PredictMed: A logistic regression-based model to predict health conditions in cerebral palsy

Does a Community Care programme reach a high health need population and high users of acute care hospital services in Melbourne, Australia? An observational cohort study

OBJECTIVE

The Community Care programme is an initiative aimed at reducing hospitalisations and emergency department (ED) presentations among patients with complex needs. We aimed to describe the characteristics of the programme participants and identify factors associated with enrolment into the programme.

DESIGN

This observational cohort study was conducted using routinely collected data from the National Centre for Healthy Ageing data platform.

SETTING

The study was carried out at Peninsula Health, a health service provider serving a population in Melbourne, Victoria, Australia.

PARTICIPANTS

We included all adults with unplanned ED presentation or hospital admission to Peninsula Health between 1 November 2016 and 31 October 2017, the programme's first operational year.

OUTCOME MEASURES

Community Care programme enrolment was the primary outcome. Participants' demographics, health factors and enrolment influences were analysed using a staged multivariable logistic regression.

RESULTS

We included 47 148 adults, of these, 914 were enrolled in the Community Care programme. Participants were older (median 66 vs 51 years), less likely to have a partner (34% vs 57%) and had more frequent hospitalisations and ED visits. In the multivariable analysis, factors most strongly associated with enrolment included not having a partner (adjusted OR (aOR) 1.83, 95% CI 1.57 to 2.12), increasing age (aOR 1.01, 95% CI 1.01 to 1.02), frequent hospitalisations (aOR 7.32, 95% CI 5.78 to 9.24), frequent ED visits (aOR 2.0, 95% CI 1.37 to 2.85) and having chronic diseases, such as chronic pulmonary disease (aOR 2.48, 95% CI 2.06 to 2.98), obesity (aOR 2.06, 95% CI 1.39 to 2.99) and diabetes mellitus (complicated) (aOR 1.75, 95% CI 1.44 to 2.13). Residing in aged care home and having high socioeconomic status) independently associated with reduced odds of enrolment.

CONCLUSIONS

The Community Care programme targets patients with high-readmission risks under-representation of individuals residing in residential aged care homes warrants further investigation. This study aids service planning and offers valuable feedback to clinicians about programme beneficiaries.

Identifying Postural Instability in Children with Cerebral Palsy Using a Predictive Model: A Longitudinal Multicenter Study

Insufficient postural control and trunk instability are serious concerns in children with cerebral palsy (CP). We implemented a predictive model to identify factors associated with postural impairments such as spastic or hypotonic truncal tone (TT) in children with CP. We conducted a longitudinal, double-blinded, multicenter, descriptive study of 102 teenagers with CP with cognitive impairment and severe motor disorders with and without truncal tone impairments treated in two specialized hospitals (60 inpatients and 42 outpatients; 60 males, mean age 16.5 ± 1.2 years, range 12 to 18 yrs). Clinical and functional data were collected between 2006 and 2021. TT-PredictMed, a multiple logistic regression prediction model, was developed to identify factors associated with hypotonic or spastic TT following the guidelines of "Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis". Predictors of hypotonic TT were hip dysplasia (p = 0.01), type of etiology (postnatal > perinatal > prenatal causes; p = 0.05), male gender, and poor manual (p = 0.01) and gross motor function (p = 0.05). Predictors of spastic TT were neuromuscular scoliosis (p = 0.03), type of etiology (prenatal > perinatal > postnatal causes; p < 0.001), spasticity (quadri/triplegia > diplegia > hemiplegia; p = 0.05), presence of dystonia (p = 0.001), and epilepsy (refractory > controlled, p = 0.009). The predictive model's average accuracy, sensitivity, and specificity reached 82%. The model's accuracy aligns with recent studies on applying machine learning models in the clinical field.

PredictMed-epilepsy: A multi-agent based system for epilepsy detection and prediction in neuropediatrics

BACKGROUND AND OBJECTIVE

Epileptic seizures are associated with a higher incidence of Developmental Disabilities and Cerebral Palsy. Early evaluation and management of epilepsy is strongly recommended. We propose and discuss an application to predict epilespy (PredictMed-Epilepsy) and seizures via a deep-learning module (PredictMed-Seizures) encompassed within a multi-agent based healthcare system (PredictMed-MHS); this system is meant, in perspective, to be integrated into a clinical decision support system (PredictMed-CDSS). PredictMed-Epilespy, in particular, aims to identify factors associated with epilepsy in children with Developmental Disabilities and Cerebral Palsy by using a prediction-learning model named PredictMed. PredictMed-epilespy methods: We performed a longitudinal, multicenter, double-blinded, descriptive study of one hundred and two children with Developmental Disabilities and Cerebral Palsy (58 males, 44 females; 65 inpatients, 37 outpatients; 72 had epilepsy - 22 of intractable epilepsy, age: 16.6±1.2y, range: 12-18y). Data from 2005 to 2021 on Cerebral Palsy etiology, diagnosis, type of epilepsy and spasticity, clinical history, communication abilities, behaviors, intellectual disability, motor skills, and eating and drinking abilities were collected. The machine-learning model PredictMed was exploited to identify factors associated with epilepsy. The guidelines of the "Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis" Statement (TRIPOD) were followed. PredictMed-epilepsy results: Cerebral Palsy etiology [(prenatal > perinatal > postnatal causes) p=0.036], scoliosis (p=0.048), communication (p=0.018) and feeding disorders (p=0.002), poor motor function (p<0.001), intellectual disabilities (p=0.007), and type of spasticity [(quadriplegia/triplegia > diplegia > hemiplegia), p=0.002)] were associated with having epilepsy. The prediction model scored an average of 82% of accuracy, sensitivity, and specificity. Thus, PredictMed defined the computational phenotype of children with Developmental Disabilities/Cerebral Palsy at risk of epilepsy. Novel contribution of the work: We have been developing and we have prototypically implemented a Multi-Agent Systems (MAS) that encapsulates the PredictMed-Epilepsy module. More specifically, we have implemented the Patient Observing MAS (PoMAS), which, as a novelty w.r.t. the existing literature, includes a complex event processing module that provides real-time detention of short- and long-term events related to the patient's condition.

Prediction Model for Identifying Computational Phenotypes of Children with Cerebral Palsy Needing Neurotoxin Treatments

Factors associated with neurotoxin treatments in children with cerebral palsy (CP) are poorly studied. We developed and externally validated a prediction model to identify the prognostic phenotype of children with CP who require neurotoxin injections. We conducted a longitudinal, international, multicenter, double-blind descriptive study of 165 children with CP (mean age 16.5 ± 1.2 years, range 12−18 years) with and without neurotoxin treatments. We collected functional and clinical data from 2005 to 2020, entered them into the BTX-PredictMed machine-learning model, and followed the guidelines, “Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis”. In the univariate analysis, neuromuscular scoliosis (p = 0.0014), equines foot (p < 0.001) and type of etiology (prenatal > peri/postnatal causes, p = 0.05) were linked with neurotoxin treatments. In the multivariate analysis, upper limbs (p < 0.001) and trunk muscle tone disorders (p = 0.02), the presence of spasticity (p = 0.01), dystonia (p = 0.004), and hip dysplasia (p = 0.005) were strongly associated with neurotoxin injections; and the average accuracy, sensitivity, and specificity was 75%. These results have helped us identify, with good accuracy, the clinical features of prognostic phenotypes of subjects likely to require neurotoxin injections.

Prediction Model for Identifying Factors Associated with Epilepsy in Children with Cerebral Palsy

(1) Background: Cerebral palsy (CP) is associated with a higher incidence of epileptic seizures. This study uses a prediction model to identify the factors associated with epilepsy in children with CP. (2) Methods: This is a retrospective longitudinal study of the clinical characteristics of 102 children with CP. In the study, there were 58 males and 44 females, 65 inpatients and 37 outpatients, 72 had epilepsy, and 22 had intractable epilepsy. The mean age was 16.6 ± 1.2 years, and the age range for this study was 12−18 years. Data were collected on the CP etiology, diagnosis, type of epilepsy and spasticity, clinical history, communication abilities, behaviors, intellectual disability, motor function, and feeding abilities from 2005 to 2020. A prediction model, Epi-PredictMed, was implemented to forecast the factors associated with epilepsy. We used the guidelines of “Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis” (TRIPOD). (3) Results: CP etiology [(prenatal > perinatal > postnatal causes) p = 0.036], scoliosis (p = 0.048), communication (p = 0.018), feeding disorders (p = 0.002), poor motor function (p < 0.001), intellectual disabilities (p = 0.007), and the type of spasticity [(quadriplegia/triplegia > diplegia > hemiplegia), p = 0.002)] were associated with having epilepsy. The model scored an average of 82% for accuracy, sensitivity, and specificity. (4) Conclusion: Prenatal CP etiology, spasticity, scoliosis, severe intellectual disabilities, poor motor skills, and communication and feeding disorders were associated with epilepsy in children with CP. To implement preventive and/or management measures, caregivers and families of children with CP and epilepsy should be aware of the likelihood that these children will develop these conditions.

Ensemble learning based functional independence ability estimator for pediatric brain tumor survivors

A history of brain tumor strongly affects children’s cognitive abilities, performance of daily activities, quality of life, and functional outcomes. In light of the difficulties in cognition, communication, physical skills, and behavior that these patients may encounter, occupational therapists should perform a comprehensive needs-led assessment of their global functioning after recovery. Such an assessment would ensure that the patients receive adequate support and services at school, at home, and in the community. By predicting the functional activity performance of children with a history of brain tumor, clinical workers can determine the progress of their ability recovery and the optimal treatment plan. We selected several features for testing and employed common machine learning models to predict Functional Independence Measure (WeeFIM) scores. The ensemble learning models exhibited stronger predictive performance than did the individual machine learning models. The ensemble learning models effectively predicted WeeFIM scores. Machine learning models can help clinical workers predict the functional assessment scores of patients with childhood brain tumors. This study used machine learning models to predict the WeeFIM scores of patients with childhood brain tumors and to demonstrate that ensemble machine learning models are more suitable for this task than are individual machine learning models.

Predicting osteoarthritis in adults using statistical data mining and machine learning

Background:

Osteoarthritis (OA) has traditionally been considered a disease of older adults (⩾65 years old), but it may appear in younger adults. However, the risk factors for OA in younger adults need to be further evaluated.

Objectives:

To develop a prediction model for identifying risk factors of OA in subjects aged 20–50 years and compare the performance of different machine learning models.

Methods:

We included data from 52,512 participants of the National Health and Nutrition Examination Survey; of those, we analyzed only subjects aged 20–50 years (n = 19,133), with or without OA. The supervised machine learning model ‘Deep PredictMed’ based on logistic regression, deep neural network (DNN), and support vector machine was used for identifying demographic and personal characteristics that are associated with OA. Finally, we compared the performance of the different models.

Results:

Being a female (p < 0.001), older age (p < 0.001), a smoker (p < 0.001), higher body mass index (p < 0.001), high blood pressure (p < 0.001), race/ethnicity (lowest risk among Mexican Americans, p = 0.01), and physical and mental limitations (p < 0.001) were associated with having OA. Best predictive performance yielded a 75% area under the receiver operating characteristic curve.

Conclusion:

Sex (female), age (older), smoking (yes), body mass index (higher), blood pressure (high), race/ethnicity, and physical and mental limitations are risk factors for having OA in adults aged 20–50 years. The best predictive performance was achieved using DNN algorithms.

Customer churn prediction for telecommunication industry: A Malaysian Case Study

Background: Customer churn is a term that refers to the rate at which customers leave the business. Churn could be due to various factors, including switching to a competitor, cancelling their subscription because of poor customer service, or discontinuing all contact with a brand due to insufficient touchpoints. Long-term relationships with customers are more effective than trying to attract new customers. A rise of 5% in customer satisfaction is followed by a 95% increase in sales. By analysing past behaviour, companies can anticipate future revenue. This article will look at which variables in the Net Promoter Score (NPS) dataset influence customer churn in Malaysia's telecommunications industry.  The aim of This study was to identify the factors behind customer churn and propose a churn prediction framework currently lacking in the telecommunications industry.   Methods: This study applied data mining techniques to the NPS dataset from a Malaysian telecommunications company in September 2019 and September 2020, analysing 7776 records with 30 fields to determine which variables were significant for the churn prediction model. We developed a propensity for customer churn using the Logistic Regression, Linear Discriminant Analysis, K-Nearest Neighbours Classifier, Classification and Regression Trees (CART), Gaussian Naïve Bayes, and Support Vector Machine using 33 variables.   Results: Customer churn is elevated for customers with a low NPS. However, an immediate helpdesk can act as a neutral party to ensure that the customer needs are met and to determine an employee's ability to obtain customer satisfaction.   Conclusions: It can be concluded that CART has the most accurate churn prediction (98%). However, the research is prohibited from accessing personal customer information under Malaysia's data protection policy. Results are expected for other businesses to measure potential customer churn using NPS scores to gather customer feedback.

Clinical predictive model of lumbar curve Cobb angle below selective fusion for thoracic adolescent idiopathic scoliosis: a longitudinal multicenter descriptive study

PURPOSE

To implement a clinically applicable, predictive model for the lumbar Cobb angle below a selective thoracic fusion in adolescent idiopathic scoliosis.

METHODS

A series of 146 adolescents with Lenke 1 or 2 idiopathic scoliosis, surgically treated with posterior selective fusion, and minimum follow-up of 5 years (average 7) was analyzed. The cohort was divided in 2 groups: if lumbar Cobb angle at last follow-up was, respectively, ≥ or < 10°. A logistic regression-based prediction model (PredictMed) was implemented to identify variables associated with the group ≥ 10°. The guidelines of the TRIPOD statement were followed.

RESULTS

Mean Cobb angle of thoracic main curve was 56° preoperatively and 25° at last follow-up. Mean lumbar Cobb angle was 33° (20; 59) preoperatively and 11° (0; 35) at last follow-up. 53 patients were in group ≥ 10°. The 2 groups had similar demographics, flexibility of both main and lumbar curves, and magnitude of the preoperative main curve, p > 0.1. From univariate analysis, mean magnitude of preoperative lumbar curves (35° vs. 30°), mean correction of main curve (65% vs. 58%), mean ratio of main curve/distal curve (1.9 vs. 1.6) and distribution of lumbar modifiers were statistically different between groups (p < 0.05). PredictMed identified the following variables significantly associated with the group ≥ 10°: main curve % correction at last follow-up (p = 0.01) and distal curve angle (p = 0.04) with a prediction accuracy of 71%.

CONCLUSION

The main modifiable factor influencing uninstrumented lumbar curve was the correction of main curve. The clinical model PredictMed showed an accuracy of 71% in prediction of lumbar Cobb angle ≥ 10° at last follow-up.

LEVEL OF EVIDENCE IV

Longitudinal comparative study.