Tamanho do efeito em estudos observacionais na área de Saúde Bucal Coletiva: importância, cálculo e interpretação / Size effect in observational studies in Public Oral Health: importance, calculation and interpretation

Resumo O objetivo deste estudo foi analisar a literatura científica da área de saúde bucal coletiva quanto ao cálculo, apresentação e discussão do tamanho do efeito em estudos observacionais. A literatura cientifica na área (2015 a 2019) foi analisada quanto: a) informações gerais (periódico e diretrizes aos autores, número de variáveis e desfechos), b) objetivo e coerência com o cálculo amostral apresentado; c) tamanho do efeito (apresentação, medida utilizada e coerência com a discussão dos dados e conclusão). Foram analisados 123 artigos, de 66 periódicos. A maioria dos artigos avaliados apresenta um único desfecho (74%) e não menciona a realização de cálculo amostral (69,9%). Dentre os que realizaram, para 70,3% havia coerência entre o cálculo amostral utilizado e o objetivo. Apenas 3,3% dos artigos mencionam o termo tamanho do efeito e 24,4% não o consideram na discussão dos resultados, apesar de terem calculado. A regressão logística foi a metodologia estatística mais utilizada (98,4%) e o Odds Ratio a medida de tamanho do efeito mais utilizada (94,3%), embora não tenha sido citada e discutida como uma medida de tamanho do efeito na maioria dos estudos (96,7%). Os pesquisadores, em sua maioria, restringiram a discussão dos resultados apenas à significância estatística encontrada nas associações testadas.

Abstract The objective of this study was to analyze the scientific literature in public oral health regarding calculation, presentation, and discussion of the effect size in observational studies. The scientific literature (2015 to 2019) was analyzed regarding: a) general information (journal and guidelines to authors, number of variables and outcomes), b) objective and consistency with sample calculation presentation; c) effect size (presentation, measure used and consistency with data discussion and conclusion). A total of 123 articles from 66 journals were analyzed. Most articles analyzed presented a single outcome (74%) and did not mention sample size calculation (69.9%). Among those who did, 70.3% showed consistency between sample calculation used and the objective. Only 3.3% of articles mentioned the term effect size and 24.4% did not consider that in the discussion of results, despite showing effect size calculation. Logistic regression was the most commonly used statistical methodology (98.4%) and Odds Ratio was the most commonly used effect size measure (94.3%), although it was not cited and discussed as an effect size measure in most studies (96.7%). It could be concluded that most researchers restrict the discussion of their results only to the statistical significance found in associations under study.

A Statistical Analysis on Forensic Autopsies Performed in Korea in 2017 / 대한법의학회지

Statistical analysis was performed on national forensic autopsy data collected in the Republic of Korea, with the exception of Ulsan, during 2017. A total of 8,777 cases were categorized based on the region; information was provided by the Police Agency and the Coast Guard regarding sex, age, manner of death, and cause of death. Analysis of the manner of death revealed that 3,971 cases (45.2%) were unnatural deaths, 3,679 cases (41.9%) were natural deaths, and 1,127 cases (12.8%) were unknown deaths. Among the unnatural deaths, the majority of the cases (1,740 cases, 43.8%) were accidents, 1,316 cases (33.1%) were suicide, 385 cases (9.7%) were homicide, and 530 cases (13.3%) were undetermined deaths. Among the unnatural deaths, the majority of the cases (1,575 cases, 39.7%) were trauma, followed by 793 cases (20.0%) of poisoning and 689 cases (17.4%) of asphyxia. Falling down was the major cause of death by trauma (737 cases, 46.8%). As a result of the classification of asphyxia based on previous study, strangulation was the major cause, with 538 cases (78.1%). Among the natural deaths, heart disease was the major cause (1,790 cases, 48.7%), followed by vascular disease (697 cases, 18.9%).

Study for Lists of Forensic Autopsy Request: Quantitative Analysis about Documents Relating to Autopsy Request / 대한법의학회지

In the Republic of Korea, relevant documents are submitted to forensic doctors or agencies when courts grant confiscation warrants for autopsy. If the essential data on unusual death are not submitted at the time of the autopsy, it may be difficult to properly understand the situation relating to an unusual death prior to the autopsy, thus reducing the accuracy of the autopsy. As many as 6,133 out of 6,610 autopsy data (92.8%) in the Republic of Korea in 2015 were analyzed. Most autopsy appraisal requests (99.8%) were submitted. Unusual death occurrence reports (86.0%) and command recommendations of unusual death (70.3%) were submitted in many cases. However, prosecutor commands on unusual death were submitted only in 27.8% cases, and confiscation warrants were not submitted in 7.4% cases. As for postmortem inspection and death scene investigation reports, 29.3% and 34.1% cases were submitted, respectively. In addition to the above two documents, death certificates and records of statement of a relative had significant regional variations (0.3%–80.1%, 3.1%–64.7%, 27.8%–81.3%, and 40.8%–96.8%, respectively). For postmortem inspection and death scene photos, 2.7% and 3.2% were submitted in black-and-white photographs, respectively. The authors propose a list of forensic autopsy requests including autopsy appraisal requests, unusual death occurrence reports, command recommendations of unusual deaths, prosecutor commands on unusual death, and confiscation warrants unconditionally, as an essential document reflecting the progress of investigations. We suggest that postmortem inspection reports and photos, death scene investigation reports and photos, and death certificates should be included as part of postmortem investigation data.

A Statistical Analysis on Forensic Autopsies Performed in Korea in 2016 / 대한법의학회지

A statistical analysis was performed on national forensic autopsy data collected in the Republic of Korea during 2016 to overcome regional limitations and limitations from the number of unusual deaths as reported in the literature over the previous year. A total of 8,335 cases were categorized based on the region, based on requests by the Police Agency and the Coast Guard, gender, age, manner of death, and cause of death. Analysis of the manner of death revealed that 4,028 cases (48.3%) were of unnatural death, 3,447 cases (41.4%) were of natural death, and 860 cases (10.3%) were of unknown death. Among the unnatural deaths, the majority of the manner of deaths (1,584 cases, 39.3%) was accidents, 1,378 cases (34.2%) were suicides, 428 cases (10.6%) were homicides, and 638 cases (15.8%) were undetermined deaths. Among the unnatural deaths, the majority of the cause of deaths (1,518 cases, 37.7%) was due to trauma, followed by 827 cases (20.5%) of poisoning and 732 cases (18.2%) of asphyxia. Falling was the major cause of death by trauma (668 cases, 44.0%). Based on a previous study about asphyxia, strangulation was the major cause, with 569 cases (77.7%). Among the natural deaths, heart disease was the major cause (1,727 cases, 50.1%), followed by vascular disease (587 cases, 17.0%).

Analysis of periodontal data using mixed effects models

A fundamental problem in analyzing complex multilevel-structured periodontal data is the violation of independency among the observations, which is an assumption in traditional statistical models (e.g., analysis of variance and ordinary least squares regression). In many cases, aggregation (i.e., mean or sum scores) has been employed to overcome this problem. However, the aggregation approach still exhibits certain limitations, such as a loss of power and detailed information, no cross-level relationship analysis, and the potential for creating an ecological fallacy. In order to handle multilevel-structured data appropriately, mixed effects models have been introduced and employed in dental research using periodontal data. The use of mixed effects models might account for the potential bias due to the violation of the independency assumption as well as provide accurate estimates.

Análise de Correspondência: bases teóricas na interpretação de dados categóricos em Ciências da Saúde / Correspondence Analysis: a theoretical basis for categorical data interpretation in Health Sciences / Análisis de Correspondencia: bases teóricas para la interpretación de datos categóricos en Ciencias de Salud

Na área biomédica, a ocorrência de dados categóricos é comum, e métodos de análise específicos para este tipo de dado são usados para revelar padrões existentes. A Análise de Correspondência é uma dessas técnicas, utilizada na análise de tabelas de contingência de grande porte. A maioria dos trabalhos publicados em periódicos brasileiros foca apenas na sua interpretação gráfica, não abordando outras potencialidades da técnica. O objetivo do trabalho é mostrar a técnica não limitada à análise gráfica, mas também utilizar estatísticas que permitem sua análise quantitativa. Exemplo mostra que a análise gráfica é enriquecida com a utilização dessas estatísticas, e que a inclusão de uma categoria com baixa ocorrência pode ser considerada como categoria suplementar devido à sua baixa contribuição à inércia. Assim, diminui-se a subjetividade na análise, sendo possível revelar a relação entre as categorias com a análise de resíduos, aspecto este não facilmente observado graficamente. Comparação com a Análise de Componentes Principais mostrou a vantagem da técnica.

Categorical variables are common in the biomedical field, and many descriptive methods have been proposed for revealing intrinsic patterns in data. Correspondence Analysis is an especially useful method for categorical data analysis of large contingency tables. Although numerous studies have been published on this method, most Portuguese-language articles have failed to explore its full potential, focusing only on graphical interpretation. The current paper reviews the method, showing that graphical analysis can be enriched by the right statistics. The article presents the mathematical basis for correspondence analysis and its most frequently used statistics. The procedure has shown that such statistics enrich symmetric map evaluation, that a low relative frequency category can be represented by supplementary category points, and that inertia contributions are highly related to residual analysis of contingency tables, not easily visualized by symmetric maps. Correspondence Analysis has proven advantageous when compared to principal components analysis.

En el campo biomédico, los datos categóricos son frecuentemente utilizados y los métodos de análisis específicos son empleados para revelar patrones intrínsecamente existentes en los mismos. El Análisis de Correspondencias es una de estas técnicas, siendo útil en el análisis de tablas de contingencia con un gran número de clases. A pesar de que muchos artículos han explorado esta técnica, la mayoría de trabajos en revistas brasileñas se centra sólo en su interpretación gráfica. El objetivo de este trabajo es incluir estadísticas que permitan la interpretación cuantitativa de la técnica. Como ejemplo, tenemos el análisis de un mapa simétrico enriquecido con el uso de estadísticas, en el cual la inclusión de una clase de baja ocurrencia puede ser considerada como una categoría suplementaria, debido a su baja contribución a la inercia de datos. Por lo tanto, disminuye la subjetividad en el análisis, siendo posible ahora revelar la relación entre las categorías con el análisis residual, lo que no es fácil observar en los gráficos. La comparación con el análisis de componentes principales mostró sus ventajas.

The Statistical Analysis on Legal Autopsy Performed in Korea during 2013 Year / 대한법의학회지

Medicolegal autopsy is a vital tool for obtaining reliable injury mortality data. This study statistically analyzed data obtained from medicolegal autopsies performed in Korea in 2013. The aim of this study was to analyze various aspects of the 4,861 deaths that were categorized as unusual in Korea in 2013. A total of 4,861 deaths were analyzed by gender, age, manner of death, and cause of death. Of the 4,861 deaths, 3,542 (73.3%) were of men and 1,302 (26.7%) were of women. With respect to the manner of death, 54.4% were recorded as unnatural deaths, 38.8% were natural deaths, and 6.9% had unknown causes. Of the 2,642 unnatural deaths, 45.0% were determined to be accidental deaths, 26.3% suicidal, 16.9% homicidal, and 11.8% undetermined. Of the total number of unnatural deaths, 42.1% were trauma-related deaths, for which falling down accounted for 33.8%. Asphyxiation accounted for 16.0% of unnatural deaths, and of these, the predominant cause was hanging (58.8%). In addition, 14.4% of deaths were due to drowning, 12.9% poisoning, 11.0% thermal injuries, 1.8% complications arising from medical procedures, and 0.8% electrocution, starvation, or neglect. Among the 1,886 natural deaths, heart diseases accounted for 52.0% and vascular diseases accounted for 16.9%. Of the 196 deaths among children under the age of 10 years, 41.8% were recorded as unnatural deaths, 45.1% were natural deaths, and 1.5% had unknown causes.

Statistical methods used in articles published by the Journal of Periodontal and Implant Science

PURPOSE: The purposes of this study were to assess the trend of use of statistical methods including parametric and nonparametric methods and to evaluate the use of complex statistical methodology in recent periodontal studies. METHODS: This study analyzed 123 articles published in the Journal of Periodontal & Implant Science (JPIS) between 2010 and 2014. Frequencies and percentages were calculated according to the number of statistical methods used, the type of statistical method applied, and the type of statistical software used. RESULTS: Most of the published articles considered (64.4%) used statistical methods. Since 2011, the percentage of JPIS articles using statistics has increased. On the basis of multiple counting, we found that the percentage of studies in JPIS using parametric methods was 61.1%. Further, complex statistical methods were applied in only 6 of the published studies (5.0%), and nonparametric statistical methods were applied in 77 of the published studies (38.9% of a total of 198 studies considered). CONCLUSIONS: We found an increasing trend towards the application of statistical methods and nonparametric methods in recent periodontal studies and thus, concluded that increased use of complex statistical methodology might be preferred by the researchers in the fields of study covered by JPIS.

Big Data Analysis Using Modern Statistical and Machine Learning Methods in Medicine / 대한배뇨장애요실금학회지

In this article we introduce modern statistical machine learning and bioinformatics approaches that have been used in learning statistical relationships from big data in medicine and behavioral science that typically include clinical, genomic (and proteomic) and environmental variables. Every year, data collected from biomedical and behavioral science is getting larger and more complicated. Thus, in medicine, we also need to be aware of this trend and understand the statistical tools that are available to analyze these datasets. Many statistical analyses that are aimed to analyze such big datasets have been introduced recently. However, given many different types of clinical, genomic, and environmental data, it is rather uncommon to see statistical methods that combine knowledge resulting from those different data types. To this extent, we will introduce big data in terms of clinical data, single nucleotide polymorphism and gene expression studies and their interactions with environment. In this article, we will introduce the concept of well-known regression analyses such as linear and logistic regressions that has been widely used in clinical data analyses and modern statistical models such as Bayesian networks that has been introduced to analyze more complicated data. Also we will discuss how to represent the interaction among clinical, genomic, and environmental data in using modern statistical models. We conclude this article with a promising modern statistical method called Bayesian networks that is suitable in analyzing big data sets that consists with different type of large data from clinical, genomic, and environmental data. Such statistical model form big data will provide us with more comprehensive understanding of human physiology and disease.

Inadequate Post-hoc Statistical Power Analysis in Journal of the Korean Society of Emergency Medicine

PURPOSE: Negative studies provide valuable information. However, conducting studies with inadequate power is unethical and an inefficient use of resources. Moreover, inaccurate interpretations from underpowered studies result in false conclusions that alter clinical interventions and deter further research. The purpose of this study was to determine the prevalence of negative studies with inadequate power in the Journal of the Korean Society of Emergency Medicine (JKEM). METHODS: We assessed all papers in JKEM from 2009 to 2012. We sought published evidence that a post-hoc power analysis had been performed in association with the main hypothesis of the paper. All clinical research studies containing the phrase "no difference" were identified. Data necessary for power calculation were extracted from applicable studies. RESULTS: There were a total of 351 papers in which a statistical comparison was undertaken. Out of 351 original articles, 170(48.4%) were negative studies that contained enough information for analysis. Out of 126 negative studies in JKEM, only 21((16.7%) had performed a power analysis demonstrating adequate sample size. In addition, only 6.3% of dichotomous variable articles and 10.3% of continuous variable articles had adequate power. Levels of adeadequate power in negative studies did not improve over time (p=0.148). CONCLUSION: Many negative studies in JKEM are inconclusive because they lack the adequate power to detect even large differences between groups. Therefore, it is imperative to consider power when interpreting the literature. When designing future research, power calculations should be performed to ensure sufficient patient recruitment to attain clinically meaningful results.

Uso da imputação múltipla de dados faltantes: uma simulação utilizando dados epidemiológicos / Multiple imputations for missing data: a simulation with epidemiological data

Em situações com dados faltantes, é comum restringir-se à análise dos sujeitos com dados completos. Porém, as estimativas com apenas esses sujeitos podem tornar-se viesadas. A prática de preenchimento de dados faltantes é a chamada técnica de imputação. Este trabalho tem como objetivo divulgar o método de imputação múltipla. Em um conjunto de dados de 470 pacientes cirúrgicos, foram ajustados modelos logísticos para o desfecho óbito. Foram gerados dois conjuntos de dados incompletos: um com 5 por cento e outro com 20 por cento de dados faltantes para uma variável. Foram ajustados modelos para o conjunto completo, com dados faltantes e para o conjunto completado por imputação múltipla. As estimativas obtidas pela análise dos conjuntos com dados faltantes e com o conjunto completo foram diferentes, principalmente as do conjunto com 20 por cento de dados faltantes. A imputação múltipla utilizada pareceu eficiente, pois os resultados conseguidos com o banco completado por imputações foram próximos dos obtidos com o conjunto completo. Porém, um coeficiente deixou de ser estatisticamente significativo. A imputação múltipla se mostrou superior à análise do conjunto com dados faltantes, que desconsiderou os casos incompletos.

In situations with missing data, statistical analyses are usually limited to subjects with complete data. However, such estimates may be biased. The method of "filling in" missing data is called imputation. This article aimed to present a multiple imputation method. From a data set of 470 surgical patients, logistic models were developed for death as the outcome. Two incomplete data sets were generated: one with 5 percent and another with 20 percent of missing data in a single variable. Logistic models were fitted for the complete and incomplete data sets and for the data set completed by multiple imputations. Estimates obtained for the data set with missing data were different from those observed in the complete data set, mainly in the situation with 20 percent of missing data. The multiple imputation used here appeared efficient, producing very similar results to those obtained with the complete data set. However, one coefficient became non-significant. The analysis using multiple imputations was considered superior to using the data sets that excluded incomplete cases from the analysis.

Alternatives in modeling of body mass index as a continuous response variable and relevance of residual analysis

This article presents alternatives for modeling body mass index (BMI) as a continuous variable and the role of residual analysis. We sought strategies for the application of generalized linear models with appropriate statistical adjustment and easy interpretation of results. The analysis included 2,060 participants in Phase 1 of a longitudinal study (Pró-Saúde Study) with complete data on weight, height, age, race, family income, and schooling. In our study, the residual analysis of models estimated by maximum likelihood methods yielded inadequate adjustment. The transformed response variable resulted in a good fit but did not lead to estimates with straightforward interpretation. The best alternative was to apply quasi-likelihood as the estimation method, presenting a better adjustment and constant variance. In epidemiological data modeling, researchers should always take trade-offs into account between adequate statistical techniques and interpretability of results.

Neste artigo, discutem-se alternativas de modelagem do índice de massa corporal (IMC), analisado como variável contínua, e a análise de resíduos. Buscaram-se estratégias de aplicação dos modelos lineares generalizados adequadas tanto do ponto de vista do ajuste estatístico quanto da facilidade de interpretação dos resultados. Nestas análises, foram incluídos dados relativos a 2.060 participantes da Fase 1 de estudo longitudinal (Estudo Pró-Saúde), com informação completa de peso, estatura, idade, raça/cor, renda familiar e escolaridade. Em nosso estudo, a análise de resíduos dos modelos estimados pelo método da máxima verossimilhança, amplamente utilizado, não possibilitou ajuste adequado dos modelos aos dados. A transformação da variável resposta, apesar de resultar em um bom ajuste, não conduziu a estimativas de fácil interpretação. Considerou-se como melhor alternativa a mudança do método de estimação para quase-verossimilhança. Assim, melhor ajuste foi alcançado e a variância permaneceu constante. Na modelagem de dados epidemiológicos, cabe aos pesquisadores buscarem o melhor equilíbrio entre a aplicação adequada de técnicas estatísticas e a facilidade de interpretação dos dados.