Dissertations / Theses on the topic 'Clinical risk prediction'

This work presents doctors with a model of the estimated degree of risk of rare and important neonatal outcomes to aid in better decisions and improved allocation of equipment and resources. An extensive list of admission day parameters is reduced to minimum variable sets to create models for outcomes that are relevant to decision-making in the neonatal intensive care unit. Models are applied to a special collection of cases and compared to neonatologists' risk estimates. A comparative analysis of physician's predictions and the models' discrimination abilities highlights areas of success and areas that can be improved for future trials. Doctors responded positively to the prediction interface concept and to the estimated risk stratification models. Physicians' strengths identified outcomes that could benefit from increased sensitivity. A substantial effort was made to conduct the usability and performance evaluations within the ethical standards that are especially important for engineering healthcare management applications.

Assessing offenders' risk of future violent behavior continues to be an important yet controversial role of forensic psychologists. A key debate is the relative effectiveness of assessment methods. Specifically, actuarial methods (see Quinsey et al., 1998 for a review) have been compared and contrasted to clinical and structured clinical methods (see e.g. Hart, 1998; Webster et al., 1997). Proponents of each approach argue for its superiority, yet validity studies have made few formal comparisons. In advancing the available research, the present study examines systematically the type of forensic case (i.e., sexual violence versus nonsexual violence) and type of assessment method (i.e., actuarial, structured clinical, and unstructured clinical). As observed by Borum, Otto, and Golding (1993), forensic decision making can also be influenced by the presence of certain extraneous clinical data. To address these issues, psychologists and doctoral students attending the American Psychology Law Society conference were asked to make several ratings regarding the likelihood of future sexual and nonsexual violence based on data derived from actual defendants with known outcomes. Using a mixed factorial design, each of these assessment methods were investigated for its influence on decision-makers regarding likelihood of future violence and sexually violent predator commitments. Finally, the potentially biasing effects of victim impact statements on resultant decisions were also explored.

Objectives: Cardiovascular disease is the leading cause of mortality and morbidity in the developed world. Surgery can improve prognosis and relieve symptoms. Risk prediction models are increasingly being used to inform clinicians and patients about the risks of surgery, to facilitate clinical decision making and for the risk-adjustment of surgical outcome data. The importance of risk prediction models in cardiovascular surgery has been highlighted by the publication of cardiovascular surgery outcome data and the need for risk-adjustment. The overall objective of this thesis is to advance risk prediction modelling in cardiovascular surgery with a focus on the development of models for elective AAA repair and assessment of models for cardiac surgery. Methods: Three large clinical databases (two elective AAA repair and one cardiac surgery) were utilised. Each database was cleaned prior to analysis. Logistic regression was used to develop both regional and national risk prediction models for mortality following elective AAA repair. A regional model to identify the risk of developing renal failure following elective AAA repair was also developed. The performance of a widely used cardiac surgery risk prediction model (the logistic EuroSCORE) over time was evaluated using a national cardiac database. In addition an updated model version (EuroSCORE II) was validated and both models’ performance in emergency cardiac surgery was evaluated. Results: Regional risk models for mortality following elective AAA repair (VGNW model) and a model to predict post-operative renal failure were developed. Validation of the model for mortality using a national dataset demonstrated good performance compared to other available risk models. To improve generalisability a national model (the BAR score) with better discriminatory ability was developed. In a prospective validation of both models using regional data, the BAR score demonstrated excellent discrimination overall and good discrimination in procedural sub-groups. The EuroSCORE was found to have lost calibration over time due to a fall in observed mortality despite an increase in the predicted mortality of patients undergoing cardiac surgery. The EuroSCORE II demonstrated good performance for contemporary cardiac surgery. Both EuroSCORE models demonstrated inadequate performance for emergency cardiac surgery. Conclusions: Risk prediction models play an important role in cardiovascular surgery. Two accurate risk prediction models for mortality following elective AAA repair have been developed and can be used to risk-adjust surgical outcomes and facilitate clinical decision making. As surgical practice changes over time risk prediction models may lose accuracy which has implications for their application. Cardiac risk models may not be sufficiently accurate for high-risk patient groups such as those undergoing emergency surgery and specific emergency models may be required. Continuing research into new risk factors and model outcomes is needed and risk prediction models may play an increasing role in clinical decision making in the future.

Gestational diabetes mellitus (GDM) is the most common metabolic complication in pregnancy and is associated with substantial maternal and neonatal morbidity. The standard of care for GDM in most developed countries is universal mid- to late- pregnancy (24-28 weeks gestation) glucose testing. While earlier diagnosis and treatment could improve pregnancy outcomes, tools for early identification of risk for GDM are not commonly used in practice. Existing models for predicting GDM risk within the first trimester of pregnancy based on maternal risk factors perform only modestly in the clinical setting. Heavy reliance on history of GDM to predict GDM development in the current pregnancy prevents these tools from being applicable to nulliparous women (i.e., women who have never given birth). In order to offer timely preventive intervention and enhanced antenatal care to nulliparous women, we need to be able to accurately identify those at high risk for GDM early in pregnancy. Data from the California Office of Statewide Health Planning and Development Linked Birth File was used to address three aims: 1) improve early pregnancy prediction of GDM risk in nulliparous women through development of a risk factor-based model, 2) conduct a systematic review and meta-analysis assessing the relationship between first trimester prenatal screening biomarker levels and development of GDM, and 3) determine if the addition of first and second trimester prenatal screening biomarkers to risk factor-based models will improve early prediction of GDM in nulliparous women. We developed a clinical prediction model including five well-established risk factors for GDM (race/ethnicity, age at delivery, pre-pregnancy body mass index, family history of diabetes, and pre-existing hypertension). Our model had moderate predictive performance among all nulliparous women, and performed particularly well among Hispanic and Black women when assessed within specific racial/ethnic groups. Our risk prediction model also showed superior performance over the commonly used American College of Obstetricians and Gynecologists (ACOG) screening guidelines, encouraging the prompt incorporation of this tool into preconception and prenatal care. Biomarkers commonly assessed in prenatal screening have been associated with a number of adverse perinatal and birth outcomes. However, reports on the relationship between first trimester measurements of prenatal screening biomarkers and GDM development are inconsistent. Our meta-analysis demonstrated that women who are diagnosed with GDM have lower first trimester multiple of the median (MoM) levels of both pregnancy associated plasma protein-A (PAPP-A) and free β-human chorionic gonadotropin (free β-hCG) than women who remain normoglycemic throughout pregnancy. Findings from our meta-analysis suggested that incorporation of prenatal screening biomarkers in clinical risk prediction models could aid in earlier identification of women at risk of developing GDM. Upon linkage of California Office of Statewide Health Planning and Development Linked Birth File and California Prenatal Screening Program records, we found that decreased levels of first trimester PAPP-A, increased second trimester unconjugated estriol, and increased second trimester dimeric inhibin A were associated with GDM development in nulliparous women. However, the addition of these biomarkers in clinical models did not offer improvements to the clinical utility (i.e., risk stratification) of models including maternal risk factors alone. Our findings demonstrate that incorporation of maternal risk factors in a clinical risk prediction model can more accurately identify nulliparous women at high risk for GDM early in pregnancy compared to current standard practice. The maternal characteristics model we developed is based on clinical history and demographic variables that are already routinely collected by clinicians in the United States so that it may be easily adapted into existing prenatal care practice and screening programs. Future work should focus on evaluating the clinical impact of model implementation on maternal and infant outcomes as well as financial costs to the health care system.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017.
This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.
Cataloged from student-submitted PDF version of thesis.
Includes bibliographical references (pages 99-108).
There are major practical and technical barriers to understanding human health, and therefore a need for methods that thrive on large, complex, noisy data. In this work, we present machine learning methods that distill large amounts of heterogeneous health data into latent state representations. These representations are then used to estimate risks of poor outcomes, and response to intervention in multivariate physiological signals. We evaluate the reduced latent representations by 1) establishing their predictive value in important clinical tasks and 2) showing that the latent space representations themselves provide useful insight into underlying systems. In particular, we focus on case studies that can provide evidence-based risk assessment and forecasting in settings with guidelines that have not traditionally been data-driven. In this thesis we evaluate several methods to create patient representations, and use these features to predict important outcomes. Representation learning can be thought of as a form of phenotype discovery, where we attempt to discover spaces in the new representation that are markers of important events. We argue that these latent representations are useful markers when they 1) create better prediction results on outcomes of interest, and 2) do not duplicate features that are currently known bio-markers. We present four case studies of learning representations, and evaluate the representations on real predictive tasks. First, we create forward-facing prediction models using baseline clinical features, and those from a Latent Dirichlet Allocation (LDA) model trained with clinical progress notes. We then evaluate the per-patient latent state membership to predict mortality in an intensive care setting as time moves forward. Second, we use non-parametric Multi-task Gaussian Process (MTGP) hyper-parameters as latent features to estimate correlations within and between signals in sparse, heterogeneous time series data. We evaluate the hyper-parameters for forecasting missing signals in traumatic brain injury patients, and predicting mortality in intensive care unit patients. Third, we train switching-state autoregressive models (SSAMs) to model the underlying states that emit patient vital signs over time. We evaluate the time-specific latent state distributions as features to predict vasopressor onset and weaning in intensive care unit patients. Finally, we use statistical and symbolic features extracted from wearable ambulatory accelerometers (ACC) mounted to the neck to classify patient pathology, and stratify patients' risk of voice misuse. We evaluate the utility of both statistically generated features and symbolic representations of glottal pulses towards patient classification.
by Marzyeh Ghassemi.
Ph. D.

The hypertensive disorders of pregnancy (HDPs) are one of the leading causes of maternal death and morbidity in low-resourced countries due to delays in case identification and a shortage of health workers trained to manage these disorders. The objective of this thesis was to develop an evidence-based tool that could aid community-based health workers in decision making around the care of women with the HDPs. This objective was achieved using a prospective cohort of data collected in five low and middle income countries (LMICs) to: (1) develop a clinical risk predication model using logistic regression (the “miniPIERS” model); (2) validate the miniPIERS model through bootstrapping and by applying the model to a second cohort of women with HDP; (3) extend and recalibrate the model to include the novel biomarker, pulse oximetry (SpO₂); and (4) translate the miniPIERS model into a decision rule for final creation of the PIERS on the Move decision algorithm. All stages of development of the PIERS on the Move tool included input from stakeholders in low-resourced countries. The miniPIERS model, based on demographics, symptoms and clinical signs, accurately identified women who were at greatest risk of complications from the HDP (AUC ROC 0.77 [95% CI 0.74 – 0.80]). Internal validation demonstrated minimal overfitting with an average optimism of 0.037. Addition of SpO2 to the miniPIERS model resulted in a 20% increase in classification accuracy of high-risk women. Using an iterative review and feedback process including stakeholders from our partner low-resourced countries, decision points defined by the miniPIERS model were combined with the WHO recommendations for treatment of women with HDP to create a novel decision algorithm for population level risk screening. This decision algorithm identified high-risk women in the miniPIERS cohort with a sensitivity of 74.1% and specificity of 51.4%. Pilot testing of this tool in South Africa demonstrated potential impact but the true impact of use of the PIERS on the Move tool on maternal outcome rates requires assessment through an implementation study.
Medicine, Faculty of
Obstetrics and Gynaecology, Department of
Graduate

Family-based and genome-wide association studies (GWAS) of alcohol dependence (AD) have reported numerous associated variants. The clinical validity of these variants for predicting AD compared to family history has not yet been reported. These studies aim to explore the aggregate impact of multiple genetic variants with small effect sizes on risk prediction in order to provide a clinical interpretation of genetic contributions to AD. Data simulations showed that given AD’s prevalence and heritability, a risk prediction model incorporating all genetic contributions would have an area under the receiver operating characteristic curve (AUC) approaching 0.80, which is often a target AUC for screening. Adding additional environmental factors could increase the AUC to 0.95. Using the Collaborative Study on the Genetics of Alcoholism (COGA) and the Study of Addiction: Genes and Environment (SAGE) GWAS samples, we used several different sources to capture genetic information associated with AD in discovery samples, and then tested genetic sum scores created based on this information for predictive accuracy in validation samples. Scores were assessed separately for single nucleotide polymorphisms (SNPs) associated in candidate gene studies and in GWAS analyses. Candidate gene sum scores did not exhibit significant predictive accuracy, but SNPs meeting less stringent p-value thresholds in GWAS analyses did, ranging from mean estimates of 0.549 for SNPs meeting p<0.01 to 0.565 for SNPs meeting p<0.50. Variants associated with subtypes of AD showed that there is similarly modest and significant predictive ability for an externalizing subtype. Scores created based on all individual SNP effects in aggregate across the entire genome accounted for 0.46%-0.57% of the variance in AD symptom count, and have AUCs of 0.527 to 0.549. Additional covariates and environmental factors that are correlated with AD increased the AUC to 0.865. Family history was a better classifier of case-control status than genetic sum scores, with an AUC of 0.686 in COGA and 0.614 in SAGE. This project suggests that SNPs from candidate gene studies and genome-wide association studies currently have limited clinical validity, but there is potential for enhanced predictive ability with better detection of genetic factors contributing to AD.

Health and sociological indicators confirm that life expectancy is increasing, and so, the years that patients have to live with chronic diseases and co-morbidities. Type 2 Diabetes is one of the most common chronic diseases, specially linked to overweight and ages over sixty. As a metabolic disease, Type 2 Diabetes affects multiple organs by causing damage in blood vessels and nervous system at micro and macro scale. Mortality of subjects with diabetes is three times higher than the mortality for subjects with other chronic diseases. On the one hand, the management of diabetes is focused on the maintenance of the blood glucose levels under a threshold by the prescription of anti-diabetic drugs and a combination of healthy food habits and moderate physical activity. Recent studies have demonstrated the effectiveness of new strategies to delay and even prevent the onset of Type 2 Diabetes by a combination of active and healthy lifestyle on cohorts of mid to high risk subjects. On the other hand, prospective research has been driven on large groups of population to build risk scores which aim to obtain a rule for the classification of patients according to the odds for developing the disease. Currently there are more than two hundred models and risk scores for doing this, but a few have been properly evaluated in external groups and, to date, none of them has been tested on a population based study. The research study presented in this doctoral thesis strives to use externally validated risk scores for the prediction and detection of Type 2 Diabetes on a population data base in Hospital La Fe (Valencia, Spain). The study hypothesis is that the integration of existing prediction and detection risk scores on Electronic Health Records increases the early-detection of high risk cases. To evaluate this hypothesis three studies on the clinical, user and technology dimensions have been driven to evaluate the extent to which the models and the hospital is ready to exploit such models to identify high risk groups and drive efficient preventive strategies. The findings presented in this thesis suggest that Electronic Health Records are not prepared to massively feed risk models. Some of the evaluated models have shown a good classification performance, which accompanied to the well-acceptance of web-based tools and the acceptable technical performance of the information and communication technology system, suggests that after some work these models can effectively drive a new paradigm of active screening for Type 2 Diabetes.
Los indicadores de salud y sociológicos confirman que la esperanza de vida está aumentando, y por lo tanto, los años que los pacientes tienen que vivir con enfermedades crónicas y comorbilidades. Diabetes tipo 2 es una de las enfermedades crónicas más comunes, especialmente relacionadas con el sobrepeso y edades superiores a los sesenta años. Como enfermedad metabólica, la diabetes tipo 2 afecta a múltiples órganos causando daño en los vasos sanguíneos y el sistema nervioso a escala micro y macro. La mortalidad de sujetos con diabetes es tres veces mayor que la mortalidad de sujetos con otras enfermedades crónicas. Por un lado, la estrategia de manejo se centra en el mantenimiento de los niveles de glucosa en sangre bajo un umbral mediante la prescripción de fármacos antidiabéticos y una combinación de hábitos alimentarios saludables y actividad física moderada. Estudios recientes han demostrado la eficacia de nuevas estrategias para retrasar e incluso prevenir la aparición de la diabetes tipo 2 mediante una combinación de estilo de vida activo y saludable en cohortes de sujetos de riesgo medio a alto. Por otro lado, la investigación prospectiva se ha dirigido a grupos de la población para construir modelos de riesgo que pretenden obtener una regla para la clasificación de las personas según las probabilidades de desarrollar la enfermedad. Actualmente hay más de doscientos modelos de riesgo para hacer esta identificación, no obstante la inmensa mayoría no han sido debidamente evaluados en grupos externos y, hasta la fecha, ninguno de ellos ha sido probado en un estudio poblacional. El estudio de investigación presentado en esta tesis doctoral pretende utilizar modelos riesgo validados externamente para la predicción y detección de la Diabetes Tipo 2 en una base de datos poblacional del Hospital La Fe de Valencia (España). La hipótesis del estudio es que la integración de los modelos de riesgo de predicción y detección existentes la práctica clínica aumenta la detección temprana de casos de alto riesgo. Para evaluar esta hipótesis, se han realizado tres estudios sobre las dimensiones clínicas, del usuario y de la tecnología para evaluar hasta qué punto los modelos y el hospital están dispuestos a explotar dichos modelos para identificar grupos de alto riesgo y conducir estrategias preventivas eficaces. Los hallazgos presentados en esta tesis sugieren que los registros de salud electrónicos no están preparados para alimentar masivamente modelos de riesgo. Algunos de los modelos evaluados han demostrado un buen desempeño de clasificación, lo que acompañó a la buena aceptación de herramientas basadas en la web y el desempeño técnico aceptable del sistema de tecnología de información y comunicación, sugiere que después de algún trabajo estos modelos pueden conducir un nuevo paradigma de la detección activa de la Diabetes Tipo 2.
Els indicadors sociològics i de salut confirmen un augment en l'esperança de vida, i per tant, dels anys que les persones han de viure amb malalties cròniques i comorbiditats. la diabetis de tipus 2 és una de les malalties cròniques més comunes, especialment relacionades amb l'excés de pes i edats superiors als seixanta anys. Com a malaltia metabòlica, la diabetis de tipus 2 afecta múltiples òrgans causant dany als vasos sanguinis i el sistema nerviós a escala micro i macro. La mortalitat de subjectes amb diabetis és tres vegades superior a la mortalitat de subjectes amb altres malalties cròniques. D'una banda, l'estratègia de maneig se centra en el manteniment dels nivells de glucosa en sang sota un llindar mitjançant la prescripció de fàrmacs antidiabètics i una combinació d'hàbits alimentaris saludables i activitat física moderada. Estudis recents han demostrat l'eficàcia de noves estratègies per a retardar i fins i tot prevenir l'aparició de la diabetis de tipus 2 mitjançant una combinació d'estil de vida actiu i saludable en cohorts de subjectes de risc mitjà a alt. D'altra banda, la investigació prospectiva s'ha dirigit a grups específics de la població per construir models de risc que pretenen obtenir una regla per a la classificació de les persones segons les probabilitats de desenvolupar la malaltia. Actualment hi ha més de dos-cents models de risc per fer aquesta identificació, però la immensa majoria no han estat degudament avaluats en grups externs i, fins ara, cap d'ells ha estat provat en un estudi poblacional. L'estudi d'investigació presentat en aquesta tesi doctoral utilitza models de risc validats externament per a la predicció i detecció de diabetis de tipus 2 en una base de dades poblacional de l'Hospital La Fe de València (Espanya). La hipòtesi de l'estudi és que la integració dels models de risc de predicció i detecció existents la pràctica clínica augmenta la detecció de casos d'alt risc. Per avaluar aquesta hipòtesi, s'han realitzat tres estudis sobre les dimensions clíniques, de l'usuari i de la tecnologia per avaluar fins a quin punt els models i l'hospital estan disposats a explotar aquests models per identificar grups d'alt risc i conduir estratègies preventives. Les troballes presentades sugereixen que els registres de salut electrònics no estan preparats per alimentar massivament models de risc. Alguns dels models avaluats han demostrat una bona classificació, el que va acompanyar a la bona acceptació d'eines basades en el web i el rendiment tècnic acceptable del sistema de tecnologia d'informació i comunicacions implementat. La conclusió es que encara es necesari treball per que aquests models poden conduir un nou paradigma de la detecció activa de la diabetis de tipus 2.
Martínez Millana, A. (2017). ASSESSMENT OF RISK SCORES FOR THE PREDICTION AND DETECTION OF TYPE 2 DIABETES MELLITUS IN CLINICAL SETTINGS [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/86209
TESIS

In clinical settings, the diagnosis of medical conditions is often aided by measurement of various serum biomarkers through the use of laboratory tests. These biomarkers provide information about different aspects of a patient’s health and the overall function of different organs. In this dissertation, we develop and validate a weighted composite index that aggregates the information from a variety of health biomarkers covering multiple organ systems. The index can be used for predicting all-cause mortality and could also be used as a holistic measure of overall physiological health status. We refer to it as the Health Status Metric (HSM). Validation analysis shows that the HSM is predictive of long-term mortality risk and exhibits a robust association with concurrent chronic conditions, recent hospital utilization, and self-rated health. We develop the HSM using Weighted Quantile Sum (WQS) regression (Gennings et al., 2013; Carrico, 2013), a novel penalized regression technique that imposes nonnegativity and unit-sum constraints on the coefficients used to weight index components. In this dissertation, we develop a number of extensions to the WQS regression technique and apply them to the construction of the HSM. We introduce a new guided approach for the standardization of index components which accounts for potential nonlinear relationships with the outcome of interest. An extended version of the WQS that accommodates interaction effects among index components is also developed and implemented. In addition, we demonstrate that ensemble learning methods borrowed from the field of machine learning can be used to improve the predictive power of the WQS index. Specifically, we show that the use of techniques such as weighted bagging, the random subspace method and stacked generalization in conjunction with the WQS model can produce an index with substantially enhanced predictive accuracy. Finally, practical applications of the HSM are explored. A comparative study is performed to evaluate the feasibility and effectiveness of a number of ‘real-time’ imputation strategies in potential software applications for computing the HSM. In addition, the efficacy of the HSM as a predictor of hospital readmission is assessed in a cohort of emergency department patients.

Mammographic density, the amount of radiodense tissue on a mammogram, is a strong risk factor for breast cancer, with properties that could be an asset in screening and prevention programmes. Its use in risk prediction contexts is currently limited, however, mainly due to di culties in measuring and interpreting density. This research investigates rstly, the properties of density as an independent marker of breast cancer risk and secondly, how density should be measured. The rst question was addressed by analysing data from a chemoprevention trial, a trial of hormonal treatment, and a cohort study of women with a family history of breast cancer . Tamoxifen-induced density reduction was observed to be a good predictor of breast cancer risk reduction in high-risk una ected subjects. Density and its changes did not predict risk or treatment outcome in subjects with a primary invasive breast tumour. Finally absolute density predicted risk better than percent density and showed a potential to improve existing risk-prediction models, even in a population at enhanced familial risk of breast cancer. The second part of thesis focuses on density measurement and in particular evaluates two fully-automated volumetric methods, Quantra and Volpara. These two methods are highly correlated and in both cases absolute density (cm3) discriminated cases from controls better than percent density. Finally, we evaluated and compared di erent measurement methods. Our ndings suggested good reliability of the Cumulus and visual assessments. Quantra volumetric estimates appeared negligibly a ected by measurement error, but were less variable than visual bi-dimensional ones, a ecting their ability to discriminate cases from controls. Overall, visual assessments showed the strongest association with breast cancer risk in comparison to computerised methods. Our research supports the hypothesis that density should have a role in personalising screening programs and risk management. Volumetric density measuring methods, though promising, could be improved.

Background - Current guidelines suggest that all cancer patients with venous thrombosis be treated with long-term low molecular weight heparin. Whether treatment strategies should vary according to clinical characteristics remains unknown. // Systematic review - A systematic review was performed to determine current understanding of the association between malignancy characteristics in patients with cancer-associated VTE and the risk of VTE recurrence. Four retrospective and 6 prospective studies were included. They suggest that lung cancer, metastases, and adenocarcinomas confer an increased the risk of recurrence and breast cancer a low risk. // Survey - I performed survey to evaluate thrombosis experts’ opinion about the low risk of VTE recurrence they would consider acceptable for patients with cancer- associated thrombosis 103 specialists participated. 80% of respondents agreed that a risk of recurrent VTE during anticoagulation below 7% is low enough. 92% agreed that a CPR that categorizes risk of recurrence is relevant. // Retrospective Study - I performed a single retrospective cohort study to assess the feasibility of derivation of a CPR that stratifies VTE recurrence risk in patients with cancer–associated thrombosis. The study included 543 patients. A multivariate analysis selected female, lung cancer and prior history of VTE as high risk predictors and breast cancer and stage I disease as low risk. // Conclusion - Patients with cancer-associated thrombosis do have varying risks of recurrent VTE depending on clinical characteristics.

Background: Cesarean delivery rates have been increasing since 1996, and Cesarean delivery is now the most common major operative procedure performed in the United States. Identifying risk factors for wound complications following Cesarean delivery is necessary to prevent unnecessary maternal morbidity. Methods: A case-control study was carried out and data was collected via a medical record review for patients undergoing a Cesarean delivery at the UIHC between 10/1/2011 and 12/31/2012. Results: Several modifiable risk factors were identified, and models based on patient and surgical factors performed better than the current standard NHSN risk index model. Conclusion: More robust prediction models can be created using patient and surgical factors.

This research critically analyses the different types of clinical data representation used in modelling Clinical Information Systems (CIS) and their limitations. It identifies space complexity, information overload, performance degradation, erroneous data retrieval and transmission as some of the main challenges caused by inappropriate data representation. Literature reviewed, indicated that object-oriented Health Level 7 (HL7), Entity Attribute Value (EAV), Advanced ERD with XML, and ERD –FOL (First Order Logic) are some of the contemporary methods used in modelling and optimising CIS. However, these approaches do not address the space complexity and information overload issues because of the multi-dimensional, complex large-scale nature of clinical datasets. Therefore, this research proposes a unique framework that uses object-oriented (UML) technique and combinatorial multiple attribute utility theory (CMAUT) as a new clinical data re-representation. In the CMAUT framework, the human organs, their multiple attributes and relationships are modelled using classes. The attributes of each organ class are written as logical expressions using CMAUT concepts, which are linked to each other with logical connectors AND for complementary organs such as cardiovascular and OR for substitutable organs like kidneys. The logical expressions are converted into mathematical format, which serves as the utility objective function that is optimised using linear programming method subject to a set of constraint matrix. The constraint matrix is generated by transforming the multiple attributes in the CMAUT expressions into algebraic expressions by applying an algorithm that uses unit matrix and Raman transformation table. The output of the framework gives a set of attribute values, which optimal value maximises the overall utility of the objective function in the combinatorial organs. The algorithm maps the resultant attribute values to the appropriate attributes of the organs to determine the optimal amount of data required to be retrieved for primary health care investigation. The framework retrieves and transmits only needed data for investigation thus reducing the information overload and space complexity in the CIS. The framework was implemented using the MATLAB software and validated with clinical data from the cardiovascular disease survey in England report. Functionality test conducted, revealed that for complementary organs the space complexity is θ (n + 1) using the framework and θ (2n) without the framework. Substitutable organs gave an exponential expansion of θ (2n) in both cases. Simulation conducted showed that the mean size of the data retrieved for investigation using the framework is 463.5 bytes as compared to 1216.6 bytes without it. Statistical tests carried out using the output data from the framework gave a p-value of 0.000. Hence the hypothesis that the amount of data required for primary care health investigation can be reduced when the clinical data is re-represented with UML/CMAUT and optimised using LP based algorithm is statistically significant. For hypertension disease, by converting the optimal values from the framework into percentages give results similar to the percentage risk of the user been hypertensive. The output values were benchmarked against Framingham web based heart risk calculators and statistically analysed. Hence, the novelty of the framework is that it can be used for optimising CIS, as a multi-attribute decision tool and as an epidemiological prediction model for detecting high blood pressure diseases.

Background: Venous thromboembolism (VTE), comprising the conditions of deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common acute cardiovascular event associated with increased long-term morbidity, functional disability, all-cause mortality, and high rates of recurrence. Major advances in identification, prophylaxis, and treatment over the past 3-decades have likely changed its clinical epidemiology. However, there are little published data describing contemporary, population-based, trends in VTE prevention and management. Objectives: To examine recent trends in the epidemiology of clinically recognized VTE and assess the risk of recurrence after a first acute episode of VTE. Methods: We used population-based surveillance to monitor trends in acute VTE among residents of the Worcester, Massachusetts, metropolitan statistical area (WMSA) from 1985 through 2009, including in-hospital and ambulatory settings. Results: Among 5,025 WMSA residents diagnosed with acute PE and/or lower-extremity DVT between 1985 and 2009 (mean age = 65 years), 46% were men and 95% were white. Age- and sex-adjusted annual event rates (per 100, 000) of clinically recognized acute first-time and recurrent VTE was 142 overall, increasing from 112 in 1985/86 to 168 in 2009, due primarily to increases in PE occurrence. During this period, non-invasive diagnostic VTE testing increased, vi while treatment shifted from the in-hospital (chiefly with warfarin and unfractionated heparin) to out-patient setting (chiefly with low-molecular-weight heparins and newer anticoagulants). Among those with community-presenting first-time VTE, subsequent 3-year cumulative event rates of key outcomes decreased from 1999 to 2009, including all-cause mortality (41% to 26%), major bleeding episodes (12% to 6%), and recurrent VTE (17% to 9%). Active-cancer (with or without chemotherapy), a hypercoagulable state, varicose vein stripping, and Inferior vena cava filter placement were independent predictors of recurrence during short- (3-month) and long-term (3-year) follow-up after a first acute episode of VTE. We developed risk score calculators for VTE recurrence based on a 3-month prognostic model for all patients and separately for patients without active cancer. Conclusions: Despite advances in identification, prophylaxis, and treatment between 1985 and 2009, the disease burden from VTE in residents of central Massachusetts remains high, with increasing annual events. Declines in the frequency of major adverse outcomes between 1999 and 2009 were reassuring. Still, mortality, major bleeding, and recurrence rates remained high, suggesting opportunities for improved prevention and treatment. Clinicians may be able to use the identified predictors of recurrence and risk score calculators to estimate the risk of VTE recurrence and tailor outpatient treatments to individual patients.

Background: Venous thromboembolism (VTE), comprising the conditions of deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common acute cardiovascular event associated with increased long-term morbidity, functional disability, all-cause mortality, and high rates of recurrence. Major advances in identification, prophylaxis, and treatment over the past 3-decades have likely changed its clinical epidemiology. However, there are little published data describing contemporary, population-based, trends in VTE prevention and management. Objectives: To examine recent trends in the epidemiology of clinically recognized VTE and assess the risk of recurrence after a first acute episode of VTE. Methods: We used population-based surveillance to monitor trends in acute VTE among residents of the Worcester, Massachusetts, metropolitan statistical area (WMSA) from 1985 through 2009, including in-hospital and ambulatory settings. Results: Among 5,025 WMSA residents diagnosed with acute PE and/or lower-extremity DVT between 1985 and 2009 (mean age = 65 years), 46% were men and 95% were white. Age- and sex-adjusted annual event rates (per 100, 000) of clinically recognized acute first-time and recurrent VTE was 142 overall, increasing from 112 in 1985/86 to 168 in 2009, due primarily to increases in PE occurrence. During this period, non-invasive diagnostic VTE testing increased, vi while treatment shifted from the in-hospital (chiefly with warfarin and unfractionated heparin) to out-patient setting (chiefly with low-molecular-weight heparins and newer anticoagulants). Among those with community-presenting first-time VTE, subsequent 3-year cumulative event rates of key outcomes decreased from 1999 to 2009, including all-cause mortality (41% to 26%), major bleeding episodes (12% to 6%), and recurrent VTE (17% to 9%). Active-cancer (with or without chemotherapy), a hypercoagulable state, varicose vein stripping, and Inferior vena cava filter placement were independent predictors of recurrence during short- (3-month) and long-term (3-year) follow-up after a first acute episode of VTE. We developed risk score calculators for VTE recurrence based on a 3-month prognostic model for all patients and separately for patients without active cancer. Conclusions: Despite advances in identification, prophylaxis, and treatment between 1985 and 2009, the disease burden from VTE in residents of central Massachusetts remains high, with increasing annual events. Declines in the frequency of major adverse outcomes between 1999 and 2009 were reassuring. Still, mortality, major bleeding, and recurrence rates remained high, suggesting opportunities for improved prevention and treatment. Clinicians may be able to use the identified predictors of recurrence and risk score calculators to estimate the risk of VTE recurrence and tailor outpatient treatments to individual patients.

Older people admitted to hospital are at high risk of rehospitalization and medication errors. We have demonstrated, in a randomized controlled trial, that a clinical pharmacist intervention reduces the incidence of revisits to hospital for patients aged 80 years or older admitted to an acute internal medicine ward. The aims of this thesis were to further study the effects of the intervention and to investigate possibilities of targeting the intervention by identifying predictors of treatment response or adverse health outcomes. The effect of the pharmacist intervention on the appropriateness of prescribing was assessed, by using three validated tools. This study showed that the quality of prescribing was improved for the patients in the intervention group but not for those in the control group. However, no association between the appropriateness of prescribing at discharge and revisits to hospital was observed. Subgroup analyses explored whether the clinical pharmacist intervention was equally effective in preventing emergency department visits in patients with few or many prescribed drugs and in those with different levels of inappropriate prescribing on admission. The intervention appeared to be most effective in patients taking fewer drugs, but the treatment effect was not altered by appropriateness of prescribing. The most relevant risk factors for rehospitalization and mortality were identified for the same study population, and a score for risk-estimation was constructed and internally validated (the 80+ score). Seven variables were selected. Impaired renal function, pulmonary disease, malignant disease, living in a nursing home, being prescribed an opioid and being prescribed a drug for peptic ulcer or gastroesophageal reflux disease were associated with an increased risk, while being prescribed an antidepressant drug (tricyclic antidepressants not included) was linked with a lower risk. These variables made up the components of the 80+ score. Pending external validation, this score has potential to aid identification of high-risk patients. The last study investigated the occurrence of prescription errors when patients with multi-dose dispensed (MDD) drugs were discharged from hospital. Twenty-five percent of the MDD orders contained at least one medication prescription error. Almost half of the errors were of moderate or major severity, with potential to cause increased health-care utilization.

A parent screening instrument for learning difficulties (PSILD) was investigated clinically over three years with children referred for learning and behaviour problems. This established face validity for an instrument with eight sub-areas (i) SocioDemographic, (ii) Genetic, (iii) Pregnancy, (iv) Birth, (v) Early Illness, (vi) SpeechLanguage, (vii) Movement, (viii) Behaviour (Withdrawal - Social Competence, Anxiety-Depression, Conduct-Oppositional Disorder, Hyperactivity, Attention). To these were added the sub-areas of Early Education and Social Strength (resilience). PSILD was then trialled at school entry for usability, acceptability, cost effectiveness, and ease of administration by school personnel, in one private primary school, two state schools from low socio-economic areas, and a child health clinic. The clinic results showed a higher incidence of problems, providing further face validity. Usability, acceptability, cost effectiveness and ease of administration were seen to be satisfactory. Parent feedback led to deletions and modifications of the items and an A3 form was found to be quicker and easier to complete than an A4 form. A brief introductory letter to parents was also seen to help completion of the instrument. A computer-scored Feedback Sheet was constructed providing teachers with an overall risk level (1 - 9 scale; 1 =high risk; 9 =no risk) and sub-area risk levels (1 - 3 scale; risk, borderline, no risk). This laid the basis for the early identification of academic problems, besides remedial procedures and professional referrals. PSILD, based on 305 items, was then completed by 422 parents and factor analysed by sub-area. The factors obtained were found to logically define the previously trialled PSILD sub-areas, providing construct validity. Items with a loading ofless than .3, were eliminated, reducing the total from 305 to 202 which enhanced content validity. Items were also eliminated which did not contribute to the Cronbach Alpha coefficient per sub-area. The resultant 170 items gave an average time of completion of 12.7 minutes. The overall Alpha coefficient (internal consistency) was .83 (.8295) (N = 215). Test-retest stability after a median period of 127 days was also .83 (.8329) (N = 81 ). Both these measures showed PSILD to have a high level of reliability. Employing 215 children from six schools, a teacher instrument, the Pupil Rating Scale Revised (PRSR) and the Letter Identification Test (CLAY), given after nine months of schooling, were used as predictive instruments with PSILD and chronological age, against the Wechsler Individual Achievement Test (Reading Sub-Test) (WIAT), given 21 months after school entry. PSILD was found to have a modest but significant correlation with WIAT (r = .34). When PSILD was reduced to five components (SUBPSILD) (Genetic, Pregnancy, Withdrawal, Early Education and parents' educational level), the level rose tor= .44. The PRSR and CLAY levels were .45 and .51 respectively. When SUBPSILD was combined with Auditory Comprehension and Memory (PRSRAUD), CLAY and age, multiple R reached .76 (variance 57 percent). Entering these components into a discriminant function analysis produced a hit rate of 91.2 percent and an odds ratio of 12.6 (12 < .0001). This result was superior to those of seven recent studies which also used multiple measures (Scarborough, 1998). SUBPSILD's level of prediction was shown to be equivalent to that of the teacher measure (PRSRAUD). A .25 correlation between the PSILD and PRSR behaviour subscales was similar to that of Achenbach et al. (1987) of .27 for 41 parent-teacher samples. This provided a potential basis for pervasive behavioural evaluation at nine months of schooling. Thus PSILD was shown to combine the roles of educational prediction with that of the early identification of academic and behavioural problems.

Background. Whether to continue or to discontinue oral anticoagulation therapy (OAT) after 6 months of treatment to prevent recurrent or fatal events in unprovoked VTE patients is currently controversial. We sought to develop and internally validate a clinical prediction rule (CPR) to identify patients at low risk of recurrent VTE (at most 3% annual risk) for whom OAT could be safely discontinued. Methods. Univariate and multivariate analysis techniques were used to identify the best set of predictor variables. Results and conclusions. We derived and internally validated a CPR for females comprised of D-Dimer over 250 ug/L, post-thrombotic signs present, older age (over 65 years) and obesity (BMI over 30 kg/m2). Women with one or none of the four aforementioned clinical predictors had an annual risk of recurrent VTE of 1.6% and may be able to discontinue OAT. None of the models for males was shown to be safe.

Los trastornos de control de impulsos (TCI) son una complicación frecuente del tratamiento de la enfermedad de Parkinson (EP), particularmente del uso de agonistas de dopamina (AD). Los TCI en la EP se han estudiado durante dos décadas. Sin embargo, la evidencia prospectiva aún es escasa y no hay modelos predictivos disponibles. Esta tesis consta de cuatro trabajos que abordan estas lagunas del conocimiento. En el primer trabajo evaluamos la asociación entre impulsividad y TCI incidente. No confirmamos la asociación sospechada, aunque encontramos una asociación significativa entre la impulsividad y la severidad de los TCI (p = 0.001). La falta de asociación entre la impulsividad y la presencia de TCI se confirmó longitudinalmente en el cuarto trabajo. El segundo trabajo fue un estudio de supervivencia prospectivo longitudinal para evaluar la causalidad de la asociación entre la depresión y el TCI. Los pacientes con EP que estaban deprimidos tenían aproximadamente el doble de riesgo de desarrollar TCI (p <0,001). Esta asociación fue específica e independiente del uso de AD y otros posibles factores de confusión. Además, los pacientes con depresión refractaria tenían un riesgo aún mayor (p = 0,001). El tercero fue un estudio de casos y controles anidado en un estudio prospectivo longitudinal. Estudiamos el metabolismo cerebral a usando PET de 18 FDG. Comparamos pacientes con TCI de nuevo inicio con pacientes apareados sin TCI de la misma cohorte. Los pacientes con TCI mostraron un metabolismo más alto en amplias áreas corticales (p <0.05 corregido para FWE) Los resultados fueron los mismos utilizando el análisis voxelwise y el análisis intracortical. También demostramos que no había diferencias estructurales tanto en el grosor cortical y como en la segmentación subcortical. Usando un grupo de controles sanos apareados, descubrimos que el metabolismo más alto encontrado en los pacientes con TCI es en realidad preservación, dado que no se encontraron diferencias del metabolismo cortical entre los pacientes con EP y TCI y los controles sanos, mientras que los pacientes con EP sin TCI mostraron hipometabolismo en comparación con los controles sanos. El objetivo del cuarto estudio fueron marcadores que pudieran diferenciar a los pacientes con EP con riesgo alto de TCI y aquellos con riesgo bajo. En particular, estudiamos la “feedback related negativity” (FRN), un marcador neurofisiológico del procesamiento de recompensas. Encontramos que este marcador era diferente en los pacientes que desarrollaron TCI en los siguientes tres años en comparación con aquellos que no lo hicieron (p = 0.001). Además, desarrollamos dos modelos para la predicción de TCI: uno utilizando sólo datos clínicos y el otro incluyendo también la FRN basal. El modelo que incluyó la FRN desempeñó significativamente mejor (p = 0.003). Los pacientes identificados gracias a la FRN como de alto riesgo tenían un riesgo diez veces mayor de desarrollar TCI durante los tres años siguientes que los identificados como de bajo riesgo. En conclusión, encontramos evidencias que respalda la depresión como un factor de riesgo de TCI en la EP, evidencias de preservación metabólica cerebral entre los pacientes con EP que tienen TCI, evidencias que respalda el papel del procesamiento de recompensas para el desarrollo de TCI y evidencia que sugiere que la impulsividad debe descartarse como un factor de riesgo para TCI. Por último, demostramos que en el contexto del Parkinson, el desarrollo de TCI puede predecirse, y, por lo tanto, probablemente puede evitarse.
Impulse control disorders (ICD) are a common complication of Parkinson’s disease (PD) treatment, particularly of dopamine agonist (DA) use. ICD in PD have been studied for two decades. Nonetheless, prospective evidence is still scarce and predictive models are lacking. This thesis consists of four works addressing these gaps. In the first work we evaluated the association between impulsivity and incident ICD. We did not confirm the suspected association although we found a significant association between impulsivity and ICD severity (p=0.001). The lack of association between impulsivity and ICD presence was confirmed longitudinally in the fourth work. The second work was a longitudinal prospective survival study to evaluate whether the association between depression and ICD was causal. We found that depressed PD patients had approximately double risk of developing ICD (p<0.001). This association was specific and independent from DA use and other potential confounders. Besides, patients with refractory depression had an even higher risk (p=0.001). The third was a case-control study nested in a longitudinal prospective study. We studied brain metabolism via 18 FDG PET. We compared patients with new onset ICD with matched patients free of ICD from the same cohort. ICD patients showed higher metabolism in widespread cortical areas (p<0.05 FWE corrected). The results were the same using voxelwise analysis and intracortical analysis. We also showed that there were no structural differences using cortical thickness and subcortical segmentation. Using a group of matched healthy controls, we found that the higher metabolism found in ICD patients should be regarded as preservation because no cortical metabolic differences were found between PD‑ICD patients and healthy controls, while PD‑nonICD patients showed hypometabolism when compared with healthy controls. The fourth study targeted markers that could differentiate PD patients at high risk of ICD and those at low risk. Particularly we targeted the feedback related negativity (FRN) a neurophysiological marker of reward processing. We found this marker to be different in patients who developed ICD within the subsequent three years compared with those who did not (p=0.001). Furthermore, we developed two models for ICD prediction: one used only clinical data and the other also included the baseline FRN. The model including the FRN performed significantly better (p=0.003). Patients identified using the FRN as high risk had a risk ten times higher for the next three years than those identified as low risk. In conclusion, we found evidence that backs depression as a risk factor for ICD in PD, evidence of brain metabolic preservation among PD patients with ICD, evidence that supports the role of reward processing for ICD development and evidence that suggests that impulsivity should be discarded as a risk factor for ICD. Lastly, we showed that the development of ICD can be predicted in PD patients and therefore, probably can be avoided.

The high cost, complexity and multimodality of clinical data collection restrain the datasets available for predictive modelling using machine learning (ML), thus necessitating new data-efficient approaches specifically for limited datasets. This interdisciplinary thesis focuses on clinical outcome modelling using a range of ML techniques, including artificial neural networks (NNs) and their ensembles, decision trees (DTs) and random forests (RFs), as well as classical logistic regression (LR) and Cox proportional hazards (Cox PH) models. The utility of ML for data-efficient regression, classification and survival analyses was investigated in three clinical applications, whereby exposing the common limitations inherent in patient data, such as class imbalance, incomplete samples, and, in particular, limited dataset size. The latter problem was addressed by developing a methodological framework for learning from datasets with less than 10 observations per predictor variable. A novel method of multiple runs overcame the volatility of NN and DT models due to limited training samples, while a surrogate data test allowed for regression model evaluation in the presence of noise due to limited dataset size. When applied to hard tissue engineering for predicting femoral fracture risk, the framework resulted in 98.3% accurate regression NN. The framework was used to detect early rejection in antibody- incompatible kidney transplantation, achieving 85% accurate classification DT. The third clinical task – that of predicting 10-year incidence of type 2 diabetes in the UK population – resulted in 70-85% accurate classification and survival models, whilst highlighting the challenges of learning with the limited information characteristic of routinely collected data. By discovering unintuitive patterns, supporting existing hypotheses and generating novel insight, the ML models developed in this research contributed meaningfully to clinical research and paved the way for data-efficient applications of ML in engineering and clinical practice.

Psychological trauma may affect higher-order executive functions, which include selective attention, inhibition, and task-switching processes. Difficulty in these executive processes can in turn influence individuals' daily functioning and may also negatively affect the psychological treatment of post-trauma symptoms. Women may be most at risk for developing problems with executive functioning following trauma, consistent with their overall greater risk of developing post-trauma symptoms. Yet, little is understood about the influence of psychological variables, premorbid functioning, and specific trauma factors in determining post-trauma cognitive functioning in women. Additionally, individual variability in susceptibility to psychological distress and neuropsychological deficits following trauma remains an open area of study. The present study investigated the relationship between psychological and trauma factors with neuropsychological outcomes in women with trauma histories as well as individual variability in risk for poor neuropsychological outcomes. In total, 60 participants' data (age M = 29.73, SD = 10.91) were included in analyses. The final sample consisted of 33 community members recruited from the UNT Psychology Clinic and the UNT student body and 27 veterans recruited from the Veterans Affairs North Texas Healthcare System (VANTHCS). Regression and path analysis identified premorbid intellectual functioning as a predictor of better neuropsychological outcomes and anxiety and depression symptoms as risk factors for worse neuropsychological functioning. Person-centered cluster analyses focused on individual differences in outcomes identified three groups differing in psychological distress and neuropsychological functioning. Additional analyses identified differences in trauma exposure, psychological functioning, and neuropsychological performance between subgroups of civilians and veterans and those with and without a history of PTSD.

Lower gastrointestinal bleeding (LGIB) can result in serious adverse events. Appropriate risk stratification of LGIB patients can improve their care. Previous risk scores to identify severe LGIB patients have limitations, therefore we developed clinical decision tool to accurately identify LGIB patients presenting to the emergency department (ED) who are at risk for 30-day adverse outcomes that would overcome these limitations. We conducted a health records review and compared two methods of regression analysis on our data in order to develop a clinical decision tool. We identified five risk factors that have a high sensitivity and good predictive value for identifying low risk LGIB patients: age ≥ 75 years, INR ≥2.0, hemoglobin ≤ 100 g/l, ongoing bleeding in the ED and a medical history of colorectal polyps. Future, large, prospective studies should be done to validate the results, after which implementation studies should be conducted.

Falls are a significant problem in acute care, hospital settings, and can have serious consequences, especially for older patients. Fall prevention has therefore been recognised as an important area for research and intervention. In order to target interventions and use resources effectively, a major strategy of many fall prevention programmes has been the development and/or use of risk assessment tools to identify patients who are at high risk of falling. Although many tools have been developed, few have been rigorously tested, and there is currently no evidence to support the clinical utility of fall risk assessment tools. There is a need to conduct further research to establish the efficacy of fall risk assessment tools for inpatient populations. Additionally, nurses clinical judgement in assessing fall risk may aid the development of fall risk assessment protocols and further research is needed to build on limited knowledge in this area. A prospective cohort study was used to evaluate two fall risk assessment tools and nurses' clinical judgement in predicting patient falls. Each patient was assessed for fall risk by the clinical judgement of the nurse caring for the patient and by the researcher using a data collection form containing the two fall risk assessment tools. The study wards comprised two aged care and rehabilitation wards, within a 570 bed acute care tertiary teaching hospital facility in Western Australia. Test-retest reliability of the two fall risk assessment tools and nurses' clinical judgement was established over a twenty four hour period. The ability of the fall risk assessment tools, and nurses' clinical judgements to discriminate between patients with a high probability of falling and , patients with a low probability of falling; was determined by calculating the sensitivity, specificity, positive predictive value and negative-predictive value for each method. The reference criterion used for these calculations was whether or not the patient fell within the hospitalisation period in which they were admitted to the study. In addition, the accuracy of each method was determined by calculating the number of times the risk assessment tool or clinical judgement classified the patient into the correct, fall risk category, expressed as a percentage. The same reference criterion was used for this calculation. Both the fall risk assessment tools and nurses' clinical judgement had good test-retest reliability. When assessing validity, all three methods of determining fall risk showed good sensitivity, ranging from 88% to 91 %,but poor specificity, ranging from 25% to 26%. This meant that the risk assessment methods classified too many patients who did not fall as at high risk for falling. All methods also had limited accuracy, ranging from 35% to 36%, and overall exhibited an inability to adequately discriminate between patient populations at risk of falling and those not at risk of falling. Consequently, neither nurses' clinical judgement nor the fall risk assessment tools could be recommended for assessing fall risk in the clinical setting. In addition, results indicated that there was a large difference between the accuracy of first year enrolled and registered nurses in assessing patient fall risk. First year enrolled nurses accurately predicted fall risk 44.4% of the time while first year registered nurses achieved an accuracy level of only 8.6%. These results are potentially biased, as measuring differences in accuracy between types of nurses was not a main focus of this study and in many cases the same nurse gave multiple judgements about patients' fall risk. The results however, provide an indication that further study is warranted using a specifically methodology to explore this issue. There are a number of specific recommendations arising from the results of this study. It is recommended that further studies be undertaken to assess the reliability and validity of current fall risk assessment tools in inpatient populations. If no valid and reliable fall risk assessment tool can be identified, research should be undertaken to develop such a tool. It is also recommended that studies be conducted to assess changes in fall risk profiles over time to determine if the sensitivity and specificity of instruments changes depending on the timing of the risk assessment. Differentiating between stable and transient risk factors should be an integral component of these types of studies. Further research is also required to determine if there are differences in fall risk factors between different specialties or if a generic risk assessment tool can be used for all inpatient populations. Additionally, further investigation into the clinical judgement of registered and enrolled nurses in .their first year of clinical practice should be undertaken and results reported to appropriate educational institutions. Changes in accuracy of clinical judgement in the first five years of clinical practice should also be measured.

Doutoramento em Ciências e Tecnologias da Saúde - Decisão Clínica
A histeroscopia é hoje uma técnica imprescindível em ginecologia quer no diagnóstico de alterações genitais superiores, quer no tratamento minimamente invasivo, sendo segura, fiável e com poucos efeitos secundários e complicações. A miniaturização dos aparelhos juntamente com aperfeiçoamento técnico (nomeadamente com a abordagem vaginoscópica, sem recurso a espéculo, nem a tração do colo) vieram permitir o seu uso em consultório. Usado sem anestesia reduz os riscos do internamento e tornam o exame acessível; tem contudo a limitação da dor provocada pela instrumentação. Apesar de muitas tentativas analgésicas e anestésicas, o controlo da dor não é satisfatória em algumas doentes nas quais a histeroscopia é difícil de suportar. Este trabalho pretende estabelecer se a dor é menor com os mini histeroscópios do que com os instrumentos convencionais, avaliar quão grave é o problema da dor (quantificando a proporção de mulheres que se queixam) mesmo com este aparelhos mais delgados e tentar saber se existem fatores de risco que favoreçam a dor, ou que pelo contrário protejam a doente. Também se pretende estabelecer se os inquéritos de satisfação às doentes se correlacionam com a pontuação de dor e se a ansiedade interfere com as queixas álgicas. Finalmente também tentámos investigar se uma técnica nova de anestesia local, administrada através do histeroscópio com recurso a uma agulha cistoscópica, reduz a dor e torna o exame mais tolerável. Os resultados mostraram haver redução estatisticamente significativa da perceção da dor com mini histeroscópios em relação a aparelhos convencionais. Mostraram ainda que mesmo com calibres finos há uma proporção de doentes entre 13 e 30% que ainda refere dor moderada a severa e que reduzir o calibre abaixo dos 3,5mm pode não resultar numa redução maior da dor. Quanto a fatores de risco para a dor os nossos resultados não encontraram relação, exceto uma proteção na dor para as doentes obesas, que relacionamos com uma maior impregnação hormonal (androgénica e por via da aromatase, estrogénica). A ansiedade não parece ser importante na dor sentida, ainda que exista uma pouco significativa relação entre traço ansioso e intensidade da dor relatada. No que se refere aos questionários de satisfação, correlacionam-se muito bem com a dor reportada, tendo uma boa sensibilidade e especificidade; sendo simples de administrar e fáceis de interpretar poderiam provavelmente substituir as escalas da dor e ser úteis para eventual seleção das doentes a quem administrar a anestesia local histeroscópica. Finamente a técnica histeroscópica de injeção local de anestésico reduz significativamente a dor e poderá ser uma solução para tornar a intervenção suportável em ambulatório.
Hysteroscopy today is an essential tool in gynaecology both for diagnosis of female upper genital tract abnormalities and for minimally invasive surgery procedures. It is safe, reliable and has few side effects and complications. Diminishment of instrument diameter together with technical improvements (such as the vaginal “no touch” approach without use of speculum or cervical traction) has allowed procedures in office environment. Used without analgesia or aneasthesia it has reduced hospitalization risks and made the examination affordable; it has a drawback which is the level of pain some women refer with instrumentation. Although many attempts with the use of analgesics and anaesthetics have been made, pain control is not satisfactory in some patients for whom hysteroscopy is hard to endure. In this work we aim to establish if pain reporting is lower with mini-hysteroscopes compared to conventional scopes, how big is the problem “pain” (quantifying the proportion of women still complaining) even when using the slender mini-scopes and try to establish if there are risk factors which might influence pain reporting, or on the contrary protect, women from agony. We also explored if satisfaction questionnaires correlate well with pain scores and if patient anxiety interferes with pain. Finally we have investigated if a new local anaesthetic administration technique, with the use of a cystoscopic needle through the hysteroscope, can reduce pain and make the procedure more tolerable. Our results show there is a statistically significant reduction of pain scores when using mini-hysteroscope compared to conventional instruments. They also show that even using smaller caliber scopes there is a proportion of women varying from 13 to 30% who will still complain of moderate to severe pain and that reduction of scopes below 3.5mm diameter may not reduce pain scores any further. As to risk factors for pain, our results have not found relation to pain with risk factors except for some protection for pain in women with high body mass index, and we relate this finding with elevated circulating hormones (androgens which are peripherally converted to estrogens via aromatase in adipose tissue). Anxiety does not seem important in pain reporting, even if a slight statistical significance was found between anxiety trait and pain reporting. As to satisfaction questionnaires, they seem to correlate well with the pain experience and show and excellent sensitivity and specificity: simple to administer and easy to interpret, they could probably replace more complicated pain rating scales and be useful in selecting women who would benefit from local anaesthesia. Finally as to the new technique of applying local anaesthetics “hysteroscopic anaesthesia”, results show a statistical reduction of pain after injection and it could become a practical solution in making hysteroscopy bearable in an office setting.

While the overall stability of personality across the lifespan has been well-documented, there is also evidence of meaningful personality change. This is particularly true when individuals are going through periods of developmental transition. Over time, one sees incremental changes not just in behavior but in basic personality as well. 1,906 early adolescents were assessed for urgency scores, levels of maladaptive behavior engagement (drinking, smoking, and binge eating), and pubertal status every six months for four years. Zero-Inflated Poisson structural equation modeling (SEM) was used to test the model of reciprocal influence between behavior and personality. Across most six-month intervals over the course of the four-year study, urgency predicted increased engagement in the maladaptive behaviors. Strikingly, the reverse was true as well: engagement in behaviors predicted subsequent increases in urgency, which is otherwise a stable personality trait. This study is the first to find reciprocal prediction between engagement in maladaptive, risky behaviors and endorsement of the maladaptive personality trait of urgency during the early adolescent years. One implication of these findings is the apparent presence of a positive feedback loop of risk, in which maladaptive behaviors increase high-risk personality traits, which in turn further increase the likelihood of maladaptive behaviors.

Background: Overtreatment of localized prostate cancer (PCa) is a concern as many men die of other causes prior to experiencing a treatment benefit. This dissertation characterizes the need for assessing other cause mortality (OCM) risk in older men with PCa and informs efforts to identify patients most likely to benefit from definitive PCa treatment. Methods: Using the linked Surveillance Epidemiology and End Results-Medicare Health Outcomes Survey database, 2,931 men (mean age=75) newly diagnosed with clinical stage T1a-T3a PCa from 1998-2009 were identified. Survival analysis methods were used to compare observed 10-year OCM by primary treatment type. Age and health factors predictive of primary treatment type were assessed with multinomial logistic regression. Predicted mortality estimates from Social Security life tables (recommended for life expectancy evaluation) and two OCM risk estimation tools were compared to observed rates. An improved OCM prediction model was developed fitting Fine and Gray competing risks models for 10-year OCM with age, sociodemographic, comorbidity, activities of daily living, and patient-reported health data as predictors. The tools’ ability to discriminate between patients who died and those who did not was evaluated with Harrell’s c-index (range 0.5-1), which also guided new model selection. Results: Fifty-four percent of older men with localized PCa underwent radiotherapy while 13% underwent prostatectomy. Twenty-three percent of those treated with radiotherapy and 12% of those undergoing prostatectomy experienced OCM within 10 years of treatment and thus were considered overtreated. Health factors indicative of a shorter life expectancy (increased comorbidity, worse physical health, smoking) had little to no association with radiotherapy assignment but were significantly related to reductions in the likelihood of undergoing prostatectomy. Social Security life tables overestimated mortality risk and discriminated poorly between men who died and those who did not over 10 years (c-index=0.59). Existing OCM risk estimation tools were less likely to overestimate OCM rates and had limited but improved discrimination (c-index=0.64). A risk model developed with self-reported age, Charlson comorbidity index score, overall health (excellent-good/fair/poor), smoking, and marital status predictors had improved discrimination (c-index=0.70). Conclusions: Overtreatment of older men with PCa is primarily attributable to radiotherapy and may be reduced by pretreatment assessment of mortality-related health factors. This dissertation provides a prognostic model which utilizes a set of five self-reported characteristics that better identify patients likely to die of OCM within 10 years of diagnosis than age and comorbidity-based assessments alone.

OBJECTIVE To determine the prevalence of patients with diabetes mellitus (DM2) in Valencia. Moreover, the analysis of population characteristics, disease profile, multimorbidity, pharmaceutical costs, resource consumption and complications We want to establish a prediction model resource consumption of patients with type II diabetes mellitus (DM 2). STUDY DESIGN Methodology Observational, descriptive, retrospective and cross-sectional study. First of all, a database is prepared with patients diagnosed with DM2 in Valencia in 2012. They are rank by their state health system using the Clinical Risk Groups (CRG). From this database, comorbidities and specific complications associated with the disease and pharmaceutical expenditure are analyzed. This is analyzed from different perspectives: comorbidities, CRG health conditions and drugs used to treat the disease. As a result, a multivariate regression models arises to predict pharmaceutical expenditure depending on the different available variables in which the dependent variable is the total pharmaceutical expenditure per patient per year and the explanatory variables will be selected from among those available in the data base developed. Finally, a factorial analysis and principal component analysis is presented to study differences among diabetic patients analyzed in terms of pharmaceutical expenditure, comorbidities not related to DM2, DM2-related comorbidities and complications. Study Population Population diagnosed with type II diabetes of Valencia in the year 2012. Of the 491,854 patients diagnosed with diabetes. We excluded all patients with juvenile diabetes, gestational diabetes and those who received no drug treatment, reducing the group to 350,015 patients. Sources of information and study variables: The basic sources for obtaining data was: - The electronic medical record, ABUCASIS. - The module pharmaceutical services, GAIA. - The Minimum Data Set, CMBD. - The catalog of corporate resources, CRC. - Population Information System, SIP. - Clinical Risk Groups (Clinical Risk Groups, CRGs). Limitations of the study 1. State of coding diagnoses, both in quantitative and qualitative terms. 2. Imbalance in the use of corporate systems outpatient, which enables errors in classification information (SIA-GAIA) RESULTS The estimated gross rate of diabetes was 7.72%, with an average consumption per patient per year was approximately € 1,330. Considering health states, about 48% of patients are classified in health status 6. As for comorbidities of patients with DM2, 68% also suffer from hypertension and 53% of dyslipidemia. The prevalence of comorbid increases with CRG health status and the severity of the patient; and older ages. As expected, also it increases with the number of comorbidity (from 592 € per patient per year with one comorbidity to 3,825 € to six and more comorbidity). By type of pharmaceutical treatment, consumption is lower in biguanides and sulfonylureas, with annual amounts lower to 50 € and achieves amounts over 600 €. Finally models predictive for diabetic patients are raised, with the proposed model considering multivariate linear regression health status and severity of the patient according to the CRG system model. Moreover, the proposed analysis identifies six factors explaining 49.3% of the variance in the pharmaceutical annual consumption of diabetic patients. CONCLUSIONS DM2 shows a high prevalence and is characterized by the coexistence of other diseases, which leads to significant increases in the cost of treatment and complicates the management of the disease. Per patient consumption increases with worsening state of health of the patient and older age, and is also determined by the type of pharmaceutical treatment following by patients. Linear regression and factor analysis allow to establish models to predict and control the pharmaceutical consumption of diabetic patients.
OBJETIVO Determinación de la prevalencia y análisis de las características de la población, perfil de morbilidad, multimorbilidad, costes farmacéuticos, consumo de recursos y complicaciones de los pacientes con Diabetes mellitus (DM2) en la Comunidad Valenciana. Con ello, se pretende establecer un modelo de predicción del consumo de recursos de los pacientes con Diabetes mellitus tipo 2 (DM 2). DISEÑO DEL ESTUDIO Metodología Estudio observacional, descriptivo, retrospectivo y de corte transversal. Se elabora una base de datos con los pacientes diagnosticados de DM2 en la Comunidad Valenciana en el año 2012 y se clasifica a los mismos por estados de salud, utilizando el sistema Clinical Risk Groups (CRG). Se analizan las comorbilidades y complicaciones específicas asociadas a la enfermedad así como el gasto farmacéutico. Este se analiza desde diferentes puntos de vista: comorbilidades, estados de salud CRG y fármacos utilizados para el tratamiento de la enfermedad. Con todo ello, se plantea modelos de regresión multivariante para predecir el gasto farmacéutico en función de las diferentes variables disponibles en los que la variable dependiente es el gasto farmacéutico total por paciente y año y las variables explicativas se seleccionarán, de entre las disponibles en la base de datos elaborada. Finalmente, se presenta un análisis factorial y de componentes principales para estudiar diferencias entre los pacientes diabéticos analizados, en cuanto a gasto farmacéutico. Población de estudio Población con diagnóstico de diabetes tipo 2 de la Comunidad Valenciana en el 2012. De los 491.854 pacientes con diagnóstico de diabetes, se excluyen todos aquellos pacientes con diabetes juvenil, diabetes gestacional y los que no recibieron tratamiento farmacológico, reduciendo el grupo a 350.015 pacientes. Fuentes de información y variables de estudio Las fuentes fundamentales para la obtención de los datos: -La historia clínica electrónica, ABUCASIS. -El módulo de prestación farmacéutica, GAIA. -El Conjunto Mínimo Básico de Datos, CMBD. -El catálogo de recursos corporativos, CRC. -Sistema de Información Poblacional, SIP. -Los Grupos de Riesgo Clínico (Clinical Risk Groups, CRGs). Limitaciones del estudio 1.Estado de la codificación de los diagnósticos, tanto en términos cuantitativos como cualitativos. 2. Desequilibrio en la utilización de los sistemas de información ambulatorios corporativos con posibilidad de clasificaciones erróneas (SIA-GAIA). RESULTADOS La prevalencia bruta estimada de la diabetes fue del 7,72%, con un consumo medio por paciente y año de cerca de 1.330€. Por estados de salud, cerca de 48% de los pacientes se clasifican en el estado de salud 6. En cuanto a las comorbilidades de los pacientes con DM2, el 68% sufren además hipertensión y un 53% dislipemia. El gasto farmacéutico aumenta con la presencia de comorbilidades (de 592€ a 3.825€ anuales por paciente, con una comorbilidad a seis y más, respectivamente)., el estado de salud CRG y la severidad del paciente; y con edades más avanzadas. Por tipo de fármaco, el consumo varía entre importes anuales inferiores a 50€ hasta cifras que superan los 600€. Un modelo de regresión lineal multivariante que considera el estado de salud y severidad del paciente alcanza un coeficiente de determinación corregido¿R2 de un 39,3% de la variabilidad. Finalmente, el análisis factorial identifica 6 factores que explican el 49,3% de la varianza en el consumo anual farmacéutico de los pacientes diabéticos. CONCLUSIONES La diabetes presenta una elevada prevalencia y se caracteriza por la coexistencia de otras enfermedades, lo cual conlleva incrementos considerables en el coste del tratamiento. El consumo por paciente aumenta con el empeoramiento del estado de salud del paciente y la edad más avanzada, así como por el tipo de tratamiento farmacológico del paciente. La r
OBJECTIU Determinació de la prevalença i anàlisi de les característiques de la població, perfil de morbiditat, multimorbilitat, costos farmacèutics, consum de recursos i complicacions dels pacients amb Diabetis Mellitus (DM2) a la Comunitat Valenciana (CV). Es pretén establir un model de predicció del Consum de recursos dels Pacients amb Diabetis Mellitus tipus II (DM 2). DISSENY DE L'ESTUDI Metodologia Estudi observacional, descriptiu, retrospectiu i de tall transversal. En primer s'elabora una base de dades amb els pacients amb diagnòstic de DM2 a la Comunitat Valenciana en l'any 2012 i es classifica als pacients per estats de salut, utilitzant del sistema de Grups de Risc Clínic (CRG). S'analitzen les comorbiditats i complicacions associades a la infermetat, així com la despesa farmacèutica des de diferents perspectives: comorbiditats, estats de salut i fàrmacs utilitzats per al tractament de la malaltia. Es plantegen models de regressió multivariant per predir la despesa farmacèutica en funció de les diferents variables disponibles. La variable dependent es la despesa farmacèutica total per pacient i any, i les variables explicatives es seleccionaran, d'entre les disponibles a la base de dades elaborada. Finalment, a la tesi, es presenta l'anàlisi factorial i de components principals que ens permet estudiar les diferències entre els pacients diabètics analitzats, des del punt de vista de la despesa farmacèutica, comorbiditats relacionades i no relacionades amb la DM2 i complicacions. Població d'estudi Població amb Diagnòstic de diabetis (DM 2) de la Comunitat Valenciana en l'any 2012. Dels 491.854 pacients amb diagnòstic de la diabetis, s'exclouen de els pacients amb diabetis juvenil, diabetis gestacional i els que no van rebre tractament farmacològic, reduint el grup a 350.015. Fonts d'Informació i variables d'estudi - La història clínica electrònica, ABUCASIS. - El mòdul de prestació farmacèutica, GAIA. - El Conjunt Mínim de Dades Bàsiques, CMBD. - El catàleg de recursos corporatius, CRC. - Sistema d'Informació Poblacional, SIP. - Els Grups de Risc Clínic (grups de risc clínics, CRGs). Limitacions de l'estudi 1. Estat de codificació dels diagnòstics, tant en termes quantitatius com qualitatius. 2. Desequilibri en la utilització dels sistemes d'informació corporatius ambulatoris amb possibilitat de classificació. RESULTATS La prevalença bruta estimada de la diabetis va ser del 7,72, i el consum mitjà per pacient i any de uns 1.330€. Per estats de salut, prop de 48% dels pacients es classifiquen en l'estat de salut 6, i al voltant del 29% en el 5. En quant a les comorbiditats dels Pacients amb DM2, el 68% pateixen hipertensió, i un 53% dislipèmia. La despesa augmenta amb la presencia de comorbiditats (de 592 € a 3.825 € per pacient anualment, amb una comorbiditat i amb una sis i més, respectivament), així com amb l'estat de salut del CRG i la severitat del pacient. Per tipus de fàrmac, el consum és més baix en biguanides i sulfonilurees, amb imports anuals inferiors a 50 € fins a més de 600€ en el cas de combinacions. Es plantegen models per predicció del consum farmacèutic ambulatori de pacients diabètics del tipus 2, proposant com a resultat un model de regressió lineal multivariant que considera l'estat de salut i severitat del pacient segons el sistema CRG. D'altra banda, l'anàlisi factorial identifica 6 factors que expliquen de l'49,3% de la variància a el consum anuals farmacèutic dels pacients. CONCLUSIONS La diabetis presenta una elevada prevalença i es caracteritza per la coexistència d'altres malalties, la qual cosa comporta increments considerables en el cost del tractament i complica la gestió de la malaltia. El consum per pacient augmenta amb l'empitjorament de l'estat de salut del pacient i l'edat avançada, i a més també ve determinat pel tipus de tractament farmacològic que segueixen e
Sancho Mestre, C. (2017). MODELO PREDICTIVO DE AJUSTE DE RIESGO Y COSTES EN PACIENTES CON DIABETES MELLITUS EN LA COMUNIDAD VALENCIANA [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/83483
TESIS

The worldwide increasing number of persons affected by largely preventable diseases like diabetes demands better prevention and treatment. Insulin is required for effective utilisation of circulating nutrients. Impaired responsiveness to insulin (insulin resistance, IR) is a hallmark of type 2 diabetes and independently raises the risk of heart attack and stroke. The pathophysiology of IR is incompletely understood. High-throughput measurement of large numbers of circulating biomarkers may provide new insights beyond established risk factors. The aims of this thesis were to (i) use proteomics, metabolomics and genomics methods in large community samples to identify biomarkers of IR; (ii) assess biomarkers for risk prediction and insights into aetiology and consequences of IR; and (iii) use Mendelian randomisation analysis to assess causality. In Study I, analysis of 80 circulating proteins in 70-to-77-year-old Swedes identified cathepsin D as a biomarker for IR and highlighted a tentative causal effect of IR on raised plasma tissue plasminogen activator levels. In Study II, nontargeted fasting plasma metabolomics was used to discover 52 metabolites associated with glycaemic traits in non-diabetic 70-year-old men. Replication in independent samples of several thousand persons provided evidence for a causal effect of IR on reduced plasma oleic acid and palmitoleic acid levels. In Study III, nontargeted metabolomics in plasma samples obtained at three time points during an oral glucose challenge in 70-year-old men identified associations between a physiologic measure of IR and concentration changes in medium-chain acylcarnitines, monounsaturated fatty acids, bile acids and lysophosphatidylethanolamines. Study IV provided evidence in two large longitudinal cohorts for causal effects of type 2 diabetes and impaired insulin secretion on raised coronary artery disease risk. In conclusion, the Studies in this thesis provide new insights into the pathophysiology and adverse health consequences of IR and illustrate the value of combining traditional epidemiologic designs with recent molecular techniques and bioinformatics methods. The results provide limited evidence for the role of circulating proteins and small molecules in IR and require replication in separate studies and validation in experimental designs.

The utilization of in vitro tests with a tiered testing strategy for detection of mild ocular irritants can reduce the use of animals for testing, provide mechanistic data on toxic effects, and reduce the uncertainty associated with dose selection for clinical trials. The first section of this thesis describes how in vitro methods can be used to improve the prediction of the toxicity of chemicals and ophthalmic products. The proper utilization of in vitro methods can accurately predict toxic threshold levels and reduce animal use in product development. Sections two, three and four describe the development of new sensitive in vitro methods for predicting ocular toxicity. Maintaining the barrier function of the cornea is critical for the prevention of the penetration of infections microorganisms and irritating chemicals into the eye. Chapter 2 describes the development of a method for assessing the effects of chemicals on tight junctions using a human corneal epithelial and canine kidney epithelial cell line. In Chapter 3 a method that uses a primary organ culture for assessing single instillation and multiple instillation toxic effects is described. The ScanTox system was shown to be an ideal system to monitor the toxic effects over time as multiple readings can be taken of treated bovine lenses using the nondestructive method of assessing for the lens optical quality. Confirmations of toxic effects were made with the utilization of the viability dye alamarBlue. Chapter 4 describes the development of sensitive in vitro assays for detecting ocular toxicity by measuring the effects of chemicals on the mitochondrial integrity of bovine cornea, bovine lens epithelium and corneal epithelial cells, using fluorescent dyes. The goal of this research was to develop an in vitro test battery that can be used to accurately predict the ocular toxicity of new chemicals and ophthalmic formulations. By comparing the toxicity seen in vivo animals and humans with the toxicity response in these new in vitro methods, it was demonstrated that these in vitro methods can be utilized in a tiered testing strategy in the development of new chemicals and ophthalmic formulations.