OBJECTIVE—To test the hypothesis that patient readiness to change (RTC) predicts future changes in glycemic control in adults with diabetes.

RESEARCH DESIGN AND METHODS—We linked survey data with HbA1c data for a stratified random sample of consenting adults with diabetes. Change in HbA1c from baseline to the 1-year follow-up was computed and used as a dependent variable. Linear regression models assessed RTC and other patient variables as predictors of HbA1c change.

RESULTS—Among 617 patients with baseline HbA1c ≥7% and complete data for analysis, RTC predicted subsequent improvement in HbA1c for those with higher physical functioning (interaction t = −2.45, P < 0.05). Other factors that predicted HbA1c improvement in multivariate linear regression models included higher self-reported medication adherence (t = −4.41, P < 0.01), higher baseline HbA1c (t = −15.08, P < 0.01), and older age (t = −2.61, P < 0.01).

CONCLUSIONS— Diabetes RTC independently predicts change in HbA1c for patients with high but not for patients with low functional health status. Customized use of RTC assessment may have potential to improve care.

The large gap between recommended levels of diabetes care and levels actually achieved in primary care or subspecialty practice is well recognized (1). Recent studies suggest that patient education and activation may be an especially important factor related to the control of glucose in those with diabetes. For example, improvement initiatives that include a patient-activation component are generally more effective than those that do not (2).

A patient’s “readiness to change” (RTC) diabetes care provides a potentially useful metric of patient activation (37). RTC, which can be assessed by two brief questions, may indicate whether a patient is actively planning to make behavioral changes to improve their diabetes care. Behaviors related to medication adherence, nutrition therapy, physical activity, and home glucose monitoring are strongly related to the state of glycemic control in patients with diabetes. Moreover, a patient who is ready to change may be an ideal candidate for intensification of diabetes therapy.

Thus, if RTC is a valid predictor of improved glycemic control, its wider use in clinical practice could provide information useful to clinicians at the time of a diabetes office visit. This work tests the hypothesis that patient RTC is a significant predictor of subsequent change in glycemic control, measured by change in HbA1c over time. The relationship of RTC to adherence to therapy and treatment intensification is also assessed.

This report is based on data from a prospective cohort study, Project QUEST, designed to identify patient, physician, and other factors that predict quality of care provided to adults with diabetes or heart disease. This work focuses on patient-related factors that predict subsequent change in HbA1c level in adults with diabetes.

To be included in this analysis, potential study subjects had to be enrolled in HealthPartners, a Minnesota health plan, and have an established diagnosis of diabetes in 1998 based on meeting at least one of the following two criteria: 1) have one or more inpatient or two or more outpatient ICD-9 (International Classification of Diseases) diagnostic codes 250.xx for diabetes or 2) have a filled prescription for a diabetes-specific drug, including insulins, sulfonylureas, biguanides, α-glucosidase inhibitors, thiazolidinediones, or meglitamides. This method of diabetes identification has been previously validated in the study population and has an estimated sensitivity of 0.91, a specificity of 0.99, and a positive predictive value of 0.94 (8).

In addition, study subjects had to meet all the following eligibility criteria: 1) have health care insurance through HealthPartners in 1998, 2) be at least 19 years old in 1998, 3) receive care from a clinic that had a minimum of 10 HealthPartners-insured diabetic patients per primary care provider in 1998, and 4) receive care from a clinic and medical group whose leaders were willing to participate in the study. We sought to recruit patients from 21 medical groups; 18 of these participated in the study. These 18 medical groups had 84 eligible clinics; 83 of these participated in the study. Because all study subjects had similar health insurance coverage, all had access to diabetes education, endocrinology referrals, and other diabetes-related services. We then randomly selected a sample of 10 adults with diabetes per primary care provider at each clinic. The complex sampling strategy was designed to power the study to identify organizational characteristics of clinics and medical groups that may affect quality of care.

Following identification of subjects, a patient survey was sent to 4,780 adults with diabetes and returned by 2,832, for a response rate of 59.2%. To further screen out case subjects without diabetes, 83 patients who responded “no” to the baseline patient survey item “Has a doctor ever told you that you had diabetes?” were excluded from the study. Of the 2,749 survey respondents who reported having diabetes, 2,056 (74.8%) gave written informed consent for a medical record review, which was completed for 2,016 (98.1%). This report is based on the 1,794 adults with diabetes who had a medical record review and an HbA1c measure in the baseline year. Predictive analyses were further limited to 746 of these patients who had a baseline HbA1c ≥7%, which is above the currently recommended level of glycemic control (9,10), and who had HbA1c measures available at both baseline and follow-up as well as the RTC measure.

Variable definition and measurement

Administrative data on health plan members was initially gathered to determine eligibility for the study. Information collected for this purpose included medical group and clinic membership, ICD-9 codes, CPT-4 (Current Procedural Terminology, Edition 4) codes, and pharmacy data.

Patients were surveyed in 2001 to obtain data on duration of diabetes, RTC for diabetes care, indicators of microvascular (numbness/tingling in feet or hands, foot ulcers/infections, retinopathy) and macrovascular (heart trouble, heart attack, stroke) complications, self-reported medication changes for diabetes in the past 12 months, perception of whether diabetes has been in good control for >6 months, self-reported medication adherence, belief that following the doctor’s orders will help delay future problems, SF-12 physical function and mental function measures, height, weight, employment status, education level, and other information. Whenever possible, items and scales used were drawn from validated scales or from previously published studies (1113). Medication adherence was assessed based on the response to a survey question: “In the past week how often have you not taken your medications as prescribed?” Those reporting not taking medications as directed once a week or more often were classified as having low adherence.

Patients were assigned to an RTC category on the basis of their responses to two items used in previous studies: “I am intending to make changes in my diabetes self-care in the (next 6 months)/(next month)” (14,15). Precontemplators were defined as those not intending to make changes in the next 6 months or in the next month. Contemplators were defined as those who intended to make changes in the next 6 months but not in the next month. Preparers were defined as those who intended to make changes in the next month. Due to the small percentage of patients in the contemplation stage (5.6%), individuals in this stage were combined with those in the preparation stage for all analyses. Thus, a single dummy code for stage of change used in the regression analyses contrasted patients in the contemplation and preparation stages with those in the precontemplation stage. The analysis was limited to the three stages of change relevant to patients with HbA1c ≥7%.

HbA1c levels done in the 12-month period before and the 12-month period after the patient survey were obtained from chart audits. When more than one HbA1c measurement was available in a particular 12-month chart period, the HbA1c test value closest to the end of the 12-month period was identified and used in the analysis. Although laboratories used by various clinics may have slightly different scales for HbA1c assays, the normal range of HbA1c was uniformly from a lower bound of 4.4 to 4.6% to an upper bound of 6.0 to 6.1%. We monitored laboratory arrangements of study clinics and did not identify any changes in laboratory assay for HbA1c during the period of data collection. Although more formal standardization of HbA1c could be done, the HbA1c values reported here are those used to guide diabetes care in the community. The dependent variable used in the analysis consists of the difference score of HbA1c level taken in the year after the survey minus the level in the year before the survey, so that in multivariate models a negative score represents an improvement in glycemic control over time.

Statistical analysis

Frequencies and χ2 tests were used to describe sample attributes and to compare individuals with HbA1c values <7% and those with values ≥7% as well as to compare individuals by RTC category. Ordinary least-squares methods were used throughout this work rather than a multilevel/nested approach due to the findings in previous work on this sample that clinic and medical group levels accounted for a negligible fraction of variance in HbA1c change over time. Paired t tests were used to compare HbA1c at baseline and follow-up for the entire sample of diabetic patients and for those with HbA1c at baseline ≥7%.

To test the hypothesis that RTC category predicts HbA1c change, a series of ordinary least-squares multiple linear regression equations was estimated. All equations included an a priori set of control variables that consisted of sex, age, education (college graduate or not), and duration of diabetes.

Multivariate models included predictor variables that showed empirical evidence of zero order correlation with change in HbA1c at an α < 0.10. These predictors included patient report of whether diabetes has been in good control in the past 6 months, self-reported medication compliance, and patient belief that following a doctor’s advice will delay future problems. Other predictors evaluated included those considered to have theoretical importance, including presence of microvascular or macrovascular problems, patient-reported changes to diabetes medication in the past 12 months, patient-reported visits with a diabetes educator, satisfaction with medical care, trust of doctor, and mental and physical function scores from the SF-12.

The model-building strategy involved testing the predictive value of the dummy code for RTC in models containing one or more of the sets of predictors previously identified (a priori covariates, empirically promising, theoretically important). A priori covariates were included in all models. Items from the two other sets of predictors having at least marginally significant (P < 0.10) model parameters in the initial analysis were retained for entry into the final model. In order to test for potential moderators of the effect of RTC category on HbA1c change, all predictors were individually entered in equations along with the a priori covariates, the stage of change dummy code, and the two-way interaction of the predictor and the dummy code. Significant interactions from these single interaction term models were retained for entry into the final model.

The final model was first estimated with the set of a priori covariates, significant or marginally significant empirical and theoretical predictors, significant interaction terms, and the constituent items of these interaction terms. For the last estimation of the final model, the a priori covariates were retained as well as the dummy code for RTC category and all terms that were statistically significant from the initial estimation of this model.

Human subjects protection

The study protocol was reviewed in advance, approved, and monitored by the HealthPartners Institutional Review Board.

Some small but statistically significant differences exist by survey responder status among the 4,780 case subjects identified by administrative data as having diabetes. Survey responders were more likely than nonresponders to have heart disease according to administrative data (23.0 vs. 18.7%, P = 0.0004), to be women (47.9 vs. 43.9%, P = 0.006), and to be older (mean age 61.2 vs. 57.0 years, P < 0.0001). No other comparisons were made because additional data elements were not available from nonresponders.

Few differences were found by chart audit consent status among the 2,749 individuals who completed a survey and self-reported having diabetes. Individuals who consented to chart audits (n = 2056) and those who did not (n = 693) showed no differences in sex, living situation, education level, full-time work status, heart disease status, duration of diabetes, self-report of out of control diabetes, or stage of change. However, those who consented to a chart audit were older than those who did not consent (61.4 vs. 60.2 years, P = 0.03).

Some differences were found by baseline HbA1c level. Individuals with HbA1c ≥7% are more likely than those with HbA1c <7% to have a longer duration of diabetes, are less likely to say their diabetes has been in good control for >6 months, and are more likely to be in the contemplation or preparation stage.

Characteristics of study participants by RTC category are shown in Table 1 for patients with HbA1c ≥7%. Note that of 983 subjects with HbA1c ≥7%, RTC category is classified for 881 with available data for RTC classification. When restricting the sample to just those patients with HbA1c ≥7% in the baseline year of the study, differences by stage of change are numerous. Those in the contemplation or preparation stage are more likely than those in precontemplation to be women, younger, employed full-time, to have a shorter duration of diabetes, and higher BMI. Among those with HbA1c ≥7%, 82.2% of precontemplators compared with 43.9% of contemplator/preparation patients reported that their diabetes is already in good control.

Among patients with a baseline HbA1c ≥7%, mean HbA1c was 8.23 ± 1.19% (±SD) at baseline and 7.86 ± 1.23% at follow-up. The mean change of −0.37% was statistically significant (t829 = −8.79, P < 0.0001).

When considering the entire sample of adults with diabetes and HbA1c ≥7%, RTC was not a statistically significant predictor of change in HbA1c. Specifically, RTC did not predict change in HbA1c in preliminary regression models when entered along with a priori covariates (RTC β = −0.012, P = 0.89), with a priori and empirical covariates (β = −0.027, P = 0.77), or a priori and theoretical covariates (β = −0.045, P = 0.65). However, a significant interaction term in the final multiple regression model (described below) indicates that RTC relates significantly to reduction in HbA1c, but only for the subset of individuals with higher physical functioning at baseline.

The final multiple regression model incorporating a priori covariates is shown in Table 2 for adults with diabetes and baseline HbA1c ≥7%. Negative model parameters indicate a reduction in HbA1c from baseline to follow-up and improved glycemic control. Significant improvement in HbA1c was found for individuals with higher HbA1c at baseline, those who were older, and those who took medications as prescribed.

The significant effect for RTC must be interpreted in light of the significant interaction of RTC and SF-12 physical score. Figure 1 illustrates the form of this interaction in the prediction of HbA1c improvement. For individuals with poorer health (SF-12 score <45, which is the 50th percentile of this distribution), RTC does not significantly predict improvement in HbA1c (β = 0.249, P = 0.07). However, for individuals with higher physical functioning scores (SF score of ≥45), higher RTC significantly predicted improvement in HbA1c (β = −0.245, P = 0.046).

Results indicate that RTC is a significant predictor of future HbA1c change in those with HbA1c ≥7% with high functional status at baseline. In these patients, assessing RTC using the question “Are you considering making changes in your diabetes self-care in the next 6 months” may be a good indicator of future improvement in HbA1c. The fact that adherence to medication is an independent predictor of change in HbA1c suggests that RTC and adherence may be complementary but distinct domains (1620).

In an earlier study, Peterson et al. (15) found RTC to be strongly predictive of HbA1c change in all diabetic patients. However, the earlier study included few study subjects, and multivariate covariate adjustment was precluded by the small sample size. Also, the functional health status of the patients in the Peterson study was not specified.

Our results partially support the hypothesis that RTC carries significant predictive information and can be used to maximize the effectiveness and efficiency of diabetes care delivered in office practice (18,2123). Our data also suggest that RTC exerts effects beyond what may be mediated by adherence to medications, beliefs about diabetes, or even intensification of diabetes therapy, factors shown in previous reports to predict future changes in HbA1c (2426).

The factors that influence RTC, and its variation over time, are a topic of intensive research across a wide range of behavioral domains, many of them closely related to preventive care or health-related behaviors. A number of studies suggest that RTC for diabetes care and other behavioral domains is complexly determined by the balance of barriers and facilitators, some amenable to personal control, others environmentally determined (13,27,28). Moreover, there is evidence that diabetes care is a complex adaptive system in which both patient mental model and patient RTC evolve over time (13,2729). For this reason, ongoing assessment of RTC and other patient-specific variables is important.

A number of factors limit the interpretation of our results. First, the analysis was limited to those with HbA1c ≥7% and some of the observed improvement may have been related to regression to the mean. However, we have included baseline HbA1c in the analysis to control for regression to the mean. Second, the data are inherently clustered at the physician and clinic level. However, other analyses suggest that the intraclass correlation coefficients related to the clustering are small, and thus we present results using standard multivariate methods. Third, some experts contend that RTC should be applied only to specific behaviors, such as smoking or exercise, and is not well suited for application to more complex patterns of behavior. Finally, there are limits on generalizability of results based on the design of the study and the community in which it was conducted.

Nonetheless, the results we observed are both interesting and important and have clinical implications that extend to the heart of how diabetes care is conceptualized and delivered.

Based on these data, we conclude that RTC and other factors, such as medication adherence, which are related to a patient mental model of diabetes, may act as important facilitators or barriers to improved diabetes care.

This research was supported by grant RO1 HS09946 from the U.S. Agency for Healthcare Research and Quality.

The authors thank the leaders of the 19 Minnesota medical groups who voluntarily participated in Project QUEST, motivated by the desire to improve the care they provide to their patients with diabetes.

1.
Saaddine JB, Engelgau MM, Beckles GL, Gregg EW, Thompson TJ, Narayan KM: A diabetes report card for the United States: quality of care in the 1990s.
Ann Intern Med
136
:
565
–574,
2002
2.
Renders CM, Valk GD, Griffin SJ, Wagner EH, Eijk Van JT, Assendelft WJ: Interventions to improve the management of diabetes in primary care, outpatient, and community settings: a systematic review.
Diabetes Care
24
:
1821
–1833,
2001
3.
Ruggiero L, Rossi JS, Prochaska JO, Glasgow RE, de Groot M, Dryfoos JM, Reed GR, Orleans CT, Prokhorov AV, Kelly K: Smoking and diabetes: readiness for change and provider advice.
Addict Behav
24
:
573
–578,
1999
4.
Velicer WF, Rossi JS, Diclemente CC, Prochaska JO: A criterion measurement model for health behavior change.
Addict Behav
21
:
555
–584,
1996
5.
Prochaska J, DiClemente C: Stages of change in the modification of problem behaviors.
Prog Behav Modif
28
:
183
–218,
1992
6.
Sperl-Hillen J, O’Connor PJ, Carlson RR, Lawson TB, Halstenson C, Crowson T, Wuorenma J: Improving diabetes care in a large health care system: an enhanced primary care approach.
Jt Comm J Qual Improv
26
:
615
–622,
2000
7.
Glasgow RE, Anderson RM: In diabetes care, moving from compliance to adherence is not enough: something entirely different is needed (Commentary).
Diabetes Care
22
:
2090
–2092,
1999
8.
O’Connor P, Rush W, Pronk N, Cherney L: Identifying diabetes mellitus or heart disease among health maintenance organization members: sensitivity, specificity, predictive value and cost of survey and database methods.
Am J Manag Care
4
:
335
–342,
1998
9.
American Diabetes Association: Clinical practice recommendations
2003
.
Diabetes Care
26 (Suppl. 1)
:
S1
–S156,
2003
10.
Institute for Clinical Systems Improvement (ICSI):
Management of Type 2 Diabetes Mellitus
. Bloomington, MN, ICSI,
2002
11.
Stein AD, Lederman RI, Shea S: The Behavioral Risk Factor Surveillance System questionnaire: its reliability in a statewide sample.
Am J Public Health
83
:
1768
–1772,
1993
12.
Ware J Jr, Kosinski M, Keller SD: A 12-item short-form health survey: construction of scales and preliminary tests of reliability and validity.
Med Care Mar
34
:
220
–233,
1996
13.
Greenfield S, Kaplan SH, Ware JE Jr, Yano EM, Frank HJ: Patients’ participation in medical care: effects on blood sugar control and quality of life in diabetes.
J Gen Intern Med
3
:
448
–457,
1988
14.
Ruggiero L, Prochaska JO: Readiness for change: application of the transtheoretical model to diabetes.
Diabetes Spectrum
6
:
21
–60,
1993
15.
Peterson KA, Hughes M: Readiness to change and clinical success in a diabetes educational program.
J Am Board Fam Pract
15
:
266
–271,
2002
16.
O’Connor PJ: From blame to understanding: moving diabetes care forward (Editorial).
J Fam Pract
46
:
205
–206,
1998
17.
O’Connor PJ, Pronk NP: Integrating population health concepts, clinical guidelines and ambulatory medical systems to improve diabetes care.
J Ambul Care Manage
21
:
67
–73,
1998
18.
O’Connor PJ: Patient archetypes, physician archetypes, and tailored diabetes care.
J Am Board Fam Pract
15
:
334
–337,
2002
19.
Prochaska JO, Velicer WF: The transtheoretical model of health behavior change.
Am J Health Promot
12
:
38
–48,
1997
20.
Prochaska JO, Norcross JC, DiClemente CC:
Changing for Good
. New York, Avon Books,
1995
21.
Velicer WF, Prochaska JO, Fava JL, Laforge RG, Rossi JS: Interactive versus noninteractive interventions and dose-response relationships for stage-matched smoking cessation programs in a managed care setting.
Health Psychol
18
:
21
–28,
1999
22.
O’Connor PJ: Organizing diabetes care: identify, monitor, prioritize, intensify (Editorial).
Diabetes Care
24
:
1515
–1516,
2001
23.
Boyle RG, O’Connor PJ, Pronk NP, Tan A: Stages of change for physical activity, diet, and smoking among HMO members with chronic conditions.
Am J Health Promot
12
:
170
–175,
1998
24.
Johnson PE, Veazie PJ, Kochevar L, O’Connor PJ, Potthoff SJ, Verma D, Dutta P: Understanding variation in chronic disease outcomes.
Health Care Manag Sci
5
:
175
–189,
2002
25.
UK Prospective Diabetes Study (UKPDS) Group: Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33).
Lancet
352
:
837
–853,
1998
26.
Campen RB, DiLoreto D: Here to stay: health care compliance: an overview.
Med Group Manage J
47
:
30
–33,
2000
27.
Kleinman A:
The Illness Narratives
. New York, Basic Books,
1988
28.
O’Connor PJ, Crabtree BF, Yanoshik MK: Differences between diabetic patients who do and do not respond to a diabetes care intervention: a qualitative analysis.
Fam Med
29
:
424
–428,
1997
29.
Wilson T, Holt T, Greenhalgh T: Complexity science: complexity and clinical care.
BMJ
323
:
685
–688,
2001

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