Objective: To determine whether an intervention focusing clinicianattention on drug choice for hypertension treatment improvesconcordance between drug regimens and guidelines.
Study Design: Cluster-randomized controlled trial comparingan individualized intervention with a general guideline implementationin geographically diverse primary care clinics of a university-affiliated Department of Veterans Affairs healthcare system.
Methods: Participants were 36 attending physicians and nursepractitioners (16 in the general group and 20 in the individualizedgroup), with findings based on 4500 hypertensive patients. A generalguideline implementation for all clinicians, including educationabout guideline-based drug recommendations and goals foradequacy of blood pressure control, was compared with additionof a printed individualized advisory sent to clinicians at eachpatient visit, indicating whether or not the patient's antihypertensivedrug regimen was guideline concordant. We measured changefrom baseline to end point in the proportion of clinicians' patientswhose drug therapy was guideline concordant.
Results: The individualized intervention resulted in an improvementin guideline concordance more than twice that observed forthe general intervention (10.9% vs 3.8%, = 2.796, = .008).Bootstrap analysis showed that being in the individualized groupincreased the odds of concordance 1.5-fold (= .025). The proportionof patients with adequate blood pressure control increasedwithin each study group; however, the difference between groupswas not significant.
Conclusion: An individualized advisory regarding drug therapyfor hypertension given to the clinician at each patient visit wasmore effective in changing clinician prescribing behavior thanimplementation of a general guideline.
(Am J Manag Care. 2005;11:677-685)
Active steps are required to translate clinical practiceguidelines into practice.1-4 Provision of educationand feedback to clinicians based on drugprofiling of panels of patients effectively changes practicein various settings,5 including hypertension management.6 The influence of feedback can be extended byreminder messages.7 An intervention that coincideswith a clinic visit is particularly effective in improvingphysician compliance with guidelines for preventivecare.8 The consequences of implementing evidence-basedguidelines have been studied mainly for preventivecare7,9-12 rather than chronic disease management.
We studied implementation of hypertension guidelinesbecause drug prescribing for hypertension frequentlydeviates from evidence-based clinical practiceguidelines.13-17 Clinicians may not recognize their lackof guideline adherence.18 We designed an interventionto address known barriers to clinician guideline adherence,2,19-21 using education, performance feedback, recommendationsabout drug therapy for specific patientsdelivered to the clinician during the clinic visit, andrequests for a response from the clinician. Multifacetedquality improvements have been found more effectivethan single-component strategies in changing clinicianbehavior.4
We hypothesized that an intervention with informationindividualized to particular patients would increaseclinician adherence to drug-therapy guidelines in primarycare clinics of a large healthcare system.
Sites and Subjects
The study was conducted at the Department ofVeterans Affairs Palo Alto Health Care System(VAPAHCS), which has rural, suburban, and urban sitesin the San Francisco Bay and Central Valley areas ofCalifornia. All sites with established clinics at the startof the study were included: Palo Alto, San Jose,Monterey, Livermore, Menlo Park, and Stockton. Allattending physicians and nurse practitioners with primarycare clinics in which they provided direct patientcare (rather than providing only supervision of residents,which takes place at 1 site) were included.Because the intervention was provided to clinicians, werandomized by clinicians. The randomization, performedby the biostatistician (Dr Lavori) using the SPlus2000 statistical program (Insightful Corporation,Seattle, Wash), was done for all clinicians at the sametime, and was stratified by physician versus nurse practitioner.Of 42 clinicians randomized, 4 left the primarycare clinics before the intervention began and 2 moreleft shortly after, all for reasons unrelated to this study;their patients were reassigned to the 36 remaining clinicians.Participants were not informed of their studygroup assignment.
Study patients had a diagnosis of hypertension ontheir problem list, an active prescription for an antihypertensivedrug at the start of the study, and at least 1primary care clinic visit with a study clinician duringthe study period; and they were not dually followed inthe Hypertension Clinic. The 36 study clinicians had atotal of 4533 patients meeting the study criteria. Ofthese, 33 patients (19 in general intervention and 14 inindividualized intervention) died during the study period,leaving 4500 patients. At the end point, 124 patientsno longer had active prescriptions for antihypertensivedrugs; these patients were not included in the concordanceanalysis, which was based on drugs prescribed,but they were included in the blood pressure (BP)analyses.
Sixth Report of the Joint National Committee on
Prevention, Detection, Evaluation, and Treatment of
High Blood Pressure
We implemented the guidelines described in the(JNC 6),13 the version of JNC thatwas in effect at that time, and the Department ofVeterans Affairs (VA) national guidelines.22 To operationalizethese guidelines and set up measurable performancestandards, Veterans Integrated ServiceNetwork (VISN) 21, the region of VA where VAPAHCS islocated, had adopted performance standards for drugtherapy of hypertension.23
Because VISN 21 mandated a hypertension guidelinefor all patients, it was not possible to have a "usual care"comparison group, so the study was designed to compare2 interventions. The trial compared a general interventionwith an individualized intervention that supplementedthe general intervention. All clinicians receivedthe general intervention, including a printed copy of JNC6 guidelines, VA national hypertension guidelines, andthe VISN 21 performance standards; all physicians alsoreceived information about the proportion of hypertensivepatients at VAPAHCS whose drug therapy was concordantwith the performance standards. In addition, theVA national guidelines and the VISN 21 guidelines formanagement of hypertension were reviewed and discussedin small-group workshops held 2 months after thestart of the study period. In the encounter packet was aform for each scheduled clinic visit of a study patient,listing the antihypertensive drugs prescribed for thatpatient. This general intervention was designed toimprove concordance with guidelines for drug therapy ofhypertension and also to increase awareness of theimportance of adequacy of BP control.
Clinicians in the individualized-intervention groupreceived all the general interventions describedabove. In addition, their version of the encounterpacket form included an advisory about guidelineconcordance of the patient's antihypertensive drugregimen (Table 1). We asked clinicians to sign andreturn the forms, as a measure to enhance the effectivenessof the advisory.21
All clinicians received a form at each visit of a studypatient; except for the presence or absence of the advisory;these forms were similar in appearance. We sentclinic clerks 11 056 forms for separate clinic appointments,with an average of 2.5 (median = 2) forms(scheduled visits) per patient. We estimated that personneltime required to compute, print, and mail all theforms for both the general-intervention and individualized-intervention groups was approximately one halfday per week (10% full-time equivalent) for the durationof the intervention.
We also sent the clinicians in the individualized-interventiongroup drug profiles for their own panel ofpatients. These drug profiles were sent at the beginningof the study period as a "priming" strategy to makethem aware of their own rates, and at 3 months into thestudy period as a "reinforcing" strategy.20
The intervention occurred from February throughNovember 1999.
Concordance With Drug Guidelines.
The major outcomemeasure for the study was the change from baselineto end point in the proportion of patients in theclinicians' panels whose drug regimen was guidelineconcordant. This measure takes into account the baselineconcordance.
The concordance of each patient's drug regimen wascalculated as follows:
Patients were classified into the VISN 21 hypertensionpatient categories (Table 2).
The patient's active prescriptionswere analyzed todetermine whether the recommendeddrug was part of thepatient's regimen.
For example, if a category 1patient had an active prescriptionfor a thiazide diuretic, thepatient was classified as "concordant"at that time point.Each patient's category wasdetermined at baseline and atthe end point, using the same 2calendar dates for all patients.Using the same end point datefor all patients allowed us tocapture medication changesmade at the last visit.
The guideline-concordancedeterminations described abovewere done by a computer programapplied to the patients'electronic medical record datawithout regard to the studygroup assignment. Hence, theassessment was blinded togroup assignment.
Blood pressure measurements weredone as part of routine clinical care using IVAC automatedmachines (ALARIS Medical Systems, San Diego,Calif). Before the start of the study, the VISN 21 statementon accurate measurement of BP was reviewedwith nurse managers in each clinic, with the support ofnursing administration.
To obtain BP data for each study clinician, wereviewed charts of a random sample of 10 studypatients who had at least 2 clinic visits with BP measurementsat least 30 days apart (to allow time for anydrug changes to take effect), for a total of 350 patients.(If the clinician had fewer than 10 study patients, allcharts were reviewed.) All recorded BPs taken at baseline(first primary care clinic visit during study period)and outcome (last visit) were noted. Measurementswere averaged if there were multiple readings at a visit.Adequacy of BP control had been defined dichotomouslyby VA national quality assurance criteria assystolic BP less than 140 and diastolic BP less than 90.Criteria were not defined for subgroups of patients suchas diabetics; for comparability with other studies,24 weused the VA criteria. For a second sample, we analyzedthe BPs that had been entered by the clinic nurses intothe computerized BP records. Because not all clinicsites were entering these data at the time of the study,these BPs were available for only 1817 (40%) of thepatients, of which 829 (46%) were in the general-interventiongroup and 988 (54%) were in the individualized-intervention group.
We summarized each clinician's adherence to thedrug guidelines by a single measure: namely, the changein percent guideline concordance, reflecting the averagechange in concordance from the beginning to the end ofthe study period for that clinician's patient panel.Patients who had no active prescriptions for antihypertensivedrugs at the end point (124 patients) were noteligible for guideline drug concordance scoring, whichapplied only to patients prescribed antihypertensivetherapy; therefore, those patients were excluded fromthe concordance reports. Average systolic BPs and diastolicBPs for clinician panels were compared from baselineto outcome by paired tests and across interventiongroups by independent tests. The outcome measureswere analyzed in clinician-level tests, the most conservativetype of analysis. Primary analyses were done withSPSS statistical software (SPSS Inc, Chicago, Ill). As asecondary form of analysis, the "bootstrap" function inS-Plus 2000 (5000 bootstrap samples) was used toassess the concordance with drug guidelines by usingpatient-level data analyzed in a logistic regression ofposttreatment concordance of the patients' drugs on 2clinician-level variables (physician vs nurse practitioner;clinic site). We also analyzed patient-level data withthe general estimating equation25,26 in S-Plus 2000 totake account of clustering by clinician.
These were intention-to-treat analyses with per-protocolexclusions. The study protocol was approved bythe Medical Human Subjects Panel of StanfordUniversity, with waiver of individual informed consent.
At baseline, the clinicians in the 2 interventiongroups were similar with respect to panel size, clinicsite, and guideline concordance; their patients had similaraverage systolic BPs and diastolic BPs, and adequacyof BP control (Table 3 and Table 4). The clinicians' patients were distributed similarly across the patientcategories at baseline. More than half of the patientswere classified into patient category 1 (hypertensivewithout diabetes or heart failure); more than one quarterof the patients had diabetes mellitus. Guideline concordanceat baseline was generally higher for thecategories in which angiotensin-converting enzymeinhibitors were the recommended drugs (ie, patientswith diabetes and heart failure) compared with the categoriesfor which thiazides or beta-adrenergic receptorantagonists were the recommended drugs.
The mean number of primary care clinic visits duringthe study period was 2.4 for the general-interventiongroup and 2.2 (= .52) for the individualized-interventiongroup; these visits may or may not have been arranged onaccount of the patients' hypertension. Overall, 1264(28%), 1810 (40%), 1018 (23%), and 408 (9%) patientshad, respectively, 1, 2, 3, and 4 or more clinic visits. Theaverage number of antihypertensive drugs per patientper clinician at baseline was 2.0 for the general groupand 1.9 for the individualized group (= .08). At theend point, the average number of antihypertensive drugsper patient was 2.0 in both groups (= .75).
Guideline Concordance of Clinician Prescribing
Concordance with the drug therapy guidelines forhypertension improved in both study groups, withsubstantially more improvement in the individualized-interventiongroup compared with the general-interventiongroup. Concordance improved almost 11% withthe individualized intervention compared with 4% with thegeneral intervention (= 2.796, = .008; Table 5); thisabsolute increase of 11% represents a 26% relativeimprovement over baseline nonconcordance in theindividualized group versus 7% in the general group.The Figure shows the change for each clinician: in theindividualized-intervention group, every clinicianexcept 1 improved his/her guideline drug concordance.
Bootstrap analysis showed that being in the individualized-intervention group increased the odds of concordance1.5-fold (2-sided = .03, z = 2.23; 95%confidence interval = 0.05, 2.12). There was no significantphysician versus nurse practitioner effect. In thelogistic regression model with general-estimating-equationcorrection for clustering by clinician, the odds ofconcordance after the intervention were increased by1.69-fold if the patient's clinician was in the individualized-intervention group (z = 2.78, < .01) comparedwith the general-intervention group.
Changes in average BP per clinician panel from therandom sample of abstracted charts are shown in Table6. Blood pressure measurements decreased during thecourse of the study within each group; the differencebetween the general-intervention group and the individualized-intervention group was not significant. The proportionof patients with adequate BP control changedfrom 43% to 45% and from 39% to 47% in the general-interventionand individualized-intervention groups,respectively. These analyses were repeated using thedata from 1817 patients with computer-recorded BPs;the findings were similar.
Implementation of a hypertension clinical practiceguideline via an individualized intervention, with recommendationsabout individual patients presented toclinicians at the time of scheduled clinic appointmentsand drug profiling of the clinicians' patientpanel, led to a significantly greater proportion ofpatients receiving prescriptions for guideline-concordantdrugs than a more general intervention.Clinicians receiving the individualized interventionhad a percent change in guideline concordance twiceas large as that seen in the comparison group. Theseresults were found with clinician-level statisticalanalysis, demonstrating that the effect of the interventionis robust enough to be detected using themost conservative analysis. This improvement inguideline concordance was achieved without worseningof BP control.
We selected hypertension as a model for study ofguideline implementation because of its prevalence andseriousness, and because, despite widely promoted evidence-based guidelines,22 hypertension managementoften diverges from guideline-recommended drugselection and BP goal values.15-17,24 Our finding that theindividualized intervention improved guideline concordanceis encouraging both for treatment of hypertensionand also for the potential extension of thisapproach to other chronic diseases. A managed carepractice could set up the organizational and paymentsystem for the personnel time required by implementationof our approach.
Changing Clinician Behavior
Clinical practice often deviates from guidelines.1,27Methods of implementation such as handing out guidelines(passive dissemination) are generally ineffective inchanging practice.2,3 Our intervention builds on previousapproaches to changing physician behavior.3,28 Itincorporated features advocated by the "awareness-to-adherence"model20: a priming strategy of drug profilingwith feedback to the clinician to demonstrate lack ofguideline concordance, an educational interventionoffered to all clinicians at a workshop, and a reinforcingstrategy of recommendations delivered at the time ofclinical decision making. We took advantage of theavailability of a computerized record system to performautomated analysis of patient records and generation ofindividualized advisories. Computerized record systemswith the data needed to carry out our intervention—diagnoses, drug lists, and appointment schedules—aremuch more widely available than complete electronicmedical records; hence, our approach has wideapplicability.
Reminder systems have been used extensively andeffectively to foster adherence to preventive medicineguidelines.7,9,11,12,29-31 Reminder systems and feedbacksystems also have been used for hypertension; most ofthese systems, identified in a systematic review,32 havefocused on BP measurement and follow-up.6,33-35 A studyof a computerized clinical reminder system for standardsof care in ambulatory practice found that thehypertension reminder had no effect.31
A recent systematic review of quality improvementstrategies for hypertension found that the medianimprovement in the proportion of provider adherenceto recommended practices was 3.3%.36 Our interventionled to more substantial improvement. A UnitedKingdom National Health Service study concluded thata greater health gain could theoretically be obtained byimproving professional standards for the treatment ofhypertension than by improving patient adherence totreatment; our intervention offers 1 method for improvingclinician management of hypertension.37
Effect on Blood Pressure
The aim of the individualized intervention was tomodify clinician prescribing to make drug selectionmore concordant with guidelines. The guideline drugsare recommended on the basis of improved cardiovascularor renal outcomes rather than inherentlygreater efficacy in lowering BP. Interestingly, wefound that both groups showed improved BP control.Indeed, we may be underestimating the impact of ourinterventions on BP because the short duration of thestudy probably did not allow sufficient time for cliniciansto titrate drugs to optimally effective doses.Because the drop in BP was seen only in observationaldata, and is not a finding from the randomized trial,it could be due to a secular trend. Consequently, wecannot conclude that either intervention broughtabout improved BP; nonetheless, we are reassured thatthe intervention did not worsen BP.
This study was conducted in a healthcare systemthat was implementing clinical practice guidelines,including hypertension guidelines. As there was nogroup without a general guideline intervention, it wasnot possible to determine the full impact of the individualizedintervention compared with usual care. Thegeneral-intervention clinicians could have been movedin the direction of increased concordance by discussionwith individualized-intervention clinicians. It is possiblethat the impact of the individualized intervention wouldhave been greater if it could have been compared withusual care.
This study was conducted within a single healthcaresystem; however, VAPAHCS is geographically dispersed.Indeed, the largest numbers of patients in thisstudy group were seen at sites remote from the PaloAlto campus, which is a tertiary-care hospital staffedlargely by university-affiliated physicians. The othersites, where many physicians have no academic affiliation,include small primary care clinics in remote sitesand intermediate-sized facilities with urgent care, generalmedical clinics, and visiting subspecialists. Thereare many nurse practitioners. Consequently, this systemincludes components found in many healthcaresystems nationally; it is likely that the findings are generalizableto primary care clinics in other large healthcaresystems.
Because the VA patient population is largely male,we have limited understanding of how our interventionaffects women. However, it is difficult to imagine amechanism by which our approach would amplify anyexisting treatment disparities between men and women,or create new ones.
We lack data on treatment of hypertension beforeimplementation of the general and individualized interventions.It is possible that guideline concordance wassteadily increasing among clinicians prior to the study.Even so, our results demonstrate that the individualizedintervention produced greater guideline concordancethan the general intervention—regardless of whetherboth study arms were "riding" an upward concordancetrend at baseline.
By having only a few clinical categories of patientswith hypertension, the VISN 21 guidelines did nottake into account some important patient variables.Future interventions would preferably include(where electronic medical records allow it to be doneefficiently) more detailed information about patients.Systems with more sophisticated recommendations38-41 may be able to achieve even more impressiveoutcomes.
Individualized recommendations about drug therapyfor hypertension presented to clinicians at the timeof a patient visit are effective in changing prescribingto achieve higher rates of guideline adherence.Providing individualized recommendations to clinicianscan be done in healthcare systems with electronicpharmacy and diagnostic data, even in theabsence of a complete electronic health record.Generation and distribution of recommendations canbe done efficiently if these activities are integratedwith existing procedures.
We would like to thank Parisa Gholami, MPH, for project coordination;Martin O'Connor, MS, Stanford Medical Informatics, for assistance withautomating the generation of the patient-specific advisories; JudithThielen, RNP, assistant chief, Office of Ambulatory Care at VA Palo Alto,for assistance in delivery of the clinic hypertension forms; Susana Martins,MD, for assistance with chart review and analysis of the blood pressuredata; and Arnold Saha, BA, for assistance with final preparation of the manuscript.We also would like to thank Dr David Siegel, chief of the MedicalService at VA NCHCS, and the rest of the VISN 21 Hypertension GuidelineCommittee for their support and encouragement of operationalizing the VAhypertension guidelines in VISN 21.
From the Geriatrics Research Education and Clinical Center (GRECC), VA Palo AltoHealth Care System, Palo Alto, Calif (MKG, RC); the Center for Primary Care and OutcomesResearch (MKG) and Stanford Medical Informatics (AA), Stanford University School ofMedicine, Stanford, Calif; the Department of Health Research and Policy, StanfordUniversity (PL); and the VA Boston Health Care System-West Roxbury and Harvard MedicalSchool, Boston, Mass (BBH).
This research was supported by Department of Veterans Affairs Health ServicesResearch and Development Service (VA HSR&D) grant CPG-97006. Dr Goldstein wassupported by VA HSR&D grant RCD-96301. Dr. Advani is supported by grant NationalLibrary of Medicine (NLM) 5-T15-LM-7033.
The views expressed in this article are those of the authors and do not necessarily representthe views of the Department of Veterans Affairs.
Address correspondence to: Mary K. Goldstein, MD, VA Palo Alto Health Care SystemGRECC 182B, 3801 Miranda Ave, Palo Alto, CA 94304. E-mail: firstname.lastname@example.org.
N Engl J Med.
1. Lomas J, Anderson GM, Domnick-Pierre K, Vayda E, Enkin MW, Hannah WJ.Do practice guidelines guide practice? The effect of a consensus statement on thepractice of physicians. 1989;321:1306-1311.
2. Cabana MD, Rand CS, Powe NR, et al. Why don't physicians follow clinicalpractice guidelines? A framework for improvement. 1999;282:1458-1465.
3. Grimshaw JM, Russell IT. Effect of clinical guidelines on medical practice: asystematic review of rigorous evaluations. 1993;342(8883):1317-1322.
4. Grimshaw JM, Shirran L, Thomas R, et al. Changing provider behavior: anoverview of systematic reviews of interventions. 2001;39(8 suppl 2):II2-45.
5. Soumerai SB, McLaughlin TJ, Gurwitz JH, et al. Effect of local medical opinionleaders on quality of care for acute myocardial infarction: a randomized controlledtrial. 1998;279:1358-1363.
6. Dickinson JC, Warshaw GA, Gehlbach SH, Bobula JA, Muhlbaier LH,Parkerson GR Jr. Improving hypertension control: impact of computer feedbackand physician education. 1981;19:843-854.
Arch Intern Med.
7. Balas EA, Weingarten S, Garb CT, Blumenthal D, Boren SA, Brown GD.Improving preventive care by prompting physicians. 2000;160:301-308.
8. Tierney WM, Hui SL, McDonald CJ. Delayed feedback of physician performanceversus immediate reminders to perform preventive care. 1986;24:659-666.
J Am Med Inform Assoc.
9. Shea S, DuMouchel W, Bahamonde L. A meta-analysis of 16 randomized controlledtrials to evaluate computer-based clinical reminder systems for preventivecare in the ambulatory setting. 1996;3:399-409.
Arch Fam Med.
10. Balas EA, Austin SM, Mitchell JA, Ewigman BG, Bopp KD, Brown GD. Theclinical value of computerized information services. A review of 98 randomizedclinical trials. 1996;5:271-278.
Arch Intern Med.
11. McPhee SJ, Bird JA, Jenkins CN, Fordham D. Promoting cancer screening. Arandomized, controlled trial of three interventions. 1989;149:1866-1872.
J Gen Intern Med.
12. McPhee SJ, Bird JA. Implementation of cancer prevention guidelines in clinicalpractice. 1990;5(5 suppl):S116-122.
The Sixth Report of the
Joint National Committee on Prevention, Detection, Evaluation, and Treatment of
High Blood Pressure.
13. National High Blood Pressure Education Program. Washington, DC: National Institutes of Health; 1997.
Am Heart J.
14. Swales JD. Current clinical practice in hypertension: the EISBERG (Evaluationand Interventions for Systolic Blood pressure Elevation-Regional and Global) project.1999;138(3 pt 2):231-237.
Am J Hypertens.
15. Mehta SS, Wilcox CS, Schulman KA. Treatment of hypertension in patientswith comorbidities: results from the Study of Hypertensive Prescribing Practices(SHyPP). 1999;12(4 pt 1):333-340.
Arch Intern Med.
16. Hyman DJ, Pavlike VN. Self-reported hypertension treatment practices amongprimary care physicians. 2000;160:2281-2286.
Can Fam Physician.
17. Beaulieu MD, Dufresne L, LeBlanc D. Treating hypertension. Are the rightdrugs given to the right patients? 1998;44:294-288, 301, 302.
Am J Med.
18. Steinman MA, Fischer MA, Shlipak MG, et al. Clinician awareness of adherenceto hypertension guidelines. 2004;117:747-754.
19. Lomas J, Enkin M, Anderson GM, Hannah WJ, Vayda E, Singer J. Opinionleaders vs audit and feedback to implement practice guidelines. 1991;265:2202-2207.
20. Pathman DE, Konard TR, Freed GL, Freeman VA, Koch GG. The awarenessto-adherence model of the steps to clinical guideline compliance. 1996;34:873-889.
J Gen Intern Med.
21. Litzelman DK, Dittus RS, Miller ME, Tierney WM. Requiring physicians torespond to computerized reminders improves their compliance with preventivecare protocols. 1993;8:311-317.
The Pharmacologic Management of Hypertension.
22. Medical Advisory Panel for the Pharmacy Benefits Management StrategicHealth Group. Washington,DC: Department of Veterans Affairs, Veterans Health Administration; 1996.
Am J Hypertens.
23. Siegel D, Lopez J, Meier J, et al. Academic detailing to improve antihypertensiveprescribing patterns, part 1. 2003;16:508-511.
N Engl J Med.
24. Berlowitz DR, Ash AS, Hickey EC, et al. Inadequate management of bloodpressure in a hypertensive population. 1998;339:1957-1963.
Analysis of Longitudinal Data.
25. Diggle PJ, Liang KY, Zeger SL. Oxford,England: Clarendon Press; 1994.
26. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models.1986;73:13-22.
N Engl J Med.
27. Greco PJ, Eisenberg JM. Changing physicians' practices. 1993;329:1271-1274.
Br Med J.
28. Bero LA, Grilli R, Grimshaw JM, Harvey E, Oxman AD, Thomson MA, andthe Cochrane Effective Practice and Organisation of Care Review Group.Closing the gap between research and practice: an overview of systematic reviewsof interventions to promote the implementation of research findings. 1998;317:465-468.
29. Hunt DL, Haynes RB, Hanna SE, Smith K. Effects of computer-based clinicaldecision support systems on physician performance and patient outcomes: a systematicreview. 1998;280:1339-1346.
30. McPhee SJ, Bird JA, Fordham D, Rodnick JE, Osborn EH. Promoting cancerprevention activities by primary care physicians. Results of a randomized, controlledtrial. 1991;266:538-544.
31. Demakis JG, Beauchamp C, Cull WL, et al. Improving residents' compliancewith standards of ambulatory care: results from the VA Cooperative Study onComputerized Reminders. 2000;284:1411-1416.
J Epidemiol Community Health.
32. Montgomery AA, Fahey T. A systematic review of the use of computers in themanagement of hypertension. 1998;52:520-525.
33. McDowell I, Newell C, Rosser W. A randomized trial of computerizedreminders for blood pressure screening in primary care. 1989;27:297-305.
34. Barnett GO, Winickoff RN, Morgan MM, Zielstorff RD. A computer-basedmonitoring system for follow-up of elevated blood pressure. 1983;21:400-409.
Br Med J
(Clin Res Ed).
35. McAlister NH, Covvey HD, Tong C, Lee A, Wigle ED. Randomised controlledtrial of computer assisted management of hypertension in primary care. 1986;293(6548):670-674.
Quality Gap: A Critical Analysis of Quality Improvement Strategies.
36. Walsh J, McDonald KM, Shojania KG,McDonald KM, et al. Hypertension, 3.In: Shojania KG, McDonald KM, Wachter RM, Owens DK, eds. TechnicalReview 9 (prepared by the Stanford University-UCSF Evidence-Based PracticeCenter, under contract 290-02-0017). Rockville, Md: Agency for HealthcareResearch and Quality (AHRQ); 2005. AHRQ publication 04-0051-3.
J Public Health Med.
37. Ebrahim S, Smith GD. Lowering blood pressure: a systematic review of sustainedeffects of non-pharmacological interventions. 1998;20:441-448.
38. Goldstein MK, Hoffman BB, Coleman RW, et al. Implementing clinical practiceguidelines while taking account of changing evidence: ATHENA DSS, an easilymodifiable decision-support system for managing hypertension in primary care.Proceedings of the American Medical Informatics Association Symposium. 2000;7(20 suppl):300-304.
J Am Med
39. Goldstein MK, Hoffman BB, Coleman RW, et al. Patient safety in guideline-baseddecision support for hypertension management: ATHENA DSS. 2002;9(6 suppl):S11-16.
40. Shankar RD, Martins SB, Tu SW, Goldstein MK, Musen MA. Building anexplanation function for a hypertension decision-support system. 2001;10(pt 1):538-542.
J Am Med Inform Assoc.
41. Goldstein M, Coleman R, Tu S, et al. Translating research into practice: organizationalissues in implementing automated decision support for hypertension inthree medical centers. 2004;11:368-376.