Patient-level Dataset Research Articles

Following hip fracture, hospital organisational factors associated with prescription of anti-osteoporosis medication on discharge, to address imminent refracture risk: a record-linkage study.

Patients who sustain a hip fracture are known to be at imminent refracture risk. Their complex multidisciplinary rehabilitation needs to include falls prevention and anti-osteoporosis medication (AOM) to prevent such fractures. This study aimed to determine which hospital-level organisational factors predict prescription of post-hip fracture AOM, and refracture risk. A cohort of 178 757 patients aged ≥60years who sustained a hip fracture in England and Wales (2016-19) was examined and followed for 1year. Patient-level hospital admission datasets from 172 hospitals, the National Hip Fracture Database, and mortality data were linked to 71 metrics extracted from 18 hospital-level organisational reports. Multilevel models determined organisational factors, independent of patient case-mix, associated with (i) AOM prescription, (ii) refracture (by ICD10 coding). Patients were mean (SD) 82.7 (8.6) years old, 71% female, with 18% admitted from care homes. Overall, 101 735 (57%) were prescribed AOM during admission; while 50 354 (28%) died during 1-year follow-up, 12 240 (7%) refractured. Twelve organisational factors were associated with AOM prescription, e.g., orthogeriatrician-led care compared to traditional care models (OR 4.65 [95%CI: 2.25-9.59]); AOM was 9% (95%CI: 6%-13%) more likely to be prescribed in hospitals providing routine bone health assessment to all patients. Refracture occurred at median 126days (IQR 59-234). Eight organisational factors were associated with refracture risk; hospitals providing orthogeriatrician assessment to all patients within 72-hours of admission had an 18% (95%CI: 2-31%) lower refracture risk, weekend physiotherapy provision an 8% (95%CI: 3-14%) lower risk, and where occupational therapists attended clinical governance meetings, a 7% (95%CI: 2-12%) lower risk. Delays initiating post-discharge community rehabilitation were associated with a 15% (95%CI: 3-29%) greater refracture risk. These novel, national findings highlight the importance of orthogeriatrician, physiotherapist and occupational therapist involvement in secondary fracture prevention post hip fracture; notably fracture risk reductions were seen within 12months of hip fracture.

Differences in maternal-newborn ABO blood groups and risk of serious infant infection.

During pregnancy, various maternal IgG antibodies are transferred to the developing fetus, some of which may protect the newborn against infection. If a mother and her fetus have different A, B or O (ABO) blood groups, then transferred maternal antibodies may plausibly protect the infant against infection. To determine if maternal-newborn ABO blood group incongruence vs. congruence is associated with a lower risk of serious infection in the infant. Retrospective population-based cohort. We used linked patient-level datasets for all singleton hospital livebirths from 2008 to 2022 in Ontario, Canada, with known maternal and newborn ABO blood groups. We used a dichotomous exposure state, either ABO blood group congruent (N = 114 507) or incongruent (N = 43 074). The main outcome of interest was the risk of serious infant infection within 27 days, and from 28 to 365 days, after birth. Cox proportional hazard models generated hazard ratios and 95% confidence intervals, and were adjusted for maternal age, world region of origin, residential income quintile and gestational age at birth. Relative to maternal-newborn congruency, incongruent ABO blood group was associated with an adjusted hazard ratio of 0.88 (95% CI: 0.80-0.97) for serious neonatal infection within 27 days of birth, and 0.93 (95% CI: 0.90-0.96) for serious infection between 28 and 365 days after birth. Maternal-newborn ABO incongruence may be associated with a lower relative risk of a serious infant infection within 27 days, and from 28 to 365 days, after birth.

A deep learning approach for projection and body-side classification in musculoskeletal radiographs

BackgroundThe growing prevalence of musculoskeletal diseases increases radiologic workload, highlighting the need for optimized workflow management and automated metadata classification systems. We developed a large-scale, well-characterized dataset of musculoskeletal radiographs and trained deep learning neural networks to classify radiographic projection and body side.MethodsIn this IRB-approved retrospective single-center study, a dataset of musculoskeletal radiographs from 2011 to 2019 was retrieved and manually labeled for one of 45 possible radiographic projections and the depicted body side. Two classification networks were trained for the respective tasks using the Xception architecture with a custom network top and pretrained weights. Performance was evaluated on a hold-out test sample, and gradient-weighted class activation mapping (Grad-CAM) heatmaps were computed to visualize the influential image regions for network predictions.ResultsA total of 13,098 studies comprising 23,663 radiographs were included with a patient-level dataset split, resulting in 19,183 training, 2,145 validation, and 2,335 test images. Focusing on paired body regions, training for side detection included 16,319 radiographs (13,284 training, 1,443 validation, and 1,592 test images). The models achieved an overall accuracy of 0.975 for projection and 0.976 for body-side classification on the respective hold-out test sample. Errors were primarily observed in projections with seamless anatomical transitions or non-orthograde adjustment techniques.ConclusionsThe deep learning neural networks demonstrated excellent performance in classifying radiographic projection and body side across a wide range of musculoskeletal radiographs. These networks have the potential to serve as presorting algorithms, optimizing radiologic workflow and enhancing patient care.Relevance statementThe developed networks excel at classifying musculoskeletal radiographs, providing valuable tools for research data extraction, standardized image sorting, and minimizing misclassifications in artificial intelligence systems, ultimately enhancing radiology workflow efficiency and patient care.Key points• A large-scale, well-characterized dataset was developed, covering a broad spectrum of musculoskeletal radiographs.• Deep learning neural networks achieved high accuracy in classifying radiographic projection and body side.• Grad-CAM heatmaps provided insight into network decisions, contributing to their interpretability and trustworthiness.• The trained models can help optimize radiologic workflow and manage large amounts of data.Graphical

Patient level dataset to study the effect of COVID-19 in people with Multiple Sclerosis

Multiple Sclerosis (MS) is an inflammatory autoimmune disease of the central nervous system, causing increased vulnerability to infections and disability among young adults. Ever since the outbreak of coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 infections, there have been concerns among people with MS (PwMS) about the potential interactions between various disease-modifying therapies and COVID-19. The COVID-19 in MS Global Data Sharing Initiative (GDSI) was initiated in 2020 with the aim of addressing these concerns. This paper focuses on the anonymisation and publicly releasing of a GDSI sub-dataset, comprising data entered by PwMS and clinicians using a fast data entry tool. The dataset includes information on demographics, comorbidities and hospital stay and COVID-19 symptoms of PwMS. The dataset can be used to perform different statistical analyses to improve our understanding of COVID-19 in MS. Furthermore, this dataset can also be used within the context of educational activities to educate different stakeholders on the complex data science topics that were used within the GDSI.

A large-scale dataset of patient summaries for retrieval-based clinical decision support systems

Retrieval-based Clinical Decision Support (ReCDS) can aid clinical workflow by providing relevant literature and similar patients for a given patient. However, the development of ReCDS systems has been severely obstructed by the lack of diverse patient collections and publicly available large-scale patient-level annotation datasets. In this paper, we collect a novel dataset of patient summaries and relations called PMC-Patients to benchmark two ReCDS tasks: Patient-to-Article Retrieval (ReCDS-PAR) and Patient-to-Patient Retrieval (ReCDS-PPR). Specifically, we extract patient summaries from PubMed Central articles using simple heuristics and utilize the PubMed citation graph to define patient-article relevance and patient-patient similarity. PMC-Patients contains 167k patient summaries with 3.1 M patient-article relevance annotations and 293k patient-patient similarity annotations, which is the largest-scale resource for ReCDS and also one of the largest patient collections. Human evaluation and analysis show that PMC-Patients is a diverse dataset with high-quality annotations. We also implement and evaluate several ReCDS systems on the PMC-Patients benchmarks to show its challenges and conduct several case studies to show the clinical utility of PMC-Patients.

Prevalence of multiple long-term conditions (multimorbidity) in England: a whole population study of over 60 million people.

To determine the prevalence of multiple long-term conditions (MLTC) at whole English population level, stratifying by age, sex, socioeconomic status and ethnicity. A whole population study. Individuals registered with a general practice in England and alive on 31 March 2020. 60,004,883 individuals. MLTC prevalence, defined as two or more of 35 conditions derived from a number of national patient-level datasets. Multivariable logistic regression was used to assess the independent associations of age, sex, ethnicity and deprivation decile with odds of MLTC. The overall prevalence of MLTC was 14.8% (8,878,231), varying from 0.9% (125,159) in those aged 0-19 years to 68.2% (1,905,979) in those aged 80 years and over. In multivariable regression analyses, compared with the 50-59 reference group, the odds ratio was 0.04 (95% confidence interval (CI): 0.04-0.04; p < 0.001) for those aged 0-19 years and 10.21 (10.18-10.24; p < 0.001) for those aged 80 years and over. Odds were higher for men compared with women, 1.02 (1.02-1.02; p < 0.001), for the most deprived decile compared with the least deprived, 2.26 (2.25-2.27; p < 0.001), and for Asian ethnicity compared with those of white ethnicity, 1.05 (1.04-1.05; p < 0.001). Odds were lower for black, mixed and other ethnicities (0.94 (0.94-0.95) p < 0.001, 0.87 (0.87-0.88) p < 0.001 and 0.57 (0.56-0.57) p < 0.001, respectively). MLTC for persons aged 0-19 years were dominated by asthma, autism and epilepsy, for persons aged 20-49 years by depression and asthma, for persons aged 50-59 years by hypertension and depression and for those aged 60 years and older, by cardiometabolic factors and osteoarthritis. There were large numbers of combinations of conditions in each age group ranging from 5936 in those aged 0-19 years to 205,534 in those aged 80 years and over. While this study provides useful insight into the burden across the English population to assist health service delivery planning, the heterogeneity of MLTC presents challenges for delivery optimisation.

Cancer data quality and harmonization in Europe: the experience of the BENCHISTA Project - international benchmarking of childhood cancer survival by stage.

Variation in stage at diagnosis of childhood cancers (CC) may explain differences in survival rates observed across geographical regions. The BENCHISTA project aims to understand these differences and to encourage the application of the Toronto Staging Guidelines (TG) by Population-Based Cancer Registries (PBCRs) to the most common solid paediatric cancers. PBCRs within and outside Europe were invited to participate and identify all cases of Neuroblastoma, Wilms Tumour, Medulloblastoma, Ewing Sarcoma, Rhabdomyosarcoma and Osteosarcoma diagnosed in a consecutive three-year period (2014-2017) and apply TG at diagnosis. Other non-stage prognostic factors, treatment, progression/recurrence, and cause of death information were collected as optional variables. A minimum of three-year follow-up was required. To standardise TG application by PBCRs, on-line workshops led by six tumour-specific clinical experts were held. To understand the role of data availability and quality, a survey focused on data collection/sharing processes and a quality assurance exercise were generated. To support data harmonization and query resolution a dedicated email and a question-and-answers bank were created. 67 PBCRs from 28 countries participated and provided a maximally de-personalized, patient-level dataset. For 26 PBCRs, data format and ethical approval obtained by the two sponsoring institutions (UCL and INT) was sufficient for data sharing. 41 participating PBCRs required a Data Transfer Agreement (DTA) to comply with data protection regulations. Due to heterogeneity found in legal aspects, 18 months were spent on finalizing the DTA. The data collection survey was answered by 68 respondents from 63 PBCRs; 44% of them confirmed the ability to re-consult a clinician in cases where stage ascertainment was difficult/uncertain. Of the total participating PBCRs, 75% completed the staging quality assurance exercise, with a median correct answer proportion of 92% [range: 70% (rhabdomyosarcoma) to 100% (Wilms tumour)]. Differences in interpretation and processes required to harmonize general data protection regulations across countries were encountered causing delays in data transfer. Despite challenges, the BENCHISTA Project has established a large collaboration between PBCRs and clinicians to collect detailed and standardised TG at a population-level enhancing the understanding of the reasons for variation in overall survival rates for CC, stimulate research and improve national/regional child health plans.

What are the treatment remission, response and extent of improvement rates after up to four trials of antidepressant therapies in real-world depressed patients? A reanalysis of the STAR*D study’s patient-level data with fidelity to the original research protocol

ObjectiveReanalyse the patient-level data set of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study with fidelity to the original research protocol and related publications.DesignThe study was open label and...

Feasibility and validity of using healthcare databases to conduct cross-national comparative studies of opioid use, its determinants and consequences.

A cross-national comparative (CNC) study about opioid utilization would allow the identification of strategies to improve pain management and mitigate risk. However, little is known about the accessibility and validity of information in healthcare databases internationally. This study aimed to identify the feasibility of using healthcare databases to conduct a CNC study of opioid utilization and its associated consequences. A cross-sectional survey was launched in March 2018, including experts interested in CNC studies comparing opioid utilization by purposeful sampling. An electronic survey was used to collect database characteristics, medicine information, and linkage information of each aggregate-level dataset (AD) and individual patient-level dataset (IPD). Overall, participants from 21 geographical regions reported 18 ADs and 19 IPDs. Information on dispensed medications is available from 17 ADs and 17 IPDs. Of the 16 ADs that include primary care settings, only 9 ADs can obtain information from secondary care settings. Fourteen IPDs included patients' characteristics or could be retrieved from linkage databases. Although most ADs are publicly accessible (n = 13), only five IPDs can be accessed without extra cost. Most ADs could be used to report opioid utilization in a primary care setting. IPDs with linkage databases should be applied to identify potential determinants, clinical outcomes, and policy impact. Data access restrictions and governance policies across jurisdictions can be challenging for timely analysis and require further collaboration.

Probabilistic Forecasting of Patient Waiting Times in an Emergency Department

Problem definition: We study the estimation of the probability distribution of individual patient waiting times in an emergency department (ED). Whereas it is known that waiting-time estimates can help improve patients’ overall satisfaction and prevent abandonment, existing methods focus on point forecasts, thereby completely ignoring the underlying uncertainty. Communicating only a point forecast to patients can be uninformative and potentially misleading. Methodology/results: We use the machine learning approach of quantile regression forest to produce probabilistic forecasts. Using a large patient-level data set, we extract the following categories of predictor variables: (1) calendar effects, (2) demographics, (3) staff count, (4) ED workload resulting from patient volumes, and (5) the severity of the patient condition. Our feature-rich modeling allows for dynamic updating and refinement of waiting-time estimates as patient- and ED-specific information (e.g., patient condition, ED congestion levels) is revealed during the waiting process. The proposed approach generates more accurate probabilistic and point forecasts when compared with methods proposed in the literature for modeling waiting times and rolling average benchmarks typically used in practice. Managerial implications: By providing personalized probabilistic forecasts, our approach gives low-acuity patients and first responders a more comprehensive picture of the possible waiting trajectory and provides more reliable inputs to inform prescriptive modeling of ED operations. We demonstrate that publishing probabilistic waiting-time estimates can inform patients and ambulance staff in selecting an ED from a network of EDs, which can lead to a more uniform spread of patient load across the network. Aspects relating to communicating forecast uncertainty to patients and implementing this methodology in practice are also discussed. For emergency healthcare service providers, probabilistic waiting-time estimates could assist in ambulance routing, staff allocation, and managing patient flow, which could facilitate efficient operations and cost savings and aid in better patient care and outcomes.Supplemental Material: The online supplement is available at https://doi.org/10.1287/msom.2023.1210 .

Oral buprenorphine utilization, concomitant benzodiazepines and opioid analgesics, and payment source: Trends from 2015 to 2019

IntroductionThis study describes utilization patterns of oral buprenorphine products indicated for the treatment of opioid use disorder with a focus on patterns consistent with prescribing guidelines and the safe use conditions during induction and maintenance treatment outlined by the Risk Evaluation and Mitigation Strategy (REMS) Program, including trends over time. MethodsUsing an anonymized longitudinal patient-level dataset that captures information on medical and pharmacy claims in the United States (October 1, 2014 through March 31, 2020), buprenorphine prescriptions, days' supply, and daily dose were described overall and stratified by induction (month 1) vs. maintenance (month 2–6) treatment, along with duration of concomitant benzodiazepines or opioid analgesics. ResultsOverall, there were 1.5 million buprenorphine treatment episodes initiated between January 1, 2015 to September 30, 2019 (2015: 258,899; 2019: 351,378). Treatment episodes included an average of 6.8 buprenorphine prescriptions (standard deviation [SD] = 6.7), 16.8 days' supply per prescription (SD = 10.5), 94.2 total days' supply per treatment episode (SD = 71.4), and a mean daily dose of 13.6 mg (SD = 6.3), with the number of prescriptions and total days' supply per treatment episode declining over the study period. There was a lower mean number of days' supply per prescription in the first month of treatment compared to months 2–6 (month 1: 15.8 [SD = 11.0]; month 2–6: 19.0 [SD = 10.1]) and daily dose per prescription (month 1: 13.3 mg [SD = 6.4]; months 2–6: 14.3 mg [SD = 6.2]), and a higher mean number of prescriptions per month (month 1: 2.5 per month [SD = 1.7]; months 2–6: 1.8 per month [SD = 1.2]). From 2015 to 2019, there appeared to be a shift in payer mix, with increases in Medicaid/Medicare and declines in cash and commercial insurance. Concomitant benzodiazepine and opioid analgesic use declined over time; in 2019, 16.6 % and 14.3 % of treatment episodes had any concomitant benzodiazepine or opioid analgesic, respectively, and <5 % had chronic (>90 overlapping days) concomitant use (3.0 % and 0.4 %, respectively). ConclusionsThe number of oral buprenorphine treatment episodes increased over the study period, and prescribing was generally consistent with the REMS and other treatment guidelines. There was a decline in concomitant buprenorphine and benzodiazepine or opioid analgesics, and chronic concomitant use was rare.

Systems of Care Factors Should Be Considered in Regionalization of Congenital Cardiac Surgery

BackgroundRegionalization of care has been proposed to optimize outcomes in congenital cardiac surgery (CCS). We hypothesized that hospital infrastructure and systems of care factors could also be considered in regionalization efforts. MethodsObserved-to-expected (O/E) mortality ratio and hospital volumes were obtained between 2015 and 2018 from public reporting data. Using a resource dependence framework, we examined factors obtained from American Hospital Association, Children’s Hospital Association, and hospital websites. Linear regression models were estimated with volume only, then with hospital factors, stratified by procedural complexity. Robust regression models were reestimated to assess the impact of outliers. ResultsWe found wide variation in the volume of congenital cardiac surgeries performed (89-3920) and in the surgical outcomes (O/E ratio range, 0.3-3.1). Six outlier hospitals performed few high-complexity cases with high mortality. Univariate analysis including all cases indicated that higher volume predicted lower O/E ratio (β = −0.02; SE = 0.008; P = .011). However, this effect was driven by the most complex cases. Models stratified by The Society of Thoracic Surgeons–European Association for Cardio-Thoracic Surgery category show that volume is a significant predictor only in category 5 cases (β = −1.707; SE = 0.663; P = .012). Robust univariate regression accounting for outliers found no effect of volume on O/E ratio (β = 0.005; SE = 0.002; P = .975). Elimination of outliers through robust multivariate regression decreased the volume-outcome relationship and found a modest relationship between health plan ownership and outcomes. ConclusionsSystems of care factors should be considered in addition to volume in designing regionalization in CCS. Patient-level data sets will better define these factors.

International benchmarking of childhood cancer survival by stage at diagnosis: The BENCHISTA project protocol.

Several studies have shown significant variation in overall survival rates from childhood cancer between countries, using population-based cancer registry (PBCR) data for all cancers combined and for many individual tumour types among children. Without accurate and comparable data on Tumour stage at diagnosis, it is difficult to define the reasons for these survival differences. This is because measurement systems designed for adult cancers do not apply to children's cancers and cancer registries often hold limited information on paediatric tumour stage and the data sources used to define it. The BENCHISTA project aims to test the application of the international consensus "Toronto Staging Guidelines" (TG) for paediatric tumours by European and non-European PBCRs for six common paediatric solid tumours so that reliable comparisons of stage at diagnosis and survival rates by stage can be made to understand any differences. A secondary aim is to test the data availability and completeness of collection of several 'Toronto' consensus non-stage prognostic factors, treatment types given, occurrence of relapse/progression and cause of death as a descriptive feasibility study. PBCRs will use their permitted data access channels to apply the Toronto staging guidelines to all incident cases of six solid childhood cancers (medulloblastoma, osteosarcoma, Ewings sarcoma, rhabdomyosarcoma, neuroblastoma and Wilms tumour) diagnosed in a consecutive three-year period within 2014-2017 in their population. Each registry will provide a de-identified patient-level dataset including tumour stage at diagnosis, with only the contributing registry holding the information that would be needed to re-identify the patients. Where available to the registry, patient-level data on 'Toronto' non-stage prognostic factors, treatments given and clinical outcomes (relapse/progression/cause of death) will be included. More than 60 PBCRs have been involved in defining the patient-level dataset items and intend to participate by contributing their population-level data. Tumour-specific on-line training workshops with clinical experts are available to cancer registry staff to assist them in applying the Toronto staging guidelines in a consistent manner. There is also a project-specific help desk for discussion of difficult cases and promotion of the CanStaging online tools, developed through the International Association of Cancer Registries, to further ensure standardisation of data collection. Country-specific stage distribution and observed survival by stage at diagnosis will be calculated for each tumour type to compare survival between countries or large geographical regions. This study will be promote and enhance the collection of standardized staging data for childhood cancer by European and non-European population-based cancer registries. Therefore, this project can be seen as a feasibility project of widespread use of Toronto Staging at a population-level by cancer registries, specifying the data sources used and testing how well standardized the processes can be. Variation in tumour stage distribution could be due to real differences, to different diagnostic practices between countries and/or to variability in how cancer registries assign Toronto stage. This work also aims to strengthen working relationships between cancer registries, clinical services and cancer-specific clinical study groups, which is important for improving patient outcomes and stimulating research.

How Did Zero-Markup Medicines Policy Change Prescriptions in the Eyes of Patients?-A Retrospective Quasi-Experimental Analysis.

Background: China implemented the zero-markup medicines policy to reverse the overuse of medicine in public health institutions, by changing the distorted financing mechanism, which heavily relies on revenue generated from medicines. The zero-markup medicines policy was progressively implemented in city public hospitals from 2015 to 2017. Objective: This study is expected to generate convincing evidence with subjective measurements and contribute to a more comprehensive evaluation of the policy from both objective and subjective perspectives. Methods: This study was based on a large patient-level dataset with a quasi-experimental design. We employed the difference-in-difference (DID) method, combined with propensity score matching methods, to estimate the causal effect of the policy in reducing overprescriptions from the patient perspective. Results: The study estimated a statistically significant increased probability that the responded outpatients denied overprescription in their visiting hospitals. The mean interacted policy effect, in percentage points, of all observations were positive (logit DID model: 0.15, z = 10.27, SE = 0.01; PSM logit DID model: 0.15, z = 10.26, SE = 0.01; PSM logit DID hospital fixed-effect model: 0.12, z = 3.00, SE = 0.04). Discussion: The policy might reduce overprescription in public hospitals from the patient’s perspective. The patient’s attitude is one aspect of a comprehensive policy evaluation. The final concrete conclusion of the policy evaluation can only be made through a systematic review of the studies with rigorous design and with both objective and subjective measurements.

Notre nom, un marqueur d’identité

Vaccinations are considered the major tool to curb the current SARS-CoV-2 pandemic. A randomized placebo-controlled trial of the BNT162b2 vaccine has demonstrated a 95% efficacy in preventing COVID-19 disease. These results are now corroborated with statistical analyses of real-world vaccination rollouts, but resolving vaccine effectiveness across demographic groups is challenging. Here, applying a multivariable logistic regression analysis approach to a large patient-level dataset, including SARS-CoV-2 tests, vaccine inoculations and personalized demographics, we model vaccine effectiveness at daily resolution and its interaction with sex, age and comorbidities. Vaccine effectiveness gradually increased post day 12 of inoculation, then plateaued, around 35 days, reaching 91.2% [CI 88.8%-93.1%] for all infections and 99.3% [CI 95.3%-99.9%] for symptomatic infections. Effectiveness was uniform for men and women yet declined mildly but significantly with age and for patients with specific chronic comorbidities, most notably type 2 diabetes. Quantifying real-world vaccine effectiveness, including both biological and behavioral effects, our analysis provides initial measurement of vaccine effectiveness across demographic groups.

Multiple hospital organisational factors are associated with adverse patient outcomes post-hip fracture in England and Wales: the REDUCE record-linkage cohort study.

ObjectivesDespite established standards and guidelines, substantial variation remains in the delivery of hip fracture care across the United Kingdom. We aimed to determine which hospital-level organisational factors predict adverse patient outcomes in the months following hip fracture.MethodsWe examined a national record-linkage cohort of 178,757 patients aged ≥60 years who sustained a hip fracture in England and Wales in 2016–19. Patient-level hospital admissions datasets, National Hip Fracture Database and mortality data were linked to metrics from 18 hospital-level organisational-level audits and reports. Multilevel models identified organisational factors, independent of patient case-mix, associated with three patient outcomes: length of hospital stay (LOS), 30-day all-cause mortality and emergency 30-day readmission.ResultsAcross hospitals mean LOS ranged from 12 to 41.9 days, mean 30-day mortality from 3.7 to 10.4% and mean readmission rates from 3.7 to 30.3%, overall means were 21.4 days, 7.3% and 15.3%, respectively. In all, 22 organisational factors were independently associated with LOS; e.g. a hospital’s ability to mobilise >90% of patients promptly after surgery predicted a 2-day shorter LOS (95% confidence interval [CI]: 1.2–2.6). Ten organisational factors were independently associated with 30-day mortality; e.g. discussion of patient experience feedback at clinical governance meetings and provision of prompt surgery to >80% of patients were each associated with 10% lower mortality (95%CI: 5–15%). Nine organisational factors were independently associated with readmissions; e.g. readmissions were 17% lower if hospitals reported how soon community therapy would start after discharge (95%CI: 9–24%).ConclusionsReceipt of hip fracture care should be reliable and equitable across the country. We have identified multiple, potentially modifiable, organisational factors associated with important patient outcomes following hip fracture.

The Effects of Health Information Exchange Access on Healthcare Quality and Efficiency: An Empirical Investigation

Health information exchanges (HIEs) are designed to improve the quality and efficiency of healthcare by facilitating improved information sharing between health entities. This study systematically examines the impact of HIE use in emergency departments (EDs) on the quality and efficiency of medical care. We focus on the length of stay (LOS) and the 30-day readmission rate to capture healthcare efficiency and quality, respectively. We also examine whether the breadth of patient health information and physicians’ experience with the HIE moderates these effects. We leverage a unique panel data that tracks actual HIE access by physicians who practice in a set of hospitals that participate in the focal HIE. The patient-level encounter data set—which involves more than 80,000 ED encounters attended by more than 300 physicians over a 19-month period—comprises detailed medical provider information, patient-level medical information, and various other information related to procedures that were performed. After controlling for a battery of patient-specific, physician-specific, disease-specific, and ED visit-specific variables, our results show that HIE access in information-intensive environments (such as EDs) reduces LOS and 30-day readmission rate. We find that breadth of patient health information and physicians’ HIE experience amplify these benefits. We account for endogeneity issues and perform additional falsification tests and robustness checks. We document that benefits of HIE access are amplified for noninjury, chronic condition, and uncommon diagnoses related patient visits. Based on our results, we offer insights to practitioners and academicians alike on how HIEs can yield better patient-level and provider-level outcomes.This paper was accepted by Anandhi Bharadwaj, information systems.Supplemental Material: The web appendix is available at https://doi.org/10.1287/mnsc.2022.4378 .

Application of transfer learning and ensemble learning in image-level classification for breast histopathology

BackgroundBreast cancer has the highest prevalence among all cancers in women globally. The classification of histopathological images in the diagnosis of breast cancers is an area of clinical concern. In computer-aided diagnosis, most traditional classification models use a single network to extract features, although this approach has significant limitations. Moreover, many networks are trained and optimized on patient-level datasets, ignoring lower-level data labels. MethodsThis paper proposed a deep ensemble model based on image-level labels for the binary classification of breast histopathological images of benign and malignant lesions. First, the BreaKHis dataset was randomly divided into training, validation, and test sets. Then, data augmentation techniques were used to balance the numbers of benign and malignant samples. Third, based on their transfer learning performance and the complementarity between networks, VGG16, Xception, ResNet50, and DenseNet201 were selected as base classifiers. ResultsIn a ensemble network model with accuracy as the weight, the image-level binary classification achieved an accuracy of 98.90%. To verify the capabilities of our method, it was experimentally compared with the latest transformer and multilayer perception (MLP) models on the same dataset. Our ensemble model showed a 5%–20% advantage, emphasizing its far-reaching abilities in classification tasks. ConclusionsThis research focuses on improving the performance of a classification model with an ensemble algorithm. Transfer learning has an essential role in classification of small datasets, improving training speed and accuracy. Our model may outperform many existing approaches with respect to accuracy and has applications in the field of auxiliary medical diagnosis.

Abstract 4094: WAYFIND-R: Delivering a high-quality real-world data (RWD) global registry of patients diagnosed with a solid tumor and profiled with next-generation sequencing (NGS)

Abstract NGS-based, molecularly guided therapy is becoming a pivotal part of precision oncology. Rare molecular cancer subtypes pose difficulties in traditional clinical trials; however, collecting patient RWD can help elucidate the clinical utility of targeted therapies, molecular tumor boards (MTBs), and NGS-based genomic profiling. WAYFIND-R (NCT04529122) is a global registry that: 1) characterizes the treatments and clinical course of patients with cancer; 2) provides a data research platform to evaluate real-world treatment decisions and outcomes; 3) supports the design/conduct of epidemiological research and clinical trials. WAYFIND-R will provide key learnings on data quality, sharing, and privacy for contemporary disease registries. Unlike many disease registries, WAYFIND-R is global, prospective, longitudinal, generalizable (all clinically validated NGS-based tests of any panel size are permitted), and includes NGS testing and standardized data collection to enhance data quality and minimize missing data. WAYFIND-R includes adults with any type/stage of solid tumor profiled with NGS ≤3 months before enrollment. NGS details, patient sociodemographics, cancer characteristics, patient-relevant safety data, treatment decisions, and outcomes are collected at baseline and prospectively every ≥6 months. WAYFIND-R will initially operate in ~12 countries, followed by further global expansion, in academic centers, community hospitals, and private clinics with regular MTBs. By 2026, WAYFIND-R aims to enroll ~15,000 patients. A key component is the potential to explore global and local data within a secure analytics platform (also known as a trusted research environment), accessed through a visual dashboard to find treatment decisions for similar patients elsewhere. Researchers can access the patient-level dataset for correlative analyses, after independent scientific committee approval. WAYFIND-R’s methodological/operational aspects were guided by external experts and the European Medicines Agency. As of October 2021, 11 patients have been enrolled. WAYFIND-R will inform on best practice for NGS-based treatment decisions by clinicians, foster global collaborations between cancer centers (enabling robust conclusions regarding outcome data to be drawn), aid understanding of disparities in patients’ access to advanced diagnostics and therapies, and ultimately drive advances in precision oncology. Citation Format: Christophe Le Tourneau, Allan Hackshaw, Jean-Yves Blay, Jan Geissler, Clare Turnbull, Camille Perret, Olga Skatkova, Martina von Meyenn, Rodrigo Dienstmann. WAYFIND-R: Delivering a high-quality real-world data (RWD) global registry of patients diagnosed with a solid tumor and profiled with next-generation sequencing (NGS) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4094.

Global Impact of Different Blood Pressure Thresholds in 4 021 690 Participants of the May Measurement Month Initiative.

Influential guidelines currently define hypertension at different thresholds of blood pressure (BP). The global May Measurement Month initiative provides a unique opportunity to estimate the potential consequences of universal lowering of BP thresholds on the prevalence of hypertension based on a large, real-world, patient-level data set. The average of the second and third of 3 attended BP readings after 5 minutes of rest from 4 021 690 standardized, opportunistic BP screenings in various settings of the 2017 to 2019 May Measurement Month initiatives from 104 countries were analyzed to assess the demographic and phenotypic impact of 3 defined BP thresholds. The age- and sex-standardized median of the relative increase in rate of hypertension with a change of thresholds from ≥140/≥90 to ≥130/≥80 mm Hg was 72.3% (interquartile range, 59.3%-91.3%) among the participating countries. With a change from ≥140/≥90 to ≥120/≥70 mm Hg, the median range was 162.6% (interquartile range, 132.8%-202.6%). The increase in rate of hypertension was most pronounced in low-income (95.3% and 203.9%) and least pronounced in high-income countries (71.6% and 167.1%). The impact of a universal application of lower BP thresholds for the definition of hypertension would have an enormous impact on the prevalence of hypertension globally with large disparities between countries and substantial heterogeneity of demographic and phenotypic patterns, which should be confirmed in dedicated population studies. While focusing on lower BP targets may constitute an opportunity for early intervention, the potential socioeconomic consequences may pose unsurmountable obstacles for most health care systems worldwide.