Patient-centered medical homes and oral anticoagulation therapy initiation.

510563 research-article2013

MCR71210.1177/1077558713510563Medical Care Research and ReviewBeadles et al.

Empirical Research

Patient-Centered Medical Homes and Oral Anticoagulation Therapy Initiation

Medical Care Research and Review 2014, Vol. 71(2) 174–191 © The Author(s) 2013 Reprints and permissions: sagepub.com/journalsPermissions.nav DOI: 10.1177/1077558713510563 mcr.sagepub.com

Christopher A. Beadles1,2, Kristen Hassmiller Lich2, Anthony J. Viera2, Sandra B. Greene2, M. Alan Brookhart2, and Morris Weinberger1,2

Abstract Despite evidence-based guidelines, oral anticoagulation therapy (OAT) initiation is low among incident atrial fibrillation (AF) patients. Patient-centered medical homes (PCMHs) may increase access, quality, and value through coordinating care. As such, PCMHs hold potential for improving OAT initiation among AF patients. We estimated the effect of receiving care in accredited PCMHs on OAT initiation for incident AF patients compared with those not receiving care in accredited PCMHs. Our study, a retrospective cohort new user design, included privately insured patients in North Carolina during years 2006 to 2010. We developed propensity scores for PCMH exposure, performed inverse probability of treatment weighting, and estimated effects with generalized estimating equations. We found a positive association between PCMH exposure and OAT initiation in unadjusted (6.78%; p < .001) and adjusted (6.25%; p < .001) models. Greater implementation and optimization of PCMH model principles may enhance this association, reducing AF-related stroke morbidity and mortality. Keywords atrial fibrillation, anticoagulation, guideline adherence, patient-centered care This article, submitted to Medical Care Research and Review on July 25, 2013, was revised and accepted for publication on October 3, 2013. 1Center

for Health Services Research in Primary Care, Durham, VA, USA of North Carolina at Chapel Hill, NC, USA

2University

Corresponding Author: Christopher A. Beadles, Center for Health Services Research in Primary Care (152), Durham VA Medical Center, 508 Fulton Street, Durham, NC 27705, USA. Email: [email protected]

Downloaded from mcr.sagepub.com at EASTERN KENTUCKY UNIV on May 23, 2015

175

Beadles et al.

Introduction Acute stroke is the fourth leading cause of death in the United States (Go et al., 2013). Patients with atrial fibrillation (AF) have an average fivefold increase in ischemic stroke risk (Roger et al., 2012) that increases with age from 1.5% in individuals aged 50 to 59 to 23.5% for those aged 80 to 89 (Wang et al., 2003; Wolf, Abbott, & Kannel, 1991). Moreover, severity, disability, mortality, and frequency of stroke recurrence are greater with AF associated stroke (Dulli, Stanko, & Levine, 2003; Jorgensen, Nakayama, Reith, Raaschou, & Olsen, 1996; Lin et al., 1996). Since the first randomized clinical trial found that warfarin reduced stroke risk for patients with AF (Petersen, Boysen, Godtfredsen, Andersen, & Andersen, 1989), guidelines from major organizations and professional societies have recommended oral anticoagulation therapy (OAT) for AF patients at moderate and high risk of ischemic stroke (American College of Chest Physicians, 1995; Goldstein et al., 2006; Goldstein et al., 2011; Hart & Bailey, 2002). Yet studies repeatedly report OAT underutilization in eligible AF patients (Beyth et al., 1996; Birman-Deych, Radford, Nilasena, & Gage, 2006; Bungard, Ghali, Teo, McAlister, & Tsuyuki, 2000; Choudhry et al., 2006; Gage et al., 2000; Go et al., 1999; Harley et al., 2005; Hart, 1999; Johnston et al., 2003; Ogilvie, Newton, Welner, Cowell, & Lip, 2010; Zimetbaum et al., 2010). Physician and health care system factors contribute to OAT underutilization (Buckingham & Hatala, 2002; Bungard et al., 2003; Choudhry et al., 2006; Dantas, Thompson, Manson, Tracy, & Upshur, 2004; Johnston et al., 2003; Lane & Lip, 2008). Physician-level barriers to OAT initiation occur in at least three identifiable ways. First physicians may perceive that the risks of OAT exceed benefits for their patients (Beyth et al., 1996; Brodsky et al., 1996; Chang, Bell, Deroo, Kirk, & Wasson, 1990; Kutner, Nixon, & Silverstone, 1991; McCrory, Matchar, Samsa, Sanders, & Pritchett, 1995; Rodgers, Sudlow, Dobson, Kenny, & Thomson, 1997). But physicians frequently underestimate the relative stroke risk reduction of OAT and overestimate the risk of adverse events such as major gastrointestinal bleeding or intracranial hemorrhage (Chang et al., 1990). Second, physicians may be unfamiliar with guidelines recommending OAT for AF patients with moderate or high risk of ischemic stroke (Chang et al., 1990; Kutner et al., 1991). Finally, physicians may elect not to initiate OAT if they believe patients will be nonadherent (Beyth et al., 1996; Chang et al., 1990; Kutner et al., 1991). Health care system barriers to OAT initiation include poor access to appropriate laboratory facilities for necessary blood work, difficulty with provider schedule availability, and difficulty with followup on out of range laboratory values for OAT monitoring and dosage titration (Kutner et al., 1991; McCrory et al., 1995; Rodgers et al., 1997). Physician and health care system barriers may be mitigated by patient-centered medical homes (PCMHs). The PCMH is intended to increase access, quality, and value by coordinating patient-centered care to enhance population health. Despite the potential of the PCMH, many decision makers and policymakers are awaiting empirical evidence supporting its effectiveness. Although there is empirical support for individual components of PCMHs (Rosenthal, 2008), there is little evidence concerning


176

Medical Care Research and Review 71(2)

their ability to improve quality and health outcomes while reducing costs (DePalma, 2007; Homer et al., 2008; Peikes, Zutshi, Genevro, Parchman, & Meyers, 2012; Robert Graham Center, 2007). Two recent large systematic reviews of PCMHs found some favorable effects on quality, patient experience, and caregiver experience; a few unfavorable effects on costs; and mostly inconclusive results (Jackson et al., 2013; Peikes et al., 2012).

New Contribution We examined the association between receiving care in a National Committee on Quality Assurance (NCQA)–accredited PCMH and OAT initiation among incident AF patients in a large, insured population observed for 5 years. This research contributes to existing literature in three ways. First, our study examines receipt of evidence-based care for incident AF patients, OAT initiation. OAT initiation requires high-level performance in multiple practice processes of care (e.g., strong therapeutic relationship, enhanced access, care coordination, follow-up and monitoring, practice commitment to improving care quality and safety). While all of these processes may simultaneously occur in any practice, they are more likely to occur in a PCMH. This contribution is significant because it rigorously evaluates the PCMH model requiring these multiple component principles. Second, by creating a unique linked data set, we could consider the influence of patient, provider, practice, and county level variables, which allowed a more robust examination of our question. Additionally, the data set includes individuals with Medicare coverage. The 5-year observation period also allowed us a longer duration of assessment in the PCMH model, which has thus far been somewhat limited (typically 12-24 months). Finally, the study was conducted in North Carolina, which has a robust PCMH movement: during the study period, North Carolina had more accredited PCMHs than any other state except New York (NCQA, 2012). This resulted in our including more than 100 independent PCMHs, far more than any other study known to us. This heterogeneity allows a greater generalizability to the PCMH experience as a whole rather than the selected experience of a few PCMHs.

Conceptual Framework PCMHs may address multiple barriers and positively influence OAT initiation through several mechanisms, including a stronger therapeutic relationship with a personal physician, enhanced access to care (e.g., obtaining prescription refills, laboratory collections), greater follow-up and monitoring (e.g., laboratory values, unfilled prescriptions), coordination of care with specialists (e.g., cardiologists), and a practice commitment to improving quality and patient safety. These mechanisms are the standards espoused by the Joint Principles of the PCMH, which are individually supported in the literature but have not been well demonstrated as part of a multicomponent intervention. OAT initiation by a provider requires careful consideration of a patient’s ability to maintain adherence, undergo frequent laboratory monitoring, change dosing regimen as recommended,


177

Beadles et al. Table 1. Matching Determinants of the Problem to Principles of the Intervention. Barriers to OAT initiation Patient level 1. Perceived stroke risk 2. Perceived adverse event risk 3. Difficulty with access 4. Difficulty with adherence Provider level 1. Perceived risk/benefit ratio 2. Lack of guideline awareness 3. Discrepancy between believed and actual practice patterns 4. Guidelines inappropriate for patient 5. Perceive patient will be nonadherent Health care system level 1. Access to laboratory facilities 2. Access to provider 3. Follow-up of out of range lab values

Principles of PCMH that address barriers 1. Personal relationship with physician 2. Personal relationship with physician 3. Enhanced access/coordinated care 4. Personal relationship with physician/wholeperson orientation 1. Coordinated care/commitment to quality improvement 2. Coordinated care/commitment to quality improvement 3. Commitment to quality improvement (audit-feedback) 4. Coordinated care/commitment to quality improvement 5. Personal relationship with physician/wholeperson orientation 1. Enhanced access/coordinated care 2. Enhanced access 3. Enhanced access/coordinated care/ commitment to quality improvement

Note. PCMH = patient-centered medical home; OAT = oral anticoagulation therapy.

and continued coordination of care with other providers. These challenges, along with the aforementioned barriers to OAT initiation, have the potential to be addressed by PCMH principles. A PCMH provider following these principles is committed to evidence-based high-quality care and has developed a sustained therapeutic relationship with patients. In a practice guided by PCMH principles, patients have greater access to filling prescriptions, laboratory monitoring, follow-up care, medication management, and care coordination with specialists. We posit that patients receiving care from this type of provider in this type of practice setting are more likely to begin OAT. Our conceptual framework very simply couples principles of the intervention, the PCMH, with determinants of the problem, suboptimal OAT initiation among incident AF patients. In this case, barriers at the level of patients, providers, and the health care system comprise the determinants of the problem. The specific barriers and PCMH principles that address them are summarized in Table 1. This conceptual framework is analogous to intervention mapping, a framework for health education intervention development that couples program objectives addressing specific problems with specific theory-based intervention methods and practical strategies while


178


providing specific guidance in designing and organizing a program, specifying adoption/implementation plans, and generating program evaluation plans (Bartholomew, Parcel, & Kok, 1998). Intervention mapping has been successfully applied in designing interventions to address complex health problems (Belansky et al., 2013; Cherrington et al., 2012). The NCQA has created an accreditation process designed to facilitate high-level performance on PCMH principles. Currently, it is the most widely used formal evaluation program for the PCMH. Therefore, we examined whether receiving care from an NCQA-accredited PCMH is associated with increased OAT initiation compared with practices that do not have PCMH accreditation.

Method Data Source We used data from January 1, 2006, to December 31, 2010, from the North Carolina State Health Plan (NCSHP), a large self-funded insurance plan for the study. The NCSHP, administered by Blue Cross and Blue Shield of North Carolina, includes almost 700,000 state employees, teachers, retirees, and their dependents at any given time and approximately 1 million individuals are included in the 5-year study window. Approximately 10% of enrollees are retired non-Medicare participants, and 16% are retired Medicare beneficiaries. This claims structured database contains inpatient, outpatient, and pharmacy records. Enrollee descriptors include unique encrypted member identification numbers, basic demographic information including age, gender, county, and zip code of primary residence. Records include information about diagnoses, procedures, providers, charges, and payments. The database also contains physician-level characteristics, which include provider zip code, type of provider, and provider specialty if applicable. We also linked individual and facility counties with additional variables from the Area Health Resource File (Health Resources and Services Administration, 2012), which allowed us to consider patient, provider, and county characteristics concurrently.

Study Design We created three cohorts of patients who varied by thromboembolism risk levels and strength of OAT indications: (a) new onset AF patients, (b) perceived very high-risk controls (mechanical heart valve or significant venous embolism), and (c) perceived low-risk controls (paroxysmal AF). For all three cohorts, patients needed to be continuously enrolled in the NCSHP for a minimum of 6 months prior to and 6 months following the qualifying index claim. Individuals with a prescription claim for warfarin more than 30 days prior to an index claim for any of the three cohorts were excluded because of a high probability of representing prevalent rather than incident conditions. Eligibility criteria for each cohort are described below.


179

Beadles et al.

New Onset AF Patients. We used either one inpatient diagnosis or two outpatient diagnoses within 12 months (International Classification of Disease, 9th edition–Clinical Modification [ICD-9-CM] code 427.31) to identify individuals with AF. We designated the first outpatient AF claim or hospital admission as the date of entry into the cohort. The American College of Cardiology, The American College of Chest Physicians, and the American Heart Association endorse the use of a risk-based score to identify individuals who will benefit from receiving OAT. CHADS2 is a commonly employed scoring system (Gage et al., 2001). CHADS2 scores are readily generated using claims data. Individuals receive 1 point for congestive heart failure (ICD-9-CM 425.4, 428.x, 429.4), hypertension (ICD-9-CM 362.11, 401.x-405.xx, 437.2), age > 75 years, and diabetes mellitus (ICD-9-CM 250.xx 357.2, 362.0x, 366.41); they receive 2 points for any prior stroke or stroke symptoms (325, 362.3x, 415.1x, 433.xx-434.xx, 435.x, 436, 437.1, 438, 444.xx, 451.xx, 452, 453.x). Based on current recommendations (Goldstein et al., 2011), we included all individuals meeting criteria for incident AF with a CHADS2 score ≥2. To increase the precision of our estimates, we used ICD9-CM codes to exclude individuals with ≥1 relative contraindication to OAT from the incident AF or intervention cohort (e.g., Prior: hemorrhagic stroke, gastrointestinal or other major bleed, fall risk—table available from lead author on request). Finally, to reduce the probability of including individuals with prevalent, rather than incident, AF, we excluded individuals with any AF-related claim in the 6 months preceding the index claim. Perceived Very High-Risk Controls. This cohort includes individuals with at least one inpatient diagnosis or two outpatient diagnoses within 12 months indicating a mechanical heart valve or significant thromboembolism by ICD-9-CM codes (table available from lead author on request). This cohort has an extremely high thromboembolism risk, and in ideal settings, all or nearly all receive OAT. The CHADS2 score was not applied to this cohort, as it is only validated in patients with AF. Finally, to reduce the probability of including individuals with prevalent conditions rather than incident mechanical heart valves or significant thromboembolism, we excluded individuals with any condition-related claim in the 6 months preceding the index claim. Perceived Low-Risk Controls. Individuals with at least one inpatient diagnosis or two outpatient diagnoses within 12 months for paroxysmal AF (ICD-9-CM code 427.21) were included in the guideline negative cohort. This cohort has equivalent thromboembolism risk as incident AF but is often perceived as having a lower risk. In ideal settings, this group has similar OAT initiation as the incident AF cohort, but practically is often substantially lower. Again we excluded individuals with one or more relative contraindications to receiving OAT to create cohorts similar in bleeding risk that differed only in their risk of acute ischemic stroke.

Measures The dependent variable was OAT initiation, defined by a prescription claim for Warfarin within 30 days from AF date of onset. The exposure of interest was receipt of


180


care in an NCQA-accredited PCMH during the 30 days prior to the index event. During 2006 to 2007, the NCQA PCMH accreditation process was an earlier version known as Physician Practice Connections (PPC), which emphasized the use of health information systematically to enhance the quality of patient care. In our sample, almost every PPC practice became a PCMH when they were reevaluated in or after 2008. Therefore, we include PPC practices that later became PCMHs at the time of their PPC accreditation. Control variables, identified a priori, were created at the patient, county, and provider levels. To mitigate potential confounding, all control variables were measured in the baseline period or immediately prior to the index AF claim. Patient-level control variables include age, gender, Charlson Comorbidity Index, CHADS2 score, rurality/urbanicity of residence, and number of outpatient visits to a primary care provider in the 90 days prior to the index claim. Charlson Comorbidity Index was categorized as 0, 1 to 2, and 3 or more. CHADS2 score was categorized as 2, 3 to 4, and ≥5. Rurality/urbanicity of residence was categorized into rural, micropolitan, or metropolitan as defined by the Area Health Resource File (Health Resources and Services Administration, 2012). County-level control variables included race and ethnicity demographics, unemployment rate, percentage of persons below poverty line, median household income, a rolling 3-year average number of deaths from cerebrovascular disease, the county’s Health Professional Shortage Area status (yes, partial, no), number of general practitioners, and number of cardiovascular subspecialists in county, which were extracted from the Area Health Resource File (Health Resources and Services Administration, 2012). Provider-level control variables included rurality/urbanicity of practice location and a binary variable indicating participation with the North Carolina Medicaid’s medical home program (Steiner et al., 2008). This program, an enhanced primary care case-management model, has driven improvements in quality of care for several diseases (e.g., asthma, diabetes) in the Medicaid population, and many participating practices became NCQA-accredited PCMHs. Although we investigated a privately insured population, we desired to control for any potential spillover effects.

Propensity Score Implementation To reduce possible confounding with PCMH exposure, we used propensity scores to balance the two groups on pretreatment covariates. We estimated, conditional on baseline covariates, an individual’s probability of receiving care in a PCMH during the 30 days prior to the index event. In our propensity score models, we included variables that, based on prior theory or evidence, were associated with OAT initiation or PCMH status (Brookhart et al., 2006). To estimate the propensity score, we employed generalized boosted regression using the “twang” package in R (Version 2.15.1). This package uses an iterative algorithm to construct multiple classification and regression trees, leading to selection of a model that optimizes covariate balance based on the average standardized absolute mean difference across covariates (McCaffrey, Ridgeway, & Morral, 2004). We then used stabilized inverse-probability-of-treatment weights


181

Beadles et al.

(IPTW) to estimate the average treatment effect (Robins, Hernan & Brumback, 2000). Comparison of standardized difference between PCMH and non-PCMH exposure groups before and after IPTW was performed to assess adequacy of exchangeability on observed pretreatment covariates. A priori, we decided that a standardized difference ≤10% would be the threshold for adequate balance between exposure groups (Austin, 2009).

Statistical Analysis To estimate the association between PCMH exposure and OAT initiation within each cohort, we utilized a generalized estimating equation model with a Poisson distribution, log link, and exchangeable within-group correlation structure. Robust standard errors were included to correct for under dispersion of the binary dependent variable in the context of a count distribution. We grouped at the practice level using practice identification fields in the database. This model accounted for within-practice correlation of outcomes among the accredited PCMH and non-PCMH practices across separate individuals. We estimated average marginal effects (absolute risk differences) for ease of interpretation. All models were estimated in Stata 11 (StataCorp, College Station, TX).

Sensitivity Analysis We first examined variations in the definition of OAT initiation. To test and compare with our OAT initiation definition based on pharmacy claims, we created a second nonpharmacy claims definition of OAT initiation. The second definition used ICD-9 CM codes for anticoagulation management (V58.61) and procedure codes for corresponding blood tests (CPT codes: 99363, 99364, 85610, 85730, 85732) within 30 days of index event. A third definition of OAT initiation was a combination of the first two: positive if either of the first two definitions were positive. Next, we examined variations in the definition of PCMH exposure based on the percentage of all primary care visits in the 30 days prior to index event that were to a PCMH. We used 30% and 50% thresholds for these two additional definitions of PCMH exposure (Mehrotra, Adams, Thomas, & McGlynn, 2010).

Results Descriptive Baseline characteristics of the AF (n = 4,424), perceived very high-risk (n = 6,530), and perceived low-risk cohorts (n = 618) before IPTW are presented in Table 2. Standardized differences between PCMH users and non-PCMH users for each cohort are also shown. Baseline characteristics between PCMH and non-PCMH groups for each cohort following IPTW are shown in Table 3. Standardized difference was below the suggested threshold of 10% for all variables in both the AF and the perceived very


182


75.2 1.9 2.9 0.41 4.6 75.0% 16.4% 26.3% 2.7% 71.0% 95.9 27 141.2 9.0% 22.4% 25.5% 22.1% 21.0%

76.3 1.9 2.9 0.45 3.3 63.4% 24.2% 36.4% 8.6% 54.9% 61.3 16 102.9 19.6% 25.7% 20.7% 18.6% 15.4% 45.0% 44.5% 38.9% 30.0%

28.8%

9.0% 2.7% 1.2% 8.3% 27.0% 7.4%

61.0 1.4 N/A 0.40 5.7 79.4% 14.5% 21.7% 3.1% 75.7% 102.2 28 146.3 11.7% 17.7% 22.7% 26.3% 21.7%

65.1 1.7 N/A 0.41 4.9 63.9% 25.6% 36.1% 8.3% 55.6% 66.1 17 108.5 16.5% 21.5% 20.5% 20.8% 20.7% 42.7% 42.1% 32.3% 15.7%

38.5%

26.3% 14.1% N/A 2.3% 9.9% 12.9%

PCMH Non-PCMH Standardized PCMH Non-PCMH Standardized (n = 466) (n = 3,957) difference, % (n = 863) (n = 5,667) difference, %

Perceived very high-risk cohort

69.6 1.8 2.2 0.55 4.37 75.3% 19.4% 22.6% 5.4% 72.0% 96.3 27 140.6 8.6% 16.1% 33.3% 23.7% 18.3%

67.7 1.7 2.0 0.63 3.43 65.5% 22.5% 36.0% 8.1% 56.0% 68 18 109.6 13.9% 26.3% 19.8% 23.0% 17.1%

36.9% 37.2% 31.6% 18.8%

33.1%

14.4% 2.0% 15.9% 16.9% 18.0% 6.7%

PCMH Non-PCMH Standardized (n = 69) (n = 549) difference, %

Perceived low-risk cohort

Note. HPSA = Health Professional Shortage Area; GPs = number of office-based general practitioners in county; CV specialist = number of office-based cardiovascular specialists in county; Dth_CV_DZ = 3-year average number of cerebrovascular (stroke) deaths in county; N/A = not applicable; PCMH = patientcentered medical home.

Age Charlson CHADS2 score Male Pre-index visits Metropolitan Micropolitan HPSA—No HPSA—Yes HPSA—Partial No. of GPs No. of CV specialist Dth_CV_DZ 2006 2007 2008 2009 2010

Covariate

Atrial fibrillation cohort

Table 2. Baseline Characteristics (Unadjusted).

183


63.2 1.5 N/A 39.1% 4.6 70.8% 22.5% 31.4% 6.4% 62.2% 77.7 21 118.9 12.6% 20.7% 20.6% 24.3% 21.9%

64.4 1.6 N/A 41.0% 5.0 66.3% 23.8% 33.9% 7.5% 58.5% 71.6 19.1 114.4 15.8% 21.0% 20.9% 21.5% 20.8% 7% 7% 3.9% 8%

6.4%

7.1% 7.8% N/A 4.4% 5.3% 3.1%

68.3 1.7 2.1 59.0% 3.6 63.8% 27.4% 30.1% 7.7% 62.8% 79.7 22.4 124.0 9.4% 19.9% 32.2% 23.9% 15.1%

PCMH (n = 69) 67.8 1.7 2.0 63.1% 3.6 67.1% 21.7% 35.2% 7.7% 57.2% 71.2 19.2 113.3 13.3% 25.7% 20.5% 23.6% 17.0%

10.7% 12.7% 10.5% 7.7%

10.7%

3.6% 0.5% 7.5% 7.1% 0.1% 13.2%

Non-PCMH Standardized (n = 549) difference, %


Note. HPSA = Health Professional Shortage Area; GPs = number of office-based general practitioners in county; CV specialist = number of office-based cardiovascular specialists in county; Dth_CV_DZ = 3-year average number of cerebrovascular (stroke) deaths in county; N/A = not applicable; PCMH = patientcentered medical home.

6.2% 6.5% 3.4% 6.3%

6.4%

0.7% 1.2% 3.3% 4.8% 1.4% 6.7%

76.3 1.9 2.9 42.0% 3.5 66.9% 24.1% 32.6% 6.9% 60.4% 70.3 19 111.1 15.0% 26.9% 20.9% 19.9% 17.3%

Age Charlson CHADS2 score Male Pre-index visits Metropolitan Micropolitan HPSA—No HPSA—Yes HPSA—Partial No. of GPs No. of CV specialist Dth_CV_DZ 2006 2007 2008 2009 2010

76.2 1.9 2.9 45.1% 3.5 64.9% 23.1% 35.3% 8.0% 56.7% 65.5 17.5 107.7 18.4% 25.4% 21.1% 19.1% 16.0%

PCMH Non-PCMH Standardized PCMH Non-PCMH Standardized (n = 466) (n = 3,957) difference, % (n = 863) (n = 5,667) difference, %


Covariate


Table 3. Baseline Characteristics Following Inverse Probability of Treatment Weighting For Each Cohort.

184


high-risk cohorts, indicating sufficient baseline covariate balance by PCMH status. However, the perceived low-risk cohort contained multiple baseline covariates that remained above the 10% threshold in absolute standardized mean difference following optimized generation of propensity scores and IPTW (Austin, 2009). This cohort was also quite small in comparison with the other cohorts. Thus, although we report data for this cohort, we urge caution in interpretation of results because of the small sample size and failure to balance on observed covariates. The AF cohort had a greater mean age and Charlson Comorbidity Index score than the perceived very high-risk cohort. The number of pre-index outpatient visits was highest in the perceived very high-risk cohort. The distribution of individuals by geographic residence, health professional shortage areas, generalist and specialist availability, and year of entry into the cohort was comparable among the cohorts.

Multivariable Results Unadjusted bivariate results and multivariable adjusted results for exposure to an NCQA PCMH are detailed in Table 4. Model 1 for each cohort adjusts for patient characteristics, location of residence, use of health care in the baseline period, and year of cohort entry. Model 2 adjusts for the aforementioned characteristics and additional practice characteristics. Unadjusted and adjusted estimates for Model 1 and Model 2 of the effect of PCMH use on OAT initiation are comparable for both the AF (6% to 7%) and perceived very high-risk (14%) cohorts but not for the perceived low-risk cohort (6.5% vs. 10% to 12%). In other words, patients exposed to an NCQAaccredited PCMH were 6 to 7 percentage points more likely to initiate OAT than patients not exposed to an NCQA-accredited PCMH. In addition to our primary results concerning PCMH exposure, we note several additional salient findings from our fully adjusted models. Men were more likely to initiate OAT in the AF cohort, but equally as likely to initiate OAT in the perceived very high-risk cohort. More than two outpatient visits in the 90 days preceding the index event was also associated with an increased likelihood of OAT initiation. We found a strong positive time association with initiating OAT in the AF cohort that was not observed (as expected) in the perceived very high-risk cohort. Finally practice participation in the North Carolina Medicaid medical home program (CCNC practice) was weakly suggestive of a positive association with OAT initiation in both AF and perceived very high-risk cohorts.

Sensitivity Analysis In the sensitivity analyses (table available from lead author on request), we implemented the fully adjusted models (Model 2). Generally, the average marginal effect of PCMH on OAT initiation was robust to alternative definitions of OAT initiation and PCMH status. The magnitude of the association decreased with increasing sensitivity and decreasing specificity of the OAT initiation measure, but in all cases remained


185


6.78%***

4,423

7.04%*** −3.89% 3.15% −1.68% 6.67%*** −0.55% −1.05% −2.02% −8.63%*** −4.14%** −5.44%* 4.17%* 2.6% 1.53% 5.19%*** 10.82%*** 13.35%*** 15.77%*** 12.86%***

Model 1 6.25%*** −1.75% 4.29%* −1.41% 6.70%*** −0.02% −0.82% -1.43% −8.54%*** −4.26%** −5.31%* −1.59% −0.22% 1.92% 4.88%*** 11.04%*** 13.43%*** 16.94%*** 13.10%*** 10.98%*** 4.3% 3.96%* 4,423

Model 2


Unadjusted estimate

Note. CCNC = Community Care North Carolina (Medicaid Medical Home). *p < .05. **p < .01. ***p < .001.

Medical home user Age 18-40 (ref. group 41-60) Age 61-70 Age 71+ Male Charlson 0 (ref. group Charlson 1) Charlson 2 Charlson 3-4 Charlson 5-12 CHADS2 3-4 (ref. group CHADS2 2) CHADS2 5-6 Rural (ref. group Metropolitan) Micropolitan Zero prior visits (ref. group 1 prior visit) 2+ prior visits Year = 2007 (ref. group 2006) Year = 2008 Year = 2009 Year = 2010 Rural practice (ref. group Metropolitan) Micropolitan practice CCNC Practice n

Covariate 14.4%***

6,530

14.71%*** −1.67% −2.04% −7.69%*** 1.63% 4.97%** −6.07%*** −6.09%*** −14.59%*** 0.74% −1.05% 4.23%** 2.05% −6.0%*** 6.73%*** 0.66% 0.66% −2.42% −0.91%

Model 1 13.95%*** −2.21% −2.01% −8.19%*** 1.80% 5.03%** −5.79%** −5.87%** −14.07%*** 1.36% −0.96% −1.52% −3.73%** −5.15%** 7.17%*** 1.21% 0.52% −2.25% −0.84% 12.4%*** 9.93%*** 3.12% 6,530

Model 2

Adjusted models


Unadjusted estimate

Table 4. Unadjusted Difference in Percent OAT Initiation and Marginal Effect of Explanatory Variables. 6.53%

610

11.67%** −14.51%* 12.91%*** 5.17% 15.48%*** 0.13% −4.84% −5.85% −11.04%** −5.06% 4.73% −4.36% 4.83% 7.28% 12.36%** 18.04%*** −1.4% 6.45% 10.05%*

Model 1

9.98%** −14.7%** 15.1%*** 5.53% 15.33%** −0.9% −4.5% −5.71% −12.86%** −7.13% −7.49% −3.43% 4.64% 7.84% 12.14%** 18.12%*** −1.81% 7.18% 9.22% −8.18% −0.87% −5.78% 610

Model 2


Unadjusted estimate

186


positive. Similarly varying the definition of PCMH user resulted in a smaller magnitude positive association of PCMH with OAT initiation.

Discussion Despite evidence-based guidelines, OAT initiation rates are low for patients with AF. Along with the rise of Accountable Care Organizations, PCMHs hold potential for improving OAT initiation for AF patients. We found a modest increase in OAT initiation associated with PCMH exposure for patients with incident AF, and a somewhat larger increase for the perceived very high-risk comparison cohort. We hypothesize several possible explanations for these findings. First, there are numerous provider and healthcare system barriers to OAT initiation (Beyth et al., 1996; Brodsky et al., 1996; Bungard et al., 2003; Choudhry et al., 2006, Lane & Lip, 2008).The PCMH may only partially mitigate these barriers for the AF cohort, or the effect of some barriers may be less substantial for the younger perceived very high-risk cohort. For example, patient-level barriers addressed (perhaps only partially) by enhanced access and coordinated care principles of the PCMH model may be critical for the AF cohort but less influential for the perceived very high-risk cohort. Similarly, the perceived very high-risk cohort may value personal relationship with their physician and a whole-person orientation more than the AF cohort. Second, our results may reflect greater PCMH OAT use regardless of indication. PCMHs may be more likely to prescribe OAT because of geographic location, previous practice patterns, or patient preference. PCMH status may proxy for practices inclined toward OAT use, even when guidelines recommend against it (e.g., AF/paroxysmal AF with CHADS2 < 1). Given the breadth of scope and principles of the PCMH model (i.e., whole-person orientation, commitment to continuous evaluation, and improvement of care quality and patient safety), we believe this explanation unlikely. Finally, practices with PCMH accreditation may reflect higher quality practices or practices that are more predisposed to quality improvement initiatives at baseline. While paroxysmal AF is equivalent to AF in thromboembolism risk, it is often misperceived as a lower risk. Greater attention to guidelines for paroxysmal AF and anticoagulation could result in more OAT initiation in this cohort. Thus, the positive association between PCMHs and paroxysmal AF could be consistent with guideline appropriate care. Ultimately, we believe our findings represent a mixture of the first and third explanations. PCMH status does appear associated with increased OAT initiation in eligible AF patients. Similar larger results in the perceived very high-risk cohort may result from PCMH status reflecting higher baseline quality or from partial implementation of all PCMH model components (i.e., enhanced access); providing less benefit to older frailer patients (AF cohort) than younger more robust patients (perceived very highrisk cohort). The positive time trend associated with OAT initiation in only the AF cohort also merits discussion. We believe this finding represents, in part, a secular trend toward increasing OAT initiation in eligible AF patients. However, adequate rationales for this trend are likely complex, multidimensional and speculative.


187

Beadles et al.

There are several limitations of our study. First, because we used claims data, we may have misclassified OAT initiation, especially with the popularity of low-cost generic OAT that could be purchased without billing the insurer (Lauffenburger et al., 2013). We do not have reason to suspect a differential error in pharmacy claims for warfarin across the three cohorts regarding OAT initiation. Moreover, because we control for time in our models, we would have expected this trend, decreasing OAT initiation in later years, to appear in the year indicators. However, we find the opposite trend in our year indicators, suggesting this effect is small or nonexistent. Importantly, our sensitivity analyses suggested that findings were robust to alternative definitions of OAT initiation. Second, though we controlled for county-level indicators of race and socioeconomic status, we could not do so at the patient level. Third, we do not control for levels (1-4) of PCMH accreditation attained by practices or changes in levels across time. This may introduce heterogeneity into our results. However, by our review, a substantial majority of PCMHs were accredited at the 3 or 4 level and maintained their initial accreditation level for the duration of the study. Finally, most members of our cohorts had employee-sponsored group health insurance; results may not generalize to publicly insured populations. Notably, our cohorts included a large proportion of retired Medicare enrollees utilizing both their Medicare and NCSHP benefits, which serves to enhance generalizability to Medicare beneficiaries. Despite these limitations, our findings demonstrate that PCMH status is associated with greater OAT initiation among eligible AF patients, a key factor in reducing stroke risk and its sizeable morbidity/mortality burden. Our findings have several implications for future research, policy development, and practice. OAT-associated stroke risk reduction accrues cumulatively with time; therefore, future research should evaluate the association between PCMHs and duration of OAT. Additionally, OAT quality across settings (e.g., traditional community practices, specialized anticoagulation clinics) has received considerable attention. Future research should correlate PCMH administered OAT with quality of achieved anticoagulation and stroke as outcomes. From a policy perspective, our findings suggest PCMHs may improve processes of care, which may support enhanced payment structures to PCMHs. Similarly, our findings inform the development of Accountable Care Organizations (ACOs). Specifically, incorporating PCMHs or their principles within an ACO may greatly facilitate the delivery of high-quality care. This consideration for ACOs is particularly relevant in chronic conditions and their dynamic management plans, which require tight synchronicity of care processes. Finally, for practice, these findings suggest that improvements in processes of care in outpatient primary care are associated with NCQA accreditation, which emphasizes achieving the principles of the PCMH model through increased access and the utilization of health information technology. Declaration of Conflicting Interests The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of AHRQ, UNC, the Department of Veterans Affairs or the United States government.


188


Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: AHRQ: 1T32 HS019442; VA Office of Academic Affiliations Postdoctoral fellowship (TPP 21-023); and VA Senior Research Career Scientist Award (RCS 91-408).

References American College of Chest Physicians. (1995). 4th American College of Chest Physicians Consensus Conference on Antithrombotic Therapy. Tucson, Arizona, April 1995. Proceedings. Chest, 108(Suppl. 4), 225S-522S. Austin, P. C. (2009). Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Statistics in Medicine, 28, 3083-3107. Bartholomew, L. K., Parcel, G. S., & Kok, G. (1998). Intervention mapping: A process for developing theory- and evidence-based health education programs. Health Education & Behavior, 25, 545-563. Belansky, E. S., Cutforth, N., Chavez, R., Crane, L. A., Waters, E., & Marshall, J. A. (2013). Adapted intervention mapping: A strategic planning process for increasing physical activity and healthy eating opportunities in schools via environment and policy change. Journal of School Health, 83, 194-205. Beyth, R. J., Antani, M. R., Covinsky, K. E., Miller, D. G., Chren, M. M., Quinn, L. M., & Landefeld, C. S. (1996). Why isn’t warfarin prescribed to patients with nonrheumatic atrial fibrillation? Journal of General Internal Medicine, 11, 721-728. Birman-Deych, E., Radford, M. J., Nilasena, D. S., & Gage, B. F. (2006). Use and effectiveness of warfarin in Medicare beneficiaries with atrial fibrillation. Stroke, 37, 1070-1074. Brodsky, M. A., Chun, J. G., Podrid, P. J., Douban, S., Allen, B. J., & Cygan, R. (1996). Regional attitudes of generalists, specialists, and subspecialists about management of atrial fibrillation. Archives of Internal Medicine, 156, 2553-2562. Brookhart, M. A., Schneeweiss, S., Rothman, K. J., Glynn, R. J., Avorn, J., & Sturmer, T. (2006). Variable selection for propensity score models. American Journal of Epidemiology, 163, 1149-1156. Buckingham, T. A., & Hatala, R. (2002). Anticoagulants for atrial fibrillation: Why is the treatment rate so low? Clinical Cardiology, 25, 447-454. Bungard, T. J., Ghali, W. A., McAlister, F. A., Buchan, A. M., Cave, A. J., Hamilton, P. G., . . . Tsuyuki, R.T. (2003). The relative importance of barriers to the prescription of warfarin for nonvalvular atrial fibrillation. Canadian Journal of Cardiology, 19, 280-284. Bungard, T. J., Ghali, W. A., Teo, K. K., McAlister, F. A., & Tsuyuki, R. T. (2000). Why do patients with atrial fibrillation not receive warfarin? Archives of Internal Medicine, 160, 41-46. Chang, H. J., Bell, J. R., Deroo, D. B., Kirk, J. W., & Wasson, J. H. (1990). Physician variation in anticoagulating patients with atrial fibrillation. Dartmouth Primary Care COOP Project. Archives of Internal Medicine, 150, 83-86. Cherrington, A., Martin, M. Y., Hayes, M., Halanych, J. H., Wright, M. A., Appel, S. J., . . . Safford, M. (2012). Intervention mapping as a guide for the development of a diabetes peer support intervention in rural Alabama. Preventing Chronic Disease, 9, E36.


189

Beadles et al.

Choudhry, N. K., Soumerai, S. B., Normand, S. L., Ross-Degnan, D., Laupacis, A., & Anderson, G. M. (2006). Warfarin prescribing in atrial fibrillation: The impact of physician, patient, and hospital characteristics. American Journal of Medicine, 119, 607-615. Dantas, G. C., Thompson, B. V., Manson, J. A., Tracy, C. S., & Upshur, R. E. (2004). Patients’ perspectives on taking warfarin: Qualitative study in family practice. BMC Family Practice, 5, 15. DePalma, J. (2007). Evidence to support medical home concept for children with special health care needs. Home Health Care Management Practice, 19, 473-475. Dulli, D. A., Stanko, H., & Levine, R. L. (2003). Atrial fibrillation is associated with severe acute ischemic stroke. Neuroepidemiology, 22, 118-123. Gage, B. F., Boechler, M., Doggette, A. L., Fortune, G., Flaker, G. C., Rich, M. W., & Radford, M. J. (2000). Adverse outcomes and predictors of underuse of antithrombotic therapy in Medicare beneficiaries with chronic atrial fibrillation. Stroke, 31, 822-827. Gage, B. F., Waterman, A. D., Shannon, W., Boechler, M., Rich, M. W., & Radford, M. J. (2001). Validation of clinical classification schemes for predicting stroke: Results from the National Registry of Atrial Fibrillation. Journal of the American Medical Association, 285, 2864-2870. Go, A. S., Hylek, E. M., Borowsky, L. H., Phillips, K. A., Selby, J. V., & Singer, D. E. (1999). Warfarin use among ambulatory patients with nonvalvular atrial fibrillation: The anticoagulation and risk factors in atrial fibrillation (ATRIA) study. Annals of Internal Medicine, 131, 927-934. Go, A. S., Mozaffarian, D., Roger, V. L., Benjamin, E. J., Berry, J. D., Borden, W. B., . . . Turner, M. B. (2013). Heart disease and stroke statistics—2013 update: A report from the American Heart Association. Circulation, 127(1), e6-e245. Goldstein, L. B., Adams, R., Alberts, M. J., Appel, L. J., Brass, L. M., Bushnell, C. D., . . . Sacco, R. L. (2006). Primary prevention of ischemic stroke: A guideline from the American Heart Association/American Stroke Association Stroke Council. Circulation, 113, e873-923. Goldstein, L. B., Bushnell, C. D., Adams, R. J., Appel, L. J., Braun, L. T., Chaturvedi, S., . . . Person, T. A. (2011). Guidelines for the primary prevention of stroke: A guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke, 42, 517-584. Harley, C. R., Riedel, A. A., Hauch, O., Nelson, M., Wygant, G., & Reynolds, M. (2005). Anticoagulation therapy in patients with chronic atrial fibrillation: A retrospective claims data analysis. Current Medical Research and Opinion, 21, 215-222. Hart, R. G. (1999). Warfarin in atrial fibrillation: underused in the elderly, often inappropriately used in the young. Heart, 82, 539-540. Hart, R. G., & Bailey, R. D. (2002). An assessment of guidelines for prevention of ischemic stroke. Neurology, 59, 977-982. Health Resources and Services Administration. (2012). Area Health Resource File 2011-2012. Rockville, MD: Author. Homer, C. J., Klatka, K., Romm, D., Kuhlthau, K., Bloom, S., Newacheck, P., . . . Perrin, J. M. (2008). A review of the evidence for the medical home for children with special health care needs. Pediatrics, 122, e922-937. Jackson, G. L., Powers, B. J., Chatterjee, R., Bettger, J. P., Kemper, A. R., Hasselblad, V., . . . Williams J. W. (2013). The patient-centered medical home. Annals of Internal Medicine, 158, 169-178.


190


Johnston, J. A., Cluxton, R. J., Jr., Heaton, P. C., Guo, J. J., Moomaw, C. J., & Eckman, M. H. (2003). Predictors of warfarin use among Ohio Medicaid patients with new-onset nonvalvular atrial fibrillation. Archives of Internal Medicine, 163, 1705-1710. Jorgensen, H. S., Nakayama, H., Reith, J., Raaschou, H. O., & Olsen, T. S. (1996). Acute stroke with atrial fibrillation. The Copenhagen Stroke Study. Stroke, 27, 1765-1769. Kutner, M., Nixon, G., & Silverstone, F. (1991). Physicians’ attitudes toward oral anticoagulants and antiplatelet agents for stroke prevention in elderly patients with atrial fibrillation. Archives of Internal Medicine, 151, 1950-1953. Lane, D. A., & Lip, G. Y. (2008). Barriers to anticoagulation in patients with atrial fibrillation: Changing physician-related factors. Stroke, 39, 7-9. Lauffenburger, J. C., Balasubramanian, A., Farley, J. F., Critchlow, C. W., O’Malley, C. D., Roth, M. T., . . . Brookhart, M. A. (2013). Completeness of prescription information in U.S. commercial claims databases. Pharmacoepidemiology and Drug Safety, 22, 899-906. Lin, H. J., Wolf, P. A., Kelly-Hayes, M., Beiser, A. S., Kase, C. S., Benjamin, E. J., & D’Agostino, R. B. (1996). Stroke severity in atrial fibrillation. The Framingham Study. Stroke, 27, 1760-1764. McCaffrey, D. F., Ridgeway, G., & Morral, A. R. (2004). Propensity score estimation with boosted regression for evaluating causal effects in observational studies. Psychological Methods, 9, 403-425. McCrory, D. C., Matchar, D. B., Samsa, G., Sanders, L. L., & Pritchett, E. L. (1995). Physician attitudes about anticoagulation for nonvalvular atrial fibrillation in the elderly. Archives of Internal Medicine, 155, 277-281. Mehrotra, A., Adams, J. L., Thomas, J. W., & McGlynn, E. A. (2010). The effect of different attribution rules on individual physician cost profiles. Annals of Internal Medicine, 152, 649-654. National Committee on Quality Assurance. (2012). Patient-centered medical home. Retrieved from http://www.ncqa.org Ogilvie, I. M., Newton, N., Welner, S. A., Cowell, W., & Lip, G. Y. (2010). Underuse of oral anticoagulants in atrial fibrillation: A systematic review. American Journal of Medicine, 123, 638-645. Peikes, D., Zutshi, A., Genevro, J. L., Parchman, M. L., & Meyers, D. S. (2012). Early evaluations of the medical home: Building on a promising start. American Journal of Managed Care, 18, 105-116. Petersen, P., Boysen, G., Godtfredsen, J., Andersen, E. D., & Andersen, B. (1989). Placebocontrolled, randomised trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation. The Copenhagen AFASAK study. Lancet, 1, 175-179. Robert Graham Center. (2007, November). The patient centered medical home: History, seven core features evidence, and transformational change. Retrieved from http://www.aafp.org Robins, J. M., Hernan, M. A., & Brumback, B. (2000). Marginal structural models and causal inference in epidemiology. Epidemiology, 11, 550-560. Rodgers, H., Sudlow, M., Dobson, R., Kenny, R. A., & Thomson, R. G. (1997). Warfarin anticoagulation in primary care: A regional survey of present practice and clinicians’ views. British Journal of General Practice, 47, 309-310. Roger, V. L., Go, A. S., Lloyd-Jones, D. M., Benjamin, E. J., Berry, J. D., Borden, W. B., . . . Turner, M. B. (2012). Heart disease and stroke statistics–2012 update: A report from the American Heart Association. Circulation, 125(1), e2-e220.


191

Beadles et al.

Rosenthal, T. C. (2008). The medical home: Growing evidence to support a new approach to primary care. Journal of the American Board of Family Medicine, 21, 427-440. Steiner, B. D., Denham, A. C., Ashkin, E., Newton, W. P., Wroth, T., & Dobson, L. A., Jr. (2008). Community care of North Carolina: Improving care through community health networks. Annals of Family Medicine, 6, 361-367. Wang, T. J., Massaro, J. M., Levy, D., Vasan, R. S., Wolf, P. A., D’Agostino, R. B., . . . Benjamin, E. J. (2003). A risk score for predicting stroke or death in individuals with newonset atrial fibrillation in the community: The Framingham Heart Study. Journal of the American Medical Association, 290, 1049-1056. Wolf, P. A., Abbott, R. D., & Kannel, W. B. (1991). Atrial fibrillation as an independent risk factor for stroke: The Framingham Study. Stroke, 22, 983-988. Zimetbaum, P. J., Thosani, A., Yu, H. T., Xiong, Y., Lin, J., Kothawala, P., & Emons, M. (2010). Are atrial fibrillation patients receiving warfarin in accordance with stroke risk? American Journal of Medicine, 123, 446-453.


A non-experimental study of oral anticoagulation therapy initiation before and after national patient safety goals.

Complications associated with oral anticoagulation therapy.

Management of Venous Thromboembolism: Recent Advances in Oral Anticoagulation Therapy.

IOLs and anticoagulation therapy.

[Bleeding complications under oral anticoagulation].

[Bleeding complications under oral anticoagulation].

[Bleeding complications under oral anticoagulation].

Self-management of oral anticoagulation.

The Patient and Patient-Centered Medical Homes.

Implementing evidence-based patient and family education on oral anticoagulation therapy: a community-based participatory project.

Legal and professional implications of shared care: a case study in oral anticoagulation stroke prevention therapy.

Evaluation of oral anticoagulation therapy: rationale and design of the thrombEVAL study programme.

Anticoagulation drug therapy: a review.

Prevalence of oral anticoagulation in atrial fibrillation.

Continued anticoagulation in oral surgery procedures.

A Multilevel Analysis of Real-World Variations in Oral Anticoagulation Initiation for Atrial Fibrillation in Valencia, a European Region.

Guidelines on oral anticoagulation: second edition.

Structuring payments to patient-centered medical homes.

Patient Experience in Health Center Medical Homes.

Medical homes in Louisiana: the authors reply.

Dental procedures in patients treated with antiplatelet or oral anticoagulation therapy - an anonymous survey.

Patient Preferences for Oral Anticoagulation Therapy in Atrial Fibrillation: A Systematic Literature Review.

Patient-centered medical homes in Louisiana.

Pharmacists' roles in patient-centered medical homes.