Class of 2015 Abstract / Objectives: To assess the efficacy of fluconazole prophylaxis in the prevention of coccidioidomycosis in the post-heart transplant patient and to identify risk factors for coccidioidomycosis infection. Methods: Heart transplant patients with ICD-9 code V42.1 from October 2001 to October 2013, were selected and electronic medical records were retrospectively reviewed for coccidioidomycosis history, Coccidiodes serologies, reason for transplantation, immunosuppressive drug therapy regimens, rejection treatment course, fluconazole dose, and demographics. Negative Coccidiodes serology results post transplantation relative to negative Coccidiodes serology results prior to transplantation will be determined using a Chi Square test. Risk factors for disease contraction will be analyzed using multivariate logistic regression. Results: Between October 2001 and October 2013, 244 patients received a heart transplant at this institution. Fourteen (5.7%) heart transplant recipients with a negative Coccidiodes serology pre-transplantation had a positive Coccidiodes serology post-transplantation. Nine (64.2%) of those recipients received antifungal prophylaxis (p=0.16). Risk factors for developing a positive Coccidiodes serology included using tacrolimus (p=0.05) and non-ischemic cardiomyopathy (p=0.04). Conclusions: Antifungal prophylaxis does not reduce the risk of developing a positive Coccidiodes serology after heart transplantation. Risk factors for developing a positive Coccidiodes serology include the use of tacrolimus and having non-ischemic cardiomyopathy prior to transplant.

No description available.

Over the past several decades rates of spine surgeries in the U.S. have increased dramatically. Spinal fusion surgery rates, in particular, have grown exponentially despite being one of the most costly, invasive, and controversial methods for treating patients suffering from back conditions. Furthermore, lumbar fusion surgeries continue to be performed at increasing rates despite a lack of scientific evidence and consensus that they are cost-effective and produce better clinical outcomes than less radical treatment of lower back pain. As a result, large amounts of healthcare dollars continue to be invested in these costly procedures which are potentially dangerous and have questionable efficacy in terms of improving patient outcomes. Importantly, there is a lack of population studies in the literature on spinal fusion surgeries from a health services research perspective. Therefore, the present research is a population based study using an administrative database and includes patients of all ages and payer types. The data used in the present study come from the Florida Agency for Health Care Administration (AHCA) and include all hospitalizations in Florida in 2010. The objective of the study is to analyze the incidence of spinal fusion surgeries in Florida hospitals for patients of all ages and payer types by demographic variables to understand who gets these surgeries and for which conditions. The first null hypothesis is that there are no statistically significant predictors of the incidence of lumbar/lumbosacral, dorsal/dorsolumbar spinal fusion surgeries in Florida hospitals. Logistic regression was used to analyze the incidence of fusion surgeries. The binary dependent variable was coded as a "1" for all patients who were a case (i.e. they received one of the five procedure codes being studied in the present research) and a "0" for all patients who were controls (meaning they did not receive any of the five fusion procedure codes). Logistic regression was used to predict the probability of an observation being a "1" given the independent variables included in the model. Additionally, hospital charges were analyzed to understand the associated hospital charges with these surgeries. The second null hypothesis is that there are no statistically significant predictors of the charges of Lumbar/Lumbosacral, Dorsal/Dorsolumbar spinal fusion surgeries in Florida Hospitals. A mixed effects model was used to test this hypothesis and the fixed effects which were included in the model were gender, age, race, principal payer, and principal procedure. A mixed effects model was chosen due to the fact that cases who had surgeries performed at the same hospital are not independent and therefore the data were clustered on hospitals. A random intercept term was used to address this fact. SAS software was used to complete all of the analyses. In 2010, there were 16,236 Lumbar/Lumbosacral, Dorsal/Dorsolumbar fusion surgery cases in Florida hospitals that were included in the case population and 21,856 individuals included in the control population for a total of 38,092 included in the study population. An understanding of who is most likely to receive a fusion surgery, at what age, and for which diagnoses, as has been done here, is extremely important. This knowledge can help researchers, policy makers, and physicians alike. Comprehensive physician practice guidelines for performing fusion surgeries still do not exist in the year 2013; therefore, in order to have the greatest impact, the efforts for creating the guidelines should be focused on those individuals who are most likely to receive fusions as shown for the first time by the data analyzed here. Given the high incidence of these surgeries in Florida alone, the need for practice guidelines cannot be overstated. The total hospital charges in Florida hospitals for the 16,236 cases were $2,095,413,584. Despite having the same principal diagnoses and a similar number of additional diagnoses, patients who received a fusion surgery resulted in approximately three times the charges as those incurred by the controls. Overall, the high incidence and charges for fusion surgeries shown in this study emphasize the importance of having a better understanding of when these surgeries are justified and for which patients. Without comprehensive practice guidelines established through evidence-based research this is difficult, if not impossible, to accomplish. The diagnoses which are most prevalent and show the most inconsistencies between cases may be a good starting point for such guidelines.

Many manual biomedical annotation tasks can be categorized as instances of the typical multi-label classification problem where several categories or labels from a fixed set need to assigned to an input instance. MeSH term assignment to biomedical articles and diagnosis code extraction from medical records are two such tasks. To address this problem automatically, in this thesis, we present a way to utilize latent associations between labels based on output label sets. We used random indexing as a method to determine latent associations and use the associations as a novel feature in a learning-to-rank algorithm that reranks candidate labels selected based on either k-NN or binary relevance approach. Using this new feature as part of other features, for MeSH term assignment, we train our ranking model on a set of 200 documents, test it on two public datasets, and obtain new state-of-the-art results in precision, recall, and mean average precision. In diagnosis code extraction, we reach an average micro F-score of 0.478 based on a large EMR dataset from the University of Kentucky Medical Center, the first study of its kind to our knowledge. Our study shows the advantages and potential of random indexing method in determining and utilizing implicit relationships between labels in multi-label classification problems.

Diagnosis code assignment is a field that looks at automatically assigning diagnosis codes to free-text clinical notes. Assigning a diagnosis code to clinical notes manually needs expertise and time. Being able to do this automatically makes getting structured data from free-text clinical notes in Electronic Health Records easier. Furthermore, it can also be used as decision support for clinicians where they can input their notes and get back diagnosis codes as a second opinion. This project investigates the effects of using the hierarchies the diagnosis codes are structured in when training the diagnosis code assignment models compared to models trained with a standard loss function, binary cross-entropy. This has been done by using the hierarchy of two systems of diagnosis codes, ICD-9 and SNOMED CT, where one hierarchy is more detailed than the other. The results showed that hierarchical training increased the recall of the models regardless of what hierarchy was used. The more detailed hierarchy, SNOMED CT, increased the recall more than what the use of the less detailed ICD-9 hierarchy did. However, when using the more detailed SNOMED CT hierarchy the precision of the models decreased while the differences in precision when using the ICD-9 hierarchy was not statistically significant. The increase in recall did not make up for the decrease in precision when training with the SNOMED CT hierarchy when looking at the F1-score that is the harmonic mean of the two metrics. The conclusions from these results are that using a more detailed hierarchy increased the recall of the model more than when using a less detailed hierarchy. However, the overall performance measured in F1-score decreased when using a more detailed hierarchy since the other metric, precision, decreased by more than what recall increased. The use of a less detailed hierarchy maintained its precision giving an increase in overall performance. / Diagnoskodstilldeling är ett fält som undersöker hur man automatiskt kan tilldela diagnoskoder till fri-text läkaranteckningar. En manuell tildeling kräver expertis och mycket tid. Förmågan att göra detta automatiskt förenklar utvinning av strukturerad data från fri-text läkaranteckningar i elektroniska patientjournaler. Det kan även användas som ett hjälpverktyg för läkare där de kan skriva in sina läkaranteckningar och få tillbaka diagnoskoder som en andra åsikt. Detta arbete undersöker effekterna av att ta användning av hierarkierna diagnoskoderna är strukturerade i när man tränar modeller för diagnoskodstilldelning jämfört med att träna modellerna med en vanlig loss-funktion. Det här kommer att göras genom att använda hierarkierna av två diagnoskod-system, SNOMED CT och ICD-9, där en av hierarkierna är mer detaljerad. Resultaten visade att hierarkisk träning ökade recall för modellerna med båda hierarkierna. Den mer detaljerade hierarkien, SNOMED CT, gav en högre ökning än vad träningen med ICD-9 gjorde. Trots detta minskade precision av modellen när man den tränades med SNOMED CT hierarkin medan skillnaderna i precision när man tränade hierarkiskt med ICD-9 jämfört med vanligt inte var statistiskt signifikanta. Ökningen i recall kompenserade inte för minskningen i precision när modellen tränades med SNOMED CT hierarkien som man kan see på F1-score vilket är det harmoniska medelvärdet av de recall och precision. Slutsatserna man kan dra från de här resultaten är att en mer detaljerad hierarki kommer att öka recall mer än en mindre detaljerad hierarki ökar recall. Trots detta kommer den totala prestandan, som mäts av F1-score, försämras med en mer detaljerad hierarki eftersom att recall minskar mer än vad precision ökar. En mindre detaljerad hierarki i träning kommer bibehålla precision så att dens totala prestandan förbättras.

Millions of people in the United States alone suffer from undiagnosed or late-diagnosed chronic diseases such as Chronic Kidney Disease and Type II Diabetes. Catching these diseases earlier facilitates preventive healthcare interventions, which in turn can lead to tremendous cost savings and improved health outcomes. We develop algorithms for predicting disease occurrence by drawing from ideas and techniques in the field of machine learning. We explore standard classification methods such as logistic regression and random forest, as well as more sophisticated sequence models, including recurrent neural networks. We focus especially on the use of medical code data for disease prediction, and explore different ways for representing such data in our prediction algorithms.

No description available.

Évaluer la sécurité des patients hospitalisés constitue un enjeu majeur de la gestion des risques pour les services de santé. Le développement d’indicateurs destinés à mesurer les événements indésirables liés aux soins (EIS) est une étape cruciale dont le défi principal repose sur la performance des données utilisées. Le développement d’indicateurs de la sécurité des patients – les Patient Safety Indicators (PSIs) – par l’Agency for Healthcare Research and Quality (AHRQ) aux Etats Unis, utilisant des codes de la 9ème révision (cliniquement modifiée) de la Classification Internationale des Maladies (CIM) présente des perspectives intéressantes. Nos travaux ont abordé cinq questions fondamentales liées au développement de ces indicateurs : la définition du cadre nosologique, la faisabilité de calcul des algorithmes et leur validité, la qualité des données pour coder les diagnostics médicaux à partir de la CIM et leur performance pour comparer plusieurs pays, et la possibilité d’établir une valeur de référence pour comparer ces indicateurs. Certaines questions demeurent cependant et nous proposons des pistes de recherche pour améliorer les PSIs : une meilleure définition des algorithmes et l’utilisation d’autres sources de données pour les valider (i.e., données de registre), ainsi que l’utilisation de modèles d’ajustement utilisant l’index de Charlson, le nombre moyen de diagnostics codés et une variable de la valeur prédictive positive, afin de contrôler les variations du case-mix et les différences de qualité du codage entre hôpitaux et pays. / Assessing safety among hospitalized patients is a major issue for health services. The development of indicators to measure adverse events related to health care (HAE) is a crucial step, for which the main challenge lies on the performance of the data used for this approach. Based on the limitations of the measurement in terms of reproducibility and on the high cost of studies conducted using medical records audit, the development of Patient Safety Indicators (PSI) by the Agency for Healthcare Research and Quality (AHRQ) in the United States, using codes from the clinically modified 9th revision of the International Classification of Diseases (ICD) shows interesting prospects. Our work addressed five key issues related to the development of these indicators: nosological definition; feasibility and validity of codes based algorithms; quality of medical diagnoses coding using ICD codes, comparability across countries; and possibility of establishing a benchmark to compare these indicators. Some questions remain, and we suggest several research pathways regarding possible improvements of PSI based on a better definition of PSI algorithms and the use of other data sources to validate PSI (i.e., registry data). Thus, the use of adjustment models including the Charlson index, the average number of diagnoses coded and a variable of the positive predictive value should be considered to control the case-mix variations and differences of quality of coding for comparisons between hospitals or countries.

No description available.

Résumé : La douleur lombaire (DL) est l’une des conditions musculosquelettiques les plus fréquentes et coûteuses au Canada. La prévalence annuelle de DL aigüe varierait de 19 % à 57 %, et un patient sur quatre souffrirait de récurrence dans la même année. La présente étude vise donc à produire une analyse descriptive de l’épidémiologie de la DL récurrente à l’échelle de la population. Une nouvelle approche méthodologique est proposée afin d’optimiser l’identification de vrais cas incidents de DL récurrente à partir d’une analyse secondaire de données administratives. Puisque 10 % des patients ayant de la DL seraient responsables de 80 % des coûts qui y sont associés, nous avons également déterminé la tendance séculaire des coûts d’interventions médicales des patients récurrents incidents entre 2003 et 2008. En utilisant le fichier des services médicaux rémunérés à l’acte de la Régie de l’assurance maladie du Québec, des cohortes prévalentes ont été construites à partir de 401 264 dossiers de patients ayant consulté au moins trois fois pour de la DL entre 1999 et 2008. Onze ans d’historique médical des 81 329 patients de la cohorte de 2007 ont ensuite été analysés afin d’exclure les patients ayant eu des consultations antérieures de DL. Une valeur prédictive positive et un coefficient de Kappa élevés ont permis d’identifier une clairance optimale pour récupérer les cas véritablement incidents. Les coûts de consultations ont ensuite été calculés pour tous les patients incidents de 2003 à 2007 à partir des manuels de facturation. Nous avons observé une prévalence annuelle de la DL récurrente de 1,64 % en 2000 chez les hommes diminuant à 1,33 % en 2007. Cette baisse a majoritairement eu lieu dans le groupe d’âge des 35-59 ans. Les femmes âgées (> 65 ans) étaient 1,4 fois plus à risque de consulter un médecin de manière récurrente que les hommes du même âge. L’incidence annuelle de la DL en 2007 était de 242 par 100 000 personnes. Les hommes de 18 à 34 ans étaient 1,2 fois plus à risque que les femmes de développer un premier épisode récurrent et les personnes âgées 1,9 fois plus à risque que les jeunes. L’incidence annuelle a diminué de 12 % entre 2003 et 2007 pendant que les coûts totaux augmentaient de 1,4 %. La médiane des coûts était la plus élevée chez les femmes âgées et tendait à augmenter dans le temps. Ces analyses secondaires suggèrent de s’intéresser particulièrement à la DL chez les personnes très âgées, et de déterminer si la baisse de fréquence de consultations récurrentes observée dans le temps est liée à une meilleure gestion de la DL ou à un problème d’accessibilité. Les coûts devraient faire l’objet d’un suivi continu pour limiter les hausses. // Abstract : Low back pain (LBP) is one of the most frequent and costly musculoskeletal health conditions in Canada. Annual prevalence was found to vary between 19 % and 57 % and likely one out of four patients experience a LBP recurrence within one year. The body of knowledge on the prevalence of recurrent LBP is still limited. This study sought to present a descriptive analysis on the epidemiology of recurrent LBP in a medical population. A new methodology is also proposed to identify true cases of incident recurrent LBP. Since 10 % of LBP patients have been reported to generate 80 % of the costs, we will sought to determine the secular trend of medical costs for the incident cohorts of 2003 to 2008. Using the Canadian province of Quebec medical administrative physicians’ claims database, 401 264 prevalent claims-based recurrent LBP patients were identified between 1999 to 2008 for having consulted at least three times for LBP in a period of 365 days. The medical history of 81 329 prevalent patients in 2007 was screened for a retrospective period of 11 years. High positive predictive values and Kappa statistics were used to determine the optimal clearance period for capturing true incidence cases among patients with no prior encounters for LBP. Physicians’ claims manuals were then used to apply a price for every intervention provided to LBP incident patients in their index year and follow-up years. We observed a decrease from 1.64 % to 1.33 % in the LBP annual prevalence between 2000 and 2007 for men. This decrease was mostly observed between 35 and 59 years of age. Older women (≥ 65 years) were 1.4 times more at risk to consult a physician for LBP in a recurrent manner than older men. The annual incidence in 2007 of adult claims-based recurrent LBP was 242 per 100 000 persons. Males of 18 to 34 years of age were found 1.2 times more at risk than their counterparts. Altogether, elderlies were 1.9 times more at risk than young adults to consult in a recurrent manner for LBP. The annual incidence decreased by 12 % between 2003 and 2007, while the direct costs increase by 1.4 %. The median cost for consultations was highest for elder women and increasing in time. These secondary analyses emphasize the importance to keep the watch on the elders in regards to LBP, and to determine if the timely decrease in morbidity is related to improvements in LBP management or to a medical accessibility issue. Also, costs will need to be surveyed on a regular basis to limit the impact of future increases.