Differential Effects of the Two Types of Information Systems: A HospitalBased Study

Journal of Management Information Systems Summer 2009, Vol 26, No 1, pp 297–316 © 2009 M E Sharpe, Inc 0742–1222 2009 9 50 + 0 00 DOI 10 2753MIS0742 1222260111 Differential Effects of the Two Ty.Nirup M. MeNoN is an associate professor of Information Systems at george Mason university and a Visiting professor at the Instituto de Empresa Business School, Madrid, Spain. He has previously held teaching appointments at the university of Texas at Dallas and Texas Tech university. His research interests include economics of information systems, business value of information technology, enterprise systems, information privacy, and information security. Dr. Menon has published in several leading information systems journals, including Management Science, Information Systems Research, and Journal of Management Information Sy

Differential Effects of the Two Types of Information Systems: A Hospital-Based Study Nirup M Menon, Ulku Yaylacicegi, and Asunur Cezar Nirup M Menon is an Associate Professor of Information Systems at George Mason University and a Visiting Professor at the Instituto de Empresa Business School, Madrid, Spain He has previously held teaching appointments at the University of Texas at Dallas and Texas Tech University His research interests include economics of information systems, business value of information technology, enterprise systems, information privacy, and information security Dr Menon has published in several leading information systems journals, including Management Science, Information Systems Research, and Journal of Management Information Systems Ulku Yaylacicegi is an Assistant Professor of Information Systems at Cameron Business School, University of North Carolina at Wilmington She obtained her Ph.D from the University of Texas at Dallas Her research interests span information communications technologies, telecommunications policy, information security, IT productivity, health-care IT, quality management, and innovative education Her research has appeared in, or is forthcoming in, Technology in Society, Industrial Management and Data Systems, Journal of Information Systems Applied Research, and International Journal of Innovation and Learning Asunur Cezar is a Ph.D candidate in the School of Management at the University of Texas at Dallas She has received degrees in computer engineering, financial engineering, and supply-chain management Her research focuses on the economics of privacy and security, and economics of information systems Having more than eight years of experience in IT industry, she has participated in the design, development, and launch of leading-edge technology solutions Abstract: A new empirical model for the production function of the hospital incorporating two types of information systems (IS) is developed One type of IS is representative of information technology (IT) used in primary, clinical, value-chain activities, and the other is representative of the IT used in support (administrative) value-chain activities The model innovation is that it accommodates up to a seven-year lag for each type of IS The output variables for the production function are hospital output and medical labor productivity Using data spanning from 1979 to 2006 from several hospitals, it was found that clinical IS improve hospital output in the short run (of two years) Administrative IS were found to be negatively associated with organizational performance in the short run, but positively associated with these performance measures over the long run (over four years) These results highlight the importance of timing IT investments and the sequencing chosen for the implementation of IS presenting various value-chain activities, and the resulting pattern of business value over time Journal of Management Information Systems / Summer 2009, Vol 26, No 1, pp 297–316 © 2009 M.E Sharpe, Inc 0742–1222 / 2009 $9.50 + 0.00 DOI 10.2753/MIS0742-1222260111 298 Menon, Yaylacicegi, and Cezar Differential lag length of the types of IS is to be considered in estimating the rate of return of new IT projects Key words model and phrases : health-care informatics, IT productivity, value-chain Despite the downturn in the information technology (IT) industry since the late 1990s, investments in IT have not decreased [28] IT adoption in health care has been increasing rapidly U.S health-care costs, as a percentage of gross domestic product (GDP), continue to grow, but hospitals have been laggard in adopting IT [22] The rate of growth of expenditures in IT by health-care organizations is gradually catching up with the rate of growth of health-care costs [44] Industry analysts report that IT spending at hospitals was $15.4 billion out of the $23.6 billion that the U.S economy spent on health-care IT in 2003 [18, 33] Current forecasts indicate that health-care IT spending will surpass $39.5 billion by 2010 [40] The benefits of IT in hospitals are manifold Hospitals and Health Networks’ 2005 Most Wired Survey and Benchmarking Study reported significantly lower mortality rates for the most wired hospitals (i.e., those with the highest rate of investment in IT) [46] Preventable medical errors rank as the sixth leading cause of death in America [51] The mortality rank of medical errors is ahead of those for diabetes, liver disease, and pneumonia The lower mortality rate in the most wired hospitals is said to be due to the greater adoption of computerized physician order entry, more accurate medication order and delivery, and better decision support (e.g., information is maintained on duplicate orders, drug–drug interaction, dose checking, and allergy alerts) Hospitals classified as the most wired environments not only excel in lowering the mortality rates but also generate 5.4 percent more net patient revenues per fulltime employee compared to the U.S hospital average of 2.9 percent [46] The most wired hospitals also recorded a 3.8 percent decrease in paid staff hours per hospital discharge, compared to the U.S hospital average decrease of 1.8 percent Despite the potential benefits of IT, it is not clear whether hospitals are investing in the appropriate IT, how different types of IT are timed and sequenced, and what returns are obtained over time as a result [5, 23] One reason for the lack of clarity is that hospitals possess a dichotomous structure whereby administrators and medical staff form two “branches,” contending for resources, working under different constraints, and aiming for different objectives [26] There is some evidence that the use of different ITs affects contemporary hospital performance [5, 7, 17] The performance of hospitals with specific administrative IT applications (accounts payable and receivable, payroll, personnel administration) has been found to be different from those with specific clinical IT (e.g., electronic health records, electronic lab results, electronic clinical notes systems, electronic images throughout the hospital, Differential Effects of the Two Types of Information Systems 299 electronic lab orders, electronic reminders for guideline-based interventions, and e‑prescribing) [23] The value-chain model is often used to determine the business value of activities in an organization [43] One estimate puts the contribution of primary activities (inventory, production, sales, marketing, customer service) to value-added at 65.2 percent and the contribution of support activities (firm infrastructure, human resources, technology, research and development) at 34.8 percent [1, p 22] From an IT perspective, it is instructive to ascertain the business value of information systems (IS) following this nomenclature of activities Administrative IS in this paper represent IT in support value-chain activities, and clinical IS are representative of the IT in primary value-chain activities in a hospital It is argued that information transactions generated by primary activities possess higher value because of their revenue potential [32] IT for value-chain activities integrates product and customer data across various departments and sites, thereby affecting output directly, and has also been termed information technology for core competence [19, 39, 48, 58] Although competitive advantage is not the focus of this paper, productivity enhancements over time are considered necessary for competitiveness [4, 15, 45] Primary activities are dependent not only on other primary activities but also on many support activities This type of interdependence of processes in an organization must guide the timing and sequence of acquisition and upgrading of IT Support activity IT is also useful for generating long-term benefits because, for example, post hoc (after-sales) analysis of accounting data can provide unique insights for future value-chain activities Long-term value of IT comes from the long-term value of information Once created, databases such as customer records provide value even when transported from an old system to a new one [32] As a result, the maturation rate—that is, the pattern of business value of IS over time—of different types of IS will differ One way to find this pattern is to measure the effect of costs of different types of IS on performance over several years [10, 55] An elaborate model of hospital production function is presented, which helps to unravel patterns in the business value of types of IS Evaluating differential effects of types of IS has implications for IS project management and IS policy in accounting, among others With respect to IS project management, the empirical evidence that different types of IT applications have varying patterns in how returns on investment surface is enlightening The long-term association between the costs for two major types of IT, and hospital output and labor productivity measures is estimated for a sample of hospitals spanning from 1979 to 2006 The extensive panel data analysis, because it encompasses many generations of technology, enables more generality in interpreting the results, for a broader understanding of the impact of IT investment in the health-care industry Current and lagged (past years) capital depreciation amounts on two types of IT served as independent variables that approximate the long-term capabilities of each type of technology The two kinds of IT were found to exhibit different effects on the dependent variables over time Lags of clinical IS boosted hospital output 300 Menon, Yaylacicegi, and Cezar Administrative IS were related to an initial drop in organizational performance but improved performance in the long term Hospital IT and Business Value The electronic health record, a clinical IT application, contains information regarding a patient’s medical history—illnesses, digital radiology images, a list of allergies, billing records, and so on Electronic medical records have numerous advantages over paper records, including increased accuracy; decreased medical errors and mortality rates; improved efficiency; lowered costs; and better, safer, and more equitable care [2, 3] Electronic records also allow improved privacy and thus better compliance with privacy regulation [13] Policymakers call for the universal adoption of electronic health records by 2014, but the industry is unlikely to meet this deadline given its current adoption rates [30, 36] In a study by Jorgenson and colleagues [30], health care’s weighted average contribution to the annual U.S productivity growth rate between 1997 and 2000 ranked within the bottom two, with a –10 percent growth rate, compared to a 15 percent growth rate in the field of computers and office equipment The Healthcare Information and Management Systems Society’s 2002 leadership survey listed the barriers to implementing IT in health care as limited vendor ability and end-user acceptance, difficulty in providing return on investment, lack of IT strategy, and insufficient financial and management support [50] Productivity lags in labor-intensive service industries, such as health care [6] The health-care IT literature has sought to determine the value of IT in this costconscious industry (Table 1) Teplensky et al [49] focused on technology adoption Teplensky et al investigated hospitals’ reasons for investing in new technology (in their study, magnetic resonance imaging technology), and found that technology adoption was associated with the strategic orientation of health-care organizations Unlike the single technology focus of Teplensky et al., Burke and Menachemi [11] and Menachemi et al [38] looked at overall technology diffusion in hospitals Burke et al [12] associated IT adoption diffusion with various characteristics of hospitals, such as size and competition They also found that hospitals with high technology diffusion adopted strategic IT more readily, whereas hospitals with less technology diffusion were mostly limited to adoption of administrative IT Menachemi [37] found that hospitals with high technology diffusion also excelled in quality In a cover story, BusinessWeek reported on a hospital that, after spending over $72 million in IT projects over six years, entered electronically only 10 percent of its tests and orders [41] Other studies looked at postadoption effect of technology For example, Burke et al [12] observed operational financial improvements following health-care IT investments Walker et al [52] pointed out the financial value enhancement effects of standardized health-care systems Some studies such as Devaraj and Kohli [17] focused on postadoption effects of a single technology Devaraj and Kohli found that a decision support system had a positive effect on performance in health-care settings with a three-month lag, using a data set of eight hospitals over three years Kramer et al Objective Investigate financial benefits/costs of electronic medical records Wang et al [53] Investigates IT munificence as a measure of IT capability Studies the electronic health records functionality use Burke et al [12] Menachemi et al [38] Technology diffusion in hospitals Study the investment motivations of health institutions in new technology Teplensky et al [49] Technology adoption in hospitals Studies The black box of IT is quantified with IT munificence measure Hospitals with high technology diffusion also excel in quality Electronic health records adoption Primary care utilizes improved benefits (reduced drug expenditures, decreased billing errors, etc.) with use of more ambulatory electronic medical records features and as more time elapses after implementation Net financial costs/ benefits IT munificence Hospitals’ adoption decisions are primarily based on their strategic orientation Main findings Technological preeminence, profit maximization, and clinical excellence Major constructs Table Hospital Information Technology Literature Summary The data set used is limited to surveys The data set used is limited to surveys Only impact of electronic medical records adoption is investigated for one hospital This study is limited to adoption of only one capitalintensive medical technology Limitations (continues) The studies are limited to one type of technology adoption or a single hospital By including different types of IT expenditures, the differential effects over time can be explained Implications for this study Differential Effects of the Two Types of Information Systems 301 Analyzes savings due to electronic medical records Examine IT performance effects over time Evaluates the impact of physician order entry and electronic medication administration record adoption on health care Devaraj and Kohli [17] Mekhjian et al [36] Objective Kramer et al [31] One type of technology Studies Table Continued Significant monetary savings followed by electronic medical records implementation is reported IT investments have a positive profitability effect with three months or more lag Improved patient safety and timeliness are observed postimplementation Decision support systems Length of stay and cost Main findings Return on investment of electronic medical records Major constructs It is limited to one center Only 10–12 month period following investment is studied The study is limited to one type of IT, decision support systems The sample is limited to one psychiatric hospital Limitations Disaggregating the IT investment into clinical IT and administrative IT is sufficient granularity to capture to value chain effect of IT investments in hospitals Implications for this study 302 Menon, Yaylacicegi, and Cezar Explores the clinical and financial benefits of electronic data exchange Examines the relation between IT adoption and financial performance in hospitals Walker et al [52] Menachemi [37] The results cannot be generalized Data is mostly from expert estimates The results cannot be generalized Hospitals with high technology diffusion adopted strategic IT more readily Standardized health-care systems lead to enhanced financial value IT adoption leads to improved financial outcomes both overall and operationally Clinical IT, administrative IT, strategic IT, all-IT Cost and benefits of health-care information exchange and interoperability Administrative IT, clinical IT, strategic IT Count of applications used as measure of IT Count of applications used as measure of IT Clinical health-care IT has positive effect on hospital operational performance measured using the performance score as performance variable IT adoption lowers the costs at most automated hospitals Administrative IT, clinical IT, strategic IT Administrative IT, clinical IT, strategic IT Measures the effect of healthcare information technology adoption on hospital operational performance Examines how IT use affects the hospital operating costs using a longitudinal data Bhattacherjee et al [5] Borzekowski [7] Association between types of health-care IT and hospital performance Studies hospitals’ IT adoption strategies Burke and Menachemi [11] Financial gain associated with health-care IT investments In these studies, even the longitudinal one, the count of applications does not elicit the pattern of the effect of IT investments over time across many hospitals A study comparing the effect of IT investment on value-chain activities versus IT investment on support activities can be more useful explaining IT black box Differential Effects of the Two Types of Information Systems 303 304 Menon, Yaylacicegi, and Cezar [31] examined one psychiatric hospital and reported significant positive performance consequences of investment in electronic medical record technology Mekhjian et al [36] observed efficiency improvements along with reduced medical errors in an academic medical center after the implementation of physician order entry and electronic medical administration record systems Wang et al [53] studied the financial impact of electronic medical record systems Two works that looked at how the presence of types of IT affects performance are Bhattacherjee et al [5] and Borzekowski [7] The independent variable is the count of the number of applications of each type of IT—administrative, clinical, and strategic Borzekowski [7] used hospital costs as the performance variable, whereas Bhattacherjee et al [5] used a performance score variable collected from an accreditation organization for hospitals Both found support for a positive association between an index for clinical IT and performance, but not for other types of IT Both studies did not analyze the pattern of costs of different IS and the resulting pattern of returns obtained from these IS The reasons the costs and values of administrative and clinical IS differ over time stem from differences in types of value-chain activities, acquisition pattern, and usage pattern Usage and adoption of administrative IT is likely to be high, because administration’s tasks for communicating, coordinating, controlling, and planning are enhanced through IT [24] Physicians possess considerable influence in the clinical side, and have resisted adopting clinical ITs [7, 24] To date, no study has compared the patterns of costs and value of the two types of IS over time Methodology To show that two types of IS have different productivity effects, we conducted an empirical study of the relationship between different kinds of IT and the contribution to hospital performance at the hospital unit of analysis The data from 1979 to 2006 were obtained from the Washington State Department of Health (WaDoH) hospital database Data Overview The annual cost accounting data for each hospital are available at the subunit level in each hospital Costs comprise salaries, capital depreciation, supplies, purchased services, rental and lease expenses, and miscellaneous expenses For each hospital, the data contain inpatient and outpatient revenues, inpatient days, outpatient visits, and number of beds To homogenize the hospital population, we included only general medical and surgical hospitals in the sample.1 Because the data spanned several years, during which the U.S economy and the medical environment underwent several changes, we deflated all monetary variables to remove macroeconomic and time effects across observations: (1) salaries, employee benefits, and professional fees in all accounts for each hospital were deflated by the employment cost index for healthcare services, obtained from the Web site of the Bureau of Labor Statistics [42]; Differential Effects of the Two Types of Information Systems 305 (2) other expenses—depreciation, rental and lease, supplies, purchased services, and miscellaneous expenses—were deflated by the Producer Price Index for hospitals; (3) revenues—inpatient and outpatient—were deflated by the Consumer Price Index for health-care services The hospitals followed the same reporting format2 so that the department name and definition provided by WaDoH were useful in classifying the major capital depreciation Examples of departments in a hospital are data processing, admitting, patient records, hospital administration, accounting, and personnel Independent Variables: Two Types of IT Capital depreciation in departments identified as clinical or administrative was approximated as a measure of the department’s IT capability The variance in capital depreciation across hospitals and over time is a proxy for IT infrastructure variance across hospitals and time IT infrastructure is defined as “the base foundation of the information technology portfolio, which is shared throughout the firm in the form of reliable services” [1, p. 258] Thus, even without application-level data, we expected that the capital depreciation of infrastructure would provide a reasonable proxy for IT acquisition pattern, and perhaps to some degree, the adoption and usage patterns Aggregating capital over several departments has the advantage over a finer grain for IT in that the latter would require an inventory of all hardware and software technologies in all hospitals The quantity of data (the number of hospitals and years) prevented such a detailed analysis It is useful to measure IT in terms of capital depreciation (rather than presence or absence of application), because more costly applications require better quality infrastructure Some have argued that superior infrastructure has direct organizational benefits [8, 31, 39, 48], whereas others have regarded infrastructure as a competitive necessity [1] Departments in which capital is primarily clinical IT are patient accounts, patient records, admitting, and pharmacy [29].3 Again, the underlying assumption is that this is the infrastructure that most affects clinical applications An objective of these departments is to reduce errors in treatment by better maintaining the medical history of patients, helping to reduce the confusion that can be caused when patient care is provided by multiple entities, including the hospital and physicians Proper use of clinical IT avoids redundant tests, which lowers operating costs [17] Clinical IT helps maintain records of billings and payments to physicians and other health-care providers For administrative IT, we summed the capital depreciation from utilization management, management engineering, hospital administration, nursing administration, data processing, personnel, purchasing, accounting, and communications Dependent Variables Most hospitals in the state of Washington are not-for-profit organizations and strive to reduce costs Hospital output and labor productivity are important outcomes in the health-care industry Hospital output is measured by the number of adjusted patient 306 Menon, Yaylacicegi, and Cezar Table Descriptive Statistics and Correlations Correlation matrix Dependent and independent Standard variables Mean deviation Hospital charges   per labor cost Adjusted patient   days per   employee Administrative IT Clinical IT Hospital Adjusted charges patient per days per Administrative labor cost employee IT 0.42 0.44 1.0 4.18 0.48 0.70 1.0 8.85 7.17 1.34 1.36 –0.30 –0.30 –0.59 –0.55 1.0 0.81 Note: All Pearson correlation coefficients significant at < 0.0001 days of patient care The adjusted patient days measure was the sum of inpatient days and outpatient “days” (i.e., visits converted into a days measure according to the proportion of inpatient revenues and outpatient revenues).4 Labor productivity was calculated in two ways: (1) hospital charges divided by total salaries, and (2) adjusted patient days divided by total number of employees The former measure can be thought of as a price-weighted average of the latter measure Accordingly, the former measure accounts, to some degree, for the differences in the types of cases a hospital faces, as well as the skill levels of employees Descriptive statistics (mean, standard deviation, and the correlation matrix) of the log-transformed dependent variable and the main independent variables are provided in Table All Pearson correlation coefficients are significant Control Variables We used variables to control production-specific effects, hospital-specific fixed effects, and time-specific effects in the model for testing hypotheses To control productionspecific effects, we added other input factors used by a hospital to produce services to the model These were medical capital Medical capital explains both hospital output and medical labor productivity Lags of medical capital were used following the logic that non-IT capital affect productivity in a lagged manner [10, 27] Hospital-specific effects could occur as a result of differences in production processes and organizational incentives [16, 20] These effects were captured by government status, profit status, teaching status, and urban status [57] The government status of hospitals was a binary indicator, coded for government hospitals Profit status was also a binary indicator, with for-profit hospitals coded as The teaching status of hospitals also contributed to hospital-specific differences.5 Another aspect that contributed to hospital-specific production process and to the levels of capital was urban status [9].6 In addition, the Differential Effects of the Two Types of Information Systems 307 degree of care provided to the older adult and indigent population was captured by a variable called Medicare status, which was the ratio of Medicare and Medicaid charges to total hospital charges Finally, the severity of cases status of a hospital was captured by the ratio of the charges of intensive care, acute care, and semi-intensive care units to the total charges This control variable replaces the case-mix variable often used in hospital-based studies, because hospitals began computing case-mix index in the early 1990s, while our data begins from the late 1970s To control for size of the hospital, we used the log-transformed number of beds as a control variable for each hospitalyear observation We also used fixed effects binary variables, coded for a particular hospital and for other hospitals, to capture the effects of any remaining effects (e.g., epidemiologic) on the dependent variables Finally, the lag of the dependent variable is also used as a control variable so that the independent variables are explaining the incremental variance in each year given last year’s performance This helps to control for some unobservable factors in panel data (similar to the first-difference method) Empirical Model To test the longitudinal relationship between the two dependent variables (hospital output and medical labor productivity) and types of IT (clinical IT, administrative IT), we used a log-linear formulation function resembling the Cobb–Douglas function Use of lagged values of independent variables in an ordinary least squares model leads to two problems First, there will be high multicollinearity among the regressors that will inflate the standard errors and increase p‑values of individual estimates, leading to the rejection of otherwise significant estimates Second, the degrees of freedom decrease with the inclusion of the lagged regressors The Almon polynomial-distributed lag model alleviates these problems [25] In this procedure, the coefficients of the lagged values of the regressors are assumed to be a polynomial function of the lag numbers (see Equation (1) where lag is a variable representing the lag number, and p is the polynomial power) The parameters are estimated by maximum likelihood through gradient methods Similar models have been used in several applications, such as in determining the lagged effect of advertising [54] and in accounting [35] The lengths of the lag for each type of IT tested were four, five, six, and seven years There is empirical evidence that aggregated IT spending has an impact for about six years [10] Other estimates of the “life” of IT are seven years [42], five to seven years [56], and 12 years [47] These examples are used for accounting purposes and may not reflect the length of time that economic value is derived from an asset Borzekowski [7] found that the presence of clinical applications affected hospital costs in a favorable manner for a lag of five years (however, he did not perform an analysis with lags of costs of clinical applications) It follows that IT, on which many applications run, could show an impact for four to seven years The degree of the polynomial function of the lag number for the coefficients of lagged variables is in most cases 2, which is followed here (parameter p in Equation (1)) The formal regression models (for four lags: 4, 5, 6, and years) are given in Equation (1): 308 Menon, Yaylacicegi, and Cezar log ( yit ) =     ∑   ∑ aip lag p  log (CNITt −lag ) lag =   p =1   + ∑   ∑ aip lag  log ADITt − lag  + coontrol + ε it ,  lag =   p =1  4,5,6,7   4,5,6,7   p ( (1) ) where yit is the dependent variable and i is i ∈ {o, l} (o means that the dependent variable is hospital output, and l is labor productivity) CNIT stands for clinical IT, and ADIT stands for administrative IT The coefficients of IT variables are in the Almon polynomial-distributed lag form S p=1aiplagp, where p is the degree of the polynomial The control variables are medical capital (with same lags as IT) and labor (with three lags), government status, teaching status, profit status, severity status, Medicare status, log of beds, and year-fixed effects The combination of these several control variables ensures that hospital and time-specific variance in the dependent variable are accounted for and that the variance explained by the IT variables and their lags can be generalized Results Equation (1) was tested for first- and second-order autocorrelation of the error terms What this means is that the one-year and two-year lags of the dependent variables were added as control variables in the corresponding models The first lags of the dependent variables were significant, but the second-year lags were not, and so one-year lag of the dependent variable were put as a control variables in each model Thus, the independent variables explain variance in hospital performance after controlling for the previous year’s performance As mentioned earlier, different lag structures including four, five, six, and seven lags we tested The Akaike information criterion was useful to balance models’ precision (how well data are explained) and complexity (degree of parameterization) [25].7 We report only the seven lags model for both IT variables in Tables and (the dependent variable is hospital charges per labor cost), and Tables 5 and (the dependent variable is adjusted patient days per employee) For hospital charges per labor cost, administrative IT did not show a significant association until the sixth year (Table 3) Clinical IT showed a negative association in the current year (–0.02) and the past year (–0.01) The coefficient (0.01) for the past-year lag of medical capital was significant The coefficients of other control variables were also significant, with expected signs for the coefficients For example, profit status, urban status, and number of beds were positively associated with hospital charges per labor cost, and government status was negatively associated There were some exceptions to the expected signs of coefficients, however Teaching status was not significantly associated, and severity status was negatively associated with charges per labor cost The model fit in Table does suggest a reasonable model For adjusted patient days per employee, administrative IT showed negative associations in the current and past year and positive associations with three- to five-year lags Differential Effects of the Two Types of Information Systems 309 Table Hospital Charges per Labor Cost Estimates Estimate Lag for IT and capital variables ADIT CNIT Medical capital Current year Past year Past two years Past three years Past four years Past five years Past six years Past seven years 0.015 Intercept (0.01) 0.010* Teaching (0.004) status 0.005 Government (0.003) status 0.002 Profit (0.004) status –0.0006 Urban (0.004) status –0.003 Severity (0.003) status –0.004 Medicare (0.003) status –0.005 Log of (0.008) beds 0.002 (0.006) 0.001 (0.003) 0.0009 (0.002) 0.0010 (0.003) 0.001 (0.003) 0.003 (0.002) 0.0056* (0.003) 0.008 (0.006) –0.019*** (0.005) –0.012*** (0.002) –0.006** (0.002) –0.002 (0.003) 0.0003 (0.003) 0.001 (0.002) 0.0005 (0.002) –0.001 (0.005) Other controls Estimate 0.572*** (0.09) –0.005 (0.01) –0.114*** (0.01) 0.184*** (0.02) 0.030* (0.01) –0.028** (0.01) 0.082 (0.05) 0.041** (0.01) Notes: Model corresponds to Equation (1) with seven lags for IT in the following form: log (Charges per Labor Cost ) = +   p   ∑ aiplag  log CNITt − lag   lag =   p =1 ( ∑   p   ∑ aiplag  log ADITt − lag   lag =   p =1 ( ∑  )   ) + control + ε  it Standard errors are shown in parentheses Fixed effects for hospital not reported here * p