Discovering unknown response patterns in progress test data to improve the estimation of student performance.

BACKGROUND: The Progress Test Medizin (PTM) is a 200-question formative test that is administered to approximately 11,000 students at medical universities (Germany, Austria, Switzerland) each term. Students receive feedback on their knowledge (development) mostly in comparison to their own cohort. In this study, we use the data of the PTM to find groups with similar response patterns. METHODS: We performed k-means clustering with a dataset of 5,444 students, selected cluster number k = 5, and answers as features. Subsequently, the data was passed to XGBoost with the cluster assignment as target enabling the identification of cluster-relevant questions for each cluster with SHAP. Clusters were examined by total scores, response patterns, and confidence level. Relevant questions were evaluated for difficulty index, discriminatory index, and competence levels. RESULTS: Three of the five clusters can be seen as "performance" clusters: cluster 0 (n = 761) consisted predominantly of students close to graduation. Relevant questions tend to be difficult, but students answered confidently and correctly. Students in cluster 1 (n = 1,357) were advanced, cluster 3 (n = 1,453) consisted mainly of beginners. Relevant questions for these clusters were rather easy. The number of guessed answers increased. There were two "drop-out" clusters: students in cluster 2 (n = 384) dropped out of the test about halfway through after initially performing well; cluster 4 (n = 1,489) included students from the first semesters as well as "non-serious" students both with mostly incorrect guesses or no answers. CONCLUSION: Clusters placed performance in the context of participating universities. Relevant questions served as good cluster separators and further supported our "performance" cluster groupings.

Hospitalization Before Liver Transplantation Predicts Posttransplant Patient Survival: A Propensity Score-Matched Analysis.

In contrast to donor factors predicting outcomes of liver transplantation (LT), few suitable recipient parameters have been identified. To this end, we performed an in-depth analysis of hospitalization status and duration prior to LT as a potential risk factor for posttransplant outcome. The pretransplant hospitalization status of all patients undergoing LT between 2005 and 2016 at the Charité-Universitätsmedizin Berlin was analyzed retrospectively using propensity score matching. At the time of organ acceptance, 226 of 1134 (19.9%) recipients were hospitalized in an intensive care unit (ICU), 146 (12.9%) in a regular ward (RW) and 762 patients (67.2%) were at home. Hospitalized patients (RW and ICU) compared with patients from home showed a dramatically shorter 3-month survival (78.7% versus 94.4%), 1-year survival (66.3% versus 87.3%), and 3-year survival (61.7% versus 81.7%; all P < 0.001), whereas no significant difference was detected for 3-year survival between ICU and RW patients (61.5% versus 62.3%; P = 0.60). These results remained significant after propensity score matching. Furthermore, in ICU patients, but not in RW patients, survival correlated with days spent in the ICU before LT (1-year survival: 1-6 versus 7-14 days: 73.7% versus 60.5%, P = 0.04; 7-14 days versus >14 days, 60.5% versus 51.0%, P = 0.006). In conclusion, hospitalization status before transplantation is a valuable predictor of patient survival following LT.

Clinical interpretation of CNVs with cross-species phenotype data.

BACKGROUND: Clinical evaluation of CNVs identified via techniques such as array comparative genome hybridisation (aCGH) involves the inspection of lists of known and unknown duplications and deletions with the goal of distinguishing pathogenic from benign CNVs. A key step in this process is the comparison of the individual's phenotypic abnormalities with those associated with Mendelian disorders of the genes affected by the CNV. However, because often there is not much known about these human genes, an additional source of data that could be used is model organism phenotype data. Currently, almost 6000 genes in mouse and zebrafish are, when knocked out, associated with a phenotype in the model organism, but no disease is known to be caused by mutations in the human ortholog. Yet, searching model organism databases and comparing model organism phenotypes with patient phenotypes for identifying novel disease genes and medical evaluation of CNVs is hindered by the difficulty in integrating phenotype information across species and the lack of appropriate software tools. METHODS: Here, we present an integrated ranking scheme based on phenotypic matching, degree of overlap with known benign or pathogenic CNVs and the haploinsufficiency score for the prioritisation of CNVs responsible for a patient's clinical findings. RESULTS: We show that this scheme leads to significant improvements compared with rankings that do not exploit phenotypic information. We provide a software tool called PhenogramViz, which supports phenotype-driven interpretation of aCGH findings based on multiple data sources, including the integrated cross-species phenotype ontology Uberpheno, in order to visualise gene-to-phenotype relations. CONCLUSIONS: Integrating and visualising cross-species phenotype information on the affected genes may help in routine diagnostics of CNVs.

The Effects of MELD-Based Liver Allocation on Patient Survival and Waiting List Mortality in a Country with a Low Donation Rate.

The Model for End-Stage Liver Disease (MELD)-based allocation system was implemented in Germany in 2006 in order to reduce waiting list mortality. The purpose of this study was to evaluate post-transplant results and waiting list mortality since the introduction of MELD-based allocation in our center and in Germany. Adult liver transplantation at the Charité-Universitätsmedizin Berlin was assessed retrospectively between 2005 and 2012. In addition, open access data from Eurotransplant (ET) and the German Organ Transplantation Foundation (DSO) were evaluated. In our department, 861 liver transplantations were performed from 2005 to 2012. The mean MELD score calculated with the laboratory values last transmitted to ET before organ offer (labMELD) at time of transplantation increased to 20.1 from 15.8 (Pearson's R = 0.121, p < 0.001, confidence interval (CI) = 0.053-0.187). Simultaneously, the number of transplantations per year decreased from 139 in 2005 to 68 in 2012. In order to overcome this organ shortage the relative number of utilized liver donors in Germany has increased (85% versus 75% in non-German ET countries). Concomitantly, 5-year patient survival decreased from 79.9% in 2005 to 60.3% in 2012 (p = 0.048). At the same time, the ratio of waiting list mortality vs. active-listed patients nearly doubled in Germany (Spearman's rho = 0.903, p < 0.001, CI = 0.634-0.977). In low-donation areas, MELD-based liver allocation may require reconsideration and inclusion of prognostic outcome factors.