The use of prior authorizations (PA) by state Medicaid

Transcription

1 RESEARCH Impact of a Prior Authorization Policy for Montelukast on Clinical Outcomes for Asthma and Allergic Rhinitis Among Children and Adolescents in a State Medicaid Program Shellie L. Keast, PhD; David Thompson, PhD; Kevin Farmer, PhD; Michael Smith, PhD; Nancy Nesser, PharmD, JD; and Donald Harrison, PhD ABSTRACT BACKGROUND: Public policymakers often struggle with increased membership and limited budgets. Restrictions, commonly in the form of prior authorizations, are often placed on more costly pharmaceuticals, especially when lower cost or more effective products are available. Restrictions placed on products for difficult-to-manage disease states must be reviewed in order to ensure that unintended clinical consequences do not occur. OBJECTIVE: To assess the impact of a prior authorization policy for montelukast on clinical outcomes for asthma and allergic rhinitis among children and adolescent members of Oklahoma Medicaid (MOK) from 2007 through METHODS: Monthly individual-level utilization data were collected from MOK paid pharmacy and medical claims from January 1, 2007, through December 31, 2010, for members with asthma and/or allergic rhinitis. Members who were continuously eligible for the entire 48-month review period were included. The effect of a prior authorization policy for montelukast on emergency room (ER) utilization, disease-related physician office visits (DRV), and antibiotic prescription utilization (ABX) was analyzed using segmented logistic regression. RESULTS: For all 3 outcomes, decreases in mean number of claims per member per month were detected when comparing the pre-implementation and post-implementation prior authorization periods for all 3 disease states of asthma, allergic rhinitis, or both. Odds of having an ER event at the point of prior authorization implementation were 0.71 (P < 0.001) and were 1.29 (P < 0.001) and 1.26 (P < 0.001) for DRV and ABX, respectively. Overall trend in odds was 1.02 (P < 0.001), 0.93 (P < 0.001), and 0.95 (P < 0.001) for ER, DRV, and ABX, but during the post-implementation period, the odds were 0.92 (P < 0.001) for ER and 1.03 (P < 0.001) for both DRV and ABX. The final result was an increasing trend prior to implementation for ER, a decrease at implementation, and a continued decrease in odds of an event in the post-implementation period. However, for DRV and ABX, there was an overall decrease in trend regardless of period, with a small increase in odds at the point of implementation. CONCLUSIONS: While there was a point increase at implementation for DRV and ABX, the overall trend remained negative, indicating that no unexpected adverse clinical outcomes occurred. Additionally, no signal was found in ER use after implementation to indicate that unintended consequences occurred, particularly for those patients with asthma. J Manag Care Pharm. 2014;20(6): Copyright 2014, Academy of Managed Care Pharmacy. All rights reserved. What is already known about this subject While prior authorization policies in Medicaid programs have been shown to save money, unintended clinical consequences are often not monitored or detected. Asthma is a complex disease state that can have significant financial impact on the health care system if maximum disease control is not achieved. Policies that affect difficult-to-manage disease states should be carefully monitored. What this study adds A prior authorization process for a single pharmaceutical product with multiple indications can be successfully implemented based on different diagnoses, severities, and age groups. Isolating the effects of an individual policy can be difficult when reviewing a complex and dynamic health care system, and caution should be used when evaluating results. Point increases in physician office visits may occur at the initiation of a prior authorization due to adjustments needed in overall therapy. The use of prior authorizations (PA) by state Medicaid policymakers, as a strategy to help control drug expenditures, has been widely researched over the past 2 decades. However, continuous review of the strategy is necessary, particularly in Medicaid populations, to ensure that best practices are applied and that the most cost-effective use of resources is achieved. The review of PA policies in Medicaid has often focused on costs associated with newly implemented restrictions, but policymakers should also review these policies for the occurrence of unintended clinical consequences. 1 Polinski et al. (2007), in their review of access to atypical antipsychotic medication in state Medicaid programs, discussed the problem with policies that are based on costs alone and do not distinguish between appropriate clinical use for different disease states and their relevant costs. 2 Further, as more products gain U.S. Food and Drug Administration (FDA) approval for additional indications that are different from their original applications, policymakers must develop drug use policies that 612 Journal of Managed Care Pharmacy JMCP June 2014 Vol. 20, No. 6

2 Children 11 years Children 12 and adults Children < 2 years Children 2 years TABLE 1 Prior Authorization Approval Criteria for Montelukast in Oklahoma Medicaid Asthma Diagnosis of asthma, or A claim for inhaled corticosteroid, or Use of 3 or more rescue medications Diagnosis of mild or moderate persistent asthma, and/or exercise induced asthma, and Trial of inhaled corticosteroid and corticosteroid/laba therapy within the previous 6 months and reason for trial failure Allergic Rhinitis Trial of an oral antihistamine, 14 days in duration, that has failed to relieve allergic rhinitis symptoms within the past 30 days Trial of an antihistamine and nasal corticosteroid, each 14 days in duration, that failed to relieve allergic rhinitis symptoms Antihistamine and nasal corticosteroids may be used concomitantly or consecutively within the past 30 days account for the clinical features of each disease state. Such was the problem confronted by Oklahoma Medicaid (MOK) when developing and implementing its PA policy for montelukast. Montelukast is a leukotriene receptor antagonist originally approved in 1998 for prophylaxis and treatment of chronic asthma (AX) in adults and pediatric patients 6 years of age. The FDA later granted an additional indication for the relief of symptoms of seasonal allergic rhinitis (AR) in adults and children 2 years of age on December 31, 2002, and for perennial AR on July 27, In 2008, MOK spent $15.9 million dollars on montelukast (Singulair), or approximately 5% of the total reimbursement to pharmacies for all prescription products for that year. It was also in the top 5 prescription drug products for MOK. In January 2009, because of concerns for potentially high and inappropriate prescribing of montelukast for members diagnosed with AR, MOK implemented a PA on montelukast. Previously, MOK had adopted PA guidelines for antihistamine products and nasal allergy products but had not required a PA for montelukast out of concern about restricting the product s use for patients with AX. Because of the dual indications for montelukast, difficulties existed in attributing drug use between the 2 indications. MOK fashioned the new PA so that criteria for approval were applied appropriately to each of the 2 disease states for which the FDA had approved indications (Table 1). These criteria were in line with the contemporary National Asthma Education and Prevention Program Expert Panel guidelines for management of AX. 4 The guidelines outline treatment of AX based on a child s age (0-4 years, 5-11 years, and 12 years). Contemporary guidelines for AR were similar for all ages, with nasal corticosteroids considered as first-line therapy. However, none of those products were FDA approved for children less than 2 years at the time of the PA implementation. 5 TABLE 2 Diagnosis Assignment Criteria Disease State Inferred Diagnosis ICD-9-CM Codes Allergic 2 AR-related prescriptions, or 477.xx rhinitis (AR) 2 montelukast prescriptions with no AX-related prescriptions or AX ICD-9-CM codes Asthma (AX) 2 AX-related prescriptions, or 493.xx 2 prescriptions for a leukotriene modifier other than montelukast, or 2 prescriptions for montelukast and 2 prescriptions for a beta-2 agonist Both (ARX) Presence of both AR and AX identified by above criteria 477.xx and 493.xx ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification. Policymakers need to understand whether drug use strategies, such as that implemented by MOK, for a drug that has indications for different but related disease states, produce unintended clinical consequences, including the need for additional therapies and poor clinical outcomes. Therefore, the objective of this research was to assess the impact of a PA policy for montelukast on the clinical outcomes for asthma and allergic rhinitis among children and adolescent members of MOK from 2007 through Methods Study Design Overview This research used retrospective longitudinal administrative claims data from the Oklahoma Health Care Authority (OHCA), drawing on its pharmacy and medical Medicaid paid claims database. The study assessed the probability of obtaining health care services or prescriptions for individuals diagnosed with AR, AX, or both (ARX) before and after the intervention of a PA policy for montelukast, using segmented logistic regression. This method was used in order to review the changes for each individual over time as well as for the group. OHCA is the single state agency designated to administer the Medicaid program in Oklahoma. The study was approved by the University of Oklahoma Institutional Review Board and OHCA. Data Source This research used data on pharmacy and medical paid claims maintained by Pharmacy Management Consultants, the pharmacy benefit manager for OHCA, in a secure server environment. Each pharmacy claim consisted of a single observation and contained information regarding the medication, reimbursement amount, quantity, number of days supplied, date filled, and prescriber. Medical claims consisted of outpatient physician visits, inpatient and outpatient hospital visits, and other nonpharmacy billable services. The medical claims contained diagnoses codes, procedure codes, dates of service, and reimbursement amount. Additional records containing Vol. 20, No. 6 June 2014 JMCP Journal of Managed Care Pharmacy 613

3 FIGURE 1 Eligible for TXIX benefits January 2007-December 2010 n = 1,392,353 Age 18 years n = 755,722 Diagnosed with allergic rhinitis, asthma, or both n = 245,918 Final study population n = 15,903 Inclusion and Exclusion Flowchart Exclude Medicare eligibles n = 156,226 Exclude > 18 years n = 636,631 Exclude diagnoses of COPD, emphysema, or CF n = 9,247 Exclude lack of continuous eligibility n = 230,015 CF = cystic fibrosis; COPD = chronic obstructive pulmonary disease; TXIX = Title 19. beginning and ending eligibility dates and patient demographic information were also included. While all pharmacy and medical claims data can be linked via an OHCA member identification number, all personally identifiable elements were removed prior to data analysis. Data Inclusion and Exclusion Criteria There were a total of 1,392,353 MOK members eligible for Title 19 (TXIX) benefits from January 1, 2007, through December 31, TXIX eligibility is the standard criterion for Medicaid eligibility used by the states. Any member with Medicare eligibility at any time during the study period was excluded from the study, as these members had incomplete Medicaid pharmacy and medical claims data. Because of the multiyear longitudinal nature of the study, only members who were less than 19 years of age at study start (January 1, 2007) were included in the study, since MOK children under the age of 21 have a different prescription benefit with no copay requirements and no monthly limit. The selection criteria of less than 19 years of age at study start also helped prevent members from aging out before the study end date (December 31, 2010). This inclusion criterion ensured that members who were eligible at study start would still be eligible at study end. Members were selected who had either an inferred (Table 2) or an actual diagnosis with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code, indicating a diagnosis of AR, AX, or ARX (Table 2) at any time during the study period. Members were excluded if they had a diagnosis of chronic obstructive pulmonary disease (COPD; ICD-9-CM 491.2x, 493.2x, 496, 506.4), emphysema (ICD-9-CM 492.x, 518.1, or 518.2), or cystic fibrosis (CF; ICD-9-CM 277.0x). Additionally, TABLE 3 Product SubGPI 10 Allergy-related prescription products Intranasal , , , corticosteroids , , , , , Nonsedating , , , , antihistamines Oral decongestants , , , , , First-generation antihistamines Leukotriene inhibitors (montelukast) , , , , , , , , , Ocular antiallergy , , , , , , , Nasal , , antihistamines Combinations , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Asthma-related prescription products B2 agonist , , , , , , , Oral asthma , , , products , Combination , inhalers Corticosteroid inhalers Prescription Product Classifications , , , , , , , , , Leukotriene inhibitors Biological SubGPI 2 or SubGPI 4 Antibiotics 01, 02, 03, 04, 05, 07, 1600, 1615, 1621, 1622, 1623, 1625, 1629, 1699 SubGPI = Medi-Span s generic product identifier therapeutic classification system. only members who were eligible for the entire 48-month study period were included in the final study population (Figure 1). Medications were determined to be AR or AX related based on Medi-Span s generic product identifier (GPI) therapeutic classification system at various levels (SubGPI 2, 4, or 10, Table 3). 6 Study Variables We chose 3 dependent variables for analysis based on their potential to indicate unintended clinical consequences (i.e., loss of disease control) possibly linked to the implementation of the PA. The 3 variables were emergency room (ER) visits, 614 Journal of Managed Care Pharmacy JMCP June 2014 Vol. 20, No. 6

4 TABLE 4 General Description of Total Included Population Category N Percent Total study population 15, a Male 9, Age (years) 0-4 6, , , Race White 10, Black 2, Other 2, Unknown Urban 12, Allergic rhinitis (age, years) 4, , , Asthma (age, years) 2, , Allergic rhinitis and asthma (age, years) 9, , , , Montelukast use (entire study population) 8, a Percentage of total MOK population under 21 years of age for entire study period. disease-related outpatient physician office visits (DRV), and antibiotic prescription utilization (ABX). An ER visit identified in the included member s claim history was considered related to AR or AX if an appropriate ICD-9-CM code was listed in the first or second ICD-9-CM code fields. Likewise for DRV, only those with a diagnosis of AR or AX in the first or second ICD-9-CM field were included. This requirement was to ensure that the visits were indeed related to the disease being studied and not to an unrelated condition. As previously mentioned, for AR- or AX-related medications, the SubGPI 10s for antibiotic prescriptions were used to indicate that an ABX was written for the member (Table 3). Data were collected by calendar month time periods. Months when a patient experienced a claim were coded as 1, and all months without claims were coded as 0. In order to model trends in use before and after the PA, the model included 3 independent variables related to time: time (in months), pre- and post-pa, and post-time trend. 7 The other independent variables included in the model were disease state (AR, AX, ARX) and age group (0-4, 5-11, and years). The covariates chosen for this research were based on the Andersen s Framework of Health Determinants and were the predisposing and enabling characteristics available in the administrative claims database. 8 These variables included age, gender, race, and urban designation. Age was recorded at study entry and used throughout the study period. Race was collapsed into a 3-level variable defined as white, black, and all others. Residence was designated as urban or rural by matching the member s county of residence with the U.S. Office of Management and Budget statistical areas. 9 Those living in metropolitan or micropolitan counties were regarded as urban. Groups were also defined according to whether the member received a prescription for montelukast. The following comorbidities were chosen based on their relationship to AR: sinusitis, nasal polyposis, conjunctivitis, otitis media with effusion, and upper respiratory infections (ICD-9-CM 473, 471, 372, 381, 382, 465). These comorbidities (along with asthma, breathing through the mouth, and decreased quality of life) are considered to be more frequent in patients with AR. 5 For AX, the following comorbidities were chosen: sinusitis, gastroesophageal reflux disease, depression, anxiety, and obesity (ICD-9-CM 473, , 296.2x, 296.3x, 311, 300, 278). 10,11 These comorbidities (and rhinitis) were some of the most commonly reported comorbidities of AX, especially childhood AX. AR and AX were not counted as comorbidities of each other. Analysis To determine whether there were changes in use of services before and after the policy implementation, both bivariate and multivariate analyses were performed to characterize the associations between the 3 main outcome variables and the 3 study groups (AR, AX, and ARX). Changes in subject-specific log odds of using services before and after the policy implementation were determined using segmented logistic regression analysis, applying generalized linear mixed models for each of the 3 dependent variables assessed (SAS PROC GLIMMIX). 12 The segmented logistic regression equation, including key covariates and interactions (see definitions of equation elements in the Appendix), was as follows: log (1-π) π =β 0+β 1 x 1 +β 2 x 2 +β 3 x 3 +β 4 x 4 +β 5 x 5 +β 6 x 6 +β 7 x7+ε i Gender, race, urban designations, and comorbidities were also controlled for in the model. Because of the large study sample, a more restrictive a priori alpha (P < 0.001) was employed. Data management and analysis were conducted using SAS software, version Results There was a total of 1,392,353 MOK members who were TXIX eligible between January 1, 2007, and December 31, From that total, 156,226 members were excluded due to Medicare eligibility. The population was further limited to those members who were less than 19 years of age at the study start. There were 245,918 members further identified as having either AR, AX, or ARX during this period. A total of Vol. 20, No. 6 June 2014 JMCP Journal of Managed Care Pharmacy 615

5 TABLE 5 Pre- and Post-Implementation Changes by Disease Category a Pre (n = 4,439) Allergic Rhinitis Asthma Allergic Rhinitis and Asthma Post (n = 4,071) P Value b Pre (n = 3,586) Post (n = 2,495) P Value b Pre (n = 7,878) Post (n = 9,337) P Value b Mean age (years) Urban, % With comorbidity, % < White, % Montelukast claims PMPM < < < ER visits PMPM < < < Antibiotic claims PMPM < < < DRV PMPM < < < a Pre-implementation period was January 2007 through December Post-implementation period was January 2009 through December b Chi-square tests were used to determine associations between variables and time periods. ER = emergency room; DRV = diagnosis-related outpatient visit; PMPM = per member per month. 9,247 members were excluded due to the presence of COPD, emphysema, CF, or use of an anticholinergic product. After excluding members who were not continuously eligible, the final study population consisted of 15,903 members (Figure 1). Among the 15,903 included members, 4,071 were categorized as having only AR throughout the study period; 2,495 were categorized as having only AX; and 9,337 categorized as having both AR and AX at some point during the study period (Table 4). The majority of the included population lived in an urban designated area (80.2%), with 69.1% of the population being considered white. Approximately 40% of the population was in each of the 0-4 and 5-11 year age ranges. The remaining 20% were in the year range. Finally, the study population was just over 56% male. Table 5 demonstrates the associations between select variables and the 2 time periods defined by implementation of the PA for montelukast. Among those diagnosed with ARX, there was a significant association between members with at least 1 comorbidity and time period. The per member per month (PMPM) claims for all 4 outcomes (montelukast prescriptions, ABX, DRV, and ER visits) decreased significantly in the post-implementation period when compared with the pre-implementation period. Estimates from the segmented logistical model are depicted in Table 6. All nonsignificant (P 0.001) interactions were removed from each model. For both DRV and ABX, the overall trend in log odds was negative ( , , P < 0.001), with a slightly more positive statistically significant trend in the post-implementation period ( , , P < 0.001). For ER visits, the overall trend was positive ( , P < 0.001), with a negative post-implementation trend ( , P < 0.001). Those diagnosed with the more complicated disease states, AX or ARX, had higher odds of experiencing any of the outcomes than those diagnosed with AR (P < 0.001). Compared with the younger age group, both older groups were negative. For DRV and ABX, interactions between disease states (AR, AX, ARX) and the point of PA or the post-implementation trend (PA Time) were not significant. Figure 2 displays the trend in predicted probability of occurrence of each outcome measure over the study period. Discussion This research attempted to describe potential clinically important unintended outcomes for patients with AR or AX subjected to a montelukast PA. Unlike other studies that have reviewed the effect of PA implementation on patients compared with others with no exposure, this study attempts to determine whether there were any unintended consequences on the entire population with the disease of interest. This study is also unique because it compared the related, yet distinct, disease states and age groups for which the product is indicated. Demographics Table 4 displays basic demographic estimates for the entire study population. Generally, the proportions are similar to the entire MOK population, with males being slightly higher in the study population. Previous research with patients in this age range indicated that 30.1% of the total eligible MOK population under 19 years of age during the study period had a diagnosis of AR. 13 National averages report that 40% of children in the United States have AR. 14 The lower rate for this research could be attributed to potential errors in coding or to the conservative nature of the inferred disease classification method used for this research. However, the comorbidity of AR with AX is reported to occur in approximately 60% to 70% of patients. 14,15 In this study, the comorbidity of AR and AX (ARX) appeared to be very similar to the national average, with 58.7% of the study population categorized as ARX. Because a member could change from AR or AX to ARX at any time in the study, the population classified as only AR or AX during the pre- and post-implementation periods was examined and displayed in Table 5. In both the AR-only and 616 Journal of Managed Care Pharmacy JMCP June 2014 Vol. 20, No. 6

6 TABLE 6 Estimates from Segmented Logistic Regression Model Emergency Room Visits Disease-Related Visits Antibiotic Prescriptions Variable Estimate P Value a Estimate P Value a Estimate P Value a Intercept < < < Time b < < < PA time b < < < PA b < < < AR reference reference reference AX < < < ARX < < < Age 0-4 reference reference reference Age < < Age < < < AR PA reference reference reference AX PA ARX PA AR time reference reference reference AX time < ARX time < < AR PA time reference reference reference AX PA time ARX PA time < Age 0-4 PA reference reference reference Age 5-11 PA Age 12 PA Age 0-4 time reference reference reference Age 5-11 time < < < Age 12 time < < < Age 0-4 PA time reference reference reference Age 5-11 PA time < < Age 12 PA time < < a Separate segmented logistic regression models were run for each outcome variable. b Pre- and post-trends in time and change in level at point of prior authorization. AR = allergic rhinitis; AX = asthma; ARX = asthma and allergic rhinitis; PA = prior authorization. AX-only populations, the number of members decreased after the policy implementation in This decrease coincided with an increase in the number of those classified as having ARX (Table 5). The shift in disease classification also produced shifts in mean ages among the 3 diagnostic groups, but the groups mean ages did not differ (P > 0.05) either before or after the PA. Additionally, the proportions of patients living in urban and rural counties did not differ before and after the PA. The only significant difference when time periods were compared was an increase in the proportion of patients with a comorbidity of interest in those patients categorized as ARX. The difference may be caused by an increasing likelihood of becoming diagnosed with both AR and AX as time progressed, resulting in more complicated patients comprising the ARX category by the end of the study. Montelukast Prescribing Overall, montelukast was prescribed at least once for 55% of the total study population, indicating that more than half of all study participants were exposed to montelukast at some point during the 4-year period (Table 4). When the mean PMPM claims for montelukast were compared, there was a significant decrease (P < 0.001) post-implementation for all diagnoses groups (Table 5). This decrease is consistent with the intent of implementing the PA, as well as with previously reported data on the effect of PA on prescription use Emergency Room Visits The ER visit PMPM decreased after the PA implementation for participants in all 3 disease groups (P < 0.001; Table 5). Members classified as AX had a 0.02 PMPM drop in level of the mean number of monthly visits in the post-implementation period, while AR and ARX had a drop in level of 0.01 PMPM. Table 6 indicates that after controlling for potentially confounding variables, the effect of the PA was to reduce the number of ER visits. The results of multivariate analyses demonstrate that there was both a decrease in level (y-intercept) Vol. 20, No. 6 June 2014 JMCP Journal of Managed Care Pharmacy 617

7 FIGURE 2 Predicted Event Probability Outcome Measures Probability a Study Month ER DRV ABX a Monthly predictions from segmented logistic regression for each outcome variable. ABX = antibiotic prescription; DRV = diagnosis-related outpatient visit; ER = emergency room. and trend (slope) in ER visits reported at the point of PA and post-implementation period, when compared with the time period prior to implementation. No statistical interactions were detected between age and time or between disease category and time. However, the decrease in ER visits may also be related to an effort initiated by MOK during this time period to reduce inappropriate ER visits. 20 Because the PA was designed to improve the rational use of montelukast, we speculate that the policy may have had the effect of improved prescribing of other medications (e.g., inhaled corticosteroids) that are at least as effective in controlling asthma; however, this assumption was not part of the research, and no other efforts were made during this time to educate patients, caregivers, or providers on the use of alternate therapy. The American Lung Association found in a randomized comparison of AX treatment strategies that patients with mild AX who used montelukast had a higher rate of treatment failure than those on inhaled corticosteroids. 21 Finally, these results do demonstrate that implementation of the PA for montelukast did not appear to negatively affect the control of patients with any of the disease states of interest. Antibiotic Prescribing Table 5 indicates that there was a significant reduction in mean PMPM antibiotic claims after implementation of the PA across all disease categories (P < 0.001). Interestingly, the reduction in PMPM antibiotic claims was highest for those with ARX (0.08 PMPM). Although the multivariate analyses, when considered as a main effect, revealed apparent increased odds of having an antibiotic claim at the point of PA (y-intercept), when the point estimate was interacted with age and diagnosis, the change was not significant (Table 6). Overall, these results also demonstrate that implementation of the PA for montelukast did not appear to negatively affect the control of patients with any of the disease states of interest. The increase in level at the point of PA could indicate a transient loss of disease stability or may be reflective of the peak in the influenza and respiratory syncytial virus (RSV) seasons for , which was the highest year for the entire study period. 22 Additionally, an informational intervention was attempted by OHCA with the highest antibiotic prescribers in late 2009, but the results of that intervention were not evaluated. Disease-Related Physician Office Visits Again, Table 5 indicates that the mean PMPM claims for DRV were significantly reduced after PA implementation, with patients categorized as ARX having the largest decrease of 0.07 PMPM (P < 0.001). The multivariate analyses depicted in Table 6 also indicate that, similar to antibiotic prescription claims, there was an overall decreasing trend in the odds of having a DRV during the entire study period, but there was an increase in the odds at the point of the PA implementation. Influenza and RSV seasons may be a possible factor in this increase, or the transient increase may be because of an increase in office visits as providers and members initially encountered the PA for montelukast and either received a PA or were prescribed an alternate therapy. Another factor that may have caused a transient increase in office visits in the first part of 2009 was the conversion from the albuterol inhalers that used chlorofluorocarbon propellants to the inhaler products that used hydrofluoroalkane propellants. 23 Also in January 2009, OHCA changed the method for payment of physician services for its largest benefit program. Prior to 2009, most members were enrolled in a partially capitated physician payment program, where most 618 Journal of Managed Care Pharmacy JMCP June 2014 Vol. 20, No. 6

8 services for outpatient physician visits were included in the capitation payment and therefore not billed for payment to OHCA. After January 2009, all visits had to be billed to OHCA in order to receive payment. The result would have been an immediate increase in services billed, but not necessarily services received from 2009 forward. Finally, and importantly, there was no statistically significant interaction between the point of PA and either age or disease category. These results indicate that regardless of what caused the increase at the point of implementation, the increase did not differentially impact any particular disease state or age group. Further, these results also depict that there did not appear to be any lasting adverse clinical consequences of implementing the PA for montelukast. Limitations This research used administrative claims data to determine whether potential unintended consequences occurred as a result of a PA policy. These data are submitted for payment of services and not for research purpose; therefore, there are inherent risks and limitations in any research using this type of data. There is also a possibility that diagnoses codes were not included on service claims, so individuals with the studied diseases were not included. The choice of the 3 clinical outcome variables may not be reliable determinants of unintended clinical consequences in this population. While we assumed that ER visits and outpatient visits are accepted surrogates, antibiotic prescribing may not be a reliable surrogate. Also, this research was performed in a pre/post quasi-experimental fashion. Even though each diagnosis category was compared with the others, there was not a similar group that was not exposed to the policy to serve as a comparator. This lack of a control group could limit the robustness and potential generalizability of the research. However, the methods used were sufficient to achieve the objectives of this research. While we determined that no suitable control or comparator group existed for this research, inclusion of a control group or analysis of newly diagnosed members before and after the intervention would possibly increase the robustness and generalizability of the research. Another possible limitation is that decreased service use might be a factor of increasing age. Since the population was continuously eligible for the entire study period, the fact that members were 2 years older at the time of the implementation and 4 years older at the end of the study may have biased the results. The inclusion of comorbidities is important to any research of this nature. Because the research was performed with children and adolescents, comorbidities unrelated to the disease itself were expected to be rare. Also, comorbidities for AX in children are minimal. Inclusion of additional comorbidities of other disease states possibly could influence the costs and clinical events experienced by the individuals in this research. Further research should consider inclusion of additional comorbidity information. The modeling of the count data in a binomial fashion could have created a loss of information regarding the combination of having an event and the magnitude of the count of events. However, because of the extremely small occurrence of greater than 1 event monthly for any individual, this loss of information is considered minimal. The results of this research may not be generalizable to the population as a whole or even to other Medicaid populations without careful review of the demographics of the comparison population. Also, because of the continuous eligibility requirement for the entire 4-year study period, the final study population may not be representative of the full population. While the age, gender, and race percentages remain comparable to the larger population regardless of continuous eligibility, a larger proportion of the study population was diagnosed with ARX. Thus, the study population may represent a subset of the overall population who have more severe disease. Other methods for inclusion based on enrollment might have been attempted in order to have a more generalizable study population. ER visits were analyzed using the entire population with at least 1 month of eligibility, and the results were similar. The reported results do indicate that disease-specific or group-specific PA criteria can be applied and avoid unintended clinical consequences. However, each program should develop criteria based on its unique population and conduct population-specific reviews. Conclusions The PA policy implemented for montelukast by MOK in 2009 appears to have achieved its goal of reduced use of the drug without adverse effects on disease control. Because this policy was implemented on 2 related diseases, with different perceived clinical severity, we wanted to determine whether a PA could be successfully implemented on a product that was frequently used for both diseases without causing unintended clinical consequences. While we cannot be completely certain that there were no unintended consequences, the results are consistent with positive overall effects. This conclusion is particularly true when considering ER visits, where there was a decrease at the point of the implementation followed by a continued decline in the post-implementation period. There was no increase in the odds of incurring an ER visits at the point of PA for montelukast, which was followed by declining odds of an event in the post-implementation period. For both DRV and ABX, there was an increase in the odds of having a claim at the point of PA, but both outcomes continued their negative trend in odds of occurrence after implementation, although the trend was less steep. Overall, the implementation of the PA policy for montelukast in MOK did not appear to cause unintended clinical consequences for AR, AX, or ARX patients; however, other factors not included in the research may also have influenced the studied outcomes. Vol. 20, No. 6 June 2014 JMCP Journal of Managed Care Pharmacy 619

9 Authors SHELLIE L. KEAST, PhD, is Assistant Professor; KEVIN FARMER, PhD, is Associate Professor; MICHAEL SMITH, PhD, is Assistant Dean for Tulsa Operations and Associate Professor; DONALD HARRISON, PhD, is Associate Professor, University of Oklahoma College of Pharmacy, Oklahoma City, Oklahoma. DAVID THOMPSON, PhD, is Associate Professor, University of Oklahoma College of Public Health, Oklahoma City, Oklahoma. NANCY NESSER, PharmD, JD, is Pharmacy Director, Oklahoma Health Care Authority, Oklahoma City, Oklahoma. AUTHOR CORRESPONDENCE: Shellie L. Keast, PhD, Assistant Professor, University of Oklahoma College of Pharmacy, CPB-222, P.O. Box 26901, Oklahoma City, OK Tel.: ext ; Fax: ; DISCLOSURES Harrison, Farmer, Smith, and Thompson are employees of the University of Oklahoma College of Pharmacy and report no financial disclosures. Keast was an employee of the University of Oklahoma under contract with the Oklahoma Health Care Authority. Nesser is an employee of the Oklahoma Health Care Authority. Study concept and design were primarily contributed by Keast and Harrison, with input from the other authors. Data collection and analyses were performed by Keast, with statistical assistance provided by Thompson. The manuscript was primarily written by Keast, with input from Harrison and Nesser. Review and revisions were performed by the other authors. REFERENCES 1. Soumerai SB. Benefits and risks of increasing restrictions on access to costly drugs in Medicaid. Health Aff (Millwood). 2004;23(1): Polinski JM, Wang PS, Fischer MA. Medicaid s prior authorization program and access to atypical antipsychotic medications. Health Aff (Millwood). 2007;26(3): U.S. Food and Drug Administration. SINGULAIR label and approval history. Available at: index.cfm?fuseaction=search.overview&drugname=singulair. Accessed April 23, National Asthma Education and Prevention Program. Expert Panel Report 3: Guidelines for the Diagnosis and Management of Asthma. Bethesda, MD: National Heart, Lung, and Blood Institute (NIH); Pawankar R, Holgate ST, Canonica GW, Lockey RF, eds. WAO White Book on Allergy. Milwaukee, WI: World Allergy Organization; Medi-Span. Master Drug Data Base Documentation Manual. Indianapolis, IN: Medi-Span; Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther. 2002;27(4): Andersen R, Newman JF. Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc. 1973;51(1): U.S. Census Bureau. Metropolitan and micropolitan statistical areas main. Available at: Accessed April 1, Boulet LP, Boulay MÈ. Asthma-related comorbidities. Expert Rev Respir Med. 2011;5(3): Gershon AS, Wang C, Guan J, To T. Burden of comorbidity in individuals with asthma. Thorax. 2010;65(7): SAS software. Version 9.3. Cary, NC: SAS Institute Inc. 13. Keast SL. Impact of prior authorization of montelukast on resources and outcomes for asthma and allergic rhinitis amongst children and adolescents in a state Medicaid program [dissertation]. Oklahoma City, OK: University of Oklahoma Health Sciences Center; Wallace DV, Dykewicz MS, Bernstein DI, et al. The diagnosis and management of rhinitis: an updated practice parameter. J Allergy Clinical Immunol. 2008;122(Suppl 2):S1-S Berger WE. Overview of allergic rhinitis. Ann Allergy Asthma Immunol. 2003;90(6 Suppl 3): Adams AS, Zhang F, LeCates RF, et al. Prior authorization for antidepressants in Medicaid: effects among disabled dual enrollees. Arch Intern Med. 2009;169(8): Law MR, Ross-Degnan D, Soumerai SB. Effect of prior authorization of second-generation antipsychotic agents on pharmacy utilization and reimbursements. Psychiatr Serv. 2008;59(5): Fischer MA, Choudhry NK, Winkelmayer WC. Impact of Medicaid prior authorization on angiotensin-receptor blockers: can policy promote rational prescribing? Health Aff (Millwood). 2007;26(3): Kloepfer KM, Helm ME, Perry TT, Hu P, Jones SM, Vargas PA. Preferred drug utilization: treating allergic rhinitis with less-sedating antihistamines. Am J Pharm Benefits. 2012;4(5):e130-e Oklahoma Health Care Authority. Minding our P s and Q s: the OHCA SFY 2008 performance and quality report Available at: org/workarea/downloadasset.aspx?id= Accessed April 23, American Lung Association Asthma Clinical Research Centers. Randomized comparison of strategies for reducing treatment in mild persistent asthma. N Engl J Med. 2007;356(20): Centers for Disease Control and Prevention. Update: influenza activity United States, season, and composition of the influenza vaccine. MMWR Morb Mortal Wkly Rep ;60(21): Available at: Accessed April 1, U.S. Food and Drug Administration. Transition from CFC propelled albuterol inhalers to HFA propelled albuterol inhalers: questions and answers Available at: QuestionsAnswers/ucm htm. Accessed April 1, Journal of Managed Care Pharmacy JMCP June 2014 Vol. 20, No. 6

10 Appendix Definitions for Segmented Logistic Regression Equation log π/(1-π) = log odds of receiving a service for a given month β 0 = intercept β 1 = estimate for monthly trend (or slope) before the intervention x 1 = month number (January 2007 = 1, December 2010 = 48) β 2 = estimate for the change in level at the point of intervention x 2 = PA variable (0 = pre-implementation period, 1 = post-implementation period) β 3 = estimate for change in trend (or slope) after the intervention x 3 = post-implementation period (0 = pre-implementation period, 1-24 = post-implementation period) β 4 = estimate for the effect of AX compared with AR (reference group) x 4 = dummy variable for presence of AX β 5 = estimate for the effect of ARX compared with AR (reference group) x 5 = dummy variable for presence of ARX β 6 = estimate for the effect of age between 5 and 11 compared with the youngest age group (reference group) x 6 = dummy variable for the effect of age between 5 and 11 β 7 = estimate for the effect of age greater than 12 compared with the youngest age group (reference group) x 7 = dummy variable for the effect of age greater than 12 AR = allergic rhinitis; ARX = allergic rhinitis and asthma; AX = asthma; PA = prior authorization. Vol. 20, No. 6 June 2014 JMCP Journal of Managed Care Pharmacy 621