Search

In clinical practice, it is often the case where the association between the occurrence of events and time-to-event outcomes is of interest; thus, it can be modeled within the framework of recurrent events. The purpose of our study is to enrich the information available for modeling survival with relevant dynamic features, properly taking into account their possibly time-varying nature, as well as to provide a new setting for quantifying the association between time-varying processes and time-to-event outcomes. We propose an innovative methodology to model information carried out by time-varying processes by means of functional data, modeling each time-varying variable as the compensator of marked point process the recurrent events are supposed to derive from. By means of Functional Principal Component Analysis, a suitable dimensional reduction of these objects is carried out in order to plug them into a Cox-type functional regression model for overall survival. We applied our methodology to data retrieved from the administrative databases of Lombardy Region (Italy), related to patients hospitalized for Heart Failure (HF) between 2000 and 2012. We focused on time-varying processes of HF hospitalizations and multiple drugs consumption and we studied how they influence patients' overall survival. This novel way to account for time-varying variables allowed to model self-exciting behaviors, for which the occurrence of events in the past increases the probability of a new event, and to quantify the effect of personal behaviors and therapeutic patterns on survival, giving new insights into the direction of personalized treatment.

OBJECTIVES Although it has been postulated that patients might benefit from the centralization of high-volume specialized centres, conflicting results have been reported on the relationship between the number of lung resections performed and the long-term, all-cause mortality rates among patients who underwent surgery for lung cancer. A population-based observational study was performed to contribute to the ongoing debate. METHODS The 2613 patients, all residents of the Lombardy region (Italy), who underwent lung resection for lung cancer from 2012 to 2014 were entered into the cohort and were followed until 2018. The hospitals were classified according to the annual number of pulmonary resections performed. Three categories of lung resection cases were identified: low (≤30), intermediate (31–95) and high (>95). The outcome of interest was all-cause death. A frailty model was used to estimate the death risk associated with the categories of numbers of lung resections performed, taking into account the multilevel structure of the data. A set of sensitivity analyses was performed to account for sources of systematic uncertainty. RESULTS The 1-year and 5-year survival rates of cohort members were 90% and 63%. Patients operated on in high-volume centres were on average younger and more often women. Compared to patients operated on in a low-volume centre, the mortality risk exhibited a significant, progressive reduction as the numbers of lung resections performed increased to intermediate (−13%; 95% confidence interval +10% to −31%) and high (−26%; 0% to −45%). Sensitivity analyses revealed that the association was consistent. CONCLUSIONS Further evidence that the volume of lung resection cases performed strongly affects the long-term survival of lung cancer patients has been supplied.

Background The role of image-derived biomarkers in recurrent oligometastatic Prostate Cancer (PCa) is unexplored. This paper aimed to evaluate [18F]FMCH PET/CT radiomic analysis in patients with recurrent PCa after primary radical therapy. Specifically, we tested intra-patient lesions similarity in oligometastatic and plurimetastatic PCa, comparing the two most used definitions of oligometastatic disease. Methods PCa patients eligible for [18F]FMCH PET/CT presenting biochemical failure after first-line curative treatments were invited to participate in this prospective observational trial. PET/CT images of 92 patients were visually and quantitatively analyzed. Each patient was classified as oligometastatic or plurimetastatic according to the total number of detected lesions (up to 3 and up to 5 or > 3 and > 5, respectively). Univariate and intra-patient lesions' similarity analysis were performed. Results [18F]FMCH PET/CT identified 370 lesions, anatomically classified as regional lymph nodes and distant metastases. Thirty-eight and 54 patients were designed oligometastatic and plurimetastatic, respectively, using a 3-lesion threshold. The number of oligometastic scaled up to 60 patients (thus 32 plurimetastatic patients) with a 5-lesion threshold. Similarity analysis showed high lesions' heterogeneity. Grouping patients according to the number of metastases, patients with oligometastatic PCa defined with a 5-lesion threshold presented lesions heterogeneity comparable to plurimetastic patients. Lesions within patients having a limited tumor burden as defined by three lesions were characterized by less heterogeneity. Conclusions We found a comparable heterogeneity between patients with up to five lesions and plurimetastic patients, while patients with up to three lesions were less heterogeneous than plurimetastatic patients, featuring different cells phenotypes in the two groups. Our results supported the use of a 3-lesion threshold to define oligometastatic PCa.

AIM: To identify the effect of single nucleotide polymorphism (SNP) interactions on the risk of toxicity following radiotherapy (RT) for prostate cancer (PCa) and propose a new method for polygenic risk score incorporating SNP-SNP interactions (PRSi). MATERIALS AND METHODS: Analysis included the REQUITE PCa cohort that received external beam RT and was followed for 2 years. Late toxicity endpoints were: rectal bleeding, urinary frequency, haematuria, nocturia, decreased urinary stream. Among 43 literature-identified SNPs, the 30% most strongly associated with each toxicity were tested. SNP-SNP combinations (named SNP-allele sets) seen in ≥10% of the cohort were condensed into risk (RS) and protection (PS) scores, respectively indicating increased or decreased toxicity risk. Performance of RS and PS was evaluated by logistic regression. RS and PS were then combined into a single PRSi evaluated by area under the receiver operating characteristic curve (AUC). RESULTS: Among 1,387 analysed patients, toxicity rates were 11.7% (rectal bleeding), 4.0% (urinary frequency), 5.5% (haematuria), 7.8% (nocturia) and 17.1% (decreased urinary stream). RS and PS combined 8 to 15 different SNP-allele sets, depending on the toxicity endpoint. Distributions of PRSi differed significantly in patients with/without toxicity with AUCs ranging from 0.61 to 0.78. PRSi was better than the classical summed PRS, particularly for the urinary frequency, haematuria and decreased urinary stream endpoints. CONCLUSIONS: Our method incorporates SNP-SNP interactions when calculating PRS for radiotherapy toxicity. Our approach is better than classical summation in discriminating patients with toxicity and should enable incorporating genetic information to improve normal tissue complication probability models. ispartof: RADIOTHERAPY AND ONCOLOGY vol:159 pages:241-248 ispartof: location:Ireland status: published

Simple Summary Patients receiving chemotherapy for liver metastases from colorectal cancer may develop liver injuries that impair hepatic function and postoperative outcome. The non-invasive diagnosis of these damages is still an unmet need. Recently, advanced imaging analysis techniques, including the so-called “radiomics”, achieved adequate prediction of pathology data. The present study demonstrated that radiomic analysis of liver parenchyma in combination with clinical and laboratory data improves non-invasive diagnosis of chemotherapy-related liver injuries. Abstract Non-invasive diagnosis of chemotherapy-associated liver injuries (CALI) is still an unmet need. The present study aims to elucidate the contribution of radiomics to the diagnosis of sinusoidal dilatation (SinDil), nodular regenerative hyperplasia (NRH), and non-alcoholic steatohepatitis (NASH). Patients undergoing hepatectomy for colorectal metastases after chemotherapy (January 2018-February 2020) were retrospectively analyzed. Radiomic features were extracted from a standardized volume of non-tumoral liver parenchyma outlined in the portal phase of preoperative post-chemotherapy computed tomography. Seventy-eight patients were analyzed: 25 had grade 2–3 SinDil, 27 NRH, and 14 NASH. Three radiomic fingerprints independently predicted SinDil: GLRLM_f3 (OR = 12.25), NGLDM_f1 (OR = 7.77), and GLZLM_f2 (OR = 0.53). Combining clinical, laboratory, and radiomic data, the predictive model had accuracy = 82%, sensitivity = 64%, and specificity = 91% (AUC = 0.87 vs. AUC = 0.77 of the model without radiomics). Three radiomic parameters predicted NRH: conventional_HUQ2 (OR = 0.76), GLZLM_f2 (OR = 0.05), and GLZLM_f3 (OR = 7.97). The combined clinical/laboratory/radiomic model had accuracy = 85%, sensitivity = 81%, and specificity = 86% (AUC = 0.91 vs. AUC = 0.85 without radiomics). NASH was predicted by conventional_HUQ2 (OR = 0.79) with accuracy = 91%, sensitivity = 86%, and specificity = 92% (AUC = 0.93 vs. AUC = 0.83 without radiomics). In the validation set, accuracy was 72%, 71%, and 91% for SinDil, NRH, and NASH. Radiomic analysis of liver parenchyma may provide a signature that, in combination with clinical and laboratory data, improves the diagnosis of CALI.

EEG technology finds applications in several domains. Currently, most EEG systems require subjects to wear several electrodes on the scalp to be effective. However, several channels might include noisy information, redundant signals, induce longer preparation times and increase computational times of any automated system for EEG decoding. One way to reduce the signal-to-noise ratio and improve classification accuracy is to combine channel selection with feature extraction, but EEG signals are known to present high inter-subject variability. In this work we introduce a novel algorithm for subject-independent channel selection of EEG recordings. Considering multi-channel trial recordings as statistical units and the EEG decoding task as the class of reference, the algorithm (i) exploits channel-specific 1D-Convolutional Neural Networks (1D-CNNs) as feature extractors in a supervised fashion to maximize class separability; (ii) it reduces a high dimensional multi-channel trial representation into a unique trial vector by concatenating the channels' embeddings and (iii) recovers the complex inter-channel relationships during channel selection, by exploiting an ensemble of AutoEncoders (AE) to identify from these vectors the most relevant channels to perform classification. After training, the algorithm can be exploited by transferring only the parametrized subgroup of selected channel-specific 1D-CNNs to new signals from new subjects and obtain low-dimensional and highly informative trial vectors to be fed to any classifier.

Nowadays, the importance of educational data mining and learning analytics in higher education institutions is being recognised. The analysis of university careers and of student dropout prediction is one of the most studied topics in the area of learning analytics. From the perspective of estimating the likelihood of a student dropping out, we propose an innovative statistical method that is a generalisation of mixed-effects trees for a response variable in the exponential family: generalised mixed-effects trees (GMET). We performed a simulation study in order to validate the performance of our proposed method and to compare GMET to classical models. In the case study, we applied GMET to model undergraduate student dropout in different courses at Politecnico di Milano. The model was able to identify discriminating student characteristics and estimate the effect of each degree-based course on the probability of student dropout.

Adherence to medication is the process by which patients take their drugs as prescribed, and represents an issue in pharmacoepidemiological studies. Poor adherence is often associated with adverse health conditions and outcomes, especially in case of chronic diseases such as heart failure (HF). This turns out in an increased request for health care services, and in a greater burden for the health care system. In recent years, there has been a substantial growth in pharmacotherapy research, aimed at studying effects and consequences of proper/improper adherence to medication both for the increasing awareness of the problem and for the pervasiveness of poor adherence among patients. However, the way adherence is computed and accounted for into predictive models is far from being informative as it may be. In fact, it is usually analyzed as a fixed baseline covariate, without considering its time-varying behavior. The purpose and novelty of this study is to define a new personalized monitoring tool exploiting time-varying definition of adherence to medication, within a joint modeling approach. In doing so, we are able to capture and quantify the association between the longitudinal process of dynamic adherence to medication with the long-term survival outcome. Another novelty of this approach consists of exploiting the potential of health care administrative databases in order to reconstruct the dynamics of drugs consumption through pharmaceutical administrative registries. In particular, we analyzed administrative data provided by Regione Lombardia - Healthcare Division related to patients hospitalized for HF between 2000 and 2012.

Background: REQUITE (validating pREdictive models and biomarkers of radiotherapy toxicity to reduce side effects and improve QUalITy of lifE in cancer survivors) is an international prospective cohort study. The purpose of this project was to analyse a cohort of patients recruited into REQUITE using a deep learning algorithm to identify patient-specific features associated with the development of toxicity, and test the approach by attempting to validate previously published genetic risk factors. Methods: The study involved REQUITE prostate cancer patients treated with external beam radiotherapy who had complete 2-year follow-up. We used five separate late toxicity endpoints: ≥grade 1 late rectal bleeding, ≥grade 2 urinary frequency, ≥grade 1 haematuria, ≥ grade 2 nocturia, ≥ grade 1 decreased urinary stream. Forty-three single nucleotide polymorphisms (SNPs) already reported in the literature to be associated with the toxicity endpoints were included in the analysis. No SNP had been studied before in the REQUITE cohort. Deep Sparse AutoEncoders (DSAE) were trained to recognize features (SNPs) identifying patients with no toxicity and tested on a different independent mixed population including patients without and with toxicity. Results: One thousand, four hundred and one patients were included, and toxicity rates were: rectal bleeding 11.7%, urinary frequency 4%, haematuria 5.5%, nocturia 7.8%, decreased urinary stream 17.1%. Twenty-four of the 43 SNPs that were associated with the toxicity endpoints were validated as identifying patients with toxicity. Twenty of the 24 SNPs were associated with the same toxicity endpoint as reported in the literature: 9 SNPs for urinary symptoms and 11 SNPs for overall toxicity. The other 4 SNPs were associated with a different endpoint. Conclusion: Deep learning algorithms can validate SNPs associated with toxicity after radiotherapy for prostate cancer. The method should be studied further to identify polygenic SNP risk signatures for radiotherapy toxicity. The signatures could then be included in integrated normal tissue complication probability models and tested for their ability to personalize radiotherapy treatment planning. ispartof: FRONTIERS IN ONCOLOGY vol:10 ispartof: location:Switzerland status: published

Background Investigating similarities and differences among healthcare providers, on the basis of patient healthcare experience, is of interest for policy making. Availability of high quality, routine health databases allows a more detailed analysis of performance across multiple outcomes, but requires appropriate statistical methodology. Methods Motivated by analysis of a clinical administrative database of 42,871 Heart Failure patients, we develop a semi-Markov, illness-death, multi-state model of repeated admissions to hospital, subsequent discharge and death. Transition times between these health states each have a flexible baseline hazard, with proportional hazards for patient characteristics (case-mix adjustment) and a discrete distribution for frailty terms representing clusters of providers. Models were estimated using an Expectation-Maximization algorithm and the number of clusters was based on the Bayesian Information Criterion. Results We are able to identify clusters of providers for each transition, via the inclusion of a nonparametric discrete frailty. Specifically, we detect 5 latent populations (clusters of providers) for the discharge transition, 3 for the in-hospital to death transition and 4 for the readmission transition. Out of hospital death rates are similar across all providers in this dataset. Adjusting for case-mix, we could detect those providers that show extreme behaviour patterns across different transitions (readmission, discharge and death). Conclusions The proposed statistical method incorporates both multiple time-to-event outcomes and identification of clusters of providers with extreme behaviour simultaneously. In this way, the whole patient pathway can be considered, which should help healthcare managers to make a more comprehensive assessment of performance.

Background According to published data, radiomics features differ between lesions of refractory/relapsing HL patients from those of long-term responders. However, several methodological aspects have not been elucidated yet. Purpose The study aimed at setting up a methodological framework in radiomics applications in Hodgkin’s lymphoma (HL), especially at (a) developing a novel feature selection approach, (b) evaluating radiomic intra-patient lesions’ similarity, and (c) classifying relapsing refractory (R/R) vs non-(R/R) patients. Methods We retrospectively included 85 patients (male:female = 52:33; median age 35 years, range 19–74). LIFEx (www.lifexsoft.org) was used for [18F]FDG-PET/CT segmentation and feature extraction. Features were a-priori selected if they were highly correlated or uncorrelated to the volume. Principal component analysis-transformed features were used to build the fingerprints that were tested to assess lesions’ similarity, using the silhouette. For intra-patient similarity analysis, we used patients having multiple lesions only. To classify patients as non-R/R and R/R, the fingerprint considering one single lesion (fingerprint_One) and all lesions (fingerprint_All) was tested using Random Undersampling Boosting of Tree Ensemble (RUBTE). Results HL fingerprints included up to 15 features. Intra-patient lesion similarity analysis resulted in mean/median silhouette values below 0.5 (low similarity especially in the non-R/R group). In the test set, the fingerprint_One classification accuracy was 62% (78% sensitivity and 53% specificity); the classification by RUBTE using fingerprint_All resulted in 82% accuracy (70% sensitivity and 88% specificity). Conclusions Lesion similarity analysis was developed, and it allowed to demonstrate that HL lesions were not homogeneous within patients in terms of radiomics signature. Therefore, a random target lesion selection should not be adopted for radiomics applications. Moreover, the classifier to predict R/R vs non-R/R performed the best when all the lesions were used.

BACKGROUND: Much data about prescription adherence in patients with heart failure (HF) are available, but few exist about the evaluation of true patient adherence. Further, methods for analyzing this issue are poorly known. OBJECTIVES: Our objective was to evaluate the impact of patient adherence to disease-modifying drugs after HF hospitalization in a community-based cohort. METHODS AND RESULTS: Patients hospitalized with first diagnostic HF code and at least one post-discharge purchase of evidence-based drugs for HF between 2009 and 2015 were included (12,938 patients). A new method for measuring adherence to polypharmacy (patient adherence indicator [PAI]) was introduced, based on proportion of days covered (PDC) and medication possession ratio (MPR). The investigated drugs were β-blockers (BBs), angiotensin-converting enzyme inhibitors (ACEIs), angiotensin-receptor blockers (ARBs), and anti-aldosterone agents (AAs). Regional administrative databases were analyzed. RESULTS: The mean age of the cohort was 80 years; 53% was female; the median Charlson Comorbidity Index score was 2, and the overall death rate was 60%. PAI based on PDC estimated a nonadherence rate of 47%. Median daily dosages were well below target dosages for all drugs considered. A good PAI significantly lowered the mortality risk, irrespective of the computational method used: PDC (PAI adjusted hazard ratio [HR] 0.93; 95% confidence interval [CI] 0.88-0.97; p = 0.001) or MPR (PAI adjusted HR 0.93; 95% CI 0.89-0.98; p = 0.004). CONCLUSIONS: In a real-world setting, medication adherence of patients with HF remains unsatisfactory, especially when in a polypharmacy setting. Irrespective of PDC and MPR, good patient adherence to polypharmacy was associated with a lower death rate.

Background The healthcare sector is an interesting target for fraudsters. The availability of a great amount of data makes it possible to tackle this issue with the adoption of data mining techniques, making the auditing process more efficient and effective. This research has the objective of developing a novel data mining model devoted to fraud detection among hospitals using Hospital Discharge Charts (HDC) in Administrative Databases. In particular, it is focused on the DRG upcoding practice, i.e., the tendency of registering codes for provided services and inpatients health status so to make the hospitalization fall within a more remunerative DRG class. Methods We propose a two-step algorithm: the first step entails kmeans clustering of providers to identify locally consistent and locally similar groups of hospitals, according to their characteristics and behavior treating a specific disease, in order to spot outliers within this groups of peers. An initial grid search for the best number of features to be selected (through Principal Feature Analysis) and the best number of local groups makes the algorithm extremely flexible. In the second step, we propose a human-decision support system that helps auditors cross-validating the identified outliers, analyzing them w.r.t. fraud-related variables, and the complexity of patients’ casemix they treated. The proposed algorithm was tested on a database relative to HDC collected by Regione Lombardia (Italy) in a time period of three years (2013-2015), focusing on the treatment of Heart Failure. Results The model identified 6 clusters of hospitals and 10 outliers among the 183 units. Out of those providers, we report the in depth the application of Step Two on three Hospitals (two private and one public). Cross-validating with the patients’ population and the hospitals’ characteristics, the public hospital seemed justified in its outlierness, while the two private providers were deemed interesting for a further investigation by auditors. Conclusions The proposed model is promising in identifying anomalous DRG coding behavior and it is easily transferrable to all diseases and contexts of interest. Our proposal contributes to the limited literature regarding behavioral models for fraud detection, identifying the most ’cautious’ fraudsters. The results of the first and the second Steps together represent a valuable set of information for auditors in their preliminary investigation.

Health care expenditures constitute a significant portion of governmental budgets. The percentage of fraud, waste and abuse within that spending has increased over years. This paper introduces the emerging area of statistical medical fraud assessment, which becomes crucial to handle the increasing size and complexity of the medical programmes. An overview of fraud types and detection is followed by the description of medical claims data. The utilisation of sampling, overpayment estimation and data mining methods in medical fraud assessment are presented. Recent unsupervised methods are illustrated with real world data. Finally, the paper introduces potential future research areas such as integrated decision making approaches and Bayesian methods and concludes with an overall discussion. The main goal of this exposition is to increase awareness about this important area among a broader audience of statisticians.

We propose a new method for detecting outliers in multivariate functional data. We exploit the joint use of two different depth measures, and generalize the outliergram to the multivariate functional framework, aiming at detecting and discarding both shape and magnitude outliers. The main application consists in robustifying the reference samples of data, composed by G different known groups to be used, for example, in classification procedures in order to make them more robust. We asses by means of a simulation study the method’s performance in comparison with different outlier detection methods. Finally we consider a real dataset: we classify data minimizing a suitable distance from the center of reference groups. We compare performance of supervised classification on test sets training the algorithm on original dataset and on the robustified one, respectively.

Healthcare administrative databases are becoming more and more important and reliable sources of clinical and epidemiological information. They are able to track several interactions between a patient and the public healthcare system. In the present study, we make use of data extracted from the administrative data warehouse of Regione Lombardia, a region located in the northern part of Italy whose capital is Milan. Data are within a project aiming at providing a description of the epidemiology of Heart Failure (HF) patients at regional level, to profile health service utilization over time, and to investigate variations in patient care according to geographic area, socio-demographic characteristic and other clinical variables. We use multi-state models to estimate the probability of transition from (re)admission to discharge and death adjusting for covariates which are state dependent. To the best of our knowledge, this is the first Italian attempt of investigating which are the effects of pharmacological and outpatient cares covariates on patient's readmissions and death. This allows to better characterise disease progression and possibly identify what are the main determinants of a hospital admission and death in patients with Heart Failure.

With the impressive growth of available data and the flexibility of network modelling, the problem of devising effective quantitative methods for the comparison of networks arises. Plenty of such methods have been designed to accomplish this task: most of them deal with undirected and unweighted networks only, but a few are capable of handling directed and/or weighted networks too, thus properly exploiting richer information. In this work, we contribute to the effort of comparing the different methods for comparing networks and providing a guide for the selection of an appropriate one. First, we review and classify a collection of network comparison methods, highlighting the criteria they are based on and their advantages and drawbacks. The set includes methods requiring known node-correspondence, such as DeltaCon and Cut Distance, as well as methods not requiring a priori known node-correspondence, such as alignment-based, graphlet-based, and spectral methods, and the recently proposed Portrait Divergence and NetLSD. We test the above methods on synthetic networks and we assess their usability and the meaningfulness of the results they provide. Finally, we apply the methods to two real-world datasets, the European Air Transportation Network and the FAO Trade Network, in order to discuss the results that can be drawn from this type of analysis.

This paper proposes a clustering procedure for samples of multivariate functions in $(L^2(I))^{J}$, with $J\geq1$. This method is based on a k-means algorithm in which the distance between the curves is measured with a metrics that generalizes the Mahalanobis distance in Hilbert spaces, considering the correlation and the variability along all the components of the functional data. The proposed procedure has been studied in simulation and compared with the k-means based on other distances typically adopted for clustering multivariate functional data. In these simulations, it is shown that the k-means algorithm with the generalized Mahalanobis distance provides the best clustering performances, both in terms of mean and standard deviation of the number of misclassified curves. Finally, the proposed method has been applied to two real cases studies, concerning ECG signals and growth curves, where the results obtained in simulation are confirmed and strengthened., 22 pages

Objective: To determine whether out-of-hospital healthcare and adverse outcomes are better in stroke patients admitted to a neurology ward compared with those admitted to general wards. Methods: Beneficiaries of the National Health Service from the Italian Lombardy Region who were discharged alive after hospital admission during the year 2009 for ischemic stroke (9776 patients) or intracerebral or subarachnoid hemorrhage (1102 patients) entered into the cohort and were followed until 2012. Exposure of interest was the ward type where inpatients were admitted (neuro vs. general wards). Outcomes were out-of-hospital healthcare (i.e., drug prescriptions, diagnostic procedures, and laboratory clinical evaluations) and adverse clinical outcomes (i.e., all-cause death and hospital readmission). Exposure-outcome associations were investigated. High-dimensional propensity score methodology was used for taking into account confounders. Mediation analysis was used to verify whether the association between ward type and clinical outcomes is mediated by out-of-hospital adherence to healthcare. Results: Better adherence to out-of-hospital healthcare received from patients discharged from neuro, rather than general, wards was observed being the proportions of adherent patients 42.4% and 39.5%, respectively. Compared with general wards, discharge from neuro was associated with reduced 3-year emergency admissions (from 50.1 to 47.5% among ischemic stroke patients) and reduced 3-year mortality (from 37.5 to 27.0% among hemorrhagic stroke patients). From 10 to 15% of outcome risk, reductions were mediated by better adherence to out-of-hospital healthcare. Conclusions: For patients with acute ischemic and hemorrhagic stroke, admission to neuro vs. general wards is associated with better out-of-hospital healthcare and long-term adverse outcomes.

In recent years, there has been a widespread cross-fertilization between Medical Statistics and Machine Learning (ML) techniques.

We analyse data collected from the administrative datawarehouse of an Italian regional district (Lombardia) concerning patients affected by Chronic Heart Failure. The longitudinal data gathering for each patient hospital readmissions in time, as well as patient-specific covariates, is studied as a realization of non homogeneous Poisson process. Since the aim behind this study is to identify groups of patients behaving similarly in terms of disease progression and then healthcare consumption, we conjectured the time segments between two consecutive hospitalizations to be Weibull distributed in each hidden cluster. Adding a frailty term to take into account the within subjects unknown variability, the corresponding patient-specific hazard functions are reconstructed. Therefore, the comprehensive distribution for each time to event variable is modelled as a Weibull Mixture. We are then able to easily interpret the related hidden groups as healthy, sick, and terminally ill subjects.

Background: Congestive Heart Failure (HF) is a widespread chronic disease characterized by a very high incidence in elder people. The high mortality and readmission rate of HF strongly depends on the complicated morbidity scenario often characterising it. Methods: Data were retrieved from the healthcare administrative datawarehouse of Lombardy, the most populated regional district in Italy. Network analysis techniques and community detection algorithms are applied to comorbidities registered in hospital discharge papers of HF patients, in 7 cohorts between 2006 and 2012.Results: The relevance network indexes applied to the 7 cohorts identified death, ipertension, arrythmia, renal and pulmonary diseases as the most relevant nodes related to HF, in terms of prevalence and closeness/strenght of the relationship. Moreover, 3 clusters of nodes have been identified in all the cohorts, i.e. those related to cancer, lung diseases and heart/circulation related problems.Conclusions: Network analysis can be a useful tool in epidemiologic framework when relational data are the objective of the investigation, since it allows to visualize and make inference on patterns of association among nodes (here HF comorbidities) by means of both qualitative indexes and clustering techniques.

Collagen content quantified through 7-wavelength (635-1060 nm) time domain diffuse optical mammography in 200 women proved key to discriminate malignant from benign breast lesions, to measure breast density, and to estimate breast cancer risk.

Heart failure (HF) is a common, serious chronic condition with high morbidity, hospitalisation and mortality. The healthcare systems of England and the northern Italian region of Lombardy share important similarities and have comprehensive hospital administrative databases linked to the death register. We used them to compare admission for HF and mortality for patients between 2006 and 2012 (n = 37,185 for Lombardy, 234,719 for England) with multistate models. Despite close similarities in age, sex and common comorbidities of the two sets of patients, in Lombardy, HF admissions were longer and more frequent per patient than in England, but short- and medium-term mortality was much lower. English patients had more very short stays, but their very elderly also had longer stays than their Lombardy counterparts. Using a three-state model, the predicted total time spent in hospital showed large differences between the countries: women in England spent an average of 24 days if aged 65 at first admission and 19 days if aged 85; in Lombardy these figures were 68 and 27 days respectively. Eight-state models suggested disease progression that appeared similar in each country. Differences by region within England were modest, with London patients spending more time in hospital and having lower mortality than the rest of England. Whilst clinical practice differences plausibly explain these patterns, we cannot confidently disentangle the impact of alternatives such as coding, casemix, and the availability and use of non-hospital settings. We need to better understand the links between rehospitalisation frequency and mortality.

This paper focuses on the pivotal role that statisticians are challenged to undertake in the Big Data era. Their traditional work of managing variability, complexity, and hidden information is indeed made extremely more complex by the enormous volume of a large variety of data that new technologies are able to provide at high velocity. In detail, the paper briefly discusses few paradigmatic cases of analysis of Big Data in which theoretical, methodological and computational aspects have been fruitfully integrated with specific competences from industry, biology, and finance.

Several techniques are being investigated as a complement to screening mammography, to reduce its false-positive rate, but results are still insufficient to draw conclusions. This initial study explores time domain diffuse optical imaging as an adjunct method to classify non-invasively malignant vs benign breast lesions. We estimated differences in tissue composition (oxy- and deoxyhemoglobin, lipid, water, collagen) and absorption properties between lesion and average healthy tissue in the same breast applying a perturbative approach to optical images collected at 7 red-near infrared wavelengths (635–1060 nm) from subjects bearing breast lesions. The Discrete AdaBoost procedure, a machine-learning algorithm, was then exploited to classify lesions based on optically derived information (either tissue composition or absorption) and risk factors obtained from patient’s anamnesis (age, body mass index, familiarity, parity, use of oral contraceptives, and use of Tamoxifen). Collagen content, in particular, turned out to be the most important parameter for discrimination. Based on the initial results of this study the proposed method deserves further investigation.

Background How different risk profiles of heart failure (HF) patients can influence multiple readmissions and outpatient management is largely unknown. We propose the application of two multi-state models in real world setting to jointly evaluate the impact of different risk factors on multiple hospital admissions, Integrated Home Care (IHC) activations, Intermediate Care Unit (ICU) admissions and death. Methods and findings The first model (model 1) concerns only hospitalizations as possible events and aims at detecting the determinants of repeated hospitalizations. The second model (model 2) considers both hospitalizations and ICU/IHC events and aims at evaluating which profiles are associated with transitions in intermediate care with respect to repeated hospitalizations or death. Both are characterized by transition specific covariates, adjusting for risk factors. We identified 4,904 patients (4,129 de novo and 775 worsening heart failure, WHF) hospitalized for HF from 2009 to 2014. 2,714 (55%) patients died. Advanced age and higher morbidity load increased the rate of dying and of being rehospitalized (model 1), decreased the rate of being discharged from hospital (models 1 and 2) and increased the rate of inactivation of IHC (model 2). WHF was an important risk factor associated with hospital readmission. Conclusion Multi-state models enable a better identification of two patterns of HF patients. Once adjusted for age and comorbidity load, the WHF condition identifies patients who are more likely to be readmitted to hospital, but does not represent an increasing risk factor for activating ICU/IHC. This highlights different ways to manage specific patients' patterns of care. These results provide useful healthcare support to patients' management in real world context. Our study suggests that the epidemiology of the considered clinical characteristics is more nuanced than traditionally presented through a single event.

In chronic diseases like heart failure (HF), the disease course and associated clinical event histories for the patient population vary widely. To improve understanding of the prognosis of patients and enable health care providers to assess and manage resources, we wish to jointly model disease progression, mortality and their relation with patient characteristics. We show how episodes of hospitalisation for disease-related events, obtained from administrative data, can be used as a surrogate for disease status. We propose flexible multi-state models for serial hospital admissions and death in HF patients, that are able to accommodate important features of disease progression, such as multiple ordered events and competing risks. Fully parametric and semi-parametric semi-Markov models are implemented using freely available software in R. The models were applied to a dataset from the administrative data bank of the Lombardia region in Northern Italy, which included 15,298 patients who had a first hospitalisation ending in 2006 and 4 years of follow-up thereafter. This provided estimates of the associations of age and gender with rates of hospital admission and length of stay in hospital, and estimates of the expected total time spent in hospital over five years. For example, older patients and men were readmitted more frequently, though the total time in hospital was roughly constant with age. We also discuss the relative merits of parametric and semi-parametric multi-state models, and model assessment and comparison.

This study was undertaken to evaluate trends in heat failure hospitalizations (HFHs) and 1-year mortality of HFH in Lombardy, the largest Italian region, from 2000 to 2012.Hospital discharge forms with HF-related ICD-9 CM codes collected from 2000 to 2012 by the regional healthcare service (n=699797 in 370538 adult patients), were analyzed with respect to in-hospital and 1-year mortality; Group (G) 1 included most acute HF episodes with primary cardiac diagnosis (70%); G2 included cardiomyopathies without acute HF codes (17%); and G3 included non-cardiac conditions with HF as secondary diagnosis (13%). Patients experiencing their first HFH since 2005 were analyzed as incident cases (n=216782).Annual HFHs number (mean 53830) and in-hospital mortality (9.4%) did not change over the years, the latter being associated with increasing age (p0.0001) and diagnosis Group (G1 9.1%, G2 5.6%, G3 15.9%, p0.0001). Incidence of new cases decreased over the years (3.62 [CI 3.58-3.67] in 2005 to 3.13 [CI 3.09-3.17] in 2012, per 1000 adult inhabitants/year, p0.0001), with an increasing proportion of patients aged ≥85y (22.3% to 31.4%, p0.0001). Mortality lowered over time in75y incident cases, both in-hospital (5.15% to 4.36%, p0.0001) and at 1-year (14.8% to 12.9%, p=0.0006).The overall burden and mortality of HFH appear stable for more than a decade. However, from 2005 to 2012, there was a reduction of new, incident cases, with increasing age at first hospitalization. Meanwhile, both in-hospital and 1-year mortality decreased in patients aged75y, possibly due to improved prevention and treatment.

Background Administrative data are increasingly used in healthcare research. However, in order to avoid biases, their use requires careful study planning. This paper describes the methodological principles and criteria used in a study on epidemiology, outcomes and process of care of patients hospitalized for heart failure (HF) in the largest Italian Region, from 2000 to 2012. Methods Data were extracted from the administrative data warehouse of the healthcare system of Lombardy, Italy. Hospital discharge forms with HF-related diagnosis codes were the basis for identifying HF hospitalizations as clinical events, or episodes. In patients experiencing at least one HF event, hospitalizations for any cause, outpatient services utilization, and drug prescriptions were also analyzed. Results Seven hundred one thousand, seven hundred one heart failure events involving 371,766 patients were recorded from 2000 to 2012. Once all the healthcare services provided to these patients after the first HF event had been joined together, the study database totalled about 91 million records. Principles, criteria and tips utilized in order to minimize errors and characterize some relevant subgroups are described. Conclusions The methodology of this study could represent the basis for future research and could be applied in similar studies concerning epidemiology, trend analysis, and healthcare resources utilization.

Top 15% age-matched collagen content (estimated on 109 subjects by optical mammography) and high mammographic density identify different subgroups of high cancer occurrence, suggesting collagen is an independent breast cancer risk factor.

Studies of variations in healthcare utilization and outcome involve the analysis of multilevel,clustered data, considering in particular the estimation of a cluster-specific adjusted response,covariate effects and components of variance. Besides reporting on the extent of observedvariations, these studies quantify the role of contributing factors including patients’ andproviders’ characteristics. In addition, they may assess the relationship between healthcareprocess and outcomes. We consider Bayesian generalized linear mixed models to analyzeMOMI2 (Month MOnitoring Myocardial Infarction in MIlan) data on patients admittedwith ST-elevation myocardial infarction (STEMI) diagnosis in the hospitals belonging to theMilano Cardiological Network. Both clinical registries and administrative databanks wereused to predict survival probabilities. We fit a logit model for the survival probability withone random effect (the hospital), under a semiparametric prior. We take advantage of thein-built clustering property of the Dirichlet process prior assumed for the random-effectsparameters to obtain a classification of providers.

Since 2001, the urban area of Milan has been operating a network among 23 cardiac care units, the 118 dispatch centre (national free number for medical emergencies), and the county government health agency called Group for Prehospital Cardiac Emergency.In order to monitor the network activity, time to treatment, and clinical outcome, a periodic survey, called MOMI(2), was repeated two or three times a year. Each survey lasted 30 days and was repeated in comparable periods. Data were stratified for hospital admission mode. We collected data concerning 708 consecutive ST-elevation myocardial infarction (STEMI) patients (male 72.6%; mean age 64.4 years). In these six surveys, we observed a high rate of primary percutaneous coronary intervention (73.2%) and a mortality rate of 6.3%. Using advanced statistical models, we identified age, Killip class, and the symptom onset-to-balloon time as most relevant prognostic factors. Nonparametric test showed that the modality of hospital admittance was the most critical determinant of door-to-balloon time. 12-lead ECG tele-transmission and activation of a fast track directly to the catheterization laboratory are easy action to reduce time to treatment.The experience of the Milan network for cardiac emergency shows how a network coordinating the community, rescue units, and hospitals in a complex urban area and making use of medical technology contributes to the health care of patients with STEMI.

In this work we propose the use of a graphical diagnostic tool (the funnel plot) to detect outliers among hospitals that treat patients affected by Acute Myocardial Infarction (AMI). We consider an application to data on AMI hospitalizations recorded in the administrative databases of our regional district. The outcome of interest is the in-hospital mortality, a variable indicating if the patient has been discharged dead or alive. We then compare the results obtained by graphical diagnostic tools with those arising from fitting parametric mixed effects models to the same data.

We propose a new algorithm, based on a linear regression model, to statistically estimate the hemodynamic activations in fNIRS data sets. The main concern guiding the algorithm development was the minimization of assumptions and approximations made on the data set for the application of statistical tests. Further, we propose a K-means method to cluster fNIRS data (i.e. channels) as activated or not activated. The methods were validated both on simulated and in vivo fNIRS data. A time domain (TD) fNIRS technique was preferred because of its high performances in discriminating cortical activation and superficial physiological changes. However, the proposed method is also applicable to continuous wave or frequency domain fNIRS data sets. (C) 2015 Optical Society of America

Clinical registries are established as tools for auditing clinical standards and benchmarking quality improvement initiatives. They also have an emerging role (as electronic health records) in cardiovascular research and, in particular, the conduct of RCTs. While the RCT is accepted as the most robust experimental design, observational data from clinical registries has become increasingly valuable for RCTs. Data from clinical registries may be used to augment results from RCTs, identify patients for recruitment and as an alternative when randomization is not practically possible or ethically desirable. Here the authors appraise the advantages and disadvantages of both methodologies, with the aim of clarifying when their joint use may be successful.

In this work, a Bayesian semiparametric multivariate model is fitted to study data related to in-hospital and 60-day survival probabilities of patients admitted to a hospital with ST-elevation myocardial infarction diagnosis. We consider a hierarchical generalized linear model to predict survival probabilities and a process indicator (time of intervention). Poisson-Dirichlet process priors, generalizing the well-known Dirichlet process, are considered for modeling the random-effect distribution of the grouping factor which is the hospital of admission.

In this work we propose a novel EM method for the estimation of nonlinear nonparametric mixed-effects models, aimed at unsupervised classification. We perform simulation studies in order to evaluate the algorithm performance and we apply this new procedure to a real dataset.

Direzione Generale Sanita, Regione Lombardia, funded in January 2009 the Strategic Program “Exploitation, integration and study of current and future health databases in Lombardia for Acute Myocardial Infarction.” The aim of this project was the integration and the statistical analysis of complex clinical and administrative databases. In this context the STEMI Archive was created in collaboration with Politecnico di Milano and several hospital structures. This archive contains clinical data about patients affected by acute myocardial infarction with ST elevation (STEMI) admitted in any hospital of Regione Lombardia. In this chapter we will discuss the representativeness of the sample of patients registered in the archive compared with the overall population of STEMI patients affected by acute myocardial infarction, extracted from the administrative database of the Regione Lombardia.

A time-domain multiwavelength (635 to 1060 nm) optical mammography was performed on 147 subjects with recent X-ray mammograms available, and average breast tissue composition (water, lipid, collagen, oxy-and deoxyhemoglobin) and scattering parameters (amplitude a and slope b) were estimated. Correlation was observed between optically derived parameters and mammographic density [Breast Imaging and Reporting Data System (BI-RADS) categories], which is a strong risk factor for breast cancer. A regression logistic model was obtained to best identify high-risk (BI-RADS 4) subjects, based on collagen content and scattering parameters. The model presents a total misclassification error of 12.3%, sensitivity of 69%, specificity of 94%, and simple kappa of 0.84, which compares favorably even with intraradiologist assignments of BI-RADS categories. © The Authors.

In this work we describe design, aims, and contents of the ST-segment Elevation Myocardial Infarction (STEMI) Archive, which is a multicenter observational clinical registry planned within the Strategic Program “Exploitation, integration and study of current and future health databases in Lombardia for Acute Myocardial Infarction.” This is an observational clinical registry that collects clinical indicators, process indicators, and outcomes concerning STEMI patients admitted to any hospital of the regional district, one of the most advanced and intensive-care area in Italy. This registry is arranged to be automatically linked to the Public Health Database, the ongoing administrative datawarehouse of Regione Lombardia. Aims and perspectives of this innovative project are discussed, together with feasibility and statistical analyses which are to be performed on it, in order to monitor and evaluate the patterns of care of cardiovascular patients.

In this work, we develop Bayes rules for several families of loss functions for hospital report cards under a Bayesian semiparametric hierarchical model. Moreover, we present some robustness analysis with respect to the choice of the loss function, focusing on the number of hospitals our procedure identifies as "unacceptably performing". The analysis is carried out on a case study dataset arising from MOMI2 (Month MOnitoring Myocardial Infarction in MIlan) survey on patients admitted with ST-Elevation Myocardial Infarction to the hospitals of Milan Cardiological Network. The major aim of this work is the ranking of the health-care providers performances, together with the assessment of the role of patients' and providers' characteristics on survival outcome. © Springer International Publishing Switzerland 2013.

In this work we propose a novel unsupervised classification technique based on the estimation of nonlinear nonparametric mixed-effects models. The proposed method is an iterative algorithm that alternates a nonparametric EM step and a nonlinear Maximum Likelihood step. We apply this new procedure to perform an unsupervised clustering of longitudinal data in two different case studies.

Provider profiling is the process of evaluation of the performance of hospitals, doctors and other medical practitioners in order to increase the quality of medical care. This paper reports statistical analyses carried out on data arising from a regular survey concerning patients admitted with an ST-elevation myocardial infarction diagnosis in one of a number of hospitals in the Milan area. The main aim is to determine process indicators to be used in health-care evaluation. Effective statistical support for clinical and organizational governance is obtained by analysing and modelling data from clinical registries. A grouping structure and a consequent ranking of hospitals is investigated, taking into account the fact that this kind of survey data are intrinsically grouped at first level by where patients are hospitalized. We compare three different techniques for hospital classification based, respectively, on: (a) traditional comparison of survival rates; (b) analysis of variance components in fitted generalized linear mixed effects models; and (c) non-parametric random effects estimation.