Skip to main content

Transcript: Certain Disorders, Education Level Predict High Utilization of Inpatient Psychiatric Services

April 07, 2021

In a podcast for the Psychiatry & Behavioral Health Learning Network, Jane Hamilton, PhD, MPH, LCSW, and David S. Buck, MD, MPH, discussed their research on the factors that increase the likelihood of becoming a “high utilizer” of inpatient psychiatric hospital services.

Read the transcript of the podcast below.

Dr. Jane Hamilton: My name is Jane Hamilton. I'm an Assistant Professor at the University of Texas Health Science Center at Houston's McGovern Medical School. I'm in the Louis A. Faillace Department of Psychiatry and Behavioral Sciences, where I conduct mental health disparities and outcomes research.

I've been conducting research at the University of Texas Harris County Psychiatric Center since 2013, focusing on service delivery for vulnerable populations and examining high utilization and trying to gain knowledge on how to develop better care transitions to community‑based services.

Dr. David Buck: Hi. I'm Dave Buck, Associate Dean for Community Health at the University of Houston College of Medicine, Texas, and a practicing family physician and have worked with the homeless since 1984. My main focus has been working developing care systems for the homeless and in community engagement in general.

We're thrilled to be here. The context for this paper emerges from one of the biggest challenges I think we have in healthcare today, the $3.8 trillion that's spent in healthcare. Half of the money, or $1.9 trillion, is spent by five percent of the population. They experience some of the worst health outcomes of any population group in the developed world.

We want to understand, “what is the cause of these health outcomes?” We need to understand the needs and utilization and service delivery for this subgroup, perhaps the most challenging subgroup with severe mental illness, and how we can begin to improve their care.

Dr. Hamilton: Our study addresses a gap in the research by examining the relative predictive importance of a comprehensive set of factors that have been extracted from electronic health record at the psychiatric hospital where the study was conducted.

We wanted to identify factors that were most likely to predict high utilization, both in a single year and over multiple years, to gain more knowledge, to tailor interventions, and to improve outcomes for this population.

Our study population included all individuals ages 18 and older that were treated at the psychiatric safety net hospital from 2014 to 2016 in Harris County, Texas. Our study included 9,840 unique patients, which included 15,558 visits.

The majority of the patients treated at the safety net psychiatric hospital have higher rates of adverse social determinants of health that have been previously associated with high utilization in inpatient psychiatric hospitals.

For our methods, we looked at the outcome, as I said, of single versus multi‑year high utilization. That's three or more psychiatric hospitalizations within a 365‑day period. We used a machine learning method called the elastic net algorithm because it's been showed to be more stable in estimating the strength of a predictor's relationship to the health outcome.

Our most significant findings in this study were three factors that were found to be highly predictive of high utilization within 365 days. The first was having a diagnosis of schizophrenia.

The second finding, the second significant predictor, was having a co‑occurring personality disorder diagnosis along with a primary psychiatric diagnosis of either schizophrenia, major depression, or bipolar disorder. Third, having less than a high school education.

These findings were what we expected overall, for several reasons.

Both schizophrenia and personality disorders can be very difficult to treat. Many patients with these diagnoses are disadvantaged and vulnerable to health disparities. We feel that we really need to link people with these conditions to more evidence‑based treatment to help avoid repeat hospitalizations.

This study is a great first step in helping to identify the appropriate outpatient resources to help these patients remain stable in the community and avoid repeat hospitalizations.

Also, learning that people with less than a high school education was also a significant finding because we have often suspected that there were major health literacy issues among this population.

Dr. Buck has been studying this for years. He's been a strong advocate to improve the way we discharge patients from the psychiatric hospital. Oftentimes, they're given complex discharge instructions written on pieces of paper with small print that they can't read. We often suspected that health literacy barriers might be driving hospitalization.

Finding that limited education was a significant predictor has provided more evidence and really points to a need for us to do a better job of addressing health literacy barriers through interventions like teach‑back, shared decision‑making, providing more appropriate written feedback at discharge, and other sorts of strategies to improve health literacy.

Dr. Buck, do you have anything you might want to add?

Dr. Buck: Yes. One of the studies we did that you alluded to was looking at the discharges that result from our county jail. In those discharges, people immediately, upon discharge, would lapse in their care because they ran out of their medication. They're discharged at 12:01 in the morning, so they don't get their medications for the next day and they already are lapsing in their care.

Just by linking people to services on the day of discharge reduced rearrest rates by 62%. Part of what we wanted to do in this study is to look at what are these service factors and coordination of the right resource that might be able to help.

Dr. Hamilton: Absolutely. Dr. Buck and I study social determinants of health. What we are learning is that many high utilizers have complex health needs. They're more likely to be from socially disadvantaged groups. They have limited access to community‑based healthcare and social services.

Through our future research, we really hope to identify the right mix of services and the right kind of services to help this population improve in their outcomes.

Really, the outcomes that we're looking for are social determinants of health, such as having better access to income and benefits, better access to primary care, better access to community‑based mental healthcare, as well as housing stability. We believe that these factors, once achieved by the high utilizer population, will result in fewer hospitalizations.

Dr. Buck: Some of the future work that could be directed as a result of this include looking at pathways to care, how we could begin to improve specific pathways, say, for those with co‑occurrings, behavioral health and a severe mental illness or comorbid medical and psychiatric conditions, and then also concomitant social factors.

Looking at COVID-19, since that's a relevant factor today, the number one risk factor are people that are over 25 years of age but did not graduate from high school. Can we begin to look at different people and their risks and target interventions based on the cluster of sociodemographics and risks and be more effective?

Building these pathways takes time. It's really one at a time, trying to build it, refine it, and ultimately to scale it.

Dr. Hamilton: At the hospital where we conduct research, approximately 9000 patients are served per year. Among these, a third are experiencing homelessness. One of the new interventions in our region is to provide a post‑discharge psychiatric respite.

In contrast to discharging the patients experiencing homelessness to the streets—really to shelters, not the streets—there is a new intervention in place to provide residential rehabilitation following psychiatric hospital discharge.

Dr. Buck and I have approached the county. We're looking at trying to conduct a rigorous evaluation of this intervention to see if this particular care transition works and there are certain pathways that are really meaningful for this population to improve outcomes.

Another outcome of our study is that we've developed a framework for analyzing electronic health record data. What we're hoping is to develop collaborations across the country and to conduct work with colleagues in other safety net hospitals and psychiatric hospitals, to see if we can find similar predictors or ways that we can leverage electronic health record data to tailor treatment.

One of the benefits of this study is that psychiatric hospitals are historically have limited resources to conduct research and to implement new interventions. A benefit of our study is we leveraged existing data to conduct research. We didn't have to conduct new assessments. We didn't have to hire additional staff to go out onto the units and assess people.

If this turns out to be beneficial for the Harris County Psychiatric Center, this is a study that we hope to replicate in other regions to show the effectiveness of using electronic health record data to improve patient outcomes, specifically in psychiatric care, where this type of study is less utilized compared to acute care hospitals.


Hamilton J, Suchting R, Buck DS, et al. Determinants of high psychiatric utilization at a large urban safety-net hospital. Journal of Health Care for the Poor and Underserved. 2021;32(1):296-320.

Back to Top