Reliability and feasibility of a near patient test for C-reactive protein in primary care.

BACKGROUND: The applications of new diagnostic technologies such as near patient tests are relevant to the further development and potential of primary care. Through their use, doctors in the community may increase the accuracy of their diagnoses and improve their ability to monitor disease. A reliable indicator of disease activity in various clinical conditions is C-reactive protein (CRP) and a near patient test for this is now available, although there is little information on its use outside hospitals. AIM: A study was set up to evaluate the feasibility of using a novel near patient test for CRP in primary care to validate the results against the laboratory "gold standard' for CRP (Beckman Array) and to compare results with the usual inflammation test used in general practice.
METHOD: Prospective recording of CRP as a near patient test on an "intention to investigate' basis, with validation of results against the Beckman Array system for CRP and hospital laboratory erythrocyte sedimentation rate results, in six general medical practices in Birmingham. Main outcome measures were change in local laboratory usage, characteristics of patients chosen for testing, use of quality control, and comparison of readings with results from the same sample sent to an independent laboratory.
RESULTS: Tests of CRP levels were rarely requested before the study was undertaken. During the 3-month study period, 181 near patient tests were carried out, 146 (81%) to establish a diagnosis and the remainder for disease monitoring. Out of the tests, 67% were performed by general practitioners, mostly during the consultation itself. Using a cut-off level of 10 mg I-1, the near patient test and the Beckman Array gave results which agreed in 84% of cases. The sensitivity and specificity of the near patient test results were 97 and 79%, respectively. The predictive value of a positive result was 59% and that of a negative result was 99%. Cohen's Kappa was 62% and the overall mean bias for results in the range of the test was 6.11 mg I-1 (SE = 3.07 mg I-1). Each test took 6 min on average to perform, including all preparations, blood letting, performing the test and averaging the time for quality control estimations. The cost per test averaged pounds 1.72, rising to pounds 4.17 including labour, capital costs, quality controls and consumables (general practitioner performing the assay at average frequency found in this study).
CONCLUSIONS: Measurement of CRP is rarely used in primary care and awareness of its value could be raised. This near patient test proved feasible for use by general practitioners and practice nurses. Its reliability compared with a laboratory result was satisfactory overall, and excellent with adequate operator technique.