Machine Learning Links Depression, Quality of Life to ALS Caregiver Burden

Iqra Mumal, MSc avatar

by Iqra Mumal, MSc |

Share this article:

Share article via email
ALS Focus survey

A machine learning algorithm has shown that depression and a perceived lower quality of life are significant predictors of high caregiver burden among those who care for patients with amyotrophic lateral sclerosis (ALS), a study has found.

The study, “Prediction of caregiver burden in amyotrophic lateral sclerosis: a machine learning approach using random forests applied to a cohort study,” was published in the journal BMJ Open.

ALS is a rare neurodegenerative disease that causes motor disability in patients, who increasingly need the assistance of caregivers as their disease worsens. In turn, caregivers, often family members, face high levels of burden due to their increased responsibilities for patients.

“The term ‘caregiver burden’ represents the decline of the caregiver’s emotional or physical health, social life and financial status,” the study stated.

In the study, a group of researchers in Ireland set out to determine the factors that better predict caregiver burden by using machine learning, which is an application of artificial intelligence that can be used to build predictive models.

The researchers also set out to build a clinical decision support system (CDSS), which is software that incorporates patient data to help healthcare practitioners make better, faster, and more personalized care decisions about their patients.

In this case, the CDSS used patient and caregiver characteristics to determine if a caregiver required additional support to help alleviate their burden.

The study had two goals: the first was to identify caregiver and patient characteristics associated with caregiver burden in ALS; the second was to group these characteristics together and build a model that could alert doctors when there’s a risk of high caregiver burden.

The researchers examined data from 89 patient-caregiver pairs followed at the National ALS/Motor Neuron Disease Multidisciplinary Clinic at Beaumont Hospital, in Dublin. Patients were assessed at three different time points at 4–6 month intervals.

Information was collected from patients and their caregivers, including demographics, socioeconomic data, quality of life, and anxiety and depression.

Most caregivers were women (70%), with a mean age of 56 years, and were either spouses (70%), children (21%), parents (2.2%), siblings (4.4%), or friends (1.1%) of ALS patients. Caregiving hours per week ranged from zero to 168 hours, with a median of 28 hours.

The researchers built several predictive models using a total of 232 predictive factors. The most accurate took into account 25 variables and had 92% sensitivity and 78% specificity, meaning it correctly identified 92% of caregivers with high burden, and 78% of those with low burden.

In particular, the model revealed that the caregiver’s perceived quality of life, their depression score, and the amount of control caregivers felt they had over their lives were the most predictive factors for burden.

Next, the researchers developed a CDSS that could alert when a caregiver is at risk for high burden, using data that may be routinely collected. Hence, the caregiver’s quality of life, as well as anxiety and depression scores, were excluded from the analysis.

The most predictive features for CDSS were weekly caregiving duties of the primary caregiver, his or her age and health, as well as the patient’s physical functioning and age of disease onset.

This model had a lower sensitivity (84%) and specificity (72%), but while caregivers with low burden were misclassified as having high burden in nearly 30% of cases, the algorithms captured most of the caregivers who experienced high burden (84%).

The researchers also found that the inability of patients to cut their own food and handle utensils was also highly predictive of caregiver burden.

“While additional work will be required to refine the model, the work demonstrates a proof of concept of an informatics solution to identifying caregivers at risk that can be incorporated into future care pathways,” the researchers said.