Raymond E. Lenhard, Jr., M.D.

Introduction The Oncology Clinical Information System (OCIS) is a comprehensive computing system that provides organized and complete medical information to physicians, nurses, medical students, and other health care providers at logical times in the patient’s management. Although implemented in a cancer center, the principles that have been used in its development and many of the applications that are in daily use are equally applicable to general medicine. Physicians caring for patients with cancer face common medical problems that are not specific to patients with cancer. Conversely, physicians in other medical specialties are concerned with the same data management and decision-making issues that confront oncologists.

Decisions in medicine should be based on information that is not only reliable, complete, and timely, but also presented in a way that leads to the correct decision. Many models of computer-assisted decision support have been explored over the last several decades. The differences in these systems seem primarily to be how the responsibility for processing data are distributed between man and machine.

In designing an information system that supports medical decision making, there are several philosophical issues that must be decided. First, one must determine the primary unit of concern. For example, one can design a system around admissions, outpatient visits, billing sequences, or patients. In the episodic care of nonreturning patients, there is no loss of information when each treatment sequence is independent. However, for long-term follow-up and continuing care, as one typically finds in a cancer center setting, any distinction between events and data resulting from admissions and outpatient encounters is artificial. Therefore, the OCIS was designed to manage the information about a patient in an integrated unit so that it could report on findings and therapy regardless of episode, location, or source of care.

The next philosophic issue to be resolved involves the kinds of knowledge to be used by the system. In an artificial intelligence paradigm, the goal is to have the necessary knowledge stored in the computer to guide the diagnostic or patient management decision making. Heuristics provide satisfactory (as opposed to optimal) advice, and much of the effort in developing a knowledge-based system goes into the process of collecting and formalizing that knowledge base (an activity called knowledge engineering). For complex medical domains, however, these artificial intelligence applications are still research programs.

If one recognizes, therefore, that it is premature to organize most of the medical knowledge used to diagnose and manage patient care in a knowledge base, then alternative forms of knowledge organization are required. In the design of the OCIS we have divided the responsibility of knowledge management between the physician and the system. All knowledge organization activities that can be described with some precision (i.e., are routine or can be stated algorithmically) are relegated to the system. Examples include the organization of clinical data in medically meaningful ways* the capture of routine therapy clusters and formal protocol sequences, and the association of data displays with clinical situations. Clearly, these system data organization tasks represent functions that should be both natural and designed to assist the provider in applying his or her knowledge. In this case the OCIS exhibits its intelligence by reducing the information overload and presenting the data in forms that help in rapidly finding the best solutions to the current medical problems.

The third philosophical issue to be decided relates to system closure. That is, how closely linked should the system be to the operational process. For example, a laboratory system is closed with respect to orders, result reporting, and charge determination; every order that comes in must produce a result and some charge action. However, the laboratory system is not closed in the care process; its reports may be unread, and there may be no response to abnormal results. Of course, systems with alerts attempt to close this processing loop by demanding a response when certain events are recognized. In the case of OCIS, the goal is for the system to behave in an intelligent manner and complement the physician’s knowledge. This implies that the system must be integrated into the care process so that it can respond appropriately. Thus, it must know something about the patient’s status to produce the most effective reports.

In summary, then, we see that the OCIS was designed to store and organize all patient data in an integrated fashion. It also manages well-understood knowledge to assist a broad range of medical providers in the various aspects of patient care. To be effective, it is necessary that the OCIS be integrated into the care process. Where recommendations are made, it is essential that the OCIS be informed of the resulting action. However, it is equally important that the OCIS not intrude on the process. It should be perceived to be a useful tool. For some providers, there may not even be an awareness of its presence—only a recognition that the Oncology Center always manages its medical records in this way. In these situations the OCIS is a clear success; it invisibly supplies an intelligent data management facility that naturally supports the physician’s decision making.

The remainder of this chapter illustrates how clinical data are related to patient management and how OCIS, designed to meet the preceding philosophical objectives, supports the process. The organization of the automated medical record is described first. It is presented as a layered structure in which the patient’s name and identifier are at the top; below it are the census data that record major events, and below that is the abstract that contains an overview of the medical record narrative. The clinical data provide additional information, and a variety of formats and reports is available to complement the physician’s knowledge. Several clinical examples are given; these illustrate how this intelligent display of data improves patient care. More detailed descriptions of the OCIS tools are contained in the chapters of Part II.

Patient Identification and Census In a hospital or group practice there is considerable cross-coverage of patients by physicians and office nurses. Patients who come to the outpatient clinic or emergency room, or telephone a general number for the practice need to be identified as belonging to the practice group. They can be served best if the person responding to the patient has rapid access to information about their diagnosis and current treatment.

First, the patient must be identified as a participant in that care system. The OCIS census provides the correct name, history number, age, and diagnosis. In addition, the date on which the patient was first seen and the last outpatient visit are immediately available. All hospital stays with admission and discharge data and reason for admission are also displayed. This information is available on line, 24 hours a day, for 56,000 patients. It is an example of administrative information that is collected as a by-product of each encounter and not specifically entered into the system for this purpose. Figure 1 shows a typical encounter screen as it might be used by a registrar, primary nurse, or physician to identify the patient.

The Abstract The abstract is, as the name implies, a collection of pertinent information presented in abstracted form for rapid reference during the course of diagnosis and treatment (Figure 2). An OCIS abstract is completed on all patients at The Johns Hopkins Oncology Center and is available on line, as is all other information, 24 hours a day, 7 days a week.

Based on the data collected on all patients by The Johns Hopkins Hospital Tumor Registry, the abstract contains medical information presented in chronological order in a sequence similar to that used by physicians to describe a patient to a colleague. The demographics are commonly used by care providers to contact the patient or their relatives. The diagnosis section provides ICD-9 coded diagnoses

Figure 1. A typical encounter screen.

and other Tumor Registry information. In addition, the free text report of the pathology or cytopathology diagnosis is collected on each patient as part of the Tumor Registry data entry function.

Following the diagnostic information, a table of chronological events is presented. These data are entered by data coordinators and summarize each hospitalization and each change in treatment. Each entry is a single line of text showing pertinent diagnostic tests, procedures, and treatments. Surgical procedures, such as biopsies and their results, are included, as are major surgical therapeutic procedures. In addition, tests, such as estrogen receptors in breast cancer and the presence of hormone or tumor markers such as CEA, AFP, or myeloma protein, are shown with their numerical results.

Treatments also are shown in the abstract. Radiation therapy is described by site, total dose, and the duration of treatment, but not in such detail as specific field size, dose fractions, or treatment machine source. A typical radiation therapy statement is shown in Figure 2, recording that 5000 rads were given to the right chest wall and 4500 rads to the right supraclavicular area and showing the associated dates of treatment. Chemotherapy and hormone administration are similarly collected, showing what drug >vas given, the dose fraction, and duration, but not cumulative dose. Figure 2 shows that Tamoxifen was administered beginning in June 1987.

The next section in the abstract is a short history and physical examination section with information about the patient’s occupations, toxic exposures, and cancer-related personal lifestyle factors, such as smoking history.

Finally, the abstract provides updated information on which physicians and nurses are responsible for the care of the patient. Both the doctor following the patient at Johns Hopkins and the personal physician and referring physician in the community are listed. This information provides an important link to the patient’s primary physician and shows where follow-up care will be given and follow-up information should be sent.

Inpatient Data Management OCIS was primarily intended to help support information management of complex patients with chronic medical problems who have repeated admissions to the hospital and frequent outpatient visits.

Patients on the acute inpatient nursing unit present a special problem in data management. This is because of the large volume of information that is returned and the way in which that information is collected and reported. Requests for laboratory services are sent in chronological order, but each laboratory test has a different time required for completion and return of results to the requesting physician. In a high-data-volume environment, the physician finds it difficult to recall which tests have been reported and which are still pending. In addition, when the tests results return, they are not delivered in the same chronological order in which they were sent. Physicians spend considerable time organizing this information either on paper or in their memory. Frequently, they are required to telephone several laboratories to get recent information that may not have been reported to them. This manual time-consuming effort is driven by the level of illness of the patient and the urgency for getting information as quickly as possible. As physicians have only a limited amount of time to allocate to each patient, time spent correcting errors of omission and processing data must compete for time spent in direct patient care and family interactions and in making decisions based on available information. OCIS is designed to substitute for this manual effort and to free the physician’s time for patient care and decision making.

At Johns Hopkins and many other hospitals, a common organizational plan is the clustering of patients with similar medical problems on a single inpatient unit. This is done as a logical extension of medical specialization to provide optimal medical care. In the Oncology Center, acute leukemia, bone marrow transplantation, and other specialized diseases are clustered on a single nursing unit. Of course, this is not unique to oncology as renal dialysis units, coronary care units, and other facilities are also designed around specific illnesses. Although this makes the management of patients more effective, using nurse specialists and specially trained medical support personnel who will respond to a medical emergency in a reliable and consistent way, it makes data management more difficult. Not only do these patients have large amounts of infor mation returning to the physician for evaluation, but the problem of assessing the importance of each individual value is accentuated because of the similarity of the patients housed on that unit. Patients in kidney failure clustered on a dialysis unit all have very similar problems, and each has a similar limited set of laboratory results that define his or her disease and treatment status. Gathering a full set of laboratory values, ensuring that there are no missing data, and then weighing this information against other related data and against itself as it changes over time constitute a major source of information overload for physicians.

Figure 2. An OCIS abstract.

An Example A good example of how OCIS was designed to manage the problem of information processing can be illustrated by a clinical inpatient unit containing ten patients with acute leukemia. We shall follow this scenario of acute leukemia management throughout the next several sections. In these illustrative cases dates have been artificially compressed and several weeks of hospitalization are truncated for ease of discussion.

Patients with leukemia have indicators of their primary illness that can be quantified in the laboratory and reported as a numerical value. These markers;are the white blood cell count and the white blood cell differential. These cells, which both define the presence of the disease and quantify its extent, are both enumerated and examined for morphological characteristics. Enumeration generally is done by an automated particle counter linked to a computer. The results are expressed as particles (cells) per cubic millimeter (mm3) of fluid (blood). For example, the white blood cell count (wbc) may be reported as 7000/mm3.

Leukemia is characterized by both increased numbers of cells circulating in the blood and the abnormal morphology of these cells, determined by examining stained fixed blood smears under a microscope. Treatment strategy is based on the administration of medications or groups of medications from different chemical classes to destroy the leukemic cells and their precursors, but leave the normal cell precursors relatively intact. Successful treatment outcome is assessed by a return of normal white blood cells to normal numbers and the absence of all abnormal leukemic cells. To reach this successful outcome (commonly referred to as a complete remission), patients must be supported through at least three major medical complications of the disease and its treatment. These are:

Infection. There is an increase in the risk of infection because of the disease. This risk is made worse by chemotherapy, which further lowers the remaining normal white blood cells, making the patient highly susceptible to infection and at high risk of dying, unless the presence of infection is detected and treated in a timely and appropriate fashion.

Hemorrhage. Because there is a lowering of the patient’s platelet count by both the disease and the chemotherapy, patients are at risk of bleeding because of their inability to carry out normal blood clotting. The support strategy for this complication is the administration of transfusions of platelets derived from normal donors, to provide the necessary numbers of platelets needed for normal clotting. Platelet transfusions are administered daily until the patient’s normal platelets recover.

Complex medical problems. Fluid and electrolyte balance is an example of the difficult medical problems requiring attention to detail and repetitive calculations. A balance must be maintained between the amount of salt and fluids taken in by mouth or by intravenous drip and the amount of fluid that is put out through urine, perspiration, and bowel movements.

Figure 3. Semilog plot of white cell and platelet data.

to Each

Figure 4. Comp flow screen.

of the above demonstrates a need for rapid, accurate, and well-correlated data presented in formats that show related information from tests done in several laboratories displayed relative to each other. Figure 3 shows a typical, relatively uncomplicated, clinical course of a single patient and is described in some detail.

Infection. Just before treatment on August 1, the white blood cell count (w) is elevated to 40,000/mm.3 This is significantly higher than normal (between 4500 and 11,000/mm.3) In addition, the platelet count (p) is depressed to fewer than 20,000 because of the leukemia (normal platelet count 150-350,000/mm3). Two treatments (chemotherapy and platelet transfusion) are therefore shown along the bottom of the plot. The first is chemotherapy for the primary disease to destroy the leukemic cells. This is administered on August 1 (treatment day 1) and is shown at the bottom of the figure as days that Daunomycin (daunorubicin) and Cytosine A (cytosine arabinoside) are given. Neither dose nor schedule is shown on the graph, but only the fact that on that day treatment was administered. This same information is shown in detail in Figure 4, where the actual doses and blood counts are displayed.

Chemotherapy drug doses are frequently calculated on the basis of milligrams per square meter of the patient’s body surface area. OCIS assists in chemotherapy administration by providing a facility for doctors, nurses, and other staff to calculate the patient’s body surface as a function of height and weight. This calculation can be used by the physician to plan the original dose, by the nurse as a verification before administering the medication, and by the pharmacist in preparing the medication for administration. These internal checks lessen the risk of dose miscalculation for medications that are used at the high levels of tolerance at which overdose can be a risk to the patient’s life.

As the administration of chemotherapy is a major event, it is often used as the starting point for measuring the time to the occurrence of many other events. OCIS provides a chronological “event counter,” which can be started at the time of a treatment or other specified event. This is seen at the bottom of Figure 3. It shows that the day on which the two chemotherapy treatments were administered is day “one” (August 1, 1987).

The toxic results of chemotherapy on the white blood cell count can be recognized in Figures 3 and 4. The disease responds as expected with the white blood cell count falling from 40,000/mm3 to 800/mm3 by the eighth day following chemotherapy. This is a graphic representation of a successful response to treatment and shows that the appropriate medication was selected.

As noted above, successful treatment requires major support for infectious disease, bleeding, and fluid balance. On day 11 after chemotherapy, the anticipated infection emerges. This can be seen in the OCIS displays in two ways. First, on the standard white blood cell and platelet count graph, an asterisk is displayed along the bottom line to indicate that the patient has a normal temperature on that date. When the asterisk is replaced by a number, it indicates the number of degrees above 38.3°C, a threshold that requires that the patient be started on antibiotics. Actual temperature values are displayed in Figure 4, and graphically (T) in Figure 5. In this figure another important feature of OCIS is shown, the use of threshold lines. A horizontal line has been drawn across the entire graph at the level of a temperature of 38.3°C. In the medical setting described here, all numbers below this temperature are normal, and all above are abnormal and require immediate antibiotic administration following a predetermined treatment plan. These guides to treatment decisions have high visual impact and help the physician, nurse, and other care providers to separate at a glance patients who are normal from those who have abnormal findings.

Medical response to common clinical problems have been reduced to a standard regimen to help doctors, nurses, and other technical personnel respond rapidly in a preplanned fashion. It is known that all patients with low white blood counts have a risk of infection. Therefore, it is the physician’s job to detect the problem promptly, to weigh successfully the importance of fever in the clinical setting, and to react appropriately. This commonly happens late at night when a full staff of senior physicians is not immediately available for consultation. The algorithm for infection states that when a patient’s white blood cell count is less than 1000/mm3, and the temperature is greater than 38.3°C, then antibiotics must be started, as the patient is presumed to have a life-threatening infection that must be treated immediately. This protocol-driven medical approach has been shown to prevent sudden death from acute overwhelming bacterial infection in this class of patients. Rapid initiation of treatment with antibiotics that are lethal to bacteria is credited as one of the most important support strategies in this disease. This has allowed patients to be entered safely into high-risk chemotherapy treatment protocols that are designed to be curative, but require meticulous attention to detail.

In a clinical setting where there are multiple patients with leukemia in an inpatient unit, each is at a different point in time in his or her treatment course. The sorting and weighting of information are time-consuming manual tasks that are not easily done by the least senior member of the medical staff in an emergency situation. OCIS displays link medical treatment to specific indicators. In this way OCIS

Figure 5. Temperature and white cell plot.

applies the experience and knowledge of senior staff to influencing decisions that are made in a defined clinical setting without requiring senior staff members to be physically present. This use of routine support plans is part of the goal of OCIS and has proved to be one of its most successful features.

In this illustrative case antibiotics were begun on day 11 when the patient’s temperature rose to 39.0°C. The correctness of the selection of antibiotics and the timeliness of the decision were verified by a prompt return of the temperature to below the fever threshold level. Antibiotics were then continued until both the patient remained free of fever and his white blood cell count (composed of normal rather than leukemic cells) returned to safe levels above 1000/mm3. Then, antibiotics were carefully withdrawn. The patient’s temperature remained normal and the risk of acute sepsis was judged to be over.

Hemorrhage. Platelets are the necessary normal blood component for allowing blood to clot to prevent or stop bleeding. Returning to Figure 3, one can see that the patient had a platelet count on admission of 24,000/mrh3. Our experience has taught us that the risk of hemorrhage rises above an acceptable level when the platelet count is less than 20,000. Therefore, on this graph, a second limit line is drawn at 20,000 using the scale on the right of the graph to assist in blood product support. Blood Bank transfusion specialists monitor the status of all patients being treated for any cancer in the Oncology Center and, by following the slope and direction of the plot, they can anticipate the need for transfusion and have preplanned blood products ready for transfusion of patients. Using OCIS, the Blood Bank has developed a creative prospective inventory and donor management program to provide high-quality and timely platelet transfusion services, while minimizing operating overhead and the need to respond to unexpected emergency demands. As shown on the graph, on August 1 the patient’s platelets were below 20,000 and platelet transfusions were correctly administered, resulting in a rise of the platelets above 20,000 on the following day.

The administration of life-saving human-derived platelet products that are in short supply (as well as the quality control review of our support system) can be monitored by senior staff members using this tool. Each transfusion is shown as a number showing how many units of platelets were given. This also is presented in tabular form in Figure 4. As the patient’s leukemia improves, the platelet counts advance toward normal. The fact that the counts are maintained spontaneously, without the need for additional platelet transfusions, is evidence for the success of the treatment.

Management of Fluid Balance. The third support problem to be discussed is the management of fluid balance. This problem is common to the management of heart, liver, and kidney disease, and it serves as a good example of the generalization of OCIS applications to the management of a broad range of medical problems.

The related data in the management of this problem are patient body weight, amount of fluid taken in as oral and intravenous fluids, fluid lost in urine, bowel movements and from the skin, and proportions of salts taken in and lost. These relations change with activity and temperature, and both foods and medications may contain large amounts of salt. In the ill patient, the normal mechanisms for regulating this complicated equilibrium are frequently dysfunctional. The physician needs to help manage these factors and relies heavily on measurement of the daily weight and the input and loss of fluids. Graphic representation of these relations are helpful. On such a plot, a line can be drawn from the baseline weight and used to show change up or down from that number. Fluid and salt equilibrium should correlate with body weight. The persistence of an input/out-put imbalance, with more fluid taken in than is lost, can place a strain on the heart and lead to accumulation of fluid in the lungs and resultant respiratory distress, an emergency that requires prompt action.

Sodium salt is closely linked with fluid balance in this medical problem. Salt intake contributes to fluid retention and its balance must also be calculated, adding further to the complexity of this management problem. With the amount of sodium salt administered to the patient coming from a variety of sources, such as food, medications, and intravenous fluids, the calculation of sodium input may be difficult to obtain. The most important measure may be a global evaluation of the dynamic relationship between the patient’s weight and fluid input/output balance. Further sophistication of this measurement, to provide a more accurate clinical management plot, can be built into the system in models that account for insensible loss of fluid through the skin relative to patient temperature and loss of sodium via other body fluids.

Assessment of Clinical Data—Implied Intelligence Laboratory and clinical data are usually not assessed as abstract numbers in the practice of medicine. Most data must be clustered with related data items to be clinically relevant. The magnitude of difference between sequential values and the rate of change between these values are both critical for medical decisions. A scenario using a kidney function test, such as serum creatinine, illustrates this point.

The chemistry laboratory has reported an abnormal value of 4 mg/dl. This value is well above the normal range (<1.2 mg/dl) and should be reported immediately to the physician as emergency intervention may be needed.

The physician’s responsibility is to respond appropriately to this value. There are several possible clinical settings that would lead to completely different interpretations of this abnormal value.

Two weeks ago the patient was seen as an outpatient and the serum creatinine was 1.2 mg/dl. A rapid increase to 4.0 mg/dl represents a serious emergency. The patient needs to be seen immediately and admitted to the hospital for acute medical management.

Two weeks ago the patient’s creatinine was 4 mg/dl. The patient has chronic kidney failure and nothing has changed over the last two weeks. The result of 4 mg, although abnormal, is stable and represents a chronic condition calling for no new medical treatment. The patient does not need emergency care.

Two weeks ago the serum creatinine was 7.5 mg/dl. Over the last two weeks medical management has succeeded in decreasing the creatinine to 4.0. This represents not an emergency, but a sign of improvement and successful therapeutic response to current management. In this instance the result viewed by the laboratory as an emergency is actually an indication that the current treatments are successful.

A laboratory value, therefore, cannot be considered an absolute number and cannot be evaluated without knowing the last value and the rate and direction of change between these numbers. Clinical information systems should take these simple principles into consideration. Modern automated systems should include this level of “implied intelligence.”

The OCIS organizes the patient data so that they may be viewed in medically meaningful presentations. Examples of groupings that present data from more than one database include:

Serum calcium displayed with the serum albumin so that bound and unbound serum calcium can be considered.

Cerebral spinal fluid glucose, simultaneous serum glucose, and CSF white blood cell count displayed together. This example is also an educational comment from the system for, if one or more of these values is not present in the system, the physician can be made aware that evaluation of the CSF glucose, in the absence of the other two values, may be unreliable.

The use of tabular and graphic information and meaningful groupings are two examples of implied intelligence that appear throughout the OCIS system. Many of the tables and graphs are prepared by specialists on the senior faculty and help to give less experienced doctors and nurses displays of data that not only show the results, but also suggest an appropriate action that might be taken in response to those results. The OCIS developers believe strongly that information systems have a responsibility for displaying data in a way that increases the likelihood of the physician’s arriving at the correct decision in the most timely fashion. Therefore, the expansion of this function is a key goal in all subsequent phases of development.

Outpatients Medicine is increasingly being practiced in an outpatient setting. Although many of the data items are the same, the information-processing environment is entirely different from the acute inpatient setting described above. Patient visits to the outpatient clinic are episodic, with irregular intervals of time between encounters. The physician sees many other patients in the time between outpatient visits, and it is often difficult to recall pertinent facts for any given patient and to assess accurately which laboratory results are changing and how rapidly change is occurring. In addition, many laboratory tests are either carried out at the completion of a visit or requested to be done at the next visit. Consequently,

Figure 6. Plasma cell dyscrasia.

there are long delays between the time an action is initiated and the results of the action are reported.

When the laboratory or radiographic results are reported, correlation with previous values necessitates the retrieval of a chart that has been returned to the file. This process of outpatient results reporting and the frequency of delays in retrieving the paper chart lead to a dependence on the physician’s memory for recall of previous laboratory values and tend to focus the physician on those results that are significantly abnormal. Subtle, slowing changing trends are easily overlooked until the absolute value becomes sufficiently abnormal to call attention to itself. Slowly progressive anemia, weight loss, or an increasing laboratory marker of extent of disease, such as the CEA test, are all examples of this clinical problem.

A good example of the worth of an information system in monitoring long-term trends in chronic disease is illustrated by the protein marker of multiple myeloma. Figure 6 shows the M-protein measured over a "six-month period and demonstrates a gradual worsening of the disease that takes place in weekly or monthly intervals rather than the hourly or daily dramatic changes illustrated previously in the inpatient setting. If early intervention is important in reversing a disease process before symptoms appear or serious organ dysfunction occurs, then long-term trend analysis by either graphic or mathematical means requires an information system that takes into consideration outpatient events and integrates inpatient and outpatient data. It also emphasizes the need for prospective reminders for collection of information at appropriate intervals, such as routine follow-up tkpap” smears, chest x-rays, or measurement of a blood test marker of disease. Modern information systems need to support the practice of anticipatory medicine by showing early trends that can be modified by outpatient care, thus avoiding emergencies that could have been anticipated by the careful analysis of trends of multiple variables. This approach is medically important but difficult to achieve in an outpatient setting that relies exclusively on a paper record. As outpatient medicine becomes the standard of care, outpatient information systems are becoming a necessity.

Research Support A by-product of the OCIS system is the support of clinical research. All information collected for medical management can be used both for prospective as well as retrospective analysis. The data are easily accessed and organized into treatment groups. The use of this information for research also emphasizes the need for careful control of the quality of data collection and enhances the credibility of the system for clinical care. The system has been invaluable for detecting unexpected toxicities by collecting individual data and grouping them into research clusters. Such a grouping can help the investigator to recognize an abnormality that occurs in several patients at a specific period of time after a new treatment is given, instead of interpreting it as an unusual event in an individual patient.

To support this level of analysis, OCIS allows down-loading of information into a clinical research database for further statistical analysis using other computers and statistical tools. It also assists the Cancer Center Research Office, which collects and enters into the general system dates of entry for a study, completion of the study for a particular patient, and assessment of results of treatment. By using electronic mail, notification and updating of information between the clinical areas and the research office are facilitated. Investigators are able to search the database and derive lists of patients who are in their studies and to monitor centrally the collection of information and its quality. The database can be searched periodically by the study chairman to monitor the progress of patients in the study and to determine whether appropriate testing was done by the staff collaborating on the study.

Research reminders can be linked to the appointment system. Much clinical research is directed by prospective protocols. In this environment a group of patients to be studied is defined, treatments to be studied are described, and a set of parameters to be measured to assess the effects of the treatment are selected. These need to be collected at predefined intervals to allow analysis of similar values collected at similar intervals from a time of intervention. The data collection system can be supported by a simple scheduling system, and compliance with the plan can be monitored by a patient query shortly after the scheduled visit to ensure correct and complete collection and to detect errors in a timely enough way to allow for corrective action to take place. If a well-directed medical information system is in place, the additional costs of clinical research can be minimized and the power of the entire system augmented by incremental additions of special research information to the basic medical data set, providing additional benefit to the clinician while serving the needs of clinical research.

Summary OCIS was originally designed as a medical support system functioning in the environment of a cancer center, but our experience with it is generalizable to the management of medical problems seen in a community hospital. The system’s success is due in large part to the accuracy, timeliness, and relevance of the information contained. It is important that clinicians perceive that it provides valuable information and that by using OCIS, the quality of medical care is improved. With the focus of medical care moving from the inpatient to the outpatient setting, OCIS becomes an increasingly powerful tool for integrating medical management of inpatients and outpatients. This patient-oriented approach can limit the cost of medical care by supporting planned collection of information before a patient is admitted to the hospital and carrying inpatient information forward into the outpatient clinic for continuity of care. OCIS is a good model for hospital systems of the 1990s.