Developing and validating an explainable digital mortality prediction tool for extremely preterm infants

preprint OA: closed
📄 Open PDF Full text JSON View at publisher
AI-generated deep summary by claude@2026-07, 2026-07-06 · read from full text

The paper develops and internally validates an explainable online tool to predict mortality before neonatal discharge among 25,902 extremely preterm infants (23+0 to 27+6 weeks’ gestation) admitted to 185 English and Welsh neonatal units from 2010–2020, using routinely entered electronic patient record data. The authors compared nine machine learning approaches and selected an explainable model based on stepwise backward logistic regression, reporting good discrimination (AUROC 0.746, 95% CI 0.729–0.762) and superior calibration and net benefit across probability thresholds (10%–70%), outperforming previously published models; performance was also acceptable in an external multinational cohort (Australasia). The main limitation explicitly emphasized is that further evaluation is needed before routine use. This paper does not explicitly discuss endometriosis or adenomyosis; it was included in the corpus via a keyword match in the upstream search index.

Read from the paper's body, not the abstract. Not a substitute for reading the paper. No clinical advice. How this works

Abstract

Decision-making in perinatal management of extremely preterm infants is challenging. Mortality prediction tools may support decision-making. We used population-based routinely entered electronic patient record data from 25,902 infants born between 23 +0 – 27 +6 weeks’ gestation and admitted to 185 English and Welsh neonatal units from 2010– 2020 to develop and internally validate an online tool to predict mortality before neonatal discharge. Comparing nine machine learning approaches, we developed an explainable tool based on stepwise backward logistic regression ( https://premoutcome.shinyapps.io/Death/ ). The tool demonstrated good discrimination (area under the receiver operating characteristics curve (95% confidence interval) of 0.746 (0.729–0.762)) and calibration with superior net benefit across probability thresholds of 10%–70%. Our tool also demonstrated superior calibration and utility performance than previously published models. Acceptable performance was demonstrated in a multinational, external validation cohort of preterm infants. This tool may be useful to support high-risk perinatal decision-making following further evaluation. Author Summary Increasingly, more premature babies are being born even earlier and surviving. Each premature baby is unique, with different combinations of factors affecting their chances of survival. An individualised approach is needed to support discussions with parents in creating a care plan for the baby before birth. Prediction tools can help support this discussion and reduce variation in the care delivered by providing an objective measure after considering important risk factors. We used artificial intelligence to analyse the electronic health records of 25,902 premature babies born between 23 and 27 completed weeks of pregnancy from 2010 to 2020 in England and Wales. We worked with parent groups to use the data pattern identified by artificial intelligence to develop an online tool ( https://premoutcome.shinyapps.io/Death/ ) to predict the risk of premature babies dying. The tool demonstrated how the risk factors contributed to the prediction, explaining how the predicted risk was derived. The tool developed demonstrated better performance than previously developed tools in our cohort of babies in England and Wales. The tool also showed good performance when tested in a separate cohort of babies in Australasia. The tool developed could support parental discussion and decision-making following further evaluation.
Full text 57,387 characters · extracted from preprint-html · click to expand
Developing and validating an explainable digital mortality prediction tool for extremely preterm infants | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Developing and validating an explainable digital mortality prediction tool for extremely preterm infants View ORCID Profile T’ng Chang Kwok , Chao Chen , Jayaprakash Veeravalli , Carol AC Coupland , Edmund Juszczak , Jonathan Garibaldi , Kirsten Mitchell , View ORCID Profile Kate L Francis , View ORCID Profile Christopher J D McKinlay , Brett J Manley , View ORCID Profile Don Sharkey doi: https://doi.org/10.1101/2025.07.09.25331175 T’ng Chang Kwok 1 Centre for Perinatal Research, Lifespan and Population Health, School of Medicine, University of Nottingham , Nottingham, UK 2 Nottingham Neonatal Service, Nottingham Univesity Hospitals NHS Trust , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for T’ng Chang Kwok Chao Chen 3 School of Computer Science, University of Nottingham , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jayaprakash Veeravalli 3 School of Computer Science, University of Nottingham , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Carol AC Coupland 4 Centre for Academic Primary Care, Lifespan and Population Health, School of Medicine, University of Nottingham , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Edmund Juszczak 5 Nottingham Clinical Trials Unit, School of Medicine, University of Nottingham , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jonathan Garibaldi 3 School of Computer Science, University of Nottingham , Nottingham, UK 6 Provost Office, University of Nottingham , Ningbo, China Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kirsten Mitchell 7 Spoons Charity , Manchester, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kate L Francis 8 Murdoch Children’s Research Institute , Melbourne, Victoria, Australia 9 Department of Paediatrics, The University of Melbourne , Melbourne, Victoria, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Kate L Francis Christopher J D McKinlay 10 Kidz First Neonatal Care, Te Whatu Ora Counties Manukau , Auckland, New Zealand 11 Department of Paediatrics: Child and Youth Health, University of Auckland , Auckland, New Zealand Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Christopher J D McKinlay Brett J Manley 8 Murdoch Children’s Research Institute , Melbourne, Victoria, Australia 12 Newborn Research, The Royal Women’s Hospital , Melbourne, Victoria, Australia 13 Department of Obstetrics , Gynaecology and Newborn Health, The University of Melbourne , Melbourne, Victoria, Australia Find this author on Google Scholar Find this author on PubMed Search for this author on this site Don Sharkey 1 Centre for Perinatal Research, Lifespan and Population Health, School of Medicine, University of Nottingham , Nottingham, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Don Sharkey For correspondence: don.sharkey{at}nottingham.ac.uk Abstract Full Text Info/History Metrics Preview PDF Abstract Decision-making in perinatal management of extremely preterm infants is challenging. Mortality prediction tools may support decision-making. We used population-based routinely entered electronic patient record data from 25,902 infants born between 23 +0 – 27 +6 weeks’ gestation and admitted to 185 English and Welsh neonatal units from 2010– 2020 to develop and internally validate an online tool to predict mortality before neonatal discharge. Comparing nine machine learning approaches, we developed an explainable tool based on stepwise backward logistic regression ( https://premoutcome.shinyapps.io/Death/ ). The tool demonstrated good discrimination (area under the receiver operating characteristics curve (95% confidence interval) of 0.746 (0.729–0.762)) and calibration with superior net benefit across probability thresholds of 10%–70%. Our tool also demonstrated superior calibration and utility performance than previously published models. Acceptable performance was demonstrated in a multinational, external validation cohort of preterm infants. This tool may be useful to support high-risk perinatal decision-making following further evaluation. Author Summary Increasingly, more premature babies are being born even earlier and surviving. Each premature baby is unique, with different combinations of factors affecting their chances of survival. An individualised approach is needed to support discussions with parents in creating a care plan for the baby before birth. Prediction tools can help support this discussion and reduce variation in the care delivered by providing an objective measure after considering important risk factors. We used artificial intelligence to analyse the electronic health records of 25,902 premature babies born between 23 and 27 completed weeks of pregnancy from 2010 to 2020 in England and Wales. We worked with parent groups to use the data pattern identified by artificial intelligence to develop an online tool ( https://premoutcome.shinyapps.io/Death/ ) to predict the risk of premature babies dying. The tool demonstrated how the risk factors contributed to the prediction, explaining how the predicted risk was derived. The tool developed demonstrated better performance than previously developed tools in our cohort of babies in England and Wales. The tool also showed good performance when tested in a separate cohort of babies in Australasia. The tool developed could support parental discussion and decision-making following further evaluation. Introduction Preterm births are increasing in industrialised countries and are the leading cause of mortality in children under five( 1 ). Over the last 20 years, survival of extremely preterm infants born before 28 weeks’ gestation has improved across the UK( 2 , 3 ) and other developed countries( 4 , 5 ). Approximately 50% of neonatal unit deaths of extreme preterm infants occur in the first week of life( 3 ). Parental counselling and preterm birth decision-making surrounding the perinatal management of infants born at the extremes of prematurity can be very challenging( 6 ). Parents prefer counselling to be personalised, to balance hope and realistic expectations, and to offer shared decision-making( 7 ). Perinatal clinical teams may lack recent risk-based data to allow more individualised counselling and decision-making, potentially introducing bias and personal experience into the discussion, resulting in variability with subsequent care of the infant. Internationally, there is marked variation in active treatment of extremely preterm infants, especially those born less than 25 weeks’ gestation, which can be dependent on the gestational days of the pregnancy( 8 ) and can influence care pathways with variation with in-utero transfer to a tertiary care centre or initiation of antenatal corticosteroids( 9 – 11 ). Difficult decisions around active treatment or the need to consider transfer could be influenced by outcome data, which will differ based on the characteristics of the pregnancy. To try and address these issues, national bodies have written guidance( 12 , 13 ) on the perinatal management of extremely preterm infants with emphasis on two key areas. Firstly, the need for an individualised approach in decision-making, taking into account multiple factors, moving away from gestational age or birth weight cut-offs alone. Secondly, the importance of supporting parental involvement in decision-making by providing clear antenatal information. The current recommended risk-based approach to perinatal decision-making( 14 ) is still subjective and open to misinterpretation. Although risk factors are listed in the recommendations( 12 , 13 ), it is unclear how these factors interact with one another to determine infant outcomes. Hence, a multivariable prediction tool combining key perinatal characteristics could provide an objective measure to support complex shared decision-making. Although several mortality prediction models have been developed( 2 , 15 – 17 ), none are used in routine clinical practice to predict individual outcomes or provide users with estimates of the importance of characteristics for that infant. Using population-based, routinely entered electronic patient record data in England and Wales and nine machine learning approaches, we aimed to develop and internally validate a mortality prediction tool for extremely preterm infants. We planned to compare our developed prediction tool with previously published models( 2 , 15 – 17 ) in a separate ‘test’ cohort, and externally validated the tool in an international cohort of preterm infants. Results Mortality prediction tool development Study population A total of 25,902 infants born between 23 +0 and 27 +6 weeks’ gestation were included in the study cohorts to develop and internally validate the prediction tool. Of these, 5,550 (21%) infants died before neonatal discharge. 1,642 (6%) infants with missing data for the nine predictors used in the modelling were excluded from the analysis ( S1 Table ). The proportion of deaths in infants in the ‘test’ cohort was lower than that in the ‘training and validation’ cohort (20% vs 22%). The infants in the ‘test’ cohort were also slightly more preterm and had a lower median birth weight z score. Mothers of infants in the ‘test’ cohort were more likely to have received a complete course of antenatal corticosteroids (ANC) and delivered in a centre with a co-located neonatal intensive care unit (NICU) ( Table 1 ). View this table: View inline View popup Table 1 Characteristics of infants in the cohorts used to develop, internally and externally validate the prediction model. IQR denotes interquartile range (25 th to 75 th percentile). NICU denotes Neonatal Intensive Care Unit. 1 Complete course of antenatal corticosteroids was defined as at least two doses of corticosteroids given before delivery. Algorithm development The nine machine learning approaches are described in S1 File . Across all machine learning approaches used, gestational age and birth weight z-score were consistently the two most important predictors based on the mean SHapley Additive exPlanations (SHAP) values ( Fig 1 ). Chorioamnionitis and congenital anomalies were the least important predictors ( S1 Fig, S2 Table ). Download figure Open in new tab Fig 1 Beeswarm plot showing the SHapley Additive exPlanations (SHAP) value of the predictors (ranked in the y-axis by their mean absolute SHAP values) for the overall ‘test’ cohort and by weeks of gestation based on the logistic regression model. Chorioamnionitis and congenital anomalies predictors were dropped from the logistic regression model after backward stepwise selection. Internal validation (discrimination, calibration and utility) Discrimination (Area under the receiver operating characteristics curve) There was little difference in the discrimination performance in the ‘test’ cohort in terms of area under the receiver operating characteristics curve (AUROC) between the seven approaches of extreme gradient boosting, feedforward neural network, random forest, long short-term memory, adaptive neuro-fuzzy inference system, AutoPrognosis 2.0, and logistic regression with AUROCs ranging from 0.746 to 0.757 ( Table 2 ). View this table: View inline View popup Download powerpoint Table 2 Model performance of the machine learning approaches and previously published models in the ‘test’ cohort (n=5,879). ROC denotes Receiver Operating Characteristic. Machine learning approaches are arranged in descending order of area under the ROC curve. Calibration (Calibration plot, calibration-in-the-large and calibration slope) Extreme gradient boosting, feedforward neural network, long short-term memory, and logistic regression had good calibration across the range of predicted mortality risk ( Table 2 , Fig 2 ). Download figure Open in new tab Fig 2 Calibration plot of the machine learning approaches in the ‘test’ cohort (N = 5,879). Perfect calibration is seen when the predicted value (x-axis) matches the observed average of mortality (y-axis), ie when the machine learning approach follows the diagonal grey line. Utility (Decision curve analysis) All the approaches, except for support vector machine, also had a similar superior net benefit compared with providing palliative/comfort care for all or no infants across a reasonable range of mortality threshold probabilities between 10% to 70% when deciding perinatal management in the decision curve analysis ( Fig 3 ). Download figure Open in new tab Fig 3 Decision curve analysis of the machine learning approaches in the ‘test’ cohort (N = 5,879). The threshold probability denotes the mortality risk threshold above which healthcare professionals would provide palliative/comfort care. Net benefit denotes the ‘net’ mortality cases detected per 100 infants. Net benefit is calculated as a weighted combination of true and false positives based on the threshold probability. The logistic regression model ( S3 Table ) was further developed in preference to the other machine learning approaches due to its familiarity among healthcare professionals and ease of interpretation, while achieving similar model performance in the ‘test’ cohort. Comparison with other published models Although the discrimination performance of the Manktelow 2013( 17 ) and Santhakumaran 2018( 2 ) models (AUROC 0.758 and 0.754, respectively) were similar to our developed logistic regression mortality prediction model (AUROC 0.746) in the ‘test’ cohort, they had worse calibration performance ( Table 2 and Fig 4 ). Our logistic regression model was also found to have superior net benefit across the range of mortality probability thresholds of 20% to 80% than the four previously published models ( Fig 4 ). Download figure Open in new tab Fig 4 (A) Calibration plot and (B) decision curve analysis comparing the logistic regression model with four previously published models in the ‘test’ cohort (N = 5,879). Subgroup analysis Of the included infants, 1,500 (26%) and 55 (1%) in the ‘test’ cohort had missing data on maternal ethnicity and neonatal network, respectively. The performance of the logistic regression mortality prediction model in infants born to mothers who identified as Black was worse than in infants born to mothers of other ethnicities. The AUROC was 0.713 with a 95% confidence interval (CI) of 0.655 to 0.770 ( S4 Table ) with an overestimation of mortality risk ( S2 Fig ) in infants born to mothers who identified as Black. However, the logistic regression model demonstrated superior net benefit across a range of mortality threshold probabilities between 10% to 50% in all four maternal ethnicities ( S3 Fig ). There was also variation in the logistic regression model performance across the 13 neonatal networks. In networks with lower neonatal mortality risk, such as Networks 3 and 10, an overestimation of neonatal mortality risk was observed ( S4 Table and S2 Fig ), and the net benefit of using the model was only found across a narrower range of mortality threshold probabilities ( S3 Fig ). External validation Study population Our logistic regression model was externally validated on the 1,051 infants recruited to the international, multicentre PLUSS trial( 19 ), after excluding seven infants born before 23 weeks’ gestation and one infant with missing data on prolonged rupture of membranes and chorioamnionitis. In this PLUSS cohort, 206 (20%) infants died before neonatal discharge. Infant characteristics are reported in Table 1 . Model performance The logistic regression model had similar discrimination in the ‘external validation’ cohort with an AUROC of 0.728 (95% CI 0.700 to 0.767) with an overestimation of mortality risk (calibration-in-the-large of −0.27 (95% CI −0.52 to −0.01) and calibration slope of 0.94 (95% CI 0.75 to 1.13)) ( Fig 5A ). The logistic regression model demonstrated superior net benefit across a range of mortality threshold probabilities between 10% to 40% ( Fig 5B ). Download figure Open in new tab Fig 5 (A) Calibration plot and (B) decision curve analysis of the developed logistic regression model in the ‘external validation’ cohort (N = 1,051). Online tool An online prediction tool based on the logistic regression algorithm was developed and can be accessed at https://premoutcome.shinyapps.io/Death/ ( Fig 6 ). Download figure Open in new tab Fig 6 Online prediction tool ( https://premoutcome.shinyapps.io/Death/ ) based on the logistic regression model with explainability provided using Local Interpretable Model-agnostic Explanations (LIME). In this example, the predicted risk (95% confidence interval) of mortality is 70% (67% to 72%) in a male infant born at 24 +0 weeks’ gestation, birth weight 600 grams, with no exposure to antenatal corticosteroids, and born in a centre with no co-located NICU. The predicted risk is driven by the gestation and no exposure to antenatal corticosteroids. The predicted risk decreases to 45% (43% to 47%) if the mother had received a complete course of antenatal corticosteroids (two doses of corticosteroids before birth) and given birth in a centre with a co-located NICU. Discussion We have developed and validated a risk prediction model, and produced an online mortality prediction tool for extremely preterm infants born 23 +0 to 27 +6 weeks’ gestation. We used perinatal factors based on the largest and most recent population-based real-world clinical data. Our tool is novel in providing information on how risk predictions are made for individual infants. The tool based on logistic regression demonstrates good model performance in the temporal internal validation cohort. The performance in the ‘external validation’ cohort was acceptable considering it was a cohort of infants from four different countries with strict eligibility criteria for clinical trial recruitment, and likely represents a slightly more restricted population with more immature and sicker infants than the whole extremely preterm population. Our tool also demonstrates a superior net benefit to previously published models( 2 , 15 – 17 ), especially across the mortality threshold probabilities of between 10 and 70%. This threshold range is likely to represent the circumstances when decision-making and parental discussions are more challenging( 13 ). Survival-focused care, i.e. resuscitation and subsequent intensive care, is likely to be provided to infants with a very low mortality risk, whereas comfort care may be provided to those infants with an extremely high mortality risk and low chance of survival. The logistic regression algorithm had similar overall model performance to the other machine learning approaches but offers a greater level of interpretability. The use of SHAP values to quantify predictors offers insights into the changing impact they have across the range of gestations. As expected, gestational age, birth weight z score and exposure to antenatal corticosteroids had the greatest impact on survival. Male sex has a negative effect on survival. Birth in a centre without an NICU, multiple pregnancy and prolonged rupture of membranes also negatively influence survival. The online tool allows users to explore the impact of modifiable factors, i.e. provision of antenatal corticosteroids and transfer to a centre with an NICU, to understand their influence on survival and so potentially support decision-making. The tool allows more individualised mortality prediction in the perinatal period. In addition to the support it could offer for decision-making, it could also have a role to play in research. Using the key characteristics which are important to the risk of infant death in this population could allow a more personalised approach to trial recruitment based on this risk and could be enhanced with the addition of other multi-omic biomarkers( 20 ). This would be particularly useful for high-risk treatments by allowing identification of patients most likely to benefit, whilst at the same time avoiding exposing those least likely to benefit, but where the adverse effects may carry significant risk. Mortality is predicted for the whole neonatal admission, but is based on early perinatal characteristics. The accuracy is therefore unlikely to significantly improve beyond the current routinely collected baseline data without the inclusion of important postnatal events during the NICU stay, and this likely explains why the different models show similar results. Further development of the tool will explore the addition of other important morbidities, such as severe bronchopulmonary dysplasia or intraventricular haemorrhage, and neurodevelopmental outcomes, which are of prime importance to parents and are often discussed during the perinatal counselling process( 13 , 21 , 22 ). Limitations Despite our tool demonstrating promising findings, further external validation and impact studies are needed before consideration for use in clinical practice to support shared clinical decision-making. At present, the dataset used to develop the tool does not include infants who died in the labour suite or obstetric theatre before admission to neonatal units. We used birth weight z scores, which are not known prior to birth, rather than estimated fetal weight z scores, which may be quite different to the actual birth weight, potentially leading to inaccurate mortality estimates. Temporal internal validation was used as it was computationally expensive to analyse the entire dataset using a bootstrapping approach. Our models did not account for birth year as a predictor, as we anticipated that these would be accounted for by the changes in perinatal predictors across the birth years. Furthermore, our tool demonstrated good model performance in the temporal internal validation despite the changes in perinatal characteristics and neonatal outcomes with time. Data inaccuracies and missing data could not be controlled for as the data were entered at the point of care. However, the dataset used to develop our tool encompasses all neonatal admissions of extremely preterm infants in England and Wales from 2012 and 2014, respectively. Hence, it reflected current perinatal practice with a diverse infant cohort representative of the national population. The variation in the model performance of our developed tool in different maternal ethnicities and neonatal networks must be interpreted with caution due to the smaller sample size and number of deaths within each subgroup, and the high missing data of 26% in maternal ethnicities. The variation in neonatal outcomes across different maternal ethnicities( 23 ) and neonatal networks( 24 ) may also partly explain some of the differences in model performance found. However, maternal ethnicity and region of care were not included as predictors, as our tool was intended to be used widely across high-income countries. Furthermore, the performance of our tool in different ethnicities and regions of care needs to be continuously monitored to avoid exacerbating ingrained inequalities in the current healthcare system. It is unclear how our tool performs compared to clinical judgement alone in predicting mortality in extremely preterm infants. Previously published studies demonstrated the complexity and variation of mortality risk assessment in extremely preterm infants among perinatal healthcare professionals. Healthcare professionals were found to underestimate ( 25 , 26 ) or overestimate ( 27 , 28 ) the mortality risk, demonstrating a need for an objective measure in risk assessment and consideration of care plans for extremely preterm infants. Conclusion Our individualised, online mortality prediction tool demonstrated the potential of applying machine learning approaches to population-based routinely recorded electronic patient data to predict neonatal outcomes in preterm infants. Further external validation and impact analysis studies are needed to understand if this could support parental discussion and shared perinatal clinical decision-making. Materials and methods Study population and data source Model development and internal validation Routinely entered electronic patient record data from a retrospective population-based cohort of extremely preterm infants born between 23 +0 and 27 +6 weeks’ gestation and admitted to all 185 neonatal units in England and Wales between January 2010 and December 2020 were extracted from the National Neonatal Research Database( 29 ). This represented over 90% of neonatal units in England in 2010 with complete coverage in England from 2012 and Wales from 2014. Infants with birth weight z score 4 based on the UK WHO preterm growth reference( 30 ) or discharged to non-participating neonatal units were excluded as these were potentially erroneous or incomplete entries. The study population was split into three cohorts depending on the birth year of the infants. Firstly, infants born from 2010 to 2015 (‘training’ cohort) were used to develop the mortality prediction models using different hyperparameters for each of the nine machine learning approaches. Secondly, infants born from 2016 to 2017 (‘validation’ cohort) were used to identify the optimal hyperparameters that produced the best discrimination performance for each of the machine learning approaches. Lastly, infants born from 2018 to 2020 (‘test’ cohort) were used to internally validate the performance of each of the machine learning approaches using the optimal hyperparameters identified previously( 31 ). The ‘test’ cohort was also used to compare the performance of our preferred final model implemented in our developed online mortality prediction tool with previously published models( 2 , 15 – 17 ). External validation The developed prediction tool was externally validated on infants born between 23 +0 and 27 +6 weeks’ gestation enrolled in the PLUSS trial( 19 ) from 21 NICUs in four countries (Australia, New Zealand, Singapore and Canada) from January 2018 to March 2023. Infants were recruited in the first 48 hours of age if they were either invasively ventilated or receiving non-invasive ventilation with a decision to treat with surfactant. Predictors Nine perinatal predictors were determined a-priori based on their clinical significance and association with mortality from the literature review( 32 , 33 ), framework of practice( 13 ) and national guidance( 12 ). No postnatal predictors nor biomarkers were included. The predictors recorded at or shortly after delivery were: (i) gestational age determined using the best obstetric estimate( 34 ); (ii) birth weight z-score as a measure of fetal growth; (iii) sex; (iv) multiple pregnancy; (v) exposure to complete (defined as at least two doses of ANC) or incomplete ANC; (vi) prolonged rupture of membranes of more than 18 hours; (vii) chorioamnionitis; (viii) major congenital anomaly based on the European Surveillance of Congenital Anomalies (EUROCAT) registry( 35 ); and (ix) born in a maternity centre with a co-located Level 3 NICU( 36 ). Definitions of these predictors are described in S5 Table . Due to the retrospective nature of the study, the assessment of the predictors and outcome was not blinded. Outcome Mortality was defined as death of the infant before discharge from neonatal units. Statistical Analysis The extracted dataset was cleaned using STATA 15( 37 ). Machine learning approaches The prediction models were developed in the ‘training and validation’ cohort using R version 4.3.2 in Rstudio( 38 ) and Python version 3.11.7 in Jupyter Notebook( 39 ) for nine machine learning approaches. The nine approaches were Logistic Regression, AutoPrognosis 2.0, Adaptive Neuro-Fuzzy Inference System, Extreme Gradient Boosting, Feedforward Neural Network, K-Nearest Neighbour, Long Short-Term Memory, Random Forest and Radial Kernel Support Vector Machine ( Table 2 ). For the latter seven machine learning approaches, the dataset was z-score normalised based on the training data values to ensure stability and speed up the convergence of the algorithms. A complete case analysis was used for all approaches, excluding infants with missing data for the predictor variables. Further details of the modelling approaches are described in S1 File . For each of the nine machine learning approaches, the importance of each predictor to the prediction in the ‘test’ cohort was determined using the mean SHAP values. These were derived from Kernel SHAP( 40 ) using the kernelshap package in R( 41 ) and a sample of 100 infants to calculate the marginal expectation. As a subgroup analysis, the importance of the predictors to the prediction of the best-performing algorithm was determined for each gestational week in the ‘test’ cohort using mean SHAP values. Model Performance Performance of the nine machine learning approaches was assessed by temporal internal validation in the ‘test’ cohort across the three domains of (i) discrimination (AUROC)( 42 ), (ii) calibration (calibration plot, calibration-in-the-large and calibration slope)( 43 ), and (iii) utility measures (decision curve analysis)( 44 ). Decision curve analysis is a decision-analytic measure to summarise the ‘net’ benefit of the model in detecting neonatal death across a range of mortality risk thresholds above which healthcare professionals would provide palliative/comfort care (threshold probability). The performance of the previously developed mortality prediction models (Tyson 2008( 18 ), Manktelow 2013( 17 ), Santhakumaran 2018( 2 ) and NICHD 2020( 16 )) was also assessed in the ‘test’ cohort using published regression coefficients for the first three models and the freely available online NICHD 2020( 16 ) tool. Only infants within the ‘test’ cohort who fulfilled the inclusion criteria for each of the previously published models were used to assess their performance. Subgroup analysis was also performed to assess the performance of the best-performing algorithm across different maternal ethnicities and regions where perinatal care was provided (neonatal networks). The performance of the best-performing algorithm was then externally validated on the PLUSS trial( 19 ) cohort of infants. Online prediction tool An online prediction tool was developed based on the best-performing algorithm using the shiny package in R( 45 ) and deployed on shinyapps.io. The tool displays the predicted mortality risk for each infant alongside the predicted mortality risk if the modifiable predictors of exposure to antenatal corticosteroids and being born in a centre with a co-located NICU were optimised. Explainability of the individual predictions made was provided using the Local Interpretable Model-Agnostic Explanation (LIME) package in R( 46 ). The ‘training and validation’ cohort was used to train the LIME explainer. Once the LIME explainer is created, 5,000 permutations using forward selection in a ridge regression model and Gower’s distance were used to explain how the individual predictions were made. The tool was co-designed with parents of ex-preterm infants through two focus group sessions organised with the Spoons charity. Ethics and Dissemination Ethical approval was granted by the Sheffield Research Ethics Committee (REC reference 19/YH/0115). The study was reported using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis AI (TRIPOD+AI) checklist( 47 ) and the Transparency, Reproducibility, Ethics and Effectiveness (TREE) framework( 48 ). The PLUSS Trial Steering Committee approved the use of grouped trial data for the external validation. Data availability Professor Don Sharkey had full access to the NNRD data in the study. The data that support the findings of this study are available from the Neonatal Data Analysis Unit, but restrictions apply to the availability of these data, which were used under license for the current study and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the Neonatal Data Analysis Unit. The full study protocol and analytical code are also available from the authors upon reasonable request. Professor Brett Manley and Dr Kate Francis had full access to the PLUSS trial data. Funding TCK received the Action Medical Research training fellowship ( https://action.org.uk/ ) supported by the Albert Gubay Foundation, as part of this study. DS was funded by the National Institute for Health and Care Research (NIHR) Children and Young People MedTech Co-operative (CYP MedTech) ( https://hrc-children.nihr.ac.uk/ ). BJM was funded by the National Health and Medical Research Council (Australia) ( https://www.nhmrc.gov.au/ ). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. The funders play no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Author contributions TCK and DS conceptualised the study. TCK, CC and JV analysed and performed the machine learning approaches. KM, TCK and DS ran the two parent focus group sessions. KLF performed the external validation analysis on the PLUSS trial cohort of infants. CJDM and BJM are the chief investigators for the PLUSS trial. All authors read and approved the final manuscript. Competing interest All authors declare no financial or non-financial competing interests. Supporting Information S1 Table Table demonstrating the characteristics of infants with complete data included in developing and validating the prediction model, and excluded infants with missing data for birth weight z score, multiple pregnancy, exposure to antenatal corticosteroids or born in a centre with neonatal intensive care unit (NICU). IQR denotes interquartile range. S1 File File detailing the model development process for each of the nine machine learning approaches. S2 Table Table demonstrating the importance of the predictors to the predictions in the testing dataset based on mean SHapley Additive exPlanations (SHAP) values. NICU denotes Neonatal Intensive Care Unit. PROM denotes Prolonged rupture of membranes. 1 Chorioamnionitis and congenital anomalies predictors were dropped from the logistic regression model after backward stepwise selection 2 Sex and Multiple pregnancy predictors were used to group the dataset in the Adaptive Neuro-Fuzzy Inference System. S1 Fig Figure demonstrating the importance of the predictors to the predictions of the logistic regression model in the ‘test’ cohort based on mean SHapley Additive exPlanations (SHAP) values for the overall cohort and by gestational week groups. S3 Table Table presenting the regression coefficients with 95% confidence intervals (CI) of the final logistic regression model to predict death before neonatal discharge. Regression coefficient denotes estimated change in the log odds of death when the associated predictor increases by one unit or in comparison with a reference category. NICU denotes Neonatal Intensive Care Unit. S4 Table Table describing the incidence of death and the model performance of the logistic regression approach in the ‘test’ cohort (N = 5,879) stratified by (A) maternal ethnicity (N = 4,379) and (B) neonatal network (N = 5,824). ROC = Receiver Operating Characteristic. 1,500 (26%) and 55 (1%) of infants with missing data on maternal ethnicity and neonatal network were excluded. S2 Fig Calibration plot of the logistic regression approach in the ‘test’ cohort (N = 5,879) stratified by (A) maternal ethnicity (N = 4,379) and (B) neonatal network (N = 5,824). S3 Fig Decision curve analysis of the logistic regression approach in the ‘test’ cohort (N = 5,879) stratified by (A) maternal ethnicity (N = 4,379) and (B) neonatal network (N = 5,824). S5 Table Table describing the definition of variables extracted from the National Neonatal Research Database. S6 Table Participating neonatal units in England and Wales and their respective lead clinicians. The list was accessed from https://www.imperial.ac.uk/neonatal-data-analysis-unit/neonatal-data-analysis-unit/list-of-national-neonatal-units/ on 06/01/2022. Acknowledgements Electronic patient data recorded at participating neonatal units are transmitted to the Neonatal Data Analysis Unit (NDAU) to form the NNRD ( S6 Table ). We are grateful to all the families that agreed to include their infant’s data in the NNRD, the health professionals who recorded data and the NDAU team. We are also grateful to the Spoons charity and all parents who participated in the focus group sessions in co-designing the online tool. We would also like to thank all the families and infants who participated in the PLUSS trial and the health professionals who cared for them during their neonatal stay. We would like to acknowledge the significant contribution of the PLUSS trial investigators (Omar F. Kamlin, Jeanie L. Y. Cheong, Jennifer A. Dawson, Susan E. Jacobs, Lex W. Doyle, Peter G. Davis (Royal Women’s Hospital, Melbourne, Australia), Susan M. Donath (Murdoch Children’s Research Institute, Melbourne, Australia), Peter A. Dargaville (Menzies Institute for Medical Research, Hobart, Australia), Pita Birch (Mater Mother’s Hospitals, Brisbane, Australia), Steven M. Resnick (King Edward Memorial Hospital, Perth, Australia), Georg M. Schmölzer, Brenda Law (Centre for the Studies of Asphyxia and Resuscitation, Edmonton, Canada), Risha Bhatia (MonashChildren’sHospital, Melbourne, Australia), Katinka P. Bach (Newborn Services, Starship Child Health, Auckland, New Zealand), Koertde Waal, Javeed N. Travadi (John Hunter Children’s Hospital, Newcastle, Australia), Pieter J. Koorts (Royal Brisbane and Women’s Hospital, Brisbane, Australia), Mary J. Berry (University of Otago, Wellington, New Zealand), Kei Lui (Royal Hospital for Women, New South Wales, Australia), Victor S. Rajadurai, Suresh Chandran (KK Women’s and Children’s Hospital, Singapore) Martin Kluckow (Royal North Shore Hospital, New South Wales, Australia), Elza Cloete (Christchurch Women’s Hospital, Christchurch, New Zealand), Margaret M. Broom (Canberra Hospital, Canberra, Australian), Michael J. Stark (The Women’s and Children’s Hospital, Adelaide, Australia), Adrienne Gordon (Royal Prince Alfred Hospital, New South Wales, Australia), Vinayak Kodur (Te Whatu Ora Waikato, Hamilton, New Zealand)). References 1. ↵ Cao G , Liu J , Liu M . Global, Regional, and National Incidence and Mortality of Neonatal Preterm Birth, 1990-2019 . JAMA Pediatrics . 2022 ; 176 ( 8 ): 787 - 96 . OpenUrl PubMed 2. ↵ Santhakumaran S , Statnikov Y , Gray D , Battersby C , Ashby D , Modi N , et al. Survival of very preterm infants admitted to neonatal care in England 2008-2014: time trends and regional variation . Arch Dis Child Fetal Neonatal Ed. 2018 ; 103 ( 3 ): F208 - F15 . OpenUrl Abstract / FREE Full Text 3. ↵ Kwok TC , Poulter C , Algarni S , Szatkowski L , Sharkey D . Respiratory management and outcomes in high-risk preterm infants with development of a population outcome dashboard . Thorax . 2023 . 4. ↵ Lui K , Lee SK , Kusuda S , Adams M , Vento M , Reichman B , et al. Trends in Outcomes for Neonates Born Very Preterm and Very Low Birth Weight in 11 High-Income Countries . J Pediatr . 2019 . 5. ↵ Norman M , Hallberg B , Abrahamsson T , Björklund LJ , Domellöf M , Farooqi A , et al. Association Between Year of Birth and 1-Year Survival Among Extremely Preterm Infants in Sweden During 2004-2007 and 2014-2016 . JAMA . 2019 ; 321 ( 12 ): 1188 – 99 . OpenUrl CrossRef PubMed 6. ↵ Nuffield Council on Bioethics. Critical care decisions in fetal and neonatal medicine: ethical issues . November 2006 . https://www.nuffieldbioethics.org/publications/neonatal-medicine-and-care . [Date last accessed: October 12 2020 ]. 7. ↵ Sullivan A , Arzuaga B , Luff D , Ward E , Williams DN , Cummings C . Advice to Clinicians From Expectant Parents at Extreme Prematurity: A Multimethod Study . Pediatrics . 2024 ; 153 ( 3 ): e2023062178 . OpenUrl PubMed 8. ↵ Rysavy MA , Li L , Bell EF , Das A , Hintz SR , Stoll BJ , et al. Between-Hospital Variation in Treatment and Outcomes in Extremely Preterm Infants . New England Journal of Medicine . 2015 ; 372 ( 19 ): 1801 – 11 . OpenUrl CrossRef PubMed 9. ↵ Hannah GG , Alexis S , Michael JS , Stefan CK , Jeanie LYC , Calum TR , et al. In-utero transfer, survival-focused care and survival to 28-days at 22-24 weeks’ gestation pre- and post-implementation of an extreme prematurity management guideline in Victoria , Australia. BMJ Paediatrics Open . 2024 ; 8 ( 1 ): e002462 . OpenUrl PubMed 10. Beltempo , M. , Mukerji , A. , Yoon , E. W. , Goswami , N. , and Members of the Annual Report Review Committee ( 2024 ) The Canadian Neonatal Network 2023 Annual Report . Available at: https://www.canadianneonatalnetwork.org/portal/Portals/0/Annual%20Reports/2023%20CNN%20Annual%20Report.pdf [Accessed 08 May 2025 ]. 11. ↵ Smith LK , van Blankenstein E , Fox G , Seaton SE , Martínez-Jiménez M , Petrou S , et al. Effect of national guidance on survival for babies born at 22 weeks’ gestation in England and Wales: population based cohort study . BMJ Medicine . 2023 ; 2 ( 1 ): e000579 . OpenUrl Abstract / FREE Full Text 12. ↵ Lemyre B , Moore G . Counselling and management for anticipated extremely preterm birth . Paediatrics and Child Health (Canada) . 2017 ; 22 ( 6 ): 334 – 50 . OpenUrl 13. ↵ Mactier H , Bates SE , Johnston T , Lee-Davey C , Marlow N , Mulley K , et al. Perinatal management of extreme preterm birth before 27 weeks of gestation: a framework for practice . Arch Dis Child Fetal Neonatal Ed. 2020 ; 105 ( 3 ): 232 - 9 . OpenUrl FREE Full Text 14. ↵ Cummings J , Committee On F , Newborn , Watterberg K , Eichenwald E , Poindexter B , et al. Antenatal Counseling Regarding Resuscitation and Intensive Care Before 25 Weeks of Gestation . Pediatrics . 2015 ; 136 ( 3 ): 588 – 95 . OpenUrl CrossRef PubMed 15. ↵ National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN). Extremely preterm birth outcome data . https://www.nichd.nih.gov/about/org/der/branches/ppb/programs/epbo/Pages/epbo_case.aspx . [Date last accessed: October 12 2020 ]. 16. ↵ Rysavy MA , Horbar JD , Bell EF , Li L , Greenberg LT , Tyson JE , et al. Assessment of an Updated Neonatal Research Network Extremely Preterm Birth Outcome Model in the Vermont Oxford Network . JAMA Pediatr . 2020 : e196294 . 17. ↵ Manktelow BN , Seaton SE , Field DJ , Draper ES . Population-Based Estimates of In-Unit Survival for Very Preterm Infants . Pediatrics . 2013 ; 131 ( 2 ): E425 – E32 . OpenUrl PubMed 18. ↵ Tyson JE , Parikh NA , Langer J , Green C , Higgins RD , Natl Inst Child H , et al. Intensive care for extreme prematurity - Moving beyond gestational age. New England Journal of Medicine . 2008 ; 358 ( 16 ): 1672 – 81 . OpenUrl PubMed 19. ↵ Manley BJ , Kamlin COF , Donath SM , Francis KL , Cheong JLY , Dargaville PA , et al. Intratracheal Budesonide Mixed With Surfactant for Extremely Preterm Infants: The PLUSS Randomized Clinical Trial . Jama . 2024 . 20. ↵ Newnham JP , Kemp MW , White SW , Arrese CA , Hart RJ , Keelan JA . Applying Precision Public Health to Prevent Preterm Birth . Front Public Health . 2017 ; 5 : 66 . 21. ↵ Gallagher K , Shaw C , Parisaei M , Marlow N , Aladangady N . Attitudes About Extremely Preterm Birth Among Obstetric and Neonatal Health Care Professionals in England: A Qualitative Study . JAMA Network Open . 2022 ; 5 ( 11 ): e2241802-e . OpenUrl 22. ↵ Wilkinson D. Chapter 4. Who should decide for critically ill neonates and how? The grey zone in neonatal treatment decisions. In: McDougall R, Delany C, Gillam L, editors . When Doctors and Parents Disagree: Ethics, Paediatrics & the Zone of Parental Discretion . Sydney, Australia : The Federation Press ; 2016 . 23. ↵ Draper ES , Gallimore ID , Smith LK , Matthews RJ , Fenton AC , Kurinczuk JJ , Smith PW , Manktelow BN , on behalf of the MBRRACE-UK Collaboration . MBRRACE-UK Perinatal Mortality Surveillance, UK Perinatal Deaths for Births from January to December 2021: State of the Nation Report . Leicester : The Infant Mortality and Morbidity Studies, Department of Population Health Sciences, University of Leicester . 2023 . 24. ↵ National Neonatal Audit Programme (NNAP) Project Board. National Neonatal Audit Programme Annual Report 2023 - on 2022 data. Royal College of Paediatrics and Child Health . October 2023 . 25. ↵ Wood K , Di Stefano LM , Mactier H , Bates SE , Wilkinson D . Individualised decision making: interpretation of risk for extremely preterm infants—a survey of UK neonatal professionals . Archives of Disease in Childhood - Fetal and Neonatal Edition . 2022 ; 107 ( 3 ): 281 – 8 . OpenUrl 26. ↵ Chan KL , Kean LH , Marlow N . Staff views on the management of the extremely preterm infant . European Journal of Obstetrics & Gynecology and Reproductive Biology . 2006 ; 128 ( 1 ): 142 – 7 . OpenUrl CrossRef PubMed Web of Science 27. ↵ Rosemarie AB , Peter GD , Jennifer AD , Lex WD , for the Victorian Infant Collaborative Study G. Predicting death or major neurodevelopmental disability in extremely preterm infants born in Australia . Archives of Disease in Childhood - Fetal and Neonatal Edition . 2013 ; 98 ( 3 ): F201 . OpenUrl 28. ↵ Blanco F , Suresh G , Howard D , Soll RF . Ensuring accurate knowledge of prematurity outcomes for prenatal counseling . Pediatrics . 2005 ; 115 ( 4 ): e478 – 87 . OpenUrl CrossRef PubMed 29. ↵ Gale C , Morris I , Board NDAUNS. The UK National Neonatal Research Database: using neonatal data for research, quality improvement and more . Arch Dis Child Educ Pract Ed. 2016 ; 101 ( 4 ): 216 - 8 . OpenUrl FREE Full Text 30. ↵ Scientific Advisory Committee on Nutrition (SACN) . Application of WHO growth standards in the UK 2007 . London : Stationery Office . 2008 . 31. ↵ Steyerberg EW , Harrell FE , Jr . . Prediction models need appropriate internal, internal-external, and external validation . J Clin Epidemiol . 2016 ; 69 : 245 – 7 . OpenUrl CrossRef PubMed 32. ↵ Medlock S , Ravelli AC , Tamminga P , Mol BW , Abu-Hanna A . Prediction of mortality in very premature infants: a systematic review of prediction models . PLoS One . 2011 ; 6 ( 9 ): e23441 . OpenUrl CrossRef PubMed 33. ↵ Del Río R , Thió M , Bosio M , Figueras J , Iriondo M . [Prediction of mortality in premature neonates. An updated systematic review] . An Pediatr (Barc) . 2020 . 34. ↵ National Institute for Health and Care Excellence (NICE). Antenatal care for uncomplicated pregnancies (Clinical Guideline CG62) . February 2019 . https://www.nice.org.uk/guidance/cg62 . [Date last accessed: December 1 2020 ]. 35. ↵ European Surveillance of Congenital Anomalies (EUROCAT). EUROCAT Guide 1.4 and Reference Documents . 2013 . http://www.eurocat-network.eu/ . [Date last accessed: February 12 2020 ]. 36. ↵ NHS England. Neonatal Critical Care . 2015 . https://www.england.nhs.uk/commissioning/spec-services/npc-crg/group-e/e08/ . [Date last accessed: January 30 2021 ]. 37. ↵ StataCorp . 2017 . Stata Statistical Software: Release 15 . College Station, TX : StataCorp LLC . 38. ↵ R Core Team ( 2020 ). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria . URL https://www.R-project.org/ . 39. ↵ Kluyver T , Ragan-Kelley B , Perez F , Granger B , Bussonnier M , Frederic J , et al. Jupyter Notebooks - a publishing format for reproducible computational workflows. Loizides F, Scmidt B, editors . Netherlands : IOS Press ; 2016 . 40. ↵ Lundberg SM , Lee S-I , editors. A Unified Approach to Interpreting Model Predictions . Neural Information Processing Systems ; 2017 . 41. ↵ Mayer M , Watson D. kernelshap: Kernel SHAP. R package version 0.4.1 . https://CRAN.R-project.org/package=kernelshap 2023. 42. ↵ Hanley JA , McNeil BJ . The meaning and use of the area under a receiver operating characteristic (ROC) curve . Radiology . 1982 ; 143 ( 1 ): 29 – 36 . OpenUrl CrossRef PubMed Web of Science 43. ↵ Huang Y , Li W , Macheret F , Gabriel RA , Ohno-Machado L . A tutorial on calibration measurements and calibration models for clinical prediction models . J Am Med Inform Assoc . 2020 ; 27 ( 4 ): 621 – 33 . OpenUrl PubMed 44. ↵ Vickers AJ , Elkin EB . Decision curve analysis: a novel method for evaluating prediction models . Med Decis Making . 2006 ; 26 ( 6 ): 565 – 74 . OpenUrl CrossRef PubMed Web of Science 45. ↵ Chang W , Cheng J , Allaire J , Sievert C , Schloerke B , Xie Y , et al. shiny: Web Application Framework for R. R package version 1.8.1.9001 , https://github.com/rstudio/shiny , https://shiny.posit.co/ . 2024. 46. ↵ Hvitfeldt E , Pedersen T , Benesty M. lime: Local Interpretable Model-Agnostic Explanations. R package version 0.5.3 , https://CRAN.R-project.org/package=lime . 2022 . 47. ↵ Gary SC , Karel GMM , Paula D , Richard DR , Andrew LB , Ben Van C , et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods . BMJ . 2024 ; 385 : e078378 . OpenUrl FREE Full Text 48. ↵ Vollmer S , Mateen BA , Bohner G , Király FJ , Ghani R , Jonsson P , et al. Machine learning and artificial intelligence research for patient benefit: 20 critical questions on transparency, replicability, ethics, and effectiveness . BMJ . 2020 ; 368 : l6927 . View the discussion thread. Back to top Previous Next Posted July 10, 2025. Download PDF Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Developing and validating an explainable digital mortality prediction tool for extremely preterm infants Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Developing and validating an explainable digital mortality prediction tool for extremely preterm infants T’ng Chang Kwok , Chao Chen , Jayaprakash Veeravalli , Carol AC Coupland , Edmund Juszczak , Jonathan Garibaldi , Kirsten Mitchell , Kate L Francis , Christopher J D McKinlay , Brett J Manley , Don Sharkey medRxiv 2025.07.09.25331175; doi: https://doi.org/10.1101/2025.07.09.25331175 Share This Article: Copy Citation Tools Developing and validating an explainable digital mortality prediction tool for extremely preterm infants T’ng Chang Kwok , Chao Chen , Jayaprakash Veeravalli , Carol AC Coupland , Edmund Juszczak , Jonathan Garibaldi , Kirsten Mitchell , Kate L Francis , Christopher J D McKinlay , Brett J Manley , Don Sharkey medRxiv 2025.07.09.25331175; doi: https://doi.org/10.1101/2025.07.09.25331175 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Pediatrics Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4425) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15221) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6588) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1368) Health Systems and Quality Improvement (1612) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6588) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5440) Public and Global Health (9219) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (710) Sports Medicine (529) Surgery (710) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffb738feb44dfa9',t:'MTc3OTQ0OTAzNA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();

Text is read by the "Ask this paper" AI Q&A widget below. Extraction quality varies by source — PMC NXML preserves structure cleanly, OA-HTML may include some navigation residue, and OA-PDF can have broken hyphenation. The publisher copy (via DOI) is the canonical version.

My notes (saved in your browser only)

Ask this paper AI returns verbatim quotes from the full text · source: preprint-html

Answers must be backed by verbatim quotes from this paper's full text. Hallucinated quotes are dropped automatically; if no verbatim passage answers the question, we say so. How this works

Citation neighborhood (no data yet)

We don't have any in-corpus citations linked to this paper yet. This is a recent paper (2025) — citers typically take a year or two to land, and the OpenAlex reference graph may still be filling in.

Source provenance

europepmc
last seen: 2026-05-20T01:45:00.602351+00:00