Falls in the elderly were predicted opportunistically using a decision tree and systematically using a database-driven screening tool

A systematic review of fall prediction models for community-dwelling older adults: comparison between models based on research cohorts and models based on routinely collected data

BACKGROUND

Prediction models can identify fall-prone individuals. Prediction models can be based on either data from research cohorts (cohort-based) or routinely collected data (RCD-based). We review and compare cohort-based and RCD-based studies describing the development and/or validation of fall prediction models for community-dwelling older adults.

METHODS

Medline and Embase were searched via Ovid until January 2023. We included studies describing the development or validation of multivariable prediction models of falls in older adults (60+). Both risk of bias and reporting quality were assessed using the PROBAST and TRIPOD, respectively.

RESULTS

We included and reviewed 28 relevant studies, describing 30 prediction models (23 cohort-based and 7 RCD-based), and external validation of two existing models (one cohort-based and one RCD-based). The median sample sizes for cohort-based and RCD-based studies were 1365 [interquartile range (IQR) 426-2766] versus 90 441 (IQR 56 442-128 157), and the ranges of fall rates were 5.4% to 60.4% versus 1.6% to 13.1%, respectively. Discrimination performance was comparable between cohort-based and RCD-based models, with the respective area under the receiver operating characteristic curves ranging from 0.65 to 0.88 versus 0.71 to 0.81. The median number of predictors in cohort-based final models was 6 (IQR 5-11); for RCD-based models, it was 16 (IQR 11-26). All but one cohort-based model had high bias risks, primarily due to deficiencies in statistical analysis and outcome determination.

CONCLUSIONS

Cohort-based models to predict falls in older adults in the community are plentiful. RCD-based models are yet in their infancy but provide comparable predictive performance with no additional data collection efforts. Future studies should focus on methodological and reporting quality.

Factors associated with fall risk of community-dwelling older people: A decision tree analysis

Objective

To examine the predictive attributes for accidental falls in community-dwelling older people in Hong Kong using decision tree analysis.

Methods

We recruited 1151 participants with an average age of 74.8 years by convenience sampling from a primary healthcare setting to carry out the cross-sectional study over 6 months. The whole dataset was divided into two sets, namely training set and test set, which respectively occupied 70% and 30% of the whole dataset. The training dataset was used first; decision tree analysis was used to identify possible stratifying variables that could help to generate separate decision models.

Results

The number of fallers was 230 with 20% 1-year prevalence. There were significant differences in gender, use of walking aids, presence of chronic diseases, and co-morbidities including osteoporosis, depression, and previous upper limb fractures, and performance in the Timed Up and Go test and the Functional Reach test among the baselines between the faller and non-faller groups. Three decision tree models for the dependent dichotomous variables (fallers, indoor fallers, and outdoor fallers) were generated, with overall accuracy rates of the models of 77.40%, 89.44% and 85.76%, respectively. Timed Up and Go, Functional Reach, body mass index, high blood pressure, osteoporosis, and number of drugs taken were identified as stratifying variables in the decision tree models for fall screening.

Conclusion

The use of decision tree analysis for clinical algorithms for accidental falls in community-dwelling older people creates patterns for decision-making in fall screening, which also paves the way for utility-based decision-making using supervised machine learning in fall risk detection.

Predictive Model of Functional Exercise Compliance of Patients with Breast Cancer Based on Decision Tree

OBJECTIVE

Regular functional exercise can help recover the functions of upper limb for patients with breast cancer. By finding the influencing factors of functional exercise compliance and constructing a predictive model, patients with a poor functional exercise compliance can be better identified. This study aims to find out the factors influencing the functional exercise compliance of patients with breast cancer and build a predictive model based on decision tree.

METHODS

Convenience sampling was used at two tertiary hospitals in Shantou from August 2020 to March 2021. Data of patients with breast cancer patient was obtained from questionnaires and based on demographics, Constant-Murley Score, Functional Exercise Compliance Scale for Postoperative Breast Cancer Patients, Champion Health Belief Model Scale, Social Support Rating Scale, Disease Perception Questionnaire and Family Care Index Questionnaire. Possible influencing factors of functional exercise compliance were analyzed using correlation analysis as well as univariate and binary logistic regression analysis through SPSS v25, and a CHAID decision tree was used to construct a predictive model on training, validation and test sets via SPSS Modeler v18 at a ratio of 6:2:2. Prediction accuracy, sensitivity, specificity and AUC were used to analyze the efficacy of the predictive model.

RESULTS

A total of 227 valid samples were collected, of which 145 were assessed with a poor compliance (63.9%). According to a logistic regression analysis, perceived benefits, time after surgery and self-efficacy were influencing factors. The prediction accuracy, sensitivity, specificity and AUC of the predictive model, based on a CHAID decision tree analysis, were 70.73%, 57.1%, 77.8% and 0.81 respectively.

CONCLUSION

The predictive model, based on a CHAID decision tree analysis, had a moderate predictive efficacy, which could be used as a clinical auxiliary tool for clinical nurses to predict patients' functional exercise compliance.

Development of the ADFICE_IT models for predicting falls and recurrent falls in community-dwelling older adults: pooled analyses of European cohorts with special attention to medication

Background

Use of fall prevention strategies requires detection of high-risk patients. Our goal was to develop prediction models for falls and recurrent falls in community-dwelling older adults and to improve upon previous models by using a large, pooled sample and by considering a wide range of candidate predictors, including medications.

Methods

Harmonized data from 2 Dutch (LASA, B-PROOF) and 1 German cohort (ActiFE Ulm) of adults aged ≥65 years were used to fit 2 logistic regression models: one for predicting any fall and another for predicting recurrent falls over 1 year. Model generalizability was assessed using internal–external cross-validation.

Results

Data of 5 722 participants were included in the analyses, of whom 1 868 (34.7%) endured at least 1 fall and 702 (13.8%) endured a recurrent fall. Positive predictors for any fall were: educational status, depression, verbal fluency, functional limitations, falls history, and use of antiepileptics and drugs for urinary frequency and incontinence; negative predictors were: body mass index (BMI), grip strength, systolic blood pressure, and smoking. Positive predictors for recurrent falls were: educational status, visual impairment, functional limitations, urinary incontinence, falls history, and use of anti-Parkinson drugs, antihistamines, and drugs for urinary frequency and incontinence; BMI was a negative predictor. The average C-statistic value was 0.65 for the model for any fall and 0.70 for the model for recurrent falls.

Conclusion

Compared with previous models, the model for recurrent falls performed favorably while the model for any fall performed similarly. Validation and optimization of the models in other populations are warranted.

The use of predictive fall models for older adults receiving aged care, using routinely collected electronic health record data: a systematic review

BACKGROUND

Falls in older adults remain a pressing health concern. With advancements in data analytics and increasing uptake of electronic health records, developing comprehensive predictive models for fall risk is now possible. We aimed to systematically identify studies involving the development and implementation of predictive falls models which used routinely collected electronic health record data in home-based, community and residential aged care settings.

METHODS

A systematic search of entries in Cochrane Library, CINAHL, MEDLINE, Scopus, and Web of Science was conducted in July 2020 using search terms relevant to aged care, prediction, and falls. Selection criteria included English-language studies, published in peer-reviewed journals, had an outcome of falls, and involved fall risk modelling using routinely collected electronic health record data. Screening, data extraction and quality appraisal using the Critical Appraisal Skills Program for Clinical Prediction Rule Studies were conducted. Study content was synthesised and reported narratively.

RESULTS

From 7,329 unique entries, four relevant studies were identified. All predictive models were built using different statistical techniques. Predictors across seven categories were used: demographics, assessments of care, fall history, medication use, health conditions, physical abilities, and environmental factors. Only one of the four studies had been validated externally. Three studies reported on the performance of the models.

CONCLUSIONS

Adopting predictive modelling in aged care services for adverse events, such as falls, is in its infancy. The increased availability of electronic health record data and the potential of predictive modelling to document fall risk and inform appropriate interventions is making use of such models achievable. Having a dynamic prediction model that reflects the changing status of an aged care client is key to this moving forward for fall prevention interventions.

Low Light Exposure and Physical Activity in Older Adults With and Without Age-Related Macular Degeneration

Purpose

To investigate the extent of low light exposure and associated physical activity in older adults with and without age-related macular degeneration (AMD).

Methods

Light exposure (lux) and physical activity (counts per minute, CPM) were measured in 28 older adults (14 bilateral AMD and 14 normally sighted controls) using a wrist-worn actigraphy device (Actiwatch) for 7 days and nights. Exposure to low light levels (≤10 lux) and physical activity during waking hours were determined, as well as number of brief active periods during sleeping hours (e.g., going to the bathroom). Assessments included visual acuity and the Low Luminance Questionnaire (LLQ).

Results

No significant differences were found in low light exposure (39 ± 14% vs. 34 ± 10%) or physical activity (200 ± 82 CPM vs. 226 ± 55 CPM) during waking hours between the AMD and control group. However, the AMD group had more brief active periods during sleeping hours than controls (1.8 ± 1.3 vs. 1.1 ± 0.4; P = 0.007). Reduced physical activity under low light levels was significantly associated with lower LLQ scores (P = 0.012).

Conclusions

Exposure to low light levels and associated physical activity were similar in older adults with and without AMD. This has important implications for older adults with AMD, given the impact of low light levels on visual function and mobility, suggesting the need for including lighting advice in rehabilitation programs for this population.

Translational Relevance

Older adults with and without AMD spend over a third of waking hours under low light levels, which are an environmental falls hazard. Findings suggest the need for interventions to improve lighting levels for older adults.

Development and internal validation of a risk prediction model for falls among older people using primary care electronic health records

Background

Currently used prediction tools have limited ability to identify community-dwelling older people at high risk for falls. Prediction models utilizing electronic health records (EHRs) provide opportunities but up to now showed limited clinical value as risk stratification tool, because of among others the underestimation of falls prevalence. The aim of this study was to develop a fall prediction model for community-dwelling older people using a combination of structured data and free text of primary care EHRs and to internally validate its predictive performance.

Methods

We used EHR data of individuals aged 65 or older. Age, sex, history of falls, medications, and medical conditions were included as potential predictors. Falls were ascertained from the free text. We employed the Bootstrap-enhanced penalized logistic regression with the least absolute shrinkage and selection operator to develop the prediction model. We used 10-fold cross-validation to internally validate the prediction strategy. Model performance was assessed in terms of discrimination and calibration.

Results

Data of 36 470 eligible participants were extracted from the data set. The number of participants who fell at least once was 4 778 (13.1%). The final prediction model included age, sex, history of falls, 2 medications, and 5 medical conditions. The model had a median area under the receiver operating curve of 0.705 (interquartile range 0.700–0.714).

Conclusions

Our prediction model to identify older people at high risk for falls achieved fair discrimination and had reasonable calibration. It can be applied in clinical practice as it relies on routinely collected variables and does not require mobility assessment tests.

Early Recognition of the Preference for Exclusive Breastfeeding in Current China: A Prediction Model based on Decision Trees

Exclusive breastfeeding (EBF) is affected by multiple risk factors. Therefore, it is difficult for clinical professionals to identify women who will not practice EBF well and provide subsequent medical suggestions and treatments. This study aimed to apply a decision tree (DT) model to predict EBF at two months postpartum. The socio-demographic, clinical and breastfeeding parameters of 1,141 breastfeeding women from Nanjing were evaluated. Decision tree modelling was used to analyse and screen EBF factors and establish a risk assessment model of EBF. The Chinese version of the Breastfeeding Self-Efficacy Scale (CV-BSES) score, early formula supplementation, abnormal nipples, mastitis, neonatal jaundice, cracked or sore nipples and intended duration of breastfeeding were significant risk factors associated with EBF in the DT model. The accuracy, sensitivity and specificity of the DT model were 73.1%, 75.5% and 66.3%, respectively. The DT model showed similar or better performance than the logistic regression model in assessing the risk of early cessation of EBF before two months postpartum. The DT model has potential for application in clinical practice and identifies high-risk subpopulations that need specific prevention.

The association between chronic kidney disease, falls, and fractures: a systematic review and meta-analysis

Patients with chronic kidney disease (CKD) are more likely to experience falls and fractures due to renal osteodystrophy and the high prevalence of risk factors for falls. However, it is not well established how great the risk is for falls and fractures for the different stages of CKD compared to the general population. The objective of this systematic review and meta-analysis was to assess whether, and in which degree, CKD was associated with falls and fractures in adults. A systematic search in PubMed, Embase, CINAHL, and The Cochrane Library was performed on 7 September 2018. All retrospective, cross-sectional, and longitudinal studies of adults (18 years of older) that studied the association between CKD, fractures, and falls were included. Additional studies were identified by cross-referencing. A total of 39 publications were included, of which two publications assessed three types of outcome and four publications assessed two types of outcome. Ten studies focused on accidental falling; seventeen studies focused on hip, femur, and pelvis fractures; seven studies focused on vertebral fractures; and thirteen studies focused on any type of fracture without further specification. Generally, the risk of fractures increased when kidney function worsened, with the highest risks in the patients with stage 5 CKD or dialysis. This effect was most pronounced for hip fractures and any type of fractures. Furthermore, results on the association between CKD and accidental falling were contradictory. Compared to the general population, fractures are highly prevalent in patients with CKD. Besides more awareness of timely fracture risk assessment, there also should be more focus on fall prevention.

Association Between Chronic or Acute Use of Antihypertensive Class of Medications and Falls in Older Adults. A Systematic Review and Meta-Analysis

BACKGROUND

Evaluating effect of acute or chronic use of antihypertensives on risk of falls in older adults.

METHODS

Data sources: Systematic search of primary research articles in CINAHL, Cochrane, EBM, EMBASE, and MEDLINE databases from January 1 2007 to June 1 2017. Study selection: Research studies of cohort, case-control, case-crossover, cross-sectional, or randomized controlled trial (RCT) design examining association between antihypertensives and falls in people older than 60 years were evaluated. Data synthesis: Twenty-nine studies (N = 1,234,667 participants) were included. Study quality was assessed using the Newcastle-Ottawa Scale (NOS). PRISMA and MOOSE guidelines were used for abstracting data and random-effects inverse-variance meta-analysis was conducted on 26 articles examining chronic antihypertensive use, with odds ratios (ORs) and hazards ratios (HRs) analyzed separately. Time-risk analysis was performed on 5 articles examining acute use of antihypertensives. Outcomes: Pooled ORs and HRs were calculated to determine the association between chronic antihypertensive use and falls. For time-risk analysis, OR was plotted with respect to number of days since antihypertensive commencement, change, or dose increase.

RESULTS

There was no significant association between risk of falling and chronic antihypertensive medication use (OR = 0.97, 95% confidence interval [CI] 0.93-1.01, I2 = 64.1%, P = 0.000; and HR = 0.96, 95% CI 0.92-1.00, I2 = 0.0%, P = 0.706). The time-risk analysis demonstrated a significantly elevated risk of falling 0-24 hours after antihypertensive initiation, change, or dose increase. When diuretics were used, the risk remained significantly elevated till day 21.

CONCLUSIONS

There is no significant association between chronic use of antihypertensives and falls in older adults. Risk of falls is highest on day zero for all antihypertensive medications.

Using Patient Flow Information to Determine Risk of Hospital Presentation: Protocol for a Proof-of-Concept Study

Background

Every day, patients are admitted to the hospital with conditions that could have been effectively managed in the primary care sector. These admissions are expensive and in many cases are possible to avoid if early intervention occurs. General practitioners are in the best position to identify those at risk of imminent hospital presentation and admission; however, it is not always possible for all the factors to be considered. A lack of shared information contributes significantly to the challenge of understanding a patient’s full medical history. Some health care systems around the world use algorithms to analyze patient data in order to predict events such as emergency presentation; however, those responsible for the design and use of such systems readily admit that the algorithms can only be used to assess the populations used to design the algorithm in the first place. The United Kingdom health care system has contributed data toward algorithm development, which is possible through the unified health care system in place there. The lack of unified patient records in Australia has made building an algorithm for local use a significant challenge.

Objective

Our objective is to use linked patient records to track patient flow through primary and secondary health care in order to develop a tool that can be applied in real time at the general practice level. This algorithm will allow the generation of reports for general practitioners that indicate the relative risk of patients presenting to an emergency department.

Methods

A previously designed tool was used to deidentify the general practice and hospital records of approximately 100,000 patients. Records were pooled for patients who had attended emergency departments within the Eastern Health Network of hospitals and general practices within the Eastern Health Network catchment. The next phase will involve development of a model using a predictive analytic machine learning algorithm. The model will be developed iteratively, testing the combination of variables that will provide the best predictive model.

Results

Records of approximately 97,000 patients who have attended both a general practice and an emergency department have been identified within the database. These records are currently being used to develop the predictive model.

Conclusions

Records from general practice and emergency department visits have been identified and pooled for development of the algorithm. The next phase in the project will see validation and live testing of the algorithm in a practice setting. The algorithm will underpin a clinical decision support tool for general practitioners which will be tested for face validity in this initial study into its efficacy.

Continuous quality improvement in nephrology: a systematic review

BACKGROUND

Continuous quality improvement (CQI) has been successfully applied in business and engineering for over 60 years. While using CQI techniques within nephrology has received increased attention, little is known about where, and with what measure of success, CQI can be attributed to improving outcomes within nephrology care. This is particularly important as payors' focus on value-based healthcare and reimbursement is tied to achieving quality improvement thresholds. We conducted a systematic review of CQI applications in nephrology.

METHODS

Studies were identified from PubMed, MEDLINE, Scopus, Web of Science, CINAHL, Google Scholar, ProQuest Dissertation Abstracts and sources of grey literature (i.e., available in print/electronic format but not controlled by commercial publishers) between January 1, 2004 and October 13, 2014. We developed a systematic evaluation protocol and pre-defined criteria for review. All citations were reviewed by two reviewers with disagreements resolved by consensus.

RESULTS

We initially identified 468 publications; 40 were excluded as duplicates or not available/not in English. An additional 352 did not meet criteria for full review due to: 1. Not meeting criteria for inclusion = 196 (e.g., reviews, news articles, editorials) 2. Not nephrology-specific = 153, 3. Only available as abstracts = 3. Of 76 publications meeting criteria for full review, the majority [45 (61%)] focused on ESRD care. 74% explicitly stated use of specific CQI tools in their methods. The highest number of publications in a given year occurred in 2011 with 12 (16%) articles. 89% of studies were found in biomedical and allied health journals and most studies were performed in North America (52%). Only one was randomized and controlled although not blinded.

CONCLUSIONS

Despite calls for healthcare reform and funding to inspire innovative research, we found few high quality studies either rigorously evaluating the use of CQI in nephrology or reporting best practices. More rigorous research is needed to assess the mechanisms and attributes by which CQI impacts outcomes before there is further promotion of its use for improvement and reimbursement purposes.

Standing Balance and Spatiotemporal Aspects of Gait Are Impaired Upon Nocturnal Awakening in Healthy Late Middle-Aged and Older Adults

STUDY OBJECTIVES

Nocturnal awakenings may constitute a unique risk for falls among older adults. We describe differences in gait and balance between presleep and midsleep testing, and whether changes in the lighting environment during the midsleep testing further affect gait and balance.

METHODS

Twenty-one healthy, late middle-aged and older (64.7 ± 8.0 y) adults participated in this repeated-measures design consisting of four overnight laboratory stays. Each night, participants completed baseline visual acuity, gait, and balance testing. After a 2-h sleep opportunity, they were awakened for 13 min into one of four lighting conditions: very dim white light (< 0.5 lux); dim white light (∼28.0 lux); dim orange light (∼28.0 lux); and white room-level light (∼200 lux). During this awakening, participants completed the same sequence of testing as at baseline.

RESULTS

Low-contrast visual acuity significantly decreased with decreasing illuminance conditions (F(3,45) = 98.26, p < 0.001). Our a priori hypothesis was confirmed in that variation in stride velocity and center of pressure path length were significantly worse during the mid-sleep awakening compared to presleep baseline. Lighting conditions during the awakening, however, did not influence these parameters. In exploratory analyses, we found that over one-third of the tested gait and balance parameters were significantly worse at the midsleep awakening as compared to baseline (p < 0.05), and nearly one-quarter had medium to large effect sizes (Cohen d ≥ 0.5; r ≥ 0.3).

CONCLUSIONS

Balance and gait are impaired during midsleep awakenings among healthy, late middle-aged and older adults. This impairment is not ameliorated by exposure to room lighting, when compared to dim lights.

Predicting Falls and When to Intervene in Older People: A Multilevel Logistical Regression Model and Cost Analysis

Introduction

Falls are the leading cause of injury in older people. Reducing falls could reduce financial pressures on health services. We carried out this research to develop a falls risk model, using routine primary care and hospital data to identify those at risk of falls, and apply a cost analysis to enable commissioners of health services to identify those in whom savings can be made through referral to a falls prevention service.

Methods

Multilevel logistical regression was performed on routinely collected general practice and hospital data from 74751 over 65’s, to produce a risk model for falls. Validation measures were carried out. A cost-analysis was performed to identify at which level of risk it would be cost-effective to refer patients to a falls prevention service. 95% confidence intervals were calculated using a Monte Carlo Model (MCM), allowing us to adjust for uncertainty in the estimates of these variables.

Results

A risk model for falls was produced with an area under the curve of the receiver operating characteristics curve of 0.87. The risk cut-off with the highest combination of sensitivity and specificity was at p = 0.07 (sensitivity of 81% and specificity of 78%). The risk cut-off at which savings outweigh costs was p = 0.27 and the risk cut-off with the maximum savings was p = 0.53, which would result in referral of 1.8% and 0.45% of the over 65’s population respectively. Above a risk cut-off of p = 0.27, costs do not exceed savings.

Conclusions

This model is the best performing falls predictive tool developed to date; it has been developed on a large UK city population; can be readily run from routine data; and can be implemented in a way that optimises the use of health service resources. Commissioners of health services should use this model to flag and refer patients at risk to their falls service and save resources.

Using entropy measures to characterize human locomotion

Entropy measures have been widely used to quantify the complexity of theoretical and experimental dynamical systems. In this paper, the value of using entropy measures to characterize human locomotion is demonstrated based on their construct validity, predictive validity in a simple model of human walking and convergent validity in an experimental study. Results show that four of the five considered entropy measures increase meaningfully with the increased probability of falling in a simple passive bipedal walker model. The same four entropy measures also experienced statistically significant increases in response to increasing age and gait impairment caused by cognitive interference in an experimental study. Of the considered entropy measures, the proposed quantized dynamical entropy (QDE) and quantization-based approximation of sample entropy (QASE) offered the best combination of sensitivity to changes in gait dynamics and computational efficiency. Based on these results, entropy appears to be a viable candidate for assessing the stability of human locomotion.