Palo Alto Medical Foundation Research Institute, Ames Building, 795 El Camino Real, Palo Alto,A 94301, United StatesPalo Alto Medical Foundation Research Institute, Palo Alto, CA, United StatesUniversity of Chicago, Department of Health Studies, Chicago, IL, United StatesPalo Alto Medical Foundation, Endocrinology Department, Palo Alto, CA, United StatesStanford University, Prevention Research Center, Palo Alto, CA, United StatesPalo Alto Medical Foundation Research Institute and Stanford University, Prevention Research Center, Palo Alto,A, United States
r t i c l e i n f o
rticle history:
eceived 2 December 2013
eceived in revised form
April 2014
ccepted 7 April 2014
vailable online xxx
eywords:
iabetes treatment
ncident diabetes
mbulatory care
a b s t r a c t
Objective: We examined patterns and predictors of initiation of treatment for incident dia-
betes in an ambulatory care setting in the US.
Methods: Data were extracted from electronic health records (EHR) for active patients ≥35
years in a multispecialty, multiclinic ambulatory care organization with 1000 providers. New
onset type 2 diabetes and subsequent treatment were identified using lab, diagnosis, medi-
cation prescription, and service use data. Time from the first evidence of diabetes until initial
treatment, either medication or education/counseling, was examined using a Kaplan–Meier
hazards curve. Potential predictors of initial treatment were examined using multinomial
logistic models accounting for physician random effects.
Results: Of 2258 patients with incident diabetes, 55% received either medication or edu-
cation/counseling (20% received both) during the first year. Of the treated patients, 68%
received a treatment within the first four weeks, and 13% after initial 16 weeks. Strong pos-
itive predictors (P < 0.01) of combined treatment were younger age, higher fasting glucose at
diagnosis, obesity, and visits with an endocrinologist.
Conclusions: Among insured patients who have a primary care provider in a multispecialty
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
health care system, incident diabetes is treated only half the time. Improved algorithms
for identifying incident diabetes from the EHR and team approach for monitoring may help
Guidelines for type 2 diabetes recommend early initiationof therapy, with lifestyle counseling and medication offeredconcomitantly [1,2]. Studies have shown the importance ofearly identification and treatment of diabetes for preventinglong-term complications [3]. Even in the pre-diabetic stage,initiation of lifestyle changes and medications are beneficial[4–8]. Existing studies suggest an epidemic of under-diagnosisand under-treatment of diabetes and its complications in theUS [9–12].
Little is known about how incident diabetes is treated inambulatory care practices. It is unknown, for example, howlong it takes to receive treatment after the identification ofdiabetes, and patient and practice factors contributing to thetiming of treatment initiation. Addressing such questionsrequires observation of real-world practices serving patientswith varying demographics and clinical conditions, and lon-gitudinally linked data on patient clinical history, utilizationpattern and provider practice style [13,14].
We examined initial treatment of incident diabetes, uti-lizing data from electronic health records (EHRs), reflectingclinical practice in typical ambulatory care setting. We firstdescribe timing and initial treatment choices for patients whoare newly identified as having diabetes. We then evaluatedemographic and clinical risk factors and service use patternthat are associated with initial treatment of diabetes.
2. Methods
2.1. Study setting and data source
The study used demographic and clinical data, includ-ing anthropometric measures, physician diagnoses, labora-tory results and prescription medications, extracted fromEpicCare® (Epic Systems, Verona WI) EHRs for the patients at alarge multi-specialty, mixed-payer, outpatient, group practicewith approximately 1000 physicians in northern California.The EHR system has been in use since 2000 across all theclinics and providers included in the study. The demographiccharacteristics of the patients are similar to that of residentsin the surrounding service area [15].
The study population included active patients of the healthcare system (i.e., those who visited primary care or endocrinol-ogy departments) who were age 35 years or older, never hadtype 1 diabetes and were not pregnant during the surveillanceperiod: 1/1/2007–6/30/2010 (n = 254,259). We then identifiedpatients with type 2 diabetes. Evidence for type 2 diabetesincluded (1) two physician diagnosis of diabetes (ICD-9 codes250.xx) in the EHR Problem List or (2) two abnormal laboratoryvalues (fasting glucose, random glucose, oral glucose toler-ance or HbA1c tests) according to the 2005 American DiabetesAssociation (ADA) guideline [16]. Among patients with type 2diabetes (n = 20,341), we excluded patients who had evidence
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
of diabetes or diabetes treatment before the study entry. Toascertain “new” diabetes status, we further excluded patientswho had been in the health care system less than one yearbefore the first evidence of diabetes.
x x x ( 2 0 1 4 ) xxx–xxx
Among patients identified as having incident diabetes(n = 3237), we excluded patients with active cancer (554patients) or serious kidney or liver disease (239 patients),based on clinical encounter diagnosis. Of the remaining 2429patients, we then excluded 171 patients who did not makecontacts with a primary care physician during follow up period(i.e., from the initial evidence to 6/30/2011) to ensure that thestudy sample consisted of active primary care patients andthus did not receive care exclusively outside the health caresystem during the follow-up, leaving 2258 patients includedin the study. All analytical data sets were HIPAA de-identifiedand no patients were contacted for the study.
2.2. Treatment options
Treatment options for diabetes include medication pre-scription or education/counseling. Education/counseling wasdefined to include attendance in classes focused on diabetes-related lifestyle management, physician-led shared medicalappointments on diabetes, or individual counseling with dieti-tians or nutritionists (see http://www.pamf.org/diabetes/forexamples of classes). Participation in education/counselingsessions (whether it is reimbursed by payers, self-paid orfree of charge) was documented in the EHR. All treatmentsoccurred on or after confirmatory evidence of diabetes. Weassessed guideline-adherent treatment options classified into:(1) both education/counseling and medication prescription,(2) medication prescription only, and (3) education/counselingonly.
2.3. Covariates in multivariate models of predictors oftreatment
For clinical risk factors for treatment, we examined base-line values of fasting glucose, HbA1c, obesity (BMI ≥ 30 kg/m2),overweight (BMI ≥ 25 kg/m2 and <30 kg/m2), systolic bloodpressure, triglyceride, and low density lipoprotein (LDL)cholesterol. For all the clinical values, we used the value mea-sured and recorded on the date nearest (within the windowof 365 days prior to 30 days post) to the first evidence of dia-betes. Patient demographic characteristics included were age,sex, insurance type, race/ethnicity, and limited English profi-ciency (i.e., primary language not English). We also includedthree indicators of service use during the first year followinginitial diagnosis of diabetes: (1) any visit to an endocrinologydepartment, (2) a visit to other specialty or urgent care, and (3)no physician office visit.
2.4. Statistical analysis
To describe time to initial treatment, we computed aKaplan–Meier cumulative treatment curve. The starting pointwas the date of first evidence of incident type 2 diabetes andthe end point was the date of initial treatment or, if not treated,at the 12 months of follow-up.
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
We examined summary statistics and bivariate relation-ships of patient and physician characteristics with treatmentoptions. Fisher’s exact test and X2 test were used for the bivari-ate comparisons, as appropriate.
For multivariate models, we used multinomial logis-ic model with four exclusive categories: both medicationrescription and education/counseling, medication only,ducation/counseling only, and no treatment (base cate-ory). Physician (primary care provider of each patient)andom effects model was used to take into accounton-systematic variation within and across physicians.he results presented are exponentiated coefficients (eb)r relative risk ratios (RRR). All statistical analyses wereonducted using Stata 11.2 (StataCorp, College Station,exas).
. Results
.1. Demographic, clinical and service useharacteristics
he average age of 2258 patients with incident diabetes was6.9 years; less than half (43%) were female (Table 1). Theajority were non-Hispanic white (53%) followed by Asian
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
35%). About 9% had limited English proficiency. PPO (61%) wasost common insurance type.On average, fasting glucose value (the closest to diabetes
nset) was 149 mg/dL (or 8.3 mmol/L) and HbA1c was 7.3%
Table 1 – Demographic and clinical characteristics and initial tr
Variable (unit) [range]
Patient demographic and clinical characteristicsAge [35, 88 (top-coded)]
Female
Race/ethnicityNon-Hispanic white
Asian
African American or Black
Hispanic or Latino
Limited English proficiency
Insurance typePPO
HMO
Other insurance
Clinical risk factors at baselinea
Fasting glucose [58–771 mg/dL or 3.2–42.8 mmol/L]
Glycosylated hemoglobin (HbA1C) [5.0–16.5% or 5.4–23.6 mmol/L]
Obese
Overweight
High systolic blood pressure (≥140 mmHg)
High triglyceride (≥150 mg/dL or ≥1.7 mmol/L)
High LDL cholesterol (≥130 mg/dL or ≥3.4 mmol/L)
Service use during first follow up yearAny primary care visit
Any endocrinology visit
Other specialty or urgent care visit only
No clinic visit
Treatment in the first 12 months of diabetes incidenceBoth medication and education/counseling
Medication only
Education/counseling only
No treatment
a Fasting glucose: n = 2009 (89.0%); HbA1c: n = 1714 (75.9%); for other clinictriglyceride and high LDL cholesterol), missing baseline measurement w
x x ( 2 0 1 4 ) xxx–xxx 3
(or 9.0 mmol/L). Note that some patients did not have a posi-tive lab value, either fasting glucose or HbA1c near baseline,because they were identified as having incident diabetesbased solely on 2 or more physician diagnoses. Among thepatients whose pre-baseline fasting glucose value was lessthan 126 mg/dL (or 7.0 mmol/L) or HbA1c value was less than6.5% (or 7.8 mmol/L) (n = 1003), however, a majority of them(n = 608) did have a positive lab value, had anti-diabetic med-ication prescription, or received education/counseling duringthe post-12 months study period. The remainder probably hada sign of diabetes or lab values taken elsewhere on whichphysician diagnoses and treatment were based on. Such infor-mation may have been recorded in the EHR text fields or asa scanned document attachment, but we could not retrievethe textual data for this study due to potential identificationissues. More than two thirds of patients were obese (40%) oroverweight (30%).
Most patients (74%) made a visit to the clinic to see a physi-cian during the first year after diagnosis. The remainder (26%)did not visit the clinic in the first year after diagnosis, but didvisit the clinic in subsequent years. Most patients (72%) madea visit to a primary care physician, only 2.7% made a visit to an
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
endocrinologist, and 2.5% made other specialty care (e.g., oph-thalmology, immunology, rheumatology) or urgent care visitsonly.
eatment. Numbers represent mean [SD] or frequency (%).
All patients with incident diabetes (n = 2258)
56.9 [13.9]974 (43.1)
1188 (52.6)779 (34.5)47 (2.1)244 (10.8)197 (8.7)
1383 (61.3)807 (35.7)20 (0.9)
149.1 mg/dL or 8.3 mmol/L [60.5 mg/dL]7.3% or 9.0 mmol/L [1.8%]898 (39.8)670 (29.7)527 [23.3]959 [42.5]558 [24.7]
1615 (71.5)58 (2.6)57 (2.5)580 (25.7)
447 (19.8)451 (20.0)346 (15.3)1014 (44.9)
al indicators (overweight, obese, high systolic blood pressure, highas coded as 0.
Fig. 1 – Time to initial treatment (in weeks, up to 52 weeks)
care or endocrinology visit during the first follow-up year (all
– Kaplan–Meier cumulative hazard estimates.
3.2. Time to initial treatment of type 2 diabetes
About half (55%, n = 1244) of patients were treated with eithermedication or education/counseling during the first yearfollow-up period. When treated, the treatment tended to occur
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
relatively quickly (Fig. 1). Among those who were treated dur-ing the first year, 570 (46%) initiated the treatment in the firstweek, 840 (68%) initiated the treatment within the first monthof diabetes incidence. Conversely, if not treated in the first
Table 2 – Comparison of patients with and without a treatment
High systolic blood pressure (≥140 mmHg) 283 (22.8)
High triglyceride (≥150 mg/dL or 1.7 mmol/L) 592 (47.6)
High LDL cholesterol (≥130 mg/dL or 3.4 mmol/L) 322 (25.9)
Service use during first follow up yearAny primary care visit 931 (74.8)
Any endocrinology visit 46 (3.7)
Specialty or urgent care visit only 18 (1.5)
No clinic visit, 1st follow up year 291 (23.4)
For the bivariate comparisons, Fisher’s exact tests were used for dichotomoua Any medication prescription or attendance to an education/counseling sb Fasting glucose: n = 2009 (89.0%); HbA1c: n = 1714 (75.9%); for other clinic
triglyceride and high LDL cholesterol), missing baseline measurement w
x x x ( 2 0 1 4 ) xxx–xxx
few weeks, most remained untreated throughout the follow-up period. Only 158 patients (13% of treated or 7% of overallpatients) received a treatment after initial 16 weeks.
3.3. Treatment choices during the first year of diabetesincidence
About half of the patients were treated in the first year of dia-betes incidence. A fifth (20%) received medication only, 15%received education/counseling, and 20% both (Table 1). Whena patient received both medication and education/counseling(n = 447), the treatments were likely to be concurrent (84%).Among those who received medication prescription (n = 898),the first choice of medication was Metformin for the vastmajority (87%).
3.4. Comparison of treated and not treated patients
There were several significant differences in bivariate analy-sis between the patients who were treated and untreated inthe first year of follow-up (Table 2). Patients who were treated,compared to non-treated, were more likely to be young orobese, have higher levels of fasting blood glucose, triglycerideand LDL cholesterol, and were more likely to have a primary
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
at P < 0.001). The treated were less likely to be female or over-weight, to have limited English proficiency and to have madeother specialty or urgent care visits only (all at P < 0.01).
s variables and Chi-squared tests were used for continuous variables.ession.al indicators (overweight, obese, high systolic blood pressure, highas coded as 0.
(0.85–1.60) (0.83–1.53) (0.73–1.37)Service use during the first follow up year (ref: No physician office visit)c
Any primary care visit 1.13 1.56* 0.79(0.79–1.61) (1.11–2.21) (0.57–1.09)
Any endocrinology visit 3.55** 1.12 1.63(1.45–8.68) (0.39–3.19) (0.56–4.71)
Other specialty or urgent care visit only 0.52 0.37 0.62(0.19–1.47) (0.10–1.30) (0.27–1.43)
Relative risk ratios (95% CIs) are presented.Log likelihood = −2231; n = 2009.Coefficient for physician random intercepts (95% CI) = 1.70***(1.44–2.02).a For ease of interpretation, the units of clinical indicators are set to the difference between 25 percentile and median values, as indicated in the
parentheses. For example, the coefficient for fasting blood glucose indicates the odds ratio of 15 mg/dL or 0.8 mmol/L increase in the value.b The overall regression result was similar when HbA1c, instead of fasting blood glucose, was used.c Also included but not presented in the table are indicators of clinic location.∗ P < 0.05.
∗∗ P < 0.01.
3
MtTcc
∗∗∗ P < 0.001.
.5. Predictors of treatment for incident diabetes
ultivariate analyses results show that the likelihood of get-
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
ing any treatment decreases with advanced age (Table 3).he relative risk ratio (RRR) of patients age 55–64 asompared to age 35–44 to receive both medication and edu-ation/counseling was 0.54 (95% CI = 0.35–0.84), and the RRR
of patients age 75 or older was only 0.13 (95% CI = 0.07–0.25).Patients who have HMO insurance, compared to otherinsurance, mostly PPO, were more likely to receive educa-
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
tion/counseling alone (RRR = 1.48; 95% CI = 1.12–1.96) or alongwith medication (RRR = 1.43; 95% CI = 1.07–1.92). The likeli-hood of receiving a treatment did not differ based on patientrace/ethnicity.
Fasting blood glucose level was the strongest, indepen-dent positive predictor for receiving a treatment. For example,a 15 mg/dL (or 0.83 mmol/L) increase in the baseline fas-ting blood glucose was associated with 48% increase in RRR(95% CI = 1.40–1.57) to receive both medication and educa-tion/counseling, 38% increase in RRR (95% CI = 1.30–1.46) toreceive medication prescription only. The overall regressionresults were similar when using HbA1c (not reported), ratherthan fasting blood glucose. Obesity was a positive predictorof treatment either education/counseling alone (RRR = 1.66;95% CI = 1.14–2.42) or along with medication (RRR = 1.78; 95%CI = 1.21–2.63). Increasing level of triglyceride was also apositive predictor of combined treatment (RRR = 1.45; 95%CI = 1.08–1.95).
Visit to a primary care physician during the first follow upyear, as compared to no physician office visit, was a strongpredictor for receiving medication treatment (RRR = 1.56; 95%CI = 1.11–2.21) but not for education/counseling. Visit to anendocrinologist, on the other hand, increased the RRR toreceive combined treatment more than 3-fold (RRR = 3.55; 95%CI = 1.45–8.68).
4. Discussion
Despite the known importance of early identification andtreatment of diabetes for the prevention of diabetic compli-cations, incident diabetes is suboptimally treated even in agroup of insured, active primary care patients. In our study,only half of patients were treated during the first year fol-lowing diabetes incidence, and only 20% of patients receivedboth medication prescription and lifestyle modification inter-ventions. When treated, treatment often occurred promptly.Almost half of treated patients received their treatment dur-ing the first week after diabetes incidence, and thereafter thecurve for treatment initiation is relatively flat. When a patientreceived both medication and lifestyle interventions, the twotreatments were concurrent most of time, as guidelines rec-ommend [16]. Given the extensive literature on the benefitsand safety of starting both treatments even in the pre-diabeticstage, more emphasis should be placed on earlier treatmentinitiation [4–7,17,18].
The UCLA-CHAMP program showed that immediate ini-tiation of all effective therapies for secondary preventionof coronary artery disease was more effective than a moreconservative outpatient follow-up approach [19]. A similarprogram for patients with incident diabetes may proposeimmediate initiation of therapy with education/counselingin conjunction with medication, with appropriate follow-up.More research is needed to assess the specific reasons thatpatients and providers do not initiate treatment immediately,to inform effective interventions.
The observed suboptimal initial treatment rates may reflectpatient preference. As our study cohort consisted of activeprimary care patients of the health care system, most (74%)patients in this study did see a physician during the first year
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
after diagnosis. Among those patients who were not treatedin the first year after diagnosis, 30% made a clinic visit tospecifically address their diabetes (as evidenced by a visit diag-nosis of diabetes). Two thirds (68%) of patients who were not
x x x ( 2 0 1 4 ) xxx–xxx
treated actually came into the clinic to see a physician. Allthe patients included in the study were active patients of thehealth care system and were eventually seen in clinic over thesubsequent years of follow up (up to 4.5 years after initial diag-nosis). Further years of follow up after the first year, however,did not increase treatment rates in these patients (results notreported), confirming that most initial treatment takes placeat diagnosis.
Within the first year after diagnosis, patients who saw aprimary care physician or endocrinologist were more likelyto receive medication and/or education/counseling treatment.Seeing a physician at urgent care or a specialist other thanendocrinologist, mainly for reasons other than diabetes, how-ever, did not make a difference in treatment rates, comparedto those who did not see a physician. Our findings argue for theneed for a team approach for diabetes management. Promptsin the EHR could be included to flag patients with diabetes sothat even when unrelated clinic visits are made, follow-up isappropriately pursued.
The organization studied has shown high performance indiabetes management indicators in the state-wide pay-for-performance program [20]. Even in this high resource setting,treatment of initial diabetes seems to be suboptimal. Diabetesregistries often use physician diagnosis of diabetes only, andas diabetes in its early stage is often underdiagnosed [21], inci-dent diabetes cases that are not documented tend not to beactively treated. Improved algorithms for identifying diabetescases using laboratory values (as we have done) now availablein EHR, in addition to physician diagnosis (traditionally avail-able in administrative data), may help monitor early initiationof treatment.
Our data indicate that incident diabetes cases are typicallymanaged by primary care providers rather than immedi-ately referred to an endocrinologist (only 2.6% of primarycare patients saw an endocrinologist during the first follow-up year). For patients who made a visit to primary caredepartment, particular physician demographic or practicecharacteristics (e.g., gender, years of practice, panel size, dia-betes panel size) did not predict treatment initiation (resultsnot reported). This is consistent with existing evidence show-ing that majority of new diagnosis of diabetes occurred inprimary care setting, and 65% of patients with diabetes inthe United Stated are cared for by primary care physicians[22]. The ADA guidelines recommend uncomplicated cases ofdiabetes to be managed by primary care physicians [16].
The lower treatment rate in older patients is of con-cern. Clinical guidelines generally recommend less aggressiveglycemic control for frail older adults [23,24], although theevidence regarding the recommendation is tenuous becausefrail older patients are usually excluded from clinical trials[23,24]. On the other hand, in our diverse insured population,race/ethnicity or limited English proficiency was not a signif-icant predictor of treatment, which may suggest that accessmay underlie some of the racial/ethnic disparities in diabetestreatment reported in the literature [25]. Insurance type didmatter in receiving lifestyle interventions, which may be more
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
available to HMO patients who are covered by capitation. Therecent trends toward bundled payment or Accountable CareOrganization could have a positive influence on providingmore comprehensive diabetes treatment.
Several limitations, mostly due to the defined scope ofur study, merit discussion. First, adherence to medicationas not examined, as we did not have access to prescrip-
ion dispensing data. Therefore, actual rate of medicationse by patients may be lower. On the other hand, educa-ion/counseling treatment was based on attendance, whichequires both physician referral and patient adherence. Sec-nd, some patients who were not treated nor whose diabetesas not actively monitored may have had other serious
omorbidities, so onset of diabetes may not have been a prior-ty in their treatment plan. We excluded patients with illnesseshat prohibit active diabetes treatment such as cancer, kidneyisease and liver disease, but there could be still many otheratients whose conditions may preclude active diabetes treat-ent. Finally, as our study focuses on treatment options in
linical settings, we did not examine patient-initiated lifestyleodifications outside of clinical care realm.It is important to note that the study is conducted in the US
here providers are mostly paid on a fee-for-services basis.ack of care coordination, in part due to the payment sys-em, may be an important contributor of delayed initiationf diabetes treatment. Many patients (39%) did not make aollow up visit to a primary care provider during the firstear of diabetes diagnosis. Furthermore, there are multipleuidelines for diabetes treatment, which often conflict eachther. The American Diabetes Association guideline, which
s most widely adopted by clinicians, is not strictly enforcedy the health care organizations. Treatment cost to patient,owever, may not be a large inhibiting factor because all
he patients in the study were insured, though co-pays andrug plans likely differed across patients, and we were unableo examine this factor as a predictor of initiation of ther-py. Medication and education/counseling treatment, withrovider diagnosis, are usually covered by most insurance. Theost commonly prescribed diabetes medication, Metformin,
s generically available and is very inexpensive (e.g., $4 perwo months’ supply), so direct-to-consumer advertising is notikely to have played a large role in prescribing patterns in thistudy.
. Conclusions
ewly identified diabetes is undertreated, even in a highesource setting where all patients have health insurancend access to care. When treated, treatment occurs soonfter diagnosis. If not treated immediately, incident dia-etes is rarely infrequently treated after the first few weeks,hile patients continue seeking care for other reasons. Team
pproaches across specialties to monitor patients with inci-ent diabetes, taking advantage of shared EHR system, mayacilitate early initiation of diabetes treatment.
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
onflict of interest statement
he authors declare that they have no conflicts of interest inonnection with this article.
x x ( 2 0 1 4 ) xxx–xxx 7
Funding source
This study was supported by the National Institute of Diabetesand Digestive and Kidney Diseases (1R01DK081371) and theAgency for Healthcare Research and Quality (K01HS019815).The funding sources had no role in the study’s design, conductor reporting.
e f e r e n c e s
[1] D.M. Nathan, J.B. Buse, M.B. Davidson, et al., Medicalmanagement of hyperglycemia in type 2 diabetes: aconsensus algorithm for the initiation and adjustment oftherapy, Diabetes Care 32 (2009) 193–203.
[2] ADA, Executive summary: standards of medical care indiabetes—2010, Diabetes Care 33 (Suppl. 1) (2010) S4–S10.
[3] S. Colagiuri, C.A. Cull, R.R. Holman, Are lower fasting plasmaglucose levels at diagnosis of type 2 diabetes associated withimproved outcomes? Diabetes Care 25 (2002) 1410–1417.
[4] W.C. Knowler, S.E. Fowler, R.F. Hamman, et al., 10-Yearfollow-up of diabetes incidence and weight loss in theDiabetes Prevention Program Outcomes Study, Lancet 374(2009) 1677–1686.
[5] W.C. Knowler, E. Barrett-Connor, S.E. Fowler, et al., Reductionin the incidence of type 2 diabetes with lifestyle interventionor metformin, N. Engl. J. Med. 346 (2002) 393–403.
[6] X.R. Pan, G.W. Li, Y.H. Hu, et al., Effects of diet and exercisein preventing NIDDM in people with impaired glucosetolerance. The Da Qing IGT and Diabetes Study, DiabetesCare 20 (1997) 537–544.
[7] G. Li, P. Zhang, J. Wang, et al., The long-term effect oflifestyle interventions to prevent diabetes in the China DaQing Diabetes Prevention Study: a 20-year follow-up study,Lancet 371 (2008) 1783–1789.
[8] M. Espeland, Reduction in weight and cardiovasculardisease risk factors in individuals with type 2 diabetes:one-year results of the Look AHEAD trial, Diabetes Care 30(2007) 1374–1383.
[9] S. Malik, V. Lopez, R. Chen, et al., Undertreatment ofcardiovascular risk factors among persons with diabetes inthe United States, Diabetes Res. Clin. Pract. 77 (2007)126–133.
[10] M.J. Jacobs, T. Kleisli, J.R. Pio, et al., Prevalence and control ofdyslipidemia among persons with diabetes in the UnitedStates, Diabetes Res. Clin. Pract. 70 (2005) 263–269.
[11] K.J. Hunt, M. Gebregziabher, L.E. Egede, Racial and ethnicdifferences in cardio-metabolic risk in individuals withundiagnosed diabetes: National Health and NutritionExamination Survey 1999–2008, J. Gen. Intern. Med. 27 (2012)893–900.
[12] J.A. Schmittdiel, C.S. Uratsu, A.J. Karter, et al., Why don’tdiabetes patients achieve recommended risk factor targets?Poor adherence versus lack of treatment intensification, J.Gen. Intern. Med. 23 (2008) 588–594.
[13] S.D. Horn, J. Gassaway, Practice-based evidence study designfor comparative effectiveness research, Med. Care 45 (2007)S50–S57.
[14] J.M. Westfall, J. Mold, L. Fagnan, Practice-basedresearch—“Blue Highways” on the NIH roadmap, JAMA 297(2007) 403–406.
[15] US Census Bureau, Profile of General Demographic
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
Characteristics: 2000 Census Summary File 2, 100-percentdata, 2000.
[16] ADA, Standards of medical care in diabetes, Diabetes Care28 (Suppl. 1) (2005) S4–S36.
patients with diabetes, JAMA 305 (2011) 1350–1351.
8 p r i m a r y c a r e d i a b e
[17] A. Ramachandran, C. Snehalatha, S. Mary, et al., The IndianDiabetes Prevention Programme shows that lifestylemodification and metformin prevent type 2 diabetes inAsian Indian subjects with impaired glucose tolerance(IDPP-1), Diabetologia 49 (2006) 289–297.
[18] J. Tuomilehto, J. Lindstrom, J.G. Eriksson, et al., Prevention oftype 2 diabetes mellitus by changes in lifestyle amongsubjects with impaired glucose tolerance, N. Engl. J. Med. 344(2001) 1343–1350.
[19] G.C. Fonarow, A. Gawlinski, S. Moughrabi, J.H. Tillisch,Improved treatment of coronary heart disease byimplementation of a Cardiac Hospitalization AtherosclerosisManagement Program (CHAMP), Am. J. Cardiol. 87 (2001)
Please cite this article in press as: S. Chung, et al., Initiation of treatment
an ambulatory care setting, Prim. Care Diab. (2014), http://dx.doi.org/10.10
819–822.[20] IHA, Pay for Performance-Manuals and Program Operations,
[21] M.A. Stone, J. Camosso-Stefinovic, J. Wilkinson, et al.,Incorrect and incomplete coding and classification ofdiabetes: a systematic review, Diabetic Med. 27 (2010)491–497.
[22] S.M.R.E. Schappert, Ambulatory Medical Care UtilizationEstimates for 2007, National Center for Health Statistics,2011.
[23] A.F. Brown, C.M. Mangione, D. Saliba, C.A. Sarkisian,Guidelines for improving the care of the older person withdiabetes mellitus, J. Am. Geriatr. Soc. 51 (5 Suppl. Guidelines)(2003) S265–S280.
[24] S.J. Lee, C. Eng, Goals of glycemic control in frail older
for incident diabetes: Evidence from the electronic health records in16/j.pcd.2014.04.005
[25] M. Heisler, D.M. Smith, R.A. Hayward, et al., Racialdisparities in diabetes care processes, outcomes, andtreatment intensity, Med. Care 41 (2003) 1221–1232.
