Create an annotated bibliography
babwyomi8c6
contentserver1.pdf
Payer Mix and EHR Adoption in Hospitals Dong Yeong Shin, doctoral student. Department of Health Services Administration, University of Alabama at Birmingham; Nir Menachemi, PhD, professor, health care organization and policy. University of Alabama at Birmingham; Mark Diana, PhD, assistant professor. School of Public Health, Tulane University, New Orleans; Abby Swanson Kazley, PhD, associate professor. Medical University of South Carolina, Charleston; and Eric W. Ford, PhD, distinguished professor of healthcare, Bryan School of Business, University of North Carolina at Creensboro
E X E C U T I V E S U M M A R Y Payers are known to influence the adoption of health informarion technology (HIT) among hospitals. However, previous studies examining the relationship between payer mix and HIT have not focused specifically on electronic health record systems (EHRs). Using data from the Nationwide Inpatient Sample and the American Hos- pital Association Annual Survey, we examine how Medicare, Medicaid, commercial insurance, and managed care caseloads are associated with EHR adoption in hospi- tals. Overall, we found a weak relationship between payer mix and EHR adoption. Medicare and, separately, Medicaid volumes were not associated with EHR adoption. Furthermore, commercial insurance volume was not associated with EHR adoption; however, a hospital located in the third quartile of managed care caseloads had a decreased likelihood of EHR adoption. We did not find empirical evidence to sup- port the hypothesis that payer generosity and other indirect mechanisms influence EHR adoption in hospitals. The direct incentives embedded in the Health Informa- tion Technology for Economic and Clinical Health Act may have a positive influence on EHR adoption—especially for hospitals with high Medicare and/or Medicaid caseloads. However, it is still uncertain whether the available incentives will offset the barriers many hospitals face in achieving meaningfijl use of EHRs.
For more information about the concepts in this article, contact Dr. Menachemi at [email protected].
435
JOURNAL O F HEALTHCARE M A N A G E M E N T 5 7 : 6 N O V E M B E R / D E C E M B E R 2 0 1 2
I N T R O D U C T I O N Research has shown that payer mix, defined as the combination of third- party payers that makes up a hospital's book of business, can influence hospi- tals' strategic behaviors. Studies have found that higher percentages of Medic- aid (Cleverley and Harvey 1992; McKay and Deily 2005) or Medicare patients (Rosko 2001) are negatively associated with financial performance. Further- more, given that varying reimburse- ment rates are negotiated in the private insurance book of business, hospital revenue per admission has been demon- strated to predict operational efficiency (Dor and Fariey 1996; McKay and Deily 2005) and clinical performance (Clem- ent and Crazier 2001; Menachemi et al. 2007). Attempting to leverage the influ- ence that the public insurance programs have on hospitals, the federal govern- ment, through the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, aims to increase the adoption and "meaningful use" of electronic health record systems (EHRs) by providing incentives and pen- alties to hospitals through the Medicare and Medicaid programs (CMS 2010).
The HITECH Act, part of federal stimulus legislation, allocated billions of dollars in incentive payments to providers to facilitate the adoption and use of EHR technology (Blumenthal and Tavenner 2010). The incentive payments, made to hospitals that adopt, imple- ment, upgrade, or successfully demon- strate their meaningful use of certified EHRs, are available as of fiscal year 2011. Hospitals that do not achieve meaning- ful use by 2015 are potentially subject to Medicare and Medicaid payment
penalties of up to 2 percent in later years (CMS 2010).
The relationship between hospitals' EHR adoption rates and payer mix is not fully understood. However, several researchers have found that payer mix is correlated with health information tech- nology (HIT) adoption. Based on 2004 data, Furukawa and colleagues (2008) found that the adoption of computer- ized physician order entry systems, but not EHR systems, was associated with an increased rate of Medicare patients as a percentage of all discharges. Simi- larly, McCullough (2008) found that the adoption of pharmacy informa- tion systems was positively related to Medicare caseload. The McCullough study also found that the adoption of laboratory information systems and radiology information systems was negatively associated with higher levels of Medicaid caseloads. Lastly, evidence from Florida hospitals suggests that an increase in privately insured patients is positively associated with the number of HIT applications adopted (Menachemi et al. 2007). While the Florida study did not examine EHR adoption per se, differences in public insurance (e.g.. Medicare, Medicaid) as a percentage of a hospital's discharges were not correlated with overall HIT adoption in that study. Therefore, the relationship between Medicare and Medicaid programs and EHR adoption rates is unclear.
The purpose of this article is to examine the relationship between acute care hospitals' payer mix and their EHR adoption rates. The study uses national data and a series of statistical analyses including chi-square, logistic regression, and multinomial regression to explore
436
PAYER MIX AND EHR ADOPTION IN HOSPITALS
these relationships. A better understand- ing of how payer mix and EHR adoption are related will allow hospital leaders and public policjmiakers to capitalize on the legislative intent of the HITECH Act, which utilizes the relationship between payer mix (e.g.. Medicare and/or Med- icaid payer mix) and HIT adoption to directly motivate hospitals to use EHR technology. Our findings will also ben- efit those trying to better understand the impact that the meaningful use program may have on hospitals' EHR use. More- over, our study will help hospital deci- sion makers to better gauge their overall progress on EHR adoption relative to a national cohort of hospitals with similar payer mix combinations.
CONCEPTUAL FRAMEWORK Payer mix may influence EHR adop- tion in two ways. The first is related to the concept of payer generosity; the second is related to certain reimburse- ment mechanisms inherent to some payers. Payer generosity refers to the relative payments a given insurance company provides for a given proce- dure or diagnosis. In general, certain payers are believed to reimburse less generously than others (Friedman et al. 2004). Specifically, government pay- ers, such as Medicare and Medicaid, tend to provide lower reimbursement rates than do private payers, such as traditional indemnity insurance plans (Dobson, Davanzo, and Sen 2006). Civen the negative relationship between the proportion of Medicare and Med- icaid patients and hospital operating margin (e.g., Rosko 2001), the lower reimbursement rates of public insurance programs may not cover the entire cost
of patient care (AHA 2010). This short- fall could lead to decreased operating margin in hospitals, thus leaving them with less financial flexibility to consider major capital investments. Researchers have noted that upftont capital require- ments and high maintenance costs are the primary barriers to EHR adoption in acute care US hospitals (Jha et al. 2009). All other factors being equal, hospitals with higher proportions of Medicare and Medicaid patients are less likely to have the financial wherewithal to make the large capital investments neces- sary to buy an EHR system compared to facilities with higher percentages of private-pay patients. Thus, we would expect that a high amount of publicly insured patients is associated with a deaeased likelihood of EHR adoption in hospitals.
Second, reimbursement mechanisms may play a role in EHR adoption. Under capitation arrangements and other prospective payment contracts, hospitals are financially motivated to improve efficiencies and lower costs in order to maximize profits. Under capitated reimbursement conditions, common among health maintenance organiza- tions (HMOs), hospitals are paid a set amount for each enrolled person assigned regardless of the number or type of services provided to the person in a given time period (Mello, Steams, and Norton 2002; Miller and Luft 2002). Many proponents of EHRs have claimed that EHRs will increase organizational efficiency while reducing duplication of eftort (Brailer 2005). To the extent that EHR adoption is viewed as a strategy for improving efficiencies and lower- ing costs (Ali et al. 2005; Boger 2003;
437
JOURNAL OF HEALTHCARE M A N A G E M E N T 5 7 : 6 N O V E M B E R / D E G E M B E R 2 0 1 2
Carrido et al. 2004), it may be pursued at higher rates by hospitals with a rela- tively higher number of HMQ patients.
METHODS We used a cross-sectional design with secondary data and with the acute care hospital as the unit of analysis. The analysis combines data ftom HCUP's (Healthcare Cost and Utilization Proj- ect) Nationwide Inpatient Sample (NIS) for 2007, the 2008 American Hospi- tal Association (AHA) Annual Survey database, and the 2007 Medicare Cost Reports. We drew hospital discharge data from the 2007 NIS, while the data indicating EHR adoption status and organizational characteristics of hos- pitals were drawn from the 2008 AHA database. In addition, because of the relatively small sample size we obtained from the NIS, we performed a paral- lel analysis using the limited payer mix variables, but with a much larger sample size, from the AHA Annual Survey. Doing so allows us to robustly examine the relationship between payer mix and EHR adoption.
The AHA database contains organi- zational information on US hospitals and on EHR systems use, which the AHA added to its annual assessment beginning in 2007 (Jha et al. 2009). The dependent variable is a categorical measure representing (a) fully imple- mented EHR, (b) partially implemented EHR, and (c) no EHR implemented in response to the following AHA survey question: "Does you hospital have an electronic health record?" Data mea- suring payer mix were extracted from the NIS database, which is the larg- est all-payer inpatient database in the
United States and contains all discharge data ftom approximately a 20 percent stratified sample of community hos- pitals. The NIS database is frequently employed by health services researchers interested in hospital management and related issues (Boxer et al. 2003; LaPar et al. 2010; Russell et al. 2006).
To operationalize payer mix, we cal- culated the percentage of discharges for each hospital that were covered by each of the following primary payers: (1) Medicare, (2) Medicaid, (3) commercial indemnity insurance, (4) managed care organizations, and (5) all other payers. Qf the 646 hospitals in the 2007 NIS data set that include AHA identifiers, 392 hospitals (60.7 percent) included all the information necessary to con- struct our main independent variables. We compared the excluded hospitals to our sample for validity purposes.
Given the loss in sample size that occurred as a result of merging with the NIS data, we also examined the rela- tionship between Medicare caseloads and Medicaid caseloads using the AHA Annual Survey data. The AHA sample is much larger than the NIS but is limited to information on only two payers' caseloads (i.e.. Medicare and Medicaid). By examining the relationships of inter- est with the NIS (comprehensive payer data limited to a small sample) and the AHA survey (limited payer data on a comprehensive sample size), we believe we were able to robustly examine the relationship between EHR adoption and payer mix.
Variables To compute each payer mix variable in the NIS, we conducted a series of
438
PAYER MIX AND EHR ADOPTION IN HOSPITALS
aggregations and calculations. First, we aggregated from tbe patient level to the hospital level the total number of discharges paid in each category of primary payers in each hospital ftom the Inpatient Core Files. Next, we merged this aggregated data set with the Hospital Weights Files to compute the proportion of discharges paid for each payer type for each hospital (i.e., the total number of discharges paid by each payer divided by the total number of discharges in a hospital). Each payer mix variable in the data set was con- verted into quartiles to facilitate ease of interpretation. The result is four payer mix variables, one for each primary payer category (i.e.. Medicare, Medicaid, commercial, and managed care). Lastly, hospitals were assigned a value from 1 to 4 depending on the quartile in which they resided for the distribution of each payer mix variable.
The AHA Annual Survey data include measures representing the percentage of each hospital's discharges that are Medicare and, separately, Med- icaid. These variables were each broken into quartiles to align them with the NIS data we prepared. Because vari- ous organizational factors are associ- ated with HIT adoption (Furukawa et al. 2008; Hikmet et al. 2008; Wang et al. 2005), each analysis we performed included control variables for the fol- lowing hospital characteristics: bed size (measured as the natural log of the number of staff beds), system affilia- tion (yes or no), teaching status, geo- graphic location (urban or rural), and tax status (for-profit or not-for-profit). These variables were derived from the AHA data. In addition, we controlled for
case mix (defined as the average severity of patients treated at a given hospital) using the case mix index ftom the Medi- care Cost Reports.
Oata Analyses All variables were examined for their distribution, suitability for analysis, and the existence of any potential data anomalies based on descriptive statis- tics. Next, we conducted chi-square anal- yses and independent-samples t-tests or analysis of variance (as appropriate) to identify any organizational differences between included and excluded hospi- tals, and we explored the univariate rela- tionships between full EHR adoption and each variable measuring payer mix. Finally, using the NIS data, we exam- ined the relationship between Medicare, Medicaid, commercial insurance, and managed care mix and EHR adoption while controlling for hospital character- istics and "other" types of discharges in a logistic regression model. We used the AHA data to examine a similar model with the obvious exclusion of the payer mix variables that do not appear in this data set (i.e., commercial insurance and managed care). In both the NIS and AHA models, we present EHR adoption as a binary variable (full EHR and par- tial EHR versus no EHR). In addition, we present the results of the NIS analysis specified as a multinomial regression that takes advantage of the categorical nature of the EHR variable. The results we present include adjusted odds ratios (ORs) and 95 percent confidence inter- vals for the logistic models, and beta coefficients for the multinomial regres- sion model. Multiyariate results are flagged for significance at the p < 0.05,
439
JOURNAL OF HEALTHGARE MANAGEMENT 57:6 NOVEMBER/DECEMBER 2012
p < 0.01, and p < 0.001 levels, respec- tively. In all regression models, we also controlled for the nested nature of hospitals clustered within states using Huber-White adjustments (Wooldridge 2003, 255) to our standard errors using the "clustered robust" command in STATA version 11.1.
R E S U L T S Organizational characteristics of included and excluded hospitals are displayed in Exhibit 1. Overall, NIS study hospitals (n = 392) had an aver- age Medicare caseload of 47.9 percent, Medicaid caseload of 20.1 percent, commercial insurance caseload of 14.2 percent, and managed care caseload of 16.5 percent. Excluded NIS hospitals (n = 220) and AHA study hospitals (n = 4,095) did not differ with respect to these payer mix variables (see Exhibit 1). Mean bed size for the NIS study sample was 205, significantly different than the excluded NIS hospitals (mean 143.2) and AHA study sample (mean 163.7; p < 0.001). Overall, the NIS sample had a higher proportion of urban hospitals than excluded NIS hospitals and the AHA hospitals (62.4 percent vs. 41.6 vs. 49.9; p < 0.001), but the three groups did not differ with respect to tax status, system affiliation, teaching status, or case mix index (see Exhibit 1).
In univariate analysis of the NIS data. Medicare discharges as a percent- age of a hospital's payer mix was associ- ated with EHR adoprion (see Exhibit 2). Specifically, hospitals in the lowest quartile of Medicare discharges as a percentage of all discharges were more likely to have implemented a full EHR (19.1 percent vs. 3.1 percent; p < 0.001)
or a partial EHR (55.9 percent vs. 42.2 percent; p < 0.001). No differences were observed in EHR adoption quartiles by Medicaid, commercial insurance, or managed care caseload. Increased hos- pital size, urban location, not-for-profit tax status, and teaching hospital status were positively associated with EHR adoption in univariate analyses (see Exhibit 2).
In multivariate analysis of the NIS sample and, separately, the AHA sample, controlling for payer mix, geographic location, tax status, bed size, system affiliation, teaching status, and case mix, virtually none of the payer mix variables were related to EHR adoption (see Exhibit 3). The only exception was that hospitals in the third quartile for managed care discharges (NIS sample only) were significantly less likely than those in the bottom quartile to report having an EHR (OR = 0.18, p = 0.026). In the AHA model, several control vari- ables, including geographic location, tax status, bed size, and system affiliation, were significantly associated with EHR adoption (see Exhibit 3).
Lastly, in the multinomial regression that examines the EHR adoption vari- able with three categories using the NIS sample, most of the payer mix variables were still not related to EHR adoption (see Exhibit 4). The only exception was the third quartile of managed care dis- charges, where hospitals in this category were again less likely to have adopted an EHR system (no EHR vs. fijll EHR, beta = 2.289, p = 0.038).
D I S C U S S I O N While researchers have found that payer mix is associated with HIT adoption in
440
PAYER MIX AND EHR ADOPTION IN HOSPITALS
E X H I B I T 1 Organizational Characteristics of Included and Excluded Hospital Samples
Included NIS Hospitals
(NIS 2007) n = 392
Excluded NIS Hospitals
(NIS 2007) n = 220
AHA Hospitals Used in
Sensitivity Analyses
(AHA 2007) n = 4,095 p-Value
Mean percent Medicare discharges (sd)
Mean percent Medicaid discharges (sd)
Mean-percent commercial insurance discharges (sd)
Mean percent managed care
discharges (sd) Geographic location
Rural
Urban
Tax status For-profit
Not-for-profit
System affiliation Yes
No
Teaching hospital Yes
No
Mean bed size (sd)
47.9(16.6) 50.6(17.5) 49.9(18.4) 0.086
20.1(15.7) 19.0(15.4) 18.8(15.7) 0.293
14.2 (12.2)
16.5 (15.4)
N/A
N/A
147 (37.6%) 128 (58.4%) 2021 (50.1%) 0.000 244 (62.4%) 91 (41.6%) 2015 (49.9%)
56(14.3%) 30(13.6%) 706(17.2%) 0.142 336 (85.7%) 190 (86.4%) 3389 (82.8%)
222 (56.6%) 130 (59.1%) 2159 (52.7%) 0.072 170 (43.4%) 90 (40.9%) 1936 (47.3%)
33 (8.4%) 10 (4.5%) 233 (5.7%) 0.062 359 (91.6%) 210 (95.5%) 3862 (94.3%)
205.0(211.9) 143.2(150.6) 163.7(182.1) 0.000
Unadjusted case mix index (sd) 1.40(0.25) 1.35(0.26) 1.37(0.28) 0.129
Note: Numbers may not add to 100% due to rounding. N/A is not applicable because information about commercial insurance
and managed care discharges is not provided for excluded hospitals in the NIS 2007 data set or for all hospitals in the AHA
2007 data set.
441
JOURNAL OF HEALTHCARE MANAGEMENT 57:6 NOVEMBER/DECEMBER 2012
E X H I B I T 2 Univariate Reiationship Between Payer Mix and EHR Adoption in Hospitals Using the Nationai Inpatient Sample (/; = 392)
Medicare discharges 1st quartile (<33% of discharge) 2nd quartile (34-42% of discharge) 3rd quartile (43-53% of discharge) 4th quartile (54+% of discharge)
Medicaid discharges 1st quartile (<8% of discharge) 2nd quartile (9-15% of discharge) 3rd quartile (16-21% of discharge) 4th quartile (22+% of discharge)
Commercial insurance discharges 1st quartile (<4% of discharge) 2nd quartile (5-12% of discharge) 3rd quartile (13-21% of discharge) 4th quartile (22+% of discharge)
Managed care discharges 1st quartile (<2% of discharge) 2nd quartile (3-14% of discharge) 3rd quartile (15-26% of discharge) 4th quartile (27+% of discharge)
Bed size Small (<125 beds) Medium (126-399 beds) Large (400+ beds)
Geographic location Rural Urban
Tax status For-profit Not-for-profit
System affiliation Yes
No
Teaching hospital Yes N o
Partial EHR Adoption (%)
55.9 57.7 63.2 42.2
44.4 62.9 58.3 54.3
64.3 43.6 53.4 62.0
52.5 50.8 62.0 54.1
49.6 61.4 55.6
49.1 58.9
36.4 57.7
55.9 54.5
58.6 54.9
Fuli EHR Adoption (%)
19.1 15.4 15.8 3.1
11.1 15.7
9.7
17.3
12.5 14.1
15.1 12.7
12.5 13.8
8.9
21.3
6.3
14.9 31.1
5.7
18.3
12.1
13.8
14.5 12.7
34.5 11.3
p-Vaiue
<0.001
0.115
0.185
0.346
<0.001
<0.001
0.024
0.806
<0.001
4 4 2
PAYER MIX AND EHR ADOPTION IN HOSPITALS
E X H I B I T 3 Multivariate Relationship Between Payer Mix and EHR Adoption in Hospitals
Payer Mix: Independent Variables
Medicare caseload
Quartile 1 (low volume)
Quartile 2
Quartile 3 Quartile 4 (high volume)
Medicaid caseload
Quartile 1 (low volume) Quartile 2
Quartile 3 Quartile 4 (high volume)
Commercial insurance caseload
Quartile 1 (low volume)
Quartile 2 Quartile 3
Quartile 4 (high volume)
Managed care insurance caseload
Quartile 1 (low volume)
Quartile 2 Quartile 3
Quartile 4 (high volume)
Other payers caseload''
Control Variables
Geographic location: Urban
Tax status: For-profit
System affiliation: Yes
Teaching hospital: Yes
Natural log of bed size
Unadjusted case mix index
EHR Adoption
Results Using the NIS 2007 Data
(n = 392)
1.00
0.84 (0.30-2.37)
1.34 (0.36-4.98) 0.19 (0.02-2.20)
1.00 1.00 (0.27-3.71)
0.61 (0.14-2.64)
1.39 (0.33-5.88)
1.00
1.68 (0.50-5.69) 1.71 (0.44-6.70)
1.13(0.19-6.73)
1.00
0.32 (0.08-1.28) 0.18(0.04-0.82)*
0.34 (0.06-2.04)
0.00 (0.00-2.32)
2.36 (0.61-9.14)
1.33 (0.36-5.00) 1.03 (0.45-2.37)
1.30(0.38-4.49) 1.70(0.82-3.55)
1.71 (0.24-11.96)
Odds Ratio (95% CI)
Results Using the AHA 2007 Data
(/J = 4,707)
1.00
1.04 (0.76-1.44)
1.12(0.80-1.56)
1.01 (0.69-1.46)
1.00
1.34(0.99-1.79)
1.06 (0.76-1.48)
0.82 (0.56-1.22)
N/A»
N/A"
N/A''
1.54(1.15-2.08)**
0.54(0.38-0.75)***
1.65(1.31-2.07)***
1.40 (0.98-2.01)
1.20(1.01-1.43)*
1.18(0.72-1.92)
Note: Logistic regression used to compute adjusted odds ratio controlling for variables listed in table. CI = confidence interval.
N/A = not applicable.
"Not applicable because the AHA data set does not include information on private payers' caseload.
'Measured continuously.
•p < 0.05, "p< 0.01, • • ' p < 0.001.
443
JOURNAL OF HEALTHGARE MANAGEMENT 57:6 NOVEMBER/DECEMBER 2012
E X H I B I T 4 Relationship Between Payer Mix and EHR Adoption in Hospitais: Results ot a Multinomial Regression Whereby "Full EHR Adoption" Is the Reference Category
EHR Adoption (coef.)
Payer Mix: Independent Variables
Medicare caseload Quartile 1 (<33% of discharge) Quartile 2 (34-42% of discharge) Quartile 3 (43-53% of discharge)
Quartile 4 (54+% of discharge)
Medicaid caseload Quartile 1 (<8% of discharge) Quartile 2 (9-15% of discharge) Quartile 3 (16-21% of discharge)
Quartile 4 (22+% of discharge)
Commercial insurance caseload Quartile 1 (<4% of discharge) Quartile 2 (5-12% of discharge)
Quartile 3 (13-21% of discharge) Quartile 4 (22+% of discharge)
Managed care insurance caseload Quartile 1 (<2% of discharge) Quartile 2 (3-14% of discharge)
Quartile 3 (15-26% of discharge) Quartile 4 (27+% of discharge)
Other payers caseload""
Control Variables
Geographic location: Urban Tax status: For-profit
Systeni affiliation: Yes Teaching hospital: Yes Natural log of bed size Unadjusted case mix index
'Measured continuously.
*p<0.05, ••p<0.01, •••p<0.001.
Partial
-0.278 -0.613
1.038
0.319 0.827
-0.575
-0.818 -0.715
-0.458
0.834
1.440 0.586
5.025
-0.829 • -0.617
-0.180
0.005 -0.346
-1.010
No
0.378
-0.925
2.210
-0.879 0.424
-0.390
0.232
0.160 0.416
2.036
2.289* 1.970
10.580*
-1.129
0.278 0.552
-1.282
-0.922*
0.268
4 4 4
PAYER MIX AND EHR ADOPTION IN HOSPITALS
hospitals, less attention has been paid to the association of payer mix and hos- pital EHR adoption. Given the federal government's provision of financial incentives to promote the adoption and meaningful use of EHR through the HITECH Act, we suggest that poli- cymakers and hospital decision makers need a better understanding about the influence of payer mix on EHR adoption to fully realize the benefit of the legisla- tion. In the absence of a previous analy- sis utilizing national data, we examine how the proportion of discharges paid by each payer is associated with EHR adoption by hospitals.
The findings in our current study suggest a weak relationship between payer mix and hospital EHR adoption. Even though we found that certain increases in Medicare caseloads were generally negatively associated with EHR adoption in univariate analysis, overall, these differences disappeared in our adjusted models, furthermore, most of the other payer mix variables were not associated with our outcome measure in any systematic way. These findings are in conflict with the existing studies that have found a relationship between individual HI T applications and payer mix (Furukawa et al. 20Ó8). One possible explanation for this result may be that the adoption of an EHR system is differentially influenced by payer mix relative to the adoption of other types of HIT applications. Full EHR adoption may be the final phase of HIT adoption in hospitals. Thus, it is possible that payer mix influences the adoption of infrastructure-related HIT applications such as pharmacy, laboratory, and radiology systems, but
not complete EHR systems. If so, it is possible that the infrastructure-related HIT applications are less sensitive to resource availability. Perhaps the financial flexibility arising from serving patients from relatively generous pay- ers may only apply to certain, not all, HIT decisions. This possibility would explain why we found a weak relation- ship between payer mix and hospi- tal EHR adoption, whereas previous studies found a significant relationship between payer mix and the adoption of pharmacy and laboratory information systems (McCullough 2008) as well as other clinical, administrative, or stra- tegic HIT applications (Furukawa et al. 2008; Menachemi et al. 2007).
We expected that hospitals with higher proportions of public payer caseloads would be less likely to have an EHR system. We found no evidence to support this hypothesis in either the model using the NIS data or the model using the AHA payer mix data. Overall, this lack of evidence suggests the pos- sibility that indirect incentives generated by payer generosity (as discussed earlier) may not be a strong factor influencing EHR adoption in hospitals. Under the HITECH program's direct incentives for EHR adoption, hospitals with com- parably larger Medicare and Medicaid patient caseloads will be compensated proportionately higher for achieving meaningful use of EHRs. Thus, the HITECH Act has the potential, with some provisos, to assist in motivating hospitals that disproportionately serve Medicare and/or Medicaid patients to adopt an EHR and achieve meaningful use, thus fulfilling the intended objec- tive of the legislation.
445
JOURNAL OE HEALTHCARE M A N A G E M E N T 5 7 : 6 N O V E M B E R / D E C E M B E R 2 0 1 2
The first consideration for hospitals with high Medicare and/or Medicaid case mix is the level of HITECH pro- gram reward versus the total cost of ownership for an EHR. If the program's payback exceeds the financial cost of the EHR, then adopting a system is a rational choice. However, the program is designed to offset the EHR's purchase price and does not take into account the expenses associated with workflow redesign, temporary losses in produc- tivity, and so forth (i.e., the total cost of ownership). Such analysis also may not consider potential quality gains associated with EHR use. Therefore, the decision to adopt an EHR is more complicated than merely to pursue the HITECH rewards. Considering discus- sions in a recent study that focused on ambulatory EHR adoption (Song et al. 2011), hospitals could benefit from con- sidering the financial and nonfinancial benefits of EHRs when calculating the expected cost-benefit ratio of pursing the incentive payments.
A second concern regarding hospi- tals with high Medicare and/or Medicaid caseloads relates to unintended conse- quences fi'om the policymaker's perspec- tive. The EHR incentive program may induce already undercapitalized hospi- tals to adopt a more leveraged position and face an increased risk of failure as they strive to meet the goals. Even if facilities choose not to invest in an EHR, the penalty phase of the meaning- ful use program may further cut into already faltering budgets. If these facili- ties are safety-net hospitals, the program may have the unintended consequence of hastening the failure of some of the very hospitals it is intended to assist and
the concomitant impact on the most vulnerable populations in the United States. Of particular concern is the pos- sibility that small, rural facilities will be adversely affected.
Our study also found that hospitals in the third highest category (out of four) as measured on the basis of managed care insurance caseloads were less likely than those with the lowest managed care caseloads to adopt EHR. While this find- ing is inconsistent with our hypotheses, it does not represent a systematic rela- tionship between managed care case- loads and EHR adoption in hospitals. Thus, more research is needed to fiirther understand this finding. On the other hand, several of our control variables were associated with EHR adoption in ways consistent with expectations based on previous HIT research. For example, in our study, urban hospitals (Burke et al. 2002; Furukawa et al. 2008), non- profit hospitals (Menachemi et al. 2007), system-affiliated hospitals (Wang et al. 2005), and larger hospitals (Burke et al. 2002; Furukawa et al. 2008) were all more likely to have adopted EHRs.
The findings of this study offer practical implications for hospital decision makers and raise an impor- tant issue regarding narional efforts imbedded in the HITECH Act. If payer generosity, or the indirect influence of payers, does not spur EHR adoption in hospitals, then the direct incentives in HITECH may represent the needed policy lever to influence EHR adop- tion. The important question becomes whether the direct incentives in the EHR adoption program will be motivation enough to overcome the resistance from some hospitals to begin the process
446
PAYER MIX AND EHR ADOPTION IN HOSPITALS
of achieving meaningful use. Future research is needed to determine the full impact of the HITECH Act. Such research can utilize either the NIS or the AHA data to examine how Medicare and Medicaid caseloads are associated with EHR adoption after the HITECH Act has had more time to take hold. In the meantime, hospital decision makers should be aware that while the financial flexibility aftorded by catering to more privately insured patients may enable the adoption of certain infrastructure HIT applications, such changes may not enable the adoption of EHRs.
The current study has several strengths. First, our topic is concerned with an important contemporary issue and makes use of a relatively large sample of hospitals potentially repre- sentative of US community hospitals. Furthermore, we make use of multiple data sources, which may help overcome common methods bias that negatively impacts the internal validity of studies using data extracted ftom a single source (Iezzoni 2003). Despite these strengths, our analysis is limited in some aspects. First, given the cross-sectional obser- vational nature of our analyses, we are unable to infer any causal relation- ship between payer mix and hospital EHR adoption. Therefore, our findings should be interpreted as associations only. Second, our sample size was inevi- tably decreased in the process of merg- ing data sets and operationalizing our dependent and independent variables. We tried to overcome this limitation by running parallel analyses using a larger sample from the AHA Annual Survey, which contains less detailed payer infor- mation. The results of both analyses
were similar. Lastly, our work is limited by the possibility of data entry and cod- ing errors that can occur in secondary databases.
R E F E R E N C E S Ali, N. A., H. S. Mekhjian, P L. Kuehn, T. D.
Bentley, R. Kumar, A. K. Ferketich, and S. P. Hofftnann. 2005. "Specificity of Computerized Physician Order Entry Has a Significant Effect on the Efficiency of Workflow for Critically 111 Patients." Criti- cal Care Medicine 33 (1): 110-14.
American Hospital Association (AHA). 2010. "Underpayment by Medicare and Medicaid Fart Sheet." Accessed Janu- ary 7, 2012. www.aha.org/content/00- 10/ 10medunderpayment.pdf
Blumenthal, D., and M. Tavenner. 2010. "The 'Meaningful Use' Regulation for Elertronic Health Records." New England Journal of Medicine 365 [6): 501-4.
Boger, E. 2003. "ElectronicTracking Board Reduces ED Patient Length of Stay at Indi- ana Hospital." Journal of Emergency Nursing 2 9 ( 1 ) : 3 9 - 4 3 .
Boxer, L K., J. B. Dimick, R. M. Wainess, J. A. Cowan, P K. Henke, I. C. Stanley, and C. R. Upchurch Jr. 2003. "Payer Status Is Related to Differences in Access and Outcomes of Abdorriinal Aordc Aneurysm Repair in the United States." Surgery 134 (2): 142-45.
Brailer, D. I. 2005. "Interoperability: The Key to the Future Health Care System." Health Affairs (Millwood) Web Exclusives, W5-19-W5-21.
Burke, D. E., B. B. Wang, T. T. Wan, and M. L. Diana. 2002. "Exploring Hospitals' Adop- tion of Information Technology." Journal of Medical Systems 26 (4): 349-55.
Centers for Medicare & Medicaid Services (CMS). 2010. "CMS Finalizes Require- ments for the Medicare Electronic Health Record (EHR) Incentive Program." U.S. Department of Health and Human Services. Published July 16. wv«v.cms.gov/Apps/ Media/Press/Fartsheet.Asp?Counter=3792 &Intnumperpage= 10&Checkdate=&Check key=&Srchtype=l&Numdays=3500&Srcho pt=0&Srchdata=&Keywordtype=All&Chkn ewstype=6&Intpage=&Showall=&Pyear=& Year=&Desc=&Cboorder=Date.
447
louRNAL OF HEALTHCARE MANAGEMENT 5 7 : 6 N O V E M B E R / D E C E M B E R 2 0 1 2
Clemem, J. P., and K. L Grazier. 2001. "HMO Penetration: Has It Hurt Public Hospi- tals?" Journal of Health Care Finance 28 (1): 2 5 - 3 8 .
Cleverley, W. O., and R. K. Harvey. 1992. "Competitive Strategy for Successful Hospital Management." Hospital & Health Services Administration 37 (1): 53-69.
bobson. A., I. Davanzo, and N. Sen. 2006. "The Cost-Shift Payment 'Hydraulic': Foundation, History, and Implications." Health Affairs (Millwood) 25 (1): 2 2 - 3 3 .
Dor, A., and D. E. Farley. 1996. "Payment Source and the Cost of Hospital Care: Evi- dence from a Multiprodua Cost Function with Multiple Payers." Journal of Health Economics 15 (1): 1-21.
Friedman, B., N. Sood, K. Engstrom, and D. McKenzie. 2004. "New Evidence on Hos- pital Profitability by Payer Croup and the Effects of Payer Cenerosity." International Journal of Health Care Finance and Econom- ics 4 (3): 231-46.
Furukawa, M. E, T. S. Raghu, T. I. Spaulding, and A. Vinze. 2008. "Adoption of Health Information Technology for Medication Safety in U.S. Hospitals, 2006." Health Affairs (Millwood) 27 (3): 865-75.
Carrido, T, B. Raymond, L. Jamieson, L. Liang, and A. Wiesenthal. 2004. "Making the Business Case for Hospital Information Systems—A Kaiser Permanente Investment Decision." Journal of Health Care Finance 31 (2): 16-25.
Hikmet, N., A. Bhattacherjee, N. Menachemi, V. O. Kayhan, and R. C. Brook. 2008. "The Role of Organizational Fartors in the Adoption of Healthcare Information Tech- nology in Florida Hospitals." Health Care Management Science 11 (1): 1-9.
Iezzoni, L. I. 2003. Risk Adjustment for Measur- ing Health Care Outcomes, 3rd ed. Chicago: Health Administration Press.
Jha, A. K., C. M. Desroches, E. C. Campbell, K. Donelan, S. R. Rao, T. C. Ferris, A. Shields, S. Rosenbaum, and D. Blumenthal. 2009. "Use of Electronic Health Records in U.S. Hospitals." New England Journal of Medi- cine 360 (16): 1628-38.
LaPar, D. J., C. M. Bhamidipati, C. M. Mery, C. I. Stukenborg, D. R. Iones, B. D. Schirmer,
I. L. Kron, and C. Ailawadi. 2010. "Primary Payer Status Affeas Mortality for Major Surgical Operations." Annals of Surgery 252 (3): 544-50; discussion 550-51.
McCullough, I. S. 2008. "The Adoption of Hospital Information Systems." Health Economics 17 (5): 649-64.
McKay, N. L, and M. E. Deily. 2005. "Compar- ing High- and Low-Performing Hospitals Using Risk-Adjusted Excess Mortality and Cost Inefficiency." Health Care Management Review 30 (4): 347-60.
Mello, M. M., S. C. Stearns, and E. C. Norton. 2002. "Do Medicare HMOs Still Reduce Health Services Use After Controlling for Selection Bias?" Health Economics 11 (4): 323-40.
Menachemi, N., N. Hikmet, A. Bhattacherjee, A. Chukmaitov, and R. C. Brooks. 2007. "The Effect of Payer Mix on the Adoption of Information Technologies by Hospi- tals." Health Care Management Review 32 (2): 102-10.
Miller, R. H., and H. S. Luft. 2002. "HMO Plan Performance Update: An Analysis of the Literature, 1997-2001." Health Affairs (Millwood) 21 (4): 63-86.
Rosko, M. D. 2001. "Factors Associated with the Provision of Uncompensated Care in Pennsylvania Hospitals." Journal of Health and Human Services Administration 24 (3): 352-79.
Russell, M. W., A. V. Joshi, P J. Neumann, L. Boulanger, and J. Menzin. 2006. "Predic- tors of Hospital Length of Stay and Cost in Patients with Intracerebral Hemorrhage." Neurology 67 (7): 1279-81.
Song, P. H., A. S. McAleamey, I. Robbins, and I. S. McCullough. 2011. "Exploring the Business Case for Ambulatory Electronic Health Record System Adoption." Journal of Healthcare Management 56 (3): 169-80; discussion 181-82.
Wang, B. B., T. T. H. Wan, D. E. Burke, C. I. Bazzoli, and B. Y. J. Lin. 2005. "Faaors Influencing Health Information System Adoption in American Hospitals." Health Care Management Review 30 (1): 44.
Wooldridge, J. M. 2003. Introductory Economet- rics: A Modem Approach, 2nd ed. Cincin- nati, OH: South-Western.
448
Copyright of Journal of Healthcare Management is the property of American College of Healthcare Executives
and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.
