This study showed the median survival time of nursing home residents in Iceland to be 31 months (2.6 years) with a stable death rate over the period of the study. Almost a third of the residents had died within a year from admission; a majority had died within 3 years, and less than half of the residents lived longer than 3 years. Those dying within the first year had less stable health, worse ADL performance, more pain, more depression and were less involved in social engagement. Significant predictors of mortality were age, gender, where admitted from, ADL functioning, health stability and social engagement.
The reported survival time in this study is similar to two recent studies with a 5 year follow-up time where the median survival of nursing homes was 2.3 years (N. Irel.; n = 2.112)  (US; n = 468) . Other studies have reported higher  or lower  mean survival times. However, any cross-country comparison of survival times must take into account the availability of home care services and the criteria for nursing home placement in the respective countries; admission criteria for nursing home placement especially may complicate comparison.
Combined health stability and ADL performance seem to be valid predictors of mortality and should thus be considered when selecting a preferable type of service for older persons. These findings resemble those in other studies with regard to health stability [2, 34] as well as ADL capacity [1, 35]. In this study health stability, in particular a score of 2 or higher, was a strong predictor of mortality. Residents with this score were 1.61 times likelier to die during the investigation period, and those with a score of 5 were 16.12 times likelier to die. Also, a score of 10 or higher in ADL performance significantly predicted a higher risk of mortality than for those with a lower score. For instance those with a score 10-17 were 1.33 times likelier to die during the investigation period, and those with a score of 18-28 were 1.80 times likelier to die. Thus assessment of ADL and health stability seems to be helpful in selecting the most appropriate type of service. It may well be that older persons having a health stability score lower than 2 and an ADL score below 10 are better off in home care than nursing home placement. However, using only ADL capacity and health stability as a reference may be too narrow an approach. There may be other reasons for deciding on nursing home placement apart from those with increasing risk of mortality, such as difficult social circumstances or the person's mental health. Still it turned out that unstable health and low ADL capacity should be considered as important indicators of death and, in turn, more nursing care needs, such as services available at a nursing home.
It was noteworthy that low social engagement seems to be an important variable to take into account when predicting mortality. As a concept it may be viewed as the opposite of unstable health and low ADL capacity, as such debilitation would hinder a person from seeking or developing effective social engagement. The level of ADL capacity, however, does have some effect on the relationship of social engagement and survival time. In this study those with the least social engagement had an increased risk of death compared with the reference group who were deemed to have high initiative and participated in social activities. Those with a score of 2 were 1.51 times likelier to die than the reference group, and those with a score of 0, i.e. demonstrating severe withdrawal from social engagement, were 1.65 times likelier to die. Other studies have reported decreased social engagement to be a predictor of mortality for residents already living in nursing homes [3, 36], rather than at admission. A study of one-year mortality of residents (US; n = 30.070) showed that greater levels of social engagement (scores 0-6 on the same scale as the present study) were associated with longer survival (p = 0.0001), and a one-point decrease in the index of social engagement meant that residents were 1.16 times as likely to die during the follow-up period . The present study, however, revealed that only a score of 2 and lower in social engagement significantly predicted mortality, and the risk decreased with higher levels of engagement (Table 4). Causality cannot be established in the present study although it has been stated that social engagement influences residents well-being, and that social isolation may increase mortality and morbidity . The nature of the relationship between social engagement and survival is complex. Social engagement may be hindered by disease and disabilities or other factors. Furthermore, environmental factors, activity and action by the individual may influence a person's health status . Thus, it may well be that stimulating social engagement and individual activity may increase survival time.
The high percentage of residents dying in the first to third years of living in a nursing home suggests that the concept of palliative care may be a useful model for care in a nursing home. Research has furthermore indicated that increasing numbers of residents are dying in nursing homes instead of hospitals . The findings of the present study suggest that one third of those admitted were already in a palliative stage at admission. Thus, the focus of nursing care in nursing homes needs to be on palliative care as much as restorative care. However, knowledge of palliative care and symptom management adapted to older people  as well as to those suffering from dementia is lacking in nursing homes .
The death rate was stable between cohorts, and in the first year after admission, it was 28.8% despite the fact that resources for nursing home care have decreased over the years. Findings from other studies differ and have reported both lower (17.5%)  and higher rates(34%) . In a Swedish study on two cohorts (2001 and 2002) of old people (N = 626; 65-98 years) receiving public long-term care, the two-year mortality rate was 30% and 31%, respectively . This was considerably lower than in the present study (43.4%). It should be noted that the Swedish subjects were receiving care at home as well as in nursing homes. However, where people were living was not an independent predictor of mortality . Almost a third of the residents in the present study may have needed palliative care within a year of admission. These residents had less stable health, more ADL dependency, pain and depression and were less engaged socially - needs well within the concept of palliative care. Thus dying is a central issue in nursing care in nursing homes.
Although a majority died within a year in this study, 46.9% of the residents (n = 1035) lived longer than 3 years. They may have been detected prior to nursing home placement by systematic assessment of ADL capacity and health stability. Some of them may have benefitted from receiving a type of service other than nursing home placement. For instance, home care and rehabilitation might have delayed entry into nursing homes. Such an approach would have been more in line with the official policy of enabling older people to stay at home as long as possible. Enabling old people in relatively stable health and needing low levels of ADL assistance to stay at home longer would probably decrease the demand for nursing home placement.
The strength of this study is the inclusion of 11 cohorts and data based on residents' admission status. Registered nurses trained for the purpose performed the assessments, and only a valid instrument was used [22, 40]. Nevertheless, this study has some limitations, such as variation in the sample of 13% to 84% of the total residents admitted each year to nursing homes . The low percentage of the sample in the early years stems from the fact that these were the first years for mandatory assessment in all nursing homes in Iceland. It took some years for the assessment to be fully implemented, and in the early years residents were often not assessed until they had spent considerable time in the nursing home. Another limitation of concern is that the residents in the sample may have suffered some changes to their health after admittance and before being assessed. The delay in assessment is probably mostly related to workload and the absence of staff due to sickness or leaves rather than characteristics of the residents. The error should therefore be random rather than systematic. However, this can not be substantiated. Researchers have reported, however, a decline among nursing home residents over a six-month period  and a lower mortality risk of recently admitted residents compared to others . The researchers' position, however, was that data from assessments within 90 days would sufficiently reflect the admission status of the residents.
Difference in mortality rates between nursing homes cannot be ruled out. It would have been preferable to investigate this, of course, but information on placement within individual nursing homes was not available. The reported significance of predictors of mortality may therefore vary in relation to nursing homes and this needs to be considered a limitation. Nursing homes in Iceland have however the same admission criteria and any difference in mortality rates are unlikely to have had a powerful effect.