Early prediction of in-hospital mortality utilizing multivariate predictive modelling of electronic medical records and socio-determinants of health of the first day of hospitalization.

BACKGROUND: In France an average of 4% of hospitalized patients die during their hospital stay. To aid medical decision making and the attribution of resources, within a few days of admission the identification of patients at high risk of dying in hospital is essential. METHODS: We used de-identified routine patient data available in the first 2 days of hospitalization in a French University Hospital (between 2016 and 2018) to build models predicting in-hospital mortality (at ≥ 2 and ≤ 30 days after admission). We tested nine different machine learning algorithms with repeated 10-fold cross-validation. Models were trained with 283 variables including age, sex, socio-determinants of health, laboratory test results, procedures (Classification of Medical Acts), medications (Anatomical Therapeutic Chemical code), hospital department/unit and home address (urban, rural etc.). The models were evaluated using various performance metrics. The dataset contained 123,729 admissions, of which the outcome for 3542 was all-cause in-hospital mortality and 120,187 admissions (no death reported within 30 days) were controls. RESULTS: The support vector machine, logistic regression and Xgboost algorithms demonstrated high discrimination with a balanced accuracy of 0.81 (95%CI 0.80-0.82), 0.82 (95%CI 0.80-0.83) and 0.83 (95%CI 0.80-0.83) and AUC of 0.90 (95%CI 0.88-0.91), 0.90 (95%CI 0.89-0.91) and 0.90 (95%CI 0.89-0.91) respectively. The most predictive variables for in-hospital mortality in all three models were older age (greater risk), and admission with a confirmed appointment (reduced risk). CONCLUSION: We propose three highly discriminating machine-learning models that could improve clinical and organizational decision making for adult patients at hospital admission.

Cohort profile for development of machine learning models to predict healthcare-related adverse events (Demeter): clinical objectives, data requirements for modelling and overview of data set for 2016-2018.

PURPOSE: In-hospital health-related adverse events (HAEs) are a major concern for hospitals worldwide. In high-income countries, approximately 1 in 10 patients experience HAEs associated with their hospital stay. Estimating the risk of an HAE at the individual patient level as accurately as possible is one of the first steps towards improving patient outcomes. Risk assessment can enable healthcare providers to target resources to patients in greatest need through adaptations in processes and procedures. Electronic health data facilitates the application of machine-learning methods for risk analysis. We aim, first to reveal correlations between HAE occurrence and patients' characteristics and/or the procedures they undergo during their hospitalisation, and second, to build models that allow the early identification of patients at an elevated risk of HAE. PARTICIPANTS: 143 865 adult patients hospitalised at Grenoble Alpes University Hospital (France) between 1 January 2016 and 31 December 2018. FINDINGS TO DATE: In this set-up phase of the project, we describe the preconditions for big data analysis using machine-learning methods. We present an overview of the retrospective de-identified multisource data for a 2-year period extracted from the hospital's Clinical Data Warehouse, along with social determinants of health data from the National Institute of Statistics and Economic Studies, to be used in machine learning (artificial intelligence) training and validation. No supplementary information or evaluation on the part of medical staff will be required by the information system for risk assessment. FUTURE PLANS: We are using this data set to develop predictive models for several general HAEs including secondary intensive care admission, prolonged hospital stay, 7-day and 30-day re-hospitalisation, nosocomial bacterial infection, hospital-acquired venous thromboembolism, and in-hospital mortality.

The burden of chronic pain for patients with osteoarthritis in Germany: a retrospective cohort study of claims data.

BACKGROUND: Osteoarthritis (OA) is a common condition that is often associated with chronic pain. Pain often leads patients to seek healthcare advice and treatment. In this retrospective cohort analysis of German longitudinal healthcare claims data, we aimed to explore the healthcare resource utilisation (HRU) and related healthcare costs for patients with OA who develop chronic pain. METHODS: Patient-level data was extracted from the German Institut für Angewandte Gesundheitsforschung (InGef) database. Insured persons (≥18 years) were indexed between January 2015 and December 2017 with a recent (none in the last 2 years) diagnosis of OA. HRU and costs were compared between patients categorised as with (identified via diagnosis or opioid prescription) and without chronic pain. Unweighted HRU (outpatient physician contacts, hospitalisations, prescriptions for physical therapy or psychotherapy, and incapacity to work) and healthcare costs (medication, medical aid/remedy, psychotherapy, inpatient and outpatient and sick pay in Euros [quartile 1, quartile 3]) were calculated per patient for the year following index. Due to potential demographic and comorbidity differences between the groups, inverse probability of treatment weighting (IPTW) was used to estimate weighted costs and rate ratio (RR; 95% confidence interval) of HRU by negative binomial regression modelling. RESULTS: Of 4,932,543 individuals sampled, 238,306 patients with OA were included in the analysis: 80,055 (34%) categorised as having chronic pain (24,463 via opioid prescription) and 158,251 (66%) categorised as not having chronic pain. The chronic pain cohort was slightly older, more likely to be female, and had more comorbidities. During the year following index, unweighted and IPTW-weighted HRU risk and healthcare costs were higher in patients with chronic pain vs those without for all categories. This led to a substantially higher total annual healthcare cost ─ observed mean; €6801 (1439, 8153) vs €3682 (791, 3787); estimated RR = 1.51 (1.36, 1.66). CONCLUSIONS: German patients with chronic pain and OA have higher healthcare costs and HRU than those with OA alone. Our findings suggest the need for better prevention and treatment of OA in order to reduce the incidence of chronic pain, and the resultant increase in disease burden experienced by patients.

Carotid IMT and Stiffness in the KiGGS 2 National Survey: Third-Generation Measurement, Quality Algorithms and Determinants of Completeness.

Carotid intima-media thickness (cIMT) and carotid stiffness (CS) are important markers of atherosclerotic risk in the young. We assessed a novel third-generation method for its applicability in large population-based epidemiologic studies to determine strengths, limitations, completeness and predictors of unsuccessful measurement. Four thousand seven hundred ninety-eight 14- to 31-y-old participants of the German KiGGS cohort, which is based on a nationally representative sample with 11-y follow-up, underwent B-mode ultrasound examinations of the left and right common carotid artery with semi-automatic edge detection and automatic electrocardiogram-gated real-time quality control based on a sophisticated snake algorithm and subpixel interpolation. Overall completeness was 98% for far wall cIMT and 89% for CS parameters. Plane-specific completeness varied from 92%-96% for far wall and from 64%-69% for near-wall cIMT. Obesity independently predicted unsuccessful cIMT and CS measurements with odds ratios of 12.67 (95% confidence interval: 5.50-29.19) and 7.30 (4.87-10.94) compared with non-overweight after adjustment for blood pressure, cholesterol, smoking, hazardous drinking, age, sex and sonographer. Inter- and intra-rater reliabilities of cIMT and CS parameters in a sample of 15 young adults were good or excellent. Third-generation cIMT and CS measurements in the young with semi-automatic edge-detection and automatic real-time quality control has been successfully standardized with high reliability and very high completeness in a national survey setting. This provides a strong methodological foundation for further validation of the predictive value of cIMT and CS for atherosclerotic risk in the young.

New data for action. Data collection for KiGGS Wave 2 has been completed.

The fieldwork of the second follow-up to the German Health Interview and Examination Survey for Children and Adolescents (KiGGS) was completed in August 2017. KiGGS is part of the Robert Koch Institute's Federal Health Monitoring. The study consists of the KiGGS cross-sectional component (a nationally representative, periodic cross-sectional survey of children and adolescents aged between 0 and 17) and the KiGGS cohort (the follow-up into adulthood of participants who took part in the KiGGS baseline study). KiGGS collects data on health status, health-related behaviour, psychosocial risk and protective factors, health care and the living conditions of children and adolescents in Germany. The first interview and examination survey (the KiGGS baseline study; undertaken between 2003 and 2006; n=17,641; age range: 0-17) was carried out in a total of 167 sample points in Germany. Physical examinations, laboratory analyses of blood and urine samples and various physical tests were conducted with the participants and, in addition, all parents and participants aged 11 or above were interviewed. The first follow-up was conducted via telephone-based interviews (KiGGS Wave 1 2009-2012; n=11,992; age range: 6-24) and an additional sample was included (n=4,455; age range: 0-6). KiGGS Wave 2 (2014-2017) was conducted as an interview and examination survey and consisted of a new, nationwide, representative cross-sectional sample of 0- to 17-year-old children and adolescents in Germany, and the second KiGGS cohort follow-up. The completion of the cross-sectional component of KiGGS Wave 2 means that the health of children and adolescents in Germany can now be assessed using representative data gained from three study waves. Trends can therefore be analysed over a period stretching to over ten years now. As the data collected from participants of the KiGGS cohort can be individually linked across the various surveys, in-depth analyses can be conducted for a period ranging from childhood to young adulthood and developmental processes associated with physical and mental health and the associated risk and protective factors can be explored. As such, KiGGS Wave 2 expands the resources available to health reporting, as well as policy planning and research, with regard to assessing the health of children and adolescents in Germany.