Integrating Behavioral Health in Primary Care Using Lean Workflow Analysis: A Case Study

Abstract

Background

Primary care offices are integrating behavioral health (BH) clinicians into their practices. Implementing such a change is complex, difficult, and time consuming. Lean workflow analysis may be an efficient, effective, and acceptable method for integration.

Objective

Observe BH integration into primary care and measure its impact.

Design

Prospective, mixed methods case study in a primary care practice.

Measurements

Change in treatment initiation (referrals generating BH visits within the system). Secondary measures: primary care visits resulting in BH referrals, referrals resulting in scheduled appointments, time from referral to scheduled appointment, and time from referral to first visit. Providers and staff were surveyed on the Lean method.

Results

Referrals increased from 23 to 37/1000 visits (P<.001). Referrals resulted in more scheduled (60% to 74%, P<.001) and arrived visits (44% to 53%, P=.025). Time from referral to first scheduled visit decreased (Hazard Ratio (HR) 1.60; 95% Confidence Interval (CI) 1.37, 1.88; P<0.001) as did time to first arrived visit (HR 1.36; 95% CI 1.14, 1.62; P=0.001). Surveys and comments were positive.

Conclusions

This pilot integration of BH showed significant improvements in treatment initiation and other measures. Strengths of Lean included workflow improvement, system perspective, and project success. Further evaluation is indicated.

INTRODUCTION

Primary care offices are struggling to meet the behavioral health (BH) needs of their patients. Forty percent of primary care patients need behavioral health services¹ and 43–60% of those patients are treated solely in primary medicine^2,3,4. Of those referred to specialty mental health services by their primary care providers, 50–90% fail to receive services⁵. A survey of 6,600 primary care physicians found that appropriate mental health services are the most difficult of sub-specialties to access⁶. Most external referrals for specialty mental health and substance abuse services never generate an appointment; most patients who make an appointment never initiate care⁷. A strategy to meet these needs has emerged over the past 20 years in the form of behavioral health services integrated with and co-located in primary care practices⁸.

Integration of behavioral health is effective in improving medical outcomes for patients with acute or chronic diseases as well as responding to mental health and substance use issues that are generally untreated or inadequately treated in primary care⁹. However, a recent survey of all patient centered medical homes certified by the National Committee for Quality Assurance found that less than 40% have any mental health/substance abuse/behavioral health (BH) clinicians as part of their practices. When available, those clinicians are generally not integrated into practice flow⁹. Although the operational and clinical components needed to integrate BH care are well documented¹⁰ (see Table 1, Primary Care Behavioral Health Intervention Components), lack of knowledge about how to implement BH services into primary care is a major obstacle¹¹. Implementation must address several key issues, including how deeply the BH clinician is involved in the practice and the processes of care, strategies for organizational and financial barriers, and opportunities to use health information technology^12,13.

Table 1.

Primary Care Behavioral Health Intervention Components (per Hunter CL & Goodie JL, 2010)

Clinical	Full time on-site primary care behavioral health clinician (1 per 7500 patients23) Clinician availability for personal, face-to-face introductions (“warm handoffs”) and consultation Brief evidence-supported treatment interventions; other clinical care responsibilities Intensive training of primary care behavioral health clinicians, using standardized protocols for a broad range of psychological and medical problems amenable to behavioral health treatment Population (panel) based care using measurement-based, stepped treatment and other resources
Operational	Screening for mental health, substance abuse, and health behavior issues and provider decision support seamlessly integrated into patient flow Reengineering of practice processes, e.g. “warm handoffs”, automated scheduling, referrals, etc. Training providers and staff in behavioral care procedures Appointment frequency and interval of behavioral health clinician consistent with primary care Shared, transparent EHR with two-way notes and access to information Care management coordination of referrals and information with specialty care as needed
Finance	Brief interventions, which are lower cost services, provided over shorter episodes of care Coordination of services and finances to optimize sustainability Negotiation of appropriate reimbursement Regular reports of performance, RVU and financial data

Lean (also known as Lean Manufacturing¹⁴ or Lean Thinking¹⁵) uses a tightly structured approach to analyze workflow and is known to produce sustained, effective change in large health care organizations^16,17. Lean is conducted by a front-line team representing different work functions. The team produces a systems diagram that leads to problem identification, root cause analysis, and agreement on redesigned office systems and processes¹⁴. Team members collaborate to develop a report and implementation plan, using a pre-specified format that incorporates patient and practice data. The team is responsible for producing a redesigned systems diagram, along with planned measures and expected counter-measures for anticipated problems. The team develops strategies for adapting to organizational culture and completes an action plan, with defined deliverables and dates. The Lean approach was modified for use in a primary care office practice, using a relatively brief period of time (eight hours) for team meetings that were organized around provider and staff scheduling constraints¹⁸.

Earlier efforts to integrate BH in primary care¹⁹ suggested the need for a systematic approach to implementation. Therefore, we sought to observe the process of integration into primary care and measure its impact on referrals, treatment initiation rate, efficiency, and acceptance.

BACKGROUND

Innovations in health care should fit the local context of each unique organization²⁰. This case study example of integration took place in an academic internal medicine practice in northern Vermont. In 2010–11, the practice provided 21,200 primary care patient visits/year to cover 7,500 patients who were served by 12 attending providers totaling 6 FTEs, 36 residents (9 FTEs), and 42 staff (12 clinical and 30 non-clinical; 27 FTEs). The practice provided care six days/week, 47 hours/week for both walk-ins and scheduled visits.

The practice served adults. Eighty percent were 19 to 64 years, 19% age 65 and over, and 1% 18 years or less. Most (63%) were covered by commercial insurance, with another 22% covered by Medicare, 12% Medicaid, and 3% self-pay, workers’ compensation, or other public agencies. The top eight diagnoses/conditions for these patients were hypertension, hyperlipidemia, back pain, depression, diabetes, anxiety, cough, and arthritis.

The practice had no integration of case or care managers, clinical pharmacists, or psychiatrists. Adjacent to the practice, a Master’s level psychologist provided long-term psychotherapy for mental health issues only. The first integrated behavioral health clinician was introduced to the practice immediately after the completion of this integration project, working full-time (40 hours/week) from start up.

All clinicians in the practice were paid on a salary basis; most were assigned to 50% or more of clinic time and demonstrated a high degree of commitment to the success of this organization. Although no financial incentives were provided for improving care quality or exploring innovations, providers demonstrated a long-term commitment to quality improvement projects in the practice¹⁸. All clinicians were expected to generate revenues above their salary expense. The integration of primary care was expected to improve care quality by increasing access to behavioral health services and decreasing wait times for primary care providers, while also improving the practice’s finances.

Recent regulatory and reimbursement changes in Vermont also played a role in this study. Prior to 2010, state and national leaders began to explore integrated care as a new model for behavioral and primary care services through special tasks forces on workforce, reimbursement, and population-based service needs, some of which included one of the authors. State policy makers and leading local health insurers started to support changes in traditional behavioral health programs at this time, modestly supporting behavioral health interventions in adult primary care, obstetrics, neurology, and gastroenterology^7,21,22. These changes allowed primary care services to bill directly for behavioral health services using traditional psychotherapy codes and reduced prior approval requirements for these services. While an innovation at the time, such reimbursement procedures using standard codes are now common.

METHODS

Study Design

This prospective, mixed methods case study observed a single primary care practice as it used Lean to integrate a new behavioral health services.

The Intervention

The model of integration of behavioral health services suggests placing a full time behavioral health clinician (BHC) inside the clinic for every 7,500 patients²³ to work alongside primary care providers. The BHC, a master’s level provider or a Ph.D. psychologist, receives referrals for patients with specific behavioral health needs and, when possible, meets them during the visit at which the referral is made (a “warm handoff”²¹). The BHC consults directly with the provider on co-existing medical conditions and provides a bridge to community mental health specialists. After conducting an assessment, the BHC may handoff the care to other specialists or may fully engage the patient in treatment for a defined period of time. The BHC documents patient care in the electronic health record (EHR), alongside other clinicians’ notes. Office systems support referral, scheduling, documentation, consults with providers, team education, and billing. The BHC’s scope of practice includes Cognitive Behavioral Therapy and other evidence-based behavioral interventions for depression, anxiety, insomnia, substance abuse, smoking cessation and other non-psychotic mental health and behavioral disorders. It also supports the diet, exercise, medication adherence, stress and behavioral aspects of medical conditions such as diabetes, lung disease, heart disease, asthma and others. The BHC can also provide acute stabilization and triaging to specialty mental health services.

A seven-member Quality Improvement (QI) team of physicians, mid-level providers, nurses, and office staff studied the current workflow from a patient’s first telephone call requesting a primary care appointment for an unstated behavioral health need to a completed behavioral health referral, i.e., the workflow without internal BH services available. The team, supported by a QI facilitator, created a systems diagram of this process, identifying problems with their current process and changes needed to integrate the BHC into the future patient care process²⁴. This workflow analysis was accomplished in 8 one-hour segments scheduled to accommodate members’ work schedules. The process was based on structured analysis, the requirement that all team members individually draw their own analytic diagrams and build on them from one meeting to the next, and supportive clinical and administrative leadership that championed the process with the team and the rest of the practice.

The team collectively redesigned its office system over the course of 8 meetings to include six specific changes and a new approach to provide a clinical service related to BH:

1. Direct scheduling for an appointment with the BHC based on triage nurse screening (without a physician referral);

2. Immediate face to face introduction of the BHC when a primary care provider (PCP) generates a mental health referral (the “warm handoff”);

3. Creation of an appointment with the BHC on the day that the referral is made;

4. Pre-authorization of insurance coverage for BH services by front office staff;

5. Expedited scheduling of external mental health visits; and

6. Coordination and follow-up communication by the BHC with external mental health specialists.

7. A new approach to patient care service in this practice: providing acute stabilization behavioral health services by the BHC when patients have to wait for mental health specialist services.

The Lean approach required the team to identify the needs of patients, clinicians and non-clinicians. The team members actively facilitated these recommendations by working with colleagues and managers inside the clinic and across the institution in monthly meetings and other communication forums.

Implementation

At the end of eight hours of team meetings, implementation moved forward through a series of faculty, resident, and staff meetings, each focused on engaging all members of the practice in a successful outcome. Residents, for example, received formal training in which they met regularly with the BHC in educational sessions and case presentations. Attending physicians worked with the BHC to develop a “playbook,” or written protocols, for two specific medical conditions (ADHD and chronic pain coupled with opioid therapy) to guide interactions with the BHC. The BHC, in turn, worked with medical providers to keep them aware of patient progress between their primary care visits and to support their understanding and use of the service. This iterative process of learning and fine-tuning continued through the implementation stage.

The BHC served patients with medical needs (e.g. smoking, headache, back pain, sleep disturbance) as well as those needing screening and referral for substance abuse and brief interventions for mental health issues resulting in referral to community providers. In the relatively few cases for which referral sources were not immediately available, the BHC provided acute care services beyond the 4–6 visits anticipated for most patients.

Data Collection and Review

An electronic health record (EHR) provided patient-specific data about referrals made during primary care appointments for the 12 months prior to BHC integration and five months following. The data extraction process sought to include all patients who were seen by a PCP in the target clinic during the study periods, all referrals for ambulatory BH, mental health, or substance abuse services, and all appointments for such services that followed the identified referrals for at least two years after the primary care visit. Referrals that were not followed by specialty appointments within the academic medical center are represented in the data analysis as patients with no treatment initiated, although some of these may have had treatment by community mental health specialists. Conversely, some specialty referrals may not appear in the EHR at all, especially in the pre-intervention period when robust office systems were not in place.

The primary outcome was treatment initiation rate, defined as the proportion of referrals resulting in ambulatory behavioral health visits. The EHR was also used to measure the proportion of primary care visits resulting in BH referrals, the proportion of BH referrals resulting in scheduled appointments, days from referral order date to scheduled BH appointment date, and the days from referral order date to first actual BH visit. Provider and staff surveys were conducted prior and subsequent to the intervention to measure perceptions of the QI method.

All practice members were verbally informed of this research study and seven individuals were invited to join the team based on role and formal or informal leadership within the practice, an opportunity they were free to accept or decline. The Committees on Human Research at the University of Vermont reviewed and approved the study protocol.

Pre- and post-intervention surveys were developed to collect confidential provider and staff opinions from all practice members, excepting residents due to timing reasons, on the degree to which the Lean process was acceptable, effective at changing office processes, and effective at changing the clinic’s outcome, e.g. more successful BH referrals. The authors used a model of technology acceptance to develop questions regarding the Lean QI method²⁵. Questions on workflow and clinical effectiveness were tested for face validity in preliminary trials of the survey. Survey scores were constructed on 7-point Likert scales. The pre-intervention survey was administered six months prior to BHC integration; the post-intervention survey was administered two months following BHC integration.

The results and conclusions of this study were presented to the participating providers and staff for confirmation and comments 10 months following BHC integration. Feedback was collected both orally and in writing on anonymous feedback forms. This feedback was used to re-review and confirm or modify conclusions.

Statistical Analysis

The authors used STATA 13.1 (Stata Corporation, College Station, Texas) for data management and descriptive statistics. These included comparisons across two time periods (pre-intervention and post-intervention) using Wilcoxon Rank Sum tests for ordinal data, Fisher’s Exact test for categorical data, and the log rank test for time to event data. All analyses were two-tailed with P<.05 required for statistical significance. Because 54 (8%) of the 652 referred patients received BH referrals in both periods, the analyses were repeated without these patients.

RESULTS

Practice Member Characteristics

Twelve providers and 30 staff were present in the practice at the start of the integration project, of which 17 responded to the study survey (40%). Eight months later, after integration had been completed, 12 providers and 23 staff were present, of which 14 responded (40%). (Table 2: Responding Practice Member Characteristics). There were no significant differences between these time periods in respondent age, gender, years employed in the practice, or proportion of provider respondents relative to staff respondents.

Table 2.

Responding Practice Member Characteristics, including providers and staff active in the practice in the Pre-Intervention Period (September-October 2010) and the Post-Intervention Period (April-May 2011). Response rates in both periods were 40%.

	Pre-Intervention	Post-Intervention	P-value
	Respondents (n=17)	Respondents (n=14)
Age, Mean (SD)	45 (11)	51 (7)	0.10
Female (%)	14 (82)	10 (59)	>0.99
Providers (%)	5 (29)	6 (35)	0.45
Years at the Practice, Mean (SD)	10 (7)	11 (8)	0.74

Patient Characteristics

The practice treated 7,515 unique patients in the pre-intervention time period (12 months) and 4,962 in the post-intervention period (5 months). Of these, 4,051 appeared in both time periods for a net total of 8,426 patients. 401 patients received BH referrals pre-intervention and 305 post-intervention; 54 patients received referrals in both time periods for a net total of 652 patients (Table 3: Patient Characteristics). Patients referred to BH services were more likely to be depressed than the general primary care population (78% compared to 38%), more anxious (62% compared to 28%), and visited their primary care practice more often (6.8 visits per patient per year compared to 2.9). As expected, these chronically ill workforce-aged adults were younger than the primary care population overall (45.6 years of age compared to 50.4) and had lower rates of commercial insurance (44% compared to 63%). In all other respects, the two populations were alike.

Table 3.

Patient Characteristics for all patients presenting for primary care visits and for those referred to behavioral health services over 17 Months (Feb 2010 – June 2011)

	Patients Presenting for Primary Care Visits	Patients Referred to Behavioral Health Services
Patients	8,426	652
Age, Mean (SD)	50.5 (16.9)	45.2 (16.6)
Female (%)	4,753 (56)	363 (56)
White (%)	7,709 (91)	589 (90)
Commercial Insurance (%)	5,273 (63)	284 (44)
Vermont Resident (%)	7,977 (95)	617 (95)
Chittenden County (%)	6,796 (81)	533 (82)
Behavioral Issues
Depression (%)	3,189 (38)	512 (79)
Anxiety (%)	2,390 (28)	403 (62)
Medical Co-Morbidities
Hypertension (%)	3,450 (41)	279 (43)
Diabetes (%)	1,172 (14)	106 (16)
Hyperlipidemia (%)	3,731 (44)	261 (40)
Arthritis (%)	1,979 (23)	173 (27)
Total #PCP Visits/Patient/Year (SD)	2.9 (2.5)	5.2 (4.3)

Patient Outcomes

The practice generated 480 BH referrals during 21,219 primary care visits prior to the intervention and 342 referrals during 9,180 visits post-intervention (Table 4: Results of Behavioral Integration Project). The referral rate increased from 23 referrals/1000 primary care visits to 37 (P<0.001) and the rate at which referrals resulted in scheduled mental health appointments increased from 60% to 74% (P<0.001). The rate of treatment initiation (scheduled appointments that the patient kept) increased from 44% to 53% (P=0.025). There were 54 duplicate patients in each period whose experiences cannot be considered independent of each other. When the analysis was repeated without these patients, referral scheduling increased from 59% to 73% (P=0.001) and treatment initiation increased from 44% to 54% (P=0.014). We retained the duplicate patients for the remainder of the analysis.

Table 4.

Results of Behavioral Integration Project for all patients presenting for primary care and those referred to behavioral health services before the intervention (Feb 2010 – Jan 2011) and after the intervention (Feb 2011 – June 2011).

Measures	Pre-Intervention	Post-Intervention	P-value
Duration	12 months	5 months
Primary care visits	21,219	9,180
Referrals for ambulatory MH/BH	480	342
Referrals per 1,000 primary care visits	23	37	<0.001
BH visits scheduled	278	243
Scheduling Rate	60%	74%	<0.001
Initial BH visits	205	173
Treatment Initiation Rate	44%	53%	0.03

In unadjusted Cox proportional hazards analysis, time to scheduling was significantly lower after the intervention than before (hazard ratio=1.60; 95% confidence interval = 1.37,1.88; P<0.001). This relationship was essentially unchanged when using a multivariate analysis to correct for age, sex, insurance, place of residence, race and medical problems (hazard ratio=1.68; 95% confidence interval = 1.43,1.98; P<0.001). The intervention was also significantly associated with time to initiation of therapy (hazard ratio=1.36; 95% confidence interval = 1.14,1.62; P=0.001). Again, this relationship was virtually unchanged after correcting for the potential confounders described above (hazard ratio=1.35; 95% confidence interval = 1.13,1.62; P=0.001). The fraction of patients that successfully scheduled behavioral care visits increased at all time points after referral in the post-intervention period when compared to pre-intervention (P<0.001). A similar effect was observed for the fraction of patients arriving for care (P=0.001). This was apparent for all time periods through 90 days (See Table 5: Fraction of Patients Scheduled and Arrived for Behavioral Care).

Table 5.

Fraction of Patients Scheduled and Arrived for Behavioral Care for all patients referred to behavioral health services before the intervention (Feb 2010 – Jan 2011) and after the intervention (Feb 2011 – June 2011).

Days from Referral	Fraction of Patients Scheduled		Fraction of Patients Arrived
	Pre-Intervention	Post-Intervention	Pre-Intervention	Post-Intervention
30 Days	0.39	0.63	0.29	0.42
60 Days	0.52	0.70	0.39	0.49
90 Days	0.58	0.72	0.44	0.52
P-value by log rank test	<0.001		0.001

Survey Results

Survey response rates were 40% prior to implementation and 40% afterward. The median scores for the domain “Acceptance of Lean as a QI Method” rose from 4.0 to 6.0 (P=0.35) for one statement (“The QI study was easy to do”) and was unchanged at 6.0 for the other (“If I were asked to be part of a QI project in the future, I would accept”) (P=0.95). The median scores for the domain of “Workflow Effectiveness” rose for both statements, the first (“The QI study made my job easier”) from 5.0 to 7.0 (P=0.11) and the second (“The QI project made us more efficient as a practice”) from 5 to 6 (P=0.61). The median score for the domain of “Clinical Effectiveness” (for the statement “The QI project was successful in improving the quality of patient care”) rose from 5.5 to 6.0 (P=0.77).

Feedback from Participants

The researchers presented the results of the EHR data and survey responses to study participants ten months after implementation of integrated behavioral health services. Providers and staff responded with agreement on the improved clinical results of the project, its positive effect on workflow, and the usability of Lean as a QI method. Sample comments written by participants or recorded during the presentation:

• “It was the ‘mapping’ process (system diagrams) that worked so well… If you can see it, you can understand it.” (Medical Office Assistant)

• “You don’t see what’s happening to each other when you’re working – we’re so wrapped up in our own work. We could see this.” (Advanced Practice Nurse)

• “I can tell this is happening, that it’s working, just by the calls I get now. Patients are getting the care they need.” (Triage Nurse)

• “It made sense out of chaos.” (Provider)

DISCUSSION

Office practices typically have limited time to invest in complex change. An effective QI method in health care is characterized by two key strategies: engaging front line workers directly²⁶ and eliminating wasteful work processes²⁷. As suggested by Fischer over a decade ago, and illustrated by Sinksy in a provider office practice, healthcare workers are able to redesign key office systems and processes efficiently and successfully^28,29.

The value of Lean lies in its systematic ability to address clinical and operational requirements while improving patient care. With the advance of health care reform confronting health care providers and leaders with a constant flow of change, a model for implementing effective, efficient, and acceptable change methods is of value. The use of Lean to integrate a BH clinician and change the BH referral process showed significant improvements in care, especially in treatment initiation. As a QI process, Lean had strengths in its focus on improving work processes, system perspective, and success in achieving the team’s objective. Although this application of Lean was limited to one primary care setting and intervention, the documented application of Lean in a range of health care systems and change efforts^16,17,24,26 indicates that it is likely to have broad application in practice sites.

Limitations

As a single case study, generalizability to other clinical practices is not expected. Providers and staff on the QI team were not blinded to the purpose of this study and were likely to have developed an attachment to the results of their own work, biasing some survey results and possibly affecting the responses of other clinic providers and staff. It is unlikely that team members’ opinions affected the data collected by the EHR system regarding scheduling, treatment, and the days elapsed since referral date. Although survey responses tended to show improvement after the intervention, the sample was too small to achieve statistical significance.

Key Lessons

The discovery of a better model of care (integrating behavioral health in primary care) calls for an effective method of implementing it. Using Lean, this BH implementation project was associated with significant improvements in most measures of performance. The Lean process itself was brief (8 hours), adaptable to an important and complex intervention (behavioral health integration), improved workflow design (e.g. direct scheduling of BH appointments), provided a system perspective (“it was the mapping process”), and achieved the team’s objectives (improved treatment initiation). Lean is a promising implementation strategy that may offer tangible benefits and find acceptance among providers and staff. It warrants more extensive evaluation in a variety of settings.