Understanding tuberculosis epidemiology using structured statistical models

Identifying patient-level risk factors associated with non-β-lactam resistance outcomes in invasive MRSA infections in the United States using chain graphs

Background

MRSA is one of the most common causes of hospital- and community-acquired infections. MRSA is resistant to many antibiotics, including β-lactam antibiotics, fluoroquinolones, lincosamides, macrolides, aminoglycosides, tetracyclines and chloramphenicol.

Objectives

To identify patient-level characteristics that may be associated with phenotype variations and that may help improve prescribing practice and antimicrobial stewardship.

Methods

Chain graphs for resistance phenotypes were learned from invasive MRSA surveillance data collected by the CDC as part of the Emerging Infections Program to identify patient level risk factors for individual resistance outcomes reported as MIC while accounting for the correlations among the resistance traits. These chain graphs are multilevel probabilistic graphical models (PGMs) that can be used to quantify and visualize the complex associations among multiple resistance outcomes and their explanatory variables.

Results

Some phenotypic resistances had low connectivity to other outcomes or predictors (e.g. tetracycline, vancomycin, doxycycline and rifampicin). Only levofloxacin susceptibility was associated with healthcare-associated infections. Blood culture was the most common predictor of MIC. Patients with positive blood culture had significantly increased MIC of chloramphenicol, erythromycin, gentamicin, lincomycin and mupirocin, and decreased daptomycin and rifampicin MICs. Some regional variations were also observed.

Conclusions

The differences in resistance phenotypes between patients with previous healthcare use or positive blood cultures, or from different states, may be useful to inform first-choice antibiotics to treat clinical MRSA cases. Additionally, we demonstrated multilevel PGMs are useful to quantify and visualize interactions among multiple resistance outcomes and their explanatory variables.

Multiplicative Holts Winter Model for Trend Analysis and Forecasting of COVID-19 Spread in India

The surge of the novel COVID-19 caused a tremendous effect on the health and life of the people resulting in more than 4.4 million confirmed cases in 213 countries of the world as of May 14, 2020. In India, the number of cases is constantly increasing since the first case reported on January 30, 2020, resulting in a total of 81,997 cases including 2649 deaths as of May 14, 2020. To assist the government and healthcare sector in preventing the transmission of disease, it is necessary to predict the future confirmed cases. To predict the dynamics of COVID-19 cases, in this paper, we project the forecast of COVID-19 for five most affected states of India such as Maharashtra, Tamil Nadu, Delhi, Gujarat, and Andhra Pradesh using the real-time data. Using Holt–Winters method, a forecast of the number of confirmed cases in these states has been generated. Further, the performance of the method has been determined using RMSE, MSE, MAPE, MAE and compared with other standard algorithms. The analysis shows that the proposed Holt–Winters model generates RMSE value of 76.0, 338.4, 141.5, 425.9, 1991.5 for Andhra Pradesh, Maharashtra, Gujarat, Delhi and Tamil Nadu, which results in more accurate predictions over Holt’s Linear, Auto-regression (AR), Moving Average (MA) and Autoregressive Integrated Moving Average (ARIMA) model. These estimations may further assist the government in employing strong policies and strategies for enhancing healthcare support all over India.

Bayesian generalized linear mixed modeling of Tuberculosis using informative priors

TB is rated as one of the world’s deadliest diseases and South Africa ranks 9th out of the 22 countries with hardest hit of TB. Although many pieces of research have been carried out on this subject, this paper steps further by inculcating past knowledge into the model, using Bayesian approach with informative prior. Bayesian statistics approach is getting popular in data analyses. But, most applications of Bayesian inference technique are limited to situations of non-informative prior, where there is no solid external information about the distribution of the parameter of interest. The main aim of this study is to profile people living with TB in South Africa. In this paper, identical regression models are fitted for classical and Bayesian approach both with non-informative and informative prior, using South Africa General Household Survey (GHS) data for the year 2014. For the Bayesian model with informative prior, South Africa General Household Survey dataset for the year 2011 to 2013 are used to set up priors for the model 2014.

Systems Modelling of the Socio-Technical Aspects of Residential Electricity Use and Network Peak Demand

Provision of network infrastructure to meet rising network peak demand is increasing the cost of electricity. Addressing this demand is a major imperative for Australian electricity agencies. The network peak demand model reported in this paper provides a quantified decision support tool and a means of understanding the key influences and impacts on network peak demand. An investigation of the system factors impacting residential consumers' peak demand for electricity was undertaken in Queensland, Australia. Technical factors, such as the customers' location, housing construction and appliances, were combined with social factors, such as household demographics, culture, trust and knowledge, and Change Management Options (CMOs) such as tariffs, price, managed supply, etc., in a conceptual 'map' of the system. A Bayesian network was used to quantify the model and provide insights into the major influential factors and their interactions. The model was also used to examine the reduction in network peak demand with different market-based and government interventions in various customer locations of interest and investigate the relative importance of instituting programs that build trust and knowledge through well designed customer-industry engagement activities. The Bayesian network was implemented via a spreadsheet with a tickbox interface. The model combined available data from industry-specific and public sources with relevant expert opinion. The results revealed that the most effective intervention strategies involve combining particular CMOs with associated education and engagement activities. The model demonstrated the importance of designing interventions that take into account the interactions of the various elements of the socio-technical system. The options that provided the greatest impact on peak demand were Off-Peak Tariffs and Managed Supply and increases in the price of electricity. The impact in peak demand reduction differed for each of the locations and highlighted that household numbers, demographics as well as the different climates were significant factors. It presented possible network peak demand reductions which would delay any upgrade of networks, resulting in savings for Queensland utilities and ultimately for households. The use of this systems approach using Bayesian networks to assist the management of peak demand in different modelled locations in Queensland provided insights about the most important elements in the system and the intervention strategies that could be tailored to the targeted customer segments.

Prediction and Informative Risk Factor Selection of Bone Diseases

With the booming of healthcare industry and the overwhelming amount of electronic health records (EHRs) shared by healthcare institutions and practitioners, we take advantage of EHR data to develop an effective disease risk management model that not only models the progression of the disease, but also predicts the risk of the disease for early disease control or prevention. Existing models for answering these questions usually fall into two categories: the expert knowledge based model or the handcrafted feature set based model. To fully utilize the whole EHR data, we will build a framework to construct an integrated representation of features from all available risk factors in the EHR data and use these integrated features to effectively predict osteoporosis and bone fractures. We will also develop a framework for informative risk factor selection of bone diseases. A pair of models for two contrast cohorts (e.g., diseased patients versus non-diseased patients) will be established to discriminate their characteristics and find the most informative risk factors. Several empirical results on a real bone disease data set show that the proposed framework can successfully predict bone diseases and select informative risk factors that are beneficial and useful to guide clinical decisions.

Identifying informative risk factors and predicting bone disease progression via deep belief networks

Osteoporosis is a common disease which frequently causes death, permanent disability, and loss of quality of life in the geriatric population. Identifying risk factors for the disease progression and capturing the disease characteristics have received increasing attentions in the health informatics research. In data mining area, risk factors are features of the data and diagnostic results can be regarded as the labels to train a model for a regression or classification task. We develop a general framework based on the heterogeneous electronic health records (EHRs) for the risk factor (RF) analysis that can be used for informative RF selection and the prediction of osteoporosis. The RF selection is a task designed for ranking and explaining the semantics of informative RFs for preventing the disease and improving the understanding of the disease. Predicting the risk of osteoporosis in a prospective and population-based study is a task for monitoring the bone disease progression. We apply a variety of well-trained deep belief network (DBN) models which inherit the following good properties: (1) pinpointing the underlying causes of the disease in order to assess the risk of a patient in developing a target disease, and (2) discriminating between patients suffering from the disease and without the disease for the purpose of selecting RFs of the disease. A variety of DBN models can capture characteristics for different patient groups via a training procedure with the use of different samples. The case study shows that the proposed method can be efficiently used to select the informative RFs. Most of the selected RFs are validated by the medical literature and some new RFs will attract interests across the medical research. Moreover, the experimental analysis on a real bone disease data set shows that the proposed framework can successfully predict the progression of osteoporosis. The stable and promising performance on the evaluation metrics confirms the effectiveness of our model.

Incorporating expert knowledge when learning Bayesian network structure: a medical case study

OBJECTIVES

Bayesian networks (BNs) are rapidly becoming a leading technology in applied Artificial Intelligence, with many applications in medicine. Both automated learning of BNs and expert elicitation have been used to build these networks, but the potentially more useful combination of these two methods remains underexplored. In this paper we examine a number of approaches to their combination when learning structure and present new techniques for assessing their results.

METHODS AND MATERIALS

Using public-domain medical data, we run an automated causal discovery system, CaMML, which allows the incorporation of multiple kinds of prior expert knowledge into its search, to test and compare unbiased discovery with discovery biased with different kinds of expert opinion. We use adjacency matrices enhanced with numerical and colour labels to assist with the interpretation of the results. We present an algorithm for generating a single BN from a set of learned BNs that incorporates user preferences regarding complexity vs completeness. These techniques are presented as part of the first detailed workflow for hybrid structure learning within the broader knowledge engineering process.

RESULTS

The detailed knowledge engineering workflow is shown to be useful for structuring a complex iterative BN development process. The adjacency matrices make it clear that for our medical case study using the IOWA dataset, the simplest kind of prior information (partially sorting variables into tiers) was more effective in aiding model discovery than either using no prior information or using more sophisticated and detailed expert priors. The method for generating a single BN captures relationships that would be overlooked by other approaches in the literature.

CONCLUSION

Hybrid causal learning of BNs is an important emerging technology. We present methods for incorporating it into the knowledge engineering process, including visualisation and analysis of the learned networks.

Web tools for molecular epidemiology of tuberculosis

In this study we explore publicly available web tools designed to use molecular epidemiological data to extract information that can be employed for the effective tracking and control of tuberculosis (TB). The application of molecular methods for the epidemiology of TB complement traditional approaches used in public health. DNA fingerprinting methods are now routinely employed in TB surveillance programs and are primarily used to detect recent transmissions and in outbreak investigations. Here we present web tools that facilitate systematic analysis of Mycobacterium tuberculosis complex (MTBC) genotype information and provide a view of the genetic diversity in the MTBC population. These tools help answer questions about the characteristics of MTBC strains, such as their pathogenicity, virulence, immunogenicity, transmissibility, drug-resistance profiles and host-pathogen associativity. They provide an integrated platform for researchers to use molecular epidemiological data to address current challenges in the understanding of TB dynamics and the characteristics of MTBC.

Conditional independence relations among biological markers may improve clinical decision as in the case of triple negative breast cancers

The associations existing among different biomarkers are important in clinical settings because they contribute to the characterisation of specific pathways related to the natural history of the disease, genetic and environmental determinants. Despite the availability of binary/linear (or at least monotonic) correlation indices, the full exploitation of molecular information depends on the knowledge of direct/indirect conditional independence (and eventually causal) relationships among biomarkers, and with target variables in the population of interest. In other words, that depends on inferences which are performed on the joint multivariate distribution of markers and target variables. Graphical models, such as Bayesian Networks, are well suited to this purpose. Therefore, we reconsidered a previously published case study on classical biomarkers in breast cancer, namely estrogen receptor (ER), progesterone receptor (PR), a proliferative index (Ki67/MIB-1) and to protein HER2/neu (NEU) and p53, to infer conditional independence relations existing in the joint distribution by inferring (learning) the structure of graphs entailing those relations of independence. We also examined the conditional distribution of a special molecular phenotype, called triple-negative, in which ER, PR and NEU were absent. We confirmed that ER is a key marker and we found that it was able to define subpopulations of patients characterized by different conditional independence relations among biomarkers. We also found a preliminary evidence that, given a triple-negative profile, the distribution of p53 protein is mostly supported in 'zero' and 'high' states providing useful information in selecting patients that could benefit from an adjuvant anthracyclines/alkylating agent-based chemotherapy.

Inductive reasoning about causally transmitted properties

Different intuitive theories constrain and guide inferences in different contexts. Formalizing simple intuitive theories as probabilistic processes operating over structured representations, we present a new computational model of category-based induction about causally transmitted properties. A first experiment demonstrates undergraduates' context-sensitive use of taxonomic and food web knowledge to guide reasoning about causal transmission and shows good qualitative agreement between model predictions and human inferences. A second experiment demonstrates strong quantitative and qualitative fits to inferences about a more complex artificial food web. A third experiment investigates human reasoning about complex novel food webs where species have known taxonomic relations. Results demonstrate a double-dissociation between the predictions of our causal model and a related taxonomic model [Kemp, C., & Tenenbaum, J. B. (2003). Learning domain structures. In Proceedings of the 25th annual conference of the cognitive science society]: the causal model predicts human inferences about diseases but not genes, while the taxonomic model predicts human inferences about genes but not diseases. We contrast our framework with previous models of category-based induction and previous formal instantiations of intuitive theories, and outline challenges in developing a complete model of context-sensitive reasoning.