15. Mustard CA, Frohlich N. Socioeconomic status
and the health of the population. Med Care.
16. Geronimus AT, Bound J, Waidmann TA, Hille-
meier MM, Bums PB. Excess mortality among
blacks and whites in the United States. N Engl
17. Gornick ME, Eggers PW, Reilly TW, et al.
Effects of race on mortality and use of services
among Medicare beneficiaries. N Engl J Med.
18. Woolhandler S, Himmelstein D. Reverse tar-
geting of preventive care due to lack of health
insurance. JAMA. 1988;259:2872-2874.
19. National Center for Health Statistics. Health,
United States, 1995. Hyattsville, Md: Public
Health Service; 1996:193-194.
20. Ahluwalia JS, McNagny SE, Rask KJ. Corre-
lates of controlled hypertension in indigent,
inner-city hypertensive patients. J Gen Intern
21. Shea S, Misra D, Ehrlich MH, Field L, Francis
CK. Correlates of nonadherence to hyperten-
Public Health Briefs
sion treatment in an inner-city minority popula-
tion. Am JPublic Health. 1992;82: 1607-1612.
22. Epstein AM, Taylor WC, Seage GR. Effects of
patients' socioeconomic status and physicians'
training and practice on patient-doctor commu-
nication. Am JMed. 1985;78:101-106.
23. Brook RH, Kamberg CJ, Lohr KN, Goldberg
GA, Keeler EB, Newhouse JP. Quality ofambu-
latory care. Epidemiology and comparison by
insurance status and income. Med Care. 1990;28:
S " ' t S .
Li . . S.W..;
Physician Financial Incentives and
Feedback: Failure to Increase Cancer
Screening in Medicaid Managed Care
Alan L. Hillman, MD, MBA, Kimberly Ripley, MAS, Neil Goldfarb, BS,
Isaac Nuamah, PhD, Janet Weiner, MPH, and Edward Lusk, PhD
Although early detection methods have
proven effective in preventing breast, cervi-
cal, and colorectal cancer in women more
than 50 years of age, such methods remain
underused.' In particular, low-income
women use them least, resulting in lower
survival and higher mortality in this group
than in the remainder ofthe population.2
Many studies have identified physician
recommendation as an important determi-
nant ofa patient's decision to undergo cancer
screening.3-9 Despite this key role, many
physicians do not adhere to national guide-
lines for cancer screening with their patients.
Many methods (singly and in combina-
tion) have been used to improve physicians'
delivery of preventive care, including com-
puter-generated reminders, medical record
checklists, continuing education, and chart
audits with feedback.1019 Financial incentives
can influence physician treatment behavior,
but their ability to affect the delivery of pre-
ventive care has not been documented.'6"1>2'
We studied whether a combination of finan-
cial and nonfinancial incentives for physi-
cians could improve compliance with cancer
screening guidelines in a Medicaid health
maintenance organization (HMO).
This randomized controlled trial evalu-
ated an intervention designed to improve
physician compliance with cancer screening
guidelines for women 50 years of age and
older. We conducted the study from 1993 to
1995 with Healthcare Management Alterna-
tives mIc, a Medicaid managed care organiza-
tion located in Philadelphia. Healthcare
Management Alternatives is structured like an
independent practice association, with
provider sites paid by capitation. Its patient
population is 76% Black, 13% White, 8%
Asian, and 3% other.
Cancer screening guidelines were
adapted from national recommendations that
women 50 years of age and older receive an
annual breast examination, mammogram,
Pap smear, and colorectal screening (fecal
occult testing or sigmoidoscopy).
We randomly assigned the largest 52
primary care sites to the intervention or usual
care. We stratified the randomization by
The authors are all with the University of Penn-
sylvania, Philadelphia. Alan L. Hillman, Kimberly
Ripley, and Janet Weiner are with the Division of
Medicine, School of Medicine. Alan L. Hillman is
also with the Health Care Systems Department, the
Wharton School, and the Center for Health Policy,
Leonard Davis Institute of Health Economics.
Kimberly Ripley is also with the Center for Health
Economics. Neil Goldfarb is with the Center for
Health Policy, Leonard Davis Institute of Health
Economics, and Health Management Alternatives,
Philadelphia. Isaac Nuamah is with the School of
Nursing. Edward Lusk is with the Department of
Statistics, the Wharton School.
Requests for reprints should be sent to Alan
L. Hillman, MD, MBA, Center for Health Policy,
Leonard Davis Institute of Health Economics,
3641 Locust Walk, Philadelphia, PA 19104-6218.
This article was accepted April 23, 1998.
American Journal of Public Health
Public Health Briefs
practice type (solo/group) to ensure suffi-
cient representation ofeach. Given low base-
line scores (24%), we estimated that compli-
ance scores in the intervention group could
be doubled during the 18-month study
period. This sample was adequate to detect
an effect size of0.40 (large in Cohen's termi-
nology), with 80% power, at the .05 signifi-
The intervention included semiannual
feedback to primary care providers regarding
compliance with cancer screening guidelines
and financial bonuses for "good" performers.
Feedback reports documented a site's scores
on each screening measure and a total score
across all measures, as well as planwide
scores for comparison.
We performed chart audits at baseline
and every 6 months for 1.5 years. At each
audit, we reviewed approximately 1200
charts from more than 100 physicians. For
each site, we calculated the percentage of
charts in compliance with each indicator,
excluding charts in which screening was not
clinically indicated. Consistent with the
Healthcare Management Altematives quality
assurance policy, we considered charts docu-
menting a physician referral for screening
(with or without actual test results) as being
We defined aggregate compliance
scores as the number ofindicators in compli-
ance divided by the number of applicable
charts. Prior Healthcare Management Alter-
natives audit data indicated that mammogra-
phy, Pap smear, and colorectal screening
were independent variables, whereas breast
examination correlated with Pap smear.
Therefore, breast examination was not
included in aggregate compliance scores
(although it was included in feedback reports
We based eligibility for bonuses on
aggregate compliance scores and improve-
ment in scores over time. The 3 intervention
sites with the highest compliance scores
received a "full" bonus (20% of capitation
for all female Healthcare Management Alter-
natives members 50 years of age and older);
the 3 with the next highest scores and the 3
improving the most from the previous audit
both received "partial" bonuses (10% ofcap-
itation). Bonuses ranged from $570 to $1260
per site, with an average of $775 per audit.
Seventeen (of 26) sites received at least
bonus throughout the course ofthe study.
Although randomization, auditing, and
bonus distribution were site specific, we
directed all correspondence and feedback to
individual physicians to maximize their
awareness ofthe study. At the start, all inter-
vention group physicians received a mailing
explaining the bonus program. We pilot
tested this letter with Healthcare Manage-
ment Alternatives physicians not involved in
the study to ensure clarity and readability.
Statistical testing includedX2 and t tests
for descriptive comparisons and repeated
measures analysis of variance for between-
group effects, time effects, and the Group X
We surveyed intervention group physi-
cians to determine levels of awareness of the
program. Two mailings were sent before and
after the first audit feedback. The second
mailing targeted nonresponders and respond-
ers who were unaware of the study. Eighteen
ofthe 26 sites (69%) responded to the survey.
If at least half of the physicians at a site
responded and indicated awareness of the
study, the site was considered "aware." Ofthe
18 responding sites, 12 (67%) were aware of
the study by the end of the second mailing.
Responders did not differ significantly from
nonresponders with regard to practice type
(P= .09) or site specialty (P= .10).
1 presents descriptive data, by
study group, on practice type, specialty, and
patient panel size. No significant differences
were found between the study groups.
Table 2 presents the results of audits
(baseline) through 4. Baseline compliance
scores were relatively low and did not differ
significantly between study groups. Two
sites had missing values in audit 1; we
imputed these values using group means.
Repeated measures analysis of variance
demonstrated a significant time effect for all
4 indicators (P<.001) but no significant
between-group effect or Group x Time
effect. These results did not change when we
excluded the 2 sites with missing data. We
November 1998, Vol. 88,No. II
American Journal of Public Health
TABLE 1-Characteristics of Study Sites, Philadelphia, Pa, 1993-1995
(n = 52)
Practice type,a %
Site specialty,b %
FP & IM both
Patient panel size (mean no. of members)
Target population (women.50 years of age)c
ap= .78 (Pearson chi-square test).
bp= .21 (Pearson chi-square test).
cP= .63 (Student ttest).
dp= .35 (Student ttest).
TABLE 2-Mean Compliance Scores by Study Group, Philadelphia, Pa, 1993-1995
Audit 1 (Baseline)
(n = 26)
(n = 26)
(n = 26)
(n = 26)
(n = 26)
(n = 26)
found no significant differences by type of
practice, specialty, or patient panel size.
However, group practices had consistently
higher compliance scores than solo practices
A subanalysis comparing aware and
unaware intervention sites showed no signif-
icant between-group differences (P= .175).
This study failed to demonstrate the
effectiveness of a combination of financial
and nonfinancial incentives. Our findings
should not be interpreted to mean that finan-
cial incentives cannot capture physicians'
attention toward preventive care.22 Other
possible explanations are described subse-
The first possibility is the magnitude of
the financial incentive. Prior research sug-
gests that as little as an additional 5% of a
physician's capitation income could influ-
ence his or her behavior.'7 The bonus
amounts also seemed feasible for a managed
care organization to maintain in the long run.
The incentive provided 10% to 20% addi-
tional capitation for each site's population of
female members 50 years of age and older.
However, because physicians participated in
many plans, the bonus for Healthcare Man-
agement Alternatives members alone may
not have had the necessary impact on physi-
cians' overall income.
The second possible explanation is lack
of physician awareness. Despite repeated
mailings, awareness of the study was low
(67% of intervention sites, with 30% not
responding). Many physicians may not read
mailings from the HMOs in which they par-
ticipate and may not remember which HMO
offers which programs and incentives. Other
methods to reach physicians (e.g., in-person
detailing, group meetings) may have pro-
duced better results.
The third possibility is the context and
length of the intervention. The dramatic
increase in preventive care in both groups
mirrored national trends during the 18-
month study period. Public and professional
education campaigns to improve preventive
care took place, which might have overshad-
owed the impact of our incentive. A longer
study may have produced different results.
Although screening rates improved dra-
matically in the study period, they remained
lower than desired, with 50% of all women
in this population not receiving annual can-
cer screening. Much work remains to be
done to reach national goals for cancer
This study was conducted with funding received
from the Agency for Health Care Policy and
(grant ROI HS
Healthcare Management Alternatives.
We would like to thank Christopher Schmitt
and Danita Joell for their administrative and research
assistance throughout the study. We also appreciate
the assistance provided by Dr Denise Hamilton
Ross, chiefexecutive officer and medical director of
Healthcare Management Alternatives, and the staff
of the Healthcare Management Alternatives Quality
07720) and from
1. Breen N, Kessler L. Trends in cancer screen-
ing-United States, 1987 and 1992. JAMA.
2. Anderson LM, May DS. Has the use of cervi-
cal, breast, and colorectal cancer screening
increased in the United States? Am J Public
3. Bloom JR, Grazier K, Hodge F, Hayes WA.
Factors affecting the use of screening mam-
mography among African American women.
Cancer Epidemiol Biomarkers Prev. 1991;
4. Price JH, Desmond SM, Slenker S, Smith D,
Stewart PW. Urban black women's perceptions
ofbreast cancer and mammography. J Commu-
nity Health. 1992;17:191-204.
5. Fox SA, Murata PJ, Stein JA. The impact of
physician compliance on screening mammog-
raphy for older women. Arch Intern Med.
6. Schapira DV, Pamies RJ, Kumar NB, et al.
Cancer screening. Cancer. 1993;71 :839-843.
7. Zapka JG, Stoddard AM, Costanza ME,
Greene HL. Breast cancer screening by mam-
Public Health Briefs
mography: utilization and associated factors.
Am JPublic Health. 1989;79:1499-1502.
8. NCI Breast Cancer Screening Consortium.
Screening mammography: a missed clinical
opportunity? JAMA. 1990;264:54-58.
9. Myers R, Trock B, Lerman C. Adherence to
colorectal cancer screening in an HMO popula-
tion. Prev Med. 1990;19:502-514.
10. Cheney C, Ramsdell JW. Effect of medical
records' checklists on implementation of peri-
odic health measures. Am JMed. 1987;83:
11. Chambers CV, Balaban DJ, Carlson BL,
Ungemack JA, Grasberger DM. Microcom-
puter-generated reminders: improving the com-
pliance of primary care physicians with mam-
mography screening guidelines. J Fam Pract.
12. McPhee SJ, Bird JA, Jenkins CNH, Fordham
D. Promoting cancer screening: a randomized
controlled trial of three interventions. Arch
Intern Med. 1989;149:1866-1872.
13. McPhee SJ, Detmer WM. Office-based inter-
ventions to improve delivery of cancer preven-
tion services by primary care physicians. Can-
cer. 1993;72(suppl):I 100-1112.
14. Tiemey WM, Hui SL, McDonald CJ. Delayed
feedback of physician performance vs. imme-
diate reminders to perform preventive care.
Med Care. 1986;24:659-666.
15. Winickoff RN, Coltin KL, Morgan MM,
Buxbaum RC, Barnett GO. Improving physi-
cian performance through peer comparison
feedback. Med Care. 1984;22:527-534.
16. Egdahl RH, Taft CH. Financial incentives to
physicians. NEnglJMed. 1986;315:59-61.
17. Hillman AL, Pauly MV, Kennan K, Martinek
CR. HMO managers' views on financial incen-
tives and quality. Health Aff 1991;10:207-219.
18. Hillman AL, Pauly MV, Kerstein JJ. How do
financial incentives affect physician clinical
decisions and the financial performance of
health maintenance organizations? N Engl J
Med. 1989;321 :86-92.
19. Hillman AL. Financial incentives for physi-
cians in HMOs: is there a conflict of interest?
20. Greenfield S, Nelson EC, Zubkoff M, et al.
Variations in resource utilization among med-
ical specialties and systems of care. JAMA.
21. Manning W, Liebowitz A, Goldberg G, Rogers
WH, Newhouse JP. A controlled trial of the
effect ofa prepaid group practice on use of ser-
vices. NEnglJMed. 1984;310:505-510.
22. Hillman AL. Managing the physician: rules ver-
sus incentives. Health Aff 1991;10:207-2 19.
November 1998, Vol. 88, No. 11
American Journal of Public Health