The added value of health indicators to mortality predictions in old age: A systematic review

Machine learning models of healthcare expenditures predicting mortality: A cohort study of spousal bereaved Danish individuals

BACKGROUND

The ability to accurately predict survival in older adults is crucial as it guides clinical decision making. The added value of using health care usage for predicting mortality remains unexplored. The aim of this study was to investigate if temporal patterns of healthcare expenditures, can improve the predictive performance for mortality, in spousal bereaved older adults, next to other widely used sociodemographic variables.

METHODS

This is a population-based cohort study of 48,944 Danish citizens 65 years of age and older suffering bereavement within 2013-2016. Individuals were followed from date of spousal loss until death from all causes or 31st of December 2016, whichever came first. Healthcare expenditures were available on weekly basis for each person during the follow-up and used as predictors for mortality risk in Extreme Gradient Boosting models. The extent to which medical spending trajectories improved mortality predictions compared to models with sociodemographics, was assessed with respect to discrimination (AUC), overall prediction error (Brier score), calibration, and clinical benefit (decision curve analysis).

RESULTS

The AUC of age and sex for mortality the year after spousal loss was 70.8% [95% CI 68.8, 72.8]. The addition of sociodemographic variables led to an increase of AUC ranging from 0.9% to 3.1% but did not significantly reduce the overall prediction error. The AUC of the model combining the variables above plus medical spending usage was 80.8% [79.3, 82.4] also exhibiting smaller Brier score and better calibration. Overall, patterns of healthcare expenditures improved mortality predictions the most, also exhibiting the highest clinical benefit among the rest of the models.

CONCLUSION

Temporal patterns of medical spending have the potential to significantly improve our assessment on who is at high risk of dying after suffering spousal loss. The proposed methodology can assist in a more efficient risk profiling and prognosis of bereaved individuals.

Predicting mortality and visualizing health care spending by predicted mortality in Danes over age 65

Health care expenditure in the last year of life makes up a high proportion of medical spending across the world. This is often framed as waste, but this framing is only meaningful if it is known at the time of treatment who will go on to die. We analyze the distribution of health care spending by predicted mortality for the Danish population over age 65 over the year 2016, with one-year mortality predicted by a machine learning model based on sociodemographics and use of health care services for the two years before entry into follow-up. While a reasonably good model can be built, extremely few individuals have high ex-ante probability of dying, and those with a predicted mortality of more than 50% account for only 2.8% of total health care expenditure. Decedents outspent survivors by a factor of more than ten, but compared to survivors with similar predicted mortality they spent only 2.5 times as much. Our results suggest that while spending in the last year of life is indeed high, this is nearly all spent in situations where there is a reasonable expectation that the patient can survive.

Resting respiration rate predicts all-cause mortality in older outpatients

BACKGROUND

Although respiratory rate has been a sensitive predictor for prognosis in acute settings, resting respiratory rate (RRR) is undervalued in chronic care. The clinical significance of RRR among older people is not well documented.

AIM

We investigated the association between RRR and all-cause mortality among older outpatients.

METHODS

A retrospective cohort study exhaustively included patients who had undergone medical checkups in a facility between April 2017 and March 2018 and followed up for at least 2 years. We excluded patients who were less than 60 years of age or had not undergone regular outpatient appointments. Sex, age, smoking habits, history of hospitalization, polypharmacy, long-term care insurance certification status, Mazzaglia index, pulse rate, systolic blood pressure, and Charlson Comorbidity Index were measured at the baseline medical checkup. Survival was confirmed by chart review and by contacting physicians in charge. The risk ratios were estimated by converting the odds ratios derived from the multivariable logistic regression models.

RESULTS

Of the 853 patients who underwent baseline checkups, 749 were enrolled in the analyses; death occurred in 53 patients (7.1%), with no loss to follow-up. The RRR was independently associated with all-cause mortality after adjusting for covariates [adjusted risk ratio of RRR per 1 bpm = 1.14, 95% confidence interval (CI): 1.06 - 1.22].

DISCUSSION

Given the independent association of RRR for existing predictors, this simple index seems worthy of consideration in further studies aimed at defining its predictive role in older people and in different settings.

CONCLUSION

RRR was independently associated with all-cause mortality.

Dynamical indicators in time series of healthcare expenditures predict mortality risk of older adults following spousal bereavement

BACKGROUND

The process of aging renders older people susceptible for adverse outcomes upon stress. Various indicators derived from complex systems theory have been proposed for quantifying resilience in living organisms, including humans. We investigated the ability of system-based indicators in capturing the dynamics of resilience in humans who suffer the adversity of spousal bereavement and tested their predictive power in mortality as a finite health transition.

METHODS

Using longitudinal register data on weekly healthcare consumption of all Danish citizens over the age of 65 from January 1st, 2011, throughout December 31st, 2016, we performed statistical comparisons of the indicators 'average', 'slope', 'mean squared error', and 'lag-1 autocorrelation' one year before and after spousal bereavement, stratified for age and sex. The relation between levels of these indicators before bereavement and mortality hazards thereafter was determined by time to event analysis. We assessed the added value for mortality prediction via the time dependent area (AUC) under the receiver operating characteristic curve.

RESULTS

The study included 934,003 citizens of whom 51,890 experienced spousal bereavement and 2862 died in the first year thereafter. Healthcare consumption is increased, more volatile and accelerating with aging and in men compared to women (all p-values < 0.001). All dynamic indicators before bereavement were positively related with mortality hazards thereafter (all p-values < 0.001). The average discriminative performance for the 1-year mortality risk of the model with only age as a predictor (AUC: 68.9% and 70.2%) was significantly increased with the addition of dynamical indicators (78.5% and 82.4%) for males and females, respectively.

CONCLUSIONS

Dynamic indicators in time series of health care expenditures are strong predictors of mortality risk and could be part of predictive models for prognosis after life stressors, such as bereavement.

Do changes in frailty, physical functioning, and cognitive functioning predict mortality in old age? Results from the Longitudinal Aging Study Amsterdam

Background

The ability to accurately predict survival in older adults is crucial as it guides clinical decision making. The added value of using various health indicators as well as changes in these indicators for predicting mortality remains unclear. The aim of this study was to investigate whether changes in health indicators such as frailty and physical performance improve mortality predictions in old age.

Methods

This is a population based prospective cohort study on 995 community-dwelling people aged 68–92 years from the Longitudinal Aging Study Amsterdam. Two measurements at a three-year interval (1995/1996 and 1998/1999) were available for the frailty index, frailty phenotype, grip strength, walking speed, and Mini-Mental State Examination (MMSE). Cox regression was used to analyze mortality risks associated with the current health status and changes in health, with mortality data up to 2017. The extent to which these health indicators improved mortality predictions compared to models with age and sex only was assessed by the area under the receiver operating characteristic curve (AUC).

Results

The AUC of age and sex for five-year mortality was 72.8% (95% CI 69.0 – 76.5) and was the lowest in the oldest old (age > 80.5 years). The added AUC of the current status of health indicators ranged from 0.7 to 3.3%. The added AUC of the three-year change was lower, ranging from -0.0 to 1.1%, whereas the added AUC of three-year change and current status combined was similar to current status alone, ranging from 0.6 to 3.2%. Across age, the added AUC of current status was highest in the oldest old, however there was no such pattern using three-year change. Overall, the frailty index appeared to improve mortality predictions the most, followed by the frailty phenotype, MMSE, grip strength, and walking speed.

Conclusions

Current health status improved mortality predictions better than changes in health. Its contribution was highest in the oldest old, but the added value to models with age and sex only was limited.

Mortality in hospitalised older patients: the WHALES short-term predictive score

OBJECTIVE

To develop and validate the WHALES screening tool predicting short-term mortality (3 months) in older patients hospitalised in an acute geriatric unit.

METHODS

Older patients transferred to an acute geriatric ward from June 2017 to December 2018 were included. The cohort was divided into two groups: derivation (n=664) and validation (n=332) cohorts. Cause for admission in emergency room, hospitalisation history within the previous year, ongoing medical conditions, cognitive impairment, frailty status, living conditions, presence of proteinuria on a urine strip or urine albumin-to-creatinine ratio and abnormalities on an ECG were collected at baseline. Multiple logistic regressions were performed to identify independent variables associated with mortality at 3 months in the derivation cohort. The prediction score was then validated in the validation cohort.

RESULTS

Five independent variables available from medical history and clinical data were strongly predictive of short-term mortality in older adults including age, sex, living in a nursing home, unintentional weight loss and self-reported exhaustion. The screening tool was discriminative (C-statistic=0.74 (95% CI: 0.67 to 0.82)) and had a good fit (Hosmer-Lemeshow goodness-of-fit test (X2 (3)=0.55, p=0.908)). The area under the curve value for the final model was 0.74 (95% CI: 0.67 to 0.82).

CONCLUSIONS AND IMPLICATIONS

The WHALES screening tool is a short and rapid tool predicting 3-month mortality among hospitalised older patients. Early identification of end of life may help appropriate timing and implementation of palliative care.

Tipping Points - Do the Prognostic Values of Multimorbidity and Functional Status Vary with Age?

Aging of the population is a pressing challenge for healthcare systems and knowledge of a patient’s prognosis is a key to shaping effective interventions. As the prevalence of multimorbidity strongly increases with age, the prognostic value of multiple disease diagnoses for survival among older people may diminish, whereas other measures of health, such as functional status (defined as a measure of an individual’s ability to perform activities of daily living), may become more important. In this commentary, the impact of age on the prognostic value of multimorbidity is discussed, with the aim of identifying relevant alternative risk indicators for different age groups. The key question is to determine at what age the prognostic value of multimorbidity for meaningful clinical outcomes decreases and is overridden by the prognostic value of functional status. This tipping point likely depends on age, calendar time, and birth cohort. The public health and clinical implications of these tipping points are important. Among younger and middle-aged persons, interventions could be directed towards prevention and treatment of specific diseases, while among older persons efforts should focus more on improving functional levels that include physical, emotional, and social dimensions.

Prediction of 11-year incidence of psychophysically dependent status or death among community-dwelling younger elderlies: from an age-specified community-based cohort study (the NISSIN project)

Background

Predicting adverse health events and implementing preventative measures are a necessary challenge. It is important for healthcare planners and policymakers to allocate the limited resource to high-risk persons. Prediction is also important for older individuals, their family members, and clinicians to prepare mentally and financially. The aim of this study is to develop a prediction model for within 11-year dependent status requiring long-term nursing care or death in older adults for each sex.

Methods

We carried out age-specified cohort study of community dwellers in Nisshin City, Japan. The older adults aged 64 years who underwent medical check-up between 1996 and 2000 were included in the study. The primary outcome was the incidence of the psychophysically dependent status or death or by the end of the year of age 75 years. Univariable logistic regression analyses were performed to assess the associations between candidate predictors and the outcome. Using the variables with p-values less than 0.1, multivariable logistic regression analyses were then performed with backward stepwise elimination to determine the final predictors for the model.

Results

Of the 1525 female participants at baseline, 105 had an incidence of the study outcome. The final prediction model consisted of 15 variables, and the c-statistics for predicting the outcome was 0.763 (95% confidence interval [CI] 0.714–0.813). Of the 1548 male participants at baseline, 211 had incidence of the study outcome. The final prediction model consisted of 16 variables, and the c-statistics for predicting the outcome was 0.735 (95% CI 0.699–0.771).

Conclusions

We developed a prediction model for older adults to forecast 11-year incidence of dependent status requiring nursing care or death in each sex. The predictability was fair, but we could not evaluate the external validity of this model. It could be of some help for healthcare planners, policy makers, clinicians, older individuals, and their family members to weigh the priority of support.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12199-021-00968-8.

The Impact of Frailty on COVID-19 Outcomes: A Systematic Review and Meta-analysis of 16 Cohort Studies

BACKGROUND

Frail patients are increasingly vulnerable to stress, which is mainly manifested by a reduced physiologic reserve in metabolic and immune systems and neuromuscular system. Several studies found a significant association of frailty with COVID-19 severity to support the evidence for the application of frailty assessment. However, there were contradictory results in other studies. Thus we conducted a systematic review and meta-analysis to synthesize the current studies to investigate impact of frailty on COVID-19 outcomes and provide evidence-based decisions in clinical practice.

OBJECTIVE

We aimed to synthesize the current studies to investigate impact of frailty on COVID-19 outcomes and provide evidence-based decisions in clinical practice.

DESIGN

A systematic review and Meta-analysis of 16 cohort studies.

PARTICIPANTS

Patients with COVID-19.

METHODS

A systematic retrieving for potential literature was conducted in several public electronic databases, including Medline(OvidSP), EMBASE, Pubmed and Chinese databases(China National Knowledge Infrastructure,Wanfang and Weipu) on August 1, 2020.The literature research was updated on October 26, 2020. Newcastle Ottawa Scale for cohort studies was used for quality assessment. RevMan (Version 5.3) and Stata 14.0 were used to synthesize the pooled effects.

RESULTS

According to the predefined inclusion and exclusion criteria, sixteen studies of 4324 patients were included in the final analysis. Frailty was significantly associated with increased risk of all-cause mortality among patients with COVID-19, with pooled adjusted odds ratios of 1.81 (95% confidence intervals:1.48,2.21, I2=87.0%, P<0.001). The result was consistent in stratified analysis to according to age, patient source, definitions of frailty, study quality, and adjustment method. Frailty was significant associated with an increased risk of COVID-19 severity, admission to intensive care unit, application of invasive mechanical ventilation, long-length stay.

CONCLUSIONS

In this meta-analysis, we found frailty was significantly associated with an increased risk of clinical adverse events (all- cause mortality, COVID-19 severity, admission to the intensive care unit, application of invasive mechanical ventilation, long-length stay). Given the epidemic of COVID-19 and shortage of medical resources, paying more attention to screening frailty would contribute to disease management and resource allocation among patients with COVID-19.

Development and validation of multivariable mortality risk-prediction models in older people undergoing an interRAI home-care assessment (RiskOP)

BACKGROUND

Currently, one-year survival of older people with complex co-morbidities is unpredictable. Identifying older adults with a reduced life expectancy will lead to more targeted care and better healthcare resource allocation.

METHODS

Development and validation of one-year and three-month mortality risks in people aged ≥65 years who had completed an International Resident Assessment Instrument-Home Care (interRAI-HC) assessment between July 2012 and March 2018. Data was split into development (90%) and validation data sets (10%). A multivariable logistic regression model using data from 108 interRAI questions across multiple domains was developed and validated using discrimination metrics and calibration curves. Variables each explaining at least 1% of the model were then used to develop and validate a parsimonious model. Subgroups by sex, age, ethnicity, and comorbidities were evaluated.

FINDINGS

There were 104,436 persons (60.2% female; mean age 82.1 years) in the study cohort of whom 20,972 (20.1%) died within one year. The full multivariable model had area under the curves (AUCs) of 0.778 to 0.795 in the 5 validation datasets and was well calibrated. After variable reduction a parsimonious model consisted of 16 variables and was well calibrated and the AUC remained high: 0.773 (0.769 to 0.777). The three-month parsimonious model comprised 22 variables and was well calibrated with an AUC of 0.843 (95%CI: 0.839 to 0.848).

INTERPRETATION

These community-based risk prediction models accurately predict mortality in older people with complex co-morbidities. They may contribute to both forecasting for policy making and clinical decision making regarding an individual's needs.

FUNDING

The New Zealand Health Research Council.