Background Dementia Care Mapping (DCM) is a widely used observational method for evaluating the service quality provided to people with dementia. However, there is little evidence concerning its reliability and validity when used by routine care staff for whom it was designed.

Method The study evaluated levels of inter-observer agreement; The ability of the five-minute time frame to reflect the,actual passing of time'; And the nature of the relationship between individual Well/III-Being values (WIB) and dependency levels.

Data collected using DCM and continuous time sampling (CTS) were compared. The methods were used in parallel where the CTS coder and the DCM mapper(s) observed the same participants.

Observations were carried out with 64 people with dementia within a day hospital and a continuing care ward. Inter-observer agreement was calculated across 20 participants.

Dependency levels were measured using the Clifton Assessment Procedure for the Elderly (CAPE) (Pattie and Gilleard, 1979).

Results Low levels of inter-observer agreement were found where 11 of the 25 Behaviour Category codes and all six Well/III-being Codes produced unacceptable kappas (<0.6).

The Behaviour coding frame provided a meaningful picture of activities participants engaged in, but significantly underestimated participant levels of inactivity.

A strong relationship was demonstrated between participants' WIB score and levels of dependency, thus DCM was unable to measure well-/ill-being as a separate construct from participants' levels of dependency.

Conclusions Questions were raised regarding the reliability and validity of DCM as used by routine care staff. Possible reasons for this, and suggestions for amendments are made. Copyright (C) 2004 John Wiley Sons, Ltd.