Effect of Physician Pay-for-Performance (P4P) Incentives in a Large Primary Care Group Practice
Presenter: Harold Luft, PhD1,2
Collaborators: Sukyung Chung, PhD1,2,
Latha Palaniappan, MD, MS1 Haya Rubin, MD, PhD
Laurel Trujillo, MD3 Eric Wong, MS1
1Palo Alto Medical Foundation Research Institute2Institute for Health Policy Studies, UCSF
3Palo Alto Medical Foundation
Supported by AHRQ Task Order HHSA290200600023
2
Empirical Evidence of P4P
• Recent studies of P4P show modest effects– Group level incentives
• Rosenthal et al. (2005): increase in cervical cancer screening, but no effect on mammography and HbA1c testing
• Roski et al. (2003): better documentation of tobacco use, but no change in provision of quitting advice
– Physician-specific (vs. no) financial incentives)• Levin-Scherz et al. (2006): increased diabetes screening, but
no effect on asthma controller prescription• Beaulieu & Horrigan (2005): improvement in most of the
process and outcome measures of diabetes care• Gilmore et al. (2007): improvement in most process of care
measures (e.g. cancer screening, diabetes care)• Financial incentives were generally accompanied by other
quality improvement efforts such as performance reporting
3
Empirical Evidence of P4P (cont.)
• Limitations of previous studies:– Payer-driven initiatives
• Quality measures and incentive schemes were given to, rather than chosen by, physicians or physician groups.
• Only part of the physicians’ patients were eligible for incentives.
– Incentives paid annually or at the end of the study• Effect of timing of receipt of payment, in addition to the provision
of performance reporting, is unknown.
– Based on claims data• Limited physician-level information; no opportunity to investigate
specific physician characteristics associated with incentives
4
Research Questions• Does a P4P program with physician-specific incentives
implemented in a large primary care group practice improve quality of care provided?
• Are there associations across measures (within physicians) in the effect of the incentive program?
• What physician characteristics affect variations in performance across physicians?
• Does the frequency of payment (quarterly vs. end-of-year) make a difference in performance?
5
Study Setting
• Palo Alto Medical Foundation (PAMF)– Non-profit organization – Contracts with 3 physician groups in Northern California
• Palo Alto Division (PAMF/PAD)– 5 sites: Palo Alto, Los Altos, Fremont, Redwood City, Redwood
Shores– Physician payment: based on relative value units of service– Electronic health records since 2000– Implemented physician-specific financial incentives in 2007
6
The Incentive Program• Physician-specific incentives based on own performance
• Comprehensive – All the primary care physicians (N = 179) and all their patients
regardless of insurance type– Family Medicine, Internal Medicine, Pediatrics
• Physician participation – in determining performance measures and incentive formula
• Frequency and amount of bonus payment – Physicians were randomly assigned to quarterly or year-end payment– Maximum bonus: $1250/qtr or $5000/yr (~2-3% of salary)– Payment delivered about 6 weeks following the evaluation quarter
7
The Incentive Program (cont.)• Various quality measures
– Both outcome and process measures– 10 were selected from a set of existing measures used for quality
assessment for several years– 5 new pediatrics-specific measures were selected based on AAP
guidelines; some were further modified during the year– These pediatric measures are excluded in our analyses
• Quarterly performance reporting– All the physicians were alerted by quarterly email with an electronic
link to quality workbook (a process in place for several years)– In 2007, the report was to be sent on the 24th day after the quarter
• Funds– IHA P4P incentives were supplemented by the organizational fund– Allowed application to all patients, not just those in IHA plans
8
Incentivized Quality MeasuresMeasure Description Category
Diabetes HbA1c control* HbA1c <=7 (diabetes patients) Outcome
Diabetes BP control Blood pressure <=130/80 (diabetes patients) Outcome
Diabetes LDL control* LDL <=100 (diabetes patients) Outcome
Asthma Rx*† Long-term controller prescribed (asthma patients) Process
Ht & Wt measured Height and weight measured for BMI calculation Process
Chlamydia screening*† Chlamydia testing done (eligible women) Process
Colon cancer screening Colon cancer screening complete (adults age 50+) Process
Cervical cancer screening Pap smear done (eligible women) Process
Tobacco Hx entered† History of tobacco use was asked and recorded Process
Percent score = [numerator (i.e. patients who met the guideline) / denominator (i.e. patients who were eligible for the recommended care)] X100* Similar measures (with different targets and population) were included in the IHA P4P program.†These measures apply to some pediatrics patients.
9
Other Quality Measures: Examples*Measure Description Category
Diabetes HbA1c control* HbA1c <=8 (diabetes patients) Outcome
Diabetes BP control Blood pressure <=140/90 (diabetes patients) Outcome
Diabetes LDL control* LDL <=130 (diabetes patients) Outcome
Hypertension BP control Blood pressure <=140/90 (hypertension patients) Outcome
Diabetes HbA1c check HbA1c was measured within the past 6 months Process
Diabetes BP check BP was measured within the past 12 months Process
Diabetes LDL check LDL was measured within the past 12 months Process
Hypertension BP check BP was measured within the past 12 months Process
Alcohol Hx entered History of alcohol use was asked and recorded Process
These were not incentivized, but were reported in the quality workbook.
Example: Quality Workbook for “Diabetes HbA1c Control”
6mGly7 Score-FAMP
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P1 P4 P7 P10 P13 P16 P19 P22 P25 P28 P31 P34 P37 P40 P43 P46 P49 P52 P55 P58 P61 P64 P67Provider
% score
Stretch goal (point=3)
Intermediate goal (point=2)
Minimum goal(point=1)
11
Incentive Formula
• Incentive payment = composite score * maximum amount
{=$1250/quarter}
• Composite score = (∑ achieved points) / (3 * #qualifying measures)
{Measures with <6 eligible patients for a physician in a quarter were not counted as a qualifying measure}
• Physicians with <4 qualifying measures in a quarter were not paid for the quarter
12
Number of Participating PhysiciansCategory N=167* Frequency (%)Incentive frequency Quarterly paid 77 46.1 Yearly paid 90 53.9Location Fremont 44 26.4 Los Altos 26 15.6 Palo Alto 76 45.5 Redwood City 9 5.4 Redwood Shores 12 7.2Department Family medicine 68 40.7 General internal medicine 56 33.5 Pediatrics 43 25.8
*Among the initial sample (n=179), 12 physicians did not participate in the program due to various reasons (e.g. lack of number of patients, medical/sabbatical leave, etc.).
13
Quality Scores, Number of Patients and Physicians at Quarter I, 2007
Quality measure #physicians with 6+ eligible patients at quarter 1(N=167)
Average # eligible patients/
physician(denominator)
Average # patients
meeting the guideline / MD
(numerator)
Percent score (numerator /
denominator x 100)
Clinical outcomesDiabetes HbA1c control 122 39 24 60%Diabetes BP control 122 49 24 51%Diabetes LDL control 122 43 25 57%Clinical processCervical cancer screen 123 529 418 77%Chlamydia screen 138 41 16 36%Colon cancer screen 122 315 153 45%Asthma Rx 136 21 19 92%Ht & Wt measured 152 926 747 71%Tobacco Hx entered 161 328 290 77%
14
Analyses • Does the 2006-07 change differ from 2005-06?
– H: (p2007 – p2006) – (p2006 – p2005) =0 – Outcome variables: Percent scores for incentivized and not-
incentivized measures– Unit of analysis: physician, each measure, each year (2005-2007)
• Does the trend in Palo Alto Division differ from the trend in other PAMF divisions? – H: PAD [(p2007– p2006)–(p2006–p2005)] – Other [(p2007– p2006)–
(p2006–p2005)] =0 – Outcome variables: Percent scores for quality measures similar to the
incentivized ones, but that were applied only to HMO patients– Unit of analysis: medical group, each measure, each year (2005-2007)
• Does the frequency of payment make a difference in quality?– H: p(Quarterly-paid, 2007) – p(Annually-paid, 2007) =0 – Outcome variables: Percent scores for incentivized measures– Unit of analysis: physician, each measure, four quarters of 2007
15
Quality Scores: Four Quarters, 200740
6080
100
perc
ent s
core
. Diabetes Diabetes Diabetes Asthma Cerv.cancer Chlamydia Colon cancer Ht Wt HbA1c ctrl BP ctrl LDL ctrl Rx screening screening screening measured
* **
** **
** **
* ** * *
*
*p<0.05; ** p<0.01Ref.cat.: Q1
Q1 Q2 Q3 Q4
16
Percent Scores: 2005-2007(incentivized measures)
Measures
Percent score (year)Diff.
pscore [P2006-P2005]
†‡
Diff. pscore [P2007-P2006]
†‡
Diff-in-diff pscore
[P2007-P2006]-[P2006-P2005]
‡P2005 P2006 P2007
Diabetes HbA1ccontrol (<=7) 58% 60% 62% **
Diabetes BP control (<=130/80) 47% 49% 53% ** ** **Diabetes LDL control (<=100) 60% 63% 60% ** ** (**)Cervical cancer screening 75% 77% 79% ** **
Chlamydia screening 36% 37% 38% *
Colon cancer screening 38% 40% 47% ** ** **Asthma Rx 91% 92% 92% *
Ht & Wt measured 68% 70% 73% ** **
Tobacco Hx entered 72% 75% 79% ** ** ***p<0.05; **:p<0.01†Statistics based on the results from the multilevel mixed-effects linear regression (z-statistics).‡Parentheses are used when the difference ((p2007 – p2006) or (p2006 – p2005)) is negative.
17
Percent Scores: 2005-2007 (reporting only measures)
Measures
Percent score (year)Diff.
pscore [P2006-P2005]
† ‡
Diff. pscore [P2007-P2006]
† ‡
Diff-in-diff pscore
[P2007-P2006]-[P2006-P2005]
‡P2005 P2006 P2007
Diabetes HbA1c control (<=8) 81% 81% 83% * **
Diabetes BP control (<=140/90) 77% 78% 81% ** **
Diabetes LDL control (<=130) 86% 88% 87% ** (**)Hypertension BP ctl (<=140/90) 64% 67% 72% ** ** **Hypertension BP check 90% 90% 90%
Alcohol Hx entered 67% 69% 73% ** ** **
*p<0.05; **:p<0.01†Statistics based on the results from the multilevel mixed-effects linear regression (z-statistics).‡Parentheses are used when the difference ((p2007 – p2006) or (p2006 – p2005)) is negative.
18
Effects of the Physician Incentive Program†
Measures
OLS Quantile regression
10th percentile
50th percentile
90th percentile
Interquantile difference90th vs.
10th 50th vs.
10th
Diabetes HbA1c control (<=7) 1.6 (1.2) 1.6 (2.8) 2.5 (1.7) 0.7 (1.9) -2.3 (3.2) 1.0 (2.7)Diabetes BP control (<=130/80)
2.9* (1.3) 4.9** (1.9) 1.1 (1.6) 4.7* (2.4) -0.2 (3.0) -3.8 (2.2)
Diabetes LDL control (<=100) -6.7** (1.2) -8.3** (2.8) -5.1* (2.1) -8.7** (2.5) -0.4 (3.6) 3.2 (2.9)Cervical cancer screening 1.0 (0.7) 2.2 (4.6) 0.6 (1.5) -0.9 (1.2) -3.1 (4.7) -1.6 (4.3)Chlamydia screening 1.1 (1.5) -2.7 (2.8) 3.0 (1.6) -0.4 (3.3) 2.3 (4.0) 5.7* (2.8)
Colon cancer screening 3.9** (1.0) 7.9* (3.3) 4.5* (1.9) -0.8 (3.0) -8.7* (4.1) -3.4 (3.1)
Asthma Rx 0.6 (0.9) 4.2 (4.0) -0.2 (1.0) ‡ -4.2 (4.0) -4.3 (3.9)
Height & weight measured 0.9 (1.4) 5.2 (4.6) 2.15 (3.06) 0.2 (1.3) -4.9 (4.7) -3.0 (4.8)
Tobacco Hx entered 0.7 (1.0) 5.5 (6.4) -0.06 (1.6) 0.05 (0.55) -5.4 (6.4) -5.5 (6.1)
† Difference-in-differences in pscore (p2007 – p2006) – (p2006 – p2005)‡ 90th percentile coefficient could not be estimated because there is no variation in the scores.
19
Effects of Physician Characteristics
Dependent variables:Percent score at 2007 Q1 (0-100)
(n=1179)
Percent score at 2007 Q1 (0-
100) (n=1179)
Improved between 2006-2007 (0/1)
(N=1142)
Average score in 2006 (0-100) 0.90*** -0.01***(0.01) (0.001)
Female 2.78*** -0.48 0.02(0.95) (0.44) (0.03)
Foreign graduate 0.80 -1.36 -0.08(2.29) (1.05) (0.06)
Years of practice 0.16*** -0.01 -0.002(0.05) (0.02) (0.001)
Internal medicine 1.68* -0.15 -0.01(0.94) (0.43) (0.03)
Pediatrics -31.39*** -3.22*** -0.31***(1.55) (0.85) (0.05)
R-squared 0.64 0.92 0.23
* p<0.05; ** p<0.01Linear regression; other covariates included are indicators of each quality measure and practice site.
20
Comparison to Other Groups’ Scores (2005-2007)*
These are IHA P4P measure scores. Definitions of the measures were similar to those incentivized at PAD, but the eligible patients for the IHA measures are limited to HMO patients.
Asthma Rx
0.50
0.60
0.70
0.80
0.90
1.00
2005 2006 2007measurement year
PACMGSCZ
Controlling Blood Sugar for Diabetes Patients
0.50
0.60
0.70
0.80
0.90
1.00
2005 2006 2007measurement year
PACMGSCZ
21
Comparison to Other Groups’ Scores (2005-2007) (Cont.)
These are IHA P4P measure scores. Definitions of the measures were similar to those incentivized at PAD, but the eligible patients for the IHA measures are limited to HMO patients.
Cervical Cancer Screening
0.50
0.60
0.70
0.80
0.90
1.00
2005 2006 2007measurement year
PACMGSCZ
Chlamydia Screening
0.00
0.20
0.40
0.60
0.80
1.00
2005 2006 2007measurement year
PACMGSCZ
22
Correlation in Scores Across Measures (within physicians)
0.2
.4.6
.81
HxT
obac
co_E
ntrd
avg
scor
e
0 .2 .4 .6 .8 1HxAlcohol_Entrd avgscore
0.2
.4.6
.81
BP
Avg
130_
80 a
vgsc
ore
0 .2 .4 .6 .8 16mGly7 avgscore
0.2
.4.6
.81
Col
onC
A_F
ullS
crn
avgs
core
0 .2 .4 .6 .8 16mGly7 avgscore
Y: Diabetes BP control (P4P)X: Diabetes HbA1c control (P4P)
Y: Colon cancer screening (P4P)X: Diabetes HbA1c control (P4P)
Y: Hx tobacco entered (P4P)X: Hx alcohol entered (non-P4P)
23
Does the Frequency of Payment Matter?
Dependent variable: percent score
Independent variables Coefficients (SE)
Quarterly-paid (ref. cat.: paid end-of-year) 0.50 (1.14)
Quarterly-paid* Quarter
Quarterly-paid * Quarter 2 -0.73 (0.92)
Quarterly-paid * Quarter 3 0.34 (0.92)
Quarterly-paid * Quarter 4 -0.09 (0.93)
Observations 3,767
* p<0.05; ** p<0.01Estimation methods: random effect linear regressionOther covariates included are indicators of each quarter, quality measure, practice site and department.
24
Bonus Amount by Study ArmStudy arm*
Quarterly paid Annually paid
Quarter 1, 2007(quality report date)
(payment date)
$670(6/20/07) †
(7/5/07) †
$718(6/20/07) †
-
Quarter 2, 2007(quality report date)
(payment date)
$670(7/24/07) (8/6/07)
$697(7/24/07)
-
Quarter 3, 2007(quality report date)
(payment date)
$741(10/24/07)
(11/6/07)
$758(10/24/07)
-
Quarter 4, 2007(quality report date)
(payment date)
$751(1/24/08)
(2/6/08)
$760(1/24/08)
(2/6/08)
Total $2705 $2760
*No statistical difference between two study arms.†For the first quarter, there was two months delay in the reporting and payment.
25
Summary of Findings• Physician incentives have a modest effect on the
improvement of some measures– Improvement in quality scores over the past three years for
the incentivized and other related measures. – For three measures (BP control of diabetes patients, colon
cancer screening, tobacco Hx documentation), the improvement accelerated with the incentive program.
– Similar trend is observed in a measure (BP control of hypertension patients) that was not incentivized, but was reported to the physicians.
– The trend is not distinctively different from trends of two groups which did not have the same incentive program, but also underwent various quality improvement efforts.
– Other organizational or regional quality improvement effort may have confounded the effect of P4P.
26
Summary of Findings• Within- and across- physician variations
– For each measure, within physician scores are consistent over time
– No strong correlation across measures within a physician– More improvement among physicians whose score was
middle or lowest in the previous year than those with highest score (data not shown)
• Frequency of incentive payment (quarterly vs. end- of-year) does not make a difference– No difference in scores or changes in scores over time
between the two groups based on frequency of payment.– Similar improvement in both arms for most measures.– The effect of quarterly (vs. end-of-year) incentive
payment may have been mitigated by the quarterly report sent to both arms.
27
Conclusions
• Physician-specific incentives appear associated with modest acceleration in improvement in some targeted measures.
• The frequency of payment by itself does not make a difference in performance in response to the P4P program.
• In the context of other organizational-level quality improvement efforts, relatively small financial incentives to individual physicians have limited incremental effects on well-established measures.
• Other incentives (e.g. increasing coverage of staff hours for quality improvement) should be explored.
