No national studies have examined the interaction effect of the electronic health record (EHR) and hospitalist care on length of stay (LOS). Thus, we examine the combined effect of the EHR and hospitalist care on LOS.
Objectives: To assess the combined impact of electronic health re-cord (EHR) adoption and hospitalist care on length of stay (LOS).
Study Design: Retrospective cohort study using data from the
Healthcare Information and Management Systems Society and a
5% national sample of Medicare beneficiaries. Patients included 20,862 admissions for respiratory disease cared for by hospital-ists, and 28,714 admissions for respiratory disease cared for by nonhospitalists in 2985 general and surgical hospitals in the United States.
Methods: The interaction effect of EHR and hospitalist care on LOS was evaluated using generalized linear models with log-link normal distribution after controlling for patient and hospital characteristics.
Results: In multivariable analyses controlling for patient and hospital characteristics, we found that the reduced LOS associ-ated with complete EHR was 0.166 days and was statistically significant in the hospitals in which 50% or less of patients were cared for by hospitalists. Moreover, we found that reductions in LOS associated with hospitalist care were greater in hospitals that had not adopted a complete EHR. LOS was 0.599 days shorter for patients cared for by hospitalists versus nonhospitalists in hospi-tals with incomplete EHR adoption; in hospitals with complete EHR adoption, the stay was 0.433 days shorter.
Conclusions: The reduced LOS associated with hospitalist care is greater than that associated with EHR adoption. However, the combined reduction in LOS with both EHR adoption and hospital-ist care may be substantial.
Am J Manag Care. 2015;21(3):e215-e221
The electronic health record (EHR) and hospitalist care were associated with reduced length of stay (LOS).
The percentage of hospitalized patients cared for by hospitalists—physicians who specialize in the care of this type of patient—jumped from 5.9% in 1995, to 19% in 2006.1 The justification for the increased use of hospitalists stems from both the growing complex-ity of inpatient care and its increasing costs. A physician specializing in hospital medicine should be more efficient, and also perhaps more effective in initiating rapid and ap-propriate treatment, as well as in recognizing and prevent-ing complications of hospitalization.2-4 Hospitalists can also participate in multidisciplinary teams that address patient safety, quality, and early discharge planning.5,6
Another approach to improving efficiency is the electron-ic health record (EHR). EHR use grew in US hospitals during the same time period that the use of hospitalists increased. A complete EHR contains tools to increase efficiency and to increase communication among providers. Some studies have found that the EHR is associated with lower length of stay (LOS),7,8 and the EHR has also been reported to improve the quality of care by increasing adherence to guideline- or protocol-based care, identifying adverse drug events, reduc-ing medical error, and providing decision support.7 In addi-tion, the EHR allows tracking of therapy in detail, enabling the physician to address adherence and compliance issues.9-11 Thus, the EHR can improve efficiency by improving conti-nuity in information transfer and in communication among health providers.
We know of no prior studies examining the interaction of the EHR and hospitalists on LOS. We had 2 closely related hypotheses: 1) that the effect of EHR on LOS would be greater in hospitals with few hospitalists; and 2) that the effect of hos-pitalists on LOS would be less in hospitals with a complete EHR. We examined the association of EHR and hospitalist care on LOS, focusing on 49,576 admissions with respiratory disease from a 5% national sample of Medicare data in 2009, linked to data on EHR adoption from the Healthcare Infor-mation and Management System Society (HIMSS).
Sources of Data
Claims from a 5% national sample of Medicare benefi-ciaries in 2009 were used, as well as Medicare enrollment files, Medicare Provider Analysis and Review (MEDPAR) files, Outpatient Statistical Analysis files, Medicare Car-rier files, Provider of Services (POS) files, diagnosis-related group (DRG) weight files, and HIMSS data files.
Establishment of the Study Cohort
Beneficiaries enrolled in health maintenance organiza-tions or those without both Medicare Parts A and B for the entire 12 months before admission were excluded. Be-cause the study focused on hospitalist care, and more than 95% of hospitalists treating adult patients are generalist physicians, admissions not involving an evaluation and management charge by a general internist, family physi-cian, general practitioner, or geriatrician were excluded.
We selected respiratory disease because continuity of care is very important for chronic pulmonary diseases such as chronic obstructive pulmonary disease, pneumo-nia, and bronchitis.12,13 We selected 34 DRGs related to respiratory disease using the DRG codes from 175 to 208. Patients with respiratory disease accounted for 20,862 ad-missions cared for by hospitalists and 28,714 admissions cared for by nonhospitalists.
For hospitals, we selected only general and surgical hospitals (n = 3452). However, 467 of these were excluded because they could not be merged with POS and MED-PAR files. This left 2985 general and surgical hospitals in 2009 for study.
Hospitalists were defined as generalist physicians who had more than 5 Medicare evaluations and management claims and generated more than 90% of these claims from care provided to hospitalized patients in the year studied. This definition was validated at 7 hospitals, and had a sensitivity of 84.2% and a specificity of 96.5%.1 The effect of hospitalist care was ex-amined at the individual patient level and at the hospital level as the percent of pa-tients cared for by hospitalists in a hospital.
Medicare enrollment files were used to categorize patients according to age, sex, and race (ie, white, black, and other). A Medicaid indication in the enrollment file was used as a proxy for low socioeconomic status. Information regarding origin of ad-mission (emergency department vs other), weekend versus weekday admission, admission with intensive care unit (ICU) stay, and discharge DRG were obtained from the MEDPAR files. DRG weight reflects the average amount of resources used for each DRG and was used to adjust severity across different hospitals/patients.
Elixhauser comorbidity measures were generated using inpatient and physician claims from the MEDPAR, Out-patient Statistical Analysis files, and Carrier files.14 The total number of hospitalizations and total number of out-patient visits in the year before the index hospitalization were generated from the MEDPAR and Carrier files. Hos-pital information (ie, zip code, county, state, total number of hospital beds, type of hospital, and medical school affil-iation) was obtained from the POS file. Metropolitan area size was generated from 2010 Census data. Metropolitan area size and total number of hospital beds were catego-rized into quartiles; states were grouped by census region; type of hospital was categorized as nonprofit, for-profit, or public; and medical school affiliation was categorized as none, minor, or major.
The EHR adoption information was generated using HIMSS data, which provide the adoption information for health information technology (HIT) systems. The EHR was defined as including a clinical decision support system, clinical data repository, computerized physician order entry, and physician document.6,15-17 Hospitals that adopted these 4 HIT systems were regarded as adopting the complete EHR system.
The study outcome was hospital LOS for admissions with respiratory disease, obtained from the MEDPAR files. The relationship of EHR use and hospitalist care on LOS was evaluated using generalized linear models (GLMs) with log link and normal distribution. GLMs are commonly used to analyze outcomes that are not norm-ally distributed; they apply a transformation, known as the link function, to the mean of data. Furthermore, rather than assume that a transformation of the data leads to normally distributed data to which standard linear modeling techniques can be applied, GLMs take the dis-tribution of the data into account.18 To adjust for patients clustering within hospitals, the parameters were estimated by generalized estimating equation with an exchange-able working correlation matrix. In the GLM models, we included the interaction between EHR and hospital-ist care, and controlled for patient characteristics (ie, age, sex, race, socioeconomic status, DRG group, emergency department [ED] admission, admission with ICU stay, weekend admission, comorbidity, DRG weight, and total number of hospitalizations and provider visits in the 12 months before the index admission) and hospital charac-teristics (ie, region, metropolitan area size, total number of beds, type of hospital, and teaching affiliation). We re-ported the difference on adjusted average LOS associated with EHR and hospitalist care by taking exponential to the predicted LOS from the GLM models. Analyses were performed using STATA version 10.1 (StataCorp LP, Col-lege Station, Texas).19
and provide patient and hospital char-acteristics stratified according to whether patients re-ceived care from a hospitalist or a nonhospitalist and by whether or not hospitals had a complete EHR in 2009. The 4 groups differed by socioeconomic status, number of comorbidities, average number of hospitalizations in the year before index admission, average number of doctor visits in the year before index admission, DRG weight, percentages with ED admission, ICU use, hospital teach-ing status, hospital size, hospital ownership, size of the metropolitan area, and geographic region (all P values <.001). Seventeen percent of the 2985 hospitals had ad-opted a complete EHR by 2009.
In a multivariable model, we found a significant inter-action between whether a hospital had a complete EHR and whether it was a high versus a low user of hospital-ists in terms of patient LOS. Therefore, in the analyses, we stratified hospitals by those 2 characteristics. shows the differences in adjusted LOS between hospitals with ≤ or >50% of patients cared for by hospitalists. Hospi-tals in which ≤50% of patients receive hospitalist care have slightly longer LOS than those in which more patients receive hospitalist care (4.97 vs 4.61 days, not reported). In hospitals in which ≤50% of patients receive hospitalist care, a complete EHR was associated with an average LOS 0.17 days shorter than that of similar hospitals without a complete EHR, a statistically significant result. However, a complete EHR was not associated with reduced LOS in hospitals in which most patients received hospitalist care. Moreover, hospitals in which ≤50% of patients received care by hospitalists are more likely to be smaller (222 vs 272 beds), nonprofit (63.1% vs 39.9%), and nonteaching hospitals (71.1% vs 39.9%) than those in which >50% of patients were cared for by hospitalists.
Second, we tested the hypothesis that reductions in LOS associated with hospitalist care would be greater in hospitals without a complete EHR. presents the differences in adjusted LOS between patients cared for by either hospitalists or nonhospitalists in hospitals with or without EHR adoption. LOS was 0.60 days shorter for patients cared for by a hospitalist versus a nonhospital-ist in hospitals without a complete EHR adoption. In hospitals with a complete EHR, the patient stay was 0.43 days shorter for patients cared for by a hospitalist versus a nonhospitalist. The difference of LOS decrease between those receiving nonhospitalist and hospitalist care in a hospital without and with a complete EHR is 0.17 days, which is significantly different from zero (P ≤.01).
Our definition of complete EHR is less restrictive than the EHR adoption model used in another study.9 In a sen-
eAppendix Tables 1
sitivity analysis, we also defined EHR more restrictively, including clinical decision support system, clinical data repository, computerized physician order entry, physician notes, Electronic Medication Administration Record, and nursing notes to test our hypothesis. The regression results were similar to previous ones and are reported in and (available at www.ajmc.com).
Many studies report shorter LOS associated with hos-pitalist care1-3,20; moreover, the EHR has been reported to lower LOS by increasing physician efficiency and commu-nication of medical information. However, to our knowl-edge, no study has been done on the combined impact of EHR and hospitalist care on efficiency. We found that EHR adoption improves efficiency in hospitals in which fewer pa-tients were cared for by hospitalists. Thus, EHR adoption may complement hospitalist care in hospitals that have a relatively small number of hospitalists among its physicians. Hospitals with fewer patients cared for by hospitalists are more likely to be small, nonprofit, and/or nonteaching hospitals. While these hospitals might benefit most from EHR in improving efficiency, they may have limited access to capital and infrastructure,21 which makes adoption more difficult. The incentive program of the Health Information Technology for Economic and Clinical Health (HITECH) Act may increase EHR adoption in those hospitals, result-ing in improved efficiency.
We found a significant interaction of EHR and hospi-talists in terms of length of stay, such that the presence of one lessened the effect of the other. Prior studies of the impact of EHR or of hospitalists on length of stay have studied just one of those in isolation. What this means, is that the magnitude of the impact of hospitalist on length of stay found in studies in hospitals without an EHR would not translate to a similar impact in hospitals with an EHR. Even though the reduced LOS associated with hospitalist care in hospitals with a complete EHR was smaller than that in hospitals without a complete EHR, the total reduction in LOS associated with both hospital-ists and a complete EHR was 0.61. A change of 0.61 days translates to 613 bed days saved for 1000 discharges.
First, hospital costs were not analyzed. Medicare claims data provide information on charges only. The correlation between total charges and LOS was 0.65 in our sample. However, we focused on LOS because it was a more direct measure of resource use than charge. Second, a hospitalist was defined as a doctor who derived 90% or more of evalu-ation and management charges from hospitalized patients.1 Previous studies have defined hospitalist differently22-25— our definition of hospitalist does not include the physi-cian’s years of EHR and medical subspecialty experience. Third, we included only patients admitted with respiratory disease, so caution should be used when generalizing the results of this study. Fourth, payer mix may be associated with LOS and should be included in our model; however, that information is not available in our data, and so our re-sults may be biased. Fifth, we looked at only cross-sectional data. The effect of EHR adoption on reduced LOS asso-ciated with hospitalist care could be larger if we consider longitudinal data, because EHR adoption has a learning spillover effect.26,27 Last, the combined effect of EHR and hospitalist care on outcomes, such as readmission rates, was not addressed. Future study in this area is needed.
We tested the association of hospitalist care and EHR adoption on LOS using a 5% national sample of Medicare beneficiaries and HIMSS in 2009. We used a retrospective cohort study and found that the reduced LOS associated with hospitalist care is greater than that associated with EHR adoption. However, the combined reduction in LOS with both EHR adoption and hospitalist care may be sub-stantial (0.61 days). Thus, the interaction effect of EHR adoption and hospitalist care may be of interest to hospi-tal administrators.
Author Affiliations: Sungkyunkwan University, Department of Eco-nomics (JL), Seoul, South Korea; Department of Internal Medicine (Y-FK, JSG), and Sealy Center on Aging (Y-LL), University of Texas Medical Branch, Galveston, TX.
Source of Funding: The National Institute on Aging (grants 1R01- AG033134).
Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (JL, Y-FK, JSG); acqui-sition of data (JL); analysis and interpretation of data (JL, Y-FK, Y-LL); drafting of the manuscript (JL, JSG); critical revision of the manuscript for important intellectual content (JL, Y-FK, JSG); statistical analysis (JL, Y-LL); provision of patients or study materials (JL); obtaining funding (JSG); supervision (JL, JSG); data management (JL, Y-LL).
Address correspondence to: Jinhyung Lee, PhD, Department of Eco-nomics, Sungkyunkwan University, 53 Myeongnun-Dong 3-Ga, Jongno-Gu, Seoul, Korea 110-745. E-mail: email@example.com.
1. Kuo YF, Sharma G, Freeman JL, Goodwin JS. Growth in the care of older patients by hospitalists in the United States. N Engl J Med. 2009; 360(11):1102-1112.
2. Coffman J, Rundall TG. The impact of hospitalists on the cost and quality of inpatient care in the United States: a research synthesis. Med Care Res Rev. 2005;62(4):379-406.
3. Wachter RM. Reflections: the hospitalist movement a decade later. J Hosp Med. 2006;1(4):248-252.
4. Freed DH. Hospitalists: evolution, evidence, and eventualities. Health Care Manag (Frederick). 2004;23(3):238-256.
5. Rohde JM, Jacobsen D, Rosenberg DJ. Role of the hospitalist in an-timicrobial stewardship: a review of work completed and description of a multisite collaborative. Clin Ther. 2013;35(6):751-757.
6. Whelan CT. The role of the hospitalist in quality improvement: sys-tems for improving the care of patients with acute coronary syndrome. J Hosp Med. 2010;5(suppl 4):S1-S7.
7. Furukawa MF, Raghu TS, Shao BB. Electronic medical records, nurse staffing, and nurse-sensitive patient outcomes: evidence from Califor-nia hospitals, 1998-2007. Health Serv Res. 2010;45(4):941-962.
8. Lee J, Kuo YF, Goodwin JS. The effect of electronic medical record adoption on outcomes in US hospitals. BMC Health Serv Res. 2013;13:39.
9. Jha AK, DesRoches CM, Campbell EG, et al. Use of electronic health records in U.S. hospitals. N Engl J Med. 2009;360(16):1628-1638.
10. Teich JM, Merchia PR, Schmiz JL, Kuperman GJ, Spurr CD, Bates DW. Effects of computerized physician order entry on prescribing practices. Arch Intern Med. 2000;160(18):2741-2747.
11. Wang SJ, Middleton B, Prosser LA, et al. A cost-benefit analysis of electronic medical records in primary care. Am J Med. 2003;114(5): 397-403.
12. Love MM, Mainous AG 3rd, Talbert JC, Hager GL. Continuity of care and the physician-patient relationship: the importance of continuity for adult patients with asthma. J Fam Pract. 2000;49(11):998-1004.
13. Leine S; Family Health International. Managing chronic conditions: optimizing health throughout life. http://www.c-hubonline.org/sites/ default/files/resources/main/Managing%20Chronic%20Conditions.pdf. Published December 2008. Accessed August 28, 2013.
14. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity mea-sures for use with administrative data. Med Care. 1998;36(1):8-27.
15. Miller AR, Tucker CE. Can health care information technology save babies? J Polit Econ. 2011;119(2):289-324.
16. McCullough J, Casey M, Moscovice I, Prasad S. The effect of health information technology on quality in U.S. hospitals. Health Aff (Mill-wood). 2010;29(4):647-654.
17. Fonkych K, Taylor R. The State and Pattern of Health Information Technology Adoption. Santa Monica, CA: RAND Corporation; 2005.
18. McCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. Lon-don: Chapman & Hall; 1989.
19. StataCorp. 2007. Stata Statistical Software: Release 10. College Sta-tion, TX: StataCorp LP.
20. Kuo YF, Goodwin JS. Impact of hospitalists on length of stay in the Medicare population: variation by hospital and patient characteristics, J Am Geriatr Soc. 2010;58(9):1649-1657.
21. Altarum Institute. Overcoming challenges to health IT adoption in small, rural hospitals. HealthIT.gov website. http://www.healthit.gov/ sites/default/files/pdf/OvercomingChallenges_in_SmallRuralHospitals .pdf. Published October 2011. Accessed August 28, 2013.
22. Wachter RM, Katz P, Showstack J, Bindman AB, Goldman L. Reor-ganizing an academic medical service: impact on cost, quality, patient satisfaction, and education. JAMA. 1998;279(19):1560-1565.
23. Palmer HC, Armistead NS, Elnicki DM, et al. The effect of a hospit-alist service with nurse discharge planner on patient care in an aca-demic teaching hospital. Am J Med. 2001;111(8):627-632.
24. Kearns PJ, Wang CC, Morris WJ, et al. Hospital care by hospital-based and clinic-based faculty: a prospective, controlled trial. Arch Intern Med. 2001;161(2):235-241.
25. Meltzer D, Manning WG, Morrison J, et al. Effects of physician experience on costs and outcomes on an academic general medicine service: results of a trial of hospitalists. Ann Intern Med. 2002; 137(11):866-874.
26. Patel V, Kushniruk AW, Yang S, Yale JF. Impact of a computer-based patient record system on data collection, knowledge organization, and reasoning. J Am Med Inform Assoc. 2000;7(6):569-585.
27. Borzekowski R. Measuring the cost impact of hospital information systems: 1987-1994. J Health Econ. 2009;28(5):938-949. n