introduction the question: is hmo market share associated with adoption of cardiac-care...
TRANSCRIPT
IntroductionIntroduction The Question: Is HMO market share The Question: Is HMO market share
associated with adoption of cardiac-care associated with adoption of cardiac-care technologies, and, in turn with treatments technologies, and, in turn with treatments and outcomes for heart attack patientsand outcomes for heart attack patients
Approach:Approach: Use hospital-level hazard models to study Use hospital-level hazard models to study
relationships between HMO market share and relationships between HMO market share and adoptionadoption
Use patient-level models to study relationship Use patient-level models to study relationship between availability and treatments and between availability and treatments and outcomesoutcomes
Cardiac Care Cardiac Care TechnologiesTechnologies
We focus on three cardiac technologiesWe focus on three cardiac technologies Diagnostic: Cardiac catheterizationDiagnostic: Cardiac catheterization Therapeutic: PTCATherapeutic: PTCA Therapeutic: CABGTherapeutic: CABG
All involve the adoption of equipment and All involve the adoption of equipment and staffstaff
Catheterization and CABG first developed in Catheterization and CABG first developed in the 1960s; PTCA in the 1970s the 1960s; PTCA in the 1970s
Catheterization equipment is used to do PTCACatheterization equipment is used to do PTCA PTCA and CABG are usually adopted togetherPTCA and CABG are usually adopted together
Hospital-Level DataHospital-Level Data
We focus on 2,873 hospitals in MSAs in We focus on 2,873 hospitals in MSAs in operation in 1985operation in 1985
We use Medicare Claims data from 1985-We use Medicare Claims data from 1985-2000 to identify hospitals that adopt these 2000 to identify hospitals that adopt these technologies and the year of adoptiontechnologies and the year of adoption Hospitals with 10 claims for a given service in a Hospitals with 10 claims for a given service in a
calendar year are defined as having the calendar year are defined as having the technology in that yeartechnology in that year
Based on patterns in the data, we study 3 Based on patterns in the data, we study 3 adoption states: none, catheterization only, adoption states: none, catheterization only, and all techologiesand all techologies
Hospital-Level DataHospital-Level Data
We classify hospitals according to the We classify hospitals according to the average 1990-1999 HMO market share average 1990-1999 HMO market share in their MSAin their MSA Low: <10%Low: <10% Medium: 10-30%Medium: 10-30% High: >30%High: >30%
Hospital-Level Adoption Hospital-Level Adoption ModelingModeling
Discrete-time hazard modelsDiscrete-time hazard models Competing risks for probability of moving Competing risks for probability of moving
from none to cath only or none to allfrom none to cath only or none to all Standard hazard model for probability of Standard hazard model for probability of
moving from cath only to allmoving from cath only to all Controls include a range of potential Controls include a range of potential
confounders, including urbanization, confounders, including urbanization, demographics, hospital characteristicsdemographics, hospital characteristics
Hazard Model ResultsHazard Model ResultsNone to Cath OnlyNone to Cath Only
Standard errors in parentheses. Relative Hazards in brackets. Models include additional covariates. * denotes p<0.05; ** denotes p<0.01
Medium HMO 0.174 0.170 0.175(0.119) (0.120) (0.120)[1.190] [1.186] [1.191]
High HMO 0.616 ** 0.611 ** 0.611 **(0.213) (0.215) (0.215)[1.851] [1.843] [1.842]
CICU in 1982 --- 0.711 ** 0.713 **(0.127) (0.127)
MSA AMI mortality --- --- 0.703(1.381)
State dummies yes yes yes
Hazard Model ResultsHazard Model ResultsCath Only to AllCath Only to All
Standard errors in parentheses. Relative Hazards in brackets. Models include additional covariates. * denotes p<0.05; ** denotes p<0.01
Medium HMO -0.202 -0.209 -0.223(0.134) (0.134) (0.136)[0.817] [0.812] [0.800]
High HMO -0.431 # -0.426 # -0.440 #(0.231) (0.231) (0.231)[0.650] [0.653] [0.644]
CICU in 1982 --- 0.501 * 0.505 *(0.225) (0.227)
MSA AMI mortality --- --- -0.815(1.330)
State dummies yes yes yes
Technology availability, Technology availability, treatments, and outcomestreatments, and outcomes
HMO activity affects the probability HMO activity affects the probability a heart attack patient will be treated a heart attack patient will be treated in a hospital with the technologyin a hospital with the technology
Whether or not the hospital of Whether or not the hospital of treatment has the technology affects treatment has the technology affects the probability of actually receiving the probability of actually receiving a treatmenta treatment
Receiving treatments affects Receiving treatments affects mortality ratesmortality rates
Medicare AMI DataMedicare AMI Data
Claims data on a 20% sample of FFS Claims data on a 20% sample of FFS Medicare patients in MSAs with a new Medicare patients in MSAs with a new AMI between 1996 and 2000AMI between 1996 and 2000
N=148,170N=148,170 Measure technology status of index Measure technology status of index
hospital, treatment receipt within 90 hospital, treatment receipt within 90 days of initial admission, and 1 year days of initial admission, and 1 year mortalitymortality
Data also contain detailed data on Data also contain detailed data on comorbidities and other characteristicscomorbidities and other characteristics
StatisticsStatistics
Estimate individual-level modelsEstimate individual-level models Control for a range of characteristicsControl for a range of characteristics
sex; race; age; admission in the prior 2 years for sex; race; age; admission in the prior 2 years for IHD, CHF, VA, or any other cause; conditions at IHD, CHF, VA, or any other cause; conditions at admission: cancer, diabetes, dementia, heart admission: cancer, diabetes, dementia, heart failure, hypertension, stroke, peripheral vascular failure, hypertension, stroke, peripheral vascular disease, chronic obstructive pulmonary disease, disease, chronic obstructive pulmonary disease, respiratory failure, renal failure, or hip fracture; respiratory failure, renal failure, or hip fracture; area per capita income, total area population and area per capita income, total area population and population density; % population graduated high population density; % population graduated high school/college; % of the work force white collar; school/college; % of the work force white collar; squared terms for area characteristics; yearsquared terms for area characteristics; year
Index Hospital CapabilitiesIndex Hospital CapabilitiesMultinomial LogitMultinomial Logit
(results relative to all technologies)(results relative to all technologies)
Models are multinomial logit regressions and include additional covariates and state dummies as well as interactions between HMO variables and year. * denotes p<0.05; ** denotes p<0.01
Probability that index hospital has
Nothing Cath only
Medium HMO -0.004 0.220 **(0.068) (0.050)
High HMO -0.586 ** 0.430 **(0.089) (0.071)
% of cases 12.5 29.4
Treatments ReceivedTreatments ReceivedMultinomial LogitMultinomial Logit
(relative to medical management)(relative to medical management)
Models are multinomial logistic regressions of the probability of receiving cath, PTCA, CABG, or medical management within 90days of initial hospitalization. Models include additional covariates, state dummies, and interactions between tech variables and year. # denotes p<0.10, * denotes p<0.05; ** denotes p<0.01
Cath PTCA CABG
Index Facility Cath Only -0.474 ** -0.174 ** -0.109 #0.059 0.058 0.057
Index Facility All 0.475 ** 0.979 ** 0.603 **0.040 0.040 0.041
N 148,170 148,170 148,170Percentage of all cases 0.149 0.225 0.154
Treatments and 1-year Treatments and 1-year MortalityMortality
Logistic RegressionLogistic Regression
From logistic regression of the probability of 1 year mortality. Models include additional covariates, state dummies, and interactions betweentech variables and year. # denotes p<0.10, * denotes p<0.05; ** denotes p<0.01
Patient got Catheterization -0.222 **(0.007)
Patient got PTCA -0.292 **(0.007)
Patient got CABG -0.295 **(0.007)
N 148,170DV mean 0.339
HMO Market Share and 1-HMO Market Share and 1-Year MortalityYear Mortality
Logistic RegressionLogistic Regression
Coefficient (SE) P-value
medium HMO 0.013 0.128(0.009)
high HMO 0.017 0.188(0.013)
N 148,170DV mean 0.339
From logistic regression of the probability of 1 year mortality. Models include additional covariates, state dummies, and interactions betweenHMO variables and year. # denotes p<0.10, * denotes p<0.05; ** denotes p<0.01
ConclusionsConclusions Managed care activity affected the Managed care activity affected the
adoption of cardiac technologiesadoption of cardiac technologies This could well be associated with This could well be associated with
worse outcomes for AMI patientsworse outcomes for AMI patients impacts on other patients, and other impacts on other patients, and other
outcomes, are unknownoutcomes, are unknown
Means of Hospital Level Means of Hospital Level VariablesVariables
Mean Standard DeviationLow HMO prevalence 0.148 0.355Medium HMO prevalence 0.653 0.476High HMO prevalence 0.199 0.400Teaching hospital 0.289 0.453Medical school affiliation 0.265 0.442Total beds, hundreds 2.63 2.03MSA per capita income, $ thousands 19.4 3.49MSA percentage of population - urban 0.842 0.129MSA percentage of population - urban squared 0.725 0.202MSA percentage of population - 65 and older 0.121 0.030MSA percentage of population - high school 0.766 0.058MSA percentage of population - college 0.217 0.057MSA population, millions 2.01 2.35MSA population, millions squared 9.56 20.2MSA population density, millions/mile 0.968 1.49MSA population density squared 3.15 12.10MSA hospitals per thousand population 0.023 0.008MSA generalists per thousand population 0.082 0.033MSA specialists per thousand population 0.722 0.371MSA cardiologists per thousand population 0.065 0.036
N 2,873
Kaplan-Meier Adoption Probabilities for Kaplan-Meier Adoption Probabilities for PTCA and CABG, 1985-2000PTCA and CABG, 1985-2000
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
Cu
mu
lati
ve
Ad
op
tio
n P
rob
ab
ility
PTCA
CABG
MortalityMortality
Models are OLS regressions of the probability of 1 year all-cause mortality. Models include additional covariates. * denotes p<0.05; ** denotes p<0.01
coefficient SEpatient got catheterization -0.218 0.007 **catheterization*1997 -0.002 0.011catheterization*1998 -0.004 0.011catheterization*1999 0.008 0.011catheterization*2000 0.002 0.011patient got PTCA -0.287 0.007 **PTCA*1997 0.000 0.010PTCA*1998 -0.003 0.010PTCA*1999 -0.003 0.010PTCA*2000 0.010 0.010patient got CABG -0.292 0.007 **CABG*1997 0.005 0.011CABG*1998 -0.001 0.011CABG*1999 -0.002 0.011CABG*2000 0.014 0.011
Predicted PTCA adoption Predicted PTCA adoption probability in low, medium, probability in low, medium,
and high HMO marketsand high HMO marketsCumulative Probability of PTCA adoption
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1.000
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
Cu
mu
lati
ve A
do
pti
on
Pro
bab
ilit
y
low HMO
medium HMO
high HMO
Predicted CABG adoption Predicted CABG adoption probability in low, medium, probability in low, medium,
and high HMO marketsand high HMO marketsCumulative Probability of CABG adoption
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1.000
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
Cu
mu
lati
ve A
do
pti
on
Pro
bab
ilit
y
low HMO
medium HMO
high HMO