Abstract
Medicare began reimbursing for outpatient diabetes self-management training (DSMT) in 2000; however, little is known about program utilization. Individuals diagnosed with diabetes in 2010 were identified from a 20% random selection of the Medicare fee-for-service population (N = 110,064). Medicare administrative and claims files were used to determine DSMT utilization. Multivariate logistic regression analyses evaluated the association of demographic, health status, and provider availability factors with DSMT utilization. Approximately 5% of Medicare beneficiaries with newly diagnosed diabetes used DSMT services. The adjusted odds of any utilization were lower among men compared with women, older individuals compared with younger, non-Whites compared with Whites, people dually eligible for Medicare and Medicaid compared with nondual eligibles, and patients with comorbidities compared with individuals without those conditions. Additionally, the adjusted odds of utilizing DSMT increased as the availability of providers who offered DSMT services increased and varied by Census region. Utilization of DSMT among Medicare beneficiaries with newly diagnosed diabetes is low. There appear to be marked disparities in access to DSMT by demographic and health status factors and availability of DSMT providers. In light of the increasing prevalence of diabetes, future research should identify barriers to DSMT access, describe DSMT providers, and explore the impact of DSMT services. With preventive services being increasingly covered by insurers, the low utilization of DSMT, a preventive service benefit that has existed for almost 15 years, highlights the challenges that may be encountered to achieve widespread dissemination and uptake of the new services.
Diabetes affects a rapidly growing number of people in the United States (Centers for Disease Control and Prevention [CDC], 2011), often in concert with a range of comorbidities, and it can lead to serious complications resulting in poor health outcomes and reduced quality of life (Narayan, Gregg, Fagot-Campagna, Engelgau, & Vinicor, 2000). Diabetes is particularly prevalent among Medicare beneficiaries, who are more than 65 years old and/or are disabled: more than 1 in 4 adults enrolled in Medicare fee-for-service (FFS) insurance have diabetes (Health Indicators Warehouse, 2013). Millions more have prediabetes, with nearly 400,000 new cases of diabetes per year in older adults alone (CDC, 2011). Caring for diabetes is costly, both for the individual and for the Centers for Medicare & Medicaid Services (CMS), with 1 in 3 Medicare dollars being spent on diabetes-related health services (CMS, 2012a).
Generally, research on self-management programs for adults with diabetes has indicated that these programs can improve health outcomes and reduce costs (Boren, Fitzner, Panhalkar, & Specker, 2009; Duncan et al., 2011; Rosenzweig et al., 2010; Siminerio et al., 2006). Diabetes self-management interventions have varied in approach (e.g., telephone based; Rosenzweig et al., 2010]), theory (e.g., using elements of the Chronic Care Model; Siminerio et al., 2006), and target population (e.g., Medicare and commercial populations; Duncan et al., 2011), but overall despite their variation, most interventions were associated with positive benefits that outweighed the costs (Boren et al., 2009). Despite some limitations in study design, such as small sample size, limited number of outcomes, and generalizability to the Medicare population, initial evidence indicates positive effects of diabetes self-management programs.
In 2000, Medicare began reimbursing for outpatient diabetes self-management training (DSMT) by certified health care providers. Currently the American Diabetes Association and American Association of Diabetes Educators (AADE) are authorized by CMS to certify providers’ eligibility for Medicare reimbursement. For providers to be certified, they must meet the National Standards for Diabetes Self-Management Education, and compliance is monitored over time (Indian Health Service, Division of Diabetes Treatment and Prevention, 2011). DSMT programs are available in a variety of settings, including hospitals, physician practices, diabetes centers, and other community sites such as pharmacies, and are given on either an individual or group basis. Patients qualify for DSMT if, within the past year, they were diagnosed with diabetes, began taking medicine or insulin, were considered to be at risk for diabetes-related complications, or had diabetes when they became eligible for Medicare (CMS, 2012b). Medicare beneficiaries are eligible for up to 10 hours of DSMT for the initial year, including 1 hour of individual training, and up to 2 hours per year afterward. The services are reimbursable only if individuals have an order from their health care provider, which specifies the format (individual or group; CMS, 2014). Under the law, Medicare will make payment for DSMT services only to providers who (a) have received accreditation or represent an accredited entity and (b) are already established as Medicare providers in some other capacity (e.g., physicians, registered dieticians, hospitals, pharmacies).
DSMT programs use a curriculum based on current evidence and practice guidelines to target patients’ skills and knowledge about diabetes, nutrition, and exercise to promote effective self-management of their condition and prevent complications. However, little is known about the extent of DSMT utilization among Medicare beneficiaries, or the factors associated with whether or not an individual uses DSMT services. Literature on DSMT services for Medicare beneficiaries is limited but indicates that utilization of these services has been low (Czarnowski-Hill, 2007; Siminerio et al., 2006). The AADE reported that about 1% of Medicare beneficiaries with diabetes received DSMT services in 2004 and 2005 (AADE, 2010).
Individuals with diabetes are at risk for poor health outcomes and costly care, and DSMT services could improve their ability to self-manage their condition and avoid complications. Given the current evidence gap and initial reports that DSMT services may be underutilized, this study aims to examine recent DSMT utilization among Medicare beneficiaries and identify specific populations who may face barriers to accessing DSMT programs.
Method
Study Sample and Data
This analysis focused on patients who were newly diagnosed with diabetes to ensure the study population was eligible for DSMT services. Individuals were identified using an administrative file representing a 20% random selection of Medicare FFS beneficiaries in 2010, which is housed in the CMS Chronic Condition Warehouse, along with claims files. The file contains a diagnosis flag for diabetes with the date on which the coding definition was first met since either (a) the patient enrolled in Medicare or (b) January 1, 1999, whichever is more recent. The diabetes flag requires at least one diabetes-related inpatient, skilled nursing facility, or home health agency claim or two diabetes-related outpatient or health care provider claims in a 2-year period. Patients whose flag indicated the algorithm was met for the first time in 2010 were considered to be newly diagnosed. The study population was further limited to individuals living in the United States (i.e., the 50 states and Washington, D.C.) who were alive and continuously enrolled in Medicare FFS Parts A and B during all of 2009-2011. Since ESRD facilities cannot bill separately for DSMT services (CMS, 2013a), individuals with ESRD were excluded from the analysis. The resulting study sample was comprised of 110,064 individuals.
DSMT Utilization
Outpatient and carrier claims from 2009-2011 were searched for DSMT codes (Healthcare Common Procedure Coding System [HCPCS] codes G0108 and G0109); DSMT codes that occurred between 2 months before the diabetes flag’s date through 1 year after it were defined as utilization of DSMT services. A window around the flag’s diagnosis date was included since the diabetes coding definition requires multiple outpatient or provider claims, yet the patient may have been diagnosed and referred to DSMT on the date of the first diabetes-related claim.
Factors Associated With DSMT Utilization
The authors examined a range of factors for their association with DSMT utilization. Variables were selected from those available in Medicare FFS claims files, based primarily on the behavioral model of health service use developed by Andersen (1995), including individual-level demographic factors (gender, age, and race/ethnicity), geographic residence (Census region; U.S. Census Bureau, n.d.), Medicare-Medicaid dual eligibility status, and comorbidities (cancer, heart disease, Alzheimer’s/dementia, chronic kidney disease, chronic obstructive pulmonary disease [COPD], depression, arthritis, and stroke). The indicator variable for race/ethnicity included categories for White, Black, Hispanic, Asian, North American Native, other, and unknown race/ethnicity based on administrative Medicare enrollment data. Individuals who were dually eligible for Medicare and Medicaid in the month of their diabetes diagnosis flag were categorized as a partial or full dual eligible. Major comorbid chronic conditions were identified using the Chronic Condition Warehouse flags; individuals were considered to have the condition if the coding definition was met in 2010. Many of these factors are associated with poor glycemic control or may make diabetes self-management difficult (Ali, Bullard, Imperatore, Barker, & Gregg, 2012; Blaum et al., 2010).
Provider availability was analyzed at the county level as the number of unique providers billing Medicare for DSMT services in 2010, which was derived using 100% Medicare claims files, divided by the study population. The distribution of the rate per county was used to create a categorical variable based on quartile of provider availability per individuals with newly diagnosed diabetes.
Statistical Analysis
To describe utilization of DSMT among the study population, the overall number of DSMT codes per beneficiary, and the mean count of DSMT codes among those who used the services, were calculated. Each code occurrence represents 30 minutes of DSMT services; for example, if an individual received 1 hour of individual training on a given day, then 2 codes were counted. Individuals with at least one DSMT code were considered users of DSMT, and this binary variable was used to determine the overall utilization rate as well as the rate of utilization among specific subpopulations. A multivariate logistic regression model was estimated to analyze the association of various demographic and health status factors with DSMT utilization. Model covariates included the variables described above and model coefficients were exponentiated to produce adjusted odds ratios. Statistical analyses were performed using Stata 12.1 (StataCorp, College Station, TX).
Results
Overall, approximately 5% of Medicare beneficiaries with newly diagnosed diabetes used DSMT services between 2 months prior to and 1 year following their diagnosis. The mean count of DSMT codes among users was 2.78 (SD = 1.73), representing nearly 1.5 hours of services, with a range of 1-19, or 0.5-9.5 hours (Table 1). More than half (57.4%) of the billed codes were for group sessions (G0109), with the remainder (42.6%) billed for individual training (G0108). Unadjusted, descriptive results (Table 2) did not show a substantially different rate of utilization by gender, with women (5.50%) having slightly higher utilization than men (5.27%). However, there were differences in DSMT utilization by age, race/ethnicity, dual eligibility status, the presence of comorbidities, provider availability, and geographic residence. DSMT utilization was highest among the 65 to 69 and 70 to 74 year old age ranges (6.44% for both categories) compared with younger, disabled (<65 years: 5.91%) and older age groups (75-79 years: 5.60%; 80-84 years: 4.07%; 85+ years: 1.61%). White and Black individuals had the highest prevalence of DSMT utilization (5.78% and 4.38%, respectively), followed by North American Native individuals (3.84%), individuals with unknown (3.67%) and other race/ethnicity (3.66%), Hispanic individuals (2.23%), and Asian individuals (1.51%). Dual eligible beneficiaries had lower DSMT utilization (ranging from 3.41% to 4.48%) than nondual eligible beneficiaries (5.94%). DSMT utilization was lower among individuals with comorbidities, but varied depending on the condition, with a higher proportion of DSMT utilization among individuals who had depression (4.66%), arthritis (4.56%), and cancer (4.78%), and low utilization among individuals with Alzheimer’s/dementia (1.37%).
Distribution of Hours of Diabetes Self-Management Training in Newly Diagnosed Medicare Diabetes Patients a .
Hours were computed from claims where each diabetes self-management training code represents 30 minutes of services.
Utilization of Diabetes Self-Management Training Services by Demographic, Health Status, and Provider Availability Factors.
Note. COPD = chronic obstructive pulmonary disease.
14 missing (<0.01%). bIndividuals can have multiple comorbidities, so percentages do not sum to 100. c3,393 missing (3.08%) based on availability of geographic data. d4 missing (<0.01%) based on availability of geographic data.
The cutoffs for the provider availability quartiles were: first quartile—0.19 to 4.92 unique providers per 100 beneficiaries from the study population; second quartile—4.92 to 8.73 unique providers per 100 beneficiaries; third quartile—8.73 to 13.03 unique providers per 100 beneficiaries; and fourth quartile—more than 13.03 unique providers per 100 beneficiaries from the study population. DSMT utilization increased as the availability of providers who offered DSMT services increased from the first quartile (2.05%) to the fourth quartile (10.16%). There was mixed DSMT utilization across Census regions, with the highest utilization in the West North Central region (13.49%) and the lowest in the Middle Atlantic (2.78%).
Adjusted regression results reveal patterns in utilization that were not all apparent in the descriptive statistics (Table 3). Males had lower adjusted odds of DSMT utilization compared with females (adjusted odds ratio [AOR] = 0.85, 95% confidence interval [CI] = 0.80-0.89). Adjusted odds of DSMT utilization were lower for older Medicare beneficiaries compared with younger, disabled beneficiaries (less than 65 years of age). Additionally, adjusted odds decreased as age increased so that the oldest age group (85+ years) had the lowest odds of DSMT utilization (AOR = 0.29, 95% CI = 0.25-0.34). Minority groups had lower DSMT utilization compared with White beneficiaries. For example, Black beneficiaries had 19% lower adjusted odds of DSMT utilization compared with White beneficiaries (95% CI = 0.74-0.89), and Asian beneficiaries had the lowest adjusted odds of DSMT utilization (AOR = 0.31, 95% CI = 0.22-0.44). Either type of dual eligibility status was associated with lower adjusted odds of DSMT utilization, compared with individuals with no dual eligibility, with the lowest utilization among individuals with full dual eligible status (AOR = 0.66, 95% CI = 0.60-0.73). Indication of comorbidities was associated with lower DSMT utilization compared with no comorbidities. Individuals with Alzheimer’s/dementia had 64% lower adjusted odds (95% CI = 0.30-0.42) of DSMT utilization in comparison to individuals without Alzheimer/dementia. Those with COPD also had low rates of utilization (AOR = 0.72, 95% CI = 0.66-0.79). Similar to the unadjusted results, as the concentration of providers offering DSMT services increased, so did the adjusted odds of DSMT utilization, with the highest utilization in the top quartile of availability (AOR compared with the lowest quartile = 4.35, 95% CI = 3.95-4.79). In the adjusted results, the West North Central region continued to have the highest odds of DSMT utilization, with 236% increased odds of utilization in comparison with the South Atlantic region (95% CI = 2.14-2.61); in contrast, the Middle Atlantic region still had the lowest odds of DSMT utilization, with 31% reduced odds of utilization compared with the South Atlantic region (95% CI = 0.62-0.77).
Logistic Regression for Factors Associated With Utilization of Diabetes Self-Management Training Services.
Note. OR = odds ratio; CI = confidence interval.
Reference for each comorbid condition is the group without that condition.
p < .05. **p < .01. ***p < .001.
Discussion
In this study of a random sample of FFS Medicare beneficiaries who were newly diagnosed with diabetes, utilization of DSMT services was low, with approximately 5% using these services. The average count of less than 3 DSMT codes (equivalent to less than 1.5 hours) per user was low as well, which indicates that—among those using the services—many are not receiving them as frequently as is reimbursable by CMS (up to 10 hours, or 20 DSMT codes, in the first year). As such, they may not be receiving the full benefit of the DSMT education service. However, delivery of the entire DSMT education service may not be warranted for all individuals; for some, specific modules (and fewer hours of training) may be sufficient for their health needs.
Since individuals are typically eligible for only 1 hour of individual training (unless there are specific circumstances, such as lack of availability of group sessions or disability [Indian Health Service, Division of Diabetes Treatment and Prevention, 2011]), versus 9 hours of group sessions, it is notable that nearly half of the billed codes were for individual training. This high proportion of individual to group session codes, relative to the total amount of group sessions that beneficiaries are eligible to receive, may be a function of the low utilization rates for this service, which could limit the availability of group sessions. Although group sessions may also include non-Medicare beneficiaries (CMS, 2014), broadening the pool of potential participants, other challenges such as scheduling and logistic issues could make attending group sessions difficult. Additionally, individual training sessions may be more feasible to conduct in the primary care setting (Siminerio, Ruppert, & Gabbay, 2013). This finding of higher utilization of individual than group training may also reflect the higher reimbursement rates for these trainings—the 2011 Medicare National Fee schedule rates were approximately three times higher for individual (G0108) versus group (G0109; CMS, 2013b). Research could explore whether allowing more individual sessions increases utilization of DSMT.
Our findings confirm the anecdotal reports and initial research indicating low utilization of DSMT services by Medicare beneficiaries (AADE, 2010; Czarnowski-Hill, 2007; Siminerio et al., 2006). One potential barrier to accessing DSMT services includes whether patients and providers are aware of the benefit. Additionally, access relies on whether the services are available from health care providers, who may perceive the process of getting approval to bill for DSMT, and then getting reimbursed, as being burdensome and complex (Czarnowski-Hill, 2007). A 2003 survey of DSMT programs found that most programs billed to Medicare and were paid the established fees (Pearson, Mensing, & Anderson, 2004), although the survey was small (N = 122) and had a low response rate of 13%. Other barriers to receiving DSMT services may include the distance to a program and the preponderance of delivery of DSMT in hospitals rather than physician offices (Siminerio et al., 2006). Of note, Medicare beneficiaries with diabetes may receive self-management education and training from other sources beyond DSMT. A study using data from the Medicare Current Beneficiary Survey noted that approximately one third of individuals with diabetes who had prescriptions for blood pressure or cholesterol reported that they had received training or participated in a class on diabetes self-management (Stuart et al., 2011). However, it is unclear how respondents interpreted the survey question; they may be attending other self-management programs, or individuals may interpret conversations with their health care providers about managing their diabetes as training. Additional research could clarify these issues.
In the multivariate model exploring the associations between various factors and DSMT utilization, the odds of utilization were lower among men compared with women, older individuals compared with younger, non-Whites compared with Whites, people dually eligible for Medicare and Medicaid compared with nondual eligibles, and patients with comorbidities compared with individuals without those conditions. Odds of individuals’ utilizing DSMT increased as the availability of providers who offered DSMT services increased and varied by Census region. The underlying association between diabetes and vulnerable populations may be driving these findings (Black, 2002; Narayan, Boyle, Thompson, Sorensen, & Williamson, 2003). These vulnerable populations are also less likely to receive timely and adequate health care (Black, 2002) and more likely to have limited access to a usual source of care and lower preventive health care utilization (McCall, Sauaia, Hamman, Reusch, & Barton, 2004; Peek, Cargill, & Huang, 2007). Although this study attempted to control for access to services, the measures available in the claims-based data set were limited, and access to services may be driving low utilization, especially for vulnerable populations. These findings suggest that there are disparities in DSMT utilization, and barriers to access may exist. Notably, most of these factors are fixed and not attributes that can be changed; however, environmental-level factors, such as availability of providers who offer DSMT programs, may be mutable.
Our findings, which indicate disparities by sex, age, race/ethnicity, dual eligible status, health status, geographic location, and provider availability, are in sync with the extensive research literature that documents health disparities in the United States, including among people with diabetes (Black, 2002; CDC, 2011; Miller et al., 2004; Narayan et al., 2003; National Institutes of Health, National Diabetes Data Group, 1995; Peek et al., 2007). DSMT services aim to prevent complications from diabetes, yet specific populations such as Blacks, Hispanics, and Asians, who are more likely to experience diabetes-related complications (Cowie et al., 1989; Karter et al., 2002), have reduced odds of utilizing these services according to our study. Thus, there appears to be a disconnect between the patients who may need DSMT services the most and the services themselves. Research has shown that provider factors—including provider bias, cultural competence, communication, and shared decision making—impact provider referral to preventive health services for diabetes (Peek et al., 2007). Additionally, patient characteristics related to behavioral health (e.g., exercise, diet) and health literacy may impact patient utilization of preventive services such as DSMT (McCall et al., 2004; Miller et al., 2004).
For older individuals or those with conditions such as COPD, heart disease, or Alzheimer’s/dementia, functional or cognitive limitations may be a barrier to service use. Modifications in the provision and delivery of DSMT could potentially address these barriers. Research findings on the association of multimorbidity and use of preventive services among older adults are mixed (Heflin, Oddone, Pieper, Burchett, & Cohen, 2002; Hutter, Schnurr, & Baumeister, 2010; Min et al., 2005; Min et al., 2007). Individuals with diabetes and comorbidities may be managed by specialists who are reluctant to make recommendations outside of their specialty (Barnett et al., 2012), or who may be focused on a more serious comorbidity than newly diagnosed diabetes without complications (Fenton, Korff, Lin, Ciechanowski, & Young, 2006; Redelmeier, Tan, & Booth, 1998; Wolff, Starfield, & Anderson, 2002). Additionally, providers may refer individuals with multiple chronic conditions to more general chronic disease self-management programs (e.g., Lorig et al., 1999). These general, rather than disease specific, self-management programs focus on developing knowledge and skills that all individuals with chronic conditions might benefit from (Clark et al., 1991). Future research could compare general and disease-specific self-management programs to evaluate which are more effective, including among a Medicare population with diabetes and comorbidities.
For full dual eligibles, among whom 15% are institutionalized—compared with 1% of nondual Medicare beneficiaries (Congressional Budget Office, 2013)—the facility may play a significant role in determining diet and managing medications, which may limit or change their specific needs for self-management training. However, it seems likely that factors other than patient’s domicile contribute to our results. McCall et al. (2004) found reduced use of other types of preventive services for diabetes (e.g., annual hemoglobin A1c test) was independently associated with dual eligible status. Better understanding is needed of these factors and their associated barriers to access to improve DSMT utilization.
Limitations
In considering these findings, several limitations should be noted. There may be unmeasured confounders that influence our findings, such as the behavioral characteristics of the study population, including individuals’ engagement with health care and their readiness to change. Additionally, individual-level income and education variables are not available in claims data, and while dual eligibility acts as a partial proxy for income, it may not fully capture variation in DSMT utilization by socioeconomic status. The California Rural Indian Health Board (2012) reported that Medicare data are likely to identify a percentage (14% to 52%) of the North American Native population as other or unknown race/ethnicity. This study’s North American Native population may be an underestimate, though the other and unknown race/ethnicity groups have similar utilization rates. Although the geographic covariate captures some of the variation in availability of DSMT services, the Census region level may not be granular enough to fully capture the range of variability. However, state-level models were considered in a sensitivity analysis, with similar results. Although the provider availability variable captures county-level availability, it does not capture specific provider-level characteristics, and counties vary in size and transportation options, which is likely to affect provider accessibility.
This study is based on claims data and does not capture DSMT service use or provider availability if the provider does not file a Medicare claim. For example, as with ESRD facilities, Rural Health Clinics cannot bill separately for DSMT services, since these services are part of the all-inclusive encounter rate (CMS, 2014). Similarly, use of DSMT services by North American Native individuals whose health care is billed and paid through the Indian Health Service would not be captured in this study. Thus, our findings may represent a conservative estimate of DSMT utilization. Additionally, this analysis may underestimate utilization among older adults by focusing on the Medicare FFS population, who may be sicker and less likely to participate in a self-management program than individuals enrolled in a Medicare Advantage plan. Given the cross-sectional nature of this analysis, causality cannot be attributed. Since claims data are available only for time periods where individuals were enrolled in Medicare, it is possible that some of the included individuals had preexisting, rather than newly diagnosed, diabetes. However, individuals who are newly eligible for Medicare are also eligible for DSMT services, so this distinction should not influence the results—although it could affect the demand for these services. A sensitivity analysis extended the look-back period by requiring continuous enrollment in Medicare Parts A and B in 2008, and the results were consistent with those presented here.
Conclusion
In summary, utilization of DSMT services among Medicare beneficiaries with diabetes is low, but more specifically, there appear to be marked disparities in access to DSMT by a number of factors, including male sex, older age, non-White race/ethnicity, dual eligibility, the presence of comorbidities, geographic residence, and provider availability. In light of the increasing prevalence of diabetes, future research should seek to elucidate the barriers to accessing DSMT services through surveys or focus groups, describe DSMT provider characteristics and feedback on the DSMT benefit, and explore the impact of DSMT services on health outcomes and costs among Medicare diabetes patients. DSMT is one model of care that can contribute to the secondary prevention of diabetes, and policies that support education and outreach to increase awareness about the DSMT benefit among health care providers may offer opportunities for improving availability of and access to these services and, ultimately, better health outcomes.
With the implementation of the Affordable Care Act, preventive services are increasingly covered by insurers, including Medicare. In 2013 alone, depression screenings, behavioral therapy for cardiovascular disease, and screenings and counseling for alcohol misuse, obesity, and sexually transmitted infections were added to the list of Medicare Part B-covered preventive services (CMS, 2013c). However, the low utilization of DSMT, a preventive service benefit that has existed for almost 15 years, highlights the challenges that may be encountered to achieve widespread dissemination and uptake of the new services.
Footnotes
Acknowledgements
We are most appreciative of the following individuals for their support of this study: Kristin Shifflett, Pamela West, and Kathy Bryant. The findings and conclusions contained in this article are those of the authors and do not necessarily reflect the official position of the Centers for Medicare & Medicaid Services.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.