Medication noncompliance is considered to be the single most important factor leading to relapse among patients with schizophrenia.1 One meta-analysis of maintenance antipsychotic treatment found that noncompliance causes about 40% of relapses.2 The advent of newer, second-generation oral antipsychotics has had a greater impact on efficacy than on adherence. For example, patients maintained on second-generation antipsychotics are about one third less likely to relapse compared with those receiving first-generation medications,3 but the overall gains in preventing nonadherence by moving from first-generation to second-generation medications have been much more modest.4
Schizophrenia is a very debilitating and expensive disease, resulting in $62.7 billion in direct and indirect costs in 2002 in the United States, of which $32.4 billion was indirect costs, including unemployment, reduced workplace productivity, family caregiving, and suicide. Direct healthcare costs totaled $22.8 billion ($8.0 billion in long-term care, $7.0 billion in outpatient costs, $2.8 billion inpatient, and $5.0 billion in drug therapy). In 2002, direct healthcare costs in terms of per member per year equaled $79 per US resident.5
Copay and Compliance
Studies in other therapeutic areas have demonstrated that increases in prescription drug copayments decrease compliance rates. For newly initiated renin angiotensin system blockers, every $1 increase in copayments per 30-day supply resulted in a 1.9% increased gap between refills, and a 2.8% increase in the odds of being nonpersistent (discontinued treatment).6 In a study of antihypertensive therapies, patients with a $20 copay had 76% of the compliance of those with a $5 copay; and those with a $20 to $165 copay had 48% of the compliance of those with a $5 copay.7 With oral hypoglycemic therapies, patients with a $10 increase in copay had an 18.5% reduction in utilization from baseline.8 An increase in copayment from $10 to $20 for statin therapy resulted in a 6% to 10% reduction in the number of fully compliant patients.9 When copayments were doubled across all therapeutic categories in one health plan, reductions in utilization ranged from 45% for nonsteroidal antiinflammatory drugs to 25% for diabetes medications.10 Conversely, a decrease in copayment resulted in a 7% to 14% increase in compliance in 4 of the 5 chronic medication categories studied.11 The relationship between compliance and copayment in schizophrenia has not been reported.
Other Factors Influencing Compliance In real-world practice, medication-taking behavior is complex and influenced by factors other than copayment. In a published review of 39 studies, the patientlevel factors most often associated with nonadherence in patients with schizophrenia included poor insight into the disease, negative attitude toward the medication, previous nonadherence, substance abuse, short illness duration, poor discharge planning, and poor therapeutic alliance.12
Representation of compliance as an "all-or-none" behavior, as implied by noncompliance versus compliance, does not accurately describe patient behaviors. In fact, several studies have found that "partial compliance" (taking some, but not all, prescribed medication) more accurately describes the pattern for most patients with schizophrenia.13,14 Partial compliance can take several forms, including taking an amount consistently less than that recommended, displaying irregular ("on" and "off") dosing behavior, and having discrete gaps in antipsychotic therapy (eg, patients unwilling or unable to refill prescriptions). Therefore, a new research emphasis has been placed on assessing the pattern and extent of noncompliant behaviors, and the degree of relapse risk associated with various levels of partial compliance.15,16
Administrative pharmacy claims data have become an increasingly important tool in compliance pattern evaluations,17,18 predicated on the general principle that the upper bound of patient compliance is limited by the availability of sufficient medication. Prescription claims have been used to estimate compliance in such chronic disease states as hypertension,19 congestive heart failure,20 and glaucoma.21 In combination with medical records, pharmacy claims data can also be used to demonstrate relationships between compliance behavior and outcomes (eg, hospitalization risk and increased healthcare costs) and are useful in examining patient response to disease management programs designed to enhance compliance behavior.22
We utilized 4 measures of partial compliance: medication possession ratio (MPR), medication consistency, gaps in medication therapy, and medication persistence.13,23,24 MPR, the most frequently used assessment in compliance evaluation, is a composite measure that incorporates assessment of both "skipping" and "quitting." 13,23 Consistency and medication gaps estimate compliance with reference to "skipped" doses. Persistence identifies discontinuation (quitting) behavior or a gap at the end of the observation period. A detailed description of each measure with an example calculation is provided in Table 1.
Utilizing all these 4 definitions, a recent study by Weiden and colleagues examined the relationship between partial compliance with antipsychotic treatment (MPR <0.7) and mental health hospitalization risk among 4325 outpatients diagnosed with schizophrenia in a California Medicaid database.24 The investigators found that lower compliance (across all 4 definitions) was associated with increased odds of hospitalization. Another study showed a similar relationship between partial or nonadherence to antipsychotic medication and rates of psychiatric hospitalizations in a Medicaid database of patients diagnosed with schizophrenia.25 Partial compliance has also been shown to be a risk factor for hospitalization in the Veterans Administration, with patients having lower compliance (MPR <0.8) almost 2.5 times more likely to be admitted than patients with higher compliance (MPR 0.8-1.1).23
Previous research has not determined if the relationship between compliance and hospitalization exists in patients with schizophrenia in a managed care plan. The purpose of our analysis was to investigate if the relationship between partial compliance and increased hospital risk exists in a national managed care population.
At the time of our data extraction, the PHARMetrics data set contained claims data from more than 57 managed care organizations and 33 million patients, representing a cross section of the United States. This mental health subset was extracted from the larger PHARMetrics managed care database encompassing data from January 1, 2000, through February 28, 2002. Internal Review Board assessment was unnecessary because all personal identifiers were removed and at no time did investigators have knowledge of patients' identities. Patients meeting all the following criteria were included:
- A diagnosis of schizophrenia or schizoaffective disorder (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes of 295.XX, 296.0X, 296.1X, 296.4X, or 296.9X) at any point during the period of available data
- At least 1 oral antipsychotic claim in 1999
- At least 2 dispensing events for any antipsychotic medication during the specified enrollment period (January 1, 2000-December 31, 2001)
- Continuous enrollment in the same plan for at least 13 months after the index date (defined as the date of the patient's first antipsychotic prescription during the enrollment period).
As in the Weiden study,24 patients were excluded if they resided in a long-term care setting, received decanoate forms of antipsychotics, had a ratio of days dispensed to days' supply of ≥10 (an indication of potential problems with the days' supply field), had any missing or zero values in the quantity dispensed or days' supply field for any prescription event, had a bipolar diagnosis (ICD-9-CM codes of 296.0, 296.1, 296.4-296.8), or were more than 18 years of age. This type of exclusion ensures that medication gaps accurately depict partial compliance in patients with schizophrenia.24
The primary dependent measure was a dichotomous variable indicating the presence of at least 1 "mental health–related" hospitalization in the postindex year. Each unique hospitalization event was identified using an algorithm supplied by PHARMetrics. ICD-9-CM codes were selected for diagnoses for which antipsychotic use might prevent hospitalization.24
Compliance results were analyzed as both categorical and continuous measures. Using chi-square analysis, compliance measures were categorized according to predefined ranges to assess the relationship between categories and risk of hospitalization. Maximum gap categories were defined as 0, 1–10, 11–30, or >30 days; medication consistency and MPR categories, ≥70% compliance as good compliance and <70% compliance as partial compliance; medication persistence categories, ≥90% compliant and <90% as partial compliance or discontinued therapy. The literature suggests that a cutoff at 70% for MPR compliance measurements is reasonable. These levels were selected to maintain consistency in reporting the findings between this managed care population and the Medicaid population.24
Continuous measures were used for the logistic regression analyses to model relationships between various compliance measures and the presence of a mental health hospitalization during the 12-month follow-up. Independent models were used to analyze maximum gap, consistency and persistence, persistence alone, and MPR. Age and gender were also included in the logistic regression models as independent factors. Interactions were retained in the model if they added significantly to the explanatory power; variables were eliminated if they were insignificant and had a negative impact on model fit.
A total of 1499 patients met study inclusion and exclusion criteria, of which 699 (46.6%) were men and 800 (53.4%) were women (Table 2). The mean patient age (years ± SD) was 45.1 ± 12.4, ranging from 18 to 95 (median 46). Some 89 patients (5.9%) had at least 1 mental health hospitalization during the 12 months after the index date. The mean number ± SD of mental health hospitalizations per patient was 0.08 ± 0.35; the maximum number of mental health hospitalizations was 4. The mean number of hospitalized days per patient was 0.68 ± 3.94 (range 0-87).
The mean number ± SD of dispensing events (prescription claims) in this population was 12.9 ± 9.3, and there was an average ± SD of 1.34 ± 0.63 different antipsychotic drug entities dispensed per patient. A total of 1399 patients had at least 3 dispensing events for a single drug, permitting assessment of medication consistency. On average, the study population seemed compliant, with mean values ± SD of 0.833 ± 0.206 for consistency, 0.926 ± 0.193 for persistence, and 0.761 ± 0.247 for MPR. The average theoretical maximum gap in therapy ± SD was 50.8 ± 72.5 days (median 19 days).
Regarding medication gap analysis, the percentage of patients with at least 1 mental health hospitalization rose with each increasing increment of maximum gap (Figure 1). A significantly higher proportion of patients with a maximum gap >30 days (10.1%) experienced a mental health hospitalization than did those with maximum medication gaps of 1 to 10 days (2.3%; P <.001 vs >30 days) and 11 to 30 days (4.6%; P = .0016 vs >30 days). Comparisons to patients with a maximum gap of 0 days are noteworthy despite the relatively small number of patients (n = 33). No patients with 0 days of maximum gap were hospitalized.
A 0- to 10-day gap category (combining the 0 and 1-10–day groups) was used as the reference group in the logistic regression since there were no hospitalizations for patients in the 0 days of maximum gap category (ie, the logistic regression will not converge). Compared with patients in the 0- to 10-day group, the odds of hospitalization were 2.099 times greater (P = .0569) and 4.659 times greater (P <.001) for the 11- to 30-day and >30-day groups, respectively (Table 3).
Maximum gap length was also evaluated as a continuous measure using logistic regression analysis. In this model, each increase of 5 days in maximum potential gap per prescription at any time during the 1-year study was associated with a 2.1% increase (P = .004) in the odds of hospitalization. In addition, a 10-year increase in age was associated with reductions in the risk of hospitalization in all maximum gap models described above (25.0%-26.7%; largest P =.018).
Partial compliance, assessed using consistency and MPR, was significantly associated with increased risk of hospitalization (Figure 2). Patients who were ≥70% compliant, as measured by the MPR, were significantly less likely to have been hospitalized than those who were <70% compliant (4% vs 10.4%; P <.001), as were those with consistency ≥70% versus <70% (4.0% vs 9.6%; P <.001). Using the more common 80% criterion for MPR and consistency, similar results were found. Patients who were ≥80% compliant, as measured by the MPR, were significantly less likely to have been hospitalized than those who were <80% compliant (3% vs 10.4%; P <.001), as were those with consistency ≥80% versus <80% (3.8% vs 8.5%; P <.001). The difference in hospitalization risk among patients who were <90% compliant by the persistence measure was not statistically significant when compared with patients who were =90% compliant (8.2% vs 5.5%; P = .1096).
Logistic regression analyses supported the categorical analysis findings. The odds of hospitalization were significantly affected by both compliance and age when using either the MPR or the consistency models. A 10% increase in compliance was associated with reductions in the odds of hospitalization of 16.9% for MPR (P <.001) and 18.8% for consistency (P <.001). A 10-year increase in age was associated with reductions in the risk of hospitalization, by 31.1% in the consistency model (P = .003), by 28.4% in the persistence model (P = .004), and by 26.1% in the MPR model (P = .012). Similar to the categorical analysis findings, medication persistence did not reach statistical significance as a predictor of hospitalization (P = .1334). Patient gender was not significant in the logistic regression models.
The results of this retrospective study of pharmacy claims data from a large, national managed care database support the relationship, among patients with schizophrenia, between partial compliance with antipsychotic medication regimens and increased risk of mental health hospitalization. The relationship was statistically significant across 3 of the 4 definitions of compliance measured: maximum gap, medication consistency, and MPR. The results from categorical and logistic regression analyses were also similar across 3 of the 4 definitions. Although there is ample literature documenting noncompliance and increased relapse risk, many previous studies had only one measure for assessing noncompliance. Few studies have analyzed the impact of a comprehensive range of compliance behaviors.
The conceptual framework and the compliance variables employed in our study are similar to those used in a recent study of patients with schizophrenia in a Medicaid population.23 The Medicaid study, based on a larger patient population (4325 vs 1499), produced results remarkably similar to those in the present study with regard to the relationship between compliance and hospitalization and the statistical significance of the findings. Both the Medicaid study and the current analysis were based on patient populations that had low discontinuation rates: 97.0% and 84.6% of the respective populations achieved persistence for >90% of the 1-year study period. Even within these relatively compliant, yet different patient populations, there was a significant and consistent relationship between multiple measures of partial compliance and increased hospitalization risk.
The maximum gap findings, which defined 4 different categories of compliance as independent variables, are particularly compelling in this regard: the odds of hospitalization rose with each increment of partial compliance. Each 5-day increase in the maximum potential gap in therapy was associated with a 2.1% increase in the odds of hospitalization (eg, from 5.9% to 6.02% for a 5-day gap, or a 50-day gap increases the patients' baseline risk level by 21%). If their baseline risk is 5.9%, the risk of hospitalization is 7.1% (5.9 X 1.21) after a maximum potential 50-day gap in therapy. The average maximum gap in this population was 50.8 days between 2 prescriptions.
This managed care study and the Weiden Medicaid study had the same directional relationship between persistence and hospitalization risk; however, the relationship was not statistically significant in the managed care population. The difference can be partially attributed to the different study population sizes. A relationship was expected because persistence is the measurement of the maximum gap that exists (if any) at the end of the study period.
The most significant difference between the 2 study populations is the overall rate of psychiatric hospitalization; 5.9% in the managed care study versus 15.1% in the Medicaid study.17 It is difficult to identify a single factor that explains the 2.5-fold difference. It is possible that Medicaid patients may represent a more severe illness spectrum than do those in the PHARMetrics database. Although hospitalization rates differed in the 2 studies, both analyses indicated a significant relationship between partial compliance and risk of mental health hospitalization. Copayments were likely to be higher in the managed care (commercial) population versus the Medicaid population, and their hospitalization rates were lower.
This study was subject to all of the limitations of a nonexperimental study. The presence of an association does not necessarily indicate causation. It is possible, and perhaps likely, that patients with poor compliance may also have other poor health habits that might be related to hospitalization; however, given the theoretical link between compliance and relapse in schizophrenia, the results are logical from a clinical standpoint. A double-blind randomized trial in which patients are randomly assigned to a noncompliant or subtherapeutic regimen with gaps in therapy is not likely to be seen as ethical, so nonexperimental designs are necessary.
Other limitations in our study that may restrict the generalizability of these findings include the retrospective nature of the analysis and the use of inclusion/exclusion criteria that resulted in a population with slightly higher levels of compliance than those in the general population (ie, patients who had <2 dispensing events/claims [ultra-low compliance, tolerability, or lack of efficacy] during the study period were excluded). Our data analysis was restricted to variables consistently represented in the PHARMetrics database and to plans that included "days' supply" and "quantity dispensed" as part of pharmacy records. These data were necessary to calculate the partial compliance measures. Patient samples are not included in the claims database.
Future research using medical records could be conducted to assess the association between compliance and more intermediate outcomes of relapse that occur before hospitalization. This would provide a more sensitive understanding of the negative patient outcomes prehospitalization and their relationship to compliance. Furthermore, a study conducted with intermediate negative outcomes from medical records could explore the potential modifiers of hospitalization risk that are not reflected in a claims database, including baseline disease severity, duration of schizophrenia, socioeconomic status, and the availability and use of nonpharmacologic (psychosocial) therapies.
Increased attention to improving compliance across the entire range of partial compliance behavior and removing barriers to compliance (by using behavioral interventions, removing or lowering economic barriers such as copayments, or using new compliance-friendly drug delivery systems) may improve outcomes and lower costs in schizophrenia treatment. Future research is needed to determine if removal of economic barriers (copayments) directly results in a reduction in relapse costs through an improvement in compliance.
The findings of this study, based on a managed care population of patients with schizophrenia, indicate that partial compliance with oral antipsychotic treatment is associated with a significantly increased risk of hospitalization. These results are consistent with a previous analysis of compliance measures in a Medicaid population of patients with schizophrenia. This present study highlights the central role of medication compliance in long-term treatment of schizophrenia.
This study was supported by Ortho-McNeil Janssen Medical Affairs.
Dr Kozman is a consultant for Ortho-McNeil Janssen. Dr Weiden receives grant support from AstraZeneca, Bristol-Myers Squibb/Otsuka America Pharmaceutical, and Ortho-McNeil Janssen, and is a consultant for and on the Speaker's Bureau for AstraZeneca, Bristol-Myers Squibb/Otsuka America Pharmaceutical, Ortho-McNeil Janssen, Organon, Pfizer, Shire, Vanda, and Wyeth.
1. Ayuso-Gutierrez JL, del Rio Vega JM. Factors influencing relapse in the long-term course of schizophrenia. Schizophr Res. 1997;28:199-206.
2.Weiden PJ, Olfson M. Cost of relapse in schizophrenia. Schizophr Bull. 1995;21:419-429.
3. Csernansky JG, Mahmoud R, Brenner R; Risperidone-USA-79 Study Group. A comparison of risperidone and haloperidol for the prevention of relapse in patients with schizophrenia. N Engl J Med. 2002;346:16-22.
4. Velligan DI, Lam F, Ereshefsky L, et al. Psychopharmacology: perspectives on medication adherence and atypical antipsychotic medications. Psychiatr Serv. 2003;54:665-667.
5. Wu EQ, Birnbaum HG, Shi L, et al. The economic burden of schizophrenia in the United States in 2002. J Clin Psychiatry. 2005;66:1122-1129.
6. Zhang D, Carlson AM, Gleason PP, et al. Relationship of the magnitude of member cost-share and medication persistence with newly initiated renin angiotensin system blockers. J Manag Care Pharm. 2007;13: 664-676.
7. Taira DA, Wong KS, Frech-Tamas F, et al. Copayment level and compliance with antihypertensive medication: analysis and policy implications for managed care. Am J Manag Care. 2006;12:678-683.
8. Roblin DW, Platt R, Goodman MJ, et al. Effect of increased cost-sharing on oral hypoglycemic use in five managed care organizations: how much is too much? Med Care. 2005;43:951-959.
9. Goldman DP, Joyce GF, Karaca-Mandic P. Varying pharmacy benefits with clinical status: the case of cholesterol-lowering therapy. Am J Manag Care. 2006;12:21-28.
10. Goldman DP, Joyce GF, Escarce JJ, et al. Pharmacy benefits and the use of drugs by the chronically ill. JAMA. 2004;29:2344-2350.
11. Chernew ME, Shah MR, Wegh A, et al. Impact of decreasing copayments on medication adherence within a disease management environment. Health Aff (Millwood). 2008;27:103-112.
12. Lacro JP, Dunn LB, Dolder CR, et al. Prevalence of and risk factors for medication nonadherence in patients with schizophrenia: a comprehensive review of recent literature. J Clin Psychiatry. 2002;63:892-909.
13. Docherty JP, Grogg AL, Kozman C, et al. Antipsychotic maintenance in schizophrenia: partial compliance and clinical outcome. Poster presented at: the American College of Neuropsychopharmacology 41st Annual Meeting; December 8-12, 2002; San Juan, Puerto Rico.
14. McCombs JS, Nichol MB, Stimmel GL, et al. Use patterns for antipsychotic medications in Medicaid patients with schizophrenia. J Clin Psychiatry. 1999;60(suppl 19):5-11; discussion 12-13.
15. Keith SJ, Kane JM. Partial compliance and patient consequences in schizophrenia: our patients can do better. J Clin Psychiatry. 2003;64: 1308-1315.
16. Marder SR. Overview of partial compliance. J Clin Psychiatry. 2003;64(suppl 16):3-9.
17. Steiner JF, Koepsell TD, Fihn SD, et al. A general method of compliance assessment using centralized pharmacy records: description and validation. Med Care. 1988;26:814-823.
18. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol. 1997;50:105-116.
19. Maronde RF, Chan LS, Larsen FJ, et al. Underutilization of antihypertensive drugs and associated hospitalization. Med Care. 1989;27:1159-1166.
20. Monane M, Bohn RL, Gurwitz JH, et al. Noncompliance with congestive heart failure therapy in the elderly. Arch Intern Med. 1994;154:433-437.
21. Gurwitz JH, Glynn RJ, Monane M, et al. Treatment for glaucoma: adherence by the elderly. Am J Public Health. 1993;83:711-716.
22. Skaer TL, Sclar DA, Markowski DJ, et al. Effect of value-added utilities on prescription refill compliance and Medicaid health care expenditures—a study of patients with non-insulin-dependent diabetes mellitus. J Clin Pharm Ther. 1993;18:295-299.
23. Valenstein M, Copeland LA, Blow FC, et al. Pharmacy data identify poorly adherent patients with schizophrenia at increased risk for admission. Med Care. 2002;40:630-639.
24. Weiden PJ, Kozma C, Grogg A, et al. Partial compliance and risk of rehospitalization among California Medicaid patients with schizophrenia. Psychiatr Serv. 2004;55:886-891.
25. Gilmer TP, Dolder CR, Lacro JP, et al. Adherence to treatment with antipsychotic medication and health care costs among Medicaid beneficiaries with schizophrenia. Am J Psychiatry. 2004;161:692-699.