Risk groups in children under six months of age using self-organizing maps.

Fetal and infant growth tends to follow irregular patterns and, particularly in developing countries, these patterns are greatly influenced by unfavorable living conditions and interactions with complications during pregnancy. The aim of this study was to identify groups of children with different risk profiles for growth development. The study sample comprised 496 girls and 508 boys under six months of age from 27 pediatric primary health care units in the city of Rio de Janeiro, Brazil. Data were obtained through interviews with the mothers and by reviewing each child's health card. An unsupervised learning, know as a self-organizing map (SOM) and a K-means algorithm were used for cluster analysis to identify groups of children. Four groups of infants were identified. The first (139) consisted of infants born exclusively by cesarean delivery, and their mothers were exclusively multiparous; the highest prevalences of prematurity and low birthweight, a high prevalence of exclusive breastfeeding and a low proportion of hospitalization were observed for this group. The second (247 infants) and the third (298 infants) groups had the best and worst perinatal and infant health indicators, respectively. The infants of the fourth group (318) were born heavier, had a low prevalence of exclusive breastfeeding, and had a higher rate of hospitalization. Using a SOM, it was possible to identify children with common features, although no differences between groups were found with respect to the adequacy of postnatal weight. Pregnant women and children with characteristics similar to those of group 3 require early intervention and more attention in public policy.

Comparison among probabilistic neural network, support vector machine and logistic regression for evaluating the effect of subthalamic stimulation in Parkinson disease on ground reaction force during gait.

Deep brain stimulation of the subthalamic nucleus (DBS-STN) is an approved treatment for advanced Parkinson disease (PD) patients; however, there is a need to further evaluate its effect on gait. This study compares logistic regression (LR), probabilistic neural network (PNN) and support vector machine (SVM) classifiers for discriminating between normal and PD subjects in assessing the effects of DBS-STN on ground reaction force (GRF) with and without medication. Gait analysis of 45 subjects (30 normal and 15 PD subjects who underwent bilateral DBS-STN) was performed. PD subjects were assessed under four test conditions: without treatment (mof-sof), with stimulation alone (mof-son), with medication alone (mon-sof), and with medication and stimulation (mon-son). Principal component (PC) analysis was applied to the three components of GRF separately, where six PC scores from vertical, one from anterior-posterior and one from medial-lateral were chosen by the broken stick test. Stepwise LR analysis employed the first two and fifth vertical PC scores as input variables. Using the bootstrap approach to compare model performances for classifying GRF patterns from normal and untreated PD subjects, the first three and the fifth vertical PCs were attained as SVM input variables, while the same ones plus the first anterior-posterior were selected as PNN input variables. PNN performed better than LR and SVM according to area under the receiver operating characteristic curve and the negative likelihood ratio. When evaluating treatment effects, the classifiers indicated that DBS-STN alone was more effective than medication alone, but the greatest improvements occurred with both treatments together.

Assessment of the effects of subthalamic stimulation in Parkinson disease patients by artificial neural network.

This study aims at using a probabilistic neural network (PNN) for discriminating between normal and Parkinson disease (PD) subjects using as input the principal components (PCs) derived from vertical component of the ground reaction force (vGRF). The trained PNN was further used for evaluating the effects of deep brain stimulation of the subthalamic nucleus (STN DBS) on PD, with and without medication. A sample of 45 subjects (30 normal and 15 PD subjects who underwent STN DBS) was evaluated by gait analysis. PD subjects were assessed under four test conditions: without treatment (mof-sof), only with stimulation (mof-son) or medication (mon-sof), and with combined treatments (mon-son). PC analysis was applied on vGRF, where six PC scores were chosen by the broken stick test. Using a bootstrap approach for the PNN model, and the area under the receiver operating characteristic curve (AUC) as performance measurement, the first three and fifth PCs were selected as input variables. The PNN presented AUC = 0.995 for classifying controls and PD subjects in the mof-sof condition. When applied to classify the PD subjects under treatment, the PNN indicated that STN DBS alone is more effective than medication, and further vGRF enhancement is obtained with combined therapies.

Drug resistance mutation profile and accumulation kinetics in human immunodeficiency virus-positive individuals infected with subtypes B and F failing highly active antiretroviral therapy are influenced by different viral codon usage patterns.

The major human immunodeficiency virus type 1 subtype circulating in Brazil is B, followed by F and C. We have genotyped 882 samples from Brazilian patients for whom highly active antiretroviral therapy failed, and we found subtype B and the unique recombinant B/F1 forms circulating. Due to codon usage variation, there is a significantly lower incidence of the substitutions L210W, Q151M, and F116Y in subtype F1 isolates than in the subtype B counterparts.

Assessment of the effects of subthalamic stimulation in Parkinson disease patients by artificial neural network.

This study aims at applying an artificial neural network for the evaluation of the effects of deep brain stimulation (DBS) of the subthalamic nucleus (STN) on Parkinson disease (PD) patients with and without medication. A sample of 15 PD patients who have undergone STN DBS were evaluated under four test conditions: medication off and stimulation off (mof-sof), medication off and stimulation on (mof-son), medication on and stimulation off (mon-sof) and medication on and stimulation on (mon-son). A control group with 30 subjects was also evaluated. Principal component analysis (PCA) was applied on vertical ground reaction force (vGRF) and the first six principal component scores (PC score) were obtained in both groups. Those PCs scores were used as input in a probabilistic neural network (PNN). PNN presented satisfactory classification performance in the separation of controls and PD with 90.1% accuracy, 69.2% sensitivity and 100% specificity. The stimulation mof-son and mon-son conditions presented better results compared to mon-sof. In the mof-son condition, 41.7% were classified as normal, while further enhancement (63.3%) was given by the mon-son condition. These results indicated the potentiality of PNN to quantitatively evaluate treatment effects. Furthermore, STN DBS shows improvement on vGRF pattern in PD patients, most substantially when used with medication.

[Analysis of spatial diffusion patterns for AIDS cases in some Brazilian States]. / Análise dos padrões de difusão espacial dos casos de AIDS por estados brasileiros.

Despite new improvements in AIDS treatment, preventive measures are still essential to control the epidemic. Effective programs almost always depend on correct and efficient allocation of scarce health resources. Detailed information on the epidemic, such as where, when, and how the epidemic will spread are of great value. This study was conducted to obtain a better understanding of the dissemination of AIDS cases in four important Brazilian States. Spatial diffusion patterns were evaluated qualitatively by studying sequential maps and quantitatively by analyzing spatial correlograms. Ten years were analyzed, grouped in three periods (1987-1989, 1990-1992, and 1993-1996). The diffusion process was studied for both total AIDS cases and male and female cases. Diffusion of AIDS cases presented specific characteristics for each of the four States. Information derived from the study, especially the results of the correlogram analysis, improve our understanding of the epidemic's spatial diffusion in different parts of the country and can also be used to determine parameters for other AIDS epidemiological models.

Dynamic linear model and SARIMA: a comparison of their forecasting performance in epidemiology.

One goal of a public health surveillance system is to provide a reliable forecast of epidemiological time series. This paper describes a study that used data collected through a national public health surveillance system in the United States to evaluate and compare the performances of a seasonal autoregressive integrated moving average (SARIMA) and a dynamic linear model (DLM) for estimating case occurrence of two notifiable diseases. The comparison uses reported cases of malaria and hepatitis A from January 1980 to June 1995 for the United States. The residuals for both predictor models show that they were adequate tools for use in epidemiological surveillance. Qualitative aspects were considered for both models to improve the comparison of their usefulness in public health. Our comparison found that the two forecasting modelling techniques (SARIMA and DLM) are comparable when long historical data are available (at least 52 reporting periods). However, the DLM approach has some advantages, such as being more easily applied to different types of time series and not requiring a new cycle of identification and modelling when new data become available.

Path analysis and multi-criteria decision making: an approach for multivariate model selection and analysis in health.

PURPOSE: This paper introduces an approach that includes non-quantitative factors for the selection and assessment of multivariate complex models in health. METHODS: A goodness-of-fit based methodology combined with fuzzy multi-criteria decision-making approach is proposed for model selection. Models were obtained using the Path Analysis (PA) methodology in order to explain the interrelationship between health determinants and the post-neonatal component of infant mortality in 59 municipalities of Brazil in the year 1991. Socioeconomic and demographic factors were used as exogenous variables, and environmental, health service and agglomeration as endogenous variables. Five PA models were developed and accepted by statistical criteria of goodness-of fit. These models were then submitted to a group of experts, seeking to characterize their preferences, according to predefined criteria that tried to evaluate model relevance and plausibility. Fuzzy set techniques were used to rank the alternative models according to the number of times a model was superior to ("dominated") the others. RESULTS: The best-ranked model explained above 90% of the endogenous variables variation, and showed the favorable influences of income and education levels on post-neonatal mortality. It also showed the unfavorable effect on mortality of fast population growth, through precarious dwelling conditions and decreased access to sanitation. CONCLUSIONS: It was possible to aggregate expert opinions in model evaluation. The proposed procedure for model selection allowed the inclusion of subjective information in a clear and systematic manner.

[Voronoi s Diagram for defining catchment areas for public hospitals in the Municipality of Rio de Janeiro]. / Diagramas de Voronoi para a definição de áreas de abrangência de hospitais públicos no Município do Rio de Janeiro.

One of the most important pieces of information for health resources planning is the definition of catchment areas for health units. Voronoi Diagrams are a potential technique for this purpose. They are polygons with the property whereby adjacent polygons have their borders located within the same distance of the respective generator points. One possible adjustment to the catchment areas thus defined is the use of weighted Voronoi Diagrams, which result in an improved representation of a health unit's actual capacity. In this study, the 21 public general hospitals in the city of Rio de Janeiro, Brazil, were used as generator points for Voronoi Diagrams. Non-weighted Voronoi Diagrams were initially implemented and then used as the basis for obtaining weighted Voronoi Diagrams, using as weights the annual admission rates estimated for each unit. In the classic Voronoi Diagram case, some catchment areas had similar sizes, although their respective health units had different characteristics. In the weighted case the areas were modified in a way that appeared closer to the actual functioning of the units. The method appeared simple to implement, used easy-to-access data, and did not rely on geopolitical considerations such as existing administrative areas. It thus provided a more realistic picture of a unit's capacity to support basic health programs.

Multi-criteria decision making--an approach to setting priorities in health care.

The objective of this paper is to present a multi-criteria decision making (MCDM) approach to support public health decision making that takes into consideration the fuzziness of the decision goals and the behavioural aspect of the decision maker. The approach is used to analyse the process of health technology procurement in a University Hospital in Rio de Janeiro, Brazil. The method, known as TODIM, relies on evaluating alternatives with a set of decision criteria assessed using an ordinal scale. Fuzziness in generating criteria scores and weights or conflicts caused by dealing with different viewpoints of a group of decision makers (DMs) are solved using fuzzy set aggregation rules. The results suggested that MCDM models, incorporating fuzzy set approaches, should form a set of tools for public health decision making analysis, particularly when there are polarized opinions and conflicting objectives from the DM group.

[Mathematical location models applied in the spatial organization of health units]. / Modelos matemáticos de localização aplicados à organização espacial de unidades de saúde.

Mathematical location models have been increasingly applied in the health services at the international level. In Brazil, although incipient, there exists an enormous potential for the use of such models in the area of public health. In this paper several location models that can be applied to public health are presented initially, and the location of non-emergency services, of emergency services and of services hierarchically related are analysed. A hierarchical model is then applied to the location of maternal and perinatal assistance in the municipality of Rio de Janeiro. In this part, after presenting some related data for the municipality, a four-level hierarchical model (location of out-patient units, maternity hospitals, neonatal hospitals and general hospitals) is proposed and the impact that the adoption of this methodology would have as compared with that of the present system is analysed.

The path analysis approach for the multivariate analysis of infant mortality data.

PURPOSE: This paper reviews the use of the Path Analysis (PA) methodology in health determinants modeling, with special reference to infant mortality modeling. METHODS: A review of the literature on PA applications in the modeling of infant mortality and similar problems is presented, together with a discussion of the conceptual basis of PA and its relation to other multivariate statistical techniques. Important aspects of the technique are discussed: 1) criteria for path formulation; 2) parameter estimation methods; 3) direct, indirect, spurious, and joint effects; and 4) goodness-of-fit and modification indices. RESULTS AND CONCLUSION: The review of the literature suggests that PA represents a methodological improvement regarding multivariate techniques used in modeling some health-related issues. PA allows investigation of more complex models, providing information that could have been previously overlooked, such as how the interrelations among independent variables in a model affect the dependent ones.

GISEpi: a simple geographical information system to support public health surveillance and epidemiological investigations.

One important question for the implementation of a surveillance system concern the type of instrument that can provide timely information on the course of diseases and other health events. This may facilitate prompt implementation of prevention and intervention efforts, such as strengthening control action in one specific area or initiation of epidemiological investigation. Since health related variables of interest are often spatially distributed they require special tools for representation and analysis. Owing to their inherent ability to manage spatial information, geographical information systems (GIS) provide an excellent framework for the design of surveillance systems. This paper presents a simple information system, based on the concepts of GIS, designed for representation and elementary analysis of epidemiological data. An example of its potential use to support malaria control activities in Brazil is discussed.

Spatial partitioning using multivariate cluster analysis and a contiguity algorithm.

Spatial analysis of epidemiological data can be a useful tool for identifying patterns of disease occurrence and can provide substantial support for prevention and control strategies. To obtain the greatest spatial resolution, it is important to use the smallest available areal units with homogeneous population. However, small areas usually have a small population, introducing spurious variability in the chosen indicators of disease occurrence. This paper describes an approach for combining small geographical units to stabilize mortality rates by pooling information across areas according to specified risk profiles. The procedure is based on a principal component analysis, followed by a cluster analysis of social-economic indicators to classify the risk profile of each small area. The classification is used in an algorithm to join neighbouring areas with similar profiles until an estimated population size is achieved. We applied this method to two Administrative Regions of the city of Rio de Janeiro, Brazil, using the census tracts as the basic areal unit. Census tracts were classified according to four socioeconomic categories distributed spatially as a mosaic, where tracts of differing categories neighbour each other. The aggregation algorithm produced a new partition of the region studied, with the created areal units preserving the internal socioeconomic homogeneity.

HIPERSIG: um tutor inteligente sobre técnicas de mapeamento espacial e sistemas de informaçöes geográficas / HIPERSIG: intelligent tutor about technique of spatial mapping and geographic information systems

O HIPERSIG é um sistema tutorial inteligente hipermídia, baseado na técnica de Sistemas Especialistas aliados a Hipertextos, que se propõe a auxiliar o pessoal ligado à área da saúde no que se refere às técnicas de representação espacial de dados e a tulização de sistemas de informações geográficas.

Feasibility of contour mapping epidemiological data with missing values.

Data of epidemiologic interest often occur as spatial information during each of several time periods. In most cases data are available from a set of regions or localities which can be viewed as points in a plane. Although contour mapping is useful for displaying these data, the lack of data for all data points in a region may lead to erroneous interpretation. In this paper we use stimulation to investigate the impact of missing data points for contour mapping using two distinct simulated spatial-time distributions for epidemiologic variables. A model for the occurrence of malaria in localities randomly distributed in one region is chosen as the prototype for data generation.

A monitoring system to detect changes in public health surveillance data.

One task faced by public health surveillance practitioners is the timely identification of data patterns that might suggest the onset of an epidemic period. Many available techniques for analysis of surveillance data are based on sequential procedures, which predict expected numbers of cases and compare this estimate with observed values. To detect changes in the reported occurrence of a disease (increase, decrease, or change in trend), we used exponential smoothing and transformation of the difference between the observed and estimated data to calculate a function called the probability index. We illustrate this procedure using weekly provisional data for measles cases in the US reported through the National Notifiable Diseases Surveillance System to the Centers for Disease Control and Prevention (CDC). The method is potentially useful in public health surveillance to facilitate prompt intervention and prevention efforts, since it can be used at the national and regional levels without the requirement for sophisticated computing.

[Occurrence of meningococcal meningitis in a Southern region of Brazil, from 1974 to 1980, using the point event model]. / Estudo da ocorrência de meningite meningocócica em região meridional do Brasil, de 1974 a 1980, utilizando o modelo ponto evento.

Descriptive statistical techniques and point event model methods were used to investigate the temporal series of cases of meningococcal meningitis which occurred in 100 municipalities in the Rio Grande do Sul State, Brazil, during the period 1974-1980. The data were grouped by epidemiological state (epidemic or endemic), and separated into 5 groups according to the municipal population. The number of cases of the disease notified weekly was analysed by means of incidence coefficients, with the purpose of studying the epidemic threshold for the state. The time interval between events was analysed in the light of their probability density functions and expected density functions, with the objective of studying the relationship and dependences among events. The analysis of the epidemic threshold suggests that there should not be only one threshold value for detection of outbreak of the disease throughout the state. Analysis of the expected density function extracted from inter-event intervals of the epidemic state showed a correlational structure indicating dependence between events occurring up to 14 weeks apart. No significant correlation for the endemic state, taking as reference model the shuffled version of the original intervals, was observed.

Estudo preliminar de dados para análise de determinantes de saúde no Brasil / Preliminary analisis of determinants of health in Brazil

A disponibilidade de dados confiáveis sobre o sistema de Saúde é indispensável para o desenvolvimento de trabalhos na área de planejamento em saúde. O presente trabalho descreve a implementaçäo de um banco de dados relativos aos setores Saúde, Social, Econômico e Ambiental. As dificuldades de seleçäo dos dados säo apresentados bem como um método exclusivamente matemático para interpolaçäo de populaçöes. Um estudo do erro quando se utiliza interpolaçäo linear para as variáveis näo demográficas, indica um erro médio de 12% para interpolaçäo de uma amostra em variáveis do setor saúde, mostrando a viabilidade deste procedimento quando necessário. A caracterizaçäo da funçäo de probabilidade das principais variáveis sugere uma distribuiçäo gaussiana na maioria dos casos

Caracterizaçäo dinâmica de séries históricas de determinantes de saúde / Dinamic characterization of health determinants historical series

Uma metodologia foi desenvolvida para a caracterizaçäo da ordem e do Número de Graus de Liberdade de séries temporais de curta duraçäo (número de pontos menor que 30). O problema dessa caracterizaçäo aparece com frequência em Epidemiologia, na modelagem de surtos epidemiológicos ou determintes de saúde com base em indicadores coletados anualmente. Utilizando-se os cinco primeiros coeficientes da funçäo de autocorrelaçäo, a ordem de um grupo de séries fica determinada pela distância mínima entre os coeficientes das séries e os de conjuntos de séries-padröes previamente construidos a partir da simulaçäo de processos com ordem conhecida. Para as séries de mortalidade das 60 maiores cidades brasileiras no período 1960-1983 verificou-se que a ordem é superior ou igual a 5, o que corresponde a no máximo 0,2 graus de liberdade por ponto. O erro desta estimativa é uma funçäo do tamanho e do número de séries empregados. Estes resultados possuem aplicaçäo no desenvolvimento de critérios de interpolaçäo para séries semelhantes com ausência de dados e na escolha da estrutura de modelos matemáticos de determinantes de saúde