Abstract
Longitudinal studies of people with mental disorder are important in understanding outcome and intervention effects but attrition rates can be high. This study aimed to evaluate use of multiple record sources to trace, over 12 years, a one-year discharge cohort of high-security hospital patients. Everyone leaving such a hospital in 1984 was traced until a census date of 31 December 1995. Data were collected from several national databases (Office for National Statistics (ONS), Home Office (HO) Offenders' Index, Police National Computer Records, the Electoral Roll) and by hand-searching responsible agency records (HO, National Health Service). Using all methods, only three of the 204 patients had no follow-up information. Home Office Mental Health Unit data were an excellent source, but only for people still under discharge restrictions (<50% after eight years). Sequential tracing of hospital placements for people never or no longer under such restrictions was laborious and also produced only group-specific yield. The best indicator of community residence was ONS information on general practitioner (GP/primary care) registration. The electoral roll was useful when other sources were exhausted. Follow-up of offenders/offender-patients has generally focused on event data, such as re-offending. People untraced by that method alone, however, are unlikely to be lost to follow-up on casting a wider records net. Using multiple records, attrition at the census was 38%, but, after certain assumptions, reduced further to 5%.