The ICD-coding quality for outpatients' diagnoses by German physicians was analysed in a sample of five million members of the German Statutary Health Insurance System. New federal legislation coming into effect in 2009 for the reimbursement of physicians is based on patients' morbidity risks and thus on the quality of a provider's ICD coding. A sample of physicians' billing data for 2001-2003 containing ICD codes for patients' morbidity and the billed services was linked with outpatients' prescription data for the time period 2002- 2003. As in 2001-2003 information on the certainty of diagnosis was not yet mandatory, only 7.4% of all diagnoses were labelled as either "suspected diagnosis", "excluded diagnosis" or "history of diagnosis", hampering coding validity measurements. Chronic disease persisted in the time window analysed showing only minor successive prevalence decreases after an initial dip of at least 6% in the calendar term following the index term. The immediate decrease following the initial term may be due to initially suspected disease not confirmed until the work up at subsequent visits is completed. The slight downward slope after six months may indicate minor undercoding of chronic diagnoses. Few acute diagnoses persisted for longer than two calendar terms making it unlikely that acute diagnoses were erroneously maintained for repetitive reimbursement. Undercoding of diagnosis was abundant in patients receiving insulin prescriptions, as a diagnosis of diabetes was often missing. Numerous drugs prescribed could not be associated with a corresponding diagnosis coded by physicians. We suggest that before reimbursements to physicians are based on ICD-coded morbidity, a re-analysis of the data should be performed containing information on diagnosis certainty (mandatory since 2004) and the recently updated catalogue on fees for medical procedures provided "Einheitlicher Bewertungsmassstab" (EBM).