Diabetes Quality Improvement in Department of Veterans Affairs Ambulatory Care Clinics : A group-randomized clinical trial

The ACQUIP trial was a group-randomized trial in General Internal Medicine (primary care) clinics at seven VA facilities: Birmingham, Alabama; Little Rock, Arkansas; Richmond, Virginia; San Francisco, California; Seattle, Washington; West Los Angeles, California; and White River Junction, Vermont. Firms at each site (established teams staffed by a group of physicians caring for a panel of patients) were randomized by the study statistician to intervention or control status. Patients were identified through the Veterans Health Information Systems and Technology Architecture (VistA), a comprehensive patient and system database that included automated information on inpatient stays, outpatient visits, medications, laboratory and radiology findings, pharmaceuticals, and mortality. Eligibility was restricted to veterans who were assigned to a primary care provider and who had at least one visit in the prior year. Human subjects approval was obtained at each of the seven sites.

Included in the study were all patients from intervention firms and a random sample of patients in control firms to ensure at least 1,500 completed disease-specific questionnaires for each target condition. Potential participants were mailed a self-administered health screening questionnaire and requested to identify their major chronic medical conditions, including diabetes. Those identifying a history of diabetes received the Seattle Diabetes Questionnaire, a 29-item instrument with the following seven domains: symptoms, complications, education, diabetes self-care, emotional function, control, and satisfaction (11). A number of clinical diabetes experts participated in the design and testing of the diabetes intervention.

Participants were also requested to complete the 36-item Medical Outcomes Study short form (SF-36) to map eight domains for the impact of disease on health status. The SF-36 physical component scale and the mental component scale were also used to summarize overall physical and mental function (12). Participants also completed the 21-item Seattle Outpatient Satisfaction Questionnaire, which includes 12 items from the American Board of Internal Medicine Humanistic Scale to measure providers’ interpersonal skills and additional questions, as adapted from the RAND questionnaire to measure organizational aspects of care, including waiting time and ease of getting appointments (13).

Patient self-reported data were collected every 4 months for intervention patients and annually for control patients. Before primary care visits, intervention firm health care providers received a one-page, graphical, user-friendly, patient-feedback report (Fig. 1). Included were pertinent patient self-reported data and clinical, pharmacy, laboratory, and radiology findings. The report also included rotating practice tips derived from state-of-the-art clinical studies. Control firm providers did not receive patient summaries or practice tips.

The diabetes entry cohort represented patients returning the first health screen, indicating that they had been told by a health provider that they had diabetes. Included in this analysis are only those who completed baseline and 2-year diabetes questionnaires. Participant enrollment was from 1 February 1997 to 30 December 1999, and participants were followed for 2 years.

The firm was the unit of randomization and analysis (n = 14 firms, 7 pairs). The analysis compared demographic findings at baseline between the intervention and control firms. The primary analysis examined the change between 2-year and baseline scores (T2 − T0) by firm. We tested whether the change in intervention firms was significantly different from the change in control firms, using both paired and unpaired t tests, with the two firms within a clinic as a pair. Two-year comparisons were adjusted for age, sex, race, site, income, education, marital status, employment status, years of VA care, and baseline chronic conditions.

A total of 62,847 eligible subjects from seven VA General Internal Medicine Clinics were enrolled between 1 February 1997 and 30 December 1999 and were mailed a health screening questionnaire. Of this group, 34,103 were enrolled before 1 July 1997 and comprised the entry cohort. Of these subjects, 22,645 returned the health screening questionnaire. Diabetes was reported by 5,721 of the 22,645 patients (25.2%). Randomization by firm at seven sites resulted in 3,701 patients in the seven intervention firms and 2,020 patients in the control firms. Figure 2 shows the proportion of patients who responded to the diabetes survey at baseline and at 2 years for the intervention and control firms, together with reasons for patient loss to follow-up. Patient response was sufficient for analysis by firm; however, a majority of patients in both groups completed only one survey and were therefore not eligible for the final analysis. Loss to follow-up included patient withdrawal, moving, and unable to complete the study due to poor health or other reasons. The 2-year mortality in both groups was ∼12%.

At baseline, demographic and health history characteristics were similarly distributed between patients in the intervention and the control firms, as shown in Table 1. The average veteran’s age was 65.7 years. Approximately one-third of these veterans had received VA care for <5 years. One-third of participants received both VA and private health care. Only 16% of participants in these firms had <12 years of education. The majority of participants were married and retired. Approximately 41% reported that they were also disabled. The control firms had a significantly higher proportion of African Americans, Hispanics, and patients of other ethnic origins than did the intervention firm, and ∼25% of these veterans reported an annual income of <$10,000.

Health history findings identified that 67% of these participants had hypertension, 24% had a prior myocardial infarctions or coronary artery bypass surgery, and ∼15% had a history of stroke. The two health conditions with significant baseline differences between groups were a higher percentage of patients with a history of laser therapy in the control firm (22.1 vs. 15.9%, P < 0.003) and a higher percentage of intervention firm patients with congestive heart failure (12.6 vs. 8.7%, P = 0.02). Lower-extremity findings were similar and indicated prior foot sores in 13%, peripheral vascular surgery in 6%, and lower-limb amputations in 5%. A diagnosis of cancer was reported by 10% of participants, and 22% reported a history of depression. Smoking history was reported by 63%, although only 16% were current smokers.

Changes in physical and self-care findings are presented in Table 2. In terms of clinical improvement, the proportion of patients with an average HbA_1c ≤8% declined by 10% in both groups. The frequency of patients with a diastolic blood pressure <90 mmHg decreased in intervention firms but not in the control firms. Systolic blood pressure <140 mmHg and LDL >130 mg/dl actually increased in both groups. Approximately 8% of intervention firm patients improved their HDL levels over 2 years compared with baseline findings, while control firm patients stayed the same. Small improvements in patients checking feet and following meal plans were noted comparing patients in intervention and control firms, but they did not reach statistical significance.

Self-management awareness improved to a greater extent in intervention than in control firm patients, but only to a statistically significant extent on awareness for selecting shoes (Table 3). More intervention firm patients than control firm patients gained confidence that they knew which clinician to call (change = 0.08 vs. 0.01) and when to call with health care issues (change = 0.10 vs. 0.05). A majority of patients felt comfortable with their knowledge and awareness at baseline in terms of taking medications, meal planning, controlling blood glucose, and weight control.

Participants in the intervention firm and control firm were similarly satisfied with their provider’s explanations and diabetes management (Table 4). The score indicating satisfaction with providers’ interpersonal skills and organization score remained relatively constant in both firms.

Analysis of the SF-36 findings identified that the intervention firm scores showed minor improvement on six of the eight subscales and on the physical summary composite score. However, comparing intervention and control firms over 2 years showed no overall statistically or clinically significant differences (data not shown).

This study was designed to test whether providing primary care providers in intervention firms with timely patient information before visits would significantly improve outcomes, as compared between intervention firms and control firms. Despite the minimal clinician burden accompanying this 2-year intervention, patients in intervention firms fared no better with their diabetes outcomes than did those in control firms.

Several studies have addressed the likelihood of physicians’ implementing guidelines or recommendations. Streja and Rabkin (14) examined physician characteristics associated with implementation of preventive care measures in patients with diabetes. They found that the most important predictor for implementation of recommendations was a practice style reflecting lower numbers of patients seen per unit time. The study reported that most physicians believed they implemented preventive care measures at a >90% level, which in reality happened infrequently. These authors recommended computer reminders as a solution.

Findings from the Cochrane Effective Practice and Organization of Care Review Group (5) indicated that in studies of providers and organizations that also facilitated patient education and an expanded nursing role, there were improvements in patient care processes and outcomes. O’Connor et al. (15) also identified patient involvement as an important factor and further targeted those indicating a readiness to change in an intervention designed to improve diabetes processes and outcomes.

There were multiple steps needed to effect a change in patients’ diabetes outcomes. However, this intervention focused on provider, not patient actions. Diabetes control is strongly influenced by the quality of the partnership between the clinician and patient; thus, the lack of direct patient involvement in goal setting and developing strategies for goal achievement may have been a limitation in this study. Successful models for diabetes care usually include strategies that promote and maintain improved self-care behaviors. Had the same serial information been provided to patients, their awareness and involvement in the process of changing outcomes may have been more substantial.

Wagner (16) addressed the role of patient-care teams in chronic disease management and identified that interventions generally work better when delivered by a multidisciplinary team. Often nurses and pharmacists with clinical and behavioral skills can ensure delivery of critical elements of care that physicians may not have the training or time to do, including medication regulation, self-management support, and intensive follow-up.

Although we were not able to demonstrate it in our trial, quality of care in the VA has improved since 1995, but there is still room for further improvement. A recent article described the proportion of VA patients with diabetes who received quality of care as defined by three indicators: measuring HbA_1c, annual eye exams, and semiannual lipid screening (17). From FY1995 to FY2000, performance on these indicators improved significantly comparing VA patients with Medicare patients (P < 0.001).

A review by Balas (18) of 98 randomized clinical trials addressed clinical information systems largely in outpatient care. Provider prompts and reminders were related to a significant increase in process measures. However, when outcome measures were assessed, statistically significant findings were reported only for patient knowledge and attitude and limited physiologic and psychological measures.

There are several explanations for negative findings. The involved clinicians may have been unable to interpret the data and/or simply ignored it. This is possible given the many quality-improvement initiatives concurrently conducted in the VA during the study period. Providers may have been overwhelmed. It is also possible that more initial provider education and ongoing problem solving by firms delivering the intervention could have had a beneficial effect on patient outcome.

It is also possible that, while the provider acted on recommendations and feedback provided, patient nonadherence to provided recommendations influenced the outcome. Adherence was indirectly measured on the self-reported questionnaires. Many of the participating providers were residents and fellows rotating through the VA with small patient panels. Thus, their exposure to the intervention may have been too limited.

Another potential limitation is that this trial had inadequate statistical power to detect a true difference. We think that this is unlikely, even though by employing the clinic firm as the unit of analysis we had only 13° of freedom (6° for paired analyses). Despite the large number of comparisons made, they were largely negative, and chance alone could account for the few findings with P < 0.05.

In summary, this group-randomized trial providing intervention firms with a comprehensive and sustained program of collecting and reporting patient data did not improve outcomes in patients with diabetes. Effective multicomponent (combined goal setting, provider prompts, and feedback) interventions with more patient involvement in goal setting and feedback may be indicated. Providing patient feedback reports to both patients and providers could foster beneficial joint problem solving.

This research was supported by the Department of Veterans Affairs, Health Services Research and Development Service (SDR 96-002 and IIR 99-376).

The authors would like to recognize the efforts of the clinicians and patients who participated in the Ambulatory Care Quality Improvement Project (ACQUIP).

The views expressed are those of the authors and do not necessarily represent the views of the Department of Veteran Affairs.