Impact of a Telepsychiatry Program at Emergency Departments Statewide on the Quality, Utilization, and Costs of Mental Health Services

Emergency department (ED) use for routine care strains the health system by increasing costs, raising pressures to admit patients for inpatient care, and reducing the ED’s capacity to care for true emergencies (1). Persons with mental disorders account for a large and growing portion of ED visits in the United States. Between 1992 and 2001, there was a 29% increase in the proportion of ED visits related to mental health (2). Estimates from nationally representative data suggest that poor or fair mental health increases one’s odds of being a frequent user of EDs (≥4 visits) by a factor of 1.7 (3).

Persons seeking mental health services in EDs face challenges in obtaining both high-quality care in the initial ED visit and follow-up care after discharge. A 2005 survey by the American College of Emergency Physicians found that ED personnel feel unequipped and insufficiently trained to address mental health issues (4). Lack of access to mental health personnel limits ED capacity to manage complex behavioral problems, which has consequences for the patient and the system.

To address these issues, the South Carolina Department of Mental Health (SCDMH) launched a telepsychiatry intervention program in 2009. Previous research had demonstrated that only 32% of South Carolina’s EDs had a psychiatric emergency service (5). The overarching goal of this initiative was to ensure that EDs can reach an available psychiatrist whenever needed. The program uses telecommunications technology to provide emergency facilities with access to a larger pool of psychiatrists for consultation, addressing physician and psychiatrist shortages that had limited access to psychiatric services in many of the state’s EDs.

In this study, we evaluated the impact of the SCDMH telepsychiatry intervention by comparing outcomes among the telepsychiatry recipients and a group of patients who had identical mental health conditions and demographic characteristics and who were treated in hospitals that did not have telepsychiatry. Our hypotheses were that the intervention would increase access to outpatient follow-up care while reducing overall hospital-based service use through improved triage and mental health treatment.

Methods

Telepsychiatry Program Description

In 2009, 18 hospitals enrolled their EDs in a telepsychiatry program coordinated by the SCDMH. Participating hospital EDs were recruited by personal communication and by distributing DVDs that explained the videoconferencing system, the training and credentials of the telepsychiatrists, and the goal of the collaborative treatment. Members of the project’s leadership team conducted site visits to provide opportunities to discuss the program and resolve any concerns with ED staff.

The telepsychiatry program employs six full-time telepsychiatrists and one part-time telepsychiatrist, all supervised by a lead psychiatrist. From a dedicated telepsychiatry hub in Columbia, this staff provides 24-hour coverage, seven days per week, for all of the participating EDs in the state. Each telepsychiatrist is board certified in the state of South Carolina, has at least one year of clinical experience in emergency psychiatry, and is credentialed and privileged in accordance with Joint Commission standards.

Telepsychiatrists were recruited throughout the state by contacting schools of medicine. DVDs are used for recruitment and training. The recruitment DVD discusses advantages and disadvantages (such as isolation and working at night and on weekends and holidays) of being a telepsychiatrist, explains how the system worked, and explains privileges and credentialing requirements for different hospitals and availability of the lead psychiatrist. Once recruited, telepsychiatrists are required to complete training DVDs containing six hours of clinical review, which were supplemented with handouts, before they began seeing telepsychiatry patients. The training program addresses how to adapt basic ED psychiatry skills to a distance telepsychiatry intervention modality.

The course was developed on the basis of standard competencies for telepsychiatry described by Shore and colleagues (6). Topics included suicide assessment; addressing violent or agitated patients; overdose; medical issues; substance abuse; management of emergencies in special populations (children and geriatric); and medical-legal issues, including civil commitment and duty-to-warn regulations (6). The training also addresses how to use telehealth equipment, document the case via electronic medical record, and remotely access laboratory values and mental health clinic databases. Subsequently, peer review was conducted every two weeks for the telepsychiatrist’s first three months; the lead psychiatrist met with other physicians to discuss the telepsychiatrist’s consultation, diagnosis, and treatment recommendations.

The ED telepsychiatry system used real-time video telecommunications bandwidth to transmit information without loss or distortion. Information technology support, also under the purview of the steering committee, ensured that the telepsychiatry equipment was optimized for picture quality and sound to facilitate the best clinical experience.

Telepsychiatry ED Visit Process

When a psychiatric consultation was needed at participating EDs, the ED triage nurse or physician on duty completed an online ED intake form, describing the patient’s diagnoses, address, current lab values, vital signs, and reason for the consultation. In addition to the information from the ED intake form, consulting telepsychiatrists had information from the medical health record, which provides real-time information on diagnoses, general medical health issues, past emergency behavioral services, and information on filled prescriptions.

Prior to the patient interview, the telepsychiatrist reviewed the ED intake form and information from the health record and then conducted a standard history and mental status examination with the patient via teleconferencing. Standard lab tests, such as CBC and electrolytes and a substance abuse screen, were recommended for all patients. Family members or caretakers were allowed to participate in visits with the patient’s permission; both the ED physician and the telepsychiatrist involved families whenever possible. For visits occurring during regular hours of work, the psychiatrist obtained permission to contact the patient’s primary mental health care provider (if applicable) to obtain information about the patient’s current mental health treatment and health status.

The information collected was synthesized into a treatment plan that included recommendations for acute management in the ED. Upon assessment completion, the telepsychiatrist transmitted a signed consultation recommendation to the ED. After the on-site ED staff instituted the treatment plan, the telepsychiatrist could have had a second visit or consultation with the patient if warranted. Generally a second visit was used to reevaluate any subsequent lab values and notes written in the ED. On the basis of any new information, the telepsychiatrist helped set a disposition plan with the ED and outpatient mental health team. The telepsychiatrist was then responsible for providing a full written assessment along with initial medication treatment recommendations, which were included as part of the patient’s chart after admission.

If the plan was for discharge to the community, the psychiatrist notified the facility that was the usual source of the patient’s mental health care. If the patient did not have a usual source of mental health care, the telepsychiatrist worked with the ED team to schedule an appointment for the patient at the nearest public facility that accepted the patient’s insurance, for example, a community mental health center. The telepsychiatrist’s formal consultation form included the following: multiaxial diagnosis and reason for presentation, history and mental status examination, course of care and treatment during the ED stay, medications on discharge, recommendations for ongoing psychopharmacological treatment, recommendations for other psychological or rehabilitative therapies, and recommendations for any social or logistical support (for example, transportation) needed to ensure appropriate access to outpatient services.

Statistical Analysis

The data for patients who received telepsychiatry services were linked to patient data from the South Carolina Office of Research and Statistics (ORS) data warehouse to assess health care utilization. State legislation requiring reporting of both private-sector and public-sector client-level data, including all-payer health care claims data, ensures that the data are comprehensive and complete. The data warehouse includes information from private-sector systems, such as hospitals, surgery centers, and home health; state agency systems, such as Medicaid and mental health, substance abuse, and criminal justice systems; and not-for-profit systems, such as free clinics and community health centers. Data usage approvals for the project were obtained from providers who contribute data to the warehouse. All data linkages were performed by ORS, and a deidentified data set was used for the analyses.

From the ORS administrative claims database, we also created a matched sample of patients who were seen in EDs in South Carolina that did not participate in the telepsychiatry intervention. We used an exact-matching approach to match the patients on the basis of age at the time of visit, sex, race, psychiatric diagnosis, whether they had visited an ED for a psychiatric condition in the preceding year, and whether the index visit occurred on a weekend. In matching samples, visits to nonparticipating EDs were considered only if the patient had a psychiatric diagnosis in the first two diagnosis fields on the administrative claim record. The exact-matching approach and the restriction on the fields used to document a psychiatric diagnosis were employed to ensure that the patients’ psychiatric treatment needs were similar to those of patients who were treated using telepsychiatry.

We performed both bivariate and adjusted regression analyses to test for differences in several resource use and quality outcomes. These outcomes included outpatient psychiatric follow-up at 30 and 90 days (an outpatient visit to any provider for psychiatric care, excluding emergency departments); probability of hospital admission from the emergency department, including transfer to another facility, and whether the patient was admitted within 24 hours; length of inpatient stay if admitted; all inpatient charges; and all hospital-based charges (ED and inpatient) incurred in treating patients within 30 days of the initial hospital visit in which they received telepsychiatry or—for control group patients—were identified as a match.

We used logistic regression models to examine outpatient follow-up at 30 and 90 days after the index visit, and these models included hospital random effects to account for the hierarchical nature of the data. We modeled inpatient admission and length of stay and inpatient cost models by using a two-part regression modeling approach. This approach has become standard in the health services literature for outcomes with a high proportion of zeros (7–9). When examining admission and length of stay, the first part of our two-part model was a logistic regression predicting admission immediately subsequent to the ED visit, and the second part was modeled by using a Poisson distribution with a log-canonical link. When inpatient costs were examined, the first part of the model was the same, but the second part was modeled by using a generalized linear model with a gamma distribution and a log link (9). When looking at total 30-day total health care costs, we simply used a generalized linear model with a gamma distribution and log link because there were no zeros, given that at a minimum, each patient incurred costs for the initial ED visit (9).

All regression models were adjusted for whether the visit was a weekday or weekend visit. Standard errors in all models were clustered at the hospital level to account for potential intracluster correlation, which could give rise to incorrect statistical inference. All analyses were performed by using Stata, version 13. This study was reviewed and approved by the institutional review boards at the University of South Carolina and Emory University.

Results

Between March 2009 and June 2013, there were 9,066 patients with at least one telepsychiatry visit. For 7,261 of these patients, we were able to successfully find a matching patient with a psychiatry-related visit to an ED at a nonintervention hospital in South Carolina. For patients with more than one telepsychiatry visit, only the first visit was matched (index telepsychiatry visit). The descriptive characteristics of the telepsychiatry patients for whom we were able to identify matches (N=7,261), the matched control group (N=7,261), and the telepsychiatry patients for whom we were unable to identify matches (N=1,805) are shown in Table 1. The average age of participants in the matched sample was 35.7±14.9 years; 49.8% were female, 73.1% were white, 23.8% were black, and 38.6% of the admissions occurred on a weekend. In the unmatched group, the average age was 37.5±13.6 years, 47.5% were female, 71.1% were white, 28.0% were black, and 38.2% of the admissions occurred on a weekend.

Bivariate results for the outcomes in the matched sample are shown in Table 2. Telepsychiatry recipients were more likely than the matched control group to receive 30-day outpatient follow-up (46% versus 16%, p<.001) and 90-day outpatient follow-up (54% versus 20%, p<.001). Telepsychiatry recipients were less likely than the matched control group to be admitted (11% versus 22%, p<.001). Total 30-day inpatient charges were lower among telepsychiatry recipients compared with the matched control group ($8,290 versus $11,224, p<.001). Total 30-day health care charges were also lower among telepsychiatry recipients versus the matched control group ($12,634 versus $14,052, p=.001).

Adjusted results for the utilization outcomes in the matched sample are shown in Table 3. The models were adjusted for patient age, sex, and race and included hospital random effects. The odds of 30-day outpatient follow-up were higher among telepsychiatry recipients compared with the matched control group (odds ratio [OR]=5.44, p<.001). Similarly, the odds of 90-day outpatient follow-up were higher among telepsychiatry recipients compared with the matched control group (OR=5.65, p<.001).

As noted above, the inpatient length of stay for the index visit was estimated by using a two-part model. Telepsychiatry recipients were less likely to be admitted at the index visit compared with the matched control group (OR=.41, p=.022). Length of stay, which was conditional on being admitted in the two-part model, was .43 days lower (p=.003) among telepsychiatry recipients versus the matched control group. Thus the combined effect of having a telepsychiatry consultation during the index ED visit was a reduction of .86 days in length of stay following inpatient admission.

Inpatient costs in the 30 days following the ED index visit were $2,338 lower (p=.041) among telepsychiatry recipients versus the matched control group (Table 4). However, the total health care costs of the two groups in the 30 days following the index visit were not statistically different, although the point estimate of the effect was negative.

Supplemental analyses comparing all 9,066 patients with the matched control group revealed little difference from the estimates noted here. [The analyses are available as an online supplement to this article.]

Discussion

In this study, we tested the hypothesis that a statewide, ED-based telepsychiatry program would result in better rates of outpatient follow-up and reduced inpatient service use and cost. Specifically, we examined whether recipients of a telepsychiatry consultation at participating EDs were more likely to receive follow-up outpatient care and have lower hospitalization rates and 30-day costs of care compared with a matched sample of individuals treated in nonparticipating EDs. We found that telepsychiatry recipients were more likely to get follow-up, were less likely to be admitted to the inpatient setting at their ED visit, had a shorter length of stay if admitted, and had lower overall 30-day inpatient costs.

These findings are important for two major reasons. First, from a systemwide perspective, persons with mental disorders account for a large and growing portion of ED visits in the United States (2). Lack of access to mental health personnel may particularly limit an ED’s capacity to manage complex behavioral problems and ensure effective mental health follow-up after discharge (1,5). Our results suggest that telepsychiatry may be an effective way to address this problem in the ED.

Second, and arguably more important for patients in need of mental health care, shortages of psychiatric practitioners and lack of expertise in mental health at EDs may reduce opportunities for appropriate evaluation and follow-up (4). Given challenges in recruiting mental health providers to live and work in underserved and rural communities, telepsychiatry has been proposed as a promising approach to expanding the capacity of EDs to treat mental disorders (6,10). However, little research has examined the potential impact of emergency telepsychiatry on costs and outcomes of care. A recent review of the telepsychiatry literature (11) found only one case study (12) and one state foundation report (13) describing telepsychiatry programs in EDs; each found that the programs could be feasible and acceptable to clinicians and patients but did not report clinical or health services outcomes of the programs. Our findings indicate that ED-based telepsychiatry has the potential to improve the quality of emergency mental health care and reduce, or at most yield no difference, in health care resource use.

Our study had limitations. The intervention was not randomized by patient, nor were hospitals randomized to participation. We sought to address this issue by selecting a control group whose members were matched on key characteristics; however, the control group may have differed from the intervention group on the basis of unmeasured factors. In addition, there may be hospital characteristics that both drive participation in telepsychiatry and lead to better performance and lower costs, which could bias our estimates. We assessed resource use by calculating utilization of services and charges for services, and the latter are typically higher than actual costs. Charges for the telepsychiatry consult were not available, although it is unlikely that they would have fully offset the inpatient cost or resulted in statistically higher health care charges in the 30 days following an index ED visit. In addition, there may be other types of unbilled care that we failed to capture. Finally, we were unable to match 1,805 of the 9,066 telepsychiatry patients to a patient treated at a nonparticipating hospital on the basis of our preferred method. When we included these individuals in sensitivity analysis of adjusted models, the results were qualitatively the same and quantitatively very similar, except the difference in inpatient costs was smaller by $752, which was not statistically distinguishable from zero.

Notwithstanding these limitations, how was this ED-based telepsychiatry intervention able to reduce both the probability of admission and the length of stay among admitted patients?

Recent research from the general medical field suggests that rather straightforward changes in the organization and delivery of ED-based care can lead to reduced inpatient admissions and more rapid care delivery, thus reaping substantial savings among general medicine (as opposed to surgical) patients (14). Much of this change involves treating the right patient with the right service at the right time through the adoption of dedicated protocol-driven services (14). Telepsychiatry represents a similar change in thinking regarding the process of care for patients in the ED. The program was centrally controlled and thus allowed for some degree of treatment protocol use among providers. Also, the telepsychiatry program represented a dedicated service, with physical space carved out and specific information technology resources made available at each participating hospital, thus emphasizing an organizational commitment to improving care. Finally, the program took advantage of economies of scale and of knowledge of system resources that was embedded in the coordinating agency—in this case, the SCDMH.

Conclusions

We examined the potential for a statewide, ED-based telepsychiatry program to improve the quality of emergency mental health care and reduce resource use in South Carolina. We found that having a telepsychiatry visit increased outpatient follow-up and may have even reduced resource use. More research is needed into the potential for telepsychiatry to improve ED-based behavioral care, the sustainability of such initiatives, and key implementation issues.