The general model of the ICU. 

Context in source publication

Context 1

... department, and a novel flexible bed allocation Ž FBA . scheme that reserves beds for ES patients. The strategies are evaluated on various performance criteria both for ICU patients as a group and for patients delineated by each of four distinct sources from which the unit receives its referrals. The model’s parameters are determined empirically from historical data. An earlier version of the model, which was ex- pressly built to analyse this ICU’s operations, is described in detail in Kim et al. Ž 1999 . , where we confirm its validity for the unit. Our simulations alone led us to infer that ‘‘insofar as there are serious issues relating to the managerial aspects of this particular ICU, these emanate solely from elective surgery’’, and that ‘‘the current ICU capacity of 14 beds is sufficient to handle patients at the current arrival rates’’ Ž Kim et al., 1999, p. 45 . . These inferences and the expressed concerns of the protagonists have prompted the present extension of that earlier work. In this extension, we first refine the basic model by introducing and verifying a new assumption as to the distribution of service times for the ES patients. Second, we explore the implications of various FBA strategies in a multiple-objective setting. Third, we introduce the administrator to the concept of an efficient frontier both as an immediate decision aid and as a visual means to help explain to the physicians and surgeons the preferred choice among those strategies and the status quo. There is a considerable literature on hospital capacity and bed allocation. Some of this literature relies on queuing theory Ž Bailey, 1954; Cooper and Corcoran, 1974 . . Elsewhere, simulation is preferred Ž Goldman et al., 1968; Blewett et al., 1972; Kuzdrall et al., 1974; Kwak et al., 1975; Kao and Tung, 1981; Williams, 1983; Dumas, 1984; Dumas, 1985; Hashimoto et al., 1987; Rakich et al., 1991; Parry and Petroda, 1992; Ridge et al., 1998 . . Simulation has also been widely used in attempts to improve other aspects of hospital performance. Some recent studies in point are McAleer et al. Ž 1995 . , Wharton Ž 1996 . , Klassen and Rohleder Ž 1996 . and Gonzalez et al. Ž 1997 . .Butler et al. Ž 1996 . contains a survey of the literature on operations management in the strategic planning process in hospitals. The present paper adds to that specific literature directed at bed allocation and scheduling in the ICU. Previous research specific to ICUs focused pri- marily on the clinical aspects of their operations Ž Teres et al., 1991; Zussman, 1992; Oh et al., 1993; Bein et al., 1995 . , or used the data from ICU log books Ž Williams, 1983 Ridge et al., 1998 . . The latter data often show only admitted patients and do not detail the number of patients that arrive and are admitted or rejected, on a daily basis. We are the first to develop and evaluate a bed-allocation and scheduling method for an ICU via a validated simulation model that is based on around-the-clock data that were specifically collected by the hospital for capacity planning purposes. We are also the first to incorporate the specific objective of reducing the cancellation of elective surgeries, considered as one of a pair of objectives in a multiple-objective problem. Our principal conclusion is that FBA will greatly facilitate the scheduling of elective surgeries and ES patient care, without adding to the hospital’s costs. The price to be paid is principally in the added queuing time endured by other patients. This is just one of the unavoidable tradeoffs that must be made. It falls to the administrator to evaluate the tradeoffs and determine the system to put in place. As the triage officer, the administrator must also communicate with the various attending physicians and com- ponents of the hospital that are involved in ICU patient care Ž Teres, 1993, p. 600 . . In identifying and quantifying the tradeoffs, and determining the systems that are dominated, the simulation’s output should greatly assist the administrator in the managerial communication and conflict-resolution process. The proffered means for helping to accomplish this is an efficient frontier that is derived from the simulations. The general principles we infer and the approach we take in the inferential process should interest hospital administrators globally, too, since this ICU and its multiple-objective capacity management problems are very much like those faced else- where. We study a multi-disciplinary ICU that receives patients from five sources: ward; accidents and emergency Ž A & E . ; operating theater Ž OT . -emergency; operating theater Ž OT . -elective; and other hospitals. As there was only one hospital referral in the 6-month sample period, we treat the unit as if all of its patients come from the four basic sources. The ICU physicians’ review of a referral’s dossier might take from a few hours to a few days. Qualified referrals from any group join a single queue for a bed, as shown in Fig. 1. Admission is on a first-come, first-served basis. When there are empty beds, all qualifying patients are admitted immediately. When there is no empty bed, a sufficiently recovered patient is given an expedited discharge to a general ward. When the severity of a referral’s illness requires immediate admission, an expedited discharge might become a virtual necessity. If no bed is empty and an expedited discharge is infeasible, the new referral joins the queue; in the case of ES patients, the operation is cancelled and rescheduled. Fig. 2 shows the admission process for Ž OT . -elective patients. Although the administrator is the triage officer in the prioritization process, the physicians have the final say as to admission or rejection. Referrals are not admitted if they die prematurely, if they are judged to be either too well or too ill to qualify for intensive care, or if the ICU is full. Table 1 shows the average admission and survival rates from each source. As Table 1 shows, the patient groups differ in arrival frequency and pattern, admission and survival rates, and treatment times. These differences are especially acute for ES patients. First, their surgeries are scheduled in advance in accordance with a sur- geon’s schedule, given their condition and ICU bed availability. For other patients, the need for intensive care is the unwelcome outcome of a stochastic pro- cess. Second, an ES patient’s medical history and condition is well known. This simplifies the review- and-admission decision. Third, the medical conditions of these patients are usually less severe and require less intensive monitoring during the postop- erative recovery period than those of their ICU roommates. Fourth, the average ES patient requires less ICU time than does the average patient from any source other than A & E; most ES patients are discharged within 48 h. Lastly, their paths of recovery are more predictable and their mortality rate is one fourth of the rate of the others. Because of these characteristics, during high-demand periods ES patients are often asked to take a back seat in favour of patients from the other groups who are in more immediate dire straits. Table 2 shows the abnormally higher one-in-four rejection rate for OT-elective patients due to a full ICU. A DICU is attached to a functional unit such as the surgery department and limits its service exclusively to patients in the post-operative recovery stage. This type of ICU is widely implemented in the United States. A pediatric DICU designed to meet the special needs of seriously ill children has a staff trained in both pediatric and critical care. The creation of a DICU for ES patients by physi- cally locating some beds in a place other than the current ICU should improve the service to those patients, because they will no longer compete for a bed with other patients. The performances of both units are likely to be less than optimal, however, because each one, and particularly the DICU, is subject to larger fluctuations in overall patient arrivals. Further, the remaining ICU will have fewer beds to allocate to referrals from the other sources during periods when fewer scheduled elective surgeries might otherwise result in slack. Thus, the current ICU may enjoy the advantages of ‘‘diversifica- tion’’, which in this context implies that the fluctuations of patient arrivals from one source might offset the fluctuations from another source. Such offsets stabilize the pattern of arrivals and demands and will have a positive impact on the ICU’s performance. FBA, which to the best of our knowledge is being proposed here for the first time, maintains all beds within the current ICU, reserving one or more for the exclusive use of ES patients. This is a one-way stream, since the FBA scheme that the surgeons favour assigns any empty unreserved bed to an ES referral when the reserved beds are fully occupied. Four bed-reservation schemes were tested under the basic FBA umbrella as alternatives to the current system, labelled Strategy 1. The first, Strategy 2, reserves some beds exclusively for ES patients. The other three schemes, which run counter to the surgeons preferences, seek to exploit the fact that elective surgeries are seldom performed on weekends. Under these circumstances, it seems reasonable to consider allowing patients who would not otherwise qualify for a reserved bed to occupy an empty one on weekends, when the unreserved beds are fully occupied. An additional condition is that the latter patients should come from groups that require relatively little ICU time. This restriction would nor- mally release the reallocated reserved beds for ES patients on the following Monday. A look at the A & E patient-arrival and service- time patterns encouraged consideration of a Strategy 3 that opens up the reserved beds to admit A & E patients on Fridays and Saturdays. During our sample period on average there were 1 to 1.5 A & E arrivals on Fridays and ...