Three measures of physical rehabilitation effectiveness in elderly patients: a prospective, longitudinal, comparative analysis

National and international research ethics guidelines were followed, including the Deontological Code of Ethics, Declaration of Helsinki (Fortaleza 2013) and Spain’s confidentiality law concerning personal data (Ley Orgánica 15/1999, 13 December). Detailed, understandable information was provided to patients and family members, and oral informed consent to participate was obtained from all potential participants during their hospital stay, before beginning the post-acute unit’s comprehensive geriatric assessment. In patients with dementia, oral informed consent was obtained from the main caregiver. The institution’s Clinical Ethics Committee approved the study and the informed consent process used (Comité Etico de Investigación Clínica Parc de Salut Mar: reference number 6370/I). STROBE Guidelines for reporting observational research cohorts were followed [26] (Additional file 1).

After completion of the comprehensive assessment, 68 patients were excluded from analysis because there was no evidence of functional decline. The characteristics of the 685 included patients are shown in Table 1.

At discharge (Table 1), patients had achieved major improvements in functional capacity (BI-adm = 28.00 ± 19.90; BI-disch = 57.80 ± 27.31), with a mean AFG increase of 25 BI points and a mean value of 61.7 % on the RFG and 2 points on the REI. At discharge, 21.2 % of patients were classified as having severe dependency; it should be noted that patients who died, were rehospitalized due to complications or were discharged to long-term care facilities were included in this group. Table 2 shows the patient groups by RFG and the respective AFG and REI values. Mean AFG and REI values differed significantly between the four patient groups, classified according to the level of change achieved in the RFG; the exception was the AFG comparison in Groups III and IV, which showed similar mean values.

As summarized in Table 3 by RII and in Table 4 by primary diagnosis, when effectiveness of the rehabilitation process was evaluated using the AFG, 14 variables were associated with greater AFG, eight of them (57.1 %) from the PHSV group (age < 85y, absence of previous history of heart failure, absence of previous history of pulmonary disease, absence of previous history of dementia, absence of comorbidity; higher premorbid Lawton and BI scores, living alone) and six (42.8 %) from the AHSV group (referral source different than acute care geriatric unit, lower score in BI-adm, better cognitive function, absence of dysphagia, absence of respiratory diagnosis, main diagnosis of endocrinopathy at admission). When increased RFG values were used to evaluate effectiveness, nine variables were significantly related, three of them (33.3 %) from the PHSV group (age under 85y, absence of previous history of dementia, higher scores in previous Lawton index) and six (66.6 %) from the AHSV group (higher BI-adm scores, better cognitive function, absence of dysphagia, absence of undernutrition, absence of respiratory diagnosis, main diagnosis of endocrinopathy at admission). Finally, the highest REI values also were associated with nine variables, four (44.4 %) from the PHSV group (absence of previous history of dementia, absence of comorbidity, higher BI-premorbid scores, living alone) and five (55.5 %) from the AHSV group (BI-adm score 41–60, absence of dysphagia, absence of undernutrition, absence of respiratory diagnosis, main diagnosis of endocrinopathy at admission). As shown in Tables 3 and 4, only five of the total 24 variables studied (20.8 %) were related to all three RIIs (history of dementia, respiratory disease, dysphagia, main diagnosis of endocrinopathy, functional capacity at admission).

Differences between groups for each of the three RIIs are shown in Table 3. On the AFG, the BI-adm 61–100 group had a significantly lower mean score than the other three groups; on the RFG, the BI-adm 0–20 group differed significantly from the other three, and on the REI the only significant difference was observed between the BI-adm 0–20 and 41–60 groups.

In multivariate analysis, four variables were related to achieving a minimal clinically important difference in AFG (≥20): Charlson index ≤3, absence of dysphagia, instrumental activities of daily living (Lawton index) and a BI-adm score between 11 and 50. In patients with an AFG >50, a ceiling effect was observed. These variables were included in both the PHSV and AHSV analysis. In multivariate results, absence of a respiratory diagnosis and a BI-adm score >11 were associated with a higher probability of high, or even complete, functional recovery (RFG ≥35 %) and the same two variables together with the absence of dysphagia were associated with a clinically important difference in REI (≥0.50). Therefore, BI-adm was the only variable related to all three RIIs (Table 5).

Although prognostic factors affecting rehabilitation outcomes and RII scores are known and have been studied by others, clear guidelines on the use and choice of the available RIIs are lacking. The novelty of the present study is its focus on the indications for using these indices in geriatric patients. We would highlight the different prognostic factors identified in the three different RIIs studied, and specifically, their advantages and limitations in elderly patients.

All three RIIs were useful to evaluate changes in functional capacity; however, each uses different formulas that can be affected by certain variables but not by others (Tables 3 and 4). The AFG was associated with a high number of variables in both the PHSV and AHSV groups; most of these associations have been reported by other authors [9, 24, 25]. In addition, functional capacity at admission was inversely associated with AFG (Table 3). In other words, less functional capacity was related to higher AFG, and vice versa. This must be taken into account when analysing rehabilitation outcomes in patients with slight or moderate disability because, in these cases, the AFG is limited by the maximum possible score on the scale that is used (100 points in the case of the BI). For example, the maximum AFG that a patient with a BI-adm value of 80 points can achieve is 20 (the ceiling effect of a 100-point scale). Therefore, the use of AFG as the only parameter of rehabilitation effectiveness in patients with light-moderate dependency could underestimate the results [9, 10, 27].

Given that the BI is not a continuous interval scale, AFG calculation using this index can be affected because a change in score does not have a consistent meaning in all patients. A 20-point AFG score is excellent for a patient who had lost 20 points prior to hospital admission; it represents 100 % recovery. However, it could be a disastrous outcome for a person who had lost 80 points from their BI-premorbid score. On the other hand, AFG only reflects an increase in “points on the scale” but does not provide any information about the maximum possible score on that scale; therefore, it is impossible to know whether the patient has achieved the maximum possible improvement. Therefore, some authors have suggested that AFG only indicates rehabilitation effectiveness in “reducing the level of dependency” [25].

Finally, it is difficult to establish the detected level of change in BI scores (AFG) that can be considered clinically relevant. Some authors have shown that a 20-point increase in the AFG has a favourable prognostic effect on long-term functional capacity and survival [12]; this limit has been used in other studies (including our own) as the threshold for clinically significant change [12, 23‐25]. For the REI, according to other authors [17], ≥0.50 points can be considered clinically significant.

The RFG was significantly related with more factors from the AHSV group than from the PHSV, perhaps because the RFG minimizes the effect of prior conditions on rehabilitation outcome due to its reliance on previous functional capacity as the maximum potential to be reached. This reduces “the goal” to the level of the patient’s approximate situation before the hospital admission, independent of medical history, comorbidities or previous level of independence. At the same time, RFG shows the proportion of functional capacity that has been lost recently, due to the acute process that led to hospitalization, without taking into account whether the disability existed before hospital admission. In other words, it shows the portion of disability that is potentially reversible. On the other hand, the RFG has no “ceiling effect”; AFG values decrease as BI-adm increases, while RFG increases according to BI-adm (Table 3). Finally, patients in Group IV, who achieved complete recovery (RFG = 100 %), had the same mean AFG values as others (Group III) who did not reach the same level of recovery but did recover more than one third of their lost functional capacity (RFG = 35–99 %) (Table 2). This suggests a lack of information from the AFG about the extent to which the patient fully recovers premorbid functional status. Taking all of these considerations into account, we conclude that RFG contributes more valuable –and more qualitative– information to evaluating the rehabilitation process in elderly patients, who often present with chronic diseases and prior disability.

Our study has several limitations that should be taken into account. A RFG score ≥35 % has been described as a threshold of clinical rehabilitation effectiveness [10, 17, 23] and no higher distribution has been published to date. In our sample, however, most of the patients (72.9 %) achieved a higher RFG, and a substantial percentage (27.5 %) scored 100 %; therefore, the authors followed clinical criteria to establish the RFG score groups in the present sample. As this stratification had not been previously described and could lead to non-homogeneous groups, this could be considered a limitation of our study. In addition, RFG values can be difficult to compare between studies because they vary according to whether or not the values were adjusted for the BI-premorbid score using the theoretical maximum score (100 points) or the actual (individualized) BI-premorbid score. Obviously, this has the potential to generate confusion in attempts to generalize the use of this parameter and may be a confounding factor for comparative analysis.

Mathematically, it is debatable whether a variable (BI-adm) already included in the formula of the same outcome variable (RFG) can be considered as a predictor. Nonetheless, we must take into account the RFG’s inclusion of the BI-disch score in addition to BI-adm and BI-premorbid; inclusion of BI-adm as an independent variable was based on a reasonable suspicion (supported by the modelling) that a patient’s capacity for recovery could be influenced by the functional status at admission. It is logical to think that patients in a worse initial functional state would have less capacity for recovery. For this reason, we included the BI-adm in the model as an independent variable. A new assessment after functional capacity has improved could be of interest to test the power of these RII scores in comparison with other performance scales, such as Tinetti score, Short Physical Performance Battery or Gait Speed.

REI was related to nine variables, four from the PHSV group and five from the AHSV group. In the present study, mean REI was 2.0, which is higher than that reported in other similar settings [16, 17, 24, 28]. A possible explanation is the low prevalence of neurological diagnoses in our study population (Table 1). It has been demonstrated that patients with neurological disorders have lower REI values than patients with orthogeriatric or other diagnoses [9, 28]. Otherwise, REI is clearly influenced by length of stay, which was shorter in the present study than in others [24, 25]; this could be due to the inclusion criterion requiring that participants meet conditions amenable to discharge within two weeks; these individuals may have had better prognosis than other study populations. Another aspect that could have contributed to the high REI observed is that the functional status of patients admitted to an acute care unit may be underestimated because of barriers such as catheters, intravenous lines, and bed-rails that are prevalent in conventional hospitalization. In the first days after patients are transferred to a geriatric rehabilitation unit, rapid functional improvement tends to occur as hospital care can be combined with integrated care designed to promote autonomy and minimize dependency (common dining, social spaces, etc.). In the hospital setting, REI value is highly dependent on the length of the hospital stay. This must be taken into account because length of stay can be influenced by comorbid conditions that may interrupt rehabilitation [16, 29] and by individual variables (social, personal, or family issues) that are not directly related to rehabilitation therapy [25, 28]. To avoid this potentially confounding factor, some authors have suggested that REI should be calculated on the basis of the entire period during which the patient receives rehabilitation therapy. Using this approach, the denominator is the number of days of rehabilitation therapy, from start to finish, regardless of where it is provided [9]. In our sample, BI-adm had a mean value of 28 points, similar to that obtained by other authors in similar settings [24, 25, 29].

As shown in Table 5, PHSV (comorbidity and Lawton index) were only related with AFG and not with the rest of the RIIs. Similarly, two of the AHSV variables (absence of dysphagia and absence of respiratory diagnosis) were related to both AFG and REI in multivariate analysis. Evidence of the AFG ceiling effect was obtained in the BI-adm >50 group. Only two variables were predictors of functional recovery (RFG ≥35 %): functional capacity at admission and the absence of a diagnosis of respiratory disease (both from the AHSV group). Only BI-adm was related to all three RIIs. The BI cut-off values that had predictive value reflect severe dependency. This may be due to the study population, which was severely incapacitated upon admission but had a better prognosis than populations in other studies, which would reduce the cut-off values used to discriminate worse functional prognosis (BI ≤10). A large proportion of the patients came from an acute care geriatric unit, where respiratory diagnoses are very frequent; a poor functional prognosis for these patients has been described [30], and supports our observation of a better functional recovery in the absence of respiratory disease.

Finally, we would note that AFG was the parameter that was correlated with the greatest number of prognostic variables, but has the disadvantage of a ceiling effect and also does not provide evidence of the final functional outcome of rehabilitation therapy. REI is highly conditioned by the mean length of stay, which could be influenced by multiple factors that do not depend on the rehabilitation process. RFG seems to be more associated with variables that reflect health status at admission (severity of the recent acute process) and less affected by previous health status. Also, RFG provides more precise information about the degree to which a patient returns to his or her premorbid status. A future implication of the present study is that these considerations should be taken into account when selecting parameters to determine rehabilitation effectiveness in elderly patients.