STUDY OBJECTIVE: The aim was to report on the extent to which death certificates which specify that death occurred in hospital can be matched and linked with routine hospital inpatient information systems. DESIGN: The study involved linkage of hospital records which specified that death occurred in hospital to corresponding death certificates; and linkage of death certificates which specified that death occurred in hospital to corresponding hospital records. SETTING: Six health districts in southern England covered by medical record linkage. SUBJECTS: Records were examined of patients aged 65 years and over, which specified that death occurred in hospital between 1979 and 1985. MAIN RESULTS: 98.2% of hospital record abstracts which specified that death occurred in hospital were linked by our standard computer-based techniques to death certificates. Conversely, however, only 94.4% of death certificates which specified that death occurred in hospital could be linked to the abstracts of corresponding hospital inpatient records. A major factor contributing to the latter failures may be a difference of definition of what constitutes a death "following hospital admission" in patients who die shortly after arrival at hospital. CONCLUSIONS: Linkage of hospital records to death certificates is both feasible and desirable. Error rates are generally small; but hospital inpatient record abstracts corresponding to death certificates for deaths in hospital may not invariably exist when death occurs shortly after the arrival of the patient at hospital.
STUDY OBJECTIVE: The aim was to report on the extent to which death certificates which specify that death occurred in hospital can be matched and linked with routine hospital inpatient information systems. DESIGN: The study involved linkage of hospital records which specified that death occurred in hospital to corresponding death certificates; and linkage of death certificates which specified that death occurred in hospital to corresponding hospital records. SETTING: Six health districts in southern England covered by medical record linkage. SUBJECTS: Records were examined of patients aged 65 years and over, which specified that death occurred in hospital between 1979 and 1985. MAIN RESULTS: 98.2% of hospital record abstracts which specified that death occurred in hospital were linked by our standard computer-based techniques to death certificates. Conversely, however, only 94.4% of death certificates which specified that death occurred in hospital could be linked to the abstracts of corresponding hospital inpatient records. A major factor contributing to the latter failures may be a difference of definition of what constitutes a death "following hospital admission" in patients who die shortly after arrival at hospital. CONCLUSIONS: Linkage of hospital records to death certificates is both feasible and desirable. Error rates are generally small; but hospital inpatient record abstracts corresponding to death certificates for deaths in hospital may not invariably exist when death occurs shortly after the arrival of the patient at hospital.