A cross-sectional study of interobserver variability in primary care patients.
To investigate the consistency of signs and symptoms of nerve root compression in primary care patients with pain irradiating pain into the leg (sciatica).
The literature does not report on all the clinical tests for nerve root compression. In previous studies, most patients had low back pain with no irradiation. Often, little information on examination technique, proportion of positive test results, or clinical patient characteristics was provided.
A random selection of 91 patients was investigated by a neurologist-resident couple. Agreement percentages, proportions of positive test results, and kappas were calculated.
The kappa of the overall conclusion after the history taking was 0.40, increasing to 0.66 after physical examination. Kappas were good for decreased muscle strength and sensory loss (0.57–0.82), intermediate for reflex changes (0.42–0.53), and poor for the examination of the lumbar spine (0.16–0.33). The straight leg raising, crossed straight leg raising, Bragard’s sign, and Naffziger’s sign were the most consistent nerve root tension signs (> 0.66).
Two clinicians disagreed on the presence of nerve root involvement in one of four patients after history taking, and in one of five patients after physical examination. For a more consistent overall diagnosis, the physician probably should put more emphasis on the history of pain on coughing–straining–sneezing, a feeling of coldness in the legs, and urinary incontinence. The investigation of paresis, sensory loss, reflex changes, straight leg raising, and Bragard’s sign provide the most consistent results.
From the Departments of *Neurology and †General Practice, Maastricht University Hospital, Maastricht, The Netherlands.
Acknowledgment date: January 21, 1998.
First revision date: November 12, 1998.
Acceptance date: February 18, 1999.
Address reprint requests to
Patrick C. A. J. Vroomen, MD, PhD
Maastricht University Hospital
Department of Neurology
P. Debyeplein 25
P.O. Box 5800
6202 AZ Maastricht
Suppose two clinicians would agree on the positivity of the SLR in 77 of 91 cases as did the two observers in Table 2, but because of a different prevalence of disease they would agree on 10 positive cases and on 67 negative cases. The chance agreement between the two would rise from 52% to 70%, and the κ would decline from 0.68 to 0.50.
Suppose that all the disagreement occurred in those cases the clinicians labeled as difficult, and that they agreed on all easy cases. In this study, there then would be 30 easy normal cases, 47 easy abnormal cases, and 14 difficult cases. If a general population of 1000 subjects had been investigated, this case mix might have been 939 easy normal cases, 47 easy abnormal cases, and 14 difficult cases. For this general population, the agreement would be 95% with a κ of 0.73 compared with the 85% agreement and κ of 0.68 in this study. This spectrum effect seems to oppose the mathematical effect in the previous footnote.