In India, nearly 5.8 million people (WHO report, 2015) die from NCDs (heart and lung diseases, stroke, cancer and diabetes) every year or in other words 1 in 4 Indians has a risk of dying from an NCD before they reach the age of 70. Health data contain both a time and a space component. This clearly impacts on spatial coverage and could potentially lead to biased statistical inference if the data gaps are clustered in space and/or if they differentially affect specific population groups (e.g. Topics Covered: Issues relating to development and management of Social Sector/Services relating to Health, Education, Human Resources, issues relating to poverty and hunger. This is due to the excessive smoothing following the assumption of a common variance across all the areas and time points. Hierarchical models (HM) are able to deal with complex data structures, to exploit dependencies between data sources and to propagate the inherent uncertainties that are present in both the data and the modelling process. Centers for Disease Control and Prevention. Leroux BG, Lei X, Breslow N. Estimation of disease rates in small areas: a new mixed model for spatial dependence. Overview . Comparison of two methods for cell count determination in the course of biocide susceptibility testing. These are chronic diseases of long duration, and generally slow progression and are the result of a combination of genetic, physiological, environmental and behaviours factors. Benidorm, Spain. The standard disease mapping approach has been used informally to detect anomalies (unusual observations) in space and time, i.e. The UK Small Area Health Statistics Unit (SAHSU) is part of the MRC-PHE Centre for Environment and Health, which is supported by the Medical Research Council (MR/L01341X/1) and Public Health England (PHE). The Department of Health produces timely and comprehensive statistics on health status and disease surveillance of local population. Benjamini and Hochberg first introduced an alternative index, the false-discovery rate (FDR)50 as the expected value of the rate of false-positive findings among all rejected hypotheses, and used it in a frequentist approach. Disease mapping models have been extensively used to estimate and visualize the spatial or spatiotemporal distribution of a disease (see for instance Diggle and Giorgi, Adam and Fenton, and WHO9,21,28). This report provides a summary of the burden of the key NCDs and their risk factors. To know more about NCDs and National Programme Guidelines-, This question is for preventing automated spam submissions. Some work in this area includes Foreman et al.59 who, using annual vital statistics for 1974–2011 at the US state spatial resolution, forecasted mortality up to 2024; and Ugarte et al.67 used P-splines to forecast cancer mortality counts in Spanish regions for 2009–11 using data from 1975–2008. Here we focus on current and future trends for some of the most prevalent non-communicable diseases. At the First and Second UN High-level Meetings on Noncommunicable Diseases (NCDs) in 2011 and 2014, the World Health Organization released Country Profiles, highlighting the latest data on NCDs in each WHO Member State. The global action plan has suggested 9 targets for countries to set. https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases Such continuous surveys are an invaluable source of data, but researchers face issues related to population representativeness. The authors claimed that the inclusion of social media data could be a cost-effective real-time health detection system. NCD Countdown 2030 is an independ … Recently, it was further extended to jointly model age- and gender-specific diseases.40, An alternative multivariate specification considers spatial and temporal terms explicitly, modelling the correlation among the outcomes in space/time. The majority of premature NCD deaths are preventable. World Health Organization. Traditionally, the methodological and conceptual frameworks for surveillance have been designed for infectious diseases, but the rising burden of non-communicable diseases (NCDs) worldwide suggests a pressing need for surveillance strategies to detect unusual patterns in the data and to help unveil important risk factors in this setting. Part of their popularity lies in the availability of free user-friendly SaTScanTM software [https://www.satscan.org/]. This will potentially lead to bias in population representativeness due to non-random missingness12 which will need to be addressed using advanced statistical methods, for instance through the integration of data from appropriate surveys/cohorts, as proposed in the context of residual confounding.73, An important issue with surveillance studies is that of the spatial resolution and the type of geographical areas considered; modifying these might lead to different results, as the spatial distribution of the outcome will depend on these choices. In particular, two alternative models are considered: the first one assumes a global time trend for all areas (common trend), and the second estimates a time trend for each area independently (area-specific trend). Country Office for Bangladesh. In this way, we ensured that a realistic scenario was used (for more details see Boulieri et al.,47 and Supplementary materials, Scenario 1, available as Supplementary data at IJE online). A common choice for capturing temporal dependence is a random walk prior,29 but extensions to incorporate spatiotemporal interactions among neighbouring areas and time points have also been developed.30 This framework can also account for factors known to modify spatial and temporal trends that, in the context of NCDs, will include demographic variables (e.g. Finally, in the Discussion, we conclude with a summary and some remaining discussion points. A joint model allows information to be borrowed across the outcomes, thus helping stabilize estimates, particularly when the outcomes are rare. As an example, road traffic accidents characterized by different severity were analysed over the period 2005–11 at the ward level in England while detection of high-risk areas was performed using exceedance probabilities of the area ranks based on accident rates.41. Parameters and latent stochastic processes are fundamentally different things, but within the Bayesian paradigm they are both treated as unobserved sets of random variables, and the operational calculus of estimation and prediction coalesces. Other applications of this method include burglary data,45 grey whale abundance46 and mammography data.35. National HIV Surveillance System (NHSS). More. Tackling the epidemic of chronic diseases – or non-communicable diseases (NCDs) – is at the heart of this agenda, and it’s a major challenge. Recent applications of SaTScan include the identification of signals for colorectal cancer,18 drug activity,19 criminality20 and bat activity.21, A further development has been the detection of spatial variations in temporal trends (SVTT). As an example, the Rapid Inquiry Facility (RIF) which is currently being redeveloped within SAHSU, is designed to facilitate disease mapping and risk analysis studies and has been employed by more than 45 institutions in a number of countries.78 A more recent example is the SpatialEpiApp that integrates two methods for disease mapping and cluster detection.79. Relevant references are cited on each page. This Portal is designed, developed and hosted by Centre for Health Informatics (CHI), set up at National Institute of Health and Family Welfare (NIHFW), by the Ministry of Health and Family Welfare (MoHFW), Government of India. We selected 15 areas to deviate from the overall time trend over the last five time points. Hierarchical modelling provides a coherent framework within which spatiotemporal dependencies can be explicitly modelled with integration of the uncertainties associated with both the data and the modelling process. Clayton DG. Mauritius Non Communicable Diseases Survey 2015 5 Executive summary A non-communicable disease (NCD) survey employing similar methodologies and criteria to surveys undertaken in Mauritius in previous years (1987, 1992, 1999, 2004 and 2009), was carried out in 2015. In particular, Abellan et al.42 developed a BHM model (termed STmix) where a mixture of two normal distributions characterized by different variances is specified for the space-time interaction. (b) Relative risks and 95% credible intervals of hospital admissions for asthma and COPD for the national (common) temporal trend and for Hillingdon CCG, classified as unusual. Bonferroni correction has been extensively used in epidemiology to correct for multiple testing, particularly in omics studies,48,49 but it is well known that this approach leads to conservative results. Surveillance methods need to be able to detect meaningful departures from expectation and exploit dependencies within such data to produce unbiased estimates of risk as well as future forecasts. Chronic Non Communicable Diseases (NCDs) in the Caribbean: THE FACTS • Globally and in the Caribbean, the chronic diseases of concern are heart disease, stroke, cancer, diabetes and chronic respiratory diseases. For example, if the public health question is whether current risk exceeds an agreed acceptable level in all areas that do, and in no areas that do not, meet a particular criterion such as adherence to a particular advisory policy, the correct predictive probability to attach to this statement can be calculated. Computationally intensive BHMs benefit from high-performance computing clusters to speed up computation times, but these are not necessarily required. Integration of NPCDCS with the National Health Mission (NHM) resulted into augmented infrastructure and human resources particularly in the form of frontline workers- the ANM and the ASHA. With the active participation of these frontline workers the population-based periodic screening of hypertension, diabetes, and common cancers (oral, breast, cervical cancers) is initiated to facilitate the early detection of common NCDs. Inpatient Burden and Mortality of Heat Stroke in the United States. One of the biggest challenges researchers face when analysing large and complex space-time datasets is their computational burden. (‎2018)‎. NCD deaths worldwide now exceed all communicable, maternal and perinatal nutrition-related deaths combined and represent an emerging global health threat. Cause of death, by non-communicable diseases (% of total) from The World Bank: Data Learn how the World Bank Group is helping countries with COVID-19 (coronavirus). Prevention and management of chronic obstructive pulmonary disease (COPD) and chronic Kidney disease (CKD); and better management of co-morbidities such as diabetes and tuberculosis are also considered under the programme. Fixing a 0.9 threshold (DM2), FDR decreases, despite still being above the standard threshold of 0.05, while at the same time sensitivity also decreases (0.660) (Table 1). In this way disease mapping, though not formally a surveillance method, can be used as a descriptive tool for the identification of areas and/or time periods with marked deviation from expectation. HIV Surveillance Systems. Test-based methods such as scan statistics can only answer questions related to the deviation from the null hypothesis. An example is provided by Goicoa et al.31 who proposed a space-time-age model to study prostate cancer incidence across 50 provinces in Spain for nine age groups over 25 years, accounting for all pair-wise interactions. For instance, if within-area variability is substantial, results from statistical inference might suffer from false-negative observations, as potentially high-risk places are aggregated with low-risk ones. It is important to note that the strong smoothing effect of disease mapping models leads to conservative risk estimates, hence to a small number of false-positive findings, at the expense of a low power for detecting high-risk areas with low signal. Perhaps the most popular test-based methods used for NCD surveillance are the scan statistics. Syndromic data, such as primary care data, drug prescriptions, nurse calls and home visits, which are indicative of a potential anomaly, may provide an additional level of information leading to a detection event before the data aberration occurs.68 Diggle et al.69,70 analysed NHS non-emergency telephone calls reporting symptoms of gastrointestinal diseases. Although morbidity and mortality from NCDs mainly occur in adulthood, exposure to risk factors begins in early life. An additional version of spatial scan statistic was proposed to account for correlation across spatial units, which was not considered before.17 Scan statistics have been extensively applied to numerous health care applications. BHMs have been suggested as a way to, at least partially, deal with MAUP. This can be particularly challenging for rare diseases where the numbers of cases at small-area level are very low. Another is that the Monte Carlo sampling method allows the computation of whatever joint probability statements are required. The user-friendly software BUGS (Bayesian inference Using Gibbs Sampling)60 has been traditionally used for Bayesian inference using MCMC methods; however, it can be slow when high-dimensional data and/or complex models are used. In this section, we first discuss how data availability is one of the key challenges in surveillance studies, before giving a generic overview of test-based approaches for NCD surveillance. India’s National Monitoring Framework for Prevention and Control of NCDs has committed for a 50% relative reduction in household use of solid fuel and a 30% relative reduction in prevalence of current tobacco use by 2025. Differences across the competing models were observed in terms of computation time, an important factor in assessing their performance. The shared component model,37 originally developed for two diseases, includes a common component (likely to reflect common risk factors) and a disease-specific one, which can point towards specific risk factors otherwise masked in a single disease model. Molecular Detection and Phylogenetic Analysis of. A key characteristic of BHMs is the ready availability of joint posterior/predictive distributions for parameters/latent processes and whatever of their properties are relevant to the public health questions of interest. Published by Oxford University Press on behalf of the International Epidemiological Association. Furthermore, the choice of spatial resolution is mostly dependent on data availability and sparsity. The authors performed a simulation study to compare the method against the standard disease mapping approach. Some of the reasons for the non-infectious disease are genetics, nutritional deficiency, age and sex of the individual and so on. Health promotion through social media is also being used to generate awareness about prevention and control of NCDs, such as use of mobile technology in applications called mDiabetes for diabetes control, mCessation to help for quit tobacco, and no more tension as a support for mental stress management. Both elements are specified up to the values of a set of unknown parameters, which can be estimated by Bayesian or non-Bayesian versions of likelihood-based inference, typically implemented using Markov chain Monte Carlo integration and Monte Carlo likelihood maximization methods, respectively. Non-communicable diseases are the diseases that are not transferred from an infected person to another via any means and are mostly caused by factors like improper lifestyle and eating habits. Spiegelhalter DJ, Thomas A, Best NG, Gilks W, Lunn D. BUGS: Bayesian Inference Using Gibbs Sampling. Even in high-income settings where administrative resources are available for the entire population, there may be issues regarding the population at risk, used as denominator in the risk estimates. Welfare schemes for vulnerable sections of the population by the Centre and States and the performance of these schemes. The simulation results suggest that using disease mapping (DM) for surveillance purposes is not appropriate and that one of the mixture models designed for detection should be used instead. © The Author(s) 2020. Here, we present an overview of the approaches developed for spatiotemporal disease surveillance of NCDs. In addition, it is not straightforward to define the denominator where the interest is for less-defined geographies, such as the catchment areas of clinical centres (e.g. arXiv preprint arXiv. © 2016 MoHFW, Government of India, All rights reserved. Part of this work was supported by an Early Career MRC-PHE Fellowship awarded to A.B. Through the exceedance probabilities, these maps give a perception of the uncertainty around the area-level relative risks estimates. Non-communicable Disease. to exhibit a risk pattern not deviating from the expected one.43. Resident Physician in Cardio-Thoracic and Vascular Surgery, Research Assistant Professor of Epidemiology, Copyright © 2020 International Epidemiological Association. The asthma dataset was obtained from SAHSU, Imperial College London, and consisted of disease counts across 211 Clinical Commissioning Groups (CCG) in England for 15 months, from January 2010 to March 2011. Spatiotemporal variation in health outcomes and lifestyle and environmental exposures needs to be explicitly modelled in order to reduce bias and uncertainty. At the same time there will be the need to reduce the computational burden of increasingly complex models applied to large datasets, in order to provide timely results for decision making. We compared the detection performance of disease mapping (DM1, DM2), the mixture model on the spatiotemporal interaction (STmix1, STmix2) and the mixture model on the spatiotemporal rates (FlexDetect). Mixture models have been proposed as a formal approach to anomaly detection. According to the World Health Organization (WHO), surveillance is the ‘ongoing systematic data collection, analysis and interpretation and dissemination of information in order for action to be taken’.1 National public health agencies, such as the US Centers for Disease Control and Prevention (CDC) and Public Health England (PHE), routinely carry out surveillance data analysis to provide early warnings of unexplained changes in incidence patterns of diseases as well as to aid policy formation and resource allocation.2 Specific examples include the international influenza monitoring system which started in 1948 and is now distributed in 82 countries,3 the HIV and AIDS Reporting System (HARS) used by PHE4 and the National HIV Surveillance System used by CDC.4,5, To date, the majority of methods and models commonly used in public health surveillance are designed for monitoring cases of infectious diseases.6 Due to the rising burden of non-communicable diseases (NCDs) worldwide, there is a pressing need to implement surveillance strategies to detect trends, highlight unusual changes and consequently assist in outlining emerging NCD risk factors. If the interest lies in detection in the presence of spatial proximity, recent methods have been developed to combine clustering with spatial smoothing, see for example Anderson et al.65 and Adin et al.66. The results of the simulation study showed that the standard approach was not able to capture the variability in the spatiotemporal interactions and therefore it was not able to distinguish between common and unusual areas. A number of patients living with non-communicable diseases cannot easily access treatment as the services have been scaled down and some of … There are a number of advantages to adopting a model-based approach over a test-based method, including the ability to: (i) have more statistical power to handle sparsity in the observed disease counts; (ii) explore more subtle departures from the expectation; (iii) account for the spatial and temporal correlation that is typically evident in health data; (iv) ‘borrow’ information over space and time, therefore increasing the precision of the estimates generated; and (v) include covariates that might explain some of the spatiotemporal variability. is Director of SAHSU and Director of the MRC-PHE Centre for Environment and Health. NCDs are rapidly increasing globally and reached epidemic proportions in many countries, largely due to globalization, industrialization, and rapid urbanization with demographic and lifestyle changes. STmix1 gave no false-positive results (FDR = 0) and a sensitivity of 0.773, whereas for STmix2 sensitivity increased to 0.969, but at the same time a much higher proportion of false-positives was detected (FDR = 0.220). These were developed originally in the temporal setting only13; here, a fixed length ‘scanning window’ is passed over the time-series data with the number of cases in the window being recorded. The Non-communicable Disease Unit . At a recent press conference, World Health Organization (WHO) Director-General, Dr Margaret Chan, reported unprecedented progress against neglected tropical diseases. This is particularly challenging as the statistical modelling of surveillance data becomes more sophisticated. An alternative way of dealing with computational limitations is to use approximative methods; for instance INLA (Integrated Nested Laplace Approximations)63 has been successfully used for running space-time disease mapping models (e.g.31,64) however, this method is somewhat less flexible than the aforementioned ones and, as it relies on Normality of the latent process, is not able to deal with mixture distributions. Non-Communicable Diseases as the name suggests are not transmitted from person to person but is a result of combination of genetic, physiological, environmental and behavioral factors. This improved method, termed FlexDetect, had a better performance when compared with the original method through an extensive simulation study.47, As surveillance studies involve evaluating trends for different health outcomes, many areas and different time periods at the same time, false detections are likely to occur by chance. A Survey on Cardiovascular Nursing Occupational Standard: Meeting the Needs of Employers. The quadratic SVTT method has, for example, been applied to cervical cancer data in women in the USA from 1969 to 1995, highlighting areas where the risk was significantly different from the rest.22. Surveillance methods must be able to capture spatial and temporal patterns in both lifestyle/environmental exposures and health outcomes. Physical inactivity, unhealthy diets (diets low in fruit, vegetables, and whole grains, but high in salt and fat), tobacco use (smoking, secondhand smoke, and smokeless tobacco), and the harmful use of alcohol are the main behavioural risk factors for NCDs. The epidemic of NCDs cannot be halted simply by treating the sick, healthy persons have to be protected by addressing the root causes. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide, This PDF is available to Subscribers Only. mortality/cancer registries) over an entire country infeasible.9, In the past 15 years, a number of Health and Demographic Surveillance Systems (HDSS) have been established in low-income settings to provide a reliable source of health data, and are now linked together through the International Network for the continuous Demographic Evaluation of Populations and Their Health (INDEPTH10). A log likelihood ratio (LLR)14 is calculated for each interval, and the test statistic is defined as the maximum LLR over all intervals. The Environment and Health Atlas for England and Wales36 (typical output from the Atlas was presented in Figure 1) is an example of work in this direction, providing stakeholders and the general public with a collection of maps to inform on the spatial distribution of environmental factors and diseases. Actions to beat non-communicable diseases. Furthermore, the availability of administrative or health data may become more limited: for instance, within the UK National Health Service, patients can now decide not to share their medical records for research purposes. Morin A, Laviolette M, Pastinen T, Boulet LP, Laprise C. Newton MA, Noueiry A, Sarkar D, Ahlquist P. Muller P, Parmigiani G, Rice K. FDR and Bayesian multiple comparisons rules. STmix was applied to mammography screening data in Brisbane, Australia, at the statistical local area (SLA) level from 1997 to 2008, in order to identify SLAs whose temporal trend exhibited volatility.35 A well-known drawback of this approach is its limitation in incorporating specific time patterns, for example step changes that could signal the emergence of a new risk factor. The R code used for the data simulation, together with the three models written in BUGS, can be found on [https://github.com/aretib/bayesSTmodels.git]. According to the World Health Organization (WHO), surveillance is the ‘ongoing systematic data collection, analysis and interpretation and dissemination of information in order for action to be taken’. diseases (NCDs). This shows a rapid epidemiological transition with a shift in disease burden to NCDs. This has led to the increasing development of a range of space-time methods specifically designed for NCD surveillance. They contribute to raised blood pressure (hypertension); raised blood sugar (diabetes); raised and abnormal blood lipids (dyslipidaemia); and obesity.