Skip to main content
Download PDF

Predictive model to identify multiple synergistic effects of geriatric syndromes on quality of life in older adults: a hospital-based pilot study

BMC Geriatrics volume 25, Article number: 283 (2025) Cite this article

Abstract

Background

Quality of life (QOL) has been reported to be associated with sociodemographic characteristics and geriatric syndromes in older adults, but the impact of interactions among multiple geriatric syndromes on QOL remains unexplored. We aimed to apply a machine learning method to evaluate the effects of interactions among multiple geriatric syndromes on QOL in older adults.

Methods

We recruited adults aged ≥ 65 years admitted to a tertiary medical center from June 2018 to September 2018. The main outcome was the three-level five-dimensional Euro-Quality of Life tool (EQ-5D-3 L) utility value. The random forest algorithm was used to identify and rank the strongest predictors of geriatric syndromes. The relation between predictors and outcomes was visualized with accumulated local effects plots and interaction effects. Model performance was evaluated by 5-fold cross-validation with metrics of R-square, the mean square error of estimation and the mean absolute error of estimation.

Results

The study included 160 older adults with a mean age of 79 years. The top ten features that significantly influenced the utility prediction were activities of daily living (ADL), frailty, pain, the number of medications used, age, depression, the Charlson Comorbidity Index (CCI), body mass index (BMI), peptic ulcer, and emotional loneliness. The two-way interactions between ADL, frailty, and pain significantly interacted with other predictors.

Conclusion

ADL, frailty, and pain are important factors to be considered when assessing QOL in older adults. It is important for clinicians to consider them together in clinical decision-making.

Peer Review reports

Introduction

With the aging of the population and advances in medicine, the goal of health outcomes has shifted from morbidity and mortality to well-being and quality of life (QOL). It is a multidimensional construct complexly influenced by an individual’s physical health, psychological state, level of independence, social relationships, and environmental interactions [1].

Better QOL has been reported to be associated with higher education level [2, 3], higher income [2, 4], being engaged in social activities [2, 3], living with spouse [3], and having social support [4]. Poorer QOL has been reported to be associated with older age [2, 4], female gender [2, 5], social isolation [2, 6], and multiple morbidities [2, 7]. In addition, previous studies have indicated that poor QOL is associated with the presence of geriatric syndromes, including cognitive impairment [6, 8], functional impairment [3, 6,7,8], depression [4, 6, 8], falls [7], and polypharmacy [2].

However, the interactions among multiple geriatric syndromes on QOL remain unexplored. According to previous studies, multiple factors influence QOL in older adults, and these factors may exhibit nonlinearity and interact with each other [9]. It is difficult to use a traditional linear model to investigate the association between effects of interactions among multiple geriatric syndromes and QOL [10]. Traditional linear regression requires a priori identification of predictor interactions, a process that becomes increasingly complex and unreliable as the number of predictors grows. This approach is vulnerable to multicollinearity, particularly in high-dimensional spaces, potentially leading to biased estimates when interaction terms are incorporated. In contrast, machine learning algorithms are well-suited for handling high-dimensional, nonlinear data with multiple interactions, thereby overcoming the inherent limitations of classical linear regression techniques [11]. The objective of the present study was to apply a machine learning approach to assess the interactions among multiple geriatric syndromes on QOL in older adults.

Materials and methods

Participants

We recruited individuals aged ≥ 65 years admitted to medical wards at a tertiary medical center from June to September 2018. Individuals with hospital stay < 72 h, delirium, critical or terminal conditions, impaired communication abilities, or long-term bedridden were excluded. Impaired communication abilities were defined as the inability of older adults to respond to our questionnaire due to severe cognitive impairment, or severe hearing impairment even with the use of hearing aids. We excluded older adults with acute delirium or impaired communication abilities because they may not respond to our questionnaire correctly. We excluded older adults with long-term bedridden because those with long-term bed-ridden were associated with poor quality of life [12]. This investigation was conducted in strict accordance with established ethical guidelines, and informed consent was procured from all study participants following a comprehensive disclosure of the research protocol. Approval for the study was granted by the Institutional Review Board of National Cheng Kung University Hospital (A-ER-106-261).

Upon enrollment, eligible participants underwent a face-to-face comprehensive structural questionnaire interview within 72 h after their admission. The collected data included pertinent participant information, including demographic characteristics, comorbidities, and various geriatric syndromes.

Demographic characteristics

We collected data on the participants’ demographic attributes, including their age, sex, body mass index (BMI), marital status, family support, educational attainment, household income, and institutionalization status. Marital status was further stratified into two categories: married or living with a partner and living alone due to separation, divorce, widowhood, or never married [13]. Educational level was classified into three distinct categories: illiterate, elementary school, and high school or above.

Comorbidity

To collect data on comorbid diseases, we conducted a review of the electronic medical records and determined the Charlson Comorbidity Index (CCI) [14]. As a participant’s CCI increases, they face a higher risk of mortality within the next decade [15]. In the current study, we used the CCI to show the comorbidity status.

Comprehensive geriatric assessment (CGA)

CGA was first developed in the United Kingdom, and its concepts, indications, and applications evolved over time [16]. It is a multidomain, multidisciplinary diagnostic and therapeutic process performed to assess the physical, mental, social, and functional domains of older adults [17]. The physical domain includes the number of drugs used, nutrition status, urinary incontinence, and falls; the mental domain includes cognition and mood; the social domain includes living condition, marital status, and educational level; and the functional domain includes a frailty assessment. All the components of the CGA mentioned above were completed within 72 h after admission. For the physical domain, the number of medications used was reported by the individuals themselves. The assessment of activities of daily living was conducted using the Minimum Data Set-Activities of Daily Living (MDS-ADL) scale [18], with higher scores indicating a greater degree of dependence. Pain was defined by asking the individuals if they had pain or not, and it was confirmed if the answer was yes. Nutrition status was assessed by the Malnutrition Universal Screening Tool (MUST) [19]. It is a screening tool used to categorize adults into low-risk (MUST score = 0), medium-risk (MUST score = 1), and high-risk categories for malnutrition (MUST score ≥ 2) [19]. The participants with MUST scores ≥ 2 were defined as having malnutrition in the current study. Urinary incontinence was defined if the individuals reported leakage of urine on \(\:\ge\:\) 6 days within the past year. Fall was defined according to 1987 the Kellogg International Working Group on the prevention of falls in the older adults as ‘unintentionally coming to the ground or some lower level and other than as a consequence of sustaining a violent blow, loss of consciousness, sudden onset of paralysis as in stroke or an epileptic seizure‘ [20]. A fall was indicated if fall episodes occurred ≥ 2 times in the past 6 months before admission. For the mental domain, cognition was measured by the Short Portable Mental Status Questionnaire, which is a 10-item screening tool that was developed in 1975 [21]. Incorrect answers on more than 2 questions indicated impaired cognitive function. One more incorrect answer was allowed for the participants with a grade school education or lower, and one less incorrect answer was allowed for the participants with a high school education or higher [21]. Mood state was evaluated by the geriatric depression scale-5 (GDS-5), which is a 5-item screening tool that was developed in 1999 [22]. The GDS scores ranged from 0–5, with a sensitivity of 0.97 and a specificity of 0.85 using a score ≥ 2 as a cutoff point to define the presence of depressive symptoms [22]. For the functional domain, frailty was assessed by the clinical frailty scale (CFS) [23]. The CFS is a simple and intuitive clinical examination with a 9-point scale, and older adults with CFS ≥ 4 were classified as frail within 72 hours after admission.

Quality of life and utility value

QOL was assessed with the three-level five-dimensional Euro-Quality of Life tool (EQ-5D-3 L) [24]. It was selected for its widespread use and ease of administration. It demonstrates satisfactory discriminatory power with respect to construct properties, and is a feasible and valid measure in the older population [25]. The EQ-5D contains five dimensions, including mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The assessment of each dimension involved categorization into three levels: no problems, moderate problems, and severe problems. To standardize the data, utility was estimated using Taiwan population weights [26]. The utility value ranged from 0 to 1, where 0 indicated death and 1 indicated perfect health. Notably, negative values indicated health statuses considered worse than death.

Statistical analysis

Descriptive statistics were used to summarize the baseline characteristics of the study cohort. Continuous variables are described using the mean with the standardized deviation (SD) or the median with 25th and 75th percentiles. Categorical variables are described using frequencies and proportions. The features (or predictors) included age, sex, marital status, family support, household income, education level, BMI, the number of medications used, CCI, physical illness, cognitive impairment, depression, pain, urinary incontinence, falls, activities of daily living, frailty, social and emotional loneliness, and malnutrition. Two variables, metastatic solid tumor and human immunodeficiency virus/acquired immunodeficiency syndrome, were excluded from the analysis due to a 0% prevalence rate. In total, thirty-six features were included for modeling. The target variable was utility, which was converted using the EQ-5D scale.

Multiple imputation

Missing values appear in the BMI and CCI variables, and their percentages of missing was 2.5% and 0.6%. To reduce the bias of missing, we used multiple imputation by chained equations (MICE) to impute missing data [27, 28]. MICE is a multiple imputation technique that is a robust and informative method to deal with missing data in datasets [29]. We assumed that given the features used in the imputation procedure were missing at random. The MICE procedure runs a series of regression models in which each feature with missing data is modeled conditional on the other variables in the data set. This process continues until all specified variables have been imputed.

Choosing a machine learning algorithm

The random forest algorithm was employed to assess feature importance in predicting utility. This algorithm was selected for its utilization of bootstrap aggregation (bagging), an ensemble meta-algorithm that enhances model stability, accuracy, and robustness [30]. Hyperparameters, including the number of features for potential node splitting, tree quantity, and maximum tree depth, were optimized through sensitivity analyses to maximize predictive performance. Results are presented in supplementary Figs. 1 and 2.

Model performance

Model performance was evaluated using 5-fold cross-validation, a resampling procedure designed for limited data samples. The dataset was randomly partitioned into five subsets, with each subset serving as test data once while the remainder functioned as training data. This process was iterated five times to mitigate variability. Performance metrics, including R-squared (R2), mean square error (MSE), and mean absolute error (MAE), were averaged across iterations to estimate the model’s predictive capacity on novel data. Superior model performance was indicated by larger R2 values and smaller MSE and MAE values.

Machine learning interpretation: ranking variable importance, accumulated local effects, and feature interaction

Despite its power and robustness, the random forest algorithm lacks intuitive interpretability due to its black-box nature. To address this limitation, model-agnostic methods were employed to elucidate feature-target relationships:

  1. (1)

    Permutation feature importance: Quantifies the increase in prediction error after feature value permutation, with higher errors indicating greater importance [31].

  2. (2)

    Accumulated local effects (ALE) plots: Illustrate average feature effects on model predictions [32].

  3. (3)

    Feature interactions: Friedman’s H-statistic measures interaction strength, with values > 0.1 considered significant [33, 34].

These methods bridge the gap between random forest’s robust, personalized forecasting and the interpretability of traditional regression tools, which provide clear coefficient interpretations for average effects, strength, and directionality. Analysis focused on two-way interactions among the top three most important features. Implementations utilized R (version 4.2.0) and Quanta for Medical Care AI: AI Medical Platform (QOCA aim) v. 2.4.9 (Quanta Computer Inc., Taoyuan, Taiwan), with the R-package ranger for random forest and R-package interpretable machine learning (iml) for ALE and interaction plots.

Results

From June to September 2018, there were 283 adults aged ≥ 65 years admitted to medical wards. After excluding those with critical or terminal illness (n = 3), those with hospital stay < 72 h (n = 2), and those with impaired communication abilities (n = 46), remaining 232 participants. Of 232 individuals, 72 refused to participate the study, resulting in a total of 160 older adults included in the final study cohort (Supplementary Fig. 3). The mean age was 79 years (SD 8 years). There were 52 individuals (32%) aged 65 to 74 years, 64 individuals (40%) aged 75 to 84 years, and 44 individuals (28%) aged ≧ 85 years. The proportion of males (55%) was higher than that of females (45%). The older adults who were married or cohabiting was 34%. 93% of the older adults were satisfied with their family support. 11% of the older adults identified their household income as wealthy or well-off. 41% of the older adults had a high school education. 62% of the older adults had a BMI in the range of 18–24 kg/m2. The five most common physical comorbidities were renal disease (28%), any malignancy (19%), cerebrovascular disease (18%), congestive heart failure (14%), and peptic ulcer (12%). 49% of the older adults had a pain problem, and 41% had a history of falls. The average ADL scale was 7 (SD 10), and the average clinical frailty scale was 3.5 (SD 1.8) (Table 1).

Table 1 The baseline characteristics of the study population
Full size table

The predictive performance of the random forest model was evaluated, with an average R2 value of 0.73, MSE of 0.018, and MAE of 0.093 obtained from 5-fold cross-validation. The R2 is closer to 1 which indicates a better fit of the model. Lower MSE and MAE values indicate a model with better predictive accuracy, as it signifies that the predicted values are close to the actual values. Notably, during each round of cross-validation, little variance was observed for R2, MSE, and MAE when compared to the average values. This indicated that the predictions of the random forest model were robust and reliable (Supplementary Table 1).

The top predictors of utility, in descending order of importance, were ADL, frailty, pain, medication count, age, depression, CCI, BMI, peptic ulcer, and emotional loneliness. ADL, frailty, and pain exhibited the most significant influence on utility prediction (Fig. 1).

Fig. 1
figure 1

Permutation feature importance of the top 10 features for quality of life prediction. Abbreviations: ADL: activities of daily living; CCI: Charlson Comorbidity Index, CFS: clinical frailty scale; MSE: mean square error of estimation

Full size image

The ALE plot analysis revealed that CFS > 4 was associated with decreased utility, which aligns with the expert-based decision logic. A non-linear relationship was observed between BMI and utility, with values less than 20 kg/m² or greater than 28 kg/m² correlating with reduced QOL. ADL and emotional loneliness demonstrated negative linear correlations with utility. The presence of pain, depression, or peptic ulcer was associated with average QOL decreases of 0.03, 0.015, and 0.02, respectively (Fig. 2).

Fig. 2
figure 2

Accumulated local effects for the predicted change in quality of life of the prediction model with changes in features. Abbreviations: ADL: activities of daily living; BMI: body mass index, CCI: Charlson Comorbidity Index, CFS: clinical frailty scale

Full size image

Figures 3, 4 and 5 show the two-way interactions for ADL, frailty, and pain, each interacting with other characteristics. Regarding ADL, 14.7% of the effect of ADL came from its interaction with CFS. Frailty significantly interacted with pain and ADL. 18% of the effect of frailty came from its interaction with pain. Pain significantly interacted with frailty, the number of medications used, social loneliness, peptic ulcer, ADL, and depression. Among these features, pain and frailty had significantly higher strengths of interaction.

Fig. 3
figure 3

Two-way interactions of the feature of activities of daily living with other features in the prediction model. Abbreviations: ADL: activities of daily living; BMI: body mass index, CCI: Charlson Comorbidity Index, CFS: clinical frailty scale, MUST: malnutrition universal screening tool

Full size image
Fig. 4
figure 4

Two-way interactions of the feature of frailty with other features in the prediction model. Abbreviations: ADL: activities of daily living; BMI: body mass index, CCI: Charlson Comorbidity Index, CFS: clinical frailty scale, MUST: malnutrition universal screening tool

Full size image
Fig. 5
figure 5

Two-way interactions of the feature of pain with other features in the prediction model. Abbreviations: ADL: activities of daily living; BMI: body mass index, CCI: Charlson Comorbidity Index, CFS: clinical frailty scale, MUST: malnutrition universal screening tool

Full size image

Discussion

We demonstrated that ADL, frailty and pain were significantly associated with the prediction of QOL. In light of previous reports, one study showed frailty and depression were associated with poor QOL in older adults receiving home-based health care [35], and the other two studies revealed a negative relationship between frailty and QOL [36, 37]. The discrepancies between our findings and those of previous studies may be attributed to several factors. First, the previous studies included community-dwelling older adults [35,36,37], whereas the current study population comprised hospitalized older adults. Second, the measurement tools differed, including the World Health Organization Quality of Life-BREF by Tasioudi et al. [35] and Papathanasiou et al. [37], principal component analysis by Goyal & Mohanty [36], and EQ-5D in the current study.

In the current study, machine learning indicated that frailty with a CFS ≥ 4 was associated with worse quality of life (QOL) (Fig. 2B). It is important to note that CFS ≥ 4 is relevant to the clinical definition of living with frailty [23]. Frailty significantly diminishes muscle strength and endurance in older adults, leading to functional disability, and further compromises their QOL [38].

We also showed an inverted U-shaped relationship between BMI and QOL using machine learning, which is consistent with previous studies [39,40,41,42]. In addition, the current study showed that BMI of approximately 25–26 kg/m2 was associated with the best QOL. A previous study [41] reported that the best QOL was observed among males with BMI of approximately 22–29 kg/m2 and among females with BMI of approximately 22 kg/m2. Another study [39] proposed that an increase in the change of BMI by 2.50 kg/m2 in females (median was 26.10 kg/m2) and 5.19 kg/m2 in individuals with at least one disease was associated with the highest QOL. Our findings were consistent with those of previous studies that employed a machine learning model. Individuals with obesity or underweight may experience adverse effects on social and mental domains, which can result in negative self-perception and social interactions, ultimately leading to a decline in their QOL [39].

We demonstrated pain was associated with poor QOL, compatible with previous report [36]. Individuals aged 65 and above who experience pain tend to report lower QOL, which can be attributed severe health and environmental challenges [43]. The management of pain in older adults requires special attention due to the presence of various physiological, cognitive, functional, and social factors that may undergo change with advancing age, necessitating a tailored approach [44].

We revealed depression was associated with poor QOL, consistent with previous report [35]. Depression in older adults is not a normal part of aging process [45]. Late-life depression is distinguished by an atypical presentation of symptoms, including a greater prevalence of somatic symptoms relative to mood symptoms, rendering the condition challenging to detect and treat [46]. Nevertheless, older adults who are aware of their health status may have the potential to enhance their QOL [47].

In addition to the association between multiple geriatric syndromes and QOL, we further investigated the effect of the interactions among multiple geriatric syndromes on QOL in older adults. Our results indicated that 14.7% of the effect of ADL on QOL came from its interaction with frailty (Fig. 3), and 18% of the effect of frailty on QOL came from its interaction with pain (Fig. 4). Previous studies have reported that geriatric syndromes may interact with each other, chronic diseases or medications used; for example, one study reported that the interactions among geriatric syndromes predicted 3-month mortality [48], another study investigated the interactions between geriatric syndromes and chronic diseases [9], and another study showed the interaction between drugs and geriatric syndromes [49].

The ALE plot, a model-agnostic method, elucidates feature impacts on predictions in complex machine learning models. It operates by segmenting feature values and examining localized prediction changes, subsequently accumulating these effects to demonstrate overall feature influence. In ALE plots, the X-axis represents feature values, while the Y-axis depicts accumulated effects, with positive values indicating increased prediction and negative values indicating decreased prediction. For slope characteristics, steepness indicates effect strength, direction reveals effect nature, flatness suggests minimal impact, and non-linearity (e.g., U-shape) implies complex relationships. This approach enhances model transparency, facilitating comprehensive understanding of feature effects and supporting informed decision-making.

Friedman’s H-statistic quantifies feature interaction effects on prediction variance. It ranges from 0 (no interaction) to 1 (all variance explained by partial dependence function sums). Interactions with H > 0.1 are considered significant, with the magnitude indicating the proportion of a feature’s effect attributable to its interaction with another feature.

The current study has several strengths. First, to our knowledge, it is the first study to demonstrate the effects of interactions among multiple geriatric syndromes on QOL. Second, machine learning can deal with the high-dimensional exploration of complex data and can identify nonlinear and interactive effects [50]. When the underlying mechanism of a research question is complex, machine learning-based methodologies can be used to investigate the associations among the factors of interest. Third, machine learning models’ capacity to capture complex interactions among geriatric syndromes offers insights for personalized interventions. For example, strong frailty-pain interaction in QOL prediction suggests a need for integrated clinical approaches, while non-linear BMI-QOL relationship underscores the importance of individualized weight management strategies for older adults. Despite the incomplete understanding of the relationship between frailty and pain, recent evidence [51] has indicated the presence of a bidirectional relationship between the two. This relationship has the potential to create a vicious cycle, whereby each serves to accelerate the progression of the other [51]. It underscores the importance of pain assessment and management in the recognition and prevention of frailty [52]. Integration of these algorithms into clinical workflows could facilitate risk stratification based on unique combinations of geriatric syndromes, enabling timely, targeted interventions.

The study has inherent limitations. First, this was a single-center study with a small sample size. This could have led to variability in the data, which we addressed by using cross-validation to assess model performance. Second, we recruited only older adults hospitalized with medical conditions, so the results may not be generalizable to community-dwelling older adults. Third, the number of medications used was self-reported or proxy-reported, which may have introduced recall bias. Fourth, although the random forest algorithm offers robustness and excellent predictive performance, it may be susceptible to bias when dealing with categorical features with multiple levels, leading to an overemphasis on the importance of these features. However, all categorical features were binary in the current study, which would reduce the bias. Fifth, the overlap in domains of EQ-5D may lead to overestimation of relationships, and difficulty in isolating unique contributions to QOL. Ultimately, given the current study’s cross-sectional design and lack of environmental factors, further research is necessary to elucidate the longitudinal trajectories of geriatric syndromes and environmental factors with their impact on QOL over time.

Conclusion

The machine-learning model was able to identify a number of factors that were important in the prediction of QOL, including ADL, frailty, pain, the number of medications used, age, depression, CCI, BMI, peptic ulcer, and emotional loneliness. Among these factors, ADL, frailty, and pain significantly interacted with the other risk factors. These interactions suggest that ADL, frailty, and pain have complex effects on QOL in older adults. These interactions need to be considered in clinical decision-making.

Data availability

No datasets were generated or analysed during the current study.

Abbreviations

ADL:

Activities of daily living

ALE:

Accumulated local effects

BMI:

Body mass index

CCI:

Charlson Comorbidity Index

CFS:

clinical frailty scale

CGA:

Comprehensive geriatric assessment

EQ-5D-3L:

Three-level five-dimensional Euro-Quality of Life tool

GDS-5:

Geriatric depression scale-5

MAE:

Mean absolute error of estimation

MICE:

Multiple imputation by chained equations

MSE:

Mean square error of estimation

MUST:

Malnutrition universal screening tool

QOL:

Quality of life

SD:

Standardized deviation

References

  1. Study protocol for the World Health. Organization project to develop a quality of life assessment instrument (WHOQOL). Qual Life Res. 1993;2(2):153–9.

    Article  Google Scholar 

  2. Siqeca F, Yip O, Mendieta MJ, et al. Factors associated with health-related quality of life among home-dwelling older adults aged 75 or older in Switzerland: a cross-sectional study. Health Qual Life Outcomes. 2022;20(1):166.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Wijesiri H, Wasalathanthri S, De Silva Weliange S, Wijeyaratne CN. Quality of life and its associated factors among home-dwelling older people residing in the district of Colombo, Sri Lanka: a community-based cross-sectional study. BMJ Open. 2023;13(4):e068773.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Samadarshi SCA, Taechaboonsermsak P, Tipayamongkholgul M, Yodmai K. Quality of life and associated factors amongst older adults in a remote community, Nepal. J Health Res. 2021;36(1):56–67.

    Article  Google Scholar 

  5. Attafuah PYA, Everink I, Abuosi AA, Lohrmann C, Schols J. Quality of life of older adults and associated factors in Ghanaian urban slums: a cross-sectional study. BMJ Open. 2022;12(2):e057264.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Chen H-M, Chen C-M. Factors associated with quality of life among older adults with chronic disease in Taiwan. Int J Gerontol. 2017;11(1):12–5.

    Article  Google Scholar 

  7. Li PS, Hsieh CJ, Miao NF. A study of physical activity, frailty, and health-related quality of life among community-dwelling older adults in Taiwan. J Nurs Res. 2020;28(6):e124.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Chang HK, Gil CR, Kim HJ, Bea HJ. Factors affecting quality of life among the elderly in long-term care hospitals. J Nurs Res. 2020;29(1):e134.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Wu J, Kang L, Yang M, Rossi AP. Impacts of common geriatric syndromes and their interaction with chronic diseases on health. Front Med. 2022;9:1029246.

    Article  Google Scholar 

  10. Fried LP, Xue QL, Cappola AR, et al. Nonlinear multisystem physiological dysregulation associated with frailty in older women: implications for etiology and treatment. J Gerontol Biol Sci Med Sci. 2009;64(10):1049–57.

    Article  Google Scholar 

  11. Sidey-Gibbons JAM, Sidey-Gibbons CJ. Machine learning in medicine: a practical introduction. BMC Med Res Methodol. 2019;19(1):64.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Huang CH, Umegaki H, Kamitani H, et al. Change in quality of life and potentially associated factors in patients receiving home-based primary care: a prospective cohort study. BMC Geriatr. 2019;19(1):21.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Hsu HC. Trajectories and covariates of life satisfaction among older adults in Taiwan. Arch Gerontol Geriatr. 2012;55(1):210–6.

    Article  PubMed  Google Scholar 

  14. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–83.

    Article  CAS  PubMed  Google Scholar 

  15. Charlson ME, Carrozzino D, Guidi J, Patierno C. Charlson comorbidity index: a critical review of clinimetric properties. Psychother Psychosom. 2022;91(1):8–35.

    Article  PubMed  Google Scholar 

  16. Matthews DA. Dr. Marjory Warren and the origin of British geriatrics. J Am Geriatr Soc. 1984;32(4):253–8.

    Article  CAS  PubMed  Google Scholar 

  17. Ellis G, Gardner M, Tsiachristas A, et al. Comprehensive geriatric assessment for older adults admitted to hospital. Cochrane Database Syst Rev. 2017;9(9):Cd006211.

    PubMed  Google Scholar 

  18. Morris JN, Fries BE, Morris SA. Scaling ADLs within the MDS. J Gerontol Biol Sci Med Sci. 1999;54(11):M546–53.

    Article  CAS  Google Scholar 

  19. Stratton RJ, Hackston A, Longmore D, et al. Malnutrition in hospital outpatients and inpatients: prevalence, concurrent validity and ease of use of the ‘malnutrition universal screening tool’ (‘MUST’) for adults. Br J Nutr. 2004;92(5):799–808.

    Article  CAS  PubMed  Google Scholar 

  20. The prevention of falls in later life. A report of the Kellogg international work group on the prevention of falls by the elderly. Dan Med Bull. 1987;34(Suppl 4):1–24.

    Google Scholar 

  21. Pfeiffer E. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. J Am Geriatr Soc. 1975;23(10):433–41.

    Article  CAS  PubMed  Google Scholar 

  22. Hoyl MT, Alessi CA, Harker JO, et al. Development and testing of a five-item version of the geriatric depression scale. J Am Geriatr Soc. 1999;47(7):873–8.

    Article  CAS  PubMed  Google Scholar 

  23. Rockwood K, Theou O. Using the clinical frailty scale in allocating scarce health care resources. Can Geriatr J. 2020;23(3):210–5.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Rabin R, de Charro F. EQ-5D: a measure of health status from the EuroQol group. Ann Med. 2001;33(5):337–43.

    Article  CAS  PubMed  Google Scholar 

  25. Marten O, Brand L, Greiner W. Feasibility of the EQ-5D in the elderly population: a systematic review of the literature. Qual Life Res. 2022;31(6):1621–37.

    Article  PubMed  Google Scholar 

  26. Lee HY, Hung MC, Hu FC, Chang YY, Hsieh CL, Wang JD. Estimating quality weights for EQ-5D (EuroQol-5 dimensions) health States with the time trade-off method in Taiwan. J Formos Med Assoc. 2013;112(11):699–706.

    Article  PubMed  Google Scholar 

  27. Azur MJ, Stuart EA, Frangakis C, Leaf PJ. Multiple imputation by chained equations: what is it and how does it work? Int J Methods Psychiatr Res. 2011;20(1):40–9.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Jannat-Khah DP, Unterbrink M, McNairy M, Pierre S, Fitzgerald DW, Pape J, et al. Treating loss-to-follow-up as a missing data problem: a case study using a longitudinal cohort of HIV-infected patients in Haiti. BMC Public Health. 2018;18(1):1269.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med. 2016;4(2):30.

    PubMed  PubMed Central  Google Scholar 

  30. Fernández-Delgado M, Cernadas E, Barro S, Amorim D. Do we need hundreds of classifiers to solve real world classification problems? J Mach Learn Res. 2014;15(1):3133–81.

    Google Scholar 

  31. Breiman L. Random forests. Mach Learn. 2001;45:5–32.

    Article  Google Scholar 

  32. Apley DW, Zhu J. Visualizing the effects of predictor variables in black box supervised learning models. J R Stat Soc B Stat Methodol. 2020;82(4):1059–86.

    Article  Google Scholar 

  33. Friedman JH, Popescu BE. Predictive learning via rule ensembles. 2008.

  34. Molnar C. Interpretable machine learning. Lulu. com; 2020.

  35. Tasioudi L, Aravantinou-Karlatou A, Karavasileiadou S, Almegewly WH, Androulakis E, Kleisiaris C. The impact of frailty and geriatric syndromes on the quality of life of older adults receiving home-based healthcare: a cross-sectional survey. Healthcare. 2022;11:82.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Goyal AK, Mohanty SK. Association of pain and quality of life among middle-aged and older adults of India. BMC Geriatr. 2022;22(1):939.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Papathanasiou IV, Rammogianni A, Papagiannis D, et al. Frailty and quality of life among community-dwelling older adults. Cureus. 2021;13(2):e13049.

    PubMed  PubMed Central  Google Scholar 

  38. Ruiz JG, Dent E, Morley JE, et al. Screening for and managing the person with frailty in primary care: ICFSR consensus guidelines. J Nutr Health Aging. 2020;24(9):920–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Zawisza K, Tobiasz-Adamczyk B, Galas A, Jabłońska K, Grodzicki T. Changes in body mass index and quality of life-Population-based follow-up study COURAGE and COURAGE-POLFUS, Poland. Appl Res Qual. 2021;16:501–26.

    Article  Google Scholar 

  40. Apple R, Samuels LR, Fonnesbeck C, et al. Body mass index and health-related quality of life. Obes Sci Pract. 2018;4(5):417–26.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Laxy M, Teuner C, Holle R, Kurz C. The association between BMI and health-related quality of life in the US population: sex, age and ethnicity matters. Int J Obes. 2018;42(3):318–26.

    Article  CAS  Google Scholar 

  42. Xu Y, Zhou Z, Li Y, et al. Exploring the nonlinear relationship between body mass index and health-related quality of life among adults: a cross-sectional study in Shaanxi Province, China. Health Qual Life Outcomes. 2015;13:153.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Dong HJ, Larsson B, Dragioti E, Bernfort L, Levin L, Gerdle B. Factors associated with life satisfaction in older adults with chronic pain (PainS65+). J Pain Res. 2020;13:475–89.

    Article  PubMed  PubMed Central  Google Scholar 

  44. Rajan J, Behrends M. Acute pain in older adults: recommendations for assessment and treatment. Anesthesiol Clin. 2019;37(3):507–20.

    Article  PubMed  Google Scholar 

  45. Casey DA. Depression in older adults: a treatable medical condition. Prim Care. 2017;44(3):499–510.

    Article  PubMed  Google Scholar 

  46. Devita M, De Salvo R, Ravelli A, et al. Recognizing depression in the elderly: practical guidance and challenges for clinical management. Neuropsychiatr Dis Treat. 2022;18:2867–80.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Han TC, Lin HS, Chen CM. Association between chronic disease self-management, health status, and quality of life in older Taiwanese adults with chronic illnesses. Healthcare. 2022;10(4):609.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Oud FMM, Schut MC, Spies PE, et al. Interaction between geriatric syndromes in predicting three months mortality risk. Arch Gerontol Geriatr. 2022;103:104774.

    Article  CAS  PubMed  Google Scholar 

  49. Onder G, Giovannini S, Sganga F, et al. Interactions between drugs and geriatric syndromes in nursing home and home care: results from shelter and IBenC projects. Aging Clin Exp Res. 2018;30(9):1015–21.

    Article  PubMed  Google Scholar 

  50. Rod NH, Broadbent A, Rod MH, Russo F, Arah OA, Stronks K. Complexity in epidemiology and public health. Addressing complex health problems through a mix of epidemiologic methods and data. Epidemiology. 2023;34(4):505–14.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Chaplin WJ, McWilliams DF, Millar BS, Gladman JRF, Walsh DA. The bidirectional relationship between chronic joint pain and frailty: data from the investigating musculoskeletal health and wellbeing cohort. BMC Geriatr. 2023;23(1):273.

    Article  PubMed  PubMed Central  Google Scholar 

  52. Chaplin WJ, Lewis HR, Shahtaheri SM, et al. The association of painful and non-painful morbidities with frailty: a cross sectional analysis of a cohort of community dwelling older people in England. BMC Geriatr. 2024;24(1):158.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We would like to show appreciation to American Journal Experts for English editing.

Funding

Funding for this study was provided by the National Cheng Kung University Hospital (Intramural grant: NCKUH-V101-10). The sponsors had no involvements in the study results.

Author information

Authors and Affiliations

  1. Clinical Innovation and Research Center, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan, Taiwan

    Chien-Chou Su

  2. Department of Nursing, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan, Taiwan

    Yung-Chen Yu

  3. Department of Geriatrics and Gerontology, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, No. 138 Sheng Li Road, Tainan City, 704, Tainan, Taiwan

    Deng-Chi Yang

  4. School of Medicine, College of Medicine, National Cheng Kung University, Tainan, Taiwan

    Deng-Chi Yang

  5. Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan

    Deng-Chi Yang

Authors
  1. Chien-Chou Su
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Yung-Chen Yu
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Deng-Chi Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

D.C.Y. and C.C.S. were responsible for the study design, interpretation of data, writing, and reviewing the manuscript. Y.C.Y. was responsible for collection of data, and writing the manuscript. All authors contributed to and approved the final manuscript.

Corresponding author

Correspondence to Deng-Chi Yang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consents to participate

The study was approved by the Institutional Review Board of National Cheng Kung University Hospital (A-ER-106-261). Each participant provided the written informed consent to participate in the current study. Because the identification number of each participant was transformed and encrypted, each participant’s anonymity was fully preserved.

Consent of publication

Not requires.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Su, CC., Yu, YC. & Yang, DC. Predictive model to identify multiple synergistic effects of geriatric syndromes on quality of life in older adults: a hospital-based pilot study. BMC Geriatr 25, 283 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12877-025-05931-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12877-025-05931-8

Keywords

BMC Geriatrics

ISSN: 1471-2318

Contact us