Background: The aim of this study was to validate a new method of quality control of medical records of the "Medicalisation of Information System Program" (P.M.S.I.) in Limoges university hospital.
Methods: Lot quality assurance sampling (L.Q.A.S.), developed to meet industrial quality control needs, has been applied. This method allows to give an opinion on the quality of a lot from a sample randomised in this lot. Ten hospital units have been randomised. During three months, these medical units have recoded ten medical units abstracts (R.U.M.) which were randomised each month within their stays. If five or more R.U.M. had been modified on Principal Diagnosis (D.P.), Associated Significant Diagnosis (D.A.S.) or Diagnosis Related Group (D.R.G.), the lot was rejected. The hospital unit had then less 70% R.U.M. of good quality (alpha=5%). The global errors rates were compared to the results of usually performed quality control (1998).
Results: Nine hospital units were included. Hospital units which already had a low quality rate during the 1998 quality control had also an insufficient quality by L.Q.A.S. method. Among these, 33.7 [CI(95%): 21.6-45.8] R.U.M. have been modified. Medical units had a poorer data quality than surgical ones. No statistically significant difference was shown between 1998 errors rates and L.Q.A.S. rates, except for D.P. (p=0.002).
Conclusion: L.Q.A.S. is easy to apply, could be repeated more frequently than usual controls, and allows to detect hospital units which require a faster action. This process could be applied to P.M.S.I., in which the search of quality is necessary because of the new cost model which includes secondary D.R.G.