Docstoc

Spatial Analysis of Healthcare Markets

Document Sample
Spatial Analysis of Healthcare Markets Powered By Docstoc
					Spatial Analysis of Healthcare
Markets: Separating the Signal
from the Noise in ACSC
Admission Rates
Presented at
the AcademyHealth 2004 Annual Research Meeting,
San Diego, CA, June 6, 2004
Presented by: Lee R. Mobley, PhD
Co-Authors: Elisabeth Root, MA
              Nancy McCall, ScD
              Sujha Subramanian, PhD
              Mary Kapp, MPhil
              Barbara Gage, PhD
P.O. Box 12194 · 3040 Cornwallis Road · Research Triangle Park, NC 27709
Phone: 919-541-7195 · Fax: 919-541-7384 · lmobley@rti.org · www.rti.org
RTI International is a trade name of Research Triangle Institute.          1
Overview

   This paper examines the association between
    geographic or market-level supply and demand
    factors and market-level rates of three ambulatory
    care sensitive conditions (ACSCs):
       chronic obstructive pulmonary disease (COPD)
       congestive heart failure (CHF), and
       lower limb peripheral vascular diseases (PVD)
Study Population: Medicare FFS beneficiaries over
  two time periods: mid nineties and latter nineties.
Markets Definition: 306 Hospital Referral Regions
 from the Dartmouth Atlas Project
                                                         2
Spatial Market Analysis:
Motivation
   In the previous market-level analysis, we found
    that Census division effects were significant
    even after controlling for demographic and
    disease severity factors, which suggests that
    there may be important characteristics of places
    that are omitted in the analysis.
   Market-specific factors that impact access or
    continuity of care may be important variables to
    examine in explaining ACSC admission rates.
   Medicare policy changes in the latter nineties
    may have impacted these market-specific
    effects.
                                                       3
             CHF Rates in HRRs,
             1995–1997




CHF_EARLY
    0.014 - 0.024
    0.024 - 0.032
    0.032 - 0.051
    NO_HRR
                                  4
           CHF Rates in HRRs,
           1998–2000




CHF_LATE
    0.014 - 0.024
    0.024 - 0.032
    0.032 - 0.056
    NO_HRR
                                5
           COPD Rates in HRRs,
           1995–1997




COPD_EARLY
    0.005 - 0.014
    0.014 - 0.02
    0.02 - 0.038
    NO_HRR
                                 6
           COPD Rates in HRRs,
           1998–2000




COPD_LATE
    0.006 - 0.014
    0.014 - 0.02
    0.02 - 0.04
    NO_HRR
                                 7
           PVD Rates in HRRs,
           1995–1997




PVD_EARLY
    0 - 0.002
    0.002 - 0.003
    0.003 - 0.008
    NO_HRR
                                8
           PVD Rates in HRRs,
           1998–2000




PVD_LATE
    0 - 0.002
    0.002 - 0.003
    0.003 - 0.007
    NO_HRR
                                9
Access and Utilization

   Access to care and utilization of health
    services is impacted by many factors:
      Population characteristics

      Health system characteristics

      Local area market characteristics



   Our conceptual model combines these three
    elements to show the pathways to realized
    demand (actual utilization)

                                                10
                Methods:
                Conceptual Model of Access
                (Khan and Bhardwaj, 1994; WHO, 2000)

X1: Characteristics of Health Care                 X2: Characteristics of Health Care
System                                             System
Place Specific Variables                           Person-specific Variables
•(Supply/Attraction)                               •Age, gender
•-Number/Location of Facilities                    •Race, ethnicity
•-Number/Location of Nurses                        •Education
•-HMO Penetration                                  •Morbidity
                                                   •Income, Poverty



                                X3: Barriers/Facilitators
                                (Intervening Factors)
                                •Transportation systems
                                •Traffic congestion
Potential Access                •Distance to facilities                Utilization of Health
                                •Climate                               Care Services
                                •Safety
                                                                                               11
Methods: Empirical Challenge

   The intervening factors that may be so
    important in determining elderly utilization are
    difficult to measure (proxy: population density).
   Other important missing variables are practice
    patterns and/or health behaviors that may vary
    significantly from place to place, yet may be
    similar in local regions (spillovers).
   The empirical challenge is to find a model
    specification that accounts for these missing
    variables so that their omission does not impart
    bias on other parameters of interest.

                                                        12
Methods

   Three ACSCs were examined: PVD, COPD,
    CHF; 3-year rates were constructed for 1995–
    1997 and 1998–2000 (ACSC admission/all
    beneficiaries in 1,000s).
   We combined inpatient and ER/observation bed
    stays.
   Beneficiary and county-level data were
    aggregated to HRRs, yielding 306 observations
    in an early period and 306 in a late period.
   SpaceStat software was used to estimate a
    spillovers spatial regression model

                                                    13
Methods

   Primary Data Sources for explanatory variables
      CMS Enrollment Data File and claims Files

      CMS Provider of Service (OSCAR) file

      CMS Medicare Managed Care Penetration file

      AMA Physician Masterfile

      AHA County Hospital File

      US Census of Populations

      Interstudy

      AARP




                                                     14
Methods

   Sociodemographic Characteristics
      Proportion of elderly in poverty (1989,1999)

      Proportion of county that are elderly

      Proportion of sample Black

      Proportion of sample male

      Proportion of sample >80

      Proportion of sample dual enrolled

      Proportion of sample who died

      Proportion of elderly with supplemental insurance




                                                           15
Methods

   Health Status Characteristics
      Median PIP-DCG score

      Proportion of sample with ESRD

      Proportion of sample with diabetes




                                            16
Methods

   Market Characteristics
      M+C Penetration

      Proportion of population in private HMOs

      Home health visits per Medicare insured

      Medicare admissions to SNFs

      Hospital inpatient occupancy of staffed beds

      Number of non-Federal practicing MDs

      Number of FTE, hospital-based RNs

      Number of SNFs

      Number of HHAs
      Number of Hospices

      Number of Rural Health Clinics
                                                      17
Methods

   Market Characteristics
      Number of hospitals with outreach programs

      Number of hospitals with assisted living programs

      Number of hospitals with rehabilitation programs

      Number of hospitals with transportation

      Number of hospitals with home health services




                                                           18
Methods

   Access Proxies
      Proportion of the population who said they didn’t
       visit a physician due to cost
      Proportion of the population who reported
       problems accessing a primary care provider




                                                           19
Methods

   We estimate the ecological model on data
    from two separate cross sections to assess
    whether factor effects changed over a time
   We indirectly examine the influence of SNF
    and HH payment reforms on the market rate
    of ACSC hospitalizations




                                                 20
Empirical Findings

   Beneficiary characteristics explain most of
    the market-level variation in ACSC
    admissions.
   Poverty among the elderly has become an
    increasingly important predictor of all three
    ACSCs over time — the large, positive
    association with COPD and CHF increases
    (doubles in magnitude) over time.
       The mean proportion of the elderly in
        poverty declined nationally between
        1989 and 1999.
                                                    21
Empirical Findings:
Demographic Factors
   Places with higher proportions of the oldest-old
    have lower COPD rates (and increasingly so over
    time) and lower CHF rates (and diminishing over
    time)
   Places with higher proportions of beneficiaries who
    died had higher COPD and CHF rates, fairly stable
    over time.
   Places with higher proportions of black beneficiaries
    had lower PVD and COPD rates
   Places with greater proportions of diabetics and
    ESRD had higher PVD rates and more home health
    visits per beneficiary (magnitude doubled over time)
                                                            22
Empirical Findings:
Market Factors
   Places with more SNFs exhibit higher COPD and
    CHF admit rates – but stable over time.
   No association between number of HHAs and ACSC
    admit rates but number of HHA visits positively
    associated with PVD hospitalizations and
    increasingly so over time.
   Places with more hospital-based rehabilitation
    programs have lower CHF and COPD rates – smaller
    effect over time.
   Managed care penetration of Medicare market has
    no change in effect over time for CHF and PVD and
    modest negative effect for COPD in later period

                                                        23
Other Factors

   Inpatient hospital occupancy rate was positively
    associated with PVD rates in the early period,
    and negatively associated with CHF rates in the
    later period.
   Places with higher HMO penetration in the
    private market show lower PVD and COPD
    hospitalization rates in the early period.
   Places with higher proportions of the elderly
    holding supplemental insurances show lower
    COPD hospitalization rates in the early period
    and higher CHF rates in the later period.

                                                       24
Other Factors

   Places where higher proportions of the
    population ‘didn’t visit a doctor because of cost’
    showed positive association with COPD rates
    in the later period.
   The numbers of physicians and registered
    nurses, and the statewide measure of
    physician shortage, were surprisingly
    insignificant in these models.
   Other supply variables such as hospital
    services and other post-acute care services
    were also surprisingly silent.
                                                         25
Empirical Findings:
Spatial Spillovers
Spatial Lag Model: Ri = jI wijRj +  Xi + ui
 Is the measure of strength of spatial spillovers
The estimate of  is 0.164 for PVD, >0.50 for
COPD, and >0.40 for CHF
For PVD, spillovers are local (2 closest HRRs)
while for COPD and CHF, spillovers are regional
(6 closest HRRs)
 Given the spatial clustering observed in COPD
and CHF, geographically-targeted interventions
may be possible.

                                                     26
CHF & COPD Rates ('98–'00) and Elderly in
Poverty (1999)




                                            27
Change in CHF & COPD Rates ('95–'97) to
('98–'00), and Change in Elderly Poverty Rate




                                                28

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:61
posted:4/14/2008
language:English
pages:28