Evaluation of the Relationship Between a Chronic Disease Care Management Program and California Pay-for-Performance Diabetes Care Cholesterol Measures in One Medical Group

BACKGROUND: Pay for performance (P4P) is a business model in which health plans pay provider organizations (medical groups) financial incentives based on attainment of clinical quality, patient experience, and use of information technology. The California P4P program is the largest P4P program in the United States and represents a potential revenue source for all participating medical groups. The clinical specifications for the California P4P program are based on the national Committee for Quality assurance (nCQa), Health Plan Employer Data, and information set (HEDis), and each clinical measure has its own benchmark. in 2005, participating medical groups were paid on the basis of 9 clinical measures that were evaluated in the 2004 measurement year. The cholesterol testing measure represented 4.44%-7.14% of the total P4P dollars available to participating medical groups from the health plans. OBJECTIVES: To (1) compare the percentage of medical group members aged 18 to 75 years with diabetes (type 1 or type 2) who received a low density lipoprotein cholesterol (lDl-C) test and attained lDl-C control (less than130 mg per dl) after enrolling in a chronic disease care management (CDCM) program with similar members managed by routine care, and to (2) assess the potential effect of CDCM on the quality performance ranking and financial reimbursement of a medical group reporting these measures in the 2004 California P4P measurement year. METHODS: This is a retrospective database review of electronic laboratory (lab) values, medical and hospital claims, and encounter data collected between January 1, 2003 and December 31, 2004 at 1 California medical group comprising 160 multispecialty providers. Requirements were continuous patient enrollment in 1 of the 7 health plans participating in P4P during the measurement year (2004) with no more than 1 gap in enrollment of up to 45 days. Patients aged 18 to 75 years were included in the diabetes cholesterol measure (denominator) if they had at least 2 outpatient encounters coded for a primary, secondary, or tertiary diagnosis of diabetes (International Classification of Diseases, Ninth Revision, Clinical Modification code 250.xx, 357.2, 362.0, 366.41, 648.0) or 1 acute inpatient (Diagnosis Related Group code 294 or 295) or emergency room visit for diabetes. Lab values were obtained from multiple sources, including archived lab databases during the same measurement period (numerator). The CDCM program provided education and recommendations for diet, lifestyle, and medication modification delivered by a multidisciplinary team of nurses, pharmacists, and dieticians, and this intervention was compared with routine care for patients not enrolled in the CDCM program. RESULTS: Of the 54,000 health plan members enrolled in this medical group under capitated reimbursement, 1,859 patients (3.4%) met the California P4P specifications for eligibility for the diabetes cholesterol measures and were evaluated. Of these, 8.9% (165/1,859) were followed by the CDCM program and 91.1% (1,694/1,859) by routine care. The lDl-C lab testing rate for patients in the CDCM program was 91.5% (151/165), and the lDl-C goal rate was 78.2% (129/165) compared with 67.8% (1,148/1,694) and 55.7%, respectively, for routine care (P less than 0.001 for both comparisons). if the lDl-C lab testing and goal attainment rates for the CDCM group were compared with rates for peer medical groups, this medical group would have scored in the 75th and 90th percentiles, respectively, corresponding to an annual revenue potential of $28,512 for this medical group if the total incentive payment from the health plan was $1 per member per month (PMPM), or $57,024 if the total incentive P4P payment was $2 PMPM. CONCLUSIONS: Preliminary data from 165 patients with diabetes managed in a CDCM program in a medical group operating under a small P4P financial incentive showed higher rates of lDl-C lab testing and goal.


What is already known about this subject
•P ay for Performance (P4P) programs alone have not been shown to improve quality of care.

What this study adds
•T his is the first study to show the potential impact that CDCM may have on the P4P rankings and financial payouts for amedical group. P ay for performance (P4P) is ab usiness model by which health plans pay provider organizations (medical groups) for consistently demonstrating high levels of quality performance based on established criteria. [1][2][3][4] The results of the P4P clinical measures arep ublicly available and areo ften discussed during contract negotiations between medical groups and health plans. California has the largest and most comprehensive P4P program; however,P 4P programs exist nationwide. [1][2][3][4][5][6] In fact, ar ecent survey demonstrated that most health plans that offer commercial health maintenance organization (HMO) products BaCKGROunD: Pay for performance (P4P) is abusiness model in whichhealth plans pay provider organizations (medical groups) financial incentives based on attainment of clinical quality,patient experience,and use of information technology.The California P4P programisthe largest P4P programinthe united states and represents apotential revenue source for all participating medical groups. The clinical specifications for the California P4P programare based on the national Committee for Quality assurance (nCQa), Health Plan EmployerData, and information set (HEDis), and eachclinical measurehas its ownbenchmark. in 2005, participating medical groups werepaid on the basis of 9clinical measures that wereevaluated in the 2004 measurement year. The cholesterol testing measurerepresented 4.44%-7.14% of the total P4P dollars available to participating medical groups from the health plans.
OBjECTivEs: To (1) comparethe percentage of medical group members aged 18 to 75 years with diabetes (type 1ortype 2) who received alowdensity lipoprotein cholesterol (lDl-C) test and attained lDl-C control (<130 mg per dl) after enrolling in achronic disease caremanagement (CDCM) programwith similar members managed by routine care, and to (2) assess the potential effect of CDCM on the quality performance ranking and financial reimbursement of amedical group reporting these measures in the 2004 California P4P measurement year.
METHODs: This is aretrospective database review of electronic laboratory (lab) values, medical and hospital claims, and encounter data collected between january1,2003 and December 31, 2004 at 1California medical group comprising 160 multispecialty providers. Requirements were continuous patient enrollment in 1ofthe 7health plans participating in P4P during the measurement year (2004) with no morethan 1gap in enrollment of up to 45 days. Patients aged 18 to 75 years wereincluded in the diabetes cholesterol measure(denominator) if they had at least 2outpatient encounters coded for aprimary, secondary, or tertiarydiagnosis of diabetes ( International Classification of Diseases, Ninth Revision, Clinical Modification code 250.xx, 357.2, 362.0, 366.41, 648.0) or 1acute inpatient (Diagnosis Related Group code 294 or 295) or emergencyroom visit for diabetes. lab values wereobtained from multiple sources, including archived lab databases during the same measurement period (numerator). The CDCM program provided education and recommendations for diet, lifestyle,and medication modification delivered by amultidisciplinaryteam of nurses, pharmacists, and dieticians, and this intervention wascompared with routine carefor patients not enrolled in the CDCM program.
REsulTs: Of the 54,000 health plan members enrolled in this medical group under capitated reimbursement, 1,859 patients (3.4%) met the California P4P specifications for eligibility for the diabetes cholesterol measures and wereevaluated. Of these,8.9% (165/1,859) werefollowed by the CDCM programand 91.1% (1,694/1,859) by routine care. The lDl-C lab testing rate for patients in the CDCM programwas 91.5% (151/165), and the lDl-C goal rate was78.2% (129/165) compared with 67.8% (1,148/1,694) and 55.7%, respectively,for routine care(P <0.001 for both comparisons). if the lDl-C lab testing and goal attainment rates for the CDCM group werecompared with rates for peer medical groups, this medical group would have scored in the 75th and 90th percentiles, respectively,corresponding to an annual revenue potential of $28,512 for this medical group if the total incentive Evaluation of the Relationship Between aChronic Disease Care Management Program and California Pay-for-Performance Diabetes Care Cholesterol Measures in One Medical Group in metropolitan areas use P4P in their medical group contracts. 5 The Integrated HealthcareA ssociation (IHA) is the statewide leadership group that developed and coordinates the California P4P process. According to IHA, the purpose of the California P4P initiative is to create ab usiness model that financially rewards quality performance through as tandardized measurement set. 6 Clinical measures, patient satisfaction, and the ability of the medical group to integrate information technology into patient careare the criteria for payment in the California P4P model. 1,4 Currently,A etna, Blue Cross of California, Blue Shield of California, CIGNA, Health Net, PacifiCare, and WesternHealth Advantage participate in the California P4P initiative. 1,4,7 In 2004, the California P4P program payments wereweighted 40% for up to 9clinical measures; 40% for patient satisfaction, as determined by the consumer assessment survey comprising 16 primaryq uestions; and 20% for information technology, assessed by the ability of each medical group to integrate clinical electronic datasets for both population management and clinical decision making at the point of care. 4,7 The health plan provides payment on the basis of ad ollar amount per member per month (PMPM) that is calculated in addition to the contracted PMPM amount for the deliveryo fc aret om embers in the medical group. 7 The medical group may receive incentive incremental P4P financial payment for scoring in any of the 3categorymeasures (clinical, patient experience, or information technology) and any number of the individual measures (i.e., as few as 1o ft he possible 9c linical measures) as long as the medical group scores in the appropriate percentile ranking as determined by the health plan. 7 The data for the 2004 measurement year weren ot reported until May 27 of the 2005 reporting year. 4 IHA then compiled the 2004 report data and submitted final information to the health plans by July 2005. Health plans had until the end of 2005 to pay the participating medical groups on the basis of their performance in 2004. This timeline contributed to significant delays in the availability of final P4P reporting numbers and final payment to the medical group.
The amount paid to the medical group varies by each health plan that participates in the California P4P program. The payment amount is determined from the medical group' ss core in the 20th to 50th percentile (depending on the health plan) or higher when compared with astatewide benchmark of California medical groups. 6 In the first 5y ears of the P4P program, more than $145 million in incentive payments was distributed to 210 medical groups, representing 35,000 providers and more than 7m illion commercial HMO members in California, or an amount that averaged less than $1 PMPM. 1,6,8 Few published studies have evaluated the effect of P4P incentive payments on improvements in clinical quality. 9,10 Medical groups participating in the California P4P initiative may increase revenue if consistent quality performance can be demonstrated for any of the 9clinical measures and/or the nonclinical measures. The clinical specifications for the California P4P program areb ased on the National Committee for Quality Assurance (NCQA) Health Plan Employer Data and Information Set (HEDIS), and each clinical measureh as its own benchmark (see Table 1). 1,4 Of the 11 possible clinical measures in the 2004 California P4P initiative, 4werebased on comprehensive diabetes care, making diabetes carethe most heavily weighted of the clinical measures. 4 Although 11 clinical measures arel isted in IHA specifications for the 2004 measurement year,t he cholesterol measures for cardiovascular disease and for diabetes low-density lipoprotein cholesterol (LDL-C) measures arec ombined for reporting and payment purposes (see Table 1; measures 1and 3are combined for payment purposes). 4 The reason for the combined reporting of the diabetes cholesterol measurew ith the cardiovascular cholesterol measurei st he extremely low reported numbers for the cardiovascular cholesterol clinical specifications. 4 All 11 clinical measures areincluded in the reporting of clinical quality to IHA, but plans may pay the medical group on only some or none of the clinical measures. For example, Health Net paid participating medical groups on 9ofthe 11 possible clinical measures, while the remaining 6h ealth plans that participate in the California P4P initiative provided payment to medical groups on 7ofthe 11 possible clinical measures. 7 The diabetes carem easures not only evaluate the medical group' sa bility to obtain laboratoryv alues for glycosylated hemoglobin (A1C) and LDL-C for target patients, but also the medical group' sa bility to attain goal A1C and LDL-C levels for its health plan members, as established by the California P4P initiative and HEDIS criteria. 4 Under the comprehensive diabetes careclinical measures for the 2004 measurement year, goal levels for A1C wered efined as <9%a nd for LDL-C as <130 mg per dL. 4 Current practice guidelines from the American Diabetes Association and the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATPI II) established goal A1C levels of <7% and LDL-C levels of <100 mg per dL for the treatment of diabetes and dyslipidemia, respectively. 11 MMG has been using the Mercy Heart Institute' schronic disease caremanagement (CDCM) program since its inception in 1997. The institute' sCDCM program is anot-for-profit, hospital-based, telephone surveillance disease management (DM) program that supports patients throughout the greater Sacramento area.
One facet of this program is ad yslipidemia risk-reduction program designed to educate and treat patients who arediagnosed with or who area tr isk for developing cardiovascular disease. This program was developed collaboratively with MMG. The Evaluation of the Relationship Between aChronic Disease Care Management Program and California Pay-for-Performance Diabetes Care Cholesterol Measures in One Medical Group first patients weree nrolled in 1999, and approximately 1,000 patients arec urrently enrolled in the program. Physicians from the medical group make up the largest source of patient referrals and enrollment for the CDCM program. Initially,patients referred to the program areo ffered a1 -time patient education class, during which they receive information on cardiovascular disease, such as pathophysiology,risk factors, and ways to minimize risk, including dietarymodifications and exercise recommendations. Follow-up includes ordering laboratoryv isits and prescribing lipid-lowering therapy as dictated by CDCM treatment protocol approved jointly by the medical group and the health plan.
The treatment protocol has 3t reatment tracks to which the patients area ssigned. Approximately 35% of the patients are identified as cases for primaryp revention (i.e., no historyo f coronarya rteryd isease [CAD] or diabetes or stroke), 25% as cases for secondaryprevention (i.e., historyofCAD, myocardial infarction, stroke, etc.), and 40% as diabetes risk-equivalent prevention. This treatment protocol was developed using the NCEP ATPI II as its framework. Am ultidisciplinarya pproach was chosen to ensuret he best possible caref or the patient. Physicians, dietitians, nurses, pharmacists, exercise physiologists, and social workers wereall involved in the development of the treatment protocol. The protocol methodology was reviewed and approved by various hospital and physician committees. Additionally,t he Joint Commission on the Accreditation of HealthcareOrganizations (JCAHO) reviewed the program and as of the fall of 2005, the program has been recognized by JCAHO as ahyperlipidemia-certified program.
The treatment protocol includes both pharmacologic (medication algorithm) and nonpharmacologic (therapeutic lifestyle management education) methods for reducing risk. Nurse and pharmacist caremanagers areused to carefor patients using the protocol and areallowed, under the physician' ssigned order,toprescribe medications and order laboratoryassessments as described in the protocol. Quality assessment takes place at regular intervals to assurethe success of the program in achieving LDL-C targets while avoiding medication misadventure.
Enrollment and participation in the CDCM program in this medical group is voluntary. However,the referring physician in the medical group must provide as igned order that allows the CDCM program to caref or the patient under the standardized treatment protocol. Adiagnosis of dyslipidemia and aphysician referral aret he only criteria for enrollment in the CDCM lipid management program. Thereisnopatient recruitment or other selection criteria. Once enrolled, patients area ssessed, and an individualized careplan is established.
The purpose of this study was to (1) comparethe percentage of medical group members aged 18 to 75 years with diabetes (type 1o rt ype 2) who received an LDL-C test and attained LDL-C control (<130 mg per dL) after enrolling in the CDCM program with similar members managed by routine care, and (2) assess the potential effect of CDCM on the quality performance ranking and financial reimbursement of am edical group reporting the diabetes cholesterol management measureofthe 2004 California P4P initiative.

■■ Methods
This study was approved by the Committee on Human Research at the University of California, San Francisco, and the Institutional Review Boarda tC atholic HealthcareW est. MMG is a1 60-physician, multispecialty medical group in the greater Sacramento area that participates in the California P4P program. MMG uses at eam-based CDCM program to help manage cholesterol for patients in the medical group. Am edical group can use 2m ethods to report P4P clinical data-passive and active reporting. Passive reporting involves using the claims and encounter data already submitted to the health plan to calculate and report the clinical measures. With this health plan reporting method, the medical group is completely reliant on the plan to report data to IHA on the medical group' sb ehalf. In active reporting, the medical group uses internal records, including its claims, encounter,a nd electronic laboratorya nd pharmacy claims data to self-report the clinical measures to IHA.
MMG uses active (self-) reporting of P4P clinical measures. In order for amedical group to self-report clinical measures for the California P4P program, the medical group must undergo ar igorous audit process. An outside auditor approved by NCQA and IHA validates the self-reporting process used by the medical group. This audit process verifies that the medical group is accurately capturing data for the purposes of reporting P4P clinical measures by evaluating membership data, claims processing systems, data analysis, decision support processes, and the data linkage between electronic databases. This audit is performed everyyear.
Data collection and compilation for the California P4P clinical measures at this medical group area ll electronic, using acombination of claims, encounter,and actual laboratorydata. In 2004, MMG medical records weres tored in paper charts, but encounter forms, claims, and some laboratoryv alues were processed and stored electronically.T he medical group used these electronic systems for P4P self-reporting. MMG started using an electronic medical recordi n2 006, but the P4P selfreporting process has not changed as ar esult. Laboratoryd ata for LDL-C and A1C ares tored electronically on al aboratory server and sent to the medical group on am onthly basis from the lab database vendor.Laboratoryvalues may be obtained by the provider for individual patients in real time, but the group is dependent on the monthly laboratoryreports for the purpose of P4P data aggregation.
New P4P measures for 2005 prompted targeted educational interventions and financial incentives by MMG to improve the medical group' sp erformance for these measures. 1 Because of the substantive changes in physician education and medical group interventions that took place in 2005 and 2006, it was determined that 2004 data werea ssociated with fewer confounding variables and werep erhaps the most valid for evaluating the efficacy of the CDCM program in achieving P4P target goals compared with usual careprovided during in-office physician visits.

Pay for Performance Clinical Measure: Diabetes Care-LDL-C Testing and Control
The purpose of the diabetes carec linical (2-part) measurei s to determine the percentage of medical group members aged 18 to 75 years with diabetes (type 1or2)who received an LDL-C test and had an LDL-C <130 mg per dL. 4

Inclusion Criteria
Patients included in this measurew erec ommercial HMO members continuously enrolled in MMG and aP4P participating health plan during the measurement year (2004), with no moret han 1g ap in enrollment of up to 45 days during that time. 4 Patients with diabetes who werea ged 18 to 75 years as of December 31, 2004 werei ncluded in the study.T he eligible population for the diabetes careclinical measurewas identified using pharmacy data and/or hospital/medical claims and encounter data. 4 Complete pharmacy data weren ot available for MMG during the 2004 measurement year,a nd thus only the claims/encounter method was used to identify the eligible diabetes population. On the basis of the claims/encounter data, patients werei dentified as having diabetes if they had 2 face-to-face outpatient encounters for diabetes with different dates of service in an ambulatorysetting or 1face-to-face acute (inpatient or emergency department) visit for diabetes during 2003 or 2004. 4 International Classification of Diseases, Ninth Revision, Clinical Modification codes, Diagnosis Related Group codes, and Current Procedural Terminology codes wereused to identify diabetes as the primary, secondary, or tertiaryreason for the visit and arelisted in Table 2. 4 These criteria determined the denominator for this measure. See Figuref or afl ow diagram describing patient selection.
The serum LDL-C value for the most recent cholesterol test and the number of patients with aLDL-C laboratoryvalue in 2003 or 2004 wereused for the numerator for this clinical measure. 4 The medical group could use claims/encounter data to determine if the LDL-C was performed, or electronic laboratorydata if at least the date and result of the LDL-C wereincluded in the electronic information. 4 Because the actual laboratoryvalue is required to report "good" control, this medical group used the electronic laboratorydata to determine the LDL-C level for patients in the eligible population. According to the 2005 Clinical Measure Specifications in the P4P program, patients who had no LDL-C performed in 2003 or 2004 wereassumed to be in poor control. 4 At riglyceride level >400 mg per dL would lead to inaccurate LDL-C calculations and, therefore, patients with at riglyceride level >400 mg per dL wereconsidered to be in poor control but would meet the criteria for inclusion in the numerator for the proportion of patients who received LDL-C testing. 4

Exclusion Criteria
MedicareA dvantage members weren ot included in any of the clinical measures for the 2004 measurement year and weren ot included in the P4P population for the diabetes caremeasures. 4 Patients not assigned to ap rimaryc arep hysician, and patients enrolled in ah ealth plan that did not participate in P4P,w ere also excluded from the study population.
The same denominator value was used for the LDL-C testing and control measures (the eligible population of patients with diabetes), but the numerator was reported separately for each measure. Apatient may appear in the numerator for both LDL-C testing and LDL-C good control.
Health plans that participate in the California P4P program make incentive payment to the participating medical group based on that medical group' squality performance measures compared with peer medical groups within the state. Each clinical measure

Evaluation of the Relationship Between aChronic Disease Care Management Program and California Pay-for-Performance Diabetes Care Cholesterol Measures in One Medical Group
receives ap ercentile ranking. The LDL-C testing and control measures for cardiovascular carea nd diabetes aree valuated separately by IHA. Because of the potentially small number of patients in the cardiovascular group, the cardiovascular and diabetes numbers area dded together for the cholesterol measures for the purposes of payment to the medical group. 4 The cardiovascular cholesterol numbers for this medical group wereas mall fraction (<2% [37/1,896]) of total patients in the LDL-C P4P measure, and they wereexcluded from the final P4P numbers for the purposes of this paper.The total reported rates (%) for LDL-C testing and control did not change as ar esult of excluding the cardiovascular patients from the LDL-C P4P measure (Table 3).

Diabetes Care-LDL-C Testing and Control 2005 Incentive Payment Amounts
Payment is generally based on the medical group' sa bility to scorei nt he 50th percentile or higher when compared with other (peer) medical groups. Generally,i ft he medical group scores between the 50th and 74th percentile, it is paid one half of the total possible incentive payment, and groups in the 75th to 99th percentile receive the full incentive payment amount. 6 As previously mentioned, each of the 7health plans determines the percentile ranking for incentive payment. 6 No health plan provides payment for percentile rankings below the 20th percentile. 6 Payment is commonly provided on aP MPM basis and is determined by each health plan. Health plan payout for the testing and goal attainment for the LDL-C measures varied for the 2005 report year (2004 measurement year). All participating plans paid medical groups for LDL-C laboratoryt esting (from al ow of 4.44% to ah igh of 7.14% of total available P4P dollars), but only 1p lan paid medical groups for the proportion of patients with LDL-C <130 mg per dL. 7 Both LDL-C testing and goal attainment rates werereported to IHA in 2005. On the basis of the 2005 payment amounts, all potential P4P payouts for the LDL-C testing measure in the data analysis wereb ased on the lowest plan payout of 4.44% of the total available P4P dollars. This method reduced the potential for overestimation of missed revenue opportunity for calculations involving the LDL-C testing measure. Because only 1h ealth plan out of 7p aid on the LDL-C goal rate <130 mg per dL, this measurewas assumed to have insignificant reimbursement to the medical group and thereforew as not considered in the potential payment calculations.

Data Analysis
Data werec ollected retrospectively from claims and encounter forms to determine the eligible population, as defined by the diabetes carec holesterol measures. 4 Laboratoryd ata were obtained from the Mercy Laboratories electronic database. The entiree ligible population (denominator) was identified, and then patients followed by the CDCM program weree xtracted from the whole P4P denominator and evaluated separately,using the available laboratorydata. It was assumed that if apatient was not enrolled in the CDCM program, the patient received routine care. Patients enrolled in the CDCM program received care from their primaryc arep roviders as well as from CDCM team members. Queries using the Microsoft Access database were used to identify patients who had an LDL-C test performed and to determine the LDL-C value. This process was verified for accuracy by the NCQA-and IHA-approved P4P self-report auditor.
Statistical analysis was performed with Minitab release 14 statistical software. Achi-squared statistical test was used for all discrete data and 2-sample t test was used for continuous data to determine statistical differences between the CDCM program and the routine careg roup ( P value was set to be <0.05 for statistical significance).

LDL-C Testing Rates
The LDL-C testing rate for all P4P eligible patients (both CDCM program and routine care) in this medical group in 2004 was 69.9%. The LDL-C testing rate for patients in the CDCM program was 91.5% versus 67.8% for the routine caregroup ( P <0.001, Table 4). When compared with peer groups, the patients in this medical group overall scored less than the 20th percentile for the lab testing rate (70% raw score), representing no revenue for this portion of the LDL-C testing measure. However,i ft he CDCM program patients' LDL-C testing rate (92%) had been compared with peer groups, MMG would have scored higher than the 75th percentile. This scorewould have resulted in full payment from the health plans for this measure(see Table 5).

LDL-C Goal Attainment Rates
The LDL-C goal attainment rate (LDL-C <130 mg per dL) for all medical group patients (both CDCM and routine carepatients) in 2004 was 57.7% (1,072/1,859). The LDL-C goal attainment rate for the CDCM program was 78.2%, significantly higher than the 55.7% rate for the routine careg roup ( P <0.001; see Table  4). When compared with peer medical groups, MMG scored in the 75th percentile for the goal attainment rate, representing full revenue for this portion of the diabetes careLDL-C clinical measuref romt he 1h ealth plan that provided payout for the LDL-C goal attainment rate. However,i ft he CDCM goal attainment rate had been compared with peer groups, MMG would have scored in the 90th percentile, which would have improved publicly reported scores for the medical group but would not have resulted in higher revenue for the medical group (see Table 5).
Although no financial awardorpublic reporting exists in the P4P measures for medical groups that have LDL-C goal attainment at the higher standardo f< 100 mg per dL for patients with cardiovascular disease or cardiovascular disease risk-equivalent (diabetes), this rate was evaluated in the present analysis, since 100 mg per dL is the standarddescribed in national guidelines. 11 The medical groups' overall rate for LDL-C <100 mg per dL was 36.3%. The CDCM program had an LDL-C <100 mg per dL rate of 46.7%, which was astatistically significant higher achievement than 35.2% in the routine caregroup ( P =0.004, Table 4).

Potential Financial Payouts
Because of confidentiality issues, the actual payment amount to the medical group cannot be disclosed. The contracted rates for P4P payment from each health plan aren egotiated directly between the medical group and health plan and varya mong the health plans. As previously discussed, all health plans pay the medical group adollar amount PMPM based on their percentile rankings for each measure, with 100% payment for placement in the 75th percentile or higher among all medical groups, 50% payment for placement between the 50th and 74th percentiles, and minimal or no payment if lower than the 50th percentile (only Blue Cross of California provided payments on as liding scale to medical groups that scored between the 20th and 49th percentiles). 7 All health plans made incentive payment to medical groups for LDL-C lab testing for the 2005 P4P reporting year,w ith a minimum of 4.44% of the total available PMPM dollar amount. Therefore, if $1 PMPM werep rovided to the medical group for all clinical measures, the payment amount for am edical group scoring in the 75th percentile or higher for the LDL-C testing rate would be $0.044 PMPM. However,i ft he medical group scored between the 50th and 74th percentiles for the LDL-C testing rate, the total payment would only be 2.22%, or $0.022 PMPM. If the medical group scored lower than the 20th percentile, then it would receive no payment for that clinical measure. If the contracted rate were$2PMPM, then the diabetes LDL-C testing measurewould represent amaximum payment of $0.088 PMPM.
The PMPM payout to the medical group is based on the number of commercial members that aree nrolled in P4P participating health plans. This medical group had approximately 54,000 commercial members in 2004 that werem embers of P4P participating health plans, and thereforet he payout for this LDL-C testing measurew ould be $0.044 PMPM x 54,000 members x 12 months, or $28,512 for a$ 1P MPM payment incentive, and $0.088 PMPM x 54,000 members x 12 months, or $57,024 for a$2PMPM payment incentive.
This medical group scored lower than the 20th percentile for the LDL-C testing rate in 2005 for the 2004 measurement period, and, therefore, no P4P revenue was earned for this part of the measure. If all 1,859 patients with diabetes in this medical group had been enrolled in the CDCM program in 2004, the LDL-C testing rate for this medical group could have scored in the 75th percentile. This rate represents ad ifference of 4.44% of the total contracted rate, or $28,512 for a$1PMPM payment incentive or $57,024 for a$ 2P MPM payment incentive (see Table 5).

■■ Discussion
In this medical group, patients with diabetes who werefollowed by the CDCM program had significantly higher rates for both LDL-C testing and goal attainment than did those in routine care. Results of other studies on the effectiveness of team-based lipid management have also shown high LDL-C lab testing rates (96.7%-97.3%) and LDL-C <100 mg per dL goal attainment (56.5%-83.3%) in lipid management. [13][14][15] Limitations of those studies include no patient randomization, no comparator-control group, and no evaluation of end point outcomes such as morbidity or mortality.T his is the first study to pair the clinical results of team-based CDCM with emerging P4P initiatives showing both improved quality reporting and potential increased revenue for the medical group.
Financially,h ad this medical group scored in the 75th percentile for both the LDL-C lab testing and goal attainment portions of the diabetes LDL-C measure, revenue would have increased from $0 payout to ahypothetical $28,512 for a$ 1P MPM payment incentive. The results of the present study suggest that referring morepatients to the CDCM program may be aviable way to improve the overall performance scores for the medical group.
Revenue sharing is one way to provide afinancial incentive to increase medical group intervention in targeted DM programs. Currently,this hospital-based CDCM program does not generate revenue for the services it provides, but instead this not-forprofit hospital declares the expenses as acommunity benefit and part of its charity careobligation. This schema is consistent with the organization' smission to provide direct services to the poor and to partner with others in the community to improve the quality of life in the community it serves.
Because the CDCM program receives referrals from outside MMG, as hared revenue model is not viable at this time since the program would then need to bill all referring providers in order to be compliant with the Omnibus Reconciliation Act of 1993. This law precludes preferential pricing for services to certain groups of physicians to avoid the potential for garnering referrals. As contracts with other medical groups arereviewed (or renegotiated) by the CDCM program, it might be worthwhile to develop ar eturn-on-investment model to explorew hether P4P revenue sharing would be beneficial to the program. Likewise, other medical groups wishing to contract for DM services may explores hared revenue as an incentive in contracting with a program that is structured to bill for services.
In 2006, all but 1h ealth plan provided payment for both LDL-C testing and LDL-C goal attainment <130 mg per dL for the 2005 measurey ear,m aking it even morei mportant for medical groups to improve this clinical measure. 16 Therefore, programs like the CDCM program that increase LDL-C testing and goal attainment <130 mg per dL arei mportant from both aquality and financial perspective to P4P participating medical groups.
As previously described, al ong time separates the reporting of the data to the health plans and the date that the payouts arer eceived by the medical groups. 4 Medical groups do not know their percentile rankings and actual payment amount from the health plan for up to 12 months after the close of the measurement year.This makes financial projections for the P4P initiative difficult to predict. However,medical groups who selfreport do know the patients who aree ligible for P4P reporting throughout the measurement year,allowing the groups to target the noncompliant patients in time to make apositive difference in the reported rate (e.g., referring noncompliant diabetes patients who miss an LDL-C lab test or arenot at LDL-C goal to aCDCM program).
Reporting P4P clinical measures is ac omplicated and timeconsuming process. At eam of experts comprising physicians, data analysts, pharmacists, and other key organizational personnel is critical to successful self-reporting. Although selfreporting for MMG made the reporting process much morelabor intensive, self-reporting was associated with improved MMG clinical measures compared with the passive report method that was used in the prior year for 2003. The self-report process also allows regular audits of the data so discrepancies can be clarified and gaps in carecan be targeted at the organizational level.
The CDCM program described herein was an existing program and therefored id not incur any up-front costs to the medical group. Other medical groups would need to allocate funds to initiate targeted patient care( DM) programs. Since this P4P program does not pay for improvement but only for percentile ranking, poor-performing groups cannot earnP 4P revenue to fund quality-improvement activities. This results in ap otential situation in which the good get better and the poor do not improve enough to earnrevenue. This situation has been previously described in the literatureb yR osenthal et al. for the cervical cancer,b reast cancer,a nd diabetes A1C clinical measures. 9 The P4P process is dynamic. In the 2005 and 2006 measurement years, MMG targeted multiple clinical measures for improvement, including breast cancer testing, cervical cancer testing, childhood immunizations, LDL-C testing rate and LDL-C goal attainment, A1C measurement, and A1C goal attainment. In targeting these measures, new strategies weree mployed to improve disease prevention and quality of care, including physician detailing (modeled after the CDCM program), patient letter campaigns (modeled after the CDCM program), use of patient advocates, increase in the number of patients enrolled in the CDCM program, and other chronic DM programs, such as the Diabetes Careprogram. Furthermore, physicians at MMG wereg iven financial incentives for their ability to meet certain quality measurement markers, including LDL-C testing and goal attainment rates for the diabetes carem easure. This targeted approach was associated with significant improvement for this medical group, which was recognized as atop 20% performing medical group in the subsequent (2005 measurement) year.W e chose the 2004 measurement year for the present study in an effort to reduce the confounding effects of these other medical group interventions that did not exist in 2004.
P4P programs have gained international attention, and there is great interest in whether P4P measures improve the quality of health carep rovided to patients at the medical group level. 2,5,10 Until further studies arep ublished evaluating the effect of P4P, medical groups must develop creative ways to improve their quality measures in order to keep pace with peer organizations. Team-based CDCM programs may become sources of revenue as well as am eans to avoid costs associated with lower-quality care.
Evaluation of the Relationship Between aChronic Disease Care Management Program and California Pay-for-Performance Diabetes Care Cholesterol Measures in One Medical Group rankings that would generate P4P incentive payments. Fifth, since not all medical groups have access to team-based CDCM, these CDCM program results may not be generalizable to other medical groups.

■■ Conclusions
This CDCM intervention in amedical group participating in P4P had ahigher rate of LDL-C testing (92%) compared with routine care( 68%) and ah igher proportion of patients who attained LDL-C goal (<130 mg per dL) than did those treated in routine care(78% versus 56%, respectively). If all patients with diabetes who werem easured by this P4P program had been enrolled in the CDCM intervention, this medical group would have attained the highest payout amount for the LDL-C testing measurea nd would have improved scores in public reporting of both P4P measures for LDL-C testing and LDL-C goal attainment.
Our results suggest but do not prove that af ocused CDCM intervention is effective in generating incentive revenue in aP4P program. Alarge opportunity also remains to show that a focused CDCM intervention could generate sufficient incentive payment to cover its costs. The literatureisspeckled with mostly anecdotes of success, including the results of aP 4P program established to improve diabetes carethat reportedly had areturn on investment of 1.6 to 1i nt he first year and 2.5 to 1i nt he second year of the sponsoring health plan. 17 This descriptive report without acontrol group was conducted in the Rochester area of upstate New York during 2003 to 2004.
LDL-C <100 mg per dL for patients with diabetes was not a clinical measurei nt his P4P program in 2004 but is the widely accepted LDL-C goal rate, according to NCEP ATPI II. 11 The CDCM program showed asignificantly higher rate of attainment of the LDL-C <100 mg per dL goal than did routine care. The Mercy Heart Institute' sC DCM program described in this paper was previously shown to improve LDL-C goal attainment for diabetes patients from 23.2% beforet he intervention to 56% after only 6m onths in the program. 14 This 56% rate of LDL-C goal attainment compares with 47% in the present study.

Limitations
First and foremost among the study limitations is the small sample size, with only 165 patients in the CDCM program. Second, an opportunity for selection bias exists since this was not ap rospective study with random assignment. We also did not control for confounding factors such as comorbid conditions, type of lipid-lowering agents, patient financial status, motivation, or intervention from other specialists (e.g., endocrinologist or cardiologist). The intervention and comparison groups in the present study weres tatistically different for age and gender. However,w hile the CDCM group might be different from the comparison, the practical significance of any differences would pertain to the goal attainment rate and not the testing rate because the patients in both groups needed lipid testing.
Third, we did not measurethe amount of time and resources required to operate the CDCM program and thereforecould not perform acalculation of returnoninvestment. We aretherefore unable to determine if the estimated payout opportunity for this medical group under the P4P program could have covered the cost of the CDCM program. While the CDCM program described in this paper does not represent ac ost to the medical group, thereare real costs that include salaryand benefits. Because the CDCM program is based in an ot-for-profit hospital, however, the program has been justified, in part, by the community benefit required by California State Senate Bill 697 for it to maintain not-for-profit tax-exempt status. 18 Fourth, the CDCM program is not am andatoryp rogram in this medical group, and therem ay be patients who choose not to participate, which would thereby reduce the opportunity for the CDCM program to elevate the entiremedical group into the