Seediscussions,stats,andauthorprofilesforthispublicationat:http://www.researchgate.net/publication/271594060

AnInternet-BasedDiabetesManagementPlatformImprovesTeamCareandOutcomesinanUrbanLatinoPopulation

ARTICLEinDIABETESCARE·JANUARY2015

ImpactFactor:8.57·DOI:10.2337/dc14-1412·Source:PubMed

CITATION

1

DOWNLOADS

35

VIEWS

111

7AUTHORS,INCLUDING:

GarryWelch

betweenMDLLC

65PUBLICATIONS2,052CITATIONS

SEEPROFILE

Sven-ErikBursell

UniversityofMelbourne

32PUBLICATIONS1,697CITATIONS

SEEPROFILE

MilagrosC.Rosal

UniversityofMassachusettsMedicalSchool

112PUBLICATIONS2,241CITATIONS

SEEPROFILE

RobertAGabbay

PennsylvaniaStateUniversity

42PUBLICATIONS989CITATIONS

SEEPROFILE

Availablefrom:GarryWelch

Retrievedon:08September2015

An Internet-Based DiabetesManagement Platform ImprovesTeam Care and Outcomes in anUrban Latino PopulationDOI: 10.2337/dc14-1412

OBJECTIVE

To compare usual diabetes care (UDC) to a comprehensive diabetes care inter-vention condition (IC) involving an Internet-based “diabetes dashboard”manage-ment tool used by clinicians.

RESEARCH DESIGN AND METHODS

We used a parallel-groups randomized design. Diabetes nurses, diabetes dieti-tians, and providers used the diabetes dashboard as a clinical decision supportsystem to deliver a five-visit, 6-month intervention to n = 199 poorly controlled(HbA1c >7.5% [58 mmol/mol]) Latino type 2 diabetic (T2D) patients (mean age55 years, 60% female) at urban community health centers. We compared thisintervention to an established, in-house UDC program (n = 200) for its impacton blood glucose control and psychosocial outcomes.

RESULTS

Recruitment and retention rates were 79.0 and 88.5%, respectively. Comparedwith UDC, more IC patients reached HbA1c targets of <7% (53 mmol/mol; 15.8 vs.7.0%, respectively, P < 0.01) and <8% (64 mmol/mol; 45.2 vs. 25.3%, respectively,P < 0.001). In multiple linear regression adjusting for baseline HbA1c, adjustedmean6 SE HbA1c at follow-up was significantly lower in the IC compared with theUDC group (P < 0.001; IC 8.46 0.10%; UDC 9.26 0.10%). The results showed lowerdiabetes distress at follow-up for IC patients (40.4 6 2.1) as compared with UDCpatients (48.36 2.0) (P < 0.01), and also lower social distress (32.26 1.3 vs. 27.261.4, P < 0.01). There was a similar, statistically significant (P < 0.01) improvementfor both groups in the proportion of patients moving from depressed status atbaseline to nondepressed at follow-up (41.8 vs. 40%; no significance betweengroups).

CONCLUSIONS

The diabetes dashboard intervention significantly improved diabetes-related out-comes among Latinos with poorly controlled T2D compared with a similar diabe-tes team condition without access to the diabetes dashboard.

Type 2 diabetes (T2D) is a rapidly growing epidemic in the U.S., currently affecting29.1 million Americans (1) and projected to impact.40 million individuals by 2034(2). National surveys show that the large majority of individuals with T2D are notat recommended treatment goals for its underlying risk factors, namely,

1Behavioral Medicine Research, Baystate Medi-cal Center, Springfield, MA2University of Hawaii at Manoa, Honolulu, HI3Division of Preventive and Behavioral Medicine,University of Massachusetts Medical School,Worcester, MA4Joslin Diabetes Center, Boston, MA

Corresponding author: GarryWelch, garry.welch@bhs.org.

Received 5 June 2014 and accepted 19 December2014.

Clinical trial reg. no. NCT02156037, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-1412/-/DC1.

© 2015 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

Garry Welch,1 Sofija E. Zagarins,1

Paula Santiago-Kelly,1 Zoraida Rodriguez,1

Sven-Erik Bursell,2 Milagros C. Rosal,3 and

Robert A. Gabbay4

Diabetes Care 1

CLIN

CARE/ED

UCATIO

N/N

UTR

ITION/PSYC

HOSO

CIAL

Diabetes Care Publish Ahead of Print, published online January 29, 2015

hyperglycemia, hypertension, and dysli-pidemia (3). These combine to promoteserious and costly complications of thecardiovascular system, eyes, kidneys,and feet (3). Healthcare delivery factors,such as lack of care coordination andprovider clinical inertia (i.e., slownessto appropriately intensify diabetestreatment) are significant contributingfactors to poor metabolic control seenin T2D (4,5). Also, patient psychosocialfactors, such as diabetes distress, socialdistress, and depression, that impactpatient engagement and treatment adher-ence are not systematically managed aspart of routine medical care in T2D (6,7).The Affordable Care Act is a landmark

piece of healthcare legislation that pro-motes more proactive and patient-centered management of T2D and otherchronic diseases and, to accomplish this,promotes significant clinical care deliv-ery and provider payment reforms.Other significant national healthcarelegislation has mandated the nationaladoption of electronic medical records(EMRs) in clinical care to allow more ef-ficient capture and “meaningful use” ofpatient clinical data across providersand clinical settings to facilitate greaterpatient engagement in self-care (8). ForU.S. healthcare system reforms to suc-ceed, it will be critical that providers areequipped with well-designed clinical de-cision support (CDS) tools that can facil-itate patient-centered care and improveteam communication and efficiency(9,10). CDS tools typically include clinicalalerts and reminders, order sets, and drug-dose calculators that automatically promptthe clinician to implement a specific actionand include care summary dashboardsthat provide performance feedback on im-portant quality indicators. Although signif-icant recent progress has been made inthe creation of CDS applications for T2D(11–14), they remain at an early stage ofdevelopment and evaluation.We report here on the results of a

randomized clinical trial that examinedthe clinical effectiveness of a compre-hensive diabetes care intervention inwhich an Internet-based “diabetesdashboard” disease management appli-cation was used as a CDS system forteam care delivered at urban poor safetynet clinics. We compared the clinical ben-efit of the diabetes dashboard interven-tion with that of a control conditionproviding usual diabetes care (UDC).

RESEARCH DESIGN AND METHODS

We used a parallel-groups randomizeddesign for this clinical trial. Eligible pa-tients were randomized either to thediabetes dashboard intervention condi-tion (IC) or to an in-house UDC programdelivered without access to the diabetesdashboard. The study was conductedat two affiliated Federally QualifiedHealthcare Centers (FQHCs) located inWestern Massachusetts in an areawhere .30% of families locally live be-low the federal poverty line (15). Theclinics are located in a medically under-served and health professional shortagearea. The 29 clinic providers serve a pre-dominantly (;80%) Latino urban poorcommunity including .2,400 diabeticpatients.

For this study, eligible T2D patientswere recruited from December 2010 toDecember 2012. Patients were identi-fied from a clinic diabetes registry andusing referrals from an ophthalmologypractice affiliated with the participatingclinics. Patient inclusion criteria were as fol-lows: age 18 years or older, self-identifiedHispanic ethnicity, diagnosis of T2D,HbA1c .7.5% (58 mmol/mol), and pro-vider approval given for patient par-ticipation. Exclusion criteria includedinability to consent, pregnant or plan-ning to become pregnant in the nextyear, taking glucocorticoid therapy, orhaving serious psychiatric or medicalcomplications (e.g., late-stage diabetescomplications, seizures, dementia, or psy-chiatric hospitalization) that would pre-vent participation in study activities.Patients were paid a stipend for comple-tion of baseline and 6-month follow-upresearch assessments ($25 each). The in-tervention was implemented at medicaloffices locatedwithin the FQHCs. The pro-tocol was approved by the BaystateMed-ical Center Institutional Review Board.

Diabetes Dashboard ICThe IC involved a program of five, in-person, one-on-one diabetes educationvisits with a diabetes nurse or diabetesdietitian, scheduled at baseline, 2 weeks,1 month, 3 months, and 6 months post-enrollment. The initial visit was an hourlong, and the remaining visits were a halfhour long each. The IC was deliveredby a team of four bicultural, bilingualdiabetes educators (two diabetes nursesand two diabetes dietitians), with pa-tients scheduled to see specific educators

by request or based on availability (e.g.,patients could request to see the sameeducator for repeated visits or could seeall four educators over the course oftheir study participation).

The diabetes nurse and diabetes die-titian interventionists used an Internet-based “diabetes dashboard” diseasemanagement tool (see SupplementaryFig. 1) to structure each education visitand to share information collected dur-ing each visit with each other and withclinic providers. This dashboard, re-ferred to during this study as the Com-prehensive Diabetes ManagementProgram, has been described previously(16,17) and combines existing clinicaldata obtained from paper chart–basedand electronic health records (i.e., vitalsigns, laboratories, medications, admis-sions, procedures, and diagnoses) withadditional patient data gathered usingintegrated surveys (described below)and during the course of ongoing care.Two of the diabetes educators (P.S.-K.and Z.R.) had extensive experience usingthe diabetes dashboard in an earlier pi-lot study (16).

The diabetes dashboard provides thefollowing: 1) a system of individual clin-ical alerts and reminders (e.g., missingor elevated HbA1c) and a diabetes com-plications risk profile (five compositerisks of glycemia, retinopathy, cardiac,peripheral vascular disease/peripheralneuropathy, and nephropathy) that sup-ports the delivery of evidence-basedtreatment protocols (18,19) (for exam-ple, the glycemia risk complicationsalert reflects the current level ofHbA1c, annual frequency of testing ofHbA1c, and diagnoses hypoglycemia);2) a set of nursing, medical nutritiontherapy, and physical activity treatmentplan encounter forms involving drop-down menus and a structured data col-lection process; 3) a library of diabeteseducation teaching resources based onAmerican Association of Diabetes Edu-cator guidelines (AADE7) (20); and 4) aseries of clinical reports, including a pro-vider summary (see SupplementaryFig. 2) generated after each interventionvisit that is emailed to the provider tosupport clinical decision making and in-cludes recommendations for changes inmedication management for hypergly-cemia, hypertension, and dyslipidemia.

For the current study, each educationvisit with the diabetes nurse or diabetes

2 Diabetes Dashboard and Team Care for Latinos Diabetes Care

dietitian interventionists began with areview based on a summary of patient-reported self-management behaviorsand barriers (i.e., blood glucose testing,diet, physical activity, and medicationadherence) and psychosocial challenges(i.e., diabetes distress, social distress,depression, hypoglycemia, binge eating,alcohol abuse, and low social support)collected using an established survey in-tegrated within the dashboard (i.e., theDiabetes Self-Care Profile [21]). Next,the interventionist reviewed the pa-tient’s vital signs and laboratory data,conducted a medication review and rec-onciliation process and updated themedication list, reviewed clinical alertsand reminders generated by the system,and updated the nursing or dietetictreatment plan using encounter forms.Following these steps, the interventionistdelivered diabetes education tailored tothe patient’s individual clinical, behav-ioral, and psychosocial profile and re-ferred the patient for psychosocialservices (e.g., adjacent mental healthclinic for depression) as needed andwith notification to the primary care pro-vider. Interventionists recorded clinicalnotes for each visit by free text using a“whiteboard” panel on the dashboard tofacilitate internal team communicationand patient hand off between sessions.The diabetes nurse and diabetes die-

titian interventionists created clinicalcare recommendations for providerson pharmacological management of ab-normal blood glucose, blood pressure(BP), and lipid levels (e.g., Supplemen-tary Fig. 2) after several initial diabeteseducation evaluation and education ses-sions to develop rapport, assess currentmedication adherence, and provide in-dividualized diabetes education andsupport. A patient safety and triageplan refined by the primary care providerswas used for patients who presented atintervention visits as symptomatic forshortness of breath, chest pain, headache,BP.180 mmHg, or BG.350 mg/dL withpresence of ketones. Presence of thesesymptoms triggered a notification to theprovider, covering physician, or clinicalnurse for action.To address the cultural needs of the

Latino patients that were the focus ofthis study, the intervention included thefollowing: 1) delivery of the interventionand diabetes education materials in thepatient’s preferred language (Spanish or

English), 2) literacy and numeracyscreening using a brief, practical assess-ment tool we had used in prior research(22), 3) encouragement of attendance byfamilymembers in intervention sessions,4) inclusion of ethnic foods andmodifiedethnic recipes in the provision of medicalnutrition therapy, and 5) assessment ofalternative healers and home remediesby patients and encouragement of pa-tients to discuss these alternative practi-ces for their safety and risk with theirprimary care provider.

Training for the diabetes interven-tionist team included training in theuse of the dashboard as well as a diabe-tes medication treatment protocol pro-vided for the management of bloodglucose, BP, and blood lipid medicationsin T2D based on national guidelines (19).Three hours of trainingwere provided tothe diabetes team as one in-person ses-sion and two conference calls by studyMDs with expertise in the clinical man-agement of diabetes, hypertension, andhyperlipidemia. FQHC providers re-ceived three 1-hour informational andeducational sessions conducted byG.W., R.A.G., and P.S.-K. on the diabetesprogram and CDS reports they would re-ceive during the study.

UDCThe UDC conditionwas delivered by fouradditional bicultural, bilingual diabetesnurses and diabetes dietitians who com-prised the clinical site’s long-standing,in-house diabetes program. This pro-gram was designed as part of the RobertWood Johnson Foundation Diabetes Ini-tiative to advance the delivery of cultur-ally sensitive care for patients with T2Din primary care (23,24). The UDC condi-tion involved a series of individual pa-tient visits with education content.Visit frequency was based on individualpatient needs as determined by programclinicians. Patients also had access to life-style and diabetes self-management sup-port groups run at the clinics by peervolunteers and clinical staff. Patients inthe UDC condition completed the sameassessment battery (i.e., Diabetes Self-Care Profile [18]) as that completed bypatients in the IC. However, data fromthis assessment was used only for re-search purposes and was not used toguide clinical care delivered within theUDC condition. Both IC and UDC patientsreceived routine medical care from their

healthcare providers for any acute andemergent problems based on estab-lished clinic standards and procedures.

Measurements

Clinical Measures

Patients attended a 1-hour baseline re-search assessment and a 30-min follow-up assessment at 6 months. The primarystudy outcome was defined as the per-centage of patients achieving goodblood glucose control (i.e., HbA1c ,7%[53 mmol/mol]). HbA1c was obtainedusing a validated finger stick blood testkit (Appraise Home HbA1c Kit; HeritageLaboratories International LLC). HeritageLaboratories is certified by the NationalGlycohemoglobin Standardization Pro-gram. The Appraise Home HbA1c Kit pro-duces accurate and reliable test resultsequivalent to whole blood tests col-lected in physicians’ offices. Other clini-cal variables assessed the percentage ofpatients at target BP (,130/80 mmHg)and BMI. Systolic and diastolic BP mea-surements were obtained by researchstaff during baseline and follow-up re-search visits based on a single seatedassessment using an automatic digitalBP monitor (Omron model HEM-705CP).BMIwas calculated asweight in kilogramsdivided by the square of height in meters.Hypoglycemia was defined in the Diabe-tes Self-Care Profile as any “low bloodsugars or sweating, nausea, heart pound-ing, trembling, cold and clammy skin, dif-ficulty concentrating, and irritability”overthe past month.

Psychosocial Measures

We used the Diabetes Self-Care Profilesurvey (18) to assess diabetes distress,social distress, and depression (secondarystudy outcomes). Assessment of diabetesdistress involved the short (five-item)version of the Problem Areas In Diabetes(PAID) questionnaire that assesses theemotional burden of diabetes and itstreatment. PAID is a valid and widelyused measure that uses a 0–100 scale,with higher scores denoting greaterdistress (25,26). Wemeasured social dis-tress on a 0–100 scale using the 20-itemTool for Assessing Patients’ Stress (TAPS)questionnaire, a measure with evidenceof internal reliability and construct valid-ity and found acceptable to urban poorT2D patients (6,16,27). TAPS assesses re-cent distress related to taking care offamily needs and problems, lack ofmoney for basic living needs or having

care.diabetesjournals.org Welch and Associates 3

family conflicts, legal problems, overcrowd-ing, living in an unsafe neighborhood, phys-ical or mental abuse, discrimination, andjob loss or underemployment, amongother significant social and family issuestargeted. We measured depression us-ing the Patient Health Questionnaire, avalidated, widely used nine-item self-report measure of depression (28). Otherpatient data collected at baseline in-cluded age, sex, race, ethnicity, and dura-tion of diabetes in years.

Data AnalysesWe described characteristics of the studypopulation using means and SDs for con-tinuous covariates and Student t test toassess whether differences in means be-tween the two treatment conditionswerestatistically significant. For categorical co-variates, we reported the number andpercentage of patients within each cate-gory and examined differences betweentreatment groups using Fisher exact test,which is more conservative than the x2

and is appropriate for both large andsmall cell frequencies.We conducted outcome analyses as

intention to treat, such that we analyzedpatients with the group they were ran-domized to regardless of how many in-tervention visits they completed. Weconducted an efficacy subset analysisapproach to address missing researchdata, as loss to follow-up was small,with ,10% of patients having missingresearch data at follow-up. For compar-isons of outcomes by treatment status,we conducted a sensitivity analysis inwhich we used multiple imputationmethods to address missing data.We evaluated associations between

treatment group and the dichotomousHbA1c control status variables using un-adjusted and multiple logistic regres-sion, with HbA1c control status as thedependent variable. We evaluated asso-ciations between treatment group andthe continuous outcome variables usingunadjusted and multiple linear regres-sion. We adjusted models for baselinevalues and considered covariates thatwere associated with treatment statusor HbA1c at P, 0.20 on univariable anal-ysis for inclusion in our final multiple re-gression models. We included covariatesin the final multiple regression models iftheir addition resulted in at least a 10%change in the b coefficient for the treat-ment status variable.

We performed analyses using SASsoftware version 9.3 (SAS Institute,Cary, NC) and Stata (version 12.0; Stata-Corp, College Station, TX) (29). The SAScommands we used included proc freqfor categorical comparisons and procGLM for modeling continuous variables.To examine the influence of missingdata, we used multiple imputation toreplace missing values (i.e., Stata’s “miimpute mvn” command). Assuming anunderlying multivariate normal distribu-tion, the command imputes missingvalues through an iterative MCMC ap-proach. We created 20 imputed data-sets to reduce sampling variabilityfrom the imputation process.

RESULTS

Screening, Recruitment, andRetentionFigure 1 shows the recruitment and re-tention of patients into this clinical trial.In brief, 75.4% of eligible patients in-vited to participate in the study weresubsequently enrolled and randomizedto either the intervention (n = 199) orcontrol (n = 200) study conditions.IC patients completed an average of3.86 1.5 visits with the study interven-tionists. Although data were not col-lected on the number of clinic visits forindividual UDC patients, patients receiv-ing UDC at our clinical site attended anaverage of 5.2 visits with clinic providersduring a 5–6-month time frame basedon an unpublished internal clinic reportand from interviews of diabetes staffmembers following the interventionphase. Follow-up research visits werecompleted by 86.4% of IC patients and90.5% of UDC patients.

Sample CharacteristicsParticipant baseline characteristics areshown in Table 1. There were no signif-icant differences between the studyconditions in terms of demographic(sex, age, and race), clinical (HbA1c, BP,and BMI), andmost psychosocial (depres-sion status, social distress, and perceivedsocial support) variables. However, a sig-nificant baseline difference between thegroups was observed in diabetes distress(62.9% for IC vs. 50.5% forUDC,P, 0.03),although both groups were at clinicallyhigh levels.

Clinical OutcomesRates of HbA1c control were higher amongIC patients at follow-up, such that 15.8%of

IC patients were at the treatment goal ofHbA1c ,7% (53 mmol/mol), as comparedwith 7.0%ofUDCpatients (P,0.01). In ananalysis of patients with an HbA1c .8.0%(64 mmol/mol) at baseline, 45.2% of ICpatients vs. 25.3% of UDC patients metthe goal of HbA1c ,8.0% at follow-up(P , 0.001). In multiple linear regressionadjusting for baseline HbA1c, adjustedmean6 SE HbA1c at follow-up was signif-icantly lower by 0.816 0.15% units in theIC group as compared with the UDC group(P , 0.001; IC 8.4 6 0.10%; UDC 9.2 60.10%) (Table 2). Results for mean HbA1cat follow-up were similar in our sensitivityanalysis based on imputed data, such thatHbA1c at follow-up was 0.82 6 0.15%units lower in IC versus UDC participants(P, 0.001).

We also examined descriptive clinicaldata on BP and BMI at follow-up usingmultiple linear regression adjusted forbaseline values and found no significantdifference between the groups in termsof these variables (Table 2). Resultswere similar when multiple imputationmethodswere used to fill inmissing data(data not shown).

Self-reported hypoglycemia symp-toms improved in both groups. In theIC group, 34.7% of patients reportedhaving hypoglycemia symptoms inthe prior month to baseline. Of thosepatients, only 49.3% reported hypogly-cemia symptoms at follow-up. In theUDC group, 38% of patients reported

Figure 1—Flowchart showing participantenrollment and retention rates. Pt, patient;R2, research visit number two at six monthsfollow-up.

4 Diabetes Dashboard and Team Care for Latinos Diabetes Care

hypoglycemia symptoms at baseline,and only 44.7% of those patients re-ported symptoms at follow-up. Therewere no statistical differences betweenthe IC and UDC conditions. There werealso no differences between the twoconditions in new reports of hypogly-cemia at follow-up (22 vs. 20.6%, nosignificance).

Psychosocial OutcomesThe results showed lower diabetes dis-tress at follow-up for IC patients (40.4 62.1) as compared with UDC patients(48.3 6 2.0) (P , 0.01) and also lowersocial distress (32.26 1.3 vs. 27.26 1.4,

P , 0.01) (Table 2). There was a similar,statistically significant (P , 0.01) im-provement for both groups in the pro-portion of patients moving fromdepressed status at baseline to nonde-pressed at follow-up (i.e., 41.8 vs. 40%),with no significant difference betweengroups in terms of change in depressionstatus.

CONCLUSIONS

This clinical trial conducted at two affil-iated urban safety net clinics focused onLatino T2D patients in poor glycemiccontrol and demonstrated the clinicaleffectiveness of a diabetes care program

enriched by use of a diabetes dashboardapplication to support team care. Thedashboard provided the diabetes teamwith timely clinical alerts and remindersof diabetes-specific medical and psycho-social issues, encounter and treatmentplan templates, and diabetes educationresources and generated summary re-ports of intervention sessions to sharewith providers. Despite the artificialityinherent in delivering a time-limited(6-month) clinical research interventionwithin a busy primary care setting, thediabetes dashboard helped organize thework of the diabetes educator team(i.e., diabetes nurses and diabetes dieti-tians) and supported the provision ofpatient-centered and evidence-baseddiabetes care. The diabetes dashboardalso created a bridge to the clinic pro-viders via the individual session sum-mary reports and medication changerecommendations sent to the providersfollowing intervention sessions. TheUDC control condition consisted of along-standing comprehensive diabetescare program designed as part of aRobert Wood Johnson Foundation Dia-betes Initiative to advance the deliveryof culturally sensitive care for patientswith T2D (23,24).

The study findings showed that twiceas many IC patients achieved a goal ofHbA1c ,7% (53 mmol/mol) comparedwith the UDC condition (i.e., 15.8 vs.7.0%, respectively). For an HbA1c cutoffof,8% (64mmol/mol), the results were45.2 vs. 25.3%, respectively. The in-tervention provided a statisticallyand clinically significant mean HbA1c

improvement (reduction) of 20.6%(26.6 mmol/mol) compared with aworsening for the UDC condition of+0.2% (+2.2 mmol/mol). As a benchmarkto interpret this difference in intermedi-ate diabetes outcomes, landmark na-tional studies have shown that for every1% (10.9 mmol/mol) reduction in HbA1c,the risk of developing eye, kidney, andnerve disease is reduced by 40% whilethe risk of heart attack is reduced by14% (30).

Analysis of our secondary psychoso-cial outcomes showed a significant re-duction in both diabetes distress andsocial distress for the IC comparedwith the UDC condition. Both conditionsshowed high baseline levels of distress,consistent with findings from prior stud-ies of urban poor T2D populations

Table 1—Comparison of intervention groups at baseline

Usual diabetescare, n = 200

Interventioncondition, n = 199

Pvalue1

Continuous variables Mean 6 SD Mean 6 SDAge (years) 55.2 6 11.9 54.8 6 10.3 0.72BMI (kg/m2) 33.9 6 7.5 35.4 6 7.7 0.06HbA1c (% units) 9.0 6 1.5 8.9 6 1.4 0.74HbA1c (mmol/mol) 75.0 6 16.4 74.0 6 5.3 0.74Systolic BP (mmHg) 136.2 6 19.4 135.3 6 21.3 0.68Diastolic BP (mmHg) 77.0 6 10.4 78.3 6 11.3 0.22Diabetes distress2 51.9 6 32.3 59.0 6 30.5 0.03Social distress3 34.5 6 1.6 35.8 6 1.6 0.55

Categorical variables (%) % %Female 59.0 60.8 0.71White race4 98.5 98.0 0.69Hispanic ethnicity 100.0 100.0 d

High diabetes distress2 50.5 62.9 0.01Major depression5 41.2 32.7 0.09

1Based on Student t test for continuous variables and Fisher exact test for categorical variables.2Measured using the PAID questionnaire, scored from 0 to 100, with higher scores indicative ofgreater diabetes distress; a score of.50 is indicative of high diabetes distress. 3Measured usingTAPS, scored from 0 to 100, with higher scores indicative of greater social distress. 4Remainingpatients self-identified as black/African American. 5Measured using the Patient HealthQuestionnaire nine-item depression measure; patients endorsing five or more items arecategorized as having major depression

Table 2—Clinical and psychosocial outcomes by IC

Usual diabetescare, n = 200(mean 6 SE)

Interventioncondition, n = 199

(mean 6 SE)P

value1

Clinical outcomesBMI (kg/m2) 35.0 6 0.1 34.9 6 0.1 0.50HbA1c (% units) 9.2 6 0.10 8.4 6 0.10 ,0.001HbA1c (mmol/mol) 77.0 6 1.1 68.0 6 1.1 ,0.001Systolic BP (mmHg) 137.0 6 1.3 137.2 6 1.3 0.93Diastolic BP (mmHg) 76.9 6 0.7 77.5 6 0.7 0.54

Psychosocial outcomesDiabetes distress2 48.3 6 2.0 40.4 6 2.1 ,0.01Social distress3 32.2 6 1.3 27.2 6 1.4 ,0.01

1Adjusted P values based on linear regression; models are adjusted for baseline values, with noadditional variables retained in these final models. 2Measured using the PAID questionnaire,scored from 0 to 100, with higher scores indicative of greater diabetes distress; a score of.50 isindicative of high diabetes distress. 3Measured using TAPS, scored from 0 to 100, with higherscores indicative of greater social distress.

care.diabetesjournals.org Welch and Associates 5

(16,31,32). Improvement in depressionstatus was seen among patients in bothstudy conditions (;40% of those screen-ing positive for major depression at base-line were subsequently in remission atfollow-up) but with no statistically signif-icant difference found between the con-ditions at follow-up.These results for psychosocial out-

comes provide empirical support forthe value of systematically assessingand actively managing T2D patientswho report diabetes-related psychoso-cial challenges, as has been recommen-ded in prior reviews (7,33). It is notablethat the Institute of Medicine has re-cently recommended that patient-reported assessments capturing apatient’s experience of illness shouldbe routinely incorporated into theEMRs, including emotional distress anddepression (34). The diabetes dash-board thus provides a strategy for pri-mary care clinics to meet these newrecommendations, with modificationsand updates over time, as appropriate.We explored the effect of interven-

tion treatment dose on outcomes inpoststudy sensitivity analyses and foundthat greater exposure produced greaterclinical benefit for HbA1c, diabetes dis-tress, and social distress. Future studiescould therefore consider the implemen-tation of practical strategies to enhancepatient engagement over the full courseof the intervention. For example, recentevidence supports the value of integrat-ing community health workers into thediabetes team to improve patientengagement (35,36). Also, the replace-ment of some face-to-face visits deliveredin the clinic with low-cost telehealthstrategies, including brief telephone callscombinedwith remote homemonitoringof diabetes vital signs and medicationadherence, may improve patient engage-ment and access to care among urbanpoor T2D patient groups, and may alsoovercome common barriers to regularclinic attendance, including lack of reliabletransportation, adverse weather, andcompeting family and work demands.There were several strengths of the

study, including a high patient retentionrate (88.0%) in the research follow-upvisits that involved use of a bicultural,bilingual research team as well as strongpatient participation in the interventionprogram (i.e., 78.8% of patients at-tended three visits and 48.2% attended

all five) that similarly involved use of abilingual, bicultural clinical team.

There were several weaknesses of thestudy, including our inability to track thefrequency and content of UDC clinic vis-its that could have provided a more ac-curate description of the study controlgroup and allowed adjustments for anypotential differences between studyconditions in terms of exposure to treat-ment (e.g., number of individual patienteducation sessions during the studytime period). Future research could alsoextend our outcome tracking to include aformal assessment of BP and blood lipidlevels over time and also explore differ-ences in diabetes medication manage-ment by providers taking part in theintervention and control conditionsusing a validated research protocol tocapture the necessary granularity and ac-curacy of the structured information thatwould be needed for this future goal.

It is notable that our diabetes dash-board was used as a stand-alone clinicalapplication by the diabetes team, withthe application hosted on a secureserver separate from the clinic’s EMR.As is the case for any new CDS tool,wider adoption of our diabetes dash-board will require the provision of clinicleadership support, adequate providerand support staff training and their in-put to allow successful adaptation to lo-cal clinical care processes, as well asavailability of sufficient IT and changemanagement support similar to thatseen for the current national EMR roll-out and meaningful use of patient dataas part of the HITECH Act (8).

In conclusion, thediabetes dashboard in-tervention significantly reduced diabetes-relatedmedical andpsychosocial disparitiesamong Latinoswith poorly controlled T2Dcompared with a similar diabetes teamcondition without access to the diabetesdashboard. The use of a disease-specificclinical dashboard that addresses medicaland psychosocial aspects of T2D treat-ment has broad applicability to othercommon chronic diseases that alsorequire a focus on patient-centered, com-prehensive, and efficient team care.

Acknowledgments. The authors thankGbenga Ogedegbe of NYU Langone MedicalCenter and Ana Ronderos and Kathy Berdeciaof Holyoke Health Center for their clinicalexpertise and support during the completionof the study.

Funding. This project was supported by theNational Institute of Diabetes and Digestive andKidney Diseases, National Institutes of Health,through grant 5R01-DK-084325-04.Duality of Interest. G.W. is the Chief ScientificOfficer of Silver Fern Healthcare. No otherpotential conflicts of interest relevant to thisarticle were reported.Author Contributions. G.W. designed thestudy, oversaw the study conduct as principleinvestigator, and wrote the manuscript. S.E.Z.oversaw data collection and management, con-ducted the data analysis, and edited the man-uscript. P.S.-K. and Z.R. developed andimplemented the intervention and assisted inmanuscript development. S.-E.B. assisted inintervention planning and edited the manu-script. M.C.R. assisted in assessing interventionfidelity and edited the manuscript. R.A.G. actedas medical supervisor, conducted the training ofproviders and diabetes educators, and edited themanuscript. G.W. is the guarantor of this workand, as such, had full access to all the data in thestudy and takes responsibility for the integrity ofthe data and the accuracy of the data analysis.

References1. Centers for Disease Control and Prevention(CDC). National diabetes statistics report: esti-mates of diabetes and its burden in the UnitedStates, 2014. Atlanta, GA, U.S. Department ofHealth and Human Services, 20142. Huang ES, Basu A, O’Grady M, Capretta JC.Projecting the future diabetes population sizeand related costs for the U.S. Diabetes Care2009;32:2225–22293. Stark Casagrande S, Fradkin JE, Saydah SH,Rust KF, Cowie CC. The prevalence of meetingA1C, blood pressure, and LDL goals among peo-ple with diabetes, 1988-2010. Diabetes Care2013;36:2271–22794. O’Connor PJ, Sperl-Hillen JAM, Johnson PE,Rush WA, Biltz G. Clinical inertia and outpatientmedical errors. In Advances in Patient Safety:From Research to Implementation (Volume 2:Concepts and Methodology). Henriksen K,Battles JB, Marks ES, Lewin DI, Eds. Agencyfor Healthcare Research and Quality, Rockville,MD, 20055. Khunti K,WoldenML, Thorsted BL, AndersenM,Davies MJ. Clinical inertia in people with type 2 di-abetes: a retrospective cohort study of more than80,000 people. Diabetes Care 2013;36:3411–34176. Osborn CY, Mayberry LS, Wagner JA, WelchGW. Stressors may compromise medication ad-herence among adults with diabetes and lowsocioeconomic status. West J Nurs Res 2014;36:1091–11107. Peyrot M, Burns KK, Davies M, et al. DiabetesAttitudes Wishes and Needs 2 (DAWN2): a multi-national, multi-stakeholder study of psychosocialissues in diabetes and person-centered diabetescare. Diabetes Res Clin Pract 2013;99:174–1848. Healthcare Information Technology for Eco-nomic and Clinical Health (HITECH) Act, Title XIIIof Division A and Title IV of Division B of theAmerican Recovery and Reinvestment Act of2009 (ARRA), Pub. L. No. 111-5, 123 Stat. 226(Feb. 17, 2009), codified at 42 U.S.C. §§300jj etseq.; §§17901 et seq.9. Bodenheimer TS, Smith MD. Primary care:proposed solutions to the physician shortage

6 Diabetes Dashboard and Team Care for Latinos Diabetes Care

without training more physicians. Health Aff(Millwood) 2013;32:1881–188610. Berner ES, McGowan JJ. Use of diagnosticdecision support systems in medical education.Methods Inf Med 2010;49:412–41711. Ogunyemi O, Mukherjee S, Ani C, et al.CEDRIC: a computerized chronic disease man-agement system for urban, safety net clinics.Stud Health Technol Inform 2010;160:208–21212. Koopman RJ, Kochendorfer KM, Moore JL,et al. A diabetes dashboard and physician effi-ciency and accuracy in accessing data neededfor high-quality diabetes care. Ann Fam Med2011;9:398–40513. O’Connor PJ, Sperl-Hillen JM, Rush WA,et al. Impact of electronic health record clinicaldecision support on diabetes care: a randomizedtrial. Ann Fam Med 2011;9:12–2114. Stuckey HL, Dellasega C, Graber NJ, MaugerDT, Lendel I, Gabbay RA. Diabetes nurse casemanagement and motivational interviewingfor change (DYNAMIC): study design and base-line characteristics in the Chronic Care Modelfor type 2 diabetes. Contemp Clin Trials 2009;30:366–37415. United States Census Bureau. State &County QuickFacts for Holyoke, Massachusetts[article online], 2014. Available from http://quickfacts.census.gov/qfd/states/25/2530840.html. Accessed 28 April 201416. Welch G, Allen NA, Zagarins SE, Stamp KD,Bursell SE, Kedziora RJ. Comprehensive diabetesmanagement program for poorly controlled His-panic type 2 patients at a community healthcenter. Diabetes Educ 2011;37:680–68817. Fonda SJ, Paulsen CA, Perkins J, Kedziora RJ,Rodbard D, Bursell SE. Usability test of an internet-based informatics tool for diabetes care pro-viders: the comprehensive diabetesmanagementprogram. Diabetes Technol Ther 2008;10:16–24

18. U.S. Department of Health and Human Serv-ices. Seventh report of the joint national commit-tee on prevention, detection, evaluation, andtreatment of high blood pressure (JNC7). Be-thesda, MD, National Institutes of Health, Na-tional Heart, Lung, and Blood Institute, 2004(NIH publ. no. 04-5230)19. American Diabetes Association. Clinicalpractice recommendations. Diabetes Care2014;37(Suppl. 1):S14–S8020. Tomky D, Cypress M, Dang D, Maryniuk M,Peyrot M. AADE position statement: AADE7TMself-care behaviors. Diab Educ 2008;34:445–44921. Welch G, Zagarins SE. Assessment of diabe-tes knowledge, self-care skills, and self-care be-haviors by questionnaire. In Psychosocial Carefor People with Diabetes. 1st ed. Young-HymanD, Peyrot M, Eds. Alexandria, Virginia, AmericanDiabetes Association, 2012, p. 101–11622. WelchG, Santiago-Kelly P, CrespoZ, Zagarins S,Garb J, Kedziora R. Pilot study of a comprehensive,culturally sensitive diabetes case management in-tervention for poorly controlled Hispanic type 2patients. Presented at the 67th Scientific Sessionsof the American Diabetes Association, 22–26 June2007, Chicago, IL23. Liebman J, Heffernan D. Quality improve-ment in diabetes care using community healthworkers. Clin Diabetes 2008;26:75–7624. Liebman J, HeffernanD, Sarvela P. Establishingdiabetes self-management in a community healthcenter serving low-income Latinos. Diabetes Educ2007;33(Suppl. 6):132S–138S25. Welch G, Weinger K, Anderson B, PolonskyWH. Responsiveness of the problem areas in di-abetes (PAID) questionnaire. Diabet Med 2003;20:69–7226. Welch G, Schwartz CE, Santiago-Kelly P,Garb J, Shayne R, Bode R. Disease-related emo-tional distress of Hispanic and non-Hispanic

type 2 diabetes patients. Ethn Dis 2007;17:541–54727. Welch GW, Osborne CY, Rothberg M,Zagarins SE, White R, Friderici J. Level of socialdistress among Mexican and Puerto RicanAmerican diabetes patients and relationship todiabetes control. Diabetes 2011;60:A573–A58128. Kroenke K, Spitzer RL. The PHQ-9: A newdepression diagnostic and severity measure.Psychiatr Ann 2002;32:1–729. StataCorp. Stata Statistical Software: Re-lease 12. College Station, TX, StataCorp, 201130. American Diabetes Association. Implica-tions of the United Kingdom prospective diabe-tes study. Diabetes Care 1998;21:2180–218431. Fisher L, Mullan JT, Arean P, Glasgow RE,Hessler D, Masharani U. Diabetes distress butnot clinical depression or depressive symptomsis associated with glycemic control in bothcross-sectional and longitudinal analyses. Dia-betes Care 2010;33:23–2832. Rosal MC, Ockene IS, Restrepo A, et al. Ran-domized trial of a literacy-sensitive, culturallytailored diabetes self-management interven-tion for low-income latinos: latinos en control.Diabetes Care 2011;34:838–84433. Marrero DG, Ard J, Delamater AM, et al.Twenty-first century behavioralmedicine: a con-text for empowering clinicians and patients withdiabetes: a consensus report. Diabetes Care2013;36:463–47034. Institute of Medicine. Capturing social andbehavioral domains in electronic health records:Phase 1. Washington, D.C., The National Acad-emies Press, 201435. Rosenthal EL, Brownstein JN, Rush CH, et al.Community health workers: part of the solution.Health Aff (Millwood) 2010;29:1338–134236. ShahM, Kaselitz E, HeislerM. The role of com-munity health workers in diabetes: update on cur-rent literature. Curr Diab Rep 2013;13:163–171

care.diabetesjournals.org Welch and Associates 7