Search

Prediction models fitted with logistic regression often show poor performance when applied in populations other than the development population. Model updating may improve predictions. Previously suggested methods vary in their extensiveness of updating the model. We aim to define a strategy in selecting an appropriate update method that considers the balance between the amount of evidence for updating in the new patient sample and the danger of overfitting. We consider recalibration in the large (re-estimation of model intercept); recalibration (re-estimation of intercept and slope) and model revision (re-estimation of all coefficients) as update methods. We propose a closed testing procedure that allows the extensiveness of the updating to increase progressively from a minimum (the original model) to a maximum (a completely revised model). The procedure involves multiple testing with maintaining approximately the chosen type I error rate. We illustrate this approach with three clinical examples: patients with prostate cancer, traumatic brain injury and children presenting with fever. The need for updating the prostate cancer model was completely driven by a different model intercept in the update sample (adjustment: 2.58). Separate testing of model revision against the original model showed statistically significant results, but led to overfitting (calibration slope at internal validation = 0.86). The closed testing procedure selected recalibration in the large as update method, without overfitting. The advantage of the closed testing procedure was confirmed by the other two examples. We conclude that the proposed closed testing procedure may be useful in selecting appropriate update methods for previously developed prediction models

Background Transmural support by a haemophilia nurse may improve treatment and may empower parents and patients. Aim To measure the effect of structured home visits by a haemophilia nurse in (parents of) patient on aspects of prophylactic home treatment. Methods A multicentre intervention study in two paediatric haemophilia treatment centres was performed. Primary outcome measures were: adherence to prescribed treatment, health-related quality of life and behavioural scores. Secondary outcome measures were: total clotting factor consumption, self-efficacy and number of joint bleeds. Results Over a period of 22 months (median, IQR 21–23), four to seven home visits in 46 patients (mean age 9.4 ± 4.2 years) were made. No difference in adherence to prescribed treatment was seen after the home visits when compared to baseline measurements. Both the Child Health Questionnaire (CHQ) scales on ‘Role functioning – Emotional/Behavioural’ (P = 0.02, d = 0.53) and ‘Parental Time Impact’ (P = 0.04, d = 0.33) were reduced after intervention. The disease-specific Haemo-QoL questionnaire showed improvement in domains: ‘Family’ (P = 0.04, d = −0.14), ‘Friends’ (P = 0.03, d = −0.29) and ‘Perceived support’ (P = 0.03, d = −0.37). Significant improvement was observed with regard to domain ‘Communication’ of the VERITAS-Pro scale (P = 0.03, d = −0.28). Conclusions After a period of transmural care by a haemophilia nurse, significant but small positive effects were demonstrated with regard to communication and increase of perceived support between parents and haemophilia treatment centre. No improvement was observed in other outcome measures.

UNLABELLED Physicians often have to perform a lumbar puncture to ascertain the diagnosis in patients with meningeal signs, because of the serious consequences of missing bacterial meningitis. The aim of this study was to derive and validate a clinical rule to predict bacterial meningitis in children with meningeal signs, to guide decisions on the performance of lumbar punctures. Information was collected from records of patients (aged 1 mo to 15 y) consulting the emergency department of the Sophia Children's Hospital between 1988 and 1998 with meningeal signs. Bacterial meningitis was defined as cerebrospinal fluid (CSF) leucocyte count >5 cells microl(-1) with a positive bacterial culture of CSF or blood. The diagnostic value of predictors was judged using multivariate logistic modelling and area under the receiver operating characteristic curves (ROC area). In the derivation set (286 patients, years 1988-1995) the duration of the main complaint, vomiting, meningeal irritation, cyanosis, petechiae and disturbed consciousness were independent clinical predictors of bacterial meningitis. The ROC area of this model was 0.92. The only independent predictor from subsequent laboratory tests was the serum C-reactive protein concentration, increasing the ROC area to 0.95. Without missing a single case, this final model identified 99 patients (35%) without bacterial meningitis. Validation on 74 consecutive patients in 3 subsequent years (1996-1998) yielded similar results. CONCLUSION This prediction rule identifies about 35% of the patients with meningeal signs in whom a lumbar puncture can be withheld without missing a single case of bacterial meningitis. For the individual patient this prediction rule is valuable in deciding whether or not to perform a lumbar puncture.

Pre-school children are frequently affected by wheezing illness, with substantial influences on their health-related quality of life (HRQL). The Infant/Toddler Quality of Life Questionnaire (ITQOL) is the only generic health status measure for children aged 2 months up to 5 years. In this present study we evaluated the impact of wheezing illness in pre-school children on the HRQL, using the ITQOL. A questionnaire including the ITQOL and ISAAC questions on frequency and severity of respiratory complaints were sent to parents of patients aged 6 months-5 years visiting the outpatient department with wheezing illness. Scale scores of ITQOL of the included children were compared with general population scores. Using multivariate analysis, the influence of general and clinical characteristics on ITQOL scale scores was evaluated. Results are based on 138 children, 59% male, mean age 34 months. Children with wheezing illness scored differently to the general population sample on 8 of the 11 ITQOL scales. Age, comorbidity and employment of the respondent affected scales on child's physical and emotional functioning and parental functioning. Severity of dyspnoea and wheezing, presence of cough, corticosteroid use, and number of GP visits negatively affected scales on the child's physical and emotional functioning. In conclusion, the HRQL as measured by the ITQOL was lower in a group of Dutch pre-school children with wheezing illness compared to a general population sample. The scale scores were sensitive to age, co morbidity, socio-economic situation, and disease severity.

UNLABELLED In this study, independent predictors obtained from patient history, physical examination and laboratory results for vesico-ureteric reflux (VUR) in children of 0-5 y with a first urinary tract infection (UTI) were assessed and the added value of renal ultrasound (US) investigated. Information was collected from children visiting the paediatric outpatient department with a first proven UTI, defined as a urine monoculture with > or = 10(5) organism/ml, with clinical symptoms and possible white cell count > or = 20 per high-power field of spun fresh urine. Children with neurologic bladder dysfunction were excluded. VUR was determined by voiding cystourethrography (VCUG) and graded from I to V. The diagnostic value of predictors was judged using multivariate logistic modelling with the area under the receiver operating characteristic (ROC area). A risk score was derived based on the regression coefficients of the independent predictors in the logistic model. In 140 children (51 boys and 89 girls) VUR was diagnosed in 37. Independent predictors for VUR were male gender, age, family history for uropathology, serum C-reactive protein level (CRP) and dilatation of the urinary tract on US. The ROC area of this model was 0.78 (95% CI: 0.69-0.87). This prediction model identified 12% (95% CI: 7-18) of the patients without VUR without missing one case of VUR. If we used VUR > or = grade 3 as a threshold, the model assessed VUR to be absent in 34% (95% CI: 26-42). CONCLUSION A prediction rule based on age, gender, family history, CRP and US results is useful in assessing the probability of VUR in the individual child with a first UTI and may help the physician to make decisions about performing additional imaging techniques. Prospective validation of the model in future patients, however, will be necessary before applying the rule in practice.