Risk Management and Measuring
Productivity with POAS - Point of Action System -
Masanori Akiyama M.D.,Ph.D.
Center for eBusiness, Massachusetts Institute of Technology Sloan School of Management, Cambridge, MA, USA, Department of Medical Informatics, International Medical Center of Japan,
Abstract— The concept of our system is not only to manage material flows, but also to provide an integrated
management resource, a means of correcting errors in medical treatment, and applications to EBM through the data
mining of medical records. Prior to the development of this system, electronic processing systems in hospitals did a
poor job of accurately grasping medical practice and medical material flows. With POAS (Point of Act System),
hospital managers can solve the so-called, ―man, money, material, and information‖ issues inherent in the costs of
healthcare. The POAS system synchronizes with each department system, from finance and accounting, to pharmacy,
to imaging, and allows information exchange. We can manage Man (Business Process), Material (Medical Materials
and Medicine), Money (Expenditure for purchase and Receipt), and Information (Medical Records) completely by this
system. Our analysis has shown that this system has a remarkable investment effect – saving over four million dollars
per year – through cost savings in logistics and business process efficiencies. In addition, the quality of care has been
improved dramatically while error rates have been reduced – nearly to zero in some cases.
Index Terms— POAS (point of act system), hospital management, ERP (enterprise resource Planning), financial
management, CORBA (Common Object Request Broker Architecture
There has been a tendency in medical care to give low priority to management processes and the improvement of efficiency; medicine has been regarded as a sacred area exempt from such changes. However, in September 2001 the Japanese Ministry of Health, Labor and Welfare made public a draft plan of medical system reform because of the need to seriously review the country's medical services. This was brought about both by the harsh economic conditions existing after the collapse of the asset-inflated bubble economy in the early 1990s and the aging of society accompanied by declines in the birthrate. The plan, which not only visualizes reform of the medical insurance system but also pictures an ideal system of medical care for the future, is a comprehensive draft for institutional reform in Japan. In concrete terms, the plan calls on medical professionals to respect their patients' points of view and allow patients to take responsibility for decisions regarding their own care; to improve the environment within which information is supplied; to provide high-quality, efficient medical care; to improve the quality of medical service and regional medical care security; and to introduce the use of information
Masanori Akiyama M.D., Ph.D is with Center for eBusiness, Massachusetts Institute of Technology Sloan School of Management and the department of Medical Informatics / Internal Medicine, International Medical Center of Japan. E-mail:email@example.com, firstname.lastname@example.org
systems in providing medical services. The point of these suggestions is to foster respect for the options chosen by patients, to provide the information necessary for informed decision making, to establish a system that provides high quality, efficient medical service and to build a foundation for public confidence. Because of these proposals, economic efficiency in medical care is becoming an important public issue. In this context, information technology (IT) can serve as a helpful tool. When the improvement of efficiency is stressed, the quality of medical care may tend to be sacrificed. We have developed a system that, utilizing IT, can accurately calculate costs in a bid to maintaining a balance between efficiency and quality. At the same time, the system can also be used as a yardstick for the measurement and improvement of efficiency. 2 MATERIALS AND METHODS
2.1 POINTS THAT NEED TO BE ADDRESSED
The traditional hospital information system (HIS), built by connecting order entries and the medical clerical system, takes in information about orders and outputs medical payment requests via a medical accounting system, which is actually a payment system. However, this kind of system has the following problems:
• Although physicians are supposed to enter correct payment information, the
information is often incomplete (occurrence of uncollected balance).
• The data terminals within divisions and those at the HIS are not integrated. As a result, duplicate entries are required, resulting in unnecessary extra work.
• While data held in the HIS can be sent to the medical financial system, divisional data necessary for payment cannot be entered due to inconsistencies in the master system. • It is difficult or impossible to search the information held by the medical financial or divisional systems via the order systems.
• A most important problem is that the existing systems have been used primarily for preparing medical payment requests. As a result, data on clinical activities, which have nothing to do with medical insurance, are not received (and could not be handled anyway) by the existing systems.
In these circumstances, when certain expenses are not covered by medical insurance, it has not been possible to make accurate assessments of expenses for materials and personnel through cost calculations based on the data held in the medical financial systems.
2.2 OUTLINE OF THE SYSTEM
To deal with these problems, we have designed a three-tier model . The middle-tier application server is located at the center. We use a Common Object Request Broker Architecture (CORBA) on this application server. A standardized middleware server links all the components of each system to one another. The role of the application server is to mediate among the components of the various systems. Data and the events generated by medical activities, which take place in different components of the various systems, are sent to the application server. The original data itself is not transmitted; rather it is registered for management purposes in a repository. Queries for system data are made to the application server, not to the server of each division. The application server then collects the required data from the appropriate divisions, and sends it to the client that requested it. Using this ―wrapping‖ technology one can connect specialized legacy-based systems
which are customized for each corporation or hospital. The International Medical Center of Japan has integrated it’s existing medical financial systems by routing them through the application server and the CORBA middleware .
With the use of three new functions, the collection of data, secondary use of data and improvement of the precision of data has become possible. First, the Clinical Data Repository (CDR) is a large-capacity database that manages problem-oriented data structures and houses all clinical data so that clinical records can be accessed. Data not housed in any other component will be maintained in the CDR. All system data is stored in the CDR in order to guarantee that all data can
be accessed from the clinical front line. Secondly, the Act Management System (AMS) has made it possible to support decision-making and manage work on a knowledge basis. The result is that the guidelines and protocols of clinical studies can be executed and managed. The AMS also records all changes in the condition of data, and all accesses to clinical data. This feature can be utilized to discover patterns of use by improving guidelines or recording diagnostic processes by analyzing detailed access logs for the systems. Thirdly, the Resource Management System (RMS) manages all the system resources that are normally available to a corporation. It can keep track of people and organizations – actors – connected to each system, fixed assets and equipment, and such resources as pharmaceuticals, film stock, contrast media and meals. Information obtained from the AMS can be invaluable when used for accurate and efficient distribution of resources. Each divisional system manages data that has resulted from that division and its clinical work processes. Each division manages and preserves detailed data, including its reports, and provides only the ―outlines‖ of the data to the application server. Thus, the actual data are not sent to or preserved in the application server. Since only outlines of data are held in the central application server, the volume of data stored there will not increase dramatically. Each client communicates with the others via the application server, and a graphic user interface (GUI) is provided for each occupational category.
2.3 TECHNOLOGIES USED FOR THE SYSTEM
The system was built using state-of-the-art technologies such as CORBA and Java . CORBA is used in the mechanism for data transfer and event distribution. We made a standardized interface using an Interface Definition Language (IDL), which was established by an object management group (OMG) to secure portability, extensibility and scalability of the components in the system. The GUI clients are implemented in Java. We used Extensible Markup Language (XML) to record variable length data. Document data is exchanged between clients and the application server. Meanwhile, CORBA Objects are exchanged between the application server and other components. The application server assembles and resolves XML documents obtained from sources in various divisions.
Using CORBA, an application server is implemented as an integrating system to link the servers in the endoscope division, the pathology division and the wrapped, legacy-based medical accounting system. It is possible to search and browse using the database on local area network (LAN) terminals. Orders, images, reports and the medical financial system are all integrated (Fig.1). MIE Equipment
Endoscope Prescription Injection Radiation
therapy Procedure, Health care MIE surgery Equipment Radiation
Respiratory function Physiology
Act information is entered on PDA Operations control PDA system system Activity based accounting
Billing Order entryElectronic Management account chart
Fig.1 Outline of ERP system of IMCJ
2.5 CALCULATING MEDICAL CARE COSTS
Calculating medical care costs, which had posed difficulties that needed to be resolved, has now become possible. POAS, which stands for the Point of Act System, is a design feature of this comprehensive medical information system. Its characteristics are as follows. 1. Information on all medical activities is collected as detailed data at major ―action‖
points, from the time orders are issued on through to their implementation.
2. The system is organically linked to various medical devices, such as medical diagnostic instruments, X-Ray equipment and equipment in the pharmaceutical division. It records information about medical activities, and their results, in a general-purpose database in various forms such as images, numerical values and text.
3. It uses a general-purpose data description method (AML) that enables flexible incorporation in response to advances in IT technologies.
4. It has a data warehouse structure, which collects and permits the analysis of detailed data at the level of individual medical activities.
5. It helps prevent medical errors – including mistakes at the stage of implementation – by
making it possible to cross-check such data as patient identification, ongoing medical activities, medicines to be used and what personnel carry out the medical activities, each time an activity is executed.
6. It can be used to calculate profits and costs, based on orders. It will total them by medical fees, sectors or patients. These figures can be utilized as management in-formation. 2.6 MECHANISMS FOR DATA COLLECTION
Data on medical activities at the points of action listed below can be collected centrally by direct connections to the order systems and the medical equipment in each division. Order is input, received, changed or cancelled, implemented (contact is made with the accounting section), and completed.
2.7 STRUCTURE OF ORDER ITEMS
Necessary units of data recorded by the system, based on the idea of 6Ws and 2Hs , are as follows: Who-
the implementer (the person who placed an order, or the person who carried it out); to Whom - the patient; How - medical activities and changes in them; What - materials used (pharmaceuticals, medical materials and others); How Much - amount of materials used and the number of applications; For What - name of the disease subject to these medical activities; When - date when the order was placed, when it was implemented, and when it was discontinued; and Where - place of implementation (department, hospital ward, and equipment used). We have made it possible to calculate the costs related to each type of disease by entering the name of the disease along with each order.
3.1 OPERATIONAL TRACK RECORDS
The underlying concept of this system is POAS, which enables records of ―who did what to
whom, where, when, using what, and for what reason‖ . In short, real-time input becomes
possible at the point of action. Logs, including inventories, are created. It becomes possible to reduce to a minimum the difference between expenses from medical activities and the amount claimed as lost by adopting the ―accrued basis of corporate accounting‖ concept. In short, the management of divisions and their work, using a corporate financial/accounting system, has become possible by identifying the divisions that are incurring losses. The system operates continuously at the International Medical Center of Japan, handling 100 transactions per second, or more than 360,000 transactions per hour. It has been in continuous operation for four years.
3.2 LINKING THE HOSPITAL INFORMATION SYSTEM (POAS) WITH THE MANAGEMENT INFORMATION
The hospital information system concerning diagnosis and treatment (POAS) and the management information system, centered on accounting, are separate systems. Data collected as described above are compiled at midnight each day in the clinical database and then sent to the management information system. It calculates all costs in the early morning, using batch processing. As a result, management information from the previous day is available by 6:00 a.m.
3.3 POSITIVE MANAGEMENT ANALYSIS
The use of POAS makes possible management analyses based on objective data. The following kinds of analyses can be performed.
3.3.1 PROFIT-AND-LOSS CALCULATIONS FOR MEDICAL TREATMENT DEPARTMENTS/DIVISIONS
A) THE DIFFERENCE BETWEEN THE NEW SYSTEM AND TRADITIONAL “DIVISION COST CALCULATIONS”
Under the old method, the medical treatment division is regarded as the profit center and the central medical treatment division is seen as a supplementary division. The complete personnel expenses of the hospital are distributed across the medical divisions and the central medical treatment division according to the ratio of their payrolls. The hospital's overall expenses are apportioned to the medical treatment division and the central medical treatment division according to the ratio of personnel costs (primary distribution). Then the expenses of the central medical treatment division are distributed to the medical treatment departments in proportion to their medical treatment earnings (secondary distribution), including those from radiology and other examinations.
Under POAS, the expenses of the central medical treatment division are not distributed, but scored as false earnings, called ―in-house earnings,‖ making the division a quasi-profit center.
That is, the central medical treatment division posts appropriate earnings to the medical treat-mint departments for the medical activities they carried out, based on orders. (The medical treatment departments use a method of scoring their expenses as in-house expenses). The earnings and expenses of the medical treatment departments and the central medical treatment division are calculated on the basis of individual orders.
Costs are made clear not only in the medical treatment departments, but also in the central medical treatment division. By comparing earnings with expenses – a since profits and losses
can be calculated – management can also check the appropriateness of these costs.
Defining the expense structures of money-losing divisions reveals divisions and expenses that should be subject to cost reduction.
The extent that an effort for improvement should be made is clarified by revealing the size of a unit’s losses.
It becomes possible to assess the results of management efforts in the medical treatment departments and the central medical treatment division by comparing their track records in a time series analysis.
Using the profit-and-loss calculation of the central medical treatment division, the efficiency of that division can be judged through a comparison of earnings with expenses. It should become possible in the future to establish an annual program to set profit-and-loss targets for each medical treatment department and division.
Forms for the profit-and-loss statements issued by the medical treatment departments and divisions are automatically generated by POAS.
3.3.2 PROFIT-AND-LOSS CALCULATION BY PATIENT CATEGORY
Earnings and expenses by patient category are calculated on the basis of orders. After the shift to a fixed-fee system, costs should also be compared with earnings under the fixed-fee system, not on the basis of orders. This comparison will become the most important source of information for management judgments after the shift to a fixed-fee system. B) EFFECTS
It has become possible to judge the appropriateness of medical activities by comparing earnings and costs for individual hospitalizations.
For patients who are in an acute stage, or patients with catastrophic health insurance coverage, it becomes possible to review the appropriateness of medical activities based on profit and loss, when the amount of variable expenses and targets for improvement are defined. These expenses include pharmaceuticals, medical treatment, materials and examinations. For patients in a chronic state, it is possible to review the number of hospitalization days and the amount of fixed expenses, including hospital ward expenses.
Again, the forms for profit-and-loss statements by patient category are automatically generated by POAS.
3.3.3 CALCULATING COST BY DISEASE
Using the disease group codes, input for each order datum can total cost by disease. It is possible to identify patients who belong to a specific disease group and total their costs. B) EFFECTS
Now that it has become possible to develop statistics on the cost structure of specific disease groups, such statistics have become information sources when the levels of fees under a fixed-fee system are decided. (However, cost structures at other hospitals should be verified when the fee levels are determined, since cost structures differ from hospital to hospital.) 3.3.4 PROFIT-AND-LOSS CALCULATION BY PHYSICIAN
Profit-and-loss can also be calculated for each attending physician or each physician who places an order (the physician in charge). The results of this calculation are referred to when the trends of the medical activities of each physician is assessed, based on detailed medical treatment data. However, it is dangerous to appraise physicians only on a profit-and-loss basis, since not only financial management factors are necessary, but also medical analyses are required for the qualitative appraisal of medical care.
3.4 RISK MANAGEMENT
The difference between POAS and conventional systems is that POAS is not based on orders but on actions. Essentially, traditional systems were expanded versions of the medical accounting systems that were brought into nurse stations and outpatient departments. This means they were only capable of processing orders by day. As a result, these systems can cause time lags of anywhere between 10 minutes and up to several hours, posing a major problem for the medical workplace. To shorten the time-lag to meet the requirements of medical workers at the patient’s bedside -about 2 to 3 seconds- data granularity must be based on single vials. It is important to recognize at the outset that the Medical Affairs Section and the sections responsible for executing actual medical actions require different data granularities. If a system’s granularity were to be based on single items to begin with, its data could be easily compiled to derive the data required by the Medical Affairs Section as well. This is the reason why
conventional systems have not been useful for improving productivity, gathering clinical data or improving management efficiency. While manufacturers conduct production control on their drugs and medical supplies by single types, by the time these products reach the hospital through the wholesaler, they are batched together into units of boxes or purchase orders. As a result, conventional material flow systems process these items by the shipping slip and not by single types. Even if these products were checked by shipping slip or per day, once an accident occurs, it would be too late to prepare electronic medical charts. To prevent accidents, these products must be controlled as single items from the outset. When the shipment is received, POAS controls these items as single types, not by shipping slips. This helps prevent accidents since it allows hospital operators to implement the same level of quality control as the manufacturers.
According to a survey of injection prescriptions previously conducted at the International Medical Center, changes were ordered for 20% of these prescriptions at one time or another. Since then, the average hospital stay has been halved to 15 days, and we used POAS to calculate the rate of injection instruction changes for a one-year period ending October 2004. Changes were ordered for 24% of the ―injection prescriptions‖ between the time they were issued and the ―injection mixing‖ step. In addition, changes were ordered for 15% of the instructions after ―injection mixing‖ (Fig.2). This shows that changes were ordered for a total of about 40% of the instructions. These changes should have doubled the amount of work for nurses and pharmacists, but their actual workloads did not increase. There was a reduction in nurse overtime and the number of accidents fell to zero. Similar improvements were seen at the Morioka Red Cross Hospital after they began using POAS. This was because automation eliminated tasks such as filling out and transferring slips, which previously took up most of the nurses’ time.
Action entryPerform Issue injection mixed injectionprescription Neither No change to route cancellations speed About 37 thousand/monthAbout 37 thousand/month nor changes 454 orders/day 570 orders/day (802 Rb/day) IV drip bottle
Ward / Bedside Outpatient (HIS) (Mobile terminal) Nurse station (HIS) 750 orders/day 1006 Rp/day 2,329 drugs/day 570 orders/day (1006 Rp/day) 1,770 drugs/dayChanges made to route speed Changes made to route speed 116 orders/day 116 orders/day ??? Canceled or changed orders Canceled or changed orders (204 Rb/day)(204 Rb/day)180 orders/day 180 orders/day (318Rp/day) (318Rp/day)???
There is a possibility of misadministration of about 40% There is a possibility of misadministration of about 40%
if the change of order is not communicated in real-time. if the change of order is not communicated in real-time.
Fig.2 The effects of making injection action entries (calculated from performance data)
This table shows a map of nurses’ actions from midnight to midnight. POAS also records nursing and care procedures. The logged records are 400 thousands / month, then about 80 million logs and 18 million records accumulated over two years. We can see that a variety of workloads are
concentrated in the 9:00 AM to noon timeframe. This is because the morning shift starts at 8:30 for types of work. Most of the important medical actions are carried out before noon and 40% of the prescription instruction changes also procedures during this time. This is a hazard-prone timeframe that produces the most accidents and incurs the most wasted.
8.0Fig.3 Basic analysis: From the frequency distribution of variables
The number of check actions and the error rate have a slightly negative correlation. 7.0The smaller the number of injections a nurse performs that day, the higher the alarm rate. 6.0
Coefficient of correlation ?，0.6 5.0
Number of times mixed injection was checked Fig.4 Comparison of the number of times mixed injections were checked and error rate (%)
(by different time segments)
The grayed area (upper left) of this graph shows the nurses’ total workload. The bar graph with the red frame shows the frequency at which a nurse triggered an alarm and is scaled to proportion. There were a large number of alarms in April, May and June. Obviously, none of these resulted in accidents since alarms were triggered. Since the workload is constant throughout the year, this increase is likely to be due to new nurses joining the workforce in April (Fig.3).
Injection accidents are most likely to result in personal injury. Therefore, here we analyze the causes of injection alarms. The horizontal axis shows the total number of injections performed and the vertical axis shows the alarm rates (Fig.4). Each point corresponds to a single sample with a duration of 30 minutes each. The values are for the entire hospital for a period of one year. It shows that alarm rates were lower during time segments in which a large number of injections were performed and were higher during time segments in which fewer injections were performed. This indicates that accidents do not necessarily occur because nurses are busy. Fig.5 shows the number of errors and error rates in 30-minute increments. You can see that the alarm rate increases immediately after a shift change. Additionally, you can see that the execution of instructions that were specified for 6:00 a.m. were scattered over several hours between 4:30 and 7:30. Conventional electronic medical chart systems will show these as being administered at 6:00 a.m. and there will be no way of getting a picture of the actual situation. With the POAS system, users can not only track this information, but also analyze the 8.01,200effectiveness with pharmoco-kinetics and blood kinetics, as well as efficacy for different administration times. The height of the curve becomes progressively lower for the second, third and fourth injections, and the dispersion becomes more evident. This is likely due to the fact that the first injection is started at around 6:00 a.m. and the second and later injections are IV 7.0replacements performed in response to nurse calls. The important point here is that the 1,000timeframe before 10:00 a.m. is potentially extremely hazardous. This is the timeframe during which powerful drugs are used the most often, with a corresponding spike in the number of 6.0incident and accident reports.
Time segments with higher alarm rates become even clearer when seen in 30-minute increments. 時間帯別 エラー発生率?年間?800
5.53.04.75.05.04004.99904.74.7Error rate(%)4.5Number of errors4.54.57374.32.04.26914.14.14.06494.04.03.93.93.85715563.75435373.72003.63.55003.53.53.44634624583.43.44453.34421.04384203.34153.23733683653633603573543.13433.03293233053012972942922.82602382292222.61811732.40.002.3124125120880:0062 2.047463333400:30エラー件数エラー発生率(%)1:001:302:002:303:003:304:004:30 5:005:306:006:307:007:308:008:309:009:3010:0010:3011:0011:3012:0012:3013:0013:3014:0014:3015:0015:3016:0016:3017:0017:3018:0018:3019:0019:3020:0020:3021:0021:3022:0022:3023:0023:30