Implementation of a computerized system in pediatric wards to improve nutritional care: a cluster randomized trial.

BACKGROUND/OBJECTIVES: Malnutrition occurs frequently in hospitalized children. We aimed to assess whether a computerized system could lead to improved clinical practices in malnourished children. SUBJECTS/METHODS: Healthcare workers (242) from six departments in a pediatric university hospital participated in a cluster randomized trial, studying 1457 malnourished children hospitalized from September 2009 to August 2011. Following a baseline observational pre-intervention period, all departments were randomized into either intervention or control arms. A computerized malnutrition-screening system was implemented in the intervention group to automatically trigger a dietetic referral in real time. Furthermore, the nutrition support team conducted an awareness campaign with healthcare workers and a leadership-based strategy to reinforce the message during the entire study period. Adherence to practice guidelines (daily weights, investigation of etiology for malnutrition, management by a dietitian and application of refeeding protocols) was compared between pre- and post-intervention periods in both the intervention and trial arms. RESULTS: When compared with the pre-intervention period, the clinical practices were significantly improved within the intervention arm for every outcome (P<0.01), whereas remained unchanged in the control arm. In addition, during the post-intervention period, malnutrition etiology investigation by physicians (adjusted odds ratio (OR) of 4.4, 95% confidence interval (CI) 1.7-11.8, P=0.003) and management by a dietitian (OR 2.7, 95% CI 1.0-6.9, P=0.046) occurred more frequently in the intervention clusters. CONCLUSIONS: Implementation of an electronic system to detect malnutrition in real time was associated with a rapid improvement in clinical practices for better care of hospitalized children.

[Determination of the 120-day post prostatic biopsy mortality rate]. / Mesure de la mortalité à 120 jours après biopsie de la prostate.

OBJECTIVE: Concerning death-rates were reported following prostate biopsy but the lack of contexts in which event occurred makes it difficult to take any position. Therefore, we aimed to determine the 120-day post-biopsy mortality rate. MATERIAL AND METHODS: Between 2000 and 2011, 8804 men underwent prostate biopsy in the hospice civils de Lyon. We studied retrospectively, the mortality rate after each of the 11,816 procedures. Biopsies imputability was assessed by examining all medical records. Dates of death were extracted from our local patient management database, which is updated trimestrially with death notifications from the French National Institute for Statistics and Economic Studies. RESULTS: In our study 42 deaths occurred within 120days after 11,816 prostate biopsies (0.36%). Of the 42 records: 9 were lost to follow-up, 3 had no identifiable cause of death, 28 had an intercurrent event ruling out prostate biopsy as a cause of death. Only 2 deaths could be linked to biopsy. CONCLUSIONS: We reported at most 2 deaths possibly related to prostate biopsy over 11,816 procedures (0.02%). We confirmed the fact that prostate biopsies can be lethal but this rare outcome should not be considered as an argument against prostate screening given the circumstances in which it occurs. LEVEL OF EVIDENCE: 5.

[Critical analysis of French DRG based information system (PMSI) databases for the epidemiology of cancer: a longitudinal approach becomes possible]. / Analyse critique des données du PMSI pour l'épidémiologie des cancers : une approche longitudinale devient possible.

BACKGROUND: Use of French Diagnosis Related Groups (DRGs) program databases, apart from financial purposes, has recently been improved since a unique anonymous patient identification number has been created for each inpatient in administrative case mix database. Based on the work of the group for cancer epidemiological observation in the Rhône-Alpes area, (ONC-EPI group), we review the remaining difficulties in the use of DRG data for epidemiological purposes and we consider a longitudinal approach based on analysis of database over several years. We also discuss limitations of this approach. DIFFICULTIES: The main problems are related to a lack of quality of administrative data, especially coding of diagnoses. These errors come from missing or inappropriate codes, or not being in accordance with prioritization rules (causing an over- or under-reporting or inconsistencies in coding over time). One difficulty, partly due to the hierarchy of coding and the type of cancer, is the choice of an extraction algorithm. In two studies designed to estimate the incidence of cancer cared in hospitals (breast, colon-rectum, kidney, ovaries), a first algorithm, including a code of cancer as principal diagnosis with a selection of surgical procedures less performed than the second one including a code of cancer as principal diagnosis only, for which the number of hospitalizations per patient ratio was stable across time and space. The chaining over several years allows, by tracing the trajectory of the patient, to detect and correct inaccuracies, errors and missing values, and for incidence studies, to correct incident cases by removing prevalent cases. DISCUSSION: However, linkage, complete only since 2007, does not correct data in all cases. Ways of future improvement certainly pass through improved algorithms for case identification and especially by linking DRG data with other databases.

Identifying prevalent cases of breast cancer in the French case-mix databases.

OBJECTIVES: Little is known about cancer prevalence due to a lack of systematic recording of cancer patient follow-up data. To estimate the annual hospital prevalence of breast cancer in the general population of the Isère department (1.1 million inhabitants) in the Rhône-Alpes region, the second largest region in France (6 million inhabitants), we used the inpatient case-mix data, available in most European countries, to develop a method of cancer case identification. METHODS: A selection process was applied to the acute care hospital datasets among women aged 18 years or older, living in the Isère department and treated for breast cancer between 2004 and 2007. The first step in case selection was based on the national anonymous unique patient identifier. The second step consisted of retrieving all hospital stays for each case. The third step was designed to detect inconsistencies in the coding of the primary localization. An algorithm based on ICD-10 code for the hospital admission diagnosis was used to rule out hospitalizations unrelated to breast cancer. Five possible models for estimating prevalence were created combining selection steps with the admission diagnosis algorithm. RESULTS: Hospital prevalence over the four-year period varied from 6073 breast cancer cases for the simplest model (first selection step without the admission diagnosis algorithm) to 4951 when the first selection step was associated with the breast cancer code as admission diagnosis. The model combining the third selection step with a breast cancer-specific admission reason provided 5275 prevalent cases. CONCLUSION: The last model seems more appropriate for case-mix-data coding. Selecting admission diagnosis improved specificity. Combining all hospital stays for each patient has improved diagnostic sensitivity.

Is it possible to estimate the incidence of breast cancer from medico-administrative databases?

One approach to estimate cancer incidence in the French Départements is to quantify the relationship between data in cancer registries and data obtained from the PMSI (Programme de Médicalisation des Systèmes d'Information Médicale). This relationship may then be used in Départements without registries to infer the incidence from local PMSI data. We present here some methodological solutions to apply this approach. Data on invasive breast cancer for 2002 were obtained from 12 Départemental registries. The number of hospital stays was obtained from the National PMSI using two different algorithms based on the main diagnosis only (Algorithm 1) or on that diagnosis associated to a mention of "resection" (Algorithm 2). Considering registry data as gold standard, a calibration approach was used to model the ratio of the number of hospital stays to the number of incident cases. In Départements with registries, validation of the predictions was done through cross-validation. In Départements without registries, validation was done through a study of homogeneity of the mean number of hospital stays per patient. Cross-validation showed that the estimates predicted by the model were true with data extracted by Algorithm 1 but not by Algorithm 2. However, with Algorithm 1, there was an important heterogeneity between French Départements as to the mean number of hospital stays per patient, which had an important impact on the estimations. In the near future, the method will allow using medico-administrative data (after calibration with registry data) to estimate Départemental incidence of selected cancers.