Objective: To describe the Lundby Study and the difficulties in doing repeated surveys.
Method: Best-estimate consensus diagnoses have been used since 1957 together with DSM-IV and ICD-10 in 1997.
Results: The Lundby population consisting of 3563 probands was investigated in 1947, 1957 and 1972. Sufficient information was available for 98-99%. In 1997-2000 a fourth field investigation was carried out. Attrition rate for the interviews was 13% (238/1797). About 36% (1030/2827) had died between 1972 and 1997, but data from registers, case notes and key-informants for the period 1972 and 1997 completed the information for 94% (2659/2827). The population has followed the same pattern of development as many rural populations in Sweden since the 1940s. Multiple sources of information are preferable in longitudinal studies in order to tackle the problem of changing diagnostic systems.
Conclusion: Low attrition rates over 50 years and reasonable diagnostic uniformity make comparisons over time justifiable.