News Briefs – MAY 2022
by INELDA
CANADIAN NURSING HOME STUDY | AI AND PANCREATIC CANCER | SCREENING APP FOR DEPRESSION
Canada’s New Approach to Advance Care Planning in Nursing Homes
Advance care planning allows people to establish their wishes and preferences in regards to treatment at end of life. Yet for nursing home patients in the United States, only 59% have advance directives and only 17% have a living will, according to the Journal of Palliative Medicine. The situation is similar in Canada, where a group of researchers trialed a new approach called Better tArgeting, Better outcomes for frail ELderly patients (BABEL). Residents who participated in the study avoided unhelpful and possibly harmful treatments that affect the quality of life and had a fivefold improvement in how they rated the comprehensiveness of their advance care planning.
The BABEL approach was developed by researchers at the Universities of Waterloo, Manitoba, and Calgary. The nursing home residents selected for the study were assessed as being at risk of dying within the next 6 to 12 months. The study involved a 30-minute preliminary conversation to identify the nursing home residents’ substitute decision makers (SDM) and their preliminary wishes regarding resuscitation (CPR) and transfer to the hospital. The preliminary discussion was followed by a discussion of one hour or longer with each resident, each SDM, and staff, sometimes including the physician. The study involved randomized recruitment for both intervention and control groups involving 29 nursing homes and 713 residents.
The advance care planning confirmed the SDM’s identity and role, preparing each SDM for medical emergencies, clarifying the resident’s medical situation, ascertaining the resident’s preferred philosophy to guide decision-making, and determining the resident’s preferred options for specific medical emergencies most relevant to their health. One of the primary principles of the BABEL approach is a recognition that even residents with poor cognitive function can and must be allowed to express their wishes about their care and end-of-life preferences, even if they are ultimately found to be incapable of understanding or making known their wishes. In other words, the resident’s wishes always matter.
Study strengths highlighted by the researchers included the fact that BABEL was more comprehensive than previous approaches and that its intervention design embedded best practices into the usual care processes. The study findings indicated that advance care planning needs to be seen as a core expectation and competency of nursing home staff because it is fundamental to nursing home care.
AI May Detect Early Signs of Pancreatic Cancer
Pancreatic ductal adenocarcinoma (PDAC) is the most common form of pancreatic cancer, accounting for over 90% of cases. It is also one of the most challenging cancers to treat, making early detection critical. The overall five-year survival rate of PDAC barely exceeds 10%. But now an artificial intelligence (AI) tool developed at Cedars-Sinai Medical Center in Los Angeles holds the promise of earlier detection, which could result in a survival rate of up to 50%.
Although PDAC accounts for only about 3% of all cancers, it causes 8% of cancer deaths, making it the third most deadly cancer, with nearly 47,000 deaths a year. One of the reasons for this high death rate compared with the frequency of the disease is that PDAC aggressively develops metastases quickly. Compounding that problem is the fact that in the early stages of PDAC, people don’t have symptoms. The result: 80% of people with PDAC have advanced disease when they are first diagnosed. This makes surgical removal of the entire tumor very unlikely.
The main symptoms of PDAC are abdominal pain and unexplained weight loss, which are associated with a broad range of diseases. So when a person presents these symptoms to a doctor or goes to the hospital, the initial examinations look for illness other than pancreatic cancer. By the time the cancer is visible on CT scans and diagnosis is definitive, the disease is usually advanced.
Using the naive Bayes AI classifier tool, the researchers at Cedars-Sinai were able to identify a large number of unique features potentially predictive of PDAC in prediagnostic CT scans from patients who later developed the disease. These features, which can’t be detected by the human eye, were found through radiomic analysis. By identifying patients with the potential to develop the disease before there are visible signs of it, doctors can carefully monitor them. This will allow doctors to detect PDAC in its earliest stages, when removal of the tumor is still possible.
The researchers believe this is the first time that irregularities at the tissue microlevel of the pancreas predictive of PDAC have been identified with CT scans analyzed through an automated system. Although this work seems very promising, the researchers also cautioned that the amount of eligible data for the study was low, as the prediagnostic scans for people with PDAC are rarely available. They consider their work a proof of concept that will encourage researchers to establish a large data set to validate the results.
Screening Voice Recordings for Depression
A group of researchers at Worcester Polytechnic Institute have developed a computer program that screens voice recordings to identify people who are depressed. This highly effective program can help alert physicians and other clinicians that a patient needs help. The program analyzes the words a person uses as well as tone of voice. It was presented at the Association for Computing Machinery Conference on Information and Knowledge Management in November 2021, where it received the award for best applied research.
According to the researchers, if a person is depressed, that person’s vocal tone becomes a monotone and might jitter or shake a little bit. The program, Audio-assisted Bidirectional Encoder Representations from Transformers (AudiBERT), automates the detection of these signs in the human voice through machine learning models. The researchers experimented with 15 voice data sets of clinical interviews in which a virtual agent asked patients different questions, such as “How are you doing today?” The data sets were scored according to the depression status of each participant based on a depression screening questionnaire. Even though there were a limited number of data sets, AudiBERT accurately detected depression in the voice recordings.
AudiBERT could be deployed by doctors to screen for signs of depression and to monitor patients over time for signals that they need additional support. The researchers envision a day when patients visiting a doctor’s office or being seen at a virtual appointment could seamlessly be screened for depression based on real-time analysis of their voice as recorded during the visit. It is easy to see how this kind of technology could be used in the future to identify depression in patients receiving end-of-life care. It also opens up the possibility of identifying other mental health issues for patients undergoing treatment for a terminal illness.