A simple scoring algorithm based on intrinsic capacity for functional ability in community-dwelling older adults in Taiwan

This study of healthy ageing concept inspires a new sight for health care in older people to focus on optimizing IC for further prevention the functional disability as they age. In the present study, we used a cross-sectional sample of older community-dwellers to explore the prevalence of impaired IC varied from 16.6 to 44.6% while the proportion of disabled ADL, IADL, and worsening frailty were 12.0 to 16.4%. To the best of our knowledge, this is the first study to survey the develop a model for predicting the functional ability of community-dwelling older adults in Taiwan. The algorithm-based model was well calibrated by integrating 7-domain of ICOPES-TW which was useful for prediction advanced daily function and detection worsening frailty among older adults. In addition, this model exhibited a moderate ability for discriminating between community-dwelling older adults with versus without IADL disability and worsening frailty.

Our study showed that age, gender, living alone, educational status, and financial independence were significantly different between older adults with or without BADL disability, but not in IADL disability or worsening frailty. Our result revealed IC scores independently predicted impairment of IADL and frailty, but for BADL, which is similar as a previous study in China [35]. Consistent with previous literature, the IC can provide valuable predictive information on an individual’s function [36, 37] and frailty status [13, 37], even after accounting for the number of personal characteristics and multimorbidity. However, our result was inconsistent with a structural equation analyses research showed IC had directed and indirect relationships with BADL [36]. The reason might because that our study sample was relative health and independence who were recruited from either outpatient clinics or health check-up community centers. Our study also found personal characteristics (e.g., age, gender, education, and financial independence) could significantly predict BADL disability; but failed to predict IADL disability and worsening frailty in community-dwelling older adults, which is similar as previous studies [38, 39].

In this study, we integrated IC-related variables from the aforementioned WHO conceptual framework [40] to propose a simple scoring algorithm. The scoring system dichotomizes older adults into high-risk (weighted cutoff ≥ 13 for IADL and ≥ 16 for frailty; non-weighted ≥ 2) and low-risk (weighted cutoff < 13 for IADL and < 16 for frailty; non-weighted < 2) populations to maximize the sensitivity and specificity of the prediction of IADL disability and worsening frailty. Therefore, early identification using four or more factors from a total of seven significant IC predictors among older adults can be an effective strategy for future clinical practice. Nonetheless, under the low prevalence of functional disability among the 1152 older adults included in this study, 861 (179 true positive plus 682 true negative) and 918 (174 true positive plus 744 true negative) were correctly functioning predicted with the overall accuracy of 74.7% and 79.7% for IADL disability and frailty worse respectively on the basis of proposed weighted cut-off points. Therefore, 408 and 393 older adults were recommended to undergo further IADL and frailty intervention respectively, although only 179 or 174 actually had IADL disability or frailty worse, resulting in a PPV of 43.9% versus 44.3%. Moreover, the overall accuracy for predicting IADL disability and worsening frailty was 76.6% and 79.7% respectively, based on the proposed non-weighted cutoff points. Out of the 882 older adults correctly predicted to be functioning well, 158 had actual IADL disability and 724 did not. Similarly, out of the 918 older adults correctly predicted to be functioning well, 150 had actual worsening frailty and 768 did not. In total, 345 older adults were recommended for IADL intervention and 210 for frailty intervention. However, only 158 and 150 of them respectively actually had IADL disability or worsening frailty. This results in a PPV of 45.8% for IADL disability and 71.4% for worsening frailty. Given the importance of early identification and strategy provision to prevent functional deterioration among high-risk community-dwelling older adults, our proposed scoring algorithm could be still considered useful in community practice even though several older adults could be over-diagnosed and intervened.

Although both weighted algorithms developed in this study had satisfactory predicted values with the 13 and 16 cut-off points for older adults’ functional ability, our result showed IC better predicted for worsening frailty compared to the IADL disability. Frailty, defined as the progressive decline of physiological systems with increasing vulnerability to stressors and exposing risks of adverse health outcomes, exhibits many similarities as the IC (i.e., composite of all mental and physical capacities) [41]. A Korean study has mentioned IC and frailty represented the two faces of the same coin, with one indicating the reserves of the individual and the other denoting the deficits that accumulate with aging [42]. Therefore, under certain aspects, IC might be considered as a sort of evolution of frailty; both attempts to anticipate the importance and necessity of comprehensively evaluated and adequately managed the aging individual based on integration and multidisciplinary of services [15]. Similarity with previous studies conducted in developing and developed country, our study revealed that older adults with impaired IC reported higher odds of functional disability [43]; as well as the other evidence that showed locomotion (e.g., grip strength and gait speed) could significantly predict IADL disability, but not other evaluated outcomes, such as BADL difficulty and frailty [44].

Furthermore, our study classified the locomotion domain of IC as either impaired or normal based on the time it takes for adults to transition from a seated to a standing position and back five times. If this duration exceeds 12 s, it indicates impaired locomotion in older individuals [45]. Frailty is assessed using CFS, a scale from 1 (very fit) to 9 (terminally ill), which evaluates mobility, energy, physical activity, and overall function based on the interviewer’s clinical judgment of older people [30]. In our study, frailty scores ranging from 4 to 9 were considered as cutoff points for worse frailty. Since a frailty score of four or above indicates that older people often experience limited activity symptoms and commonly complain of “slowing up,” individuals with this issue may have a significantly higher likelihood of failing the evaluation to stand and perform five back positions within 12 s. Due to the commonality of the contents existed within the two measures, our study has a strong association between locomotion impairment and increased frailty. Therefore, we suggested that the locomotion status could impact the progression of frailty.

Our simple scoring algorithms integrating the WHO ICOPE framework IC measures (i.e., cognition, locomotion, vitality, vision, hearing, psychological, and medication) is suitable for determining the most relevant determinants of disability and frailty among community-dwelling older adults. Among the seven IC-related components, six predictors expect vision capacity were identified to be significantly associated with the functional abilities in varied contribution. The results are consistent with a study assessing the performance of diagnostic measures of the ICOPE screening tool in European community-dwelling older adults, which showed only vision, but not other IC-related domains vision, had the lowest sensitivity among all IC domains between normal and altered population [46]. Among all IC predictors in the scoring algorithm, locomotion weighted most in predicting IADL and frailty. Previous studies had similar result; one research revealed impaired locomotion significant predicted frailty in 3 years [47] while the other French study had reported each IC domains demonstrated positive association with the risks of incident frailty and ADL disability, especially the limited mobility played the most critical role with 2.97 hazard ratio to predict worsening frailty [48]. Furthermore, we have identified polypharmacy is the second strong predictor for IADL and frailty among community-dwelling older adults in Taiwan. Several studies have added some relevant components, not only medication [49, 50] but also fall or urinary incontinence [42], for the comprehensiveness of the WHO ICOPE framework measure.

Although several articles investigated a similar prediction question, they were not identical in methodology [43, 47, 48, 51]. Jia et al. [47] conducted the longitudinal cohort study using the 3-year transition (keep well, improved, worsened, and kept poor) of IC and found impaired vitality and locomotion were associated with worsened or kept frail. In contrast, our study is a cross-sectional design. Shen et al. [51] employed a cross-section study design similar to ours, but they defined each domain on a 0–2 score range to represent the functional status to predict frail risk and frailty. They found vision impairment increased the risk of frailty after adjusting for the related potential confounders; however, the presented study showed impaired vision is not a significant predictor for either IADL or frailty and has been excluded from the scoring algorithm. Zeng et al’s study [43] determined the normal as 1 and decline as 0 in each domain and combined the vision and hearing as one score to represent the sensory domain status, which differs from our approach. Multivariate regression models showed only cognition, but no other IC domains showed significance to the 1-year follow-up IADL function. The other study by González-Bautista et al. [48] calculated the total score by adding the number of IC impairments (score range 0–6, with higher scores indicating greater impairment) as a marker of a higher risk of frailty and disability. The results showed limited mobility imposed the highest risk of incident frailty among IC domains over five years, and visual impairment cut points for the ICOPE sum score were ≥ 3 for incident frailty and ≥ 2 for incident IADL; both method and result are very close to our study. However, our novelty is that we used the regression coefficient weighting method to give each domain’s point of risk score. Thus, our study finding provided an evidence-based recommendation to incorporate the medication domain into the ICOPE-TW screening.

This study proposed that both weighting and non-weighting scoring systems yield similar AUCs for IADL (0.80 vs. 0.78) and frailty (0.90 vs. 0.86), indicating that both can be employed. Nonetheless, the scoring algorithm sensitivity in weighting is noticeably higher (0.91 vs. 0.79), resulting in fewer false negatives. This improvement in sensitivity helps identify elderly individuals who truly require subsequent interventions.

Our study has some limitations. First, participants were recruited from residents in southern Taiwan by using the convenience sampling method, potential selection bias restricted the generalizability might exist. The cross-sectional design also restricted our ability to investigate the temporal relationships between IC-based score and IADL and frailty. In terms of making predictive interpretations of our findings, we should be cautious. Second, although we had considered several confounders and adjusted in our prediction models; however, the potential of residual confounding cannot be completely eliminated. Third, for determining functional ability, the frailty measure in this study relied on the single item assessment while both BAAL and IADL assessments were self-reported. Thus, additional multi-item questionnaires or objective measures might be required to recognize functioning to prevent the potential for outcome misclassification bias. In addition, most measures in the present study were based on the nature of self-reports; therefore, reporting bias due to social desirability responding in self-reported measures may exist. Fourth, we evaluate the predictive performance of a model with the same data we used for training. It might be an exaggerated model performance. However, some argue that randomly split sample data into training and test data might ignore autocorrelation in data, still leading to an over-optimistic assessment of model predictive power [52]. Furthermore, the low prevalence rate of IADL disability and frailty (20.9% vs. 16.4%) in our sample may affect the prediction ability (i.e., NPV) when applied in other populations. For example, as for nursing home residents with higher prevalence of functional disability or frailty, there could be higher probability of positive results from ICOPE screening and in fact have functional disability. Therefore, further large-scale study with broader spectrum population such as residents in nursing home and is necessary to increase the utility of this algorithm. Thus, larger population studies with prospective longer-term outcome measures are necessary to validate our study findings.