Characteristics Associated with Hospital Health IT Vendor Switching and Dropping

Abstract

While a growing body of research has investigated the diffusion of health IT among providers, no empirical research has yet focused on health IT vendor switching by hospitals. Vendor switching is one indicator of a competitive commercial vendor market, and competition among vendors can spur innovations which contribute to better products over time. This study examines the interaction of hospitals with commercial vendors in the recent past to serve as a baseline for future investigations into how the federal health IT incentive program influences changes in the vendor market and vendor-provider relationships. We find that there has been considerable switching between vendors by hospitals, including some hospitals switching away from automated systems all together. Furthermore, our descriptive cross-sectional analysis reveals various hospital characteristics which are associated with vendor switching and dropping, including lower constraints on hospitals’ financial resources, nonprofit ownership, and having some form of integrated arrangement with physicians.

Introduction

In 2009, Congress passed the American Recovery and Reinvestment Act which includes $22 billion to promote the adoption and use of health IT, including EHRs. Eighteen billion dollars of these funds will be distributed as rewards to providers for meeting various criteria for “meaningful use” of health IT through the Medicare and Medicaid insurance programs. The emphasis on meaningful use stems from concerns about using the funds to do more than merely influence providers to purchase health IT equipment, but moreover to make full use of the capabilities of this equipment and software to improve the quality of care. Criteria to assess meaningful use include use of EHRs which meet particular certification standards, and vendors will certainly update their products to meet these criteria by including the various functions called for in the regulations. However, other characteristics of EHRs which have a significant impact on their ability to contribute to improved quality of care are not addressed by the certification criteria, and incentives for continuing innovation and improvement are left to competition in the commercial marketplace. Usability of the technology is one example of such a characteristic for which no particular standards are endorsed by the meaningful use certification criteria.

While a growing body of research has investigated the diffusion of health IT among providers^4–6, to the best of our knowledge no empirical research has yet focused on hospital switching behavior among health IT vendors. This represents a significant gap in the research literature on health IT as vendor switching is one indicator of a competitive commercial vendor market, and competition among vendors can spur innovations which contribute to better products over time. At the very least, vendor switching indicates some degree of dissatisfaction with earlier IT systems even if the newly chosen vendor fails to surpass the quality of the old system. Given the widely acknowledged importance of vendor efforts for achieving effective implementation of health IT in health delivery settings^1,7, a clearer understanding of the interaction of hospitals with commercial vendors in the recent past can serve as a useful baseline for future investigations into how the federal health IT incentive program influences changes in the vendor market and vendor-provider relationships. Towards that end, with this study we address the question of which hospital characteristics are determinants of hospital vendor switching and dropping behavior.

We find that there has been considerable vendor switching by hospitals, including some hospitals switching away from automated systems all together. These results may reflect what some have characterized as the immaturity of the technology during the early part of the period which we examine⁷. Furthermore, our descriptive cross-sectional analysis reveals various hospital characteristics which are associated with vendor switching and dropping, including constraints on hospitals’ financial resources, non-profit ownership, and not having some form of integrated arrangement with physicians.

Data

We used the HIMSS Analytics surveys from 2003 and 2008, which provide information on health information technology use among a large national sample of non-federal acute care hospitals. In particular this survey includes names of vendors for health IT applications in use by hospitals. Other hospital characteristics, including not-for-profit, for-profit or government ownership, metropolitan location, membership in a multi-hospital system, and total admissions come from the American Hospital Association’s annual survey. For most analyses we excluded all hospitals that reported having a health IT system in place but did not report a vendor name (a small portion of all reported systems). Several mergers among vendors took place during the study period (2003–2008) and we treated these cases as the hospital maintaining the same vendor.ⁱ

We examine reported vendor changes for two particular health IT applications in the HIMSS survey: enterprise-wide electronic medical records (henceforth designated as EMR) and computerized provider order entry (CPOE). These two health IT applications were chosen because of their fundamental importance with respect to current policy initiatives aimed at promoting health IT adoption and meaningful use by hospitals and other providers².

Methods

We examined two main dependent variables: 1) whether the hospital switched from one named commercial vendor or self-developed system to another named commercial vendor; and 2) whether the hospital switched from a named commercial vendor to no reported operational EMR or CPOE system.ⁱⁱ

Predictors of vendor switching and vendor dropping

We selected hospital characteristics which we expected a priori to have some bearing on the likelihood that a hospital which already had an EMR or CPOE in place would switch vendors or drop the technology during the 2003 through 2008 period. These characteristics included whether or not the hospital is an academic hospital, it’s ownership structure (for-profit, not-for-profit, or government), size, financial constraints, membership in a multi-hospital system, metropolitan location, and competition from other hospitals in the same market.

We expected that financial constraints, such as having a higher share of patients covered by a public payer (Medicaid), would hinder a hospital’s ability to incur the costs of changing its system once in place and thus have a negative relationship with vendor switching and a positive relationship with vendor dropping. Membership in a multi-hospital system may reflect greater access to resources to assist with upgrades by switching to different vendors’ products and we expected it to have a positive relationship with vendor switching and a negative relationship with vendor dropping. Being acquired by a multi-hospital system may also lead to vendor switching as a newly acquired hospital may be required to switch to the prevailing vendor in the acquiring system, but with access to additional resources through the acquiring system, we expected a reduction in the probability of vendor dropping. Hospital size may also reflect relative resources to some extent, and we expected it to have a relationship to vendor switching and dropping similar to system membership. On the other hand, size may also indicate larger fixed costs of installation, and if switching entails similarly large costs, may be negatively associated with switching. Not-for-profit ownership status has been argued to signal a commitment to quality of care above the profit-maximizing level,⁹ and therefore should be associated with a greater propensity to search for higher quality products and a lower likelihood of dropping the technology all together relative to for-profits. However, for-profits may also participate in vendor switching if initial poor investments in HIT prove to be more costly than anticipated and other products offer substantial cost-reductions. Market characteristics such as metropolitan location and the number of other hospitals in the same market (defined here as a Hospital Referral Region—HRR) indicate the degree of competitive pressure experienced by hospitals. We expected such competitive pressures to be positively associated with the propensity to switch vendors in response to initially sup-optimal choices or to desirable innovations produced by different vendors, and negatively associated with the propensity to drop the technology assuming that at least some health IT vendors sell products that are better than a paper-based system. We also examined the relationship of hospital-physician organizational arrangements, including joint ventures (e.g. Physician Hospital Organizations) and full integration (e.g. Integrated Salary Model) as possible shifters of the outcomes. We included an indicator variable for hospital participation in at least one such hospital-physician arrangement as reported in the AHA data. These arrangements change the relative bargaining strength between hospital administrators and the medical staff and may or may not mitigate the problem of physician resistance to health IT which has been identified in prior research^4,7–8 by encouraging greater participation of physicians in vendor selection and implementation.

Analysis

We first examined the overall landscape of commercial vendor and self-developed systems by examining the number of commercial vendors serving hospitals with EMR and CPOE systems as well as the overall number of vendor switches that took place from 2003 to 2008. We used 2003 as the start of our study period because this is the first year during which the HIMSS survey collected information on CPOE use. We then inspected the mean levels of hospital characteristics which we selected as possible predictors of vendor switching and dropping within the overall sample, and in the particular subsamples of vendor switchers and droppers. We subsequently used a multivariable probit regression model to estimate the probability of vendor switching and vendor dropping as functions of various hospital and hospital market characteristics described above. We clustered the standard errors at the market level in order to account for correlations in the choices of hospitals in the same market due to market-level disturbances. Our analysis is cross-sectional in nature using hospital characteristics reported in 2008 as predictors of whether or not a hospital with a live and operational EMR or CPOE system switched vendors during the preceding 5 years.

Results

Table 1 provides summary statistics on the numbers of hospitals which reported operational CPOE and EMR systems in 2003 and 2008 as well as breakdowns of the numbers reporting named commercial vendors, self-developed systems, and no reported vendor name.

Table 1.

Summary statistics for commercial CPOE and EMR markets in 2003 and 2008

	2003				2008
	Live and operational	Named commercial vendor	Self-developed	Unreported vendor	Live and operational	Named commercial vendor	Self-developed	Unreported vendor
CPOE	297 (100%)	221 (74%)	53 (17.8 %)	23 (7.7%)	1,280 (100%)	1,151 (98.6%)	119 (9.3%)	10 (1%)
EMR	1,839 (100%)	1,583 (86.1%)	215 (11.7%)	41 (2.2%)	2,397 (100%)	2,252 (94%)	135 (5.6%)	10 (0%)

Note: Row percentages of all “Live and operational” systems for each year are given in parentheses

Summary of EMR vendor market

While the hospital EMR adoption rate has steadily grown, the EMR vendor market has consolidated somewhat over the five year period we examined. In 2003, 46 named commercial vendors accounted for 1,583 EMR systems, while in 2008, the number of named commercial vendors shrank to 36, accounting for 2,252 systems. The share of the market occupied by self-developed EMRs shrank from 11.7 to 5.6 percent.

Summary of CPOE vendor market

In general, the reporting rate for vendor names was quite high in both 2003 and 2008. In 2003, of 297 respondents who indicated an operational CPOE system, 274 reported a named commercial vendor or a self-developed system, while in 2008 1,270 hospitals reported a named commercial vendor or self-developed CPOE system out of 1,280 reported operational CPOE systems. Fifteen commercial vendors were reported who accounted for 221 CPOE systems in 2003, while in 2008 twenty-six commercial vendors accounted for 1,151 systems. Thus, while the growing market for EMRs consolidated into fewer vendors, the number of vendors serving the commercial CPOE market increased. The share of the market occupied by self-developed CPOE systems shrank from 17.8 to 9.3 percent during this period.

Summary of commercial vendor switches

Many hospitals made changes in their HIT vendors during our study period. A substantial proportion of hospitals which had already invested in commercial clinical information systems by 2003 switched to different commercial vendors or turned off their systems all together by 2008. Of the 1,579 hospitals reporting a live and operational EMR from a named commercial vendor in 2003, twenty-eight percent switched to another named commercial vendor by 2008 and an additional eighteen percent reported no automated EMR in 2008. Of the 221 hospitals reporting a live and operational CPOE system from a named commercial vendor in 2003, eighteen percent switched to another named commercial vendor by 2008, and twenty-four percent reported no automated CPOE in 2008. The large volume of switching between vendors or away from the technology all together highlights some concerns about the immaturity of the commercial health IT products during the earlier part of last decade⁷. The incidence of vendor switching and vendor dropping seems surprisingly high but is consistent with other estimates produced by industry experts³.

Summary of hospital characteristics by descriptive statistics

Descriptive statistics for all hospitals which reported operational EMR or CPOE systems in 2003 as well as the subset of hospitals that switched vendors by 2008 and the subset that dropped the technology over this period are reported in Table 1. For our initial bivariate analysis we found that, relative to all hospitals with operational EMR or CPOE systems in 2003, larger proportions of vendor switchers were teaching hospitals, not-for-profits, had some formal organizational arrangement with physician staff, were in metropolitan areas and in markets with a few more competitors on average. Relatively smaller proportions of vendor switchers were for-profits and government owned. Vendor switchers had a slightly lower proportion of patients covered by Medicaid and were slightly larger in size.

Among those who reported dropping the technology, a smaller proportion were teaching hospitals, not-for-profits, in metropolitan areas, and possessed some formal organizational arrangement with physicians relative to all hospitals initially reporting operational EMR or CPOE systems in 2003. Those hospitals who dropped the technology consisted of larger proportions of for-profits, and members of multi-hospital systems relative to all hospitals with operational systems in 2003. Hospitals acquired by multi-hospital systems were represented in equal proportions between Technology droppers also were on average smaller, and had slightly fewer competitors in the same market.

Summary of hospitals characteristics from multivariable analysis

In our multivariable analysis (see Table 3), we found that for-profit status had a significant negative relationship with vendor switching (p < 0.1) and percentage of hospital discharges covered by Medicaid were significant negative predictors of switching behavior (p < 0.1). Contrary to our initial expectations, we did not find any significant relationship between our measures of competition (metropolitan location and number of other hospitals in the same market) and switching behavior, nor was there any significant relationship with respect to system membership, acquisition by a system, hospital size, and teaching status.

Table 3.

Health IT vendor switching 2003–2008 estimated by probit—average marginal effects reported

Variable	Switched vendors	Switched from vendor to no operational system
All organizational arrangements with physicians	−0.050 (0.031)	0.010 (0.021)
Hospital characteristics
Member of the council of teaching hospitals	0.004 (0.045)	−0.111*** (0.040)
For profit	−0.103* (0.053)	0.305*** (0.023)
Government owned	−0.020 (0.044)	0.069*** (0.027)
Total admissions	0.002 (0.002)	−0.005*** (0.002)
Member of a health care system	0.057 (0.036)	−0.047** (0.023)
Hospital acquired by a system	−0.052 (0.071)	0.042 (0.048)
Percent Medicaid discharges	−0.36*** (0.13)	0.130 (0.097)
Market characteristics
Metropolitan or Division CBSA	0.047 (0.038)	−0.091*** (0.022)
Number of other hospitals in the same market (HRR)	0.00081 (0.00078)	−0.00063* (0.00037)
N	1364	1688

Notes: Standard errors calculated by the delta method, and clustered by HRR market, given in parentheses.

^*Statistically significant at the 10% level.

^**Statistically significant at the 5% level.

^***Statistically significant at the 1% level.

In contrast, our multivariable analysis of vendor dropping by hospitals revealed a number of patterns consistent with our expectations. Teaching hospitals, larger hospitals in terms of admissions, hospitals in metropolitan areas and hospitals in multi-hospital systems were less likely to drop the technology (p < 0.01), as were hospitals in more competitive environments (p < 0.1). For-profits and non-federal government hospitals were more likely to drop the technology (p < 0.01) relative to not-for-profits, but the marginal effect of government ownership on dropping is much less than that of for-profit ownership.

Discussion

In this first study of the vendor switching and dropping practices of hospital health IT users, we have uncovered several notable findings. First, hospitals in significant numbers have made changes in the vendors with whom they do business. More than thirty percent of hospitals reporting an operational EMR or CPOE system in 2003 switched to a different commercial vendor or away from a self-developed system to a commercial system by 2008, while more than twenty percent reported no longer having an operational system by 2008. While providing some evidence of competition in the vendor market, these striking findings may also reflect the relative immaturity of the technology and the commercial health IT vendor market during the earlier part of last decade.

Furthermore, we have identified a number of hospital characteristics which are significant predictors of health IT vendor switching and dropping. Consistent with our hypotheses, we find that larger portions of patients covered by the public insurer Medicaid (an indicator of financial resource constraints) are associated with a significant reduction in the probability of switching vendors. This provides evidence consistent with the presence of important financial costs associated with vendor switching. Other indicators of access to resources for vendor switching such as multi-hospital system membership and size displayed no significant relationship with switching contrary to our expectations, however. For-profit ownership is associated with a significantly lower probability of vendor switching consistent with the hypothesis that not-for-profits may be more inclined to incur the costs of searching for higher quality systems over time. Competition does not appear to play a significant role in vendor switching according to our analysis, although our two indicators of a competitive environment (metropolitan location and number of competitors) had the predicted positive sign.

More significant patterns emerged from our analysis of technology dropping behavior. Teaching hospitals were significantly less likely to drop the technology, consistent with the conventional understanding that such hospitals are more sophisticated users and often employ highly customized home-grown systems. Ownership emerged as a significant determinant of technology dropping, as for-profits were considerably more likely than private not-for-profits to drop the technology over this period. This is consistent with the hypothesis that not-for-profits may be more inclined to reinvest net revenue in quality improvement as opposed to distributing it to stockholders⁹. Therefore they should be more likely to seek a strategy of selecting a new vendor in an attempt to re-optimize their technology selection rather than giving up on it all together. The significant negative relationship between our indicators of hospital market competition and vendor dropping provide evidence that use of this technology may offer a competitive advantage and serve as a useful competitive signal to patients and payers regarding quality in order to attract more business to the hospital. These competitive pressures may mitigate unobserved factors which contribute to dropping in less competitive markets. For example competition may stimulate greater effort during the implementation process which ensures more effective integration of the technology into clinician workflow.

Limitations

Our study has a number of limitations. First, because of its cross-sectional nature and because we do not use instrumental variables, we cannot assert a causal relationship between the various hospital characteristics and the vendor switching and technology dropping outcomes. Our goal with this initial study is to broadly identify the patterns of HIT vendor switching and dropping among hospitals and some key characteristics of hospitals which engage in these behaviors, thereby setting the stage for subsequent research which may establish causal patterns. Furthermore, the nature of the data which we use—self-reported survey data—may introduce reporting biases such as misreporting of health IT functionalities based on the social desirability of health IT use among hospitals. However, we sought to minimize the impact of such reporting errors by removing observations without reported vendor names in either period from our vendor-switching analysis. Our analysis therefore relies on more specific information about health IT applications contained in the HIMSS database than does much previous research on health IT applications using these data. Furthermore, the HIMSS Analytics survey provides the most comprehensive annual look at health IT use within hospitals over many years. While one can imagine how misreporting the use of specific health IT functions may be systematically related to hospital characteristics, it is more difficult to imagine how changes in named vendors would be systematically related to hospital characteristics.

Conclusion

The results of this initial analysis indicate substantial hospital health IT vendor switching and dropping and provide some evidence of competitive pressures at work in the commercial health IT vendor market. However, the high level of dropped health IT applications may also point to significant implementation problems in the industry during this period. The evidence uncovered in this descriptive analysis indicates that financial resource constraints may hinder the ability of hospitals to switch vendors, while various hospital characteristics including not-for-profit ownership, size, teaching status, system membership, and hospital market competition are associated with a lower propensity for failed health IT implementation. Further research is needed to establish causal relationships between some of the characteristics identified in this study as associated with vendor switching and dropping and the outcomes of interest.

Table 2.

Descriptive statistics (means) for EMR and CPOE vendor switching (2003–2008) and hospital characteristics in 2008

Variable	All hospitals with live and operational systems in 2003		Switched vendors		Switched from vendor to no operational system
	Mean	Standard deviation	Mean	Standard deviation	Mean	Standard deviation
Switched vendors	0.34	0.47
Switched from vendor to no operational system	0.23	0.42
Hospital characteristics
All organizational arrangements with physicians	0.51	0.50	0.51	0.50	0.44	0.50
Member of the council of teaching hospitals	0.12	0.32	0.15	0.36	0.03	0.16
For profit	0.17	0.37	0.06	0.23	0.41	0.49
Private not for profit	0.68	0.46	0.81	0.39	0.43	0.49
Government owned	0.15	0.36	0.13	0.34	0.16	0.37
Total admissions (10,000s)	1.1	1.1	1.3	1.2	0.61	0.64
Member of a health care system	0.66	0.47	0.69	0.46	0.68	0.47
Acquired by a health care system	0.054	0.23	0.05	0.22	0.05	0.23
Percent Medicaid discharges	0.18	0.11	0.17	0.10	0.181	0.098
Market characteristics
Metropolitan or Division CBSA	0.71	0.45	0.79	0.41	0.55	0.50
Number of other hospitals in the same market (HRR)	35.9	26.8	37.5	29.4	35.7	26.3
N	1648		439		391