UNSUPERVISED INDEXING OF MEDLINE ARTICLES THROUGH GRAPH-BASED RANKING

The biomedical literature is extensively catalogued and indexed in MEDLINE. MEDLINE indexing is done by trained human indexers, who identify the most important concepts in each article, and is expensive and inconsistent. Automating the indexing task is difficult: the National Library of Medicine produces the Medical Text Indexer (MTI), which suggests potential indexing terms to the indexers. MTI’s output is not good enough to work unattended. In my thesis, I propose a different way to approach the indexing task called MEDRank. MEDRank creates graphs representing the concepts in biomedical articles and their relationships within the text, and applies graph-based ranking algorithms to identify the most important concepts in each article. I evaluate the performance of several automated indexing solutions, including my own, by comparing their output to the indexing terms selected by the human indexers. MEDRank outperformed all other evaluated indexing solutions, including MTI, in general indexing performance and precision. MEDRank can be used to cluster documents, index any kind of biomedical text with standard vocabularies, or could become part of MTI itself.

Using incomplete citation data for MEDLINE results ranking.

Information overload is a significant problem for modern medicine. Searching MEDLINE for common topics often retrieves more relevant documents than users can review. Therefore, we must identify documents that are not only relevant, but also important. Our system ranks articles using citation counts and the PageRank algorithm, incorporating data from the Science Citation Index. However, citation data is usually incomplete. Therefore, we explore the relationship between the quantity of citation information available to the system and the quality of the result ranking. Specifically, we test the ability of citation count and PageRank to identify "important articles" as defined by experts from large result sets with decreasing citation information. We found that PageRank performs better than simple citation counts, but both algorithms are surprisingly robust to information loss. We conclude that even an incomplete citation database is likely to be effective for importance ranking.

Using incomplete citation data for MEDLINE results ranking.

Information overload is a significant problem for modern medicine. Searching MEDLINE for common topics often retrieves more relevant documents than users can review. Therefore, we must identify documents that are not only relevant, but also important. Our system ranks articles using citation counts and the PageRank algorithm, incorporating data from the Science Citation Index. However, citation data is usually incomplete. Therefore, we explore the relationship between the quantity of citation information available to the system and the quality of the result ranking. Specifically, we test the ability of citation count and PageRank to identify "important articles" as defined by experts from large result sets with decreasing citation information. We found that PageRank performs better than simple citation counts, but both algorithms are surprisingly robust to information loss. We conclude that even an incomplete citation database is likely to be effective for importance ranking.

Case-control study of factors associated with virologic failure in HIV-infected children on HAART in Botswana, a developing African country

According to the United Nations Program on HIV/AIDS (UNAIDS, 2008), in 2007 about 67 per cent of all HIV-infected patients in the world were in Sub-Saharan Africa, with 35% of new infections and 38% of the AIDS deaths occurring in Southern Africa. Globally, the number of children younger than 15 years of age infected with HIV increased from 1.6 million in 2001 to 2.0 million in 2007 and almost 90% of these were in Sub-Saharan Africa. (UNAIDS, 2008).^ Both clinical and laboratory monitoring of children on Highly Active Anti-Retroviral Therapy (HAART) are important and necessary to optimize outcomes. Laboratory monitoring of HIV viral load and genotype resistance testing, which are important in patient follow-up to optimize treatment success, are both generally expensive and beyond the healthcare budgets of most developing countries. This is especially true for the impoverished Sub-Saharan African nations. It is therefore important to identify those factors that are associated with virologic failure in HIV-infected Sub-Saharan African children. This will inform practitioners in these countries so that they can predict which patients are more likely to develop virologic failure and therefore target the limited laboratory monitoring budgets towards these at-risk patients. The objective of this study was to examine those factors that are associated with virologic failure in HIV-infected children taking Highly Active Anti-retroviral Therapy in Botswana, a developing Sub-Saharan African country. We examined these factors in a Case-Control study using medical records of HIV-infected children and adolescents on HAART at the Botswana-Baylor Children's Clinical Center of Excellence (BBCCCOE) in Gaborone, Botswana. Univariate and Multivariate Regression Analyses were performed to identify predictors of virologic failure in these children.^ The study population comprised of 197 cases (those with virologic failure) and 544 controls (those with virologic success) with ages ranging from 3 months to 16 years at baseline. Poor adherence (pill count <95% on at least 3 consecutive occasions) was the strongest independent predictor of virologic failure (adjusted OR = 269.97, 95% CI = 104.13 to 699.92; P < 0.001). Other independent predictors of virologic failure identified were: First Line NNRTI with Nevirapine (OR = 2.99, 95% CI = 1.19 to7.54; P = 0.020), Baseline HIV-1 Viral Load >750,000/ml (OR = 257, 95% CI = 1.47 to 8.63; P = 0.005), Positive History of PMTCT (OR = 11.65, 95% CI = 3.04-44.57; P < 0.001), Multiple Care-givers (>=3) (OR = 2.56, 95% CI = 1.06 to 6.19; P = 0.036) and Residence in a Village (OR = 2.85, 95% CI = 1.36 to 5.97; P = 0.005).^ The results of this study may help to improve virologic outcomes and reduce the costs of caring for HIV-infected children in resource-limited settings. ^ Keywords: Virologic Failure, Highly Active Anti-Retroviral Therapy, Sub-Saharan Africa, Children, Adherence.^

Lifetime prevalence and risk factors for cancer in a Mexican American population with severe health disparities: The Cameron County Hispanic Cohort

Retrospective data from the Cameron Country Hispanic Cohort (1) were analyzed to assess the burden of cancer in the Mexican American population living in Brownsville TX. Data provided by the study participants for themselves and their parents and other extended relatives on cancer and related risk factors were used to determine both the prevalence of cancer and these risk factors as well as any associations between them. Lifetime incidence of cancer among the study participants was of 2.8%. Lifetime incidence of cancer among the parents of the study population was calculated for cancer in general and for specific cancer sites to determine the ranking of occurrence of each type of cancer. Some cancer types in this population were ranked higher than what would be expected when compared with national data from Hispanics in the U.S, these were: Liver cancer (3rd vs. 7th nationally in males and 6th vs. 13th nationally in females), stomach cancer (4th vs. 8th nationally in males and 5th vs. 11th nationally in females) and ovarian cancer (3rd vs. 8th nationally in females). A significant association with cancer was found for being born in the United States compared to being born elsewhere (O.R. 1.62, 95% C.I. 1.01–2.60) among study participants and the same association was also found between birth of parents in the United States regardless of gender for cancers in general (O.R. 1.38 95% C.I. 1.12–1.70), stomach cancer (O.R. 1.92 95% C.I. 1.01–3.67) and colorectal cancer (O.R. 2.93 95% C.I. 1.28–6.72). Having been born in the United States and having a family history of cancer was also found to be significantly associated with other risk factors for cancer such as obesity, diabetes and insulin resistance, both among the parents and the participant population, suggesting these interactions are complex. These high rates of cancer and particular prominence of less usual cancer such as liver and ovary in health disparities warrant evaluation of early detection strategies.^

The Relationship of Provincial Economic Level and Child Injury Mortality in Thailand: A Cross-Sectional Analysis in a Middle Income Country

The objective of this cross-sectional study was to examine the relationship of provincial economic development indices with incidences of child injury mortality in Thailand from 1999 - 2001. All injury deaths among children age 1-14 years were included. The independent variables included gross provincial product per capita (GPP/c), poverty and inequality indices, material and social deprivation indices, population in rural/ urban areas, and migration. Due to multicollinearity of such variables, the 76 provinces were categorized by GPP/c quartile, and means of overall injury, drowning, and transport-related mortality rates were compared among quartile groups. Spearman’s rho correlation between GPP/c and injury mortality rates was also performed. Finally, factor analysis was employed to create a set of factors to be treated as uncorrelated variables and stepwise multiple regression was carried out for the effects of the factors on injury mortality rates. A significant direct relationship was observed between GPP/c and overall injury mortality among children age 1-4 years, and 10-14 year-olds of both genders. Drowning was the main cause of this relationship among children age 1-4 years, and transport-related injury was the principle cause among children age 10-14 years. Conversely, provinces with lower GPP/c experienced higher injury mortality rates among school-age children 5-9 years old for both genders, mostly due to drowning. Factor analysis, and multiple regression results confirmed the relationships between economic development and injury mortality rates. These findings revealed that economic development had an adverse impact on injury-related mortality among children 1 to 4 and 10 to14 in Thailand.

Editorial: Take Me Home, Down Country Roads

Those of us committed to the tenets of Family Preservation must advocate for increased awareness and attention to the needs of children and their families in rural America. "Country roads" and the rural spaces they traverse have been eulogized by many poets and song writers as ideal places to live. But they may not be ideal for everyone. The past few months, it has become all too evident that rural America is not immune to acts of extreme violence by troubled children. Even though almost 1/3 of American youth live in rural areas, they have been "virtually ignored by mental health service planners and providers"(Cutrona, Halvorson, & Russell, 1996, p. 217). Mental health risk factors such as poverty, parental alcohol abuse, and family instability are on the rise in rural areas, and there has been an increase in suicide attempts, family violence, depression, and alcohol abuse (Cutrona, Halvorson, & Russell, 1996; Petti & Leviton, 1986; National Mental Health Association, 1988). Native Americans are especially concerned about the increases in child abuse and neglect, depression, substance abuse, and suicide in their communities.