Managed care and technology adoption in health care: evidence from magnetic resonance imaging

https://doi.org/10.1016/S0167-6296(01)00072-8Get rights and content

Abstract

This paper empirically examines the relationship between HMO market share and the diffusion of magnetic resonance imaging (MRI) equipment. Across markets, increases in HMO market share are associated with slower diffusion of MRI into hospitals between 1983 and 1993, and with substantially lower overall MRI availability in the mid- and later 1990s. High managed care areas also had markedly lower rates of MRI procedure use. These results suggest that technology adoption in health care can respond to changes in financial and other incentives associated with managed care, which may have implications for health care costs and patient welfare.

Introduction

Among the many questions raised by growth in managed care is the impact it will have on the development, adoption, and use of new medical technologies. The US health care system has been characterized for many decades by rapid technological progress, fueled in part by a reimbursement system that generously payed for the development and use of new advanced therapies (Weisbrod, 1991). Now, some fear that growth in managed care is eroding the pillars that supported this system. Managed care has reduced physician and hospital reimbursement in many parts of the US and moved to limit use of expensive tests and procedures by enrollees, often focusing most intently on spending for new high cost technologies at the most advanced institutions. These kinds of changes could reduce the profitability of new innovations, and slow adoption and reduce the overall availability of technologies. This, in turn, could have ripple effects throughout the process of technology development if researchers and developers perceive changes in the markets for new products and scale back their efforts or alter their research and development strategies.

Changes in the availability of new technologies and the rate of technological change could have important implications for the health care system. It is widely believed that the majority of health care cost growth over the past 50 years is due to technological change (Newhouse, 1992, Fuchs, 1996). Rapid technological change also appears to have significantly improved the capacity of medicine to treat disease and thereby substantially improved the well-being of patients (e.g. Cutler et al., 1998). Understanding the impact of managed care on technology change is thus important for assessing its impact on spending and patient welfare, as well as for evaluating and optimally designing policies that would influence the future development of managed care.

This paper investigates the relationship between managed care activity and the adoption of magnetic resonance imaging (MRI) equipment, a good example of a technology that could have been influenced by managed care. MRI is a diagnostic tool for producing high resolution images of body tissues, most frequently the brain and spinal cord. The first prototype MRI machines for human imaging were installed in the United States in 1980 and MRI entered general clinical use in 1982 (Baltaxe and Geokas, 1983, Hillman and Schwartz, 1985). It diffused during the 1980s and 1990s, the period in which managed care came to play a significant role in the US health care system.

MRI has attracted attention, from managed care plans among others, partly because of its high cost. Hospitals or other health care providers that wish to offer MRI typically purchase and install an MRI scanner.1 The machines themselves are costly, typically running more than US$ 1.5 million for a new imager, and there are often substantial facility modifications that must be made to provide coolant for the magnets and shield surrounding equipment from powerful magnetic fields (Bell, 1996, Steinberg and Evens, 1988). The operating costs of MRI can also be significant. Personnel and maintenance costs can easily amount to US$ 200 000 per year for most facilities, and the cost of supplies were estimated to be more than US$ 100 per scan in 1995 (Bell, 1996). Patient charges associated with MRI procedures vary widely depending on things like the complexity of the images required and physician charges for interpretation, but it is not difficult to run up total charges of more than US$ 1000 for many MRI procedures.

MRI can improve the speed and accuracy with which many diagnoses can be made. For some conditions and patients, though, MRI scans can be viewed as helpful but not necessary. It is sometimes possible to substitute other less expensive procedures like computed tomography (CT) or ultrasound for MRI with arguably limited effects on the quality of the diagnosis. Virtually, all insurers cover medically indicated MRI procedures, but health plans concerned about costs have been particularly vigilant about the use of MRI in cases where its benefits are not immediately clear or where other procedures could be substituted.

Section 2 briefly reviews mechanisms by which managed care could have influenced adoption of MRI. I then turn to the empirical work. Using data on hospital adoption of MRI and data on the overall availability of MRI, I examine MRI diffusion and availability in markets with varying levels of HMO market share. I conclude that managed care was associated with slower diffusion of MRI, particularly in hospitals. With expenditures for the purchase of MRI equipment typically running in the millions of dollars, reductions in adoption should have produced substantial savings. But, achieving these savings may also have entailed reductions in the use of services that benefit patients, which would leave questions about the net welfare effects open. After studying the effects of managed care on adoption patterns, I briefly examine the relationship between HMO market share and the number of MRI procedures performed. The results suggest that increases in HMO market share are strongly associated with declines in the utilization of MRI procedures.

The general patterns and determinants of technology adoption in health care, including the case of MRI, have been widely studied.2 The overall impacts of managed care on spending and premiums have also been previously examined.3 However, there is relatively little work that examines the effects of managed care on technology adoption and use in detail. Cutler and Sheiner (1998) investigated the relationship between state-level HMO market share and the availability of a range of services in hospitals, finding evidence that managed care slowed the diffusion of technologies that were diffusing recently. Baker and Brown (1999) report that managed care reduced the number of mammography providers, but increased the volume of procedures performed at the remaining sites. Baker and Phibbs (2000) report that high HMO market share was associated with slower adoption of neonatal intensive care units between 1980 and 1996. The literature is not unanimous, however, Baker and Spetz (1999) report that managed care activity was not associated with changes in an index of hospital technologies, suggesting that managed care may not slow overall adoption, and Hill and Wolfe (1997) reported mixed effects of managed care on diffusion of a range of technologies in Wisconsin during and after rapid growth in managed care activity.4

Section snippets

Managed care activity and technology adoption

“Managed care”, for purposes of discussion here, describes a collection of activities health plans can undertake that are designed to reduce the high levels of utilization and spending that accompanied unfettered fee-for-service medicine, and improve the efficiency of health care delivery. These activities can include a range of things like the use of financial incentives to influence utilization patterns, direct oversight of utilization decisions, selective contracting with preferred

Empirical approach and data

I study the impact of managed care on MRI by comparing MRI diffusion and availability in markets with varying levels of managed care activity. I define markets as health care service areas (HCSAs). HCSAs are groups of counties constructed to approximate markets for health care services based on Medicare patient flow data (Makuc et al., 1991). There are 802 HCSAs covering the entire continental United States.

I use data on MRI diffusion and availability from three complementary sources. I begin

Overall MRI diffusion and availability

Table 1 summarizes adoption of MRI by hospitals in the 1983–1993 AHA survey sample. Over this time period, 1176 hospitals adopted MRI. Kaplan-Meier estimates of the cumulative adoption probability rise from 0 to nearly 25% by 1993. A total of 90 of the original 5344 hospitals are censored before 1993, when all remaining hospitals are censored. The majority of these are censored because they close or merge during the time period, and the large number of hospitals in this category is consistent

HMOs and hospital MRI, 1983–1993

Hazard models provide a natural framework for studying technology adoption by a well defined set of candidate adopters (e.g. Rose and Joskow, 1990, Cutler and McClellan, 1996). Denoting the cumulative probability that hospital i has MRI at time t by Fi(t) and the density function as fi(t), the hazard is defined as the probability that hospital i acquires MRI at time t conditional on not having acquired MRI up to that point: λi(t)=fi(t)/[1−Fi(t)]. I parameterize the hazard using a proportional

HMOs and MRI machines in and out of hospitals, 1993 and 1995

The AHA data provide only part of the picture of MRI availability in the United States since many MRI machines are not in hospitals, and even for hospitals the AHA surveys do not provide information about the number of installed units. The 1993 and 1995 MRI censuses provide data on the number of magnets operated in and outside of hospitals.

I used MRI census data for 707 HCSAs with populations over 25 000 to estimate regressions of the formMAGj,1993POPj,199301HMOj2XjjMAGj,1995POPj,19950

HMOs and hospital MRI After 1993

A natural question is the extent to which these effects have persisted into the later 1990s. AHA data from 1994 to 1998 can provide information about more recent trends in hospital MRI adoption. I used this data to compute the number of hospitals that indicated having MRI “owned or provided by the hospital or a subsidiary” per 100 000 population in each area, which should serve as a general measure of MRI availability in the area. I then estimated annual cross-sectional OLS models of the formMRI

MRI utilization

Reductions in adoptions could reduce patient access to MRI equipment.

Conclusions

Data from AHA surveys and from two MRI censuses indicate that increases in HMO activity are associated with slower diffusion of MRI equipment and lower overall MRI availability. This is consistent with the view that changes in incentives brought about by managed care can have important impacts on technology diffusion in health care.

Since MRI is still diffusing, it is probably too early to say for sure that the equilibrium levels of MRI will be lower in high managed care areas, although that

Acknowledgements

I thank David Cutler, Randall Ellis, Alan Garber, William Vogt, anonymous referees, and many seminar participants for helpful comments. Susan Wheeler provided invaluable assistance with data analysis. This project was supported by a grant from the Robert Wood Johnson Foundation’s Health Care Financing and Organization Program.

References (55)

  • L.C Baker et al.

    HMO penetration and the cost of health care: market discipline or market segmentation?

    American Economic Review

    (1996)
  • Baker, L.C., Phibbs, C.S., 2000. Managed care, technology adoption, and health care: the adoption of neonatal intensive...
  • Baker, L.C., Spetz, J., 1999. Managed care and medical technology growth. In: Garber, A.M. (Ed.), Frontiers in Health...
  • H.A Baltaxe et al.

    Nuclear magnetic resonance

    Annals of Internal Medicine

    (1983)
  • H.D Banta

    The diffusion of the computed technology (CT) scanner in the United States

    International Journal of Health Services

    (1980)
  • R.A Bell

    Economics of MRI technology

    Journal of Magnetic Resonance Imaging

    (1996)
  • M Chernew

    The impact of non-IPA HMOs on the number of hospitals and hospital capacity

    Inquiry

    (1995)
  • Clement, D.G., Gleason, P.M., Brown, R.S., 1992. The effects of risk contract hmo market penetration on medicare...
  • Cutler, D.M., Sheiner, L., 1998. Managed care and the growth of medical expenditures. In: Garber, A.M. (Ed.), Frontiers...
  • Cutler, D.M., McClellan, M.B., 1996. The determinants of technological change in heart attack treatment. National...
  • Cutler, D.M., McClellan, M.B., Newhouse, J.P., 1998. The costs and benefits of intensive treatments for cardiovascular...
  • R Feldman et al.

    The effect of HMOs on premiums in employment-based health plans

    Health Services Research

    (1993)
  • R Feldman et al.

    The competitive impact of health maintenance organizations on hospital finances: an exploratory study

    Journal of Health Politics, Policy, and Law

    (1986)
  • A.M Fendrick et al.

    Hospital adoption of laparoscopic cholecystectomy

    Medical Care

    (1994)
  • R.G Frank et al.

    The competitive effects of HMOs: a review of the evidence

    Inquiry

    (1985)
  • V Fuchs

    Economics, values, and health care reform

    American Economic Review

    (1996)
  • GHAA, various years. National Directory of HMOs. Group Health Association of America, Washington,...
  • Cited by (111)

    • Competition-driven physician-induced demand

      2021, Journal of Health Economics
    • Utility and Access to 3-Dimensional Printing in the Context of Congenital Heart Disease: An International Physician Survey Study

      2020, CJC Open
      Citation Excerpt :

      In the late 1980s, the United States had approximately 8 times more MRI units per capita than Canada.26 The combination of capital equipment and skilled human resources needed to initiate an MRI facility is similar to a 3D printing program.27 Our study found that in practice, the most common use for 3D models of CHD lesions is for procedural planning, specifically for surgical planning.

    View all citing articles on Scopus
    View full text