Article Text

Download PDFPDF

A pilot study testing a medication algorithm to reduce polypharmacy
  1. L A Mistler1,2,
  2. T A Mellman3,
  3. R E Drake1,4
  1. 1
    Department of Psychiatry, Dartmouth Medical School, Hanover, New Hampshire, USA
  2. 2
    VA National Quality Scholars Fellowship Program, VA Medical Center, White River Junction, Vermont, USA
  3. 3
    Department of Psychiatry, Howard University, Washington, DC, USA
  4. 4
    Departments of Community & Family Medicine and of Psychiatry, Dartmouth Medical School, Hanover, New Hampshire, USA
  1. Dr L A Mistler, VA Quality Scholars Program (11Q), VA Medical Center, White River Junction, VT 05009, USA; lisa.a.mistler{at}


Background: Polypharmacy is common in the treatment of persons with severe mental illness, yet it is not an evidence-based practice. To address this, an attempt was made to reduce medications for patients already receiving polypharmacy during an episode of acute psychiatric hospitalization.

Methods: A medication-reduction algorithm was developed , based on the best available evidence regarding indications for and efficacy of medications and principles of collaborative care. A feasibility pilot study was conducted using a matched case-control design for 12 patients treated with the algorithm and 12 patients treated as usual.

Results: The intervention patients were discharged on significantly fewer medications than controls; symptom reduction and length of stay did not differ significantly.

Conclusion: A collaborative approach to reducing polypharmacy may reverse the trend to add medications during hospitalization.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The trend toward increased coprescribing of psychotropic medications, which is often termed polypharmacy, is well documented.115 For example, Clark et al.16 studied the prescription trends for a cohort of New Hampshire Medicaid and Medicare recipients diagnosed with schizophrenia and schizoaffective disorders, and found that the likelihood of being prescribed more than one antipsychotic medication concurrently increased from 4% of patients in 1995 to 24% of patients in 1999. While individual patients may benefit from specific combination strategies, the practice of coprescription is not supported by current research and probably exposes patients to greater risks of non-adherence, side effects and toxicity.1013

Reasons for the increases in coprescription have been speculated to include the comfort of clinicians in combining newer agents, greater emphasis on comorbidities, enhanced goals for recovery, increase in number of approved psychoactive medications, increase in patient use of non-prescription medications, increased internet availability of medications, the syndromic nature of psychiatric disorders, treatment of side effects, increased insurance coverage for medications, lack of communication between prescriber and patient, decreased use of behavioral and social techniques, and increase in age of patients.17 Pressures for brief hospital stays also contribute, as new medications are often added during hospitalization.18

Treatment guidelines and medication algorithms have been proposed to encourage systematic and evidence-based medication management.19 For example, the Texas Medication Algorithm Project provides and regularly updates algorithms for the treatment of adults with schizophrenia, bipolar disorder and major depression.2021 With few exceptions, the Texas Medication Algorithm Project recommendations for initial medication interventions and second-line alternatives are single agents. As far as we are aware, none of the current guidelines and algorithms addresses procedures for reducing polypharmacy.

The purpose of this pilot project was to evaluate the feasibility and usefulness of reducing redundant medications during acute admissions to a state hospital. We hypothesized that eligible patients would be interested in participating and that algorithmic treatment would result in reduced psychotropic medications at discharge, without negative effects on symptom management and length of stay.



Based on published treatment guidelines and clinical research, we developed a collaborative medication reduction algorithm (Appendix A). The algorithm attempts to optimize single drug regimens where appropriate and to eliminate redundant or ineffective medications whenever possible. Information from patients, their families and their practitioners is used to review adequacy of previous trials and the degree of improvement, tolerability and adherence associated with each current medication. Following the evidence-based medicine paradigm,22 the clinical procedure emphasizes collaboration between prescriber and patient to document prior responses to medications, to develop an evidence-based plan for medication changes and to assess the response systematically.

Monitoring of symptoms and side effects is designed to be practical and not to increase clinician burden. For example, specific items from the Brief Psychiatric Rating Scale (BPRS)23 and nursing observations are used to assess the target symptoms (e.g., suspiciousness, hostility) and side effects (e.g., weight gain). Systematic ratings and collaborative decision-making based on five underlying principles outlined by the authors were used to determine all medication changes, dose adjustments, discontinuations and substitutions (Appendix A). The principles include discontinuing a medication if it has not proven effective for the patient, choosing the more effective of two or more medications in the same class and discontinuing the rest, and titrating a medication to the optimal tolerable dose for an adequate duration. At each evaluation, available options indicated in the algorithm were discussed with each patient; patients had the final say in what the next step would be, emphasizing their control in the decision-making process.

To evaluate the effectiveness of the medication-reduction algorithm, we used a matched case-control design. The first author enrolled and treated 12 consecutive consenting patients who met eligibility criteria. The controls were retrospective, as a prospective match for diagnosis, number of medications on admission and age within 5 years would have been virtually impossible in this small hospital within a reasonable time frame. After each intervention patient was identified and treated, a patient from the hospital’s medical records not treated by the first author was identified to match diagnosis and number of psychotropic medications at admission, age within 5 years and admission time within 5 years. The first author and a medical student performed these searches, purposely not looking at medications at discharge to minimize selection bias. Given the lack of electronic records, it was difficult to hand-search paper records to find matches for both diagnosis and number of psychotropic medications on admission.

The Dartmouth Medical School and New Hampshire State Committees for the Protection of Human Subjects approved the project. Patients treated with the medication-reduction algorithm gave written informed consent prior to enrollment in the study.


Adult, non-geriatric psychiatric patients admitted to New Hampshire Hospital with pre-admission medication regimens consisting of three or more psychotropic agents and/or two within the same therapeutic class were included. Patients under 18 and over 65 were excluded, since they were generally admitted to specialty adolescent and geriatric units in the hospital, and not to the treating psychiatrist’s unit. Patients were excluded if they were pregnant women or were discharged within 3 days of admission. From our initial observations, it appeared that approximately 30% of new admissions met these criteria. We invited 16 patients to participate in the medication reduction plan. Four patients declined because they were concerned they would have to stay in the hospital longer. Table 1 shows the baseline characteristics and outcomes of the study group.

Table 1 Characteristics of intervention and control groups


Patients treated with the medication-reduction algorithm were assessed by the first author and a research assistant. Control-group patients were identified by the first author and a medical student, who also reviewed medical records for initial and final symptom scores, medication changes and lengths of stay. Outcome measures included the number of medications, degree of symptom reduction from admission to discharge and length of hospitalization. The overall symptom status was determined by use of the BPRS23 at admission and at discharge, routinely done by a social worker or psychologist on the patient’s team. The BPRS is a 16-item assessment instrument for positive and negative symptoms as well as hostility and aggression. These are rated from 1 (no evidence) to 7 (all the time) by a clinician. We followed changes in the sum of BPRS items that were initially scored moderate or greater (⩾4), and we subsequently refer to this as a modified BPRS score.

In order to incorporate shared decision-making into the algorithm, decisions about medication changes were informed by a systematic review of common side effects and categorization of the patients’ tolerance of the medication. Based on a patient’s history of response to, tolerability of and strength of trial of a medication, the algorithm suggests an action to be taken next: continue, optimize, reduce or eliminate. This is done as a shared decision-making process with the patient, respecting the patient’s right to choose a different option than the algorithm indicates.

Data analyses

Due to the non-normal distribution of the data and the small number of participants (n = 24), Wilcoxon rank sum tests were used to analyse demographic data as well as number of medications, symptoms and lengths of stay. The statistical package was Intercooled STATA 9.


Table 1 summarizes the results. At baseline, the two groups were well matched on demographic and clinical features. All 12 patients who entered the medication-reduction algorithm adhered to the protocol and were included in the intent-to-treat analysis. As the table shows, the medication-reduction algorithm group decreased their overall psychiatric medications significantly compared with the control group, with similar reductions in psychiatric symptoms and no significant difference in length of stay. Modified BPRS scores are included as an indication of how symptoms changed for each group, based on blinded-rater assessment.


Our results suggest that it is possible to counter the trend toward polypharmacy in a state hospital setting by explicitly focusing on reducing the number of psychotropic medications in the same class using a collaborative, evidence-based algorithm for patients who are already receiving coprescriptions. Several findings from the pilot deserve emphasis. Patients were interested in reducing medications, were able to participate in the protocol, were able to reduce the number of medications during brief hospitalizations and experienced symptom reductions comparable with those of a matched control group. The more distal goals of reducing unnecessary medication exposure and related side effects need to be examined longitudinally. Importantly, however, in this brief pilot study, patients on the medication-reduction algorithm experienced similar symptom reductions to those of control patients, suggesting that the hospital setting and an evidence-based approach to pharmacology may supersede the need to add medications. The medication-reduction group stayed in the hospital slightly longer, but the difference was not significant, perhaps due to the small sample size and large variance. Our findings are consistent with those reported in Glick et al.24 In their study of hospitalized patients with stabilized schizophrenia, they were able to taper all other psychotropic medications not indicated by a comorbid illness without any apparent change in patients’ symptoms.

This pilot study warrants several caveats. The matched case-control design is appropriate for a pilot study, but further investigation will require a randomized controlled trial. Similarly, though the small sample size was adequate for a pilot study, future studies should have sufficient power to examine high-variance outcomes, such as length of stay. This pilot was also limited by the author knowing which patients were in the protocol, and therefore could be subject to bias. Fortunately, the BPRS ratings were done independently of the author, and the status of the patient as a control or subject was not known to the rater. However, we did not test for inter-rater reliability. Using a medical record review to find matched control subjects is not as elegant as matching to patients on other treatment teams as they are admitted. We did not perform a long-term follow-up, which might highlight some confounding differences between patients. Finally, there is possible confounding due to unobserved variance in doctors, treatment units and other factors.

Notwithstanding these limitations, we conclude that a medication-reduction algorithm deserves further study. Psychotropic medications can be expensive, and the systematic reduction of coprescribed medications could potentially save hospitals, insurance companies and patients money. Limiting polypharmacy is a critical clinical issue. This pilot study establishes the feasibility of reducing the number of medications during brief hospital admissions without producing adverse effects.


The authors thank the Dartmouth Medical School Quality Research Grant Program for financial support. In addition, we thank L Leary, the QRGP coordinator, J Marshall, who helped with chart reviews, and T Onega, who performed the statistical analysis.


Algorithm for simplifying medication regimens

Principle I

See table A1.

Table A1 Medication prescribed to target principal disorder, for which it has an established indication, and comorbid psychiatric disorders that are persistent and/or independent of principal diagnosis

Principle II

If medication is used or considered for a target that is not at the level of an Axis I diagnosis (e.g. aggressiveness secondary to a personality disorder) then a single agent is chosen based on considerations of the relationship of the target to a diagnosis and existing research evidence. The medication is not continued if evidence for efficacy is questionable. Non-pharmacological approaches are also considered.

Principle III

Medication to be titrated to optimal, tolerable dose and response observed for adequate duration prior to combining medications to enhance efficacy.

Principle IV

When an individual presents with more than one agent for the same target indication, the medication judged to be the most likely to be effective for the individual based on their treatment history and relevant research evidence is titrated to an optimal dose, or a new medication from the therapeutic class is initiated, and the other(s) are discontinued.

Principle V

Medications that are not resulting in significant improvement to be discontinued (substituted if appropriate). While inadequate for assessing a full response, improvement should begin to be evident within 1 week of achieving and an adequate dose of an antipsychotic medication and 2 weeks for a thymoleptic medication.


  1. Maintain combination or augmentation strategy if clear and compelling history of incremental benefit after observing response to simpler regimen provided by patient or treatment provider.

  2. Reinstitute if clear exacerbation follows tapering or discontinuation (and exacerbation is not due to acute withdrawal effect that is likely to be transient).

  3. Depot antipsychotic medications can be used with oral agents if there is a significant risk of medication non-adherence after discharge, and there is a clearly established added benefit of the oral agent.



  • Funding: The Dartmouth Medical School Quality Research Grant Program provided LAM with support to present this project as a poster at the Fifth Annual Summer Institute on Evidence Based Practices in San Antonio, TX, 29 June to 1 July 2006.

  • Competing interests: None for the first and third authors; speakers’ bureau for Takeda Pharmaceuticals Limited for the second author.

  • Ethics approval: Ethics approval was obtained from The Dartmouth Medical School and New Hampshire State Committees for the Protection of Human Subjects.

  • Patient consent: Obtained.