From the Editor

Imagine that you are asked to design a program to prevent depression in a population at risk. Would you hire psychiatrists? Look to nurses? Tap the expertise of psychologists? All three?

In the first selection from Nature Medicine, Pamela J. Surkan (of Johns Hopkins University) and her co-authors describe a study that focused on prevention. As they worked in Pakistan – a nation with few mental health providers by Western standards – they chose to train lay people, teaching them to deliver CBT. In their single-blind, randomized controlled trial, 1 200 women who were pregnant and had anxiety (but not depression) were given enhanced usual care or CBT. “We found reductions of 81% and 74% in the odds of postnatal MDE and of moderate-to-severe anxiety…” We discuss the paper and its implications.

In the second selection, Joseph Friedman and Dr. Helena Hansen (both of the University of California, Los Angeles) look at deaths of despair in the United States in a research letter for JAMA Psychiatry. Their work builds on the idea that some deaths are related to the hopelessness of a person’s social or economic circumstance; past publications focused largely on White Americans. Friedman and Hansen drew on more than two decades of data, including ethnicity, from a US database, finding a different pattern and that: “Rising inequalities in deaths of despair among American Indian, Alaska Native and Black individuals were largely attributable to disproportionate early mortality from drug- and alcohol-related causes…”

A recent survey finds that psychiatrists see AI as potentially helpful with paperwork and diagnosing patients. But could AI help you keep up with the literature? In the third selection from Annals of Family Medicine, Dr. Joel Hake (of the University of Kansas) and his co-authors used ChatGPT to produce short summaries of studies, then evaluated their quality, accuracy, and bias. “We suggest that ChatGPT can help family physicians accelerate review of the scientific literature.”


Selection 1: “Anxiety-focused cognitive behavioral therapy delivered by non-specialists to prevent postnatal depression: a randomized, phase 3 trial”

Pamela J. Surkan, Abid Malik, Jamie Perin, et al.

Nature Medicine, 16 February 2024

Common mental disorders (CMDs) such as depression and anxiety occur frequently in the perinatal period, often remain untreated and constitute important global health concerns. Pooled estimates indicate a 29% and 24% self-reported prevalence of elevated prenatal and postnatal anxiety symptoms in the Global South, respectively. The high prevalence of associated symptoms in the prenatal period presents challenges for families and pregnancies, especially in resource-limited settings where mental health services are scarce. Prenatal anxiety predicts anxiety, depression and suicide risk in the postnatal period. Anxiety and depression also tend to co-occur, with one study from Pakistan reporting that 69% of pregnant women with anxiety disorders also had depression and 37% of pregnant women with depression also had an anxiety disorder. Additionally, prenatal anxiety is related to poor growth and developmental outcomes in infants. While the effects of postnatal CMDs are serious, prenatal anxiety is one of the strongest predictors of postnatal CMDs, is highly prevalent (for example, affecting between one-third and one-half of women in Pakistan) and is an understudied condition, with the preponderance of epidemiological and intervention studies focused on depression.

Effective treatments exist, but preventive approaches that could reduce the prevalence of severe postnatal depression are lacking. This is critical given the host of negative infant outcomes, including impaired physical and cognitive development, associated with the condition. Cognitive behavior therapy (CBT) is an effective treatment for both anxiety and depression and, combined with strategies to address social stressors, has been used effectively for depression in the later prenatal and postnatal period in low-resource settings; however, CBT has rarely been used in primary prevention, an approach that is especially vital in low- and middle-income countries (LMICs) where an enormous treatment gap exists and where those in greatest need often have the least access to mental healthcare.

So begins a paper by Surkan et al.

Here’s what they did:

  • They conducted a “phase 3, two-arm, single-blind, randomized controlled trial in Pakistan…”
  • Participants were women at or under 22 weeks of pregnancy who had anxiety but not depression. (!) Women needed to be at least 18 years of age and to live near the health facility. 
  • Participants were randomized to enhanced care or to the Happy Mother-Healthy Baby program, which offered cognitive behavioral therapy (six one-on-one intervention sessions delivered by non-specialist providers). 
  • The primary outcome: major depression, generalized anxiety disorder, or both at six weeks after delivery.

Here’s what they found:

  • More than 91 000 women were screened with the vast majority excluded. Of the 1 307 remaining women with symptoms of at least mild anxiety, 1 200 (92%) consented to participate and were randomized (600 to the enhanced care group and 600 to the Happy-Healthy Baby program).
  • Demographics. The average age was 25 years, with an average gestational age of 16 weeks, and about half (49%) were low income. The average Hospital Anxiety and Depression Scale (HADS) anxiety score was 11.0.
  • Results. “Examined jointly, we found 81% reduced odds of having either a major depressive episode (MDE) or moderate-to-severe anxiety for women randomized to the intervention (adjusted odds ratio (aOR) = 0.19…). Overall, 12% of women in the intervention group developed MDE at 6 weeks postpartum, versus 41% in the control group. We found reductions of 81% and 74% in the odds of postnatal MDE (aOR = 0.19…) and of moderate-to-severe anxiety (aOR = 0.26…), respectively.” 
  • Safety. “The study did not have any unexpected events or adverse events related to the intervention, thus the adverse events that occurred were unrelated to the trial.” 

A few thoughts:

1. This is an impressive study published in a big journal with a big result.

2. The main finding in a sentence: “Results of our trial show that an early prenatal intervention to treat symptoms of anxiety had strong preventive effects on postnatal depression in addition to reducing moderate-to-severe symptoms of anxiety.”

3. Wow. They sought to prevent depression in a vulnerable population, and they did.

4. Like other programs in low-income and middle-income nations, they worked with lay people who were taught CBT. It’s an impressive model, and one that is very different than what we would think of in the West. Of course, we can ask: are aspects of this study relevant here?

5. Perspective: the drop-out rate was high. “We expected a high attrition rate (30%) among trial participants; however, approximately 37% of women were lost before completing final data collection. In addition, 25% of those randomized to the intervention never received any intervention session.”

6. Global psychiatry has been strongly influenced by the work of Dr. Vikram Patel (of Harvard University) who was the first to use the lay-person-CBT model in Goa. (He’s cited three times by Surkan et al.) I interviewed him for a podcast in 2019. On the need to be creative, he noted: “There are more psychiatrists of African origin in the US than in the whole of Africa. And I could actually say similar examples from the Philippines or India or many other countries. There is an enormous shortage of mental health resource.” The podcast can be found here:—global-psychiatry

The full Nat Med paper can be found here:

Selection 2: “Trends in Deaths of Despair by Race and Ethnicity From 1999 to 2022”

Joseph Friedman and Helena Hansen

JAMA Psychiatry, 10 April 2024  Online First

The deaths of despair theory is influential in understanding the declining health status among people in the US. The seminal analysis on the topic highlighted that midlife deaths from suicide, drug overdose, and alcoholic liver disease were driving a unique decrease in life expectancy among White individuals. Deaths from these causes have been collectively referred to as deaths of despair and have been argued to be associated with declining social and economic conditions and a perceived loss of status, especially among White US individuals without a college degree. We assessed trends by race and ethnicity in deaths of despair, especially in the years following the seminal analysis, during which increases in racial and ethnic inequalities were reported for numerous causes of death.

So begins a research letter by Friedman and Hansen.

Here’s what they did:

  • Drawing from the Centers for Disease Control and Prevention WONDER database, they calculated midlife (age, 45-54 years) mortality from deaths of despair in the US from January 1999 to December 2022.
  • They used race and ethnicity data.

Here’s what they found:

  • Black individuals. “After 2015, there was a sharp increase in deaths of despair among Black individuals. By 2022, the death rate nearly tripled among Black individuals, rising to 103.81 per 100 000 compared with 102.63 per 100 000 among White individuals.”
  • American Indian or Alaska Native individuals. “The midlife mortality rate from deaths of despair among American Indian or Alaska Native individuals was significantly higher than that among White individuals from 1999 to 2022. In 2022, the midlife death rate from these causes among American Indian or Alaska Native individuals was 241.70 per 100 000, 2.36 times the rate among White individuals.”
  • Drugs. “The rising burden of deaths of despair among racial and ethnic minoritized communities reflects higher rates of drug overdose mortality among American Indian or Alaska Native (104.95 per 100 000) and Black (84.80 per 100 000) individuals in 2022 compared with White individuals (59.26 per 100 000).”

A few thoughts:

1. This presents interesting and important US data.

2. Ouch.

3. Writing on X (formerly Twitter), Friedman summarizes the big finding: “Helena Hansen and I show that deaths of despair among Black Americans overtook the rate for White Americans in 2022. This further debunks the narrative that these deaths uniquely harm White communities.” Friedman has just matched to the Psychiatry residency program at the University of California, San Diego. We look forward to more research from him in the coming years.

4. The authors point a way forward: “Interventions to address these challenges must be culturally appropriate and targeted to reduce inequalities.” Thoughtful.

The full JAMA Psych research letter can be found here:

Selection 3: “Quality, Accuracy, and Bias in ChatGPT-Based Summarization of Medical Abstracts”

Joel Hake, Miles Crowley, Allison Coy, et al.

Annals of Family Medicine, March 2024

Nearly 1 million new journal articles were indexed by PubMed in 2020, and worldwide medical knowledge now doubles approximately every 73 days. Meanwhile, care models emphasizing clinical productivity leave clinicians with scant time to review the academic literature, even within their own specialty.

Recent developments in artificial intelligence (AI) and natural language processing might offer new tools to confront this problem. Large language models (LLMs) are neural network–based computer programs that use a detailed statistical understanding of written language to perform many tasks including text generation, summarization, software development, and prediction. One LLM, Chat Generative Pretrained Transformer (ChatGPT) has recently garnered substantial attention in the popular press. We wondered if LLMs could help physicians review the medical literature more systematically and efficiently… Unfortunately, LLMs can also ‘hallucinate,’ producing text that, whereas often convincing and seemingly authoritative, is not fact based. In addition, many concerns have been raised regarding the possibility of bias in AI models including LLMs.

So begins a paper by Hake et al.

Here’s what they did:

“We evaluated ChatGPT’s ability to summarize 140 peer-reviewed abstracts from 14 journals. Physicians rated the quality, accuracy, and bias of the ChatGPT summaries. We also compared human ratings of relevance to various areas of medicine to ChatGPT relevance ratings.”

Here’s what they found:

  • Length. ChatGPT wrote summaries that were 70% shorter (mean abstract length of 2,438 characters down to 739 characters). 
  • Quality, accuracy, and bias. Summaries were scored by physician reviewers as high quality (median score 90.0), high accuracy (median 92.5), and low bias (median 0). ChatGPT self-rated the summaries as high quality, high accuracy, and low bias.
  • Hallucinations. Hallucinations were “uncommon.” 
  • Relevance. “Classification of the relevance of entire journals to various fields of medicine closely mirrored physician classifications…”

A few thoughts:

1. This is an interesting study with an unusual use for ChatGPT.

2. The big finding: ChatGPT produced summaries that scored high on quality, accuracy, and bias, according to physician reviewers. (ChatGPT agreed in its self-evaluation – a nice touch.)

3. Of course, the study focused on only a handful of journals. Another limitation: “We also focused exclusively on primary research reports, systematic reviews, and meta-analyses. We did not evaluate the performance of ChatGPT on abstracts from many other article types that are important to the scientific process including nonsystematic reviews, perspectives, commentaries, and letters to the editor.”

4. The implications? The authors argue that ChatGPT could be used as a “screening tool to help busy clinicians and scientists more rapidly evaluate whether further review of an article is likely to be worthwhile.” I’m not sure we are quite there yet – but perhaps soon. (In the meantime, I’m going to continue writing the Readings.)

The full Ann Fam Med study can be found here:

Reading of the Week. Every week I pick articles and papers from the world of Psychiatry.