This paper examines and compares the predictive value of five performance-based measures for the onset of ADL difficulties during an 18-month period. Unlike much of the previous research examining the relationship between performance-based measures and ADL dependency,5, 8-11
self-reported ADL difficulty was used as the outcome measure in this study.
Among the five performance-based measures, BBS was the most consistent and strong predictor for the onset of basic ADL difficulty over the 18-month study period. The c
-statistics suggested the predictive ability of BBS was excellent at 12 and 18 months, and the odds ratios found BBS the only significant performance measure throughout the 18-month period. From the sensitivity and specificity curves, we identified BBS scores of 46 and 53.5 achieved 90% of sensitivity and 90% of specificity, respectively, for the prediction of ADL difficulty over an 18-month period. Thus, older adults with BBS score less than 46 were more likely to develop basic ADL difficulty, and including them for early interventions may delay their onset of ADL difficulty. Older adults with BBS score greater than 53.5 can be excluded from early interventions because they were less likely to develop basic ADL difficulty. Those with BBS score between 46 and 54 can be included when the resources are available. Several studies have investigated the relationship between BBS and ADL disability among individuals post stroke, with Parkinson’s disease, traumatic brain injury and multiple sclerosis.13, 14, 18
Paltamaa et al. 16
identified BBS as the strongest predictor for both perceived difficulties and dependence in self care, mobility, and domestic life when comparing to selected performance measures including Box and Block Test, Kela Coordination test, postural stability tests, ten-meter walk test and six-minute walk among individuals with multiple sclerosis. However, no cutoff score for prediction was identified, and thus no comparison with the BBS scores identified in this study can be made. In addition, the predictive value of BBS for incident ADL difficulty in community-dwelling older adults has not been previously reported. Based on the results, we suggest the BBS can be useful in identifying community-dwelling older adults at risks for incident basic ADL difficulty over an 18-month period of time.
In the study, SPPB was an excellent predictor for onset of basic ADL difficulty at 12 months. The SPPB has been shown to predict self-reported disability, nursing home admission, and mortality.11
Previous studies have investigated the predictive value of SPPB for onset of ADL dependence, but not difficulty. Guralnik et al. in 2000 found SPPB allowed for the estimation of incident disability of ADLs at 1-year and 4-year follow-up in community-dwelling populations, and compared to older adults with a SPPB score of 9-12, those with a score of 4-6 had a relative risk ranging from 3.4 to 5.1 of developing ADL dependence.9
From the sensitivity and specificity curves for the SPPB in this study, scores of 6 and 8 represented 90% sensitivity and 90% specificity, respectively, for the prediction of ADL difficulty at 12 months. Although the two studies investigated different aspects of ADL disability, both studies suggest a SPPB score of 6 or less is highly indicative of one-year ADL disability, as indicated by difficulty or dependence. Among the five performance-based measures, the SPPB generated the highest optimal sensitivity and specificity value (6 months, 71%; 12 months, 74%; 18 months, 72%) of baseline cutoff score for onset of ADL difficulty over the 18-month period. Older adults with SPPB score less than 6 are more likely to benefit from early interventions than those with SPPB score greater than 8. SPPB, which takes only a short time to administer, can be useful clinically for screening older adults for onset of ADL difficulty.
In the current study, gait speed was an acceptable predictor (0.7 ≤ c
< 0.8) for the onset of basic ADL difficulty at 12 and 18 months. Gait speed is a common physical performance measure used in clinical practice, and it has been shown to be a good predictor of ADL dependence in previous studies. Guralnik et al. 9
in 2000 found gait speed (c
-statistic, 0.70) was nearly as good as SPPB (c
-statistic, 0.75) in predicting mobility and ADL disability at 1 year. Similar relationship between gait speed and SPPB were observed in the current study at one year (gait speed, c
=0.797; SPPB, c
=0.827). Although the predictive value of gait speed is not as good as BBS and SPPB, it can be measured in a very short time, which may be useful for very frail elderly who are unable to complete SPPB in clinical settings.
In the study, baseline grip strength was an acceptable predictor for onset of basic ADL difficulty at 12 and 18 months for female elderly. Grip strength was found to be an independent predictor of disability among older Mexican men and women.38
In a 25-year prospective study, Rantanen et al.12
found the risk of self-care disability, defined as self-reported ADL difficulty, was more than 2 times greater in male subjects with baseline grip strength in middle age less than 37 kg compared to those with grip strength greater than 42 kg. In this study, the sensitivity and specificity curves constructed identified grip strength of 12 kg and 25 kg achieved 90% of sensitivity and 90% of specificity, respectively, for the prediction of ADL difficulty in female subjects over an 18-month period (). The results suggest older female adults with grip strength less than 12 kg were more likely to develop basic ADL difficulty, and those with grip strength greater than 25 kg were less likely to develop basic ADL difficulty over an 18-month follow-up period.
Although the TUG has been used to identify older adults who are independent in transfer tasks (TUG time <20 seconds) and older adults dependent in ADLs (TUG time > 30 seconds), the predictive validity of TUG for onset of ADL difficulty was not established.17
Results from the study suggested the TUG was an acceptable (0.7 ≤ c
< 0.8) predictor for onset of basic ADL difficulty at 6, 12, or 18 months.
Many studies have investigated the predictive value of performance-based measures for onset of ADL disability. However, due to the differences in the operational definitions of disability, the results can be quite different. Jette reported 1.2 to 5 times greater estimates of the prevalence of ADL disability using difficulty scale than dependency scale.19
The onset rates of disability, defined by ADL difficulty, were much higher comparing to previous studies. In the study, the onset rates of disability among community-dwelling older adults who did not report ADL difficulties at baseline were 28%, 30%, and 32% at 6, 12, and 18 months, respectively. In two studies conducted by Gill et al. 4, 7
, ADL disability, defined as the onset of functional dependence (receiving personal assistance or being completely dependent) in one or more of the seven basic ADLs, developed in 9% and 16% of community-living older adults at one-year follow-up. In other studies investigators found a 27% onset rate of functional dependence at 6 years5
and 46% self-reported rate of at least one episode of functional dependence in basic ADLs at 3 years.6
In Guralnik’s study9
, the onset rates of functional dependence (inability to perform basic ADLs without help) for older adults who scored 4 to 6, 7 to 9, and 10 to 12 in SPPB were 8%, 2%, and 1%, respectively, at one year. In the current study, stratifying older adults by SPPB scores, the rates of ADL difficulty onset at one-year follow-up for those scored 4 to 6 and 7 to 9 in SPPB were 47% and 27%, respectively. Since the onset of ADL difficulty usually happens before dependence, and the rates of ADL disability are higher in this study. In addition to the differences between the measure of dependence and difficulty, the types of ADL tasks chosen can also generate different results and clinical implications. In Guralnik’s study9
, older adults reported mobility-related disability (inability to walk 0.5 mile or climb stairs without help) plus the inability to perform one or more of the four activities (moving from a bed to a chair, using the toilet, bathing, and walking across a small room) were recruited.9
In Shinkai’s study5
, older adults were independent in five basic ADLs at baseline. In the current study, older adults who reported difficulties in one of the seven items (bathing, dressing, eating, getting in/out of bed/chairs, personal hygiene, walking, using the toilet) were classified as having ADL disability. The greater range of types of ADL disability (any one of 7 basic activities) used in this study may have contributed to a greater number of older adults identified as having disability.
Characteristics of subjects also contribute to the differences in results between studies. Although most previous studies included older adults who were independent in ADLs at baseline, the baseline physical function levels were not similar. First, the mean age of the subjects in our study was 80 years, older than those who participated in previous studies.4, 6-8
Secondly, the subjects in our study had poorer physical function. Our participants scored 3 to 10 on the summary score of SPPB and mean gait speed for the sample was 0.68 m/s. In Guralnik’s study9
, older adults scored 4 to 12 on the summary SPPB score. In Shinkai’s study5
, older adults walked at a mean gait speed of 1.09 m/s at baseline. Older adults with gait speed of 1.0 m/s may be at the starting point of steepest decline in function.25
In the current study, older adults with a mean gait speed of 0.68 m/s, thus, the slope of decline is less likely to be as steep and lead to non-ambulatory status. It is possible that the rates of ADL disability are higher in this study due to the poorer condition of subjects.
The results of the study suggested BBS, followed by SPPB, TUG, gait speed and grip strength was the most predictive performance-based measure for onset of ADL difficulty. Since BBS encompasses some components (standing balance and sit-to-stand) of other performance-based measures, it is not surprising to find BBS the most significant predictor for ADL difficulty. Similarly, SPPB encompasses the gait speed component, and the predictive value of SPPB is higher than gait speed alone. The five performance tests examined in the study are performance-based tests that can be administered easily in the clinical settings. However, the different nature and requirement of the clinical tests are likely to influence their clinical use. For example, BBS appears to be the most consistent and strong predictor for ADL difficulty, but it takes more time to administer in clinical settings. For older adults with poor physical tolerance or in situations with time constraints, SPPB or gait speed, which also predicts ADL difficulty, can be used. Primary care physicians who have limited time in clinics may also use SPPB or gait speed to decide necessity for further referrals. On the other hand, BBS can be a very useful tool in rehabilitation settings as it provides more information about which tasks should be prioritized in treatment. TUG was an acceptable predictor for ADL difficulty, but its role in predicting disability has not been supported by as many previous studies as other tests. Grip strength is also an acceptable predictor for ADL difficulty at 12 months. However, it is a measure of upper extremity strength, which may not provide as many insights for treatment planning.
The present study has several limitations. Older adults in this study were followed for 18 months, which is shorter than the follow up period in previous studies.4-10
It is possible that some performance-based measures are more predictive of ADL disability with longer follow-up periods. Because the subjects for the current analyses were identified from older adults enrolled in a prospective cohort study which requires subjects to be under care of participating physicians, many older adults who changed their primary care physicians were lost to follow-ups at 18 months (about 10%) and the overall rate of subjects lost to follow up (31% at 18 months) was greater than in previous studies.4, 6-8
However, comparisons of baseline characteristics revealed no significant differences between subjects who were lost-to-follow-up and who were not. The results of the study cannot be generalized to all community-dwelling older adults as 90% of the subjects in the study were Caucasians with the mean age of 80 years old and mild to moderate frailty (SPPB=3-10). The number of subjects is relatively small compared to previous studies. In the current study, the number of subjects was far less than previous studies6, 8, 9
which included 754 to over 4000 subjects.
To our knowledge, this is the first study investigating and comparing the predictive value of performance-base measures for the onset of ADL difficulty. This is also the first study confirming the predictive value of BBS for incident ADL disability. The results from the study also raised questions regarding to the sensitivity of different performance-based measures. It is possible that a performance-based measure is more sensitive to certain ADL tasks in certain populations. Future studies may examine the BBS for a longer follow-up time, and include different types of ADL disability. Gill et al.6
found bathing as the most common ADL disability to develop among 754 non-disabled community-living older adults. Several common ADL disabilities assessed, such as dressing, eating, and personal hygiene, are partly or mostly upper extremity related tasks. Future studies may also investigate differences between the onset of ADL disability related to primarily upper versus lower extremity tasks.