ORIGINAL ARTICLE

Development and Validation of the Patient Assessment

of Chronic Illness Care (PACIC)
Russell E. Glasgow, PhD,* Edward H. Wagner, MD, MPH,† Judith Schaefer, MPH,†
Lisa D. Mahoney, MPH,† Robert J. Reid, MD, PhD,† and Sarah M. Greene, MPH†

Rationale: There is a need for a brief, validated patient self-report

instrument to assess the extent to which patients with chronic illness
T he increasingly well-documented gap between clinical re-
search findings and practice1–3 has spurred a number of
efforts to improve the quality of chronic illness care.4,5 It also is
receive care that aligns with the Chronic Care Model—measuring
care that is patient-centered, proactive, planned and includes collab- apparent that this problem cannot be solved by simply trying to
orative goal setting; problem-solving and follow-up support. do more; instead, fundamental changes are needed in the way
Sample: A total of 283 adults reporting one or more chronic illness that care is structured and delivered. For chronic illness care, this
from a large integrated health care delivery system were studied. change entails shifting from medical care that is reactive and
Methods: Participants completed the 20-item Patient Assessment of event-driven to care that is proactive and planned.
Chronic Illness Care (PACIC) as well as measures of demographic The Chronic Care Model, or CCM,5,6 is one approach
factors, a patient activation scale, and subscales from a primary care to improving chronic illness care that is being used increas-
assessment instrument so that we could evaluate measurement ingly to assess and improve care. The CCM is based on
performance, construct, and concurrent validity of the PACIC. evidence-based practices and reviews of the literature on
Results: The PACIC consists of 5 scales and an overall summary effective care.5–7 The model was extensively reviewed by a
score, each having good internal consistency for brief scales. As panel of experts in the area of chronic illness management
predicted, the PACIC was only slightly correlated with age and and revised accordingly. It was then used to organize data
gender, and unrelated to education. Contrary to prediction, it was collection and analysis of semistructured interviews with 72
only slightly correlated (r ⫽ 0.13) with number of chronic condi- programs and 14 site visits of innovative care delivery sys-
tions. The PACIC demonstrated moderate test-retest reliability (r ⫽
tems recommended by the experts. On the basis of these
0.58 during the course of 3 months) and was correlated moderately,
recommendations and experiences of the “best practice” sys-
as predicted (r ⫽ 0.32– 0.60, median ⫽ 0.50, P ⬍ 0.001) to
tems, the CCM proposes to change reactive acute-oriented
measures of primary care and patient activation.
Discussion: The PACIC appears to be a practical instrument that is
care to care that is planned, proactive population-based and
reliable and has face, construct, and concurrent validity. The result- patient-centered (www.improvingchroniccare.org).8
ing questionnaire is in the public domain, and recommendations for The 6 key dimensions of the CCM are organization of
its use in research and quality improvement are outlined. health care, clinical information systems, delivery system
design, decision support, self-management support, and com-
Key Words: chronic illness, survey, assessment, Chronic Care
munity resources. The CCM has been used, often in the form
Model, reliability
of “Breakthrough Series Collaboratives,” which involve dif-
(Med Care 2005;43: 436 – 444) ferent health care teams working together to improve care for
one or more chronic illnesses, to improve both processes and
outcomes of care on population or panel-wide basis (http://
www.rand.org/health/ICICE).9,10
Our group has previously developed a tool for clini-
From *Kaiser Permanente Colorado, Denver, Colorado; and the †Group
cians and health care teams to assess the extent to which they
Health Cooperative, Seattle, Washington. are employing elements of CCM in the routine care provided
Funding for this work was made available through the Robert Wood Johnson to their patients—the Assessment of Chronic Illness Care
Foundation, Grant # HS-97-112. (ACIC). This instrument has been used for quality improve-
Reprints: Russell E. Glasgow, PhD, Clinical Research Unit, Kaiser Perma- ment and found to be related to other indices of quality of
nente Colorado, 335 Road Runner Lane, Penrose, CO 81240. E-mail:
russg@ris.net.
care and sensitive to CCM-based improvement efforts.11 To
Copyright © 2005 by Lippincott Williams & Wilkins our knowledge, there are no comparable patient-report instru-
ISSN: 0025-7079/05/4305-0436 ments to assess quality of patient-centered care for chronic

436 Medical Care • Volume 43, Number 5, May 2005

Medical Care • Volume 43, Number 5, May 2005 Development and Validation of PACIC

illness consistent with the CCM. Such an instrument is METHODS

important (1) to provide convergent evidence of care deliv-
ery, (2) to understand and incorporate patient perspectives,
Recruitment and Respondents
Five hundred enrollees age 50 or older receiving care
and (3) to overcome potential over-reporting biases of clini-
from 7 primary care clinics within Group Health Cooperative,
cians in describing their care delivery.
a large managed care organization in the states of Washington
There are several validated measures of related con-
and Idaho, were sent a package containing an introductory
structs, such as patient-provider communication, provider
letter, a demographic questionnaire, the PACIC, 1 of 2 other
autonomy support, and patient empowerment or activa-
instruments described in this article, and informed consent
tion.12–14 There are also measures of the extent to which
form. As an incentive, a $5 gift card was included in the
patients receive care congruent with the Institute of Medicine packet. After 2 weeks, nonrespondents received an identical
recommendations for primary care.15–20 These measures, packet without the gift card. After 2 additional weeks, non-
however, are not directly related to the CCM, generally report respondents were contacted by phone and offered the option
on overall receipt of care, and are not specific to chronic of completing the survey by phone or having a third packet
illness. mailed. The study was approved by the Group Health Coop-
This article reports on the development and validation erative Institutional Review Board.
of a new instrument, the Patient Assessment of Chronic A total of 379 persons (76% response rate, conserva-
Illness Care (PACIC), which is designed to complement the tively assuming that all received the mailing and were alive
ACIC by providing a patient perspective on receipt of CCM- and eligible) responded to the survey, 283 (75%) having 1 or
related chronic illness care. The PACIC collects patient more chronic illness (210 after the initial mailing and 73 after
reports of the extent to which they have received specific follow-up calls). Ninety-six did not report a chronic illness
actions and care during the past 6 months that are congruent and were excluded. Of the 283 reporting a chronic illness,
with various aspects of the CCM. This scale is intended to 266 answered greater than 50% of the survey questions and
assess the receipt of patient-centered care, which emphasizes were included in the analyses. Characteristics of these 266
the key elements of modern self-management support (eg, participants are summarized in Table 1. Considerable vari-
collaborative goal settings, problem-solving and follow- ability was found on most patient characteristics. The average
up)10,13,21 and planned, proactive, and population-based age was 64, 56% were female, and most had at least 2 chronic
care.4 –7 Our goal was to develop a relatively brief instrument conditions. The most common chronic illnesses were hyper-
that could be used in a variety of health care settings and tension and arthritis, followed by depression, diabetes,
would be applicable to adult patients having one or more of asthma, and pain. Most respondents had been diagnosed with
many different chronic illnesses. chronic illness for at least 4 years (Table 1). The average
The specific purposes of this work are to (1) summarize rating on the overall health rating was 2.7 on a 4-point scale.
the development of the PACIC, (2) report initial data on Information available from automated data indicates
its psychometric characteristics (eg, scale distributions, test- that participants (n ⫽ 379) and nonparticipants (n ⫽ 122)
retest reliability, internal consistency) and relationship to were similar on characteristics available to us. No differences
patient demographic and medical characteristics, and (3) were found on gender, and nonrespondents were slightly but
assess its relationship to leading measures of other constructs, significantly older than respondents (mean age ⫽ 64.9 versus
such as patient activation12 and primary care.20 63.4, P ⬍ 0.001). One hundred persons who responded to the
Our hypotheses were that: first mailing were sent a second mailing approximately
12 weeks later, again with a second packet mailed after 2
1. The overall PACIC and its scales would reliably charac- weeks. Eighty-two completed the PACIC a second time,
terize receipt of several dimensions of CCM-related care 63 of whom reported at least one chronic condition. No
and have good psychometric characteristics (eg, internal telephone follow-up was conducted with the retest sample.
consistency and test-retest reliability).
2. The PACIC and its scales would (a) generally not be
related to patient demographics (eg, gender, age, educa- Measures
tion) but (b) would be related to disease characteristics The Patient Assessment of Chronic Illness Care
(eg, number of comorbid conditions). (PACIC) Survey is a 20-item patient report instrument that
assesses patient’s receipt of clinical services and actions
3. The PACIC would be moderately related to, but not
consistent with the CCM (Appendix, with permission from
redundant, with measures of primary care and patient
Improving Chronic Illness Care, a national program of The
activation.
Robert Wood Johnson Foundation; the instrument is also
We also developed a priori subhypotheses about the available at: http://improvingchroniccare.org/tools/pacic.htm).
PACIC subscales that are reported in the results section. Respondents rated how often they experienced the content

© 2005 Lippincott Williams & Wilkins 437

Glasgow et al Medical Care • Volume 43, Number 5, May 2005

TABLE 1. Respondent Characteristics and Relationship to PACIC Scores

Patient Characteristics Correlation to PACIC Scores

Delivery
System
Design/ Goal Problem
Mean Overall Patient Decision Setting/ Solving/ Follow-up/
or % SD PACIC Activation Support Tailoring Contextual Coordination

Age* 64.2 (10.5) 0.16† 0.01 0.09 0.11 0.23† 0.19†

% Female 56% 0.24† 0.14† 0.21† 0.25† 0.20† 0.22†
Education* (n ⫽ 251) ⫺0.05 0.05 ⫺0.03 ⫺0.04 ⫺0.06 ⫺0.05
⬍9th grade 2%
10–11th grade 4%
High school 22%
1–2 years college 37%
College degree 20%
Post-Graduate 16%
No. chronic conditions* 2.3 (1.3) 0.13† 0.07 0.02 0.09 0.13† 0.25†
Overall health (n ⫽ 245)* 2.7 (0.9) ⫺0.04 ⫺0.08 ⫺0.16† 0.00 ⫺0.07 0.09
Years diagnosed (n ⫽ 230)* 0.01 0.06 ⫺0.01 ⫺0.04 0.04 0.06
⬍1 year 6%
1–3 years 19%
4–9 years 30%
10–19 years 20%
20 ⫹ years 24%
*Spearman rank order correlations conducted for these measures to assess relationship to PACIC scores.
†
P ⬍ 0.05.

described in each item during the past 6 months. Each item (items 7–11); Problem-solving/Contextual Counseling (items
was scored on a 5-point scale ranging from 1 (no or never) to 12–15); and Follow-up/Coordination (items 16 –20). Table 2
5 (yes or always). Patients rated care received from their provides a definition of each scale and its component item
primary health care team (not just their personal physician) numbers; the corresponding items in the Appendix illustrate
for the chronic illness that they identified as most impacting how each concept was operationalized. Each scale is scored
their life. The written version of the PACIC takes 2–5 by simple averaging of items completed within that scale, and
minutes to complete. Phone administration times were highly the overall PACIC is scored by averaging scores across all 20
variable but averaged approximately 7– 8 minutes. The 20 items. These scales emphasize patient-health care team inter-
items were selected or modified from a larger pool of 46 actions and, in particular, aspects of self-management support
items generated by a national pool of experts on chronic (eg, goal setting, problem solving).21,22 The 5 PACIC scales
illness care and the CCM and pilot tested with a separate, do not map perfectly onto the 6 CCM components because
earlier sample of 130 patients. The experts suggested items we did not feel that most patients would be able to report on
for each of the PACIC domains below and reviewed and issues such as clinical information systems or organization of
provided feedback on earlier revisions of the scale. Items health care that are generally not visible to them. Delivery
retained from the larger pool were those that showed ade- System Design/Decision support maps directly onto these 2
quate variability, that patients did not have trouble under- CCM components. Patient Activation, Goal Setting, and
standing, and that best represented the underlying constructs Problem-solving/Contextual counseling all map onto self-
(see www.improvingchroniccare.org for listing of key change management support in the CCM, and Follow-up/Coordina-
concepts). tion is important for most CCM components.
We aggregated the 20 items into 5 a priori scales based The Patient Self-Activation Scale12 is a 22-item scale that
on the key components of the CCM. These subscales were assesses the extent to which patients feel able to take responsi-
Patient Activation (items 1–3 in the Appendix); Delivery bility for their care. Hibbard and colleagues12 define self-acti-
System Design/Decision Support (items 4 – 6); Goal Setting vation as having the knowledge, skill, and confidence to self-

438 © 2005 Lippincott Williams & Wilkins

Medical Care • Volume 43, Number 5, May 2005 Development and Validation of PACIC

TABLE 2. Definitions of PACIC Scale Constructs

Scale Definition

Patient Activation (items 1–3) Actions that solicit patient input and involvement in
decision-making
Delivery System Design/Decision Support Actions that organize care and provide information
(items 4–6) to patients to enhance their understanding of care
Goal Setting/Tailoring (items 7–11) Acquiring information for and setting of specific,
collaborative goals
Problem-Solving/Contextual (items 12–15) Considering potential barriers and the patient’s
social and cultural environment in making
treatment plans
Follow-up/Coordination (items 16–20) Arranging care that extends and reinforces office-
based treatment, and making proactive contact
with patients to assess progress and coordinate
care

manage and to collaborate with providers. We hypothesized that might provide alternative explanations for relationships). As-
higher PACIC scores would be moderately correlated with sociations between PACIC scores and categorical variables
higher self-activation since receipt of patient-centered, self- were evaluated using ␹2 or point-biserial correlations (for
management support should enhance self-activation. dichotomous characteristics) as appropriate. ANOVA and
Items from 4 of the 11 subscales from the primary care ANCOVA (to adjust for potential confounding variables)
module from the Ambulatory Care Experience Survey were used to evaluate potential differences among different
(PCP-ACES) (communication, integration, contextual knowl- chronic illness conditions.
edge of patient and preventive care)20 were used to assess key
aspects of primary care we thought would align with the CCM RESULTS
components. The 47-item PCP-ACES survey is a refinement of
the Primary Care Assessment Survey that has been previously Scale Characteristics
validated18,23 and widely used to predict a variety of out- Preliminary analyses indicated that there were no dif-
comes.24 –26 The PCP-ACES scales operationalize the key as- ferences on mean overall PACIC or scale scores between
pects of quality primary care identified by the Institute of these who responded to paper and pencil versus phone ad-
Medicine to include access, continuity, integration, comprehen- ministration so responses are collapsed across those subsets.
siveness, “whole-person” orientation, and sustained clinician- Also, of 6 patient characteristics analyzed, there were signif-
patient partnership.15 We hypothesized that higher PACIC icant differences between mail and phone respondents only
scores would be moderately correlated with higher levels of on overall rating of health, and this effect was not large (mean
primary care from the PCP-ACES questionnaire (see “Validity” of 2.65 for mail versus 2.92 for phone respondents, P ⬍
section of Results for specific subscale hypotheses). 0.03). There were no differences on education, race, ethnic-
ity, desire for involvement in care, or number of years with
Analyses chronic illness. Most of PACIC items demonstrated adequate
Initial descriptive analyses included means, median, variability and were strongly related to their a priori specified
standard deviation, skewedness, and distribution of scores on subscale(s). The scales all had reasonable distributional char-
individual items, scales, and the overall PACIC to evaluate acteristics (Table 3). Two hundred fifty-five of the 266
distributional characteristics. Internal consistency for the respondents had no missing data. At the item level, no items
overall PACIC and the various scales was evaluated using had ceiling effect problems. The few individual items that
coefficient alpha. We conducted a confirmatory factor anal- had a sizeable proportion at the floor or minimal level were
ysis to evaluate the extent to which the items loaded on to the activities that would be expected to be low in the absence of
hypothesized scales and the degree to which the scales were quality improvement interventions (eg, problem-solving as-
intercorrelated. Test-retest reliability and associations among sistance; follow-up coordination). As shown in Table 3, the
continuous measures were assessed using Pearson Product median scale alpha was 0.84 (range, 0.77– 0.90) and the alpha
Moment correlation coefficients; partial correlations were for internal consistency of the overall scale was 0.93.
used to control for the influence of potential confounding A confirmatory factor analysis was conducted to evaluate
variables (by partialling out scores on third variables that how well the data fit our proposed 5-scale structure. Almost all

© 2005 Lippincott Williams & Wilkins 439

Glasgow et al Medical Care • Volume 43, Number 5, May 2005

TABLE 3. Characteristics of PACIC Overall and Scale Scores

Reliability
Cronbach 3-Month
Scale Mean SD Alpha Test-Retest

Overall PACIC 2.60 (1.0) 0.93 0.58

Patient Activation 2.99 (1.3) 0.82 0.52
Delivery System Design/Decision Support 3.13 (1.1) 0.77 0.47
Goal Setting/Tailoring 2.43 (1.1) 0.84 0.60
Problem Solving/Contextual 2.87 (1.3) 0.90 0.60
Follow-up/Coordination 1.97 (1.1) 0.86 0.68
n ⫽ 255–266, except for test–retest subset, for which n ⫽ 52–57.
All correlation and alpha coefficients are significant at P ⬍ 0.01 ⬍ 0.001.

TABLE 4. Instrument Properties From Confirmatory Factor Analysis

Variance
Extracted
Scale Question Standarized t Statistic* Reliability Estimate

Patient Activation/Involvement 0.840† 0.687

Q1 0.752 11.46 0.566
Q2 0.841 13.47 0.708
Q3 0.798 12.47 0.637
Delivery System Design/Decision Support 0.784† 0.621
Q4 0.698 10.38 0.487
Q5 0.708 10.58 0.501
Q6 0.813 12.77 0.661
Goal Setting/Tailoring 0.862† 0.628
Q7 0.870 14.61 0.756
Q8 0.829 13.56 0.687
Q9 0.782 12.42 0.612
Q10 0.542 7.71 0.294
Q11 0.677 10.18 0.459
Problem Solving/Contextual 0.903† 0.737
Q12 0.769 12.13 0.591
Q13 0.813 13.18 0.661
Q14 0.888 15.15 0.788
Q15 0.874 14.76 0.764
Follow-up/Coordination 0.877† 0.651
Q16 0.651 9.61 0.424
Q17 0.678 10.13 0.460
Q18 0.702 10.62 0.493
Q19 0.889 15.04 0.791
Q20 0.890 15.06 0.792
*All t statistics are significant at P ⬍ 0.001.
†
Composite reliability for scale.

of the items loaded highly on the proposed scales (see Table 4). hypothesized model allowed for correlated factors because we
Only 3 items had standardized factor loadings less than 0.70, and felt that many of the different CCM activities were inter-related.
10 of the 20 items had factor loadings of 0.80 or greater. Our For three-quarters of the items, item reliability (defined as the

440 © 2005 Lippincott Williams & Wilkins

Medical Care • Volume 43, Number 5, May 2005 Development and Validation of PACIC

prevent variation explained by their corresponding factor) was Coordination (r ⫽ 0.25, P ⬍ 0.001). Gender was the only
greater than 0.50 (Table 4). The variance extracted estimate for patient characteristic that was consistently related to PACIC
the scales ranged from 0.62 to 0.74, with values in excess of the scales and these correlations were modest (point biserial
0.50 considered acceptable.27 correlations ⫽ 0.14 – 0.25). Women consistently reported
Finally, the goodness of the fit of the overall model was higher levels of receipt of CCM practices on both the overall
moderate. Because of the relatively small sample size, we and individual subscales. No other demographic or medical
used the non-normed fit index and the comparative fit index condition variable was consistently related to PACIC scales,
to assess model fit,28 whose values were 0.87 and 0.89, but to control for potential moderating variables, analyses
respectively. Residual analysis revealed relatively poor fit for below assessing the relationship of the PACIC to other
one item in particular (#16, “Contacted after a visit to see instruments were conducted both with and without control-
how things were going.”) However, we retained this item ling for gender, age, and number of chronic conditions.
because we feel this is an essential element of follow-up As shown in Table 5, there were few meaningful
support. The resulting 5 subscales are moderately to highly differences among patients responding based on the 6 most
intercorrelated (as would be expected, given that they are all frequently reported chronic conditions. The only significant
indices of CCM congruent care: median, r ⫽ 0.65, sharing difference revealed by one-way analysis of variance was on
42% of the variance; range, 0.49 – 0.80). The Follow-up the Follow-up/Coordination subscale. Tukey Honest Signifi-
Coordination scale was the least correlated with other scales, cant Difference follow-up tests revealed that the diabetes
and the Goal Setting and Problem-Solving scales were most patients reported higher levels of follow-up than other con-
highly correlated with each other. The PACIC summary ditions (see Table 5). It may be that diabetes patients receive
score, and most of its scales, were moderately stable during more frequent follow-up care and appointments, which could
the 3 months. The test-retest reliability for the overall PACIC account for this difference. The conclusions of these analyses
was 0.58, and individual scale reliabilities ranged from 0.47 were not changed when employing covariance analyses that
to 0.68 (Table 3). controlled for gender, age, and comorbid conditions.
Relationship to Demographic Validity
and Medical Characteristics Table 6 presents correlations of the PACIC scales and
Overall, respondents reported receiving a moderate the overall PACIC with the measures of convergent validity.
number of the services and activities that support the CCM As shown, the overall PACIC and most of the component
(mean ⫽ 2.6 of 5). Of the 5 subscales, average scores were scales were moderately correlated as predicted with both the
highest on Delivery System Design/Decision Support and Hibbard patient activation instrument and with the modified
lowest on Follow-up/Coordination (Table 3). Correlations PCP-ACES scales. The overall PACIC was correlated mod-
between PACIC scores and patient characteristics were all erately to strongly (0.42– 0.60) with 4 of the convergent
ⱕ0.25 (Table 1). Our hypothesis that patients reporting more validity measures and r ⫽ 0.32 with the final measure
chronic illnesses would have higher scores on the PACIC (Integration scale from PCP-ACES). This was true for both
than those with fewer received only weak support (r ⫽ 0.13, unadjusted correlations (above the diagonal in Table 6) and
P ⬍ 0.05). The only PACIC scale that correlated meaning- for partial correlations controlling for gender, age, and num-
fully with number of conditions reported was Follow-up ber comorbid conditions (below the diagonal).

TABLE 5. Mean (and SD) of PACIC Summary Score and Subscales by Chronic Illnesses

Delivery System Goal Problem

Overall Patient Design/Decision Setting/ Solving/ Follow-up/
Condition n PACIC Activation Support Tailoring Contextual Coordination

Hypertension 130 2.62 (1.0) 2.87 (1.3) 3.14 (1.1) 2.47 (1.1) 2.93 (1.3) 2.082 (1.2)
Arthritis 109 2.67 (1.1) 3.11 (1.3) 3.12 (1.1) 2.51 (1.1) 3.06 (1.4) 2.012 (1.1)
Depression 51 2.71 (1.1) 3.24 (1.3) 3.24 (1.2) 2.47 (1.2) 2.85 (1.4) 2.202 (1.2)
Diabetes 41 2.83 (1.0) 2.79 (1.2) 3.21 (1.0) 2.57 (1.1) 3.09 (1.2) 2.651 (1.3)
Asthma 41 2.40 (1.0) 2.51 (1.3) 3.00 (1.1) 2.30 (1.1) 2.70 (1.4) 1.882 (1.0)
Pain 41 2.64 (1.0) 3.03 (1.2) 3.01 (1.1) 2.42 (1.1) 3.00 (1.3) 2.142 (1.1)
Overall Mean (SD) 255 2.60 (1.0) 2.99 (1.3) 3.13 (1.1) 2.43 (1.1) 2.87 (1.3) 1.972 (1.1)
NOTE: The only significant difference among conditions was on the Follow-up/Coordination subscale. Tukey Honest Significant Difference follow-up tests
revealed that diabetes patient reported higher levels of follow-up than all other conditions (see superscripts in right-hand column indicating subgroups).

© 2005 Lippincott Williams & Wilkins 441

Glasgow et al Medical Care • Volume 43, Number 5, May 2005

TABLE 6. Relationship of PACIC Scores to Other Measures

Safran Assessment of Primary Care

Hibbard Interpersonal Contextual Preventive
Scale Activation Integration Communication Knowledge Care

Overall PACIC 0.42/0.40 0.32/0.36 0.57/0.55 0.60/0.56 0.50/0.59

Patient Activation 0.29/0.29 0.22/0.26 0.35/0.31 0.40/0.37 0.32/0.37
Delivery System Design/ 0.41/0.39 0.30/0.30 0.53/0.50 0.51/0.46 0.40/0.48
Decision Support
Goal Setting/Tailoring 0.43/0.41 0.31/0.33 0.49/0.46 0.49/0.45 0.48/0.51
Problem Solving/Contextual 0.38/0.36 0.32/0.36 0.54/0.53 0.59/0.53 0.45/0.55
Follow-up/ Coordination 0.32/0.30 0.16*/0.21 0.46/0.45 0.53/0.47 0.47/0.49
The first number in each cell is the unadjusted correlation; the second number is a partial correlation adjusting for gender, age, and
number of comorbid conditions.
*Only correlation not P ⬍ 0.05.

The specific a priori predictions regarding PACIC receiving care that is congruent with the CCM. When paired
subscales that we made were: with the ACIC survey, these surveys can provide comple-
mentary consumer and provider assessments of important
1. The PACIC Patient Activation scale should correlate mod-
aspects of care for chronic illness patients. This work has
erately with Safran Communication and Interpersonal
summarized the preliminary, but promising, reliability and
Care scale and with the Hibbard activation scale. The
validity results on the PACIC. The PACIC produced similar
rationale for this hypothesis was that patient-centered
communication and involvement should be related to the means and standard deviations across 6 different chronic
activation concept. This hypothesis was confirmed (r ⫽ illnesses, with the exception of diabetes patients on the
0.35 and 0.29, P ⬍ 0.001, respectively) (Table 6), al- Follow-up Coordination scale. Other clinical and research
though the correlation with the Hibbard activation scale groups are encouraged to use the instrument, which is in the
was not as high as we would have expected. public domain, to replicate our findings in other populations
2. The PACIC Goal Setting scale should correlate moder- and settings. In particular, replications of our results in
ately with Hibbard Activation scale. The rationale for this different healthcare systems, and use of the PACIC results in
hypothesis was that collaborative goal setting is a key conjunction with and to inform quality improvement pro-
activity that facilitates patient activation.9,10,21 This hy- grams are recommended.
pothesis was confirmed (r ⫽ 0.43, P ⬍ 0.001). The various scales of the PACIC, as well as the overall
3. The PACIC Problem Solving/Contextual Counseling scale score, appear both internally consistent and to be moderately
should correlate moderately with the Hibbard scale and stable over the three-month test-retest interval. The PACIC
with the Safran Contextual Knowledge scale. The ratio- includes 20 items, which should be sufficiently brief to use in
nale for this was that problem-solving is a key strategy to many settings. Given the intercorrelations among the PACIC
support patient activation and self-management;10,21 and scales and the high internal consistency of the total score,
that counseling that took into account the patient’s envi- respondents may not have recognized differences among the
ronment would need to use “contextual knowledge” of the subscale constructs. Therefore, we are most confident recom-
patient. This hypothesis was confirmed (r ⫽ 0.38 and mending use of the entire PACIC and the total score to
0.59, P ⬍ 0.001). represent CCM congruent care. If this is too long for some
4. The PACIC Follow-up/Coordination scale should corre- settings, users might consider administering only the sub-
late moderately with the Safran Integration scale. The scales most appropriate for given their program and ques-
rationale for this hypothesis was that coordination of care tions, or dropping the patient activation items (#1–3), because
and follow-up on referrals should produce perceptions of this construct is addressed in other instruments and is the
more integrated care. This hypothesis was not confirmed most highly correlated with other subscales.
(r ⫽ 0.16, P ⫽ 0.09). We recommend that users administer at least the last 3
scales of the PACIC and, at a minimum, the Problem Solving/
DISCUSSION Contextual and Follow-up Coordination scales, because those
The PACIC provides a brief, patient-reported assess- brief scales assess specific activities that form the core of
ment of the extent to which chronically ill patients report modern, patient-centered self-management support and that the

442 © 2005 Lippincott Williams & Wilkins

Medical Care • Volume 43, Number 5, May 2005 Development and Validation of PACIC

literature indicates are seldom delivered consistently.21,22,29 2. Lenfant C. Clinical research to clinical practice–lost in translation?
N Engl J Med. 2003;349:868 – 874.
Also as illustrated by our data, many of these strategies (eg, see 3. McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered
collaborative goal setting and follow-up support especially in to adults in the United States. N Eng J Med. 2003;348:2635–2645.
Table 5) are under-used.3 4. Bodenheimer TS, Wagner EH, Grumbach K. Improving primary care for
patients with chronic illness. JAMA. 2002;288:1775–1779.
The PACIC and its scales demonstrated substantial con- 5. Wagner EH, Austin BT, Von Korff M. Organizing care for patients with
struct validity by correlating moderately with predicted mea- chronic illness. Milbank Q. 1996;74:511–544.
sures including patient activation12 and selected primary care 6. Von Korff M, Gruman J, Schaefer J, et al. Collaborative management of
scales from the modified ACES.20 In contrast, PACIC scores do chronic illness. Ann Intern Med. 1997;19:1097–1102.
7. Wagner EH, Davis C, Schaefer J, et al. A survey of leading chronic
not appear strongly related to specific diseases or patient demo- disease management programs: are they consistent with the literature?
graphic characteristics, with the possible exception of gender (r Manag Care Q. 1999;7:56 – 66.
values for gender ⫽ 0.14 – 0.25). Confirmation of this latter 8. Wagner EH. Chronic disease management: what will it take to improve
care for chronic illness. Effec Clin Pract. 1998;1:1– 4.
finding is recommended, but it is reassuring that most of rela- 9. Wagner EH, Glasgow RE, Davis C, et al. Quality improvement in
tionships remained virtually unchanged after controlling for chronic illness care: a collaborative approach. Jt Comm J Qual Improv.
gender, age, and number of comorbid illnesses. Future research 2001;27:63– 80.
10. Glasgow RE, Funnell MM, Bonomi AE, et al. Self-management aspects of
is needed on the PACIC, and on gender differences in particular, the improving chronic illness care Breakthrough Series: implementation
to determine whether these effects are caused by differences in with diabetes and heart failure teams. Ann Behav Med. 2002;24:80 – 87.
perception or reporting or to actual differences in care received. 11. Bonomi AE, Glasgow RE, Wagner EH, et al. Assessment of chronic
illness care (ACIC): a practical tool for quality improvement. Health
This study has both strengths and limitations. Limitations Serv Res. 2001;37:791– 820.
include a sample from only one health care organization. Al- 12. Hibbard JH, Stockard J, Mahoney ER, et al. Development of the Patient
though this sample was relatively heterogeneous on a variety of Activation Measure (PAM): conceptualizing and measuring activation in
patients and consumers. Health Serv Res. 2004;39:1005–1026.
characteristics and included patients having a large number of 13. Williams GC, Freedman ZR, Deci EL. Supporting autonomy to motivate
different chronic illnesses, it did not include a high percentage patients with diabetes for glucose control. Diabetes Care. 1998;21:
of non-whites. Probably the single greatest limitation is that we 1644 –1651.
14. Anderson RM, Funnell MM, Fitzgerald JT, et al. The Diabetes Empow-
were not able to assess sensitivity to change as a result of erment Scale. Diabetes Care. 2000;23:739 –743.
intervention. Future research is needed to determine if interven- 15. Institute of Medicine. Primary Care: America’s Health in a New Era.
tions based on the CCM produce improvements in relevant Washington, DC: National Academy of Sciences; 1996.
16. Borowsky SJ, Nelson DB, Fortney JC, et al. VA community-based
PACIC scores, and to explore variations across medical prac- outpatient clinics: Performance measures based on patient perceptions of
tices and health care systems. care. Med Care. 2002;40:578 –586.
Strengths of the present study include a relatively large 17. Flocke SA. Measuring attributes of primary care: development of a new
instrument. J Fam Pract. 1997;45:64 –74.
sample size and inclusion of patients having a number of 18. Safran DG, Kosinski M, Tarlove AR, et al. The Primary Care Assess-
different chronic illnesses and type of comorbidity. For survey ment Survey: tests of data quality and measurement performance. Med
research conducted in the present era of telemarketing, we also Care. 1998;36:728 –739.
19. Shi L, Starfield B, Xu J. Validating the adult primary care assessment
obtained a relatively high participation rate and telephone and tool. J Fam Pract. 2001;50:161W–175W.
paper and pencil administration appear to produce equivalent 20. Safran DG, Karp M, Coltin K, et al. Measuring patients’ experiences
results. Additional research is recommended to further explore with individual physicians. J Gen Int Med. 2005, in press.
21. Glasgow RE, Davis CL, Funnell MM, et al. Implementing practical
the convergent and divergent validity of the PACIC, such as its interventions to support chronic illness self-management in health
relationship to the ACIC, self-efficacy, self-management behav- care settings: Lessons learned and recommendations. Jt Comm J Qual
iors, more sophisticated measures of comorbidity, and health Improv. 2003;29:563–574.
22. Lorig KR, Holman HR. Self-management education: history, definition,
care team reports on implementation of the CCM. outcomes, and mechanisms. Ann Behav Med. 2003;26:1–7.
23. Murray A, Safran DG. The Primary Care Assessment Survey: a tool for
measuring, monitoring, and improving primary care. In: Maruish ME,
ACKNOWLEDGMENTS ed. Handbook of Psychological Assessment in Primary Care Settings.
We wish to express our gratitude to Connie Davis, who Mahwah, NJ: Lawrence Erlbaum Associates; 2000:623– 651.
was instrumental in generating ides for the PACIC and who also 24. Safran D, Taira D, Rogers W, et al. Linking primary care performance
to outcomes of care. J Fam Pract. 1998;47:213–220.
served as a member of our expert panel. Other expert panel 25. Safran DG, Murray AM, Chang H, et al. Linking doctor-patient rela-
members who provided helpful feedback included Martha Fun- tionship quality to outcomes. J Gen Intern Med. 2000;15:116.
nell, Michael Goldstein, Michelle Heisler, Judith Hibbard, Kate 26. Safran DG, Montgomery JE, Chang H, et al. Switching doctors: deter-
minants of voluntary disenrollment from a physician’s practice. J Fam
Lorig, David McCulloch, Deborah Toobert, Geoffrey Williams,
Pract. 2001;502:136.
and Michael Von Korff. 27. Forness C, Larker DF. Evaluating structural equation models with unob-
servable variables and measurement error. J Mark Res. 1981;18:39 –50.
28. Bentler BP, Bonett DG. Significance tests and goodness-of-fit in the
REFERENCES analysis of covariance structure. Psychol Bull. 1988;88:588 – 606.
1. Institute of Medicine, Committee on Quality of Health Care in America. 29. Glasgow RE, Eakin EG, Fisher EB, et al. Physician advice and support
Crossing the Quality Chasm: A New Health System for the 21st Century. for physical activity: results from a national survey. Am J Prev Med.
Washington, DC: National Academy Press; 2001. 2001;21:189 –196.

© 2005 Lippincott Williams & Wilkins 443

Glasgow et al Medical Care • Volume 43, Number 5, May 2005

APPENDIX
PACIC Scale

FIGURE 1. PACIC Scale, with permission from Improving Chronic Illness Care, a national program of The Robert Wood Johnson
Foundation. The instrument is also available at: http://improvingchroniccare.org/tools/pacic.htm.

444 © 2005 Lippincott Williams & Wilkins