scholarly journals Developing Machine Learning Algorithms for the Prediction of Early Death in Elderly Cancer Patients: Usability Study (Preprint)

2018 ◽  
Author(s):  
Gabrielle Ribeiro Sena ◽  
Tiago Pessoa Ferreira Lima ◽  
Maria Julia Gonçalves Mello ◽  
Luiz Claudio Santos Thuler ◽  
Jurema Telles Oliveira Lima
Keyword(s):  
Machine Learning ◽  
Cancer Patients ◽  
Geriatric Assessment ◽  
Short Form ◽  
Geriatric Oncology ◽  
Depression Scale ◽  
Mini Nutritional Assessment ◽  
Early Death ◽  
Machine Learning Algorithms ◽  
Elderly Cancer Patients

BACKGROUND The importance of classifying cancer patients into high- or low-risk groups has led many research teams, from the biomedical and bioinformatics fields, to study the application of machine learning (ML) algorithms. The International Society of Geriatric Oncology recommends the use of the comprehensive geriatric assessment (CGA), a multidisciplinary tool to evaluate health domains, for the follow-up of elderly cancer patients. However, no applications of ML have been proposed using CGA to classify elderly cancer patients. OBJECTIVE The aim of this study was to propose and develop predictive models, using ML and CGA, to estimate the risk of early death in elderly cancer patients. METHODS The ability of ML algorithms to predict early mortality in a cohort involving 608 elderly cancer patients was evaluated. The CGA was conducted during admission by a multidisciplinary team and included the following questionnaires: mini-mental state examination (MMSE), geriatric depression scale-short form, international physical activity questionnaire-short form, timed up and go, Katz index of independence in activities of daily living, Charlson comorbidity index, Karnofsky performance scale (KPS), polypharmacy, and mini nutritional assessment-short form (MNA-SF). The 10-fold cross-validation algorithm was used to evaluate all possible combinations of these questionnaires to estimate the risk of early death, considered when occurring within 6 months of diagnosis, in a variety of ML classifiers, including Naive Bayes (NB), decision tree algorithm J48 (J48), and multilayer perceptron (MLP). On each fold of evaluation, tiebreaking is handled by choosing the smallest set of questionnaires. RESULTS It was possible to select CGA questionnaire subsets with high predictive capacity for early death, which were either statistically similar (NB) or higher (J48 and MLP) when compared with the use of all questionnaires investigated. These results show that CGA questionnaire selection can improve accuracy rates and decrease the time spent to evaluate elderly cancer patients. CONCLUSIONS A simplified predictive model aiming to estimate the risk of early death in elderly cancer patients is proposed herein, minimally composed by the MNA-SF and KPS. We strongly recommend that these questionnaires be incorporated into regular geriatric assessment of older patients with cancer.

scholarly journals Developing Machine Learning Algorithms for the Prediction of Early Death in Elderly Cancer Patients: Usability Study

JMIR Cancer ◽  
10.2196/12163 ◽  
2019 ◽  
Vol 5 (2) ◽  
pp. e12163 ◽  
Author(s):  
Gabrielle Ribeiro Sena ◽  
Tiago Pessoa Ferreira Lima ◽  
Maria Julia Gonçalves Mello ◽  
Luiz Claudio Santos Thuler ◽  
Jurema Telles Oliveira Lima
Keyword(s):  
Machine Learning ◽  
Cancer Patients ◽  
Geriatric Assessment ◽  
Short Form ◽  
Geriatric Oncology ◽  
Depression Scale ◽  
Mini Nutritional Assessment ◽  
Early Death ◽  
Machine Learning Algorithms ◽  
Elderly Cancer Patients

Background The importance of classifying cancer patients into high- or low-risk groups has led many research teams, from the biomedical and bioinformatics fields, to study the application of machine learning (ML) algorithms. The International Society of Geriatric Oncology recommends the use of the comprehensive geriatric assessment (CGA), a multidisciplinary tool to evaluate health domains, for the follow-up of elderly cancer patients. However, no applications of ML have been proposed using CGA to classify elderly cancer patients. Objective The aim of this study was to propose and develop predictive models, using ML and CGA, to estimate the risk of early death in elderly cancer patients. Methods The ability of ML algorithms to predict early mortality in a cohort involving 608 elderly cancer patients was evaluated. The CGA was conducted during admission by a multidisciplinary team and included the following questionnaires: mini-mental state examination (MMSE), geriatric depression scale-short form, international physical activity questionnaire-short form, timed up and go, Katz index of independence in activities of daily living, Charlson comorbidity index, Karnofsky performance scale (KPS), polypharmacy, and mini nutritional assessment-short form (MNA-SF). The 10-fold cross-validation algorithm was used to evaluate all possible combinations of these questionnaires to estimate the risk of early death, considered when occurring within 6 months of diagnosis, in a variety of ML classifiers, including Naive Bayes (NB), decision tree algorithm J48 (J48), and multilayer perceptron (MLP). On each fold of evaluation, tiebreaking is handled by choosing the smallest set of questionnaires. Results It was possible to select CGA questionnaire subsets with high predictive capacity for early death, which were either statistically similar (NB) or higher (J48 and MLP) when compared with the use of all questionnaires investigated. These results show that CGA questionnaire selection can improve accuracy rates and decrease the time spent to evaluate elderly cancer patients. Conclusions A simplified predictive model aiming to estimate the risk of early death in elderly cancer patients is proposed herein, minimally composed by the MNA-SF and KPS. We strongly recommend that these questionnaires be incorporated into regular geriatric assessment of older patients with cancer.

Machine Learning Algorithms in the Prediction of Early Death in Elderly Cancer Patients (Preprint)

2019 ◽  
Author(s):  
Gabrielle Sena ◽  
Tiago Pessoa Lima ◽  
Jurema Telles Lima ◽  
Maria Julia Mello ◽  
Luiz Claudio Thuler
Keyword(s):  
Machine Learning ◽  
Cancer Patients ◽  
Mental State ◽  
Mini Mental State Examination ◽  
Early Death ◽  
Machine Learning Algorithms ◽  
Karnofsky Performance Scale ◽  
Elderly Cancer Patients ◽  
State Examination ◽  
Performance Scale

BACKGROUND The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of Machine Learning (ML) algorithms. The International Society of Geriatric Oncology (SIOG) recommends the use of the Comprehensive Geriatric Assessment (CGA), a multidisciplinary tool to evaluate health domains, for the follow-up of elderly cancer patients. However, none ML application have been proposed using CGA to classify elderly cancer patients. OBJECTIVE To propose and develop predictive models, using ML and CGA, to estimate the risk of early death in elderly cancer patients. METHODS The ability of ML algorithms to predict early mortality in a cohort involving 608 elderly cancer patients was evaluated. The CGA was conducted during admission by a multidisciplinary team and included the following questionnaires: Mini-Mental State Examination, Geriatric Depression Scale, International Physical Activity Questionnaire, Timed Get-Up and Go, Katz Index, Charlson Comorbidity Index, Karnofsky Performance Scale, Polypharmacy, Mini Nutritional Assessment. The K-fold Cross Validation algorithm was used to evaluate all possible combinations of these questionnaires to estimate the risk of early death, considered when occurring within six months of diagnosis, in a variety of ML classifiers, including Naive Bayes (NB), Decision Tree (J48) and Multilayer Perceptron (MLP). On each fold of evaluation, tie-breaking is handled by choosing the smallest set of questionnaires. RESULTS It was possible to select CGA questionnaire subsets with high predictive capacity for early death, either statistically similar (NB) or higher (J48 and MLP) compared to the use of all questionnaires investigated. These results show that CGA questionnaire selection can improve accuracy rates and decrease the time spent to evaluate elderly cancer patients. The only questionnaire selected in all folds was the Mini Nutrition Evaluation. The Karnofsky Performance Scale was selected in all folds by the NB and MLP, while the Mini Mental State Examination was selected in all folds by the NB. CONCLUSIONS A simplified predictive model aiming to estimate the risk of early death in elderly cancer patients is proposed herein, minimally composed by the Mini Nutrition Evaluation accompanied or not by the Karnofsky Performance Scale and/or the Mini-Mental State Examination.

Validation of G8 and its acuracy as a predictor of object in oncological patients with age of ≥ 60 YEARS. prospective cohort of 889 brazilian elderly cancer patients.

2019 ◽  
Vol 37 (15_suppl) ◽  
pp. e23036-e23036
Author(s):  
Jurema Telles O Lima ◽  
Raissa Viana ◽  
Letícia telles Sales ◽  
Mirella Rebello ◽  
José Natal Figueiroa ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Geriatric Assessment ◽  
Predictive Value ◽  
Prospective Cohort ◽  
Nutritional Assessment ◽  
Geriatric Oncology ◽  
Mini Nutritional Assessment ◽  
Early Death ◽  
Solid Organ ◽  
Oncologic Patients

e23036 Background: The G8, a questionnaire with 8 questions, has been specifically developed as a screening tool for vulnerability in older patients (70+) with cancer, with results suggesting 14 points as a threshold (equivalent to 90% sensitivity and 60% specificity). Its prognostic value for early death was not evaluated in Brazilian population. Methods: Cancer patients (60 years or older) with solid organ malignancies were included in a prospective cohort in a Brazilian Geriatric Oncology Clinic, between 2015 and 2017. Before the start of any oncologic therapy, a Comprehensive Geriatric Assessment (CGA) with 12 questionnaires was performed. For this study, we analyzed the G8 (Figure 1) considering items from Mini Nutritional Assessment (MNA), polypharmacy ( > 3 medications per day), self reported health status and the age of patient. G8 score was considered abnormal if ≤14, and early death (within six months of surveillance) was the gold standard for the study accuracy. Cox proportional hazard risk and Kaplan-Meier survival curves were performed. Results: 889 patients were enrolled and there were 145 (16.3 %) early deaths. Overall, 52% were male, the extremes of age were from 60 to 97 (mean 72.5 ± 0.24), and 38% were < 70 years old. The most common cancers were prostate (31.2%), digestive tract (21.9%) and breast (16.0%); 30.4% were metastatic. 470 (52,9%) from all patients had abnormal (≤ 14) G8 score; 128 (27.2%) who had abnormal G8 score died within the first 6 months, versus 17 (4.1%) with normal score (Table 1). At cut point 14, the G8 sensitivity was 88.3% (CI95% 81.9 - 93.0); specificity 54.0% (CI95% 50.4 - 57.7); positive predictive value 27.2% (CI95% 23.3 - 31.5); negative predictive value 95.9% (CI95% 93.6 - 97.5); area under ROC curve was 0.819 (Figure 2). For those older than 70, the G8 sensitivity was 91% (CI95% 83.4 - 96.1). Abnormal scores showed significant differences in survival probability (Figure 3). Conclusions: G8 score is a strong and consistent predictor of overall 6-month survival, regardless of age, metastatic status or tumor site in Brazilian older (60 or older) oncologic patients. This finding strengthens the clinical utility of this instrument in the Geriatric Oncology and may be an option to extend the practice of Geriatric Assessment in countries with medium-low economic development, even in the younger elderly population.

Comprehensive Geriatric Assessment Adds Information to Eastern Cooperative Oncology Group Performance Status in Elderly Cancer Patients: An Italian Group for Geriatric Oncology Study

2002 ◽  
Vol 20 (2) ◽  
pp. 494-502 ◽  
Author(s):  
Lazzaro Repetto ◽  
Lucia Fratino ◽  
Riccardo A. Audisio ◽  
Antonella Venturino ◽  
Walter Gianni ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Geriatric Assessment ◽  
Comprehensive Geriatric Assessment ◽  
Daily Living ◽  
Group Performance ◽  
Performance Status ◽  
Eastern Cooperative Oncology Group ◽  
Geriatric Oncology ◽  
Oncology Group ◽  
Elderly Cancer Patients

PURPOSE: To appraise the performance of Comprehensive Geriatric Assessment (CGA) in elderly cancer patients (≥ 65 years) and to evaluate whether it could add further information with respect to the Eastern Cooperative Oncology Group performance status (PS). PATIENTS AND METHODS: We studied 363 elderly cancer patients (195 males, 168 females; median age, 72 years) with solid (n = 271) or hematologic (n = 92) tumors. In addition to PS, their physical function was assessed by means of the activity of daily living (ADL) and instrumental activities of daily living (IADL) scales. Comorbidities were categorized according to Satariano’s index. The association between PS, comorbidity, and the items of the CGA was assessed by means of logistic regression analysis. RESULTS: These 363 elderly cancer patients had a good functional and mental status: 74% had a good PS (ie, lower than 2), 86% were ADL-independent, and 52% were IADL-independent. Forty-one percent of patients had one or more comorbid conditions. Of the patients with a good PS, 13.0% had two or more comorbidities; 9.3% and 37.7% had ADL or IADL limitations, respectively. By multivariate analysis, elderly cancer patients who were ADL-dependent or IADL-dependent had a nearly two-fold higher probability of having an elevated Satariano’s index than independent patients. A strong association emerged between PS and CGA, with a nearly five-fold increased probability of having a poor PS (ie, ≥ 2) recorded in patients dependent for ADL or IADL. CONCLUSION: The CGA adds substantial information on the functional assessment of elderly cancer patients, including patients with a good PS. The role of PS as unique marker of functional status needs to be reappraised among elderly cancer patients.

Basal physical inactivity as an independent risk factor for early death in elderly cancer patients: A prospective Brazilian cohort study with 608 patients.

2017 ◽  
Vol 35 (15_suppl) ◽  
pp. e21534-e21534
Author(s):  
Jurema Telles O Lima ◽  
Maria Julia Gonçalves Mello ◽  
Anke Bergmann ◽  
Mirella Rebello Bezerra ◽  
Zilda Cavalcanti ◽  
...  
Keyword(s):  
Physical Activity ◽  
Cohort Study ◽  
Cancer Patients ◽  
Geriatric Assessment ◽  
Physical Inactivity ◽  
Early Death ◽  
Premature Death ◽  
Elderly Cancer Patients ◽  
Low Middle Income Countries ◽  
Physically Inactive

e21534 Background:With the aging of the population, cancer in the elderly people is becoming an increasingly common and complex health problem, particularly in low-middle-income countries, where physical inactivity and obesity have also been challenging and correlated with cancer and aging. The aim of this study is to determine whether the baseline physical inactivity per International Physical Activity Questionnaire(IPAQ) is an independent factor of premature death in the first six months.Methods: prospective cohort study of elderly patients (≥ 60 years). Participants with a recent diagnosis of cancer were from eight community hospitals and one cancer center in Northeast Brazil and were recruited during their first medical appointment at the outpatient oncologic clinic. Sociodemographic and clinical variables were determined and baseline comprehensive geriatric assessment (CGA) was conducted including IPAQ and Time Up and Go test. Data were analyzed using multivariate Cox proportional hazards models, overall survival was estimated using the Kaplan–Meier method and survival curves were compared using the Log rank test. Results:From 2015-2016, 608 patients, mean age 71.9 (SD ±7.4; range 60-96), 50.7% male, were enrolled. Main diagnoses were prostate cancer (29.1%), digestive system cancer (25.5%) and breast cancer (16.0%). 40.3% were considered obese/overweight (BMI ≥ 27). 42.1% were considered physically inactive per IPAQ. The mean of the Time Up and Go test on the physically inactive group by IPAQ was of 13.11s (SD ± 6.5), and of 22.72s (SD± 16.7) on the physically active one. After adjustment for age, site of cancer and cancer stage, the multivariate regression Cox model showed that being physically inactive is an independent predictor for early death (HR = 2.35, 95% CI: 1.50‒ 3.66, p < 0.001).Conclusions:Physical inactivity at admission was identified as a significant predictor of risk for premature death in elderly cancer patients. The evaluation of the physical activity should be incorporated into regular geriatric assessment of elderly cancer patients.

The comprehensive geriatric assessment in malignancies and prognostic factors: Description of a prospective cohort of 889 Brazilian elderly cancer patients.

2019 ◽  
Vol 37 (15_suppl) ◽  
pp. e23035-e23035
Author(s):  
Jurema Telles O Lima ◽  
Raissa Viana ◽  
Mirella Rebello ◽  
Maria Julia Gonçalves Mello ◽  
Letícia telles Sales ◽  
...  
Keyword(s):  
Clinical Practice ◽  
Cancer Patients ◽  
Geriatric Assessment ◽  
Comprehensive Geriatric Assessment ◽  
Prospective Cohort ◽  
Early Death ◽  
Northeastern Brazil ◽  
World Health ◽  
Elderly Cancer Patients ◽  
Risk Of Death

e23035 Background: According to the World Health Organization (WHO), the definition of "elderly" varies according to the degree of development of the country. In Low and Medium Development Countries (LMDC), a person aged 60 or older is considered elder, opposed to developed countries (65 or older). It is in LMDC that is occurring the largest relative increase in the incidence of cancer, specially those related to aging. The Comprehensive Geriatric Assessment (CGA) is still underutilized in oncological clinical practice, especially in LMDC. Objectives: To determine predictive factors for the occurrence of early death (in the first six months of surveillance) and to perform the development and temporal validation of a practical prognostic score based on the CGA to predict early death (up to 180 days) in elderly cancer patients. Methods: A prospective cohort enrolled elderly patients ≥ 60 years with a recent cancer diagnosis admitted between 2015-2017. The CGA performed at the time of admission included the following instruments: CCI; KPS; MMSE; TUG test; IPAQ; ADL; MNA; MNA-SF; GDS15; PPS and Polypharmacy. The studied outcome was early death, defined as the one that occurred within the first six months after the diagnosis. Survival analysis (Kaplan-Meier) and Cox proportional hazard regression was performed. Results: 889 patients were included in the study, performed at a referral center in cancer of a teaching hospital in Northeastern Brazil. The independent risk factors for death identified by CGA were: Mini exam of mental state (MMSE) as a continuous variable (HR 1.04 95% CI 1.00-1.07), Geriatric depression scale (GDS-15) ≥ 10 (HR 1 , 50 IC95% 1.10-2.07), Karnofsky Functional Performance Scale (KPS) < 50 (HR 1.57 IC95% 1.02-2.42), Katz Index ≤4 (HR 2.58 IC95% 1.68-3.97) and the Mininutrition assessment (MAN-SF) < 12 (HR 2.96 IC95% 2.00-4.39), with higher risk for early death amongst patients with abnormalities detected by the scales performed at admission (log rank < 0.001). Conclusions: Comprehensive geriatric assessment is an important tool to identify fragility in elderly cancer patients. Some of its scales should be incorporated into clinical practice, as they are simple and significant prognostic markers and identify patients with a higher risk of death in the first twelve months.

A practical clinical score to predict early death after a comprehensive geriatric assessment (CGA) in elderly cancer patients: A prospective cohort study with 608 patients.

2017 ◽  
Vol 35 (15_suppl) ◽  
pp. e21537-e21537
Author(s):  
Jurema Telles O Lima ◽  
Anke Bergmann ◽  
Maria Julia Gonçalves Mello ◽  
Zilda Cavalcanti ◽  
Mirella Rebello Bezerra ◽  
...  
Keyword(s):  
Risk Factors ◽  
Cohort Study ◽  
Cancer Patients ◽  
Geriatric Assessment ◽  
Comprehensive Geriatric Assessment ◽  
Prospective Cohort ◽  
Early Death ◽  
Clinical Score ◽  
Elderly Cancer Patients

e21537 Background: Components of the comprehensive geriatric assessment (CGA) correlate with risk of early mortality in elderly cancer patients (ECP). However, its complexity and the time required for its administration. The aim of this study was to determine the impact of each CGA domain on overall survival(OS) and to first step for the development of a prognostic scoring system to stratify ECP. Methods: a prospective cohort study. Participants with a recent diagnosis of cancer were from eight hospitals and one cancer center in Brazil and were recruited during their first medical appointment at the outpatient oncologic clinic. A basal CGA was done before the care decision (ADL, Charlson Comorbidity Index- CCI, Karnofsky Performance status – KPS, GDS15, IPAQ, MMSE, MNA, MNA-SF, PS, PPS, Polipharmacy, QLQc30, TUG). During the follow up of six months, information about the treatments performed and early death was collected. OS was estimated using the Kaplan–Meier method, and survival curves were compared using the Log rank test for categorical variables. A multivariate Cox proportional hazards model was used to select early death risk factors. A clinical score considering the number of risk variables was created. Results: From 2015-2016, 608 ECP, mean age 71.9 (SD ±7.4; range 60-96), 50.7% male, were enrolled. 100 (16.4%) ECP died in less than six months of follow-up. In our multivariate model, controlled by age, site of cancer and cancer stage, the remaining significant risk factors were malnutrition/nonutrition determined by MNA (HR 3.3, 95%CI 1.81-5.99, p < 0.001), KPS < 50% (HR 2.44, CI 1.56-3.81, p < 0.001) and CCI > 2 (HR 1.6, CI 1.09-2.52, p = 0.018). The risk for early death according to the number of risk variables: three (HR 12.99, CI 5.69-29.60, p < 0.001), two (HR 5.65, CI 2.61-12.24, p < 0.001) or one (HR 2.7, CI 1.28-5.87, p = 0.009). Conclusions: a practical clinical score using three instruments of the CGA (MNA, KPS and CCI) can predict independent the risk for an early death in ECP. The development of a practical system for risk scoring, incorporating few clinical prognostic factors, helps to stratify patients into risk groups and to plan a personalized care.

Sign in / Sign up

Export Citation Format

Share Document