December 1997

Preface and Acknowledgments

This briefing paper explores an emerging tool in behavioral healthcare -- the use of algorithms for the medication management of people with serious mental illnesses. Treatment algorithms, also referred to as practice guidelines, are a series of treatment steps, each of which is defined by the clinical response of the patient to the preceding step. Used traditionally to treat physical illnesses, this technology has recently emerged in behavioral healthcare settings in the context of psychiatric medications.

The goals of this paper are to provide the reader with an understanding and rationale for the appropriate use of treatment algorithms for people with serious mental illnesses. It suggests effective strategies for using treatment algorithms to improve the quality of treatment and to increase the accountability of medication treatment. The paper also addresses potential dangers in developing practice guidelines and provides advice for avoiding these pitfalls. Issues related to legal matters and managed care contracting are discussed briefly. The paper concludes by recommending principles and action steps for policymakers, practitioners and mental health consumers alike. Finally, appendices include a sample treatment algorithm and a list of mental health-related web sites.

The initial idea and support for this paper was developed by members of the Medical Directors Council of the National Association of State Mental Health Program Directors (NASMHPD). The Council has also been working in partnership with NASMHPD and the NASMHPD Research Institute on a related project focused on defining the leadership role of medical directors to provide evidence-based quality services in public mental health systems, including using treatment guidelines to accomplish this goal. This effort culminated in a recent meeting, "1997 Best Practices Symposium: Focus on Schizophrenia" held in October 1997 in Washington, DC and attended by medical directors from more than 36 states.

The paper was prepared by the Council for NASMHPD and the NASMHPD Research Institute. The Council was authorized by the NASMHPD Board of Directors in 1995 and its membership includes medical directors from state mental health authorities from across the country. Its primary mission is to identify emerging clinical issues impacting state mental health authorities and the public mental health service delivery system and to provide advice and guidance to NASMHPD members and staff alike on best practices for these issues.

The National Association of State Mental Health Program Directors represents the interests of state mental health authorities in 55 states, territories and jurisdictions. NASMHPD's primary mission is to provide members with federal advocacy, training and technical assistance designed to improve mental health, substance abuse and developmental disability systems throughout the nation.

The NASMHPD Research Institute is a non-profit research entity established to study issues in the delivery of public mental health services supported by state mental health agencies. Independent and nonpartisan, the Research Institute responds to current needs for objective analyses and basic information, and attempts to facilitate the application of research findings to management of state mental health programs.

Acknowledgment and thanks go to those individuals who contributed their expertise to the production of this paper. Steven P. Shon, M.D. and William Rago, Ph.D., M.B.A. of the Texas Department of Mental Health and Mental Retardation authored the paper and incorporated comments provided by members of the Medical Directors Council. Gail P. Hutchings, M.P.A. of NASMHPD edited the paper and prepared it for final production.

Thomas W. Hester, M.D. Chair, NASMHPD Medical Directors Council

Introduction

The responsibility for creating cost contained, quality driven, outcome oriented services is a major challenge for administrators of mental health systems in the rapidly changing current environment. One tool that behavioral healthcare systems are turning to is the use of treatment algorithms. A treatment algorithm (also known as a practice guideline) is a series of treatment steps, each of which, in turn, is defined by the clinical response of the patient to the preceding step.

While we are gaining research-based knowledge of the benefits of many treatments, (e.g., psychosocial treatment), research on medication treatment of the seriously and persistently mentally ill (SPMI) has advanced to a point where it is feasible to integrate this knowledge with expert clinical judgment in the development of algorithms. Medication treatment is only one component of psychiatric treatment. However, because medication treatment not only has significant impact on the quality of life of patients but significantly enhances the effectiveness of other treatments, medication treatment has typically been the first selected for the development of algorithms.

Treatment algorithms have been used for many years in various areas of medical treatment, e.g., diabetes, cardiac care, etc., and have been found to be extremely useful. Only recently, have treatment algorithms rapidly begun to emerge in the behavioral healthcare field for the use of psychiatric medications. (See Appendices A and B for sample algorithms.)

There are many reasons for using algorithms in the treatment of persons with serious mental illnesses.

The eleven that follow have been divided into two categories:

• improving the quality of treatment, and
• improving the accountability of medication treatment.

Improving the Quality of Treatment

Facilitate clinical decision making

Over 30,000 articles are entered into the National Library of Medicine's data base each month (Jobson & Potter, 1995). This explosion in knowledge requires clinicians to sort through the research, evaluate its quality, integrate the findings into a coherent model, and incorporate this into their practices. Algorithms, appropriately developed and regularly updated, aid clinicians in this otherwise unwieldy and unrealistic task. In this way, algorithms facilitate clinical decision-making by practitioners.

Reduce unnecessary variation in clinical practice patterns

Research has documented large inconsistencies in the rate at which specific procedures are performed by physicians in different geographic areas (Wennberg & Gittelshon, 1973; Wennberg, Freeman & Culp, 1987; Perrrin, Homer, Berwick, Woolf, Freeman & Wennberg, 1989). Such results lend credence to the argument that providing clinicians with algorithms would be useful in reducing inappropriate variation, thereby improving quality of treatment. Indeed, recognizing the clinical variation in medical practices, and the costs associated with it, was part of the impetus for Congress to create the Agency for Healthcare Policy Research which has significantly advanced the development of practice guidelines (Barlow & Barlow, 1995).

Improve outcomes

By providing clinicians with the best treatment knowledge, a well constructed clinical algorithm can achieve a faster and more complete consumer response to treatment than would occur with treatment as usual. Algorithm-based treatment can reduce symptoms and increase a consumer's psychosocial functioning faster than non-algorithm guided treatment. As such, algorithms offer the potential to reduce treatment costs while obtaining better patient outcomes (Rush & Trivedi, 1995).

Facilitate consistent treatment across different service systems

Shorter hospital stays mean that the physician is unlikely to know at discharge whether the treatment selected is the best one for the consumer. A strategy to address this increasingly prevalent occurrence is to have a consistent medication plan that moves with the consumer across different treatment venues, e.g., inpatient, day hospital, outpatient, and across doctors. Algorithms provide a basis for developing this plan and for communicating treatments across different venues and physicians.

Individualize treatment

According to Roche and Durieux (1994), the main goal of clinical algorithms is to facilitate decisions by practitioners who cannot integrate into their daily practice all the published data concerning new technologies and treatment approaches. One treatment is not best for all consumers. A well constructed medication algorithm is not a cookbook which shackles a psychiatrist into a rigid approach to treatment, but an aide to guide the clinician through the maze of multiple treatment options. By incorporating the concept of different paths depending on prior history, consumer preference and individual response (symptoms, functioning and side effects), to each step, its intention is to inform treatment decision-making with the goal of achieving full remission.

Increase the cost efficiency of treatment

A more thorough and complete clinical response may lead to reduced hospital use or better prognosis. A more rapid consumer response may cost less if fewer overall subsequent visits result. If more physician visits are necessary at the beginning of treatment but consumer improvement is more rapid and more complete, the cost efficiencies gained from the algorithm may be found in other areas. Similarly, if a medication has fewer significant side effects, consumer adherence with treatment may be enhanced, thus improving clinical outcome. These may be in direct costs such as decreased clinic, emergency room visits, and hospitalizations, or they may be related to decreased indirect costs, such as the consumer's faster return to work, and enhancing the positive impact of psychosocial interventions (Greenberg, Stiglin, Finkelstein & Berndt, 1993).

Thus, even though the immediate cost of medication treatment may increase (depending on the algorithm recommendation), there is a strong possibility that the enhanced benefits of treatment will achieve longer term cost efficiencies. This potential benefit is particularly likely to accrue to consumers with illnesses that have an otherwise unremitting, chronic course.

Increasing the Accountability of Medication Treatment

Make clinical decisions explicit

One of the important functions of algorithms is to make explicit the "traditional art" of clinical reasoning (Feinstein, 1974). Algorithms do this by allowing clinicians to identify the components and pathways of their clinical judgments which preserves the vitality of diagnostic reasoning while enhancing its clinical effectiveness. By making clinical decisions explicit, communication among physicians regarding consumers and treatments is significantly enhanced.

Provide a framework for comparing consumer progress

To the extent that algorithms associate consumer outcomes with clinical strategies, clinicians can compare the progress of consumers with that identified in the algorithm's decision points. This comparison enriches decision-making by enabling clinicians to measure individual consumer progress against what might be expected.

Provide a framework for evaluating when and whether to adopt new medications

The last decade has witnessed the introduction of six new antidepressant, three new antipsychotic, and two new antimanic mood stabilizing medications. One can expect more psychotropic agents to be forthcoming in the next decade. At least theoretically, some medications represent such dramatic advances that they replace others in priority almost immediately, e.g., a new drug that is twice as safe and twice as effective as the standard. Rarely, however, is such a clearcut case presented. Rather, new agents often have equal efficacy but may be better tolerated, safer in overdose, or effective for consumers failing to respond to other agents. A system based on a clearly articulated, multi-step medication algorithm can define empirically where, in the sequence of steps, the new agent may afford the most clinical benefit. In this way, the algorithm can inform physicians in their use of new medications.

Provide a framework for defining cost of treatment

Since algorithmic care provides a framework for clinical decision-making, it also provides a means for documenting the costs associated with care. This should allow mental health systems to delineate the costs associated with specific treatment intervention and link costs with consumer outcomes. Various consumer and treatment environmental factors can be identified that affect both costs and consumer outcome.

Provide a framework for identifying good documentation

The appropriate use of algorithms requires good documentation. Good documentation not only includes recording consumer response to treatment at specific steps in the algorithm, but also the consumer's current status regarding symptoms and functioning as assessed by reliable and valid measures. The algorithm itself provides a framework for documentation, doing so by identifying what consumer assessments are needed, at what steps in the algorithm, and tying these to usual expectations regarding consumer progress.

Potential Shortcomings with Practice Guidelines

There are potential dangers in the development of practice guidelines of which developers and clinicians must be cognizant. Among these are the following:

Insufficient evidence
Well developed guidelines should include a rigorous review of scientific literature and empirical evidence. Guidelines based upon only one aspect of evidence or a very narrow set of parameters may not lead to the best quality of care and outcomes for the consumer. For example, guidelines heavily influenced by the comparative cost of medications may sacrifice long term beneficial outcomes for short term financial savings.

Biased opinions
Guidelines developed by consensus panels may not always reflect a broader consensus of experts, depending upon which experts were selected for the panel. Also, within the panel the results may not reflect a true consensus of the group, but instead the opinions of the person or persons who were most articulate, vocal, or unyielding.

Increased costs and utilization of services
The use of guidelines may have the opposite results from those originally intended, especially if lower utilization of certain types of services is an objective of the guidelines (Katz, et al., 1996).

Substitute for clinical judgment
Psychiatrists know that every individual is unique and that, while they might fit certain categories, they will also differ to greater or lesser degrees. If a guideline is too rigid and inflexible, the psychiatrist may not be able to use appropriate expertise and judgment in making decisions in the best interest of the consumer. Guidelines that shackle the psychiatrist in such a manner could actually do more harm than good. On the other hand, if the guideline is too flexible, it will not be effective in guiding or directing care.

Standard of care
Algorithms may eventually become one of the instruments used by various entities including those that provide funding and oversight. Ill conceived algorithms may actually lead to a lower quality than treatment as usual.

Inaccurate diagnosis
The effectiveness of the algorithm relies upon an accurate diagnosis. Diagnostic accuracy becomes crucial if the application of an algorithm is to yield good results.

Outdated content
The utility and effectiveness of treatment alogithms is maximized when they are regularly revisited and updated with contemporary information on new medications and new scientific evidence.

Legal Issues

A "standard" is a rule(s) or criterion by which something is measured. Medication guidelines and algorithms can and will help set the general standard, i.e., standard of practice, by which psychiatric medication practice will be judged relative to quality care. Those algorithms that are driven by scientific evidence of efficacy will be most influential in defining the general standards related to quality care.

This raises future issues of liability if a psychiatrist does not follow a specific algorithm that is considered valid. This is especially problematic if an algorithm is rigid, inflexible, and does not allow for rational clinical judgment based upon a specific, individualized clinical situation.

In order to be both clinically sound, as well as to minimize legal exposure it is important that an algorithm allow a certain latitude of deviation, (e.g., skipping one or more steps or selecting among options within a particular step), in order to best respond to an individual, sometimes idiosyncratic, clinical situation. While this flexibility is important, it is also crucial that the algorithm require the psychiatrist to clearly document the clinical rationale for that particular deviation. With a well documented rationale, liability should be minimized.

Realistically, if psychiatrists are following a scientifically sound, clinically flexible medication algorithm, and are documenting appropriately, they, as a group, as well as the system in which they are practicing, should have increased protection from legal exposure.

Contracting Issues

Both managed behavioral healthcare organizations and public sector delivery organizations are moving toward required use of medication guidelines and algorithms for psychiatrists in their systems. State authorities may influence or require which medication algorithm is used by service providers by clearly stating in their contracts and RFPs for contracts which medication algorithm must be used. It is important to make this clear right up front as it will influence the formulary and cost calculations regarding medications.

It is also important to note that appropriate use of medication algorithms entails other costs as well. These costs are primarily training costs for:

• Training in the use of the algorithm

• Training in the use of instruments for evaluating symptoms, functioning, and side effects (these are also outcome measures)

• Training in the use of new and/or unfamiliar medications

• Timely provision of "expert" consultation

Also, ongoing consultation on the use of the algorithm is essential and requires some cost consideration as well.

Algorithm Implementation Issues:
Lessons Learned from the Texas Experience

During the past year, Texas Department of Mental Health and Mental Retardation has embarked upon a project to develop and field test the use of three psychiatric medication algorithms with the ultimate goal of implementing them throughout the state as a best practice. This project, called the Texas Medication Algorithm Project (TMAP), is divided into three phases.

• Phase 1: Development of medication algorithms for three major mental disorders
• Phase 2: Feasibility testing of the algorithms in selected sites
• Phase 3: Prospective evaluation of the clinical and economic impact of the three medication algorithms compared to treatment as usual

Phase 1 was completed in fall 1996 and Phase 2, in August, 1997. The experience with Phase 1 and partial experience with Phase 2 has led the Texas group to the following preliminary conclusions regarding implementation, which may have relevance to other states.

1. Include Stakeholders in Development and Testing
It is important to include all mental health stakeholders, e.g., consumer groups (Texas Alliance for the Mentally Ill, Depressive/Manic Depressive Association, Texas Mental Health Consumers, etc.), state psychiatric association, university research programs, administrators, practicing psychiatrists, and other clinicians from the beginning of the process. Support from all of these groups is crucial in moving implementation forward.

2. Psychopharmacology and Algorithm
In the Texas experience, most psychiatrists were not familiar or experienced with every single medication that appeared in a scientifically based medication algorithm. Therefore, education programs for physicians that include a focus on each medication contained in an algorithm (especially the newer medications) are very important for the proper implementation of a medication algorithm.

3. Medication Availability
In order for a scientifically-based medication algorithm is going to work, the medications in the algorithm must be available to prescribe. Formulary restrictions (which are frequently based on cost) will interfere with the benefit derived from an algorithm. Implicit in medication algorithm development is the assumption that formulary decisions must be based on the overall cost effectiveness of treatment.

4. Medication Visit Length of Time
In order to carry out an algorithm appropriately and to provide good medication treatment physicians require adequate time to evaluate target symptoms, functioning, side effects, etc. If the time allowed is too short, appropriate treatment under an algorithm cannot be carried out.

5. Medication Visit Frequency
The scientific literature recommends that individuals who are unstable and need medication adjustments, i.e., initiation, medication changes, medication adjustments, etc., need to be seen frequently during these times. If frequency of visits are restricted by rigidly fixed schedules, algorithms cannot be appropriately implemented.

6. Algorithm Consultation
Texas psychiatrists implementing algorithms were universally supportive of having access to regular consultation on the use of the algorithms and the medications within them. With the introduction of several new medications for treatment of schizophrenia, depression, and bipolar disorder in recent years, medication therapy has become increasingly complex. Quickly available consultation is an important part of effectively implementing a medication algorithm.

7. Consumer and Family Education
In the Texas experience, consumers and family members were very interested in the algorithm concept. Their inclusion in the process as well as their involvement in developing educational materials regarding mental illness, psychiatric medications, and algorithms has been extremely helpful.

8. Instrument Assessment of Symptomatic Response
All algorithms require the clinician to decide what to do next based on results obtained at the time of specific decision points (e.g., change medication, adjust dose, add medication). Simple symptom ratings appear to help clinicians more accurately evaluate when and whether such strategic or tactical revisions are needed.

9. Physician - Consumer Partnership
TMAP's three algorithms were developed to include "choice points" where the physician and consumer can choose among medications with similar efficacy and safety profiles, (i.e., choice rests largely on the medication's side effect profile). Such choice is predicted to increase consumer adherence with the treatment regimen, an especially important function in the long term treatment of people with serious mental illnesses. To take advantage of choice, consumer education about the algorithms, and their medication is important. Thus, in the Texas experience, the pre-defined treatment algorithm serves as the framework for structuring communication between the physician and consumer. Within this framework, choice, education, consumer progress and treatment adherence issues are discussed.

The implementation issues described above were identified by TMAP in the early stages of the project, and certainly others will emerge as the project moves forward. Many of the issues have resource and planning implications which states will need to consider as they move down the path.Principles and Recommended Actions

As medication algorithms for the SPMI population emerge, serious policy questions arise for administrators. Among them are:

• Whose algorithms should be selected for use?

• What principles drive the algorithms (cost containment vs. quality)?

• Is the algorithm valid (has it been scientifically tested)?

• What is the impact (outcomes) of the use of the medication algorithm?

• What are the legal ramifications of using treatment algorithms?

Unfortunately, several of the algorithms emerging from private managed care entities are driven primarily by short term cost considerations, i.e., the cheapest medications must be used first and failure (often multiple) documented before more effective and safer (and more expensive) medications may be used. The withholding of clearly superior medication treatment in favor of using cheaper medications first has caused great concern among consumers, family members, and treating psychiatrists in mental health systems across the country.

Because medication treatment algorithms will be used to help define standards of care in the future, the members of the NASMHPD Medical Directors Council believe that it is important that NASMHPD take a position on the development and use of such algorithms. The Council recommends that NASMHPD endorse the following principles and actions:

Principles

Efficacy and quality care (rather than cost) should be the major defining principle of psychiatric medication algorithms.

Psychiatric medication algorithms should not be static, but should regularly evolve, incorporating new medications and new scientific evidence.

Public sector psychiatric medication algorithms should be valid for a range of illness severity levels (including the most severe), and should account for diversity among populations in ethnicity, gender and socioeconomics.

Recommended Actions

State mental health authorities should collaborate with academia, professional organizations, and consumer groups to adopt and/or develop psychiatric medication algorithms which are applicable to the vast majority of individuals served in their system.

States should collaborate with each other whenever possible in the development and prospective evaluation of medication algorithms.

State mental health authorities, should review and approve the use of psychiatric medication algorithms used in their states, based upon their scientific validity and empirical evaluation.

References

Barlow DH, Barlow DG. Practice guidelines and empirically validated psychosocial Treatments: Ships passing in the night? Behavioral Health Tomorrow 1995:May/June: 25-29.

Feinstein AR. An analysis of diagnostic reasoning III: The construction of clinical algorithms. Yale J of Biology and Medicine 1974:1:5-32.

Greenberg PE, Stiglin LE, Finkelstein SN, Berndt ER. The economic burden of depression in 1990. J Clin Psychiatry 1993: 54:11:405-418.

Jobson KO, Potter, WZ. International Psychopharmacology Algorithm Project Report: Introduction. Psychopharmacology Bulletin, 1995:31(3): 457-459.

Katz DA, Griffith JL, Beshansky JR, Selker, HP. The Use of Empiric Clinical Data in the Evaluation of Practice Guidelines for Unstable Angina, JAMA, 1996, 276 (9).

Perrrin JM, Homer CJ, Berwick DM, Woolf AD, Freeman JL, Wennberg, JE. Variations in rates of hospitalization of children in three urban communities. N Engl J Med 1989:320:1183-1187.

Roche N, Durieux P. Clinical practice guidelines: From methodological to practical issues. Intensive Care Medicine 1994:20:593-601.

Rush AJ, Trivedi MH. Treating depression to remission. Psychiatric Annals 1995: 25:704-709.

Wennberg JE, Gittelshon A. Small-area variation in healthcare delivery. Science 1973:182:1102-1108.

Wennberg JE, Freeman JL, Culp WJ. Are hospital services rationed in New Haven or over-utilized in Boston? Lancet 1987:1:1185-1189.

Appendix A: Exemplar Algorithm Strategies Template

Appendix B: Strategies for the Treatement of Schizophrenia

Appendix C: Mental Health-Related Web Site Addresses

BACK TO TOP

NASMHPD Medical Directors Home Page