Originally published on Scientific American’s Guest Blog on February 28, 2017.

On January 2, 1979, Dr. Rafael Osheroff was admitted to Chestnut Lodge, an inpatient psychiatric hospital in Maryland. Osheroff had a bustling nephrology practice. He was married with three children, two from a previous marriage. Everything had been going well except his mood.

For the previous two years, Osheroff had suffered from bouts of anxiety and depression. Dr. Nathan Kline, a prominent psychopharmacologist in New York City, had begun Osheroff on a tricyclic antidepressant and, according to Kline’s notes—which were later revealed in court—he improved.

But then Osheroff decided, against Kline’s advice, to change his dose. He got worse. So much worse that he was brought to Chestnut Lodge.

For the next seven months, Osheroff was treated with intensive psychotherapy for narcissistic personality disorder and depression. It didn’t help. He lost 40 pounds, suffered from excruciating insomnia, and began pacing the floor so incessantly that his feet became swollen and blistered.

Osheroff’s family, distressed by the progressive unraveling of his mind, hired a psychiatrist in Washington D.C. to intervene. In response, Chestnut Lodge held a clinical case conference yet decided to not change treatment. Importantly, they decided to not begin medications but to continue psychotherapy. They considered themselves “traditional psychiatrists”—practitioners of psychodynamic psychotherapy, the technique used by Sigmund Freud and other pioneers.

At the end of seven months, in a worse state yet, Osheroff’s family had him transferred from Chestnut Lodge to Silver Hill in Connecticut. Silver Hill’s doctors immediately diagnosed him as having a psychotic depressive episode and began him on a combination of phenothiazine and tricyclic antidepressants—a combination that recent clinical trials had shown to be effective.

“Within weeks after his transfer,” Dr. Alan Stone later wrote in the New England Journal of Medicine, “biological treatment with antidepressants [produced] a dramatic recovery.” Three months after his transfer, Osheroff left Silver Hill with a diagnosis of manic-depression, an early name for bipolar disorder. A quick turn-around.

Yet the previous year had destroyed Osheroff’s life. Kidney patients cannot wait a year to be seen, so Osheroff lost his lucrative medical practice. Concerned about her children, Osheroff’s ex-wife gained custody of two of his children. His reputation in the community was shattered.

Osheroff sued Chestnut Lodge for not providing the latest, evidence-based treatment. He sued “for negligence, because the staff failed to prescribe drugs and instead treated him according to the psychodynamic and social model.”

As Dr. Gerald Klerman described in the American Journal of Psychiatry: at the time, there was no evidence for psychodynamic therapy for psychotic depression. “In contrast, there are numerous randomized, controlled trials of the efficacy of ECT and the combination of tricyclic and neuroleptic medications in the treatment of psychotic depression.” Klerman later notes Chestnut Lodge’s “strange clinical logic to ignore available evidence in favor of a conjecture based on doctrine.”

Osheroff won the lawsuit and, on appeal, settled with Chestnut Lodge outside of court. (Chestnut Lodge, a lovely historical landmark, eventually folded, was converted to upscale condos, and subsequently burned to the ground.)

The case sparked a decades-long debate—one with “considerable spunk”—that captured the attention of the psychiatric community: “Has psychiatry reached the point where use of the psychodynamic model is viewed as malpractice when it is the exclusive treatment for serious mental disorders?” Stone asked. Another clinician questioned, “Are psychoanalysis and medical psychiatry compatible?”

Data showing one therapy was effective could evidently legally compel clinicians to change practice to avoid claims of negligence. Furthermore, if theories about the etiology of brain diseases like depression were demonstrated and generally accepted, clinicians who guide therapy with “traditional,” nonscientific theories could also be considered negligent.

Recall that since Osheroff’s 1980s case, tens of thousands of papers and scores of books have described our ever-deepening knowledge of the neuroscience of mental illness, fixing psychiatry squarely as a medical specialty, as a specialty of brains.

Yet, as Dr. Sophia Vinogradov, Chief of Psychiatry at the University of Minnesota Medical School, recently wrote in Nature Human Behavior, “There’s a secret that we psychiatrists do not like to talk about: the abysmally primitive state of how we assess, understand, and treat mental illness.”

But many have great hope this will change.

Last year, The Lancet Psychiatry published a joint study between The University of Texas Southwestern and Yale University used a machine-learning algorithm to see which of 164 clinical measures were most predictive of treatment success with the antidepressant citalopram.

The clinical measures included well-validated scales like the Quick Inventory of Depression Symptomatology (QIDS) and the Hamilton Depression Rating scale as well as sociodemographic features, previous diagnoses and antidepressants the patient had taken, and the first 100 items on a psychiatric diagnostic symptom questionnaire.

The three best predictors of treatment success were current employment, years of education, and loss of insight into their depressive condition. The three best predictors of treatment failure were baseline depression severity, feeling restless, and reduced energy level.

The tool predicted treatment outcome with 60 percent accuracy in an independent data sample—far better than clinicians. The research group has published an online tool to predict a patient’s likelihood of success with citalopram.

This single tool is unlikely to be the answer, but it is a harbinger of data science for psychiatry. We are beginning to approach the brain as a computational organ, one to be evaluated with measurements and calculations.

Calculators of disease risk are regularly used in medicine—if you have atrial fibrillation and go to a cardiologist, she will use multiple datapoints to calculate your risk of stroke, known as a CHAD-VASC score. Depending on your risk, she might prescribe you an anticoagulant like Coumadin.

The CHAD-VASC calculator is freely available online and does not pretend to be a perfect assessment of risk. It is sometimes wrong. But it is our medical community’s best approximation of your stroke risk if you have atrial fibrillation. The calculator is not a vote of no confidence in the cardiologist’s ability. Rather, like all empirical tests, it signifies that decisions based on more data are better than those based on less.

Psychiatry remains an outlier in the medical profession regarding the use of data; even after the rigorous Osheroff v. Chestnut Lodge debate, the importance of data in practice remains unsettled. In particular, objective data and data science remain underutilized by the psychiatric community. Has your therapist ever used a predictive algorithm to guide your treatment?

As Harvard Psychiatrists John Torous and Justin Baker recently wrote in JAMA Psychiatry, “Data science and technology can provide a nearly limitless set of decision-support and self-monitoring tools. However, without individual psychiatrists and the field at large making a concerted push to drive the technology forward…these advances will likely fail to transform our troubled system of care.”

The concern is that psychiatry lacks the will to apply what is known to what is practiced. Osheroff all over again.

“The scientific knowledge base is already
in place to radically improve the clinical practice of psychiatry,” Vinogradov asserted, “what we need next is the collective vision.”

Share
About author / Daniel

I was born in Dallas and spent my childhood scampering through the countrysides of central and eastern Texas, with brief escapades in Maryland and Utah. I began medical school in San Antonio, where I met my wife and future psych co-resident Kristin Budde. After my PhD, we moved together to New Haven, where I finished med school. I enjoy writing about neuroscience as a way to think through some of the problems that come up in clinic. I spend a great chunk of my time thinking about and researching how to develop useful biomarkers of brain disease. When I'm not at the hospital or working on research stuff, I'll be fixing up my 1920s New England house. I just recently refinished an old Blue Jay sailboat, which was a great new dad project (sanding is a good activity when you're sleep deprived).

Leave a reply
Bitnami