Accountable Care Organizations (ACOs) need to reconsider their provider configuration and make it capable of managing clinical and financial risk. To that aim, their management must decide which medical procedures are done by the ACO itself, and which are contracted out to market providers. Making this decision requires a balanced treatment of market and firm organization, recognizing that each has properties that can turn into relative strengths. Such a balanced treatment is lacking in the ACO debate. Using the transaction cost theory, we provide such a balanced treatment of market and firm organization, and discuss implications for the design of ACOs and accountable care initiatives in general.  

 Market and firm organization need a balanced treatment in the accountable care debate. The urgency of such a balanced treatment is 3-fold: 

		Care providers’ lack of knowledge on the workings of market and firm organization impedes the effective implementation of accountable care.

		Accountable care initiatives will not reach their full potential when there is a mismatch between the strategy of participating care providers and their contract management. Such a mismatch may occur when market and firm organization do not get a balanced treatment in decision making.

		Balanced treatment would allow for a more refined view of coexisting volume and value incentives that characterize many accountable care initiatives.

Health system reforms in several countries have changed accountability relationships in healthcare. Traditional payers of healthcare, like governments and health insurers, have now started sharing the accountability for healthcare costs and quality with care providers. The introduction of accountable care organizations (ACOs), as laid out in the Affordable Care Act and in regulations issued by the Centers for Medicare & Medicaid Services, will change accountability relationships in a similar fashion. ACOs are held financially and clinically accountable for the costs and quality of the entire spectrum of care provided to a defined patient population, and share in the savings or losses relative to a spending benchmark. Thus, ACOs bear a financial risk similar to that of traditional payers of healthcare. In order to manage this financial risk, ACOs will have to reconsider their provider configuration.

ACO provider configurations are commonly described in terms of the medical specialties involved,

 but attention must also be paid to how ACOs plan to organize their contractual relationships with these providers. Broadly speaking, ACOs can opt for either firm organization or market organization. When an ACO opts for firm organization, it extends its organizational boundaries to include a newly employed physician or medical discipline. A typical market contract is one in which physicians remain autonomous and are paid a fee-for-service (FFS).

When should ACOs prefer firm organization over market organization and vice versa? The question we refer to as the “boundary problem” has now become more pertinent as care providers—organized into ACOs&mdash;are held financially and clinically accountable for all care provided to their patient population.

The Unbalanced Treatment of Market and Firm Organization in the ACO Debate

A solution to the boundary problem requires a balanced treatment of market and firm organization and the recognition that each has properties that can turn into relative strengths. It is unfortunate that the ACO debate is therefore largely devoted to the anticompetitive properties of firm organization alone, as is evident from discussions on the employment of primary care physicians by hospital-led ACOs and physician self-referral.

 Now that it is recognized that providers’ lack of knowledge regarding organizational structures impedes the effective implementation of accountable care,

 a balanced treatment of market and firm organization is even more urgent.

: A Balanced Treatment of Market and Firm Organization

ACOs and individual providers maintain contractual relationships in a volatile market characterized by technological change

 Turnover in assigned patient populations, as will be experienced by ACOs participating in the Medicare Shared Savings program,

 will additionally increase this volatility.

Market volatility implies that contracts must continuously be adapted to the current landscape. To that aim, ACOs’ management will need to (1) remain informed regarding the care providers’ value as contract partners, (2) resolve contractual gaps, and (3) determine a form of reimbursement such that care providers are incentivized to act in the spirit of the contract. These 3 processes have to be organized in such a way that the resulting costs—“transaction costs”&mdash;are minimized. The transaction cost theory concerns itself with the efficient organization of contractual relationships. An efficient organization of contractual relationships is one that minimizes transaction costs. The theory distinguishes 2 different methods of organizing a contractual relationship: the price system and hierarchy. When an ACO opts for the price system, its care providers remain autonomous. The ACO will go by market prices when determining a care provider’s value and reimbursements are based on output valued at market prices. Under hierarchy, care providers transfer their right to make productive decisions to the ACO’s management in exchange for receiving an income independent of volume (eg, a fixed salary).

Markets and firms are economic institutions that use a different mix of organizing methods. Markets predominantly rely on the price system and firms on hierarchy.

 The differences between the price system and hierarchy have implications for transaction costs. To understand why, one must know the 2 human traits believed to lie at the bottom of these costs. The first trait is contract parties’ limited ability to collect and analyze all information necessary to make optimal decisions. The second trait is a tendency to opportunistically exploit these situations of imperfect knowledge. This can manifest in multiple ways. ACOs may need to confront supplier-induced demand and stinting on care by some of their contracted providers. Furthermore, the FFS reimbursement system which remains in place does not by itself incentivize individual providers to practice in congruence with an ACO’s policy of cooperation, task substitution, and evidence-based care.

To minimize transaction costs, one must curb these opportunistic behaviors. As the price system rewards for volume, it addresses stinting on care. However, this volume incentive of the price system sometimes gets in the way of effective care. This happens when a lack of transparency prevents ACOs from monitoring whether demand for care has been induced and fulfilled properly. Market prices then fail to reflect the value of the underlying medical service, which may incentivize providers to induce demand. A hierarchical organization of contractual relationships lowers this incentive because itcomplicates the relationship between the volume of services and income. A range of options is available—bundled payments, global payments, and traditional capitation all complicate the relationship between volume of services and income by acting as price ceilings. However, under these payment options, volume of services and income are still linked through the marginal costs of providing care which are borne by the providers themselves. Only under a fixed salary contract are providers’ incomes independent of volumes. The fixed salary these care providers receive also make them more willing to follow directions regarding cooperation and task substitution, as these activities will no longer have an impact on their income through a lower volume of services.

Although hierarchical organization addresses several perverse incentives that failing markets for medical services introduce, they may reintroduce stinting on care and on-the-job consumption. Depending on the medical service that is contracted, however, ACOs may be better able to monitor these behaviors rather than demand inducement.

Apart from the incentives resulting from different forms of reimbursement, transaction cost theory also discusses the differential effects of organizing methods on the information collection process and bargaining under different competitive environments. The theory’s contribution to the ACO debate would be that ACOs’ provider configurations be viewed in terms of the organizing methods they employ and that each organizing method has a different impact on care providers’ behavior and eventually, transaction costs. This has several implications.

ACOs’ provider configurations are made up of contractual relationships with care providers. When deciding on ways to organize these contractual relationships, ACOs must consider the incentives that each organizing method brings along. These incentives must be compatible with ACOs’ strategies for fulfilling their financial accountability. A strategy of preventive medicine requires a different way of incentivizing care providers than a strategy of task substitution. Evidence from The Netherlands, where organizations similar to ACOs called “care groups” have been held accountable for the cost and quality of chronic care as of 2007,

 suggests that this match between contractual organization and strategy is overlooked in practice.

 For example, care groups experienced difficulties realizing task substitution in fundus examinations for diabetic patients. Care groups wanted to shift this procedure from ophthalmologists (specialist care level) to optometrists (primary care level). However, because ophthalmologists were paid on the FFS-base by the care group, they were not incentivized to cooperate, as the shift would lower the number of patients that returned to them for an extensive eye examination. This example illustrates how ACOs could potentionally miss out on savings when contractual organization is not compatible with their cost-containment strategy.

The idea that the price system and hierarchy each have a different way of incentivizing parties to a contract further refines the discussion regarding coexisting volume- and value-based reimbursement systems that characterize so many accountable care initiatives. Berenson

 approaches this issue from the perspective of risk management and concludes that direct FFS payments compare favorably with uncertain future shared savings. A transaction cost approach refines this statement in 2 ways. First, ACOs will only collect FFS payments from their employed care providers. The ACO then maintains a hierarchical relationship with its care providers which may make it problematic to order these care providers—for whom volume and income may only be weakly related&mdash;to maximize volumes and profits. Second, patients of ACOs that do not extend their organizational boundaries will sometimes need to go outside the ACO’s network. For these ACOs the value incentives of shared savings&mdash;determined over the total spectrum of care&mdash;may overrule the volume incentives, which have partially leaked onto independent providers. Contractual organization at the physician level thus plays an important role when hypothesizing about whether the value incentives of shared savings will counter any parallel volume incentives of FFS payment system.

For each contractual relationship, ACOs must determine whether or not to extend their organizational boundaries. Apart from the lack of transparency previously discussed, the instances of market failure for which firm organization could provide relief are well documented.

 Viewing firm organization solely as a means to secure patient referrals (or to secure a shift to outpatient care under population-based payment)

 does not do justice to the economics of firm organization, the hierarchical organizing method on which it relies, and the instances of market failure it will address. An unbalanced treatment of market and firm organization in either policy, practice, or the public debate could therefore limit ACOs’ potential in successfully taking on clinical and financial accountability for patient outcomes.

From Tilburg University (AH, MJVDB, BRM), Tilburg, The Netherlands; Radboud University Nijmegen Medical Centre (GPW), IQ Healthcare, Nijmegen, The Netherlands.

The authors (APH, MJVDB, BRM, GPW) 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 (APH, MJVDB, BRM, GPW); drafting of the manuscript (APH); critical revision of the manuscript for important intellectual content (APH, MJVDB, BRM, GPW); provision of study materials or patients (APH); and supervision (MJVDB, BRM, GPW).

Arthur P. Hayen, MSc, PO Box 90153, 5000 LE Tilburg, The Netherlands. E-mail: a.p.hayen@uvt.nl. 1. McClellan M, McKethan AN, Lewis JL, Roski J. A national strategy to put accountable care into practice. Health Aff (Millwood). 2010;29(5): 982-989.

2. Kocher R, Sahni NR. Hospitals’ race to employ physicians--the logic behind a money-losing proposition. N Eng J Med. 2011;364(19):1790-1793.

3. Williamson OE. Comparative economic-organization: the analysis of discrete structural alternatives. Admin Sci Quart. 1991;36(2):269-296.

4. Hennart JF. Explaining the swollen middle: why most transactions are a mix of market and hierarchy. Organ Sci. 1993;4(4):529-547.

5. Cutler DM, McClellan M. Is technological change in medicine worth it? Health Aff (Millwood). 2001;20(5):11-29.

6. Fani Marvasti F, Stafford RS. From sick care to health care: reengineering prevention into the US system. N Eng J Med. 2012;367(10):889-891.

7. Medicare program. Medicare Shared Savings Program: Accountable Care Organizations. Final rule. Federal register. 2011;76(212):67802-67990.

8. Simon HA. Organizations and Markets. J Econ Perspect. 1991;5(2): 25-44.

9. Struijs JN, Baan CA. Integrating care through bundled payments: lessons from The Netherlands. N Eng J Med. 2011;364(11):990-991. 

10. Struijs JN, van Til JT, Baan CA. Experimenting with a bundled payment system for diabetes care in the Netherlands: the first tangible effects. 2010, National Institute for Public Health and the Environment: Bilthoven, the Netherlands.

11. Berenson RA. Shared savings program for accountable care organizations: a bridge to nowhere? Am J Manag Care. 2010;16(10):721-726.

12. Williamson OE. The vertical integration of production: market failure considerations. Am Econ Rev. 1971;61(2):112-123. 

Articles

TRENDS FROM THE FIELD

Accountable care organizations will be more successful in taking on accountability for patient outcomes when market and firm organization get a balanced treatment.

Accountable Care Organizations: How to Dress for Success

 Patient education is a compulsory element of the nationwide disease management program (DMP) for type 2 diabetes mellitus in Germany. However, a considerable proportion of patients do not attend diabetes self-management education courses. Therefore, the purpose of this study was to describe patient-reported reasons for nonparticipation and to identify typical subgroups of nonparticipants in order to improve recruitment strategies.

The authors performed a cross-sectional observational study on 165 participants and 132 nonparticipants in diabetes education using a postal survey and chart review.

 Participants and nonparticipants were compared using 2-sided t tests and χ

 tests. Nonparticipants were grouped by cluster analysis based on the reasons for nonparticipation.

 A total of 95% of participants and 36% of nonparticipants reported to have received a recommendation for diabetes education from their physician. The authors identified 4 typical subgroups of nonparticipants: the “informed and responsible,” the “unconcerned without desire for more information,” the “uninformed but responsible,” and the “anxious and burdened with  psychosocial problems and functional limitations.”

 The physician’s recommendation seems to influence participation in diabetes education and should be used intentionally to increase participation rates. Also, differentiating barriers of nonparticipants can be determined by the degree of feeling informed and responsible for diabetes management. Physicians should more clearly explore patients’ perception of their knowledge of diabetes and their attribution of responsibility for diabetes management. Starting from this patient perspective might help physicians motivate patients to participate in diabetes education.

A cluster analysis identified 4 subgroups of nonparticipants in type 2 diabetes mellitus education in order to improve recruitment strategies.

The subgroups differed in respect to the perception of their knowledge of diabetes and their attribution of responsibility for diabetes management (to the physician or themselves).

A strong association between the recommendation of the attending physician and participation in diabetes education was found.

Starting from the patient’s perspective, physicians should explore the patient’s perceptions of the disease and its treatment more clearly as well as the patient’s personal goals and social context to motivate them to participate in diabetes education.

Diabetes self-management education is considered to be an effective intervention in patients with diabetes mellitus. It has been shown to lead to better glycemic control

 and better psychosocial and health outcomes.

 Today, diabetes education is an important part of the national clinical practice guidelines4 and a compulsory element of the nationwide disease  management program (DMP) for type 2 diabetes mellitus (T2DM) in Germany.

 Nonparticipation in diabetes education is not  penalized. The success of diabetes education depends essentially on the willingness of the patients to participate.

 The researchers’ own previous work has shown that self-active and motivated patients with a lower risk of diabetic complications seem to be more likely to participate in the German DMP for T2DM. As a conclusion, patients with little self activity and motivation, who might have the greatest need for assistanceand the largest degree of room for improvement, may not be  reached by the program.

 Analogously it is probable that a similar group of patients with little self activity and motivation, as well  as deficits in diabetes knowledge and self-management fails to participate in diabetes education.

An evaluation from the federal state of North Rhine-Westphalia showed that 30% of DMP patients did not attend diabetes education—neither before nor after enrolment in the DMP.8 Other studies revealed a total participation rate in diabetes education in Germany of about 50%.

 Research in other countries, eg, the United States, indicates similar findings.

 Several studies have identified individual barriers for patients regarding diabetes education including time constraints, stress, transportation problems, functional limitations, no interest,

doubt of usefulness of the education program,

 Until now there has been little research concerning the identification of patient subgroups for whom diabetes education could be tailored,

 as it is not possible to provide individual education on a large scale.

Therefore, the aim of this study was to analyze sociodemographic and illness-specific differences between participants and  nonparticipants in diabetes education and to describe the reasons for nonparticipation. We intended to identify typical groups of nonparticipants in order to help improve recruitment strategies for diabetes education as well as adjust their contents to the needs of these groups.

The study was designed as a cross-sectional, observational study comparing participants in diabetes education with nonparticipants. Patients with T2DM were randomly selected and recruited from 30 primary care practices located in and around 3 larger German cities (Hamburg, Bremen, and Berlin). In each practice we created a list of all patients with the 

International Classification of Diseases, Revision 10

) diagnosis E11 to E14 in the electronic patient charts of the general practitioner (GP). Patients with type 1 diabetes (as reported by the GP) and patients 80 years and older were excluded. From this list we randomly selected a maximum of 25 patients in each practice and contacted them for written informed consent to participate in the study. In total, 586 eligible patients received an informed consent form and the postal questionnaire.  Nonresponders to the initial request received 2 additional invitations. In total, 297 patients (51%) agreed to participate and returned a completed questionnaire.

Data collection took place between April and November 2009. The study was carried out in accordance with the Code of Ethics of the Declaration of Helsinki and approved by the Ethics Committee of the Medical Association of Hamburg (reference number OB-024/07).

) the researchers used a standardized postal patient survey and chart review. Chart review was used to assess age, gender, the duration of diabetes, the treatment of diabetes with oral medication or insulin, glycated hemoglobin (A1C) values, systolic blood pressure (BP) values, and body mass index (BMI) of the patients. The patient survey consisted of other items including 3 standardized measures: (1) the PHQ-2,

 a depression screening with 2 items concerning feelings of loneliness and loss of interest rated on a 4 point Likert scale; (2) a 7-item version of F-SOZU,

 an instrument for assessment of perceived and anticipated social support rated on a 5-point Likert scale; and (3) education classified by the international CASMIN standard

in 3 groups: a) inadequately completed general education, general elementary education, or basic vocational qualification, b) intermediate qualification or general maturity certificate, and c) lower or higher tertiary education. The authors also included a self-developed standardized instrument with reasons for nonparticipation in diabetes education courses (12 items with dichotomous response options; see Figure 1) based onexisting evidence regarding barriers to attending diabetes education.

Nonparticipants in diabetes education were grouped by cluster analysis based on their reasons for not participating by the clustering algorithm “partitioning around medoids.”

 The researchers chose the cluster solution between 2 and 5 clusters that had the maximum mean similarity within the clusters as measured by the maximum average silhouette width. The Jaccard coefficient was chosen as similarity coefficient,

 since it is suitable for asymmetric binary variables. Missing values regarding reasons for nonparticipation in diabetes education were imputed before analysis. Seven nonparticipants with more than 50% missing values were excluded, leaving a total of 125 nonparticipants for cluster analysis.

Bivariate tests of significance (2-sided t test and χ

 test) were performed for assessment of differences between 1) participants and nonparticipants in diabetes education, and 2) clusters of nonparticipants. These analyses consisted of about 150 statistical tests. To minimize the number of false positive results an α-level of 0.1% (

.001) was defined as statistically significant. Cluster analysis was performed with S-PLUS 8.1; all other statistical tests were conducted with Stata 11.0.

The comparison of participants and nonparticipants in diabetes education is based on 165 participants and 132  nonparticipants. Participants were treated with insulin more often than nonparticipants (31.1% vs 12.2%) and a much greater proportion of participants had received a recommendation for diabetes education from their attending physician (95.1% vs 35.5%) compared with nonparticipants. No differences between participants and nonparticipants were found regarding age, gender, duration of diabetes, A1C, BP, BMI, proportion of patients treated with oral medication, depressive symptoms, social support, education, or income (see Table 1).

More than half of the nonparticipants in diabetes education indicated that they did not attend diabetes education courses because they considered their knowledge regarding  diabetes to be sufficient. Another 29% stated that they did not participate because they felt their attending physician was responsible for diabetes management, not they themselves, while 21% stated that they did not understand the purpose of diabetes education. Other reasons included negative feelings regarding education courses (21%) or groups of people (20%) and fear of excessive demands of diabetes education, eg, that the content would be too difficult  to understand (13%). Finally, nonparticipants also complained about a great regional distance to the training center (21%), physical demands of diabetes education (19%), and problems with vision or hearing (10%), as well as time constraints in connection with their occupation (15%), or family (10%) and financial contributions, eg, travel expenses (12%). The reasons  for nonparticipation are shown in Figure 1.

Cluster analysis was performed on 125 nonparticipants in diabetes education. The average silhouette width was at its maximum 0.36 in a 4-cluster solution (see 

). Descriptive data for the 4 clusters can be found in 

, and statistical significance of differences between the clusters is shownin the eAppendix (available at 

). There were  no differences between the clusters in sociodemographic or illness-specific data. The 4 clusters differed most in the 2 most frequently given reasons for nonparticipation. Patients in cluster1 completely agreed with “Knowledge of diabetes is considered sufficient” and completely disagreed with “Physician is responsible for diabetes management.” In contrast, patients in cluster 3  completely disagreed with both statements. Patients in clusters 2 and 4 largely agreed with both statements (73% and 100%  or cluster 2 and 83% and 85% for cluster 4). Cluster 4 is distinct from the 3 other clusters due to the significantlyhigher rate of  agreement with “Intended purpose of diabetes education is unclear” (79%), “Negative feelings regarding education courses” (85%) and “groups of people” (77%), and “Fear of excessive demands of diabetes education, eg, difficult contents” (67%). There is also a higher rate of patients with “Problems with vision or hearing” (57%) than in clusters 1 or 3. Patients in cluster 4 have the highest agreement rate (they agree with 8.3 of the reasons on average) and patients in cluster 3 have the lowest rate (0.7 reasons on average). The difference between patients from cluster 1 (2.2 reasons) and cluster 2 (2.7 reasons) is not statistically significant.

As a first step the authors compared participants with nonparticipants in diabetes education, yielding a strong association between the recommendation of the attending physician and participation in diabetes education, which has been reported before.

The correlation between physician recommendation and patients’ attendance in diabetes education might mean that the attending physician (who is the GP in most cases) strongly influences the patient’s decision. If this association can be confirmed, participation rates in diabetes education could be increased if physicians recommended it to more patients

 However, the information on the recommendation of the attending physician is based on patients’ self reports. It could be that patients who have not participated in an education course did not remember such a recommendation. It should be assessed in future research whether this finding can be confirmed in physician interviews and/or chart reviews.

Among participants, there is a greater proportion of patients who are treated with insulin than among nonparticipants. This may be explained by the fact that insulin therapy is more invasive than oral medication or lifestyle modification, and that diabetes education is more important for everyday self-management if patients are insulin dependent. From other research it is also known that many patients with T2DM underestimate the disease as long as they do not need to inject insulin.

 This may  impact on education adherence as well. However, two-thirds of the patients attending diabetes education were not insulin dependent.

No further differences between participants and nonparticipants in other illness-specific or sociodemographic variables could be found. This is in contrast to evidence from other countries that age,

 influence participation in diabetes education. The lack of such differences in the present study may be explained by the fact that the  authors only included DMP participants, which may have homogenized the population. The rationale for this inclusion criterion was that diabetes education was funded only for DMP participants within the statutory health insurance scheme during the time period under study.

Among nonparticipants, the most important reason for not attending diabetes education was that their knowledge of diabetes was considered sufficient. There was also a rather high percentage of persons who felt that diabetes management was not their own responsibility and relied on the activity of their attending physician. Only a maximum of one-fifth of the total population of nonparticipants agreed with each of the other statements. This might indicate that beyond knowledge and the attribution of responsibility, other reasons for nonparticipation are quite individual. Four typical subgroups of nonparticipants emerged from cluster analysis:

 All patients from cluster 1, the “informed and responsible,” considered their knowledge of diabetes sufficient and they all felt responsible for diabetes management. Nonparticipation in diabetes education may be a rational decision for these patients; they may be a group of patients who are sufficiently informed from sources other than diabetes education, eg, family, friends,

 media, or their attending physician. Further research is needed to clarify whether this patient group really is so well informed that they wouldn’t benefit from diabetes education—at least as long as their diabetes does not worsen.

The largest part of patients in cluster 2, the “unconcerned without desire for more information,” also feels  sufficiently informed, but firmly believes that they are not responsible for their own diabetes management. It might be that patients from this cluster do not take diabetes seriously enough, because of low severity, for instance, or because they deny presence of the illness.

 Feeling sufficiently informed has been shown to be a major barrier to diabetes education, but it does not necessarily imply that the patients in fact have enough information. Therefore it may also be that they do not understand the importance of their contribution to diabetes treatment,

 because their attending physician does not provide enough information, for example, or because he does not sufficiently involve the patients in diabetes treatment.

 For this reason patients from cluster 2 might benefit from improvements in physician communication questioning the assignment of diabetes management to the  physician.

Cluster 4, the “anxious and burdened,” is very similar to cluster 2 regarding subjective knowledge andresponsibility, but for a large portion of these patients many other reasons are also relevant for  nonparticipation. These especially include psychosocial problems (as indicated by negative feelings regarding groups of people or education courses) and functional limitations (especially problems with vision or hearing). Itmay be that this comorbidity dominates their treatment decisions. Psychological problems in particular might account for a shortfall in regimen adherence.

 Patients with highly anxious temperament might profit from anxiety-reducing interventions in diabetes education.

 Additionally, it may be important to tailor programs for persons with functional limitations (eg, poor vision or hearing).

 Many patients from cluster 4 also state that the intended purpose of diabetes education is unclear and that they fear excessive demands, such as difficult content. Attending physicians should therefore communicate the relevance of diabetes education for everyday life and, if applicable, the existence of programs specially tailored to these patients.

Patients from cluster 3, the “uninformed but responsible,” which is the largest cluster of the sample, completely stated that they did not feel sufficiently informed though they felt completely responsible for their diabetes management. Why these patients did not participate in diabetes education remains unclear. A qualitative study has shown that many patients attend diabetes education because they are experiencing practical problems and feeling uncertain about diabetes management due to insufficient information.30 Patients from cluster 3 mentioned only a very few reasons for nonparticipation, and 49% of these patients did not agree with any of the predefined statements. In the authors’ view patients from cluster 3 may be the most accessible target group for diabetes education, but there seem to be barriers to attending diabetes education that remain unmeasured in the present study.

It is notable that, except for cluster 4, most of the suggested reasons beyond the sufficient knowledge and the attribution of responsibility to the physician were marked only to a small degree as being relevant. It is possible that these reasons really didn’t fit, but it could also be that patients have difficulty in either understanding their specific reasons or in admitting that these mostly negatively attributed reasons play a role.

In consequence it is important that the attending physician assesses the barriers of diabetes education individually during a consultation. The patients’ individual perceptions of knowledge of diabetes and sense of responsibility for diabetes management might be a good starting point for enhancing the communication on barriers to diabetes education.

This study used cluster analysis for identifying differences in barriers to attending diabetes education between patient subgroups. The cluster analyses performed well on the givendata set; ie, the clusters are clearly separated by a subset of variables with high statistical significance (

 <.001), and the analyses have a good model fit expressed by a silhouette width of 0.36.

Another strength is the use of pooled data analysis with data from a postal patient survey (which include the patient’s perspective) and data from chart reviews (which represent more objective illness-specific data). The participants in this study were not selected by the GP or patients’ healthcare utilization patterns, since a chart registry approach rather than a waiting room recruitment strategy was used. The researchers avoided errors in data input by including multiple plausibility checks for each variable during and after data entry. Interviewers for data collection of A1C, BP, and BMI at GP practices were trained and monitored, and the questionnaire for the postal survey was piloted on 84 patients from 6 GP practices. Each patient received up to 3 invitations to participate in this study, and a response rate of 51% was obtained.

A weakness of this approach is that the analyses are based on 165 participants and 132 nonparticipants in diabetes education,which seems to be a rather low sample size. This problem becomes even more prominent when one considers thatthe authors had to conduct multiple statistical tests, which led to the requirement of a very strict significance level of 

.001 in order to minimize the possibility of false positive results. Because of the low sample size and the strict significance level, some differences between clusters (especially regarding sociodemographic and illness-specific data) may have been missed. Irrespective of this problem, large differences in patient subgroups regarding personal reasons for nonparticipation in diabetes education were identified. Further research will need to show whether the hypotheses generated from these explorative analyses can be confirmed in larger populations.

There was a strong association between the recommendation of the attending physician and participation in diabetes education. For this reason the recommendation can be used intentionally to increase participation rates. Physicians should explore the patients’ perceptions of the disease and its treatment more clearly, as well as the patients’ personal goals and social context. Starting from this patient perspective might help the physicians in motivating patients to participate in diabetes education.

Among nonparticipants the authors found 4 typical subgroups: the “informed and responsible,” the “unconcerned without desire for more information,” the “uninformed but responsible,” and the “anxious and burdened.” It seems to be important that the attending physician assesses the barriers to diabetes education individually during a consultation. As the 4 groups mainly differed regarding the perception of their knowledge of diabetes and their attribution of responsibility for diabetes management (to the physician or to themselves) it might be helpful to address these 2 aspects first in the consultation in order to discuss and eventually overcome the barriers to attending diabetes education.

 From Department of Primary Medical Care (IS, CK, MP, HVDB, HK), University Medical Center Hamburg-Eppendorf, Hamburg, Germany; Institute for Biometry (BW), Hannover Medical School, Hanover, Germany.

 This study was funded by the German Federal Ministry of Education and Research (grant number 01GK0714) and the German Medical Association within their peer-reviewed Funding Initiative for Health Services Research (grant number 06-152). The funding bodies had no influence on study design, on the collection, analysis, or interpretation of data, on the writing of the manuscript, or on the decision to submit the manuscript for publication.

The authors (IS, CK, BW, MP, HK) 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 (CK, BW, HK); acquisition of data (CK, HK); analysis and interpretation of data (IS, BW, HK); drafting of the manuscript (IS, MP, HK); critical revision of the manuscript for important intellectual content (MP, HVDB, HK); statistical analysis (IS, BW); obtaining funding (CK, HVDB, HK); administrative, technical, or logistic support (MP, HVDB, HK); and supervision (HK).

 Marc Pawels, Dipl-Psych, Dept Primary Medical Care, University Medical Centre Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. E-mail: m.pawels@uke.de.1. Norris SL, Lau J, Smith SJ et al. Self-management education for adults with type 2 diabetes. Diabetes Care. 2002;25(7):1159-1171.

2. Chodosh J, Morton SC, Mojica W, et al. Meta-analysis: chronic disease self-management programs for older adults. Ann Intern Med. 2005;143(6)427-438.

3. Brown SA. Interventions to promote diabetes self-management: state of the science. Diabetes Educ. 1999;25(suppl 6):52-61.

4. German Medical Association, Drug Commission of the German Medical Association, German Diabetes Association et al. National practice guideline diabetes mellitus type 2 (in German). 1. Ed. 2002.

5. Federal Ministry of Justice, ed. Requirements for structured care programs for type 2 diabetes [in German]. BGBl I. 2002;2290-2297.

6. Norris SL, Engelgau MM, Narayan KM. Effectiveness of self-management training in type 2 diabetes: a systematic review of randomized controlled trials. Diabetes Care. 2001;24(3):561-587.

7. Schäfer I, Küver C, Gedrose B, et al. Selection effects may account for better outcomes of the German Disease Management Program for type 2 diabetes. BMC Health Serv Res. 2010;10:351.

8. Altenhofen L, Brenner G, Haß W, Siering U. Quality Assurance Report 2004, Disease-Management-Programs in North Rhine [in German]. http://www.kvno.de/downloads/qualbe_dmp04.pdf. Accessed 2010.

9. Icks A, Rathmann W, Haastert B, et al. Quality of care and prevalence of complications in a random population-based sample of patients with type 2 diabetes mellitus, based on the KORA Survey 2000 [in German]. Dtsch Med Wochenschr. 2006;131(3):73-78.

10. Mielck A, Reitmeir P, Rathmann W. Knowledge about diabetes and participation in diabetes training courses: the need for improving health care for diabetes patients with low SES. Exp Clin Endocrinol Diabetes. 2006;114(5):240-248.

11. Ruppert K, Uhler A, Siminerio L. Examining patient risk factors, comorbid conditions, participation, and physician referrals to a rural diabetes self-management education program. Diabetes Educ. 2010;36(4): 603-612.

CLINICAL

This study describes reasons for nonparticipation in type 2 diabetes mellitus education and identifies typical subgroups of nonparticipants in order to improve recruitment strategies.

Identifying Groups of Nonparticipants in Type 2 Diabetes Mellitus Education

The patient-centered medical home (PCMH) offers an innovative method of delivering primary care. However, the necessary staffing infrastructure is not well established.

To evaluate the roles of personnel within a PCMH and to propose necessary staffing ratios and associated incremental costs to implement this model of care.

 We sampled primary care clinical practices that either have successfully deployed or were in the process of implementing a PCMH practice model. We conducted targeted interviews of administrators from these practices and reviewed published literature on the personnel roles within a PCMH. Collectively, these data were compared with current staffing standards and used to inform an analytical model and sensitivity analysis.

Primary care practices that successfully transitioned to a PCMH have incorporated a range of new staff and functionalities. Based on our model, we estimated that 4.25 full-time equivalents (FTEs) should be allocated to staffing personnel per 1 physician FTE. Compared with the base-case model of current staffing in the United States of 2.68 FTEs per physician FTE, this is a 59% increase. After applying sensitivity analysis for variability in staffing and compensation, the incremental staffing FTE per physician FTE was 1.57 (range 1.41-1.73) and the incremental associated cost per member per month was $4.68 (range $3.79-$6.43).

Our study suggests that additional staff with specific expertise and training is necessary to implement a PCMH. Further study and opportunities for funding additional staffing costs will be important for realizing the potential of the PCMH model of care.

 In this study we evaluated the personnel infrastructure, staffing ratio, and associated costs needed to implement a patient-centered medical home (PCMH).

		Experience from successful PMCH practices suggests that additional staff with necessary expertise and training will be required in order to achieve the goals of improved outcomes and better care at reasonable cost that incorporate effective care coordination, medication management, and support for behavioral health and population management.

		This additional support may need to be further augmented in practices serving vulnerable populations and/or providing practice sites for resident trainees.

Amid a shortage of primary care physicians, the need to provide high-quality, affordable care to a growing population drives the patient-centered medical home (PCMH) movement as an innovative method of delivering primary care in the United States.

 The key premise is that improvements in the quality and efficiency of care will result from a interprofessional teambased approach where each member of the team works collaboratively with other team members, all performing at the top of their scope of practice.

Although implementation of a PCMH is highly contextual, there is growing evidence to suggest that certain functionalities are important for success.

 Joint Principles for the Medical Education of Physicians as Preparation for Practice in the Patient-Centered Medical Home

 summarized the overarching principles that seek to strengthen the physician-patient relationship by replacing episodic care with coordinated, proactive, team-based, and accessible care that emphasizes safety and quality.7 Practices have used the following key strategies to help transition to the PCMH model: enhanced care management

 integration of behavioral health into medical care,

 improved medication management,16-20 improved triage and appointment availability, 

population management,23,24 and engagement of patients and families including community outreach.

 All these elements require a strong health information system, the capacity to analyze data for quality improvement, and an emphasis on team-based care involving a range of personnel.

Given these new functionalities, the decision to implement a PCMH requires completing a needs assessment that examines existing staff roles and functions, considers the extent to which existing staff can meet the new and expanded roles of a PCMH, and determines whether additional personnel and financial support will be needed to fill the gaps. Thus far, there are limited data on staffing requirements for practices that are interested in transitioning to a PCMH.

 The objective of this study was to evaluate the roles of personnel within a PCMH and to propose necessary staffing ratios and associated incremental costs to implement this model of care.

This study was initiated based on the deliberation of a work group created at the PCMH Education Summit convened by the Society of General Internal Medicine in 2011. The purpose of this summit was to discuss the challenges of advancing the PCMH model within academic medical centers in the United States and to create work groups to address specific topics ranging from curriculum and scheduling to infrastructure. The summit, but not this study, was supported by the Josiah Macy Jr Foundation and others. This study was approved by the institutional review board at the University of Pennsylvania.

There is limited published literature to address our research question: what is the appropriate staffing infrastructure of a PCMH? Because limited role-specific data exist in some contexts (eg, role of pharmacists in the PCMH

), we used published data when available and conducted targeted interviews of a small sampling of practices implementing a PCMH to fill in the gaps when published data were not available. Collectively, these data were compared with current staffing standards and were used to inform an analytical model and sensitivity analysis to propose insights into the necessary staffing infrastructure for a PCMH.

To inform the staffing and cost estimates of our analytical model, we reviewed published literature on the staffing roles within Joint Principles for the Medical Education of Physicians as Preparation for Practice in the Patient-Centered Medical Home,

 conducted interviews with leaders and administrators at a small sampling of primary care practices, obtained current staffing estimates from the Medical Group Management Association (MGMA),

 and gathered staffing compensation estimates by geographic location.

After determining key functions of a PCMH based on practice principles, published literature on staffing roles within a PCMH was reviewed to determine which professionals within current PCMHs carried out those roles in practice. In 

 we display the PCMH principles based on the 2007 

 and the key strategies and components for each principle. Based on this information, we elucidated the implications for staffing and conducted a keyword search for the staffing role (eg, care manager) and combinations of the staffing role and a PCMH (eg, PCMH, patient-centered medical home, medical home).

During the Society of General Internal Medicine Education Summit, we identified a convenience sample of primary care clinical practices that were known to have either successfully deployed or were in the process of implementing a PCMH practice model. These practices were selected because they had published their results in peer-reviewed journals, they had presented their approaches at academic meetings, or their efforts to deploy the PCMH were known to members of the Society of General Internal Medicine Education Summit working group. We chose to interview administrators because they were often cited as the primary decision makers for staffing their medical practices. Administrators of eligible practices, physician or nonphysician, were interviewed over the telephone and asked the following questions: (1) What are your average attending and resident panel sizes? Do you use any risk adjustment methodologies? (2) What types of health professionals (including care managers, nurse practitioners, physician assistants, clinical pharmacists, health coaches/educators, social workers, behavioral health providers, or nutritionists) have been incorporated in the care team, and what are their roles? (3) Have any other administrative personnel (eg, data analysts) had been added to the team? (4) What are the staffing ratios between these positions and physician full-time equivalents (FTEs) in the practice? (5) Is the practice certified as a PCMH by the National Committee for Quality Assurance (or other organization)? (6) Does the manager directly overseeing the practice have a master’s degree in business, healthcare administration, public health, or a related field?  

Nine administrators with knowledge of their institution’s primary care practices and PCMH initiatives agreed to be interviewed. They represented academic medical centers, nonprofit and for-profit institutions, and government-owned health systems. Of these institutions, 7 had at least 1 practice certified as a PCMH (6 by the National Committee for Quality Assurance, 1 by a health insurer). One practice had applied for certification. One large practice group had not sought formal PCMH certification. Between January 2011 and May 2011, a single investigator (MJA) conducted the telephone interviews and followed up by phone and/or e-mail to confirm findings when additional clarification was needed. Interview results were de-identified and aggregated for review by study authors.

Data from the MGMA 2010 cost survey report

 were used to assess current staffing models in the United States. The median staffing level for all internal medicine practices owned by a hospital/integrated delivery system was used as the benchmark for comparison in our study. Data from Kenexa CompAnalyst Market Data,

 a national proprietary database, were used to determine the median annual total salary for the clinical and administrative staffing roles. Variances in compensation by geography were based on data from the Economic Research Institute’s Geographic Assessor pay survey.

Data from literature review and interviews were aggregated for role-specific comparison. A model incorporating an estimate of staffing of each role was developed as follows. Published estimates of staffing ranges were used when available. When published data were unavailable, current staffing baselines were compared with interview ranges and the median from our survey of 9 administrators was used. Using the median staffing level in the 2010 MGMA cost suvey

 as a baseline, the incremental staff necessary to meet our proposed staffing model was then determined. Data were reported as FTE requirements per physician FTE. Sensitivity analysis was applied at ±10% to each staffing estimate. Median annual compensation for clinical and administrative roles in the PCMH was estimated using the Kenexa Compensation Analysis.

 The cost of fringe benefits, estimated to be 30% of base salary, then was added to estimate total annual employee compensation. Sensitivity analysis for compensation was conducted by using the Economic Research Institute survey data to estimate the range among the 9 institutions in our study. A range of —10% to +25% was applied to represent geographic variations between 2 cities: Dubuque, Iowa (a city with lower salaries) and Manhattan, New York (an area with high salaries). Incremental cost estimates were calculated by multiplying the incremental FTEs per physician by the annual compensation estimates. The incremental reimbursement required to offset the incremental costs were estimated. Assuming a typical panel size of ~2150 patients per physician FTE (approximately the median among the 9 practices in our study), the incremental cost per member per month was computed, where each member was a patient within the physician’s panel.

The majority of practices interviewed in our study noted that the manager directly responsible for the PCMH did not have formal leadership training such as a master’s degree in public health, business administration, health administration, or another related field. A commonly expressed sentiment was that the increased complexity of a PCMH required a significant expansion of the administrative skills of the managers.

, we display a summary of the results of staffing estimates from our interviews compared with MGMA baseline estimates. Panel sizes for physician FTEs ranged from 625 to 2500 patients and varied significantly in terms of the criteria used to define a physician’s panel and the number of patients a physician was expected to manage. This wide range reflects the variability in criteria used to determine panel sizes and includes academic faculty, who carry smaller patient loads. Resident panels provided by 2 practices varied by site and years of training, ranging from 30 to 60 patients for postgraduate year 1 to 80 to 125 patients for postgraduate years 2 and 3. Only 3 sites reported risk adjustment to their patient panels, using either software imbedded within the electronic health record and practice management systems or proprietary software. Several practices noted their intent to risk-adjust patient panels in the future.

A care manager was the most consistent addition of staff to the practices in our study. A registered nurse filled this role, often in a ratio of 1 to 1 with each physician FTE. Other common findings among practices included the use of health coaches or educators (0-0.25 per physician FTE) and behavioral health providers (0-0.83) to address the importance of self-management and emotional support, and the addition of pharmacists (0-0.53) to assist in medication management, chronic disease care, and consultation practice. Several practices employed nutritionists (0-0.20) to educate patients regarding dietary issues. There were less consistent data on which professional provided information technology support, including registry management, coordinating an effective and updated electronic health record, and ongoing data reporting.

Staffing Model and Associated Cost Estimates

Based on review of estimates in published literature andbthemes within our interviews, we propose our staffing model in column 5 of Table 2. The total staffing estimate per physician FTE was 4.25, which resulted in an additional 1.57 staff per physician FTE. In comparison with baseline MGMA estimates, we recommend increased staffing in the forms of care managers, behavioral health/social workers, pharmacists, health educators, nutritionists, and data analysts. We also suggest an increase in clerical staff that is required for population management reporting.

, we report cost estimates for the roles within our proposed staffing infrastructure. We applied sensitivity analysis to the incremental staffing estimates (range 1.41-1.73) and then estimated incremental costs with sensitivity analysis to adjust for geographic variation. The incremental cost of our proposed infrastructure for staffing a PCMH was estimated to be $4.68 (range $3.79-$6.43) per member per month, where each member is a patient within the physician’s panel.

Transforming primary care practices into PCMHs will require substantial changes in work flow and for many practices, additional staff. Based on our proposed staffing model, we estimate that a 59% increase (4.25 vs 2.68) in staff per physician FTE will be needed for practices currently staffing at the MGMA median to transition successfully to a PCMH. While the exact mix of new personnel will vary by each individual practice, it is clear that in order to provide such services the new staff will require health professionals with specialized training such as nurse practitioners, care managers, data analysts, pharmacists, and behavioral health providers. Practices that may need staffing on the higher range of our estimates are those that care for patients with higher levels of illness complexity or more demanding socioeconomic environments.

Many components of our proposed incremental staffing estimates fall within the range of that suggested for specific PCMH roles in the literature.

 Our overall proposed staffing model is higher because we included improved functionality that professionals such as pharmacists, social workers, health coaches, dieticians, and clinical data analysts provide, which were not included in prior reports. Furthermore, our interviews with practice administrators provided anecdotal evidence supporting the experience of others that adding a care manager (typically a registered nurse) yielded substantial, rapid benefits.

Substantial increases in reimbursement will be required to support these additional staff. Based on our model, we estimate an extra $4.68 per member per month will be necessary to cover additional staffing compensation. It is important to note that this figure does not include new personnel costs related to reconfigured office space, new technology, and training materials or any incremental compensation for the primary care physicians. Additionally, the potential costs attributable to obtaining more experienced nonphysician managers and greater involvement of physicians in practice management are not included in this estimate.

 The augmented reimbursement necessary will vary considerably by geography. In our study, sensitivity analysis estimated a range of $3.79 to $6.43, which accounts for variation in number of staff FTEs per physician FTE and compensation by geography. It is worth noting that the estimated incremental revenue needed to support the PCMH in our model is within the range offered by many pilot projects across the United States.

Recent studies raise hope that the additional staffing costs might be offset by savings achieved through improved outcomes due to implementation of the PCMH. The Group Health Cooperative conducted a study that compared costs before and after the first 2 years of implementing their PCMH.

 After implementing their PCMH model, they observed 29% fewer emergency department visits and 6% fewer hospital admissions for the practice. This translated into a savings of $10.30 per member per month; however, this amount was not statistically significant (P = .08). Geisinger conducted a study of their PCMH that compared utilization data 2 years before and 2 years after implementation.

 They found 18% fewer hospital admissions and 36% fewer hospital readmissions after implementing the PCMH. A subsequent financial analysis of Geisinger’s Medicare Advantage patients after a median of 26 months in the PCMH demonstrated a 4.3% to 7.1% reduction in total costs (

 Other pilot projects have demonstrated significant improvement in quality and reductions in utilization and total costs.

Networking and creative partnerships could help to leverage resources across smaller practices. As indicated in Table 2, solo practitioners would require only a 0.40 FTE of a care manager and 0.25 FTE of a social worker. A large health system with smaller affiliate primary care practices might decide to hire a diabetic educator to work part-time at each practice site, in hopes of improving patient adherence and chronic disease management within each of the practices. Several practices might create a network that could hire a care manager who could be a liaison between those practices and a hospital system.

 Improved collaboration with insurance plans might allow more efficient use of the insurance plan’s management of high-risk patients. Ultimately, however, significant enhancement in reimbursement to practices is likely to be required if the full functionality and benefit of a PCMH model is to be realized.

Our study has several limitations. The range of staffing ratios between physician and clinical staff in the surveyed practices varied considerably. Determining precise ratios was challenging in some practices due to differences in position titles across practices, shared staff within a PCMH or with other non-PCMH practices, bundled staffing data that prevented determination of ratios of specific roles, the inclusion of some positions in the practice that were not integrated into the PCMH team, and the off-site locations of some staff (eg, case managers).

Staffing estimates were derived from phone interviews and written communication, and therefore are subject to reporter bias. The practices surveyed included a mix of academic and community-based offices. Only 2 practices provided specific information about intern and resident panel sizes, and only 3 practices provided specific staffing ratios for their teaching practices. As a result, the recommended staffing ratios may need to be modified for use in teaching practices. In addition, it is important to note that we sampled administrators and not front-line staff or workers within the practices, who might have different perspectives on staffing infrastructure.

The staffing ratios for some roles in the MGMA report were derived from a relatively small number of practices. The cost estimates we used reflected median compensation levels, and compensation levels will vary considerably across the country. The cost of employee benefits will also vary considerably across practice settings and may be substantially different from the estimate used in this analysis (30% of base salary). Finally, many of the practices surveyed in our study noted that their planning was influenced by the initiatives and regulations in the state in which they reside, and we were not able to account for this variation.

Implementing a PCMH requires an evaluation of the practice’s current staffing and functional roles, along with determination of the needs of its patient population. Our proposed staffing model suggests that in order to achieve the goals of improved outcomes and better care, additional staff with necessary expertise and training will be required. Although potential long-term savings from implementation of a PCMH might offset the additional up-front costs of increased staffing, further study is needed and additional funding sources will be required if we are to realize the full potential of the PCMH.

From Philadelphia Veteran Affairs Medical Center (MSP), Philadelphia, PA; Robert Wood Johnson Clinical Scholars Program (MSP), Division of General Internal Medicine (SD), University of Pennsylvania, Philadelphia, PA; Beth Israel Medical Center (MJA), New York, NY; Medical Associates (TAS), Dubuque, IA; Mercy Catholic Medical Center (EHG), Darby, PA; Department of Medicine (DWB), Northwestern University, Chicago, IL; Division of General Internal Medicine (JLB), Oregon Health and Science University, Portland, OR.

 The authors (MSP, MJA, TAS, EHG, DWB, JLB) 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 (MSP, MJA, TAS, EHG, DWB, JLB, SD); acquisition of data (MJA, TAS); analysis and interpretation of data (MSP, MJA, TAS, EHG, SD); drafting of the manuscript (MSP, MJA, TAS, EHG, DWB, JLB, SD); critical revision of the manuscript for important intellectual content (MP, MJA, TAS, EHG, DWB, JLB, SD); statistical analysis (MP); administrative, technical, or logistic support (MP, SD); and supervision (MP, JLB, SD).

 Mitesh S. Patel, MD, MBA, Robert Wood Johnson Clinical Scholars, University of Pennsylvania, 423 Guardian Dr, 13th Fl, Philadelphia, PA 19104. E-mail: mpatel@mail.med.upenn.edu. 1. Larson EB, Reid R. The patient-centered medical home movement: why now? JAMA. 2010;303(16):1644-1645.

2. Flores L, Margolis JW. The patient centered medical home: 2011 status and needs study: reestablishing primary care in an evolving healthcare marketplace. Medical Group Management Association. 2011:1-21.

This study suggests that implementing a patient-centered medical home requires additional staff with specific expertise based on the needs of the practice and its population.

June 2013

Accountable Care Organizations: How to Dress for Success

Identifying Groups of Nonparticipants in Type 2 Diabetes Mellitus Education

Estimating the Staffing Infrastructure for a Patient-Centered Medical Home

Cervical Cancer Screening Overuse and Underuse: Patient and Physician Factors

Population Health Approach for Diabetic Patients With Poor A1C Control

Shifting the Dialogue From Hospital Readmissions to Unplanned Care

Affordability in a Mandated Environment

Impact of Cardiac Telemetry on Patient Safety and Cost

Effects of Medicare Part D Coverage Gap on Medication Adherence

Medicare and Commercial Inpatient Resource Use: Impact of Hospital Competition

Consumer Cost Sharing and Use of Biopharmaceuticals for Rheumatoid Arthritis