Age effect on the prediction of risk of prolonged length hospital stay in older patients visiting the emergency department: results from a large prospective geriatric cohort study

A comprehensive review of machine learning algorithms and their application in geriatric medicine: present and future

The increasing access to health data worldwide is driving a resurgence in machine learning research, including data-hungry deep learning algorithms. More computationally efficient algorithms now offer unique opportunities to enhance diagnosis, risk stratification, and individualised approaches to patient management. Such opportunities are particularly relevant for the management of older patients, a group that is characterised by complex multimorbidity patterns and significant interindividual variability in homeostatic capacity, organ function, and response to treatment. Clinical tools that utilise machine learning algorithms to determine the optimal choice of treatment are slowly gaining the necessary approval from governing bodies and being implemented into healthcare, with significant implications for virtually all medical disciplines during the next phase of digital medicine. Beyond obtaining regulatory approval, a crucial element in implementing these tools is the trust and support of the people that use them. In this context, an increased understanding by clinicians of artificial intelligence and machine learning algorithms provides an appreciation of the possible benefits, risks, and uncertainties, and improves the chances for successful adoption. This review provides a broad taxonomy of machine learning algorithms, followed by a more detailed description of each algorithm class, their purpose and capabilities, and examples of their applications, particularly in geriatric medicine. Additional focus is given on the clinical implications and challenges involved in relying on devices with reduced interpretability and the progress made in counteracting the latter via the development of explainable machine learning.

Older patients referred for geriatric consultation in the emergency department: characteristics and healthcare utilization

BACKGROUND

Comprehensive geriatric assessment (CGA) is difficult to perform in the emergency department (ED) environment and performance of screening tools in identifying vulnerable older ED patients who are best candidates for a geriatric consultation remain questionable.

AIM

To determine the characteristics of older patients referred for a geriatric consultation by ED staff and to investigate these patients' subsequent healthcare utilization.

METHODS

Secondary analysis of data previously collected for a prospective observational study of patients aged 75 + years visiting the ED of an academic hospital in Switzerland over four months (Michalski-Monnerat et al., J Am Geriatr Soc 68(12):2914-20, 2020). Socio-demographic, health, functional (basic activities of daily living; BADL), cognitive, and affective status data were collected at admission by a research nurse using a standardized brief geriatric assessment. Information on geriatric consultations, hospitalization, discharge destination, and 30-day readmission were retrieved from hospital database. Bivariable and multivariable analyses were performed using this data set collected previously.

RESULTS

Thirty-two (15.8%) of the 202 enrolled patients were referred for a geriatric consultation. Compared to the others, they were older (84.9 ± 5.4 vs 82.9 ± 5.4 years, p = .03), more impaired in BADL (4.8 ± 1.6 vs 5.5 ± 1.0, p = .01), with more comorbid conditions (5.3 ± 1.5 vs 4.5 ± 1.9, p = .03), more frequently admitted after a fall (43.7% vs 19.4%, p = .01), and hospitalized over the previous 6-month period (53.1% vs 30.6%, p = .02). Multivariable analyses that adjusted for variables significantly associated with outcomes in bivariable analysis found that being admitted after a fall (AdjOR 4.0, 95%CI 1.7-9.4, p < .01) and previously hospitalized (AdjOR 2.7, 95% CI 1.2-6.2, p = .02) remained associated with increased odds of consultation, whereas the inverse association with BADL performance remained (AdjOR 0.7, 95%CI 0.5-0.9, p = .01). Patients referred for geriatric consultation had higher odds of hospitalization (84.4% vs 49.4%; AdjOR 5.9, 95%CI 2.1-16.8, p < .01), but similar odds of home discharge when admitted, and of 30-day readmission.

CONCLUSION

About one in six older ED patients were referred for a geriatric consultation who appeared to be those most vulnerable, as suggested by their increased hospitalization rate. Alternative strategies are needed to enhance access to geriatric consultation in the ED.

Machine learning-based prediction of hospital prolonged length of stay admission at emergency department: a Gradient Boosting algorithm analysis

Objective

This study aims to develop and compare different models to predict the Length of Stay (LoS) and the Prolonged Length of Stay (PLoS) of inpatients admitted through the emergency department (ED) in general patient settings. This aim is not only to promote any specific model but rather to suggest a decision-supporting tool (i.e., a prediction framework).

Methods

We analyzed a dataset of patients admitted through the ED to the "Sant"Orsola Malpighi University Hospital of Bologna, Italy, between January 1 and October 26, 2022. PLoS was defined as any hospitalization with LoS longer than 6 days. We deployed six classification algorithms for predicting PLoS: Random Forest (RF), Support Vector Machines (SVM), Gradient Boosting (GB), AdaBoost, K-Nearest Neighbors (KNN), and logistic regression (LoR). We evaluated the performance of these models with the Brier score, the area under the ROC curve (AUC), accuracy, sensitivity (recall), specificity, precision, and F1-score. We further developed eight regression models for LoS prediction: Linear Regression (LR), including the penalized linear models Least Absolute Shrinkage and Selection Operator (LASSO), Ridge and Elastic-net regression, Support vector regression, RF regression, KNN, and eXtreme Gradient Boosting (XGBoost) regression. The model performances were measured by their mean square error, mean absolute error, and mean relative error. The dataset was randomly split into a training set (70%) and a validation set (30%).

Results

A total of 12,858 eligible patients were included in our study, of whom 60.88% had a PloS. The GB classifier best predicted PloS (accuracy 75%, AUC 75.4%, Brier score 0.181), followed by LoR classifier (accuracy 75%, AUC 75.2%, Brier score 0.182). These models also showed to be adequately calibrated. Ridge and XGBoost regressions best predicted LoS, with the smallest total prediction error. The overall prediction error is between 6 and 7 days, meaning there is a 6-7 day mean difference between actual and predicted LoS.

Conclusion

Our results demonstrate the potential of machine learning-based methods to predict LoS and provide valuable insights into the risks behind prolonged hospitalizations. In addition to physicians' clinical expertise, the results of these models can be utilized as input to make informed decisions, such as predicting hospitalizations and enhancing the overall performance of a public healthcare system.

Association Between Preoperative Blood Glucose Level and Hospital Length of Stay in Patients With Kidney Stones Undergoing Percutaneous Nephrolithotomy

Aim

To assess the effect of preoperative blood glucose (POBG) levels on the length of stay (LOS) in patients with kidney stones undergoing percutaneous nephrolithotomy (PCNL).

Methods

We conducted a retrospective study of patients who underwent PCNL at the Zhongda Hospital of Southeast University from 2013 to 2019. The relationship between POBG level and LOS was investigated by dose-response analysis curves of restricted cubic spline function.

Results

We included 310 patients and divided them into three groups (&lt;5.04, 5.04 to &lt;6.88, ≥6.88 mmol/L) according to the POBG levels. Patients with POBG levels ≥6.88 mmol/L (adjusted odds risk [aOR] 1.67; 95% CI 0.83–3.33) had a 67% higher risk of LOS &gt; 2 weeks than patients with POBG levels &lt;5.04 mmol/L. A positive dose-response analysis curve was observed between POBG and the adjusted risk of LOS &gt;2 weeks. Similar results were observed in the subgroups analysis.

Conclusion

We demonstrated that higher POBG levels were significantly associated with longer LOS in patients with kidney stones undergoing PCNL.

The association between length of stay in the emergency department and short-term mortality

The detrimental effects of increased length of stay at the emergency department (ED-LOS) for patient outcome have been sparsely studied in the Swedish setting. Our aim was to further explore the association between ED-LOS and short-term mortality in patients admitted to the EDs of two large University hospitals in Sweden. All adult patients (> 18 years) visiting the ED at the Karolinska University Hospital, Sweden, from 1/1/2010 to 1/1/2015 (n = 639,385) were retrospectively included. Logistic regression analysis was used to determine association between ED-LOS and 7- and 30-day mortality rates. All patients were triaged according to the RETTS-A into different levels of medical urgency and subsequently separated into five quintiles of ED-LOS. Mortality rate was highest in highest triage priority level (7-day mortality 5.24%, and 30-day mortality 9.44%), and decreased by lower triage priority group. For patients with triage priority levels 2-4, prolonged ED-LOS was associated with increased mortality, especially for lowest priority level, OR for priority level 4 and highest quintile of ED-LOS 30-day mortality 1.49 (CI 95% 1.20-1.85). For patients with highest triage priority level the opposite was at hand, with decreasing mortality risk with increasing quintile of ED-LOS for 7-day mortality, and lower mortality for the two highest quintile of ED-LOS for 30-day mortality. In patients not admitted to in-hospital care higher ED-LOS was associated with higher mortality. Our data suggest that increased ED-LOS could be associated with slightly increased short-term mortality in patients with lower clinical urgency and dismissed from the ED.

Machine learning in patient flow: a review

This work is a review of the ways in which machine learning has been used in order to plan, improve or aid the problem of moving patients through healthcare services. We decompose the patient flow problem into four subcategories: prediction of demand on a healthcare institution, prediction of the demand and resource required to transfer patients from the emergency department to the hospital, prediction of potential resource required for the treatment and movement of inpatients and prediction of length-of-stay and discharge timing. We argue that there are benefits to both approaches of considering the healthcare institution as a whole as well as the patient by patient case and that ideally a combination of these would be best for improving patient flow through hospitals. We also argue that it is essential for there to be a shared dataset that will allow researchers to benchmark their algorithms on and thereby allow future researchers to build on that which has already been done. We conclude that machine learning for the improvement of patient flow is still a young field with very few papers tailor-making machine learning methods for the problem being considered. Future works should consider the need to transfer algorithms trained on a dataset to multiple hospitals and allowing for dynamic algorithms which will allow real-time decision-making to help clinical staff on the shop floor.

Assessing Older Patients' Vulnerability in the Emergency Department: A Study of InterRAI ED Screener Accuracy

BACKGROUND

Identifying vulnerable older patients admitted to an emergency department (ED) who are at increased risk for adverse events and require a comprehensive geriatric assessment remains a major challenge. The interRAI Emergency Department Screener (EDS) was developed for this specific purpose, but data regarding its validity are scarce.

OBJECTIVES

To determine (1) convergent validity of the EDS with results of a geriatrician's assessment in defining the need for prompt versus delayed/no further geriatric assessment and (2) predictive validity of the EDS for hospital admission, prolonged hospital length of stay (LOS), and 30-day readmission.

DESIGN

Prospective observational study.

SETTING

ED of an academic hospital in Switzerland.

PARTICIPANTS

Older patients, aged 75 years or older (N = 202), who visited the ED over a 4-month period. Patients with life-threatening conditions were excluded.

MEASUREMENTS

Data for EDS were collected by two clinical nurses. A brief geriatric assessment was performed separately and interpreted by a geriatrician blinded to the EDS results. Orientation after ED discharge, hospital LOS, and 30-day readmission were retrieved from the administrative database.

RESULTS

Participants were aged 83.2 ± 5.4 years, 56.9% were female, and 43.6% lived alone. Frequent findings at geriatric assessment were impairment in gait/balance (69.3%), polypharmacy (64.9%), cognitive impairment/delirium (48.2%), risk of malnutrition (46.0%), and mood impairment (38.1%). The proportions of participants who required prompt, delayed, and no further geriatric assessment, according to the EDS, were 27.2%, 29.2%, and 43.6%, respectively. The EDS had low sensitivity in predicting hospital admission (28.8%), prolonged LOS (26.3%), and 30-day readmission (26.1%), with the Area Under the Receiver Operating Characteristics (AUROC) being 51.8%, 48.1%, and 49.4%, respectively.

CONCLUSION

The EDS performed poorly in both convergent and predictive validity analyses, precluding its use as a screening tool in this ED environment. Further efforts should be undertaken to better target interventions to reduce adverse health trajectories in the older ED population.

Perceived risk of institutionalization, hospitalization, and death in oldest old primary care patients

OBJECTIVES

This study aims to analyze the accuracy and predictive ability of the Risk Instrument for Screening in the Community (RISC) scored by general practitioners (GPs) in a sample of primary care patients aged 80+ with perceived mental health concerns.

METHOD

GPs ranked the perceived risk of the three adverse outcomes (hospitalization, institutionalization and death) at 1 year in a five Likert scale (RISC score), where 1 is the lowest risk and 5 is the highest. Follow up contacts were conducted after 1 year of assessment in order to collect data on the three outcomes.

RESULTS

The 1-year proportion of institutionalization, hospitalization and death were 12.1 %, 25.2 % and 19.0 % respectively. Based upon the sensitivity and specificity from the Receiver Operating Characteristic (ROC) curves, we found an optimal cut-off point of ≥4 for the RISC. The RISC had fair accuracy for 1-year risk of institutionalization (Area Under the ROC curve (AUC) = 0.75, 95% CI 0.43-0.68) and hospitalization (AUC = 0.65, 95% CI 0.52-0.78), but not for death (AUC = 0.55, 95% CI 0.43-0.68).

CONCLUSIONS

The RISC as a short global subjective assessment is to be considered a reliable tool for use by GPs. Our results showed that RISC seems to be a good instrument to triage very old people at risk for institutionalization but with poor accuracy at predicting hospitalization and limited predictive ability for death, suggesting further research and caution on this instrument's use.