Multi-Agent-Based Simulation of a Complex Ecosystem of Mental Health Care

J Med Syst. 2016;40:39. (doi:10.1007/s10916-015-0374-4)

Agent-based models simulate the actions of individuals (“agents”) in order to get an understanding of their effects on a system.1 Although this kind of modeling has been used to inform decision making in physical health care,2 its use in mental health systems is lacking.

A project undertaken jointly by Otsuka and IBM focused on care coordination technology and used agent-based simulation modeling to see the changes that resulted from using that technology. Their project, which attracted the attention of the US House of Representatives,3 looks at how technology aimed at care coordination, and the process flows associated with it, can reduce system fragmentation and ideally prevent patient crisis situations.

The data input into the model were gathered from publicly available statistics, as well as from more than 125 hours of interviews with a variety of stakeholders, including medical personnel, criminal justice professionals, and care services personnel. The model created “patients” using randomly determined characteristics. Each patient in the model had a minimum score for mental health state and an initial score for physical health (1–100 range; the mental health score is similar to the Global Assessment of Functioning). The way in which the mental and physical health scores interact demonstrates the way in which treatment for one can impact the other.

As a given patient progresses through decision flows in the model, the patient moves between physical environments (home, prison, hospital, homelessness) based on probabilities that vary by mental state. Care transitions and changes in physical setting can, in turn, improve or worsen the patient’s mental state. Financial variables such as costs to the patient, and the ability to pay those costs, factor into the model as well. The variety of medications, services, and health care providers that a patient may encounter are rated according to their “cost to the patient,” which affects whether a patient is likely to engage with them. Government and other funding sources are incorporated, since they influence the patient’s access to resources such as care providers, medicines, and transportation.

The model was employed specifically to find out how care coordination technologies impact system performance in terms of (1) patient transfer between providers and (2) patients’ level of compliance with appointments.

An aim of care coordination technology is to facilitate successful transfer of patients between healthcare providers, or “handoffs.” The technologies can improve coordination, increase sharing of information about patients, and allow facilities to reach out to new patients. Successful handoff of patients can prevent their disappearance from the system, and the probable crisis points that would follow. The model used by Kalton and colleagues found that as successful handoffs increased, so would the percentage of patients compliant with their medication (Figure 1). Increase in the rate of successful handoffs was also associated with decrease in the costs of arrests and hospitalizations.

Figure 1. Effect of Improvement in Handoff Success Rate

As Handoff Success rates increase by 5% to 15%:

Medication compliance (patients on medication)
  improves by 3.5% to 5.5%

Hospitalization costs decrease by 2% to 5%
  Arrest costs decrease by 5% to 10%

Another aim of care coordination technology is to improve appointment compliance via reminders, improved planning processes, and transportation coordination. The model indicated that greater appointment compliance was associated with a greater percentage of patients living in private residences—versus a prison or hospital (Figure 2). Further, as appointment compliance increased, costs of arrests and hospitalizations decreased.

Figure 2. Effect of Improvement in Patient Appointment Compliance

As Appointment Compliance rates increase by 3% to 12%:

Number of patients living in private residence
  improves by 0.25% to 0.75%

Hospitalization costs decrease by 2% to 6%
  Arrest costs decrease by up to 6%

Kalton and colleagues’ project highlights the range of variables that affect how patients move through mental health ecosystems, and it also shows that increased coordination among care providers is likely to result in measurable improvements. As new technologies are introduced, the authors intend to conduct subsequent analyses that evaluate actual impact versus predictions.

References

  1. Bonabeau E. Agent-based modeling: methods and techniques for simulating human systems. Proc Natl Acad Sci U S A. 2002;99(suppl 3):7280–7287. doi:10.1073/pnas.082080899 PubMed
  2. Gill S, Paranjape R. A review of recent contribution in agent-based healthcare modeling. In: Rodrigues J, ed. Health Information Systems: Concepts, Methodologies, Tools, and Applications. Hershey, PA: IGI Global; 2010:356–373. doi:10.4018/978-1-60566-988-5.ch024
  3. Statement of Judge Steve Leifman, Chair, Supreme Court of Florida Task Force on Substance Abuse and Mental Health Issues in the Courts, before the Subcommittee on Oversight and Investigations of the Energy and Commerce Committee of the United States House Of Representatives, concerning People with Mental Illnesses Involved in the Criminal Justice System. March 16, 2014. US House of Representatives Document Repository Web site. http://docs.house.gov/meetings/IF/IF02/
    20140326/101980/
    HHRG-113-IF02-Wstate-LeifmanS-20140326.pdf
    . Accessed May 3, 2016.