was read the article
array:23 [ "pii" => "S1665268124003223" "issn" => "16652681" "doi" => "10.1016/j.aohep.2024.101528" "estado" => "S300" "fechaPublicacion" => "2024-09-01" "aid" => "101528" "copyright" => "Fundación Clínica Médica Sur, A.C." "copyrightAnyo" => "2024" "documento" => "article" "crossmark" => 1 "subdocumento" => "fla" "cita" => "Ann Hepatol. 2024;29:" "abierto" => array:3 [ "ES" => true "ES2" => true "LATM" => true ] "gratuito" => true "lecturas" => array:1 [ "total" => 0 ] "itemSiguiente" => array:18 [ "pii" => "S1665268124003132" "issn" => "16652681" "doi" => "10.1016/j.aohep.2024.101519" "estado" => "S300" "fechaPublicacion" => "2024-09-01" "aid" => "101519" "copyright" => "Fundación Clínica Médica Sur, A.C." "documento" => "article" "crossmark" => 1 "subdocumento" => "fla" "cita" => "Ann Hepatol. 2024;29:" "abierto" => array:3 [ "ES" => true "ES2" => true "LATM" => true ] "gratuito" => true "lecturas" => array:1 [ "total" => 0 ] "en" => array:11 [ "idiomaDefecto" => true "cabecera" => "<span class="elsevierStyleTextfn">Original article</span>" "titulo" => "Efficacy of blood plasma spectroscopy for early liver cancer diagnostics in obese patients" "tienePdf" => "en" "tieneTextoCompleto" => "en" "tieneResumen" => "en" "contieneResumen" => array:1 [ "en" => true ] "contieneTextoCompleto" => array:1 [ "en" => true ] "contienePdf" => array:1 [ "en" => true ] "resumenGrafico" => array:2 [ "original" => 0 "multimedia" => array:8 [ "identificador" => "fig0003" "etiqueta" => "Fig. 3" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => true "mostrarDisplay" => false "figura" => array:1 [ 0 => array:4 [ "imagen" => "gr3.jpeg" "Alto" => 1321 "Ancho" => 1667 "Tamanyo" => 127140 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0006" "detalle" => "Fig " "rol" => "short" ] ] "descripcion" => array:1 [ "en" => "<p id="spara003" class="elsevierStyleSimplePara elsevierViewall">Average electronic circular dichroism spectra of blood plasma samples of patients with HCC (red) and controls (blue). The shaded areas represent the standard deviation of intensity, and the marked spectral variables were used to build the optimized model.</p>" ] ] ] "autores" => array:1 [ 0 => array:2 [ "autoresLista" => "Petr Hříbek, Ondřej Vrtělka, Kateřina Králová, Johana Klasová, Markéta Fousková, Lucie Habartová, Kristýna Kubíčková, Tomáš Kupsa, Tomáš Tůma, Vladimír Setnička, Petr Urbánek" "autores" => array:11 [ 0 => array:2 [ "nombre" => "Petr" "apellidos" => "Hříbek" ] 1 => array:2 [ "nombre" => "Ondřej" "apellidos" => "Vrtělka" ] 2 => array:2 [ "nombre" => "Kateřina" "apellidos" => "Králová" ] 3 => array:2 [ "nombre" => "Johana" "apellidos" => "Klasová" ] 4 => array:2 [ "nombre" => "Markéta" "apellidos" => "Fousková" ] 5 => array:2 [ "nombre" => "Lucie" "apellidos" => "Habartová" ] 6 => array:2 [ "nombre" => "Kristýna" "apellidos" => "Kubíčková" ] 7 => array:2 [ "nombre" => "Tomáš" "apellidos" => "Kupsa" ] 8 => array:2 [ "nombre" => "Tomáš" "apellidos" => "Tůma" ] 9 => array:2 [ "nombre" => "Vladimír" "apellidos" => "Setnička" ] 10 => array:2 [ "nombre" => "Petr" "apellidos" => "Urbánek" ] ] ] ] ] "idiomaDefecto" => "en" "EPUB" => "https://multimedia.elsevier.es/PublicationsMultimediaV1/item/epub/S1665268124003132?idApp=UINPBA00004N" "url" => "/16652681/0000002900000005/v1_202408300521/S1665268124003132/v1_202408300521/en/main.assets" ] "itemAnterior" => array:18 [ "pii" => "S1665268124003107" "issn" => "16652681" "doi" => "10.1016/j.aohep.2024.101516" "estado" => "S300" "fechaPublicacion" => "2024-09-01" "aid" => "101516" "copyright" => "Fundación Clínica Médica Sur, A.C." "documento" => "article" "crossmark" => 1 "subdocumento" => "fla" "cita" => "Ann Hepatol. 2024;29:" "abierto" => array:3 [ "ES" => true "ES2" => true "LATM" => true ] "gratuito" => true "lecturas" => array:1 [ "total" => 0 ] "en" => array:12 [ "idiomaDefecto" => true "cabecera" => "<span class="elsevierStyleTextfn">Original article</span>" "titulo" => "The steatosis-associated fibrosis estimator (SAFE) outperformed the FIB-4 score in screening the population for liver disease" "tienePdf" => "en" "tieneTextoCompleto" => "en" "tieneResumen" => array:2 [ 0 => "en" 1 => "en" ] "contieneResumen" => array:1 [ "en" => true ] "contieneTextoCompleto" => array:1 [ "en" => true ] "contienePdf" => array:1 [ "en" => true ] "resumenGrafico" => array:2 [ "original" => 1 "multimedia" => array:6 [ "identificador" => "fig0005" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => false "mostrarDisplay" => true "figura" => array:1 [ 0 => array:4 [ "imagen" => "ga1.jpeg" "Alto" => 885 "Ancho" => 2213 "Tamanyo" => 298397 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alttx001" "detalle" => "Image, graphical abstrac" "rol" => "short" ] ] ] ] "autores" => array:1 [ 0 => array:2 [ "autoresLista" => "Mingkai Li, Ying Lin, Hongsheng Yu, Weichun Lin, Jianning Chen, Yidong Yang, Bin Wu" "autores" => array:7 [ 0 => array:2 [ "nombre" => "Mingkai" "apellidos" => "Li" ] 1 => array:2 [ "nombre" => "Ying" "apellidos" => "Lin" ] 2 => array:2 [ "nombre" => "Hongsheng" "apellidos" => "Yu" ] 3 => array:2 [ "nombre" => "Weichun" "apellidos" => "Lin" ] 4 => array:2 [ "nombre" => "Jianning" "apellidos" => "Chen" ] 5 => array:2 [ "nombre" => "Yidong" "apellidos" => "Yang" ] 6 => array:2 [ "nombre" => "Bin" "apellidos" => "Wu" ] ] ] ] "resumen" => array:1 [ 0 => array:3 [ "titulo" => "Graphical abstract" "clase" => "graphical" "resumen" => "<span id="abss0005" class="elsevierStyleSection elsevierViewall"><p id="spara028" class="elsevierStyleSimplePara elsevierViewall"><elsevierMultimedia ident="fig0005"></elsevierMultimedia></p></span>" ] ] ] "idiomaDefecto" => "en" "EPUB" => "https://multimedia.elsevier.es/PublicationsMultimediaV1/item/epub/S1665268124003107?idApp=UINPBA00004N" "url" => "/16652681/0000002900000005/v1_202408300521/S1665268124003107/v1_202408300521/en/main.assets" ] "en" => array:18 [ "idiomaDefecto" => true "cabecera" => "<span class="elsevierStyleTextfn">Original article</span>" "titulo" => "Accurate prediction of all-cause mortality in patients with metabolic dysfunction-associated steatotic liver disease using electronic health records" "tieneTextoCompleto" => true "autores" => array:1 [ 0 => array:4 [ "autoresLista" => "Ignat Drozdov, Benjamin Szubert, Ian A. Rowe, Timothy J. Kendall, Jonathan A. Fallowfield" "autores" => array:5 [ 0 => array:3 [ "nombre" => "Ignat" "apellidos" => "Drozdov" "referencia" => array:1 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">a</span>" "identificador" => "aff0001" ] ] ] 1 => array:3 [ "nombre" => "Benjamin" "apellidos" => "Szubert" "referencia" => array:1 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">a</span>" "identificador" => "aff0001" ] ] ] 2 => array:3 [ "nombre" => "Ian A." "apellidos" => "Rowe" "referencia" => array:2 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">b</span>" "identificador" => "aff0002" ] 1 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">c</span>" "identificador" => "aff0003" ] ] ] 3 => array:3 [ "nombre" => "Timothy J." "apellidos" => "Kendall" "referencia" => array:2 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">d</span>" "identificador" => "aff0004" ] 1 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">e</span>" "identificador" => "aff0005" ] ] ] 4 => array:4 [ "nombre" => "Jonathan A." "apellidos" => "Fallowfield" "email" => array:1 [ 0 => "Jonathan.Fallowfield@ed.ac.uk" ] "referencia" => array:2 [ 0 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">e</span>" "identificador" => "aff0005" ] 1 => array:2 [ "etiqueta" => "<span class="elsevierStyleSup">*</span>" "identificador" => "cor0001" ] ] ] ] "afiliaciones" => array:5 [ 0 => array:3 [ "entidad" => "Bering Limited, London, UK" "etiqueta" => "a" "identificador" => "aff0001" ] 1 => array:3 [ "entidad" => "Leeds Institute of Medical Research, University of Leeds, UK" "etiqueta" => "b" "identificador" => "aff0002" ] 2 => array:3 [ "entidad" => "Leeds Liver Unit, St James's University Hospital, Leeds Teaching Hospitals, UK" "etiqueta" => "c" "identificador" => "aff0003" ] 3 => array:3 [ "entidad" => "Edinburgh Pathology, University of Edinburgh, Edinburgh, UK" "etiqueta" => "d" "identificador" => "aff0004" ] 4 => array:3 [ "entidad" => "Centre for Inflammation Research, Institute for Regeneration and Repair, University of Edinburgh, Edinburgh, UK" "etiqueta" => "e" "identificador" => "aff0005" ] ] "correspondencia" => array:1 [ 0 => array:3 [ "identificador" => "cor0001" "etiqueta" => "⁎" "correspondencia" => "Corresponding author." ] ] ] ] "resumenGrafico" => array:2 [ "original" => 0 "multimedia" => array:8 [ "identificador" => "fig0001" "etiqueta" => "Fig. 1" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => true "mostrarDisplay" => false "figura" => array:1 [ 0 => array:4 [ "imagen" => "gr1.jpeg" "Alto" => 2748 "Ancho" => 3167 "Tamanyo" => 378771 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0005" "detalle" => "Fig " "rol" => "short" ] ] "descripcion" => array:1 [ "en" => "<p id="spara001" class="elsevierStyleSimplePara elsevierViewall"><span class="elsevierStyleBold">Schematic representation of an EHR vector.</span> The patient's timeline is represented by horizontal arrows and each data point is depicted by colour-coded tokens. Predictive models were trained on the data in the Observation Window (10 years), whilst a binary outcome of all-cause mortality was used as a ground truth. A&E, emergency department; U&Es, urea and electrolytes; FBC, full blood count, LFT, liver function tests; ACE inhibitor, angiotensin-converting enzyme inhibitor; BMI, body mass index.</p>" ] ] ] "textoCompleto" => "<span class="elsevierStyleSections"><span id="sec0001" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">1</span><span class="elsevierStyleSectionTitle" id="cesectitle0008">Introduction</span><p id="para0005" class="elsevierStylePara elsevierViewall">Metabolic dysfunction-associated steatotic liver disease (MASLD), previously termed nonalcoholic fatty liver disease (NAFLD) [<a class="elsevierStyleCrossRef" href="#bib0001">1</a>,<a class="elsevierStyleCrossRef" href="#bib0002">2</a>], is defined as the presence of hepatic steatosis in conjunction with at least one cardiometabolic risk factor (obesity, hypertension, type 2 diabetes, dyslipidemia) but no discernible secondary causes, including substantial alcohol intake, medications known to cause steatosis, or inherited metabolic conditions [<a class="elsevierStyleCrossRef" href="#bib0003">3</a>,<a class="elsevierStyleCrossRef" href="#bib0004">4</a>]. Around 25 % of people with hepatic steatosis progress to metabolic dysfunction-associated steatohepatitis (MASH), which is characterized by hepatocellular ballooning and lobular necroinflammation, and an increased risk of fibrosis, cirrhosis, hepatic decompensation, hepatocellular carcinoma (HCC), and all-cause mortality [<a class="elsevierStyleCrossRef" href="#bib0003">3</a>]. The global burden of MASLD is increasing at an alarming rate [<a class="elsevierStyleCrossRef" href="#bib0005">5</a>], with a worldwide prevalence of up to 32.4 % [<a class="elsevierStyleCrossRef" href="#bib0006">6</a>]. The health economic impacts of MASLD are considerable, including annual direct medical costs of $103 billion ($1613 per patient) in the United States (US) and €35 billion (€1163 per patient) in the Europe-4 countries (Germany, France, Italy, and the United Kingdom) [<a class="elsevierStyleCrossRef" href="#bib0007">7</a>].</p><p id="para0006" class="elsevierStylePara elsevierViewall">Although the past four decades have witnessed major advances in the biological understanding of MASLD and the mechanisms driving the pathogenesis of cirrhosis and HCC, there has been little translation to improved clinical outcomes [<a class="elsevierStyleCrossRef" href="#bib0008">8</a>]. This can be explained, in part, by the lack of approved therapy [<a class="elsevierStyleCrossRef" href="#bib0009">9</a>] and the heterogeneous natural history of the MASLD, which includes extra-hepatic manifestations such as cardiovascular disease (CVD) and chronic kidney disease (CKD) that can further increase disease burden and contribute to the risk of all-cause mortality [<a class="elsevierStyleCrossRef" href="#bib0010">10</a>]. Indeed, in many series, the leading cause of death in individuals with MASLD is CVD, followed by extra-hepatic cancers, and then liver-related mortality [<a class="elsevierStyleCrossRef" href="#bib0011">11</a>], highlighting the importance of a holistic approach to the management of this patient population. Moreover, there is an urgent need to embed effective strategies within primary care, as well as hepatology services, to enable early risk stratification and evidence-based treatment initiation to curtail MASLD-associated morbidity and mortality.</p><p id="para0007" class="elsevierStylePara elsevierViewall">The increased availability of electronic health records (EHRs) opens new opportunities to develop predictive case-finding algorithms that facilitate effective MASLD surveillance [<a class="elsevierStyleCrossRef" href="#bib0012">12</a>]. EHR adoption has reached near-universal levels in the US and the European Union in both acute care hospitals and primary care [<a class="elsevierStyleCrossRef" href="#bib0013">13</a>]. Moreover, studies have shown the potential utility of applying artificial intelligence (AI) and machine learning (ML) algorithms to EHR data to improve the early detection, diagnosis, and management of many conditions, particularly cardiometabolic diseases [<a class="elsevierStyleCrossRef" href="#bib0014">14</a>,<a class="elsevierStyleCrossRef" href="#bib0015">15</a>]. However, despite the proliferation of machine-readable datasets, the development and scaling of predictive models have been limited. The complexities of real-world clinical data, replete with thousands of potential predictor variables and missing values, are seen as the key barriers to implementation [<a class="elsevierStyleCrossRefs" href="#bib0016">16-18</a>]. Deep neural networks (DNNs) have emerged as robust tools with applications to sequence prediction within mixed modality data sets [<a class="elsevierStyleCrossRefs" href="#bib0017">17-20</a>]. The key advantages of DNN methods are their ability to handle large volumes of relatively noisy data, including errors in labels, as well as large numbers of input variables [<a class="elsevierStyleCrossRef" href="#bib0018">18</a>].</p><p id="para0008" class="elsevierStylePara elsevierViewall">Because liver-related outcomes in MASLD are strongly associated with the severity of liver fibrosis [<a class="elsevierStyleCrossRef" href="#bib0021">21</a>,<a class="elsevierStyleCrossRef" href="#bib0022">22</a>], existing risk stratification is anchored to the histological stage or non-invasive assessment of fibrosis using surrogate markers [<a class="elsevierStyleCrossRefs" href="#bib0023">23-25</a>], but such approaches may not reflect the complexity and multimorbidity of MASLD. Furthermore, despite the increasing application of AI and ML tools to MASLD disease management, predictive analysis has largely focused on diagnosis and screening [<a class="elsevierStyleCrossRef" href="#bib0026">26</a>,<a class="elsevierStyleCrossRef" href="#bib0027">27</a>], as well as disease quantification [<a class="elsevierStyleCrossRef" href="#bib0028">28</a>,<a class="elsevierStyleCrossRef" href="#bib0029">29</a>]. Therefore, there is a critical need for a reliable and accessible risk stratification approach for broad clinical outcomes of interest, such as all-cause mortality, to enable early/proactive community interventions such as lifestyle adjustments and future care planning.</p><p id="para0009" class="elsevierStylePara elsevierViewall">In this work, we test the hypothesis that a simple Transformer neural network, trained on routinely collected in-patient and out-patient data from people with MASLD, can be used effectively to predict individuals at an increased risk of all-cause mortality.</p></span><span id="sec0002" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2</span><span class="elsevierStyleSectionTitle" id="cesectitle0009">Patients and Methods</span><span id="sec0003" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2.1</span><span class="elsevierStyleSectionTitle" id="cesectitle0010">Study populations</span><p id="para0010" class="elsevierStylePara elsevierViewall">The SteatoSITE retrospective dataset was drawn from a population representing 12 of the 14 territorial Health Boards in Scotland and consists of <span class="elsevierStyleItalic">n</span> = 940 histologically defined patients (55.4 % men and 44.6 % women; median body mass index 31.3; 32 % with type 2 diabetes) covering the complete MASLD severity spectrum. Detailed characteristics of the SteatoSITE cohort have been published [<a class="elsevierStyleCrossRef" href="#bib0030">30</a>]. For the validation study, a nested case-control design was used on an independent non-biopsy MASLD patient population (<span class="elsevierStyleItalic">n</span> = 528 patients) from NHS Greater Glasgow and Clyde (GG&C) collected between 2002 and 2021.</p><p id="para0011" class="elsevierStylePara elsevierViewall">For SteatoSITE, cases with a liver tissue sample acquired between January 2000 and October 2019 and a histological diagnosis of NAFLD (MASLD) were included. The other inclusion/exclusion criteria were: men or women; >18 years of age at the time of tissue sampling; all ethnic groups, socio-economic backgrounds, and health status; dead or alive at the time of inclusion into data commons; no documented history of chronic liver disease of any non-MASLD etiology, including alcohol-related liver disease, chronic viral hepatitis, hemochromatosis, Wilson disease, autoimmune hepatitis, primary biliary cholangitis, primary sclerosing cholangitis; and patients with excessive alcohol use documented within the clinical data supplied on the specimen request form (>21 units/week for men, >14 units/week for women); or histological features suggesting a secondary non-MASLD diagnosis.</p><p id="para0012" class="elsevierStylePara elsevierViewall">The SteatoSITE dataset was used for model training and in-sample validation using five-fold cross-validation. During each cross-validation run, the dataset was partitioned into training, validation, and testing subsets such that the distribution of age, gender, and outcomes were stratified across each partition. To avoid data leakage across data partitions, we ensured that there were no overlapping patient identifiers.</p><p id="para0013" class="elsevierStylePara elsevierViewall">The NHS GG&C dataset comprised of patient EHRs obtained between 2000 and 2019 and followed similar inclusion and exclusion criteria to SteatoSITE, although the clinical diagnosis was based on the International Classification of Diseases – Tenth Revision (ICD-10) codes (K76.0 [NAFLD, all] and K75.8 NASH). The NHS GG&C dataset was used for out-of-sample model validation.</p><p id="para0014" class="elsevierStylePara elsevierViewall">Notably, ICD diagnostic coding for inpatient and outpatient episodes and procedures (OPCS Classification of Interventions and Procedures (OPCS-4)) for both study populations followed recent expert consensus guidelines for using administrative coding in EHR-based research of MASLD [<a class="elsevierStyleCrossRef" href="#bib0031">31</a>].</p><p id="para0015" class="elsevierStylePara elsevierViewall">We categorized the cause of death based on the methods used by Simon <span class="elsevierStyleItalic">et al</span>. [<a class="elsevierStyleCrossRef" href="#bib0032">32</a>]. The following ICD-10 and OPCS-4 codes were used for the cause-specific categories: HCC ('C22′, 'C220′, 'C229′, 'C2299′); cirrhosis (Y830, T864, K74, K72, K767, I8* (includes other decompensation causes and post-transplant complications)); non-HCC cancer (any C code apart from those for HCC); cardiovascular disease (any I code apart from I8* (varices)); other (none of the above). Additionally, we filtered sequentially down the hierarchy of cause of death information and used the first non-other code that appeared.</p></span><span id="sec0004" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2.2</span><span class="elsevierStyleSectionTitle" id="cesectitle0011">Data representation and ground truthing</span><p id="para0016" class="elsevierStylePara elsevierViewall">Each patient's longitudinal EHR vector was split into an Observation and Prediction Window (<a class="elsevierStyleCrossRef" href="#fig0001">Fig. 1</a>). The Index Date for case patients was calculated as the date 12, 24, or 36 months before the patient's date of death. The Index Date for controls was calculated as the date 12, 24, or 36 months before the last EHR entry. The Observation Window comprised all EHR vectors during a ten-year period in the run-up to the Index Date. Only data in the Observation Window was used to represent the patient during model training, validation, and testing.</p><elsevierMultimedia ident="fig0001"></elsevierMultimedia><p id="para0017" class="elsevierStylePara elsevierViewall">Patient features used in predictive modelling are shown in <a class="elsevierStyleCrossRef" href="#tbl0001">Table 1</a>. For each patient, two feature vector representations were generated. The first representation consisted of static features – age, gender, and ethnicity. The second representation reflected dynamic features associated with inpatient and outpatient activity over a five-year period of the Observation Window. This temporal input vector was discretized into twelve exponentially increasing time bins, such that the most recent time points were assigned to the shortest time bin. If a feature (e.g., ICD-10 codes) within an Observation Window contained multiple values, the most frequent value was retained. In cases where a numerical feature (e.g., BMI) contained multiple entries within one observation window, an average was calculated. Missing values were filled by forward propagation.</p><elsevierMultimedia ident="tbl0001"></elsevierMultimedia><p id="para0018" class="elsevierStylePara elsevierViewall">Numerical data was scaled to a range between 0 and 1, whilst categorical data was represented as 32-dimensional vectors of a large pre-trained language model trained on <span class="elsevierStyleItalic">n</span> = 2067,531 full text PubMed articles totalling <span class="elsevierStyleItalic">n</span> = 224,427,218 sentences [<a class="elsevierStyleCrossRef" href="#bib0033">33</a>,<a class="elsevierStyleCrossRef" href="#bib0034">34</a>].</p></span><span id="sec0005" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2.3</span><span class="elsevierStyleSectionTitle" id="cesectitle0012">Model training</span><p id="para0019" class="elsevierStylePara elsevierViewall">Given significant variation in length and density of patient records (e.g., vital sign measurements in an intensive care unit vs. outpatient clinic), we formulated a simple Transformer architecture with multi-head attention [<a class="elsevierStyleCrossRef" href="#bib0035">35</a>], to take advantage of such data.</p><p id="para0020" class="elsevierStylePara elsevierViewall">Input layers of the Transformer network were adjusted to concurrently use time-invariant and time-dependent features. Multiple inputs were concatenated along a horizontal axis and passed to four transformer encoder blocks with multi-head attention. Four attention heads were used with head size fixed at 256. The classification head of the network consisted of a global average pooling layer, followed by a dense layer with rectified linear unit [<a class="elsevierStyleCrossRef" href="#bib0036">36</a>] activation and a dropout layer. A softmax activation function was applied to the final dense layer.</p><p id="para0021" class="elsevierStylePara elsevierViewall">The number of neurons in the penultimate dense layer and the dropout rate were tuneable hyperparameters optimized during training using the Hyperband algorithm [<a class="elsevierStyleCrossRef" href="#bib0037">37</a>], with the best set of parameters corresponding to the lowest sparse categorical cross-entropy loss on the validation set. The number of neurons was selected from the range of [32, 512], and the dropout rate took values from the range [0, 0.2].</p><p id="para0022" class="elsevierStylePara elsevierViewall">Training was performed with batch size of 512 using an Adam optimizer with a learning rate of 1  ×  10<span class="elsevierStyleSup">−4</span> while minimizing the categorical cross-entropy loss.</p><p id="para0023" class="elsevierStylePara elsevierViewall">The network was trained to output probabilities of mortality following 12-, 24-, and 36-month prediction windows. Training was terminated early if validation loss did not improve after ten consecutive epochs.</p></span><span id="sec0006" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2.4</span><span class="elsevierStyleSectionTitle" id="cesectitle0013">Statistical analysis</span><p id="para0024" class="elsevierStylePara elsevierViewall">Model performance was assessed using the area under the receiver operating characteristic (AUROC) curve, overall accuracy, sensitivity, specificity, and positive predictive value (PPV). For AUROC measures, 95 % confidence intervals (CIs) were calculated empirically using 2000 bootstrap samples. CIs for sensitivity, specificity, and positive predictive value are exact Clopper-Pearson CIs. Patient demographics were compared across the training/validation, internal testing, and clinical evaluation sets using ANOVA for continuous variables and Chi-square for categorical variables. <span class="elsevierStyleItalic">p</span>-values < 0.05 were considered as statistically significant.</p><p id="para0025" class="elsevierStylePara elsevierViewall">Dimensionality reduction was performed using the Ivis algorithm [<a class="elsevierStyleCrossRef" href="#bib0038">38</a>]. Briefly, prior to analysis, categorical variables were one-hot encoded, whilst numerical variables were scaled to values between 0 and 1. The dataset was reduced to two components using the ‘maaten’ twin neural network architecture and default Ivis hyperparameter values. To identify the salient features captured by the Transformer model, we calculated the coefficient of determination (R<span class="elsevierStyleSup">2</span>) between low-dimensional representations of the model global average pooling layer and training set features. Where categorical features were used, their numerical representation was extracted from the model's feature embedding layer.</p><p id="para0026" class="elsevierStylePara elsevierViewall">Model probabilities were evaluated using the reliability diagram [<a class="elsevierStyleCrossRef" href="#bib0039">39</a>] and the calibration_curve function in the scikit-learn library [<a class="elsevierStyleCrossRef" href="#bib0040">40</a>]. Predicted probabilities were binned into ten discrete intervals, and the mean predicted probability and the true frequency of the positive class were plotted for each interval.</p><p id="para0027" class="elsevierStylePara elsevierViewall">All statistical tests were carried out using the SciPy module (version 1.7.3) for Python (version 3.9.14).</p></span><span id="sec0007" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">2.5</span><span class="elsevierStyleSectionTitle" id="cesectitle0014">Ethical statement</span><p id="para0028" class="elsevierStylePara elsevierViewall">Unified transparent approval for unconsented data inclusion in the multimodal pan-Scotland SteatoSITE database [<a class="elsevierStyleCrossRef" href="#bib0030">30</a>] was provided by the West of Scotland Research Ethics Committee 4 (Reference: 20/WS/0002; 18th February 2020), Public Benefit and Privacy Panel for Health and Social Care (PBPP; Reference: 1819–0091; 4th June 2021), Institutional Research & Development departments and Caldicott Guardians. Delegated research and ethics approvals for the validation cohort study were granted by the Local and Advisory Committee at NHS Greater Glasgow and Clyde (NHS GG&C). The cohort and de-identified linked data were prepared by the West of Scotland Safe Haven at NHS GG&C. In Scotland, patient consent is not required where routinely collected patient data are used for research purposes through an approved Safe Haven. For that reason, informed consent was not required and was not sought. All research was conducted following both the Declarations of Helsinki (2013) and Istanbul (2018), and Good Clinical Practice principles. This study was conducted and reported in accordance with the TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prediction or Diagnosis) guidelines.</p></span></span><span id="sec0008" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">3</span><span class="elsevierStyleSectionTitle" id="cesectitle0015">Results</span><span id="sec0009" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">3.1</span><span class="elsevierStyleSectionTitle" id="cesectitle0016">Training and testing dataset characteristics</span><p id="para0029" class="elsevierStylePara elsevierViewall">Demographic and phenotypic characteristics of the training and testing cohorts are shown in <a class="elsevierStyleCrossRef" href="#tbl0002">Table 2</a>. Patient age in the training set was significantly younger than the testing set (two-tailed unpaired <span class="elsevierStyleItalic">t-</span>test, <span class="elsevierStyleItalic">p</span> = 0.0001), whilst there were no significant differences in BMI (two-tailed unpaired <span class="elsevierStyleItalic">t</span>-test, <span class="elsevierStyleItalic">p</span> = 0.76) or the frequencies of gender or ethnicity distributions (chi-square, <span class="elsevierStyleItalic">p</span> = 0.12–0.16).</p><elsevierMultimedia ident="tbl0002"></elsevierMultimedia><p id="para0030" class="elsevierStylePara elsevierViewall">The most common cause of death in both the training and testing sets was extra-hepatic cancer (<a class="elsevierStyleCrossRef" href="#tbl0003">Table 3</a>). In the training set (SteatoSITE), liver-related mortality (cirrhosis and HCC) was the second most common cause of death, followed by cardiovascular deaths, similar to the findings of a large nationwide cohort study of over 10,000 patients with biopsy-confirmed NAFLD [<a class="elsevierStyleCrossRef" href="#bib0032">32</a>]. In contrast, in the testing set (GG&C non-biopsy cohort) cardiovascular deaths were higher. This likely reflects the different composition of the respective populations. One was a secondary care cohort based on clinically indicated tissue sampling (biopsy, resection, or explant), so there is inherent spectrum bias towards cases with more severe liver disease, whereas the other represents a more generalizable MASLD cohort, reflected in the observed cause of death frequencies that are more consistent with data from other community-diagnosed population studies [<a class="elsevierStyleCrossRef" href="#bib0041">41</a>,<a class="elsevierStyleCrossRef" href="#bib0042">42</a>]. Notably, a substantial number of deaths did not fall into any of these categories and were classified as ‘other’.</p><elsevierMultimedia ident="tbl0003"></elsevierMultimedia></span><span id="sec0010" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">3.2</span><span class="elsevierStyleSectionTitle" id="cesectitle0017">Prediction of all-cause mortality</span><p id="para0031" class="elsevierStylePara elsevierViewall">The Transformer neural network was trained and validated using five-fold cross-validation in the SteatoSITE dataset. Dimensionality reduction of the global average pooling layer confirmed the model's propensity to learn the target class (<a class="elsevierStyleCrossRef" href="#fig0002">Fig. 2</a>A). The model achieved an AUROC of 0.90 (95 % CI: 0.86–0.94), 0.85 (95 % CI: 0.79–0.90), and 0.73 (95 % CI: 0.69–0.79) for the prediction of 12-, 24, and 36-month mortality, respectively.</p><elsevierMultimedia ident="fig0002"></elsevierMultimedia><p id="para0032" class="elsevierStylePara elsevierViewall">Binarizing predicted cases and controls using an operating point of probability of mortality ≥ 50 %, resulted in sensitivity of 64 %-82 % (95 % CI: 69 %-94 %) specificity of 75 %-92 % (95 % CI 72 %-95 %), and PPV of 94–98 % (95 % CI: 91 %-100 %). Model probabilities were well calibrated, with a Pearson's R2-values of 0.94–0.99 (two-sided <span class="elsevierStyleItalic">p</span>-value = 9.9 = 10<span class="elsevierStyleSup">−4</span> – 8.3 = 10<span class="elsevierStyleSup">−2</span>, <a class="elsevierStyleCrossRef" href="#fig0002">Fig. 2</a>B).</p><p id="para0033" class="elsevierStylePara elsevierViewall">Model performance generalized well to the out-of-sample dataset, with AUROCs of 0.86 (95 % CI: 0.85–0.90), 0.80 (95 % CI: 0.79–0.88), and 0.70 (95 % CI: 0.67–0.74) for prediction of mortality after 12, 24, and 36 months respectively. Binarizing predicted cases and controls using an operating point of probability of mortality ≥ 50 %, resulted in sensitivity of 69–70 % (95 % CI: 67 %-75 %), specificity of 96–97 % (95 % CI 94 %-98 %), and PPV of 75 %-77 % (95 % CI: 74 %-81 %) (<a class="elsevierStyleCrossRef" href="#tbl0004">Table 4</a>).</p><elsevierMultimedia ident="tbl0004"></elsevierMultimedia><p id="para0034" class="elsevierStylePara elsevierViewall">Coefficients of Determination (see Methods) were calculated for every input feature. Features that correlated the most with the global average pooling layer of the model were serum albumin (R<span class="elsevierStyleSup">2</span> = 0.89), estimated glomerular filtration rate (R<span class="elsevierStyleSup">2</span> = 0.75), and aspartate aminotransferase (AST) levels (R<span class="elsevierStyleSup">2</span> = 0.67), as well as BMI (R<span class="elsevierStyleSup">2</span> = 0.63), age at index date (R<span class="elsevierStyleSup">2</span> = 0.58), and systolic blood pressure (R<span class="elsevierStyleSup">2</span> = 0.55) (<a class="elsevierStyleCrossRef" href="#fig0002">Fig. 2</a>C).</p><p id="para0035" class="elsevierStylePara elsevierViewall">Model misclassifications in the out-of-sample testing set were interpretable. For example, at 36-month probability of mortality ≥ 50 %, the model identified n = 49 false positive cases. Of these, <span class="elsevierStyleItalic">n</span> = 11 patients (22.4 %) and <span class="elsevierStyleItalic">n</span> = 13 (26.5 %) had recent diagnoses (within twelve months) of ‘acute myocardial infarction’ (I21.9) and ‘atherosclerotic heart disease of native coronary artery’ (I25.1). Furthermore, <span class="elsevierStyleItalic">n</span> = 22 (44.9 %), <span class="elsevierStyleItalic">n</span> = 9 (18.3 %), and <span class="elsevierStyleItalic">n</span> = 16 (32.6 %) patients had at least a three-year history of lipid regulators, beta adrenoreceptor blockers, and antiplatelet drug usage. Finally, <span class="elsevierStyleItalic">n</span> = 25 patients (51 %) had in-patient stays under Cardiology services as their primary specialty.</p><p id="para0036" class="elsevierStylePara elsevierViewall">Conversely, at probability of 36-month mortality ≥ 50 %, the model identified n = 38 false negative cases. Of these, the most common diagnoses upon discharge over a ten-year Observation Window were, were ‘urinary tract infection, site unspecified’ (<span class="elsevierStyleItalic">n</span> = 27 patients [71.1 %], N39.0) and ‘unspecified acute lower respiratory infection’ (<span class="elsevierStyleItalic">n</span> = 22 patients [57.9 %], J22.X). The most common primary specialty amongst the false negative cases was General Medicine (<span class="elsevierStyleItalic">n</span> = 35 patients, 92.1 %) and General Surgery (<span class="elsevierStyleItalic">n</span> = 24 patients, 63.1 %). Finally, the most frequently prescribed medication classes were non-opioid analgesics (<span class="elsevierStyleItalic">n</span> = 35 patients, 92.1 %) and antidiabetic drugs (<span class="elsevierStyleItalic">n</span> = 19 patients, 50.0 %).</p></span></span><span id="sec0011" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">4</span><span class="elsevierStyleSectionTitle" id="cesectitle0018">Discussion</span><p id="para0037" class="elsevierStylePara elsevierViewall">MASLD is a global public health problem with multisystem and multidisciplinary implications [<a class="elsevierStyleCrossRef" href="#bib0043">43</a>]. However, techniques for accurately predicting the risk of adverse clinical outcomes in patients with MASLD, such as liver biopsy [<a class="elsevierStyleCrossRef" href="#bib0044">44</a>] or dynamic changes in imaging measurements [<a class="elsevierStyleCrossRef" href="#bib0045">45</a>,<a class="elsevierStyleCrossRef" href="#bib0046">46</a>], rely on patient engagement and healthcare resource utilization. Here, we demonstrate that a simple Transformer neural network model, trained on routinely collected secondary care data, produced well-calibrated probabilities and achieved good discriminatory power in an out-of-sample dataset within a long (up to 3-year) predictive window, with AUROC of 0.70–0.86 (95 % CI: 0.67–0.90) for all-cause mortality.</p><p id="para0038" class="elsevierStylePara elsevierViewall">The training and out-of-sample testing sets were comparable in terms of gender and ethnicity distributions, as well as BMI ranges. Patients in the training set were significantly younger than the testing set (55 <span class="elsevierStyleItalic">vs</span>. 66.3 years, <a class="elsevierStyleCrossRef" href="#tbl0002">Table 2</a>). However, this discrepancy did not adversely influence model performance. Notably, the cardiovascular-specific cause of death was significantly more prevalent in the testing set, compared to the training cohort (23.9 % vs<span class="elsevierStyleItalic">.</span> 12.9 %), which may contribute to the increased overall prevalence of deaths in the out-of-sample set (23.94 % vs<span class="elsevierStyleItalic">.</span> 37.12 %).</p><p id="para0039" class="elsevierStylePara elsevierViewall">The mortality classifier used in this study is a Transformer neural network (TNN) [<a class="elsevierStyleCrossRef" href="#bib0035">35</a>]. Traditionally, the Transformer architecture was extensively applied to natural language processing, achieving state-of-the-art performance in text annotation [<a class="elsevierStyleCrossRef" href="#bib0033">33</a>], named entity recognition [<a class="elsevierStyleCrossRef" href="#bib0047">47</a>], and representation learning [<a class="elsevierStyleCrossRef" href="#bib0048">48</a>]. More recently, the utility of the Transformer architecture was explored in longitudinal EHRs, demonstrating a striking capacity to parse heterogeneous data sequences and predict multiple clinical trajectories [<a class="elsevierStyleCrossRef" href="#bib0049">49</a>,<a class="elsevierStyleCrossRef" href="#bib0050">50</a>], considerably outperforming conventional ML techniques. The propensity of this technique to handle large volumes of relatively noisy data, including errors in labels, as well as large numbers of input variables [<a class="elsevierStyleCrossRef" href="#bib0018">18</a>], makes it an attractive tool for interrogation of real-world EHRs.</p><p id="para0040" class="elsevierStylePara elsevierViewall">Machine learning models have traditionally targeted the detection of MASLD in the general population using routinely collected medical records. For example, a simple coarse trees model that utilized fasting C-peptide levels and waist circumference identified MASLD with 74.9 % accuracy in <span class="elsevierStyleItalic">n</span> = 3235 individuals [<a class="elsevierStyleCrossRef" href="#bib0026">26</a>]. Similarly, large-scale analyses (n = 1016–73,190 patients) of EHRs in secondary care settings predicted MASLD with AUROC 0.83–0.92 [<a class="elsevierStyleCrossRef" href="#bib0051">51</a>,<a class="elsevierStyleCrossRef" href="#bib0052">52</a>].</p><p id="para0041" class="elsevierStylePara elsevierViewall">The assumption behind current diagnostic algorithms in MASLD is that liver disease is the principal threat for these patients, whilst the weight of evidence indicates it is not [<a class="elsevierStyleCrossRef" href="#bib0053">53</a>]. Leveraging large, diverse, multidimensional datasets in MASLD and applying sophisticated methods such as AI/ML tools or multi-state modelling [<a class="elsevierStyleCrossRef" href="#bib0054">54</a>] will elucidate novel subphenotypes with disease trajectories reflecting variable susceptibility to liver-related and/or non-liver-related outcomes. Additionally, the availability of ‘upstream’ risk stratification tools that consider the whole patient history could assist in developing a new paradigm for community-based prognostication in MASLD that captures key demographic influences (such as age, gender, ethnicity, and deprivation index) and embraces comorbidity and polypharmacy. This strategy aligns with a growing shift in government policy in many countries towards a more preventative and anticipatory approach to the management of long-term conditions such as MASLD; the value of which is maximized when it is targeted at patients who are most likely to benefit.</p><p id="para0042" class="elsevierStylePara elsevierViewall">In this work we achieve a good balance between prediction window length (one to three years) and model performance (out-of-sample AUROC 0.70–0.86). Indeed, shorter prediction windows provide limited therapeutic benefit, with underlying disease mechanisms becoming less modifiable, whilst longer prediction windows may result in many false positives, rendering proactive therapeutic or lifestyle intervention less practicable [<a class="elsevierStyleCrossRef" href="#bib0055">55</a>,<a class="elsevierStyleCrossRef" href="#bib0056">56</a>].</p><p id="para0043" class="elsevierStylePara elsevierViewall">We utilized a data-driven strategy to delineate the salient features captured by our model by computing the coefficient of determination (R<span class="elsevierStyleSup">2</span>) between low-dimensional representations of the model global average pooling layer and input features. Although several algorithms exist to explain black box models [<a class="elsevierStyleCrossRef" href="#bib0057">57</a>,<a class="elsevierStyleCrossRef" href="#bib0058">58</a>], they are limited to lower-dimensional tabular data. Our approach, validated in medical imaging [<a class="elsevierStyleCrossRef" href="#bib0059">59</a>], attempts to explain features captured within the unstructured temporal information. Variability in albumin levels and eGFR over the ten-year observation window accounted for 90 % and 75 % of the variance in the model's global average pooling embedding layer, respectively. Recently, a multicenter study (<span class="elsevierStyleItalic">n</span> = 229 patients from 22 hospitals) demonstrated that an annual decline in serum albumin concentration in patients with MASLD is associated with adverse events, including gastroesophageal varices leading to rupture or requiring preventive intervention, hepatic failure leading to hepatic encephalopathy, HCC, other organ malignancy, and cardiovascular events [<a class="elsevierStyleCrossRef" href="#bib0060">60</a>]. Similarly, in <span class="elsevierStyleItalic">n</span> = 18,073 UK Biobank participants identified to have CKD, MASLD was associated with an increased risk of cardiovascular events, all-cause mortality, and end-stage kidney disease [<a class="elsevierStyleCrossRef" href="#bib0061">61</a>]. These studies corroborate the utility of the coefficient of determination to identify important learnt features in complex black box models.</p><p id="para0044" class="elsevierStylePara elsevierViewall">Our Transformer model presents several advantages. First, it was trained on a diverse population using routinely collected EHRs on a national scale, covering the full MASLD severity spectrum. This offers a robust inclusion criterion for a population-level risk stratification algorithm. Second, model probabilities were well calibrated and generalized well to an out-of-sample population. Finally, an accurate inference at three-year prediction window resolution offers an opportunity for a timely, low-cost preventative intervention in the general population.</p><p id="para0045" class="elsevierStylePara elsevierViewall">Our study also had limitations. First, we chose all-cause mortality as our initial clinical outcome of interest, as this is a hard endpoint that is free from bias. Future work will extend our approach to liver-specific outcomes. Second, the training set was relatively small (<span class="elsevierStyleItalic">n</span> = 940 patients) and may not represent the full breadth of clinical activity in MASLD patients, although all stages of the disease were equally represented. Furthermore, larger evaluation cohorts would provide insight into model performance and potential biases across different strata of the population (e.g., ethnicity, age groups, and deprivation indices) [<a class="elsevierStyleCrossRef" href="#bib0062">62</a>]. This presents an urgent requirement to validate model generalizability in other systems outside Scotland. This should be feasible due to the routine availability of model features and is currently our primary focus of research. Next, the retrospective nature of this study resulted in a level of class balance that may not represent real-world prevalence. Therefore, any future validation should involve a prospectively selected cohort of patients. Finally, despite our work on coefficient of determination, the black-box nature and the dimensionality of training data make interpretation of our model unintuitive. This can pose a challenge to clinical implementation.</p></span><span id="sec0012" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleLabel">5</span><span class="elsevierStyleSectionTitle" id="cesectitle0019">Conclusions</span><p id="para0046" class="elsevierStylePara elsevierViewall">In conclusion, we show that a simple Transformer model utilizing routinely collected EHRs may offer a robust tool for community-based risk stratification of MASLD patients at an increased risk of all-cause mortality. Integration of such models into health and social care systems could assist primary care physicians in the targeting of anticipatory interventions at the individual patient level, refine secondary care referral pathways, and assist more broadly in service planning. Future work will require a prospective validation study, which would allow for evaluation of the algorithm when exposed to real-world class distributions, assessing its effect on workflow safety and operational efficiency.</p></span><span id="sec0013" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0020">Funding</span><p id="para0047" class="elsevierStylePara elsevierViewall">This work was funded by a grant from <span class="elsevierStyleGrantSponsor" id="gs0001">Innovate UK EUREKA</span> (Reference: <span class="elsevierStyleGrantNumber" refid="gs0001">105976</span>) and by Bering Limited. For the purpose of open access, the author has applied a creative commons attribution (CC BY) licence to any author-accepted manuscript version arising.</p></span><span id="sec0014" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0021">Author contributions</span><p id="para0048" class="elsevierStylePara elsevierViewall">ID, JAF, TJK: conception and design, acquisition of data; ID, BS, JAF, TJK, IAR: analysis and interpretation of data; ID, JAF: drafting the article; TJK, IAR: revising the article critically for important intellectual content; all authors: approval of the final manuscript version. *ID and JAF are co-corresponding authors.</p></span></span>" "textoCompletoSecciones" => array:1 [ "secciones" => array:12 [ 0 => array:3 [ "identificador" => "xres2231018" "titulo" => "Abstract" "secciones" => array:4 [ 0 => array:2 [ "identificador" => "abss0001" "titulo" => "Introduction and Objectives" ] 1 => array:2 [ "identificador" => "abss0002" "titulo" => "Patients and Methods" ] 2 => array:2 [ "identificador" => "abss0003" "titulo" => "Results" ] 3 => array:2 [ "identificador" => "abss0004" "titulo" => "Conclusions" ] ] ] 1 => array:2 [ "identificador" => "xpalclavsec1868224" "titulo" => "Keywords" ] 2 => array:2 [ "identificador" => "xpalclavsec1868225" "titulo" => "Abbreviations" ] 3 => array:2 [ "identificador" => "sec0001" "titulo" => "Introduction" ] 4 => array:3 [ "identificador" => "sec0002" "titulo" => "Patients and Methods" "secciones" => array:5 [ 0 => array:2 [ "identificador" => "sec0003" "titulo" => "Study populations" ] 1 => array:2 [ "identificador" => "sec0004" "titulo" => "Data representation and ground truthing" ] 2 => array:2 [ "identificador" => "sec0005" "titulo" => "Model training" ] 3 => array:2 [ "identificador" => "sec0006" "titulo" => "Statistical analysis" ] 4 => array:2 [ "identificador" => "sec0007" "titulo" => "Ethical statement" ] ] ] 5 => array:3 [ "identificador" => "sec0008" "titulo" => "Results" "secciones" => array:2 [ 0 => array:2 [ "identificador" => "sec0009" "titulo" => "Training and testing dataset characteristics" ] 1 => array:2 [ "identificador" => "sec0010" "titulo" => "Prediction of all-cause mortality" ] ] ] 6 => array:2 [ "identificador" => "sec0011" "titulo" => "Discussion" ] 7 => array:2 [ "identificador" => "sec0012" "titulo" => "Conclusions" ] 8 => array:2 [ "identificador" => "sec0013" "titulo" => "Funding" ] 9 => array:2 [ "identificador" => "sec0014" "titulo" => "Author contributions" ] 10 => array:2 [ "identificador" => "xack769361" "titulo" => "Acknowledgements" ] 11 => array:1 [ "titulo" => "References" ] ] ] "pdfFichero" => "main.pdf" "tienePdf" => true "fechaRecibido" => "2024-06-10" "fechaAceptado" => "2024-06-13" "PalabrasClave" => array:1 [ "en" => array:2 [ 0 => array:4 [ "clase" => "keyword" "titulo" => "Keywords" "identificador" => "xpalclavsec1868224" "palabras" => array:5 [ 0 => "Metabolic dysfunction-associated steatotic liver disease" 1 => "Electronic health records" 2 => "Artificial intelligence" 3 => "Deep Learning" 4 => "Prognostic model" ] ] 1 => array:4 [ "clase" => "abr" "titulo" => "Abbreviations" "identificador" => "xpalclavsec1868225" "palabras" => array:14 [ 0 => "A&E" 1 => "AI" 2 => "BMI" 3 => "CKD" 4 => "DNN" 5 => "EHR" 6 => "HCC" 7 => "ICD" 8 => "MASLD" 9 => "ML" 10 => "NAFLD" 11 => "NASH" 12 => "OPCS-4" 13 => "TNN" ] ] ] ] "tieneResumen" => true "resumen" => array:1 [ "en" => array:3 [ "titulo" => "Abstract" "resumen" => "<span id="abss0001" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0002">Introduction and Objectives</span><p id="spara007" class="elsevierStyleSimplePara elsevierViewall">Despite the huge clinical burden of MASLD, validated tools for early risk stratification are lacking, and heterogeneous disease expression and a highly variable rate of progression to clinical outcomes result in prognostic uncertainty. We aimed to investigate longitudinal electronic health record-based outcome prediction in MASLD using a state-of-the-art machine learning model.</p></span> <span id="abss0002" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0003">Patients and Methods</span><p id="spara008" class="elsevierStyleSimplePara elsevierViewall"><span class="elsevierStyleItalic">n</span> = 940 patients with histologically-defined MASLD were used to develop a deep-learning model for all-cause mortality prediction. Patient timelines, spanning 12 years, were fully-annotated with demographic/clinical characteristics, ICD-9 and -10 codes, blood test results, prescribing data, and secondary care activity. A Transformer neural network (TNN) was trained to output concomitant probabilities of 12-, 24-, and 36-month all-cause mortality. In-sample performance was assessed using 5-fold cross-validation. Out-of-sample performance was assessed in an independent set of <span class="elsevierStyleItalic">n</span> = 528 MASLD patients.</p></span> <span id="abss0003" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0004">Results</span><p id="spara009" class="elsevierStyleSimplePara elsevierViewall">In-sample model performance achieved AUROC curve 0.74–0.90 (95 % CI: 0.72–0.94), sensitivity 64 %-82 %, specificity 75 %–92 % and Positive Predictive Value (PPV) 94 %-98 %. Out-of-sample model validation had AUROC 0.70–0.86 (95 % CI: 0.67–0.90), sensitivity 69 %–70 %, specificity 96 %–97 % and PPV 75 %–77 %. Key predictive factors, identified using coefficients of determination, were age, presence of type 2 diabetes, and history of hospital admissions with length of stay >14 days.</p></span> <span id="abss0004" class="elsevierStyleSection elsevierViewall"><span class="elsevierStyleSectionTitle" id="cesectitle0005">Conclusions</span><p id="spara010" class="elsevierStyleSimplePara elsevierViewall">A TNN, applied to routinely-collected longitudinal electronic health records, achieved good performance in prediction of 12-, 24-, and 36-month all-cause mortality in patients with MASLD. Extrapolation of our technique to population-level data will enable scalable and accurate risk stratification to identify people most likely to benefit from anticipatory health care and personalized interventions.</p></span>" "secciones" => array:4 [ 0 => array:2 [ "identificador" => "abss0001" "titulo" => "Introduction and Objectives" ] 1 => array:2 [ "identificador" => "abss0002" "titulo" => "Patients and Methods" ] 2 => array:2 [ "identificador" => "abss0003" "titulo" => "Results" ] 3 => array:2 [ "identificador" => "abss0004" "titulo" => "Conclusions" ] ] ] ] "multimedia" => array:6 [ 0 => array:8 [ "identificador" => "fig0001" "etiqueta" => "Fig. 1" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => true "mostrarDisplay" => false "figura" => array:1 [ 0 => array:4 [ "imagen" => "gr1.jpeg" "Alto" => 2748 "Ancho" => 3167 "Tamanyo" => 378771 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0005" "detalle" => "Fig " "rol" => "short" ] ] "descripcion" => array:1 [ "en" => "<p id="spara001" class="elsevierStyleSimplePara elsevierViewall"><span class="elsevierStyleBold">Schematic representation of an EHR vector.</span> The patient's timeline is represented by horizontal arrows and each data point is depicted by colour-coded tokens. Predictive models were trained on the data in the Observation Window (10 years), whilst a binary outcome of all-cause mortality was used as a ground truth. A&E, emergency department; U&Es, urea and electrolytes; FBC, full blood count, LFT, liver function tests; ACE inhibitor, angiotensin-converting enzyme inhibitor; BMI, body mass index.</p>" ] ] 1 => array:8 [ "identificador" => "fig0002" "etiqueta" => "Fig. 2" "tipo" => "MULTIMEDIAFIGURA" "mostrarFloat" => true "mostrarDisplay" => false "figura" => array:1 [ 0 => array:4 [ "imagen" => "gr2.jpeg" "Alto" => 3129 "Ancho" => 3167 "Tamanyo" => 490112 ] ] "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0006" "detalle" => "Fig " "rol" => "short" ] ] "descripcion" => array:1 [ "en" => "<p id="spara002" class="elsevierStyleSimplePara elsevierViewall"><span class="elsevierStyleBold">Transformer neural network performance in prediction of all-cause mortality in the SteatoSITE dataset. (A)</span> Scatterplot shows two-dimensional twin neural network (Ivis) embedding of the global average pooling layer values in the trained transformer neural network. Each point represents a single patient in the testing set. Blue and orange colours represent the presence and absence of all-cause mortality in a 12–36 month predictive window respectively. (<span class="elsevierStyleBold">B)</span> Calibration plots demonstrating the relationship between average SteatoSITE cohort mortality probabilities and proportion of true positives within each probability bin. Green, blue, and orange lines reflect model outputs for 12-, 24-, 36-month predictive windows. <span class="elsevierStyleBold">C)</span> Bar plot showing model input features and their respective coefficient of determination (R<span class="elsevierStyleSup">2</span>) values. Values reflect variance within the global average pooling layer explained by each feature. ALB, albumin: EGFR, estimated glomerular filtration rate; AST, aspartate aminotransferase; BMI, body mass index; ALT, alanine aminotransferase; BP, blood pressure; ALP, alkaline phosphatase; CREA, creatinine; HB, hemoglobin; DIAG1, code for main diagnosis; DIAG2–4, codes for other diagnosis; OP1B, code for approach site/laterality of main procedure; SPEC, code for speciality; A&E, Accident and Emergency; BNF, British National Formulary; K, potassium; LYM, lymphocytes; NEUT, neutrophils.</p>" ] ] 2 => array:8 [ "identificador" => "tbl0001" "etiqueta" => "Table 1" "tipo" => "MULTIMEDIATABLA" "mostrarFloat" => true "mostrarDisplay" => false "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0001" "detalle" => "Table " "rol" => "short" ] ] "tabla" => array:1 [ "tablatextoimagen" => array:1 [ 0 => array:2 [ "tabla" => array:1 [ 0 => """ <table border="0" frame="\n \t\t\t\t\tvoid\n \t\t\t\t" class=""><thead title="thead"><tr title="table-row"><a name="en0001"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black"><span class="elsevierStyleBold">Data Type</span> \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0002"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black"><span class="elsevierStyleBold">Description</span> \t\t\t\t\t\t\n \t\t\t\t\t\t</th></tr></thead><tbody title="tbody"><tr title="table-row"><a name="en0003"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Age \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0004"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Patient age at Index Date \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0005"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Gender \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0006"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Patient gender at Index Date \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0007"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Ethnicity \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0008"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Patient ethnicity at Index Date \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0009"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Medications \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0010"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">British National Formulary (BNF) Subsection codes \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0011"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Laboratories \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0012"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Serum analytes: urea, estimated glomerular filtration rate (eGFR), creatinine, sodium, potassium, hemoglobin, neutrophils, lymphocytes, platelets, total bilirubin, alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, gamma-glutamyltransferase \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0013"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Hospitalisations \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0014"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">All ICD-9/10 and OPCS-4 codes associated with admission, Length of stay associated with admission, Primary clinical speciality \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0015"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">A&E Attendances \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0016"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">ICD-9/10 and OPCS-4 codes associated with attendances \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0017"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Outpatients \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0018"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Appointment speciality, ICD-9/10 and OPCS-4 codes associated with each appointment \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0019"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Vital Signs \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0020"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Systolic blood pressure, body mass index (BMI) \t\t\t\t\t\t\n \t\t\t\t</td></tr></tbody></table> """ ] "imagenFichero" => array:1 [ 0 => "xTab3637370.png" ] ] ] ] "descripcion" => array:1 [ "en" => "<p id="spara003" class="elsevierStyleSimplePara elsevierViewall">Features of the patient EHR used as inputs into the predictive model.</p>" ] ] 3 => array:8 [ "identificador" => "tbl0002" "etiqueta" => "Table 2" "tipo" => "MULTIMEDIATABLA" "mostrarFloat" => true "mostrarDisplay" => false "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0002" "detalle" => "Table " "rol" => "short" ] ] "tabla" => array:2 [ "tablatextoimagen" => array:1 [ 0 => array:2 [ "tabla" => array:1 [ 0 => """ <table border="0" frame="\n \t\t\t\t\tvoid\n \t\t\t\t" class=""><thead title="thead"><tr title="table-row"><a name="en0021"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Variable \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0022"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Training Set \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0023"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Testing Set \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0024"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black"><span class="elsevierStyleItalic">p-v</span>alue \t\t\t\t\t\t\n \t\t\t\t\t\t</th></tr></thead><tbody title="tbody"><tr title="table-row"><a name="en0025"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Age \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0026"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">55 (+/- 13.5) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0027"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">66.3 (+/- 14.4) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0028"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.0001<a class="elsevierStyleCrossRef" href="#tb2fn1">*</a> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0029"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Gender \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0030"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0031"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0032"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.16<a class="elsevierStyleCrossRef" href="#tb2fn2"><span class="elsevierStyleSup">⁎⁎</span></a> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0033"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Male \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0034"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">55.4 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0035"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">45 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0036"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0037"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Female \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0038"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">44.6 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0039"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">55 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0040"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0041"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Ethnicity \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0042"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0043"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0044"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.12<a class="elsevierStyleCrossRef" href="#tb2fn2"><span class="elsevierStyleSup">⁎⁎</span></a> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0045"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Asian, Asian British, Asian Welsh \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0046"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">2.34 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0047"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">5.58 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0048"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0049"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Black, Black British, Black Welsh, Caribbean or African \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0050"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.11 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0051"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0052"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0053"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Mixed or Multiple \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0054"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0055"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.24 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0056"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0057"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>White \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0058"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">62.98 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0059"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">59.39 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0060"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0061"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Other \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0062"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0063"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">5.79 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0064"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0065"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>Unknown \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0066"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">34.57 \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0067"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">29 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0068"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0069"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Type 2 Diabetes \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0070"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">32 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0071"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">30 % \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0072"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0073"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">BMI \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0074"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">32.82 (+/-7.94) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0075"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">32.95 (+/-7.96) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0076"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.76<a class="elsevierStyleCrossRef" href="#tb2fn2"><span class="elsevierStyleSup">⁎⁎</span></a> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0077"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">Number of Deaths \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0078"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">225 (23.94 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0079"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">196 (37.12 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0080"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr></tbody></table> """ ] "imagenFichero" => array:1 [ 0 => "xTab3637371.png" ] ] ] "notaPie" => array:2 [ 0 => array:3 [ "identificador" => "tb2fn1" "etiqueta" => "⁎" "nota" => "<p class="elsevierStyleNotepara" id="notep0001">two-tailed unpaired <span class="elsevierStyleItalic">t</span>-test.</p>" ] 1 => array:3 [ "identificador" => "tb2fn2" "etiqueta" => "⁎⁎" "nota" => "<p class="elsevierStyleNotepara" id="notep0002">Chi-squared test.</p>" ] ] ] "descripcion" => array:1 [ "en" => "<p id="spara004" class="elsevierStyleSimplePara elsevierViewall">Demographic characteristics of the training and testing sets.</p>" ] ] 4 => array:8 [ "identificador" => "tbl0003" "etiqueta" => "Table 3" "tipo" => "MULTIMEDIATABLA" "mostrarFloat" => true "mostrarDisplay" => false "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0003" "detalle" => "Table " "rol" => "short" ] ] "tabla" => array:1 [ "tablatextoimagen" => array:1 [ 0 => array:2 [ "tabla" => array:1 [ 0 => """ <table border="0" frame="\n \t\t\t\t\tvoid\n \t\t\t\t" class=""><thead title="thead"><tr title="table-row"><a name="en0081"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black"> \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0082"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Training Set \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0083"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Testing Set \t\t\t\t\t\t\n \t\t\t\t\t\t</th></tr></thead><tbody title="tbody"><tr title="table-row"><a name="en0084"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top">Cardiovascular-specific (ICD10: I01-I99) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0085"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">18.1 % (12.9 % including resections) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0086"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">23.9 % \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0087"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top">Hepatocellular Carcinoma (ICD10: C22.0, C22.9, C22.99) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0088"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">6.3 % (9.3 % including resections) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0089"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">2.5 % \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0090"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top">Cirrhosis-specific (ICD10: K74) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0091"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">23.6 % (13.3 % including resections) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0092"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">2.5 % \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0093"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top">Cancer-specific (ICD10: C00-C99, excluding C22.0, C22.9, C22.99) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0094"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">33.1 % (52.0 % including resections) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0095"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">28.5 % \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0096"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top">Other causes \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0097"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">18.9 % (12.4 % including resections) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0098"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">42.6 % \t\t\t\t\t\t\n \t\t\t\t</td></tr></tbody></table> """ ] "imagenFichero" => array:1 [ 0 => "xTab3637369.png" ] ] ] ] "descripcion" => array:1 [ "en" => "<p id="spara005" class="elsevierStyleSimplePara elsevierViewall">Frequencies of the primary causes of death in the training and testing sets.</p>" ] ] 5 => array:8 [ "identificador" => "tbl0004" "etiqueta" => "Table 4" "tipo" => "MULTIMEDIATABLA" "mostrarFloat" => true "mostrarDisplay" => false "detalles" => array:1 [ 0 => array:3 [ "identificador" => "alt0004" "detalle" => "Table " "rol" => "short" ] ] "tabla" => array:1 [ "tablatextoimagen" => array:1 [ 0 => array:2 [ "tabla" => array:1 [ 0 => """ <table border="0" frame="\n \t\t\t\t\tvoid\n \t\t\t\t" class=""><thead title="thead"><tr title="table-row"><a name="en0099"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black"> \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0100"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">AUROC \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0101"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Sensitivity \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0102"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">Specificity \t\t\t\t\t\t\n \t\t\t\t\t\t</th><a name="en0103"></a><th class="td" title="\n \t\t\t\t\ttable-head\n \t\t\t\t " align="" valign="top" scope="col" style="border-bottom: 2px solid black">PPV \t\t\t\t\t\t\n \t\t\t\t\t\t</th></tr></thead><tbody title="tbody"><tr title="table-row"><a name="en0104"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleBold">SteatoSITE</span> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0105"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0106"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0107"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0108"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0109"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>1-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0110"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.90 (0.86–0.94) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0111"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">82 % (80 %-86 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0112"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">92 % (89 %-95 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0113"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">98 % (95 %-100 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0114"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>2-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0115"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.85 (0.79–0.90) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0116"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">78 % (76 %-80 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0117"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">89 % (86 %-94 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0118"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">95 % (93 %-98 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0119"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>3-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0120"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.73 (0.69–0.79) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0121"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">64 % (63 %-69 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0122"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">75 % (72 %-79 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0123"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">94 % (91 %-97 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0124"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleBold">NHS GG&C</span> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0125"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0126"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0127"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0128"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top"> \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0129"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>1-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0130"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.86 (0.85–0.90) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0131"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">70 % (69 %-73 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0132"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">97 % (95 %-98 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0133"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">77 % (75 %-81 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0134"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>2-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0135"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.80 (0.79–0.88) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0136"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">69 % (68 %-72 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0137"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">96 % (94 %-98 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0138"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">75 % (74 %-79 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr><tr title="table-row"><a name="en0139"></a><td class="td-with-role" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t ; entry_with_role_rowhead " align="" valign="top"><span class="elsevierStyleHsp" style=""></span>3-year \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0140"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">0.70 (0.67–0.74) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0141"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">69 % (67 %-71 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0142"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">96 % (94 %-98 %) \t\t\t\t\t\t\n \t\t\t\t</td><a name="en0143"></a><td class="td" title="\n \t\t\t\t\ttable-entry\n \t\t\t\t " align="" valign="top">76 % (74 %-80 %) \t\t\t\t\t\t\n \t\t\t\t</td></tr></tbody></table> """ ] "imagenFichero" => array:1 [ 0 => "xTab3637372.png" ] ] ] ] "descripcion" => array:1 [ "en" => "<p id="spara006" class="elsevierStyleSimplePara elsevierViewall"><span class="elsevierStyleBold">Model performance metrics across internal training and external testing sets</span>. Model operating point was set to ≥50 % (probability of all-cause mortality). 95 % CIs are shown in parentheses. AUROC = Area under Receiver Operating Characteristic Curve.</p>" ] ] ] "bibliografia" => array:2 [ "titulo" => "References" "seccion" => array:1 [ 0 => array:2 [ "identificador" => "cebibsec1" "bibliografiaReferencia" => array:62 [ 0 => array:3 [ "identificador" => "bib0001" "etiqueta" => "[1]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "A multisociety Delphi consensus statement on new fatty liver disease nomenclature" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "ME Rinella" 1 => "JV Lazarus" 2 => "V Ratziu" 3 => "SM Francque" 4 => "AJ Sanyal" 5 => "F Kanwal" 6 => "D Romero" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1097/HEP.0000000000000520" "Revista" => array:7 [ "tituloSerie" => "Hepatology" "fecha" => "2023" "volumen" => "78" "paginaInicial" => "1966" "paginaFinal" => "1986" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/37363821" "web" => "Medline" ] ] "itemHostRev" => array:3 [ "pii" => "S0022347614000936" "estado" => "S300" "issn" => "00223476" ] ] ] ] ] ] ] 1 => array:3 [ "identificador" => "bib0002" "etiqueta" => "[2]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "99 % of patients with NAFLD meet MASLD criteria and natural history is therefore identical" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "H Hagstrom" 1 => "J Vessby" 2 => "M Ekstedt" 3 => "Y. Shang" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:2 [ "tituloSerie" => "J Hepatol" "fecha" => "2023" ] ] ] ] ] ] 2 => array:3 [ "identificador" => "bib0003" "etiqueta" => "[3]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Mechanisms and disease consequences of nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "R Loomba" 1 => "SL Friedman" 2 => "GI. Shulman" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/j.cell.2021.04.015" "Revista" => array:6 [ "tituloSerie" => "Cell" "fecha" => "2021" "volumen" => "184" "paginaInicial" => "2537" "paginaFinal" => "2564" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/33989548" "web" => "Medline" ] ] ] ] ] ] ] ] 3 => array:3 [ "identificador" => "bib0004" "etiqueta" => "[4]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Non-alcoholic fatty liver disease in lean individuals" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "S Albhaisi" 1 => "A Chowdhury" 2 => "AJ. Sanyal" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/j.jhepr.2019.08.002" "Revista" => array:6 [ "tituloSerie" => "JHEP Rep" "fecha" => "2019" "volumen" => "1" "paginaInicial" => "329" "paginaFinal" => "341" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/32039383" "web" => "Medline" ] ] ] ] ] ] ] ] 4 => array:3 [ "identificador" => "bib0005" "etiqueta" => "[5]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Advancing the global public health agenda for NAFLD: a consensus statement" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "JV Lazarus" 1 => "HE Mark" 2 => "QM Anstee" 3 => "JP Arab" 4 => "RL Batterham" 5 => "L Castera" 6 => "H Cortez-Pinto" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41575-021-00523-4" "Revista" => array:6 [ "tituloSerie" => "Nat Rev Gastroenterol Hepatol" "fecha" => "2022" "volumen" => "19" "paginaInicial" => "60" "paginaFinal" => "78" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34707258" "web" => "Medline" ] ] ] ] ] ] ] ] 5 => array:3 [ "identificador" => "bib0006" "etiqueta" => "[6]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The prevalence and incidence of NAFLD worldwide: a systematic review and meta-analysis" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "K Riazi" 1 => "H Azhari" 2 => "JH Charette" 3 => "FE Underwood" 4 => "JA King" 5 => "EE Afshar" 6 => "MG Swain" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/S2468-1253(22)00165-0" "Revista" => array:6 [ "tituloSerie" => "Lancet Gastroenterol Hepatol" "fecha" => "2022" "volumen" => "7" "paginaInicial" => "851" "paginaFinal" => "861" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/35798021" "web" => "Medline" ] ] ] ] ] ] ] ] 6 => array:3 [ "identificador" => "bib0007" "etiqueta" => "[7]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The economic and clinical burden of nonalcoholic fatty liver disease in the United States and Europe" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "ZM Younossi" 1 => "D Blissett" 2 => "R Blissett" 3 => "L Henry" 4 => "M Stepanova" 5 => "Y Younossi" 6 => "A Racila" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1002/hep.28785" "Revista" => array:6 [ "tituloSerie" => "Hepatology" "fecha" => "2016" "volumen" => "64" "paginaInicial" => "1577" "paginaFinal" => "1586" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/27543837" "web" => "Medline" ] ] ] ] ] ] ] ] 7 => array:3 [ "identificador" => "bib0008" "etiqueta" => "[8]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Complications, morbidity and mortality of nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:6 [ 0 => "A Mantovani" 1 => "E Scorletti" 2 => "A Mosca" 3 => "A Alisi" 4 => "CD Byrne" 5 => "G. Targher" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "Metabolism" "fecha" => "2020" "volumen" => "111S" ] ] ] ] ] ] 8 => array:3 [ "identificador" => "bib0009" "etiqueta" => "[9]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Antifibrotic therapy in nonalcoholic steatohepatitis: time for a human-centric approach" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:6 [ 0 => "PN Brennan" 1 => "AM Elsharkawy" 2 => "TJ Kendall" 3 => "R Loomba" 4 => "DA Mann" 5 => "JA. Fallowfield" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41575-023-00796-x" "Revista" => array:6 [ "tituloSerie" => "Nat Rev Gastroenterol Hepatol" "fecha" => "2023" "volumen" => "20" "paginaInicial" => "679" "paginaFinal" => "688" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/37268740" "web" => "Medline" ] ] ] ] ] ] ] ] 9 => array:3 [ "identificador" => "bib0010" "etiqueta" => "[10]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Non-alcoholic fatty liver disease and its relationship with cardiovascular disease and other extrahepatic diseases" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "LA Adams" 1 => "QM Anstee" 2 => "H Tilg" 3 => "G. Targher" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1136/gutjnl-2017-313884" "Revista" => array:6 [ "tituloSerie" => "Gut" "fecha" => "2017" "volumen" => "66" "paginaInicial" => "1138" "paginaFinal" => "1153" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/28314735" "web" => "Medline" ] ] ] ] ] ] ] ] 10 => array:3 [ "identificador" => "bib0011" "etiqueta" => "[11]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Causes and risk profiles of mortality among individuals with nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "P Konyn" 1 => "A Ahmed" 2 => "D. Kim" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.3350/cmh.2022.0351" "Revista" => array:6 [ "tituloSerie" => "Clin Mol Hepatol" "fecha" => "2023" "volumen" => "29" "paginaInicial" => "S43" "paginaFinal" => "S57" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36417893" "web" => "Medline" ] ] ] ] ] ] ] ] 11 => array:3 [ "identificador" => "bib0012" "etiqueta" => "[12]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Clinical care pathway for the risk stratification and management of patients with nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "F Kanwal" 1 => "JH Shubrook" 2 => "LA Adams" 3 => "K Pfotenhauer" 4 => "V Wai-Sun Wong" 5 => "E Wright" 6 => "MF Abdelmalek" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1053/j.gastro.2021.07.049" "Revista" => array:6 [ "tituloSerie" => "Gastroenterology" "fecha" => "2021" "volumen" => "161" "paginaInicial" => "1657" "paginaFinal" => "1669" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34602251" "web" => "Medline" ] ] ] ] ] ] ] ] 12 => array:3 [ "identificador" => "bib0013" "etiqueta" => "[13]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Henry J, Pylypchuk Y, Searcy T, Patel V. Adoption of Electronic Health Record Systems among U.S. Non-Federal Acute Care Hospitals: 2008-2015. In; 2016." ] ] ] 13 => array:3 [ "identificador" => "bib0014" "etiqueta" => "[14]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Development and validation of a machine learning model using administrative health data to predict onset of type 2 diabetes" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "M Ravaut" 1 => "V Harish" 2 => "H Sadeghi" 3 => "KK Leung" 4 => "M Volkovs" 5 => "K Kornas" 6 => "T Watson" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "JAMA Netw Open" "fecha" => "2021" "volumen" => "4" ] ] ] ] ] ] 14 => array:3 [ "identificador" => "bib0015" "etiqueta" => "[15]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Can machine-learning improve cardiovascular risk prediction using routine clinical data?" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:5 [ 0 => "SF Weng" 1 => "J Reps" 2 => "J Kai" 3 => "JM Garibaldi" 4 => "N. Qureshi" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "PLoS One" "fecha" => "2017" "volumen" => "12" ] ] ] ] ] ] 15 => array:3 [ "identificador" => "bib0016" "etiqueta" => "[16]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Opportunities and challenges in developing risk prediction models with electronic health records data: a systematic review" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "BA Goldstein" 1 => "AM Navar" 2 => "MJ Pencina" 3 => "JP. Ioannidis" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1093/jamia/ocw042" "Revista" => array:6 [ "tituloSerie" => "J Am Med Inform Assoc" "fecha" => "2017" "volumen" => "24" "paginaInicial" => "198" "paginaFinal" => "208" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/27189013" "web" => "Medline" ] ] ] ] ] ] ] ] 16 => array:3 [ "identificador" => "bib0017" "etiqueta" => "[17]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Deep learning for clustering of multivariate clinical patient trajectories with missing values" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "J de Jong" 1 => "MA Emon" 2 => "P Wu" 3 => "R Karki" 4 => "M Sood" 5 => "P Godard" 6 => "A Ahmad" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "Gigascience" "fecha" => "2019" "volumen" => "8" ] ] ] ] ] ] 17 => array:3 [ "identificador" => "bib0018" "etiqueta" => "[18]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Scalable and accurate deep learning with electronic health records" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "A Rajkomar" 1 => "E Oren" 2 => "K Chen" 3 => "AM Dai" 4 => "N Hajaj" 5 => "M Hardt" 6 => "PJ Liu" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41746-018-0029-1" "Revista" => array:5 [ "tituloSerie" => "NPJ Digit Med" "fecha" => "2018" "volumen" => "1" "paginaInicial" => "18" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/31304302" "web" => "Medline" ] ] ] ] ] ] ] ] 18 => array:3 [ "identificador" => "bib0019" "etiqueta" => "[19]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Learning from longitudinal data in electronic health record and genetic data to improve cardiovascular event prediction" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "J Zhao" 1 => "Q Feng" 2 => "P Wu" 3 => "RA Lupu" 4 => "RA Wilke" 5 => "QS Wells" 6 => "JC Denny" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41598-018-36745-x" "Revista" => array:5 [ "tituloSerie" => "Sci Rep" "fecha" => "2019" "volumen" => "9" "paginaInicial" => "717" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/30679510" "web" => "Medline" ] ] ] ] ] ] ] ] 19 => array:3 [ "identificador" => "bib0020" "etiqueta" => "[20]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Predicting hypertension onset from longitudinal electronic health records with deep learning" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:7 [ 0 => "S Datta" 1 => "A Morassi Sasso" 2 => "N Kiwit" 3 => "S Bose" 4 => "G Nadkarni" 5 => "R Miotto" 6 => "EP Bottinger" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1093/jamiaopen/ooac097" "Revista" => array:5 [ "tituloSerie" => "JAMIA Open" "fecha" => "2022" "volumen" => "5" "paginaInicial" => "ooac097" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36448021" "web" => "Medline" ] ] ] ] ] ] ] ] 20 => array:3 [ "identificador" => "bib0021" "etiqueta" => "[21]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Liver fibrosis, but no other histologic features, is associated with long-term outcomes of patients with nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "P Angulo" 1 => "DE Kleiner" 2 => "S Dam-Larsen" 3 => "LA Adams" 4 => "ES Bjornsson" 5 => "P Charatcharoenwitthaya" 6 => "PR Mills" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:5 [ "tituloSerie" => "Gastroenterology" "fecha" => "2015" "volumen" => "149" "paginaInicial" => "389" "paginaFinal" => "397" ] ] ] ] ] ] 21 => array:3 [ "identificador" => "bib0022" "etiqueta" => "[22]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Prospective study of outcomes in adults with nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "AJ Sanyal" 1 => "ML Van Natta" 2 => "J Clark" 3 => "BA Neuschwander-Tetri" 4 => "A Diehl" 5 => "S Dasarathy" 6 => "R Loomba" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1056/NEJMoa2029349" "Revista" => array:6 [ "tituloSerie" => "N Engl J Med" "fecha" => "2021" "volumen" => "385" "paginaInicial" => "1559" "paginaFinal" => "1569" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34670043" "web" => "Medline" ] ] ] ] ] ] ] ] 22 => array:3 [ "identificador" => "bib0023" "etiqueta" => "[23]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Current status of imaging in nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:5 [ 0 => "Q Li" 1 => "M Dhyani" 2 => "JR Grajo" 3 => "C Sirlin" 4 => "AE. Samir" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.4254/wjh.v10.i8.530" "Revista" => array:6 [ "tituloSerie" => "World J Hepatol" "fecha" => "2018" "volumen" => "10" "paginaInicial" => "530" "paginaFinal" => "542" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/30190781" "web" => "Medline" ] ] ] ] ] ] ] ] 23 => array:3 [ "identificador" => "bib0024" "etiqueta" => "[24]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Non-invasive tests accurately stratify patients with NAFLD based on their risk of liver-related events" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "J Boursier" 1 => "H Hagstrom" 2 => "M Ekstedt" 3 => "C Moreau" 4 => "M Bonacci" 5 => "S Cure" 6 => "J Ampuero" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/j.jhep.2021.12.031" "Revista" => array:6 [ "tituloSerie" => "J Hepatol" "fecha" => "2022" "volumen" => "76" "paginaInicial" => "1013" "paginaFinal" => "1020" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/35063601" "web" => "Medline" ] ] ] ] ] ] ] ] 24 => array:3 [ "identificador" => "bib0025" "etiqueta" => "[25]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Design and validation of a histological scoring system for nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "DE Kleiner" 1 => "EM Brunt" 2 => "M Van Natta" 3 => "C Behling" 4 => "MJ Contos" 5 => "OW Cummings" 6 => "LD Ferrell" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1002/hep.20701" "Revista" => array:6 [ "tituloSerie" => "Hepatology" "fecha" => "2005" "volumen" => "41" "paginaInicial" => "1313" "paginaFinal" => "1321" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/15915461" "web" => "Medline" ] ] ] ] ] ] ] ] 25 => array:3 [ "identificador" => "bib0026" "etiqueta" => "[26]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Machine learning models for predicting non-alcoholic fatty liver disease in the general United States population: NHANES database" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "A Atsawarungruangkit" 1 => "P Laoveeravat" 2 => "K. Promrat" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.4254/wjh.v13.i10.1417" "Revista" => array:6 [ "tituloSerie" => "World J Hepatol" "fecha" => "2021" "volumen" => "13" "paginaInicial" => "1417" "paginaFinal" => "1427" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34786176" "web" => "Medline" ] ] ] ] ] ] ] ] 26 => array:3 [ "identificador" => "bib0027" "etiqueta" => "[27]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Application of machine learning in predicting non-alcoholic fatty liver disease using anthropometric and body composition indices" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "F Razmpour" 1 => "R Daryabeygi-Khotbehsara" 2 => "D Soleimani" 3 => "H Asgharnezhad" 4 => "A Shamsi" 5 => "GS Bajestani" 6 => "M Nematy" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41598-023-32129-y" "Revista" => array:5 [ "tituloSerie" => "Sci Rep" "fecha" => "2023" "volumen" => "13" "paginaInicial" => "4942" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36973382" "web" => "Medline" ] ] ] ] ] ] ] ] 27 => array:3 [ "identificador" => "bib0028" "etiqueta" => "[28]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "High-throughput, machine learning-based quantification of steatosis, inflammation, ballooning, and fibrosis in biopsies from patients with nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "R Forlano" 1 => "BH Mullish" 2 => "N Giannakeas" 3 => "JB Maurice" 4 => "N Angkathunyakul" 5 => "J Lloyd" 6 => "AT Tzallas" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/j.cgh.2019.12.025" "Revista" => array:7 [ "tituloSerie" => "Clin Gastroenterol Hepatol" "fecha" => "2020" "volumen" => "18" "paginaInicial" => "2081" "paginaFinal" => "2090" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/31887451" "web" => "Medline" ] ] "itemHostRev" => array:3 [ "pii" => "S1590865821000852" "estado" => "S300" "issn" => "15908658" ] ] ] ] ] ] ] 28 => array:3 [ "identificador" => "bib0029" "etiqueta" => "[29]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "A machine learning approach enables quantitative measurement of liver histology and disease monitoring in NASH" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "A Taylor-Weiner" 1 => "H Pokkalla" 2 => "L Han" 3 => "C Jia" 4 => "R Huss" 5 => "C Chung" 6 => "H Elliott" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1002/hep.31750" "Revista" => array:6 [ "tituloSerie" => "Hepatology" "fecha" => "2021" "volumen" => "74" "paginaInicial" => "133" "paginaFinal" => "147" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/33570776" "web" => "Medline" ] ] ] ] ] ] ] ] 29 => array:3 [ "identificador" => "bib0030" "etiqueta" => "[30]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "An integrated gene-to-outcome multimodal database for metabolic dysfunction-associated steatotic liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "TJ Kendall" 1 => "M Jimenez-Ramos" 2 => "F Turner" 3 => "P Ramachandran" 4 => "J Minnier" 5 => "MD McColgan" 6 => "M Alam" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41591-023-02602-2" "Revista" => array:6 [ "tituloSerie" => "Nat Med" "fecha" => "2023" "volumen" => "29" "paginaInicial" => "2939" "paginaFinal" => "2953" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/37903863" "web" => "Medline" ] ] ] ] ] ] ] ] 30 => array:3 [ "identificador" => "bib0031" "etiqueta" => "[31]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Administrative Coding in Electronic Health Care Record-Based Research of NAFLD: An Expert Panel Consensus Statement" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "H Hagstrom" 1 => "LA Adams" 2 => "AM Allen" 3 => "CD Byrne" 4 => "Y Chang" 5 => "H Gronbaek" 6 => "M Ismail" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1002/hep.31726" "Revista" => array:6 [ "tituloSerie" => "Hepatology" "fecha" => "2021" "volumen" => "74" "paginaInicial" => "474" "paginaFinal" => "482" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/33486773" "web" => "Medline" ] ] ] ] ] ] ] ] 31 => array:3 [ "identificador" => "bib0032" "etiqueta" => "[32]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Mortality in biopsy-confirmed nonalcoholic fatty liver disease: results from a nationwide cohort" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:5 [ 0 => "TG Simon" 1 => "B Roelstraete" 2 => "H Khalili" 3 => "H Hagstrom" 4 => "JF. Ludvigsson" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1136/gutjnl-2020-322786" "Revista" => array:6 [ "tituloSerie" => "Gut" "fecha" => "2021" "volumen" => "70" "paginaInicial" => "1375" "paginaFinal" => "1382" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/33037056" "web" => "Medline" ] ] ] ] ] ] ] ] 32 => array:3 [ "identificador" => "bib0033" "etiqueta" => "[33]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Development and prospective validation of COVID-19 chest X-ray screening model for patients attending emergency departments" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "I Drozdov" 1 => "B Szubert" 2 => "E Reda" 3 => "P Makary" 4 => "D Forbes" 5 => "SL Chang" 6 => "A Ezhil" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41598-021-99986-3" "Revista" => array:5 [ "tituloSerie" => "Sci Rep" "fecha" => "2021" "volumen" => "11" "paginaInicial" => "20384" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34650190" "web" => "Medline" ] ] ] ] ] ] ] ] 33 => array:3 [ "identificador" => "bib0034" "etiqueta" => "[34]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Supervised and unsupervised language modelling in Chest X-Ray radiological reports" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:6 [ 0 => "I Drozdov" 1 => "D Forbes" 2 => "B Szubert" 3 => "M Hall" 4 => "C Carlin" 5 => "DJ. Lowe" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "PLoS One" "fecha" => "2020" "volumen" => "15" ] ] ] ] ] ] 34 => array:3 [ "identificador" => "bib0035" "etiqueta" => "[35]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, Kaiser L, et al. Attention is all you need. In; 2017. p. arXiv:1706.03762." ] ] ] 35 => array:3 [ "identificador" => "bib0036" "etiqueta" => "[36]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Agarap AF. Deep learning using rectified linear units (ReLU). In; 2018. p. arXiv:1803.08375." ] ] ] 36 => array:3 [ "identificador" => "bib0037" "etiqueta" => "[37]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Li L, Jamieson K, DeSalvo G, Rostamizadeh A, Talwalkar A. Hyperband: A novel bandit-based approach to hyperparameter optimization. In; 2016. p. arXiv:1603.06560." ] ] ] 37 => array:3 [ "identificador" => "bib0038" "etiqueta" => "[38]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Structure-preserving visualisation of high dimensional single-cell datasets" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "B Szubert" 1 => "JE Cole" 2 => "C Monaco" 3 => "I. Drozdov" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41598-019-45301-0" "Revista" => array:5 [ "tituloSerie" => "Sci Rep" "fecha" => "2019" "volumen" => "9" "paginaInicial" => "8914" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/31222035" "web" => "Medline" ] ] ] ] ] ] ] ] 38 => array:3 [ "identificador" => "bib0039" "etiqueta" => "[39]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Predicting good probabilities with supervised learning" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:2 [ 0 => "A Niculescu-Mizil" 1 => "R. Caruana" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "LibroEditado" => array:5 [ "titulo" => "Proceedings of the 22nd international conference on Machine learning" "paginaInicial" => "625" "paginaFinal" => "632" "conferencia" => "Bonn, Germany" "serieFecha" => "2005" ] ] ] ] ] ] 39 => array:3 [ "identificador" => "bib0040" "etiqueta" => "[40]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Scikit-learn: machine learning in python" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "F Pedregosa" 1 => "G Varoquaux" 2 => "A Gramfort" 3 => "V Michel" 4 => "B Thirion" 5 => "O Grisel" 6 => "M Blondel" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:6 [ "tituloSerie" => "J Mach Learn Res" "fecha" => "2011" "volumen" => "12" "paginaInicial" => "2825" "paginaFinal" => "2830" "itemHostRev" => array:3 [ "pii" => "S002234761631037X" "estado" => "S300" "issn" => "00223476" ] ] ] ] ] ] ] 40 => array:3 [ "identificador" => "bib0041" "etiqueta" => "[41]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The natural history of nonalcoholic fatty liver disease: a population-based cohort study" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:7 [ 0 => "LA Adams" 1 => "JF Lymp" 2 => "J St Sauver" 3 => "SO Sanderson" 4 => "KD Lindor" 5 => "A Feldstein" 6 => "P Angulo" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1053/j.gastro.2005.04.014" "Revista" => array:6 [ "tituloSerie" => "Gastroenterology" "fecha" => "2005" "volumen" => "129" "paginaInicial" => "113" "paginaFinal" => "121" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/16012941" "web" => "Medline" ] ] ] ] ] ] ] ] 41 => array:3 [ "identificador" => "bib0042" "etiqueta" => "[42]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Trends in mortality from extrahepatic complications in patients with chronic liver disease, from 2007 through 2017" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "D Kim" 1 => "AC Adejumo" 2 => "ER Yoo" 3 => "U Iqbal" 4 => "AA Li" 5 => "EA Pham" 6 => "G Cholankeril" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1053/j.gastro.2019.06.026" "Revista" => array:6 [ "tituloSerie" => "Gastroenterology" "fecha" => "2019" "volumen" => "157" "paginaInicial" => "1055" "paginaFinal" => "1066" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/31251928" "web" => "Medline" ] ] ] ] ] ] ] ] 42 => array:3 [ "identificador" => "bib0043" "etiqueta" => "[43]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The EASL-lancet liver commission: protecting the next generation of Europeans against liver disease complications and premature mortality" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "TH Karlsen" 1 => "N Sheron" 2 => "S Zelber-Sagi" 3 => "P Carrieri" 4 => "G Dusheiko" 5 => "E Bugianesi" 6 => "R Pryke" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/S0140-6736(21)01701-3" "Revista" => array:6 [ "tituloSerie" => "Lancet" "fecha" => "2022" "volumen" => "399" "paginaInicial" => "61" "paginaFinal" => "116" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34863359" "web" => "Medline" ] ] ] ] ] ] ] ] 43 => array:3 [ "identificador" => "bib0044" "etiqueta" => "[44]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Fibrosis stage but not NASH predicts mortality and time to development of severe liver disease in biopsy-proven NAFLD" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:7 [ 0 => "H Hagstrom" 1 => "P Nasr" 2 => "M Ekstedt" 3 => "U Hammar" 4 => "P Stal" 5 => "R Hultcrantz" 6 => "S. Kechagias" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1016/j.jhep.2017.07.027" "Revista" => array:6 [ "tituloSerie" => "J Hepatol" "fecha" => "2017" "volumen" => "67" "paginaInicial" => "1265" "paginaFinal" => "1273" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/28803953" "web" => "Medline" ] ] ] ] ] ] ] ] 44 => array:3 [ "identificador" => "bib0045" "etiqueta" => "[45]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Change in serial liver stiffness measurement by magnetic resonance elastography and outcomes in NAFLD" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "T Gidener" 1 => "RA Dierkhising" 2 => "KC Mara" 3 => "TM Therneau" 4 => "SK Venkatesh" 5 => "RL Ehman" 6 => "M Yin" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1002/hep.32594" "Revista" => array:6 [ "tituloSerie" => "Hepatology" "fecha" => "2023" "volumen" => "77" "paginaInicial" => "268" "paginaFinal" => "274" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/35642504" "web" => "Medline" ] ] ] ] ] ] ] ] 45 => array:3 [ "identificador" => "bib0046" "etiqueta" => "[46]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Dynamics in liver stiffness measurements predict outcomes in advanced chronic liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "G Semmler" 1 => "Z Yang" 2 => "L Fritz" 3 => "F Kock" 4 => "BS Hofer" 5 => "L Balcar" 6 => "L Hartl" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1053/j.gastro.2023.06.030" "Revista" => array:6 [ "tituloSerie" => "Gastroenterology" "fecha" => "2023" "volumen" => "165" "paginaInicial" => "1041" "paginaFinal" => "1052" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/37442301" "web" => "Medline" ] ] ] ] ] ] ] ] 46 => array:3 [ "identificador" => "bib0047" "etiqueta" => "[47]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Transformer-based named entity recognition for parsing clinical trial eligibility criteria" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "S Tian" 1 => "A Erdengasileng" 2 => "X Yang" 3 => "Y Guo" 4 => "Y Wu" 5 => "J Zhang" 6 => "J Bian" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "ACM BCB" "fecha" => "2021" "paginaInicial" => "2021" ] ] ] ] ] ] 47 => array:3 [ "identificador" => "bib0048" "etiqueta" => "[48]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Huang K, Altosaar J, Ranganath R. ClinicalBERT: modeling clinical notes and predicting hospital readmission. In; 2019. p. arXiv:1904.05342." ] ] ] 48 => array:3 [ "identificador" => "bib0049" "etiqueta" => "[49]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "BEHRT: transformer for electronic health records" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "Y Li" 1 => "S Rao" 2 => "JRA Solares" 3 => "A Hassaine" 4 => "R Ramakrishnan" 5 => "D Canoy" 6 => "Y Zhu" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1038/s41598-020-62922-y" "Revista" => array:5 [ "tituloSerie" => "Sci Rep" "fecha" => "2020" "volumen" => "10" "paginaInicial" => "7155" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/32346050" "web" => "Medline" ] ] ] ] ] ] ] ] 49 => array:3 [ "identificador" => "bib0050" "etiqueta" => "[50]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Rupp M, Peter O, Pattipaka T. ExBEHRT: Extended transformer for electronic health records to predict disease subtypes & progressions. In; 2023. p. arXiv:2303.12364." ] ] ] 50 => array:3 [ "identificador" => "bib0051" "etiqueta" => "[51]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Application of machine learning methods to predict non-alcoholic steatohepatitis (NASH) in non-alcoholic fatty liver (NAFL) patients" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "S Fialoke" 1 => "A Malarstig" 2 => "MR Miller" 3 => "A. Dumitriu" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:6 [ "tituloSerie" => "AMIA Annu Symp Proc" "fecha" => "2018" "volumen" => "2018" "paginaInicial" => "430" "paginaFinal" => "439" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/30815083" "web" => "Medline" ] ] ] ] ] ] ] ] 51 => array:3 [ "identificador" => "bib0052" "etiqueta" => "[52]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Artificial Intelligence in NAFLD: Will Liver Biopsy Still Be Necessary in the Future?" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:2 [ 0 => "L Zhang" 1 => "Y. Mao" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "Healthcare (Basel)" "fecha" => "2022" "volumen" => "11" ] ] ] ] ] ] 52 => array:3 [ "identificador" => "bib0053" "etiqueta" => "[53]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Too much medicine: overdiagnosis and overtreatment of non-alcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:1 [ 0 => "IA. Rowe" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:5 [ "tituloSerie" => "Lancet Gastroenterol Hepatol" "fecha" => "2018" "volumen" => "3" "paginaInicial" => "66" "paginaFinal" => "72" ] ] ] ] ] ] 53 => array:3 [ "identificador" => "bib0054" "etiqueta" => "[54]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The utility of multistate models: a flexible framework for time-to-event data" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:3 [ 0 => "JG Le-Rademacher" 1 => "TM Therneau" 2 => "FS. Ou" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1007/s40471-022-00291-y" "Revista" => array:6 [ "tituloSerie" => "Curr Epidemiol Rep" "fecha" => "2022" "volumen" => "9" "paginaInicial" => "183" "paginaFinal" => "189" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/36003089" "web" => "Medline" ] ] ] ] ] ] ] ] 54 => array:3 [ "identificador" => "bib0055" "etiqueta" => "[55]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Alarm fatigue: a patient safety concern" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:2 [ 0 => "S Sendelbach" 1 => "M. Funk" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1097/NCI.0b013e3182a903f9" "Revista" => array:6 [ "tituloSerie" => "AACN Adv Crit Care" "fecha" => "2013" "volumen" => "24" "paginaInicial" => "378" "paginaFinal" => "386" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/24153215" "web" => "Medline" ] ] ] ] ] ] ] ] 55 => array:3 [ "identificador" => "bib0056" "etiqueta" => "[56]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Early detection of heart failure using electronic health records: practical implications for time before diagnosis, data diversity, data quantity, and data density" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:5 [ 0 => "K Ng" 1 => "SR Steinhubl" 2 => "C deFilippi" 3 => "S Dey" 4 => "WF. Stewart" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1161/CIRCOUTCOMES.116.002797" "Revista" => array:6 [ "tituloSerie" => "Circ Cardiovasc Qual Outcomes" "fecha" => "2016" "volumen" => "9" "paginaInicial" => "649" "paginaFinal" => "658" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/28263940" "web" => "Medline" ] ] ] ] ] ] ] ] 56 => array:3 [ "identificador" => "bib0057" "etiqueta" => "[57]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Tulio Ribeiro M, Singh S, Guestrin C. "Why should I trust you?": Explaining the predictions of any classifier. In; 2016. p. arXiv:1602.04938." ] ] ] 57 => array:3 [ "identificador" => "bib0058" "etiqueta" => "[58]" "referencia" => array:1 [ 0 => array:1 [ "referenciaCompleta" => "Lundberg S, Lee S-I. A unified approach to interpreting model predictions. In; 2017. p. arXiv:1705.07874." ] ] ] 58 => array:3 [ "identificador" => "bib0059" "etiqueta" => "[59]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "An artificial neural network for nasogastric tube position decision support" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "I Drozdov" 1 => "R Dixon" 2 => "B Szubert" 3 => "J Dunn" 4 => "D Green" 5 => "N Hall" 6 => "A Shirandami" ] ] ] ] ] "host" => array:1 [ 0 => array:1 [ "Revista" => array:3 [ "tituloSerie" => "Radiol Artif Intell" "fecha" => "2023" "volumen" => "5" ] ] ] ] ] ] 59 => array:3 [ "identificador" => "bib0060" "etiqueta" => "[60]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Decline in serum albumin concentration is a predictor of serious events in nonalcoholic fatty liver disease" "autores" => array:1 [ 0 => array:2 [ "etal" => true "autores" => array:7 [ 0 => "K Kawaguchi" 1 => "Y Sakai" 2 => "T Terashima" 3 => "T Shimode" 4 => "A Seki" 5 => "N Orita" 6 => "Y Takeshita" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1097/MD.0000000000026835" "Revista" => array:6 [ "tituloSerie" => "Medicine (Baltimore)" "fecha" => "2021" "volumen" => "100" "paginaInicial" => "e26835" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/34397849" "web" => "Medline" ] ] "itemHostRev" => array:3 [ "pii" => "S0268960X16300789" "estado" => "S300" "issn" => "0268960X" ] ] ] ] ] ] ] 60 => array:3 [ "identificador" => "bib0061" "etiqueta" => "[61]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "The impact of non-alcoholic fatty liver disease and liver fibrosis on adverse clinical outcomes and mortality in patients with chronic kidney disease: a prospective cohort study using the UK Biobank" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:7 [ 0 => "TJ Hydes" 1 => "OJ Kennedy" 2 => "R Buchanan" 3 => "DJ Cuthbertson" 4 => "J Parkes" 5 => "SDS Fraser" 6 => "P. Roderick" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1186/s12916-023-02891-x" "Revista" => array:5 [ "tituloSerie" => "BMC Med" "fecha" => "2023" "volumen" => "21" "paginaInicial" => "185" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/37198624" "web" => "Medline" ] ] ] ] ] ] ] ] 61 => array:3 [ "identificador" => "bib0062" "etiqueta" => "[62]" "referencia" => array:1 [ 0 => array:2 [ "contribucion" => array:1 [ 0 => array:2 [ "titulo" => "Dissecting racial bias in an algorithm used to manage the health of populations" "autores" => array:1 [ 0 => array:2 [ "etal" => false "autores" => array:4 [ 0 => "Z Obermeyer" 1 => "B Powers" 2 => "C Vogeli" 3 => "S. Mullainathan" ] ] ] ] ] "host" => array:1 [ 0 => array:2 [ "doi" => "10.1126/science.aax2342" "Revista" => array:6 [ "tituloSerie" => "Science" "fecha" => "2019" "volumen" => "366" "paginaInicial" => "447" "paginaFinal" => "453" "link" => array:1 [ 0 => array:2 [ "url" => "https://www.ncbi.nlm.nih.gov/pubmed/31649194" "web" => "Medline" ] ] ] ] ] ] ] ] ] ] ] ] "agradecimientos" => array:1 [ 0 => array:4 [ "identificador" => "xack769361" "titulo" => "Acknowledgements" "texto" => "<p id="para0050" class="elsevierStylePara elsevierViewall">We are grateful to Charlie Mayor (NHS Greater Glasgow & Clyde Safe Haven) for electronic health data extraction and provision and to Lynn McMahon and Marian McNeil (Precision Medicine Scotland-Innovation Centre) for project management, data compliance, data security, and workspace provision.</p>" "vista" => "all" ] ] ] "idiomaDefecto" => "en" "url" => "/16652681/0000002900000005/v1_202408300521/S1665268124003223/v1_202408300521/en/main.assets" "Apartado" => array:4 [ "identificador" => "78265" "tipo" => "SECCION" "en" => array:2 [ "titulo" => "Original Articles" "idiomaDefecto" => true ] "idiomaDefecto" => "en" ] "PDF" => "https://static.elsevier.es/multimedia/16652681/0000002900000005/v1_202408300521/S1665268124003223/v1_202408300521/en/main.pdf?idApp=UINPBA00004N&text.app=https://www.elsevier.es/" "EPUB" => "https://multimedia.elsevier.es/PublicationsMultimediaV1/item/epub/S1665268124003223?idApp=UINPBA00004N" ]
Year/Month | Html | Total | |
---|---|---|---|
2024 November | 2 | 0 | 2 |
2024 October | 46 | 17 | 63 |
2024 September | 51 | 18 | 69 |
2024 August | 74 | 35 | 109 |
2024 July | 85 | 33 | 118 |