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.

Joint use of epidemiological and hospital medico-administrative data to estimate prevalence. Application to French data on breast cancer.

BACKGROUND: Estimate complete, limited-duration, and hospital prevalence of breast cancer in a French Département covered by a population-based cancer registry and in whole France using complementary information sources. METHODS: Incidence data from a cancer registry, national incidence estimations for France, mortality data, and hospital medico-administrative data were used to estimate the three prevalence indices. The methods included a modelling of epidemiological data and a specific process of data extraction from medico-administrative databases. RESULTS: Limited-duration prevalence at 33 years was a proxy for complete prevalence only in patients aged less than 70 years. In 2007 and in women older than 15 years, the limited-duration prevalence at 33 years rate per 100,000 women was estimated at 2372 for Département Isère and 2354 for whole France. The latter rate corresponded to 613,000 women. The highest rate corresponded to women aged 65-74 years (6161 per 100,000 in whole France). About one third of the 33-year limited-duration prevalence cases were diagnosed five years before and about one fourth were hospitalized for breast-cancer-related care (i.e., hospital prevalence). In 2007, the rate of hospitalized women was 557 per 100,000 in whole France. Among the 120,310 women hospitalized for breast-cancer-related care in 2007, about 13% were diagnosed before 2004. CONCLUSION: Limited-duration prevalence (long- and short-term), and hospital prevalence are complementary indices of cancer prevalence. Their efficient direct or indirect estimations are essential to reflect the burden of the disease and forecast median- and long-term medical, economic, and social patient needs, especially after the initial treatment.

A Suitable Approach to Estimate Cancer Incidence in Area without Cancer Registry.

Objective. Use of cancer cases from registries and PMSI claims database to estimate Département-specific incidence of four major cancers. Methods. Case extraction used principal diagnosis then surgery codes. PMSI cases/registry cases ratios for 2004 were modelled then Département-specific incidence for 2007 estimated using these ratios and 2007 PMSI cases. Results. For 2007, only colon-rectum and breast cancer estimations were satisfactorily validated for infranational incidence not ovary and kidney cancers. For breast, the estimated national incidence was 50,578 cases and the incidence rate 98.6 cases per 100,000 person per year. For colon-rectum, incidence was 21,172 in men versus 18,327 in women and the incidence rate 38 per 100,000 versus 24.8. For ovary, the estimated incidence was 4,637 and the rate 8.6 per 100,000. For kidney, incidence was 6,775 in men versus 3,273 in women and the rate 13.3 per 100.000 versus 5.2. Conclusion. Incidence estimation using PMSI patient identifiers proved encouraging though still dependent on the assumption of uniform cancer treatments and coding.

Breast cancer incidence using administrative data: correction with sensitivity and specificity.

OBJECTIVE: To estimate breast cancer incidence in the general population using a method that corrects for lack of sensitivity and specificity in the identification of incident breast cancer in inpatient claims data. STUDY DESIGN AND SETTINGS: Two-phase study: phase 1 to identify incident cases in claims data, and phase 2 to estimate sensitivity and specificity in a subset of the population. Two algorithms (1: principal diagnosis; 2: principal diagnosis+specific surgery procedures) were used to identify incident cases in claims of women aged 20 years or older, living in a French district covered by a cancer registry. Sensitivity and specificity were estimated in one district and used to correct incident cases identified. RESULTS: The sensitivity and specificity for algorithms 1 and 2 were 69.0% and 99.89%, and 64.4% and 99.93%, respectively. In contrast to specificity, the sensitivity for both algorithms was lower for women younger than 40 years and older than 65 years. Cases reported by cancer registries were closer to cases identified with algorithm 2 (-3.2% to +20.1%) and to corrected numbers with algorithm 1 (-1% to +15%). CONCLUSION: To obtain reliable estimates of breast cancer incidence in the general population, sensitivity and specificity, which reflect medical and coding practice variations, are necessary.

Nosocomial bacteremia in very old patients: predictors of mortality.

BACKGROUND AND AIMS: Nosocomial Bacteremia (NB) is associated with high mortality in elderly patients. To determine specific prognostic factors for 7- and 30-day mortality in elderly patients with NB, we analysed the characteristics of 62 NB patients, retrospectively. METHODS: This retrospective study concerns 62 cases of NB diagnosed within a 3-year period in a geriatric department. Bacteremia is described according to CDC definitions. Epidemiological characteristics, co-morbidities, clinical (activities of daily living (ADL) before NB) and biological findings (neutrophil count, lymphocyte count, albuminemia before NB) were collected for each patient. A systemic clinical reaction was defined by the presence of one of the following parameters: chills, hypothermia <36 degrees C or hyperthermia >38.5 degrees C, or shock. Types of micro-organism and source of NB were also collected. All variables were analysed for mortality at day 7 (7-day mortality) and at day 30 (30-day mortality). RESULTS: The 7-day mortality rate was 21% and the 30-day rate was 45%. In multivariate analysis, 7-day mortality was only associated with the absence of systemic clinical reaction [OR 9.7 (3.7-25.7)]. Again, in multivariate analysis, 30-day mortality was associated with an ADL score <2 [OR 8.3 (4.3-16.4)] and cocci gram positive NB [OR= 3.6 (1.9-6.9)]. CONCLUSIONS: The absence of any systemic clinical reaction as a single independent predictor for 7-day mortality suggests either a poorer immune response to nosocomial bacteremia or a delay in diagnosis. Functional status was the strongest predictor for 30-day mortality. In this population, further prospective studies need to include these factors to evaluate predictors of mortality for serious infectious diseases.

A case of rhabdomyolysis with fatal outcome after a treatment with levofloxacin.

Fluoroquinolones are known to cause rhabdomyolysis. Levofloxacin is a recent fluoroquinolone and its muscular toxicity is not well documented. We describe the case of a 77-year-old female patient, who presented with an acute rhabdomyolysis after treatment with levofloxacin. She had a background of serious cardio-pulmonary disease. She received an oral ambulatory treatment with levofloxacin for pulmonary infection. After 6 days, she presented with severe rhabdomyolysis, resulting in complete anuria with hyperkalaemia, complicated with acute liver cytolysis and respiratory failure. The treatment was a daily repeated haemodialysis. She presented with a fatal myocardial infarction 13 days after admission. The medical history inclines us to strongly suspect levofloxacin as the cause of this severe adverse drug reaction. We also reviewed 27 other suspect cases reported in the database provided by the World Health Organization Collaborating Centre for Drug Monitoring (Uppsala, Sweden). We conclude that rhabdomyolysis can be a rare, severe adverse effect of levofloxacin, as well as the other fluoroquinolones.

[Management of acute community-acquired pneumonia in a health centre. Assessment of 101 cases using the retrospective clinical audit method]. / Prise en charge des pneumonies aiguës communautaires en établissement de santé. Evaluation de 101 cas par la méthode d'audit clinicque rétrospectif.

OBJECTIVE: To assess the conformity of practitioners' practices in the management of community acquired pneumonia with the French Agence Nationale d'Accréditation et d'Evaluation en Santé (Anaes) guidelines. METHODS: We retrospectively reviewed a random sample of 210 medical records which included a principal or associated diagnosis of pneumonia in a French university hospital. RESULTS: A hundred and one medical records were assessable. Sixty-two patients were high risk (Pneumonia Severity Index class IV or V of the prediction rule of Fine et al.), and 10 patients were admitted into an intensive care unit. The overall in-hospital mortality was 14 patients [8-22]. The level of care was appropriate according to the guidelines in 40 cases ([30-50)]. Seven patients did not require hospitalisation, 31 patients required admission into a medical department, 56 patients into an intensive care unit and 7 patients were managed in non specified conditions. Eighteen patients ([11-27]) had appropriate microbiologic investigations. Forty-three patients (([33-53]) received antibiotics within 8 hours of arrival. Empirical antibiotic treatment (dosage and molecule) was appropriate in 38 patients ([28-48]). There was no significant relationship between compliance with the guidelines and in-hospital mortality. CONCLUSION: The rate of conformity of practitioners' practices with the Anaes guidelines for management of community-acquired pneumonia is low in our hospital. It could be improved by active implementation of these guidelines.