Practical Prediction Model Research Articles

Heritage protection and tourism income: the tourism heritage Kuznets curve

Purpose Protecting a nation’s heritage is an essential element for sustainable development. Heritage attracts tourism and drives tourist income, as well as bringing wider economic, social and environmental benefits. This study aims to improve understanding of the relationship between heritage protection spending in countries and economic development via tourism revenues via economic logic and the Kuznets curve. Design/methodology/approach This study applies economic modelling to test for the presence of a Kuznets curve. This study uses rare available matched data on heritage protection spending from UNESCO and on travel and tourism total contribution to gross domestic product (GDP) from the World Travel and Tourism Council. Findings This paper finds evidence of a negative quadratic relationship between heritage protection spending per capita and tourism GDP per capita, whereby increases in heritage spending initially lead to more tourism income (typically less developed nations), but then tourism spending increases more rapidly for smaller increases heritage spending on top of this initial investment (typically more developed nations). No countries are yet at the stage where they reap high tourism income from much lower levels of heritage spending. Research limitations/implications The findings suggest that the future of heritage tourism will be very different to that of today; future tourism is likely to develop significantly through innovations in management approach and advanced technologies to manage the balance between demand and heritage degradation. Practical implications The predictive model allows us to plot future possibilities for heritage tourism, which in turn enables more informed decision-making regarding the planning and management of future heritage resources and tourism demand. Tourism heritage institutions and their benefactors should further plan and invest in innovations (such as technologies and management practices) to advanced economically and move along the tourism heritage Kuznets curve. Originality/value To the best of the author’s knowledge, this paper formally tests, for the first time, for the presence of a Kuznets curve in the relationship between heritage protection spending and tourism income. It provides original evidence of the value of applying the underlying theory of the Kuznets curve in a tourism context.

Management and Policy Modeling of the Market Using Artificial Intelligence

This paper investigates the market management and modeling based on advanced artificial intelligence. The proposed model deploys the combination of the support vector machine and fuzzy set theory to provide a practical and powerful prediction model for the market price over the next day. A realistic and effective model is then introduced to model the market players, such as the renewable energy sources of solar and wind turbines, as well as the fossil-fueled sources of micro turbines and fuel cells. In order to provide an optimal management program, it introduces a stochastic framework based on the point estimate method and adaptive grey wolf optimization algorithm (GWO). The proposed optimization methods use an adaptive strategy to choose the most fitting modification for enhancing the GWO performance. A realistic scenario is simulated to demonstrate the model’s effectiveness and impression on the real market management. The results clearly show the effectiveness of the prediction and management model. The prediction results show the superiority of the proposed model by RMSE of 2.9643 compared to the 3.217 for SVR, 3.2364 for ANN and 3.0621 for the grey model. Moreover, the optimal MAPE is 2.7453 by the proposed method, which is much better than the 3.052 by SVR, 3.1552 by ANN and 2.9285 by the grey model. From point of view of optimization, the most fitting power dispatch has been attained with the total cost of 300.8632 over 24 h.

Productivity evaluation model for multi-cluster fractured wells based on volumetric source method

Multi-cluster fracturing in horizontal wells is a widely performed key technology for the development of low permeability and tight oil reservoirs. However, it is difficult to predict the oil productivity after fracturing efficiently and accurately. A practical productivity prediction model of fractured horizontal wells based on volume source method is proposed. The model comprehensively considers the effects of nonlinear seepage, reservoir physical properties, cluster distribution mode and non-uniform fracture length on productivity. The reliability of the proposed model is verified by field data. The conclusion shows that the cumulative production of fractured horizontal wells increases with the increase of permeability and storage capacity coefficient, and decreases with the increase of starting pressure gradient (SPG) and stress sensitivity coefficient (SSC). With the increase of the number of fractures, the influence of nonlinear seepage characteristics on cumulative production (CP) intensifies. Increasing the crack length can reduce the impact of production decline caused by stress sensitivity effect, but has little impact on the starting pressure gradient. With the increase of dimensionless conductivity, the decrease of CP caused by starting pressure gradient and stress sensitivity effect gradually increases. Under the same fracture parameter, the influence of stress sensitivity effect is stronger than that of starting pressure gradient effect on CP. In the segmented clustering mode, the multi-cluster fracturing productivity decreases gradually with the decrease of segment spacing and the increase of cluster spacing. The cumulative production of multi-cluster fracturing with non-uniform spacing and non-uniform fracture length distribution mode is less than that with equal spacing and equal fracture length mode. This model eliminates the problems of singularity of point source function, slow convergence speed or non-convergence of calculation. The nonlinear percolation characteristics of tight reservoirs are taken into account in this prediction model, which has the advantages of fast convergence and high calculation efficiency. The findings of this study can help for better understanding of productivity prediction of fractured horizontal wells, which has important guiding significance for the optimization of fracturing parameters.

Development and Internal Validation of the ARISE Prediction Models for Rebleeding After Aneurysmal Subarachnoid Hemorrhage.

Aneurysmal rerupture is one of the most important determents for outcome after aneurysmal subarachnoid hemorrhage and still occurs frequently because individual risk assessment is challenging given the heterogeneity in patient characteristics and aneurysm morphology. To develop and internally validate a practical prediction model to estimate the risk of aneurysmal rerupture before aneurysm closure. We designed a multinational cohort study of 2 prospective hospital registries and 3 retrospective observational studies to predict the risk of computed tomography confirmed rebleeding within 24 and 72 hours after ictus. We assessed predictors with Cox proportional hazard regression analysis. Rerupture occurred in 269 of 2075 patients. The cumulative incidence equaled 7% and 11% at 24 and 72 hours, respectively. Our base model included hypertension, World Federation of Neurosurgical Societies scale, Fisher grade, aneurysm size, and cerebrospinal fluid drainage before aneurysm closure and showed good discrimination with an optimism corrected c-statistic of 0.77. When we extend the base model with aneurysm irregularity, the optimism-corrected c-statistic increased to 0.79. Our prediction models reliably estimate the risk of aneurysm rerupture after aneurysmal subarachnoid hemorrhage using predictor variables available upon hospital admission. An online prognostic calculator is accessible at https://www.evidencio.com/models/show/2626 .

Practical Prediction Model for Ovarian Insufficiency after Radiation.

The present study aimed to develop a useful mathematical model that predicts the age at which premature ovarian insufficiency might occur after teletherapy radiation. A diagnosis of premature or early menopause has physical and psychological consequences, so women may need support and long-term medical follow-up. To correlate ovarian radiation dose with ovarian function, we used the formula described by Wallace et al.: √g(z) = 10(2-0,15z), where "g(z)" and "z" represent oocyte survival rate and the radiation dose (in Gray), respectively. By simulating different ages and doses, we observed a pattern that could be used to simplify the relationship between radiation dose and remaining time of ovarian function. We obtained a linear function between ovarian radiation dose and loss of ovarian function (LOF) that is the percentage of decrease in the time to the ovarian failure compared with the time expected for a woman at the same age without irradiation exposition. For patients < 40 years old and with ovarian radiation doses < 5 Gy, the equation LOF = 2.70 + (11.08 x Dose) can be applied to estimate the decrease in time to premature ovarian insufficiency. The present study reports a practicable theoretical method to estimate the loss of ovarian function. These findings can potentially improve the management and counseling of young women patients submitted to radiotherapy during their reproductive years.

Use of unstructured text in prognostic clinical prediction models: a systematic review.

ObjectiveThis systematic review aims to assess how information from unstructured text is used to develop and validate clinical prognostic prediction models. We summarize the prediction problems and methodological landscape and determine whether using text data in addition to more commonly used structured data improves the prediction performance.Materials and MethodsWe searched Embase, MEDLINE, Web of Science, and Google Scholar to identify studies that developed prognostic prediction models using information extracted from unstructured text in a data-driven manner, published in the period from January 2005 to March 2021. Data items were extracted, analyzed, and a meta-analysis of the model performance was carried out to assess the added value of text to structured-data models.ResultsWe identified 126 studies that described 145 clinical prediction problems. Combining text and structured data improved model performance, compared with using only text or only structured data. In these studies, a wide variety of dense and sparse numeric text representations were combined with both deep learning and more traditional machine learning methods. External validation, public availability, and attention for the explainability of the developed models were limited.ConclusionThe use of unstructured text in the development of prognostic prediction models has been found beneficial in addition to structured data in most studies. The text data are source of valuable information for prediction model development and should not be neglected. We suggest a future focus on explainability and external validation of the developed models, promoting robust and trustworthy prediction models in clinical practice.

Prediction of hypertension using traditional regression and machine learning models: A systematic review and meta-analysis.

We aimed to identify existing hypertension risk prediction models developed using traditional regression-based or machine learning approaches and compare their predictive performance. We systematically searched MEDLINE, EMBASE, Web of Science, Scopus, and the grey literature for studies predicting the risk of hypertension among the general adult population. Summary statistics from the individual studies were the C-statistic, and a random-effects meta-analysis was used to obtain pooled estimates. The predictive performance of pooled estimates was compared between traditional regression-based models and machine learning-based models. The potential sources of heterogeneity were assessed using meta-regression, and study quality was assessed using the PROBAST (Prediction model Risk Of Bias ASsessment Tool) checklist. Of 14,778 articles, 52 articles were selected for systematic review and 32 for meta-analysis. The overall pooled C-statistics was 0.75 [0.73-0.77] for the traditional regression-based models and 0.76 [0.72-0.79] for the machine learning-based models. High heterogeneity in C-statistic was observed. The age (p = 0.011), and sex (p = 0.044) of the participants and the number of risk factors considered in the model (p = 0.001) were identified as a source of heterogeneity in traditional regression-based models. We attempted to provide a comprehensive evaluation of hypertension risk prediction models. Many models with acceptable-to-good predictive performance were identified. Only a few models were externally validated, and the risk of bias and applicability was a concern in many studies. Overall discrimination was similar between models derived from traditional regression analysis and machine learning methods. More external validation and impact studies to implement the hypertension risk prediction model in clinical practice are required.

Novel Predictors of Stroke-Associated Pneumonia: A Single Center Analysis.

Stroke-associated pneumonia (SAP) is a common cause of disability or death. Although the researches on SAP have been relatively mature, the method that can predict SAP with great accuracy has not yet been determined. It is necessary to discover new predictors to construct a more accurate predictive model for SAP. We continuously collected 2,366 patients with acute ischemic stroke, and then divided them into the SAP group and non-SAP group. Data were recorded at admission. Through univariate analyses and multivariate regression analyses of the data, the new predictive factors and the predictive model of SAP were determined. The receiver operating characteristic (ROC) curve and the corresponding area under the curve (AUC) were used to measure their predictive accuracy. Of the 2,366 patients, 459 were diagnosed with SAP. International normalized ratio (INR) (odds ratio = 37.981; 95% confidence interval, 7.487–192.665; P < 0.001), age and dysphagia were independent risk factors of SAP. However, walking ability within 48 h of admission (WA) (odds ratio = 0.395; 95% confidence interval, 0.287–0.543; P < 0.001) was a protective factor of SAP. Different predictors and the predictive model all could predict SAP (P < 0.001). The predictive power of the model (AUC: 0.851) which included age, homocysteine, INR, history of chronic obstructive pulmonary disease (COPD), dysphagia, and WA was greater than that of age (AUC: 0.738) and INR (AUC: 0.685). Finally, we found that a higher INR and no WA could predict SAP in patients with acute ischemic stroke. In addition, we designed a simple and practical predictive model for SAP, which showed relatively good accuracy. These findings might help identify high-risk patients with SAP and provide a reference for the timely use of preventive antibiotics.

Construction of a Tool to Predict Overall Survival of Patients With Primary Spinal Tumors After Surgical Resection: A Real-World Analysis Based on the Surveillance, Epidemiology, and End Results Database.

We aim to construct a practical clinical prediction model to accurately evaluate the overall survival (OS) of patients with primary spinal tumors after primary tumor resection, thereby aiding clinical decision-making. A total of 695 patients diagnosed with a primary spinal tumor, selected from the Surveillance, Epidemiology, and End Results (SEER) database, were included in this study. The Cox regression algorithm was applied to the training cohort to build the prognostic nomogram model. The nomogram's performance in terms of discrimination, calibration, and clinical usefulness was also assessed in the internal SEER validation cohort. The fitted prognostic nomogram was then used to create a web-based calculator. Four independent prognostic factors were identified to establish a nomogram model for patients with primary spinal tumors who had undergone surgical resection. The C-index (.757 for the training cohort and .681 for the validation cohort) and the area under the curve values over time (both >.68) showed that the model exhibited satisfactory discrimination in both the SEER cohort. The calibration curve revealed that the projected and actual survival rates are very similar. The decision curve analysis also revealed that the model is clinically valuable and capable of identifying high-risk patients. After developing a nomogram and a web-based calculator, we were able to reliably forecast the postoperative OS of patients with primary spinal tumors. These tools are expected to play an important role in clinical practice, informing clinicians in making decisions about patient care after surgery.

Implementing Machine Learning Models for Suicide Risk Prediction in Clinical Practice: Focus Group Study With Hospital Providers.

BackgroundInterest in developing machine learning models that use electronic health record data to predict patients’ risk of suicidal behavior has recently proliferated. However, whether and how such models might be implemented and useful in clinical practice remain unknown. To ultimately make automated suicide risk–prediction models useful in practice, and thus better prevent patient suicides, it is critical to partner with key stakeholders, including the frontline providers who will be using such tools, at each stage of the implementation process.ObjectiveThe aim of this focus group study is to inform ongoing and future efforts to deploy suicide risk–prediction models in clinical practice. The specific goals are to better understand hospital providers’ current practices for assessing and managing suicide risk; determine providers’ perspectives on using automated suicide risk–prediction models in practice; and identify barriers, facilitators, recommendations, and factors to consider.MethodsWe conducted 10 two-hour focus groups with a total of 40 providers from psychiatry, internal medicine and primary care, emergency medicine, and obstetrics and gynecology departments within an urban academic medical center. Audio recordings of open-ended group discussions were transcribed and coded for relevant and recurrent themes by 2 independent study staff members. All coded text was reviewed and discrepancies were resolved in consensus meetings with doctoral-level staff.ResultsAlthough most providers reported using standardized suicide risk assessment tools in their clinical practices, existing tools were commonly described as unhelpful and providers indicated dissatisfaction with current suicide risk assessment methods. Overall, providers’ general attitudes toward the practical use of automated suicide risk–prediction models and corresponding clinical decision support tools were positive. Providers were especially interested in the potential to identify high-risk patients who might be missed by traditional screening methods. Some expressed skepticism about the potential usefulness of these models in routine care; specific barriers included concerns about liability, alert fatigue, and increased demand on the health care system. Key facilitators included presenting specific patient-level features contributing to risk scores, emphasizing changes in risk over time, and developing systematic clinical workflows and provider training. Participants also recommended considering risk-prediction windows, timing of alerts, who will have access to model predictions, and variability across treatment settings.ConclusionsProviders were dissatisfied with current suicide risk assessment methods and were open to the use of a machine learning–based risk-prediction system to inform clinical decision-making. They also raised multiple concerns about potential barriers to the usefulness of this approach and suggested several possible facilitators. Future efforts in this area will benefit from incorporating systematic qualitative feedback from providers, patients, administrators, and payers on the use of these new approaches in routine care, especially given the complex, sensitive, and unfortunately still stigmatized nature of suicide risk.

A machine learning algorithm for predicting prolonged postoperative opioid prescription after lumbar disc herniation surgery. An external validation study using 1,316 patients from a Taiwanese cohort

Preoperative prediction of prolonged postoperative opioid prescription helps identify patients for increased surveillance after surgery. The SORG machine learning model has been developed and successfully tested using 5,413 patients from the United States (US) to predict the risk of prolonged opioid prescription after surgery for lumbar disc herniation. However, external validation is an often-overlooked element in the process of incorporating prediction models in current clinical practice. This cannot be stressed enough in prediction models where medicolegal and cultural differences may play a major role. The authors aimed to investigate the generalizability of the US citizens prediction model SORG to a Taiwanese patient cohort. Retrospective study at a large academic medical center in Taiwan. Of 1,316 patients who were 20 years or older undergoing initial operative management for lumbar disc herniation between 2010 and 2018. The primary outcome of interest was prolonged opioid prescription defined as continuing opioid prescription to at least 90 to 180 days after the first surgery for lumbar disc herniation at our institution. Baseline characteristics were compared between the external validation cohort and the original developmental cohorts. Discrimination (area under the receiver operating characteristic curve and the area under the precision-recall curve), calibration, overall performance (Brier score), and decision curve analysis were used to assess the performance of the SORG ML algorithm in the validation cohort. This study had no funding source or conflict of interests. Overall, 1,316 patients were identified with sustained postoperative opioid prescription in 41 (3.1%) patients. The validation cohort differed from the development cohort on several variables including 93% of Taiwanese patients receiving NSAIDS preoperatively compared with 22% of US citizens patients, while 30% of Taiwanese patients received opioids versus 25% in the US. Despite these differences, the SORG prediction model retained good discrimination (area under the receiver operating characteristic curve of 0.76 and the area under the precision-recall curve of 0.33) and good overall performance (Brier score of 0.028 compared with null model Brier score of 0.030) while somewhat overestimating the chance of prolonged opioid use (calibration slope of 1.07 and calibration intercept of -0.87). Decision-curve analysis showed the SORG model was suitable for clinical use. Despite differences at baseline and a very strict opioid policy, the SORG algorithm for prolonged opioid use after surgery for lumbar disc herniation has good discriminative abilities and good overall performance in a Han Chinese patient group in Taiwan. This freely available digital application can be used to identify high-risk patients and tailor prevention policies for these patients that may mitigate the long-term adverse consequence of opioid dependence: https://sorg-apps.shinyapps.io/lumbardiscopioid/.

Modification of the TRISS: simple and practical mortality prediction after trauma in an all-inclusive registry.

Numerous studies have modified the Trauma Injury and Severity Score (TRISS) to improve its predictive accuracy for specific trauma populations. The aim of this study was to develop and validate a simple and practical prediction model that accurately predicts mortality for all acute trauma admissions. This retrospective study used Dutch National Trauma Registry data recorded between 2015 and 2018. New models were developed based on nonlinear transformations of TRISS variables (age, systolic blood pressure (SBP), Glasgow Coma Score (GCS) and Injury Severity Score (ISS)), the New Injury Severity Score (NISS), the sex-age interaction, the best motor response (BMR) and the American Society of Anesthesiologists (ASA) physical status classification. The models were validated in 2018 data and for specific patient subgroups. The models' performance was assessed based on discrimination (areas under the curve (AUCs)) and by calibration plots. Multiple imputation was applied to account for missing values. The mortality rates in the development and validation datasets were 2.3% (5709/245363) and 2.5% (1959/77343), respectively. A model with sex, ASA class, and nonlinear transformations of age, SBP, the ISS and the BMR showed significantly better discrimination than the TRISS (AUC 0.915 vs. 0.861). This model was well calibrated and demonstrated good discrimination in different subsets of patients, including isolated hip fractures patients (AUC: 0.796), elderly (AUC: 0.835), less severely injured (ISS16) (AUC: 878), severely injured (ISS ≥ 16) (AUC: 0.889), traumatic brain injury (AUC: 0.910). Moreover, discrimination for patients admitted to the intensive care (AUC: s0.846), and for both non-major and major trauma center patients was excellent, with AUCs of 0.940 and 0.895, respectively. This study presents a simple and practical mortality prediction model that performed well for important subgroups of patients as well as for the heterogeneous population of all acute trauma admissions in the Netherlands. Because this model includes widely available predictors, it can also be used for international evaluations of trauma care within institutions and trauma systems.

Abstract P5-15-02: The risk for locoregional breast cancer recurrences: Significant differences in perceptions among health care professionals (HCPs)

Abstract Introduction: Unlike the highly personalized breast cancer treatment, surveillance in the Netherlands is still ‘one-size-fits-all’: annual imaging and physical examination for at least five year after treatment. To identify patients with a low or high risk for locoregional recurrences (LRRs) who might benefit from less or more intensive surveillance, risk prediction models can be used. The INFLUENCE-nomogram is an example of this and is developed using on data of the Netherlands Cancer Registry. It remains unclear to which extent health care professionals (HCPs) are able to accurately estimate the patient individual risk for locoregional recurrences. Objective: To evaluate the current awareness, knowledge and ability to estimate patients individual risks for locoregional recurrences after breast cancer - without using any prediction tools - among HCPs in the Netherlands. Methods: In this study, a cross-sectional survey was performed among general practitioners (GPs) and HCPs working in breast cancer care in six teaching hospitals in the Netherlands. In total, 85 HCPs (25 GPs and 60 hospital HCPs) participated. HCPs received descriptions of eleven breast cancer cases with different patient-, tumor- and treatment characteristics for which they were asked to estimate the risk for LRRs. The estimations were compared among different groups of HCPs and with the estimation calculated using the INFLUENCE-nomogram. Results: HCPs structurally overestimated the risk for LRRs (ranging from an average overestimation from 4% on one case to 26% on another case). In nine out of eleven cases, significant differences in risk estimates were found between groups of HCPs. In five of the eleven cases, GPs had a significantly higher median estimate than medical oncologists (respectively 10 % vs. 2.5 %; 15 % vs. 5 %; 25 % vs. 6.5 %; 25% vs. 6.5 %; 8 % vs. 1 %; and 12 % vs. 2% on these five cases). Conclusion: The structural overestimation of the risks stresses the importance to create more awareness and knowledge of the risk for LRRs after breast cancer. Medical oncologists seem to estimates risks for recurrences more accurately then GPs. This might be due to the fact that medical oncologists have more experience with the use of risk prediction models (such as Predict) in clinical practice. More awareness and knowledge of the risk for LRRs can be generated using objective estimations by prediction models. This will support the implementation of personalized surveillance after breast cancer. Table: Descriptive statistics predicted five-year risks HCPs and risk estimation INFLUENCE-nomogramCaseNMin.Max.MedianMeanStd. DeviationVarianceRisk estimation INFLUENCE-nomogramAverage overestimation HCPs1a850305***5.805.3028.040.96 %4.84 %1b850708***10.6410.07101.312.29 %8.35 %28536015***17.6712.16147.847.17 %10.50 %3a850305***6.956.8046.231.18 %5.77 %3b850.17010***12.5310.70114.482.27 %10.26 %4a8515010***12.598.3769.980.74 %11.85 %4b850.58019***19.0612.86165.441.72 %17.34 %5a851.56020***20.5713.38179.041.57 %19.00 %5b842.59025***29.6818.70349.783.6 %26.08 %6a630305***7.216.1137.310.64 %6.57 %6b620.14010***11.909.1884.272.94 %8.96 %Results of Wilcoxon one-sample signed rank test *** Median is significantly greater than risk estimation INFLUENCE-nomogram at the 0.001 level. ** Median is significantly greater than risk estimation INFLUENCE-nomogram at the 0.01 level. * Median is significantly greater than risk estimation INFLUENCE-nomogram at the 0.05 level. Citation Format: Jet W. Ankersmid, Pauline E.R. Spronk, Anneke M. Zeillemaker, Sabine Siesling. The risk for locoregional breast cancer recurrences: Significant differences in perceptions among health care professionals (HCPs) [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P5-15-02.

A Comparison of the Validity of Three Exercise Tests for Estimating Maximal Oxygen Uptake in Korean Adults Aged 19–64 Years

The purpose of this study was to compare the validity of three submaximal exercise tests (SMETs) and develop practical predictive models for the VO2max in Korean adults. A total of 541 (287 males and 254 females) adults participated in this study. Their ages ranged from 19 to 64. The VO2max was measured using the maximal-graded exercise treadmill test. The SMETs were performed by a treadmill test, the YMCA step test, and the PACER test. Regression analysis was conducted to compare the validity of the VO2max predictive equations using SMETs. The validity of the predictive models was evaluated using explanatory power, standard error of estimate (SEE), and Bland-Altman analysis. The explanatory power between the measured VO2max and the predicted VO2max was 58.0% (&lt;0.001), 59.2% (&lt;0.001), and 71.7% (&lt;0.001), respectively. The SEEs were 4.545, 4.478, and 3.732 (mL/kg/min). The models were significant predictors of VO2max and had acceptable validity in a large sample of Korean adults. Especially, among the predictive models, PACER had the highest acceptable effectiveness. Therefore, the equations developed in this study are recommended to better evaluate the cardiovascular endurance of Korean adults.

Population median imputation was noninferior to complex approaches for imputing missing values in cardiovascular prediction models in clinical practice

ObjectivesTo compare the validity and robustness of five methods for handling missing characteristics when using cardiovascular disease risk prediction models for individual patients in a real-world clinical setting. Study design and settingThe performance of the missing data methods was assessed using data from the Swedish National Diabetes Registry (n = 419,533) with external validation using the Scottish Care Information ˗ diabetes database (n = 226,953). Five methods for handling missing data were compared. Two methods using submodels for each combination of available data, two imputation methods: conditional imputation and median imputation, and one alternative modeling method, called the naïve approach, based on hazard ratios and populations statistics of known risk factors only. The validity was compared using calibration plots and c-statistics. ResultsC-statistics were similar across methods in both development and validation data sets, that is, 0.82 (95% CI 0.82–0.83) in the Swedish National Diabetes Registry and 0.74 (95% CI 0.74–0.75) in Scottish Care Information-diabetes database. Differences were only observed after random introduction of missing data in the most important predictor variable (i.e., age). ConclusionValidity and robustness of median imputation was not dissimilar to more complex methods for handling missing values, provided that the most important predictor variables, such as age, are not missing.

A Model for Identifying Optimal Patients for Primary Tumor Resection in Patients With Metastatic Bladder Cancer.

BackgroundA survival benefit was observed in metastatic bladder cancer patients who underwent primary tumor resection, but it was still confusing which patients are suitable for the surgery. For this purpose, we developed a model to screen stage M1 patients who would benefit from primary tumor resection.MethodsPatients with metastatic bladder cancer were screened from the Surveillance, Epidemiology, and End Results database (2004–2016) and then were divided into surgery (partial or complete cystectomy) group and non-surgery group. To balance the characteristics between them, a 1:1 propensity score matching analysis was applied. A hypothesis was proposed that the received primary tumor resection group has a more optimistic prognosis than the other group. The multivariable Cox model was used to explore the independent factors of survival time in two groups (beneficial and non-beneficial groups). Logistic regression was used to build a nomogram based on the significant predictive factors. Finally, a variety of methods are used to evaluate our model.ResultsA total of 7,965 patients with metastatic bladder cancer were included. And 3,314 patients met filtering standards, of which 545 (16.4%) received partial or complete cystectomy. Plots of the Kaplan–Meier and subgroup analyses confirmed our hypothesis. After propensity score matching analysis, a survival benefit was still observed that the surgery group has a longer median overall survival time (11.0 vs. 6.0 months, p < 0.001). Among the surgery cohort, 303 (65.8%) patients lived longer than 6 months (beneficial group). Differentiated characteristics included age, gender, TNM stage, histologic type, differentiation grade, and therapy, which were integrated as predictors to build a nomogram. The nomogram showed good discrimination in both training and validation cohorts (area under the receiver operating characteristic curve (AUC): 0.806 and 0.742, respectively), and the calibration curves demonstrated good consistency. Decision curve analysis showed that the nomogram was clinically useful. Compared with TNM staging, our model shows a better predictive value in identifying optimal patients for primary tumor resection.ConclusionsA practical predictive model was created and verified, which might be used to identify the optimal candidates for the partial or complete cystectomy group of the primary tumor among metastatic bladder cancer.

Identifying optimal candidates for tumor resection among borderline and locally advanced pancreatic cancer: A population-based predictive model

BackgroundWhether patients with borderline resectable and locally advanced pancreatic cancer (BR/LAPC) benefit from resection of the primary cancer is controversial. We developed a nomogram to screen who would benefit from surgery for the primary tumor. MethodsWe identified patients from the Surveillance, Epidemiology, and End Results (SEER) database and then divided them into surgical and non-surgical groups. A 1:1 propensity score matching (PSM) was used to mitigate the bias. We hypothesized that patients who underwent surgery would benefit from surgery by having a longer median overall survival (OS) than patients who did not undergo surgery. Univariate and multivariate logistic regression analyses were used to determine the variables affecting surgical outcomes, and a nomogram was created based on the multivariate logistic results. Finally, we verified the discrimination and calibration of the nomogram with receiver operator characteristic (ROC) curve and calibration plots. ResultsA total of 518 pairs of surgical and non-surgical pancreatic cancer patients were matched after PSM. Survival curves showed longer OS in the surgical group than in the non-surgical group, median survival times were 14 months versus 8 months. In the surgical group, 340 (65.63%) patients have a longer survival time than 8 months (beneficial group). Multifactorial logit regression results showed that including age, tumor size, degree of differentiation, and chemotherapy were significant influences on the benefit of surgery for primary tumors and were used as predictors to construct a nomogram. The area under the ROC curve (AUC) reached 0.747 and 0.706 in the training and validation sets. ConclusionWe developed a practical predictive model to support clinical decision-making that can be used to help clinicians determine if there is a benefit to surgical resection of the primary tumor in patients with BR/LAPC.

Impacts of data preprocessing and selection on energy consumption prediction model of HVAC systems based on deep learning

Accurate energy consumption prediction is the basis of predictive control for heating, ventilation and air conditioning (HVAC) systems. Data-driven models are widely used for energy consumption prediction. The prediction accuracy can be affected by data preprocessing or selection, which are not well studied yet. This paper attempts to study the impacts of different data processing methods and feature selection on the HVAC energy consumption prediction based on data-driven models. Long and short-term memory models are developed based on historical data to predict the day-ahead hourly energy consumption of HVAC systems. Two data smoothing methods, Gaussian kernel density estimation and Savitzky-Golay filter, are selected and compared. The impacts of feature selection, training set volumes and update frequency of models are analyzed and compared. To obtain general results of the above problems, three office buildings are selected. Results show that the smoothing methods can not ensure the improvement in accuracy by removing the exceptional data in the raw data which arise reasonably in practice. The inputs with raw 1-day historical energy consumption, the dry-bulb temperature and dew-point temperature in the next day, holiday type and day type is recommended to predict day-ahead HVAC energy consumption. It is recommended to use a larger training set if computing cost is acceptable. The model should be updated after being used for over 7 weeks. This study would guide the development of prediction models in practice.

An Improved Nomogram to Reduce False-Positive Biopsy Rates of Breast Imaging Reporting and Data System Ultrasonography Category 4A Lesions.

Background:The NCCN clinical guidelines recommended core needle biopsy for breast lesions classified as Breast Imaging Reporting and Data System (BI-RADS) 4, while category 4A lesions are only 2-10% likely to be malignant. Thus, a large number of biopsies of BI-RADS 4A lesions were ultimately determined to be benign, and those unnecessary biopsies may incur additional costs and pains. However, it is important to emphasize that the current risk prediction model focuses primarily on the details and complex risk features of US or MG findings, which may be difficult to apply in order to benefit from the model. To stratify and manage BI-RADS 4A lesions effectively and efficiently, a more effective and practical predictive model must be developed.Methods:We retrospectively analyzed 465 patients with BI-RADS ultrasonography (US) category 4A lesions, diagnosed between January 2019 and July 2019 in Tianjin Medical University Cancer Institute and Hospital and National Clinical Research Center for Cancer. Univariate and multivariate logistic regression analyses were conducted to identify risk factors. To stratify and predict the malignancy of BI-RADS 4A lesions, a nomogram combining the risk factors was constructed based on the multivariate logistic regression results. In order to determine the predictive performance of our predictive model, we used the concordance index (C-index), calibration curve, and receiver operating characteristic (ROC), and the decision curve analysis (DCA) to assess the clinical benefits.Results:Based on our analysis, 16.3% (76 out of 465) of patients were pathologically diagnosed with malignant lesions, while 83.6% (389 out of 465) were diagnosed with benign lesions. According to univariate and multivariate logistic regression analysis, age (OR = 3.414, 95%CI:1.849-6.303), nipple discharge (OR = .326, 95%CI:0.157-.835), palpable lesions (OR = 1.907, 95%CI:1.004-3.621), uncircumscribed margin (US) (OR = 1.732, 95%CI:1.033-2.905), calcification (mammography, MG) (OR = 2.384, 95%CI:1.366-4.161), BI-RADS(MG) (OR = 5.345, 95%CI:2.934-9.736) were incorporated into the predictive nomogram (C-index = .773). There was good agreement between the predicted risk and the observed probability of recurrence. Furthermore, we determined that 153 was the best cutoff score for distinguishing between patients in the low- and high-risk groups. Malignant lesions were significantly more prevalent in high-risk patients than in low-risk patients.Conclusion:Based on clinical, US, and MG features, we present a predictive nomogram to reliably predict the malignancy risk of BI-RADS(US) 4A lesions, which may assist clinicians in the selection of patients at low risk of malignancy and reduce the number of false-positive biopsies.

Mobile Network Coverage Prediction Based on Supervised Machine Learning Algorithms

The need for wider coverage and high-performance quality of mobile networks is critical due to the maturity of Internet penetration in today&#x2019;s society. One of the primary drivers of this demand is the dramatic shift toward digitalization due to the Covid-19 pandemic impact. Meanwhile, the emergence of the 5G wireless standard and the increasingly complex actual operating environment of mobile networks make the traditional prediction model less reliable. With the recent advancements and promising capabilities of machine learning (ML), it is seen as an alternative to the traditional approaches for ground to ground (G2G) mobile communication coverage prediction. In this study, various ML models have been tested and evaluated to develop an ML-based received signal strength prediction model for mobile networks. However, the challenge is to identify a practical ML model that can fulfill the computing speed criteria while still meeting the prediction accuracy. A total of six categories of ML models, namely Linear Regression (LR), Artificial Neural Network (ANN), Support Vector Machine (SVM), Regression Trees (RT), Ensembles of Trees (ET), and Gaussian Process Regression (GPR) that consists of more than 20 types of established algorithms/kernels have been tested and evaluated in this paper to identify the best contender among them, in terms of speed and accuracy. Findings from the evaluation showed that the GPR model is the most accurate model for Reference Signal Received Power (RSRP) prediction in terms of <inline-formula> <tex-math notation="LaTeX">$RMSE$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula>, followed by ET, RT, SVM, ANN and LR. Nevertheless, prediction speed and model training times are also important factors in determining the most practical model for RSRP prediction for several real-world mobile network planning applications. Finally, the ET model with Random Forest (RF) algorithm has been selected and highly recommended as the most practically employed ML model for developing rigorous RSRP predictions model in multi-frequency bands and multi-environment. The developed prediction model is capable of being utilized for the network analysis and optimization.