OBJECTIVE—We evaluated the association of different types of educational visits for diabetic patients of the eight Philadelphia Health Care Centers (PHCCs) (public safety-net primary care clinics), with hospital admission rates and charges reported to the Pennsylvania Health Care Cost Containment Council.

RESEARCH DESIGN AND METHODS—The study population included 18,404 patients who had a PHCC visit with a diabetes diagnosis recorded between 1 March 1993 and 31 December 2001 and had at least 1 month follow-up time.

RESULTS—A total of 31,657 hospitalizations were recorded for 7,839 (42.6%) patients in the cohort. After adjustment for demographic variables, baseline comorbid conditions, hospitalizations before the diabetes diagnosis, and number of other primary care visits, having had any type of educational visit was associated with 9.18 (95% CI 5.02–13.33) fewer hospitalizations per 100 person-years and $11,571 ($6,377 to $16,765) less in hospital charges per person. Each nutritionist visit was associated with 4.70 (2.23–7.16) fewer hospitalizations per 100 person-years and a$6,503 ($3,421 to$9,586) reduction in total hospital charges.

CONCLUSIONS—Any type of educational visit was associated with lower hospitalization rates and charges. Nutritionist visits were more strongly associated with reduced hospitalizations than diabetes classes. Each nutritionist visit was associated with a substantial reduction in hospital charges, suggesting that providing these services in the primary care setting may be highly cost-effective for the health care system.

Self-management is the cornerstone of modern diabetes care, and providing patients with the information, skills, and support they need to manage the disease is a critical issue for health care providers and systems (1). Self-management education can be provided in a variety of settings to patients with diabetes using any of a variety of models and methods (2). Education may be provided in a one-on-one format or in group settings. Patients may receive instruction incidentally to other health care encounters or in a single educational visit or a scheduled series of classes. Despite the variety of educational models present in the U.S. health care system, most patients with diabetes receive no formal diabetes education (3). This problem exists throughout the U.S. health care system but is exacerbated for providers who serve low-income and other disadvantaged populations (4).

Effective diabetes education in conjunction with effective medical management has been shown to improve glycemic control, which in turn reduces hospital admissions and a range of other adverse diabetes outcomes (1,5,6). Few studies (79), however, have directly examined the association between diabetes education and hospitalizations. Here, we report on analyses of the association between diabetes education visits and hospital admissions for a large, urban, safety-net primary care system.

The Urban Diabetes Study linked primary care encounter form data for patients with diabetes from the Philadelphia Health Care Centers (PHCCs) with hospital discharge data from the Pennsylvania Health Care Cost Containment Council. Death records for Philadelphia residents were examined to identify patient deaths during the study period. The PHCCs are a network of eight neighborhood health care centers operated by the Philadelphia Department of Public Health. Patients receive primary care services, including pharmacy, without reference to insurance status or ability to pay. The Urban Diabetes Study included every PHCC patient for whom a diagnosis of diabetes was recorded at any time between 1 March 1993 and 31 December 2001. Encounter form data include up to four diagnostic codes for each visit. These patients were linked with data on hospital discharges from the Pennsylvania Health Care Cost Containment Council for the same time period, using an algorithm including Social Security number, date of birth, and sex (10). The Urban Diabetes Study was approved by the institutional review board of the Philadelphia Department of Public Health.

The PHCC encounter form data include codes for nutritionist visits, diabetes classes, and health education visits. We examined these types of visits individually, as well as receipt of any of the three types of educational visit. The content of the visits and the credentials of the personnel providing education varied between individual clinics and over time during the study period. Some patients had visits of several types because they had individual assessment counseling sessions before and after scheduled diabetes classes, which were coded as health education visits. Others were scheduled for individual nutritionist or health education visits because diabetes classes were not currently available. Nutritionist visits were one-on-one and might have included nutrition counseling, instruction in managing insulin and other medications, or instruction in the use of blood glucose meters. Some nutritionists were registered dietitians (and also certified diabetes educators), while others had relevant training and experience but not professional registration or certification. Health education visits were also one-on- one and may have addressed diabetes management, smoking cessation, or other topics. Health educators had either a degree in that field or relevant training and experience. Additional information on education content is available in the online-only appendix (available at http://dx.doi.org/10.2337/dc07-1871).

Diabetes classes, covering all of these topics, were group classes led by a nurse, nutritionist, or health educator. Curricula were developed and refined during the study period under the direction of a Master's-prepared registered nurse/certified diabetes educator. Programs at six of the PHCCs received recognition through a state program in 2000, and three of the clinics were in the process of obtaining American Diabetes Association education recognition at the end of the study period in 2001.

The Pennsylvania Health Care Cost Containment Council receives data on each hospital admission from every hospital in Pennsylvania, as mandated by state law. All hospitalizations for the patients identified were included, whether or not any diabetes-related condition was coded in the hospital discharge record (10).

Demographic variables were taken from the PHCC encounter form records, with age calculated as the time from date of birth to the first diabetes diagnosis in PHCC. Comorbid conditions, including heart disease, kidney disease, lower-extremity ulcers, stroke, eye disease, and hypertension, were assessed based on diagnoses in either the primary care or hospital discharge records (11). The total number of PHCC visits other than educational visits was also calculated for each patient. Follow-up time for each patient was defined as the time from the first date of a diabetes diagnosis in the PHCCs to the date of death, if a death was recorded, or until the end of the study period.

### Analysis

Two different outcome variables were modeled in separate linear regression models: hospitalization rate (number of hospitalizations during follow-up/follow-up time) and hospital charges (total charges for all hospitalizations during follow-up). Each of six educational visit variables was tested: having had at least one educational visit of any type, the number of educational visits of any type, having had at least one diabetes class, having had at least one nutritionist visit, the number of diabetes classes, and the number of nutritionist visits.

The associations of each educational visit variable with the two outcomes were tested in unadjusted analyses and in models controlling for other factors potentially associated with hospitalizations. Adjusted models included both demographic and clinical variables. The demographic variables controlled were sex, race/ethnicity, age at baseline, date of initial diagnosis, and the interaction between sex and age, which had previously been found to be associated with hospitalization rates (10). The baseline patient clinical variables included the number of hospitalizations recorded before the first outpatient diabetes diagnosis and comorbid conditions (heart disease, kidney disease, lower-extremity ulcers, stroke, neuropathy, eye disease, and hypertension) recorded before the first outpatient diabetes diagnosis and, in order to minimize confounding by access to care and patient engagement, the number of PHCC visits other than educational visits.

Models including interactions between educational visit variables and sex or race/ethnicity were examined. Models restricted to patients who had some type of educational visit were also analyzed. All analyses were carried out using SAS 9.1 (SAS Institute, Cary, NC).

### Study population

During the study period, 19,437 patients were diagnosed with diabetes in the PHCCs; 18,404 of these patients (94.7%) had adequate information to allow linkage to hospital discharge records and follow-up time of at least 1 month. Mean follow-up time was 4.7 years. The study population was 56.7% female (Table 1). Most patients (73.0%) were African American, and 12.2% were non-Hispanic whites, 9.8% were Hispanic, and 3.0% were Asian.

Of these patients, 1,054 (5.7%) had at least one diabetes class visit, 1,683 (9.1%) had at least one nutritionist visit, and 332 (1.8%) had a health education visit. A total of 2,465 (13.4%) patients had at least one of any of these educational visits. Among the patients who had at least one diabetes class, the mean number of classes was 5.7; among those who had at least one nutritionist visit, the mean number of such visits was 1.6; and among those who had other health education visits, the mean number of such visits was 1.5.

Differences in baseline comorbidities between patients who had educational visits and those who did not were small; hypertension was more prevalent among patients who did receive educational visits, while kidney disease and lower-extremity ulcers were more prevalent among those who did not. The two groups did not differ significantly in the baseline prevalence of heart disease, stroke, or eye disease. The mean number of PHCC visits other than educational visits was 25.2. This variable was strongly correlated with the number of educational visits (r = 0.20, P < 0.001).

A total of 31,653 hospitalizations were recorded for the study population, for a rate of 36.4 hospitalizations per 100 person-years. Aggregate hospital charges were $848,103,288, an average of$46,083 per person or $9,760 per person-year averaged across all patients including the majority who were not hospitalized at any time during follow-up. The hospitalization rate for patients who had no educational visits during follow-up was 38.1 per 100 person-years (Table 2). For patients who had at least one educational visit, the hospitalization rate was 34% lower at 25.0 per 100 person-years. Patients who had at least one visit with a nutritionist had a hospitalization rate of 21.1 per 100 person-years, 45% lower than the rate for patients who had no education visits. Rates for patients who attended a diabetes class or saw a health educator were intermediate between the rates for those who saw a nutritionist and those who had no educational visits. ### Associations with educational visits of any type In both unadjusted and adjusted models, having had any educational visit was strongly associated with lower rates of hospitalization, as were the number of educational visits (Table 3). In the adjusted models, having had at least one educational visit was associated with 9 (95% CI 5–13) fewer hospitalizations per 100 person-years. Each educational visit of any type was associated with 1 (0–2) fewer hospitalization per 100 person-years. Hospital charges were substantially reduced among patients who had an educational visit. At least one educational visit was associated with hospital charges that were$11,571 lower (95% CI −$6,377 to −$16,765). Each educational visit was associated with a reduction of $1,793 (−$817 to −$2,770) in hospital charges. ### Associations with specific types of educational visits Although nutritionist visits, diabetes classes, and health educator visits were each associated with reduced hospitalizations in unadjusted models, nutritionist visits were consistently associated with the greatest reductions in hospitalizations and hospital charges. Having had at least one nutritionist visit was associated with 11 (95% CI 6–16) fewer hospitalizations per 100 person-years and$13,872 (95% CI $7,799 to$19,945) less in hospital charges in the adjusted model. Each nutritionist visit was associated with 5 (95% CI 2–7) fewer hospitalizations per 100 person-years and $6,503 (95% CI$3,421 to $9,586) less in hospital charges. Each diabetes class was associated with a reduction of$1,396 ($254-$2,538) in hospital charges. Restricting the analyses to patients who had at least one educational visit did not alter the conclusions. There was no evidence of interaction between the study variables and sex or race/ethnicity.

In this large, unselected cohort of low-income primary care patients with diabetes, attending at least one visit devoted specifically to patient education was associated with significantly lower rates of hospitalization and hospital charges. Having had any educational visit, as opposed to none, was associated with eight fewer hospitalizations per 100 person-years; each educational visit was associated with $1,684 less in total hospital charges per patient over the course of follow-up. Other investigators have found that diabetes education was associated with decreased hospital admissions in trials in managed care organizations (8) and mixed-income primary care systems (7). A small program for Medicaid recipients in Arkansas found that over a 1-year time period participants had an estimated 0.37 fewer hospital admissions after adjustment for demographic variables, prior drug treatment, and prior period Medicaid costs. Total Medicaid expenditures, including the costs of the diabetes education program, were not significantly different from those of nonparticipants (9). Our finding that individual nutritionist visits were more strongly associated with improved outcomes than were group diabetes classes differs from most other studies (12,13). Most studies that have compared group with individual diabetes education strategies have found equivalent outcomes, with some finding improved short-term intermediate outcomes for group participants (13). Greisinger et al. (7) found that diabetes classes were associated with improved outcomes, but individual sessions with a certified diabetes educator were not. The difference in our findings may reflect differences in the specific services offered or in the time period. Our study covered hospitalizations between 1993 and 2001, while Greisinger et al. (7) assessed hospitalizations in 2002. Alternatively, the inner-city, low-income patients served by the PHCCs may respond differently to these services than higher-income patient populations, or nutritionists in this public health system may have been more focused on the educational needs of patients with diabetes than most clinicians in other health care systems. A review of interventions designed to improve diabetes care for socially disadvantaged populations identified one-on-one interventions with individualized assessment and reassessment as one of the features associated with the most positive effects (14), and patients with low literacy were found to derive greater benefits in an assessment of a disease management program emphasizing one-on-one counseling (15). Our findings are consistent with the evidence that group-based diabetes education results in better outcomes than usual care or no formal diabetes education (16). Most other studies of diabetes education have included only intermediate clinical outcomes, such as glycemic control, cholesterol levels, and BMI, and have included relatively short follow-up periods. Some studies suggest that the intermediate clinical effects of diabetes education on these intermediate outcomes declined substantially within a few months after the end of the intervention (17). This study offers additional evidence that participation in diabetes education is associated with significant differences in important clinical outcomes over a mean 4.7-year follow-up period. As with any observational study, our analyses are vulnerable to confounding. Diabetic patients who attend educational visits may be systematically different from those who do not, in ways that are associated with improved outcomes. However, we did find that in this patient population, patients with comorbid conditions were more likely than other patients to have had an educational visit, suggesting that the factors influencing selection into the intervention group were not one sided. In addition, we controlled for a wide range of baseline clinical factors, including the number of hospitalizations recorded before the diabetes diagnosis as well as major comorbid diagnoses. We also controlled for the number of primary care visits other than educational visits each patient had during the course of follow-up as an indicator of patient engagement and access to care. This variable was also strongly associated with better patient outcomes and was responsible for much of the attenuation of the education effects seen in the adjusted models. Although residual confounding associated with patient characteristics could still have affected the results, inclusion of these patient variables substantially reduces the magnitude of any remaining bias (18). Patients who received educational visits were more likely to be between the ages of 45 and 64 years at baseline than were those who did not. However, restricting the analyses to this age-group did not alter the conclusions. Patients were offered one type of education or the other based on what was available at that time at the health care center they attended. Confounding by indication is unlikely to occur in this situation (19), although we cannot exclude the possibility that the availability of diabetes education classes was associated with other factors that affected outcomes. The finding that nutritionist visits had a stronger association with hospital admissions and charges than diabetes classes is strengthened by its robustness irrespective of the model used. There is no reason to believe that patients were selected into one or the other type of visit based on their preexisting characteristics. Nonetheless, we cannot rule out the possibility that our findings are vulnerable to selection bias. Our analyses are based on administrative data, which are prone to error, and did not include some important clinical variables, such as BMI, blood pressure, or glucose values. The data are, however, not subject to biases associated with self-reporting of health care use. Although the long time period covered allows us to examine outcomes over a mean 4.7 years of follow-up, the availability and content of diabetes education was not uniform throughout the period and does not reflect more recent practices. These data reflect outcomes not in a controlled clinical trial environment but in the real-world practice of a large, urban public health care system. Outcomes were hospital admissions—a clinical event representing substantial patient morbidity and burden—and hospital charges. The finding that for this large population, attendance at even one educational visit was associated with substantially reduced hospitalization rates and hospital charges provides evidence that providing formal diabetes education is an important and cost-effective imperative in primary care settings. We have no data on how many patients were offered diabetes education but did not take advantage of it. This occurs frequently (20,21), although there are few data available on patient recruitment to diabetes education outside of the research setting. This problem may be especially severe for low-income and less-educated patients (22). A program offered to 1,968 eligible Medicaid patients in Arkansas succeeded in enrolling only 212 patients (9). Retention of patients in diabetes education programs that involve multiple sessions is also problematic (9,22,23). The finding that any type of educational visit was beneficial suggests that health care providers will do well to provide diabetes education in whichever format providers can support and patients will attend. The ideal form of diabetes education may well vary among systems and among patient populations. These findings underscore the importance of focusing on both feasibility for health care systems and acceptability to patients (or “adoption” and “reach” in the RE-AIM evaluation framework) in designing and evaluating educational interventions (20). Our most important finding is that either diabetes classes or nutritionist visits in any quantity were strongly associated with reductions in patient hospitalizations and hospital charges. The average annual hospital charges for patients who received any educational visit were$6,244, 39% less than the \$10,258 per year average for patients who had no such visits. If the proportion of diabetic patients receiving education could be raised to the Healthy People 2010 goal of 60% (3), the data suggest that 10s of millions of dollars in hospital charges and 10s of thousands of hospitalizations could be avoided each year, just among patients of one large primary care system. Many safety-net providers find that obtaining payment for diabetes education from insurers is so burdensome that the payments do not effectively compensate for the effort required (24,25). Both public and private insurers would be well advised to invest in efforts to improve the reach of diabetes education in order to reduce both long-term costs and human suffering for the large and growing population of patients with diabetes.

This research was made possible by the National Institute of Diabetes and Digestive and Kidney Diseases (grant no. R21DK064201). Parts of this article were presented in abstract form at the 67th annual meeting of the American Diabetes Association, Chicago, Illinois, 25 June 2007.

1.
Funnell MM, Brown TL, Childs BP, Haas LB, Hosey GM, Jensen B, Maryniuk M, Peyrot M, Piette JD, Reader D, Siminerio LM, Weinger K, Weiss MA: National standards for diabetes self-management education.
Diabetes Care
30
:
1630
–1637,
2007
2.
Chodosh J, Morton SC, Mojica W, Maglione M, Suttorp MJ, Hilton L, Rhodes S, Sheleel P: Meta-analysis: chronic disease self-management programs for older adults.
Ann Intern Med
143
:
427
–438,
2005
3.
U.S. Department of Health and Human Services:
Healthy People 2010: Understanding and Improving Health
. 2nd ed. Washington, DC, U.S. Government Printing Office,
2000
4.
U.S. Health and Human Services: Healthy People 2010 midcourse review [article online],
2007
. Available at http://www.healthypeople.gov/data/midcourse/pdf/FA05.pdf. Accessed 18 July
2007
5.
Warsi A, Wang PS, LaValley MP, Avorn J, Solomon DH: Self-management education programs in chronic disease: a systematic review and methodological critique of the literature.
Arch Internal Med
164
:
1641
–1649,
2004
6.
Cook CB, Hentz JG, Tsui C, Ziemer DC, Naylor DB, Miller WJ: Poor glycemic control increases risk of hospitalization in urban African Americans with diabetes.
Ethn Dis
16
:
880
–885,
2006
7.
Greisinger AJ, Balkrishnan R, Shenolikar RA, Wehmanen OA, Muhammad S, Champion PK: Diabetes care management participation in a primary care setting and subsequent hospitalization risk.
Dis Manage
7
:
325
–332,
2004
8.
Sadur CN, Moline N, Costa M, Michalik D, Mendlowitz D, Roller S, Watson R, Swain BE, Selby JV, Javorski WC: Diabetes management in a health maintenance organization: efficacy of care management using cluster visits.
Diabetes Care
22
:
2011
–2017,
1999
9.
Balamurugan A, Ohsfeldt R, Hughes T, Phillips M: Diabetes self-management education programs for Medicaid recipients: a continuous quality improvement process.
Diabetes Educ
32
:
893
–900,
2006
10.
Robbins JM, Webb DA: Hospital admission rates for a racially diverse low-income cohort of patients with diabetes: the Urban Diabetes Study.
Am J Public Health
96
:
1260
–1264,
2006
11.
Robbins JM, Webb DA, Sciamanna CN: Cardiovascular comorbidities among public health clinic patients with diabetes: the Urban Diabetics Study.
BMC Public Health
5
:
15
,
2005
12.
Mullen PD, Green LW, Persinger GS: Clinical trials of patient education for chronic conditions: a comparative meta-analysis of intervention types.
Prev Med
14
:
753
–781,
1985
13.
Tang TS, Funnell MM, Anderson RM: Group education strategies for diabetes self-management.
Diabetes Spectr
19
:
99
–105,
2006
14.
Glazier RH, Bajcar J, Kennie NR, Willson K: A systematic review of interventions to improve diabetes care in socially disadvantaged populations.
Diabetes Care
29
:
1675
–1688,
2006
15.
Rothman RL, DeWalt DA, Malone R, Bryant B, Shintani A, Crigler B, Weinberger M, Pignone M: Influence of patient literacy on the effectiveness of a primary-care based diabetes disease management program.
JAMA
292
:
1711
–1716,
2004
16.
Deakin T, McShane CE, Cade JE, Williams RDRR: Group based training for self-management strategies in people with type 2 diabetes mellitus.
Cochrane Database Syst Rev
CD003417,
2007
17.
Norris SL, Lau J, Smith SJ, Schmid CH, Engelgau MM: Self-management education for adults with type 2 diabetes: a meta-analysis of the effect on glycemic control.
Diabetes Care
25
:
1159
–1171,
2002
18.
Fewell Z, Davey Smith G, Sterne JAC: The impact of residual and unmeasured confounding in epidemiologic studies: simulation study.
Am J Epidemiol
166
:
646
–655,
2007
19.
Shrier I, Boivin J-F, Steele RJ, Platt RW, Fulan A, Kakuma R, Brophy J, Rossignol M: Should meta-analyses of interventions include observational studies in addition to randomized controlled trials? A critical examination of underlying principles.
Am J Epidemiol
166
:
1203
–1209,
2007
20.
Glasgow RE, Nelson CC, Strucker LA, King DK: Using RE-AIM metrics to evaluate diabetes management support interventions.
Am J Prev Med
30
:
67
–73,
2006
21.
Toobert DJ, Strycker LA, Glasgow RE, Bagdade JD: If you build it, will they come?. Reach and adoption associated with a comprehensive lifestyle management program for women with type 2 diabetes.
Patient Educ Couns
48
:
99
–105,
2002
22.
Thoolen B, de Ridder D, Bensing J, Gorter K, Rutten G: Who participates in diabetes self-management interventions?: Issues of recruitment and retention.
Diabetes Educ
33
:
465
–474,
2007
23.
Rickheim PL, Weaver TW, Flader JL, Kendall DM: Assessment of group versus individual diabetes education: a randomized study.
Diabetes Care
25
:
269
–274,
2002
24.
Balamurugan A, Rivera M, Jack L Jr, Allen K, Morris S: Barriers to diabetes self-management education programs in underserved rural Arkansas: implications for program evaluation.
Prev Chronic Dis
3
:
A15
,
2006
25.
Czarnowski-Hill JV: Diabetes self-management education and third-party reimbursement: what are the options?
Curr Diab Rep
7
:
381
–385,
2007

Published ahead of print at http://care.diabetesjournals.org on 9 January 2007. DOI: 10.2337/dc07-1871.