Johan Bring

How to Standardize Regression Coefficients

Abstract In many regression studies, there is an ambition to compare the relative importance of different variables. One measure frequently used is standardized regression coefficients. The present article reveals an inconsistency in the definition of the standardized coefficients and demonstrates that if they were correctly defined, a comparison between standardized coefficients would be similar to comparing t values.

A Geometric Approach to Compare Variables in a Regression Model

Abstract Geometry is a very useful tool for illustrating regression analysis. Despite its merits the geometric approach is seldom used. One reason for this might be that there are very few applications at an elementary level. This article gives a brief introduction to the geometric approach in regression analysis, and then geometry is used to shed some light on the problem of comparing the “importance” of the independent variables in a multiple regression model. Even though no final answer of how to assess variable importance is given, it is still useful to illustrate the different measures geometrically to gain a better understanding of their properties.

Cervical spine fractures in the elderly: Factors influencing survival in 65 cases

To evaluate whether a cervical spine fracture increases the death risk in elderly patients, and to define risk factors, we studied the survival of 65 patients (26 women) with a mean age of 77 (66-99) years. 8 of the patients were tetraparetic. In 35 patients, the upper cervical spine was fractured. 7 patients suffered from ankylosing spondylitis. Severe co-morbidity was present in 16. Survival status and the date of death were retrieved from the government official personal registry. The expected survival was calculated from data retrieved from the Swedish National Board of Health and Welfare. Variables having a possible relation with survival (i.e., a p-value < 0.10 when entered into a Kaplan-Meier survival analysis) were used in a Cox multiple regression survival analysis. 53 (24-105) months after injury, 25 of the 65 patients had died. The survival was significantly lower than the expected values. Severe co-morbidity (risk ratio: 5,6), neurological injury (6,4), high age (1,1), and ankylosing spondylitis (5,5) proved to be significant risk factors for death. Thus, a cervical spine fracture may lead to earlier death in a patient with a severe co-morbidity. A neurological complication constitutes a risk also for a previously healthy individual. Patients having ankylosing spondylitis (with increased death risk) run a higher than normal risk of sustaining a cervical spine fracture.

Inter-examiner reliability in the assessment of low back pain (LBP) using the Kirkaldy-Willis classification (KWC)

Reliable classification systems and clinical tests are sought for the care of patients with low back pain (LBP). The objectives of this clinical study were to evaluate inter-examiner reliability in the classification of patients with LBP, the influence of radiological findings on the classification and the reliability of some clinical tests. Two examiners independently assessed 50 outpatients with LBP. Inter-examiner reliability in classification of patients with LBP using Kirkkaldy-Willis classification (KWC) system and in 30 clinical tests was calculated as percentage agreement and kappa coefficients (κ). Inter-examiner reliability was excellent (κ>0.8) for classification according to KWC. Radiological findings did not influence the reliability. Age of the patient, movement range, and pain and neurological signs seemed to guide the decision on classification. The reliability of clinical tests was good (κ>0.6) in 6 tests and moderate (κ>0.4) in 12 tests. Good inter-examiner reliability was found for the SLR test, movement range and sensibility testing with spurs in dermatome areas. We conclude that the KWC for classifying patients with LBP seems to be a reliable classification system depending on a few key observations and that moderate and good inter-examiner reliability can be achieved in several clinical tests in the assessment of LBP.

Heart failure diagnosis in primary health care: clinical characteristics of problematic patients. A clinical judgement analysis study

BackgroundEarly detection of chronic heart failure has become increasingly important since the introduction of effective treatment. However, clinical diagnosis of heart failure is known to be difficult, especially in mild cases or early in the course of the disease. The purpose of this study is to analyse how patient characteristics contribute to difficulties in diagnosing systolic heart failure.MethodsDesign: A Clinical Judgement Analysis study of 40 case vignettes based on authentic patients, including relevant clinical data except echocardiography. Setting: Primary health care and two cardiology outpatient clinics in Stockholm. Subjects: 70 participants with different types of clinical experience; 27 specialists in general practice, 22 cardiologists, and 21 medical students. Main outcome measures: The assessed probability of heart failure for each case vignette, and the disagreement between the participants. The number of clinical variables (cues) indicative of heart failure in the case vignettes.ResultsThe ten case vignettes with the least diverging assessments more often had increased relative cardiac volume and atrial fibrillation. No further specific clinical patterns could be found in subgroups of the case vignettes. The ten case vignettes with the most diverging assessments were those with an intermediate number of clinical variables. The case vignettes with the least diverging assessments more often represented patients with cardiac enlargement and atrial fibrillation.ConclusionDiagnosing mild heart failure is difficult, as these patients are not easy to characterise. In our study, a larger number of positive cues resulted in more diagnostic conformity among the participants, and the most important information was cardiac enlargement. The importance of more objective diagnostic methods in diagnosing suspected cases of heart failure should be emphasised.

GPs' decisions on drug treatment for patients with high cholesterol values: a think-aloud study.

BackgroundThe purpose was to examine how General Practitioners (GPs) use clinical information and rules from guidelines in their decisions on drug treatment for high cholesterol values.MethodsTwenty GPs were presented with six case vignettes and were instructed to think aloud while successively more information about a case was presented, and finally to decide if a drug should be prescribed or not. The statements were coded for the clinical information to which they referred and for favouring or not favouring prescription.ResultsThe evaluation of clinical information was compatible with decision-making as a search for reasons or arguments. Lifestyle-related information like smoking and overweight seemed to be evaluated from different perspectives. A patients smoking favoured treatment for some GPs and disfavoured treatment for others.ConclusionsThe method promised to be useful for understanding why doctors differ in their decisions on the same patient descriptions and why rules from the guidelines are not followed strictly.

Variable importance by partitioningR 2

One purpose of many regression studies is to compare the relative importance of the independent variables. Several different measures have been used to measure importance:t-values, standardized regression coefficients, elasticity, commonality analysis, increment inR2, correlation coefficients, hierarchical partitioning etc. Some of these measures have the common feature of partitioningR2 between the independent variables and assess their importance according to their contribution toR2. This paper is an attempt to clarify the advantages and disadvantages with these different methods and find out if any useful information can be gained by a partitioning ofR2.

Serum levels of soluble vascular cell adhesion molecule‐1 (sVCAM‐1) are elevated in advanced stages of non‐Hodgkin's lymphomas

Abstract: The serum levels of soluble vascular cell adhesion molecule‐1 (sVCAM‐1) were measured in 116 patients with non‐Hodgkins lymphomas (NHL) tested previously for soluble intercellular adhesion molecule‐1 (sICAM‐1). In contrast to Hodgkins disease and chronic lymphocytic leukaemia, the sVCAM‐1 levels in NHL patients were not significantly different from the levels of healthy controls (n = 31). However, sVCAM‐1 was elevated in advanced stage disease, i.e. stages III + IV. Elevated serum levels of sVCAM‐1 were associated with significantly poorer disease‐free (p = 0.024) and overall (p = 0.02) survival. sVCAM‐1 correlated poorly with other known prognostic variables (LDH, sTK and β2m) and with sICAM‐1. None of the tested markers added prognostic information for disease‐free survival independently of Ann Arbor stage and B‐symptoms. The expression of VCAM‐1 and ICAM‐1 in tumour biopsies from 15 patients representing 7 different histologies were examined and compared with the serum levels of the soluble adhesion molecules. No correlation was found between the adhesion molecule expression by vascular endothelium and the corresponding serum levels.

General practitioners’ coronary risk assessments and lipid-lowering treatment decisions in primary prevention : Comparison between two European areas with different cardiovascular risk levels

Aim: To investigate whether general practitioners (GPs) in countries with different levels of cardiovascular risk would make different risk estimates and choices about lipid-lowering treatment when ...