Somesh Nigam, PhD

Although pharmacy benefit carve-outs are promoted as a cost-containment tool, their impact on medical spending is not well understood. We compare the health care spending of Blue Cross and Blue Shield of Louisiana (BCBSLA) members covered by an integrated (“carved-in”) pharmacy benefit with that of members covered under a pharmacy benefit carve-out.

 We identified members with coverage through an employer contracting for administrative services only (ie, self-insured) and determined whether they received a pharmacy benefit through BCBSLA. We matched members with and without integrated benefits using a baseline year and compared their medical spending trajectories in 3 subsequent years. These comparisons were repeated in the subset of patients with chronic comorbidities.

 Among patients with chronic illnesses, relative growth in per-member per-month (PMPM) medical spending was significantly lower in the integrated benefit group by the second and third follow-up years. Neither the level nor the growth of PMPM medical spending significantly differed in the full population sample, although point estimates suggest that the integrated benefit members may be on a lower cost growth trajectory over time.

 Members with chronic illnesses receiving an integrated pharmacy benefit experienced slower medical cost growth compared with members covered by a pharmacy carve-out. Group leaders and brokers should consider the additional cost savings achieved by integrated pharmacy benefits when comparing the total costs of carve-in vs carve-out prescription drug programs.

Members with chronic illnesses receiving an integrated pharmacy benefit experienced slower medical cost growth compared with members covered by a pharmacy carve-out.

Integrated pharmacy benefits are associated with reduced medical expenditures over the long term, resulting in annual per-member per-month savings compared with a carve-out.

These savings are observable in the population as a whole but are more pronounced for patients with chronic illnesses.

Any potential savings on drug spending promised by a pharmacy benefits carve-out should be weighed against potential increases in medical spending.

An employer considering the structure of its employee health insurance plan must make several consequential decisions about benefit design. One key choice is whether to exclude (ie, “carve out”) benefits for certain services or patient populations from core medical benefits or to offer an integrated benefits package from a single vendor (ie, to “carve in” these services). Extensive debate about the relative merits of these 2 approaches followed the emergence of pharmacy benefits managers (PBMs) in the 1980s

 and mental health and substance abuse (MH/SA) carve-out plans in the 1990s.

Carve-outs are generally promoted as a mechanism for cost containment while also attempting to preserve a high standard of quality of care.

 In both the PBM and MH/SA contexts, carve-outs rely on specialization as a component of cost reductions.

 In the case of PBMs, this can take the form of utilization review and formulary management, which are used to identify cost-effective drugs and promote their utilization through physician and beneficiary education or tiered co-pays.

 Carve-out plans can also leverage greater scale to negotiate discounts more effectively. In 2017, PBMs directed prescription drug coverage for roughly 266 million Americans. Furthermore, consolidation of PBMs ensured that about 80% of American lives, in 2014, were covered by the top 3 PBMs.

In the MH/SA case, prior research has shown that carving out benefits can lead to reductions in costs in a variety of settings,

 although the evidence regarding the impact of carve-outs on quality is mixed.

 When comparing the performance of pharmaceutical carve-outs and integrated benefits, the picture is much murkier, owing in part to the opacity of negotiations between PBMs and drug manufacturers.

 Given this uncertainty, it is important to consider areas where pharmacy benefit carve-outs may not deliver savings—especially given recent commentary, which has highlighted that PBMs are not necessarily organized or incentivized to generate the most value for patients.

 Indeed, some industry reports have pointed to potential savings from the provision of an integrated pharmacy benefit.

 Moreover, several states have made sweeping changes to their Medicaid PBM contracts, ranging from imposing new disclosure requirements to terminating them entirely, indicating concerns about whether these arrangements are delivering value.

For example, one aspect of MH/SA carve-outs highlighted in evaluation is the contribution of utilization review and care coordination. Although selective contracting with specialty providers or PBMs might drive efficiencies in health care consumption 

 a given category of spending, it could also increase fragmentation across health care settings for the 

. Therefore, providing integrated pharmacy or other benefits could allow for improved care coordination. One important component of integrated benefits is the availability of prescription drug utilization data to health plans because prior work has shown that pharmacy data can be a reliable predictor of a patient’s total health care spending.

These considerations are particularly pressing given the massive contribution of prescription drug spending to health care costs and cost growth. Retail drugs account for 21% of spending by employer-sponsored health plans.

 Although expenditure growth slowed in 2016 and 2017, the previous 2 years saw 12.4% and 8.9% growth in retail drug spending, pointing to high uncertainty in these costs as well.

 High prescription drug costs will continue to exert pressure on employers to seek savings in pharmacy costs.

This study used claims data from Blue Cross and Blue Shield of Louisiana (BCBSLA) members to evaluate the association between coverage under integrated pharmacy benefits and medical spending, which may guide the choices of employers considering the design of employee benefits. We followed a matched cohort of patients with and without integrated pharmacy benefits over 3 years and assessed the potential impact of integrated pharmacy benefits on spending for the entire population and a select cohort of members with chronic conditions.

We conducted a longitudinal, matched case-control study to assess whether the provision of integrated pharmacy benefits was associated with lower levels of growth of health care spending. Our analysis used member-level medical claims data (excluding pharmacy claims) from BCBSLA covering the period from October 1, 2014, through September 30, 2018, and included members covered through employers who contracted for administrative services only (self-insured firms). Our outcome of interest was total combined inpatient and outpatient medical costs, which excluded spending on prescription drugs because these records were unavailable for members covered by a prescription drug program carved out to a PBM. Using administrative data, we partitioned our sample into 2 mutually exclusive groups: (1) members with integrated pharmacy coverage and (2) those covered by a pharmacy benefit carve-out. We then conducted a 1:1 match of integrated benefit and carve-out members on attributes in the baseline year of the data (2014-2015) and followed these members for comparison in the 3 following years (2015-2018). The details of the sample selection, matching, and comparisons follow.

Members had to have 48 months of continuous enrollment during the study period (October 1, 2014, through September 30, 2018) to be considered for inclusion in the analytic sample. All members were then required to have their primary coverage through BCBSLA, and their coverage had to be provided through a self-insured employer during all 4 years of the study. Additionally, we excluded members with missing data for key characteristics and members with extreme outlier medical spending values (the top 1%). Of the 565,385 members who met the initial criterion, a total of 77,354 (13.7%) members met the additional inclusion criteria (54,783 integrated members and 22,571 carve-out members). An attrition diagram detailing these steps appears in 

It is important to note that employers do not randomly choose to carve out pharmacy benefits; they may consider factors like recent spending and utilization trends among their employees when evaluating competing benefit designs. As a result, a naive comparison of integrated benefit and carve-out members may result in biased estimates of the effect of carving out pharmacy benefits. To account for this possibility, we used exact matching to create a matched cohort of integrated benefit and pharmacy carve-out members and followed them for 3 years post matching to compare both level and trend differences in medical spending. The matching procedure used member values for key variables during the first year (baseline year) of the study period, from October 1, 2014, through September 30, 2015. The matching criteria were selected to account for factors most likely to impact the need for medical care.

Regarding demographics and health status, we matched on sex, clinical diagnoses for chronic diseases that had the largest distributions in our study group (eg, diabetes, asthma, hypertension, and mental health conditions), and narrow bins of DxCG concurrent risk score. The DxCG risk score uses health care data to assign individual risk scores to patients for both concurrent and prospective diagnoses and estimates both retrospective and future expected health care payments for a commercially insured population. To account for the potential impact of other aspects of members’ health insurance coverage, we also included the type of insurance plan (health maintenance organization vs preferred provider organization) and insurance market region in the match. We combined all large urban areas into a single category for purposes of the matching procedure. Other parishes and out-of-state addresses remained as separate categories.

Additionally, we matched on indicators for member or provider enrollment in 1 or more of BCBSLA’s care coordination and management programs intended to reduce costs and utilization. These include Quality Blue Primary Care, a provider program similar to the patient-centered medical home concept, emphasizing population health measures and quality improvement

; Quality Blue Value Partnership, a shared savings program that builds on the primary care program

; and/or the BCBSLA care management program. Overall, the matching approach was chosen to reduce the risk of bias from baseline differences between members with integrated benefits and pharmacy carve-outs, while minimizing the loss of sample in achieving common support. This procedure yielded exact matches for 20,027 of the members with integrated pharmacy benefits. Within the matched sample, we repeated our comparisons using only members with selected chronic illnesses, which yielded samples of 7904 members in the integrated benefit group and 7926 members with pharmacy carve-outs. All analyses were conducted in SAS Enterprise Guide version 7.1 (SAS Institute).

 provides descriptive statistics on the integrated benefit and pharmacy carve-out cohorts before and after matching. Before matching, we saw qualitatively large and statistically significant differences across all categories. The carve-out group was older, more likely to be female, more likely to live in a large urban region, and in better health. These members were also more likely to be enrolled in the Quality Blue programs and less likely to be enrolled in care management programs. After matching, however, we saw that these differences were no longer significant except for age, which was not explicitly included in the matching.

We considered 2 features of costs to assess the relationship between medical costs and the provision of an integrated benefit or pharmacy carve-out. First, in the figures and tables that follow, we plotted the average 

 of per-member per-month (PMPM) medical spending in our matched cohort. Second, we compared the difference in the 

 of spending during the study years to assess whether the use of either benefit design was associated with bending the cost curve.

The numbers plotted alongside the medical spending curves reflect the average medical PMPM spending. In the tables below 

, the first 2 columns express the difference between the pharmacy carve-out and integrated benefit groups and 

 tests that compared the level of spending. In the following columns, we present the change in spending compared with the baseline year for the 2 groups, as well as a third column (Δ trend) that expresses the relative difference between the 2 groups, compared with the gap between them at baseline.

The results for the entire matched cohort are presented in Figure 2. In the baseline year, we saw a small, statistically insignificant difference in average medical spending: only $1.90 PMPM. This is to be expected, as our matching approach exactly matches on small bins of health risk, as well as selected chronic illnesses and geographic region, and this result suggests that our matching strategy was successful in accounting for important drivers of spending. In the first follow-up year, we found that both groups had experienced an average PMPM increase in medical spending of approximately 22% compared with the baseline year, but it was not statistically significant. In the second and third follow-up years, however, we observed some divergence between the 2 groups. Neither the level of medical spending nor the trend in spending growth was significantly different in any year, although the integrated benefit members did experience lower levels of spending and slower growth by the end of the study period.

The results presented in Figure 2 convey the average differences in spending across the full matched cohort. Although the overall association between pharmacy benefit structure and medical spending was not significant, different relationships may exist among important subpopulations within this group. In particular, it is essential to consider the effects of benefits design on patients with chronic illnesses. Although roughly half of American adults have a chronic condition, this group accounts for 86% of health care costs.

 Additionally, these costs may be modifiable with improved coordination and other delivery system improvements.

 With these facts in mind, we also considered the growth of PMPM medical spending among patients with selected chronic illnesses. These results are presented in Figure 3.

In the baseline year, the integrated benefit group had slightly higher spending, although it was not significantly different from the pharmacy carve-out group. Additionally, spending levels were not significantly different in any of the follow-up years. Turning to spending trends, as we observed in the full-cohort comparisons, the growth rate of spending was not significantly different moving to the first year of follow-up. However, the spending gap between the 2 groups narrowed by the first follow-up year, and we saw significant differences in the relative rate of growth in PMPM costs by the second and third follow-up years. By the last year of the study, the pharmacy carve-out members had experienced an additional $64.20 in PMPM medical spending growth—a difference of nearly 17% of the baseline spending level. By the final year of the study, average medical PMPM spending was $588 in the carve-out group, compared with $537 in the integrated benefit group.

Since their introduction in the 1980s, pharmacy benefit carve-outs have been the subject of vigorous debate. Most recently, these controversies have typically focused on the impact of PBMs on drug spending, such as the pricing models used by PBMs and the kinds of data transparency that employers might need to negotiate with them effectively.

 Our analysis took a different approach to evaluating the impact of pharmacy carve-outs relative to integrated benefits. Here, we considered whether outside contracting for a subset of health care services would lead to an inefficient provision of care and, consequently, increased costs outside prescription drug expenditures.

After matching patients on demographics, health status, and a variety of health insurance benefit characteristics, we found that overall, BCBSLA members covered by an integrated pharmacy benefit did not experience significantly different levels of medical spending or trends in spending growth compared with members covered by a pharmacy benefit carve-out. However, when we restricted our analysis to members with selected chronic illnesses, we found significant reductions in medical spending growth among those with an integrated pharmacy benefit. These benefits appeared to materialize over time, with no significant differences in the first comparison year but significant differences in the second and third year of postmatch follow-up. These results suggest that, although the cost of contracting for prescription drug benefits is an important topic for employers, brokers, government leaders, policy makers, and other stakeholders, the effect of benefit design on “whole-person” care, not just the services delegated to a carve-out, must also be considered—particularly for patients with chronic medical conditions.

Several possible mechanisms could explain the results presented here. For example, the employer or health plan with integrated pharmacy benefits, armed with new access to data on prescription drug utilization, could have observed certain physician prescribing behaviors or trends in member medication adherence and responded with increased engagement or education campaigns. Pharmacy data could have also informed risk stratification and led to better targeting of disease management programs and other programs. For example, New York state switched from a carve-out to a carve-in approach in 2011, and its costs since implementing the policy change showed a 21.4% savings on managed care organization–paid prescriptions compared with the rest of the nation during federal fiscal year 2019 by using a pharmacy carve-in model.

 We do observe significantly larger increases in care management program enrollment among members with an integrated benefit. Although both cohorts began at 3.1% enrollment, this figure had risen to 20.2% in the integrated benefit group, but only 19.4% in the carve-out group, by the third follow-up year. However, definitively resolving the question of what drove the reduction in medical spending growth is beyond the scope of this paper. Recent findings by Parekh et al showed a similar increase in health management program enrollment and decrease in PMPM spending among carve-in members, breaking down utilization and cost differences in further detail.

 More research attempting to explain the mechanisms by which the provision of an integrated benefit or carve-out can affect health care utilization across the spectrum of services is needed.

These analyses have several limitations. Our work considered only impacts on medical spending and lacked access to prescription drug claims. In many ways, this limitation is indicative of one of the challenges posed by pharmacy benefit carve-outs: Data on the outcome of interest (prescription drug spending) were not available for half of our matched cohort. As a result, our analyses should be interpreted as presenting an important caveat to any assertions that carve-outs reduce spending, as changes in spending on prescription drugs may tell only part of the story.

Another point to consider is the potential generalizability of the findings. Although we included members employed in a variety of industries and regions, Louisiana is our sample, so our results reflect only 1 state. The study population is also young, with low rates of comorbidities. Additionally, our analyses considered only self-insured employers—firms facing different financial risks in their insurance contracts may have approached these benefit design decisions differently. Finally, our results should be interpreted as describing an association between integrated benefits and medical spending growth rather than making strong causal claims. It is important to point out that unobserved heterogeneity in contracting for carve-in vs carve-out at the employer level may affect our results. Therefore, we are careful not to make causal statements. Although our matched longitudinal study was able to account for the effects of a variety of observable member and health plan attributes, we could not account for potential unobserved differences at baseline (like smoking status) or endogenous choice of benefit design by the employers.

The design of an employer-sponsored health insurance benefit can have significant impacts on the cost and quality of care that covered individuals receive. One particularly important decision is whether to offer a comprehensive, integrated benefit or to carve out certain aspects of coverage like pharmacy benefits. Our results show that members with selected chronic illnesses covered by integrated pharmacy benefits experience slower medical expenditure growth over the long term, resulting in annual PMPM savings compared with members covered by a pharmacy carve-out. We also see that the integrated benefit and carve-out groups appear to be on gradually diverging trends with respect to rising PMPM medical spending, with the integrated benefit group on a slower trajectory. Although neither the level nor trend differences were significant by the end of our study, the gap does appear to be growing over time. As noted earlier, these analyses consider only medical spending and exclude the costs of prescription drugs, so we cannot definitively say that the integrated benefit approach reduces net spending. However, the results do suggest that any potential savings resulting from reduced drug spending promised by a pharmacy benefit carve-out should be weighed against potential increases in medical spending.


 Blue Cross and Blue Shield of Louisiana (EL, ML, JO, BV, MF, BPK, SN), Baton Rouge, LA.

 Blue Cross and Blue Shield of Louisiana.

 All authors are employees of Blue Cross and Blue Shield of Louisiana, which offers integrated medical pharmacy benefits to group customers.

 Concept and design (EL, ML, JO, MF); acquisition of data (EL, ML, JO, MF, BPK); analysis and interpretation of data (EL, ML, JO, MF); drafting of the manuscript (EL, ML, JO, MF); critical revision of the manuscript for important intellectual content (EL, ML, JO, BV, MF, SN); statistical analysis (EL, ML, JO); obtaining funding (BPK); administrative, technical, or logistic support (BV, SN); and supervision (BV, BPK, SN).

 Emanuel Lucas, MPH, Blue Cross and Blue Shield of Louisiana, 5525 Reitz Ave, Baton Rouge, LA 70809. Email: Emanuel.lucas@bcbsla.com.

1. Aitken M, Berndt ER, Cutler D, Kleinrock M, Maini L. Has the era of slow growth for prescription drug spending ended? 

. 2016;35(9):1595-1603. doi:10.1377/hlthaff.2015.1636

2. Hodgkin D, Horgan CM, Garnick DW, Merrick EL, Goldin D. Why carve out? determinants of behavioral health contracting choice among large U.S. employers. 

. 2000;27(2):178-193. doi:10.1007/BF02287312

3. Ali ÖG, Mantrala M. Pharma rebates, pharmacy benefit managers and employer outcomes. 

. 2010;13(4):281-293. doi:10.1007/s10729-010-9129-4

4. Cook A, Kornfield T, Gold M. The role of PBMs in managing drug cost: implications for a Medicare drug benefit. Kaiser Family Foundation. January 2000. Accessed December 4, 2019. https://www.kff.org/wp-content/uploads/2013/01/the-role-of-pbms-in-managing-drug-costs-implications-for-a-medicare-drug-benefit.pdf

5. Frank RG, Garfield RL. Managed behavioral health care carve-outs: past performance and future prospects. 

. 2007;28:303-320. doi:10.1146/annurev.publhealth.28.021406.144029

6. Atlas RF. The role of PBMs in implementing the Medicare prescription drug benefit. 

. 2004;23(suppl 1):W4-504–W4-515. doi:10.1377/hlthaff.W4.504

7. Shrank WH, Porter ME, Jain SH, Choudhry NK. A blueprint for pharmacy benefit managers to increase value. 

8. Werble C. Health policy brief: pharmacy benefit managers. 

. September 14, 2017. Accessed December 4, 2019. https://www.healthaffairs.org/do/10.1377/hpb20171409.000178/full/

9. Goldman W, McCulloch J, Sturm R. Costs and use of mental health services before and after managed care. 

. 1998;17(2):40-52. doi:10.1377/hlthaff.17.2.40

10. Ma CA, McGuire TG. Costs and incentives in a behavioral health carve-out. 

. 1998;17(2):53-69. doi:10.1377/hlthaff.17.2.53

11. Frank RG, McGuire TG. Savings from a Medicaid carve-out for mental health and substance abuse services in Massachusetts. 

. 1997;48(9):1147-1152. doi:10.1176/ps.48.9.1147

12. Lyles A. Pharmacy benefit management companies: do they create value in the US healthcare system? 

. 2017;35(5):493-500. doi:10.1007/s40273-017-0489-1

13. Menges J, Batt J, Mouna A, Pantaleo N, Singh P. Comparison of Medicaid pharmacy costs and usage in carve-in versus carve-out states. AHIP. April 2015. Accessed June 15, 2020. https://www.ahip.org/wp-content/uploads/2015/11/Medicaid-Pharmacy-Carve-In-Final-Paper-The-Menges-Group-April-2015.pdf

14. Smith-McLallen A. Effects of pharmacy benefit carve-in on utilization and medical costs: a three-year study. 

. 2012;49(2):1-6. https://www.ifebp.org/inforequest/0161568.pdf

15. Bai G, Socal MP, Sharp M, Pachman ES. Medicaid managed care programs’ contracts for generic drugs are inefficient. 

. May 1, 2019. Accessed June 15, 2020. https://www.healthaffairs.org/do/10.1377/hblog20190426.775617/full/

16. Zhao Y, Ash AS, Ellis RP, et al. Predicting pharmacy costs and other medical costs using diagnoses and drug claims. 

17. Rae M, Cox C, Sawyer B. What are recent trends and characteristics of workers with high drug spending? Peterson-KFF Health System Tracker. September 27, 2018. Accessed June 15, 2020. https://www.healthsystemtracker.org/chart-collection/recent-trends-characteristics-workers-high-drug-spending-2016/

18. Martin AB, Hartman M, Washington B, Catlin A. National health care spending in 2017: growth slows to post-great recession rates; share of GDP stabilizes. 

. 2019;38(1):96-106. doi:10.1377/hlthaff.2018.05085

19. American Academy of Family Physicians. Joint principles of the patient-centered medical home

20. Shi Q, Yan TJ, Lee P, et al. Evaluation of the Quality Blue Primary Care Program on health outcomes. 

21. Yuan X, Guh S, Hochheiser D, Mayo T, Nigam S. Incremental effect of Quality Blue practices on disease management in Louisiana. 

. 2018;21(suppl 1):S136-S137. doi:10.1016/j.jval.2018.04.916

22. Chapel JM, Ritchey MD, Zhang D, Wang G. Prevalence and medical costs of chronic diseases among adult Medicaid beneficiaries. 

. 2017;53(6 suppl 2):S143-S154. doi:10.1016/j.amepre.2017.07.019

23. Bodenheimer T, Wagner EH, Grumbach K. Improving primary care for patients with chronic illness: the chronic care model, part 2. 

. 2002;288(15):1909-1914. doi:10.1001/jama.288.15.1909

24. Bai G, Socal MP, Anderson GF. Policy options to help self-insured employers improve PBM contracting efficiency. 

. May 29, 2019. Accessed June 15, 2020. https://www.healthaffairs.org/do/10.1377/hblog20190529.43197/full/

25. The Menges Group. Why New York should maintain its Medicaid pharmacy carve-in approach. New York Health Plan Association. October 1, 2020. Accessed November 1, 2020. https://nyhpa.org/wp-content/uploads/2020/10/Menges-Rx-Carve-Out-Report-10-1-20-1.pdf

26. Parekh N, Papa S, Drnach A, et al. Effect of carving in pharmacy benefits on utilization and costs. 

. 2020;26(10):1317-1324. doi:10.18553/jmcp.2020.26.10.1317

Articles

Members covered by an integrated pharmacy benefit (as opposed to a pharmacy carve-out) experienced slower growth in medical spending.

Reduced Medical Spending Associated With Integrated Pharmacy Benefits

The impact of primary care practices adopting the patient-centered medical home (PCMH) model is analyzed by comparing per member per month (PMPM) costs and utilization among commercial HMO members with chronic illnesses in PCMH and non-PCHM practices in the Philadelphia area. Transforming primary care practices to conform to the PCMH model has shown early promise in reducing costs and improving outcomes, and chronically ill patients’ frequent contact with the healthcare system and costly care make them ideal targets for such health system reforms.

The impact of the PCMH model on PMPM costs was analyzed using a generalized linear regression model to adjust for age, gender, and baseline cost. The impact of the PCMH model on utilization per 1000 rates was analyzed with the Poisson regression model, adjusting for baseline differences in age, gender, and risk score.

After accounting for differences at baseline, PCMH practices achieved lower total, inpatient, and specialist PMPM costs, as well as lower relative utilization of hospital admissions and specialist visits.

These findings suggest that policy makers should maintain or expand incentives to adopt PCMH reforms and that targeting chronically ill patients may be the most effective way to leverage the benefits of PCMH adoption.

Chronically ill patients enrolled in nonpediatric primary care practices that adopted the patient-centered medical home (PCMH) model had lower total, inpatient, and specialist per member per month costs compared with non-PCMH practices after adjusting for baseline characteristics.

		PCMH cost reductions appear to have been driven by lower utilization rates of hospital admissions and specialist visits.

		Inpatient admissions per 1000 patients were significantly lower for chronically ill patients in PCMH practices in each of the 3 program years analyzed.

Despite extensive evidence that quality primary care has the effect of improving health and controlling costs,

 the field is described as facing a crisis. One oft-cited source of the problem is the current fee-for-service reimbursement system, which does not adequately compensate the “cognitive” services that lie at the core of primary care.

The American College of Physicians (ACP) position paper distinguishes “cognitive” from procedural-driven services, and refers to the evaluation and management of patients by primary care physicians.

 The fee-for-service reimbursement model is credited in part with driving medical school graduates to enter better compensated specialty fields, creating the potential for a shortage in primary care practitioners.

Other concerned voices note that beyond the debate about an adequate supply of doctors, there is a need for primary care practitioners to take a more proactive role in the coordination of care across multiple sites and providers.

Poor coordination of care, as can occur when a patient lacks a usual source of care, has been shown to lead to increased costs, diminished care quality, reduction of access to preventive services, and increased usage of high-intensity care.

This relationship is of particular concern for the treatment of chronically ill patients. Primary care doctors are encountering increasingly complex patients, with the average family practice office encounter addressing 2.7 medical problems.

This complexity has very real implications for costs in the healthcare system: one study found that nearly all of the cost growth from 1987 to 2002 occurred among patients being treated for 5 or more conditions.

 Another noted that among Medicare beneficiaries, 10% of patients—typically those with multiple chronic illnesses that require numerous prescription medications, and who experience frequent hospitalization&mdash;account for 70% of healthcare costs.

The presence of chronic illness makes accurate, readily accessible medical records essential to providing quality care, and effective management may require ongoing patient contact, even outside of the doctor’s office.

The patient-centered medical home (PCMH) model offers a promising alternative to the current paradigm in primary care. The concept of the “medical home” originated in 1967 with the American Academy of Pediatrics (AAP), was initially presented as a model for addressing special healthcare needs among children. The medical home focuses on a team-based approach to primary care whereby a physician-leader coordinates care received from other providers across multiple sites and specialties. This methodology encourages increased access in terms of expanded practice hours and enabling new ways to engage with patients.

Information technology, such as electronic health records, is part of this approach, which can facilitate better identification of patient needs, more effective care management, and more efficient tracking of health outcomes. Incentivebased payment structures to better reward evaluation and management are also included. This concept gained wider attention following the 2007 statement of Joint Principles by the American Academy of Family Physicians, AAP, ACP, and American Osteopathic Association, which offered a unified vision of the central components of the PCMH model.

Some preliminary research has shown promising, though mixed, early results for practices adopting the PCMH model in terms of patient experience

 and in reducing utilization of high-cost medical interventions, such as emergency department (ED) visits and hospitalizations.

 This was corroborated by a recent review of studies of the PCMH model which showed that while individual programs have yielded successes, the overall evidence is more mixed. Adoption of the PCMH model appears to have positive effects on patient experience and provision of preventive services to patients, but its effect on ED utilization, an important driver of healthcare costs, was limited to older patients. Moreover, this review found no consistent reduction in hospital admissions or overall expenditures for patients in practices that adopted the PCMH model.

Given inconsistent definitions of the “medical home,"

 this review and others point to the difficulty in assessing the impact of the PCMH model. As much of the work to date has been inconclusive, this analysis will focus on the relevant subpopulation of chronically ill patients for whom the PCMH model would be most likely to yield benefits.

This paper aims to contribute to the growing body of literature and evidence on the PCMH model. Previous work, such as the 2010 Group Health Cooperative (GHC) study by Reid et al

 study on the Pennsylvania Chronic Care Initiative (PACCI) have evaluated the effects of the medical home on costs and utilization for all patients. This study focuses on the PCMH effect on patients with chronic illness(es). This subpopulation requires complex care and frequent contact with the healthcare system, which makes them more likely to benefit from improved coordination of care and enhanced access and communication.

 We hypothesize that implementing PCMH practices will result in lower total, inpatient, and specialist per member per month (PMPM) costs, and will lower relative utilization of hospital admissions and specialist visits over the 3-year study period.

This study evaluates early implementation of the PCMH model using data collected from the PACCI. This initiative was convened as a collaborative effort by the state’s largest commercial health plans (led by Independence Blue Cross), all 3 Medicaid managed care plans, and 32 medical practices, and was organized by the Governor’s Office of Health Care Reform to promote the adoption of the PCMH model and improve the quality of primary care for patients with chronic illnesses. Engagement by multiple stakeholders allowed for the appropriate resources, scale, and support to be provided, creating a framework for education, training, and data submission, as well as allowing for consistency in monitoring outcomes across the program.

To facilitate transition to the PCMH model, practices received supplemental financial incentives and were required to participate in a Learning Collaborative based on Wagner’s Chronic Care Model,

 working toward recognition from the National Committee for Quality Assurance (NCQA) as a PCMH by the end of the first year of participation. Each practice enrolled in the initiative was required to a) have at least 1 physician, plus a practice support leader, in attendance at each of 7 full-day sessions and to b) participate in weekly meetings. Furthermore, practices were paired with practice coaches responsible for providing additional support and training.

Data from practices in the PACCI pilot were used to determine the program’s effect on patients with chronic illness(es), a subpopulation requiring complex care. This was a longitudinal observational study conducted to assess the impact of adopting the PCMH as a model for improving primary care, focusing on PMPM healthcare costs and utilization per 1000 patients over 3 years of follow-up data.

A cohort of chronically ill members—defined as patients having asthma, coronary artery disease, congestive heart failure, chronic obstructive pulmonary disease, diabetes, and/or hypertension&mdash;was created from administrative medical claims in the baseline year, 2008. Chronic illness was identified using 

International Classification of Diseases, Ninth Revision, Clinical Modification 

) codes from medical claims data. Additionally, all patients must have been continuously enrolled in the same practice from 2008 through 2011. We also limited our analysis to practices located within Philadelphia city limits in order to limit the effect of urban-suburban disparities on our findings

 and because the majority of initial PCMH practices were located within the city.

Defining the Patient-Centered Medical Home

To provide an operational definition of a “medical home,” the PACCI relied on the definition provided by the NCQA as of 2008.26 Practices were assigned PCMH status if they received at least Level 1 NCQA recognition in 2009 and maintained recognition through at least 2011. Level 1 recognition required that practices meet 5 of 10 possible “must pass” criteria and accumulate 25 total “points.” The complete list of potential PCMH improvements recognized by the NCQA appears in

. Practices designated “non-PCMH” did not join the PCMH program at any point during years 2008 to 2011 and were used as controls. The resulting study population consisted of 2674 patients enrolled in 11 PCMH practices and 19,546 patients in 393 non-PCMH practices, all of which were located within Philadelphia city limits.

This study was conducted as a longitudinal observational study to best assess the impact of adopting the PCMH model as a vehicle for improving primary care of chronically ill members. The outcomes of interest were PMPM healthcare utilization and costs. Costs were reported separately as inpatient, ED, specialist, and outpatient care costs, as well total PMPM costs. Utilization was reported in terms of inpatient admissions, ED visits, specialist visits, and outpatient visits per 1000 patients over 3 years of follow-up. Data on costs and utilization were extracted from medical claims; patient-level covariates of age and gender were recorded from enrollment data. To control for differences between the PCMH and non-PCMH groups at baseline, patient-level covariates (age, gender, and costs at baseline) were included in regression models.

At baseline, a number of differences were observed between case and control practices in terms of patient age and risk profile, socioeconomic characteristics, and comorbid conditions, which translated into significant differences in terms of costs and utilization even before the introduction of PCMH changes in the cases. The Figure presents some of these baseline comparisons. Patients in practices adopting PCMH reforms were generally younger (aged 39.2 years vs 45.4 years; 

 <.0001), but had lower estimated risk scores (1.61 vs 1.81; 

<.05) and significantly different case mix in terms of illness profiles.

To control for variation at baseline, statistical models were employed to generate adjusted figures for the program effects on costs and utilization for each of the follow-up years. PMPM costs were adjusted using a generalized regression model, which controlled for age, gender, and costs at baseline. For utilization per 1000 patients, Poisson regression models were used to estimate utilization counts after adjustment for baseline differences in age, gender, and risk score. These methods parallel those used in the GHC study. Risk scores were calculated using Verisk Health DxCG Risk Solutions version 3.

 a commercial population risk adjustment package. The risk score used patient age, gender, and claims information from all medical encounter and enrollment data to estimate annual total medical resource use. Once adjusted cost and utilization figures were imputed for each of the program years, differences between PCMH practices and controls were compared using the t test for dependent means, using Satterthwaite or unequal variances. The results appear in 

Controlling for baseline differences, PCMH practices maintained significantly lower utilization for hospital admissions (

<.01) for each year in the follow-up period. PCMH practices also saw 0.3 fewer admissions per patient in 2009, and 0.2 fewer admissions per patient in both 2010 and 2011. Specialist visits were reduced by 12.3 visits per 1000 patients in 2009, and by more than 10 visits per 1000 patients in 2010 and 2011. However, PCMH practices observed significantly higher utilization in ED and outpatient visits, though the adjusted difference in ED visits shrank over the period from 2009 to 2011 (Table 2).

Predictably, the observed differences in utilization translated into lower adjusted total PMPM costs. Total costs were significantly lower in PCMH practices during all 3 follow-up years (

<.05). Relative to baseline, overall PMPM costs were $16.50 lower in 2009, a difference of 5.5%. Total PMPM costs were $13.00, and $13.70 lower in subsequent years as well. This reduction was driven by significantly lower inpatient (

<.0001) costs among PCMH practices over all 3 program years. The relative reduction in specialist costs was particularly pronounced: in 2009, adjusted costs for PCMH were 17.5% lower than those in non-PCMH practices. While significant relative increases in ED PMPM costs (

<.0001) partially offset these reductions, PCMH practices did not experience a significant increase in outpatient costs despite the observed increase in outpatient utilization (Table 3).

This study adds to the growing body of literature, including prior work by the GHC team, which indicates that the initial investments required for primary care practices to adopt PCMH reforms can yield successes in controlling growth in cost and utilization of high-cost, high-intensity medical services.14 GHC is a nonprofit health insurance and care delivery system based in Seattle. That study looked at cost and utilization results between 12 and 21 months after implementation in an attempt to track progress in quality improvement, in creating a sustainable work environment, and in lowering costs. The data source on utilization and costs was the GHC’s costing system, which allocates use and costs for all services provided at GHC facilities and from external claims. Utilization and costs from 7018 continuously enrolled adults at the prototype clinic were compared with those of 200,970 adults enrolled at other clinics in the Puget Sound area.

Three particularly important differences exist between the work by the GHC and this study: this study explores only cost and utilization, without commenting on patient or provider satisfaction; the end points do not match exactly (the GHC study measured cost and utilization after 12, 18, and 21 months, whereas the Pennsylvania study measured results annually over 3 years). Most importantly, this study focuses on chronically ill patients, whereas GHC analyzed all patients. Data comparing the GHC study results with this study’s results appear in 

. Variables discussed include differences in inpatient admissions, ambulatory care utilization, and ED and specialty care costs and utilization. For this study, there is also discussion on possible reasons for gaps in cost and utilization between PCMH and non-PCMH practices.

Regarding inpatient admissions, the GHC study showed that differences in all-cause inpatient admissions did not reach significance until the 21-month follow-up, whereas the differences in inpatient admissions per 1000 patients were significant in all 3 program years of the Pennsylvania study. This gap between the 2 studies, in terms of utilization, was also reflected in PMPM costs: the GHC study found a consistent relative reduction of more than $10 PMPM in total costs, but this did not meet statistical significance at α = .05. By comparison, we found statistically significant relative reductions amounting to $16.50 PMPM in the first program year and $13.00 and $13.70 PMPM in subsequent years.

In the case of ambulatory care—sensitive treatment, the medical home practices in the GHC study were able to achieve significant relative reductions in admissions when the analysis was limited to inpatient admissions for ambulatory care–sensitive conditions (

<.001). By controlling for the presence of chronic illness—either by limiting the population to such patients, as in the case of the Pennsylvania study, or by framing the outcomes to focus only on treatment for chronic illness&mdash;significant differences become apparent in the outcomes achieved by PCMH and non-PCMH practices. This suggests that, with respect to costs and utilization, the PCMH may lead to meaningful improvements only when applied to relevant subpopulations, such as chronically ill patients.

In the Pennsylvania study, the increased utilization of outpatient care is actually suggestive of further success for the PCMH model. By improving coordination of care, doctors may have been appropriately directing their patients to lower-cost, lower-intensity services, which acted as substitutes for costlier hospital admissions and other services. This finding is especially promising, given that the increased utilization in terms of visits per 1000 patients did not lead to a significant increase in PMPM costs. Similar findings of increased primary care intensity leading to lower costs via substitution have been reported elsewhere.

Also, Wang et al report using a case-control matched longitudinal cohort study in patients with type 1 and type 2 diabetes mellitus, and found that adoption of the PCMH reduced overall medical cost for diabetic members by 21% in the first year.29 This result largely reflects a drop in inpatient costs, which fell by 44%.

The 2 studies also report disparate findings concerning costs and utilization of specialty care. Patients enrolled in the GHC study’s PCMH practices exhibited fewer ED visits and more specialty care visits, whereas the opposite was true of the Pennsylvania PCMH patients. This difference may be attributable in part to the Pennsylvania study’s focus on chronically ill patients. For example, increasing focus on patient needs and satisfaction may identify unmet healthcare needs across all patients, whereas among patients with chronic illness who are likely to have frequent and ongoing contact with the healthcare system, improved disease management may reduce the need for specialized care. Alternatively, physicians in the Pennsylvania managed care practices may be under-referring patients to specialist services, but these differences in specialty care may simply reflect regional variation in practice.

Finally, with respect to the Pennsylvania study, it is worth noting that in most cases, the gap in cost and utilization between PCMH and non-PCMH practices was seen to decrease over the 3 follow-up years. Although the control practices were required to have completed all 3 program years without formally receiving NCQA recognition, it is possible that some of the innovations adopted as a package by PCMH practices were embraced piecemeal by non-PCMH practices during follow-up years, leading to a gradual flattening of the results across practice type. One possible explanation stems from the fact that non-PCMH practices tended to be in the medium- / average-practice size category, whereas PCMH practices were typically larger group practices. The scale advantages of larger practices might have made information technology or staffing improvements more affordable, with average-size practices lagging behind on the trend. Previous research has indicated that larger practice size is positively correlated with the presence of elements of the PCMH infrastructure.

Additionally, because the PCMH practices began adopting elements of the NCQA guidelines before obtaining formal recognition, the process of improving outcomes may have started ahead of the medical home pilot program. This self-selection bias may have reduced the apparent impact of PCMH adoption.

The PCMH impact may extend to smaller practices as well. A recent PCMH study that focused on small, independent primary care practices in Rhode Island found that 5 pilot practices increased their NCQA recognition score from 42 to 90 points over a 2-year study period.

 There was a significantly lower rate of ambulatory care— sensitive ED visits in the pilot practices compared with 34 nonpilot practices.

Both the GHC experience and the results of the PACCI study suggest that the introduction of medical home innovations into primary care practices may help to curb utilization of high-intensity medical services as well as overall cost growth, despite initial investments in staffing changes and technology infrastructure. The differences in results observed when analyzing the primary care population as a whole, as opposed to the chronically ill cohort tracked in the Pennsylvania study, stresses the importance of quality risk stratification of patients, not only for research purposes, but also for making fair assessments of practice quality and for targeting of resources by medical practices.

A study by the Commonwealth Fund has suggested that with the exception of information technology infrastructure, incremental costs do not rise with increasing “medical homeness” of primary care practices.

 However, both the GHC medical home practices and those in the Pennsylvania pilot made staffing changes as part of their transition, adding physicians and nurse practitioners to allow enhanced patient contact.

With this in mind, financial incentives to defray the costs of introducing care managers, electronic medical records, registries, and related improvements may be particularly crucial to promoting adoption of the medical home as a model. Cost savings may eventually make the PCMH self-sustaining, with cost savings redirected to improving compensation for primary care doctors5 and to other practice improvements.

Implications for Future Research and Limitations

Although we employed the same rigorous adjustments used by the GHC study, some differences at baseline were not specifically accounted for in the regression adjustments used. For example, while all patients in the Pennsylvania study satisfied the criterion of displaying some chronic illness or illnesses, PCMH and non-PCMH practices may have had a different case mix. Though patient-level variation was accounted for, at least in part, by use of the DxCG risk score in the regression adjustments, future research might consider identifying the most useful practice changes for specified conditions or assessing whether practice case mix drives the decision to adopt PCMH innovations.

Furthermore, observed differences may be driven in part by self-selection on the part of primary care practices; practices with a trajectory toward electronic health records, enhanced case management, and so forth may have been more likely to seek NCQA certification early in the process. Future work based on finer practice-level variation could account for these baseline differences and could also assess the relative impact of higher levels of NCQA certification and within-level variation in order to distinguish the contributions of various components of the PCMH model to improving care and controlling costs and utilization. One promising method for follow-up studies would be to pair difference-in-difference regression analysis with case-control matching in order to get equal samples from PCMH and non-PCMH practices, and to achieve improved control of practice-level differences such as size and affiliation.

This study joins a growing body of research which suggests that the PCMH model offers a promising mechanism for addressing primary care’s pressing concerns with respect to delivering quality care and facing potentially unsustainable cost growth. By supporting patient self-management and improving coordination of care, practices adopting the PCMH model were able to achieve relative decreases in cost and utilization at the highest levels of care intensity (inpatient admissions) by chronically ill patients. In light of these findings, policy makers and healthcare thought leaders should continue to emphasize the PCMH as a vehicle for improving primary care, particularly for patients with complex, chronic conditions.

The authors thank Susannah Higgins, Manu Tyagi, Sara Fritz, Qiuyan Cindy Wang, Aaron Smith-McLallen, and Pam Eckardt of Independence Blue Cross, Philadelphia, for their assistance in data collection, analysis, and review. The authors acknowledge the assistance of CTI Clinical Trial & Consulting Services, Inc, Cincinnati, a paid consultant to Independence Blue Cross, for their technical support and editing during the manuscript preparation.

 Independence Blue Cross (JN, RC, CMC, RLS, SN), Philadelphia, PA.

This study was funded by Independence Blue Cross, an independent licensee of the Blue Cross and Blue Shield Association.

A longitudinal case-control design was used to evaluate the effects of the patient-centered medical home model on medical costs and utilization among chronically ill patients.

Medical Homes: Cost Effects of Utilization by Chronically Ill Patients

 Evaluate the effects of the patient-centered medical home (PCMH) model on medical costs and utilization in the nonpediatric population, particularly among high-risk patients.

 Longitudinal case-control design, comparing per member per month (PMPM) cost and utilization per 1000 patients for members enrolled in PCMH and non-PCMH practices from 2009 to 2011.

 Commercial health maintenance organization members in nonpediatric practices that adopted the PCMH model in 2009 were matched to patients in nonpediatric practices that did not adopt the model until 2011 or later. Propensity score matching was used to identify a pool of similar controls, and difference-in-differences regression analysis was used to compare PCMH and non-PCMH patients relative to baseline. Analysis was conducted using the complete pool of matched patients (N = 6940 cases and 6940 controls), then using the 10% of patients with the highest DxCG risk scores (N = 654 cases and 734 controls).

 There were no significant cost or utilization differences for the overall population. Total cost decreased significantly more for the PCMH group than for controls in the high-risk group in years 1 and 2 (reductions of $107 and $75 PMPM), driven by lower inpatient costs. The PCMH group experienced a significantly greater reduction in inpatient admissions in all 3 years (61, 48, and 94 hospitalizations per 1000).

 PCMH practices had significantly reduced costs and utilization for the highest risk patients, particularly with respect to inpatient care. As high-risk members represent a high-cost group, the most benefit can be gained by targeting these members.

High-risk patients enrolled in nonpediatric primary care practices which adopted the patient-centered medical home (PCMH) model had significantly lower per member per month medical costs and utilization per 1000 members compared with non- PCMH practices after adjusting for baseline characteristics.

		Costs and utilization did not significantly differ between PCMH and non-PCMH groups among all patients, suggesting that the benefits of the PCMH model are concentrated among high-risk, high-cost patients.

		PMCH cost reductions appear to have been driven by lower rates of hospitalization, and total costs fell even though utilization of specialist care saw significant increases in the first 2 years.

The patient-centered medical home (PCMH) has been advanced as a promising framework for transforming primary care. In 2007, the American Academy of Family Physicians, American Academy of Pediatrics, American College of Physicians, and American Osteopathic Association issued the “Joint Principles of the Patient-Centered Medical Home,” which outlined the PCMH model. The medical home model emphasizes a team-based approach to primary care, in which a physician-leader coordinates care by other providers across multiple sites and specialties. It encourages increased access, both in terms of expanding practice hours and opening new channels of communication with patients. Organizations such as the Patient-Centered Primary Care Collaborative have initiated numerous pilot programs aimed at studying the impact of PCMH adoption,

 and the PCMH model was written into the Patient Protection and Affordable Care Act of 2010 as an area for study.

A number of previous studies have shown early promise for the PCMH model as a vehicle for controlling costs and improving the quality of healthcare delivered by primary care practices,

 including for targeting subpopulations such as children with special health needs.

 However, reviews often point to the incomplete nature of this work, citing methodological concerns,

 insufficient time for practices to implement reforms, and inadequate policy support beyond the level of individual practices.

 This study aims to contribute to this literature by comparing the effects of adopting the PCMH model on the healthcare cost and utilization in the nonpediatric population, using propensity score matching in order to reduce variability in the PCMH and non- PCMH groups studied. Additionally, the analysis employs difference-indifferences regression analysis in order to further control for remaining differences in patients’ characteristics as well as cost and utilization at baseline.

This study aims to assess the impact of PCMH adoption on the patients identified as having the greatest health risks. While the Joint Principles envision the PCMH model as being applicable to all patients, other pilots have targeted only high-risk patients with complex needs.

 The high cost of care associated with relatively few individuals makes such targeting a potentially powerful mechanism: one study noted that virtually all of the growth in expenditures for Medicare over the period from 1987 to 2002 occurred among beneficiaries with 5 or more chronic illnesses

A longitudinal case-control design was used to evaluate the effects of the patient-centered medical home model on medical costs and utilization among high-risk patients.

Somesh Nigam, PhD

Articles

Reduced Medical Spending Associated With Integrated Pharmacy Benefits

Medical Homes: Cost Effects of Utilization by Chronically Ill Patients

Medical Homes and Cost and Utilization Among High-Risk Patients