Proceedings of the 2005 Systems and Information Engineering Design Symposium Ellen J. Bass, ed.

Transcription

1 Proceedings of the 2005 Systems and Information Engineering Design Symposium Ellen J. Bass, ed. ANALYZING THE PATIENT LOAD ON THE HOSPITALS IN A METROPOLITAN AREA Barb Tawney Systems and Information Engineering School of Engineering and Applied Science University of Virginia Charlottesville, VA ABSTRACT This research documents the patient load trends on the Richmond metropolitan hospitals. The work is based on applying systems and information engineering analyses to data that hospitals routinely collect and store for billing their patients. The research is motivated by two recent events in the Richmond VA metropolitan area. On January 9, 2001, and on January 8, 2003, all of the metropolitan hospitals reached their maximum patient capacity in rapid succession. During these events, patients inbound via ambulance or rescue squad to a hospital that was full, were diverted to a hospital with remaining capacity. Patient data for 2000, 2001, 2002, and 2003 for metropolitan Richmond hospitals are analyzed to characterize the patient load on these hospitals. This type of data analysis has not been reported previously in the literature. 1 INTRODUCTION Hospitals are facilities that provide both urgent and specialty medical care. The hospital staff may provide diagnosis and treatment for those persons who arrive at the facility and then release them back to the community. Some patients may need further treatment that will require the person to stay within the hospital facility for a period of time. Most hospitals provide emergency medical care in a separate department where the incoming persons seeking medical care are evaluated by triage staff. Following the triage evaluation treatment of the person s medical needs is administered to on an assessed priority of needs basis. In the U.S., we have come to expect quality care on an as needed but immediate basis. We are not accustomed to denial of medical care (diagnosis and treatment) even on a temporary basis. 1.1 Background Hospitals admit patients who arrive from three different pathways. One source of patients arriving for hospital admission is the hospital s emergency department (ED). Patients arriving at the ED usually need urgent care. The volume and needs of the patients arriving at the ED vary. Figure 1 shows the flow of the ED patients and hospital inpatients. Some persons may use the ED as a routine physician s office for a cold or flu. Uninsured persons may use the ED as their main source of medical care. A person injured in an automobile accident may need to be diagnosed and treated in an ED. A person with severe injuries may need to become an inpatient in the hospital for diagnosis and treatment. A second source of patients being admitted to the hospital is via scheduled walk-in by physicians. These patients may be scheduled for tests or procedures such as surgery. The admitting physician has previously scheduled the patient s arrival with the hospital. This patient may be treated either as an admitted patient who will be required to stay in the facility overnight, or as an outpatient who is diagnosed and treated in a comparatively short period of time without being assigned to an overnight bed. A third source of arriving patients is the patients arriving from other hospitals. Patients being transferred from other hospitals may need specialty care not available at the other hospital. The transferred patient may need urgent, high level care such as trauma care. All of these patients compete for the available resources at the admitting hospital. Each patient who will receive treatment that extends overnight will need a bed. Each patient also needs staff and other resources devoted to caring for him. Generally, a hospital schedules staff and stores supplies based on the number of planned patients. When the number of patients needing service exceeds the planned capacity, then patients are turned away or treatment must be provided in less than ideal conditions. Internal hospital gridlock may develop. A hospital has an absolute maximum capacity of beds. The number of beds known as the maximum capacity is licensed by the state. Generally, most hospitals operate at a lower number of beds. The total number of beds being used in the hospital will be referred to as operating capacity. The number of beds within the operating capacity dictates the staffing of the hospital, the supplies that are kept on hand, and even the operating capacity of food services. It is this capacity that the hospital administration uses for its forecasts of revenues and expenses. The time to increase the patient capacity from the operating capacity to a higher level capacity nearer its absolute capacity is at least thirty days and in some cases may be sixty days or longer.

2 911 Call E.D. Presentations Call Center Hospital Policies Hospital Administration Pre-Hospital Patients Ambulance Walk - In Physician Referred Triage Directives Emergency Department Directives Admitting Clinical Units Morgue Hospital Disposition Other Hospitals Home Extended Care Facilities Non Hospital Disposition Figure 1: Hospital Patient Flow [Bennett, 2003] On two recent occasions, the hospitals in the metropolitan Richmond, VA, area reached their operating capacity limit. The ED s were saturated with patients. The ED s could not accept additional patients via rescue squad, or ambulance, or any other manner of public transport. All of the hospitals were on diversion. Diversion means that the hospital has reached or exceeded its operating capacity and that transported patients are sent to other hospital facilities to receive care. A diversion at a hospital translates to a denial of care at that hospital for transported patients. Patients who arrive at the ED via private transportation and walk into the ED may stay (by law), but their waiting time may be lengthy. Patients who are being transported via rescue squad or ambulance may be sent out of the area for care if all other local hospitals lack available capacity during a diversion event. From the Richmond metropolitan area, the next nearest medical facilities are an hour away by ground transport. Again, the waiting time for patients to receive medical attention during a diversion may be lengthy whether they are being transported out of the area to a hospital, or whether they are waiting within the hospital as just one of many persons waiting to be diagnosed and treated by the overloaded staff. One exception to denial of medical care during a diversion is trauma care. Trauma care must continue to be provided for all those persons who need it. Each of the trauma patients must be evaluated for their level of needs. The hospital then provides that diagnosed level of care. Because of the lack of an available bed, a patient may be treated in a hallway, or a trauma patient needing care in an Intensive Care Unit (ICU) may be treated in the ED until an ICU bed becomes available. When an ED bed is occupied by a patient that is waiting on an ICU bed the ED bed is unavailable to an incoming ED patient. The level of staffing required to treat an ICU patient in the ED is greater than a typical ED patient so the workload on the ED staff is increased. The workload on the ED staff when it is operating at or near capacity is demanding and stressful. 1.2 Prior Work The Executive Master s students who graduated in 2002 from the Department of Systems and Information Engineering at the University of Virginia,[Cohort, 2002] first studied the problem of all hospitals in metropolitan Richmond area being on diversion. The first event of several hospitals being on diversion simultaneously was January 9, The students from their analysis of the 2001 data recommended that the hospitals team together to communicate more effectively. The hospitals made several changes after the January 2001 event. They worked to maintain better situational awareness among the hospitals

3 through improved communication. On the day of the simultaneous diversion, they instituted the plan that had been intended for a mass criticality event where there were many more incoming ED patients that would be seen on a routine day. This plan effectively helped the ED s come back on line to normal operation. A second diversion event occurred on January 8, As on January 9, 2001, the same hospitals within the metropolitan Richmond were on diversion simultaneously. The improved communication helped shorten the second diversion event but did not prevent it. In March of 2003, a web based diversion information system was implemented for use by the metropolitan hospitals. The web-based system is secure and can only be viewed by those with a password, based on a need to know. Each hospital logs its diversion status on when it goes on partial or full diversion and logs their status off diversion when it is again receiving patients. Prior work has also been done on staffing of an ED and hospital. Dr. Ramesh Shukla of VCU Medical Center [Shukla,1985] worked extensively on medical center staffing during the 1980 s and 1990 s. His work focused on staff satisfaction with regard to their working environments. He found that if a nurse knew her team assignment for the coming shift in advance, her level of satisfaction with her job was greater. During the late 1990 s, Dr. Manuel Rossetti of the University of Arkansas and his master s degree student, Mr. Trzcinski of the University of Virginia Department of Systems Engineering studied the physician staffing for the Emergency Department at the University of Virginia [Trzcinski 1999]. Mr. Trzcinski collected and analyzed data from the UVa Medical Center. Using the information from his analysis, he built an Arena 3.0 simulation model of the ED. From the work, he recommended specific physician staff levels for the ED as it was operated during their study. They recommended future study of nursing and technician staffing at the ED level. 2 RESEARCH This research is based on existing data from the Richmond VA metropolitan hospitals. The goals of the research are to analyze the patient information from 2000, 2001, 2002, and 2003 to characterize the patient load on the hospitals for the time frame of the data sets. 2.1 Research Results The data sets that are being used for the trends analyses were extracted from existing hospital billing data. The quality and accuracy of the data, therefore, has already been assured. Billing data is routinely audited by the payer organization such as an insurance company or the government for Medicare or Medicaid payments. No new data was collected for this study. The data is transformed from the native billing format to usable patient load format. The largest data set contains close to 600,000 records and it occupies just over 455 megabytes in a database file. The data has been recoded by day, by week as needed for each individual analysis. The iterative process of this project is to analyze various subsets of the full data set. Inpatient Arrivals/Discharges 14.0% Percentage of the entire Day 12.0% 10.0% 8.0% 6.0% 4.0% Arrivals Discharges 2.0% 0.0% Hour of Day Figure 2: Inpatient Arrivals/Discharges by Time of Day 217

4 The data sets have been analyzed for daily trends, weekly trends, and event trends. The hourly trends for the data set were summarized by the hour of the day for admission and discharges as are shown in Figure 2. As is shown in Figure 3, Saturday and Sunday inpatient loads are less of the overall bed occupancy than Tuesdays, Wednesday and Thursdays. Thursdays tend to have the most inpatients followed by Wednesdays and then Tuesdays. There is a definite cyclical process for the week for patient arrivals and discharges. Other trends that exist in the data set include the trends that surround holidays. Each of the four Thanksgiving holidays track well with the other Thanksgiving holidays in the dataset. Unlike the other holidays of the year, the Thanksgiving holiday always falls on the fourth Thursday of November. Each Thanksgiving holiday in the data set shows a decrease in patient flow starting by Wednesday and continuing through the weekend. The flow rebounds the following Monday without a burdensome increase in the patient flow. Of course, each of us as a potential patient wants to be home for the holiday so it is obvious that elective inpatient admissions are not scheduled during the Thanksgiving holiday period. From year to year the other family holidays such as Fourth of July and Christmas fall on different days of the week. The patient flow for the Fourth of July tracks for July 4 of each year. Christmas holidays compare for each year of the dataset. Patient flow decreases from the 22nd of December until December 26. Due to the New Year s Holiday closely following Christmas, the patient flow during the double holidays of Christmas and New Years is less than earlier in December or in the weeks that immediately follow New Year s. Included in the data sets are the patient diagnoses for each patient. The length of stay for each patient can be calculated. Therefore, the data sets were analyzed for the patient s diagnosis related group (DRG). DRG indicates the medical condition of the patient such as cardiac failure, kidney failure or psychosis. The DRG of a patient indicates the level of services and generally indicates the length of stay that a patient may require. (see Figure 4) There are variations within the level of required services and stay depending on any additional complicating factors that the patient may exhibit. The research results show that a patient who is diagnosed with psychosis DRG (DRG =430) may require a range of services from less than a 24 hour hospital stay to the patient staying in the hospital for as much as 455 days. It is more likely that the patient will require a stay of six days rather than the extreme stay of 455 days. It is useful for the hospital administrator to know the most likely length of stay for the patient. It is also useful for the hospital administrator to know that several patients have required more than 150 days of inpatient care and the longest patient stay was 455 days. Inpatient Load 18.0% Percentage of Total Load 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 12.3% 14.3% 15.0% 15.2% 15.3% 14.9% 12.9% 2.0% 0.0% Sunday Monday Tuesday Wednesday Thursday Friday Saturday Day of Week Figure 3: Inpatient Load by Day of Week

5 Emergency Department Morgue Elective Physician Referred Urgent Inpatients DRG-> treatment time & level of care Extended Care Facilities Home Other Hospitals Figure 4: Relationship of DRG and Patient Length of Stay A patient with unexplained cardiac arrest may require a hospital bed for an average of 2.4 days with the shortest stay being less than 24 hours and the longest stay being 14 days. In the 2002 data set, the patient who stayed 14 days lived to be released to her family while the other 15 patients expired and were released to the morgue at times that varied from an hour or so after admission to the hospital to five days of inpatient stay. The data set has also been analyzed by the inpatient arrivals and discharges for the 24hour period of a day (figure 5). The data analysis shows that the patients are discharged in the mid afternoon instead of the mid morning. The impact of the timing of the afternoon discharge cycle is that there may be patients who are waiting for the beds that will come available once the current occupant is discharged. Should these patients who are waiting for an inpatient bed be occupying emergency department resources, then the emergency department becomes a holding area for an inpatient instead of maintaining its required flow of triage, treatment and discharge from the ED or admission as an inpatient. Failure for the ED patients to clear the ED beds in a timely manner can result in gridlock in the ED. As is shown in Figure 5, the patient load cycle of the ED begins just before 9am each morning and continues until late into the evening. Those early patients who arrive at the ED and who need additional inpatient care may have to wait for the inpatient to be discharged. As the patients arrive at the ED and the patients wait to be discharged from inpatient status, the hospital becomes a hub of activity. Should there be any activity that may lead to a heavier ED flow, then the ED patients may experience longer wait times to complete their ED treatments and longer wait times for those patients who are going to be admitted to the hospital. Figure 6 shows that the EDs experience heaviest patient flow on Mondays. Sunday is the second busiest day. Should there be any complicating factors in the ED on a Monday, then it is more likely that there will be more patients in the ED seeking treatment than can be provided for in a timely manner. There were several external events that occurred during the time frame of the data sets. The events of 9/11/01 are well recognized as influencing world events. The 2001 data set was analyzed for the effects of 911. There were no patient flow changes for the day of 9/11/2001, however, there were less patients seeking treatment on Wednesday, September 12, The patient flow returned to its usual patterns within the week. There was not a noticeable spike as the flow patterns returned to normal. There were two other external events that occurred during the time span of the data set that resulted in a noticeable decrease in patient flow. The first event was on Monday, February 17, It was the Federal holiday Monday and there was a heavy snow/ice event in the Richmond metropolitan area. The event was noticeable in the data trends. The possible cause of the decrease in patients seeking treatment was found in the weather records. The patient flow returned to normal trends later that week without a spike in load that had to be dealt with.

6 Emergency Department Arrivals/Discharges 7.0% 6.0% 5.0% Percentage 4.0% 3.0% 2.0% Arrivals Departures 1.0% 0.0% Hour of Day Figure 5: Emergency Department Patient Arrivals/Discharges by Hour of Day Emergency Department Load 15.5% 15.0% 14.9% 15.1% Percentage of Load 14.5% 14.0% 13.5% 14.4% 13.9% 13.8% 13.5% 14.3% 13.0% 12.5% Sunday Monday Tuesday Wednesday Thursday Friday Saturday Day of Week Figure 6: ED Load by Day or Week The second event was September 18/19, Hurricane Isabel made a direct hit on the Richmond Metropolitan area. There were widespread power outages for both the population and the medical facilities. There were high winds, heavy rains, and much resulting property damage. People were urged by Emergency Management officials to stay home and off the roads. The impact of this weather event can clearly be seen in the patient flow for the time frame. There was a decrease in the patients who sought medical care. Due to the power outages at the hospitals, elective admissions were postponed until they could safely be rescheduled. Each result that is reported in this paper has been created from a separate analysis effort. Each analysis that is reported here is a multi-step individual effort. The data sets thus far have remained stable and performed well in their data base format..

7 3 CONCLUSION The work reported in this paper is a subset of the research that was accomplished for my Ph.D. dissertation. The research is the first documented use of existing hospital data to model the patient load on the hospitals in a metropolitan area. Data analysis using existing data is not currently available to the metropolitan hospitals. The work reported in this paper shows the load trends for the ED s and the inpatient population by the day of the week and the admit and discharge activities by the hour of the day. These results can be used for the administrators to reconsider the in-hospital activities that can positively affect the admits and discharges of the inpatients. The earlier that the inpatients can be discharged in the work day, the earlier those beds can become available to incoming patients who are awaiting. Thus, the wait time for service for those incoming patients would be minimized. The earlier release of the patients who are being discharged should allow for better patient flow and less bottlenecks in the patient admit cycle. R.K. Shukla, Admissions Monitoring and Scheduling to Improve Work Flow in Hospitals,Inquiry 22, pp , Spring AUTHOR BIOGRAPHY BARBARA TAWNEY is a Ph.D. Candidate at the University of Virginia, Department of Systems and Information Engineering. Her research interests include Analysis and Modeling of Health Care Systems. Her special interests include extracting data of interest from large data sets. She may be reached at tawney@ntelos.net. ACKNOWLEDGEMENTS Special thanks go to K. Preston White, Jr., and Robert M. Bennett. REFERENCES Dr. Robert Bennett, SYS581, Lecture 5, Introduction to Health Care Systems, University of Virginia, Fall Cohort Metropolitan Richmond Hospital Diversions: A Systems Analysis and Change Proposal. Project report, University of Virginia Executive Master s Program, University of Virginia, Trzcinski, Gregory F. Optimal Staff and Chart Documentation Strategies for Emergency Medicine, M. S. Thesis, Department of Systems Engineering, University of Virginia, Manuel D. Rossetti, et al., Emergency Department Simulation and Determination of Optimal Attending Physician Staffing Schedules. In 1999 Winter Simulation Conference Proceedings, IEEE, Washington, DC,