To our knowledge this is the first study that investigated predictors of nursing home admission among a representative sample of the Belgian population aged ≥65 years.
Over the study period, 13.3% of the study population was admitted into NH and 13.2% had died without having been admitted to a NH. The overall unadjusted cumulative incidence (risk) of NHA, accounting for death as competing risk was of 5.7% at 3 years of follow-up and of 13.1% at the end of the study. This crude risk was significantly higher among individuals with severe limitations as opposed to those without limitations. After adjusting for baseline characteristics, higher age, low educational attainment, belonging to low income household, living alone, use of home care services, history of falls, suffering from urinary incontinence, depression or Alzheimer disease, having had hospitalized in the past 12 months were significantly associated with a higher risk of NHA, while female, individuals with multimorbidity and increased contacts with health care providers were significantly associated with a decreased risk of NHA. Subjective health and limitations are both significant determinants of NHA, but subjective health is an effect modifier on the effect of limitations and vice versa.
The incidence of NHA may be influenced by organizational aspects and the cultural aspects [47], but also by the availability, accessibility and affordability of home care facilities. The characteristics of the study population also play a role since the cumulative incidence of NHA could be affected by the higher risk of death in the population under study. Previous studies have investigated predictors of NHA among sub-groups of the population. For instance, Bergkamp et al. [29] investigated predictors of NHA in Cerebral Small Vessel Disease (CSVD) patients and have found that after 5-years follow-up, the cumulative incidence was 3.6% (95% CI, 2.2-5.5) and 6% (95% CI, 4.2–8.3) after 8 years of follow-up. This cumulative incidence is lower than those found in our study. A possible explanation could be that the risk of the competing event, i.e., the death is likely higher in patients suffering from CSVD than in the general population, which may affect the risk of NHA. In contrast, our cumulative incidence is lower than the cumulative incidence found in a study in USA, 16.1% in a 2-years follow-up [48].
In earlier studies, advanced age emerged as a strong predictor of NHA among the predisposing factors [9, 22, 23]. In accordance with these studies, higher age was found to be a significant predictor of NHA, event after taking into account the competing risk of death. Furthermore, Gaugler et al. [23] showed (even if slightly high, probably due to the overestimation of Cox model) comparable result as found in this study regarding gender (female: HR of risk: 0.87 (95% CI, 0.81-0.93) vs. sHR of risk 0.75 (95% CI, 0.72-0.78). Unlike the results shown in Table 2 and also in the univariate analysis, even after taking into account the competing risk of death (sHR=1.54, 95% CI, 1.53-1.55) (results not shown), women appeared less likely to enter NH than men. One possible explanation could be related to their health status. Indeed, further analysis showed that women are more likely to suffer from multimorbidity (38.92%) than men (31.67%), p = 0.0230, and as our results show, multimorbidity was associated with a low risk of NHA.
It is striking that belonging to a low income household increases the risk of NHA of 29%. Previous studies also found a higher risk of NHA for individuals belonging to a low income household [43, 49]. An in-depth analysis shows that the majority of these people benefit from specific protection measures for financially vulnerable persons introduced in the health insurance system to reduce financial barriers in using care. Indeed, 52.5% of individuals admitted into NH benefited from an increased reimbursement of healthcare expenses and 77% of them have a special status for lump sum for the chronically ill. Thus, the effect of these financial protection measures could partly explain the high risk of nursing home entry for people from low income households.
Living arrangements appeared as the strongest predictor of NHA among the enabling factors. Individuals who lived alone had nearly twice the sHR to enter a NH. Our findings are in line with those in previous studies [17, 50]. Currently, people prefer living in their own home as long as possible. However, the lack of social support, but also financial problems in meeting the costs of home-based services constitute some limitations to this system. In a context where the social welfare system could cover a large part of the cost of institutionalization for individuals, an older person who lives alone might be more prone to be admitted in an institution when functional disability or health complications occur [17].
The use of home care services in the previous year was associated with the greater risk of NHA. This result is not surprising since use of home care services is generally an expression of a need for support and therefore a first step towards possible NHA. Our result is similar to those in a study on predictors of NHA after hip fracture. The authors found that receiving home care before injury was associated with an increase in HR of 2.00 (95% CI 1.54-2.61), HR 1.64 (95% CI, 1.43-1.87), and HR 1.22 [95% CI, 1.13-1.32) for patients 60 to 69 years, 70 to 79 years, and 80 to 89 years respectively [22].
Within the need factors, if either poor subjective health or severe limitations are present there is an increased risk of NHA, but when they occur together the risk of NHA decreases, most likely because for those people the risk of dying is larger than the risk of being admitted to a NH (competing risk).
Multimorbidity was associated in reducing the risk of NHA. This result could partially be explained by the effect of competing risk of death among this group. Individuals suffering from multimorbidity have also higher mortality risk and this event may arrive before NHA. Indeed, an in-depth analysis shows that among individuals suffering from multimorbidity, 93 of them (14.3%) were admitted to a NH while 118 (18.2%) died before being admitted to a NH.
In line with earlier studies [23, 51], we found that history of falls in the past 12 months was associated with an increased risk of NHA. In fact, in some cases, falls among the older people can lead to more serious events (fractures, injuries, loss of autonomy) with adverse consequences on their health status and therefore precipitate their admission to a NH.
The presence of Alzheimer’s disease is by far the strongest predictor of NHA (individuals with Alzheimer’s disease have 5 times more risk than those without Alzheimer’s disease). In the literature, beside age, cognitive comorbidities (depression, Parkinson, dementia or Alzheimer’s disease) and functional impairment were among the strongest predictors and are associated with an increased risk of NHA [9, 12, 22, 23, 52]. For example, in a study among a general older population, the authors found that Alzheimer’s or dementia increases the hazard of nursing-home entry by 20.2 times for men and 10.0 times for women [9]. In another study of 137,000 community dwelling patients aged +65 years, Harris et al. found that depression was associated with a higher risk of nursing home admission in the general population [52]. Other need factors are of lesser importance.
Strengths and limitations
From a public health perspective, major strengths of the study include the use of a representative large sample and the use of a large number of individual-level predictors, a relatively long follow-up, and the linkage to administrative data to identify main risk factors of NHA. The use of the competing risk analysis is another strength of this study. Indeed, we performed competing risk regression to study the association between several covariates and the risk of NHA. This approach is preferred over a standard survival model because in older population, death may be compete with nursing home admission, and ignoring such competing risk may lead to biased results [30]. In competing risk situations, the cumulative incidence function is more appropriate as it takes competing events into account when estimating the incidence. We further chose the Fine and Gray model over the cause-specific hazard model as our primary interest was in predictive modelling.
The current study has some limitations that deserves to be pointed out. First, the exact dates of NHA were not available. We used the dates of the first registered care in NH based on specific nomenclature code as a proxy of dates of NHA. However, these dates are pretty accurate and deviations from the exact dates are small. Second, almost all covariates included in the analysis were measured at baseline and most of them are self-reported. Therefore, possible changes (e.g., for living arrangements or other social supports) over the course of the study are not talking into account. Third, data on local variations in supply of care and/or home care services (supply of NH beds, hospital beds, and physicians in the region of residence, waiting lists, etc.) as potential important confounders at the enabling level were unavailable and therefore not included in our analyses. A further limitation is that our indicator of Alzheimer’s disease is less sensitive due to the case definition based on prescribed specific medications. Indeed, many people with Alzheimer do not take medications. So individuals suffering from this disease might be underestimated and probably especially Alzheimer patients on medication are admitted to NH. Finally, the finding of this study may not be generalized to other areas or settings because the organization of the Belgian health care system can be very different from other countries.