The inter-rater reliability of the ICD-9-CM coding system was assessed in two university psychiatrie clinics, using two methods (discharge diagnoses only versus standardized discharge summary letters), and 3 groups of raters (4 residents, 4 supervising psychiatrists, 3 ICD-9 trained coding personnel). Each rater coded 20 patients in each method.
It was found that the reliability of discharge diagnoses only reached chance level (p-value, residents: .50 ; psychiatrists: .40 ; trained personnel: .30) and that the addition of significantly more information (discharge summary letter) did not improve the reliability of the coding. Given the fact that our data represent the true state of coding affairs in psychiatry in Holland, it was concluded that Dutch WHO statistica on epidemiology should be considered with caution ; a caution perhaps not merely confined to ICD-9-coding in psychiatry.
The question was raised whether ICD-9 coding in psychiatry should not be abandoned altogether and replaced by DSM-III, which supposedly lends itself for better classification and therefore coding. Before this question can be affirmed the DSM-III system needs to be supplemented with symptom definitions and a structured interview. Up to this point in time, however, there seems to be little evidence that better diagnostic criteria leads to better examination, diagnosis, classification and coding.