i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
journal homepage: www.ijmijournal.com
Can a health information exchange save healthcare costs? Evidence from a pilot program in South Korea Hayoung Park a , Sang-il Lee b , Hee Hwang c,∗ , Yoon Kim d , Eun-Young Heo e , Jeong-Whun Kim f , Kyooseob Ha g a
Technology Management, Economics, and Policy Graduate Program, Seoul National University, 1 Kwanak-ro, Kwanak-gu, Seoul 151-015, Republic of Korea b Department of Preventive Medicine, College of Medicine, University of Ulsan, 86 Asan Byungwon-gil, Songpa-gu, Seoul 138-736, Republic of Korea c Department of Pediatrics, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea d Department of Health Policy and Management, College of Medicine, Seoul National University, 103 Daehak-ro, Jongno-gu, Seoul 110-799, Republic of Korea e Department of Medical Informatics, Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea f Department of Otorhinolaryngology, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea g Department of Psychiatry, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objective: Governments and institutions across the world have made efforts to adopt and
Received 20 January 2015
diffuse the health information exchange (HIE) technology with the expectation that the
Received in revised form 6 May 2015
technology would improve the quality and efficiency of care by allowing providers online
Accepted 15 May 2015
access to healthcare information generated by other providers at the point of care. However, evidence concerning the effectiveness of the technology is limited hindering the wide
Keywords:
adoption of a HIE. The objective of this study was to assess impacts of a HIE on healthcare
Community networks
utilization and costs of patient episodes at a tertiary care hospital following referrals by
Electronic health records
clinic physicians.
Health information exchange
Material/methods: We studied 1265 HIE and 2702 non-HIE episodes after physicians referred
Health information technology
patients from 35 HIE and 59 non-HIE clinics to Seoul National University Bundang Hospital
Healthcare cost
(SNUBH) during a 17-month period from June 2009. We examined 9 measures of healthcare utilization and the magnitude of clinical information exchanged in 4 categories. We estimated the savings resulting from HIE use through linear regression models with dummy variables for HIE participation and patient classification codes controlling the case-mix differences between HIE and non-HIE cases.
Abbreviations: EMR, electronic medical record; HIE, health information exchange; KDRG, Korean diagnosis related group; KOPG, Korean outpatient group; KRW, Korean won; NHIS, National Health Insurance Service of Korea; SNUBH, Seoul National University Bundang Hospital; USD, US dollar. ∗ Corresponding author. Tel.: +82 10 5235 0903; fax: +82 31 787 4054. E-mail addresses: [email protected] (H. Park), [email protected] (S.-i. Lee), [email protected] (H. Hwang), [email protected] (Y. Kim), [email protected] (E.-Y. Heo), [email protected] (J.-W. Kim), [email protected] (K. Ha). http://dx.doi.org/10.1016/j.ijmedinf.2015.05.008 1386-5056/© 2015 Elsevier Ireland Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
659
Results: The total charges incurred by the HIE group during episodes at SNUBH were approximately 13% lower (P < 0.001), and the charges for clinical laboratory tests, pathological diagnosis, function tests, and diagnostic imaging were 54% (P < 0.001), 76% (P < 0.001), 73% (P < 0.001), and 80% (P < 0.001) lower for the HIE group than for the non-HIE group. SNUBH physicians had access to more clinical information for HIE than for non-HIE patients. Conclusions: HIE technology improved physicians’ access to past clinical information, which appeared to reduce diagnostic test utilization and healthcare costs. The payer was the major beneficiary of HIE cost savings whereas providers paid for the technology. Fair allocation of benefits and costs among stakeholders is needed for wide HIE adoption. © 2015 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
The adoption of a health information exchange (HIE), which is defined as the electronic transmission of healthcare information among healthcare providers, has been slow despite its highly anticipated benefits in healthcare quality and cost [1–3]. The technology enables provider online access to healthcare information generated by others at the point of care, and therefore it is expected to improve the quality and efficiency of care and reduce the operating and administrative costs of healthcare providers [4]. South Korea, in particular, needs this technology because care is fragmented. Hospitals employ their own medical staff and are closed to physicians at local clinics. However, the barriers to technology adoption are particularly high. Hospitals and clinics often compete for patients, with a large share of hospital revenue coming from outpatient care. Further complicating HIE adoption, providers are paid on a fee-for-service basis such that higher utilization of healthcare services generates higher payments for providers. As a result, providers’ financial incentives are not aligned with containment of healthcare costs through the adoption of an HIE. Previous literature has emphasized the importance of effectiveness research for wide adoption and diffusion of an HIE, particularly its value in saving healthcare costs [2,5–8]. Financial incentives that align interests of stakeholders cannot be designed without proper estimation of benefits, which previous studies have indicated as a significant factor in facilitating the adoption and sustainability of the adopted technology [4,9–12]. However, few researchers have attempted to measure the economic benefits of an HIE, and evidence quantified with empirical data obtained from an operational HIE was sketchy [13–18]. Overhage et al. found a decrease of 26 US dollars (USD) per emergency department encounter in an HIE group at 1 of 2 pilot study sites in Indianapolis, Indiana, but no significant savings in the other site [13]. Frisse et al. found that HIE access was significantly associated with a decrease in hospital admissions and utilization of head and body computed tomography as well as laboratory testing in a study of emergency department encounters in Memphis, Tennessee [16]. Magnus et al. [17] and Bell et al. [18] reported improvements in quality of care measures through the HIE technology in HIV care settings. Further studies are needed to strengthen the evidence of HIE technology impact on healthcare cost and quality.
We examined physicians’ acceptance and use of an HIE, patients’ perceptions of the technology, and the costs and benefits of the system as part of a 3-year pilot project supported by a grant from the Ministry of Health and Welfare of South Korea [19–21]. In this report of the study, we assessed the impact of an HIE on healthcare utilization and costs. We estimated cost savings by comparing healthcare utilization of 2 groups of patients receiving similar care but who differed in terms of access to HIE technology.
2.
Materials and methods
2.1.
Study setting
The Ministry of Health and Welfare of South Korea funded a 3-year pilot project to examine the feasibility of using an HIE in South Korea, operate a prototype system, obtain evidence on the clinical and economic impacts of the technology, and explore other issues that may emerge when an HIE is introduced. Seoul National University Bundang Hospital (SNUBH), a 980-bed tertiary care university teaching hospital located 20 km south of Seoul, received a grant from the ministry and launched the pilot project in November 2007. The first version of the system was deployed in June 2008 and was completed in October 2009. Terms and conditions of participation were presented to local clinics and physician practices (hereafter, clinics) with referral arrangements with the hospital, either in recruitment sessions or in visits to clinics on their request. Clinic participation in the project was on a voluntary basis, and 35 clinics participated in the project as of October 2009. SNUBH and participating clinics had been using electronic medical records (EMRs) at the time the project commenced. The HIE system was based on a federated architecture model, and in the final system, exchanged information included patients’ demographic data and health status, including diagnoses with chief complaints, prescribed medications, laboratory results, diagnostic images, duration and content of treatments, care plans, vital signs, history, and summaries [19,20]. Physicians at HIE clinics introduce the HIE system to patients being transferred to SNUBH and ask them whether they would participate in the system. Patients consent by signing with a pen and pad designed for this purpose, then the referral message is sent to the central registry server and the registry at SNUBH. Upon their first visit to SNUBH,
660
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
patients register their visit at the office of referrals and an office staff member enters the patients’ visit into the hospital’s EMR system as per the procedure for all patients. The central registry server checks information from the clinic and against that entered by the hospital, then patient care information that had been entered by the clinic is retrieved for physicians’ review. When hospital physicians first see HIE patients in their office, the system flashes a button on a display screen signaling that the patient is an HIE participant. Physicians click the button and patient care information from the clinic and sourced through the HIE is shown to the physicians through the hospital’s EMR system. The system log confirmed that physicians click the button at over 99% of first encounters to see patients’ care information exchanged in the HIE system.
2.2.
Study design and analytical model
Patients receiving care for acute or chronic conditions at a clinic with referral arrangements with SNUBH may be sent to SNUBH upon their own request or when a clinic physician deems the need for tertiary care. A patient’s care at the hospital is either completed at SNUBH or he/she is referred back to the clinic for routine care when the tertiary care ends. Hospitals in South Korea employ their own physicians and are closed to physicians employed at outside clinics, typically resulting in disconnected care experiences for patients seeking medical attention at both clinics and hospitals. In the conventional system, referring physicians issue a letter, delivered by the patient, that typically contains little information about patients’ care at the clinic, and physicians at clinics do not have access to care information logged at the hospital even after a patient is referred back to them. When a patient deems the need for the full exchange of care information generated at the clinic, he/she asks the clinic to copy medical records and/or diagnostic imaging films, pays fees for the service, and delivers copied information to the hospital. Through the HIE system, the clinician’s letter is transmitted online to the hospital with relevant patient care information, at the clinic. Under the HIE, physicians at the referring clinics can subsequently access the referred patients’ care information entered by SNUBH physicians. However, this aspect of the HIE which has been addressed in Lee et al. [20] as to physicians’ perceptions of the HIE is not included in this paper. In this study, which was approved by the Institutional Review Board of SNUBH, we examined resource utilization during episodes of care at SNUBH. An episode of care starts when a referred patient visits the hospital, and it ends when the care is completed at the hospital or when the patient is referred back to the referring clinic. An episode of care consists of outpatient visits with or without hospitalizations. We used patient classifications to consolidate episodes into groups with homogenous resource-utilization profiles. We classified episodes with hospitalizations using Korean Diagnosis Related Groups (KDRGs), an inpatient classification system similar to the Diagnosis Related Groups used in the United States, and classified outpatient episodes without hospitalizations using Korean Outpatient Groups (KOPGs), which is similar to the Ambulatory Patient Groups used in the United States [22–24]. We used the first 4-digits of the 6-digit classification codes of the KDRGs and KOPGs in this study to
avoid instability in statistical estimations caused by small cell sizes, which collapses age categories as well as complication and comorbidity categories in the classification process. We assumed that resources needed to treat patients’ conditions are homogeneous within a classification code regardless of the patients’ HIE participation status. We assigned a KDRG or KOPG code to each episode and used the codes in the analytical model to control for the difference in the case mixes of the HIE and non-HIE groups. We used a multivariate analysis of variance (MANOVA) model for unbalanced data to estimate the effects of the HIE system on resource utilization incurred during an episode of care at SNUBH following referrals by physicians at local clinics controlling for the difference in the case mixes of the HIE and non-HIE groups, which needs to be analyzed with a linear regression model: Yi = ˛ + ˇHIEi +
j PGij + εi
j
in which i = subscript for episodes; j = subscript for patient groups; Y = log-transformed measure of resource utilization; HIE = 1 if the episode is in the HIE group and 0 otherwise; PGj = 1 if the episode is associated with the jth patient group and 0 otherwise; ˇ, j = coefficients; ε = error term. We log-transformed the dependent variables, and the percentage difference of a resource utilization measure between the HIE and the non-HIE groups was derived with the estimate of the ˇ coefficient: 100 × (eˇ¨ − 1). In addition to resource utilization, we also examined the magnitude of information transmitted to SNUBH from referring clinics. We computed descriptive statistics of the magnitude of information transmitted to SNUBH from referring clinics as well as the measures of resource utilization, and we tested the statistical significance of the differences between the HIE and non-HIE groups with the Wilcoxon ranksum test.
2.3.
Study subjects and data
The initial HIE group included all episodes referred to SNUBH from 35 HIE clinics in which treatment was started at the hospital from June 2009 through October 2010. The non-HIE group included those cases referred by 59 non-HIE clinics (Fig. 1). We assigned KDRG and KOPG codes to each episode, and retained those coded groups that contained 2 or more episodes in each of the HIE and non-HIE groups: 100 KOPGs and 29 KDRGs. The final study dataset included a total of 1265 HIE episodes and 2702 non-HIE episodes. The sizes of the 35 HIE clinics measured by the number of practicing physicians ranged from 1 to 5 with a mean of 1.86, and the sizes of the 59 non-HIE clinics ranged from 1 to 14 with a mean of 1.81; the means were not significantly different. The 3 most frequent specialties among both HIE and non-HIE clinics were internal medicine (34%, 18%), otorhinolaryngology (12%, 16%), and ophthalmology (12%, 14%). Table 1 presents general characteristics of study subjects. The average age, the gender of patients and the type of episode among HIE and non-HIE groups did not significantly differ. Approximately 89% of episodes in this study reflected cases
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
661
Fig. 1 – Study subjects.
consisting of only outpatient care. The 3 most frequently assigned KOPGs among HIE episodes without hospitalizations reflect minor ophthalmological tests and procedures (6.7%), symptoms and signs involving the circulatory and respiratory systems (4.8%), and endoscopy of the upper airway (4%). Similar symptoms were among the most treated in the non-HIE episodes: minor ophthalmological tests and procedures (7.3%), superficial needle biopsy (7%), and otorhinolaryngologic function tests (6%). The 3 most frequently assigned KDRGs among HIE episodes with hospitalizations were lens procedures with large incisions (10.7%), inguinal and femoral hernia procedures without resection of intestine (7.1%), and major thyroid procedures (6.4%). Among non-HIE episodes, they included major thyroid procedures (9.7%), gastroscopy for non-major digestive disease (8.4%), and major retinal and vitreous procedures without lens procedures (6.5%). We studied 9 measures of resource utilization incurred during an episode of care at SNUBH: total charges; costs for medication as well as clinical laboratory, pathological
diagnostic, function, and diagnostic imaging tests; numbers of orders, outpatient visits, and inpatient days. The entire population of South Korea is covered either by the National Health Insurance Program or by the Medical Aid Program, which are both administered by the National Health Insurance Service (NHIS). A fee schedule for each calendar year is set by the NHIS, and we computed charges based on the fee schedule of the year 2009. Charges included both physician and hospital fees, both insurer and patient payments for covered services, and patient payment for non-covered services. Charge data were extracted from the hospital accounting system and data for the number of orders, outpatient visits, and inpatient days were extracted from EMRs using computer programs created for this study. We estimated 9 models, one for each measure, to examine the effects of the HIE technology on utilization of different types of resources. We counted the number of information items in 4 categories: prescription of medication, order and results of clinical laboratory tests, order and results of diagnostic imaging
Table 1 – General characteristics of study subjects. HIE group
Non-HIE group
Characteristic
(n = 1265)
(n = 2702)
Mean age in years (SD) Gender (%) Male Female Type of episode (%) W/o hospitalization (outpatient visits only) W/hospitalization (outpatient and inpatient care)
43.4 (25.1)
44.2 (24.5)
0.65a
611 (48%) 654 (52%)
1237 (46%) 1465 (54%)
0.14b
1125 (89%) 140 (11%)
2381 (89%) 321 (11%)
0.46b
a b
Wilcoxon test. Pearson’s chi-square test.
P-value
662
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
Table 2 – Descriptive statistics and comparison of group means of the study variables for the health information exchange (HIE) and non-HIE groups. HIE group Variable
Mean
Episodes w/o hospitalization (N for HIE group = 1125, for non-HIE group = 2381) 244,049 Total chargesb 6026 Charges for drugb 5072 Charges for clinical pathology testsb 38,112 Charges for pathological diagnosisb 63,728 Charges for function testsb 95,238 Charges for diagnostic imagingb Num. of orders 6.30 Num. of outpatient visits 1.90 Num. of info. items exchanged: drug 1.65 Num. of info. items exchanged: clinical laboratory tests 2.16 Num. of info. items exchanged: diagnostic imaging 0.01 Num. of info. items exchanged: treatment procedures 0.06 Episodes w/ hospitalization (N for HIE group = 140, for non-HIE group = 321) 2123,910 Total chargesb 138,863 Charges for drugb 92,408 Charges for clinical pathology testsb 64,624 Charges for pathologicaldiagnosisb 148,478 Charges for function testsb 85,611 Charges for diagnostic imagingb Num. of orders 53.35 Num. of outpatient visits 4.10 Length of hospital stay 3.21 Num. of info. items exchanged: drug 1.24 Num. of info. items exchanged: clinical laboratory tests 2.34 Num. of info. items exchanged: diagnostic imaging 0.03 Num. of info. items exchanged: treatment procedures 0.05 a b
P-valuea
SD
Mean
SD
281,053 66,462 18,053 62,899 105,601 207,474 6.66 0.72 5.86 7.17 0.14 0.24
274,622 7683 7418 42,059 70,165 107,540 7.00 1.89 0.18 0.95 0.02 0.02
293,267 75,510 21,537 75,045 110,019 204,890 7.13 0.68 0.75 4.69 0.05 0.15
journal homepage: www.ijmijournal.com
Can a health information exchange save healthcare costs? Evidence from a pilot program in South Korea Hayoung Park a , Sang-il Lee b , Hee Hwang c,∗ , Yoon Kim d , Eun-Young Heo e , Jeong-Whun Kim f , Kyooseob Ha g a
Technology Management, Economics, and Policy Graduate Program, Seoul National University, 1 Kwanak-ro, Kwanak-gu, Seoul 151-015, Republic of Korea b Department of Preventive Medicine, College of Medicine, University of Ulsan, 86 Asan Byungwon-gil, Songpa-gu, Seoul 138-736, Republic of Korea c Department of Pediatrics, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea d Department of Health Policy and Management, College of Medicine, Seoul National University, 103 Daehak-ro, Jongno-gu, Seoul 110-799, Republic of Korea e Department of Medical Informatics, Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea f Department of Otorhinolaryngology, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea g Department of Psychiatry, College of Medicine, Seoul National University and Seoul National University Bundang Hospital, 82 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 463-707, Republic of Korea
a r t i c l e
i n f o
a b s t r a c t
Article history:
Objective: Governments and institutions across the world have made efforts to adopt and
Received 20 January 2015
diffuse the health information exchange (HIE) technology with the expectation that the
Received in revised form 6 May 2015
technology would improve the quality and efficiency of care by allowing providers online
Accepted 15 May 2015
access to healthcare information generated by other providers at the point of care. However, evidence concerning the effectiveness of the technology is limited hindering the wide
Keywords:
adoption of a HIE. The objective of this study was to assess impacts of a HIE on healthcare
Community networks
utilization and costs of patient episodes at a tertiary care hospital following referrals by
Electronic health records
clinic physicians.
Health information exchange
Material/methods: We studied 1265 HIE and 2702 non-HIE episodes after physicians referred
Health information technology
patients from 35 HIE and 59 non-HIE clinics to Seoul National University Bundang Hospital
Healthcare cost
(SNUBH) during a 17-month period from June 2009. We examined 9 measures of healthcare utilization and the magnitude of clinical information exchanged in 4 categories. We estimated the savings resulting from HIE use through linear regression models with dummy variables for HIE participation and patient classification codes controlling the case-mix differences between HIE and non-HIE cases.
Abbreviations: EMR, electronic medical record; HIE, health information exchange; KDRG, Korean diagnosis related group; KOPG, Korean outpatient group; KRW, Korean won; NHIS, National Health Insurance Service of Korea; SNUBH, Seoul National University Bundang Hospital; USD, US dollar. ∗ Corresponding author. Tel.: +82 10 5235 0903; fax: +82 31 787 4054. E-mail addresses: [email protected] (H. Park), [email protected] (S.-i. Lee), [email protected] (H. Hwang), [email protected] (Y. Kim), [email protected] (E.-Y. Heo), [email protected] (J.-W. Kim), [email protected] (K. Ha). http://dx.doi.org/10.1016/j.ijmedinf.2015.05.008 1386-5056/© 2015 Elsevier Ireland Ltd. All rights reserved.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
659
Results: The total charges incurred by the HIE group during episodes at SNUBH were approximately 13% lower (P < 0.001), and the charges for clinical laboratory tests, pathological diagnosis, function tests, and diagnostic imaging were 54% (P < 0.001), 76% (P < 0.001), 73% (P < 0.001), and 80% (P < 0.001) lower for the HIE group than for the non-HIE group. SNUBH physicians had access to more clinical information for HIE than for non-HIE patients. Conclusions: HIE technology improved physicians’ access to past clinical information, which appeared to reduce diagnostic test utilization and healthcare costs. The payer was the major beneficiary of HIE cost savings whereas providers paid for the technology. Fair allocation of benefits and costs among stakeholders is needed for wide HIE adoption. © 2015 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
The adoption of a health information exchange (HIE), which is defined as the electronic transmission of healthcare information among healthcare providers, has been slow despite its highly anticipated benefits in healthcare quality and cost [1–3]. The technology enables provider online access to healthcare information generated by others at the point of care, and therefore it is expected to improve the quality and efficiency of care and reduce the operating and administrative costs of healthcare providers [4]. South Korea, in particular, needs this technology because care is fragmented. Hospitals employ their own medical staff and are closed to physicians at local clinics. However, the barriers to technology adoption are particularly high. Hospitals and clinics often compete for patients, with a large share of hospital revenue coming from outpatient care. Further complicating HIE adoption, providers are paid on a fee-for-service basis such that higher utilization of healthcare services generates higher payments for providers. As a result, providers’ financial incentives are not aligned with containment of healthcare costs through the adoption of an HIE. Previous literature has emphasized the importance of effectiveness research for wide adoption and diffusion of an HIE, particularly its value in saving healthcare costs [2,5–8]. Financial incentives that align interests of stakeholders cannot be designed without proper estimation of benefits, which previous studies have indicated as a significant factor in facilitating the adoption and sustainability of the adopted technology [4,9–12]. However, few researchers have attempted to measure the economic benefits of an HIE, and evidence quantified with empirical data obtained from an operational HIE was sketchy [13–18]. Overhage et al. found a decrease of 26 US dollars (USD) per emergency department encounter in an HIE group at 1 of 2 pilot study sites in Indianapolis, Indiana, but no significant savings in the other site [13]. Frisse et al. found that HIE access was significantly associated with a decrease in hospital admissions and utilization of head and body computed tomography as well as laboratory testing in a study of emergency department encounters in Memphis, Tennessee [16]. Magnus et al. [17] and Bell et al. [18] reported improvements in quality of care measures through the HIE technology in HIV care settings. Further studies are needed to strengthen the evidence of HIE technology impact on healthcare cost and quality.
We examined physicians’ acceptance and use of an HIE, patients’ perceptions of the technology, and the costs and benefits of the system as part of a 3-year pilot project supported by a grant from the Ministry of Health and Welfare of South Korea [19–21]. In this report of the study, we assessed the impact of an HIE on healthcare utilization and costs. We estimated cost savings by comparing healthcare utilization of 2 groups of patients receiving similar care but who differed in terms of access to HIE technology.
2.
Materials and methods
2.1.
Study setting
The Ministry of Health and Welfare of South Korea funded a 3-year pilot project to examine the feasibility of using an HIE in South Korea, operate a prototype system, obtain evidence on the clinical and economic impacts of the technology, and explore other issues that may emerge when an HIE is introduced. Seoul National University Bundang Hospital (SNUBH), a 980-bed tertiary care university teaching hospital located 20 km south of Seoul, received a grant from the ministry and launched the pilot project in November 2007. The first version of the system was deployed in June 2008 and was completed in October 2009. Terms and conditions of participation were presented to local clinics and physician practices (hereafter, clinics) with referral arrangements with the hospital, either in recruitment sessions or in visits to clinics on their request. Clinic participation in the project was on a voluntary basis, and 35 clinics participated in the project as of October 2009. SNUBH and participating clinics had been using electronic medical records (EMRs) at the time the project commenced. The HIE system was based on a federated architecture model, and in the final system, exchanged information included patients’ demographic data and health status, including diagnoses with chief complaints, prescribed medications, laboratory results, diagnostic images, duration and content of treatments, care plans, vital signs, history, and summaries [19,20]. Physicians at HIE clinics introduce the HIE system to patients being transferred to SNUBH and ask them whether they would participate in the system. Patients consent by signing with a pen and pad designed for this purpose, then the referral message is sent to the central registry server and the registry at SNUBH. Upon their first visit to SNUBH,
660
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
patients register their visit at the office of referrals and an office staff member enters the patients’ visit into the hospital’s EMR system as per the procedure for all patients. The central registry server checks information from the clinic and against that entered by the hospital, then patient care information that had been entered by the clinic is retrieved for physicians’ review. When hospital physicians first see HIE patients in their office, the system flashes a button on a display screen signaling that the patient is an HIE participant. Physicians click the button and patient care information from the clinic and sourced through the HIE is shown to the physicians through the hospital’s EMR system. The system log confirmed that physicians click the button at over 99% of first encounters to see patients’ care information exchanged in the HIE system.
2.2.
Study design and analytical model
Patients receiving care for acute or chronic conditions at a clinic with referral arrangements with SNUBH may be sent to SNUBH upon their own request or when a clinic physician deems the need for tertiary care. A patient’s care at the hospital is either completed at SNUBH or he/she is referred back to the clinic for routine care when the tertiary care ends. Hospitals in South Korea employ their own physicians and are closed to physicians employed at outside clinics, typically resulting in disconnected care experiences for patients seeking medical attention at both clinics and hospitals. In the conventional system, referring physicians issue a letter, delivered by the patient, that typically contains little information about patients’ care at the clinic, and physicians at clinics do not have access to care information logged at the hospital even after a patient is referred back to them. When a patient deems the need for the full exchange of care information generated at the clinic, he/she asks the clinic to copy medical records and/or diagnostic imaging films, pays fees for the service, and delivers copied information to the hospital. Through the HIE system, the clinician’s letter is transmitted online to the hospital with relevant patient care information, at the clinic. Under the HIE, physicians at the referring clinics can subsequently access the referred patients’ care information entered by SNUBH physicians. However, this aspect of the HIE which has been addressed in Lee et al. [20] as to physicians’ perceptions of the HIE is not included in this paper. In this study, which was approved by the Institutional Review Board of SNUBH, we examined resource utilization during episodes of care at SNUBH. An episode of care starts when a referred patient visits the hospital, and it ends when the care is completed at the hospital or when the patient is referred back to the referring clinic. An episode of care consists of outpatient visits with or without hospitalizations. We used patient classifications to consolidate episodes into groups with homogenous resource-utilization profiles. We classified episodes with hospitalizations using Korean Diagnosis Related Groups (KDRGs), an inpatient classification system similar to the Diagnosis Related Groups used in the United States, and classified outpatient episodes without hospitalizations using Korean Outpatient Groups (KOPGs), which is similar to the Ambulatory Patient Groups used in the United States [22–24]. We used the first 4-digits of the 6-digit classification codes of the KDRGs and KOPGs in this study to
avoid instability in statistical estimations caused by small cell sizes, which collapses age categories as well as complication and comorbidity categories in the classification process. We assumed that resources needed to treat patients’ conditions are homogeneous within a classification code regardless of the patients’ HIE participation status. We assigned a KDRG or KOPG code to each episode and used the codes in the analytical model to control for the difference in the case mixes of the HIE and non-HIE groups. We used a multivariate analysis of variance (MANOVA) model for unbalanced data to estimate the effects of the HIE system on resource utilization incurred during an episode of care at SNUBH following referrals by physicians at local clinics controlling for the difference in the case mixes of the HIE and non-HIE groups, which needs to be analyzed with a linear regression model: Yi = ˛ + ˇHIEi +
j PGij + εi
j
in which i = subscript for episodes; j = subscript for patient groups; Y = log-transformed measure of resource utilization; HIE = 1 if the episode is in the HIE group and 0 otherwise; PGj = 1 if the episode is associated with the jth patient group and 0 otherwise; ˇ, j = coefficients; ε = error term. We log-transformed the dependent variables, and the percentage difference of a resource utilization measure between the HIE and the non-HIE groups was derived with the estimate of the ˇ coefficient: 100 × (eˇ¨ − 1). In addition to resource utilization, we also examined the magnitude of information transmitted to SNUBH from referring clinics. We computed descriptive statistics of the magnitude of information transmitted to SNUBH from referring clinics as well as the measures of resource utilization, and we tested the statistical significance of the differences between the HIE and non-HIE groups with the Wilcoxon ranksum test.
2.3.
Study subjects and data
The initial HIE group included all episodes referred to SNUBH from 35 HIE clinics in which treatment was started at the hospital from June 2009 through October 2010. The non-HIE group included those cases referred by 59 non-HIE clinics (Fig. 1). We assigned KDRG and KOPG codes to each episode, and retained those coded groups that contained 2 or more episodes in each of the HIE and non-HIE groups: 100 KOPGs and 29 KDRGs. The final study dataset included a total of 1265 HIE episodes and 2702 non-HIE episodes. The sizes of the 35 HIE clinics measured by the number of practicing physicians ranged from 1 to 5 with a mean of 1.86, and the sizes of the 59 non-HIE clinics ranged from 1 to 14 with a mean of 1.81; the means were not significantly different. The 3 most frequent specialties among both HIE and non-HIE clinics were internal medicine (34%, 18%), otorhinolaryngology (12%, 16%), and ophthalmology (12%, 14%). Table 1 presents general characteristics of study subjects. The average age, the gender of patients and the type of episode among HIE and non-HIE groups did not significantly differ. Approximately 89% of episodes in this study reflected cases
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
661
Fig. 1 – Study subjects.
consisting of only outpatient care. The 3 most frequently assigned KOPGs among HIE episodes without hospitalizations reflect minor ophthalmological tests and procedures (6.7%), symptoms and signs involving the circulatory and respiratory systems (4.8%), and endoscopy of the upper airway (4%). Similar symptoms were among the most treated in the non-HIE episodes: minor ophthalmological tests and procedures (7.3%), superficial needle biopsy (7%), and otorhinolaryngologic function tests (6%). The 3 most frequently assigned KDRGs among HIE episodes with hospitalizations were lens procedures with large incisions (10.7%), inguinal and femoral hernia procedures without resection of intestine (7.1%), and major thyroid procedures (6.4%). Among non-HIE episodes, they included major thyroid procedures (9.7%), gastroscopy for non-major digestive disease (8.4%), and major retinal and vitreous procedures without lens procedures (6.5%). We studied 9 measures of resource utilization incurred during an episode of care at SNUBH: total charges; costs for medication as well as clinical laboratory, pathological
diagnostic, function, and diagnostic imaging tests; numbers of orders, outpatient visits, and inpatient days. The entire population of South Korea is covered either by the National Health Insurance Program or by the Medical Aid Program, which are both administered by the National Health Insurance Service (NHIS). A fee schedule for each calendar year is set by the NHIS, and we computed charges based on the fee schedule of the year 2009. Charges included both physician and hospital fees, both insurer and patient payments for covered services, and patient payment for non-covered services. Charge data were extracted from the hospital accounting system and data for the number of orders, outpatient visits, and inpatient days were extracted from EMRs using computer programs created for this study. We estimated 9 models, one for each measure, to examine the effects of the HIE technology on utilization of different types of resources. We counted the number of information items in 4 categories: prescription of medication, order and results of clinical laboratory tests, order and results of diagnostic imaging
Table 1 – General characteristics of study subjects. HIE group
Non-HIE group
Characteristic
(n = 1265)
(n = 2702)
Mean age in years (SD) Gender (%) Male Female Type of episode (%) W/o hospitalization (outpatient visits only) W/hospitalization (outpatient and inpatient care)
43.4 (25.1)
44.2 (24.5)
0.65a
611 (48%) 654 (52%)
1237 (46%) 1465 (54%)
0.14b
1125 (89%) 140 (11%)
2381 (89%) 321 (11%)
0.46b
a b
Wilcoxon test. Pearson’s chi-square test.
P-value
662
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 4 ( 2 0 1 5 ) 658–666
Table 2 – Descriptive statistics and comparison of group means of the study variables for the health information exchange (HIE) and non-HIE groups. HIE group Variable
Mean
Episodes w/o hospitalization (N for HIE group = 1125, for non-HIE group = 2381) 244,049 Total chargesb 6026 Charges for drugb 5072 Charges for clinical pathology testsb 38,112 Charges for pathological diagnosisb 63,728 Charges for function testsb 95,238 Charges for diagnostic imagingb Num. of orders 6.30 Num. of outpatient visits 1.90 Num. of info. items exchanged: drug 1.65 Num. of info. items exchanged: clinical laboratory tests 2.16 Num. of info. items exchanged: diagnostic imaging 0.01 Num. of info. items exchanged: treatment procedures 0.06 Episodes w/ hospitalization (N for HIE group = 140, for non-HIE group = 321) 2123,910 Total chargesb 138,863 Charges for drugb 92,408 Charges for clinical pathology testsb 64,624 Charges for pathologicaldiagnosisb 148,478 Charges for function testsb 85,611 Charges for diagnostic imagingb Num. of orders 53.35 Num. of outpatient visits 4.10 Length of hospital stay 3.21 Num. of info. items exchanged: drug 1.24 Num. of info. items exchanged: clinical laboratory tests 2.34 Num. of info. items exchanged: diagnostic imaging 0.03 Num. of info. items exchanged: treatment procedures 0.05 a b
P-valuea
SD
Mean
SD
281,053 66,462 18,053 62,899 105,601 207,474 6.66 0.72 5.86 7.17 0.14 0.24
274,622 7683 7418 42,059 70,165 107,540 7.00 1.89 0.18 0.95 0.02 0.02
293,267 75,510 21,537 75,045 110,019 204,890 7.13 0.68 0.75 4.69 0.05 0.15
