To review the literature relating to treatment intensification in diabetes care and provider traits.

 A literature search was conducted using PubMed and Google Scholar for papers published in or after 2000 that examined treatment intensification in diabetes care. Results from the searches were combined with a conventional Google search and a supplemental review of papers that were identified from reference lists of identified studies.  

 The majority of papers that were identified used administrative data to assess treatment intensification, and these showed that individuals with diabetes and elevated glucose frequently do not receive timely changes in therapy in response to hyperglycemia. Relatively few reports address provider and practice characteristics associated with these treatment decisions. Many of the studies focusing on the relationship between provider traits and treatment intensification are based on small studies in a limited number of practices.  Factors such as practice size, location, or experience in treating patients with diabetes were often not addressed in the literature despite their potentially far-reaching impact on treatment.

Our literature search on treatment intensification in diabetes care shows that the majority of papers using administrative data to assess treatment intensification suggest that care is often discordant with recommended guidelines. However, there is a dearth of literature based on large databases examining physician and practice traits related to this discordance. Better understanding physician behavior and practice traits associated with treatment intensification may permit greater targeting of interventions aimed at improving care.

	Our literature search on treatment intensification in diabetes care shows that the vast majority of papers using administrative data to assess treatment intensification suggest that care is often discordant with recommended guidelines. However, there is a dearth of literature using large databases that examine physician and practice traits related to this discordance.

		There is a great deal of evidence indicating failure to intensify treatment in diabetes care, but little is known about physician and practice traits that influence treatment decisions.

		Better understanding physician behavior and practice traits associated with treatment intensification may facilitate targeting of interventions aimed at improving care.

Clinical practice guidelines for the management of people with diabetes provide direction to clinicians on a number of key issues, including targets for glycemic (glucose) control and when to intensify an existing treatment regimen that is not meeting target.

 Despite the availability of these detailed guidelines, there is ample evidence supporting the idea that many people with diabetes have blood glucose levels and other cardiovascular risk factors that are above recommended levels,

 leading to poor clinical outcomes and avoidable health costs. A wealth of research finds that diabetes is undertreated, with many individuals with diabetes going undiagnosed, diagnosed but untreated, or treated but not adherent.

 Equally important, a great deal of research shows that despite access to healthcare, these individuals with elevated glucose—in both the inpatient and outpatient settings&mdash;do not receive timely changes in therapy in response to documented hyperglycemia.

 For example, a study of administrative data describing 16,800 patients with diabetes at 13 Veterans Administration medical centers showed that despite an average of 8.8 annual visits, glycated hemoglobin (A1C)—a key indicator of glycemic control&mdash;did not improve, with nearly 40% of patients having values of 8% or higher, which is well above normal. Intensification of diabetes treatment occurred in only 9.8% of all visits in this study.

Against this backdrop, obvious questions arise regarding why healthcare providers do not consistently intensify treatment of their patients with diabetes in situations where there is a clear indication that glucose, or other risk factors such as blood pressure and lipid measures, are inconsistent with practice guidelines. These questions are particularly salient as Medicare, Medicaid, and other payers pursue delivery reforms (eg, accountable care organizations and medical homes) intended to improve clinical and cost results through better management of chronic illnesses. A recent report by the American Diabetes Association estimated that in 2012, the total costs associated with diabetes topped $245 billion, with 43% of total costs going to inpatient care, suggesting that there may be ample opportunities for savings associated with better care.

 Specifically, achieving better performance under new health system incentives—such as those intended to reduce certain inpatient readmissions&mdash;may depend in part on improved chronic care management.

 led to numerous studies focusing on the role of “clinical inertia” in healthcare delivery, including diabetes care. As conceptualized by Phillips et al, clinical inertia is the “failure of healthcare providers to initiate or intensify treatment when indicated.” This failure is postulated to be fueled by a combination of provider, patient, and organizational factors.

 The Phillips paper was a driving force behind the publication of a large body of literature whose findings suggested that the gap between what is recommended in diabetes care and what is actually acted upon during a clinical encounter reflects widespread deficiencies in quality of care. This line of thinking equates lack of action with lack of quality. It goes on to argue that these apparent deficiencies result in a considerable amount of potentially avoidable morbidity and mortality in an ever-increasing diabetic population, and that aggressive action is needed to encourage physicians to engage in prescribing behaviors that are more effective in helping patients achieve standard clinical targets. One report even likens failure to intensify treatment to a medical error.

In contrast, Parchman et al offer an opposing viewpoint

 in which the “competing demands” that characterize routine clinical practice—particularly with older, medically complex patients&mdash;influence the likelihood of changes in therapy. In this line of thinking, physicians and patients prioritize problems, address the most pressing ones, and defer action on lower priority problems to future encounters. In an environment in which clinicians are being encouraged to engage in patient-centered decision making,

 this type of collaborative prioritization can be viewed as both reasonable and justifiable, a concept that has been acknowledged elsewhere.

Regardless of which model one ascribes to, if case-mix is adequately adjusted for, intensification care should reflect patient traits, not provider, provider group, or delivery system traits. The purpose of this literature review and synthesis is to assess what is known about the relationship between provider, practice group, or system traits and treatment intensification. Moreover, we are interested in how much of our current understanding is based on studies of a few practice settings as opposed to analyses of large databases that might capture the general patterns of care across the country. Understanding these relationships can help with the design and targeting of interventions in the new delivery models, emphasizing systems of care and improved chronic care management.

To maximize application of this review to the current clinical landscape, we limited our search to material that was published from the year 2000 onward. We identified combinations of search terms based on an initial review of recent reports on clinical inertia in diabetes care; these terms repeat with great frequency in many of the published reports. Articles were identified if they met any of these criteria. The search combinations were: “clinical inertia and type 2 diabetes,” “clinical inertia and glycemic control,” “clinical inertia and glucose control,” “clinical inertia and HbA1c,” “under treatment and type 2 diabetes,” “treatment intensification and type 2 diabetes,” “non-treatment and type 2 diabetes,” and “clinical targets and type 2 diabetes.” In addition to these PubMed searches, a Google Scholar search was conducted using the terms “clinical inertia” and “type 2 diabetes” and a conventional Google search was conducted using the terms “clinical inertia” and “type 2 diabetes.” Finally, we used the reference lists from papers that were identified in these searches to find additional literature that was relevant to our questions.

Our searches yielded a multitude of studies examining various aspects of treatment intensification using administrative databases.

These and similar reports, which use sources such as patient medical records, insurance claims, and pharmacy data, leave little doubt that diabetes treatment recommendations are often not followed in practice. The absence of qualitative factors that place objective clinical data in a real-life context was an almost universal and far-reaching limitation of these reports, and the lack of this type of information might explain appropriate decision making that results in lack of treatment intensification for some portion of those instances when intensification is called for by guidelines.

Providers’ Perceptions of Their Patients

A number of reports that collected information on providers’ perceptions of key patient characteristics suggested that these perceptions&shy; strongly influenced treatment decisions, making it important to assess and address (as appropriate), these perceptions and any other barriers to intensification as delivery systems focus on improving chronic care. For example, one study surveyed 83 primary care physicians in 3 states and provided unusually detailed information concerning qualitative factors that underpin their diabetes treatment decisions.

 Results indicated that although 69% of respondents agreed with guidelines stipulating A1C targets of <7%, more than half also reported that they would raise the target value depending on factors such as life expectancy, poor patient self-management, low educational attainment or health literacy. When asked why they would decide not to initiate insulin, 43% expressed concerns about self-management skills. Nearly half of these providers indicated that they would determine that a patient was incapable of managing insulin therapy based on their perception of the patient’s self-management skills.

Further, 77% reported that they would not initiate insulin “some” or “most” of the time if they felt the patient’s adherence to oral medication was low, and an additional 5% said they would never initiate insulin under these circumstances. Similar findings were noted in a survey of 886 members of the Society of General Internal Medicine and the American Diabetes Association that confirmed that among both generalists and specialists, physician perceptions of certain patient characteristics influence the choice of initial hypoglycemic therapy as well as the choice to initiate insulin.

 For example, 59% of specialists and 57% of generalists reported that patients’ adherence behavior influenced the choice of initial therapy, and 67% and 71% of the 2 groups reported that patients’ motivation to improve influences their choice to initiate insulin. A similar study reported that providers tended to prescribe insulin more frequently when they believed patients were more adherent to medication regimens.

Against this backdrop, it is instructive to consider the impact of patient age on physician perceptions that may influence treatment decisions. One notable study surveyed more than 500 primary care physicians and asked them to provide information on their patients with diabetes aged 65 years or more who were newly diagnosed and who had not started treatment during the past 6 months.

 Almost half of the physicians who responded to this survey reported that they had concerns about hypoglycemic treatment in the elderly. Indeed, although a large portion of the 770 patients in this study had A1C <7%, fully one-third had A1C exceeding 7%. When asked why these older patients were not started on therapy, physicians frequently reported that they were still trying to manage the patient with diet and exercise, that the patient’s hyperglycemia was “mild,” or that they had concerns about antihyperglycemic agents in older patients, as well as concerns about patient comorbidities and polypharmacy in this patient population.

Another survey of 358 general practitioners and 2028 newly diagnosed diabetes patients echoed these findings: Nearly one-fourth of the physicians in this survey reported that they would not start an older patient with diabetes on medication because of concerns about hypoglycemia, and large numbers reported that factors such as concerns about a patient’s cognitive burden associated with therapy administration and medication monitoring, as well as the likelihood of the patient’s noncompliance, all played roles in the decision not to initiate medication.

 Moreover, the physicians frequently reported that patients had “mild hyperglycemia” even when A1C levels exceeded 7%. Taken together, these results indicate that physician perceptions about a variety of their patients’ behaviors, abilities, and motivation influence treatment decisions. Further, physicians appear to have additional perceptions about older patients that result in a lower likelihood of promoting aggressive treatment, or indeed, initiating treatment at all. Forward movement in this area would likely benefit not only from a better understanding of providers’ perceptions and concerns regarding these and similar issues, but also identification and implementation of effective approaches for addressing them. Guideline writers may also want to consider evaluating and addressing common perceptions that lead to variance from guidelines.

Provider Knowledge and the Challenge of Uncertainty

Although improved physician education has been proposed as a strategy to improve treatment intensification,

 few studies directly address the question concerning exactly what type of education is needed and most effective for improving treatment intensification and outcomes. For example, our findings suggest that there is widespread knowledge concerning the diabetes practice guidelines target of A1C of 7%,

 and similarly, information concerning the recommendation to initiate metformin monotherapy as a first-line approach to pharmacologic treatment when diet and exercise fail is widespread.

 However, things begin to fall apart when metformin is no longer sufficient to control glucose—an outcome that occurs with great frequency because of the progressive nature of diabetes.

Practice guidelines diverge and considerable uncertainty is introduced regarding factors such as titration, choice of combination therapy, and when to initiate insulin.

 Not surprisingly, it has been suggested that the lack of clear guidance concerning the relative merits and liabilities of the many therapeutic options that are available to physicians inhibits effective treatment and introduces considerable uncertainty into routine clinical care.

 In particular, knowledge about proper use of insulin and the various options available for insulin use—including appropriate use in the inpatient setting&mdash;appeared to be areas where knowledge was especially sparse. One study showed that 58% of medical residents and mid-level providers reported that knowledge about what insulin type (eg, long-acting) or regimen works best was a barrier to effective inpatient management of hyperglycemia.

 Learning more about practices that can address these kinds of barriers, and also perhaps about useful ways to modify guidelines, would likely assist the new delivery systems working to improve chronic care management.

Time With the Patient and Number of Patient Concerns

An important theme that emerged from our review was that a confluence of circumstances—including larger numbers of older, medically complex patients, the availability of more diabetes treatment options, and polypharmacy&mdash;act together to reduce the amount of time that physicians have to work with their patients to modify diabetes treatment if it is indicated&shy;. Diabetes may be only 1 of a number of problems that doctors and patients must address during a single encounter.

 A survey of 108 diabetes specialists found that as time with the patient decreased, the likelihood that providers would not comply with practice guidelines increased.

 Along the same lines, a study of 211 patient encounters in 20 primary care clinics addressed the idea of competing demands and stood out as an excellent example of the type of research that could help shed light on why physicians make so many individual exceptions to standard care.

Among the notable findings of this report was the observation that among diabetic patients with A1C >7%, each additional problem that was raised in the primary care visit was associated with a 49% reduction in the likelihood that their diabetes medication would be changed. Further, no changes in medication occurred among patients who expressed more than 4 concerns during their encounter. However, among patients whose diabetes medication was not changed, each additional minute of encounter length with the physician reduced the time to the next appointment by 2.8 days. Similarly, among these patients, each 1% increase in A1C resulted in a reduction of 8.6 days to the next visit. Taken together, these data suggest that although doctors and patients prioritize problems during the encounter in ways that can be unfavorable toward immediate changes in diabetes treatment, there appears to be an acknowledgment of this decision in the form of shorter follow-up aimed at addressing the problem at the next encounter.

A key question in understanding treatment intensification is the extent to which issues that are inherent in the process of care—factors that are largely independent of the patient&mdash;play a role in what is broadly interpreted as clinical inertia. The aforementioned survey of 108 diabetes professionals that examined barriers to helping patients achieve glycemic control showed that measures such as lack of entries in the medical record, lack of teamwork among physicians and nurses, and the lack of alarms in electronic medical records that are linked to treatment goals created challenges for providers in intensifying treatment.

 This study did not examine the relationship between these measures and glycemic control or other intermediate outcomes. In contrast, a report from the Translating Research Into Action for Diabetes group shed light on the apparent lack of association between intensity of diabetes care management and intermediate clinical outcomes.

 This report examined 3 measures of the intensity of care management—physician reminders, performance feedback, and structured care&mdash;in relation to 8 process of care and outcome measures, which included lipid and urine albumin measures and foot and retinal exams, among others. Additional intermediate outcomes included A1C, LDL cholesterol, and blood pressure values&mdash;the basic clinical information intended to trigger modifications in therapy if they are indicated.

This report found that, in general, as intensity of diabetes care management increased, so did the performance on process of care measures. However, disease management intensity was not associated with intermediate outcomes. Notably, practices with the most intense diabetes care management were less likely to have patients with favorable A1C or those on 2 or more diabetes medications. These results could suggest that system-level factors other than those related to process of care may influence key outcomes, or they may reflect the impact of unmeasured confounders. The latter concept—the idea that variables that were not measured in the study could have a meaningful impact on results&mdash;is important because qualitative factors are frequently unmeasured in these types of studies despite their potential impact on treatment patterns.

Similar to results in the outpatient setting, studies of diabetes and treatment intensification indicated that despite frequent glucose readings and ready access to information, hyperglycemia is common among inpatients with diabetes,

 with 1 study indicating that 71% of inpatients had 1 or more bedside reading exceeding 200 mg/dL.

 Results of inpatient studies not only showed that treatment was not intensified when indicated, but 1 report revealed that in many cases, insulin dosing was actually decreased among patients with the highest glucose readings.

 These findings call for exploration of their relationship to readmissions, the reduction of which is a prime target for improving clinical quality and achieving cost savings, as well as longer-term outcomes.

As with most other studies of treatment intensification in diabetes, many hospital-based studies relied heavily on administrative data. However, studies that surveyed physicians, nurse practitioners, and physician assistants offered information on provider and organizational factors that supplemented administrative data by providing a more qualitative assessment of barriers to care. For example, a survey of 103 residents and midlevel practitioners showed that nearly all of these providers believed that glucose control was “very important” in critically ill patients, but only 68% believed it was “very important" in noncritically ill patients.

 Although respondents indicated that they would target glucose in a manner consistent with practice guidelines, only 43% reported that they were “very comfortable” managing hyperglycemia and 48% said they were “very comfortable” managing hypoglycemia.

 The report concludes with a number of recommendations to improve the quality of inpatient diabetes care. These include recognition of the problem, supportive infrastructure, involvement of local thought leaders, prioritization and incremental implementation, sustained commitment by stakeholders, and an iterative development process. Although these recommendations seem reasonable, the relative importance of each, and how such a quality improvement program should be implemented in different types of hospitals (eg, rural vs urban), as well as the evolving role of health information technology, are all issues that remain to be resolved in efforts to enhance inpatient diabetes care.

Much is known about optimal treatment of diabetes. While there are variations among individual patients, exceptions to rules, disagreements about specific regimens under some circumstances, and important remaining gaps in knowledge, achieving good control of blood sugar is generally recognized as yielding better clinical outcomes. Despite much evidence for and widespread agreement on this principle, there are multiple gaps in diabetes treatment that lead to poor clinical outcomes and costs that are, to a significant degree, avoidable.

One of these gaps is lack of intensification when clinically appropriate. Given general evidence about variation in treatment patterns across providers, provider groups, and healthcare systems, it is reasonable to believe that provider, group, and system traits may be associated with intensification. In fact, the movement to measure and reward quality of care generally focuses on providers, groups, and systems.

We identified 2 primary types of studies that examined intensification of diabetes treatment: those based on large administrative databases that quantify both failure to intensify treatment and patient factors related to use of prescription drugs; and those based on survey data or examination of practice patterns in relatively small samples or settings. Findings from these relatively small studies suggest that there is significant variation in the treatment of diabetes across physicians and practices.

Large studies examining physician and practice traits related to intensification are lacking. For example, existing studies do not adequately assess the relationship between intensification and factors such as practice size, region, urban location, or behavior of peer physicians. Moreover, physician traits like age, gender, and experience treating patients with diabetes, are often absent. Additionally, the answers to basic questions—such as whether failure to intensify treatment is concentrated among certain physicians or practices or if it is spread relatively evenly across providers and provider settings&mdash;are unknown. The relationship between frequency of visits or lab testing is another area that has not been fully investigated. Understanding the association between these factors and treatment intensification will permit greater targeting of interventions aimed at improving care and speak to strategies that hold providers accountable for quality when patient traits (and behaviors) are important determinants of performance. Finally, interventions that can change physician behavior when warranted (related to intensification of diabetes treatment) have often not been well studied. As delivery-system reforms, these topics should be of great interest to clinicians, patients, health systems, and policy makers seeking better results and greater value from our healthcare system. 

Articles

WEB EXCLUSIVE

The authors undertook a literature search on treatment intensification in diabetes care to identify relevant research that has been published since 2000.

Provider Behavior and Treatment Intensification in Diabetes Care

 The employment of more nurse practitioners (NPs) is one of the most promising ways to expand the capacity of medical group practices. The objective of this study was to determine the association of NPs with patient-level cost and quality of care.

 Eighty-five primary care medical group practices were matched with 315,000 Medicare patients. Per beneficiary per year total costs and quality of care were calculated from Medicare claims data. Data were analyzed using multivariate regression analysis.

 A national sample of primary care medical group practices based on responses to the 2009 Medical Group Management Association Performance Survey. The cost variable was annual risk-adjusted Medicare expenditures per capita for patients attributed to a practice. There were 5 quality of care measures.

Employing NPs in primary care practices is associated with increased risk-adjusted patient cost for up to 1 NP for every 2 physicians, but cost decreases as the number of NPs per physician increases. There was little evidence of systematic association of NPs with quality of care or the practice’s net revenue.

Primary care medical group practices need to evaluate the alternate clinical roles of their NPs and develop models that optimize cost and quality of care. Practices that have employed more than 1 NP for every 2 physicians appear to have lower per capita Medicare spending with no adverse effects on quality. Research now needs to explore these causalities. 

	Employing more nurse practitioners (NPs) to meet increased needs for primary healthcare could increase costs and have a negative effect on quality if their clinical roles are not clarified.

		Nurse practitioner staffing data indicate that employing 0.25 to 0.50 NPs per primary care physician increases patient-level costs and decreases quality.

		In this staffing range, this suggests that the clinical role of NPs has an adverse influence on performance. Lower staffing has no effect on costs or quality, but employing two NPs per physician reduces costs and improves quality. 

		Medical administrators should evaluate the influence of NP staffing ratios on their clinical roles and subsequent cost and quality performance.

The employment of more nurse practitioners (NPs) is one of the most promising ways to expand the capacity of medical group practices to meet the projected needs for primary healthcare created by an aging population and increased health insurance coverage under the Affordable Care Act. However, little is known about the optimal use of these clinicians in primary care medical practices or the effect of NPs on the costs and quality of care. In this manuscript, we report the findings from a study of the association of NPs in 85 primary care medical group practices with risk-adjusted Medicare expenditures per patient (hereafter referred to as “cost”), quality of care, and the practice’s net revenue.

Several recent studies have highlighted the pending shortage of primary care physicians (PCPs). The Health Resources and Services Administration identified more than 5000 primary care shortage areas as of January 2013. These areas affect more than 57 million people, and almost 16,000 additional practitioners are required to meet primary care needs.

 Moreover, Hofer et al estimated that an additional 4307 to 6940 PCPs would be required just to meet the needs of the newly insured under the Affordable Care Act.

 A study by Huang and Finegold expanded this analysis to identify geographic areas disproportionately affected by the combination of insurance expansion and the current shortage of PCPs.

 They estimated that 51 million Americans live in the areas that will experience the greatest effect of primary care shortage.

While numerous studies have focused on the potential contributions of and need for more NPs,

 few have examined their association with costs; and to our knowledge, none have assessed their association with practice-level costs or the relationships between those costs and quality of care. An extensive recent review of the contributions of advanced practice nurses to primary care conducted by Laurant et al (2014)

 identified over 1100 publications, but only 3 studies assessed resource utilization such as consultation rates and prescription drug use, and 2 addressed direct costs, including labor costs per patient visit.

Most of these studies found the performance of advanced practice nurses to be equal to or better than that of physicians. While these studies provide important insights into the performance of NPs, they do not address the challenges faced by primary care medical group practices when they participate in accountable care organizations or similar programs in which patient population—level costs and quality are the main performance measures. As pointed out by Roblin et al (2004),

 while labor costs per patient visit decline in managed care organizations as more PAs and NPs are employed, more research is needed on delivery system cost savings and cost-effectiveness. This is the focus of our study.

The data set for this analysis was created by identifying the adult primary care medical group practices that provided organizational and performance data for the Medical Group Management Association (MGMA) 2009 survey. Data on 85 practices were then linked to claims data on the Medicare beneficiaries who received a plurality of their nonhospital evaluation and management visits from the practices.

The practices ranged in size from 5 to 53.2 full-time equivalent (FTE) physicians, with half owned by hospitals. Seventy-three percent have at least 1 NP and 46% have electronic health records (EHRs). The practices are located in 26 states, and although this is not a nationally representative sample of primary care medical group practices, it covers many different regions of the United States and provides a relatively large and diverse convenience sample to support our analysis.

Costs are calculated from Medicare claims files and are risk-adjusted using the Medicare Hierarchical Condition Category algorithm plus patient age and gender. Differences in payment rates are adjusted using the Medicare Geographic Practice Cost Index. The resulting per beneficiary per year (PBPY) adjusted costs represent dollar-weighted resource use to provide and manage care for similar patients.

We also include a net revenue after operating costs variable in our analysis to determine the association of NPs with the profit level of the practice. This is an important measure of the relationship between practice-level revenue and cost of care, and a measure of the contribution of NPs to the efficiency of the medical group practice in providing services.

Five quality-of-care measures are included in our analysis: 3 Healthcare Effectiveness Data and Information Set (HEDIS) variables, non-emergent use of emergency departments (EDs), and ambulatory care—sensitive (ACS) hospitalization rates. These variables are calculated from the Medicare claims files and are specified as follows:

 These include a) mammography rates: mammography rates compared with clinical guidelines for breast cancer screening; b) cardiovascular disease/low-density lipoprotein (CVD/LDL) rates: LDL test rates for patients with CVD; and c) glycated hemoglobin (A1C) rates: glycated hemoglobin test rates for patients with diabetes.

 These include a) non-emergent ED rate: proportion of ED visits classified as non-emergent based on the Billings algorithm

; and b) ACS hospital rate: proportion of hospital admissions classified as resulting from inadequate ambulatory care using the algorithm available from the Agency for Healthcare Research and Quality.

The NP variable is entered into our analyses as the ratio of FTE NPs to FTE physicians and NPs in each practice. Thus, 1 NP and 2 physicians would produce a ratio of 1:3, or 0.33. Since other practice-level attributes have been shown to be associated with cost and quality variables in our previous research, we include 5 of those as control variables in this analysis.

 Those variables, taken from the MGMA survey, are specified as follows: a) practice ownership: physician-owned, hospital-owned, other-owned (eg, community health centers, health insurance plans, local government agencies); b) net revenue: net revenue after operating costs as a percentage of total net revenue; c) rural versus urban location: rural if in a community of 5000 or fewer population; d) EHR: practice has an EHR; and e) years with EHR: number of years since implementing EHR.

We compiled data on the number of FTE physicians in the practice, but dropped these data from the analyses at the suggestion of a reviewer who pointed out that it could be a mediator between the NP variables and the outcome variables. Dropping them had little effect on the NP coefficients. We also compiled data on the number of patients attributed to each practice. That variable was not significant in any analyses at the 0.05 level, and its inclusion had only a minor effect in 1 regression, as discussed in the Results section.

All of these analyses used ordinary least squares regression equations to estimate the association of the explanatory variables, including the NP/physician staffing ratios with patient-level cost and quality and practice-level economic efficiency. As in most organizational-level studies, organizational characteristics are not assigned randomly to organizations, but are “endogenously” chosen by the organization. Thus, we cannot say that a practice that hires an additional NP will experience the effect on the dependent variables that we estimate from our data. However, we can say that practices that have found a way to accommodate a higher ratio of NPs to physicians exhibit the differences (or lack of differences) in the dependent variables that we find in our analyses.

Descriptive statistics for the variables in the analysis are shown in 

. The descriptive statistics are shown separately for practices employing no NPs versus some NPs. Only 2 of the differences are statistically significant at the 0.05 level. Practices employing NPs have higher rates of non-emergent ED visits and lower rates of CVD/LDL testing than practices not employing NPs.

Multivariate regression results begin in 

. The coefficient of no nurse practitioners per physician (NPPs) shows that on average, Medicare costs are $445 higher per patient for practices that do not employ NPs than practices that employ some NPs, but the difference is not statistically significant. The coefficient of the NP staffing level in the risk-adjusted cost equation is 9374.440 and the coefficient of the squared term is —14,939.620. Both are statistically significant at close to the 0.05 level. (Including the number of patients attributed to the physician’s practice in that regression improves the precision of those estimates, resulting in a level of significance of 0.046 on the square of the ratio, even though the coefficient of the number of patients is itself not statistically significant.) Those results imply that among practices with some NPs, risk-adjusted PBPY costs are higher up to 0.32 and decline thereafter as the ratio increases. A ratio of 0.32 corresponds roughly to 1 NP for every 2 physicians: 1 NP divided by 1 NP + 2 physicians. In contrast, the mean ratio across the entire sample (including practices that employ no NPs) is 0.154, or about 1 NP per 6.5 physicians. However, among practices employing some NPs, the ratio is 0.211, or about 1 NP per 4.7 physicians. Both ratios are below the cost-maximizing ratio.

Our risk-adjusted cost variable represents a cost to the Medicare program, but it is revenue to the practice, and thus one might expect the NP staffing ratio to exhibit the same nonlinear relationship to the practice’s net revenue, but that is not the case. It is important to remember that the net revenue variable is based on all the practice’s patients, whereas the risk-adjusted PBPY cost variable is computed only for the practice’s Medicare patients.

The number of years the practice has had an EHR is positively associated with net revenue, as is rural location. More experience with EHRs might improve net revenue by improving patient scheduling and other patient flow processes or by capturing more billable services. Rural practices might benefit from lower overhead costs and support staff ratios.

The NP staffing ratio does not appear to be related to non-emergent ED use ratios, but higher ratios are associated with higher ACS admissions (

). The square of the NP staffing ratio is negative in both equations in Table 3, but not statistically significant. Increased ACS admissions may be one of the ways that use of NPs is associated with higher costs up to the ratio of 0.32. However, the extremely poor overall fit of the ACS regression results in a negative 

 after making the adjustment for the number of explanatory variables. Consequently, it is difficult to interpret these results. Having an EHR is statistically related to lower non-emergent ED rates, but the years of EHR experience is associated with increased non-emergent ED rates.

The relationships of NP staffing ratios and the HEDIS quality measures are similar to the avoidable utilization findings (

). NP staffing ratios are not statistically related to performance on any of these measures.

Physician-owned practices have higher A1C testing rates, as do practices with an EHR. Practices in the other ownership category (owned by a community health center, the government, or a health plan) have higher A1C and mammography rates. Rural practices have lower CVD/LDL rates but do not differ on the other measures.

Our data indicate that Medicare costs for primary care medical group practices are associated with the ratio of NPs to NPs plus physicians, but this relationship is nonlinear. The cost-maximizing ratio is 0.32, or about 1 NP for every 2 physicians. Current ratios for all practices and those employing some NPs lie below that level. Ratios higher than 0.32 are associated with lower patient costs, but practically not at all with quality.

The theoretical model proposed by Liu and D’Aunno (2012)

 suggests that these performance differences may result from differences in the clinical roles of NPs at various staffing levels. NPs perform a variety of clinical roles ranging from replacing a physician to assisting the physician. Unfortunately, little is known about the configuration of these clinical roles or how they influence costs and quality of care. This is an important area for future research if the full potential of these nonphysician clinicians is to be achieved.

If higher NP staffing levels are related to the development of team-based patient care, this could account for an association with decreased patient-level costs, although it does not appear that quality of care is improved, nor does it result in improved net revenue in the practices. The NP staffing ratio is unrelated to practice net revenue despite the (nonlinear) relationship of NPs to Medicare costs and the obvious reduction in patient-care costs to the practice if NPs are substituted for physician labor. As noted by Grumbach and Bodenheimer (2004)

 staffing is only 1 part of team care. Much of the effectiveness is related to organizational and cultural factors.

Certainly, more research is needed to document the factors influencing these performance measures and how the practice determines the clinical roles of the NPs once they are employed, as well as how practices bill for the time of NPs, and how the roles of NPs and physician supervision of NPs differ among practices.

Because practices are choosing their NP staffing levels in addition to many other practice characteristics, it is likely to prove difficult to estimate the true causal effect of NPs or other endogenous practice characteristics on healthcare costs. We cannot say that medical practices will experience the outcomes indicated by our estimated coefficients if they alter the number of NPs they employ. Perhaps there are unobserved variables that influence decisions to hire NPs and also influence patient-level cost and quality. For example, it is possible that practices with more costly patients, even after standard risk adjustment, hire more NPs. More research is needed to explore potential confounders and the causal pathways that link the use of NPs to cost and quality of care.

Our inability to give a causal interpretation to our coefficients makes them less useful for practices or policy makers who are attempting to change practice outcomes by changing the level of NP staffing. However, our findings alert medical group practice administrators and physicians to the need to pay more attention to the roles of nurse practitioners when adding them to their clinical teams. 

 Division of Health Policy and Management, School of Public Health, University of Minnesota (JK, BD), Minneapolis, MN; Medica Research Institute (JK, AC, MS, JT), Minnetonka MN.

 Funding provided by the Robert Wood Johnson Health Care Financing and Organization initiative.

 The authors do not have any conflict of interest relationships related to this research, and the corresponding author attests that all individuals who contributed to the manuscript have been appropriately acknowledged.

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

The potential of nurse practitioners is not being fully realized in primary care medical practices. Consequently, cost and quality gains are not being achieved.

The Role of Nurse Practitioners in Primary Healthcare

 To assess the impact of a patient engagement intervention utilizing the Medication User Self-Evaluation (MUSE) tool on the completion percentage of comprehensive medication reviews (CMRs) among Medicare Part D beneficiaries.

 Beneficiaries from 2 Medicare Part D plans were randomly assigned to 3 study arms (1 control arm plus 2 intervention arms for 2011 and 2012, respectively). Each beneficiary who participated in the MUSE intervention met 3 inclusion criteria and was matched with a single control group beneficiary based on: gender; age (within 5-year interval); plan type (ie, Medicare Prescription Drug Plan, Medicare Advantage Prescription Drug Plan); number of unique prescriptions; pharmacy medication therapy management (MTM) training status; and time period (2011, 2012). The outcome of interest was whether or not the beneficiary received a CMR in the 6 months following the index date. Generalized estimating equation (GEE) models were used to compare CMR percentages over time and between MUSE intervention groups. This study used MTM service claims data.

 The final sample of 1015 beneficiaries received MUSE intervention, of which 1007 were successfully matched to a control beneficiary. The estimated odds of having a CMR among those who received the MUSE intervention were 2 times that of their counterparts (P = .0048) across both study years. 

 Given the strong evidence found for a positive association between participation in a CMR and the MUSE intervention, Part D plans could use the MUSE to engage targeted beneficiaries in using pharmacist-provided MTM services.

	Among Medicare Part D beneficiaries, the low uptake of comprehensive medication reviews (CMRs) remains a concern. To our knowledge, there has been limited effective stimulation of beneficiaries to seek CMRs. This study assessed the impact of a Medication User Self-Evaluation (MUSE) intervention tool designed to engage beneficiaries’ use of CMRs.

		The MUSE tool was found to be associated with a higher completion rate of CMRs, providing significant empirical evidence of the importance of enhancing awareness and applying personalized information. 

		It is feasible that the MUSE tool can be used by Part D plans and other stakeholders to help engage targeted beneficiaries to use such pharmaceutical services.

Medicare Part D plans have been required to offer medication therapy management (MTM) programs to eligible enrollees since 2006.

 CMS has expanded the scope of MTM services since then, further acknowledging the benefits of avoiding drug-related problems, optimizing medication outcomes, and decreasing healthcare costs.

 Currently, the comprehensive medication review (CMR) serves as a fundamental component of that expansion, as Part D plans must offer an annual CMR to all eligible MTM program enrollees.

 An annual CMR must involve an interactive review and consultation that cover all of a beneficiary’s medications obtained through contact with a pharmacist or other qualified provider.

Despite this requirement, the participation in CMRs has been lower than anticipated. The national median CMR completion rate among MTM eligibles was only 10.9% based on the 2014 display measure report, using the 2012 performance data submitted by Part D plans.

 CMS has attempted to improve the situation by imposing additional requirements, including the use of a 3-component standardized format—a CMR cover letter, a medication action plan, and a personal medication list&mdash;for a written summary of CMR results, setting the CMR completion rate as a display quality measure for plans, and possibly applying the CMR completion rate to Star ratings for drug plans in 2016.

 Still, the policy changes alone have had limited effect on engaging Part D beneficiaries to use CMRs, because they regulate the activity of plan sponsors but not the behavior of beneficiaries.

One barrier to CMR participation revealed in previous studies is that the majority of eligible Medicare Part D beneficiaries are unaware of or do not understand MTM services. These factors result in relatively low expectations of personal benefit from such services, including CMRs, and underscore the importance of finding effective ways to encourage beneficiaries to participate in MTM services.

 Recently, in exploring Australian patients’ utilization of Home Medicines Review (HMR)—a pharmacist-provided service available in Australia for patients’ medication assessment&mdash;Carter et al

 found that increased patient expectations of receiving useful medication information can motivate eligible patients to use that service. Although many of the details of HMR differ from CMR, the two are similar in that a pharmacist conducts a comprehensive interview with the patient about their medications. The findings of Carter et al are informative to us: it is likely that highlighting the CMR as a valuable medication information source would stimulate Medicare beneficiaries’ engagement in MTM services. Also, the literature has shown that patients generally seek more health information than what they receive

; specifically, they would like information that is tailored to them during their physician visits,

 which could translate to getting medication information in a CMR during a pharmacy visit.

 These findings suggest the importance of exposing Part D beneficiaries to personalized information regarding eligibility and potential benefits from the CMR. Offering such personalized advice may require direct input from Part D beneficiaries, such as health information and behaviors. It is insufficient to rely solely on prescription claims, as Medicare Part D plans often do.

Motivated by these findings, a patient engagement intervention was implemented as part of an MTM quality improvement project. The intervention utilized the Medication User Self-Evaluation (MUSE) tool (

), which was validated to be useful in targeting MTM services among Medicare Part D beneficiaries by estimating their likelihood of benefiting from the services.

 The MUSE incorporates 7 items as a self-reported questionnaire: it asks about the number of medical conditions; whether the patient has forgotten to take medications; if cost-related issues exist regarding prescriptions; if there have been hospitalizations in the past 6 months; and the number of prescribed medications, physicians, and pharmacies utilized. Part D beneficiaries’ responses were used to provide their own likelihood of benefiting from a CMR. It was expected that calling Medicare beneficiaries to complete the MUSE would stimulate them to seek a CMR if they were given feedback to expect a benefit from a CMR.

The objective of this study was to assess the impact of the MUSE intervention on the completion percentages of CMRs among a cohort of Medicare Part D beneficiaries.

OutcomesMTM, an MTM program administrator, provided a sampling frame of eligible Medicare beneficiaries and staff for the implementation of the MUSE intervention. OutcomesMTM has extensive experience in MTM program administration, having administered pharmacist-delivered clinical intervention programs for more than 10 years, covering more than 5 million patients. The pharmacists in OutcomesMTM’s network have been trained to deliver MTM services, identify potential drug therapy problems, and take steps to resolve such issues using a secure Web-based documentation and billing platform. The participant pharmacies and beneficiaries in this study were from 2 of OutcomesMTM’s Medicare Part D clients: a Medicare Prescription Drug Plan (PDP; ie, Plan A) and a Medicare Advantage Prescription Drug plan (MA-PD; ie, Plan B). The eligibility criteria for both plans were consistent over the study period.

The study design was a case-control study implemented in 2011 and 2012. The MUSE intervention cases were 2 targeted subsets—1 in 2011 and 1 in 2012&mdash;randomly selected to be offered the MUSE intervention, while controls were matched from a control arm. Both cases and controls received the usual MTM services through OutcomesMTM’s program, but only the intervention group was offered the MUSE. Two time periods allowed separation of temporal effects from study intervention effects.

Within the larger sampling frame of beneficiaries, only those who met 3 inclusion criteria were selected randomly to receive the MUSE intervention; criteria included being 65 years or older, having received at least 12 prescription fills of any medication during a 6-month pre-intervention period (April 1 to September 30 in 2010 or 2011), and having been continuously enrolled during the time frame of interest (2 prescription fills at least 150 days apart).

 This ensured that beneficiaries were eligible for a CMR under Medicare Part D, were continuously enrolled through the monitoring window, and were taking several medications for chronic conditions. The final targeted sample sizes for the MUSE intervention were 2843 beneficiaries in 2011 and 3247 in 2012. The study was approved by the University of Iowa Institutional Review Board.

The patient engagement intervention (ie, MUSE intervention) was designed to stimulate participation in CMRs among Medicare Part D beneficiaries. Beneficiaries in the intervention arms received a letter from OutcomesMTM describing MTM services, and notification of a patient engagement telephone call. Operationally, the contact process was done on a rolling weekly basis to notify beneficiaries of the phone call that they would receive within the upcoming week. The beneficiaries in the control arm were not contacted by study personnel during the study period. The week after the mailing, outreach personnel at OutcomesMTM called beneficiaries, primarily during the day. If a contacted beneficiary agreed to participate, the outreach staff member and beneficiary worked through the MUSE tool together during the phone call.

The call script was programmed into a Microsoft Access database, which also collected the beneficiary’s responses during the call. Access was programmed to immediately predict the likelihood of benefit from receiving a CMR (ie, low, moderate, and high), using beneficiary responses and a validated prediction equation.

 The beneficiary was told their predicted category and given specific information on getting a CMR, though CMR appointments were not made during the call. It was the beneficiary’s choice to decide whether or not to take the next step to schedule a CMR on their own with their local pharmacist.

Using de-identified claims data, each beneficiary who participated in the MUSE intervention was matched to a single control beneficiary from the same time period (2011 or 2012) from the control group who did not receive the outreach phone call with the MUSE tool. Using information on prescription fills and demographics from OutcomesMTM, an exact match was required for plan, number of unique prescriptions, pharmacy training status, and gender, while age was matched to within 5 years. This process used a greedy algorithm and was conducted for each time period. When multiple matches were found, 1 was randomly selected and the remaining ones were returned to the candidate pool.

Only pharmacies in OutcomesMTM’s pool that are designated as “trained and contracted” are able to bill for CMRs; thus, MTM training is an important factor on which to be matched when evaluating the effect of the MUSE on CMRs. A beneficiary was considered associated with a trained and contracted pharmacy if they had obtained at least 1 prescription in the year of interest from such a pharmacy.

The study outcome was whether or not the beneficiary received a CMR in the 6 months following the index date. For the intervention arms, the index date was the date of the telephone contact in the MUSE intervention. Matched beneficiaries from the control arm were given the same index date so that the pair would be monitored for a CMR in the same 6-month period. The chosen observation window was assumed to give potential beneficiaries sufficient time to complete a CMR if they wanted to do so, but not so long as to become unrelated to the MUSE intervention. The intervention year was incorporated to allow assessment of change in environment over time. CMRs performed outside the 6-month window were not counted.

Generalized estimating equations were used to model CMR percentage over time and between those who did and did not receive the MUSE intervention, incorporating correlation due to the matching. The MUSE intervention indicator was 1 if the beneficiary received the intervention, and 0 if they did not. Year was also treated as a binary variable, set to 1 for 2012 and 0 for 2011. Two-way interactions were explored. The logit link and binomial distribution for the outcome were assumed. All statistical analyses were performed in SAS version 9.3 (SAS Institute, Cary, North Carolina).

In 2011, 348 MUSE records were received by the study analytic team from OutcomesMTM. This reflected a 12.2% acceptance rate of MUSE intervention (ie, responding to the MUSE questionnaire) among the initial sample (n = 2843). In 2012, 667 MUSE records were received, representing a 20.5% acceptance rate among the sample (n = 3247). Reasons for the low acceptance rate reflected several issues. According to records of those making MUSE phone calls, nearly half of the study sample could not be reached (eg, their registered telephone number was no longer correct or they were not home to answer the call). Voicemail messages were left for those who had answering machines, and if the beneficiary did not respond by returning the call after 2 messages were left, they were deemed to have opted out of the study. Some individuals who answered the phone simply chose not to participate and opted out at that time.

Excluding those who opted out or could not be reached, the final sample size of those who participated in the MUSE intervention was 1015, of whom 1007 were successfully matched to a control beneficiary (

). Relaxation of the matching criteria was only moderately successful in matching the other 8, so they were eliminated from the analyses. There were 343 and 664 matched beneficiaries in the 2011 and 2012 groups, respectively. Of those, 4.77% (n = 48) of the intervention group and 2.38% (n = 24) of the control group had a CMR in either 2011 or 2012.

Using the quasi-information criterion for model selection,

 the interaction of intervention/control with year was dropped. Thus, in the final model, the likelihood of having a CMR was modeled as a function of receiving a MUSE intervention and the intervention year, with no interaction. This implied that the interaction term was not needed, as the effect of time did not differ across the 2 groups.

Based on this model, the estimated odds of having a CMR among those who received the MUSE intervention were double that of their corresponding control beneficiaries (

). In 2011, 2.33% of the MUSE intervention participants had a CMR in the 6-month observation period, whereas 0.58% of the control group beneficiaries had one, resulting in an observed odds ratio of 4.09. In 2012, 6.02% of the beneficiaries in the MUSE intervention had a CMR, whereas 3.31% of the control group did so, resulting in an observed odds ratio of 1.87. Since the yearly odds were not significantly different, the modeled pooled estimate of the odds ratio of the MUSE intervention over control was 2.06.

A large increase in the odds of having a CMR was seen between 2011 and 2012 in both study groups. The odds of having a CMR in 2012 were estimated to be more than 3 times the odds of doing so in 2011 (

= .0004), across both groups. Members of the control group had observed odds of having a CMR in 2012 that were 5.87 times larger than the corresponding odds in 2011. In the MUSE intervention group, the observed odds of a CMR were 2.69 times higher in 2012 than in 2011. The final model estimate, which combines information in the intervention and control groups, was an odds ratio of 3.33 for the year of 2012 over 2011.

In this study, the delivery of the MUSE was associated with a higher completion percentage of CMRs, implying that the MUSE intervention engaged beneficiaries in seeking a pharmacist-provided medication review. Three possible explanations for this association can be considered: first, the MUSE intervention includes interactive communication between outreach personnel and the beneficiaries. The interactive communication is deemed to be more compelling than 1-way communication, encouraging participants to pay closer attention and to remember more of the information that was delivered.

 Similarly, the MUSE intervention would be more noteworthy to the beneficiaries than a general promotional activity such as a flyer introducing the CMR service. Interpersonal contact could increase the likelihood that beneficiaries become aware of the CMR during the MUSE intervention, which could stimulate future CMR participation.

Second, earlier research showed that interventions tailoring information to individuals are more effective than untailored ones in promoting health behavior change.

 Such tailoring could contribute to the effectiveness of the MUSE intervention, given that MUSE recipients received personalized advice about their potential to benefit from receiving a CMR. As such, the advice—versus non-personalized advice&mdash;would be more meaningful during the period when beneficiaries were deciding whether or not to obtain a CMR. This is also supported in a study by Tang and colleagues,

 suggesting that individually tailored information is most welcome to patients.

Third, through the MUSE phone call, beneficiaries’ expectations of benefiting from a CMR could enhance their willingness to receive a CMR, and in turn stimulate their engagement in such pharmacist-provided medication review services. Having perceived value from a health service mirrors the recent findings of an Australian research group.

 By introducing the CMR as a low-cost medication information resource, the MUSE intervention may have led beneficiaries to expect an outcome of reduced medication concerns through participation in a CMR in the future.

It was observed that the overall CMR percentage increased from 2011 to 2012. One explanation for this may be the influence of CMS on Medicare plans. CMS has implemented a series of policy changes for MTM programs, focusing on Medicare plans; these changes include the obligation to offer an annual CMR, the requirement of using a standardized format for giving patients a summary of the CMR, and the inclusion of the CMR completion rate as a display quality measure in Star ratings going forward. These policy changes emphasized the importance of CMRs so that plans would pay more attention to promoting them. In addition, pharmacies that have contracts with multiple Part D plans would have seen increased demand for CMRs, and consequently expanded their CMR capacity.

Another potential explanation focuses on interpersonal influence by physicians, friends, or family members. General practitioners’ social influence has been found to affect patients’ acceptance of such medication review services in Australia.

 The strong relationship between a patient and physician, involving high levels of trust, could have a direct influence on patients’ health behaviors. Specifically, if the physician promoted the usefulness of pharmacist-provided medication review services, the patient would be more likely to believe that participation in a CMR is beneficial. With the recent growth of MTM services, it is believed that physicians became more supportive, which in turn increased patients’ overall acceptance of the CMR over the 2 years. As more and more beneficiaries experience CMRs, they might talk to their family members or friends about the benefits of such services.

 These positive word-of-mouth messages could have contributed to the observed increased CMR percentages as well.

The 7-question MUSE tool was effective in collecting helpful patient-reported information within a short period of time. Having such tailored information can improve the quality and efficiency of the interactive communication between the outreach personnel and the beneficiaries. Thus, the use of tools, such as the MUSE intervention, is promising for stimulating medication reviews in the future due to their potential to improve the efficiency and personalization of interactive communication through standardized interventions.

As with any study, some limitations exist. Only 2 Medicare Part D plans were included, which restricts the generalizability of the findings, though one was a PDP and the other was an MA-PD. Also, some beneficiaries could not be invited to participate in the MUSE intervention because the available contact information was either not current or inaccurate. Such poor maintenance of accurate contact information reinforces the importance of updating patients’ details in a timely manner to support likewise timely outreach activities. Additionally, beneficiaries were responsible for scheduling a CMR themselves with the local pharmacist; even if the MUSE score indicated a high likelihood of benefiting from a CMR, while they might have fully intended to do so, some otherwise willing patients may not have made an appointment with the pharmacist after the MUSE call.

Future research is needed in several areas. More beneficiaries from a variety of Medicare Part D plans could be examined to see if a patient engagement tool, such as the MUSE, stimulates participation in CMRs. For example, this study involved community pharmacists providing the CMRs. Some Part D plans rely only on their own pharmacists providing such services over the telephone. The use of the MUSE tool likely could help these plans target their telephone outreach.

This article describes a study of an intervention to engage Medicare Part D beneficiaries in obtaining a comprehensive medication review.

Stimulating Comprehensive Medication Reviews Among Medicare Part D Beneficiaries

 To describe the lifetime outcomes and economic implications of combinatorial pharmacogenomic (CPGx) testing versus treatment as usual (TAU) psychopharmacologic medication selection for a representative major depressive disorder patient who has not responded to previous treatment(s).

Markov state-transition analysis based on clinical studies.

Clinical validity and utility were based on published findings in prospective clinical studies of a commercially available CPGx test. Data for quality of life, direct costs, and indirect costs were extracted from meta-analyses of published literature on clinical studies and claims databases. Outcomes were assessed from a societal perspective, and included differences between the CPGx and the TAU strategies in quality-adjusted life-years (QALYs), cumulative direct and indirect costs, and cost per QALY gained.

CPGx improved the treatment response rate by 70% (1.7 times as high as that with TAU), increasing QALYs by 0.316 years. With these health benefits, CPGx is expected to save $3711 in direct medical costs per patient and $2553 in work productivity costs per patient over the lifetime. The cost-effectiveness of CPGx testing was robust over a wide range of variation in the input parameters, including the scenario when testing efficacy was set to its lower limit.

CPGx testing has been shown by prospective studies to modify treatment decisions for patients nonresponsive to previous treatment(s), with increased rates of treatment response. These effects are projected to increase quality-adjusted survival, and to save both direct and indirect costs to individual patients and society generally.

	Combinatorial pharmacogenomic (CPGx) testing dominates the treatment-as-usual approach in managing major depressive disorder patients who did not respond to previous treatments.

		Prospective clinical studies across multiple settings have shown that CPGx testing helps to guide treatment decision-making, matching patients with the treatments that are most likely to be safe and effective.

		Further analysis indicates that CPGx testing is related to improved quality of life and cost savings.

Major depressive disorder (MDD) is associated with significant clinical and economic burden worldwide, with the United States

 experiencing a 16.6% lifetime prevalence rate.

 Although guidelines recommend antidepressants as well-validated treatment options, not all individuals will respond and/or achieve remission with an initial treatment and little evidence exists to guide next-step treatment, resulting in a trial and error approach to drug selection.

 In the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial, the rate of response to an initial antidepressant treatment was only 49.6%, which declined with an increasing number of treatment trials.

 A systematic review of published studies found that patients nonresponsive to 1 or more treatments have a 15% likelihood of suicide ideation, an approximately 17% likelihood of a suicide attempt, and a 10% incidence of severe adverse events; the net effect being a 0.26 absolute reduction in health utility compared with treatment-responsive patients.

 Moreover, the annual medical costs for managing a patient nonresponsive to treatments were nearly $10,000 more than those for a treatment-responsive patient, in 2012 US$.

Studies have indicated significant association of pharmacogenomic (PGx) stratification with patients’ response to treatment, quality of life (QoL), work productivity, and the related costs.

 Perlis et al found that a hypothetical PGx test for a newly diagnosed MDD patient was associated with a relative risk of recovery of 1.28, and the cost/quality-adjusted life-year (QALY) gained varied from $1000 to more than $50,000, depending on input parameters.

 Because some studies indicated a weak association between certain single-nucleotide polymorphisms and antidepressant response,

 an integrated analysis of multiple polymorphisms has been proposed as a solution to the pharmacotherapy “treatment odyssey,” because this may allow the assessment of genetic variations that influence the activity of multiple enzymes in drug metabolism. Clinical data and adjunctive analyses are needed to estimate efficacy and to enable pharmacoeconomic analysis of combinatorial tests.

Prospective studies have recently provided evidence of improved antidepressant response rates associated with the use of a multi-gene combinatorial pharmacogenomic (CPGx

Using a state-transition probability analysis, this study assessed the long-term outcomes and economic implications of combinatorial pharmacogenomic testing for managing patients with major depressive disorder who were nonresponsive to treatments.

June 2015

Provider Behavior and Treatment Intensification in Diabetes Care

The Role of Nurse Practitioners in Primary Healthcare

Stimulating Comprehensive Medication Reviews Among Medicare Part D Beneficiaries

Cost-Effectiveness of Combinatorial Pharmacogenomic Testing for Treatment-Resistant Major Depressive Disorder Patients

Medication Adherence and Measures of Health Plan Quality

How Will Provider-Focused Payment Reform Impact Geographic Variation in Medicare Spending?

Improving Diabetic Patient Transition to Home Healthcare: Leading Risk Factors for 30-Day Readmission

Impact of the Patient-Centered Medical Home on Veterans' Experience of Care

Quality of Care and Relative Resource Use for Patients With Diabetes

Higher 30-Day and 60-Day Readmissions Among Patients Who Refuse Post Acute Care Services

Moving From Healthcare to Health