Clinical Decisions Regarding HbA_1c Results in Primary Care: A report from CaReNet and HPRN

Practices in the Colorado Research Network (CaReNet) and the High Plains Research Network (HPRN) were eligible to participate in this study; provider participation was voluntary. CaReNet and HPRN are practice-based research networks—groups of clinical practices organized to efficiently conduct research studies (8–12). CaReNet primarily consists of community health centers, migrant health clinics, residency practices, and other urban practices. HPRN is a network of hospitals and clinics in rural and frontier communities of northeast Colorado. Combined, these two networks comprise a wide variety of practice settings throughout Colorado.

A brief provider survey was attached to every HbA_1c result for nonpregnant patients aged 18 years or older with type 2 diabetes. Data were collected at each site during a 3- to 6-week period between December 2001 and August 2002. We assumed each survey represented a unique patient.

We collected data on demographics and other patient characteristics, insurance, HbA_1c, and the clinical decision. Provider’s name was optional.

The options for the clinical decision on the HbA_1c result were “action taken” (change in medication or lifestyle), “no action taken,” or “decision deferred”. If no action was taken, the survey asked why. Deferred decision refers to postponing making or implementing the decision until the next clinical encounter. Several months later, we reviewed charts on over half of the deferred cases and reclassified most of these into the action and no action groups. The remaining 17 deferred cases (no decision could be determined or chart was unavailable for review) were excluded from the decision-making portion of the analysis.

To investigate whether medication costs are an important predictor of HbA_1c decision making, we compared Medicare only (lacks medication coverage) with all other insurance types combined (including patients with both Medicare and additional insurance that covers medications). HbA_1c values were classified into three categories based on ADA clinical guidelines: <7% (“no action needed”), 7–8% (“consider action”), and >8% (“recommend action”).

We used χ² tests and ANOVA to compare sociodemographic and clinical characteristics of patients by HbA_1c levels. Patient characteristics associated with action/no action were assessed using χ² tests and Mantel-Hantzel analysis, stratified by HbA_1c levels. Provider characteristics were not analyzed because of the high rate of missing information. Overall and strata-specific action rates are reported, along with crude and adjusted odds ratios.

We received 483 usable surveys from 19 clinics (16 family medicine, 1 internal medicine, and 1 geriatric). Eight sites were residencies, four were community health centers, and four were rural.

Eighty-eight providers identified themselves on 353 (72.9%) surveys; 84.1% were physicians, with the remainder mostly physician assistants and nurse practitioners. Of the physicians, 29.7% were residents.

Most patients were female (62.5%), and there was a large group of Hispanic patients (28.6%) (Table 1). Patients were generally taking at least one glycemic medication (83.5%) and had at least three other comorbidities (64.1%). Over one-third of the patients achieved the HbA_1c target (<7%).

As expected, the action rate was low (7%) for HbA_1c <7%. The action rate for the cases with an HbA_1c ≥7% and a known clinical decision (n = 294) was 70.7%; for HbA_1c >8% (n = 188), it was 89.9%. Patients who were black were less likely to have action taken (adjusted OR 0.18), and patients with Medicare only (no medication insurance) had a significantly lower rate of action (80.5 vs. 92.5%) on HbA_1c measurements >8% (Table 2). Age and ethnicity are significantly associated with action, but the relationship is no longer significant after controlling for HbA_1c. Sensitivity analyses, using models that accounted for clustering at the clinic level, yielded similar results.

For 80 patients with HbA_1c values ≥7% and no action taken, the provider gave at least one reason for no action. Over half the reasons were “improving/doing well”; “other reasons” was selected 11 times. Additional reasons listed at least twice were as follows (in order of most to least frequent):

• Competing demands

• Hypoglycemic risk

• Patient noncompliance

• Medication intolerance

• Medication costs

• Cognitive impairment

• Polypharmacy concerns

• Short life expectancy

• Provider unfamiliar with patient

• Transient event raised glucose.

In this study, the primary care rate of clinical intervention when glycemic control was suboptimal (71%) compares favorably with the rate (64%) reported from a diabetes practice (6), although different methodologies were used. This finding contrasts with other reports suggesting lower quality diabetes care in primary care compared with endocrinology (13–16). Additionally, this study is the first to describe primary care providers’ reasons for not intensifying glycemic treatment when the HbA_1c was above ADA-recommended targets.

We found disparities suggestive of less aggressive glycemic control interventions in black patients and in Medicare patients who lacked insurance for medications. Although many factors may contribute to poorer outcomes in black patients (17–23), this finding suggests that primary care providers may play a role in this disparity. An investigation of provider decisions in a population with more black patients may further clarify this disparity. Patient insurance status has been reported previously as a factor in providers’ clinical decisions (24–26). Medication coverage for Medicare beneficiaries may contribute to improved glycemic control.

More than half of the provider-reported reasons for inaction were that the patient was doing well or improving. The remaining reasons for inaction were related to a variety of clinical situations, which may reflect the complexity of diabetes management in primary care, or they may be “soft reasons” for inaction, indicative of “clinical inertia” (27). Future studies could further clarify the meaning of these reasons.

There are several limitations to this study. It is possible completing a survey at the time of clinical decision may affect the decision (Hawthorne effect) (28). In another study that examined provider glycemic-control decisions at the point of care, the rate of action increased modestly from 55% before the survey to 64% during the survey (6). In a CaReNet study that retrospectively examined HbA_1c decisions, the rate of action on HbA_1c values >8% (78%, CaReNet, unpublished data) was somewhat lower than the rate we found here (89.9%). Thus, although there is likely a modest Hawthorne effect, it probably does not substantially change the findings in the present study.

There may also be bias due to differential completion of surveys by providers. While not required, six clinics voluntarily tracked survey completion rates and found high rates (mean 97%).

There are likely other influences on the HbA_1c action rate that were not investigated in this study (e.g., provider characteristics, specific comorbidities, patient income, or the duration and severity of diabetes). In addition, there may be confounding effects that were not addressed in this analysis. The two practice-based research networks in this study consist of multiple practices and providers in a variety of urban, rural, academic, and community settings. There was a relatively high proportion of physicians-in-training and underserved patients. Our results may not generalize to other practice settings with different patient populations.

Our findings suggest that primary care providers are generally aggressive with glycemic management in patients with type 2 diabetes, comparable to providers in diabetes specialty practices. When primary care providers are not aggressive, their reasons for inaction reflect a wide variety of patient situations.

This study was supported by the American Academy of Family Physicians Foundation.

We thank the participating clinics that made this study possible: A.F. Williams Family Medicine Center; Brighton Salud Family Health Center; Colorado Springs Family Medicine; Exempla St. Joseph Family Practice; Family Medicine of Littleton; Ft. Morgan Salud Family Health Center; LaCasa-Quigg Newton Health Center; Lowry Family Health Center; Metropolitan Community Providers Network Littleton; Plains Medical Center; Rose Family Medicine Center; St. Mary’s Family Practice Center; Southern Colorado Family Practice; Swedish Family Medicine Center; University Family Medicine Park Meadows; University Family Medicine Westminster; University of Colorado Internal Medicine-AOP; University of Colorado Seniors Clinic; and Yuma Clinic.

We also thank Elizabeth Staton for assistance with the manuscript.