From the Editor
He drinks heavily, but does he have a diagnosed alcohol use disorder?
Does the answer to that question tie to ethnicity and biases? In a new American Journal of Psychiatry paper, Rachel Vickers-Smith (of the University of Kentucky) and her co-authors suggest it does. Drawing on US Veterans Affairs’ data with over 700,000 people, they analyzed the scores of a screening tool and the diagnoses with ethnicity recorded in the EMR. “We identified a large, racialized difference in AUD diagnosis, with Black and Hispanic veterans more likely than White veterans to receive the diagnosis at the same level of alcohol consumption.” We look at the paper and mull its implications.
In the second selection, Alastair C. van Heerden (of the University of the Witwatersrand) and his co-authors consider AI and its potential for global mental health services in a new JAMA Psychiatry Viewpoint. They focus on large language models (think ChatGPT) which could do several things, including helping to train and supervise humans. “Large language models and other forms of AI will fundamentally change how we treat mental disorders, allowing us to move away from the current model in which most of the world’s population does not have access to quality mental health services.”
And, in the third selection, Paula Halprin discusses her mother’s alcohol use in an essay for The Globe and Mail. In a moving piece that touches on anger, trauma, and regret, Halprin writes about her re-examination of her mother’s life. “I now understand my mother drank not because of a weak character, but to cope with a body wearing out before its time from unremitting pregnancy and as a way to swallow her anger and disappointment. It was also a way to mourn a loss of self.”
Selection 1: “Racial and Ethnic Bias in the Diagnosis of Alcohol Use Disorder in Veterans”
Rachel Vickers-Smith, Amy C. Justice, William C. Becker, et al.
The American Journal of Psychiatry, 3 May 2023 Online First
Diagnosis is a foundation of clinical decision making and treatment (1). Diagnoses, as clinical labels, can produce lasting stigma and, when inappropriate, can produce lasting damage to the individuals who receive a diagnosis (2). Diagnoses that are stigmatized, such as alcohol use disorder (AUD), can be particularly damaging. Furthermore, misdiagnosis can result in ineffective treatment, inaccurate prognostic assessments, poor outcomes, and distrust of the health care system…
Studies in the Veterans Health Administration of the Department of Veterans Affairs (VA) have shown that the rate of clinically recognized AUD is higher among Black and Hispanic veterans than among White veterans. Black veterans are also more likely than White veterans to be identified as needing an intervention and to receive psychosocial interventions but less likely to receive pharmacotherapy for AUD…
So begins a paper by Vickers-Smith et al.
Here’s what they did:
“The sample included 700,012 Black, White, and Hispanic veterans enrolled in the Million Veteran Program. Alcohol consumption was defined as an individual’s maximum score on the consumption subscale of the Alcohol Use Disorders Identification Test (AUDIT-C), a screen for unhealthy alcohol use. A diagnosis of AUD, the primary outcome, was defined by the presence of relevant ICD-9 or ICD-10 codes in electronic health records. Logistic regression with interactions was used to assess the association between race and ethnicity and AUD as a function of maximum AUDIT-C score.”
Here’s what they found:
- Demographics and men. 91% of the study sample were men. 74% were White, 19% Black, and 7% Hispanic.
- Scores and diagnosis in men. “The correlation between AUDIT-C score and AUD diagnosis was lowest among White men, followed by Hispanic and Black men. At every maximum AUDIT-C score, White men were less likely than Black men to receive an AUD diagnosis. The greatest difference was at an AUDIT-C score of 4 (i.e., the positive screening cutoff score for men), where White men were approximately one-third as likely to have an AUD diagnosis as Black men.”
- Multivariable analysis and men. “Black men had 23%-109% greater odds of an AUD diagnosis than White men at maximum AUDIT-C scores of 1 to 10…”
- Demographics and women. “62% were White, 30% Black, and 8% Hispanic.”
- Scores and diagnosis in women. “The correlation between AUDIT-C score and AUD diagnosis was lower among White women than Hispanic or Black women. Black women had a higher likelihood of an AUD diagnosis than White or Hispanic women at nearly every level of alcohol consumption, which was significant at AUDIT-C scores of 2 and 4 to 7.”
- Multivariable analysis and women. “Black women had a higher probability of an AUD diagnosis than Hispanic or White women at moderate AUDIT-C scores, and Hispanic women had a lower probability than Black or White women of an AUD diagnosis at higher AUDIT-C scores…”
A few thoughts:
1. This is a good study, with much to like: a huge sample with solid analyses, published in a major journal.
2. The main finding in a sentence: “The large discrepancy in the prevalence of AUD across groups despite a similar distribution of alcohol consumption levels suggests that there is racial and ethnic bias, with Black and Hispanic veterans more likely than White veterans to receive an AUD diagnosis.”
3. Where could bias have been seen the most? “At scores near the threshold, providers are more likely to assign a diagnosis of AUD to Black or Hispanic than White veterans…”
4. Like all studies, there are limitations, and the authors note several, including the reliance on self-reports – which could be touched by recall bias. And, of course, we can wonder about the generalizability of VA data.
5. This study has just been published and is thoroughly modern – tapping a big EMR and drawing on much data. But classic papers in psychiatry have shown that providers may minimize problems in patients that they identify with (favouring diagnoses like bipolar disorder over schizophrenia). Is there nothing new under the sun?
The full AJP paper can be found here:
Selection 2: “Global Mental Health Services and the Impact of Artificial Intelligence – Powered Large Language Models”
Alastair C. van Heerden, Julia R. Pozuelo, and Brandon A. Kohrt
JAMA Psychiatry, 17 May 2023 Online First
There is a large and growing need for mental health services worldwide, but there is a massive shortage of mental health specialists to meet these needs… One strategy that has emerged to address treatment gaps is to rely on nonspecialists (eg, lay health workers, teachers, social workers, and peer mentors) to provide mental health services. Although this approach can be effective, current strategies demand substantial training and supervision. They also require highly standardized interventions, which may paradoxically limit more person-centered treatments. Concurrently, the field of artificial intelligence (AI) is evolving rapidly and changing how we detect and treat mental health disorders. Artificial intelligence applications in psychiatry are varied and include developing prediction models for disease detection and prognosis, creating algorithms that can help clinicians choose the right treatment plan, monitoring patient progress based on data from wearable devices, building chatbots that deliver more personalized and timely interventions…
One breakthrough that may be particularly relevant for global mental health is the development of autoregressive large language models (LLMs), such as GPT (Generative Pretrained Transformer) and BERT (Bidirectional Encoder Representations from Transformers). These models use deep learning algorithms trained on big data sets scraped from the internet to predict the next word or sequence of words in a given text based on a prompt or question. Large language models are proving surprisingly capable on various tasks, including the ability to generate long, coherent, and convincing text that seems close to human quality.
So begins a paper by van Heerden et al.
They see great potential. “It seems possible that LLM-based agents could be fine-tuned on digitized texts in psychology and psychiatry, including textbooks and manuals, alongside many years’ worth of therapeutic transcripts, to offer an inexpensive tool capable of delivering complex and tailored therapeutic models with high fidelity, compassion, and perfect recall that can engage with thousands of clients simultaneously.”
Looking at low-income nations, they note: “The unavailability of supervision has been a major bottleneck in expanding mental health services in low-income countries, whether that be supervision of primary care physicians taking on diagnosis and pharmacological treatment of mental health conditions or community health workers delivering psychological services.”
“Large language models could help reduce this bottleneck by supporting the training and supervision of the human workforce. For example, LLMs could act as clients with whom nonspecialists could practice their skills, provide nonspecialists with customized learning materials, review session transcripts, and provide feedback based on competency rating tools. In other words, LLMs could assist nonspecialists in acquiring foundational knowledge in an engaging, tailored, and interactive manner. This strategy would free up specialists and allow them to focus on more complicated clinical supervision challenges and expand the human workforce for delivering quality mental health care.”
They suggest an agenda:
- “There is an urgent need to continue discussing and developing guidelines for the use of LLMs that take into consideration the needs and conditions of the population being served, [and] the availability and quality of existing mental health resources…”
- “Ethical frameworks and human oversight are essential to ensure AI-powered systems are used appropriately…”
- “Another challenge that needs addressing… is the perpetuation of existing cultural, gender, and language biases. The models are no better than the global pool of text on which they are based, and this pool is rooted in male, Western, colonial biases, thus generally perpetuating a homogeneous view of what a healthy psyche looks like and how to achieve it.”
A few thoughts:
1. This is a well-written paper.
2. AI holds much potential.
3. This topic is even more relevant and interesting since ChatGPT was released in November 2022.
4. Global mental health has been considered in past Readings. For example, we looked at a podcast interview with Dr. Vikram Patel (of Harvard University). You can find it here:
5. For the record, I asked ChatGPT about its potential for global mental health services with a focus on low-income countries.
“In low-income countries with limited access to mental health services, ChatGPT holds tremendous potential. Its AI capabilities can provide valuable information, guidance, and support to individuals in need. Available 24/7 and in multiple languages, ChatGPT offers immediate assistance, reducing stigma and empowering individuals to seek appropriate help. While it cannot replace face-to-face interactions, it complements existing services by offering preliminary assessments, educational materials, and referrals. By tailoring it to local contexts and collaborating with local organizations, ChatGPT can bridge the gap, extending mental health support to underserved regions and improving outcomes in low-income countries.”
The full JAMA Psych Viewpoint can be found here:
Selection 3: “It took some time, but I finally understand why my mother drank”
The Globe and Mail, 7 May 2023
The night before I married my first husband, 50 years ago in Dublin, I told my parents that if they drank at the wedding reception, I would never speak to them again.
I stood in the doorway of our 1950s kitchen, with its blue linoleum floor and white Formica table. I was wearing rollers in my hair, and some sort of gunk plastered on my face that I hoped would result in a glowing complexion on the big day.
So begins an essay by Halprin.
She explains her stance. “My no-booze admonition was mainly directed at my mother. At that time in Ireland, an inebriated man would have been a member in good standing of the fraternity of drinkers, particularly if his favourite tipple was a frothy-headed pint of Guinness, sucked back with the lads in his local pub.
“A publicly intoxicated woman, on the other hand, would be considered a disgrace to her family and to Catholic womankind in general. She’d be spoken of in knowing whispers.”
The essay discusses her mother’s misuse of alcohol and her own anger. “In the evenings when we gathered in the sitting room to watch the BBC, Josephine would place beside her armchair a wicker basket of the kind she kept her knitting in. This basket would contain no yarn but several bottles of Carlsberg Special Brew and a tall glass and a bottle opener… She would begin in fine form, but would slowly become incoherent as the basket emptied. I remember the irritation, bordering on contempt, that would rise in me like bile.”
The essay describes her history. “My mother took her first drink when she was in her late 30s, after she had had her children. It took another five years or so for the drinking to become a serious problem and some 20 years before she found her way out of the trap, reducing her intake to the occasional glass of wine with dinner. In that sad mid-point of her life, she used whisky and strong beer, sometimes mixed with sleeping pills, to deaden the pain of having too many children, never enough money, and knock-down fights with my father, the love of her life.”
But she also recognizes her mother’s amazing strengths. “She was an intelligent woman who read widely and grew hybrid tea roses. Those who knew her in her youth say she was a strong enough sea swimmer to have swum the English Channel. She was a canny poker player too. Who knows, she might have become a professional gambler. Or maybe a poet. She once shyly showed me a verse she had written that was printed in a national newspaper. Her brother Kevin, a doctor, used to say that Josephine would have done well in medical school. She was that smart.”
The writer shames her mother on the eve of her wedding – and is ashamed of her mother. Over time, she reconsiders her view. “The relationship between trauma and addiction was not well understood 50 years ago. Drinking to excess was seen as self-indulgent, a moral failing. I’m grateful that time and distance have combined to change my perspective.”
A few thoughts:
1. This is a well-written essay that expresses anger, yes, but also acceptance.
2. She notes her changing views – and also society’s. We don’t speak of “moral failing” as we did before. #Progress
3. Alcohol touches many, from mothers in Dublin to award-winning journalists in Toronto. The Toronto Star’s Jim Coyle discussed his alcohol use disorder and recovery in an incredible essay that was highlighted in a past Reading. You can find it here:
The full Globe essay can be found here:
Reading of the Week. Every week I pick articles and papers from the world of Psychiatry.
Leave a Reply