Daisy Ng-Mak, PhD

This study compared the risk of hospitalization among adults with schizophrenia being treated with equivalent dose ranges of lurasidone versus aripiprazole, olanzapine, quetiapine, or risperidone. Administrative claims data for this analysis came from the IBM MarketScan Commercial, Medicare Supplemental, and Multi-State Medicaid databases between January 2011 and June 2017. The study included adults with schizophrenia who initiated treatment with an antipsychotic and were continuously enrolled for 360 days prior to and following the date of the initial antipsychotic prescription. Risk of all-cause and schizophrenia-related hospitalization among patients who received lurasidone monotherapy versus aripiprazole, olanzapine, quetiapine, or risperidone in equivalent dose ranges were assessed. Marginal structural models that accounted for preindex characteristics, changes in antipsychotic treatment, and time-varying covariates assessed differences between lurasidone and other second-generation antipsychotics on all-cause and schizophrenia-related hospitalizations. A sensitivity analysis was conducted without the dose-equivalence requirement. A total of 20,212 patients met the study inclusion criteria. Compared with those treated with lurasidone monotherapy, the adjusted risk of all-cause hospitalization was significantly higher among patients treated with olanzapine (adjusted rate ratio [aRR], 1.49; 

 = .04), but not aripiprazole (aRR, 1.24; 

 = .28). A similar, non-statistically significant pattern of adjusted risks of schizophrenia-related hospitalizations was observed. A sensitivity analysis without the dose-equivalence requirement produced consistent results. As hospitalization is a major cost driver of direct healthcare cost, lurasidone may be a cost-saving treatment option for patients with schizophrenia.

Schizophrenia is a debilitating, chronic mental illness that affects approximately 0.3% to 1.0% of the US population.

 Symptoms can be divided into “positive,” such as delusions and hallucinations, and “negative,” such as lack of emotion and social withdrawal.

 Because of the severity of the symptoms, most individuals with schizophrenia are unable to engage in paid employment, maintain marriage/partner relationships, or live independently.

In the United States, the economic burden of schizophrenia in 2013 was estimated at $155.7 billion, including $37.7 billion in direct healthcare costs.

 Inpatient costs were the largest driver of direct healthcare costs, accounting for $15.2 billion (40.3%) of the direct costs.

 Different types of healthcare expenditures have different clinical implications in schizophrenia.

 Inpatient and emergency department (ED) use may reflect symptomatic relapses, while higher rates of outpatient visits and greater use of antipsychotic medications may reflect greater engagement in treatment.

 Increased adherence with antipsychotics has been linked with reduced inpatient costs

Antipsychotics are the primary treatment for acute schizophrenia

 and are recommended in the stable phase because they substantially reduce the risk of relapse.

 Second-generation antipsychotics (SGAs) are preferred over first-generation antipsychotics (FGAs) for acute schizophrenia because of a reduced risk of developing extrapyramidal symptoms.

 Meta-analyses summarizing the results of randomized controlled trials of patients who are taking antipsychotics have reported substantial variations in extrapyramidal symptoms, prolactin elevation, sedation, and weight gain.

 The metabolic burden associated with commonly used SGAs has been classified as substantial for olanzapine and clozapine, intermediate for quetiapine and risperidone, and neutral to low for aripiprazole, lurasidone, and ziprasidone.

 A patient’s past experience with antipsychotics, particularly tolerability issues, should be considered before prescribing a particular treatment.

 Study results have shown that patients prefer efficacious antipsychotics with less risk of weight gain and hyperglycemia.

Comparative Effectiveness Study Challenges

Patient preferences, adherence, real-world dosing, and other factors can affect the efficacy of treatment in usual clinical care and potentially limit the generalizability of clinical trial results to real-world effectiveness.

When designing comparative effectiveness studies using real-world data, the Agency for Healthcare Research and Quality (AHRQ) recommends, “When appropriate and possible, comparisons should be made for exposure and comparison groups at various clinically equivalent dose levels.”

 Studying medication effects across different dosage strengths is an important consideration when designing head-to-head clinical trials

; however, it is often ignored in studies using real-world data where treatment effects are studied without considering the consequences of dosage strength. Comparisons between medications used at different doses could make a lower-dosed medication appear more tolerable and a higher-dosed medication appear more efficacious. In addition, comparisons among different dose levels could result in confounding by severity, as higher doses are more likely to be given to patients with greater disease severity.

An additional challenge for comparative effectiveness research in schizophrenia is the high frequency of antipsychotic treatment switching. As such, an intent-to-treat (ITT) analysis of antipsychotic treatment in real-world clinical practice is often inappropriate. Many comparative effectiveness studies using administrative claims have used an ITT approach,

 in which patients are classified according to their first prescribed antipsychotic, and outcomes are observed regardless of whether patients discontinue or switch to a different antipsychotic treatment. However, in clinical trials using ITT, adherence to the treatment protocol is closely monitored and data are no longer collected from a patient if treatment is discontinued.

To address these challenges in real-world studies, some investigators have used treatment episodes, rather than patients, as the unit of analysis.

 Recent studies have gone a step further and used marginal structural models (MSMs),

 which account for treatment switching, incorporate time-varying covariates, and appear to more accurately estimate causal treatment effects.

 Marginal structural modeling is a multistep analytic procedure which uses weighted repeated measures and takes into consideration time-varying treatment effects. To accomplish this, weights are calculated and assigned to each observation to balance background characteristics across treatment groups. Consistent estimates of treatment effects can be generated.

The objective of this study was to compare the risk of hospitalization among patients with schizophrenia treated with equivalent doses of lurasidone versus aripiprazole, olanzapine, quetiapine, or risperidone using an MSM that accounted for changes in treatment and covariates over time.

This retrospective database study used administrative claims data from the IBM MarketScan Commercial Claims and Encounters Database, MarketScan Medicare Supplemental and Coordination of Benefits Database, and Multi-State Medicaid Database.

 The databases contained adjudicated inpatient, outpatient, and pharmacy claims data that were deidentified per the Health Insurance Portability and Accountability Act of 1996.

 For this study, all the administrative claims were dated between January 1, 2011, and June 30, 2017.

All patients were required to have at least 1 diagnosis of schizophrenia (

International Classification of Disease, 9th Revision, Clinical Modification

: F20.x; for both, x can be any value). Patients were also required to initiate an oral SGA (lurasidone, aripiprazole, asenapine, brexpiprazole, cariprazine, clozapine, iloperidone, paliperidone, olanzapine, quetiapine, risperidone, or ziprasidone) or an oral FGA (chlorpromazine, fluphenazine, haloperidol, loxapine, prochlorperazine, perphenazine, pimozide, thioridazine, thiothixene, or trifluoperazine) following their first diagnosis for schizophrenia. Patients were required to be at least 18 years of age on the index date, defined as the date of the first antipsychotic prescription. Continuous enrollment in a health plan with medical and pharmacy benefits was required for 360 days before (preindex period) and after (postindex period) the index date. Finally, patients could not have been treated with any antipsychotics during the preindex period.

This study compared hospitalization rates among patients treated with monotherapy lurasidone versus aripiprazole, olanzapine, quetiapine, or risperidone. These 5 SGAs were the focus because they were the most commonly used, accounting for more than 75% of all antipsychotic monotherapy use in our dataset. The daily dose associated with each antipsychotic claim was estimated by multiplying the dose strength (in milligrams) with the quantity of tablets or capsules supplied, then dividing by the estimated number of days the quantity should last (days supplied). In the primary analysis, dosing was restricted to a maximum daily dose equivalent to 80 mg of lurasidone.

 The dose of each medication equivalent to 80 mg of lurasidone was determined using published estimates of doses equivalent to 100 mg of chlorpromazine.

 The rationale for using lurasidone 80 mg per day as a threshold of dose equivalence was based on its inclusion in the initially approved dose range for schizophrenia and as the most common dose used in real-world settings. Lurasidone was initially approved in 2010 with a recommended dose range of 40 mg to 80 mg in adult patients with schizophrenia; this dose was updated to 40 mg to 160 mg in 2013.

 In clinical practice, the majority of adult patients with schizophrenia are treated with 40 mg to 80 mg of lurasidone.

The median daily dose is reported for each antipsychotic with and without this dose-equivalence requirement in 

The 360-day postindex period was divided into twelve 30-day intervals (“months”), such that each patient contributed 12 patient-months to our dataset. Monotherapy was defined as treatment for at least 75% of days (≥22 days) with a single antipsychotic during a 30-day interval. The comparison groups included monotherapy lurasidone, aripiprazole, olanzapine, quetiapine, or risperidone, at a dose equivalent to 80 mg per day or less of lurasidone. Patient-months that did not meet these criteria were classified as “other.” The “other” group was included because the statistical model (described below) required all patient-months to be classified and could include combination therapy, monotherapy with FGAs or SGAs other than those listed above, monotherapy with equivalent doses above 80 mg per day of lurasidone, and no antipsychotic treatment.

A sensitivity analysis that did not involve dose-equivalence comparison was conducted. In this analysis, lurasidone, aripiprazole, olanzapine, quetiapine, and risperidone monotherapy patient-months were classified as such, regardless of the daily dose. Patient-months classified as “other” in this analysis included the same treatments as in the primary analysis, with the exception of monotherapy doses above 80 mg per day of lurasidone.

All-cause hospitalizations were identified using any hospital claim, regardless of associated diagnostic codes. Schizophrenia-related hospitalizations were those in which the hospital claim contained a diagnosis of schizophrenia (

: F20.x) in any position. Rates of all-cause and schizophrenia-related hospitalization associated with lurasidone monotherapy patient-months were compared with rates of all-cause and schizophrenia-related hospitalization associated with aripiprazole, olanzapine, quetiapine, and risperidone monotherapy patient-months.

Patient preindex characteristics were calculated across the 360-day preindex period. These included age, gender, insurance type (Medicaid or commercial/Medicare Supplemental), and Charlson Comorbidity Index (CCI) score

; the proportions of patients who had at least 1 claim with a diagnosis of substance/alcohol/nicotine use, anxiety disorder, depression, bipolar disorder, diabetes, hypertension, or obesity (

); the proportions of patients with at least 1 prescription for an antidepressant, anxiolytic, or mood stabilizer (lithium, divalproex, lamotrigine, carbamazepine, oxcarbazepine, and topiramate); and numbers of all-cause hospitalizations, schizophrenia-related hospitalizations, and all-cause ED visits.

MSMs were used to examine the effects of antipsychotic treatment on hospitalization. MSMs can estimate causal treatment effects when treatment and other covariates change over the course of follow-up.

 This is accomplished by weighting the data with the probability of receiving each of the various treatments in each time period (using stabilized inverse probability weights). Time-dependent confounding variables can be controlled when outcomes of interest are modeled.

Stabilized inverse probability weights were calculated for each patient-month in the postindex period.

A multinomial logistic regression model was used to predict assignment to each treatment category using preindex (time-independent) as well as postindex (time-dependent) covariates. The time-independent covariates consisted of an indicator of whether the patient was above or below the mean age at index; gender; preindex CCI score; and diagnosis of substance/alcohol/nicotine use, anxiety disorder, bipolar disorder, depression, diabetes, hypertension, and obesity. The time-dependent covariates included treatment in the prior month, any hospitalization in the prior month, and any ED visit during the prior month.

Because there was no preindex antipsychotic treatment, stabilized inverse probability weights could not be calculated for month 1. Therefore, outcomes were modeled over month 2 through month 12, although treatment and resource utilization in month 1 were used to predict treatment category and outcomes in subsequent months. Counts of inpatient admissions during month 2 through month 12 were modeled using generalized estimating equations (GEEs) with an exchangeable correlation matrix, negative binomial distribution, and log link function. Covariates in these GEE models included the treatment category for the current patient-month and all the previously mentioned time-independent variables. All GEE models were weighted using the stabilized inverse probability weights. All statistical analyses were completed using SAS version 9.4 (SAS Institute; Cary, NC) with a prespecified 2-tailed alpha of 

A total of 20,212 patients with schizophrenia who initiated antipsychotic therapy met all inclusion criteria (

). Of these, 10,644 patients received initial monotherapy with lurasidone (2.7%), aripiprazole (8.2%), olanzapine (9.4%), quetiapine (12.7%), or risperidone (19.7%) at a dose equivalent to 80 mg or less per day of lurasidone. Over the postindex period, daily doses were low and mostly below the dose-equivalence threshold. Daily doses were similar before and after dose-equivalence restriction (

 The average age of the patients in the sample was 41.3 years, and 50.8% were female. Nearly half of the patients had a history of substance, alcohol, or nicotine use (46.9%) or depression (42.5%). Comorbidities such as diabetes (16.6%), hypertension (36.7%), and obesity (13.8%) were common. Most patients had Medicaid insurance (67.0%). Relative to patients initiating the other SGAs, patients initiating lurasidone were younger, more likely to be female, and more likely to have a history of depression, obesity, and treatment with antidepressants, anxiolytics, and mood stabilizers (

). In the preindex period, the rate of all-cause hospitalizations was approximately 0.06 per patient-month (ie, 6 hospitalizations per 100 patient-months).

The numbers of patients classified into each treatment group over follow-up are presented in 

. Use of each index monotherapy declined over time, with the greatest drop occurring between month 1 and month 2. The percentage of patients treated with each antipsychotic at month 12 relative to month 1 was descriptively highest for lurasidone (36.0% [194/539]), followed by aripiprazole (35.9%), olanzapine (33.5%), quetiapine (28.9%), and risperidone (24.5%).

Rate of Hospitalization Over Follow-Up Period

The unadjusted rate of all-cause hospitalization per 100 patient-months associated with lurasidone (3.42 [91 hospitalizations/2659 patient-months]) was lower than that of olanzapine (5.04 [493/9788]; 

 = .002), quetiapine (6.24 [750/12,015]; 

<.001), and risperidone (4.63 [798/17,233]; 

 = .01), but was not different from that of aripiprazole (3.26 [308/9436]; 

). After adjustment for preindex and time-dependent characteristics, the predicted rate of all-cause hospitalization associated with lurasidone was 2.78 per 100 patient-months, which was lower than that of olanzapine (4.15; 

). Adjusted rates of schizophrenia-related hospitalizations displayed a similar relationship.

Compared with lurasidone, the MSM-adjusted relative rates of all-cause hospitalization associated with olanzapine (adjusted rate ratio [aRR], 1.49; 

). Adjusted relative rates of schizophrenia-related hospitalizations were consistent with these estimates. Several preindex characteristics were significant predictors of hospitalizations. Male patients were 13% more likely to be hospitalized than female patients. History of anxiety, bipolar disorder, and depression was associated with 65%, 37%, and 48% increases in all-cause hospitalizations, respectively. History of substance, alcohol, or nicotine use was associated with a 61% increased risk of hospitalization, and hypertension was associated with an 18% increased risk (

Results of the sensitivity analysis without dose-equivalence restriction, presented in 

, consistently supported the findings of the primary analysis. After MSM adjustment, lurasidone monotherapy was associated with the lowest rate of all-cause hospitalizations. Differences in the rates of all-cause hospitalization reached statistical significance when lurasidone was compared against olanzapine, quetiapine, and risperidone, but not aripiprazole.

In this observational claims database study of patients with schizophrenia treated with equivalent doses of SGAs, lurasidone was associated with significantly fewer all-cause hospitalizations than were olanzapine, quetiapine, and risperidone. Lurasidone was also associated with a numerically lower rate of schizophrenia-related hospitalizations than other SGAs. We note that schizophrenia-related hospitalizations were less common than all-cause hospitalizations, resulting in lower statistical power.

The results from this study are consistent with those of 2 previous studies that compared hospitalization rates between lurasidone and other SGAs in patients with schizophrenia in real-world settings. The first study, conducted by an academic research group, used an ITT approach, did not control for dose, and used aripiprazole as the reference treatment.

 Relative to initiating treatment with aripiprazole, initiating treatment with lurasidone was associated with significantly fewer hospitalizations over 12 months (—5.98 hospitalizations per patient; 95% CI, –6.61 to –5.35]), followed by risperidone (–0.26; 95% CI, –0.34 to –0.17), and olanzapine (–0.16; 95% CI, –0.26 to –0.07). Quetiapine was associated with significantly more hospitalizations than aripiprazole (0.40; 95% CI, 0.32-0.49).

 Lurasidone was also associated with significantly fewer ED visits, lower hospital costs, and lower total costs compared with aripiprazole.

A second study that used a treatment-episode approach to compare patients with schizophrenia who switched from other SGAs to either quetiapine or lurasidone reported that patients who switched to quetiapine had significantly higher odds of all-cause (adjusted odds ratio [aOR], 1.64) and mental health-related hospitalizations (aOR, 1.74), but not schizophrenia-related hospitalizations (aOR, 1.35), compared with patients who switched to lurasidone.

 Therefore, the limited literature that has compared hospitalization risk associated with lurasidone and other SGAs in patients with schizophrenia has used different methods, but they consistently reported lower risk of hospitalization among patients treated with lurasidone.

In this study, persistence with SGAs was low. Assuming all patients using antipsychotic monotherapy in month 12 used the same antipsychotic consistently over follow-up, the ratio of patients treated with the study antipsychotic at month 12/month 1 may reflect an upper limit on persistence. This ratio ranged from 24.5% to 36.0%, and lurasidone had the highest value, suggesting greater persistence compared with other antipsychotic agents. Antipsychotic adherence and persistence have been linked to reduced hospitalization rates

Because changes in antipsychotic treatment are common, if an ITT approach had been utilized, the treatment patients were receiving may not have matched the treatment group to which they were assigned. The MSM approach appeared to better capture the effects of medication on hospitalization risk than an ITT approach because the ITT approach implicitly assumes high persistence.

An important aspect of this study was the use of equivalent doses. In usual clinical care, antipsychotics may be used at doses outside of the recommended dose range.

 However, in the primary analysis, comparison was limited to doses equivalent to 80 mg or less per day of lurasidone. Equivalent doses of antipsychotics have been estimated using multiple methods.

 One leading method used data from double-blind clinical trials to estimate a dose that had efficacy equivalent to 100 mg per day of chlorpromazine.

 The current study used chlorpromazine dose equivalents summarized from multiple studies and published by the College of Psychiatric and Neurologic Pharmacists.

 Therefore, the current study utilized an approach that is consistent with the AHRQ’s recommendation for dose equivalent comparison of treatment effects.

In the current study, dosing of antipsychotics was generally low based on recommended dose ranges for adult schizophrenia.

Use of low doses was most common for risperidone and quetiapine. However, the low doses observed were generally consistent with previously documented doses in real-world clinical practice.

The results of the current study indicated that monotherapy lurasidone was associated with fewer all-cause hospitalizations than other SGAs at doses equivalent to 80 mg or less per day of lurasidone. A sensitivity analysis without dose-equivalence restrictions reached similar conclusions. Notably, in 2006, the cost of a single schizophrenia hospitalization in the United States was estimated to be $8334 for a patient with Medicaid and $7802 for a patient with private insurance.

 A more recent estimate from the National Inpatient Sample Database showed the average cost of hospitalization for a person with schizophrenia was $9793 in 2016.

 For a large patient population, the cost savings associated with using lurasidone instead of these more commonly used antipsychotics could be substantial.

The administrative claims data were collected for administrative purposes rather than research purposes. Although the data reflected usual clinical care and provided excellent measures of healthcare resource use, the data were subject to potential coding errors, and important measures such as symptom severity were not available. Furthermore, although the MSMs adjusted for multiple preindex and time-varying covariates, unmeasured confounding variables could still exist. Statistical adjustments for multiple comparisons were not made. Finally, patients without health insurance could not be included and the results may not be generalizable to this population.

In this claims database analysis of patients with schizophrenia treated with equivalent doses of SGAs, those treated with lurasidone monotherapy had a lower risk of hospitalization than patients treated with olanzapine, quetiapine, or risperidone monotherapy. As hospitalization is a major cost driver of direct healthcare cost for individuals with schizophrenia, lurasidone may be a cost-saving treatment option.&ensp;

 Sunovion Pharmaceuticals; STATinMED Research.

 Financial support for this work was provided by Sunovion Pharmaceuticals Inc.

 Drs Ng-Mak, Messali, and Loebel are full-time employees of Sunovion Pharmaceuticals Inc. Ms Huang and Dr Wang are full-time employees of STATinMED Research.

: The authors would like to thank Alex Liu from STATinMED Research for data analytical support, and Michael Stensland from Agile Outcomes Research, Inc, for medical writing support.

 Concept and design (DNM, AM, AH, LW, AL); analysis and interpretation of data (DNM, AM, AH, LW, AL); and critical revision of the manuscript for important intellectual content (DNM, AM, AH, LW, AL).

: Daisy Ng-Mak, PhD, Sunovion Pharmaceuticals Inc, 84 Waterford Dr, Marlborough, MA 01752. Email: daisy.ng-mak@

Diagnostic and Statistical Manual of Mental Disorders

, 5th ed. Washington, DC: American Psychiatric Publishing; 2013.

Saha S, Chant D, Welham J, McGrath J. A systematic review of the prevalence of schizophrenia. 

. 2005;2(5):e141. doi: 10.1371/journal.pmed.0020141.

Karagianis J, Novick D, Pecenak J, et al. Worldwide-Schizophrenia Outpatient Health Outcomes (W-SOHO): baseline characteristics of pan-regional observational data from more than 17,000 patients. 

. 2009;63(11):1578-1588. doi: 10.1111/j.1742-1241.2009.02191.x.

Cloutier M, Aigbogun MS, Guerin A, et al. The economic burden of schizophrenia in the United States in 2013. 

. 2016;77(6):764-771. doi: 10.4088/JCP.15m10278.

Articles

Hospitalization Risk in Patients with Schizophrenia Treated with Dose-Equivalent Antipsychotics

Improvements in patient outcomes from the meaningful use of electronic health records (EHRs) have not been extensively studied among patients with schizophrenia. This study assessed the association between EHR use, provision of quality care, and patient outcomes. Providers who were at least 50% compliant with current requirements for the CMS Electronic Health Records Incentive Program were classified as EHR providers. Quality of mental health care was assessed using 4 Healthcare Effectiveness Data and Information Set indicators. Patient outcomes included inpatient admissions and emergency department visits. A total of 18,305 providers (EHR: 16.6%; non-EHR: 83.3%) treated 27,153 patients with schizophrenia. EHR use was associated with improved rates of diabetes screening (77.9% vs 72.4%; 

<.001), diabetes monitoring (72.1% vs 61.4%; 

<.001), and better antipsychotic adherence (54.7% vs 36.6%; 

<.001). EHR use was also associated with fewer inpatient admissions (15.9% vs 25.2%; 

<.001) and emergency department visits (32.2% vs 49.3%; 

<.001). These data suggest that EHR use may have a positive influence on the process and outcomes of psychiatric care when treating patients with schizophrenia. More research is needed to identify the drivers of the influence of EHRs and to develop programs that ensure all EHR users enjoy the same potential benefits as demonstrated here.

Improving healthcare quality is a top priority for CMS, which is consistent with the Triple Aim of optimizing patients’ quality of care and improving population health, while maintaining the affordability of healthcare.

 The use of electronic health records (EHRs) to appropriately diagnose, treat, and manage patient care through meaningful tracking and monitoring systems is an approach that is being widely adopted in the United States and elsewhere.

In the United States, the Health Information Technology for Economic and Clinical Health (HITECH) Act incentivized the use of healthcare technologies by providers to improve coordination of care.

 Meaningful use of EHRs, as defined in the HITECH Act, set the minimum standard for the incentives. Implementation of meaningful use occurred in 3 sequential stages with the objectives of each stage overlapping and building on those prior in a cumulative manner. Stage 1, which began in 2011, focused on capturing and sharing basic patient information through 15 core objectives.

 Stage 2, which began in 2014, focused on advanced clinical processes through 17 core objectives (eg, identifying patients who should receive preventative/follow-up care, identifying patient-specific education resources, using secure electronic messaging to communicate with patients).

 Stage 3, which began in 2017, focused on improving patient outcomes.

 The specific requirements and timelines for stages 2 and 3 have since been modified by newer legislation.

The majority of providers utilize some form of EHR, and most believe EHR use improves quality of care.

 A review of the literature identified increased productivity, better quality, and improved data management as the most frequently reasons for adoption of EHR, while the most commonly reported barriers to adoption of EHR were data errors, a lack of standards, and productivity loss.

  At one academic healthcare center, EHRs were found to be a substantial burden on primary care physicians’ time (5.9 hours per day),

 although the time to complete the same clinical process using paper charts was not estimated. Critics have argued that there is no consistent evidence that EHR use leads to improved patient outcomes.

 Although the meaningful use of EHRs has been correlated with enhanced healthcare quality and processes,

 the impact of the meaningful use of EHRs on patient outcomes, especially in treating serious mental illness such as schizophrenia, has not yet been determined.

Schizophrenia is uncommon, affecting only 0.5% to 1% of the population in the United States

; however, it is one of the most expensive and disabling mental illnesses.

 In 2013, the total cost—including direct healthcare costs, direct nonhealthcare costs (eg, law enforcement, homeless shelters, and research and training), and indirect costs (eg, unemployment, productivity loss, premature mortality, and caregiving)&mdash;of schizophrenia in the United States was estimated at $155.7 billion, with $37.7 billion reflecting the direct healthcare costs.

 Schizophrenia ranks as the sixth most disabling disease worldwide, due in part to severe psychiatric symptoms such as delusions, hallucinations, and social withdrawal.

 In addition, patients with schizophrenia also tend to have higher rates of diabetes and cardiovascular disease compared with the general population.

 Many of the atypical antipsychotics that are the first-line treatment for schizophrenia

cause weight gain, which in turn can exacerbate these comorbid conditions.

Recognizing these issues, the National Committee for Quality Assurance added Healthcare Effectiveness Data and Information Set (HEDIS) measures to specifically address metabolic care for patients with mental illness. These 4 measures include diabetes screening, diabetes monitoring, cardiovascular monitoring, as well as the assessment of antipsychotic medication adherence.

 The results of a recent retrospective insurance claims database study indicated that over 80% of patients with schizophrenia who were taking antipsychotics received the recommended HEDIS metabolic monitoring.

 However, it remains unclear if meeting these specific quality-of-care metrics actually leads to improved patient outcomes. Thus, the current study assessed the association between the meaningful use of EHRs, the quality of patient care, and patient outcomes in schizophrenia.

This retrospective database analysis examined the meaningful use of EHRs and patient outcomes in schizophrenia utilizing 3 large datasets, which were linked through physicians’ National Provider Identifier (NPI) numbers. The Inovalon “More

 Registry”—a multipayer medical and pharmacy claims database from private and public (Medicare Advantage or Managed Medicaid) sector health plans in all 50 US states that contains information on 252 million unique patients&mdash;contained all of the data with the exception of providers’ EHR use.

 All patient health data in the Inovalon More

 Registry were de-identified in compliance with the Health Insurance Portability and Accountability Act.

Meaningful use of EHR information came from 2 separate CMS datasets. Providers in the Inovalon More

 Registry were grouped into either the EHR providers group or the non-EHR providers group. Those who were listed in the

CMS Electronic Health Records Incentive Program dataset as at least 50% compliant with current requirements were identified as having meaningful use of EHR. Those who were not identified in this dataset were either not participating in meaningful use of EHR or simply not seeing Medicare or Medicaid patients. Therefore, all non-EHR providers were also required to be in CMS’s “Medicare Provider Utilization and Payment Data: Physician and Other Supplier” dataset. Less than 0.5% of providers were excluded because they were not identified in either of these 2 datasets.

Because these data were analyzed at the provider rather than patient level, the claims of all patients who met the study inclusion or exclusion criteria were linked with their providers. Based on the patient—provider linkages, the group of patients under the care of a provider during the study period was labeled the patient panel for that provider. If a single patient saw multiple providers, the patient was included on the patient-panels for each of the providers that saw the patient.

Patients were included if they were 18 years or older and diagnosed with schizophrenia (

International Classification of Diseases (ICD)-9-Clinical Modification (CM)

 295.xx). Schizophrenia diagnosis was determined by either a single inpatient admission or ED claim with a primary diagnosis of schizophrenia, or 2 outpatient claims with a primary diagnosis of schizophrenia within a 1-year period. Patients were required to have continuous enrollment with medical and pharmacy eligibility from January 1, 2014, through December 31, 2014. Patients also needed to qualify for at least 1 of the 4 HEDIS mental health quality measures (ie, diabetes screening, diabetes monitoring, cardiovascular monitoring, assessment of antipsychotic medication adherence) assessed in 2014 as part of the primary analysis. See the 

Providers were included if they met all of the following criteria: accepted commercially insured and Medicare or Medicaid members, had at least 1 patient qualifying for 1 of 4 HEDIS mental health quality measures, and were not part of a group practice. Providers with an NPI number associated with a group practice were excluded if there was no other NPI number that identified him or her as an individual provider because the care could not be attributed to a single physician.

High- vs General-Quality Provider Definition

Providers were categorized as delivering either high- or general-quality care

using CMS’ “Guide to Quality Performance Scoring Methods for Accountable Care Organizations,” which outlined

33 measures of quality care in 4 domains: Patient/Caregiver Experience, Care Coordination/Patient Safety, Preventive Health, and At-Risk Population.

 Providers were defined as high-quality providers if at least 90% of their patients achieved all measures for which they were eligible; otherwise, the providers were classified as general-quality providers.

The quality of care provided in mental health was evaluated using the proportion of 4 HEDIS quality measures met by the provider’s patient panel. The specific HEDIS measures were defined based on the National Committee for Quality Assurance definitions for diabetes screening (for those treated with antipsychotics), diabetes monitoring (for those with diabetes), cardiovascular monitoring (for those with cardiovascular disease), and antipsychotic medication adherence (for those treated with antipsychotics).

In addition, patient outcome proxy variables were computed and aggregated at the provider level. Two relapse-related proxies based on healthcare service utilization included: (1) at least 1 ED visit with a primary diagnosis of a relevant mental health condition (

 295.xx—305.xx); and (2) at least 1 inpatient admission with a primary diagnosis for schizophrenia (

Among the subset of patients who had relevant lab values, mean values of glycated hemoglobin (A1C) and low-density lipoprotein cholesterol (LDL-C), as well as the proportion of patients with elevated A1C of at least 7% and patients with LDL-C of at least 160 mg/dL, were examined.

Means and standard deviations were computed for continuous variables, and frequencies and proportions were computed for discrete or event-related variables. All indicators were computed at the provider level. A patient’s data could contribute to multiple providers’ patient-panels. Patients who saw more than 1 provider—both those who engaged in meaningful use of EHR and non-EHR&mdash;were only included in patient-panels of the most common provider type (see 

for details). The grand mean across providers in meaningful use of EHR and non-EHR groups, as wells as the CMS high- and general-quality groups, were computed and weighted by the providers’ patient-panel sizes. Outcome measures were compared between meaningful use of EHR and non-EHR providers using 

 tests for continuous variables and Chi-square tests for categorical variables with a 2-tailed alpha of 0.05. All analyses were completed using SPSS Statistics for Windows (IBM Corporation; Armonk, NY).

A total of 18,305 providers met the eligibility requirements. The meaningful use of EHR providers (n = 3046; 16.6%) had an average patient-panel size of 1.9 patients (SD = 2.3; range = 1-58), whereas the non-EHR providers (n = 15,259; 83.3%) had an average patient-panel size of 4.0 patients (SD = 14.3; range = 1-1022). About one-third of the providers (n = 5856; 32.0%) met the CMS criteria for high-quality providers with an average patient-panel size of 1.5 patients (SD = 1.0; range = 1-19), with the remaining general-quality providers (n = 12,449; 68.0%) having an average patient-panel size of 4.6 patients (SD = 15.8; range = 1-1022). CMS high-quality providers were overrepresented among the meaningful use of EHR providers (27.1%) compared with the non-EHR providers (19.4%).

The 18,305 providers treated a total of 27,153 unique patients with schizophrenia. Patients could see multiple providers and, therefore, could be included on more than 1 provider’s patient-panel. Non-EHR and meaningful use of EHR providers appeared to have patient-panels that were relatively similar in demographic characteristics (

). The patient-panels (values shown here as non-EHR vs EHR) were approximately evenly split on gender (male: 53.1% vs 49.8%), were primarily located in the Northeast (51.4% vs 46.1%) or South (37.3% vs 36.2%) regions, and were predominately covered by either Managed Medicaid (57.9% vs 52.8%) or Medicare Advantage (36.9% vs 39.4%) plans.

EHR Use and Provision of Quality Healthcare

Significantly more EHR providers delivered quality patient care as measured by HEDIS mental health quality measures (44.9%) than the non-EHR providers (29.4%; 

<.001). Specifically, meaningful use of EHR providers were more likely to deliver diabetes screening (77.9% vs 72.4%; 

<.001) and diabetes monitoring (72.1% vs 61.4%; 

<.001), and had patient-panels with better antipsychotic adherence (54.7% vs 36.6%; 

Meaningful Use of EHR and Patient Outcomes

, meaningful use of EHRs appeared to lead to better patient health outcomes. Compared with non-EHR providers, EHR providers had significantly fewer patients with relapse-related healthcare utilization, such as inpatient admissions (16.9% vs 25.2%; 

<.001). Similarly, EHR providers were less likely to have patients with elevated A1C (≥7%) values (32.9% vs 37.8%; 

 = .043) and elevated LDL-C (≥160 mg/dL) values (2.8% vs 6.8%; 

= .007) among the approximately 2000 patients with these laboratory values available.

Patient Outcomes by CMS Provider Quality and Meaningful Use of EHR

For the analysis, groups were stratified by high- and general-quality providers within their level of EHR use to assess if the associations between meaningful use of EHR and patient health outcomes were a function of the quality of the provider. Within each stratum (high- or general-quality providers), meaningful use of EHR providers had better patient health outcomes than non-EHR providers. In addition, across the 4 subgroups, the high-quality and meaningful use of EHR provider subgroup had the best patient outcomes, whereas the general-quality and non-EHR provider subgroup had the worst patient outcomes (

At the time of our research, only a single prior study evaluating the impact of EHRs on patients with schizophrenia was identified. This 2017 study, conducted in a mental health facility in Ontario, Canada, found implementation of EHRs led to an increase in metabolic monitoring (from 76.7% to 81.6%), as well as a 19.7% decrease in the use of restraints.

 To our knowledge, ours is the first study in the United States examining the association between meaningful use of EHRs and patient outcomes in schizophrenia. The current study was conducted using claims data from the calendar year 2014 when providers were incentivized for implementing stage 1 or stage 2 of meaningful EHR use. Higher quality of care was delivered by providers who used EHRs across 4 of the HEDIS process measures specific to mental health. More importantly, the patient outcomes for providers who used EHRs were significantly better, as measured by fewer patients with psychiatric inpatient admission and ED visits, as well as fewer patients with elevated laboratory values for diabetes and cardiovascular disease. For both CMS-defined high- and general-quality providers, those engaging in the meaningful use of EHRs delivered better quality of care and had better health outcomes in patients with schizophrenia.

Outside of schizophrenia, most US-based studies have examined the effects of EHRs on quality of care rather than patient outcomes; nevertheless, the results of those examining the effects of EHRs on patient outcomes have been mixed.

 In a large study of Medicare fee-for-service beneficiaries in New York State, there was no meaningful reduction in readmissions following the implementation of EHRs that electronically captured data.

 In contrast, a meta-analytic review synthesizing 24 individual studies found that the implementation of new health information technology in hospitals led to a 15% decrease in inpatient mortality.

 A 2018 study examining inpatient mortality rates among Medicare beneficiaries found more nuanced results: mortality rates decreased over time as EHRs incorporated more functionality.

 The emerging evidence surrounding patient outcomes suggests that the benefits of EHRs do not come from simply collecting data electronically,

 but from implementing additional clinical functionality, such as computerized physician orders and clinical guidelines.

 As meaningful use of EHRs progresses beyond stage 1, greater gains may be seen in patient outcomes.

Although the research linking EHR use to patient outcomes has been limited, several prior studies have associated EHR use with improved healthcare quality. A meta-analytic review of 47 studies linked EHR use to better healthcare quality, including greater guideline adherence, fewer medication errors, and fewer medication adverse events, particularly for EHR use with greater functionality, such as decision-support systems.

Improving the quality of care is likely an important intermediate step to improving patient outcomes. Past research has reported that a 10% improvement in antipsychotic adherence (based on the medication possession ratio) was associated with a 13% reduction in the odds of hospitalization.

 The 18% increase in antipsychotic adherence found in this study may be part of the mechanism that led to the 10% reduction in inpatient admissions. Similarly, increased healthcare quality through higher rates of screening and monitoring of potential metabolic adverse events associated with some antipsychotics may also contribute to the lowered inpatient admission and ED visits. Patients with schizophrenia and diabetes have been found to have more ED visits (7.5 visits/year) than those with only schizophrenia (5.3 visits/year).

 Likewise, patients with schizophrenia and cardiometabolic comorbidities have been found to have higher admission rates and healthcare costs than those with fewer or no cardiometabolic comorbidities.

 Meaningful use of EHRs likely leads to improvements in the quality of care, which in turn leads to better patient outcomes.

Schizophrenia was one of the most expensive psychiatric conditions in the United States in 2013, with $37.7 billion in direct healthcare costs.

 Inpatient admissions and ED visits were the primary cost drivers, with a combined annual cost of $17.8 billion in 2013.

 Implementing meaningful use of EHRs appears to be associated with improved metabolic monitoring, which may reduce the metabolic adverse events associated with some atypical antipsychotics and improve antipsychotic adherence. These improvements in quality of care could then possibly lead to improved patient outcomes and reduced healthcare resource use costs.

This cross-sectional study assessed the association between the meaningful use of EHRs and patient outcomes at a single point in time; therefore, future longitudinal research is warranted to understand the impact of EHR use on long-term improvements in patient outcomes. Further, more research is needed to identify the drivers of the influence of EHRs and to develop programs that ensure that all EHR users experience the same benefits as demonstrated here. Notwithstanding this limitation, to our knowledge this descriptive analysis is the first study of this kind in schizophrenia to assess the association between EHR use, healthcare quality, and patient outcomes. Providers who utilized EHR systems but were not enrolled in the CMS incentive program would have been misclassified as non-EHR providers based on the study definition, potentially underestimating the impact of EHR use on outcomes. As previously noted, only a small percentage (0.5%) of the providers were not identified in the CMS incentive program. Finally, patient outcomes data could have contributed to multiple providers, potentially violating the independence assumption of the statistical tests. The authors, however, attempted to account for repeated observations of patients across EHR and non-EHR providers by including patients who saw both EHR and non-EHR providers in patient-panels of the most common provider type only (see 

 for details and additional operational definitions).

The meaningful use of EHRs in schizophrenia was associated with better quality of care, including improved antipsychotic adherence, better diabetes screening and monitoring, as well as reduced relapse-related healthcare utilization. Although schizophrenia only affects 0.5% to 1% of the US population, it was estimated to have direct healthcare costs of $37.7 billion to the US healthcare system in 2013. For schizophrenia, greater use of EHRs may serve the Triple Aim of improving the quality of patient care, improving population health, and reducing healthcare costs.

 Further research using longitudinal studies is needed to confirm these findings and to isolate drivers linking EHR use with improved outcomes.&ensp;

Sunovion Pharmaceuticals; Health Analytics, LLC.

Dr Ng-Mak is a full-time employee of Sunovion Pharmaceuticals Inc. Dr Reutsch is a full-time employee and owner of Health Analytics, LLC, which received funding from Sunovion, Inc to complete the study analysis.

: Brenna Brady, PhD, and Joseph Tkacz, MS, CCRP, formally with Health Analytics, were involved in initial data analysis. Writing support was provided by Michael D. Stensland, PhD, and Gwendolyn F. Elphick, PhD, of Agile Outcomes, Inc.

Concept and design (DNM and CR); analysis and interpretation of data (CR); and critical revision of the manuscript for important intellectual content (DNM and CR).

Daisy Ng-Mak, PhD, 84 Waterford Dr, Marlborough, MA 01752. Email: daisy.ng-mak@sunovion.com.

Daisy Ng-Mak, PhD

Articles

Hospitalization Risk in Patients with Schizophrenia Treated with Dose-Equivalent Antipsychotics

Association Between Meaningful Use of Electronic Health Records and Patient Health Outcomes in Schizophrenia: A Retrospective Database Analysis

Measuring Migraine-Related Quality of Care Across 10 Health Plans