Frailty as a predictor of hospital length of stay after elective total joint replacements in elderly patients

Total joint replacement (TJR) procedures are increasing in number with population aging and osteoarthritis development. In its 2014 annual report, the Canadian Joint Replacement Registry recorded a 16.5% surge of hip replacements and a 21.5% proliferation of knee replacements in the last 5 years [1] paralleled by increased frailty with advancing age in the general population [2]. Frailty embodies insufficient physiological reserves and the inability to adequately respond to external stressors. Its etiology is complex but is related to aging, faster senescence and acquired illnesses [3, 4]. Its pervasiveness in community-dwelling elders is estimated to be approximately 10%, depending on the study cited [5], and is associated with increased 1-year mortality and institutionalization.

Frailty prevalence is generally considered to be even greater in surgical pre-operative settings, varying from 10% to 46%, depending on the screening instrument selected, the surgical subspecialty/setting, and patient characteristics [6‐8]. The choice of “frailty scale” may not be obvious owing to the variety of available instruments with different conceptual underpinnings [9, 10]. Nonetheless, in most studies, frailty is a better predictor of post-operative outcomes than co-morbidity scores and even some validated risk model scores specific to certain procedures [6‐8].

In orthopedic surgery, especially in patients awaiting TJR, few teams have looked at the association between frailty and post-operative outcomes, such as hospital length of stay (LOS) and disposition at discharge. Frailty may help in selecting vulnerable patients when combined with known predictors in TJR procedures (i.e., pre-operative hemoglobin (Hb) level, American Society of Anesthesiologists score, nutritional status, etc.) [11‐13]. The lack of consensus on the different available models of frailty makes it challenging to choose appropriate screening instruments.

Therefore, we undertook a study, with 2 frailty-screening instruments, to assess associations between frailty, hospital LOS and transfer to rehabilitation centre or not among elderly patients awaiting elective, primary TJR. We hypothesized that frail patients are less likely to be discharged home and will have longer LOS. As secondary objectives, we compared the prevalence of frailty with 2 instruments and evaluated their agreement.

The baseline characteristics of study subjects are presented in Table 1. Median age of our cohort was 72.0 years (interquartile range (IQR) = 9.0), ranging from 65.0 to 88.0 years, and 65.5% were female patients. More than half of our cohort were living at home with their spouses (57.5%), while 5.6% were living in a residence. The median Charlson Comorbidity Index in our cohort was 4.0 (IQR = 1.0), and 87.4% were fully independent in their daily living activities. TKA represented 51.7% of all procedures, followed by THA and bilateral TJR (39.1% and 9.2%, respectively). The mean time period between pre-operative clinic visit and surgery was 74.3 ± 34.8 days.

The Clinical Frailty Scale classified 46.0% of our cohort as robust while the FRAIL Scale grouped 15.0% as robust. Baseline characteristics (gender, disposition before surgery and BMI) differed significantly between robust and not robust patients based on the Clinical Frailty Scale (p < 0.05), while fully independent status almost reached statistical significance (p = 0.058). Robust and not robust patients did not differ significantly on the FRAIL Scale, except in terms of gender, mean BMI and median baseline Hb level (p < 0.05).

Median hospital LOS of our cohort was 5 days (IQR = 3). Eight patients (9.1%) were discharged within 48 h after surgery, 49 patients (56.3%) were discharged between 3 and 5 days, and total LOS was over 5 days in 30 patients (34.4%) (Table 2). No patient died during hospital stay. Based on the Clinical Frailty Scale, 17.5% (N = 7) of robust patients had hospital LOS over 5 days compared to 48.9% (N = 23) of not robust patients. Hospital LOS distribution differed significantly between robust and not robust patients (p = 0.005). When classified according to the FRAIL Scale, hospital LOS distribution did not differ significantly between robust and not robust patients (p = 0.700).

A total of 17 patients (19.5%) were discharged to rehabilitation center after surgery. Based on the Clinical Frailty Scale, 10.0% (N = 4) of robust patients were transferred to rehabilitation center compared to 27.7% (N = 13) of not robust patients (p = 0.038). Robust (N = 3, 23.1%) and not robust (N = 14, 23.3%) patients did not differ significantly on the FRAIL Scale (p = 0.727).

Table 3 reports patient classification according to frailty status on the Clinical Frailty Scale and the FRAIL Scale. Since both instruments are not rated by the same scaling system, the results are presented after reclassification into robust, pre-frail and frail patients. The Clinical Frailty Scale classified 40 patients (45.9%) as robust, 43 patients (49.4%) as pre-frail, and 4 patients (4.5%) as frail, while the FRAIL Scale categorized 12 patients (13.7%) as robust, 54 patients (62.0%) as pre-frail, and 20 patients (22.9%) as frail. Both scales agreed on 39 evaluations (38.0%) and disagreed on 48 evaluations (62.0%). Out of the 20 patients considered as frail by the FRAIL Scale, 5 were classified as robust with the Clinical Frailty Scale. Due to the small numbers of patients considered as frail by the latter, both scales were dichotomized into robust versus not robust (sum of pre-frail and frail) (Table 4). Cohen’s kappa agreement was 0.245 (p = 0.002) between both scales after dichotomization.

Frailty is progressively acknowledged as a marker of functional decline and a potentially modifiable risk factor for patient improvement. It is even more relevant in peri-operative patients where interventions can be applied before the surgical procedure. In our study, pre-frail and frail patients, as defined by the Clinical Frailty Scale and undergoing TKA or THA, had longer hospital LOS and higher percentage of discharge to rehabilitation center compared to robust patients. These results are consistent with current literature on the general surgical population. Most studies, using Rockwood’s or Fried’s model, have reported associations between increasing frailty and major post-operative outcomes (in-hospital mortality, hospital LOS, post-operative complications, etc.).

In orthopedic surgery, involving the Ontario Healthcare Database, McIsaac et al. [26] found an increase in hospital LOS among frail elderly undergoing THA or TKA based on Johns Hopkins Adjusted Clinical Groups frailty-defining diagnoses indicators (odds ratio: 1.78, confidence interval: 1.74–1.81). The latter model is based on another conceptual underpinning and is also difficult to apply in clinical practice as it relies on variables from multiple datasets. Cooper et al. [27] considered both Fried’s phenotypic model and the Frailty Index in a cohort of 415 orthopedic patients. Both instruments predicted discharge to post-acute institutional care and hospital LOS > 5 days. This discrepancy between results can be explained by the higher prevalence of frailty (35% with Fried’s phenotypic model and 41% with the Frailty Index) in older patients with greater functional decline. They included patients undergoing other types of surgery, such as lumbar and cervical laminectomy.

The FRAIL Scale was not associated with hospital LOS or discharge to rehabilitation center in our cohort. Both scales differed significantly in the percentage of patients classified as robust and not robust, with moderate agreement at best. Part of the difference in performance by both scales can be explained by the modest agreement between them. The 2 instruments are not conceptually identical in their definition of frailty. The Clinical Frailty Scale is derived from the Frailty Index developed by Rockwood and colleagues [17]. It is based on a mathematical model of deficit accumulation and the complex interplay between co-morbidities [15]. The FRAIL Scale is heavily rooted in Fried’s phenotypic model, a syndrome-based definition of weight loss, fatigue and diminished energy expenditure, all related to sarcopenia [28]. Cooper et al. [27] obtained similar results with the phenotypic and deficit accumulation models with moderate agreement (kappa = 0.42). Furthermore, our cohort of orthopedic patients awaiting TJR did not suffer from significant co-morbidities or disabilities, which are all related to frailty. By excluding urgent surgeries, such as hip fracture patients and those deemed too ill for surgery, we further selected a subgroup of healthier and more functional orthopedic patients. In a cohort with higher prevalence of frail patients, the FRAIL Scale may have been associated with clinical outcomes. Also, even with our small sample size, the power of our analyses with the FRAIL scale was 99% for both clinical outcomes. It is unlikely that our results were due to beta errors.

Our study has some limitations. First, it was conducted in a single center with a small sample size, and lacked power to assess other known risk factors and their interactions with frailty status. For example, Hb level was significantly higher in robust patients (with the FRAIL scale) compared to not robust patients, although a multi-centered interventional trial did not show improved outcome with a liberal transfusion strategy in hip surgery patients [29]. Hypoalbuminemia, another risk factor associated with sepsis after TJRs [30], was not captured due to a significant amount of missing data. It is not a mandatory test in our pre-operative clinic. Since it is not, validation is also needed across other cohorts of TJR patients. Second, these results cannot be applied to other surgical specialties, since frailty prevalence might differ from orthopedic patients, and it is still unclear which frailty instrument is better suited for a specific surgical subgroup. Third, as mentioned, we excluded urgent surgical procedures as the prevalence of frailty is usually higher in this group. In a previous study evaluating frailty with the Frailty Index in hip fracture patients, the prevalence of being very frail on admission was 36%. Therefore, our findings may not apply to them.

As for our study’s strengths, the recruitment process did not delay surgery and usual care, and was easily implemented in routine assessment due to the simplicity of both selected scales. Our results agree with the existing literature and, to our knowledge, it is the first time that the Clinical Frailty Scale and the FRAIL Scale have been applied in a TJR population.

In conclusion, the Clinical Frailty Scale was, while the FRAIL scale was not, significantly associated with hospital LOS and discharge to rehabilitation center in our cohort of TJR patients. Whether either instrument can be generalized to other surgical patients and used to target vulnerable elderly patients for pre-operative intervention needs to be investigated.