In this work, a practical approach to assessing bias and uncertainty using patient samples in a clinical laboratory is presented. The scheme is essentially a splitsample setup where one instrument is appointed to being the “master” instrument which other instruments are compared to. The software presented automatically collects test results from a Laboratory Information System in production and couples together the results of pairwise measurements. Partitioning of measurement results by user-defined criteria and how this can facilitate isolation of variation sources are also discussed. The logic and essential data model are described and the surrounding workflows outlined. The described software and workflow are currently in considerable practical use in several Swedish large-scale distributed laboratory organizations. With the appropriate IT-support, split-sample testing can be a powerful complement to external quality assurance.

Small cell (neuroendocrine) carcinoma of the uterine cervix (SMCC) is a rare, highly aggressive malignant neoplasm. Both conventional and liquid-based cytology (LBC) cervical smears have low sensitivity in diagnosing SMCC, requiring immunocytochemical (ICH) confirmation. We present the first series of SMCC primarily diagnosed in cytology specimens, and ICH studies performed on the residual LBC specimens with subsequent confirmation of the diagnosis on surgical pathology specimens. Immunocytochemical stains for keratin, p16INK4, and neuroendocrine markers (synaptophysin, chromogranin, CD56) were performed on additional ThinPrep slides. HPV test used chromogenic in situ hybridization high risk HPV DNA probe. The Pap smears in all three specimens were highly cellular with a mixture of squamous cells and numerous well-preserved single or small cohesive clusters of malignant epithelial cells. Tumor cells were small, monomorphic with minimal cytoplasm and high nuclear/cytoplasmic ratio. There was significant nuclear overlap, but no nuclear molding, or smudging of nuclear chromatin. The chromatin pattern was stippled. A background tumor diathesis was prominent. Atypical squamous cells of undetermined significance (ASCUS) were noted in one case, and markedly abnormal squamous cells were seen in another case. The main cytology differential diagnoses included high-grade squamous intraepithelial lesion and an endometrial adenocarcinoma. Immunocytochemical positivity for the neuroendocrine markers supported the diagnoses of SMCC in all three cases. The morphologic features of the concurrent surgical pathology specimens were typical of SMCC. The tissue diagnoses were also confirmed by immunohistochemistry. Our study allows us to conclude that SMCC can be primarily diagnosed in LBC specimens using a panel of immunocytochemical stains.

Background Our objective was to identify and validate clinical and continuity of care variables associated with Oral anticoagulant (OAC)-related adverse events within 30 days of hospital discharge amongst seniors. Methods and Analysis This was a population-based retrospective cohort study of all adults aged 66 years or older who were discharged from hospital on an OAC from September 2010 to March 2015 in Ontario, Canada. The primary outcome was a composite of the time to first hospitalization or Emergency Department visit for a hemorrhage or thromboembolic event or mortality within 30 days of hospital discharge. A Cox proportional hazards model was used to determine the association between the composite outcome and a set of prespecified covariates. A split sample method was applied to validate the final model. Results We included 120 721 Ontario seniors of which 5423 suffered one of the primary adverse events. Patients discharged on a direct-acting oral anticoagulant (DOAC); dispensed the same OAC in the past 12 months; who had a history of a thromboembolic event; had a recent joint replacement or major surgery; had a cardiologist, hematologist or orthopedic surgeon as compared to a family medicine physician as the physician prescribing the OAC at discharge had a lower risk for the composite outcome. Though continuity of care was a variable in the final multivariate Cox model, it was not significant. The Cox model was stable with acceptable discrimination but poor goodness-of-fit. Conclusion In this study, we found that continuity of care as measured by outpatient follow-up in the 7 days post-discharge was not significantly associated with the composite outcome. Further exploration to improve the current model’s calibration and interpretation are required. / Thesis / Master of Science (MSc) / Background Oral anticoagulants (OACs) are associated with serious adverse events, with high rates immediately post-hospitalization. We aimed to identify and validate clinical and continuity of care variables in seniors discharged from hospital on an OAC, which are associated with OAC-related harm in the short-term high-risk period following hospitalization. Methods Data from administrative health databases in Ontario were used to identify and validate risk factors associated with time to first OAC-related serious events including hospitalization or emergency department visit for a bleeding or thromboembolic event, and mortality. Cox proportional hazards model and split-sample methods were utilized. Results We included 120,721 seniors of which 5423 suffered one of the primary events. Patient-, physician- and index hospitalization-characteristics were all associated with time to the composite outcome. Though continuity of care risk factor was part of the final model, it was not a significant predictor for the outcome. Conclusion Exploration of this model through sensitivity analysis is required.

A changing climate can severely perturb regional hydrology and thereby affect human societies and life in general. To assess and simulate such potential hydrological climate change impacts, hydrological models require reliable meteorological variables for current and future climate conditions. Global climate models (GCMs) provide such information, but their spatial scale is too coarse for regional impact studies. Thus, GCM output needs to be downscaled to a finer scale either through statistical downscaling or through dynamic regional climate models (RCMs). However, even downscaled meteorological variables are often considerably biased and therefore not directly suitable for hydrological impact modeling. This doctoral thesis discusses biases and other challenges related to incorporating climate model output into hydrological studies and evaluates possible strategies to address them. An analysis of possible sources of uncertainty stressed the need for full ensembles approaches, which should become standard practice to obtain robust and meaningful hydrological projections under changing climate conditions. Furthermore, it was shown that substantial biases in current RCM simulations exist and that correcting them is an essential prerequisite for any subsequent impact simulation. Bias correction algorithms considerably improved RCM output and subsequent streamflow simulations under current conditions. In addition, differential split-sample testing was highlighted as a powerful tool for evaluating the transferability of bias correction algorithms to changed conditions. Finally, meaningful projections of future streamflow regimes could be realized by combining a full ensemble approach with bias correction of RCM output: Current flow regimes in Sweden with a snowmelt-driven spring flood in April will likely change to rather damped flow regimes that are dominated by large winter streamflows.

Thèse numérisée par la Direction des bibliothèques de l'Université de Montréal. / Cette thèse a pour objectif de développer des méthodes d'inférence exacte dans les modèles structurels. Elle est motivée par les résultats reportés récemment dans la littérature économétrique concernant les problèmes des méthodes d'inférence usuelles en présence d'instruments faibles. Dans le premier essai, nous étudions le problème de construction de régions de confiance pour des transformations du vecteur des paramètres inconnus dans un modèle à équations simultanées linéaires. Plusieurs tests proposés pour l'inférence en présence d'instruments faibles présentent l'inconvénient de ne tester que des hypothèses spécifiant le vecteur entier des paramètres. C'est le cas notamment de la statistique d'Andersen-Rubin (1949, Ann. Math. Stat.) et des deux statistiques proposées récemment par Wang et Zivot (1998, Econometrica). En principe, ce problème peut être résolu en utilisant la technique de projection Dufour (1997, Econometrica), Dufour and Jasiak (2001, International Economie Review)]. Cependant, cette dernière exige, de façon générale, l'utilisation des méthodes numériques. Dans cet essai, nous fournissons une solution analytique complète au problème de construction d'intervalles de confiance par projection à partir de régions de confiance obtenues par inversion de statistiques de type Anderson-Rubin. Cette solution se base sur la théorie des quadriques et peut être perçue comme une extension des intervalles et des ellipsoïdes de confiance usuels. Les calculs requis sont semblables à ceux des moindres carrés ordinaires. Nous étudions également par des simulations Monte Carlo le degré de conservatisme des régions de confiance par projection. Enfin, nous illustrons les méthodes proposées par deux applications macroéconomiques: La relation entre le commerce international et la croissance économique et le problème des rendements d'échelle dans l'industrie américaine. Dans le deuxième essai, nous proposons des méthodes d'inférence exacte dans un modèle structurel non linéaire. Nous développons une approche simple pour construire des tests exacts. Cette approche généralise celle de Hartley (1964, Biometrika) pour le test d'hypothèses et la construction de régions de confiance dans les régressions non linéaires et la procédure d'Anderson-Rubin spécifique aux modèles structurels à équations linéaires. Nous abordons par la suite le problème d'optimalité des instruments utilisés dans le test. La notion d'optimalité réfère à la maximisation de la puissance du test proposé. Ceci contraste avec la notion d'optimalité typiquement considérée dans la littérature économétrique et qui réfère à la minimisation de la variance asymptotique d'un estimateur [Amemiya (1977, Econometrica)]. Les instruments optimaux que nous dérivons dépendent de l'hypothèse alternative, et donc nous qualifions ces derniers d'instruments "point-optimaux" [King (1988, Econometric Reviews). La matrice des instruments optimaux est inconnue de façon générale, nous proposons une méthode pour son estimation basée sur la technique du split-sample. Dans le troisième essai, nous examinons le problème de sélection d'instruments dans le cas d'un modèle structurel linéaire où la méthode 2SLS pour l'estimation et/ou le test d'Anderson-Rubin sont utilisés. Nous nous concentrons sur le cas d'une seule variable explicative endogène. Nous commençons par analyser les déterminants de la performance de ces méthodes et comment la matrice des instruments affecte la qualité de l'estimation et de l'inférence. Nous proposons par la suite une méthode de sélection d'instruments sur la base de ces résultats. Cette méthode est basée sur la maximisation séquentielle du paramètre de concentration. Nous étudions si les méthodes de diagnostic et de sélection d'instruments sont utiles en pratique ou comme conclu par Hall, Rudebusch and Wilcox (1996, International Economie Review) sont inutilisables. Nous comparons également, par des simulations Monte Carlo, les principales méthodes de sélection d'instruments proposées dans la littérature.