Figure 1. Counts of patients excluded from the analysis. BP indicates blood pressure; BWH, Brigham and Women's Hospital; DM, diabetes mellitus; LDL-C, low-density lipoprotein cholesterol; MGH, Massachusetts General Hospital; and PCP, primary care physician.
Figure 2. Kaplan-Meier curves for time to treatment target from first elevated hemoglobin A1c, blood pressure (BP), or low-density lipoprotein cholesterol (LDL-C) values are plotted for different mean encounter intervals. Distinct uncontrolled periods (from the first elevated to the first normal measurement) for the same patient were analyzed separately. A, Encounter frequency and time to hemoglobin A1c level target for patients not receiving insulin. B, Encounter frequency and time to hemoglobin A1c target for patients receiving insulin. C, Encounter frequency and time to BP target. D, Encounter frequency and time to low-density lipoprotein cholesterol (LDL-C) target. (To convert LDL-C to millimoles per liter, multiply by 0.0259.) E, Encounter frequency and time to combined target. DBP, diastolic BP; and SBP, systolic BP.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Morrison F, Shubina M, Turchin A. Encounter Frequency and Serum Glucose Level, Blood Pressure, and Cholesterol Level Control in Patients With Diabetes Mellitus. Arch Intern Med. 2011;171(17):1542–1550. doi:10.1001/archinternmed.2011.400
Background More frequent patient-provider encounters may lead to faster control of hemoglobin A1c level, blood pressure (BP), and low-density lipoprotein (LDL) cholesterol (LDL-C) level (hereafter referred to as hemoglobin A1c, BP, and LDL-C) and improve outcomes, but no guidelines exist for how frequently patients with diabetes mellitus (DM) should be seen.
Methods This retrospective cohort study analyzed 26 496 patients with diabetes and elevated hemoglobin A1c, BP, and/or LDL-C treated by primary care physicians at 2 teaching hospitals between January 1, 2000, and January 1, 2009. The relationship between provider encounter (defined as a note in the medical record) frequency and time to hemoglobin A1c, BP, and LDL-C control was assessed.
Results Comparing patients who had encounters with their physicians between 1 to 2 weeks vs 3 to 6 months, median time to hemoglobin A1c less than 7.0% was 4.4 vs 24.9 months (not receiving insulin) and 10.1 vs 52.8 months (receiving insulin); median time to BP lower than 130/85 mm Hg was 1.3 vs 13.9 months; and median time to LDL-C less than 100 mg/dL was 5.1 vs 32.8 months, respectively (P < .001 for all). In multivariable analysis, doubling the time between physician encounters led to an increase in median time to hemoglobin A1c (not receiving [35%] and receiving [17%] insulin), BP (87%), and LDL-C (27%) targets (P < .001 for all). Time to control decreased progressively as encounter frequency increased up to once every 2 weeks for most targets, consistent with the pharmacodynamics of the respective medication classes.
Conclusions Primary care provider encounters every 2 weeks are associated with fastest achievement of hemoglobin A1c, BP, and LDL-C targets for patients with diabetes mellitus.
Diabetes mellitus (DM) is increasingly common in the United States and worldwide.1,2 Elevated serum glucose, blood pressure (BP), and low-density lipoprotein cholesterol (LDL-C) (hereafter referred to as hemoglobin A1c, BP, and LDL-C) values are associated with increased risk of microvascular and macrovascular complications, and their reduction decreases the risk.3-8 Nevertheless, most patients with DM do not have their hemoglobin A1c, BP, and LDL-C under control.9,10
A variety of studies11-13 have shown that patients who visit their physicians more frequently have better outcomes. Current guidelines for treatment of DM do not include recommendations for how frequently patients should be observed.14 Recommended intervals for medication adjustments and testing range from every 2 to 3 days (for insulin) to every 3 months (for measuring hemoglobin A1c)14-16; however, benefits of more frequent provider encounters may not be limited to treatment intensification and testing.
We, therefore, performed a retrospective study of more than 26 000 patients with DM and hyperglycemia, hypertension, and/or hyperlipidemia who received care in a primary care setting to test the hypothesis that higher encounter frequency is associated with better DM control.
We conducted a retrospective cohort study to determine the optimal frequency of patient-provider encounters for patients with DM. We evaluated the relationship between mean encounter frequency and time to hemoglobin A1c, BP, and LDL-C control. We also conducted a secondary analysis to examine the relationship between encounter frequency and the rate of decrease in hemoglobin A1c, BP, and LDL-C values.
Patients with DM seen by primary care physicians affiliated with Brigham and Women's Hospital (BWH) and Massachusetts General Hospital (MGH) (both in Boston) for at least 2 years between January 1, 2000, and January 1, 2009, were studied. Patients were included in the analysis if they were 18 years and older, had a documented diagnosis of DM or a hemoglobin A1c of at least 7.0%, and at least 1 instance of hemoglobin A1c, BP, or LDL-C higher than the treatment target. Patients with missing zip codes were excluded to enable adjustment for median income by zip code.
To capture both face-to-face and remote interactions between patients and providers, we defined any note in the electronic medical record (EMR) as an encounter. We used treatment goals recommended at the beginning of the study: hemoglobin A1c less than 7.0%,17 BP lower than 130/85 mm Hg,17,18 and LDL-C less than 100 mg/dL (to convert to millimoles per liter, multiply by 0.0259).17 This study was approved by the Partners HealthCare System institutional review board, and the requirement for written informed consent was waived.
A single uncontrolled period served as the unit of analysis. We conducted 4 analyses: 1 for each of the 3 treatment targets (hemoglobin A1c, BP, and LDL-C) and a combined analysis that integrated all 3 targets. For analyses of individual treatment targets, an uncontrolled period started on the day when the relevant measurement (hemoglobin A1c, BP, or LDL-C for the hyperglycemic, hypertensive, and hyperlipidemic periods, respectively) was noted to be above the treatment target for the first time. The period ended on the first subsequent date when the measurement was below the target. Each patient could contribute multiple periods if measures fluctuated above and below target levels during the 9-year study. A combined uncontrolled period started on the first date when any of the 3 measures was above the treatment target and ended on the first subsequent date when all the measures were below their targets. The last known value was carried forward if all the measurements were not available on the same date.
The lowest measurement on a given date was used in the analysis. Lowest BP was defined as the BP measurement with the lowest mean arterial pressure. Transient elevations were defined as periods that contained only a single elevated measurement that subsequently normalized without any treatment intensification, and were excluded from the analysis. Uncontrolled periods without at least 1 annual encounter with a BWH or MGH primary care physician were excluded from the analysis to eliminate patients not actively treated in these practices. Periods without any medication information available in the EMR were excluded to enable inclusion of insulin treatment as a confounder variable in the analysis. Periods that contained more than 1 encounter with an endocrinologist were excluded to focus the analysis on the primary care setting. Finally, hyperglycemic and hyperlipidemic periods in which the rate of change of hemoglobin A1c and LDL-C, respectively, was greater than 3 SDs from the mean were excluded to eliminate likely measurement errors from the analysis.
Time to normalization for hemoglobin A1c, BP, and LDL-C during their respective uncontrolled periods was the length of the uncontrolled period. The mean encounter interval was determined by dividing the period length by the number of encounters with primary care physicians during that period. In these analyses, we categorized encounter intervals as 1 week or less, greater than 1 week to 2 weeks or less, greater than 2 weeks to 3 weeks or less, greater than 3 weeks to 1 month or less, greater than 1 month to 2 months or less, greater than 2 months to 3 months or less, and greater than 3 months. Treatment intensification was defined as initiation of a new medication or an increase in the dose of an existing medication.19 The treatment intensification rate was defined as the number of unique dates per month on which at least 1 medication in the relevant class was intensified. Medication change was conservatively classified as intensification as previously described20 because there is no reliable method to estimate relative medication potency for individual patients. Drug cessations were not captured in this analysis. Mean rate of change for hemoglobin A1c, systolic BP (SBP), diastolic BP (DBP), and LDL-C was calculated by subtracting the final value from the initial value and dividing by the period length. The patient's primary care physician was defined as the physician in a primary care practice who had the most encounters with the patient during the uncontrolled period.
Demographic information, BP measurements, and medication and laboratory data were obtained from the EMR at Partners HealthCare, an integrated health care delivery network in eastern Massachusetts that includes BWH and MGH. The Partners HealthCare EMR contains all medication prescription and laboratory records starting in at least 2000, and earlier for many patients. Blood pressure was obtained from a combination of structured vital sign records in the EMR and computational processing of narrative electronic provider notes as previously described.21 Medication intensification was abstracted from a combination of structured medication records and computational analysis of electronic provider notes as previously validated.22
Summary statistics were conducted by using frequencies and proportions for categorical data and using means (SDs), medians, and ranges for continuous variables. The log-rank test was used to compare times to hemoglobin A1c, BP, and LDL-C normalization between different encounter intervals.
Marginal Cox proportional hazards regression models for clustered data23 were constructed to estimate the association between time to normalization and encounter interval while accounting for repeated events in individual patients and adjusting for demographic confounders (age, sex, race/ethnicity, primary language, health insurance, and median income by zip code), patient Charlson Comorbidity Index, treatment intensification, hemoglobin A1c and LDL-C measurement rates, and maximum hemoglobin A1c, SBP, DBP, and LDL-C values (where appropriate). To more clearly present the findings as a direct effect of encounter interval on time to normalization, we reanalyzed the data using Weibull regression models. We confirmed the equivalence of the Weibull regression models and the marginal Cox proportional hazards regression models by comparing Cox-Snell residual plots between these models using paired t tests and graphically using Nelson-Aalen plots.
To rule out ascertainment bias stemming from increased hemoglobin A1c, BP, and LDL-C measurement opportunities for patients with more frequent encounters, a sensitivity analysis was conducted at the patient level to compare the probability of target achievement at the end of 2 years from the first elevated hemoglobin A1c, BP, or LDL-C measurement with the frequency of patient-provider encounters. The logistic regression model adjusted for demographic confounders; Charlson Comorbidity Index; treatment intensification rates; maximum hemoglobin A1c, BP, and/or LDL-C measures; rates of hemoglobin A1c and LDL-C measurements, where appropriate; and clustering within providers.
To determine the relationship between encounter interval and rate of hemoglobin A1c, BP, and LDL-C change, we constructed hierarchical multivariable mixed linear regression models with random intercepts to account for clustering within individual physicians and repeated measurements with compound symmetry structure within patients.24 The models also included patient age, sex, race/ethnicity, primary language, income, health insurance, treatment intensification rate during the uncontrolled period, and insulin use for the hyperglycemic and combined periods. P values were obtained using the type III test and were adjusted for multiple hypothesis testing using the Simes-Hochberg method.25,26 All the analyses were performed using commercially available statistical software (SAS, version 9.2; SAS Institute Inc, Cary, North Carolina).
We identified 32 482 adults with DM who were regularly seen by BWH or MGH primary care physicians and had experienced at least 1 hyperglycemic, hypertensive, or hyperlipidemic period (Figure 1). We excluded 5655 hyperglycemic, 5181 hypertensive, and 5406 hyperlipidemic patients who were treated by endocrinologists; had no medication records; had only transient elevations in hemoglobin A1c, BP, and LDL-C values; had suspected hemoglobin A1c or LDL-C measurement errors; had missing demographic information; or were not regularly seen by a primary care physician associated with BWH or MGH during the study. The remaining 14 293 hyperglycemic, 26 128 hypertensive, and 15 739 hyperlipidemic patients were included in the study.
During the study, only median DBP was below the treatment target (75 mm Hg); median hemoglobin A1c was 7.4%, SBP 130 mm Hg, and LDL-C 106.7 mg/dL (Table 1). The mean number of uncontrolled periods per patient during the study ranged from 1.4 for hyperlipidemia to 3.4 for hypertension. Hyperglycemic patients had hemoglobin A1c values above target 46% of the time, hypertensive patients had uncontrolled BP 42.7% of the time, and hyperlipidemic patients had elevated LDL-C values 46.3% of the time. At least 1 of the study measurements was not under control 88.4% of the time.
Median time between encounters ranged from 1.1 months for hypertensive periods to 1.8 months for hyperlipidemic periods (Table 2). The mean rate of antihyperglycemic medication intensification was approximately once per year, antihypertensive medications once every 4 months, and antihyperlipidemic medications once every 17 months. Overall, patients with at least 1 measurement above target had their treatment intensified on average once every 2.8 months.
In all the treatment categories, time to treatment target rose progressively as the interval between encounters increased (Figure 2). Compared with patients with a mean encounter interval of 1 to 2 weeks, median time to hemoglobin A1c target for patients whose mean encounter interval was 3 to 6 months was 4.4 vs 24.9 months (those not receiving insulin) and 10.1 vs 52.8 months (those receiving insulin), time to BP target was 1.3 vs 13.9 months, and time to LDL-C target was 5.1 vs 32.8 months. For all treatment targets combined, median time to target was 1.5 vs 36.9 months for a mean encounter interval of 1 to 2 weeks vs 3 to 6 months.
As encounter intervals increased, the proportion of patients who never reached treatment targets also rose steadily. Comparing patients with a mean encounter interval of 1 to 2 weeks with those with an interval of more than 6 months, the fraction of uncontrolled periods that never reached treatment target was 35.4% vs 55.6% for hyperglycemic patients treated with insulin and 5.4% vs 15.9% for hypertensive patients. For hyperglycemic patients not receiving insulin and hyperlipidemic patients, the lowest proportion of uncontrolled periods that did not achieve treatment target was for encounter intervals of 1 to 2 weeks: 14.8% and 16.8% compared with 36.8% and 31.9% for encounter intervals longer than 6 months. For all treatment targets combined, the proportion of uncontrolled periods that never achieved all targets was 11.0% for encounter intervals of 1 week or less vs 43.4% for encounter intervals greater than 6 months.
In the multivariable Weibull regression model (Table 3) adjusted for demographic characteristics; Charlson Comorbidity Index; insulin administration (in hyperglycemic and combined uncontrolled periods); maximum hemoglobin A1c, SBP, DBP, and LDL-C (where relevant); hemoglobin A1c and LDL-C testing rates (where relevant); and treatment intensification, doubling the time between physician encounters resulted in a 35% (those not receiving insulin) and 17% (those receiving insulin) increase in median time to hemoglobin A1c normalization, an 87% increase in median time to BP normalization, and a 27% increase in median time to LDL-C normalization (P < .001 for all). Higher rates of treatment intensification; lower hemoglobin A1c, BP, and LDL-C; and not being treated with insulin (for hyperglycemic patients) were also associated with shorter periods (P < .001 for all). In a Weibull regression model of combined uncontrolled periods, doubling the time between physician encounters led to an 84% increase in the time to achievement of all treatment targets (P < .001). When treatment intensification was excluded from the model, doubling the time between physician encounters translated into 38%, 20%, 90%, 32%, and 88% increases in time to hemoglobin A1c when not receiving and receiving insulin, BP, LDL-C, and combined control, respectively (P < .001 for all). In a post hoc multivariable sensitivity analysis including periods for patients treated by endocrinologists, encounter frequency had similar effects on time to hemoglobin A1c, BP, and LDL-C normalization (results not shown).
Multivariable sequential comparison of time to treatment target for encounter interval categories adjusted for patient demographics; highest hemoglobin A1c, BP, and LDL-C (where relevant) during the uncontrolled period; rate of treatment intensification; and insulin treatment (for hyperglycemic patients) showed that differences between most consequent encounter interval categories were highly significant (P < .001). Exceptions included encounter intervals of 1 week or less vs 1 to 2 weeks for hyperglycemic patients not treated with insulin (P = .006) and hyperlipidemic patients (P = .90) and encounter intervals of 1 to 2 weeks vs 2 to 3 weeks (P = .13) and 2 to 3 weeks vs longer than 3 months (P = .68) for hyperglycemic patients treated with insulin.
In multivariable analysis adjusted for demographic characteristics; Charlson Comorbidity Index; insulin treatment (in hyperglycemic patients); highest hemoglobin A1c, BP, and LDL-C (where relevant) values during the uncontrolled period; rate of treatment intensification and hemoglobin A1c and LDL-C measurement (where relevant); and clustering within individual physicians and repeated measurements within patients, for every additional month between encounters, rate of hemoglobin A1c declined an additional 0.014% per month, rate of SBP decreased by 2.5 mm Hg per month, rate of DBP decreased by 1.0 mm Hg per month, and rate of LDL-C decreased by 0.28 mg/dL per month (P < .001 for all). More frequent treatment intensification led to faster rates of decrease for all diabetes measures (P < .001 for all).
In this large retrospective study, we found a strong association between encounter frequency and hemoglobin A1c, BP, and LDL-C control in patients with DM. This relationship was confirmed in individual and combined analyses of time to normalization, rate of measure decrease, and rate of target achievement. A strong dose-response relationship between encounter frequency and the outcomes was evident in all the associations we analyzed.
Current guidelines provide little guidance for how frequently patients with DM should be seen by their physicians, apart from the recommendation for hemoglobin A1c measurement every 3 months.14 The present findings provide evidence that for many patients with elevated hemoglobin A1c, BP, or LDL-C, more frequent patient-provider encounters were associated with a shorter time to treatment target, and control was fastest at 2-week intervals. Encounters every 2 weeks may, therefore, be appropriate for the most severely uncontrolled patients or under a different treatment care model.
More frequent opportunities for medication intensification are likely an important mediator of the encounter frequency effect. This explanation is corroborated by a decrease in the encounter frequency effect when treatment intensification rate is included in the model. Many textbooks27,28 recommend a lower limit of 4 to 6 weeks on the medication intensification frequency out of concern for a stacking effect and overdose. However, time to maximum effect for most medications is shorter than commonly believed. The majority of antihyperglycemic agents achieve most of their effect within 2 weeks,29-32 and others in less than 4 weeks33-36; antihypertensive agents (except thiazides) in less than 2 weeks37-42; and statins within 2 weeks.43 These results are consistent with the present findings that biweekly encounters are associated with fastest achievement of serum glucose, BP, and LDL control.
Although median time between patient-physician encounters was only 1.4 months for hyperglycemic patients, treatment intensification occurred just once per year. The target hemoglobin A1c is commonly reached much more slowly than recommended by guidelines; the incongruity between encounter frequency and rates of treatment intensification suggests that there are many opportunities for physicians to alter medications that may lead to faster hemoglobin A1c control during encounters.
Treatment intensification may not be the sole factor responsible for the association between encounter frequency and patient outcomes, as illustrated by the strong residual association between encounter frequency and time to normalization when controlling for treatment intensification. Other studies44,45 have shown that more frequent encounters are also associated with better medication adherence. During encounters, physicians may also be providing lifestyle coaching or other education that leads to better DM control.
There is evidence that faster control of intermediate end points (hemoglobin A1c level, BP, and LDL-C) that could be achieved by more frequent provider encounters translates into improvement in clinical outcomes. Early intensive insulin therapy in patients with newly diagnosed type 2 DM leads to more durable control and improvement in β-cell function.46 The VALUE (Valsartan Antihypertensive Long-term Use Evaluation) randomised trial found that lower BP in the first 3 months decreased rates of stroke and myocardial infarction.47 Several studies48-52 have shown that statin therapy lowered rates of cardiovascular events in high-risk patients within 3 to 6 months of initiation.
Because more frequent encounters could increase demand on health care resources, straining an already taxed53 and dwindling primary care environment,54,55 increased encounter frequency implementation may require innovative approaches to patient care delivery. Medical homes may help coordinate care of patients, and some interactions could be accomplished through group visits, telephone, fax, e-mail, or Internet communications.56 Studies56-60 have shown that midlevel providers can alleviate physician workload without any negative effect on patient outcomes.
Once a patient achieves DM control, the frequency of the encounters may be decreased to alleviate the strain on health care resources and possibly to also reward the patient.61 It has been shown that in patients with controlled hypertension, patient-provider encounters can be 6 months apart without adverse effects.62
The present study has several strengths. With access to EMRs from 2 large hospital systems, we were able to analyze more than 26 000 patients with uncontrolled DM from diverse backgrounds and health insurance coverage plans. We focused on the primary care setting, where most patients with DM are treated. Importantly, the present results were consistent with pharmacodynamic data, providing a physiologic basis for the findings.
This study also has several limitations. It was conducted in clinics affiliated with 2 academic medical centers in eastern Massachusetts and, thus, may not be generalizable to all settings. These clinics do not include many midlevel providers, primarily limiting these conclusions to primary care physicians. Uncontrolled periods were censored at the beginning of the study; however, unless encounter frequency was systematically uneven over the duration of the study, this should not have biased the results. We were unable to distinguish between routine scheduled encounters and last-minute appointments with physicians; the focus of care (routine vs urgent) probably differs between these 2 visit types. Hemoglobin A1c, BP, and LDL-C values were measured only during the course of routine care, possibly leading to an ascertainment bias because patients with shorter encounter intervals had more frequent opportunities to have measurements below target. However, a separate analysis showed that higher encounter frequency was linked to higher probability of hemoglobin A1c, BP, and LDL-C target achievement at 2 years after the first abnormal level was measured (data not shown). This finding supports our interpretation in a manner not subject to bias by the missing measurement data. The retrospective nature of these data does not allow us to assess the availability or motivation of patients to see their physicians, which may be another indicator of adherence. We were also unable to consider how individual patient-provider goals may have differed from published guidelines or which physicians may practice in clinics that institute DM management protocols; however, we did correct for clustering within providers and repeated measures within patients, which helps mitigate this confounder. There were several potential confounders we could not measure, including type of DM, face-to-face vs remote encounters, focus of treatment at an encounter, patient motivation, and medication adherence. We were also unable to measure potential costs and risks associated with higher encounter frequency, making a full risk-benefit analysis impossible. Although some clinical trials found evidence that faster attainment of intermediate measures can result in improved clinical outcomes, this study is limited to intermediate outcomes, and we do not have evidence that the association between higher encounter frequency and faster hemoglobin A1c, BP, and LDL-C control reported herein leads to improved clinical outcomes in this study population. The retrospective nature of this study prevents us from establishing a causal relationship between encounter frequency and patient outcomes. A randomized interventional study is, therefore, needed to definitively establish optimal encounter frequency for patients with DM.
Correspondence: Alexander Turchin, MD, MS, Division of Endocrinology, Brigham and Women's Hospital, 221 Longwood Ave, Boston, MA 02115 (firstname.lastname@example.org).
Accepted for Publication: May 23, 2011.
Author Contributions:Study concept and design: Shubina and Turchin. Acquisition of data: Turchin. Analysis and interpretation of data: Morrison, Shubina, and Turchin. Drafting of the manuscript: Morrison. Critical revision of the manuscript for important intellectual content: Shubina and Turchin. Statistical analysis: Shubina. Obtained funding: Turchin. Administrative, technical, and material support: Morrison. Study supervision: Turchin.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grants 5R18HS017030 from the Agency for Healthcare Research and Quality (Drs Shubina and Turchin) and 5RC1LM010460 from the National Library of Medicine (Ms Morrison and Drs Shubina and Turchin), and the Diabetes Action Research and Education Foundation (Dr Turchin).
Role of the Sponsor: 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.
Previous Presentation: This study was presented in part at the 71st Scientific Sessions of the American Diabetes Association; June 26, 2011; San Diego, California.