Hospital Ownership and Performance: An Integrative Research Review - PowerPoint PPT Presentation

Loading...

PPT – Hospital Ownership and Performance: An Integrative Research Review PowerPoint presentation | free to download - id: 1b301c-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Hospital Ownership and Performance: An Integrative Research Review

Description:

Funded by grant #050953 under the Robert Wood Johnson Foundation's Changes in ... Altruism / Quality-Quantity Maximization (Newhouse 1970; Lakdawalla & Philipson 1998) ... – PowerPoint PPT presentation

Number of Views:45
Avg rating:3.0/5.0
Slides: 50
Provided by: wendy89
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Hospital Ownership and Performance: An Integrative Research Review


1
Hospital Ownership and Performance An
Integrative Research Review
Preliminary work in progress Comments welcome
  • Research-in-Progress Seminar
  • Stanford, May 11, 2005
  • Yu-Chu Shen
  • Naval Postgraduate School and NBER
  • Karen Eggleston, Joseph Lau, Christopher Schmid
  • Tufts University
  • Funded by grant 050953 under the Robert Wood
    Johnson Foundations Changes in Health Care
    Financing and Organization (HCFO) Initiative

2
Presentation Outline
  • Research objective
  • Brief theory background
  • State of the empirical literature
  • Scope of our integrative review
  • Analytical methods to synthesize literature
  • Results
  • Discussions

3
Mixed Ownership Is an Abiding Feature of
Healthcare Delivery in the US
Source Eggleston (2004), based on Rorem (1930)
Hayes (1954) American Hospital Association
Hospital Statistics (various years).
4
Research Objective
  • Does ownership affect hospital performance
    (quality, finance, or provision of uncompensated
    care)?
  • Competing theories with contrasting predictions
  • Hundreds of empirical studies to date with
    conflicting findings
  • policymakers have little clear evidence
  • economics of ownership and behavior imperfectly
    understood

5
Taxonomy of Theories of Nonprofits
Complete Information Incomplete Information
Objectives Differ Altruism / Quality-Quantity Maximization (Newhouse 1970 Lakdawalla Philipson 1998) Physicians Cooperative (Pauly Redisch 1973) NPs help government fulfill demand for collective goods (Weisbrod 1975) NPs as for-profits in disguise (Weisbrod 1988) Government favoritism of NPs because trustworthy (James 1985)
Choice (objectives need not differ) Regulation and tax treatment Firms differ in ability to benefit from a given ownership form (David 2004 Lakdawalla Philipson 1998) Trust signal and concontractible quality (Arrow 1963 Hansmann 1980 Glaeser and Shleifer 1998) Mechanism for consumer control (Ben-Ner and Gui 1993)
(modified from Guy David 2004)
6
Empirical Predictions Consistent With Some
Ownership Theories
7
Mixed Empirical Evidence
  • Studies differ widely in analytic methods
  • Mixed and inconclusive evidence on whether
    ownership differs and the magnitude of
    differences in quality, cost, and social benefits
  • We use meta analytical methods to combine
    quantitative evidence from different studies

8
Scope of the Integrative Review
  • Synthesize the main findings of the empirical
    literature between 1990 and July 2004 on hospital
    ownership and performance (published or
    unpublished)
  • Examine multivariate empirical studies of US
    acute general short stay hospitals
  • Examine studies that compare differences between
    for-profits and nonprofits, between nonprofits
    and government, or both.

9
Scope of the Integrative Review
  • Focus on four broad categories of performance
    measures
  • financial performance (efficiency, cost, revenue,
    profit)
  • quality / patient outcomes
  • uncompensated care or community benefits
  • Staffing
  • Presenting only findings from financial
    performance measures

10
Literature Selection Process
  • 1434 potentially relevant studies from 1990 to
    2004 were identified and screened for retrieval
    through
  • search engines (EconLit, MedLine, Proquest, ABI)
  • contacting all corresponding authors of initially
    included studies

11
Defining Study Population
12
Inclusion and Study Design Criteria
13
Outcome Criteria and Other Exclusions
14
Number of Studies By Category of Hospital
Performance
Performance Measure Number of studies with this as primary measure Number of studies that analyze this measure
Cost, revenue, profits 45 47
Efficiency 21 22
Cost shifting 3 4
Staffing 8 13
Patient outcomes 44 59
Uncompensated care/community benefits 16 23
Other misc outcomes 4 4
TOTAL 141 172
15
Detailed Financial Performance Categories
of articles
Total or operating cost 21
Profit margin (total or operating) 14
Patient revenue or returns on assets 13
Overall efficiency 12

16
Detailed Financial Performance Categories
Technical efficiency--efficient levels of inputs 6
cost of a specific disease 5
Medicare cost 5
scale efficiency--efficient output/input mix 3
misc efficiency measures 3
labor or personnel cost 3
allocative inefficiency--efficient mix of inputs 2
debt/asset ratio 2
payroll/labor as a share of revenue or expense 2
changes in total or operating cost 7
changes in total or patient revenue 3
changes in other financial measures 3
changes in profit or margin 1
misc financial outcomes 14
17
How much work is a systematic review?
  • Allen and Olkin (1999) analyze 37 meta-analyses
  • Average hours were 1138 per study
  • Based on their formula, it implied 1,044 hours
    for our review

18
Analytical Methods
  • A typical study estimates the impact of ownership
    on performance as follows
  • The coefficients ß1 and ß2 capture the effect on
    Y of for-profit and public ownership,
    respectively, relative to nonprofit ownership

19
Defining Effect Size of Ownership Studies (1)
  • The goal of our integrative analysis is to answer
    the following questions
  • What is the magnitude of the relationship between
    ownership and performancewhat is the effect
    size?
  • How precise or reliable is this estimated effect
    size?
  • How do differences in analytic methods and other
    study features affect the estimates of effect
    size?

20
Defining Effect Size of Ownership Studies (2)
  • Problems with using ß1 and ß2 directly from
    studies
  • Heterogeneous dependent variables
  • Effect size can be measured in actual dollars or
    in percentage.

21
Defining Effect Size of Ownership Studies (3)
  • Partial correlation coefficient as a measure of
    effect size
  • Ythe residuals in a regression of Y on a set of
    X
  • FPthe residuals in a regression of FP ownership
    on a set of X
  • The partial correlation r is the simple
    correlation between Y and FP.
  • r measures the correlation between a given
    ownership and Y controlling for the effect of X

22
Partial Correlation Coefficient As Effect Size
  • Partial correlation coefficient, r, measures the
    correlation between a given ownership and Y
    controlling for the effect of X
  • It can be derived from commonly reported
    statistics
  • r
  • Its unit free, so comparable across a
    heterogeneous set of studies
  • Unlike t-statistics, magnitude of r does not
    depend on sample size

23
Adjusting Effect Size Estimates
  • The distribution of r becomes more skewed as the
    population value of r gets further and further
    away from zero.
  • We apply Fisher (1928) transformation that is
    distributed nearly normally

24
Estimating Confidence Intervals Around the Effect
Size
  • Adjusted effect size
  • Variance(Zr)
  • 95 Confidence interval of the adjusted effect
    size

25
Combining Effect Sizes Across Studies (1)
  • A common way to combine study results is to
    compute a weighted average effect size
  • The weight that minimizes the variance of this
    measure is the inverse of the effect size
    variance from each study

26
Combining Effect Sizes Across Studies (2)
  • Variance of weighted average effect size
  • Confidence interval of the combined effect size
    measures

27
Issues In Combining Effect Size For Hospital
Ownership Literature
  • Research questions are not homogeneous.
  • Studies vary widely in analytical methods. Need
    to categorize the methods in some ways.
  • Overlapping hospitals and data sources
  • Unlike randomized clinical trials with
    independent samples, there are fewer than 5000
    general acute hospitals in the US. Many studies
    analyze almost the entire population of hospitals
  • Furthermore, most studies use one of two common
    data sources.

28
Research Questions Vary
  • Fixed- or random-effects models?
  • When the combined studies are a homogeneous set
    designed to answer the same question in the same
    population, a fixed-effects model is appropriate.
  • When heterogeneity is detected, random-effects
    models are used, which assume that there is no
    single truth, but a distribution of such truths.

29
Random Effects Model
  • True effect size is not fixed. The variance of
    effect size from each study is assumed to have
    two components
  • Between-studies variance
  • Within-study variance
  • Because of the additional between-studies
    variance, random effects model tend to be more
    conservative than fixed-effects model.

30
Categorizing Analytical Methods
  • Three types of methodology rigor
  • Type 3 if a study meets both of the following
    conditions
  • (a) uses panel estimation or explicitly
    accounts for potential selection problem
  • (b) includes two of the following three sets of
    controls patient level, hospital level, market
    level
  • Type 2 if meets EITHER (a) or (b)
  • Type 1 if meets NEITHER (a) nor (b)

31
Overlapping Sample and Data Sources
  • Partial correlation coefficients are valid effect
    size measures when observations are correlated.
  • Meta regression has been suggested as a possible
    imperfect solution by including dummies of the
    common data sources.
  • No satisfactory solutions to date.

32
Meta Regression
  • The regression approach allows us to examine
    whether differences in effect sizes across
    studies can be explained by analytical methods,
    region studied, years covered, or other study
    features.
  • The dependent variable is the effect size from
    each study.
  • The explanatory variables are the empirical
    features of each study (differ across financial
    measures)
  • The model is necessarily parsimonious due to
    sample size issues.

33
Integrative Review of Hospital Overall Efficiency
  • Overall efficiency is usually defined as least
    cost production or least amount of input for a
    given level of output.
  • Two common ways to estimate overall efficiency
  • Stochastic frontier approach
  • Data envelopment analysis (DEA)
  • Both are controversial (e.g. Newhouse 1994)
  • 10 studies contain N-F comparison
  • 7 studies contain N-G comparison

34
Efficiency Summary of Effect Size By Methods
(N-F)
35
Efficiency Summary of Effect Size By Decade (N-F)
36
Efficiency Summary of Effect Size By Covered
Region (N-F)
37
Efficiency Potential Publication Bias?N-F
Comparison
38
Integrative Review of Hospital Cost N-F
Differences
  • Studies assume different functional forms for the
    cost model
  • Log (total cost)
  • 6 studies/10 observations
  • Log (average cost per admission)
  • 5 studies/11 observations
  • Average cost per admission/discharge
  • 3 studies/6 observations
  • Others
  • 4 studies/4 observations

39
Cost Summary of N-F Effect Size By Cost
Definition
Other Cost Def
Log(Total Cost)
Log(Avg Cost)
Avg Cost
40
Cost Summary of N-F Effect Size By Decades
Data from 1980s
Data from 1990s
41
Cost Summary of N-F Effect Size By Covered Years
1 year of data
Multiple year of data
42
Cost Summary of N-F Effect Size By Method Types
Method Type 1
Method Type 2
Method Type 3
43
Cost Potential Publication Bias?N-F Comparison
44
Integrative Review of Hospital Revenue N-F
Differences
  • Studies assume different functional forms for the
    revenue model
  • Log (average revenue)
  • 3 studies/4 observations
  • Average cost per admission)
  • 3 studies/3 observations
  • Returns on assets
  • 5 studies/5 observations
  • Others
  • 1 study/1 observations

45
Revenue Summary of N-F Effect Size By Revenue
Definition
Log(average revenue)
Average revenue
Returns on assets
46
Revenue Summary of N-F Effect Size By Covered
Region
47
Revenue Summary of N-F Effect Size By Method
Types
48
Revenue Potential Publication Bias?N-F
Comparison
49
Integrative Review of Profit Margin N-F
Differences
  • Profit margins are usually defined in the form of
    (revenue-cost)/revenue.

50
Profit Margin Summary of N-F Effect Size By
Covered Region
CA
FL
Urban
National
VA
51
Profit Margin Summary of N-F Effect Size By
Method Types
52
Profit margin Potential Publication Bias? N-F
Comparison
53
What Do We Learn? (1)
  • Evidence is pretty conclusive regarding revenue
    and profit margins
  • For-Profits tend to earn more revenue (per
    admission) and have higher profit margins
  • There is little evidence of any difference in
    cost between FP and NP hospitals
  • Evidence is inconclusive regarding efficiency.
  • Although almost all individual studies report
    significant findings, collectively their results
    are not consistent.

54
What Do We Learn? (2)
  • Functional forms and analytical methods matter
  • Weaker methods and functional forms tend to
    predict larger differences between nonprofits and
    for-profits
  • National samples tend to produce more
    conservative estimates of effect size
  • No evidence of publication bias
About PowerShow.com