Introduction

Over 1.2 million U.S. older adults aged ≥65 years reside in a nursing home (NH) [1]. Physical frailty, characterized by decreased physiologic reserve and increased vulnerability to exogenous stressors [2], and cognitive impairment, ranging from mild cognitive impairment to fully-developed dementia [3], are the two most prominent conditions in this population. Both are highly prevalent, with 30–85% of older nursing home residents experiencing physical frailty [4,5,6] and 65% moderate to severe cognitive impairment [1]. Both are associated with adverse health outcomes, including lowered quality of life and elevated risks for hospitalization and mortality [4, 7,8,Supplement Method). Because the subgroups of physical frailty did not vary across cognitive impairment levels, we included cognitive impairment as a covariate [41] to assess its association with the identified physical frailty subgroups using multinomial logistic model, adjusting for demographic and clinical characteristics. Results were presented in adjusted odds ratio (aOR) and 95% confidence interval (95%CI).

Sensitivity analysis

Consistent with the “cognitive frailty” concept by IANA and IAGG, we conducted three sets of sensitivity analysis by (A) excluding older residents with diagnosis of Alzheimer’s disease (n = 767,034); (B) excluding older residents with diagnosis of non- Alzheimer’s/other dementia (n = 529,832); (C) excluding older residents with diagnosis of Alzheimer’s disease and those with non- Alzheimer’s/other dementia (n = 460,612). In each subsample, LCA models were fit to identify the physical frailty subgroup, and the association between cognitive impairment and the identified subgroups were assessed following the same steps as the main analysis.

Results

Sample characteristics

As shown in Table 1, of the 871,801 newly-admitted older residents, 44.3% were ≥ 85 years old, 65.3% women, and 18.8% racial/ethnic minority. Approximately three quarters of residents entered urban nursing homes. Nearly two-thirds were admitted from acute hospitals and less than one in five from the community. At admission, nearly two thirds of residents were physically frail, one in four was pre-frail, and over a third had severe cognitive impairment. About 45% of older residents had more than two physician-documented active diagnoses. Two in five reported presence of pain. Receipt of antidepressants (43.9%), antipsychotics (19.2%) and antianxiety medications (18.4%) were common.

Table 1 Demographic and clinical characteristics of newly-admitted older nursing home residents (2014–2016)
Full size table

Indicators of physical frailty

Of all residents, 62.1% did not experience fatigue, 91.5% needed physical assistance to transfer between surfaces, 85.7% could not walk between locations in a room, 57.5% experienced bowel incontinence, 3.0% lost at least 5% of weight in the past 3 months or 10% of weight in the past 6 months, 67.8% were on a regular diet, and 95.2% needed help with dressing. Similar distributions were observed across cognitive impairment level except for a few items. For older adults with severe cognitive impairment, there were higher proportions who did not experience fatigue, had bowel incontinence, and needed mechanically altered diet (Table 2).

Table 2 Physical frailty indicators by cognitive impairment in newly-admitted older nursing home residents (2014–2016)
Full size table

Subgroups of physical frailty

For model selection, although entropy favored the 2-subgroup model, a clinically relevant subgroup emerged in the 3-subgroup model based on the item response probabilities. While AIC/BIC/adjusted BIC values favored models with more subgroups, for models with 4–6 subgroups, at least two of the identified subgroups largely overlapped and lacked sufficient separation. With these considerations, we chose the 3-subgroup model to represent physical frailty subgroups in nursing home residents at admission (Supplement Table S.2).

Based on the item-response probabilities, we assigned qualitative labels to the three subgroups: “mild physical frailty”, “moderate physical frailty” and “severe physical frailty”. (Table 3) About 7.6% of older residents had higher probabilities to belong to the “mild physical frailty” subgroup, 44.5% to the “moderate physical frailty” subgroup, and 47.9% to the “severe physical frailty” subgroup. The major difference between the “mild physical frailty” subgroup and the other two subgroups were reflected in the probabilities for resistance and ambulation: older adults that were likely to be in the “moderate physical frailty” or the “severe physical frailty” subgroups had high probabilities of needing physical assistance to transfer between locations and inability to walk in a room. The “moderate physical frailty” subgroup and the “severe physical frailty” subgroup were mainly distinguished by the item-response probability for the incontinence item: residents belonging to the “moderate physical frailty” subgroup had about an equal probability of having no urinary incontinence, urinary incontinence only, or urinary and bowel incontinence, while the “severe physical frailty” subgroup had a high probability of both urinary and bowel incontinence.

Table 3 Physical frailty 3-class latent class model: subgroup prevalence and item-response probabilities of indicators
Full size table

In sensitivity analysis when older residents with Alzheimer’s disease and/or those with non-Alzheimer’s/other dementia were excluded, the three-subgroup model appeared to best fit all three subpopulations (Supplement Table S.3). The overall patterns of the item-response probabilities and the respective prevalence of the physical frailty subgroups were similar and consistent with the full sample: “mild physical frailty” (prevalence range: 6.4–7.2%), “moderate physical frailty” (45.0–47.4%), and “severe physical frailty” (46.1–47.7%) (Supplement Table S.4).

Association between physical frailty subgroups and cognitive impairment

The three subgroups appeared consistent across cognitive impairment levels (Supplement Table S.5 and S.6). Cognitive impairment was thus included as a covariate in the 3-subgroup LCA model to examine its association with physical frailty subgroups, with the “mild physical frailty” subgroup as the reference, adjusting for demographic and clinical characteristics (Table 4).

Table 4 Association between physical frailty subgroups and cognitive impairment in newly-admitted older nursing home residentsa
Full size table

Compared to those with none/mild cognitive impairment, older residents with moderate impairment had similar odds to belong to the “moderate physical frailty” subgroup (aOR: 1.01, 95%: 0.99–1.03), while over twice as likely (aOR: 2.41, 95%CI: 2.35–2.47) to belong to the “severe physical frailty” subgroup; older residents with severe impairment had slightly higher odds to belong to the “moderate physical frailty” subgroup (aOR: 1.03, 95%CI: 1.01–1.05), and were close to 6 times as likely (aOR: 5.74; 95%CI: 5.58–5.90) to belong to the “severe physical frailty” subgroup.

For demographic and clinical characteristics, older age and being female were associated with higher odds of belonging to the “moderate physical frailty” or “severe physical frailty” subgroups, compared to their respective counterparts. Older residents who were racial/ethnic minorities were less likely to belong to the “moderate physical frailty” subgroup, but more likely to belong to the “severe physical frailty” subgroup. Older residents in rural nursing homes were less likely to be in the “moderate physical frailty” or “severe physical frailty” subgroups than those in urban nursing homes. Older residents admitted from acute hospitals had much higher probabilities of belonging to the “moderate physical frailty” and “severe physical frailty” subgroup than those admitted from the community.

Older residents with cancer, heart failure, diabetes mellitus, cerebrovascular accident/TIA/stroke, multiple sclerosis, Parkinson’s disease, seizure disorder/epilepsy, hip fracture, other facture, or depression had higher odds of belong to the “moderate physical frailty” or “severe physical frailty” subgroups, while those with anxiety disorder had lower odds. Older residents with hypertension, arthritis or osteoporosis were more likely to belong to the “moderate physical frailty” subgroup, but less likely to be in the “severe physical frailty” subgroup. Older residents with any pain presence at admission were more likely to be in the “moderate physical frailty” or “severe physical frailty” subgroups. Older residents who received antipsychotics were less likely to be in the “moderate physical frailty” or “severe physical frailty” subgroups, while those who received antianxiety medications or antidepressants were more likely to do so.

Findings from sensitivity analysis suggested consistent positive association between cognitive impairment and physical frailty subgroups, but the magnitude of these associations increased (Supplement Table S.7). Particularly, in the absence of Alzheimer’s disease and non-Alzheimer’s/other dementia, older residents with severe cognitive impairment were 8.55 times (95% CI: 8.18–8.92) as likely to be in the “severe physical frailty” subgroup, compared to those with none/mild cognition.

Discussion

In older adults in U.S. nursing homes, we identified three subgroups of physical frailty at nursing home admission, namely, “mild physical frailty”, “moderate physical frailty” and “severe physical frailty”. Physical frailty subgroups did not appear to differ across cognitive impairment levels. Older residents with greater levels of cognitive impairment were more likely to belong to the “moderate physical frailty” or “severe physical frailty” subgroups. Recent research has shown the possibility to reduce the prevalence or even reverse the progress of physical frailty through physical activity programs, cognitive training, nutritional supplementation, and interventions individualized to older adults’ clinical conditions [12, 13, 42]. However, these studies were conducted in community-dwelling older adults. Whether physical frailty could also serve as an intervention target for older nursing home residents warrants further exploration. To the best of our knowledge, this is the first study to provide evidence for the heterogeneity of physical frailty in older nursing home residents and its association with cognitive impairment, which can inform the development of interventions tailored to specific clinical profiles of physical frailty and cognitive impairment, while also considering the potential impact from other demographic and clinical characteristics.

The majority of the older nursing home residents in this study had high probabilities of belonging to either “moderate physical frailty” or “severe physical frailty” subgroups. This was expected as nearly two-thirds of the older nursing home residents were admitted post-hospitalization, indicating a more clinically complex group with greater care needs. The use of LCA allowed us to examine the heterogeneity of physical frailty by identifying three distinct subgroups. Regardless of the subgroups they were more likely to belong to, all residents had a high probability of requiring assistance with dressing. Besides the high probabilities of limited mobility that older residents belonging to the “moderate physical frailty” subgroup or the “severe physical frailty” subgroup were shown to have, those in the “severe physical frailty” subgroup also had particularly greater probability of bowel incontinence. Such distinctive experiences would be masked when physical frailty is measured by categorizing a total score into robust/pre-frail/frail levels. Using the LCA person-centered approach, findings not only reflected the increasing levels of physical frailty severity, but also provided a more nuanced picture of the physical frailty experience in older nursing home residents.

We note one important caveat that the characteristics of the subgroups to be identified by LCA is determined by the observed indicators, namely, FRAIL-NH items in the context of this study. Unique experiences of physical frailty in older nursing home residents that were not captured by FRAIL-NH would not be reflected in the identified subgroups. Therefore, other distinct subgroups of physical frailty may exist in older nursing home residents and future studies should consider additional metrics to provide a more comprehensive picture of the heterogeneity of physical frailty in this population.

The finding that greater levels of cognitive impairment was associated with increasingly higher odds to be in the “moderate physical frailty” and “severe physical frailty” subgroups provides additional evidence on the frequent co-occurrence of physical frailty and cognitive impairment, which has been established in older adults in the community [43, 44], but not in nursing homes. Further, in the sensitivity analysis when older residents with Alzheimer’s disease and those with non-Alzheimer’s/other dementia were excluded, the magnitude of the association between cognitive impairment and the “severe physical frailty” subgroup substantially increased, which could be indicative of “cognitive frailty”.

Regardless of older residents’ cognitive impairment levels, the characteristics of the identified physical frailty subgroups appeared to be similar, without notable differences in the patterns of the item-response probabilities. The consistent patterns of physical frailty subgroups were also observed in sensitivity analysis. These findings should be interpreted in light of the potential limitation of the instruments used to measure physical frailty and cognitive impairment. Despite several validation studies [4, 27,28,29,30,55]. Given that MDS 3.0 only documenting the receipt of psychotropic medications in the past 7 days or since nursing home admission and the cross-sectional nature of the current study, we could not ascertain the clinical indications for these prescriptions and the length of time that the older adults have been using them, nor could we establish a causal relationship between psychotropic medications and physical frailty subgroups, explicitly, whether it was the concerns for physical frailty that influenced the prescription of these medications, or the use of these medications lead to a higher probability to belong to a certain physical frailty subgroup. However, considering that physical frailty may increase older adults’ vulnerability to adverse drug effects [52], additional research to examine their long-term impact on physical frailty could further inform the consideration of psychotropic medications in managing physical frailty in this population.

Limitations should be noted. Our analysis focused on older residents who stayed for longer than 100 days in nursing homes with life expectancy at admission longer than 6 months. If residents’ length of stay and/or life expectancy were differential with regards to symptoms of physical frailty, cognitive impairment levels, or other demographic and clinical characteristics, selection bias cannot be ruled out. This was a cross-sectional study at nursing home admission. As physical frailty and cognitive impairment could change during residents’ stay, longitudinal studies may be informative in exploring if and how physical frailty subgroups and cognitive impairment change over time.

Conclusions

In summary, three subgroups of physical frailty were identified in older U.S. nursing home residents at admission, and older residents with greater levels of cognitive impairment were increasingly more likely to belong to the “moderate physical frailty” and “severe physical frailty” subgroups. Findings have implications for future efforts to tailor interventions to specific symptom profiles of physical frailty and cognitive impairment and provide new evidence for the interrelationship between these two prominent conditions in older nursing home residents.