Using machine learning to link electronic health records in cancer registries: On the tradeoff between linkage quality and manual effort.

BACKGROUND: Cancer registries link a large number of electronic health records reported by medical institutions to already registered records of the matching individual and tumor. Records are automatically linked using deterministic and probabilistic approaches; machine learning is rarely used. Records that cannot be matched automatically with sufficient accuracy are typically processed manually. For application, it is important to know how well record linkage approaches match real-world records and how much manual effort is required to achieve the desired linkage quality. We study the task of linking reported records to the matching registered tumor in cancer registries. METHODS: We compare the tradeoff between linkage quality and manual effort of five machine learning methods (logistic regression, random forest, gradient boosting, neural network, and a stacked method) to a deterministic baseline. The record linkage methods are compared in a two-class setting (no-match/ match) and a three-class setting (no-match/ undecided/ match). A cancer registry collected and linked the dataset consisting of categorical variables matching 145,755 reported records with 33,289 registered tumors. RESULTS: In the two-class setting, the gradient boosting, neural network, and stacked models have higher accuracy and F1 score (accuracy: 0.968-0.978, F1 score: 0.983-0.988) than the deterministic baseline (accuracy: 0.964, F1 score: 0.980) when the same records are manually processed (0.89% of all records). In the three-class setting, these three machine learning methods can automatically process all reported records and still have higher accuracy and F1 score than the deterministic baseline. The linkage quality of the machine learning methods studied, except for the neural network, increase as the number of manually processed records increases. CONCLUSION: Machine learning methods can significantly improve linkage quality and reduce the manual effort required by medical coders to match tumor records in cancer registries compared to a deterministic baseline. Our results help cancer registries estimate how linkage quality increases as more records are manually processed.

The impact of cross-jurisdictional patient flows on ascertainment of hospitalisations and cardiac procedures for ST-segment-elevation myocardial infarction in an Australian population.

Introduction: The patient journey for residents of New South Wales (NSW) Australia with ST-elevation myocardial infarction (STEMI) often involves transfer between hospitals and these can include stays in hospitals in other jurisdictions. Objective: To estimate the change in enumeration of STEMI hospitalisations and time to subsequent cardiac procedures for NSW residents using cross-jurisdictional linkage of administrative health data. Methods: Records for NSW residents aged 20 years and over admitted to hospitals in NSW and four adjacent jurisdictions (Australian Capital Territory, Queensland, South Australia, and Victoria) between 1 July 2013 and 30 June 2018 with a principal diagnosis of STEMI were linked with records of the Australian Government Medicare Benefits Schedule (MBS). The number of STEMI hospitalisations, and rates of angiography, percutaneous coronary intervention and coronary artery bypass graft were compared for residents of different local health districts within NSW with and without inclusion of cross-jurisdictional data. Results: Inclusion of cross-jurisdictional hospital and MBS data increased the enumeration of STEMI hospitalisations for NSW residents by 8% (from 15,420 to 16,659) and procedure rates from 85.6% to 88.2%. For NSW residents who lived adjacent to a jurisdictional border, hospitalisation counts increased by up to 210% and procedure rates by up to 70 percentage points. Conclusions: Cross-jurisdictional linked hospital data is essential to understand patient journeys of NSW residents who live in border areas and to evaluate adherence to treatment guidelines for STEMI. MBS data are useful where hospital data are not available and for procedures that may be conducted in out-patient settings.

[Methodology and Attribution Success of a Data Linkage of Clinical Registry Data with Health Insurance Data]. / Methodik und Zuordnungserfolg eines Linkage von Daten klinischer Krebsregister mit Abrechnungsdaten gesetzlicher Krankenkassen.

BACKGROUND: The aim of the project "Effectiveness of care in oncological centres" (WiZen), funded by the innovation fund of the federal joint committee, is to investigate the effectiveness of certification in oncology. The project uses nationwide data from the statuory health insurance AOK and data from clinical cancer registries from three different federal states from 2006-2017. To combine the strengths of both data sources, these will be linked for eight different cancer entities in compliance with data protection regulations. METHODS: Data linkage was performed using indirect identifiers and validated using the health insurance's patient ID ("Krankenversichertennummer") as a direct identifier and gold standard. This enables quantification of the quality of different linkage variants. Sensitivity and specificity as well as hit accuracy and a score addressing the quality of the linkage were used as evaluation criteria. The distributions of relevant variables resulting from the linkage were validated against the original distributions in the individual datasets. RESULTS: Depending on the combination of indirect identifiers, we found a range of 22,125 to 3,092,401 linkage hits. An almost perfect linkage could be achieved by combining information on cancer type, date of birth, gender and postal code. A total of 74,586 one-to-one linkages were achieved with these characteristics. The median hit quality for the different entities was more than 98%. In addition, both the age and sex distributions and the dates of death, if any, showed a high degree of agreement. DISCUSSION AND CONCLUSION: SHI and cancer registry data can be linked with high internal and external validity at the individual level. This robust linkage enables completely new possibilities for analysis through simultaneous access to variables from both data sets ("the best of both worlds"): Information on the UICC stage that stems from the registries can now be combined, for instance, with comorbidities from the SHI data at the individual level. Due to the use of readily available variables and the high success of the linkage, our procedure constitutes a promising method for future linkage processes in health care research.

Linkage of multiple electronic health record datasets using a 'spine linkage' approach compared with all 'pairwise linkages'.

BACKGROUND: Methods for linking records between two datasets are well established. However, guidance is needed for linking more than two datasets. Using all 'pairwise linkages'-linking each dataset to every other dataset-is the most inclusive, but resource-intensive, approach. The 'spine' approach links each dataset to a designated 'spine dataset', reducing the number of linkages, but potentially reducing linkage quality. METHODS: We compared the pairwise and spine linkage approaches using real-world data on patients undergoing emergency bowel cancer surgery between 31 October 2013 and 30 April 2018. We linked an administrative hospital dataset (Hospital Episode Statistics; HES) capturing patients admitted to hospitals in England, and two clinical datasets comprising patients diagnosed with bowel cancer and patients undergoing emergency bowel surgery. RESULTS: The spine linkage approach, with HES as the spine dataset, created an analysis cohort of 15 826 patients, equating to 98.3% of the 16 100 patients identified using the pairwise linkage approach. There were no systematic differences in patient characteristics between these analysis cohorts. Associations of patient and tumour characteristics with mortality, complications and length of stay were not sensitive to the linkage approach. When eligibility criteria were applied before linkage, spine linkage included 14 509 patients (90.0% compared with pairwise linkage). CONCLUSION: Spine linkage can be used as an efficient alternative to pairwise linkage if case ascertainment in the spine dataset and data quality of linkage variables are high. These aspects should be systematically evaluated in the nominated spine dataset before spine linkage is used to create the analysis cohort.

Record linkage of claims and cancer registries data-Evaluation of a deterministic linkage approach based on indirect personal identifiers.

PURPOSE: In Germany, record linkage of claims and cancer registry data is cost- and time-consuming, since up until recently no unique personal identifier was available in both data sources. The aim of this study was to evaluate the feasibility and performance of a deterministic linkage procedure based on indirect personal identifiers included in the data sources. METHODS: We identified users of glucose-lowering drugs with residence in four federal states in Northern and Southern Germany (Bavaria, Bremen, Hamburg, Lower Saxony) in the German Pharmacoepidemiological Research Database (GePaRD) and assessed colorectal and thyroid cancer cases. Cancer registries of the federal states selected all colorectal and thyroid cancer cases between 2004 and 2015. A deterministic linkage approach was performed based on indirect personal identifiers such as year of birth, sex, area of residence, type of cancer and an absolute difference between the dates of cancer diagnosis in both data sources of at most 90 days. Results were compared to a probabilistic linkage using "direct" personal identifiers (gold standard). RESULTS: The deterministic linkage procedure yielded a sensitivity of 71.8% for colorectal cancer and 66.6% for thyroid cancer. For thyroid cancer, the sensitivity improved when using only inpatient diagnosis to define cancer in GePaRD (71.4%). Specificity was always above 99%. Using the probabilistic linkage to define cancer cases, the risk for colorectal cancer was estimated 10 percentage points lower than when using the deterministic approach. CONCLUSIONS: Sensitivity of the deterministic linkage approach appears to be too low to be considered as reasonable alternative to the probabilistic linkage procedure.

Linkage of National Congenital Heart Disease Audit data to hospital, critical care and mortality national data sets to enable research focused on quality improvement.

OBJECTIVES: To link five national data sets (three registries, two administrative) and create longitudinal healthcare trajectories for patients with congenital heart disease (CHD), describing the quality and the summary statistics of the linked data set. DESIGN: Bespoke linkage of record-level patient identifiers across five national data sets. Generation of spells of care defined as periods of time-overlapping events across the data sets. SETTING: National Congenital Heart Disease Audit (NCHDA) procedures in public (National Health Service; NHS) hospitals in England and Wales, paediatric and adult intensive care data sets (Paediatric Intensive Care Audit Network; PICANet and the Case Mix Programme from the Intensive Care National Audit & Research Centre; ICNARC-CMP), administrative hospital episodes (hospital episode statistics; HES inpatient, outpatient, accident and emergency; A&E) and mortality registry data. PARTICIPANTS: Patients with any CHD procedure recorded in NCHDA between April 2000 and March 2017 from public hospitals. PRIMARY AND SECONDARY OUTCOME MEASURES: Primary: number of linked records, number of unique patients and number of generated spells of care. Secondary: quality and completeness of linkage. RESULTS: There were 143 862 records in NCHDA relating to 96 041 unique patients. We identified 65 797 linked PICANet patient admissions, 4664 linked ICNARC-CMP admissions and over 6 million linked HES episodes of care (1.1M inpatient, 4.7M outpatient). The linked data set had 4 908 153 spells of care after quality checks, with a median (IQR) of 3.4 (1.8-6.3) spells per patient-year. Where linkage was feasible (in terms of year and centre), 95.6% surgical procedure records were linked to a corresponding HES record, 93.9% paediatric (cardiac) surgery procedure records to a corresponding PICANet admission and 76.8% adult surgery procedure records to a corresponding ICNARC-CMP record. CONCLUSIONS: We successfully linked four national data sets to the core data set of all CHD procedures performed between 2000 and 2017. This will enable a much richer analysis of longitudinal patient journeys and outcomes. We hope that our detailed description of the linkage process will be useful to others looking to link national data sets to address important research priorities.

Optimizing the Retrieval of the Vital Status of Cancer Patients for Health Data Warehouses by Using Open Government Data in France.

Electronic Medical Records (EMR) and Electronic Health Records (EHR) are often missing critical information about the death of a patient, although it is an essential metric for medical research in oncology to assess survival outcomes, particularly for evaluating the efficacy of new therapeutic approaches. We used open government data in France from 1970 to September 2021 to identify deceased patients and match them with patient data collected from the Institut de Cancérologie de l'Ouest (ICO) data warehouse (Integrated Center of Oncology-the third largest cancer center in France) between January 2015 and November 2021. To meet our objective, we evaluated algorithms to perform a deterministic record linkage: an exact matching algorithm and a fuzzy matching algorithm. Because we lacked reference data, we needed to assess the algorithms by estimating the number of homonyms that could lead to false links, using the same open dataset of deceased persons in France. The exact matching algorithm allowed us to double the number of dates of death in the ICO data warehouse, and the fuzzy matching algorithm tripled it. Studying homonyms assured us that there was a low risk of misidentification, with precision values of 99.96% for the exact matching and 99.68% for the fuzzy matching. However, estimating the number of false negatives proved more difficult than anticipated. Nevertheless, using open government data can be a highly interesting way to improve the completeness of the date of death variable for oncology patients in data warehouses.

Body size in early life and the risk of postmenopausal breast cancer.

BACKGROUND: Greater early life adiposity has been reported to reduce postmenopausal breast cancer risk but it is unclear whether this association varies by tumour characteristics. We aimed to assess associations of early life body size with postmenopausal breast cancer and its subtypes, allowing for body size at other ages. METHODS: A total of 342,079 postmenopausal UK women who reported their body size at age 10, clothes size at age 20, and body mass index (BMI) at baseline (around age 60) were followed by record linkage to national databases for cancers and deaths. Cox regression yielded adjusted relative risks (RRs) of breast cancer, overall and by tumour subtype, in relation to body size at different ages. RESULTS: During an average follow-up of 14 years, 15,506 breast cancers were diagnosed. After adjustment for 15 potential confounders, greater BMI at age 60 was associated with an increased risk of postmenopausal breast cancer (RR per 5 kg/m2=1.20, 95%CI 1.18-1.22) whereas greater adiposity in childhood and, to a lesser extent, early adulthood, was associated with a reduced risk (0.70, 0.66-0.74, and 0.92, 0.89-0.96, respectively). Additional adjustment for midlife BMI strengthened associations with BMI at both age 10 (0.63, 0.60-0.68) and at age 20 (0.78, 0.75-0.81). The association with midlife adiposity was confined to hormone sensitive subtypes but early life adiposity had a similar impact on the risk of all subtypes. CONCLUSION: Early life and midlife adiposity have opposite effects on postmenopausal breast cancer risk and the biological mechanisms underlying these associations are likely to differ.

Serious infections in patients with rheumatoid arthritis and psoriatic arthritis treated with tumour necrosis factor inhibitors: data from register linkage of the NOR-DMARD study.

OBJECTIVES: To estimate the incidence of serious infections (SIs) in patients with rheumatoid arthritis (RA) and psoriatic arthritis (PsA) treated with tumour necrosis factor inhibitor (TNFi), and compare risk of SIs between patients with RA and PsA. METHODS: We included patients with RA and PsA from the NORwegian-Disease Modifying Anti-Rheumatic Drug registry starting TNFi treatment. Crude incidence rates (IRs) and IR ratio for SIs were calculated. The risk of SIs in patients with RA and PsA was compared using adjusted Cox-regression models. RESULTS: A total of 3169 TNFi treatment courses (RA/PsA: 1778/1391) were identified in 2359 patients. Patients with RA were significantly older with more extensive use of co-medication. The crude IRs for SIs were 4.17 (95% CI 3.52 to 4.95) in patients with RA and 2.16 (95% CI 1.66 to 2.81) in patients with PsA. Compared with the patients with RA, patients with PsA had a lower risk of SIs (HR 0.59, 95% CI 0.41 to 0.85, p=0.004) in complete set analysis. The reduced risk in PsA versus RA remained significant after multiple adjustments and consistent across strata based on age, gender and disease status. CONCLUSIONS: Compared with patients with RA, the risk of SIs was significantly lower in patients with PsA during TNFi exposure.

The role of all-payer claims databases to expand central cancer registries: Experience from Colorado.

OBJECTIVE: To evaluate the quality of a multiyear linkage between the Colorado all-payer claims database (APCD) and the Colorado Central Cancer Registry. DATA SOURCES: Secondary 2012-2017 data from the APCD and the Colorado Cancer Registry. STUDY DESIGN: Descriptive analysis of the proportion of cases captured by the linkage in relation to the cases reported by the registry. DATA COLLECTION/EXTRACTION METHODS: We used probabilistic linkage to combine records from both data sources for all patients diagnosed with cancer. RESULTS: We successfully linked 93% of the 146,884 patients in the registry. Approximately 63% of linked patients were perfect matches on five identifiers. Of partial matches, 81.6% were matched on four identifiers with missing or partial Social Security Numbers. The linkage rate was lower for uninsured patients at diagnosis (74.7%) or patients with private plans (89.4%) but close to 100% for Medicare and Medicaid enrollees. Most of the 29% of patients who did not have claims at the time of diagnosis were covered by private plans that may not submit claims. CONCLUSIONS: APCD-registry linkages are a promising source of data to conduct population-based research from multiple payers. However, not all payers submit claims, and the quality of the data may vary by state.

[Linkage of claims data with data from epidemiological cancer registries: possibilities and limitations in the German federal states]. / Verknüpfung von Abrechnungsdaten gesetzlicher Krankenkassen mit Daten epidemiologischer Krebsregister: länderspezifische Möglichkeiten und Limitationen.

BACKGROUND: In recent years, there has been an increasing demand for the reuse of research data in accordance with the so-called FAIR principles. This would allow researchers to conduct projects on a broader data basis and to investigate new research questions by linking different data sources. OBJECTIVES: We explored if nationwide linking of claims data from statutory health insurances (SHI) with data from population-based cancer registries can be used to obtain additional information on cancer that is missing in claims data and to assess the validity of SHI tumour diagnoses. This paper focuses on describing the specific requirements of German federal states for such data linkage. MATERIALS AND METHODS: The Pharmacoepidemiological Research Database GePaRD at the Leibniz Institute for Prevention Research and Epidemiology - BIPS and six cancer registries were used as data sources. The logistically complex direct linkage was compared with a less complex indirect linkage. For this purpose, permission had to be obtained for GePaRD and for each cancer registry from the respective responsible authority. RESULTS: Regarding the linkage of cancer registry data with GePaRD, the cancer registries showed profound differences in the modalities for data provision, ranging from a complete rejection to an uncomplicated implementation of linkage procedures. DISCUSSION: In Germany, a consistent legal framework is needed to adequately enable the reuse and record linkage of personal health data for research purposes according to the FAIR principles. The new law on the consolidation of cancer registry data could provide a remedy regarding the linkage of cancer registry data with other data sources.

An Evaluation of Dose-Related HPV Vaccine Effectiveness Using Central Registries in Michigan.

BACKGROUND: Human papillomavirus (HPV) vaccine effectiveness (VE) evaluations provide important information for vaccination programs. We established a linkage between statewide central registries in Michigan to estimate HPV VE against in situ and invasive cervical lesions (CIN3+). METHODS: We linked females in Michigan's immunization and cancer registries using birth records to establish a cohort of 773,193 women with known vaccination history, of whom 3,838 were diagnosed with CIN3+. Residential address histories from a stratified random sample were used to establish a subcohort of 1,374 women without CIN3+ and 2,900 with CIN3+ among continuous Michigan residents. VE and 95% confidence intervals (CI) were estimated using cohort and case-cohort methods for up-to-date (UTD) vaccination and incomplete vaccination with 1 and 2 doses, and stratified by age at vaccination. RESULTS: Both analytic approaches demonstrated lower CIN3+ risk with UTD and non-UTD vaccination vs. no vaccination. The cohort analysis yielded VE estimates of 66% (95% CI, 60%-71%) for UTD, 33% (95% CI, 18%-46%) for 2 doses-not UTD, and 40% (95% CI, 27%-50%) for 1 dose. The case-cohort analysis yielded VE estimates of 72% (95% CI, 64%-79%) for UTD, 39% (95% CI, 10%-58%) for 2 doses-not UTD, and 48% (95% CI, 25%-63%) for 1 dose. VE was higher for vaccination at age <20 than ≥20 years. CONCLUSIONS: The statewide registry linkage found significant VE against CIN3+ with incomplete HPV vaccination, and an even higher VE with UTD vaccination. IMPACT: Future VE evaluations by number of doses for women vaccinated at younger ages may further clarify dose-related effectiveness.

Associations of Community Water Lead Concentrations with Hemoglobin Concentrations and Erythropoietin-Stimulating Agent Use among Patients with Advanced CKD.

BACKGROUND: Although patients with kidney disease may be particularly susceptible to the adverse health effects associated with lead exposure, whether levels of lead found commonly in drinking water are associated with adverse outcomes in patients with ESKD is not known. METHODS: To investigate associations of lead in community water systems with hemoglobin concentrations and erythropoietin stimulating agent (ESA) use among incident patients with ESKD, we merged data from the Environmental Protection Agency (EPA) Safe Drinking Water Information System (documenting average 90th percentile lead concentrations in community water systems during 5 years before dialysis initiation, according to city of residence) with patient-level data from the United States Renal Data System. RESULTS: Among 597,968 patients initiating dialysis in the United States in 2005 through 2017, those in cities with detectable lead levels in community water had significantly lower pre-ESKD hemoglobin concentrations and more ESA use per 0.01 mg/L increase in 90th percentile water lead. Findings were similar for the 208,912 patients with data from the first month of ESKD therapy, with lower hemoglobin and higher ESA use per 0.01 mg/L higher lead concentration. These associations were observed at lead levels below the EPA threshold (0.015 mg/L) that mandates regulatory action. We also observed environmental inequities, finding significantly higher water lead levels and slower declines over time among Black versus White patients. CONCLUSIONS: This first nationwide analysis linking EPA water supply records to patient data shows that even low levels of lead that are commonly encountered in community water systems throughout the United States are associated with lower hemoglobin levels and higher ESA use among patients with advanced kidney disease.

Record Linkage in Clinical Cancer Registration: Experiences and Findings from Lower Saxony.

In cancer registries, record linkage procedures are used to link records of the same patient from different health care providers. In the Clinical Cancer Registry of Lower Saxony, a multi-level combination of exact assignment using the statutory health insurance number and a probabilistic procedure with control numbers and address data is applied. The procedure implemented in the register application assigns the incoming messages in this way as far as possible automatically. The aim of the observation carried out was to check the efficiency of the match variables and threshold values used, above which manual assignment is required. Weak points were identified and approaches to solutions were developed.

Identifying maternal deaths with the use of hospital data versus death certificates: a retrospective population-based study.

BACKGROUND: Accurate identification of maternal deaths is paramount for audit and policy purposes. Our aim was to determine the accuracy and completeness of data on maternal deaths in hospital and those recorded on a death certificate, and the level of agreement between the 2 data sources. METHODS: We conducted a retrospective population-based study using data for Ontario, Canada, from Apr. 1, 2002, to Dec. 31, 2015. We used Canadian Institute for Health Information (CIHI) databases to identify deaths during inpatient, emergency department and same-day surgery encounters. We captured Vital Statistics deaths in the Office of the Registrar General, Deaths (ORGD) data set. Deaths were considered within 42 days and within 365 days after a pregnancy outcome (live birth, miscarriage, ectopic pregnancy or induced abortion) for all multiple and singleton pregnancies. We calculated agreement statistics and 95% confidence intervals (CIs). RESULTS: Among 1 679 455 live births and stillbirths, 398 pregnancy-related deaths in the ORGD data set were mapped to a birth in CIHI databases, and 77 (16.2%) were not. Among 2 039 849 recognized pregnancies, 534 pregnancy-related deaths in the ORGD data set were linked to CIHI records, and 68 (11.3%) were not. Among live births and stillbirths, after pregnancy-related deaths in the ORGD data set not matched to a maternal death in the CIHI databases were removed, concordance measures between CIHI and ORGD records for maternal death within 42 days after delivery included a κ value of 0.87 (95% CI 0.82-0.91) and positive percent agreement of 0.88 (95% CI 0.83-0.94). The corresponding measures were similar for maternal death within 42 days after the end of a recognized pregnancy. When unlinked pregnancy-related deaths in the ORGD data set were retained, agreement measures declined for death within 42 days after a live birth or stillbirth (κ = 0.68, 95% CI 0.62-0.74). For maternal death within 365 days after a live birth or stillbirth, or after the end of a recognized pregnancy, the concordance statistics were generally favourable when unlinked pregnancy-related deaths in the ORGD data set were removed but were substantially declined when they were retained. INTERPRETATION: Maternal mortality cannot be ascertained solely with the use of hospital data, including beyond 42 days after the end of pregnancy. To improve linkage, we propose including health insurance numbers on provincial and territorial medical death certificates.

Probabilistic linkage without personal information successfully linked national clinical datasets.

BACKGROUND: Probabilistic linkage can link patients from different clinical databases without the need for personal information. If accurate linkage can be achieved, it would accelerate the use of linked datasets to address important clinical and public health questions. OBJECTIVE: We developed a step-by-step process for probabilistic linkage of national clinical and administrative datasets without personal information, and validated it against deterministic linkage using patient identifiers. STUDY DESIGN AND SETTING: We used electronic health records from the National Bowel Cancer Audit and Hospital Episode Statistics databases for 10,566 bowel cancer patients undergoing emergency surgery in the English National Health Service. RESULTS: Probabilistic linkage linked 81.4% of National Bowel Cancer Audit records to Hospital Episode Statistics, vs. 82.8% using deterministic linkage. No systematic differences were seen between patients that were and were not linked, and regression models for mortality and length of hospital stay according to patient and tumour characteristics were not sensitive to the linkage approach. CONCLUSION: Probabilistic linkage was successful in linking national clinical and administrative datasets for patients undergoing a major surgical procedure. It allows analysts outside highly secure data environments to undertake linkage while minimizing costs and delays, protecting data security, and maintaining linkage quality.

Integrating Electronic Health Record, Cancer Registry, and Geospatial Data to Study Lung Cancer in Asian American, Native Hawaiian, and Pacific Islander Ethnic Groups.

BACKGROUND: A relatively high proportion of Asian American, Native Hawaiian, and Pacific Islander (AANHPI) females with lung cancer have never smoked. We used an integrative data approach to assemble a large-scale cohort to study lung cancer risk among AANHPIs by smoking status with attention to representation of specific AANHPI ethnic groups. METHODS: We leveraged electronic health records (EHRs) from two healthcare systems-Sutter Health in northern California and Kaiser Permanente Hawai'i-that have high representation of AANHPI populations. We linked EHR data on lung cancer risk factors (i.e., smoking, lung diseases, infections, reproductive factors, and body size) to data on incident lung cancer diagnoses from statewide population-based cancer registries of California and Hawai'i for the period between 2000 and 2013. Geocoded address data were linked to data on neighborhood contextual factors and regional air pollutants. RESULTS: The dataset comprises over 2.2 million adult females and males of any race/ethnicity. Over 250,000 are AANHPI females (19.6% of the female study population). Smoking status is available for over 95% of individuals. The dataset includes 7,274 lung cancer cases, including 613 cases among AANHPI females. Prevalence of never-smoking status varied greatly among AANHPI females with incident lung cancer, from 85.7% among Asian Indian to 14.4% among Native Hawaiian females. CONCLUSION: We have developed a large, multilevel dataset particularly well-suited to conduct prospective studies of lung cancer risk among AANHPI females who never smoked. IMPACT: The integrative data approach is an effective way to conduct cancer research assessing multilevel factors on cancer outcomes among small populations.

Patterns of care for men with prostate cancer: the 45 and Up Study.

OBJECTIVES: To describe patterns of care in New South Wales for men with prostate cancer, and to ascertain factors associated with receiving different types of treatment. DESIGN: Individual patient data record linkage study. SETTING, PARTICIPANTS: 4003 New South Wales men aged 45 years or more enrolled in the population-based 45 and Up Study in whom prostate cancer was first diagnosed during 2006-2013. MAIN OUTCOME MEASURES: Prostate cancer treatment type received; factors statistically associated with treatment received; proportions of patients who consulted radiation oncologists prior to treatment. RESULTS: In total, 1619 of 4003 patients underwent radical prostatectomy (40%), 893 external beam radiotherapy (EBRT) (22%), 183 brachytherapy (5%), 87 chemotherapy (2%), 373 androgen deprivation therapy alone (9%), and 848 no active treatment (21%). 205 of 1628 patients who had radical prostatectomies (13%) had radiation oncology consultations prior to surgery. Radical prostatectomy was more likely for patients aged 45-59 years, with regional stage disease, living 100 km or more from the nearest radiotherapy centre, having partners, or having private health insurance, while lower physical functioning, obesity, and living in areas of greater socio-economic disadvantage reduced the likelihood. EBRT was more likely for patients aged 70-79 years, with non-localised or unknown stage disease, living less than 100 km from the nearest radiotherapy centre, or not having private health insurance, while the likelihood was lower for patients aged 45-59 years or more than 80 years and for those who had several comorbid conditions. CONCLUSIONS: Men with prostate cancer were twice as likely to have radical prostatectomy as to receive EBRT, and fewer than one in seven had consulted radiation oncologists prior to prostatectomy. The treatment received was influenced by several socio-demographic factors. Given the treatment-specific side effects and costs, policies that affect access to different treatments for prostate cancer should be reviewed.

Toward an all-inclusive trauma system in Central South Ontario: development of the Trauma-System Performance Improvement and Knowledge Exchange (T-SPIKE) project.

Background: There is currently no integrated data system to capture the true burden of injury and its management within Ontario's regional trauma networks (RTNs), largely owing to difficulties in identifying these patients across the multiple health care provider records. Our project represents an iterative effort to create the ability to chart the course of care for all injured patients within the Central South RTN. Methods: Through broad stakeholder engagement of major health care provider organizations within the Central South RTN, we obtained research ethics board approval and established data-sharing agreements with multiple agencies. We tested identification of trauma cases from Jan. 1 to Dec. 31, 2017, and methods to link patient records between the various echelons of care to identify barriers to linkage and opportunities for administrative solutions. Results: During 2017, potential trauma cases were identified within ground paramedic services (23 107 records), air medical transport services (196 records), referring hospitals (7194 records) and the lead trauma hospital trauma registry (1134 records). Linkage rates for medical records between services ranged from 49% to 92%. Conclusion: We successfully conceptualized and provided a preliminary demonstration of an initiative to collect, collate and accurately link primary data from acute trauma care providers for certain patients injured within the Central South RTN. Administration-level changes to the capture and management of trauma data represent the greatest opportunity for improvement.

Contexte: On ne dispose actuellement d'aucun système intégré de gestion des données pour évaluer le fardeau réel des traumatismes et de leur gestion dans les réseaux régionaux de traumatologie (RRT) en Ontario, en bonne partie en raison de la difficulté d'identifier les cas parmi la multiplicité des dossiers d'intervenants médicaux. Notre projet représente un effort itératif pour créer la capacité de cartographier le parcours de soin de tous les polytraumatisés du RRT de la région Centre-Sud. Méthodes: Grâce à l'engagement général des intervenants des grandes organisations de santé du RRT de la région Centre-Sud, nous avons obtenu l'approbation d'un comité d'éthique de la recherche et conclu des accords de partage des données avec plusieurs agences. Nous avons testé l'identification des cas de traumatologie du 1er janvier au 31 décembre 2017 et les méthodes de liaison des dossiers de patients entre les divers échelons de soin pour identifier les obstacles à la liaison et leurs solutions administratives possibles. Résultats: Au cours de 2017, les cas de traumatologie potentiels ont été identifiés auprès des services ambulanciers terrestres (23 107 dossiers), des services de transport médical aérien (196 dossiers), des hôpitaux référents (7194 dossiers) et du registre hospitalier principal de traumatologie (1134 dossiers). Les taux de liaison entre les différents services pour les dossiers médicaux variaient de 49 % à 92 %. Conclusion: Nous avons conceptualisé et présenté avec succès la démonstration préliminaire d'un projet visant à recueillir, colliger et relier avec justesse les données primaires des intervenants en traumatologie aiguë pour certains patients blessés du RRT du Centre-Sud. Des changements administratifs centrés sur la saisie et la gestion des données de traumatologie représentent la meilleure voie vers une amélioration.

Trends in predominant causes of death in individuals with and without diabetes in England from 2001 to 2018: an epidemiological analysis of linked primary care records.

BACKGROUND: The prevalence of diabetes has increased in the UK and other high-income countries alongside a substantial decline in cardiovascular mortality. Yet data are scarce on how these trends have changed the causes of death in people with diabetes who have traditionally died primarily of vascular causes. We estimated how all-cause mortality and cause-specific mortality in people with diabetes have changed over time, how the composition of the mortality burden has changed, and how this composition compared with that of the non-diabetes population. METHODS: In this epidemiological analysis of primary care records, we identified 313 907 individuals with diabetes in the Clinical Practice Research Datalink, a well described primary care database, between 2001 to 2018, and linked these data to UK Office for National Statistics mortality data. We assembled serial cross sections with longitudinal follow-up to generate a mixed prevalence and incidence study population of patients with diabetes. We used discretised Poisson regression models to estimate annual death rates for deaths from all causes and 12 specific causes for men and women with diabetes. We also identified age-matched and sex matched (1:1) individuals without diabetes from the same dataset and estimated mortality rates in this group. FINDINGS: Between Jan 1, 2001, and Oct 31, 2018, total mortality declined by 32% in men and 31% in women with diagnosed diabetes. Death rates declined from 40·7 deaths per 1000 person-years to 27·8 deaths per 1000 person-years in men and from 42·7 deaths per 1000 person-years to 29·5 deaths per 1000 person-years in women with diagnosed diabetes. We found similar declines in individuals without diabetes, hence the gap in mortality between those with and without diabetes was maintained over the study period. Cause-specific death rates declined in ten of the 12 cause groups, with exceptions in dementia and liver disease, which increased in both populations. The large decline in vascular disease death rates led to a transition from vascular causes to cancers as the leading contributor to death rates in individuals with diagnosed diabetes and to the gap in death rates between those with and without diabetes. INTERPRETATION: The decline in vascular death rates has been accompanied by a diversification of causes in individuals with diagnosed diabetes and a transition from vascular diseases to cancers as the leading contributor to diabetes-related death. Clinical and preventative approaches must reflect this trend to reduce the excess mortality risk in individuals with diabetes. FUNDING: Wellcome Trust.