Lifestyle Counseling and Long-term Clinical Outcomes in Patients With Diabetes

The risk of cardiovascular events is increased by 60% in patients with diabetes (1). Elevated blood glucose levels in patients with diabetes are associated with higher cardiovascular risk (2). Pharmacological reduction in blood glucose levels decreases this risk (3,4). Intensive lifestyle interventions are also effective at lowering blood glucose (5–8). However, there remains uncertainty about whether lifestyle interventions can reduce cardiovascular risk or mortality among patients with diabetes.

It was previously shown that documentation of counseling by health care providers in routine care can be computationally identified in electronic medical record (EMR) notes using natural language processing (9,10). Subsequent research demonstrated that lifestyle counseling identified in this way is associated with improvement of blood glucose control (11,12). We therefore conducted this study to determine whether lifestyle counseling in routine care is associated with lower cardiovascular risk and mortality in patients with diabetes.

Study participants included adults (≥18 years old) with uncontrolled (HbA_1c ≥7.0% [53 mmol/mol]) diabetes who were followed in primary care clinics affiliated with two academic medical centers between 2000 and 2014. The study design included three stages: a run-in period, a treatment assessment period, and an outcome ascertainment period. The run-in period began 12 months before cohort entry. The 2-year treatment assessment period started on the date of cohort entry. The outcome ascertainment period began at the end of the treatment assessment period and lasted until study exit. This study was approved by the institutional review board at the Partners HealthCare System (Boston, MA) with a waiver of informed consent.

Patients’ baseline characteristics were assessed at the end of the run-in period. The treatment assessment period was used to calculate the predictor variable (frequency of lifestyle counseling) and other treatment characteristics. Lifestyle counseling frequency was measured during the hyperglycemic periods (while HbA_1c ≥7.0% [53 mmol/mol] [13]) of the treatment assessment period.

Frequency of lifestyle counseling was calculated as the mean monthly number of primary care clinic notes during the hyperglycemic periods in which diet, exercise, or weight loss counseling were documented (11). Documentation of lifestyle counseling was identified in EMR provider notes using a previously validated natural language processing algorithm (9). Time-weighted average HbA_1c was calculated using the trapezoid rule.

The composite primary outcome was the time to the first of the following: a cardiovascular event (myocardial infarction, stroke, hospitalization for angina) or death as a result of any cause. Time to death and time to the first cardiovascular event served as secondary outcomes.

Cox proportional hazards models were used to estimate the association between lifestyle counseling frequency and patient outcomes while accounting for clustering within providers. The models were adjusted for patient demographics, comorbidities, and nonlifestyle management of diabetes during the treatment assessment period. We conducted a secondary analysis to determine whether the association between lifestyle counseling frequency and outcomes was mediated by changes in HbA_1c by adding time-weighted HbA_1c during the treatment assessment period as a covariate.

Median baseline HbA_1c of 19,293 study patients was 7.8% (62 mmol/mol), and median frequency of lifestyle counseling during hyperglycemic periods was 0.46/month. Patients were followed for a mean of 5.4 years after the end of the treatment assessment period.

The majority of patients (16,057) received lifestyle counseling less than monthly. During the 24-month treatment assessment period, HbA_1c decreased by 1.8% for patients with monthly or more versus 0.7% for less than monthly lifestyle counseling (P < 0.0001). Among patients with monthly or more lifestyle counseling (Fig. 1), the 10-year cumulative incidence rate of the primary outcome was 33.0% compared with 38.1% for less than monthly counseling (P = 0.0005).

In Cox multivariable analysis, frequency of lifestyle counseling (once per month vs. once per 3 months) was associated with hazard ratios for composite primary outcome, death, and cardiovascular events of 0.88 (95% CI 0.82–0.94; P < 0.001), 0.94 (0.86–1.01; P = 0.101), and 0.89 (0.82–0.97; P = 0.006), respectively. When time-weighted HbA_1c during the treatment assessment period was added to the model, the relationship between lifestyle counseling frequency and the primary outcome was no longer statistically significant.

In this large, real-world study, patients with diabetes who received more frequent lifestyle counseling from primary care providers were less likely to experience cardiovascular events or death. This relationship persisted after adjustment for other processes of diabetes care, including patient-provider encounters and intensification of diabetes medications. This finding was made possible by advances in both EMRs and technologies used to process EMR big data, including natural language processing.

As an observational analysis, this investigation does not provide direct evidence for a causal relationship and can only establish an association. However, mediation of the lifestyle counseling’s association with clinical outcomes by changes in HbA_1c is consistent with our understanding of the pathophysiology of complications of diabetes and provides indirect support for causality (14).

A previously published large randomized clinical trial found that a lifestyle intervention did not reduce the incidence of cardiovascular events in patients with diabetes (15). However, there were several differences between the Look AHEAD (Action for Health in Diabetes) trial and our study that could explain this discrepancy. Participants in the Look AHEAD trial did not achieve a sustained blood glucose reduction; by year 2 of the 10-year intervention, the difference in HbA_1c between the two comparison groups narrowed to 0.3% and by year 4, to <0.2%. In contrast, in our study, the difference in HbA_1c between patients who received lifestyle counseling monthly or more versus less than monthly reached nearly 1%. Furthermore, this study included nearly four times as many patients as Look AHEAD, increasing its power to detect the effect of lifestyle counseling.

The current study found that to be effective, lifestyle counseling had to be intensive: outcome benefits primarily accrued to patients who received lifestyle counseling monthly or more. This is consistent with our previous findings that intensive counseling is needed to achieve a significant improvement in glycemic control (11). This frequency of lifestyle counseling may not be universally feasible in routine care settings and may need to require different care models (e.g., group sessions, telemedicine) or be limited to selected individuals to be practical.

The results of this study should be interpreted in light of several limitations. We were unable to reliably distinguish between patients with type 1 and type 2 diabetes. The majority of patients in our study likely had type 2 diabetes; therefore, our findings may not be applicable to individuals with type 1 diabetes. Some episodes of lifestyle counseling may not have been reflected in providers’ notes, while others may not have been picked up by the natural language processing software. If these omissions were distributed nonrandomly with respect to study outcomes, the findings may have been biased. Finally, because the study was limited to practices affiliated with academic medical centers in eastern Massachusetts, its findings may not be generalizable to other settings.

In summary, this study is the first to our knowledge to show that more frequent lifestyle counseling is associated with a decreased incidence of cardiovascular events and death in patients with uncontrolled diabetes. Further research is needed to confirm the causal nature of this relationship and to establish characteristics of lifestyle counseling that might have particularly strong effects on these important outcomes.

Funding. This study was supported in part by the Patient-Centered Outcomes Research Institute (PCORI CE1304-6756), National Natural Science Foundation of China (91846106, 71432004), nonprofit Central Research Institute Fund of Chinese Academy of Medical Sciences (2017PT32020, 2018PT32001), Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (CIFMS2016-I2M-4-001), and Training Program for Excellent Talents in Dongcheng District.

The funding sources had no direct impact on the design and conduct of the study; the collection, management, analysis, and interpretation of the data; and the preparation, review, or approval of the manuscript.

Duality of Interest. A.T. declares research funding from Sanofi, Eli Lilly, and Novo Nordisk; equity in Brio Systems; and personal fees from Monarch Medical Technologies. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. H.Z., S.I.G., N.H., M.S., D.C.S., M.A.T., and A.T. critically revised the manuscript for important intellectual content. H.Z., S.I.G., N.H., M.S., and A.T. analyzed and interpreted the data. H.Z., M.S., and A.T. contributed to the study concept and design and performed the statistical analysis. H.Z. and M.A.T. obtained funding. H.Z. and A.T. drafted the manuscript. A.T. acquired data and supervised the study. A.T. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.