A pilot study of antipsychotic prescribing decisions for acutely-Ill hospitalized patients

Results Antipsychotic choice correlated significantly with ratings of predicted symptom control (OR = .92, p = 0.02) and metabolic risk (OR = .88, p = 0.01). Mean differences between the chosen and next best drugs were significant but small in predicted symptom control ( F = 2.81, df = 3, 76; p < 0.05) compared with larger differences in anticipated metabolic risk ( F = 14.80, df = 3, 76; p = 0.0001). Nevertheless, among 24 identified reasons influencing drug selection, anticipated metabolic risk of chosen antipsychotics was cited less often than efficacy measures. In contrast to psychiatrists' expectations of metabolic risk with selected treatments, we found that patients' actual baseline BMI, fasting glucose, blood pressure, and Framingham risk levels did not necessarily predict antipsychotic treatment choice independent of other factors. Conclusion In the context of an acute psychiatric hospitalization, pilot data suggest that predictions of symptom control and metabolic risk correlated significantly with antipsychotic choice, but study psychiatrists were willing to assume relative degrees of metabolic risk in favor of effective symptom control. However, prescribing decisions were influenced by numerous patient and treatment factors. These findings support the potential utility of the ATCQ questionnaire in quantifying antipsychotic prescribing decisions. Further validation studies of the ATCQ questionnaire could enhance translation of research findings and application of treatment guidelines. Research Highlights ► Antipsychotic choice correlated significantly with ratings of predicted symptom control (OR = .92, p = 0.02) and metabolic risk (OR = .88, p = 0.01). ► Study psychiatrists were willing to assume relative degrees of metabolic risk in favor of effective symptom control. ► Prescribing decisions were influenced by numerous patient and treatment factors. ► Among 24 identified reasons influencing drug selection, metabolic risk was cited less often than efficacy measures but nearly doubled in frequency among patients with elevated glucose or BMI. Abbreviations ATCQ Antipsychotic Treatment Choice Questionnaire VAS visual analogue scale APA American Psychiatric Association ADA American Diabetes Association GLM General Linear Model SAS “Statistical analysis system” software available from SAS Institute Inc. BMI Body mass index OR Odds ratio SD Standard deviation df Degrees of freedom DSM Diagnostic and Statistical Manual Keywords Antipsychotic drugs Metabolic syndrome Prescribing decisions Schizophrenia Schizoaffective disorder Bipolar disorder 1 Introduction In view of conflicting data and opinions on the relative merits of different antipsychotic drugs in efficacy, adverse effects and cost effectiveness ( Davis et al., 2003; Geddes et al., 2000; Lieberman et al., 2005; McEvoy et al., 2005; Rosenheck et al., 2003; Stroup et al., 2006 ), it is unclear to what extent evidence from clinical trials has influenced prescribing practices.( Essock, 2002; Marder, 2002 ) Surveys have shown a gap between evidence-based antipsychotic treatment algorithms and decision making in clinical practice.( Buchanan et al., 2002; Covell et al., 2002; Marder et al., 2002; Miller et al., 1999 ) For example, there has been consistent evidence on the prevalence of metabolic syndrome among patients with schizophrenia and the relative differences among antipsychotics in liability for weight gain, diabetes and dyslipidemia. ( Allison and Casey, 2001; Correll et al., 2006; Henderson, 2005; Lieberman et al., 2005; McEvoy et al., 2005; Nasrallah, 2003; Sernyak et al., 2002; Stroup et al., 2006 ) Based on these findings, a consensus statement recommending routine monitoring of weight, lipid and glucose levels in patients initiated on antipsychotics was published ( American Diabetes Association and American Psychiatric Association, 2004 ). Nevertheless, metabolic parameters are infrequently monitored, and drugs with low metabolic risk are prescribed less often than high risk drugs ( Essock, 2002; Morrato et al., 2009b ). Thus, it is unclear how metabolic risk is balanced against efficacy, costs, co-morbidities and other variables in making prescribing decisions. Few studies have examined factors influencing prescribing patterns in clinical practice. Hoblyn et al. examined predictors of antipsychotic prescribing in veterans with schizophrenia and found that hospital size, age and secondary diagnosis predicted prescription of a second rather than a first-generation drug ( Hoblyn et al., 2006 ). Hamann et al. found that older physicians were five times more likely to prescribe first-generation antipsychotics compared to younger physicians ( Hamann et al., 2004 ). Covell et al. documented the role of patient race and ethnicity influencing drug choice ( Covell et al., 2002 ). Weiden noted the importance of opinions and perceptions among psychiatrists concerning alternative treatment choices ( Weiden et al., 2006 ). Contributing to this issue is the lack of accepted methods or instruments for assessing how individual psychiatrists balance risks versus benefits in selecting antipsychotic drugs for a particular patient. Linden et al. recognized the importance of multiple physician and patient variables by developing a list of survey questions, which they then used to question psychiatrists directly about their reasons for or against switching to olanzapine albeit without specifically addressing metabolic risk ( Linden et al., 2006 ). In contrast, clinical decision making and health care utility studies in other disease states may offer a more quantifiable paradigm to measure how treatment alternatives are weighed ( Backlund et al., 2000; Fisch et al., 1981; Kaplan et al., 1993; Kirwan et al., 1990 ). For example, Backlund et al. examined factors influencing decisions to prescribe lipid-lowering drugs and how clinical decisions compared with guidelines on hyperlipidemia ( Backlund et al., 2000 ). Using laboratory records, doctors completed a survey that assessed their readiness to prescribe a lipid lowering drug in each of 40 cases on a visual analogue scale. Understanding how psychiatrists utilize evidence to make prescribing decisions has important implications for medical education, translation of research findings, application of treatment guidelines, and clinical practice. The primary objective of this preliminary study was to pilot a questionnaire designed to quantify risks and benefits considered by psychiatrists in choosing an antipsychotic medication. A secondary objective was to evaluate the extent to which antipsychotic treatment decisions are affected by metabolic and cardiovascular risk levels. We hypothesized that numerous patient and treatment variables would influence prescribing decisions, but that symptom control and metabolic risk would be major considerations in treatment choice. 2 Methods 2.1 Subjects Four psychiatrists (ages 40–70 years), with at least 10 years of patient care and teaching experience each on an acute inpatient psychiatric program located in an urban, academically-affiliated Veterans Affairs Medical Center, were recruited to participate in the study. The study focused on factors they considered in prescribing decisions choosing among antipsychotic drugs for a sample series of patients admitted under their care. Choice of antipsychotic was based entirely on clinical judgment in each case, without institutional or formulary restrictions. A consecutive sample of 80 patients admitted to the acute inpatient psychiatric program of the Veterans Affairs Medical Center was used to study prescribing decisions between November 2006 and June 2007. Patients were included if they were 18–65 years of age, diagnosed with schizophrenia, schizoaffective disorder or bipolar disorder by DSM-IV criteria, prescribed the intramuscular or oral formulations of any antipsychotic medication for reasons other than temporary management of acute agitation within seven days of the admission date, had height and weight recorded at any time during the inpatient admission, and had fasting blood glucose levels reported in the past 12-months. Other variables, such as patient demographics and psychiatric and medical diagnoses, were abstracted from medical records. Further, the investigators obtained measures from medical records that would permit calculation of Framingham risk equations of Wilson ( Wilson et al., 1998 ), such as total cholesterol, HDL, blood pressure, diabetes and smoking status. There were no additional laboratory tests or treatment changes required as part of the investigation. Patients were excluded if they had primary psychiatric diagnoses other than above, were not prescribed antipsychotic medications after admission, lacked required laboratory data in the required time frame or were women who were pregnant. The study was approved by the institutional review board and informed consent was waived for both psychiatrists and patients. 2.2 Assessment instrument The Antipsychotic Treatment Choice Questionnaire ( ATCQ; available in Supplementary Materials ) was jointly developed by the investigators through a consensus process to assess the relative importance of different factors in antipsychotic prescribing decisions. Development began by reviewing methods for health utility scaling with a search of the literature for studies on decision-making related to psychiatric prescribing. The ATCQ consisted of two sections. In the first section, psychiatrists were instructed to identify only the three most important reasons for antipsychotic drug choice for a particular patient using a simple checklist of 35 variables covering patient characteristics and treatment history, clinical assessment of symptoms, metabolic risks, and other considerations. This section of the ATCQ was changed after an interim analysis indicated that the “Patient choice or preference” and the “Caregiver choice or preference” reasons were being endorsed more often than expected, where a more specific reason for the antipsychotic choice could have been given by the psychiatrist. These categories were further clarified with the psychiatrists as default selections to be used only when more precise reasons could not be elicited, resulting in less frequent use of these non-specific reasons. In the second section of the ATCQ, psychiatrists were asked to list the antipsychotic drugs chosen for a particular patient, and then to estimate the expected degree of symptom control and metabolic risk posed by the chosen medication in that patient on separate 100 mm visual analogue scales (VAS). Finally, they were asked to list the next best alternative antipsychotic drugs, estimating the certainty of symptom control and metabolic risk compared to the chosen prescribed antipsychotic drug using the same VAS. On the VAS for achieving symptom control, ratings ranged from 0 mm (completely uncertain) to 100 mm (completely certain). The metabolic risk rating ranged from 0 mm (no risk) to 100 mm (most risk). Calculations of mean differences between drug group choices on each VAS rating scale were interpreted as a percentage difference in the degree of certainty of symptom control and metabolic risk, respectively, predicted or anticipated by the rating psychiatrists. The ATCQ required approximately 5 minutes to complete. 2.3 Study design This is a preliminary, cross-sectional, descriptive study involving the development and testing of a brief questionnaire designed to survey and quantify factors weighed by individual psychiatrists in choosing an antipsychotic medication focusing on the certainty of symptom control and degree of metabolic risk posed by a prescribing decision in a particular patient. After the ATCQ was developed as described above, it was used in assessing prescribing decisions on 80 patients consecutively admitted to the acute psychiatric inpatient program who met eligibility criteria listed above. When an eligible patient was admitted, demographic, diagnostic, treatment and metabolic data were first abstracted from the medical record to be used in the analysis of baseline metabolic parameters. Then, the ATCQ was completed anonymously within one week of admission by one of four study psychiatrists following an initial antipsychotic prescribing decision on patients assigned to their care. The psychiatrist was instructed to identify the three most important reasons for their drug selection in a particular patient among the 35 factors listed in the first section of the ATCQ. In the second section of the ATCQ, the psychiatrist was asked to indicate the antipsychotic drug chosen for the patient and to rate the degree of symptom control expected with the selected drug and the expected metabolic risk on VAS scales. Finally, they were asked to list what they considered to be the next best alternative antipsychotic drugs to achieve symptom control and minimize metabolic risk for the assigned patient, and similarly to rate expected symptom control and metabolic risk for these next best drugs on the same VAS scales used for the chosen drug. 2.4 Data analysis Chosen antipsychotic drugs were classified into treatment groups based a priori on established degrees of metabolic risk as follows: low risk-ziprasidone/aripiprazole, high risk-olanzapine/clozapine, moderate risk-risperidone/quetiapine, with first-generation antipsychotics considered separately (APA/ADA Consensus Guidelines ( American Diabetes Association and American Psychiatric Association, 2004 ). Patients were classified according to the antipsychotic with the highest metabolic risk prescribed when they received more than one antipsychotic agent. Patients in the first-generation group could not receive a second-generation antipsychotic concomitantly. The primary analysis examined the association between the actual choice of initial antipsychotic treatment and the psychiatrists' VAS ratings of certainty of symptom control and metabolic risk. A discrete choice, logistic regression tested whether VAS measures of certainty of symptom control and metabolic risk were statistically associated with choice of a low-risk antipsychotic (ziprasidone/aripiprazole) versus all other treatment groups combined. In all analyses, hypotheses were tested at an alpha level of 0.05 without adjustment for multiple comparisons. A secondary analysis comparing the four separate antipsychotic treatment groups was implemented using a General Linear Model (GLM procedure in SAS) and examined whether the mean differences in VAS ratings by the four psychiatrists between the drug chosen and the next best drug were significantly different across groups. The dependent variable for this analysis was calculated as the difference in ratings between the drug chosen and the next best drug on VAS ratings of metabolic risk and symptom control. For symptom control ratings, a positive difference was defined as indicating that the drug chosen was considered by the rating psychiatrist to be more likely to achieve symptom control compared with the next best drug. For metabolic risk ratings, a positive difference was defined as indicating that the drug chosen was considered by the rating psychiatrist to be less likely to result in adverse metabolic effects compared with the next best drug. In a third analysis, the association between antipsychotic choice and the actual, baseline metabolic status of the patient was examined. Descriptive statistics and a distribution plot of baseline metabolic parameters (BMI, fasting glucose, blood pressure, Framingham risk levels) were produced for each antipsychotic treatment group. This analysis also applied Fisher's Exact Test to test the association between the chosen treatment group and two key metabolic parameters: overweight (BMI (≥ 25 versus < 25) and elevated fasting glucose (≥ 125 mg/dL versus < 125 mg/dL). Finally, frequencies of reasons invoked by the study psychiatrists as important in their prescribing decisions were calculated for patients as described above. 3 Results 3.1 Patient characteristics A total of 80 patients fulfilled study entry criteria and had complete data abstracted from the medical record ( Table 1 ). The most common drugs prescribed were risperidone or quetiapine with 38 (48%) of the patients on either of these two medications ( Table 2 ). First-generation drugs were prescribed to 17 (21%); all but one patient in the first-generation group received haloperidol and seven of these were on haloperidol decanoate. Thirteen patients (16%) received olanzapine or clozapine, and 12 (15%) received ziprasidone or aripiprazole. The next best drug chosen for symptom control varied substantially across initial drugs chosen ( Table 2 ). The next best drug for metabolic risk was typically ziprasidone or aripiprazole, regardless of initial drug chosen. 3.2 Association between antipsychotic drug choice and VAS ratings of predicted symptom control and metabolic risk Ratings of symptom control and metabolic risk were significantly associated with choice of initial antipsychotic when jointly entered in a logistical regression model, together with age, glucose level, and BMI level (Wald Χ 2 = 29.4, df = 5, p < 0.0001). Psychiatric investigators rated metabolic risk lower when they placed a patient on ziprasidone or aripiprazole than when they chose another antipsychotic (odds ratio per unit increment in VAS (OR) = .88, p = 0.01). Similarly, VAS ratings for certainty of symptom control were significantly and independently associated with choice of antipsychotic (OR = .93, p = 0.02) such that psychiatric investigators rated certainty of symptom control lower when ziprasidone or aripiprazole were prescribed than when they chose another antipsychotic. 3.3 Mean differences between antipsychotic drug chosen and next best drug in predicted symptom control and metabolic risk The mean difference in VAS ratings of drug chosen compared with the next best drug differed significantly across treatment groups for metabolic risk ( F = 14.80, df = 3, 76; p = 0.0001) and certainty of symptom control ( F = 2.81, df = 3, 76; p < 0.05) ( Table 2 ). In pairwise comparisons between treatment groups, differences in predicted metabolic risk between the selected drug and next best choice were found, with the ziprasidone/aripiprazole group expected to have significantly less risk versus the other treatment groups ( p < 0.05) except for the first-generation antipsychotic group ( p = 0.37). Differences in the symptom control VAS ratings between the selected drug and next best choice were also significant, with the ziprasidone/aripiprazole group expected to show less certainty of symptom control compared to the second choice versus either the olanzapine/clozapine group ( p = 0.02) or the first-generation group ( p = 0.01), but not the risperidone/quetiapine group ( p = 0.12). 3.4 Relationship between antipsychotic drug choice and baseline metabolic parameters Treatment groups were compared on baseline metabolic parameters using distribution plot displays for the relationship between treatment group and BMI, fasting glucose (mg/dL), blood pressure (mm Hg) and Framingham risk levels (see Supplementary Materials). Visual examination of the displays indicated that treatment groups did not differ on mean BMI, fasting glucose, systolic or diastolic blood pressure, and Framingham risk levels calculated by the Wilson method ( Wilson et al., 1998 ). Cholesterol and triglycerides were not routinely available in patients' medical records and could not be analyzed by antipsychotic group. Fisher's Exact Tests of the relationship between treatment choice and baseline fasting glucose and BMI elevations were non-significant ( Table 3 ). 3.5 Frequency of reasons cited for antipsychotic choice In total, a broad range of 24 out of 35 unique reasons were endorsed at least once as influencing treatment choice. History of past medication responsiveness and presence or severity of positive symptoms were each invoked in about half of the patients as reasons for prescribing decisions ( Table 4 ). Metabolic reasons were cited more frequently in patients with elevated BMI and fasting glucose levels; any metabolic reason was cited in 15% of patients with BMI < 25 but in 36% of patients with BMI ≥ 25, and in 25% of patients with glucose levels < 125 mg/dL but in 41% of patients with glucose levels ≥ 125 mg/dL (see Supplementary Materials). Concerns over extrapyramidal effects were rated below metabolic risks in frequency. 4 Discussion Emerging evidence on antipsychotic effectiveness underscores the need for clinicians to individualize treatment, balancing relative efficacy and risks of antipsychotic drugs matched with patient vulnerabilities ( Lieberman et al., 2005 ). How clinicians balance competing benefits and risks in selecting among antipsychotics has not been studied, in part, due to the lack of an appropriate quantitative method. Our results provide preliminary support for the validity of the ATCQ as an instrument providing quantifiable data useful in understanding decision making processes related to prescribing practices. Although we chose to focus on metabolic risk factors in this analysis, the VAS methodology used in the ATCQ could be modified to address other questions of interest in assessing trade-offs between beneficial and adverse effects of antipsychotic drugs. Using the ATCQ, we found that ratings of both the predicted certainty of symptom control and metabolic risk correlated significantly and independently with treatment decisions indicating the relevance of these domains to antipsychotic decision making. Comparisons of ratings of selected treatments with next best alternative drugs showed that significant differences between treatment groups in predicted symptom control were small but often had greater influence on treatment choice relative to significant but larger differences in expected metabolic risk. For example, in patients for whom clozapine or olanzapine were prescribed, a 45.8% increase in expected metabolic risk was accepted by the psychiatrists to gain 6.8% more certainty in symptom control compared with the next best drugs ( Table 2 ). Decisions to prescribe other antipsychotics showed less dramatic trade-offs. The risperidone/quetiapine group was associated with a willingness to accept a 21.6% increase in metabolic risk with essentially no perceived advantage in symptom control. Decisions to prescribe a lower metabolic risk drug (ziprasidone/aripirazole) involved the opposite trade-off seen for other agents; a 7.5% loss in certainty of symptom control was accepted by the prescriber to gain risk neutrality for metabolic effects. In other words, the expectation of less likely control of psychotic symptoms was accepted for not worsening the metabolic status of the patient. Most surprisingly, the first-generation treatment group was associated with an expectation of increased certainty in symptom control (6.5%) comparable to the olanzapine/clozapine group, but with minimal difference in expected metabolic risk (5.6%) that was not significantly different than the ziprasidone/aripiprazole group. This favorable perception of the first-generation drugs is consistent with recent trial data suggesting that the older drugs may be seen as more useful in clinical practice than previously thought in comparison with second-generation drugs when considered in terms of overall effectiveness ( Geddes et al., 2000; Lieberman et al., 2005; Rosenheck et al., 2003 ). In contrast to the psychiatrists' predictions of expected risks and benefits with selected treatments for individual patients, we found that the actual metabolic status of the patient at baseline (BMI, fasting glucose, blood pressure, Framingham risk level) considered in isolation from other factors did not significantly account for the initial antipsychotic treatment choice consistent with prior research ( Morrato et al., 2009a ). Instead, the metabolic status of the patient was seriously considered but only in relation to its severity and other clinical factors. For instance, metabolic reasons were cited as factors in the prescribing decision of 15% of normal weight but 36% of overweight patients and in 25% of patients with normal glucose levels but in 41% of patients with elevated glucose levels. In addition, 24 separate clinical factors were cited in association with antipsychotic prescribing decisions, which may explain why the metabolic status of patients at baseline alone was obscured and did not correlate significantly with treatment choice. Furthermore, this suggests that treatment guidelines that focus on a single symptom dimension or adverse drug effect, e.g., baseline metabolic status ( American Diabetes Association and American Psychiatric Association, 2004 ), may be constraining and provide an incomplete or even misleading picture of the combination of variables that clinicians need to balance in prescribing decisions. There were several limitations affecting the interpretation of study results. First, the psychiatrists were aware that their prescribing decisions were being examined which could have influenced their responses in contrast to usual clinical practices, although questionnaires were completed and analyzed anonymously. In addition, the authenticity of responses may have been enhanced by having doctors respond to the ATCQ in relation to actual decisions they made for real patients under their care. We did not record independent measures of patient severity or history of treatment response, which could have been used to test the accuracy of the doctors' clinical assessments, because we were primarily interested not in the accuracy of assessment techniques or medical records, but rather in how doctors weighed and prioritized their understanding of patient characteristics and needs in deciding on treatment choice. Similarly, we did not include in the analyses the most recent antipsychotic taken by patients, or whether any antipsychotic was taken prior to admission, which could have influenced treatment choice; however, doctors were able to select history of past medication treatment failure or side effects as reasons for antipsychotic choice. Because the ATCQ questionnaire was changed during the course of the study to limit the nonspecific use of “Patient choice or preference” and the “Caregiver choice or preference”, caution should be applied when interpreting this aspect of the results. Prolonged QTc was listed among metabolic risk factors in the ATCQ, as another cardiovascular risk, but would have been less confusing if listed among “Other considerations”; however, prolonged QTc was not selected as important in prescribing in any case and had no effect on results. The small sample of patients from an inpatient program, which was composed mostly of middle-aged and chronically-ill males, with a high degree of symptom acuity and metabolic risk at baseline, may not generalize to antipsychotic prescribing decisions in other settings with different sample populations. For example, the inpatient priority of rapid control of acute symptoms at the expense of long-term metabolic risk may be reversed in outpatient or long-term care settings. Also, the cited reasons based on history of treatment response or non-adherence would not be relevant in first-episode patients. Diagnostically heterogeneous sampling complicates the interpretation of the results as patients with schizophrenia, schizoaffective or bipolar disorder were included. Decisions regarding the management of schizophrenia, mania, and other indications could not be examined separately because of the small sample size. Also, results could not be stratified by key patient characteristics which would be important in future research. Furthermore, the sample size was also limited by surveying only 4 psychiatrists on an inpatient program of a Veterans Affairs Medical Center; however, these psychiatrists were senior, academic clinicians with extensive experience treating psychosis, who were well-informed about evidence-based guidelines for prescribing antipsychotic drugs. We intended to assess reasons cited for prescribing decisions by these psychiatrists based on their standard clinical knowledge and experience; we did not propose to determine to what extent their opinions derived from evidence-based trials, facility routine, industry marketing programs or experience gained in clinical practice, though these important influences could be addressed in further studies. Future studies could aim also for increased generalizability with larger samples of psychiatrists and patients with a broader range of demographics in different clinical settings. Additional studies would be required to further examine the reliability, internal consistency and validity of the ATCQ before more widespread use. 5 Conclusions Taken together, these preliminary results suggest that antipsychotic prescribing decisions for acutely-ill hospitalized patients are determined by numerous patient and treatment factors. Predictions of both symptom control and metabolic risk were significantly associated with treatment choice, but study psychiatrists were willing to assume relative degrees of metabolic risk in favor of effective control of psychotic symptoms when choosing between different antipsychotic drugs for these hospitalized patients. In particular, a patient's history of treatment response and severity of positive symptoms were frequently cited as most important by psychiatrists in making a treatment choice. In total, 24 different reasons for medication choice were referenced on the ATCQ by the psychiatrists, indicating the complexity of decision making. This suggests that focus on a single symptom dimension or adverse drug effect may be constraining, and provide an incomplete or even misleading picture of the combination of variables that clinicians consider in prescribing decisions. Further validation and studies of prescribing decisions with larger samples in different settings using the ATCQ methodology could enhance medical education, translation of research findings and application of treatment guidelines. The following are the supplementary materials related to this article. ATCQ Boxplots of metabolic parameters ATCQ Questionnaire Reasons for drug choice by BMI and gl Supplementary materials related to this article can be found online at doi:10.1016/j.pnpbp.2010.11.018 . Acknowledgements Presented as an abstract at the American College of Neuropsychopharmacology, Scottsdale, AZ, 12-9-08. This study was sponsored by Pfizer Inc . This material is based upon work also supported in part by the Department of Veterans Affairs, Veterans Health Administration, Office of Research Development , with resources and the use of facilities at the Philadelphia Veterans Affairs Medical Center. The content of this work does not represent the views of the Department of Veterans Affairs or the United States Government. Disclosures of interests: Drs. Herman, Cuffel, Dodge and Ms. Sanders were employees of Pfizer, Inc. at the time the study was conducted. Dr. Herman is currently an employee of Sunovion Pharmaceuticals, Inc.. Drs. E. Cabrina Campbell, M.D., Melissa DeJesus, M.D., Vasant Dhopesh, M.D., and Stanley N. Caroff, M.D. were employees of the Philadelphia Veterans Affairs Medical Center which received funding from Pfizer, Inc. for this project. Dr. Caroff served as a consultant to Eli Lilly, Inc. The remaining authors have no interests to disclose. References Allison and Casey, 2001 Allison D.B. Casey D.E. Antipsychotic-induced weight gain: a review of the literature J Clin Psychiatry 62 Suppl. 7 2001 22 31 American Diabetes Association and American Psychiatric Association, 2004 American Diabetes Association American Psychiatric Association Consensus development conference on antipsychotic drugs and obesity and diabetes J Clin Psychiatry 65 2 2004 267 272 Backlund et al., 2000 L. Backlund B. Danielsson J. Bring Factors influencing GPs' decisions on the treatment of hypercholesterolaemic patients Scand J Prim Health Care 18 2 2000 87 93 Buchanan et al., 2002 R.W. Buchanan J. Kreyenbuhl J.M. Zito The schizophrenia PORT pharmacological treatment recommendations: conformance and implications for symptoms and functional outcome Schizophr Bull 28 1 2002 63 73 Correll et al., 2006 C.U. Correll A.M. Frederickson J.M. Kane Metabolic syndrome and the risk of coronary heart disease in 367 patients treated with second-generation antipsychotic drugs J Clin Psychiatry 67 4 2006 575 583 Covell et al., 2002 N.H. Covell C.T. Jackson A.C. Evans Antipsychotic prescribing practices in Connecticut's public mental health system: rates of changing medications and prescribing styles Schizophr Bull 28 1 2002 17 29 Davis et al., 2003 J.M. Davis N. Chen I.D. Glick A meta-analysis of the efficacy of second-generation antipsychotics Arch Gen Psychiatry 60 6 2003 553 564 Essock, 2002 Essock S.M. Editor's introduction: antipsychotic prescribing practices Schizophr Bull 28 1 2002 1 4 Fisch et al., 1981 H.U. Fisch K.R. Hammond C.R. Joyce An experimental study of the clinical judgment of general physicians in evaluating and prescribing for depression Br J Psychiatry 138 1981 100 109 Geddes et al., 2000 J. Geddes N. Freemantle P. Harrison Atypical antipsychotics in the treatment of schizophrenia: systematic overview and meta-regression analysis BMJ 321 7273 2000 1371 1376 Hamann et al., 2004 J. Hamann B. Langer S. Leucht Medical decision making in antipsychotic drug choice for schizophrenia Am J Psychiatry 161 7 2004 1301 1304 Henderson, 2005 Henderson D.C. Schizophrenia and comorbid metabolic disorders J Clin Psychiatry 66 Suppl. 6 2005 11 20 Hoblyn et al., 2006 J. Hoblyn A. Noda J.A. Yesavage Factors in choosing atypical antipsychotics: toward understanding the bases of physicians' prescribing decisions J Psychiatr Res 40 2 2006 160 166 Kaplan et al., 1993 R.M. Kaplan D. Feeny D.A. Revicki Methods for assessing relative importance in preference based outcome measures Qual Life Res 2 6 1993 467 475 Kirwan et al., 1990 Kirwan J.R. Chaput de Saintonge D.M. Joyce C.R. Clinical judgment analysis Q J Med 76 281 1990 935 949 Lieberman et al., 2005 J.A. Lieberman T.S. Stroup J.P. McEvoy Effectiveness of antipsychotic drugs in patients with chronic schizophrenia N Engl J Med 353 12 2005 1209 1223 Linden et al., 2006 M. Linden L. Pyrkosch R.W. Dittmann Why do physicians switch from one antipsychotic agent to another? The "physician drug stereotype" J Clin Psychopharmacol 26 3 2006 225 231 Marder, 2002 Marder S.R. Can clinical practice guide a research agenda? Schizophr Bull 28 1 2002 127 129 Marder et al., 2002 S.R. Marder S.M. Essock A.L. Miller The Mount Sinai conference on the pharmacotherapy of schizophrenia Schizophr Bull 28 1 2002 5 16 McEvoy et al., 2005 J.P. McEvoy J.M. Meyer D.C. Goff Prevalence of the metabolic syndrome in patients with schizophrenia: baseline results from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) schizophrenia trial and comparison with national estimates from NHANES III Schizophr Res 80 1 2005 19 32 Miller et al., 1999 A.L. Miller J.A. Chiles J.K. Chiles The Texas Medication Algorithm Project (TMAP) schizophrenia algorithms J Clin Psychiatry 60 10 1999 649 657 Morrato et al., 2009a E.H. Morrato B. Cuffel J.W. Newcomer Metabolic risk status and second-generation antipsychotic drug selection: a retrospective study of commercially insured patients J Clin Psychopharmacol 29 1 2009 26 32 Morrato et al., 2009b E.H. Morrato J.W. Newcomer S. Kamat Metabolic screening after the American Diabetes Association's consensus statement on antipsychotic drugs and diabetes Diab Care 32 6 2009 1037 1042 Nasrallah, 2003 Nasrallah H. A review of the effect of atypical antipsychotics on weight Psychoneuroendocrinology 28 Suppl. 1 2003 83 96 Rosenheck et al., 2003 R. Rosenheck D. Perlick S. Bingham Effectiveness and cost of olanzapine and haloperidol in the treatment of schizophrenia: a randomized controlled trial JAMA 290 20 2003 2693 2702 Sernyak et al., 2002 M.J. Sernyak D.L. Leslie R.D. Alarcon Association of diabetes mellitus with use of atypical neuroleptics in the treatment of schizophrenia Am J Psychiatry 159 4 2002 561 566 Stroup et al., 2006 T.S. Stroup J.A. Lieberman J.P. McEvoy Effectiveness of olanzapine, quetiapine, risperidone, and ziprasidone in patients with chronic schizophrenia following discontinuation of a previous atypical antipsychotic Am J Psychiatry 163 4 2006 611 622 Weiden et al., 2006 P.J. Weiden A.H. Young P.F. Buckley The art and science of switching of antipsychotic medications, part 1 J Clin Psychiatry 67 11 2006 e15 Wilson et al., 1998 P.W. Wilson R.B. D'Agostino D. Levy Prediction of coronary heart disease using risk factor categories Circulation 97 18 1998 1837 1847