Objective: To systematically evaluate and synthesize publishedevidence regarding the effect of disease management programs forpatients with diabetes mellitus on processes and outcomes of care.
Study Design: Systematic literature review and meta-analysis.
Patients and Methods: Computerized databases were searchedfor English-language controlled studies assessing the effect of diabetesdisease management programs published from 1987 to 2001.Two reviewers extracted study data using a structured abstractionform. Pooled estimates of program effects on glycated hemoglobinwere calculated using an empirical Bayes model.
Results: The pooled estimate of program effects on glycatedhemoglobin was a 0.5-percentage point reduction (95% confidenceinterval, 0.3 to 0.6 percentage points), a modest but significantimprovement. Evidence also supports program benefits inimproving screening for retinopathy and foot lesions.
Conclusions: Diabetes disease management programs canimprove glycemic control to a modest extent and can increasescreening for retinopathy and foot complications. Further effortswill be required to create more effective disease management programsfor patients with diabetes mellitus.
(Am J Manag Care. 2005;11:242-250)
Diabetes mellitus creates a significant clinical andeconomic burden on society.1-3 For 1998, directand indirect costs of diabetes mellitus and itscomplications are estimated at $98.2 billion in theUnited States.1,2 Standards of care for diabetes mellitushave been broadly disseminated since the 1980s2 in thebelief that improved processes of care can improvepatient outcomes,4,5 but primary care providers havebeen slow to implement patient care guidelines and recommendations.6,7 Several barriers to guideline adherenceand implementation have been identified,8,9including the perception that type 2 diabetes mellitus isnot serious, that aggressive treatment cannot forestallcomplications,10,11 that guidelines are not flexibleenough to be useful in patient care,8 and that patientswith diabetes mellitus are unwilling to make neededlifestyle changes.12-14
Given their numbers and economic effect, improvingcare for patients with diabetes mellitus has become apriority for health plans, payers, and patients. The numberand complexity of services required to manage suchpatients in accord with accepted guidelines have madediabetes mellitus the target of multiple disease managementefforts, as well as targeted efforts involving professionaleducation and case management.
Other reviews have addressed the effectiveness ofprofessional education and structure of care15 and casemanagement16 in improving patient outcomes. Theobjective of this study is to systematically evaluate thepublished literature on the effectiveness of diabetes diseasemanagement programs—defined as structured,multifaceted, systematic approaches to care—onglycemic control and other relevant outcomes.
team, patient care planning, primary nursing care,
case management, critical pathways, primary health
care, continuity of patient care, guidelines, practice
guidelines, disease management, comprehensive
health care, ambulatory care,
A systematic review of the medical literature wasperformed with the assistance of an expert librarian,using the computerized bibliographic databases MEDLINE,HealthSTAR, and Cochrane Database of SystematicReviews to identify assessments of diseasemanagement programs in different areas. English-languagestudies published between January 1987 andJune 2001 were identified and reviewed, with the 1987date reflecting the approximate beginning of widespreadinterest in disease management. Search terms includedthe following Medical Subject Headings: and the title words and Ahand search of bibliographies from relevant articles andreviews was also conducted, and the opinions fromexpert physicians and researchers in the field weresolicited to identify other references.
We defined disease management according to a previouslypublished definition used by Ellrodt et al.17Programs that used a systematic approach to care andincluded more than 1 intervention component (anappendix listing the classification of interventions isavailable from the authors upon request) were consideredas using disease management. Trials of pharmacologicalagents were excluded. A systematic approach tocare was defined as inclusion of any of the followingcomponents: guidelines, protocols, algorithms, careplans, or systematic patient or provider education programs.Unstructured professional education and casemanagement programs were not included.
Specific inclusion and exclusion criteria were developedfor reviewing titles, abstracts, and articles. Tworeviewers (KK, EB) trained in health services researchand in the principles of critical appraisal independentlyreviewed a 10% sample of randomly selected studies,with any discrepancies resolved by consensus. Theremaining studies were distributed among reviewersonly when a sufficient level of agreement was achieved(κ>0.7).
Titles were rejected if they did not deal with adultpatients or were reviews, case reports, editorials, letters,or meeting abstracts. Abstracts were rejected if they didnot report any objective measurements of disease management,referred to clinical trials comparing singlepharmacological agents or diagnostic procedures, or didnot use a systematic approach to care. Studies wererejected if they lacked sufficient information to measurethe effect of an intervention on at least 1 outcome ofinterest and its variance. To measure the effect of agiven intervention, studies should have used an experimentalor quasi-experimental study design meeting thecriteria established by the Cochrane Effective Practiceand Organization of Care Group and should haveincluded an appropriate comparison group. Acceptabledesigns included randomized clinical trials and controlledbefore-and-after studies (studies with a parallelnonrandomized comparison group, with baseline andfollow-up assessments of both groups).18
From the pool of accepted disease managementassessments, those aimed at the management of diabetesmellitus were selected. Study information was abstractedfrom the accepted articles using a standardized dataabstraction form and included the following: studydesign, setting, intervention strategies, and outcomesof interest. All components of the disease managementintervention were identified using classifications basedon definitions taken from the Cochrane Collaborationfor providers, patients, and organizations.18 Outcomesfrom the following 12 domains were collected: glycatedhemoglobin (GHb) levels, serum lipids (low-densitylipoprotein [LDL] cholesterol, high-density lipoprotein[HDL] cholesterol, and total cholesterol), systolic bloodpressure, hospital admissions, mean number of GHbtests per patient, screening for retinopathy, screening fornephropathy, foot screening, foot self-care, patientknowledge, self-reported health status, and patient satisfaction.If a study reported outcome measures at multipletime points for a single domain, results from thelongest follow-up were used. If the serum LDL cholesterollevel was reported, it was used instead of the totalcholesterol level. Systolic blood pressure was abstractedas it is a better predictor of cardiovascular events thandiastolic blood pressure.19 One observation per outcomewas abstracted for each intervention arm, so studiesevaluating more than 1 intervention could contributemore than 1 observation per program.
To account for baseline differences in GHb levelsbetween groups, the mean changes over baseline werecompared. Given the fact that certain data elementsmay not be provided in the identified studies, we createda list of assumptions a priori to facilitate the meta-analysis.In the 3 instances in which baseline meanswere not reported, the assumption that the treatmentand control group baseline means were equal was used.Variances of changes were rarely reported. In thoseinstances lacking reported variances, the variance ofthe change was assumed to be equal to one half of thesum of the variances of the baseline and follow-up measures.In the absence of baseline variances, it wasassumed that baseline variances in the treatment andcontrol groups were equal to the follow-up variance inthe control group.
A test for homogeneity was performed using the χ2test. The more conservative random-effects method(the empirical Bayes method proposed by Hedges andOlkin20) was used to pool the estimated program effectson GHb levels.21 Results are reported as the pooled differencesin change in GHb level in treated versus controlsubjects. Although there are subtle differencesbetween the various species of GHb, all refer to anddescribe the modification of hemoglobin by a sugar(ie, glucose).22-26
As part of an exploratory data analysis, funnel plotswere constructed by plotting each program's estimatedeffect on change in GHb against the inverse of its variance to assess potential publication bias.27,28 We usedthe trim and fill method29 to determine if adjustmentsfor publication bias were warranted and, if so, to adjustthe estimates. Stata 8 (StataCorp LP, College Station,Tex) was used to perform this analysis.
The initial search strategy identified 16 917 references.A total of 2963 titles were accepted for furtherscreening, and 581 abstracts met inclusion criteria.Eighty-five percent (n = 493) of the accepted abstractsfailed to meet inclusion criteria when the articles werereviewed. Bibliographic hand searches and expert consultationyielded an additional 53 articles for review, ofwhich 16 were accepted. Twenty-four studies30-52 dealtwith diabetes mellitus.
Of 24 studies meeting our inclusion criteria, 19 studieswere randomized clinical trials and 5 studies33,39,41,42,44were nonrandomized controlled studies.Five31,37,47,50,52 of 19 randomized trials used a clusterrandomization scheme for treatment allocation. Theaggregate sample size for the 24 studies was 6421patients, and the total number of patients studied ineach report ranged from 38 to 1939. More than half(15/24) of the studies were conducted in the UnitedStates, 4 were carried out in the United Kingdom, andthe remaining 5 were conducted in Israel, Argentina,Austria, and the Netherlands. Study duration rangedfrom 3 months38 to 30 months,46 while the duration ofthe disease management intervention ranged from severaldays37 to 30 months.46 Key characteristics of identifiedstudies are available from the author.
Most (20/24) studies were funded by research grantsprovided by governmental offices, academia, or researchfoundations. Four studies38-40,42 received additional supportfrom pharmaceutical companies. Four studies34-36,41assessed disease management costs in relation to a primarycare or health maintenance organization setting,while cost per patient43,45 and cost of program implementation31,50were assessed in 2 studies each. Differentinterventions were used, ranging from patient educationsessions to centrally administered provider remindersto integrated multidisciplinary team approaches.
Twenty studies,30,33-46,48-50,52,53 contributing 24 observationsand involving 3720 patients, were included inthe meta-analysis to evaluate the effect of disease managementon GHb level. Of these 24 treatment-controlcomparisons, 9 (38%)35,36,39-41,44,49,52,53 reported statisticallysignificant differences favoring the treatmentgroup. The rest showed no significant differences; inone52 of those studies, a treatment arm experienced anincrease in GHb level relative to the comparison arm.Significant heterogeneity in results was confirmed bythe results of our test for homogeneity (<.001).Overall, the pooled result (using a random-effectsmodel) showed that disease management programsresulted in a statistically significant reduction in GHblevel (mean reduction, 0.5 percentage point; 95% confidenceinterval [CI], 0.3 to 0.6 percentage points). Wealso calculated results based on geographic location (USvs non-US studies). The mean reduction in GHb levelfor US observations (n = 16) was 0.6 percentage point(95% CI, 0.4 to 0.9 percentage points), while the meanreduction for non-US observations (n = 8) was 0.32 percentagepoint (95% CI, 0.01 to 0.54 percentage points).However, the small number of trials in countries outsidethe United States limits conclusions.
Programs associated with the greatest estimatedchanges in GHb levels include 1 program involving pharmacistscounseling patients and medication adjustment40and 2 programs involving combined physicianand patient interventions.49,52 Small sample size limitsgeneralizations with regard to effects of different types ofprograms. A study by Vinicor et al,52 including arms withstructured physician education alone and in combinationwith patient education, demonstrated improvedresults with the combination program. The Figure is aforest plot displaying estimated program effects on GHb.
Frequency of Glycemic Monitoring
Four studies,30,41,45,46 involving 958 patients, assessedthe effect of disease management on the frequency ofglycemic monitoring (Table 1). These measured themean number of GHb tests per patient, percentage ofpatients who had a GHb test, and mean number of self-monitoredblood glucose levels. Two studies45,46 reportedsignificant increases in the frequency of GHb testsperformed in intervention patients. A third study41showed no change in the percentage of patients whoreceived at least 1 GHb test. The study by Piette et al,30involving 292 subjects, found that program patientswere more likely to perform home glucose monitoringthan control patients.
Screening for Retinopathy
Three studies,30,45,46 with 708 patients, investigatedthe effect of disease management on the frequency ofretinal examinations (Table 1). Two studies45,46 evaluatedthe mean number of retinal examinations performedper patient, while the third study30 examined the percentageof patients receiving an ophthalmologic examination.Two studies45,46 showed a small but statisticallysignificant improvement in the frequency of retinalexaminations. One study30 reported that a slightly higherproportion of patients in the intervention groupreceived an eye examination.
Screening for Nephropathy
Three studies,37,40,46 involving 447 patients, assessedprogram effects on screening for nephropathy (Table 1).One study37 reported that slightly more patients in theintervention group were screened for nephropathy.Another study46 showed that the mean number ofscreening tests for nephropathy was similar in the interventionand control groups. The last study40 demonstrateda statistically significant decrease in serumcreatinine levels in the intervention patients.
Foot Screening and Podiatrist Referral
Three studies,31,45,47 involving 1912 patients, examinedthe effects on the frequency of foot screening(Table 1). Of these, 2 studies31,47 evaluated the percentageof patients receiving foot examinations, whileanother study45 evaluated the mean number of footexaminations performed per patient. All 3 studiesshowed that programs increased the number of footexaminations performed, with the study by Naji et al45demonstrating a statistically significant improvementwith disease management.
Two studies46,47 with 547 patients evaluated the proportionof patients referred to a podiatrist. One study47showed an increase in referrals from primary carephysicians to a podiatrist, while the other study46reported fewer referrals in intervention patients.
Three studies,30,35,47 with 817 patients, evaluatedpatient foot self-care (Table 1). One study30 found thatthe control group patients examined their feet significantlymore frequently than those in the treatmentgroup. Another study35 showed that patients in theintervention group examined their feet more often thanthose in the control group. The third study47 reportedthat intervention patients practiced more appropriatefoot care behaviors compared with control patients.This difference was statistically significant.
Systolic Blood Pressure
Five studies37,39,44,45,52 (7 observations), with 1239patients, assessed program effects on systolic blood pressure(Table 2). Control group patients from the Vinicorstudy were accounted for once to avoid double counting.52 Of these 7 assessments, 5 reported a decrease insystolic blood pressure in the treatment group.37,44,52However, only 1 of these yielded a statistically significantreduction.52 Two showed a slight improvement in thecontrol group relative to the intervention group, but the resultswere not statistically significant.39,45
Low-density Lipoprotein Cholesterolor Total Cholesterol Levels
Eight studies34,36,37,39,43,44,49,51 (9 observations), with1181 patients, assessed program effects on LDL cholesterol or total cholesterol levels (Table 3). Control grouppatients from the D'Eramo-Melkus study were accountedfor once to avoid double counting.49 Three observationsevaluated LDL cholesterol levels, and 6 evaluatedtotal cholesterol levels. One study34 reported a significantreduction in LDL cholesterol levels in the treatmentgroup. Another study36 reported a larger, nonsignificantreduction in LDL cholesterol in the control group; however,the reduction did not reach statistical significance.
High-density Lipoprotein Cholesterol Levels
Five studies,34,36,39,43,51 with 822 patients, examinedprogram effects on HDL cholesterol levels (Table 3).One study51 showed a statistically significant increase inHDL cholesterol levels in program subjects. Two studies36,43reported nonsignificant improvements in HDLcholesterol levels in the treatment groups.
Other Pertinent Outcomes
The effects of disease management on quality of life(self-reported health status, physical functioning, andpatient satisfaction), healthcare use (emergency departmentvisits, hospital admissions, and visits to primarycare physicians), and condition-specific knowledge(provider and patient) were also evaluated (data notshown). Although results varied across studies, positivetrends supporting disease management were observed.
Estimates of program effects on GHb were distributedasymmetrically, suggesting publication bias.However, trim and fill analysis did not support adjustmentof the estimates.
Results of this analysis suggest that disease managementprograms on average have a modest, but clinicallyand statistically significant, effect on glycemic controlin patients with diabetes mellitus (pooled estimate,0.5-percentage point reduction; 95% CI, 0.3 to 0.6 percentagepoints). By comparison, the Diabetes Controland Complications Trial,54 an intensive glycemic controlprogram for type 1 patients, demonstrated a mean2-percentage point reduction in GHb, while the UKProspective Diabetes Study,55 an intensive program fornewly diagnosed type 2 patients, demonstrated a0.9-percentage point reduction.
The Diabetes Control and Complications Trial54 andthe UK Prospective Diabetes Study55 showed strongrelationships between the risks of developing microvascularcomplications and glycemic levels over time.Furthermore, the UK Prospective Diabetes Study establishedthat, for every percentage point decrease inhemoglobin A1c (eg, 9 to 8 percentage points), there wasa 35% reduction in the incidence of microvascular complications.Application of this result to the pooled estimateof improvement in GHb levels in the studies wereviewed implies a reduction in the incidence ofmicrovascular complications of approximately 15%.Given the prevalence of diabetes mellitus in the US populationand the poor degree of glycemic control in manypatients,56 this degree of improvement would be importantif applied to all patients with diabetes mellitus.
In addition, some programs improved other patientoutcomes (screening for retinopathy and foot complications,systolic blood pressure, and serum lipids).However, disease management did not affect screeningfor nephropathy, hospital admissions, health status, orpatient knowledge. The lack of demonstrated effect onthese outcomes may, in part, be because of the smallnumber of studies evaluating them.
The included studies varied in terms of the numberof interventions used, study setting, and outcome measuresreported. For example, 20 (83%) of the 24 studiesevaluated program effects on glycemic control, whereasonly one study46 evaluated effects on hospital admissionrates and another study36 evaluated effects on mortality.Interestingly, multiple interventions were used in allstudies. One study46 assessed the effectiveness of acombination of 8 disease management interventions,while another study38 examined the effect of 2 interventionson processes of care and patient outcomes.
For the purposes of our analysis, disease managementwas defined according to a previously published definition.17However, other definitions may have led to theinclusion of different studies and yielded different results.Therefore, it is possible that these findings depend on thespecific operational definition of disease management.
Significant heterogeneity exists among the studies inthis analysis with respect to effects on glycemic control,as the results of our test for homogeneity suggest(< .001). These findings may also be affected by publicationbias, as suggested by the funnel plot constructedfrom our exploratory data analysis. Studies, particularlysmall ones, with negative results may be less likely to bepublished, especially if they contradict prevailing opinionin the literature and in the professional community.More likely, there is incomplete representation of diseasemanagement program evaluations in journal publications;a great deal of disease management activity occursin health plans, sometimes implemented by third-partydisease management vendors for accreditation from theNational Committee for Quality Assurance and similarorganizations. These organizations may lack the fundingor expertise to properly evaluate and publish their findings,or they may choose not to publicize their results.
The argument can be made that patients' motivationto participate can affect outcomes. To mitigate this concern,we included only studies using a parallel comparisongroup design. Of 24 estimates of program effects onGHb levels, 5 were derived from studies33,39,41,42,44 usinga quasi-experimental study design (ie, controlledbefore-and-after studies) according to the CochraneCollaboration's criteria for acceptable study designs.These observations produced a mean reduction in GHblevels of 0.7 percentage point (95% CI, 0.6 to 0.8 percentagepoints). In comparison, observations from randomizedclinical trials (n = 19) produced a meanreduction of 0.4 percentage point (95% CI, 0.2 to 0.5percentage points). Although a larger reduction wasobserved for quasi-experimental studies, both reductionsattained statistical significance. While selection ofmore motivated patients into intervention groups mightbias results to some extent, the magnitude of any bias inour analysis is likely to be limited given that mostresults are from randomized trials.
Finally, these findings may reflect the fact that themethodology for implementing disease management andfor measuring its effect is in its infancy. As the field ofdisease management continues to evolve and mature,and as measurement tools are refined, views on the effectivenessof these programs are likely to change as well.
Although this analysis reveals that disease managementholds the potential to improve long-term outcomesbecause of better glycemic control and providercompliance with recommended standards, the overalleffect on glycemic control appears modest. Providercompliance with treatment recommendations improved,but outcomes such as mortality, hospitalization,patient satisfaction, patient knowledge, and patientcompliance showed no significant improvements. In ourprevious assessment of which interventions were effectivein disease management programs,57 we found thatprograms incorporating provider education, providerfeedback, provider reminders, patient education,patient reminders, and patient financial incentives wereassociated with improvements in provider adherence toguidelines and patient disease control. Further researchis needed to understand which program characteristics(eg, type and intensity) are most effective and how thescience of disease management may be improved andthe programs refined to optimize outcomes for patientswith diabetes mellitus.
From Phar, LLC (KK), and Zynx Health Incorporated, a Subsidiary of the HearstCorporation, and the Departments of Medicine and Health Services Research, Cedars-SinaiMedical Center (JJO, SRW), Los Angeles, and Cerner Health Insights, Beverly Hills (EB,ADG), Calif; TAP Pharmaceutical Products Inc, Lake Forest, Ill (JMH); and Duke ClinicalResearch Institute, Duke University, Durham, NC (VH). Dr Ofman is now with Amgen Inc,Thousand Oaks, Calif.
This investigator-initiated work was partially supported by a research grant from TAPPharmaceutical Products Inc. Dr Ofman secured funding for this project from the sponsor.Mr Henning was a paid employee of the sponsor at the time the research was conducted.He retired from TAP Pharmaceutical Products Inc effective March 31, 2003. Dr Hasselbladwas a paid consultant and provided statistical expertise on the project. Dr Knight was anemployee of Cerner Health Insights at the time the research was conducted.
Address correspondence to: Kevin Knight, MD, MPH, Phar, LLC, 1950 SawtelleBoulevard, Suite 280, Los Angeles, CA 90025. E-mail: email@example.com.
Studies on the Costs
1. Songer TJ, Ettaro L; Economics of Diabetes Project Panel. Atlanta, Ga: Centers for Disease Control and Prevention; 1998.
2. American Diabetes Association. Economic consequences of diabetes mellitus inthe US in 1997. 1998;21:296-309.
J Clin Endocrinol Metab.
3. Rubin RJ, Altman WM, Mendelson DN. Health care expenditures for peoplewith diabetes mellitus, 1992. 1994;78:809A-809F.
4. Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitusand its complications: part 1: diagnosis and classification of diabetes mellitusprovisional report of a WHO consultation. 1998;15:539-553.
5. Conseil Superieur du Diabete. Diabetes care and research in Europe: the St.Vincent Declaration Action Program [in French]. 1992;18:334-377.
6. Kenny SJ, Smith PJ, Goldschmid MG, Newman JM, Herman WH. Survey ofphysician practice behaviors related to diabetes mellitus in the US: physicianadherence to consensus recommendations. 1993;16:1507-1510.
7. Stolar MW; Endocrine Fellows Foundation Study Group. Clinical managementof the NIDDM patient: impact of the American Diabetes Association practiceguidelines, 1985-1993. 1995;18:701-707.
J Fam Pract.
8. Helseth LD, Susman JL, Crabtree BF, O'Connor PJ. Primary care physicians'perceptions of diabetes management: a balancing act. 1999;48:37-42.
9. O'Connor PJ. From blame to understanding: moving diabetes care forward. 1998;46:205-206.
10. Anderson RM, Donnelly MB, Dedrick RF, Gressard CP. The attitudes of nurses,dietitians, and physicians toward diabetes. 1991;17:261-268.
11. Anderson RM, Donnelly MB, Davis WK. Controversial beliefs about diabetesand its care. 1992;15:859-863.
12. Golin CE, DiMatteo MR, Gelberg L. The role of patient participation in thedoctor visit: implications for adherence to diabetes care. 1996;19:1153-1164.
13. Wing RR, Epstein LH, Nowalk MP, Scott N, Koeske R. Compliance to selfmonitoringof blood glucose: a marked-item technique compared with self-report.1985;8:456-460.
14. Christensen NK, Terry RD, Wyatt S, Pichert JW, Lorenz RA. Quantitativeassessment of dietary adherence in patients with insulin-dependent diabetes mellitus.1983;6:245-250.
Cochrane Database Syst Rev.
15. Renders CM, Valk GD, Griffin S, et al. Interventions to improve the managementof diabetes mellitus in primary care, outpatient and community settings.2001;CD001481.
Am J Prev Med.
16. Norris SL, Nichols PJ, Caspersen CJ, et al. The effectiveness of disease andcase management for people with diabetes: a systematic review. 2002;22:15-38.
17. Ellrodt G, Cook DJ, Lee J, et al. Evidence-based disease management. 1997;278:1687-1692.
18. Cochrane Effective Practice and Organisation of Care Group (EPOC)Web site. Available at: http://www.epoc.uottawa.ca. Accessed July 19, 2000.
Arch Intern Med.
19. Psaty BM, Furberg CD, Kuller LH, et al. Association between blood pressurelevel and the risk of myocardial infarction, stroke, and total mortality: the cardiovascularhealth study. 2001;161:1183-1192.
Statistical Methods for Meta-analysis.
20. Hedges LV, Olkin I. Orlando, Fla:Academic Press; 1985.
Control Clin Trials.
21. DerSimonian R, Laird N. Meta-analysis in clinical trials. 1986;7:177-188.
22. Goldstein DE, Little RR, Wiedmeyer HM, England JD, McKenzie EM.Glycated hemoglobin: methodologies and clinical applications. 1986;32 (suppl):B64-B70.
23. Jovanovic L, Peterson CM. The clinical utility of glycosylated hemoglobin. 1981;70:331-338.
24. Bunn HF. Nonenzymatic glycosylation of protein: relevance to diabetes. 1981;70:325-330.
25. Baynes JW, Bunn HF, Goldstein D, et al. National Diabetes Data Group:report of the expert committee on glucosylated hemoglobin. 1984;7:602-606.
The Diabetes Annual.
26. Goldstein DE, Little RR, Wiedmeyer HM, England JD, Rohlfing CL. Glycatedhaemoglobin estimation in the 1990's: a review of assay methods and clinicalinterpretation. In: Marshall SM, Home PD, eds. Amsterdam,the Netherlands: Elsevier Science BV; 1994:193-212.
27. Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detectedby a simple, graphical test. 1997;315:629-634.
Summing Up: The Science of Reviewing Research.
28. Light RJ, Pillemer DB. Cambridge, Mass: Harvard University Press; 1984.
29. Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testingand adjusting for publication bias in meta-analysis. 2000;56:455-463.
30. Piette JD, Weinberger M, Kraemer FB, McPhee SJ. Impact of automated callswith nurse follow-up on diabetes treatment outcomes in a Department of VeteransAffairs Health Care System: a randomized controlled trial. 2001;24:202-208.
31. Donohoe ME, Fletton JA, Hook A, et al. Improving foot care for people withdiabetes mellitus: a randomized controlled trial of an integrated care approach.2000;17:581-587.
32. Piette JD, Weinberger M, McPhee SJ. The effect of automated calls with telephonenurse follow-up on patient-centered outcomes of diabetes care: a randomized,controlled trial. 2000;38:218-230.
33. Benjamin EM, Schneider MS, Hinchey KT. Implementing practice guidelinesfor diabetes care using problem-based learning: a prospective controlled trial usingfirm systems. 1999;22:1672-1678.
South Med J.
34. Ridgeway NA, Harvill DR, Harvill LM, et al. Improved control of type 2 diabetesmellitus: a practical education/behavior modification program in a primarycare clinic. 1999;92:667-672.
35. Sadur CN, Moline N, Costa M, et al. Diabetes management in a health maintenanceorganization: efficacy of care management using cluster visits. 1999;22:2011-2017.
Ann Intern Med.
36. Aubert RE, Herman WH, Waters J, et al. Nurse case management to improveglycemic control in diabetic patients in a health maintenance organization: a randomized,controlled trial. 1998;129:605-612.
37. Kinmonth AL, Woodcock A, Griffin S, Spiegal N, Campbell MJ; DiabetesCare From Diagnosis Research Team. Randomised controlled trial of patient centredcare of diabetes in general practice: impact on current wellbeing and futuredisease risk. 1998;317:1202-1208.
J Am Diet Assoc.
38. Kulkarni K, Castle G, Gregory R, et al; Diabetes Care and Education DieteticPractice Group. Nutrition Practice Guidelines for Type 1 Diabetes Mellitus positivelyaffect dietitian practices and patient outcomes. 1998;98:62-70.
39. de Sonnaville JJ, Bouma M, Colly LP, Deville W, Wijkel D, Heine RJ.Sustained good glycaemic control in NIDDM patients by implementation of structuredcare in general practice: 2-year follow-up study. 1997;40:1334-1340.
40. Jaber LA, Halapy H, Fernet M, Tummalapalli S, Diwakaran H. Evaluation ofa pharmaceutical care model on diabetes management. 1996;30:238-243.
Arch Fam Med.
41. O'Connor PJ, Rush WA, Peterson J, et al. Continuous quality improvementcan improve glycemic control for HMO patients with diabetes. 1996;5:502-506.
42. Domenech MI, Assad D, Mazzei ME, Kronsbein P, Gagliardino JJ. Evaluationof the effectiveness of an ambulatory teaching/treatment programme fornon-insulin dependent (type 2) diabetic patients. 1995;32:143-147.
J Am Diet Assoc.
43. Franz MJ, Monk A, Barry B, et al. Effectiveness of medical nutrition therapyprovided by dietitians in the management of non-insulin-dependent diabetes mellitus:a randomized, controlled clinical trial. 1995;95:1009-1017.
44. Pieber TR, Holler A, Siebenhofer A, et al. Evaluation of a structured teachingand treatment programme for type 2 diabetes in general practice in a rural area ofAustria. 1995;12:349-354.
45. Naji S, Cameron I, Russell I, et al; Diabetes Integrated Care EvaluationTeam. Integrated care for diabetes: clinical, psychosocial, and economic evaluation.1994;308:1208-1212.
46. Hurwitz B, Goodman C, Yudkin J. Prompting the clinical care of non-insulindependent (type II) diabetic patients in an inner city area: one model of communitycare. 1993;306:624-630.
Ann Intern Med.
47. Litzelman DK, Slemenda CW, Langefeld CD, et al. Reduction of lowerextremity clinical abnormalities in patients with non-insulin-dependent diabetesmellitus: a randomized, controlled trial. 1993;119:36-41.
J Gen Intern Med.
48. Weinberger M, Kirkman MS, Samsa GP, et al. A nurse-coordinated interventionfor primary care patients with non-insulin-dependent diabetes mellitus:impact on glycemic control and health-related quality of life. 1995;10:59-66.
49. D'Eramo-Melkus GA, Wylie-Rosett J, Hagan JA. Metabolic impact of educationin NIDDM. 1992;15:864-869.
50. de Weerdt I, Visser AP, Kok GJ, de Weerdt O, van der Veen EA.Randomized controlled multicentre evaluation of an education programme forinsulin-treated diabetic patients: effects on metabolic control, quality of life, andcosts of therapy. 1991;8:338-345.
51. Raz I, Soskolne V, Stein P. Influence of small-group education sessions onglucose homeostasis in NIDDM. 1988;11:67-71.
52. Vinicor F, Cohen SJ, Mazzuca SA, et al. DIABEDS: a randomized trial of theeffects of physician and/or patient education on diabetes patient outcomes. 1987;40:345-356.
53. Mazzuca SA, Moorman NH, Wheeler ML, et al. The diabetes educationstudy: a controlled trial of the effects of diabetes patient education. 1986;9:1-10.
N Engl J Med.
54. Diabetes Control and Complications Trial Research Group. The effect ofintensive treatment of diabetes on the development and progression of long-termcomplications in insulin-dependent diabetes mellitus. 1993;329:977-986.
55. UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucosecontrol with sulphonylureas or insulin compared with conventional treatment andrisk of complications in patients with type 2 diabetes (UKPDS 33) [published correctionappears in 1999;354:602]. 1998;352:837-853.
56. Harris MI, Eastman RC, Cowie CC, Flegal KM, Eberhardt MS. Racial and ethnicdifferences in glycemic control of adults with type 2 diabetes. 1999;22:403-408.
57. Weingarten SR, Henning JM, Badamgarav E, et al. Interventions used in diseasemanagement programmes for patients with chronic illness: which ones work?meta-analysis of published reports. [serial online]. 2002;325:E925. Availableat: http://bmj.bmjjournals.com/cgi/content/full/325/7370/925. Accessed August 31,2004.