-resistant tuberculosis using machine learning methods

Background

Limited access to drug-susceptibility tests (DSTs) and delays in receiving DST results are challenges for timely and appropriate treatment of multi-drug resistant tuberculosis (TB) in many low-resource settings. We investigated whether data collected as part of routine, national TB surveillance could be used to develop predictive models to identify additional resistance to fluoroquinolones (FLQs), a critical second-line class of anti-TB agents, at the time of diagnosis with rifampin-resistant TB.

Methods and findings

We assessed three machine learning-based models (logistic regression, neural network, and random forest) using information from 540 patients with rifampicin-resistant TB, diagnosed using Xpert MTB/RIF and notified in the Republic of Moldova between January 2018 and December 2019. The models were trained to predict the resistance to FLQs based on demographic and TB clinical information of patients and the estimated district-level prevalence of resistance to FLQs. We compared these models based on the optimism-corrected area under the receiver operating characteristic curve (OC-AUC-ROC). The OC-AUC-ROC of all models were statistically greater than 0.5. The neural network model, which utilizes twelve features, performed best and had an estimated OC-AUC-ROC of 0.87 (0.83,0.91), which suggests reasonable discriminatory power. A limitation of our study is that our models are based only on data from the Republic of Moldova and since not externally validated, the generalizability of these models to other populations remains unknown.

Conclusions

Models trained on data from phenotypic surveillance of drug-resistant TB can predict resistance to FLQs based on patient characteristics at the time of diagnosis with rifampin-resistant TB using Xpert MTB/RIF, and information about the local prevalence of resistance to FLQs. These models may be useful for informing the selection of antibiotics while awaiting results of DSTs.

Spatio-temporal co-occurrence of hotspots of tuberculosis, poverty and air pollution in Lima, Peru

Growing evidence suggests pollution and other environmental factors have a role in the development of tuberculosis (TB), however, such studies have never been conducted in Peru. Considering the association between air pollution and specific geographic areas, our objective was to determine the spatial distribution and clustering of TB incident cases in Lima and their co-occurrence with clusters of fine particulate matter (PM2.5) and poverty. We found co-occurrences of clusters of elevated concentrations of air pollutants such as PM2.5, high poverty indexes, and high TB incidence in Lima. These findings suggest an interplay of socio-economic and environmental in driving TB incidence.

Methods used in the spatial analysis of tuberculosis epidemiology: a systematic review

BACKGROUND

Tuberculosis (TB) transmission often occurs within a household or community, leading to heterogeneous spatial patterns. However, apparent spatial clustering of TB could reflect ongoing transmission or co-location of risk factors and can vary considerably depending on the type of data available, the analysis methods employed and the dynamics of the underlying population. Thus, we aimed to review methodological approaches used in the spatial analysis of TB burden.

METHODS

We conducted a systematic literature search of spatial studies of TB published in English using Medline, Embase, PsycInfo, Scopus and Web of Science databases with no date restriction from inception to 15 February 2017. The protocol for this systematic review was prospectively registered with PROSPERO ( CRD42016036655 ).

RESULTS

We identified 168 eligible studies with spatial methods used to describe the spatial distribution (n = 154), spatial clusters (n = 73), predictors of spatial patterns (n = 64), the role of congregate settings (n = 3) and the household (n = 2) on TB transmission. Molecular techniques combined with geospatial methods were used by 25 studies to compare the role of transmission to reactivation as a driver of TB spatial distribution, finding that geospatial hotspots are not necessarily areas of recent transmission. Almost all studies used notification data for spatial analysis (161 of 168), although none accounted for undetected cases. The most common data visualisation technique was notification rate mapping, and the use of smoothing techniques was uncommon. Spatial clusters were identified using a range of methods, with the most commonly employed being Kulldorff's spatial scan statistic followed by local Moran's I and Getis and Ord's local Gi(d) tests. In the 11 papers that compared two such methods using a single dataset, the clustering patterns identified were often inconsistent. Classical regression models that did not account for spatial dependence were commonly used to predict spatial TB risk. In all included studies, TB showed a heterogeneous spatial pattern at each geographic resolution level examined.

CONCLUSIONS

A range of spatial analysis methodologies has been employed in divergent contexts, with all studies demonstrating significant heterogeneity in spatial TB distribution. Future studies are needed to define the optimal method for each context and should account for unreported cases when using notification data where possible. Future studies combining genotypic and geospatial techniques with epidemiologically linked cases have the potential to provide further insights and improve TB control.

Where is tuberculosis transmission happening? Insights from the literature, new tools to study transmission and implications for the elimination of tuberculosis

More than 10 million new cases of tuberculosis (TB) are diagnosed worldwide each year. The majority of these cases occur in low- and middle-income countries where the TB epidemic is predominantly driven by transmission. Efforts to 'end TB' will depend upon our ability to halt ongoing transmission. However, recent studies of new approaches to interrupt transmission have demonstrated inconsistent effects on reducing population-level TB incidence. TB transmission occurs across a wide range of settings, that include households and hospitals, but also community-based settings. While home-based contact investigations and infection control programmes in hospitals and clinics have a successful track record as TB control activities, there is a gap in our knowledge of where, and between whom, community-based transmission of TB occurs. Novel tools, including molecular epidemiology, geospatial analyses and ventilation studies, provide hope for improving our understanding of transmission in countries where the burden of TB is greatest. By integrating these diverse and innovative tools, we can enhance our ability to identify transmission events by documenting the opportunity for transmission-through either an epidemiologic or geospatial connection-alongside genomic evidence for transmission, based upon genetically similar TB strains. A greater understanding of locations and patterns of transmission will translate into meaningful improvements in our current TB control activities by informing targeted, evidence-based public health interventions.

Spatial patterns of multidrug resistant tuberculosis and relationships to socio-economic, demographic and household factors in northwest Ethiopia

Background

Understanding the geographical distribution of multidrug-resistant tuberculosis (MDR-TB) in high TB burden countries such as Ethiopia is crucial for effective control of TB epidemics in these countries, and thus globally. We present the first spatial analysis of multidrug resistant tuberculosis, and its relationship to socio-economic, demographic and household factors in northwest Ethiopia.

Methods

An ecological study was conducted using data on patients diagnosed with MDR-TB at the University of Gondar Hospital MDR-TB treatment centre, for the period 2010 to 2015. District level population data were extracted from the Ethiopia National and Regional Census Report. Spatial autocorrelation was explored using Moran’s I statistic, Local Indicators of Spatial Association (LISA), and the Getis-Ord statistics. A multivariate Poisson regression model was developed with a conditional autoregressive (CAR) prior structure, and with posterior parameters estimated using a Bayesian Markov chain Monte Carlo (MCMC) simulation approach with Gibbs sampling, in WinBUGS.

Results

A total of 264 MDR-TB patients were included in the analysis. The overall crude incidence rate of MDR-TB for the six-year period was 3.0 cases per 100,000 population. The highest incidence rate was observed in Metema (21 cases per 100,000 population) and Humera (18 cases per 100,000 population) districts; whereas nine districts had zero cases. Spatial clustering of MDR-TB was observed in districts located in the Ethiopia-Sudan and Ethiopia-Eritrea border regions, where large numbers of seasonal migrants live. Spatial clustering of MDR-TB was positively associated with urbanization (RR: 1.02; 95%CI: 1.01, 1.04) and the percentage of men (RR: 1.58; 95% CI: 1.26, 1.99) in the districts; after accounting for these factors there was no residual spatial clustering.

Conclusion

Spatial clustering of MDR-TB, fully explained by demographic factors (urbanization and percent male), was detected in the border regions of northwest Ethiopia, in locations where seasonal migrants live and work. Cross-border initiatives including options for mobile TB treatment and follow up are important for the effective control of MDR-TB in the region.

Identifying Hotspots of Multidrug-Resistant Tuberculosis Transmission Using Spatial and Molecular Genetic Data

BACKGROUND

We aimed to identify and determine the etiology of "hotspots" of concentrated multidrug-resistant tuberculosis (MDR-tuberculosis) risk in Lima, Peru.

METHODS

From 2009 to 2012, we conducted a prospective cohort study among households of tuberculosis cases from 106 health center (HC) areas in Lima, Peru. All notified tuberculosis cases and their household contacts were followed for 1 year. Symptomatic individuals were screened by microscopy and culture; positive cultures were tested for drug susceptibility (DST) and genotyped by 24-loci mycobacterial interspersed repetitive units-variable-number tandem repeats (MIRU-VNTR).

RESULTS

3286 individuals with culture-confirmed disease, DST, and 24-loci MIRU-VNTR were included in our analysis. Our analysis reveals: (1) heterogeneity in annual per-capita incidence of tuberculosis and MDR-tuberculosis by HC, with a rate of MDR-tuberculosis 89 times greater (95% confidence interval [CI], 54,185) in the most-affected versus the least-affected HC; (2) high risk for MDR-tuberculosis in a region spanning several HCs (odds ratio = 3.19, 95% CI, 2.33, 4.36); and (3) spatial aggregation of MDR-tuberculosis genotypes, suggesting localized transmission.

CONCLUSIONS

These findings reveal that localized transmission is an important driver of the epidemic of MDR-tuberculosis in Lima. Efforts to interrupt transmission may be most effective if targeted to this area of the city.

Sparse modeling of spatial environmental variables associated with asthma

Geographically distributed environmental factors influence the burden of diseases such as asthma. Our objective was to identify sparse environmental variables associated with asthma diagnosis gathered from a large electronic health record (EHR) dataset while controlling for spatial variation. An EHR dataset from the University of Wisconsin's Family Medicine, Internal Medicine and Pediatrics Departments was obtained for 199,220 patients aged 5-50years over a three-year period. Each patient's home address was geocoded to one of 3456 geographic census block groups. Over one thousand block group variables were obtained from a commercial database. We developed a Sparse Spatial Environmental Analysis (SASEA). Using this method, the environmental variables were first dimensionally reduced with sparse principal component analysis. Logistic thin plate regression spline modeling was then used to identify block group variables associated with asthma from sparse principal components. The addresses of patients from the EHR dataset were distributed throughout the majority of Wisconsin's geography. Logistic thin plate regression spline modeling captured spatial variation of asthma. Four sparse principal components identified via model selection consisted of food at home, dog ownership, household size, and disposable income variables. In rural areas, dog ownership and renter occupied housing units from significant sparse principal components were associated with asthma. Our main contribution is the incorporation of sparsity in spatial modeling. SASEA sequentially added sparse principal components to Logistic thin plate regression spline modeling. This method allowed association of geographically distributed environmental factors with asthma using EHR and environmental datasets. SASEA can be applied to other diseases with environmental risk factors.

On the spread and control of MDR-TB epidemics: an examination of trends in anti-tuberculosis drug resistance surveillance data

BACKGROUND

Multidrug resistant tuberculosis (MDR-TB) poses serious challenges for tuberculosis control in many settings, but trends of MDR-TB have been difficult to measure.

METHODS

We analyzed surveillance and population-representative survey data collected worldwide by the World Health Organization between 1993 and 2012. We examined setting-specific patterns associated with linear trends in the estimated per capita rate of MDR-TB among new notified TB cases to generate hypotheses about factors associated with trends in the transmission of highly drug resistant tuberculosis.

RESULTS

59 countries and 39 sub-national settings had at least three years of data, but less than 10% of the population in the WHO-designated 27-high MDR-TB burden settings were in areas with sufficient data to track trends. Among settings in which the majority of MDR-TB was autochthonous, we found 10 settings with statistically significant linear trends in per capita rates of MDR-TB among new notified TB cases. Five of these settings had declining trends (Estonia, Latvia, Macao, Hong Kong, and Portugal) ranging from decreases of 3% to 14% annually, while five had increasing trends (four individual oblasts of the Russian Federation and Botswana) ranging from 14% to 20% annually. In unadjusted analysis, better surveillance indicators and higher GDP per capita were associated with declining MDR-TB, while a higher existing absolute burden of MDR-TB was associated with an increasing trend.

CONCLUSIONS

Only a small fraction of countries in which the burden of MDR-TB is concentrated currently have sufficient surveillance data to estimate trends in drug-resistant TB. Where trend analysis was possible, smaller absolute burdens of MDR-TB and more robust surveillance systems were associated with declining per capita rates of MDR-TB among new notified cases.