Objective: To determine whether a major improvement in access(ie, implementing an open access system) in a large multispecialtymedical group during 2000 was associated with changes in utilizationor costs for patients with diabetes, coronary heart disease(CHD), or depression.
Study Design: Multilevel regression analysis of health planadministrative data.
Patients and Methods: Approximately 7000 patients with diabetes,3800 with CHD, and 6000 with depression who received allof their care in this care system served as the subjects for this study.Utilization and costs between 1999 and 2001 (before and afterimplementation of open access) were compared for these patients.The main outcome measures were rates of inpatient admissionsand various types of outpatient encounters as well as associatedcosts for these subjects.
Results: Between 1999 and 2001, total office visit changes weresmall and varied with condition, but the proportion of these visitsmade to primary care physicians increased significantly by anabsolute 5% to 9% and primary care physician continuityincreased for each condition. Urgent care visits also decreased significantlyby an absolute 5% to 9%, but there was no change inemergency department visits or hospital admissions. Total costs ofcare for these patients were much larger than those for the overallpopulation of the medical group, but increased at a similar rate.
Conclusion: A major improvement in patient access to primarycare clinics was associated with increased use and continuity ofprimary care for patients with 3 chronic conditions, but did notaffect overall resource use.
(Am J Manag Care. 2004;10:717-722)
Crossingthe Quality Chasm
The Institute of Medicine's 2001 report highlighted the chasmbetween "the care we have and the care we couldhave."1 Serious deficits in quality of healthcare havebeen further documented by McGlynn et al's study ofnational adherence to 439 indicators for 30 conditions.2The chasm report emphasized the particular need toimprove care for patients with chronic conditions andwas followed by a 2003 report identifying the 20 priorityareas for transforming care.3 Many of these 20 werecommon chronic conditions for which improving qualitynecessarily involves addressing the 6 aims or dimensionsof quality identified by the chasm report: safety,timeliness, effectiveness, efficiency, equity, and patient-centeredness.
Each of these dimensions now is receiving increasedattention, but it seems that timeliness ("reducing waitsand sometimes harmful delays") is actually beingimproved, at least in terms of access to primary care.Murray and Tantau have been major innovators withrespect to access, helping many medical groups to makesubstantial improvements in access through anapproach called Advanced Access or Second GenerationOpen Access.4-6 In this approach, the goal is to be ableto offer any patient a visit the same day that he or shecalls, with the patient's personal physician if that physicianis in the office that day.
In theory, a medical practice that can offer this typeof access might expect to see a decrease in unnecessaryoffice visits, cancellations, and no-shows; decreasedurgent-care and emergency department (ED) visits;decreased hospitalizations because serious illnesses arecaught at an earlier stage; greater continuity of care;and perhaps a decrease in total office visits.4-7 If thesechanges occur, it seems likely that another of the 6dimensions, efficiency, might be improved as well, withdecreased costs for both the care system and forpatients. Whether and to what extent these effects actuallyoccur is unknown, however, because no publishedstudies thoroughly document such changes.
Because so much of the attention to the need forquality improvement has focused on patients withchronic conditions, another open question concerns theeffect of access improvement on these frequent andhigh-cost users of the care delivery system. Murray andBerwick suggest that such patients may fare better withprescheduled visits rather than expecting them to simplycall for an appointment on the day that their routinefollow-up is needed.4 Others have been concerned thatpatients with chronic disease may fall through thecracks of a care system that becomes increasingly orientedtoward acute and same-day care.
Because our large multispecialty medical grouprecently greatly improved primary care access using theAdvanced Access model of Murray and Tantau,4-6 weconducted this study to assess the impact of increasedaccess on utilization and cost of care. Over the courseof 1 year (2000), primary care access for our patients,measured by third-next-available appointment,improved from an average of 17.8 days in 1999 to 4.2days in 2001. Murray notes that this is a better measureof real access than the first- or second-next-availableappointment because those appointments are morelikely to represent random cancellations.4 The range ofthird-next-available appointments among 17 primarycare clinics in 2001 varied from 1.7 to 6.2 days.
Because we were particularly concerned with theeffects on patients with chronic conditions (diabetes,coronary heart disease [CHD], or depression), we focusedon those patients to learn whether the overall change inprimary care access between 1999 and 2001 was associatedwith any significant changes in:
This study was conducted in a 500-physician multi-specialtymedical group that is owned by a health planwith 650 000 members. About 240 000 of these membersare cared for by the medical group, most in the 17primary care clinics included in this study. The other410 000 members receive their care through about 50medical groups in the region that contract separatelywith the health plan, and they are not part of this study.
In late 1999, the medical group leadership decided toundertake a major change in the approach to access,hoping to improve patient satisfaction as well as overallefficiency and, possibly, clinician job satisfaction.8Therefore, the leadership engaged outside consultantsto help conduct a series of full-day sessions during2000 for representatives from all of its clinics and providedconsiderable training and consultative resourcesalong with a deadline (January 1, 2001) to achieve fulladvanced access. This required marked standardizationof schedule slots and extra visit time for clinicians towork down the backlog of their appointments, but therewas no increase in care personnel or resources duringthis change. Several other major changes took placeduring this time period: the appointment-makingprocess was centralized, physician compensation wasgradually switched from salary to productivity, andmajor work flow redesign and cost restructuring wereconducted to streamline support processes and reduceoverhead.
InternationalClassification of Diseases, Ninth Revision
Adult (age >18 years) patients with either diabetes,CHD, or depression were identified from health planadministrative databases by using algorithms thatwere modified from a previously described approachand validated against chart audits.9 For CHD ordepression, these algorithms specified that patientshave at least 1 inpatient diagnosis or 2 outpatient diagnosesin a given year with specified codes (seeTable 1). For diabetes, a patient could have filled adiabetes-specific medication or have had 1 inpatient or2 outpatient diagnoses. These algorithms have estimatedpositive predictive values of .96 for diabetes, .95 forCHD, and .90-.95 for depression.
After identifying patients with each condition in eachyear from 1998 through 2001 who were enrolled for atleast 11 months of that year, their utilization and costdata were collected from health plan administrative databases.Continuity of care was calculated based on the"continuity of care" method for the distribution of visitsby a patient among different providers in each year.10 Theformula is ∑(visiti2) − ∑(visiti)/[∑(visiti)Ã—(∑visiti) − 1)](where i = number of visits to a provider). Continuity ofcare tends to increase as the total number of visitsincrease, but is unaffected by the sequencing of visits.
Multilevel linear and nonlinear (ie, logistic) regressionmodels were used to compare utilization in 1999versus 2001 (before and after the change in access) withMLwiN software version 1.10 (Multilevel Models Project,London, UK). The linear models specified a normallydistributed dependent variable and used the IterativeGeneralised Least Squares estimation method. Thenonlinear models specified a binomially or extra-binomiallydistributed dependent variable (as appropriate)and a logit link function, and used the penalized quasi-likelihoodestimation method with first-order linearization.For each dependent variable, an intercept-onlymodel identified the significant random-variancecomponents to be included in the predictive model. A3-level (time within patient within provider) random-variance structure was attempted for all variables,although the provider level was omitted if not significant.Up to 2 observations per person (eg, continuity ofcare measured in 1999 and in 2001) were included ineach model. A dummy variable for year (reference =1999) indicated whether the values for the dependentvariable were different by year, and it was the parameterof interest in all models. Sex, age in 1998, and ayear-specific Charlson score greater than or equal to 3(as a measure of disease severity and comorbidities)were included as covariates.11,12
"Costs" were measured as paid amounts from healthplan administrative data. For contracted care providers,paid amounts are those actually paid by the health plan.For providers within the staff-model medical group(studied here), the paid amounts represent approximatelywhat the health plan would have paid to a contractedprovider. Costs were adjusted to year 2000dollars by using the medical-care component of the consumerprice index for all urban consumers. All steps inthe development of the identification system, aggregationof data, and data analysis were approved in advanceand monitored by the local institutional review board.Because aggregate de-identified claims data were usedin the analysis, the institutional review board did notrequire informed consent.
Table 1 shows the number of patients in the care systemwith each of the 3 chronic conditions for the yearsstudied, using the identification algorithms described inthe Methods section. There is significant overlap for diabetes,but less for CHD and depression. Table 2 comparesoffice visit utilization rates for these populationsbetween 1999 and 2001 (ie, before and after the accesschange). Although there were only small and variablechanges in total office visits, the proportion of those visitstaking place in primary care increased for all 3 conditions,from an absolute 5% for CHD to 8% to 9% for diabetesand depression. Continuity of care with the primarycare physician also increased significantly forpatients with each condition. The observed values forthe proportion of patients with visits to disease-specificspecialists went from 11.3% to 10.2% for endocrinologyvisits and from 73.8% to 70.5% for mental health visits(data not shown). However, the adjusted model demonstratedthat 38.8% of CHD patients had a cardiology visitin 1999 and 43.3% did in 2001 ( = .03).
Table 3 shows that the proportion of patients witheach condition making urgent-care visits decreased byabout one third, but there was little change in the proportionvisiting an ED. Overall, referral of patients tourgent care from these clinics because of inability to seethem had been increasing up to 1999 (14 573 in 1997,19 904 in 1998, 21 932 in 1999), but then dropped substantially(17 172 in 2000 and 12 952 in 2001). Slightlyfewer CHD patients had hospital admissions (57.3% vs58.4% adjusted; = .002) and their length of stays wereshorter (3.76 vs 3.82 days; = .01) after accessimprovements, but no change in either parameter wasnoted for patients with diabetes or depression (data notshown). Health plan data for all adults with commercialinsurance and care from either the staff-model or allother medical groups showed a decrease of 4.4% inlength of stay between 1999 and 2001 with no change inadmission rates. These data also showed that ED visitsper 1000 population increased by 7.8% for the staff-modelmedical group and by 13.3% for all contractedmedical groups over this time period.
Real healthcare costs (adjusted for medical costinflation) increased over this time period for this groupof patients, as seen in Table 4. Total cost increasesranged between 10% and 20% by condition, withincreases for almost all cost subcategories for all conditions.The proportion of total costs represented by outpatientcare increased for each condition, the most (anabsolute 10%) for patients with depression. The percentincrease in total healthcare costs for people with diabetesor CHD was roughly similar to that experiencedby the average adult health plan patient of the medicalgroup (9%), although depression patients experiencedover twice as much increase. However, the averageinflation-adjusted total cost for the average adult healthplan member ($1413 in 1999) was far less than wasspent per person with any of these conditions.
These results suggest that the introduction of a dramaticimprovement in access to primary care clinicswas associated with relatively little change in eitheroverall utilization or overall costs of care for patientswith these chronic conditions. However, improvedaccess was related to increased continuity of care by theprimary care provider, and the proportion of office visitsoccurring in primary care increased significantly,along with primary care visits for the patients' specificchronic conditions. In addition, the proportion ofpatients with each condition making urgent-care visitsdecreased substantially, and CHD patients reducedtheir hospital admissions and length of stay.
Although we do not have enough data about theoverall patient population of the medical group to beas definitive about identifying trends, improved accessdid not appear to be associated with changes inpatients' ED visits or hospital admissions. The totalcost of care for those with chronic diseases increasedby 10% to 20% over the 3-year study period, but thecost of care for all health plan members increased proportionately.It is likely that most of these costincreases reflect national and regional healthcare costtrends, and are not attributable to increased access toprimary care.
These results may be disappointing to those who areenthusiastic about access improvement. However, theresults are actually reassuring, because some havefeared that such changes might decrease access to carefor patients with chronic conditions. The access changedid appear to be associated with these patients receivingmore of their general and disease-specific care intheir primary care clinic with the same clinician, apparentlywith less need to be deferred to urgent-care sites.Also, their overall visit frequency did not decline. Fullassessment of the effect of access changes on thesepatients must await studies of quality-of-care measures.Independently of this study, the health plan conductsyearly satisfaction surveys of a sample of patientswith diabetes. During this time period, their overall satisfactionwith quality of care and service increased significantly,from 36% to 55% reporting being verysatisfied.
The ED data deserve separate comment. Althoughthere was no change in ED use for patients with thesechronic conditions, overall ED use increased for allpatients of the staff-model medical group as well as forall contracted groups during this time. That is part of anational trend, with Centers for Disease Control andPrevention data showing an 8% increase in ED visitsper 100 population between 1997 and 2001, and agreater increase for individual EDs because many haveclosed over this time period.13 Most of this increase alsohas been demonstrated to be caused by insuredpatients, not uninsured patients.14 Thus, the lack ofchange in ED use for these patients may actually representa stabilization in the face of a secular trend toincreased use.
What might have caused the reduction in urgent-careuse, hospitalizations for CHD patients, and possibly EDuse? One possibility is the increase in continuity of care.Several studies have found that increasing continuity ofcare is associatedwith fewer ED andurgent-care visits,as well as a lowerlikelihood of hospitalization.15-17Raddish et al analyzeddata from 6health maintenanceorganizationsand foundcontinuity alsowas associatedwith a decrease inthe number of outpatientvisits, disease - specificcosts, and totalpharmacy costs for patients with each of 4 chronic diseases.16 Of course, the findings in this study also maysimply represent the result of easier access for urgentproblems. Plauth et al surveyed adult health maintenanceorganization members who sought care in itsurgent-care center.18 Of the 421 patients responding,25% said they were unable to get an appointment withtheir primary care physician and 47% said that theywould have preferred to see their primary care physicianwithin 1 or 2 days.
As in any observational study of naturalistic changesover time, we cannot separate the effects of accesschange from the effects of other concurrent changes inthe care delivery system or its patients. It is possiblethat access effects on utilization and cost were confoundedby the concomitant move to a centralizedappointment scheduling system or the transition to productivity-based physician salaries. The lack of a comparisongroup that did not experience access improvementsalso limits our ability to be sure that theaccess changes were the cause of any changes. Resultsfrom this medical group also may not be generalizable toother outpatient settings. Nevertheless, this medicalgroup was able to make a major change in patient accessthat provided an opportunity to study the effects of thatchange in access on utilization and costs of healthcare.Finally, we also are limited by not having data about anychange in quality of care over this time period.
We conclude that in this study, access improvementshad little overall impact on utilization and costs forpatients with diabetes, CHD, or depression. Fears thatimplementation of advanced access will reduce the frequencyof primary care visits or increase hospitalizationsfor patients with chronic disease appear to beunfounded. Instead, there were some potentially importantchanges in the primary care of these patients thatmight have had beneficial effects on the dimensions ofquality other than timeliness (ie, safety, effectiveness,efficiency, equity, and patient-centeredness) identifiedby the Institute of Medicine. We hope that others willadd to this evaluation with studies that provide additionalinsight about how changes in access affect healthcareprocesses and outcomes for a variety of people andconditions.
We are grateful to Mary Hroscikoski, MD, for her coordinationof this project and her many thoughtful contributions to data collectionand analysis.
From HealthPartners Research Foundation, Minneapolis, Minn.
This project was supported by grant 041868 from The Robert Wood JohnsonFoundation through the Improving Chronic Illness Care Initiative.
Address correspondence to: Leif I. Solberg, MD, HealthPartners Research Foundation,PO Box 1524, MS#21111R, Minneapolis, MN 55440-1524. Delivery address: 8100 34thAve S, 11th Fl, Bloomington, MN 55425. E-mail: firstname.lastname@example.org.
Crossing the Quality Chasm: A New Health System for the 21st Century.
1. Institute of Medicine, Committee on Quality of Health Care in America.Washington, DC: National Academy Press; 2001.
N Engl J Med.
2. McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered toadults in the United States. 2003;348:2635-2645.
Priority Areas for National Action: Transforming Health CareQuality.
3. Institute of Medicine, Committee on Identifying Priority Areas for QualityImprovement. Washington, DC: National Academies Press; 2003.
4. Murray M, Berwick DM. Advanced access: reducing waiting and delays in primarycare. 2003;289:1035-1040.
Must patients wait? Jt Comm J Qual Improv.
5. Murray M, Tantau C. 1998;24:423-425.
Manag Care Q.
6. Murray M, Tantau C. Redefining open access to primary care. 1999;7:45-55.
7. Murray M, Bodenheimer T, Rittenhouse D, Grumbach K. Improving timelyaccess to primary care: case studies of the advanced access model. 2003;289:1042-1046.
Jt Comm J Qual Safety.
8. Solberg LI, Hroscikoski MC, Sperl-Hillen JM, O'Connor PJ, Crabtree BF. Keyissues in transforming healthcare organizations for quality: the case of advancedaccess. 2004;30(1):15-24.
Am J ManagCare.
9. O'Connor PJ, Rush WA, Pronk NP, Cherney LM. Identifying diabetes mellitus orheart disease among health maintenance organization members: sensitivity, specificity,predictive value, and cost of survey and database methods. 1998;4:335-342.
J Community Health.
10. Given CW, Branson M, Zemach R. Evaluation and application of continuitymeasures in primary care settings. 1985;10:22-41.
J Clin Epidemiol.
11. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for usewith ICD-9-CM administrative databases. 1992;45:613-619.
J Chronic Dis.
12. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifyingprognostic comorbidity in longitudinal studies: development and validation. 1987;40:373-383.
13. McCaig LF, Burt CW. National Hospital Ambulatory Medical Care Survey:2001 emergency department summary. 2003;335:1-29.
Insured Americans Drive Surge in EmergencyDepartment Visits.
14. Cunningham PJ, May JH. Washington, DC: Center for Studying Health System Change;2003.
Am JPublic Health.
15. Mainous III AG, Gill JM. The importance of continuity of care in the likelihoodof future hospitalization: is site of care equivalent to a primary clinician? 1998;88:1539-1541.
Am JManag Care.
16. Raddish M, Horn SD, Sharkey PD. Continuity of care: is it cost effective? 1999;5:727-734.
Arch Fam Med.
17. Gill JM, Mainous III AG, Nsereko M. The effect of continuity of care on emergencydepartment use. 2000;9:333-338.
Am J Manag Care.
18. Plauth AE, Pearson SD. Discontinuity of care: urgent care utilization within ahealth maintenance organization. 1998;4:1531-1537.