Relationship Between Provider Type and the Attainment of Treatment Goals in Primary Care

, , , ,
The American Journal of Managed Care, September 2005, Volume 11, Issue 9

Objective: To determine the relationship between provider type(eg, resident, mid-level practitioner, attending physician) andattainment of clinical goals with respect to the treatment of dyslipidemia,diabetes mellitus, and hypertension.

International Classification of

Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)

Study Design: Using electronic medical records, we identifiedall patients seen in the Veterans Affairs Connecticut Health CareSystem in a 6-month period with an codecorresponding to a diagnosis of coronary artery disease, diabetesmellitus, or hypertension.

Methods: We recorded the most recent low-density lipoproteincholesterol level for patients with diabetes or coronary arterydisease, glycosylated hemoglobin level for patients with diabetes,and blood pressure for patients with hypertension. We thenexamined the relationship between these measures and providertype.


Results: After controlling for patient age and practice site, nosignificant differences were noted between attending physiciansand residents in attaining treatment goals for dyslipidemia (oddsratio [OR] 1.15, 95% confidence interval [CI] 0.96-1.38) or diabetes(OR 1.05, 95% CI 0.82-1.33). However, attending physicianswere significantly more likely to attain blood pressure goals thanwere residents (59% vs 54%, = .002; OR 1.23, 95% CI 1.08-1.39). Controlling for additional patient characteristics did not alterthe findings.

Conclusions: Internal medicine residents may quickly developexpertise in managing dyslipidemia and diabetes mellitus.Residents in our sample, however, were less likely than attendingphysicians to reach goal blood pressure for patients with hypertension.Educational efforts aimed at house staff to improve thetreatment of hypertension are warranted.

(Am J Manag Care. 2005;11:561-566)

Cardiovascular disease (CVD) is the leading causeof death in the United States.1 Estimates are thatnearly 50 million Americans have high bloodpressure, 42 million have elevated blood cholesterol levels,and almost 11 million have diabetes mellitus.2 Notonly does CVD account for substantial morbidity andmortality, it also exerts a substantial financial burdenon the healthcare system. In 1999, more than $26 billionwas paid to Medicare beneficiaries for hospitalexpenses due to cardiovascular problems.1 Becauseseveral of the modifiable risk factors for CVD are wellknown and prevalent (eg, hypertension, dyslipidemia,and diabetes mellitus), efforts aimed at targetingthese conditions have been promulgated by majororganizations.3-5

The Veterans Health Administration in theDepartment of Veterans Affairs (VA) is the largest fullyintegrated healthcare system in the United States, providingcare to nearly 3.5 million persons annually.6Within the VA system, primary care is provided byattending physicians, mid-level practitioners (eg, nursepractitioners and physician assistants), and internalmedicine house staff (residents).

Although residents have been shown to be deficientin their knowledge and screening practices with respectto colon cancer,7 they are more likely to prescribe narrow-spectrum antibiotics than attending physicians forsinusitis.8 In a study conducted in a large urban publichospital, diabetic patients of residents had similar ratesof goal attainment for glycosylated hemoglobin (A1C)and cholesterol when compared to diabetic patients offaculty.9

Little is known about the relative performance ofthese provider types with respect to the treatment ofcommon outpatient conditions within the VA, such asdiabetes mellitus, dyslipidemia, and hypertension. Wesought to determine whether internal medicine residentsin training were less likely to be successful inattaining recommended goals for their patients thantheir more experienced attending physicians or midlevelproviders.


Practice Setting

The VA Connecticut Health Care System cares formore than 44 000 veteran patients at 2 major academichealthcare centers (the West Haven and Newington VAMedical Centers) and 6 community-based clinics.Patients are randomly assigned a primary care provider.These primary care providers consist of internal medicineresidents, who practice under the supervision ofattending physicians; mid-level practitioners, andattending physicians. Fifty-six internal medicine residents(40 affiliated with Yale University School ofMedicine and 16 with the University of ConnecticutSchool of Medicine) have their longitudinal generalmedicine clinic within the VA Connecticut Health CareSystem and provide primary care for 8.4% of patientsenrolled in primary care. Each resident is supervised bya designated general internal medicine faculty member.Each attending physician supervises 3 to 4 residents atany 1 time and must cosign all resident notes. Residentsare expected to present each patient to the attendingphysician, during which time the patient's problems arediscussed and a plan for care is made conjointly. Whena resident completes residency training, the resident'spanel of patients is transferred to that of another incomingresident.

In the VA Connecticut Health Care System, 18 mid-levelpractitioners (13 nurse practitioners and 5 physicianassistants) practice independently, although theyare assigned designated attending physicians shouldclinical questions arise. All attending physicians forresidents or mid-level practitioners are full-or part-timestaff physicians in primary care, see their ownpanel of primary care patients, and hold academicappointments at the University of Connecticut Schoolof Medicine or at Yale University School of Medicine.When attending physicians supervise residents in primarycare, however, they do not see their ownpatients. Within the VA Connecticut Health CareSystem 39 attending physicians are engaged in thepractice of primary care.

Data Collection

The VA uses an electronic medical records system,which allowed us to compile patient lists and correspondinglaboratory values. Within the electronic medicalrecord, a patient's primary care provider is clearlyidentified, thus allowing us to identify provider types.Blood pressure measurements are obtained at all primarycare visits, and blood tests are ordered at the discretionof the providers. Blood pressure values areentered into the electronic medical record, as are theresults of all blood tests performed at the VA.

International Classification of Disease, 9th Edition,

Clinical Modification


Each patient visit is electronically assigned an() code correspondingto the diagnoses for which the patient was treated. Weidentified patients with coronary artery disease (CAD)using the codes 410 to 414.9, diabetes mellitus using thecodes 250.00 to 250.93, and hypertension using thecodes 401.0 to 401.9. Psychiatric comorbidity, that is,alcohol disorder, drug disorder, and severe mental illness(bipolar disorder, major depressive disorder, andschizophrenia), were identified using the codes 305.00,305.01, 305.02, 305.03, 303.xx, 291.xx, 305.2x, 305.99,292.xx, 304.xx, 295.xx, and 296.xx. Patients who wereseen at VA Connecticut between April 1, 2003, andSeptember 30, 2003, were identified for our study bythese codes.


For patients with codes relating to eitherdiabetes mellitus or CAD, we recorded the most recentlow-density lipoprotein (LDL) measurement obtainedwithin the last 25 months. For patients with diabetesmellitus, we also recorded the most recent A1C measurementobtained within the past 13 months. Forpatients with hypertension, we noted the most recentblood pressure obtained during the previous 13 months.These time frames were chosen because VA standardsrecommend obtaining LDL cholesterol level measurementsat least biannually for patients with diabetes mellitusor CAD, A1C at least yearly for patients withdiabetes, and blood pressures at least yearly for allpatients. Computer-generated prompts remindproviders of these standards.

Target goals for LDL, A1C, and blood pressure wereless than 100 mg/dL,5 less than 7.5%, and less than140/90 mm Hg,4 respectively. Because the AmericanDiabetes Association recommends that target A1C beless than 1% above the upper limit of normal,5 we useda target A1C of less than 7.5% since the upper limit ofnormal for our laboratory is 6.5%.

Patients without blood pressure or laboratory testvalues or without an identified provider were excludedfrom analysis. Because female veterans were treatedalmost entirely by residents and at a separate site, theywere also excluded from analysis. The study protocolwas approved by the institutional review board at theWest Haven Veterans Affairs Medical Center.

Statistical Analysis

Descriptive statistics on the sample characteristicsincluded frequencies, proportions, and ranges.Indicator variables were created for the site of care, the3 conditions of interest (hypertension, diabetes, andCAD), and for each comorbid psychiatric condition (alcoholdisorders, drug disorders, and severe mental illness).

Chi-square and Fisher's exact test, in cases in whichcell sizes were small, were used to examine the relationshipbetween provider type and dichotomous patientcharacteristics. Logistic regression was used to adjust forpotential confounding factors in the association betweenprovider type and the patients' attainment of treatmentgoals. Patient age and site of care were controlled for inall multivariate analyses. Additional analyses controlledfor comorbid psychiatric disorders and comorbid medicalconditions. Separate analyses were performed foreach goal of interest (LDL, A1C, and blood pressure).Each analysis utilized data from patients diagnosed withthe condition of interest for which a particular goal wasto be met (eg, only patients with diabetes were includedin the analysis on attaining A1C < 7.5%).


Odds ratios (ORs) and 95% confidence intervals (CIs)were calculated using the coefficients and standarderrors of the respective variable in the logistic regressionanalysis. Odds ratios from the logistic regressionswere calculated using resident as the referent group.Statistical significance was defined as a 2-tailed < .05.All statistical analyses were performed using SAS version8 (SAS Institute, Inc, Cary, North Carolina).


A total of 20 453 unique patients with a diagnosis ofhypertension, diabetes mellitus, or CAD were seen duringthe observation period. From this group, 294 wereexcluded because they had no identifiable primary careprovider and an additional 499 were excluded becausethey were female. Afterexcluding these 793 patients,19 660 unique patients metentry criteria and wereincluded in the analyses.

Attending physicians saw12 682 unique patients, mid-levelpractitioners saw 5572patients, and residents saw1406 patients. A total of15 893 patients had a diagnosisof hypertension, 6292a diagnosis of CAD, and 5830a diagnosis of diabetes mellitus;7277 patients had 2 ormore of these diagnoses.Patient characteristics, diagnoses,and comorbidity arelisted in Table 1.





A significant differencewas noted in the age of thepatients by the providertypes (< .001). Residentsand attending physicianswere more likely to seeolder patients than midlevelpractitioners. Significant differences were notedin the proportion of patients with comorbid psychiatricdisorders by provider type: mid-level practitionerssaw more patients with alcohol disorders (< .001),drug disorders (= .017), and severe mental illness(= .017); residents saw a smaller proportion ofpatients with these comorbidities than either attendingphysicians or mid-level providers.

A significant difference was noted in the proportion ofpatients with hypertension, diabetes, or CAD by providertype: mid-level practitioners were more likely to seepatients with hypertension compared to attending physicians;residents were less likely to see patients with CADcompared to attending physicians; and mid-level practitionerswere less likely to see patients with diabetes mellituscompared to attending physicians (Table 1). Whenresidents were compared with mid-level practitionersand attending physicians combined, residents saw significantlyolder patients and were less likely to see patientswith CAD or comorbidities (data not shown).

Unadjusted Analyses


For patients with CAD, 70% of attending physicians'patients had met the treatment goal of LDL less than100 mg/dL, compared with 68% of patients cared for bymid-level practitioners and 72% cared for by residents(= .331) (Table 2). No significant differences werenoted between the provider types on attainment of LDLgoals.


For diabetic patients, 71% of attending physicians'patients had A1C levels less than 7.5%, compared with74% of mid-level practitioners' patients and 71% ofresidents' patients (= .190). No statistically significantdifferences were noted in A1C attainment by providertype.




For patients with hypertension, the target bloodpressure of less than 140/90 mm Hg was attained by 59%of attending physicians' patients, compared with 57% ofmid-level practitioners' patients and 54% of residents'patients (= .001) (Table 2). Significant differences werenoted in blood pressure control between residents' andmid-level practitioners' patients (= .018) and betweenresident and attending physicians' patients ( = .002).

Multivariate Analyses

After controlling for patient age and practice location,attending physicians were significantly more likelyto attain blood pressure goals compared to internalmedicine residents (OR 1.23, 95% CI 1.08-1.39). No significantdifference was noted in blood pressure controlbetween mid-level practitioners and residents (OR 1.12,95% CI 0.98-1.28) (Table 3).

No significant difference was noted for patientattainment of LDL cholesterol goals between attendingphysicians and residents (OR 1.15, 95% CI 0.96-1.38)or between mid-level practitioners and residents (OR1.03, 95% CI 0.85-1.26). Mid-level practitioners weresignificantly more likely to attain A1C goals comparedto residents (OR 1.30, 95% CI 1.01-1.70); however,attending physicians were not more likely to attainA1C goals than residents (OR 1.05, 95% CI 0.82-1.33)(Table 3).

In separate logistic regression models controlling forthe additional effects of co-occurring medical and psychiatriccomorbidities, the pattern of ORs did notchange significantly.


Given the complexity of healthcare, use of qualitymeasures is essential to ensure the promotion of healthand prevention of disease complications. This is especiallytrue in academic medical centers, where traineesare rendering care to patients. Substandard care hasnegative implications not only to the training programbut, more importantly, to the health of individualpatients as well.

With respect to the control of 3 common outpatientproblems, we found that although attending physicianswere not more successful than residents in obtainingtreatment goals for patients with dyslipidemia and diabetesmellitus, they were more successful in attainingblood pressure goals for hypertensive patients.

Several theories may explain why the rates of goalattainment were not markedly different betweenattending physicians and residents for the treatment ofdyslipidemia or diabetes mellitus. Because these conditionsare so prevalent, perhaps residents in trainingquickly develop expertise in these common clinicalentities. Additionally, familiarity with well-promulgatedguidelines by major organizations may further improvethe quality of care.

Quality Management Quarterly

At our facilities, residents practice in the same locationas their attending physicians and have the sameaccess to ancillary support, such as clinical pharmacists,nurses, and health technicians. All practitionersreceive feedback from regarding their compliance with these guidelines, apractice that has been shown to improve the performanceof residents.10,11 It is also possible that attendingphysician input improved the performanceof the residents, and the"resident" group in our study shouldactually be thought of as "residentswith attending supervision."Similarly, some degree of crossovermay have occurred: Because residentsare in primary care clinic only1 to 2 afternoons per week and mid-levelpractitioners and attendingphysicians often have full schedules,opportunities for intervention mighthave occurred during urgent visitsthat might have been handled byother provider types in the clinic.

Why residents are less likely toreach goal attainment with respect to the treatment ofhypertension is unclear. One possibility is that becauseblood pressure is more variable over a short period oftime or during acute illness, and our methodology usedthe last recorded blood pressure in the electronic medicalrecord, residents' patients might have had moreacute illnesses than patients of other providers. Moreplausibly, it is possible that with respect to hypertension,residents might be more likely to succumb to"clinical inertia,"12 (ie, the tendancy not to make therapeuticintervention when a treatment target is not met).Additionally, it is possible that attending physicians ormid-level practitioners were more likely than residentsto involve the patient in self-management, such as utilizationof home blood pressure monitors, which areavailable at no cost to patients within VA Connecticutand which might have led to higher attainment of bloodpressure goals. Educational efforts should be aimed atresidents to improve the likelihood that they wouldundertake all available therapeutic interventions forpatients with uncontrolled hypertension. Also unclear iswhether residents would perform as well as attendingphysicians with less prevalent illnesses or illnessesunassociated with specific numerical laboratory abnormalitiesthat can be easily followed.

Interestingly, with respect to the treatment of hypertension,an area that has been well studied, all 3 groupsof healthcare providers performed far better than historicalcontrols. For example, in the 1988 to 1991National Health and Nutrition Examination Survey(NHANES III), 76% of patients with hypertension hadblood pressure measurements of 140/90 mm Hg or higher.13 Within the VA healthcare system, in a study of veteranssampled from 1990 to 1995, more than 65% hadblood pressures 140/90 mm Hg or higher.14 More recently,as of 1999, these measures have improved, althoughmost patients still exceeded recommended levels.15 It ispossible that the improving level of blood pressure controlis reflective of the ongoing system-wide efforts initiatedin the mid-1990s to improve the quality of carewithin the VA.16

With respect to the treatment of dyslipidemia, ourresults again were better than those from previous studies.For example, at a major academic medical center,only 34% of patients with CAD had LDL cholesterol levelsless than 100 mg/dL.17 Low rates of achievement ofgoal LDL were also found among patients with eitherCAD or peripheral vascular disease at another institution.18 Despite improving rates of goal attainment, thereis clearly room for continued progress.


Our study has several limitations. Becausepatients were selected by codes, we donot know if providers or specific provider subtypeswere more likely to code a visit for hypertension,CAD, or diabetes if patients were more or less likelyto meet target treatment goals' accepted guidelines.Coding disparities between groups could be apotential confounder. Another potential confounderwas the fact that a resident in primary care serves asa primary care provider for only 3 years, until thecompletion of house staff training. Thus some of theobserved differences may have been due to the lackof a prolonged relationship between patient andprovider, although it is unlikely that this would bedetrimental only to the treatment of hypertensionand not also to diabetes or dyslipidemia. We alsoused a single blood pressure measurement as anassessment of control of a chronic condition, hypertension.However, this measurement was used in allgroups and is how the VA is presently assessing thetreatment of hypertension.

Furthermore, our study population consisted of acohort of older male veterans. We do not knowwhether a study of female or predominantly youngerpatients would result in similar findings. Our methodologyalso only allowed for data collection from withinthe VA healthcare system. If patients were comanagedby community physicians, laboratory tests and bloodpressures done outside of the VA system were notincluded. However, comanagement by communityphysicians would bias the results toward the nullhypothesis.

Another potential limitation is that we analyzed datafrom only 1 point in time for each LDL value, bloodpressure measurement, or A1C: the latest data point.Although measurements change over time, providersare "graded" within the VA based on a single data point.

The major strengths of our study are that we wereable to include a large cohort of patients with retrievableblood pressure and laboratory data, as well asinclude multiple residents and attending physiciansfrom more than 1 major university.

Fortunately, the clinical "treatment gap" betweeninternal medicine trainees and their attending physicianswith respect to these common outpatient disordersis small. This finding should be reassuring toresidency training programs and patients. However,with respect to the treatment of hypertension, furthereducational efforts targeting house staff are warranted.

From the VA Connecticut Health Care System, West Haven, Conn (DGF, RK, SK, JG,AJ) and the Department of Medicine, Yale University School of Medicine, New Haven,Conn (DGF, RK, SG, AJ).

Address correspondence to: Daniel G. Federman, MD, VA Connecticut HealthCare System (11ACSL), 950 Campbell Avenue, West Haven, CT 06516.E-mail:

Heart Disease and Stroke Statistics—2003


1. American Heart Association. Dallas, Tex: American Heart Association; 2002:4-42.

Health, United States, 2001 with Urban

and Rural Health Chartbook.

2. National Center for Health Statistics. Hyattsville, Md: Dept of Health and Human Services,Centers for Disease Control and Prevention; 2001. DHHS publication 01-1232.


3. Expert Panel on Detection, Evaluation, and Treatment of High BloodCholesterol in Adults. Executive summary of the third report of the national cholesteroleducation program (NCEP) expert panel on detection, evaluation, and treatmentof high blood cholesterol in adults (adult treatment panel III). 2001;285:2486-2497.


4. Chobanian AV, Bakris GL, Black HR, Cushman WC, et al. The seventh report ofthe Joint National Committee on prevention, detection, evaluation, and treatmentof high blood pressure: the JNC 7 report. 2003;289:2560-2571.



5. American Diabetes Association. Standards of medical care in diabetes. 2004;27(suppl 1):S15-S35.

Am J Psychiatry.

6. Desai MM, Rosenheck RA, Druss BG, Perlin JB. Mental disorders and quality ofdiabetes care in the veterans health administration. 2002;159:1584-1590.

Am J Gastroenterol.

7. Barrison AF, Smith C, Oviedo J, Heeren T, Schroy PC 3rd. Colorectal cancerscreening and familial risk: a survey of internal medicine residents' knowledge andpractice patterns. 2003;98:1410-1416.

South Med J.

8. Mincey BA, Parkulo MA. Antibiotic prescribing practices in a teaching clinic:comparison of resident and staff physicians. 2001;94:365-369.

Diabetes Care.

9. Suwattee P, Lynch JC, Pendergrass ML. Quality of care for diabetic patients in alarge urban public hospital. 2003;26:563-568.

Acad Med.

10. Holmboe E, Scranton R, Sumption K, Hawkins R. Effect of medical recordaudit and feedback on residents' compliance with preventive health care guidelines.1998;73:901-903.

Acad Med.

11. Callahan M, Fein O, Battleman D. A practice-profiling system for residents.2002;77:34-39.

Arch Intern Med.

12. O'Connor PJ. Overcome clinical inertia to control systolic blood pressure.2003;163:2677-2678.


13. Burt VL, Whelton P, Roccella EJ, et al. Prevalence of hypertension in the USadult population: results from the Third National Health and Nutrition ExaminationSurvey, 1988-1991. 1995;25:305-313.

N Engl J Med.

14. Berlowitz DR, Ash AS, Hickey EC, et al. Inadequate management of bloodpressure in a hypertensive population. 1998;339:1957-1963.

Arch Intern Med.

15. Borzecki AM, Wong AT, Hickey EC. Hypertension control: how well are wedoing? 2003;163:2705-2711.

N Engl J Med.

16. Jha AK, Perlin JB, Kizer KW, Dudley RA. Effect of the transformation of theVeterans Affairs Health Care System on the quality of care. 2003;348:2218-2227.

J Clin Outcomes Manag.

17. Schnipper JL, Stafford RS. Secondary prevention of coronary artery disease: lipidmanagement as a priority for improvement. 2002;9:371-380.

s. Am J Med.

18. Schectman G, Hiatt J. Drug therapy for hypercholesterolemia in patients withcardiovascular disease: factors limiting achievement of lipid goal1996;100:197-204.