Schizophrenia is a devastating psychiatric disorder with a global median lifetime prevalence of 4 per 1000 and a lifetime morbidity risk of 7.2 per 1000.1 The treatment of schizophrenia has been focused on eliminating the symptoms of the disease and has included antipsychotic medications since the 1950s.2,3 Despite the development of new therapies for managing schizophrenia, including second-generation (or atypical) antipsychotics, approximately 66% of individuals with this serious mental illness have persisting or fluctuating symptoms even with optimal treatment.4,5 Throughout their lifetimes, most patients will experience chronic disease with repeated relapses, characterized by exacerbation of psychosis resulting in emergency department visits and costly rehospitalizations.6-8
Poor adherence to antipsychotics, particularly to oral formulations, has been associated with increased risks for disease relapse, hospitalizations, and suicide,9 and has been shown in various studies to have significant direct and indirect costs.10,11 With an annual excess direct healthcare cost in the United States estimated at $22.7 billion annually in 2002, nonadherence to medication, which affects approximately 50% of patients with schizophrenia, is a critical clinical issue and a major concern for payers.12-16
The Healthcare Effectiveness Data and Information Set (HEDIS) measures, which were developed by the National Committee for Quality Assurance (NCQA), are a key tool in evaluating the quality of patient care and are used by more than 90% of healthcare plans and by many leading employers and regulators in the United States.17 In 2013, a HEDIS quality measure for Medicaid plans was introduced that represented the proportion of a plan’s patients with schizophrenia who met a defined threshold of treatment continuity or adherence. On a patient level, treatment continuity was measured as the proportion of days covered (PDC) by antipsychotic medication, calculated from dispensed prescription claims. HEDIS considered a patient to have achieved continuity of treatment if the PDC was ≥80%.18 Treatment continuity (ie, adherence) is a major challenge in the long-term management of patients with schizophrenia, because poor patient adherence to antipsychotics has been associated with negative clinical outcomes.19,20
Understanding the determinants of adherence to antipsychotic medication is critical in the management of patients with schizophrenia. Patient-related factors related to adherence include demographic characteristics (eg, age, socioeconomic status, alcohol or drug use, ethnicity) and general cognitive functions (eg, IQ score, memory, learning abilities).21,22 The severity of the disease and the perception about illness (eg, the presence of positive or negative perceptions) may also influence patient adherence to antipsychotic medications.21,22 Medication characteristics, such as side effects and the frequency of administration, can play an important role in adherence to treatment, as well as the complexity of the healthcare system that may limit access to medications, such as a lack of continuity of care. Support from providers, family, and caregivers can tremendously improve the attitude of patients toward their illness and their need for medication, can guide them through the healthcare system, and can provide interventions quickly when patients stop taking their medications.22
With less frequent administration than oral formulations, long-acting injectable antipsychotic medications, including paliperidone palmitate, and injectable versions of risperidone and haloperidol, have been associated with lower rates of nonadherence compared with oral equivalents or with patients receiving any oral antipsychotics (eg, oral risperidone, quetiapine, or oral aripiprazole).23,24 Other potential benefits of long-acting injectable antipsychotics include minimizing disease relapses and lowering hospital admissions, findings that were documented in recent meta-analyses of nonrandomized observational studies.24-28
Approved by the US Food and Drug Administration (FDA) in 2009, paliperidone palmitate is the first once-monthly long-acting injectable antipsychotic for the acute and maintenance treatment of schizophrenia in adults.29
The current study uses important variables such as patient demographic and clinical baseline characteristics, including previous-year adherence and the use of paliperidone palmitate, on the likelihood of achieving the HEDIS antipsychotic continuity measure during the measurement year among patients with schizophrenia using real-world claims data from 5 states offering Medicaid coverage.
Methods
To control for confounding factors associated with the choice of antipsychotic treatment, inverse probability of treatment weights was used.
The study population was identified using medical and pharmacy claims from Medicaid databases for Florida, New Jersey, Iowa, Missouri, and Kansas for the years 2008 to 2011. Medicaid databases contain medical claims (eg, type of service, service unit, date, International Classification of Diseases, Ninth Revision [ICD-9] diagnoses, Current Procedural Terminology codes, physician specialty, and type of provider), prescription drug claims (eg, supply days, units, date of service, and National Drug Codes), and eligibility information (eg, age, sex, enrollment start and end dates, and date and year of death, if applicable).
Study Design
To approximate the HEDIS metric, this study used a retrospective longitudinal cohort design with calendar year baseline and measurement periods. The baseline (or reference) year, during which patient characteristics were observed to establish a reference, was 2010; the measurement (or index) year, during which the study end points were evaluated, was 2011.
For Florida only, the definitions of the baseline and measurement years were shifted by 1 quarter, because data were not available for the fourth quarter of 2011 at the time this study was conducted. Figure 1 depicts the design scheme for this study.
The study population was identified according to the specifications set by the NCQA30,31 and consisted of Medicaid patients (1) with ≥1 acute inpatient or emergency department visits or ≥2 outpatient or nonacute inpatient visits with schizophrenia as the principal diagnosis (ICD-9 code, 295.xx) during the measurement year 2011; (2) aged 25 to 64 years; (3) with continuous Medicaid enrollment for years 2010 and 2011; (4) with ≥1 dispensed antipsychotic medications (ie, typical or atypical antipsychotic with a general product identifier [GPI] code beginning with 59, and Healthcare Common Procedure Coding System [HCPCS] codes C9125, J1630, J1631, J2794, or S0163) within the first 9 months of the measurement year; and receiving an antipsychotic during the baseline year. Of note, only adults aged >25 years were analyzed in accordance with the HEDIS 2013 measures that were available at the time of the study. The patients were classified into 2 mutually exclusive cohorts, including patients with any paliperidone palmitate claim (GPI code beginning with 590700501018, and HCPCS codes C9255 or J2426) and any other antipsychotic medications usersR12;patients with a claim for an antipsychotic, including quetiapine, risperidone, and haloperidol, but no claim for paliperidone palmitate (see Appendix, Table 1, at www.AHDBonline.com for a complete list of medications). Of note, aripiprazole injectable is a once-monthly injectable antipsychotic agent approved by the FDA, but it was not yet available during this study.
Performance on the HEDIS continuity of antipsychotic medications measure was evaluated for the measurement year using the methodology developed by the NCQA, which was defined as achieving a PDC of ≥80% during the measurement year. PDC was defined as the number of nonoverlapping days of supply divided by the number of days in the measurement year (365 days).
For long-acting injectable antipsychotics reimbursed through a medical claim with no days of supply field, the days of supply was imputed from the drug label (eg, 28 days of supply were imputed for paliperidone palmitate medical claims). The baseline year and the period up to 2 years before the baseline year were used to identify potential predictive factors from those observable on medical and prescription claims.
Study End Points and Statistical Analysis
The potential predictive factors, which were evaluated during the baseline year, included age, sex, race, state, Quan Charlson Comorbidity Index, healthcare costs, type of antipsychotic treatment prescribed (eg, paliperidone palmitate or other antipsychotic medications, including quetiapine, risperidone, and haloperidol) and adherence status (PDC ≥80%). The healthcare costs and type of antipsychotic treatment received up to 2 years before baseline were also included as predictive factors. The medical and pharmacy costs were evaluated on an annual basis; the medical costs were stratified by place of service (ie, inpatient, outpatient, emergency department, home visit, long-term care, mental institute, and other), and all costs were inflated to 2011 US dollars. P values testing the difference between the paliperidone palmitate cohort and the other antipsychotics cohort were calculated using chi-square tests for categorical variables and student’s t-tests for continuous variables.
As with any real-world study, there is a risk that patients who are prescribed different medications may vary in terms of disease severity or other characteristics. To minimize the potential confounding factors between patients receiving paliperidone palmitate and those receiving other antipsychotics, and in an attempt to keep all patients in the analysis (therefore maximizing the representativeness and generalizability of the results), we used inverse probability of treatment weights based on propensity score.32-34 The propensity score, which summarizes covariate information about treatment selection into a scalar value, was calculated as the probability of being in the paliperidone palmitate cohort using a multivariate logistic regression based on age, sex, race, state, and past medical healthcare costs. These baseline variables were included as potential confounding factors between the paliperidone palmitate group and the other antipsychotics treatment group.
The weights were calculated as 1/propensity score for the paliperidone palmitate cohort and 1/(1–propensity score) for the other antipsychotics cohort and was normalized using the average weight. For example, a patient receiving paliperidone palmitate with a low predicted probability of receiving this drug (based on the propensity score) received a “high” weight, whereas a paliperidone palmitate user with a high predicted probability received a “low” weight. Weights were truncated at 1% and 99% of the distribution to limit the effect of extreme weights.
Weighted multivariate logistic regression models and odds ratios (ORs) were used to identify potential factors associated with adherence to the HEDIS continuity of antipsychotic measure, including age, sex, race, state, comorbidity index, baseline adherence, type of antipsychotics received, and past healthcare costs. The covariates considered for the logistic regression models evaluating the likelihood of achieving the HEDIS continuity of antipsychotic measure were selected from data elements available on claims from the baseline year. P values testing the difference between patients receiving paliperidone palmitate and those receiving any other antipsychotic medication and 95% confidence intervals (CIs) were calculated.
Furthermore, 2 additional multivariate logistic regression models were conducted to assess the robustness of the findings, including a weighted multivariate logistic regression with the model not truncated, and a multivariate logistic regression without inverse probability of treatment weights adjustment.
All statistical analyses were conducted using SAS version 9.3 (SAS Institute, Inc; Cary, NC).
Results
A total of 492 patients were selected for the paliperidone palmitate cohort and 12,498 were selected for the other antipsychotics (eg, quetiapine, risperidone, and haloperidol). Figure 2 outlines the sample selection process. Table 1 presents the baseline-year demographics and the clinical characteristics of the study population, as well as the stratification of the 2 treatment cohorts.
The mean age of the study population was 46.4 years, and 46.1% were women. The patients who received paliperidone palmitate were significantly different from the patients who received other antipsychotics in terms of age, race, state, and Quan Charlson Comorbidity Index (P <.001 for all). There were differences in some annual healthcare cost categories, such as pharmacy costs ($9662 vs $6738, respectively; P <.001); inpatient costs ($9061 vs $4289, respectively; P <.001); outpatient costs ($2551 vs $6463, respectively; P <.001); and mental institution costs ($5522 vs $3179, respectively; P <.001).
Of the 12,990 patients who received antipsychotics, 48.6% successfully achieved the criteria for antipsychotic drug continuity in the measurement year (Table 2). Among the adherent patients during the baseline year, 76.2% met the continuity criteria in the measurement year, as opposed to 27.3% of the nonadherent patients during the baseline year (P <.001).
The success rates on the continuity measure were similar between the paliperidone palmitate (46.3%) cohort and the other antipsychotics cohort (48.7%; P = .302).
The Figure in the Appendix (see www.AHDBonline.com) shows the distribution of the propensity score, and Table 2 in the Appendix shows the truncated and not truncated inverse probability of treatment weights. The propensity scores ranged from >0 to 0.4317 in the paliperidone palmitate cohort and from >0 to 0.4284 in the oral antipsychotics cohort; the truncated weights ranged from 1.15 to 12.97 in the paliperidone palmitate cohort and from 0.50 to 0.87 in the oral antipsychotics cohort.
After controlling for other covariates, the odds of success for the HEDIS continuity measure were increased with increased adherence during baseline (OR, 9.42; 95% CI, 8.55-10.39), female sex (OR, 1.11; 95% CI, 1.01-1.22), age 55 to 64 years (OR, 1.26; 95% CI, 1.09-1.46) relative to age 25 to 34 years old, and Hispanic ethnicity (OR, 1.37; 95% CI, 1.05-1.81) relative to white ethnicity (Table 3).
An increase in baseline inpatient admission cost was also associated with greater odds of success (increase of $10,000; OR, 1.11; 95% CI, 1.08-1.15). Accounting for the baseline differences between the 2 treatment cohorts using inverse probability of treatment weights, receiving paliperidone palmitate during the baseline year was associated with a 26% increase in the odds of achieving medication continuity compared with receiving other antipsychotics (OR, 1.26; 95% CI, 1.14-1.39). Similar trends were found when controlling with inverse probability of treatment weights but without truncating weights (sensitivity 1, OR, 1.36; 95% CI, 1.25-1.49), and when not controlling with the inverse probability of treatment weights (sensitivity 2, OR, 1.21; 95% CI, 0.98-1.49).
Discussion
This study is based on real-world claims data of patients with Medicaid coverage and using antipsychotic medication for schizophrenia. Approximately 50% of the study population successfully achieved the criteria for antipsychotic drug continuity. Among the characteristics observed in the administrative claims, the factors that were significantly associated with adherence to the HEDIS measure were identified. Older age, female sex, Hispanic ethnicity, the previous-year adherence, higher inpatient costs, and the use of paliperidone palmitate therapy were predictive factors for attaining adherence to the HEDIS measure in the measurement year.
Several studies have evaluated the consequences of nonadherence to medication among patients with schizophrenia.9,35-37 The negative impact of nonadherence on the course of illness includes increased relapses, an increase in rehospitalizations, and an increased risk for suicide attempts.36,38 The cost of nonadherence was estimated to range from $1.4 million to $1.8 million in the United States for 2005.39 Therefore, full adherence to medication is a key factor among patients with schizophrenia.
Recent meta-analyses suggest mixed results regarding the effectiveness of long-acting injectable antipsychotics compared with oral antipsychotics when analyzing randomized controlled trials,23,24 but long-acting injectable antipsychotics were superior to oral antipsychotics in observational retrospective studies.24-28 Kirson and colleagues suggest that although randomization is considered the gold standard to eliminating selection bias in estimating treatment effects, the randomized clinical trial setting does not reflect other aspects of the way therapies are used in general clinical practice.27
For example, clinical trials may have more frequent protocol-dictated medical visits and drug use controls than in the real-world setting, which may influence patients’ adherence to their treatment (ie, the Hawthorne effect).40 Real-world studies using methods to control for confounding bias, such as inverse probability of treatment weights, are warranted to assess the association between treatment and adherence in the general clinical practice, although they can only control for observable parameters.
The results of our current study add to the previous literature based on real-world data that taking paliperidone palmitate is associated with a 26% increase in the odds of achieving adherence to an antipsychotic medication (ie, the HEDIS continuity measure; OR, 1.26; 95% CI, 1.14-1.39) compared with taking other antipsychotic drugs. However, patients receiving paliperidone palmitate had higher inpatient and mental institution costs in the baseline year, which may be indicative of a different level of care in the measurement year, and this, in turn, could have affected their adherence.
Another important finding of our study is the role of baseline adherence on future adherence in a population of Medicaid-insured patients with schizophrenia. Previous studies already established this relationship in populations with and without Medicaid insurance.41-43 In a Medicaid-insured population, nonadherence to therapy was found to significantly increase the likelihood of future nonadherence by 12-fold.43 This is in line with the results of our study, in which adherence to therapy significantly increased the odds of future adherence by 9-fold, emphasizing the importance of factors other than medication type in determining adherence. This suggests that patient characteristics and other dimensions influencing medication adherence can outweigh the impact of the type of medication used to control the symptoms of psychoses.
We found little difference in the crude success rates observed for the paliperidone palmitate and the other antipsychotics cohort. Table 2 demonstrates how a single patient characteristicR12;the previous-year adherenceR12;can affect adherence, highlighting the need to control for confounding factors among treatment cohorts.
Given the variability in patient characteristics, the inverse probability of treatment weights was widely distributed. In the main analysis, the lowest and highest weights were truncated to first or 99% of the weight distribution. In sensitivity analysis 1, incorporating untruncated weights, the effects of baseline adherence and the use of paliperidone palmitate were even greater. Sensitivity analysis 2 compared the propensity score–matched cohorts without weighting to adjust for potential treatment confounders and found similar results, although the effect of the use of paliperidone palmitate was no longer significant (P = .0772).
To our knowledge, the current study is the first evaluation of the predictive factors of adherence (as defined by HEDIS) in patients identified as the target population according to HEDIS. Therefore, the results of this study provide meaningful insight for Medicaid plans in their efforts to improve the quality of care for patients with schizophrenia.
Limitations
The current study was subject to certain limitations. The Medicaid data used in this study came from only 5 states and may not be representative of the entire US population, of other states, or of nonR11;Medicaid-covered patients.
Furthermore, the data, coming from real-world claims, were subject to billing inaccuracies and missing data as all data are.
The population sample analyzed for the HEDIS measure was identified based on the NCQA guidelines, but it may not be fully representative of the general Medicaid-covered patient population, because the presence of healthcare claims limits the analysis to patients who are already engaged in the healthcare system.
Claims-based adherence measures (eg, the PDC) do not account for whether the drugs dispensed were taken as prescribed. This may overestimate patient adherence, especially for patients taking an oral antipsychotic medication, for whom we assumed that they took their medication correctly (eg, 1 pill daily), whereas for paliperidone palmitate, the duration of effect for 1 injection is independent of any further action by the patient.
The group of patients who received other antipsychotic drugs encompassed a wide variety of medications, including oral and long-acting injectable typical and atypical antipsychotics, with potentially different adherence profiles. The mix of other antipsychotic drugs could alter the effect that is specifically attributable to paliperidone palmitate.
As with any observational studies, the study groups may differ in terms of disease severity or other characteristics at baseline that are known to be prognostically important.
Finally, we used propensity scoring and inverse probability of treatment weights to control for population differences based on the information that was available in the database. Similarly, the potential predictors of the HEDIS continuity of antipsychotics measure were limited to the factors observable on the Medicaid claims. Important factors, such as socioeconomic status, alcohol or drug use, disease severity (eg, the Positive and Negative Syndrome Scale), attitude toward illness and medication, and provider or family support, may also be important determinants of treatment choice or HEDIS performance but are not observable through administrative claims databases. Nevertheless, health insurance claims data remain a valuable tool for population health decision makers because of the relevant information on patient characteristics and outcomes and the typically large volume of data from a real-world setting.
Conclusion
Of the information available in the database, the baseline factors that were associated with better performance on the continuity of antipsychotic medications measure included older age, female sex, Hispanic ethnicity, previous-year adherence, higher inpatient costs, and the use of paliperidone palmitate therapy. These findings offer insight into the effects that patient characteristics and treatment types have on adherence among patients with schizophrenia. Although more research is needed, this information may help healthcare plans that cover Medicaid populations to better address the quality of care provided to Medicaid patients with schizophrenia.
Source of Funding
This study was funded by Janssen Scientific Affairs, LLC, Titusville, NJ.
Author Disclosure Statement
Dr Clancy is an employee of Celgene Corporation. Dr Fastenau, Mr Durkin, and Dr Pesa own stocks of Johnson & Johnson. Ms Lafeuille, Dr Frois, Dr Cloutier, Ms Duh, and Mr Lefebvre are employees of Analysis Group, which has received research support grant from Janssen Scientific Affairs.
References
1. McGrath J, Saha S, Chant D, Welham J. Schizophrenia: a concise overview of incidence, prevalence, and mortality. Epidemiol Rev. 2008;30:67-76.
2. National Institute of Mental Health. Schizophrenia. Revised February 2016. www.nimh.nih.gov/health/topics/schizophrenia/index.shtml. Accessed March 25, 2016.
3. Brissos S, Veguilla MR, Taylor D, Balanzá-Martinez V. The role of long-acting injectable antipsychotics in schizophrenia: a critical appraisal. Ther Adv Psychopharmacol. 2014;4:198-219.
4. Saha S, Chant D, Welham J, McGrath J. A systematic review of the prevalence of schizophrenia. PLoS Med. 2005;2:e141.
5. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders: DSM-IV, 4th ed. Washington, DC: American Psychiatric Association; 1994:886.
6. Andreasen NC. Symptoms, signs, and diagnosis of schizophrenia. Lancet. 1995;346:477-481.
7. de Sena EP, Santos-Jesus R, Miranda-Scippa Â, et al. Relapse in patients with schizophrenia: a comparison between risperidone and haloperidol. Rev Bras Psiquiatr. 2003;25:220-223.
8. Chue P. Long-acting risperidone injection: efficacy, safety, and cost-effectiveness of the first long-acting atypical antipsychotic. Neuropsychiatr Dis Treat. 2007;3:13-39.
9. Higashi K, Medic G, Littlewood KJ, et al. Medication adherence in schizophrenia: factors influencing adherence and consequences of nonadherence, a systematic literature review. Ther Adv Psychopharmacol. 2013;3:200-218.
10. Dilla T, Ciudad A, Alvarez M. Systematic review of the economic aspects of nonadherence to antipsychotic medication in patients with schizophrenia. Patient Prefer Adherence. 2013;7:275-284.
11. Lafeuille MH, Dean J, Fastenau J, et al. Burden of schizophrenia on selected comorbidity costs. Expert Rev Pharmacoecon Outcomes Res. 2014;14:259-267.
12. Desai PR, Lawson KA, Barner JC, Rascati KL. Estimating the direct and indirect costs for community-dwelling patients with schizophrenia. J Pharm Health Serv Res. 2013;4:187-194.
13. Messias EL, Chen CY, Eaton WW. Epidemiology of schizophrenia: review of findings and myths. Psychiatr Clin North Am. 2007;30:323-338.
14. 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.
15. Regier DA, Narrow WE, Rae DS, et al. The de facto US mental and addictive disorders service system: epidemiologic catchment area prospective 1-year prevalence rates of disorders and services. Arch Gen Psychiatry. 1993;50:85-94.
16. 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.
17. National Committee for Quality Assurance. HEDIS and quality measurement: what is HEDIS? www.ncqa.org/hedis-quality-measurement/what-is-hedis. 2014. Accessed September 12, 2016.
18. National Committee for Quality Assurance. Proposed measures for HEDIS 2013: schizophrenia. 2012. https://openminds.com/wp-content/uploads/indres/030112mhcdhedis2013antipsychoticmeds.pdf. Accessed October 29, 2012.
19. Morken G, Widen JH, Grawe RW. Non-adherence to antipsychotic medication, relapse and rehospitalisation in recent-onset schizophrenia. BMC Psychiatry. 2008;8:32.
20. Sacchetti E, Vita A, Siracusano A, Fleischhacker W. Adherence to Antipsychotics in Schizophrenia. Milan, Italy: Springer-Verlag Italia; 2014.
21. El-Mallakh P, Findlay J. Strategies to improve medication adherence in patients with schizophrenia: the role of support services. Neuropsychiatr Dis Treat. 2015;11:1077-1090.
22. Correll CU. Recognition of patients who would benefit from LAI antipsychotic treatment: how to assess adherence. J Clin Psychiatry. 2014;75:e29.
23. Zhornitsky S, Stip E. Oral versus long-acting injectable antipsychotics in the treatment of schizophrenia and special populations at risk for treatment nonadherence: a systematic review. Schizophr Res Treatment. 2012;2012:407171.
24. Marcus SC, Zummo J, Pettit AR, et al. Antipsychotic adherence and rehospitalization in schizophrenia patients receiving oral versus long-acting injectable antipsychotics following hospital discharge. J Manag Care Spec Pharm. 2015;21:754-768.
25. Kane JM, Kishimoto T, Correll CU. Assessing the comparative effectiveness of long-acting injectable vs. oral antipsychotic medications in the prevention of relapse provides a case study in comparative effectiveness research in psychiatry. J Clin Epidemiol. 2013;66(8 suppl):S37-S41.
26. Kishimoto T, Robenzadeh A, Leucht C, et al. Long-acting injectable vs oral antipsychotics for relapse prevention in schizophrenia: a meta-analysis of randomized trials. Schizophr Bull. 2014;40:192-213.
27. Kirson NY, Weiden PJ, Yermakov S, et al. Efficacy and effectiveness of depot versus oral antipsychotics in schizophrenia: synthesizing results across different research designs. J Clin Psychiatry. 2013;74:568-575.
28. Lafeuille MH, Dean J, Carter V, et al. Systematic review of long-acting injectables versus oral atypical antipsychotics on hospitalization in schizophrenia. Curr Med Res Opin. 2014;30:1643-1655.
29. Johnson & Johnson. FDA approves Invega Sustenna for the acute and maintenance treatment of schizophrenia. Press release; July 31, 2009. www.investor.jnj.com/releasedetail.cfm?releaseid=400443. Accessed September 13, 2016.
30. National Committee for Quality Assurance. HEDIS 2014: summary table of measures, product lines, and changes. September 30, 2013. www.ncqa.org/Portals/0/HEDISQM/HEDIS2014/List_of_HEDIS_2014_Measures.pdf. Accessed September 13, 2016.
31. National Committee for Quality Assurance. HEDIS 2014: specification updates. www.ncqa.org/Portals/0/HEDISQM/HEDIS2014/HEDIS_2014_Volume_
2_Technical_Update_FINAL_9.30.13.pdf. Accessed September 13, 2016.
32. Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15:615-625.
33. Curtis LH, Hammill BG, Eisenstein EL, et al. Using inverse probability-weighted estimators in comparative effectiveness analyses with observational databases. Med Care. 2007;45(10 suppl 2):S103-S107.
34. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34:3661-3679.
35. Goff DC, Hill M, Freudenreich O. Strategies for improving treatment adherence in schizophrenia and schizoaffective disorder. J Clin Psychiatry. 2010;71(suppl 2):20-26.
36. Olivares JM, Alptekin K, Azorin JM, et al. Psychiatrists’ awareness of adherence to antipsychotic medication in patients with schizophrenia: results from a survey conducted across Europe, the Middle East, and Africa. Patient Prefer Adherence. 2013;7:121-132.
37. Markowitz M, Karve S, Panish J, et al. Antipsychotic adherence patterns and health care utilization and costs among patients discharged after a schizophrenia-related hospitalization. BMC Psychiatry. 2013;13:246.
38. Leucht S, Heres S. Epidemiology, clinical consequences, and psychosocial treatment of nonadherence in schizophrenia. J Clin Psychiatry. 2006;67(suppl 5):3-8.
39. Sun SX, Liu GG, Christensen DB, Fu AZ. Review and analysis of hospitalization costs associated with antipsychotic nonadherence in the treatment of schizophrenia in the United States. Curr Med Res Opin. 2007;23:2305-2312.
40. McCarney R, Warner J, Iliffe S, et al. The Hawthorne Effect: a randomised, controlled trial. BMC Med Res Methodol. 2007;7:30.
41. Novick D, Haro JM, Suarez D, et al. Predictors and clinical consequences of non-adherence with antipsychotic medication in the outpatient treatment of schizophrenia. Psychiatry Res. 2010;176:109-113.
42. Ascher-Svanum H, Furiak NM, Lawson AH, et al. Cost-effectiveness of several atypical antipsychotics in orally disintegrating tablets compared with standard oral tablets in the treatment of schizophrenia in the United States. J Med Econ. 2012;15:531-547.
43. Lang K, Federico V, Muser E, et al. Rates and predictors of antipsychotic non-adherence and hospitalization in Medicaid and commercially-insured patients with schizophrenia. J Med Econ. 2013;16:997-1006.
Appendix
