The Veterans Health Administration (VHA) is the largest integrated US health system to implement the patient-centered medical home. The Patient Aligned Care Team (PACT) initiative (implemented 2010-2014) aims to achieve teambased care, improved access, and care management for more than 5 million primary care patients nationwide.

To describe PACT and evaluate interim changes in PACT-related care processes.

Data from the VHA Corporate Data Warehouse were obtained from April 2009 (pre- PACT) to September 2012. All patients assigned to a primary care provider (PCP) at all VHA facilities were included.

Nonparametric tests of trend across time points.

VHA increased primary care staff levels from April 2010 to December 2011 (2.3 to 3.0 staff per PCP full-time equivalent). In-person PCP visit rates slightly decreased from April 2009 to April 2012 (53 to 43 per 100 patients per calendar quarter; P <.01), while in-person nurse encounter rates remained steady. Large increases were seen in phone encounters (2.7 to 28.8 per 100 patients per quarter; P <.01), enhanced personal health record use (3% to 13% of patients enrolled), and electronic messaging to providers (0.01% to 2.3% of patients per quarter). Post hospitalization follow-up improved (6.6% to 61% of VA hospital discharges), but home telemonitoring (0.8% to 1.4% of patients) and group visits (0.2 to 0.65 per 100 patients per quarter; P <.01) grew slowly.

Thirty months into PACT, primary care staff levels and phone and electronic encounters have greatly increased; other changes have been positive but slower.

The Veterans’ Health Administration (VHA) is the largest integrated health system to implement the patient-centered medical home (PCMH) to date. This initiative (implemented 2010-2014) aims to achieve team-based care, improved access, and care management for more than 5 million primary care patients nationwide. Interim changes in care processes include:

		Decreased rate of in-person primary care provider visits and increase in telephone and Internet care.

		No significant increase in shared medical appointments.

		Slight improvement in appointment access and continuity, which started at high levels.

		Improved but still suboptimal post hospitalization follow-up.

The patient-centered medical home (PCMH) model is being implemented by a growing number of health organizations, with the goal of providing more comprehensive, coordinated, and patient-centered care. The major primary care medical societies have endorsed the PCMH as the desired model for primary care.

 Large health systems and primary care practice collaboratives are implementing the PCMH model, and the Centers for Medicare & Medicaid Services is funding PCMH demonstration and innovation projects in diverse clinical sites.

 Nonetheless, published evaluations of PCMH initiatives to date have been limited to single-practice or smaller groups of practices, and often focus on limited clinical conditions.

In 2010, the Veterans Health Administration (VHA) became the largest integrated US health system to begin implementing the PCMH model at all primary care clinics throughout its nationwide system. The VHA delivers primary care to more than 5 million veterans in 16.4 million encounters annually—either at 160 large hospital-based primary care facilities, most of which are in urban areas, or 783 Community- Based Outpatient Clinics, many of which are in rural areas. Thus, VHA is facing the challenge of substantially redesigning major systems of care in an extremely large system across diverse clinical and community settings. In addition, veterans served by VHA typically have more chronic physical and mental illnesses, and are more socioeconomically vulnerable, than patients who receive care outside VHA.

The VHA’s PCMH initiative, called Patient Aligned Care Teams (PACTs), builds upon foundations established in the 1990s, when VHA undertook a major transformation from loosely organized hospitals that provided mainly inpatient and specialty care into a regionally integrated system focused on outpatient primary care.

 As part of that process, large numbers of primary care providers (PCPs) and nurses were hired by VHA, and several key elements of the PCMH model (as defined by the National Committee for Quality Assurance) were implemented, such as comprehensive electronic medical records and performance measurement and improvement programs (including programs addressing clinical outcome goals such as glycemic control and blood pressure control for patients with diabetes and cancer screening; 

). Since that time, VHA has demonstrated better clinical quality of care and outcomes in many areas than have been reported in other parts of the healthcare sector, including Medicare.

However, in the 2000s the level of primary care staffing and resources remained steady, despite steady increases in numbers of primary care patients.

 In addition, VHA identified room for improvement in care continuity and coordination (eg, by decreasing the amount that VHA patients relied on providers not part of the Department of Veterans Affairs [VA] for acute care) and in patient-centeredness of care (eg, by providing care through telephone or electronic access when patients prefer it).

 Thus, the PACT initiative aims to enhance comprehensive and coordinated care, improve patient experience, and further improve health outcomes by increasing and reorganizing primary care staffing, and introducing several PCMH components that were not already in place.

The PACT initiative began in April 2010, and full implementation is anticipated to continue through 2014. Patient Aligned Care Teams also include a concurrent plan (and budget) for nationwide evaluation, including quantitative and qualitative data collection and analysis. In this study, we describe the design of PACT and the extent of structural changes made by facilities in response to the PACT plan to date. Then, we present an interim nationwide evaluation of observed changes in patient care processes related to PACT goals.

PACT DESIGN AND STRUCTURAL CHANGES TO DATE

To enhance continuity, staff are organized into teamlets that provide care to an assigned panel of about 1200 patients.

 A teamlet consists of 1 PCP, 1 registered nurse (RN) care manager, 1 licensed practical nurse (LPN) or medical assistant, and 1 administrative clerk. Teamlets are designed to optimize work flow by enabling each member to function at the top of their expertise. For example, PACT RNs are expected to manage care of patients with multiple chronic conditions through in-person and telephone encounters, and medical assistants are expected to provide preventive health screening. To facilitate communication and planning, teamlets are expected to hold regular huddles.

 The PACT clinical pharmacists assigned to a set of teamlets also manage patients with poorly controlled chronic illnesses through independent patient encounters.

To establish the teamlet model, the VHA mandated that facilities provide 3.0 full-time equivalents (FTEs) of primary care support staff per full-time PCP by the end of 2011, with dedicated funding to facilities for this. From April 2010 to December 2011, primary care support staff increased from 10,501 FTEs to 13,742 FTEs (

), corresponding to an increase from 2.3 FTEs to 3.0 FTEs support staff per PCP FTE nationwide. In addition, 76% of facilities reported holding daily teamlet huddles in July 2011 (up from 21% pre-PACT, Table 2).

Patient Aligned Care Teams aim to improve patient access through 3 methods. First, facilities are instructed to enact advanced access scheduling,

 including increased availability of same-day appointment slots. Second, facilities are asked to conduct more appointments via phone and by shared medical appointments. In response to this directive, 40% of facilities set aside hours during the clinical workday dedicated to scheduled phone visits by July 2011 (up from 14% pre-PACT; Table 2). It should also be noted that efforts were made to more accurately code and capture clinical encounters that occurred over the phone. Third, PACT aims to increase patient access to personal health data and providers via the Internet. This goal is accomplished through (1) an enhanced Internet-based personal health record (www.MyHealtheVet.gov), which allows patients to manage prescriptions and view test results and appointments; and (2) a secure messaging website, which allows patients to send electronic messages to their teamlet. To use either service, VA patients must complete an in-person registration and identity check. The PACT initiative directed each parent facility to add a personal health record/secure messaging coordinator, who encourages and facilitates patient and staff use of these technologies (Table 2).

The VHA hired 1271 primary care RN care managers between January 2010 and December 2011 (Table 2). The role of the RN care manager is envisioned to include chronic illness management, panel management of high-risk patients, and facilitation of patient care transitions. Among these responsibilities, post hospitalization follow-up was chosen as an early focus. Some RN care managers are assigned to telehealth monitoring of patients with chronic conditions. Telehealth consoles transmit patients’ health data (eg, blood glucose, blood pressure) from the patient’s home over a phone or Internet connection to the VA. Telehealth nurses at the patient’s home facility manage clinically relevant changes over the phone.

Also included in PACT is funding for a full-time health promotion/ disease prevention specialist at each facility to oversee screening and counseling programs related to healthy behaviors, such as healthy eating and tobacco cessation. Health Behavior Coordinators (typically clinical psychologists or social workers) at each facility support this program by training clinical staff in evidence based behavior change counseling techniques. Almost all facilities now have hired these 2 staff members (Table 2).

Providing accessible and broad care for mental health conditions and coordinating that care with other aspects of primary care are essential to providing comprehensive primary care in the VHA, as approximately 15% of veterans have a mental illness or substance use disorder.

 The Primary Care-Mental Health Integration program, which began pre-PACT in 2007, provides primary care—based treatment of common mental health conditions (eg, depression) and risky health behaviors (eg, heavy alcohol use) by mental health clinicians partnering with and co-located with primary care staff. The Primary Care-Mental Health Integration program is already quite robust across the VHA, and its critical role in PACT is more fully described elsewhere.

The PACT providers and clinical staff are being trained in patient-centered communication.

 By the end of 2011, facilitators were trained at all VHA facilities, and the 2012 PACT plan was to train 50% of the primary care teamlet staff. In addition, 36% of facilities followed the PACT plan to form patient advisory councils in July 2011, an increase from 18% in October 2009.

We used data from VHA Corporate Data Warehouse databases derived from VHA operational systems. All measures except appointment wait times were calculated from data on the total number of active patients assigned to a PCP in the relevant month or quarter. All primary care patients at all VHA facilities nationally were included, regardless of the status of PACT implementation at the facility. Point measures were assessed on the last day of each month starting in December 2009. For measures compiled on a quarterly basis, data were available starting with the April through June 2009 quarter for some measures and the October through December 2009 quarter for others (

We also used data from a national, facility-level survey based on the American College of Physicians (ACP) Medical Home Builder Practice Biopsy (ACP Biopsy). The ACP Biopsy assesses whether a practice has 127 PCMH components in place through yes/no items in 7 categories: patient-centered care and communication (ie, training in communication techniques and provision of self-management support); access and scheduling (ie, flexible and non—face-to-face scheduling); use of technology (ie, e-prescribing); care coordination and transitions; organization of practice (ie, electronic records and teamwork); population management (ie, patient registries); and quality improvement. The survey was completed by ambulatory care directors at 850 facilities (out of 862 eligible) in October 2009 (pre-PACT) and by 846 facilities in July 2011 (interim PACT).

We assessed changes in key patient care processes including in-person and telephone encounter rates across teamlet members, continuity of care with the patient’s assigned PCP, appointment access, access to care through shared medical appointments and new technologies, and reach of RN care management. Details on measures are available in 

. To assess overall progress toward PCMH implementation, we compared the total percentage of yes responses in each ACP Biopsy category at both time points.

We used nonparametric tests of trend for the ranks of measures across all quarters or months measured. All P values were based on 2-sided tests. Full quarterly and monthly data are available in Appendix A. The average age and comorbidity level (Deyo score)

 of the primary care population remained stable throughout the study period, so encounter rates were not adjusted for these attributes.

Interim Changes in Patient Care Processes

The overall rate of primary care encounters (in person or by phone) per 100 patients increased from 85 in April through June 2009 (corresponding to a rate of 3.4 encounters per patient per year) to 101 in April through June 2012 (corresponding to 4.0 encounters per patient per year; 

 for trend = .01). However, in-person encounter rates declined slightly over this time period (

). Among clinical staff, PCP visit rates declined slightly (53 to 43 visits per 100 patients per quarter; 

 for trend <.01), as did clinical pharmacist visits (2.6 to 2.4 visits; 

 < .01). Encounters with RNs and LPNs spiked each fall, but did not change significantly overall (P for trend RN = .39; P for trend LPN = .84). This pattern remained after encounters for flu shots were removed from the data.

As a measure of continuity, 80% of all PCP visits in October through December 2009 were with the patient’s assigned PCP (of 2,376,003 total PCP visits), which increased to 83% in April through June 2012 (2,570,618 total PCP visits) (

Patient requests for same-day appointments were accommodated 67% of the time in December 2009 and 73% in June 2012 (

 for trend <.01, Figure 2A). The proportion of patients seen within 7 days of their desired appointment date rose from 85% to 90% during the same time period (

Phone encounter rates, although a minority of overall encounters, increased more than 10-fold from April through June 2009 to April through June 2012, increasing for each clinical staff group (

). In-person shared medical appointments increased from 0.24 to 0.65 (

 <.01) per 100 patients per quarter over the same time period (Figure 1A).

The number of patients registered for the enhanced personal health record increased from 152,416 (3% of 4,759,668 primary care patients) in December 2009 to 694,206 (13% of 5,163,531 primary care patients) in June 2012 (

). Total secure messages from patients to primary care staff increased from 9852 in 2010 to 289,519 in 2011. Patients using secure messaging (sending at least 1 message to their primary care team) increased from 0.07 per 1000 primary care patients in October through December 2009 to 22.8 per 1000 patients in April through June 2012 (

The number of patients using telehealth increased from 38,747 (0.8% of patients) in December 2009 to 70,486 (1.4% of patients) in June 2012 (

 for trend <.01; Figure 2B). Patients evaluated by primary care clinicians within 48 hours of VHA hospital discharge increased from 6% of 46,195 discharges in December 2009 to 61% of 45,068 discharges in December 2011 (

Overall Progress Toward Implementation of PCMH Components. 

Facilities’ average overall score on the ACP Biopsy increased from 69% yes in October 2009 to 80% yes in July 2011. Pre-PACT, categories with the highest average scores were organization of practice (72%), population management (75%), and quality improvement (86%) (

). Categories with the lowest initial scores were also those with the greatest improvement by 2011: patient-centered care/communication (56% to 68%) and access/scheduling (66% to 76%).

Over the first 30 months of PCMH implementation, VHA made extensive nationwide structural changes in primary care staff levels and teamlet formation and capacity for telephone care, electronic access, care management, and healthy behavior promotion. The high scores noted on the ACP Biopsy pre-PACT reflect the investments made by VHA during the previous 15 years in comprehensive electronic health records, population management tools, and quality improvement programs. Larger increases on the ACP Biopsy related to patient-centered care, communication and access, and scheduling during the initial months of PACT correspond to the main components of the PACT initiative.

Articles

WEB EXCLUSIVE

We describe the Veterans Health Administration's nationwide patient-centered medical home (PCMH) initiative and evaluate interim changes in PCMH-related patient care processes.

The Patient-Centered Medical Home in the Veterans Health Administration

To assess the level of hospital participation in the first 18 months of the Medicare and Medicaid Electronic Health Record (EHR) Incentive Programs, and to identify whether vulnerable hospitals lag behind.

 Retrospective study of participation among the 4938 Medicare-certified hospitals from the beginning of the incentive payment period (June 2011) through December 2012.

 We used multivariate models to examine which types of hospitals qualified for financial incentives either through attesting to meaningful use of EHRs or by meeting the “Adopt-Implement-Upgrade” (AIU) option that requires demonstrating progress toward achieving meaningful use. We focused on small, Critical Access, and safety-net hospitals.

 We found that more than 75% of all eligible US hospitals have qualified for financial incentives in the first 18 months of the program. Nearly two-thirds of these hospitals (52% of all hospitals) attested to meaningful use while the remaining one-third (24% of all hospitals) were paid under the AIU option only. Small hospitals were less likely than large hospitals to qualify for incentive payments (odds ratio [OR] = 0.49, 95% confidence interval [CI] 0.36-0.68; 

 <.001 across categories). Critical Access hospitals also had lower odds of incentive payment (OR = 0.69, 95% CI 0.57-0.84, 

 <.001). Safety-net hospitals were more likely to qualify for payments overall (OR = 2.51; 95% CI 1.92-3.38, 

 <.001), but did so primarily through AIU.

There is broad participation in the federally led incentive program to promote nationwide EHR uptake. Lower rates of participation among smaller hospitals and Critical Access hospitals merit close monitoring to ensure that broad adoption is achieved.

		The federal government is providing financial incentives to hospitals to encourage adoption and use of electronic health records (EHRs). The level of participation in the incentive program overall and by key groups of hospitals is not widely known.

		We found broad engagement, with more than three-fourths of hospitals receiving incentive payment. Further, most hospitals that attested to meeting federal meaningful use criteria went well beyond the minimum required thresholds and were using their EHRs broadly for all patients.

		Key policy challenges that remain include bringing small, Critical Access hospitals along and helping safety-net institutions transition to achieving meaningful use.

The 2009 Health Information Technology for Economic and Clinical Health (HITECH) Act authorizes the federal government to implement policies to incent hospitals, physicians, and other healthcare professionals to adopt electronic health records (EHRs) and use them in ways expected to improve care (“meaningful use”).

 The Act was motivated by broad consensus that persistent quality and efficiency challenges in our healthcare system stem, at least in part, from reliance on paper-based records.

The  primary mechanism to promote meaningful use of EHRs under HITECH is direct payments to eligible professionals and hospitals, for which the Act set aside $27 billion.

For hospitals, there are 2 approaches to qualify for incentives. The first is “attestation,” in which hospitals report that they have met a set of federally defined meaningful use criteria using a certified EHR technology. The alternative, which is open primarily to hospitals with at least 10% Medicaid patient volume, is known as Adopt-Implement-Upgrade (AIU) and allows eligible hospitals to receive an incentive payment before they meet meaningful use. The AIU option was created in response to data suggesting that safety-net hospitals would have trouble accessing the capital required to purchase and fully implement an EHR.

 The AIU option therefore sought to ensure that HITECH did not result in a digital divide. While AIU helps in the short term by giving financially stressed hospitals access to capital to support EHR adoption, hospitals that choose this option may struggle in the long run when they have to play catch-up and achieve the same requirements of increasingly sophisticated EHR use as other hospitals.

While the incentive program has been in place for nearly 2 years, we lack a comprehensive evaluation of the types of hospitals that have qualified for incentives, and whether the program is leaving certain groups of hospitals behind. These issues are important for 2 reasons. First, data collected prior to HITECH revealed a digital divide in which large, academic teaching institutions were far more likely to have EHRs compared with smaller, rural, and safety-net institutions.

 Second, the benefits from EHRs are substantially greater when there is broad adoption and connectivity between systems.

 These “network” effects will only be realized if the incentive program significantly increases adoption, and does not simply reward those who already had EHR systems. Therefore, empirical data  on how many hospitals have received incentives, and whether there are systematic differences in the kinds of hospitals receiving the incentives, are critically important.

To address this gap, we use data from the Centers for Medicare & Medicaid Services (CMS), the agency administering the EHR Incentive Programs, to answer 4 questions: (1) What proportion of US hospitals have qualified for incentive payments through the programs, and how many of these hospitals were capable of qualifying prior to the programs? (2) Are certain types of hospitals, specifically small, Critical Access or safety-net institutions, less likely to receive payments compared with other institutions? (3) To what extent are safety-net institutions opting for the AIU approach to receive payments? (4) And finally, to what extent are the meaningful use criteria serving as a minimum for EHR use or are hospitals going beyond the basic requirements to widely use EHR functionalities for all their patients? Our findings offer a comprehensive examination of national data on the implementation of HITECH through 2012, with a focus on how incentives are being distributed across US hospitals.

Data on hospital participation in the first stage

 of the EHR Incentive Programs as of December 31, 2012, came from CMS and were made available to our research team through the Office of the National Coordinator (ONC) for Health Information Technology at the Department of Health and Human Services. This provided unique access to the list of hospitals that qualified for payment under AIU and Medicaid Stage 1 meaningful use, which is not publicly released along with hospitals that have attested to Stage 1 meaningful use under Medicare. AIU-eligible hospitals are acute care hospitals with at least 10 percent Medicaid patient volume and children’s hospitals; to receive AIU payment, eligible hospitals must adopt, implement, or upgrade a certified EHR technology during the first year of participation in the program.

The CMS data also included the specific meaningful use criteria met by attesting hospitals, and the level of achievement for each of those criteria. Meaningful use criteria come in 2 varieties: the first type requires that the EHR perform specific tasks (ie, dichotomous measures), such as having the capability to exchange clinical data electronically. The second type, which is more common, focuses on broadbased clinical use of specific functions and is reported as a continuous variable, with the meaningful use criteria specifying a threshold that has to be achieved to receive credit. For example, the criterion for computerized provider order entry (CPOE) for medications requires that medications are ordered electronically for at least 30% of patients in a hospital.

We used the Medicare-certified list of 4938 hospitals to identify hospitals eligible for, but not participating in, the incentive program. We assigned characteristics to hospitals based on their responses to the 2011 American Hospital Association (AHA) survey. We obtained hospital financial data, including each institution’s Disproportionate Share (DSH) Index, from Medicare Cost Reports, which are publicly available for nearly all Medicare-eligible hospitals. When we merged the sources of data, 275 hospitals (5.6%) were missing 1 or more characteristics, resulting in an analytic sample of 4663 hospitals (

We classified hospitals into 1 of 3 mutually exclusive groups: (1) hospitals that attested to Stage 1 meaningful use under the Medicare or Medicaid EHR Incentive Program; (2) hospitals that chose to follow the AIU option; and (3) all remaining Medicare-certified hospitals, including those that registered for the incentive program but did not qualify for payment, as well as those that have not engaged with the program in any form. If a hospital attested to meaningful use as well as qualified for AIU payment (which is allowed under the program), we categorized it in the attested group (Group 1). To assess the proportion of Group 1 hospitals that appeared to be close to attaining (or perhaps had already attained) Stage 1 meaningful use prior to the program period, we leveraged prior work to define and determine which hospitals met a meaningful use proxy measure using mid-2010 data from the AHA Information Technology Supplement.

We also examined the degree of sophistication of EHR use among attesting hospitals (Group 1). For the subset of meaningful use criteria that are measured on a continuous scale, we focused on the degree of deployment of these meaningful use functionalities to determine whether hospitals did just enough to meet the thresholds or whether they deployed these electronic functionalities widely across the institution We defined a “sophisticated” hospital as one that met all 5 dichotomous criteria and the 9 continuous criteria at least 90% of the time in the reporting period. We also examined which hospitals were able to achieve the medication CPOE measure, arguably one of the hardest criteria, for at least 90% of patients.

 Because pre-HITECH data found that small hospitals in general, and Critical Access hospitals (CAHs) in particular, were less likely to have EHRs,

we were interested in assessing the degree to which these institutions were receiving incentives. In addition, safety-net institutions—those with a high DSH Index&mdash;were also less likely to have adopted EHR systems, due in part to greater difficulty accessing required capital.

Therefore, we sought to assess whether these institutions were receiving incentives at the same rate as non-safety-net hospitals.

We examined an additional set of hospital characteristics, either because prior data suggest that they are associated with differential EHR adoption or because they are of policy interest.These characteristics included ownership (for-profit, nonprofit private, or public), teaching status, whether the hospital was affiliated with a system, whether the hospital was part of a system that offered a health maintenance organization (HMO) product, geographic location (urban/rural and region), and 3 dimensions of financial health (operating margin, capitalization, liquidity). Financial variables were calculated using an approach8 that captures financial health over a 3-year period (2007-2009) to best reflect the ability of hospitals to invest in IT systems at the start of HITECH.

 We first examined the proportion of hospitals (and associated proportion of discharges and of beds) that attested to Stage 1 meaningful use, met AIU only (ie, did not attest), and were not participating. We then examined the bivariate relationships between these 3 groups and the hospital characteristics of interest. We subsequently built multivariable logistic regression models to identify the independent associations between each characteristic and our 4 outcomes of interest: hospitals that qualified for incentives versus those that did not, hospitals that attested to meaningful use versus those that chose AIU (among those that qualified for incentives), and 2 measures of sophistication. Finally, we generated descriptive statistics that reflect the distribution of hospital achievement for each of the 9 continuous meaningful use criteria.

Hospital Participation in the EHR Incentive Programs

 We found that 75.5% of eligible US hospitals qualified for payments under the federal EHR incentive programs through the end of 2012. Just over half of all hospitals qualified for incentive payment because they had attested to Stage 1 meaningful use (2512 hospitals, or 52% of all hospitals, representing 56% of beds and 59% of all discharges, Figure). An additional 24% of hospitals (1141 hospitals, representing 25% of beds and discharges) qualified for payment under the AIU path. Approximately 1 in 4 hospitals (1185 or 24% of hospitals, representing 18% of beds and 16% of discharges) did not participate. Among the 1573 hospitals attesting to Stage 1 meaningful use for which the meaningful use proxy measure was available, 93% (1468 hospitals) did not meet the proxy measure in 2010 (ie, appeared to be a “new” meaningful user) and the remaining 7% (105 hospitals) were at least capable of meeting Stage 1 meaningful use prior to the incentive program.

Characteristics of Hospitals Qualifying for Incentive Payments.

 In bivariate analyses, we found important differences between hospitals that attested to meaningful use, those that chose AIU, and those that did not participate. Consistent with our hypothesis, we found that small hospitals were less likely than large hospitals to have qualified for payment (69% vs 82%, 

 value for differences across groups <.001, 

). Similarly, we found that CAHs were less likely to have qualified for payment than non-CAHs (68% vs 79%, 

<.001 across categories). Most of this difference was due to lower rates of AIU (18% among CAHs compared with 26% among non-CAHs, 

<.001). We found no evidence that safety-net hospitals (those in the highest DSH quartile) were less likely to receive incentives under HITECH. In fact, safety-net hospitals had substantially higher rates of incentive payment compared withhospitals with the lowest DSH index (82% versus 65%, 

 value for differences across the 4 quartiles <.001), which was driven by higher AIU participation.

In multivariate models that adjusted for our set of hospital characteristics, we found similar results. Small hospitals had less than half the odds of qualifying for incentives compared with large institutions (OR = 0.49, 95% CI 0.36-0.68 for small,

). Even when taking size into account, we found that CAHs were less likely to have received incentives compared  with non-CAHs (OR = 0.69, 95% CI 0.57-0.84, 

<.001). In contrast, safety-net hospitals had more than twice the odds  of receiving incentives compared with hospitals in the lowest quartile of DSH (OR = 2.51, 95% CI 1.92-3.28), and odds increased for each DSH quartile (

Characteristics of Hospitals That Attested to Meaningful Use Versus Adopt-Implement-Upgrade.

 Among the 75% of hospitals that qualified for incentives, the key difference between those that attested to meaningful use versus those that met AIU was safety-net status. Hospitals in the highest DSH quartile had 42% lower odds of having attested to meaningful use compared with having chosen AIU (OR = 0.58, 95% CI 0.44-0.75, 

). The middle 2 quartiles of DSH had similarly lower odds of having attested to meaningful use versus having chosen AIU (OR = 0.61, 95% CI 0.47-0.79 for quartile 3 and 0.70, 95% CI 0.54-0.91 for quartile 2; 

<.001 across the 4 quartiles). We did not find systematic differences in whether hospitals qualified for incentives under meaningful use or AIU based on size or Critical Access status (

 Among the 52% of  hospitals that attested to meaningful use, the majority substantially exceeded the minimum threshold for each of the continuous criteria  (Table 4). For example, although meaningful use requires medication CPOE to be used for at least 30% of patients, median adherence was 92% with an interquartile range of 74% to 99%. Similarly, for the 5 measures requiring a 50% threshold, median adherence was above 95%, ranging from 96% for vital signs to 100% for electronic copy of health information and discharge instructions.

The only characteristic that distinguished sophisticated meaningful users from those that attested but did not at least achieve 90%use for all criteria was whether the hospital was part of a system that offered an HMO product (OR =1.33, 95% CI 1.02-1.74, 

). When we examined medication CPOE sophistication alone, we again found that offering an HMO product was associated with at least 90% use (OR = 1.39, 95% CI 1.08-1.78, 

 =.01). We also found that for-profit hospitals were less likely to be a sophisticated CPOE user compared with not-for-profit hospitals (OR = 0.65, 95% CI 0.51-0.83, 

). We did not find differences on either measure of sophistication for our 3 characteristics of interest: size, CAH, and safety-net status.

In an early examination of the Medicare and Medicaid EHR Incentive Programs, the centerpiece of the HITECH Act, we found that more than 3 in 4 hospitals are participating, and that more than half have achieved Stage 1 meaningful use. While small hospitals in general and Critical Access hospitals in particular are less likely to have received incentives, the vast majority of both groups are participating in the program, and half had attested to meaningful use. Further, we found that safety-net hospitals were as or even more likely to qualify for incentives. Taken as a whole, our study finds little evidence for the concern that the HITECH Act will give additional resources to wellheeled hospitals that already have EHRs while setting up struggling institutions to be penalized later. While differences exist in who is receiving incentive payments and who is not, overall the incentives seem to be flowing to a broad swath of US hospitals.

Any federal incentive program that rewards hospitals for adopting and using an expensive technology such as EHRs while penalizing those who do not (which HITECH will do starting in 2015) runs the risk of worsening the gap between wealthy and poor hospitals. Indeed, evidence prior to HITECH suggested that there were important differences in EHR adoption based on size, Critical Access status, and safety-net status.

 Based on this concern, federal policy makers created a more flexible path to incentives, the AIU  approach, which provides upfront capital to hospitals that are working toward achieving meaningful use. Our findings suggest that safety-net hospitals are availing themselves of this path, using it far more often than other institutions. Other groups of hospitals that were expected to struggle to achieve meaningful use, in particular small hospitals and CAHs, are attesting at similar rates to other institutions.

In addition, hospitals attesting to meaningful use are not simply doing the minimum necessary to receive the incentives; instead, the vast majority of patients in these hospitals are being cared for using EHRs. For policy makers, this finding should offer reassurance that they selected thresholds that were achievable. It also suggests that an approach of gradually raising the threshold may not be necessary, and future criteria could instead focus on new uses of EHRs that may deliver greater improvements in both the quality and costs of care.

As policy makers plan for future stages of the EHR incentive programs, our findings point to challenges that lie ahead. First, although small hospitals and CAHs are only modestly behind, the gap has the potential to widen as the program matures. If a subset of smaller, Critical Access hospitals take longer or fail to meet Stage 1 altogether, the financial penalties slated to begin in 2015 could make EHR adoption even more difficult for these institutions. The challenge for safety-net institutions is somewhat different. Although they are faring well under the incentive program, whether those that have gone down the AIU path will be able to transition to becoming meaningful users on the accelerated timeline is unclear and will need to be closely tracked.

Our study contributes to the growing body of work examining the impact of HITECH on EHR adoption. During the first year of HITECH, DesRoches et al found that 26.6% of hospitals had adopted at least a basic EHR and 18.4% of hospitals could meet a proxy measure of meaningful use.

They also found evidence of a digital divide with small, nonteaching, and rural hospitals adopting EHRs to a lesser extent than other types of hospitals. Compared with these data, our results suggest substantial progress. We found that in the year prior to the onset of incentives, only 7% of hospitals that eventually attested could meet the criteria for meaningful use. In the past 2 years, ONC released a report on EHR adoption trends over time,

 and CMS has issued monthly reports that include the number of hospitals that have received payment under the EHR Incentive Programs.

More detailed data from the Government Accountability Office in 2011 reported many of the same bivariate relationships that we observed (eg, higher participation among larger hospitals).

 Our study extends this literature in several ways—by examining the full set of hospitals that qualified for incentive payments under both the attestation and AIU approaches, by assessing whether key types of hospitals are falling behind, and by reporting the degree to which hospitals attesting to meaningful use are going beyond the basic thresholds required by the criteria.

Our study has important limitations. First, we were unable to assess where hospitals that did not participate in the program stood with respect to EHR adoption. It may be that these hospitals have responded, but not quickly enough to have qualified for an incentive payment through 2012, and that over the next year they will catch up to their larger and better-resourced counterparts. Alternatively, given the persistence of a digital divide in other measures of EHR adoption, these non-participating small hospitals and CAHs may still have substantial work ahead to qualify for incentives.

Second, we were unable to definitively differentiate which hospitals adopted EHRs or achieved meaningful use specifically in response to the incentives, and which hospitals had already done so or would have done so without the incentives. We therefore are not certain about the extent to which HITECH is driving EHR adoption or the achievement of meaningful use. However, electronic functionalities that have been adopted most quickly over the past 2 years are those required for meaningful use, suggesting that the program is likely playing an important role.

Finally, our  findings reflect achievement of the first stage of a 3-part program. Upcoming stages of meaningful use are expected to be more challenging and our results may not predict how hospitals will fare.

In summary, we examined the uptake of the federal incentive program designed to achieve widespread adoption and use of electronic health records across US hospitals. Our findings show broad engagement, with more than three-fourths of hospitals qualifying for incentives. Further, among the half of all hospitals that attested to meeting meaningful use through 2012, most were able to go well beyond the minimum required threshold. However, key policy challenges remain, including bringing small, Critical Access hospitals along and helping safety-net institutions transition to achieving meaningful use. Meeting these challenges will be critical to ensuring that all Americans, regardless of where they are treated, receive highquality care.

From University of Michigan School of Information and School of Public Health (JA-M), Ann Arbor, MI; Office of the NationalCoordinator for Health Information Technology at the Department of Health and Human Services (MFF, JK), Washington, DC; Harvard School of Public Health (AKJ), Boston, MA.

 Drs Adler-Milstein and Jha were supported by The Robert Wood Johnson Foundation.

The authors (JA-M, MFF, JK, AKJ) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (JA-M, MFF, JK, AKJ); acquisition of data (MFF, JK, AKJ); analysis and interpretation of data (JA-M, MFF, JK, AKJ); drafting of the manuscript (JA-M, MFF, JK, AKJ); critical revision of the manuscript for important intellectual content (JA-M, MFF, JK, AKJ); statistical analysis (JA-M, MFF); provision of study materials or patients (MFF, JK); obtaining funding (JA-M, AKJ); administrative, technical, or logistic support (JA-M, MFF, JK, AKJ); and supervision (JA-M, MFF, JK, AKJ).

Julia Adler-Milstein, PhD, University of Michigan, School of Information, 4376 North Quad, Ann Arbor, MI 48109. E-mail: juliaam@umich.edu.1. Blumenthal D. Launching HITECH. N Engl J Med. 2010;362(5): 382-385.

2. Blumenthal D. Wiring the health system — origins and provisions of a new federal program. N Engl J Med. 2011;365(24):2323-2329.

3. Jha AK, DesRoches CM, Shields AE, et al. Evidence of an emerging digital divide among hospitals that care for the poor. Health Aff. 2009;28(6):w1160-w1170.

While more than 75% of hospitals are participating in the federal electronic health record incentive program, small hospitals and Critical Access hospitals lag behind.

Early Results From the Hospital Electronic Health Record Incentive Programs

 To develop a model comparing costs of 8 hormonal contraceptives and determine whether acquisition costs for implants and intrauterinedevices (IUDs) were offset by decreased pregnancy-related costs over a 3-year time horizon from a managed care perspective.

A model was developed to assess the budget impact of branded or generic oral contraceptives (OCs), quarterly intramuscular depot medroxyprogesterone, etonogestrel/ethinyl estradiol vaginal ring, etonogestrel implant, levonorgestrel IUD, norelgestromin/ethinyl estradiol transdermal contraceptive, and ethinyl estradiol/levonorgestrel extended-cycle OC.

 Major variables included drug costs, typical use failure rates, discontinuation rates, and pregnancy costs. The base case assessed costs for 1000 women initiating each of the hormonal contraceptives.

 The etonogestrel implant and levonorgestrel IUD resulted in the fewest pregnancies,  63 and 85, respectively, and the least  cost, $1.75 million and $2.0 million, respectively. In comparison, generic OC users accounted for a total of 243 pregnancies and $3.4 million in costs. At the end of year 1, costs for the etonogestrel implant ($800,471) and levonorgestrel IUD ($949,721) were  already lower than those for generic OCs ($1,146,890). Sensitivity analysis showed that the cost of pregnancies, not product acquisition cost, was the primary cost driver.

 Higher initial acquisition costs for the etonogestrel implant and levonorgestrel IUD  were offset within 1 year by lower contraceptive failure rates and consequent pregnancy costs. Thus, after accounting for typical use failure rates of contraceptive  products, the etonogestrel implant and levonorgestrel IUD emerged as the least expensive hormonal contraceptives. 

A model was developed to assess the budget impact of 8 hormonal contraceptive options.

		Lower pregnancy rates, due to lower typical use failure rates, for the etonogestrel implant and levonorgestrel intrauterine device offset their higher initial costs (eg, acquisitionand insertion/removal costs) even in the first year of use.

		From a managed care perspective, including direct pharmacy and medical costs, branded and generic oral contraceptives were among the highest cost products overthe course of 3 years.

Many women spend more than half of their approximately 40-year reproductive life spans attempting to avoid having children.

Development of the first hormonal “pill” in the 1960s resulted in reliable, reversible, and safe contraception. After more than 40 years, hormonal contraceptives—oral contraceptives (OCs), implants, patches, vaginal rings, injectables, and intrauterine devices (IUDs)&mdash;are used by millions of women in the United States and throughout the world.

 Yet, despite the availability of a broad range of contraceptive options, almost half (49%) of all pregnancies in the United States are unintended.

The efficacy of hormonal contraceptives in averting unintended pregnancies is well established. With perfect use, failure rates are 

0.3% for hormonal contraceptives compared with 85% for no contraception.

 However, the difference between contraceptive failure rates with typical use versus perfect use, as first described in 1987,

 is recognized as a major contributor to unintended pregnancy,

and this information is now included in US Food and Drug Administration (FDA) product labeling of almost all contraceptive products.

Avoiding unintended pregnancy results in substantial cost savings. Maternal pregnancy and delivery generated the largest total  amount charged to private payers in the United States in 2006: $23 billion in charges for 2.3 million hospital stays, for an average cost of approximately $10,000 per stay.9 Neonatal care ranked third. Most healthcare insurers recognize the value of covering reversible contraceptive services and supplies to plans, members, and employers.

 However, optimal structuring of contraceptive benefits requires integration of multiple factors beyond the contraceptive drug or device per se.

We developed a transparent Excel-based model to assess the cost of 8 reversible contraceptive options: branded OCs; generic OCs; quarterly intramuscular depot medroxyprogesterone (MPA);etonogestrel/ethinyl estradiol vaginal ring (NuvaRing, Merck & Co, Inc, Whitehouse Station, NJ); etonogestrel implant (Implanon and Nexplanon, Merck & Co, Inc, Whitehouse Station, NJ); levonorgestrel-releasing intrauterine system (Mirena, Bayer HealthCare Pharmaceuticals Inc, Wayne, NJ); norelgestromin/ethinyl  estradiol transdermal contraceptive (Ortho Evra, Ortho-McNeil-Janssen Pharmaceuticals, Inc, Raritan, NJ); and ethinyl estradiol/levonorgestrel extended-cycle OC (Seasonale, Duramed Pharmaceuticals, Inc, Pomona, NY).

The objective of the present budget impact model was to determine whether initial acquisition costs of certain hormonal contraceptive methods can be offset by decreased pregnancy-related medical costs in a commercial population. We therefore applied our budget impact model to a hypothetical cohort of 1000 women initiating each of the 8 hormonal contraceptive methods. The model compared 1-, 2-, and 3-year drug and medical costs from the managed care perspective.

We evaluated the costs of 8 hormonal contraceptive methods in women aged 15 to 49 years, across a 3-year time horizon, according to a decision-analytic model (

). Costs were calculated for 1000 women initiating contraception with each of the methods at the end of year 1, year 2, after the first 2 years, at year 3, and at the end of 3 years.

The model included the following assumptions:

		Women could continue on the initial method, switch to another hormonal contraceptive, discontinue contraception, or become pregnant. It was assumed that patients switch only once over the course of the 3-year period. The costs incurred and any pregnancies after switching to another contraception method were attributed to the first method.

		Typical use failure rates, which included method and user failure, and discontinuation rates for the first year, were based on data published by Trussell

). Discontinuation rates for years 2 and 3 were assumed to be half of those in the previous year.

		All switches, dropouts, and pregnancies were assumed to occur at midyear. This assumption should be actuarially equivalent to allocating these events evenly throughout the year.

		Of women who discontinued their method of contraception, 90% were assumed to have switched to another hormonal product, and 10% were assumed to have discontinued all contraception.

		Women who were not using hormonal contraception were assumed not to be using any contraception, and were  assigned an 85% fertility rate. The effect of over-the-counter contraceptive methods was considered in the sensitivity analysis.

		For women who became pregnant in any given year, the probability of pregnancy was reduced by 25% the following year to account for the last months of pregnancy.

 Costs of contraception were assigned to the year in which the cost would occur; for example, acquisition and insertion costs for etonogestrel implant and levonorgestrel IUD were included in year 1. All prescription costs, physician visits, and pregnancy costs incurred after a switch or discontinuation were attributed to the initial method. For women who switched, the model used a weighted average for pregnancy rates, prescription costs, and medical costs based on the market share of the remaining 7 methods.

 Market share data were based on Retail and Non-Retail Hormonal Contraceptive IMS MIDAS Cycle Data in the United States for the year to date in November 2010

). These data were used in the model to determine the probability of switching to another method and to assign typical use failure rates, pharmacy, and medical costs for the remaining years following the switch. In the base case, the etonogestrel implant market share was arbitrarily set to increase to 10%, and remaining market shares were normalized to total 100% (Table 2).

Wholesale Acquisition Cost, Pharmacy Discounts, and Rebates

. Unit costs for drugs in the base case (Table 2) were wholesale acquisition cost prices as of January 2011.

 The wholesale acquisition cost prices per tablet for branded and generic OCs ($1.68 and $0.72) were weighted averages. Because rebates and maximum-allowable-cost discounts vary considerably from one managed care plan to another and are considered proprietary, they were set to $0 in the base.

Pharmacy Copayments, Coinsurance, and Dispensing Fees. 

For the base case, the model assigned 20% coinsurance for physician-administered contraceptive products (etonogestrel implant, levonorgestrel IUD, and MPA quarterly injectable), a $15 copayment for branded products (eg, branded OCs, etonogestrel/ethinyl estradiol vaginal ring, norelgestromin/ ethinyl estradiol transdermal contraceptive, ethinyl estradiol/levonorgestrel extended-cycle OC), and a $7.50 copayment for generic OCs. Mail order purchases, which allowed patients to obtain a 3-month supply of pharmacy product by mail for the equivalent of 2 retail copayments, were set to 16% of all fills.13 Dispensing fees were set to $2.44 in the base case.

. Medical costs associated with contraceptive use were based on the Current Procedural Terminology (CPT) codes and prices based on an analysis of the Thomson Reuters MarketScan Commercial Claims and Encounters Database. 15 The mean cost of all claims with CPT codes between April 2011 and September 2011 was calculated and used as the average cost of these medical procedures. Medical procedure costs comprised office visits (CPT 99213), $73.33; implant insertion (CPT 11981), $150.91; implant removal (CPT 11982), $181.28; IUD insertion (CPT 58300), $126.62; IUD removal (CPT 58301), $143.07; and quarterly injection (CPT 96372), $35.58. One doctor’s visit was included each year for each woman using products other than the IUD or injections. For women using the levonorgestrel IUD, 2 doctor’s visits were assumed for the first year and 1 visit in years 2 and 3. Women receiving injections were assumed to make quarterly physician visits. For women who switched from an implant or IUD to another product, the model added the cost for the medical office visit to remove the product and the cost for an additional medical office visit to introduce the new product. The model also assumed that there would be one medical office visit for placement of each new product.

Pregnancy Outcome Probabilities and Costs.

 The probability of pregnancy outcome for the 6 most common outcomes (full-term vaginal delivery, cesarean section, premature delivery, ectopic pregnancy, elective abortion, and miscarriage) was stratified by maternal age (

The model assumed that there was an equal distribution of women; that is, 14.3% of the cohort fell in each of the 7 age categories presented.

 Thomson Reuters MarketScan Commercial Claims and Encounters Database

 data for new events between April and September 2010; events were considered new if there were no payments with similar 

International Classification of Diseases, Ninth Revision

) or Diagnosis-Related Group (DRG) codes in the prior 3 months. All related payments with similar 

or DRG codes for the following 3 months were included in the total cost of the event. For ectopic pregnancy (

 633.xx and DRG 777) and induced abortion (

635.xx and DRG 770 and 779), inpatient   and outpatient costs were combined; miscarriage costs were outpatient only (

 634.xx); and live birth outcomes (vaginal delivery, DRG 775 and 795; cesarean section, DRG 766 and 795; and premature delivery, DRG 774, 791, and 792) were considered inpatient only (Table 4). Costs for delivery included those for newborn care in the hospital following delivery. Costs of pregnancy outcomes were assigned to the year of conception.

To determine which factors had the greatest impact on 3-year total costs for each contraceptive method compared with no contraception, a 1-way sensitivity analysis was performed. A total of 10 variables were varied one at a time by ±50% of their base case values to assess their individual impacts compared with no contraception: costs of pregnancies, proportions of patients who switch after discontinuing, discontinuation rate, unit costs of contraceptive method, yearly failure rates, average costs of medical visits, removal costs for implants, insertion costs for implants, insertion costs for IUDs, and removal costs for IUDs. In addition, the age groups were varied from 100% of women aged 15 to 19 years to 100% of women aged 45 to 49 years.

Additionally, to assess the impact of the assumption that women who drop out of hormonal contraceptive coverage experience an 85% fertility rate, this variable was set to 28% and 2%, the typical use failure rate for spermicide and the perfect use failure rate for male condoms, respectively (ie, the highest and lowest failure rates for over-the-counter contraceptive options).

Institutional review board approval was not sought for this analysis, as no direct patient data were used in this analysis.

We summarized the medical and pharmacy costs and pregnancy rates for each contraceptive option calculated for the 1000 women on each of the contraceptive options in the base case (

). The etonogestrel implant and levonorgestrel IUD were the least expensive methods for each year analyzed. In year 1, when devices were inserted, costs were $800,471 for the etonogestrel implant and $949,721 for the levonorgestrel IUD. In contrast, medical and pharmacy costs in year 1 for generic contraceptives were $1,146,890. The difference in costs between the methods increased in year 2, as the costs for acquisition and insertion for both the implant and the IUD were much reduced. For year 2, the costs dropped to $383,552 for etonogestrel and $484,863 for the levonorgestrel IUD, whereas the cost the total cost of generic contraceptives remained relatively unchanged at $1,143,432. In year 3, the cost of etonogestrel rose again to $567,729, primarily due to the added costs of removing the implants; the cost of the levon norgestrel IUD was $547,215. The cost of both of these contraceptive options in year 3 was still approximately half the estimated cost of generic OCs at $1,077,633. Over the course of the 3-year period, the least expensive contraceptive options were the etonogestrel implant and levonorgestrel IUD, costing $1,751,753 and $1,981,799, respectively. By comparison, the 3-year total costs were $3,367,955 for the generic OCs; $3,705,428 for the quarterly injectable; $3,538,401 for extended-cycle OCs; $3,708,301 for branded OCs; $4,082,093 for the vaginal ring; and $4,545,717 for the transdermal patch (Table 5).

The differences in costs were mostly attributable to differences in costs for pregnancies over the 3-year period. Among 1000 women starting on etonogestrel, the model predicted a total of 63 pregnancies over the course of 3 years: 1 on etonogestrel, 30 among those discontinuing, and 32 among those who switched to another product. In contrast, there were an expected 243 pregnancies over the course of 3 years with generic OCs: 145 on generic OCs, 59 among dropouts, and 39 among those that switched to another product. The total cost of these pregnancies was expected to be $646,103 for etonogestrel compared with $2,492,744 for generic OCs.

Of the 11 variables that were varied, costs of pregnancy outcomes were the most salient determinants of 3-year net savings compared with no contraception for all 8 contraceptive methods (

). When cost of pregnancy outcomes was varied by ±50%, the cost savings relative to no contraception varied from $2.9 million to $11.0 million for the etonogestrel implant, from $2.8 million to $10.7 million for the levonorgestrel IUD, and from $2.2 million to $8.5 million for generic OCs. In contrast, varying the unit costs of each contraceptive by ±50% led to a smaller impact on cost savings compared with no contraception: the etonogestrel implant cost savings varied from $6.7 million to $7.2 million, levonorgestrel IUD costs savings varied from $6.4 million to $7.0 million, and cost savings for branded OCs ranged from $4.3 million to $5.7 million.

When we varied the fertility rate to 28% and 2% to assess the potential impact of over-the-counter contraceptive use, the 3-year cost of contraception decreased accordingly.Over 3 years, the cost of etonogestrel implants decreased to $1,576,395 and $1,452,814, respectively. The cost of the levonorgestrel IUD decreased to $1,762,939 and $1,608,466, respectively; and the cost of generic OCs dropped to $3,051,777 and $2,825,698.

The impact of unintended pregnancy on costs to health plans, members, and employers warrants careful analysis of coverage to ensure that members have access to cost-effective contraception. This analysis demonstrated that the etonogestrel implant and levonorgestrel IUD were the least costly contraceptive methods over 3 years, despite their higher acquisition costs. At year 1, total costs for the etonogestrel implant and levonorgestrel IUD were less than those for both branded and generic OCs. The relatively high acquisition costs of the long-term methods were more than offset by their lower typical use failure rates.

A model assessing 8 hormonal contraceptive methods found that lower pregnancy costs for the etonogestrel implant and levonorgestrel intrauterine device offset their higher acquisition costs.

July 2013

The Patient-Centered Medical Home in the Veterans Health Administration

Early Results From the Hospital Electronic Health Record Incentive Programs

Budget Impact Analysis of 8 Hormonal Contraceptive Options

Physician Assistants and Their Intent to Retire

Healthcare's 'Big Data' Challenge

Previously Unrecognized Trends in Diabetes Consumption Clusters in Medicare

Healthcare System Effects of Pay-for-Performance for Smoking Status Documentation

Change in Claims-Based Diabetes Medications Is a Diabetes Improvement Indicator

Risk Bearing and Use of Fee-for-Service Billing Among Accountable Care Organizations

Encouraging Value-Based Insurance Designs in State Health Insurance Exchanges