OBJECTIVE—Although patient diabetes self-management is a key determinant of health outcomes, there is little evidence on whether patients’ own assessments of their self-management correlates with glycemic control and key aspects of high-quality diabetes care. We explored these associations in a nationwide sample of Veterans’ Affairs (VA) patients with diabetes.

RESEARCH DESIGN AND METHODS—We abstracted information on achieved level of glycemic control (HbA1c) and diabetes processes of care (receipt of HbA1c test, eye examination, and nephropathy screen) from medical records of 1,032 diabetic patients who received care from 21 VA facilities and had answered the Diabetes Quality Improvement Program survey in 2000. The survey included sociodemographic measures and a five-item scale assessing the patients’ diabetes self-management (medication use, blood glucose monitoring, diet, exercise, and foot care [α = 0.68]). Using multivariable regression, we examined the associations of patients’ reported self-management with HbA1c level and receipt of each diabetes process of care. We adjusted for diabetes severity and comorbidities, insulin use, age, ethnicity, income, education, use of VA services, and clustering at the facility level.

RESULTS—Higher patient evaluations of their diabetes self-management were significantly associated with lower HbA1c levels (P < 0.01) and receipt of diabetes services. Those in the 95th percentile for self-management had a mean HbA1c level of 7.3 (95% CI 6.4–8.3), whereas those in the 5th percentile had mean levels of 8.3 (7.4–9.2). For every 10-point increase in patients’ ratings of their diabetes self-management, even after adjusting for number of outpatient visits, the odds of receiving an HbA1c test in the past year increased by 15% (4–27%), of receiving an eye examination increased by 16% (7–27%), and of receiving a nephropathy screen increased by 13% (2–26%).

CONCLUSIONS—In this sample, patients’ assessments of their diabetes self-care using a simple five-question instrument were significantly associated both with actual HbA1c control and with receiving recommended diabetes services. These findings reinforce the usefulness of patient evaluations of their own self-management for understanding and improving glycemic control. The mechanisms by which those patients who are more actively engaged in their diabetes self-care are also more likely to receive necessary services warrant further study.

Improving clinical outcomes in diabetes requires patients to undertake and sustain a complex array of self-care behaviors, including taking medications, monitoring blood glucose levels, following a diet, engaging in regular exercise, and caring for their feet. These and other skilled behaviors to promote health and prevent complications are often called “ self-management.” Randomized controlled trials of interventions to improve patients’ diabetes self-management have led to better glycemic control (1). Moreover, some studies suggest that patients who are more actively involved in their diabetes self-care, independent of contact with health care providers, may be more likely to receive important diabetes processes of care, such as HbA1c tests and dilated eye examinations (26).

How best to evaluate and support patients’ diabetes self-management is a critically important question. A number of reliable and valid self-reported measures describing different facets of diabetes self-management through questionnaires and structured interviews have been developed in recent years (7,8). These measures are helpful tools to enable clinicians and researchers to evaluate those areas of self-management in which patients with diabetes may need additional support. However, most studies of adults with type 2 diabetes have not sought to confirm a significant association between patients’ reports of their diabetes self-management and glycemic control (8).

Although many factors influence glycemic control (e.g., genetics, physiology, and the quality of medical care), good self-management is a critical pathway to success. As we develop and evaluate initiatives to improve patients’ diabetes self-management, assessing self-management in a way that correlates with achieved glycemic control is an important clinical and methodological issue. For example, in a nationwide sample of 1,314 diabetic patients receiving diabetes care from 25 Veterans’ Affairs (VA) facilities, we found that patients who gave their health care providers higher ratings for communication effectiveness and participatory decision-making styles also evaluated their own diabetes self-management over the past year more positively (9). The scale in this study focused on a key component of patients’ diabetes self-management: their assessment of how well they had been able to adhere to recommended treatment protocols. It consisted of patients’ ratings of how difficult it had been over the past year to manage their diabetes self-care according to treatment recommendations in the areas of medication, diet, exercise, blood glucose monitoring, and foot care, a measure that incorporates both self-efficacy and understanding of provider recommendations. In determining the clinical implications of better patient-reported self-management, it is important to know whether patients’ evaluations of their own self-management, using this scale, in turn correlated with their HbA1c levels at the time they were surveyed. In addition, it would be useful to more rigorously test the hypothesis, raised in prior research (26), that “activated” patients who are more able to follow treatment protocols may also be more likely to receive recommended health care services. Accordingly, in a subsample of patients from the above study, we examined whether assessments of their diabetes self-management in five areas correlated with their level of actual glycemic control and other aspects of high-quality diabetes care (an HbA1c test, a dilated eye examination, and a nephropathy screen within the recommended time period).

### Study population

The initial sample comprised 2,000 veterans receiving diabetes care at 1 of 25 VA medical centers located in four Veterans’ Integrated Service Networks, representing three of the four census regions (9,10). Patients were identified using electronic pharmacy and laboratory information for fiscal years 1998 and 1999 from each participating VA facility and a national utilization database (11,12). Patients were eligible if they had two outpatient visits of any kind in fiscal year 1999 and were identified as having diabetes based on the following criteria: within the past 12 months they had 1) at least one prescription for a glucose control medication or monitoring supplies; 2) two or more outpatient visits with a diabetes-related ICD-9 code; or 3) one hospitalization with a diabetes-related ICD-9 code (250.x, 357.2, 362.0, or 366.41). From among those eligible, we selected a random sample of 80 patients from each of the 25 facilities for inclusion in the study.

Approval for the study was obtained from the VA Ann Arbor Institutional Review Board (IRB) and from the IRBs at 21 facilities. The survey was administered in collaboration with the VA Office of Quality and Performance as part of a broad effort to improve the quality of diabetes care in the VA Administration.

### Data collection

The self-management questions were part of a questionnaire created for the Diabetes Quality Improvement Project (DQIP) (10,13). DQIP developed a number of quality measures adopted by the National Committee for Quality Assurance for use in the Health Plan Employer Data and Information Set (HEDIS) (14,15) and a patient survey to assess issues of diabetes care quality. The survey contained ∼150 items as described elsewhere (9). A total of 1,431 participants from 25 facilities returned a survey, for an overall response rate of 72%. Administrative data showed that the respondents were older (68 vs. 65 years, P < 0.001) and had more outpatient visits than nonrespondents in the study period (4.8 vs. 4.1, P < 0.001).

As described elsewhere (16), trained experienced medical abstractors used an electronic medical record abstraction tool to collect data necessary to calculate the diabetes set of measures used for the fourth quarter 2000 VA External Peer Review Program (http://www.oqp.med.va.gov; accessed 12 February 2002), which includes the DQIP and HEDIS 2000 diabetes set measures (1517). The time period for data collection was 1 February 1999 to 31 January 2000 for most measures and 1 February 1998 to 31 January 2000 for the measures requiring 2 years of data.

We requested medical record information for the 1,174 survey respondents at the 21 facilities with IRBs (no chart reviews were conducted at four facilities without IRBs). Of these requested records, 89 (8%) were not abstracted because they could not be located (n = 28), because the record did not indicate a visit to the facility during the study period (n = 44), or because the record did not indicate that the patient had diabetes (n = 17). Of the remaining 1,085 cases, 53 were dropped before analyses because their surveys indicated that they were deceased (n = 18) or that they did not have diabetes (n = 35). This resulted in a final sample of 1,032.

### Study variables

A scale for patients’ assessment of their diabetes self-management was the principal independent variable for this study’s analyses. For five separate areas of diabetes self-care (taking medications, exercising, following an eating plan, blood glucose monitoring, and foot care), respondents were asked, “Over the past year, how difficult has it been for you to do each of the following exactly as the doctor who takes care of your diabetes suggested?” Five of the valid response categories ranged from “So difficult that I couldn’t do it at all” to “Not difficult, I got it exactly right.” Respondents could also mark that that area of diabetes care was “Not applicable” for them. Possible scores for the scale ranged from 0 to 100, with higher scores meaning greater treatment adherence in the five domains (mean 71.3, SD 17.11, Cronbach’s α 0.68). The scale was designed to reflect how well patients feel able to manage aspects of their diabetes care and has been found to be a valid reflection of self-care behaviors (13). In light of the evidence that adherence to treatment recommendations in one area of diabetes care may not correlate strongly with adherence in others (7,8,18,19), we also looked at each domain separately.

Other variables included in the analyses to adjust for factors that might influence glycemic control and receipt of diabetes services included patients’ age, sex, ethnicity (coded as white and minority), education level, household income, insulin use (vs. oral medications only or no medications), and diabetes severity and comorbidities. The number and severity of diabetes comorbidities were measured using the components of the Total Illness Burden Index directly related to diabetes, a validated scale that ranges from 0 to 100 (13,20,21). In addition, we adjusted for VA health services use with two measures: whether respondents had received >80% of their health care at that VA facility in the past year and whether they had had more than two outpatient visits at that facility in the past year.

Our principal dependent variables were patients’ most recent recorded HbA1c before survey administration (January 2000) and three DQIP process measures (receipt of an HbA1c test and dilated eye examination in the study period [1 February 1999 to 31 January 31] and nephropathy screen in the study period or year before) constructed from medical record data. Each process measure was coded 1 if eligible patients had received the service within the recommended time interval. We also gathered information on whether patients had received foot examinations and blood pressure checks. Because almost all patients in our sample had received both of these (85 and 99%, respectively), we did not include them in our analyses.

### Analyses

The data were analyzed using bivariate and multivariable methods with Stata 7 on datasets with all linkable identifying information removed (22). We used multiple linear regression to assess the relationship between patients’ reported self-management and HbA1c level and logistic regression to model the association between patients’ self-management scores and receipt of necessary diabetes services. All regression models were adjusted for the covariates described above. The amount of missing data was <7% for all variables. However, to avoid selection bias and inaccurate inferences resulting from the cumulative effects of listwise deletion, we imputed values for three independent variables that had >5% missing data: patient education, income, and age (23). Because preliminary analyses suggested significant between-facility variation, we adjusted the standard errors using the Huber-White heteroskedastic consistent estimator of the variance/covariance matrix with cluster correction (24,25). Finally, because of the nonproportional sampling strategy, we estimated all regression models with sampling weights (26,27).

To demonstrate the magnitude of the association of self-reported self-management with patient HbA1c values, we constructed a linear multivariate regression model to calculate predicted HbA1c levels and rate of receipt of each diabetes service for the 5th percentile (score = 30), 50th percentile (score = 70), and 95th percentile (score = 100). We conducted similar analyses for receipt of key diabetes services using logistic regression. For these calculations, we held all covariate values constant at their means (21,28).

Regression diagnostic procedures yielded no evidence of substantive multi-collinearity, heteroskedasticity, or overly influential outliers in any of the models. There was no evidence of significant interactions between diabetes self-management and ethnicity, income, diabetes severity, or education, or of second-order curvilinear relationships between the scale of self-reported self-management and any of the outcome variables.

Table 1 summarizes principal demographic and health characteristics of the survey sample. The overall health status of respondents was quite poor, with 60% describing their health as “poor” or “fair.” Rates of health service use were high, with 92% having had more than two outpatient visits at their VA health facilities in the past year. Most (82%) patients received >80% of their care in the past year at their VA health facilities.

### Relationship of patients’ assessments of their diabetes self-management to glycemic control

Higher patient self-ratings of diabetes self-management were significantly associated with lower HbA1c levels in bivariate analyses, and this finding persisted after adjusting for possible confounding variables. Patients’ diabetes self-management (standardized β = − 0.13, P < 0.001) and patients’ age (standardized β = −0.15, P < 0.001) were both independently associated with lower HbA1c levels, whereas being on insulin (standardized β = 0.22, P < 0.001) was associated with higher HbA1c levels. No other covariates were significantly associated with HbA1c levels in the multivariate model. In analyses looking separately at each of the five domains of the self-management scale, patients’ positive assessments of the four domains of taking medications, monitoring blood glucose levels, diet, and exercise were each independently (P < 0.05) associated with lower HbA1c levels. There was no significant association between patients’ assessment of their foot self-care and their HbA1c levels.

Predicted HbA1c levels based on patients’ ratings of their diabetes self-management are presented in Table 2. The 1-point difference in HbA1c levels between patients with self-management scores in the 5th percentile and patients with scores in the 95th percentile is both statistically and clinically significant.

### Relationship of patients’ assessments of their diabetes self-management to receipt of necessary diabetes services

A more positive self-management assessment was also associated with having received an HbA1c test, an eye examination, and a nephropathy screen within DQIP- and HEDIS-recommended time intervals. Even after adjusting for other possible confounders, including two measures of health services use, for every 10-point increase in self-rated diabetes management, the odds of having received an HbA1c lab increased by 15%, receipt of an eye examination by 16%, and receipt of a nephropathy screen by 18% (P < 0.005 for all three services). Table 2 shows the increases in mean percentages of patients who received all three of the recommended diabetes services with increases in their diabetes self-management assessment scores.

In this study, patients’ assessments of their diabetes self-management were significantly associated with their achieved glycemic control, even after controlling for possible confounding variables. The separate domains of taking medications, monitoring blood glucose levels, exercising regularly, and following a diabetes diet were each independently associated with glycemic control. In light of the complex set of factors that contribute to glycemic control, it is not surprising that our multivariable model accounted for only 11% of the variance in HbA1c values. Patient behaviors alone do not determine glycemic control. However, the magnitude of the correlation between the self-management scale and HbA1c levels was clinically significant. The difference in HbA1c between someone at the low versus the high end of the self-management scale was 1 full point. Moreover, we found that patients’ reported self-management was associated with receiving all three diabetes care processes that we examined.

Our findings suggest that patients’ evaluations of their diabetes self-management with this scale may indeed serve as a proxy for self-care behaviors that lead to improved glycemic control. These results also reinforce other researchers’ findings of the importance of assessing patients’ self-management, so as to better understand obstacles patients may face and to evaluate how educational or other care strategies may improve glycemic control (7). Finally, our findings raise intriguing questions about why patients with better self-reported diabetes management also were more likely to have received necessary diabetes processes of care. It is possible that those providers with communication and decision-making styles more supportive of patients’ diabetes self-management also are more likely to order and ensure that their patients receive necessary services (9,29,30). However, the converse may also be true: patients who are more effective in following recommended self-care protocols (and have greater confidence in their abilities as reflected in their self-assessment) may also be more effective in securing necessary diabetes services (4,29,31). For example, these patients may be more informed about the tests they need and remind their providers when certain tests are due, or they may be more likely to follow through on getting the tests and services (e.g., dilated eye examinations) their providers order. Further research, using longitudinal design and controlling for provider effects, should explore whether and how those patients who are more actively engaged in their diabetes self-care are also more likely to receive necessary services.

Our study has a number of limitations. First and most importantly, its cross-sectional design does not allow us to establish that patients’ assessment of their diabetes self-management was causally associated with glycemic control or receipt of diabetes services. Patients with better glycemic control may evaluate their self-management as better than those who have more serious disease and higher HbA1c levels, and those patients who receive more recommended services may also have better reported self-management for another reason not measured in our analyses. Regarding this point, it is worth noting that patients’ reported self-management was not associated with either of our two measures of health services use. This lack of association suggests that fewer outpatient visits is not the reason patients who evaluate their self-management poorly are less likely to receive necessary tests. Second, this study population consisted predominantly of older men. Our findings may not be generalizable to younger or predominantly female populations and should be repeated in other settings. Third, it is important to emphasize that the measure we used provides both a general assessment of how difficult patients found carrying out recommended activities in five areas of diabetes self-care and their evaluation of their level of success in undertaking these activities. In this sense, it is complementary to other self-management measures that explicitly enable patients to describe or quantify their self-care activities within a discrete recent time interval (7). These other measures may be more useful in targeting specific ways to help patients improve their current self-management. With the exception of a recent measure developed for type 1 diabetes that correlated with HbA1c levels in a sample of adolescents (8), these scales have not been shown to correlate with glycemic control. One could argue that the measure we assessed in this study serves as a marker for confidence in performance of past diabetes self-care activities. Moreover, because this scale assesses how closely patients executed treatment recommendations in five areas of diabetes self-care, it may also be tapping into how well patients understand these recommendations and whether their providers have in fact provided clear guidance in each area. Future research should explore how the scale we used in this study correlates with self-reported measures that provide more precise descriptions of the frequency with which respondents performed various self-care activities and with scales explicitly assessing patients’ “self-efficacy” (7,8,32,33). It would also be useful to evaluate the utility of this scale in measuring the impact of outpatient education programs and other interventions on patients’ assessments of their diabetes self-management.

In conclusion, we found that patients’ evaluation of their diabetes self-management was associated with both their HbA1c levels and their receipt of recommended diabetes services. A scale that correlates with glucose control can be used in two important ways: 1) to assess glucose control in settings, such as survey research and community programs, in which HbA1c values are not readily available and 2) in clinical settings to help elucidate which aspects of self-care patients may be having trouble with and thus better target educational and motivational efforts. The association of patients’ reported self-management and receipt of diabetes services is consistent with current hypotheses that “more activated” patients may interact with their providers in a way to help ensure that they receive necessary technical processes of care (4,6,34).

Table 1—

Characteristics of patient respondents

 Age (years) 67 ± 10 Years with diabetes 12 ± 10 Male sex 98 Ethnicity White 84 Minority 16 Years of education 12 ± 3 Income $10,000 or less 20$10,001 to $15,000 26$15,001 to $25,000 29$25,001 or more 25 Hypoglycemic regimen Oral medications only 59 On insulin 34 No medications 7 Health status Excellent or very good 7 Good 33 Fair 43 Poor 17 Diabetes Comorbidity (TIBI) scale* 41 ± 19 Percent with >80% of care at VA facility 82 Percent with greater than two outpatient visits at VA in past year 92
 Age (years) 67 ± 10 Years with diabetes 12 ± 10 Male sex 98 Ethnicity White 84 Minority 16 Years of education 12 ± 3 Income $10,000 or less 20$10,001 to $15,000 26$15,001 to $25,000 29$25,001 or more 25 Hypoglycemic regimen Oral medications only 59 On insulin 34 No medications 7 Health status Excellent or very good 7 Good 33 Fair 43 Poor 17 Diabetes Comorbidity (TIBI) scale* 41 ± 19 Percent with >80% of care at VA facility 82 Percent with greater than two outpatient visits at VA in past year 92

Data are means ± SD or %. n = 1,032.

*

TIBI, Total Illness Burden Index (scaled from 0 to 100, with higher scores meaning more diabetes comorbidities).

Table 2—

Predicted mean HbA1c level and percentage of patients who received diabetes services based on percentile of patients’ diabetes self-management assessment scores

Percentile of patients’ diabetes self-management score
5th (score = 30) 8.3 (7.4–9.2) 77 (55–98) 46 (34–88) 37 (22–52)
50th (score = 70) 7.7 (6.9–8.6) 85 (67–103) 60 (36–84) 55 (39–71)
95th (score = 100) 7.3 (6.4–8.2) 89 (75–103) 69 (47–91) 67 (51–83)
Percentile of patients’ diabetes self-management score
5th (score = 30) 8.3 (7.4–9.2) 77 (55–98) 46 (34–88) 37 (22–52)
50th (score = 70) 7.7 (6.9–8.6) 85 (67–103) 60 (36–84) 55 (39–71)
95th (score = 100) 7.3 (6.4–8.2) 89 (75–103) 69 (47–91) 67 (51–83)

Data in parentheses are 95% CIs. Mean HbA1c levels and mean percentages of patients who received each diabetes service were calculated for alternate values of patients’ assessments of their diabetes self-management while holding all other variables constant at the mean, using the linear or logistic regression model with all covariates. All models were adjusted for ethnicity, sex, income, education, age, diabetes comorbidities, insulin use, whether the patient had two or more outpatient visits at the VA facility in the past year, whether the patient received >80% of care at the VA facility in the past year, sampling weights, and clustering at the facility level.

This work was supported by the VA Health Services Research and Development (HSR&D) Service (DIS#99221-1), the VA Office for Quality and Performance, and the Robert Wood Johnson Foundation. Dr. Kerr is a VA HSR&D Career Development awardee.

The authors thank Mary Hogan for outstanding project management and Jill Baker for meticulously coding the survey responses.

1
Norris SL, Engelgau MM, Narayan KM: Effectiveness of self-management training in type 2 diabetes: a systematic review of randomized controlled trials.
Diabetes Care
24
:
561
–587,
2001
2
Anderson RM, Funnell MM, Butler PM, Arnold MS, Fitzgerald JT, Feste CC: Patient empowerment: results of a randomized controlled trial.
Diabetes Care
18
:
943
–949,
1995
3
Hofer TP, Katz SJ: Healthy behaviors among women in the United States and Ontario: the effect on use of preventive care.
Am J Public Health
86
:
1755
–1759,
1996
4
Smedley BD, Stith AY, Nelson AR, Institute of Medicine (US), Committee on Understanding and Eliminating Racial and Ethnic Disparities in Health Care:
Unequal Treatment Confronting Racial and Ethnic Disparities in Healthcare
. Washington, DC, National Academy Press,
2002
5
Schillinger D, Grumbach K, Piette J, Wang F, Osmond D, Daher C, Palacios J, Sullivan GD, Bindman AB: Association of health literacy with diabetes outcomes.
JAMA
288
:
475
–482,
2002
6
Walsh ME, Katz MA, Sechrest L: Unpacking cultural factors in adaptation to type 2 diabetes mellitus.
Med Care
40
:
129
–139,
2002
7
Toobert DJ, Hampson SE, Glasgow RE: The summary of diabetes self-care activities measure: results from 7 studies and a revised scale.
Diabetes Care
23
:
943
–950,
2000
8
Harris MA, Wysocki T, Sadler M, Wilkinson K, Harvey LM, Buckloh LM, Mauras N, White NH: Validation of a structured interview for the assessment of diabetes self-management.
Diabetes Care
23
:
1301
–1304,
2000
9
Heisler M, Bouknight RR, Hayward RA, Smith DM, Kerr EA: The relative importance of physician communication, participatory decision making, and patient understanding in diabetes self-management.
J Gen Intern Med
17
:
243
–252,
2002
10
Krein SL, Vijan S, Pogach L, Hogan M, Kerr EA: Aspirin use and counseling about aspirin among patients with diabetes.
Diabetes Care
25
:
965
–970,
2002
11
Pogach LM, Hawley G, Weinstock R, Sawin C, Schiebe H, Cutler F, Zieve F, Bates M, Repke D: Diabetes prevalence and hospital and pharmacy use in the Veterans Health Administration (1994): use of an ambulatory care pharmacy-derived database.
Diabetes Care
21
:
368
–373,
1998
12
Boyko EJ, Koepsell TD, Gaziano JM, Horner RD, Feussner JR: US Department of Veterans Affairs medical care system as a resource to epidemiologists.
Am J Epidemiol
151
:
307
–314,
2000
13
Kaplan SH:
Diabetes Quality Improvement Project: Patient Survey Final Report: Report to National Committee for Quality Assurance
. Washington, DC, National Committee for Quality Assurance,
2001
14
Schneider EC, Riehl V, Courte-Wienecke S, Eddy DM, Sennett C: Enhancing performance measurement: NCQA’s road map for a health information framework: National Committee for Quality Assurance.
JAMA
282
:
1184
–1190,
1999
15
National Committee for Quality Assurance:
HEDIS 2000
. Washington, DC, National Committee for Quality Assurance,
1999
16
Kerr EA, Smith DM, Hogan M, Krein SL, Pogach L, Hofer TP, Hayward RA: Comparing clinical automated, medical record, and hybrid data sources for diabetes quality measures.
Jt Comm J Qual Improv
28
:
555
–565,
2002
17
Fleming BB, Greenfield S, Engelgau MM, Pogach LM, Clauser SB, Parrott MA: The Diabetes Quality Improvement Project: moving science into health policy to gain an edge on the diabetes epidemic.
Diabetes Care
24
:
1815
–1820,
2001
18
Goodall TA, Halford WK: Self-management of diabetes mellitus: a critical review (Review).
Health Psychol
10
:
1
–8,
1991
(erratum Health Psychol 11:77, 1992)
19
Clement S: Diabetes self-management education (Review).
Diabetes Care
18
:
1204
–1214,
1995
20
Kaplan SH, Greenfield S, Gandek B, Rogers WH, Ware JE Jr: Characteristics of physicians with participatory decision-making styles.
Ann Intern Med
124
:
497
–504,
1996
21
Hayward RA, Manning WG, Kaplan SH, Wagner EH, Greenfield S: Starting insulin therapy in patients with type 2 diabetes: effectiveness, complications, and resource utilization.
JAMA
278
:
1663
–1669,
1997
22
Stata Corporation:
Stata Statistical Software: Release 7.0
. College Station, TX, Stata Corporation,
2001
23
Little RJA, Rubin DB:
Statistical Analysis With Missing Data
. New York, John Wiley & Sons,
1987
24
Huber P: The behavior of maximum likelihood estimates under non-standard conditions.
Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability
1
:
221
–223,
1967
25
White H: A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity.
Econometrica
40
:
817
–830,
1980
26
Fuller WA: Regression analysis for sample survey.
Sankha
Series
C
:
117
–132,
1975
27
Kish L, Frankel MR: Inference from complex samples.
J Royal Stat Soc B
36
:
1
–37,
1974
28
Little RJ: Direct standardization: a tool for teaching linear models for unbalanced data.
American Statistician
32
:
38
–43,
1982
29
Kaplan SH, Greenfield S, Ware JE Jr: Assessing the effects of physician-patient interactions on the outcomes of chronic disease.
Med Care
27
:
S110
–S127,
1989
(erratum Med Care 27:679, 1989)
30
Hays RD, Kravitz RL, Mazel RM, Sherbourne CD, DiMatteo MR, Rogers WH, Greenfield S: The impact of patient adherence on health outcomes for patients with chronic disease in the Medical Outcomes Study.
J Behav Med
17
:
347
–360,
1994
31
Greenfield S, Kaplan S, Ware JE Jr: Expanding patient involvement in care: effects on patient outcomes.
Ann Intern Med
102
:
520
–528,
1985
32
Williams GC, Freedman ZR, Deci EL: Supporting autonomy to motivate patients with diabetes for glucose control.
Diabetes Care
21
:
1644
–1651,
1998
33
Senecal C, Nouwen A, White D: Motivation and dietary self-care in adults with diabetes: are self-efficacy and autonomous self-regulation complementary or competing constructs?
Health Psychol
19
:
452
–457,
2000
34
Stewart AL, Napoles-Springer A, Perez-Stable EJ: Interpersonal processes of care in diverse populations.
Milbank Q
77
:
305
–339,
1999

Address correspondence and reprint requests to Michele Heisler, MD, MPA, VA HSR&D, PO Box 130170, Ann Arbor, MI 48113-0170. E-mail: mheisler@umich.edu.

Received for publication 21 August 2002 and accepted in revised form 9 November 2002.

The views expressed herein do not necessarily represent those of the Robert Wood Johnson Foundation or the Department of Veterans Affairs.

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