Much of WLES behaviour is still unknown
WLES management is a critical issue, since the demand for surgery often overwhelms supply and public systems have limited resources. The application of effective policies is still a rising concern, given no definite standard is available to measure and handle the problem. The matter is then far from being comprehended and its intrinsic complexity has been recognized for more than two decades [30
]. Likely as a consequence, detailed information about WLES is difficult to gather in most industrialized countries both from white and gray Internet literature [31
]. With very few exceptions, it is hard to know how many patients are currently waiting (or have waited) for specific diagnoses or surgical procedures. Regarding the situation in Italy, recent communications from the Ministry of Health reported scarce homogeneity and availability of waiting list data, promoting innovation by the use of information technologies [32
]. Our local context did not represent an exception to the Italian background, since waiting list dimensions and composition at the start of the project were substantially undefined in our experimental area.
A tool to handle the problem
Given the obscurity of the matter, the SWALIS model was designed with the intention of obtaining transparent data to gain a more in depth understanding of WLES, enhancing prioritization as a leveraging instrument through the use of information technologies. Feasibility and accessibility emerged early on as qualities necessary for the new model, since its application was expected to cause substantial organizational change. The model was then designed incorporating clinical urgency assessment and real time prioritization sequentially. Such a multi-module structure was chosen in order to include the whole pre-admission period, so as to adopt different clinical urgency criteria and to manage any elective surgical procedure. The system was built to provide data and prioritization in a single software environment, becoming a useful tool in the real practice. As a result, the entire demand for surgery was processed consistently, including a very heterogeneous set of clinical conditions and surgical procedures. The integration of adequate indexes within the model allowed the measure of WLES performance, obtaining the necessary information to better understand their composition and trends.
Effects in practice
The display of patients' respective priority (Figure ) allowed easy and coherent scheduling. The screenshot view of the waiting list together with the safety warnings allowed nurses and surgeons to detect when patients were at risk of exceeding their MTBT. The waiting list was checked by the waiting list manager following the priority order every time admissions were planned, i.e. at least twice a week. In addition, each surgeon usually controlled the entire list with particular attention to his/her waiting patients, logging on to the system separately and more often. As a consequence of this intense monitoring, waiting lists were kept active and clean from "ghost" patients (patients that have died or gone elsewhere for treatment) preventing evident inequities.
Admissions and surgical resources could be planned quickly by the use of an intelligible forecast, visualizing how many patients – and which of them – were to be admitted in the near future (i.e. in two, three or four weeks). It was easier to keep them in really elective conditions reducing the amount of time they waited. This in turn reduced the need for unexpected bed and operative room occupation for sudden clinical deterioration, often resulting in complicated surgery and prolonged postoperative periods as a consequence of excessive waiting times [3
]. Furthermore, on the whole, better planning could reduce postponements by reducing sudden changes in admission schedule.
The availability of waiting time data, either as waiting time distribution (Figures and ) or mean waiting times (Figure ), provided reliable and updated estimations of admission dates in real time. This information allowed a more efficient pre-admission path to be planned, avoiding redundancies and unexpected situations (i.e. in planning instrumental and clinical examinations, in the pre-surgery management of anticoagulant drugs suspension or overlap, etc.). By using waiting times indexes, waiting list trends could be analysed in depth and forecast. This allowed for action planning such as sharing reports and organizing meetings among surgeons and waiting list managers in order to solve emerging problems.
The availability of data regarding waiting list composition and dimensions (Figures , and ) allowed managers to plan the allocation of surgical resources and operative room time to each surgeon and to the surgical unit within the department, modulating supply on the base of the measured demand.
Additionally, the standard application of the model to all registrations allowed the SWALIS system to serve as the proper environment to manage the clinical added information (i.e. noting patients' clinical or private necessities). The software allowed relevant synoptic information to be retrieved (i.e. selections of patients, performance indicators, etc.) and to be delivered to respective users, such as surgeons or waiting list and health managers.
Urgency assessment is a crucial topic since it represents the main milestone of the entire prioritization process. Given the absence of great international agreement on the subject, it may represent a weak point of our study, since we have not investigated it with proper tools. With this limitation, we started from the application of the Italian implicit urgency criteria because of their statement at a national level. While the rater reliability of the Italian criteria might be further assessed and/or more objectively defined, in our experience they proved to be easily applicable to the entire pool of patients, allowing both coherent data collection and patient prioritization. In order to maintain this adaptability, our model has implied the formal separation of the three steps (urgency assessment, MTBT classification and prioritization), allowing the possible introduction of different clinical criteria if necessary.
The surgeon's role in the clinical assessment
Urgency assessment was taken into great consideration, since surgeons' clinical judgment was required to be both free and objective. As described in other experiences, a certain amount of guesswork to elevate patients' URG most likely occurred, even though the phenomenon was discouraged and ad hoc
measures were taken to prevent it from happening [33
]. In our experience, contrasts and heterogeneity in clinical evaluations were reduced to an acceptable level through the periodical audit meetings, when surgeons could compare and discuss their assessment, increasing the homogeneity of their criteria. Interestingly, there was never a major conflict of opinions between surgeons while unanimous agreement was always eventually reached. Furthermore, the surgeons' general opinion was that the URG assessment was progressively becoming simpler and more familiar, allowing the necessary individual freedom.
The patients' point of view
Patients were not asked directly about their experience under SWALIS experimentation, neither in interviews nor by questionnaires, and the impact of the model on patients might be further investigated. With this limitation, during control surveys surgeons and nurses were asked to report issues that emerged at the time of outpatient registration on the list, during the waiting period and at the time of admission.
They described no relevant problem in explaining the criteria of their urgency assessment and reported that patients generally acknowledged that waiting time should not hold the same value for them all but be based on the grade of their illness. On those occasions, patients accepted that waiting time could weigh proportionally to their respective URG and no complaints for being overtaken by more urgent patients were ever reported. We interpreted this finding as mostly due to the national statement of the URGs classification, since an understanding of the prioritization's consistency led to an increase in its level of acceptance [34
]. The positive feedback we registered might have been biased by several factors but it was perhaps influenced by the simplicity of the SWALIS model algorithm, and by the patients' perception of a strong Service's engagement in respecting their MTBT.
Regarding the patients' acceptability of waiting periods, it is known that they often suffer from consequences of extensive waiting periods, and that many of them may have different opinions of waiting times [3
]. They should therefore be informed about their expected waiting time and when their admission is most likely to occur, since they often have a better perception of their waiting period the sooner they are given this information [4
]. In the SWALIS experience, due to the availability of waiting times in real time, patients could be informed about their expected wait as early as at the time of registration.
Effects on waiting lists: treating each patient at the right moment
Assuming that free access does not automatically imply immediate healthcare delivery, our model was designed to treat patients at the right moment, satisfying their respective need. The SWALIS model adopted the waiting time as the unique non-clinical criterion, not only admitting patients by classifying them into URGs (and within their MTBT), but also by scheduling admissions using a progressive scoring system [21
Rather than in waiting list length, the effects of the SWALIS prioritization process are evident within the waiting list composition in real time, since the prioritization algorithm determines a change of their internal sorting continuously, so that every registration will reach the top of the list by its priority score (Figure ). This clear waiting list behavior allowed the selection of the patients to be admitted first at a glance, simply by looking at the top positions in the screenshot.
As expected, the application of the SWALIS model caused no evident effects in terms of reduction or increase of the overall waiting list length (demand side policy), because resources (beds and operating room availability) was set as constant. Reducing list consistency would require increasing service rate (supply side policy) and the variability shown in Figure depends on variability of arrival rate and of service rate (i.e. admitted patients).
In this study we observed no significant difference in the time series. Even so, by the physiological functioning of the SWALIS model, waiting list observation should result in API moving close to the 100% for all different URGs. We consider this point to need further investigation but this general progressive tendency appears in the latest monthly data in Figure .
As some of the authors published separately, the application of the SWALIS model brings an increase in efficiency as well as equity [27
]. According to the results of this study, data shown in Table reveal that patients were protected by horizontal and vertical inequities.
Horizontal inequities were avoided because the algorithm calculates priority proportionally to waiting time. With the exception of the few patients in category D (N = 35, 1.4%), this assumption is supported by the comparison of the overall mean and median waiting times of the respective URG, where overall cross sectional waiting times are inferior to the correspondent retrospective ones. This data give evidence to the fact that patients were admitted only after those in the same urgency category that had been waiting longer.
Vertical inequities were avoided by computing the priority on the basis of clinical urgency. The overall waiting times of each URG showed increasing values with the increase of their MTBT, meaning that the more urgent patients were not waiting as long as those in less urgent categories and were admitted earlier.
Table shows some discrepancies in single measurements at Index days in columns A1 and D, where cross-sectional mean waiting times were occasionally higher than retrospective ones, suggesting that a significant part of them were kept on the list, not being admitted before those with shorter waits. Regarding patients in URG A1 (N = 192; 7.9%), this evidence was due to the difficulties in completing the necessary pre admission diagnosis within the short MTBT (eight days). Those patients in fact often required unplanned complex evaluations before admission even while being kept under strict observation by waiting list managers. The small dimension of the A1 subset of patients often caused the single mean values to change more likely as a consequence of the delay of a few patients, as well as the variability from 0 to 100% in the API in Figure . Regarding URG D, only 35 patients (1.4%) were assessed in the lowest urgency category. The evidence of higher cross sectional mean waiting times for those patients was associated to patient initiated admission postponements, due to personal reasons unrelated to the cause of their presence in the list (i.e. other interfering therapies or illnesses). The occurrence of this phenomenon was facilitated by the long MTBT (one year) and by the possibility to delay treatment for patients who were in the least urgent conditions.
As it has happened in Victoria since 2005 [35
], proper rules for those cases should be included in the standard waiting list management policy. Our results would likely be enhanced by a more rigorous application of the prioritization algorithm, under a proper admission policy, and with the social/political acknowledgement of the process as a standard rule within the hospital.
The SWALIS model might represent a suitable test bed for different urgency criteria, more objective definitions (i.e. oriented to specific procedures), higher standards (i.e. International Classification of Functioning, Disability and Health), and rater reliability of the clinical assessment. The model might be tested in wider and more heterogeneous environments and its user pool might be expanded, including the management level. Given the model is applied by a complex technological interface, the perceived usefulness and applicability of the prioritization should be assessed together with the user ware of its software environment [36
]. The assessment of its clinical impact could include the study of pre-admission and outcome variables, both in time series and in case-control studies.
Since 2003 the SWALIS model has undergone testing in two hospitals in Northern Italy, by different software prototypes (overall 8500 registered patients). In 2007, the Administration of Liguria Region stated the progressive diffusion of the model in its territory and, as of 2009, another general public Hospital in Genoa (counting 6000 surgical admissions per year) will be adopting the model.