Introduction

Older people’s accidental falls are a priority for WHO/Europe [1] and they constitute a major public health problem. In the European Union, older adults account for half of deaths from unintentional injuries [2], despite representing one fifth of the population [3]. In older adults from developed countries, unintentional injuries are between the fifth and the seventh most common causes of death [4, 5], and the majority of unintentional injuries are caused by falls [2, 4, 6]. Furthermore, mortality due to accidental falls has recently increased in older adults [7–10], and the rapid ageing of the population makes it probable that this trend will increase in the future [7, 11]. Older adults have significantly higher mortality from minor injuries than those under 50s [12] and the interval from injury to death could be longer for minor or moderate injuries (for example rib fracture and chest or muscle contusions) than for more severe ones (hip, vertebral or skull fractures) [13]. Additionally, as the interval between injury and death increases, fall mortality could be underreported due to the death being attributed to a concomitant pathology [13, 14] (pneumonia, cerebral hemorrhage, pulmonary thromboembolism, worsening of comorbidities or other complication of the fall).

Despite the substantial number and significant consequences of aging adults’ accidental falls, it is remarkable that a relatively limited number of recent studies have analyzed the long-term mortality of accidental falls in older adults. Moreover, most of these studies come from hospital settings [15–18] rather than community settings [19] and the scant literature that compares the long-term mortality risk after a fall with an appropriate control group is either not recent [20–23] or limited to women [24]. The study of long-term survival could add further insight into the real burden of this important problem, particularly in nursing homes where falls are frequent; as far as we know, no recent study has properly focused on falls’ long-term consequences in a representative sample of nursing home residents. Hence, the aim of this study is to analyze the association between the baseline presence of a fall, and the severity of the fall, with long-term overall survival of care home residents.

Materials and methods

Results

Of the 754 participants in the baseline survey, we excluded 10 residents (1.3%) with missing information regarding falls at baseline and 55 residents (7.3%) with unknown vital status at the end of follow-up, thus leaving a final cohort of 689 residents. There were 12.6% of residents who reported at least one fall in the 30 days preceding interview, and 29.5% of those who fell suffered adverse outcomes; the severe fall group was comprised of these subjects. Fallers presented more comorbidities and prescribed medications, had higher prevalences of antidepressant use and urinary incontinence, and tended to have a higher degree of functional dependency than residents who had not fallen (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of nursing home residents by severity of fall (1998–1999)^a.

https://doi.org/10.1371/journal.pone.0231618.t001

After a median follow-up of 4.5 years, 53 fallers had died during 278 person-years of follow-up (mortality rate 18.0 per 100 person-years) and 319 non-fallers died during 2,130 person-years (mortality rate 14.7 per 100 person-years). In a model which was fully adjusted for sociodemographic and clinical characteristics, the risk for all-cause mortality was similar among fallers and non-fallers (HR = 1.03; 95% CI = 0.75–1.40) (Table 2). When falls were broken down by their severity, the hazard ratios (95% CIs) for all-cause mortality were 0.92 (0.58–1.45) for residents with non-severe falls and 1.36 (0.73–2.53) for residents with severe falls, both as compared to those without reported falls. This moderate increase in risk among severe fallers was also observed in the adjusted cumulative mortality curves by severity of fall, although the reduced number of residents with severe falls resulted in a somewhat jagged and imprecise curve (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Adjusted cumulative all-cause mortality by severity of fall among nursing home residents in Madrid, Spain.

Non-parametric cumulative mortality curves were obtained as the complement of the baseline survival functions from a Cox model stratified by severity of fall (no, non-severe, or severe), adjusted to the overall weighted percentages of baseline age (65–74, 75–79, 80–84, 85–89, or ≥ 90 years), sex (women or men), type of facility (public, subsidized, or private), dementia (yes or no), number of chronic conditions other than dementia (0–1, 2–3, or ≥ 4), use of antidepressants (yes or no), number of prescribed medications other than antidepressants (0–2, 3–4, or ≥ 5), urinary incontinence (no, mild, or severe) and functional dependency excluding (in)continence (independent, mild/moderate, or severe/total), and accounting for the stratified cluster sampling and the different selection probabilities.

https://doi.org/10.1371/journal.pone.0231618.g001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Hazard ratios for all-cause mortality by occurrence and severity of fall among nursing home residents in Madrid, Spain.

https://doi.org/10.1371/journal.pone.0231618.t002

Time-dependent effects of falls on mortality are shown in Fig 2. The cumulative hazard ratio comparing non-severe falls with no fall remained close to the null effect over the entire follow-up (P for proportional hazards = 0.94). However, residents with severe falls appeared to experience an increased risk of death during the first 2 years of follow-up, reaching a 2-year cumulative hazard ratio of 1.54 (95% CI = 0.86–2.75), with a progressive decrease in risk thereafter, albeit this time-dependent pattern was not significant (P for proportional hazards = 0.70).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Cumulative hazard ratios for all-cause mortality over time by severity of fall among nursing home residents in Madrid, Spain.

Smooth cumulative hazard ratios over time (solid curves) and 95% confidence intervals (dashed curves) were obtained from a spline-based parametric survival model with a single internal knot at the median failure time, stratified by severity of fall (no fall, non-severe, or severe), adjusted for baseline age (65–74, 75–79, 80–84, 85–89, or ≥ 90 years), sex (women or men), type of facility (public, subsidized, or private), dementia (yes or no), number of chronic conditions other than dementia (0–1, 2–3, or ≥ 4), use of antidepressants (yes or no), number of prescribed medications other than antidepressants (0–2, 3–4, or ≥ 5), urinary incontinence (no, mild, or severe) and functional dependency without (in)continence (independent, mild/moderate, or severe/total), and accounting for the cluster sampling and the different selection probabilities.

https://doi.org/10.1371/journal.pone.0231618.g002

In subgroup analyses, non-severe falls were associated with a 69% increase in mortality risk for residents in private facilities and a 37% decrease in risk for those in public institutions (Fig 3). The hazard ratios for severe falls were higher in men (2.00), residents with less comorbidities (1.87), fewer prescribed medications (2.01), and those who were functionally independent (2.10).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Hazard ratios for all-cause mortality associated with severity of fall by subgroup of nursing home residents in Madrid, Spain.

Subgroup-specific hazard ratios (squares with area inversely proportional to the variance) and 95% confidence intervals (CIs, horizontal lines) were obtained from Cox proportional hazards models with interactions between severity of fall (no, non-severe, or severe) and the corresponding subgroup indicator, adjusted for baseline age (65–74, 75–79, 80–84, 85–89, or ≥ 90 years), sex (women or men), type of facility (public, subsidized, or private), dementia (yes or no), number of chronic conditions other than dementia (0–1, 2–3, or ≥ 4), use of antidepressants (yes or no), number of prescribed medications other than antidepressants (0–2, 3–4, or ≥ 5), urinary incontinence (no, mild, or severe) and functional dependency excluding (in)continence (independent, mild/moderate, or severe/total), and accounting for the stratified cluster sampling and the different selection probabilities.

https://doi.org/10.1371/journal.pone.0231618.g003

Discussion

We found no notable association between reporting a fall of any type in the month prior to baseline and long-term overall survival among older adults residing in care homes. However, there was a small non-significant increase in risk among severe fallers. They appeared to experience a moderately increased risk of death during the first 2 years of follow-up, with a progressive decrease in risk thereafter. Furthermore, men, residents with fewer comorbidities, fewer prescribed medications and those who were functionally independent; who had suffered a severe fall, did have a noticeably increased risk of mortality compared to residents with no fall in the month prior to the baseline interview.

A limited number of studies from non-clinical settings have assessed long-term overall survival in older adults who had suffered a fall. Moreover, few of these studies compared survival against an appropriate control group of non-fallers, and those that did were not published recently or were limited to women. Sylliaas et al.’s studied women aged 75 or over living in the community [24]. They found that women who had suffered at least two falls had poorer survival rates (HR = 1.6; 95% CI = 1.1–2.4) compared to non-fallers after a 9 year follow-up. However, no associations were found between suffering only one fall and subsequent survival (HR = 1.1; 95% CI = 0.7–1.7) [24]. Donald et al. found, that all-cause mortality did not increase for single fallers compared to non-fallers (HR = 0.97; 95% CI = 0.7–1.4), in a community cohort of adults over 75 [20]. This is consistent with our results, although we did not assess the impact on survival of subsequent falls during the follow-up period. Altuough, Donald et al. did find associations between falling more than once during the 3-year follow-up and mortality (HR = 1.9; 95% CI = 1.2–3.0).[20] Along the same lines, Bath et al.’s community-based study found associations between frequent fallers (≥3) and mortality (4-year follow-up) [22]. Nevertheless, they found no significant increase in mortality for older adults who had fallen once or twice compared to non-fallers [22]. In an earlier study from 1992, Dunn et al. studied a community sample of adults ≥70 years with a 2 year follow-up [21]. They did not find any associations between suffering either one or multiple falls and mortality, after adjusting for the main covariates (including disability and chronic conditions). The HR were 1.3 (95% CI = 0.9–1.7) for a single fall and 1.3 (95% CI = 0.9–1.8) for multiple falls, as compared to an appropriate control group of non-fallers [21].

Bailly et al. studied a cohort of 329 community-based women who had suffered a fall with a 13 year follow-up. They found that elderly women who had suffered an indoor fall had shorter survival times than other kinds of falls. They considered that mortality differences were more strongly associated with factors such as frailty and comorbidities, which predisposed the subjects to falling, than to the fall itself and/or any subsequent injuries. They concluded that inside falls could be a marker of frailty. Unfortunately, they did not compare survival with a control group of non-fallers [19]. In line with this hypothesis, it is interesting to highlight that Wong et al. found that low falls (≤0.5 meters) for older adults from hospital settings were an independent predictor of 1-year mortality when compared to higher-level falls. Moreover, for low fall injuries, they considered that the long-term cause of death was unlikely to be trauma related and regarded low falls mechanism as a surrogate marker of frailty. They concluded that frailty status could explain the increased mortality found by other studies [17]. We did not adjust for frailty but, besides age and comorbidities, we included functional dependency in our analysis which could be considered a close proxy for frailty. We did not find a relevant increase in mortality risk for residents who had suffered a fall. This held even when severity of the fall was accounted for (although there was slight non-significant increase in risk for severe falls). Thus, our results are consistent with Wong et al. conclusions [17]. Moreover, in our subgroup analyses, severe falls were associated with a marked increase in mortality risk in the case of residents with apparently better baseline states (fewer comorbidities, fewer prescribed medications, functionally independent) and men. A possible explanation is that in patients with worse baseline states, mortality is primarily associated with frailty rather than with the fall itself, which is consistent with other studies’ conclusions [17, 19], and the impact of severe falls on mortality risk could be more marked on subpopulations with better baseline states. In line with this possible explanation, in those subpopulations with better baseline states, the reference group of non-fallers is expected to have good survival expectancies, so the mortality risk for those who suffered a severe fall could be remarkably higher when compared to them. On the contrary, in those populations with worse basal state (residents with frailty and more comorbidities) the reference group of non-fallers is supposed to have already bad survival expectancies. Previous studies have recommended the implementation of preventive measures aimed to attenuate frailty mainly in adults aged 75 years or older [8]. These could include, among others, the routine practice of moderate physical exercise [32] and a healthy eating pattern [33]. Our results are coherent with a possible benefit of primary preventive programs addressed to improve frailty in care homes residents. Further studies should assess the best preventive approaches of severe falls, for those in care homes residents with a better basal state.

In subgroup analyses, non-severe fallers, compared to non-fallers, presented a 69% increase in mortality risk in private facilities and a 37% decrease in risk in public institutions. However, we cannot provide a solid hypothesis for these differences in mortality risks. The important confounding factors have been controlled, so the reasons for a possible effect, if it really exists, are presumably due to differences in structural and organizational factors. Nevertheless, we do not have enough data to approach this analysis with the adequate rigor. Moreover, it is advisable to be cautious with the results from subgroup analyses since they are usually more affected by random and systematic errors. These results just suggest that future research should analyse in depth these possible associations.

Finally, in the specific case of hip fractures after an accidental fall in older populations, the hazard of death is greater during the subsequent 6–12 months and declines thereafter [34–39]. This is roughly consistent with our results which show a mild increase in risk of death for severe falls, mainly during the first year, and to a lesser extent during the second, with a progressive decrease thereafter.

A limitation of our study is that we only included residents who had suffered a fall during the month prior to the baseline interview in the analyses so we cannot rule out that some residents died shortly after the fall but before the interview (post-fall in-hospital mortality has been reported as between 3.7% and 16% [15, 16, 18, 40, 41]). Furthermore, residents who had suffered a fall and survived the first month could be stronger than non-fallers. Thus, our mortality rates after an accidental fall could be under-recorded, and the possible survival bias effect could hinder the identification of associations. However, as we were assessing long-term survival, this possible survival effect by definition must be included in the outcomes in terms of survival. Likewise, patients who survived a fall and subsequently stayed in the hospital for over a month were not included, so our figures for residents who had suffered a fall in the month prior to the baseline interview could have been underestimated. Second, some deaths might not have been reported, something that would eventually generate non-differential misclassification and, in general, lead to underestimation of the associations. Third, after a fall, mortality is higher in patients discharged into skilled nursing care or into a nursing home [15, 16]. As we studied care home residents, patients with better prognoses after a fall might not have been included in our study, but rather returned home instead of entering a care home. However, we adjusted for variables with a strong association with mortality after a fall, such as comorbid conditions, age, sex and functional dependency. Fourth, we cannot exclude that, in the last decade, the practice management regarding falls, could have changed in nursing homes leading to a possible improvement in the survival. Nonetheless, a possible recent improvement in the survival, would have been consistent with our results. Fifth, our main independent variable was to have suffered a fall within the month prior to the interview. Thus, we cannot ensure that, those classified as "non fallers", did not suffer a fall outside that period. In most of the studies from non-clinical settings with a long follow-up, the information of baseline falls was self-reported by asking the participant about having suffered a fall during the preceding 3 months or until the preceding year depending on the studies [20–22]. Donald et al. reported that they selected only the preceding three months, to self-report a previous fall, because they considered that the recall over a longer period was less accurate [20]. Anyhow, independently of the different timeframes for collecting the occurrence of a previous or subsequent falls, our results were consistent with the rest of the literature that did not find significant associations between suffering only a single fall and long term mortality [20–22, 24]. Finally, it is worth mentioning that data on falls were gathered from the facility physicians (or nurses for 8% of residents), thus notably limiting the potential information bias.

The main strengths of the study are that we selected a representative cohort of the care home resident population, with a very high response-rate and including various types of facilities. These facts enhance our study’s external validity. Moreover, the large sample size of the study limits random error. We have included the most important confounding variables in the analysis, such as urinary incontinence, comorbidities, age, sex and functional dependency [15, 40, 42–44]. Furthermore, these variables may collectively serve as proxies for unmeasured potential confounding factors, such as frailty.

Conclusions

We found no association between having suffered a fall in the month prior to baseline interview and long-term overall survival in care home residents. Additionally, we did not find a marked association when taking into account the severity of the fall, although some associations in selected subgroups warrant further attention.