Out-of-Pocket Costs and Diabetes Preventive Services

The Translating Research Into Action for Diabetes (TRIAD) study

  1. Andrew J. Karter, PHD1,
  2. Mark R. Stevens, MSPH, MA2,
  3. William H. Herman, MD, MPH3,
  4. Susan Ettner, PHD4,
  5. David G. Marrero, PHD5,
  6. Monika M. Safford, MD6,
  7. Michael M. Engelgau, MD, MS2,
  8. J. David Curb, MD, MPH7,
  9. Arleen F. Brown, MD, PHD4 and
  10. The TRIAD Study Group*
  1. 1Division of Research, Kaiser Permanente, Oakland, California
  2. 2Centers for Disease Control and Prevention, Division of Diabetes Translation, Atlanta, Georgia
  3. 3University of Michigan, School of Medicine, Ann Arbor, Michigan
  4. 4University of California, Los Angeles, School of Medicine, Los Angeles, California
  5. 5Indiana University School of Medicine, Indianapolis, Indiana
  6. 6University of Medicine and Dentistry of New Jersey, New Jersey Medical School, Newark, New Jersey
  7. 7Pacific Health Research Institute, Honolulu, Hawaii
  1. Address correspondence and reprint requests to Andrew J. Karter, PhD, Division of Research, Kaiser Permanente, 2000 Broadway, Oakland, California 94612. E-mail: andy.j.karter{at}kp.org

Abstract

OBJECTIVE—Despite the increased shifting of health care costs to consumers, little is known about the impact of financial barriers on health care utilization. This study investigated the effect of out-of-pocket expenditures on the utilization of recommended diabetes preventive services.

RESEARCH DESIGN AND METHODS—This was a survey-based observational study (2000–2001) in 10 managed care health plans and 68 provider groups across the U.S. serving ∼180,000 patients with diabetes. From 11,922 diabetic survey respondents, we studied the occurrence of self-reported annual dilated eye exams and diabetes health education and among insulin users, daily self-monitoring of blood glucose (SMBG). Conditional probabilities were estimated for outcomes at each level of self-reported out-of-pocket expenditure by using hierarchical logistic regression models with random intercepts.

RESULTS—Conditional probabilities of utilization (95% CI) varied by expenditure for dilated eye exam [no cost 78% (75–82), copay 79% (75–82), and full price 70% (64–75); P < 0.0001]; diabetes health education [no cost 29% (23–36), copay 29% (23–36), and full price 19% (14–25); P < 0.0001]; and daily SMBG [no cost 75% (68–81), copay 68% (60–75), and full price 59% (49–68); P < 0.0001]. Extensive adjustment for patient factors had no discernible effect on the estimates or their significance, and cost-utilization relationships were similar across income levels and other patient characteristics.

CONCLUSIONS—Benefit packages structured to derive greater fiscal contribution from the health plan membership result in suboptimal use of diabetes preventive services and may thus lead to poorer clinical outcomes, greater future costs, and lower health plan quality ratings.

In light of the trend toward shifting the burden of health care costs to the consumer, an understanding of how out-of-pocket expenditures impact use of diabetes preventive health care is needed. In recent years, managed care health plans are being driven by large employer purchasers toward offering lower cost products with increasing levels of cost sharing. Cost sharing, a common form of “demand management,” has been shown to reduce excessive utilization (1), yet there is relatively little known about its impact on diabetes service utilization. One concern is that the use of copays may unintentionally result in suboptimal use of essential care and medications, particularly among the poor. Thus, cost sharing may become a self-defeating strategy if it poses financial barriers that lead to poorer health outcomes and greater future health care costs.

For the most economically disadvantaged patients, cost sharing has been associated with reduced essential use of processes of care (2), medications (3), and self-care supplies (e.g., test strips for self-monitoring of blood glucose [SMBG]) (4). The magnitude of the effect of a financial barrier (price elasticity) may be modified by a patient’s socioeconomic status, because fixed copays represent a greater burden for the poorest patients. Price elasticity may also vary depending on how patients value and prioritize the importance of health care versus personal cost saving (5). Both patient attributes (e.g., educational attainment and language abilities) and provider attributes (e.g., promotion of self-care activities and quality of provider-patient communication) might further modify sensitivity to cost by imparting a better or worse understanding of the value of the processes of care and self-management.

To explore this issue, we studied whether higher out-of-pocket costs within the managed care setting (attributable to copayments, deductibles, or having to pay full price for services) were associated with lower use of three diabetes care processes that are recognized as standards of treatment: annual dilated eye exams, health education, and daily SMBG among insulin-treated diabetic patients. We hypothesized that the relationships between financial barriers and use of these processes of care may vary by patient characteristics, particularly socioeconomic position.

RESEARCH DESIGN AND METHODS

Study setting

This research is part of a larger multisite study of how managed care systems influence processes and outcomes of diabetes care called Translating Research Into Action for Diabetes (TRIAD) (6). The fundamental hypothesis of TRIAD is that structural and organizational characteristics of health systems and health care provider groups affect the processes of care, which in turn influence health and economic outcomes. Six Translational Research Centers collaborate with 10 health plans and 68 provider groups, which serve ∼180,000 patients with diabetes. The six centers are the Pacific Health Research Institute (Honolulu, HI), Indiana University Translational Research Center (Indianapolis, IN), Kaiser Foundation Research Institute (Oakland, CA), the University of California, Los Angeles School of Medicine (Los Angeles, CA), the University of Medicine and Dentistry of New Jersey (New Brunswick and Newark, NJ), and the University of Michigan Health System (Ann Arbor, MI). Managed care health plans are defined as entities that deliver, administer, or assume risk for health services in order to influence the quality, access, cost, and outcomes of health care for a defined population (7). Health plans participating in TRIAD include staff model health maintenance organizations (HMOs), network/independent practice association model HMOs, point-of-service plans, and preferred provider organizations with for-profit, not-for-profit, Medicare, and Medicaid products. The provider groups include groups of physicians with contractual arrangements with one or more health plans that provide managed care services. The groups are often engaged directly in diabetes care management and may determine compensation arrangements and financial incentives for physicians and specialty referral policies.

Study design and data collection

TRIAD study methods have been detailed previously (6). Briefly, TRIAD is a prospective cohort study that includes data collection from health plans, provider groups, and diabetic patients. Patients were eligible if they were 18 years of age or older, were not pregnant, had diabetes for at least 1 year, were continuously enrolled in the health plan for at least 1.5 years, and used services of the designated health care plans during that time. In addition, patients had to speak either English or Spanish. Patients from provider groups with <50 patients with diabetes were excluded.

The data reported here are from in-depth information from patients questioned by computer-assisted telephone interview or mailed survey during 2000–2001. Ninety-one percent (11,922) of the people who were contacted and found eligible responded to the survey. If patients unable to be contacted had the same rate of eligibility as those contacted and were counted in the denominator, the survey response rate (1) would have been 69%.

Statistical analysis

The outcomes of interest were dichotomized variables based on at least one self-reported dilated eye exam and a one-on-one or group diabetes education session during the preceding year. A dichotomized measure of daily SMBG, among insulin-treated patients, was based on self-reported usual practice of SMBG at least once daily for at least 6 days a week. Out-of-pocket expenditures for eye exams, health education, and SMBG test strips were categorized as free (“none”), partial payment (“some”), and full cost payment (“all”). Consistent with published practice recommendations by the American Diabetes Association and the American Association of Diabetes Educators (http://www.aadenet.org/index.html), we consider these outcomes normative processes of care and self-care. The American Diabetes Association Clinical Practice Recommendations (2) recommends at least one annual dilated eye exam for most diabetic patients and daily SMBG for insulin-treated patients as standards (8). Although no clearly recommended frequency of diabetes health education exists because of the prevalence of inadequate health literacy and the complexity associated with self-management of diabetes (9), periodic attendance is certainly warranted.

Hierarchical logistic regression models (SAS GLIMMIX Macro with penalized quasi-likelihood estimation method), with random intercepts for health plan, were used to account for the clustered study design (health plan, provider group, and patient levels) and dependency of patient characteristics within health plans and provider groups. When outcomes are common (e.g., >30%), logistic regression odds ratios (ORs) become poor estimates of relative risk (10). Therefore conditional probabilities were estimated for each level of out-of-pocket charge rather than relying on ORs. Whether the relationship between out-of-pocket cost and service utilization was modified by other factors was evaluated by testing the significance of cross-product terms (e.g., interaction between copay and annual household income). We focused largely on unadjusted model estimates rather than the adjusted estimates because the process measures studied are normative standards of preventive care that should be applied to all patients with diabetes, regardless of demographics or disease severity (8). For comparison, adjusted models that included sex, age, race or ethnicity, income, education, preferred language, health plan location, diabetes treatment, and the presence of comorbidities were also specified.

RESULTS

Subject characteristics

The demographic characteristics of the patients studied were reflective of the general population of individuals with diabetes in the U.S. in that they were skewed toward patients >45 years of age, were racially diverse, and had a somewhat larger representation of lower income and poorly educated individuals (Table 1). The majority of patients were treated with oral agents, insulin monotherapy, or insulin/oral agent combination therapy; whereas less than one-tenth controlled their diabetes with diet and exercise alone. Very few of the patients were unable to communicate in English. A total of 75% of subjects had a dilated eye exam within the previous year, and 70% of insulin-treated patients practiced SMBG at least once daily. In contrast, only one-fifth attended health education during the previous year. Most benefit packages covered all or part of these three outcomes.

Influence of out-of-pocket expenditures on utilization

For dilated eye exams, the conditional probability of utilization was significantly lower for individuals paying full cost (P < 0.0001): 70% (95% CI 64–75) of individuals paying full cost, while 78% (75–82) of those receiving free visits and 79% (75–82) of those paying some share of the cost received an eye exam (Fig. 1). Similarly, significantly fewer patients paying full cost reported attending diabetes health education in the previous year (P < 0.001), i.e., only 19% (14–25) of those paying full cost, but 29% (23–36) of both patients with free services and patients paying some amount received such education. Among insulin-treated patients, the conditional probability of daily SMBG decreased significantly as the out-of-pocket cost for the test strips increased (P < 0.001): 75% (68–81) of patients receiving free test strips, 68% (60–75) paying some share of the cost, and 59% (68–81) who paid the full cost of test strips reported daily SMBG. Interestingly, these relationships did not differ across income groups (no joint significance in interaction between cost sharing and income). Moreover, the relationships were not altered by adjustment for demographics (sex, age, and race or ethnicity), socioeconomic indicators (income and education), preferred language (English or Spanish), severity measures (diabetes treatment and comorbidity score), or a provider quality summary score (primary care provider discussed hypoglycemia, glycemic targets, or medication adjustment or reviewed SMBG records).

CONCLUSIONS

This is the first large multicenter study of the relation between out-of-pocket charges and use of diabetes preventive services in a managed care setting. In relation to published clinical care guidelines, we observed underutilization of recommended preventive care services, even among individuals who received free services. Copays and full-cost services were associated with even lower use of these processes of care. We speculate that this is likely because cost functioned as a disincentive. The differences in use were clinically relevant and statistically significant; relative to services with no copay, full-cost services were associated with a 12% lower use of dilated eye exams, 53% lower attendance at health education classes, and 27% lower performance of daily SMBG. Patients may perceive health education and SMBG as “optional” and are thus more sensitive to the cost than for dilated eye exams. It is also possible that physician beliefs about the efficacy of the processes of care may have influenced utilization by patients. There are clearly different levels of evidence supporting each process, particularly health education. Given the increasing emphasis on the three care processes studied, we felt that the likelihood of health care providers recommending that they not be utilized is small. In addition, extensive adjustment for patient clinical profile and provider characteristics by level of out-of-pocket expenditure had no discernible confounding effect on our point estimates or their significance. Additionally, contrary to our a priori hypothesis, the cost-utilization relations were similar across income levels and a wide array of other patient characteristics. Thus, our observed price elasticity effects should be widely generalizable.

Previous studies have also detected reduced health care utilization with increasing out-of-pocket costs. The Rand Health Insurance Experiment (1), a randomized clinical trial, demonstrated that providing free services increased utilization and also benefited health outcomes. Our study is the first to examine these relations, albeit observationally, specifically in preventive diabetes care. Previous studies have been inconsistent in terms of whether the price elasticity effect on service utilization varies across income levels. Consistent with our findings, a study (11) in the general population of the Puget Sound area of Washington found no differences in the relation between copay amount and utilization between health maintenance enrollees with higher and lower income. Karter et al. previously (4) reported a copay effect (decreasing SMBG practice with increasing out-of-pocket charge for test strips) across all levels of income but observed an accentuated effect among the poor. Additionally, previous studies have shown that patients are more price sensitive when they perceive a health service as optional and are less likely to vary utilization as a function of cost when they perceive a service as essential. For example, among patients experiencing myocardial infarction, Magid et al. (12) detected no association between copays and delays in seeking emergency care. This perception of essential versus nonessential services may account for the different utilization rates observed between the three processes of care, although we have no empirical data on patient perceptions and thus cannot rule out alternative hypotheses.

There are other limitations to this research that must be considered. Our study analyses are based on self-reported utilization that may introduce bias in the results. Studies investigating whether patients are reliable reporters of medical care processes report conflicting findings. Good correspondence between self-report data and other forms of more objective data collection have been reported. A previous study from one of the TRIAD sites (Kaiser) showed the correspondence between self-reported SMBG frequency, based on a question similar to that used by the current study, and pharmacy records of test strip utilization was quite good (13). Particularly among individuals with preexisting chronic disease, recall has been shown to be fairly accurate and without preferential recall, whereas more objective data collection such as medical charts and administrative medical records are frequently less complete for processes such as health education or SMBG frequency (1317). In this study, however, the contribution of reporting bias is assumed to be small, since it is unlikely that subjects with high copays remember or forget visits at a different rate than those with low copays. In fact, given that recall is likely to lead to a nondifferential misclassification, our findings are likely conservative (biased toward the null) (18). An additional limitation is that cross-sectional associations preclude causal inferences. We note a relation between out-of-pocket charges and utilization of some preventive care practices, although we cannot conclude that utilization would change if charges were modified. The propensity for high utilizers to purchase more comprehensive health coverage (with lower copays) could inflate the inverse association between copay and utilization. However, there are typically no within-plan coverage choices or only very limited options (19), so in order to influence their cost-sharing requirements, patients would need to switch health plans. Thus bias due to selection of benefit packages based on anticipated use should be minimal. Additionally, control for disease severity would further attenuate any bias. Previous research at one of the TRIAD sites (personal communication, W.H.H.) suggested that patients might reduce use of covered preventive services if they erroneously believed that the service was not covered. In our study, we used self-reported copay amount, which should reflect patients’ perception of the cost. But bias may be introduced if the provider takes into account copay amount when referring for services (e.g., not recommending dilated eye exams for patients who have the highest copay). This possibility is probably rare, and we assume it to not be a source of substantive bias.

The need to improve quality of care for diabetes has been a major focus in the U.S. in recent years. National studies, including the Third National Health and Nutritional Examination Survey (20,21), U.S. National Health Interview Survey (22,23), and Behavioral Risk Factor Surveillance System (24) have reported suboptimal use of most preventive diabetes services despite demonstrated health benefits. If service utilization is price elastic, out-of-pocket charges are counterproductive if they hinder quality improvement efforts. We observed underutilization of dilated eye exams, health education, and daily SMBG among diabetic health plan members receiving these services at no cost. Members who paid full price were even less likely to use these services. Benefit packages structured to derive greater fiscal contribution from the health plan membership that result in suboptimal use of preventive diabetes care services are self-defeating strategies because of the likelihood of poorer clinical outcomes, greater future costs, and lower health plan quality ratings. We are now facing an unfortunate convergence of a diabetes epidemic, rising health care costs, and rapidly evolving health insurance benefit designs with increasing cost shifting to the patients (25). With the increasing demand for quality of diabetes care, it is ever more critical that managed care health plans and large employer purchasers carefully evaluate the unintended consequences associated with imposing cost sharing for health services that patients may consider nonessential or that have demonstrated price elasticity.

APPENDIX

Members of the Translating Research Into Action for Diabetes (TRIAD) Study Group

Pacific Health Research Institute (PHRI).

Principal Investigator: J. David Curb, MD, MPH. Co-Investigators: Beth Waitzfelder, MA; Chien-Wen Tseng, MD, PhD; Richard Chung, MD (Hawaii Medical Service Association [HMSA]); Peggy Latare, MD (Kaiser Permanente Hawaii [KPH]); Lynette Honbo, MD (Hawaii State Department of Human Services [HDHS]); Adams Dudley, MD (University of California, San Francisco [UCSF]); Beatrice Rodriguez, MD, PhD; Robert Abbott, PhD; Consultant: Joseph Humphry, MD (HMSA); Analysts: Rebecca Glavan; Andrew White, PhD (HMSA); Ken Forbes (KPH); James Cooper, MA (HDHS); Administrative Assistants: Ruth Baldino; Esther Nakano.

Indiana University Translational Research Center.

Principal Investigator: David G. Marrero, PhD. Project Coordinator: Susanna R. Williams, MSPH. Co-Investigators: William M. Tierney, MD; M. Sue Kirkman, MD.

Division of Research, Kaiser Permanente.

Principal Investigator and Study Chairman: Joe V. Selby, MD, MPH.; Co-Principal Investigator: Andrew J. Karter, PhD, MS. Co-Investigators: Assiamira Ferrara, MD, PhD. Project Analyst: Bix E. Swain, MS; Tiffany Peng, MA. Project Coordinator: Marcia M. Ewing, LVN. Administrative Assistant: Carol Rabello.

University of California, Los Angeles School of Medicine.

Principal Investigator: Carol M. Mangione, MD, MSPH; Co-Principal Investigator: Arleen F. Brown, MD. Project Director: Rebecca Brusuelas. Co-Investigators: Martin F. Shapiro, MD, PhD; Susan Ettner, PhD; Sam Ho, MD (PacifiCare Health Systems). Data Analysts: Peter R. Gutierrez; Neil Steers, PhD. Senior Administrator: Carole Nagy.

University of Medicine and Dentistry of New Jersey (UMDNJ).

Principal Investigator: Monika M. Safford, MD. Co-Investigators: Dorothy A. Caputo, MA, RNC, CDE; Michael Brimacombe, PhD; Louis F. Amorosa, MD; David Hom, MS; David Kountz, MD; Leonard Pogach, MD, MBA; Louise Russell, PhD; Quanwu Zhang, PhD; David Bendich, MD (Horizon Blue Cross Blue Shield); Joseph Singer, MD; John Chard, MD; Ron Snyder, MD (Healthnet). TRIAD-wide Administrative Assistant: Gabrielle Davis, BA. Program Specialist: Patricia Prata, MPH, CHES.

The University of Michigan Health System.

Principal Investigator: William H. Herman, MD, MPH. Co-Principal Investigator: Catherine Kim, MD; Project Director: Jennifer Goewey, MHA. Programmer/Analyst: Diane Kennedy. Research Associates: Ray Burke, MA; Bahman Tabaei, MPH; Honghong Zhou, MS; Laura Mattei, MPH. Administrative Assistant: Barbara Pearlman.

TRIAD-Veterans Association.

Principal Investigator: Rodney Hayward, MD. Co-Principal Investigator: Eve Kerr, MD, MPH. Co-Investigators: Sarah Krein, PhD; John Piette, PhD. Project Managers: Fatima Makki, MPH, MSW; Jill Baker, MSW. Data Manager: Jennifer Davis, MPH.

TRIAD-wide Administrative Data Coordinator:

Barbara R.K. Smith, MHSA Social & Behavioral Research Institute, CA State University San Marcos: Richard Serpe, PhD; Allen J. Risley, MS.

National Institute of Diabetes and Digestive and Kidney Diseases: Sanford A. Garfield, PhD.

Division of Diabetes Translation; Centers for Disease Control and Prevention.

Principal Scientist: K.M. Venkat Narayan, MD, MSc, MBA. Co-Scientists: Theodore Thompson, MS; Edward W. Gregg, PhD; Robert Gerzoff, MS; Michael M. Engelgau, MD, MS; Gloria Beckles, MB, BS, MSc; Mark Stevens, MSPH, MA; Henry Kahn, MD, FACP; David F. Williamson, PhD; Patrick Boyle, PhD. Project Administrator: Bernice Moore, MBA Program Specialist: Shay Clayton.

Figure 1—

Conditional probabilities of utilization of diabetes preventive health care generated from the hierarchical logistic regression model (accounting for clustering within health plan). Bars indicate 95% CI.

Table 1—

Subject characteristics, process measures, and out-of-pocket charges for the TRIAD study population (n = 11,922)

Acknowledgments

This study was jointly funded by a grant from the Centers for Disease Control and Prevention (Division of Diabetes Translation) and the National Institute of Diabetes and Digestive and Kidney Diseases (grant U-48-CCU916373).

We acknowledge the TRIAD participants, other investigators, and staff that made this study possible.

Footnotes

  • *

    * See appendix for a complete list of members of the TRIAD Study Group.

  • A table elsewhere in this issue shows conventional and Système International (SI) units and conversion factors for many substances.

    • Accepted April 21, 2003.
    • Received February 13, 2003.

References

| Table of Contents