Full text of DOI:10.1093/gerona/56.2.M106

Journal of Gerontology: MEDICAL SCIENCES
Copyright 2001 by The Gerontological Society of America
2001, Vol. 56A, No. 2, M106–M112
The Effects of a Collaborative Model of Primary
Care on the Mortality and Hospital Use of
Community-Dwelling Older Adults
Cheryl Schraeder,1 Paul Shelton,1 and Mark Sager2
1Coordinated Care Services, Carle Clinic Association, Urbana, Illinois.
2Geriatrics Section, University of Wisconsin-Madison Medical School and Wisconsin Alzheimer’s Institute, Madison, Wisconsin.
Background. This study evaluates the ability of a model of collaborative primary care practice to reduce mortality
and hospital use in community-dwelling elderly persons.
Methods. Four rural and four urban clinic sites in east central Illinois were randomized to form treatment and com-
parison clinics from which patients were enrolled and followed prospectively for 2 years. Patients from the practices of
participating physicians were eligible if they were aged 65 and older, were living in the community, and had at least one
risk factor as determined prior to the study. Medicare hospital data were obtained from the Health Care Financing Ad-
ministration. Demographic and health status measures were obtained by telephone interview every 12 months through-
out the study.
Results. The treatment group experienced a 49% reduction in all-cause mortality during the second year of the study
(odds ratio, 0.51, 95% confidence interval, 0.29–0.91, p ϭ .02). There were no significant differences between treatment
and comparison patients in percentage of persons hospitalized, hospital length of stay, or Medicare payments. Although
measures of health status indicated that the treatment group was significantly sicker at baseline at the end of 1 year, these
differences disappeared by the end of 2 years.
Conclusions. The collaborative primary care model evaluated in this study significantly reduced mortality in the sec-
ond year, without increasing hospital use. These findings suggest that a collaborative primary care team that enhances
primary care practice can result in better patient outcomes.
HE rapid growth of the elderly population and the pro-
jected expense to our health care system have renewed
METHODS
T
Study Setting and Design
interest in primary care medicine as a means for improving
patient outcomes and controlling growth in costs (1–3). Ide-
ally, primary care practice should allow for a comprehen-
sive, integrated system of care that addresses a wide spec-
trum of acute and chronic health care needs (4). Practically,
the current scope of primary care practice is limited by fis-
cal incentives that reward disease-oriented, office-based,
and episodic care provided primarily by physicians (5). Ef-
forts to develop more broadly based systems of care that go
beyond the treatment of illness to include health mainte-
nance and prevention have often been disease specific (6,7),
limited in duration (8–10), and, at times, have not actively
involved primary care physicians in the study design (11).
The limitations of previous studies illustrate the difficulty of
evaluating new systems of primary care whose effective-
ness relies on longitudinal care provided in the context of
the doctor-patient relationship.
We conducted the current 24-month study to evaluate a
primary care practice model in which physicians, nurses,
and patients worked in collaboration to expand the scope of
primary care practice. It was our hypothesis that this collab-
orative model, which focused on comprehensive assessment
and coordinated longitudinal care management, would im-
prove patient outcomes and lower costs.
This study was conducted at the Carle Clinic Association
and Carle Foundation, Urbana, Illinois, between May 1993
and May 1996. A total of 32 family practice (FP) and 19 in-
ternal medicine (IM) physicians from eight clinics agreed to
participate in this study. These physicians were located in
four rural and four urban clinics ranging in size from 4 to 16
physicians. To ensure a balance between rural and urban FP
and IM practices, the practices of IM and FP physicians
were randomized and then combined to form treatment and
comparison groups. This resulted in 19 treatment physicians
(10 IM and 9 FP) located in four clinics (two urban, two ru-
ral) and 32 comparison physicians (9 IM and 23 FP) located
in four (two urban, two rural) clinics.
Study Population
After randomization, participating physicians were pro-
vided with patient lists that included information about pa-
tient age, gender, and the number of physician visits during
the previous year. Physicians selected potential participants
if they were aged 65 years or older, were community dwell-
ing, and had any one of the following risk factors: hospital-
ized in the previous 6 months, lived alone, lacked a care-
giver, were taking four or more prescription medications,
M106

EFFECTS OF PRIMARY CARE MODEL
M107
had difficulty walking, had limitations in activities of daily
living, had difficulties with memory, were incontinent of
urine or stool, or experienced multiple illnesses or disabili-
ties requiring special care. Those individuals who were
identified as potential study patients were contacted by mail
(1286 comparison and 1376 treatment patients). Patients
who returned an informed consent and completed a baseline
interview between May 1993 and May 1994 were included
in the study. A total of 941 individuals met the previously
described criteria; 411 in the comparison group (239 FP pa-
tients, 172 IM patients) and 530 in the treatment group (239
FP patients, 291 IM patients).
Questionnaire (HSQ) (15) and the number of restricted ac-
tivity bed days. Data were again obtained at 12 and 24
months after enrollment.
Information about hospitalizations were obtained from
the Health Care Financing Administration’s (HCFA) Medi-
care part A data. Medicare data included the total number of
hospitalizations, total number of hospital bed days, the diag-
nostic-related group (DRG) assigned to each hospital ad-
mission, and the total Medicare payment for each inpatient
stay. These data were obtained for 12 months prior to the
study and for the entire 24-month study period.
Mortality data were obtained from family members and
the patient’s medical record. Medicare part A data were
used to verify inpatient deaths, and county health depart-
ment records were used to identify deaths when survival
status was unknown.
Study Intervention
The specific details of the intervention have been pub-
lished elsewhere (12,13) and are summarized below. The
intervention included the addition of a registered nurse
(RN) and a case assistant (CA) to the primary care practice
of treatment group physicians. The collaborative team’s
goal was to enhance existing primary care by providing pa-
tient/family assessments conducted in the home or office,
flexible home or office visits, and detailed care planning, as
well as coordination and procurement of supportive ser-
vices. The intervention also included routine telephone
monitoring to identify changes in condition and adherence
to treatment regimes, proactive postillness follow-up, dis-
ease education, and wellness promotion.
To promote communication among the physician, RN,
CA, and the patient/family, the RN and CA were located in
the offices of participating physicians. Communication oc-
curred by telephone, voice mail, written notes and summa-
ries, care-plan letters to patients and families, informal of-
fice conversations, and formal meetings in which patient
treatment plans were discussed.
Statistical Analyses
Baseline comparisons between the treatment and compar-
ison groups were made using Student’s t test (two-tailed)
for normally distributed continuous variables, chi-square
tests for categorical variables, and the Wilcoxon rank sum
test for nonnormally distributed continuous variables.
The effects of the intervention on mortality, health status,
hospital utilization, and reimbursed hospital Medicare pay-
ments were evaluated using an intent-to-treat approach.
Outcome effects were evaluated for the first and second
year and for the entire 24-month study period. Sample size
calculations indicated that 410 persons were needed in each
group to detect a 25% reduction in hospitalization rates of
40% and a 45% reduction in a 10% annual mortality rate
with 80% power, assuming a two-sided significance level
of .05.
Logistic regression models, using backward stepwise se-
lection, were used to test for treatment effects on mortality
and hospitalization. Linear regression models, with step-
wise selection, were used to evaluate the treatment effects
on health status, length of stay, and Medicare expenditures.
The covariates used in all regression models to control for
differences included the following: age, gender, income, ed-
ucation, marital status, living arrangement, MMSE score,
absence of a caregiver, number of prescription medications,
the presence of comorbid illnesses, primary care physician
type, restricted activity bed days, the number of months pa-
tients were in the study, clinic site, rural/urban designation
of the clinic site, the five subscale measures of the HSQ at
baseline, and the baseline value of the outcome variable.
The regression models evaluating the amount of hospital
use and Medicare payments also included the DRGs
grouped into HCFA medical diagnostic categories (MDCs)
(16). The regression models evaluating the effectiveness of
the intervention on the five HSQ measures included the pre-
viously described covariates, but excluded the four HSQ
measures not being evaluated, the MDCs, and the time vari-
able.
Upon enrollment, all treatment patients received an initial
in-home assessment conducted by the RN. Protocols guided
areas for intervention (e.g., nutrition, medication manage-
ment, health promotion, prevention) and a team-generated
plan of care was developed. The CA provided telephone
monitoring and ensured the procurement and provision of
supportive services identified by the team. Typically, the
RN and CA were able to manage 150 patients and worked
with an average of four to five physicians. The longitudinal
nature (24 months) of the intervention was an important ele-
ment in the design of the study since it was likely that any
benefits of the intervention would accrue over time.
Data Collection and Measurement Intervals
Data were collected by telephone by trained interviewers
who were blinded to the treatment and comparison groups.
Baseline data included patient demographics, current health
status, and functional status. Demographic data included
age, gender, income, education, marital status, living ar-
rangement, and caregiver arrangement. Assessments of
health status included cognition (telephone version of the
Mini-Mental State Examination [MMSE]) (14), number of
prescription medications, and number of comorbid illnesses
(heart disease, diabetes, myocardial infarction, stroke,
chronic obstructive pulmonary disease, and cancer). Func-
tional status included five subscales of the Health Status
Due to the nonnormal distribution of the hospital and
Medicare payment data, a Cook’s D statistic Ͼ1 was used to
censure cases from the analysis allowing the parameter esti-
mates (PE) of the linear regression models to be interpreted
in their original measurement units. All statistical analyses

M108
SCHRAEDER ET AL.
were performed using the SPSS software system, version
9.0 (SPSS Inc, Chicago, IL) (17). A p value of .05 indicated
statistical significance.
phone calls. On average, each treatment patient had an addi-
tional 8.0 self-reported contacts with their primary care phy-
sicians over 2 years.
The per member per month cost of the intervention was
$38 for the 2-year study period. The average cost of the
study intervention was $906 per patient for the combined 2
years. These costs included all personnel, administrative,
and overhead expenses.
RESULTS
Baseline Characteristics of the Study Population
The baseline characteristics of the 941 patients in the
comparison (n ϭ 411) and treatment (n ϭ 530) groups are
shown in Table 1. Overall, patients in the treatment group
were older and less well educated and reported a higher
prevalence of cormobid illnesses and lower scores on the
five measures of the HSQ. The treatment group also had
higher hospital use and Medicare expenditures per person
during the year prior to the study.
Mortality
A total of 95 patients (10.1%) died during the study pe-
riod; 47 (11.4%) in the comparison group and 48 (9.1%) in
the treatment group. When the baseline differences between
the groups were controlled for in the logistic regression
analyses, the risk of death was significantly lower in the
treatment group versus the comparison group (odds ratio,
0.51, 95% confidence interval 0.29–0.91, p ϭ .02) during
the second year of the study (Table 2) and approached sig-
nificance for the 2-year study period. To assess the benefit
of the intervention, we performed a number-needed-to-treat,
or NNT, analysis (18) and found that each collaborative
team needed to treat approximately 31 patients during the
2-year study period to prevent one death.
Intervention Intensity and Cost
Each treatment patient had an average of 8.0 intervention
contacts (Ϯ9.2), totaling 4.8 hours (Ϯ5.0) in the first year
and 8.4 intervention contacts (Ϯ10.2), totaling 4.6 hours
(Ϯ5.9) in the second year. For both years, patients received
a total of 16.1 (Ϯ16.1) intervention contacts, including 7.5
(Ϯ10.9) home visits, 1.4 (Ϯ2.0) office visits, and 7.3 (Ϯ6.8)
Table 1. Characteristics of the Study Population
Comparison Group
Treatment Group
Characteristic
(n ϭ 411)
(n ϭ 530)
p Value
Primary care physician type (% IM)
Demographics
42
55
.000
Mean age, y (SD)
75.4 (6.4)
76.5 (6.7)
.01
Women (%)
Married (%)
White (%)
High school graduate or above (%)
Ͻ$20,000/y household income (%)
Lives alone (%)
75
45
96
74
58
48
71
42
97
65
64
49
.025
.43
.93
.000
.06
.76
Health Status
Ն5 prescription medications (%)
MMSE score Յ17† (%)
Restricted-activity bed days (%)
Health Conditions
32
8
24
37
12
29
.16
.10
.08
Heart disease (%)
Myocardial infarction (%)
Stroke (%)
Cancer (%)
Chronic obstructive pulmonary disease (%)
Diabetes (%)
27
14
10
9
11
20
29
21
15
7
14
23
.44
.000
.03
.17
.11
.33
Health Status Questionnaire Scores‡
Mean health perception (SD)
Mean physical health (SD)
Mean mental health (SD)
Mean pain (SD)
62 (23)
63 (29)
80 (17)
71 (27)
53 (24)
59 (24)
54 (30)
74 (20)
67 (26)
47 (24)
.06
.000
.000
.04
Mean energy/fatigue (SD)
Prior Hospital Use
.000
Hospitalized (%)
22
1.6 (.94)
5.7 (3.8)
$1705 ($4711)
28
1.5 (.94)
6.6 (5.3)
$2149 ($5104)
.02
.20
.46
.01
Mean hospitalizations/person hospitalized (SD)
Mean length of stay/person hospitalized (SD)
Mean Medicare hospital costs/person (SD)
Notes: IM ϭ internal medicine physician; MMSE ϭ Mini-Mental State Examination.
†Telephone MSE score of Յ17 out of a total score of 24.
‡Range 0–100.

EFFECTS OF PRIMARY CARE MODEL
M109
Table 2. Logistic Regression Results of the Treatment Effect on Mortality by Study Year
Year 1
OR (95% CI)
1.1 (0.47– 2.5)
Year 2
2-Year Total
Variables
OR (95% CI)
0.51* (0.29–0.91)
OR (95% CI)
Treatment effect (1 ϭ treatment)
Demographics
0.71 (0.44–1.1)
Age
Women
NS
NS
1.1** (1.1–1.2)
NS
1.1** (1.1–1.2)
0.51** (0.31–0.85)
Health Status
Ն5 prescription medications
MMSE score Յ17†
Restricted-activity bed days
Health Conditions
Heart disease
Myocardial infarction
Stroke
Cancer
NS
NS
NS
NS
2.2** (1.3–3.5)
NS
2.0** (1.2–3.2)
3.1* (1.3–7.7)
2.3* (1.0–5.3)
NS
NS
NS
NS
NS
NS
NS
2.4** (1.3–4.3)
NS
2.6** (1.2–5.6)
NS
NS
NS
NS
NS
NS
NS
Chronic obstructive pulmonary disease
Diabetes
NS
Health Status Questionnaire Scores
Health perception
Physical health
Mental health
NS
NS
0.98** (0.97–0.99)
0.98** (0.97–0.99)
NS
NS
0.98** (0.96–0.99)
0.98** (0.97–0.99)
NS
NS
NS
NS
Pain
Energy/fatigue
NS
0.032 (4 df)
NS
0.074 (8 df)
NS
0.092 (8 df)
R2
Notes: All regression models included adjustments for clinic site, rural/urban designation, primary care physician type, marital status, race, education level, house-
hold income, and living situation (alone or with others).
OR ϭ odds ratio; CI ϭ confidence interval; MMSE ϭ Mini-Mental State Examination; NS ϭ not significant.
†Telephone MMSE score of Յ17 out of a possible 24.
*p Ͻ .05; **p Ͻ .01.
Hospital Use
Medicare Payments
Eighty-nine comparison patients (21.7%) and 140 treat-
ment patients (26.4%) were hospitalized during the first
year of the study; 91 comparison group patients (22.6%)
and 128 treatment patients (25.0%) were hospitalized dur-
ing the second year of the study. A total of 149 comparison
group patients (36.3%) and 221 (41.7%) treatment group
patients were hospitalized during the entire 24-month study.
There were no significant differences in the likelihood of
hospitalization between the treatment and comparison
groups for the study period (Table 3).
The linear regression models revealed no significant dif-
ferences in hospital length of stay between the two groups
during the study periods. The average length of stay during
the first 12 months of the study for the comparison group
was 5.0 days(Ϯ4.0) versus 6.0 days (Ϯ4.0) for the treatment
group, 6.1 days (Ϯ5.0) for the comparison group versus 5.3
days (Ϯ4.0) for the treatment group in the second year of
the study, and 6.0 days (Ϯ4.0) for the comparison group and
5.4 days (Ϯ3.3) for the treatment group for the entire study
period.
The average inpatient Medicare reimbursement during
the first year of the study for the comparison group was
$1813 (Ϯ$5041) versus $2218 (Ϯ$5166) for the treatment
group. The average inpatient Medicare reimbursement dur-
ing the second year of the study for the comparison group
was $2058 (Ϯ$5056) versus $2264 (Ϯ$5500) for the treat-
ment group, and $3826 (Ϯ$7289) for the comparison group
and $4452 (Ϯ$7706) for the treatment group for the entire
study period. For those patients who had an inpatient hospi-
talization, the median Medicare reimbursement during the
first year was $5057 for the comparison group and $6139
for the treatment group. During the second year, the median
inpatient reimbursement was $6039 for the comparison
group and $6265 for the treatment group. Results of the lin-
ear regression models revealed no significant differences
between the two groups during the evaluation study periods.
DISCUSSION
This study evaluated the impact of a collaborative pri-
mary care team on the health status and hospital use of com-
munity-dwelling elderly persons living in east central Illi-
nois. The most significant finding was the reduced risk for
mortality in the treatment group during the second study
year. Although differences in mortality during the 2-year
study period only approached significance, the risk of death
in the treatment group was 49% lower in Year 2 when com-
pared with that of the comparison group. Coincident with
the lower mortality rate was a disappearance of prior health
Health Status
At the end of the first 12 months of the study period, the
linear regression analyses indicated that the treatment group
had significantly lower HSQ scores in mental health (PE, –2.2;
p Ͻ .05) and energy/fatigue (PE, –3.4; p Ͻ .05). At the end
of the study, there were no significant differences between
the two groups.

M110
SCHRAEDER ET AL.
Table 3. Logistic Regression Results Showing Treatment Effects for any Hospitalization by Study Year
Year 1
OR (95% CI)
1.1 (0.77–1.5)
Year 2
OR (95% CI)
1.0 (0.73–1.4)
2-Year Total
OR (95% CI)
1.1 (0.82–1.5)
Variables
Treatment effect (1 ϭ treatment)
Demographics
Age, y
Women
NS
NS
NS
NS
0.67* (0.47–0.96)
0.76* (0.54–0.99)
Health Status
Ն5 prescription medications
MMSE score Յ17†
Restricted-activity bed days
Health Conditions
Heart disease
Myocardial infarction
Stroke
Cancer
2.3** (1.6–3.2)
NS
NS
NS
NS
NS
NS
NS
NS
1.6* (1.1–2.2)
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Chronic obstructive pulmonary disease
Diabetes
Health Status Questionnaire Scores
Health perception
Physical health
Mental health
Pain
NS
0.99** (0.98–0.99)
NS
NS
0.99* (0.98–0.99)
1.1* (0.97–0.99)
NS
.99* (0.98–1.0)
NS
NS
NS
NS
Energy/fatigue
Any hospitalization prior to study entry
R2
NS
NS
.106 (10 df)
1.1* (1.0–1.2)
1.8** (1.3–2.6)
.051 (6 df)
NS
NS
.051 (7 df)
Notes: All regression models included adjustments for clinic site, rural/urban designation, medical diagnostic categories, months in the study, primary care physi-
cian type, marital status, race, education level, household income, and living situation (alone or with others).
OR ϭ odds ratio; CI ϭ confidence interval; MMSE ϭ Mini-Mental State Examination; NS ϭ not significant.
†Telephone Mini Mental State Exam score of Յ17 or out of a possible 24.
*p Ͻ .05; **p Ͻ .01.
status differences between the two study groups. The HSQ
measures showed significantly poorer health in the treat-
ment group at baseline and at Year 1, but this difference dis-
appeared by the end of the study.
There are two possible explanations why the reduced
mortality occurred during the second year of the study.
First, the benefits of enhanced primary care evaluated in this
study require time to develop. The potential benefits of im-
proved patient self-management, health promotion, and ed-
ucation are not immediate and must accumulate over time
with repetition and familiarity. The clinical effectiveness of
home visits and telephone monitoring likely also increase
over time as patients became more familiar and accepting of
these activities.
Second, the collaborative primary care practices were
likely better developed and more effective during the later
months of the study. It requires time to introduce new clini-
cal programs into busy office practices, build professional
relationships, establish lines of communication, and imple-
ment a team approach to primary care. These factors are
highlighted by a significant increase in both patient and
physician satisfaction that occurred during the second year
of the study (data available from the authors).
Although one effect of the intervention seems to have
been to improve the survival of what initially had been a
sicker patient population, lowered mortality was not accom-
panied by increases in hospital use. There were no signifi-
cant reductions or increases in the use of hospital care dur-
ing the study for the treatment group. There were also no
differences in rates of hospitalization, hospital length of
stay, or Medicare payments during any year of the study.
Although the costs of the intervention were quite modest
($906 per patient for both years), these costs were not offset
by measurable reductions in hospital use.
This intervention had many components designed to en-
hance primary care services and, for this reason, it is diffi-
cult to identify specific components that resulted in lowered
mortality. Although the benefits of home visits and tele-
phone monitoring are well known, we know of few studies
in which both of these methodologies have been combined
with the health promotion and prevention activities of this
collaborative primary care model. The reduced mortality is
unlikely the result of improved prescribing practices or bet-
ter physician management of specific diseases. No effort
was made to implement disease management protocols or to
enhance knowledge of geriatric medicine.
The collaborative practice model evaluated in this study
has many features that distinguish it from usual primary
care. First, the model requires that the physician and RN ac-
tively collaborate to manage the care of a well-defined
panel of high-risk patients. This is unlike typical office
practice in which the responsibilities of the RN are often
diffused among many patients, less collaborative, and more
acute care oriented. Second, the collaborative practice
model promotes shared decision making between the pa-
tient, physician, and RN. This shared decision making al-

EFFECTS OF PRIMARY CARE MODEL
M111
lowed for congruent planning and monitoring, which en-
couraged patients to adhere to health maintenance and
treatment regiments. Third, the model relied upon a stan-
dardized set of protocols and guidelines, which defined
comprehensive practice to include health maintenance and
promotion. Fourth, the model allowed for maximum flexi-
bility to manage patients outside of Medicare guidelines,
which often restrict not only what can be provided, but how
much can be provided, regardless of need. This flexibility
was essential for health monitoring and maintenance.
This study had several limitations. First, the patient as-
signment did not follow a randomized, clinical trial design,
which was not feasible in the participating primary care
practices. We were concerned that the development of col-
laborative teams would alter the care provided to all patients
in the intervention practices regardless of treatment group
assignment. Our decision to randomize practices to ensure a
balanced urban/rural mix created a disparity between the
two patient populations. Treatment patients were more
likely to be the patient of an IM physician, possibly explain-
ing why they were in poorer health. Accordingly, all analy-
ses were performed adjusting for these baseline differences
between study groups, but all adjustments were limited to
the variables available in the database. The extent to which
we were able to adequately adjust for baseline differences
between groups is unknown.
A second limitation of our study was our inability to
compare overall costs of care for the treatment and compar-
ison groups. After almost a 2-year delay in obtaining Medi-
care data from the HCFA, we had to restrict analyses to part
A data to avoid prohibitive costs and additional delay. An
unanswered question is whether the RN and CA contacts
occurred in addition to usual physician office visits or acted
as a substitute for physician visits. We also were unable to
determine the intervention’s effect on home health care,
nursing home care, and other elements of health care use.
A third limitation of the study was that we were limited in
the analyses to only those variables that had been collected
as part of the original study. We included those variables
that in previous studies have been shown to be predictors of
the dependent variables, either hospitalization or mortality.
Because this was an outpatient population, we did not have
access to inpatient records that would have allowed us to
calculate indices for severity of illness (Acute Physiology
and Chronic Health Evaluation [APACHE]) (19) or comor-
bidity (20).
collaborative practice, regardless of the potential benefits to
patients. Although the costs of this model are born by pro-
viders, the potential savings from any reduced hospitaliza-
tion accrue to the Medicare program. At present, only man-
aged care organizations who are at financial risk in the
Medicare program have both the financial incentives and
the organizational resources to develop integrated primary
care systems such as the one evaluated in this study. This
study suggests that there remain untapped efficiencies to be
gained when a more comprehensive approach to primary
care is applied to elderly patients. The reduction in mortality
suggests that our current, limited system of primary care
may be potentially detrimental to the health of our patients.
Acknowledgments
Financial support for this study was provided by Grant 92123-G from
the John A. Hartford Foundation, New York, NY.
Address correspondence to Paul Shelton, EdD, Coordinated Care Ser-
vices, P.O. Box 718, 307 East Oak #3, Mahomet, IL 61853. E-mail:
psshelton@aol.com
References
1. Mark DH, Gottlieb MS, Zellner BB, Chetty VK, Midtling JE. Medi-
care costs in urban areas and the supply of primary care physicians. J
Fam Pract. 1996;43:33–39.
2. Sofaer S. Aging and primary care: an overview of organizational and
behavioral issues in the delivery of healthcare services to older adults.
Health Serv Res. 1998(part II);33:298–321.
3. Franks P, Nutting PA, Clancy CM. Health care reform, primary care,
and the need for research. JAMA. 1993;270:1149–1453.
4. Institute of Medicine. Primary Care: America’s Health in a New Era.
Washington, DC: Academy Press; 1996.
5. Wagner EH, Austin BT, Von Korff M. Organizing care for patients
with chronic illness. Milbank Q. 1996;74:511–544.
6. The Diabetes Control and Complications Trial Research Group. The
effect of intensive treatment of diabetes on the development and
progress of long-term complications in insulin-dependent diabetes
mellitus. New Engl J Med. 1993;329:977–986.
7. Hypertension Detection and Follow-up Program Cooperative Group.
Five-year findings of the hypertension detection and follow-up pro-
gam: reduction in mortality of persons with high blood pressure,
including mild hypertension. JAMA. 1979;242:2562–2571.
8. Toseland RW, O’Donnell JC, Engelhardt JB, Hendler SA, Richie JT,
Jue D. Outpatient geriatric evaluation and management: results of a
randomized trial. Med Care. 1996;34:624–640.
9. Toseland RW, O’Donnell JC, Engelhardt JB, Hendler SA, Richie JT,
Jue D. Outpatient geriatric evaluation and management: is there an in-
vestment effect? Gerontologist. 1997;37:324–332.
10. Weinberger M, Oddone EZ, Henderson WG. Does increased access to
primary care reduce hospital admissions? New Engl J Med. 1996;334:
1441–1447.
11. Stuck AE, Aronow HU, Steiner A, et al. A trial of annual in-home
comprehensive geriatric assessments for elderly people living in the
community. New Engl J Med. 1995;333:1184–1189.
12. Schraeder C, Shelton P, Britt T, et al. Geriatric Collaborative Care:
Improving the System of Primary Care. Englewood, CO: Center for
Research in Ambulatory Health Care Administration; 1997.
13. Schraeder C, Shelton P, Britt T, et al. Collaborative care teams. In:
Netting FE, Williams FG, eds. Enhancing Primary Care of Elderly
People. New York: Garland Publishing; 1999:43–64.
14. Roccaforte WH, Burke WJ, Bayer BL, et al. Validation of a telephone
version of the Mini-Mental State Examination. J Am Geriatr Soc.
1992;40:697–702.
15. Radosevich DM, Wetzler H, Wilson SM. Health Status Questionnaire
(HSQ) 2.0: Scoring Comparisons and Reference Data. Bloomington,
MN: Health Outcomes Institute; 1994.
Finally, the results of this study may not be generalizable
to other patient populations residing in different geographi-
cal areas and served by models of primary care with differ-
ent payment mechanisms. The patients enrolled in this study
were from both urban and rural areas and were selected us-
ing broadly defined medical and/or psychosocial criteria.
Further study is needed to validate the results of this study
and to identify those patients most likely to benefit from a
more comprehensive approach to primary care.
The fundamental question raised by this study concerns
the feasibility of collaborative primary care practice in the
current health care environment. The typical primary care
practitioner does not have the resources or, under fee-for-
service reimbursement, the fiscal incentives to implement
16. Health Care Financing Administration. List of diagnosis related
groups (DRGs) by medical diagnostic categories (MDCs). Available

M112
SCHRAEDER ET AL.
at: http://www.hcfa.gov/stats/medpar/drg97dsc.txt. Accessed August
1998.
17. Statistical Package for the Social Sciences. SPSS for Windows, CD-
ROM version 9.0. Chicago: SPSS; 1999.
20. Charlson M, Pompei P, Ales K, et al. A new method of classifying
prognostic comorbidity in longitudinal studies: development and vali-
dation. J Chronic Dis. 1987;40:373–383.
18. Chatellier G, Zapletal E, Lemaitre D, et al. The number needed to
treat: a clinically useful nomogram in its proper context. Br Med J.
1996;312:426–429.
Received October 15, 1999
19. Knaus WA, Draper EA, Wagner DT, et al. APACHE II: a severity of
disease classification system. Crit Care Med. 1985;13:818–829.
Accepted March 22, 2000
Decision Editor: William B. Ershler, MD