The National Hospital Morbidity Database (NHMD) is compiled from data supplied by state and territory health authorities. It is a collection of electronic “confidentialised” summary records for separations (i.e. episodes of care) in almost all public and private hospitals in Australia. Data are currently held for the years 1993-94 to 2012–13. In 2012–13, 9.4 million separations were reported. The NHMD was not developed for research purposes and has limitations in that it is not possible to count patients individually, transfers within the hospital system may appear as separate episodes, coders extract data from the specific hospitalisation only (no historical data), and coding practices change over time. In 2004-05, the AIHW reported that 8% of hospitalisations in Australia had diabetes coded as the principal or an additional diagnosis. By 2010-11 this had dropped to 2.5%, but this was not due to improved outcomes for people with diabetes but a major change to coding rules for diabetes between 2008 and 2013. Even the 2004-05 AIHW figures are likely to be an underestimate as clinical audits from Australia and overseas have found hospital prevalence rates of diabetes of 11‐25%. The Fremantle Diabetes Study Phase I (FDS1) recruited residents with known diabetes from a postcode-defined area between 1993 and 1996. The cohort has been linked to the Western Australian Hospital Morbidity Data System (HMDS), which feeds into the NHMD, from 1970 to 2012. This presentation will explore the temporal consequences of changes to coding procedures between 1993 and 2012, the veracity of the identification of diabetes type in the HMDS, and length of stay for hospitalisations coded and not coded for diabetes. The implications for research, clinical care and reimbursement will be discussed.