ICES | Primary Care Models in Ontario English - page 16

Comparison of Primary Care Models in Ontario by Demographics, Case Mix and Emergency Department Use, 2008/09 to 2009/10
The ACG system assigns all ICD-9 and ICD-10
codes to one of 32 diagnosis clusters known
as Adjusted Diagnosis Groups (ADGs).
Individual diseases or conditions are placed
into a single ADG cluster based on five clinical
dimensions: duration of the condition, severity
of the condition, diagnostic certainty, etiology
of the condition and specialty care
involvement. In addition to ADGs, the ACG
software was used to generate Resource
Utilization Bands (RUBs) which involve
aggregations of ACGs with similar expected
utilization (1=low, 5=high) and the
Standardized ACG Morbidity Index (SAMI). The
SAMI was developed at the Manitoba Centre
for Health Policy.
This index is a set of
illness weights for the ACGs using average
provincial health care costs, and can be used
for examining differential morbidity at a
practice level and explaining variation
between practices. SAMI has been adapted by
ICES for use in Ontario and has used the full
value of in-basket FHO primary care services
to weight the ACGs.
These weights are a
measure of expected workload in a FHO
All physician diagnoses, including those made
by primary care physicians and specialists,
and all hospital discharge abstracts were
used to run the Johns Hopkins ACGs. CHC
providers can record more than one diagnosis
at each visit, but OHIP allows only a single
diagnosis per visit. In order to allow fair
comparisons across models, a random
diagnosis was chosen for each CHC visit. In
addition, analyses were limited to physicians
because nurse practitioner data were
available at the encounter level in CHCs but
not in FHTs.
Disease cohorts were used as a secondary
measure of case mix. In this study the
following cohorts were included: diabetes,
asthma, chronic obstructive pulmonary
disease, and mental illness (psychotic and
Most of these cohorts
derive from validated disease algorithms
which include hospital admission data,
require more than one physician visit and are
cumulative over time. Our approach to
producing disease cohorts that were
comparable across models was to link CHC
data with physician visits and hospital
admissions. As there were only two years of
CHC data available, we adapted these
algorithms to use a single physician visit or
hospital admission with a disease-specific
diagnosis within a two-year period. This
approach is similar to the validation used for
mental health
but would result in slightly
higher sensitivity and lower specificity for the
other validated algorithms.
Descriptive analyses were conducted to
determine the number and proportion of
people in each demographic, urban-rural
location and case mix group. The number of
ED visits and average number of ED visits
were calculated for comparisons across
models. Poisson multiple regression was
conducted to produce a risk-adjusted rate of
ED utilization per person (i.e., expected ED
visits) controlling for age, sex, SAMI, income
quintile and rurality. The observed utilization
(unadjusted) is the actual number of ED visits.
These data were used to produce the ratio of
observed to expected ED visits and 95%
confidence intervals.
This study was approved by Sunnybrook
Health Sciences Centre Research
Ethics Board.
1...,6,7,8,9,10,11,12,13,14,15 17,18,19,20,21,22,23,24,25,26,...41
Powered by FlippingBook