development of a quantitative model, which may contribute both to the planning of ... putting pressure on companies to reduce .... software must be more integrated so that facilities ...... on Facilities Management and Maintenance, Brisbane, pp.
Introduction
Development of an integrated healthcare facilities management model Igal M. Shohet and Sarel Lavy
The authors Igal M. Shohet is Senior Lecturer and Sarel Lavy is a Teaching Assistant, both at the Faculty of Civil and Environmental Engineering and National Building Research Institute, Technion – Israel Institute of Technology, Haifa, Israel.
Keywords Health services, Information systems, Maintenance, Performance levels, Risk management, Israel
Abstract The increasing competitiveness in the business sector forces facilities managers in reducing expenditure on “non-core” activities. Consequently, the integration of different domains related to facilities management (FM) motivates the development of a quantitative model, which may contribute both to the planning of FM activities and to the improved effectiveness of FM units. Three methodologies were used in the research presented in this paper: a structured FM survey conducted among acute care facilities in Israel; a statistical analysis of the data; and the conceptual development of an FM model. The proposed model is divided into three main phases that deal with the following five main FM domains. The outcomes of the model produce a set of variables, which can be analysed according to a given facility. In addition, the model provides guidelines for the methodological design and management of healthcare facilities from a life-cycle perspective.
Electronic access The Emerald Research Register for this journal is available at www.emeraldinsight.com/researchregister The current issue and full text archive of this journal is available at www.emeraldinsight.com/0263-2772.htm
Facilities Volume 22 · Number 5/6 · 2004 · pp. 129-140 q Emerald Group Publishing Limited · ISSN 0263-2772 DOI 10.1108/02632770410540342
Property, as a component of an organisation’s facilities, is regarded as an asset that, when properly managed, can add strategic value to the organisation. Added to this is the increase in competitiveness in the business sector, which is putting pressure on companies to reduce expenditure on “non-core” activities. The increasing requirements for the economic operation of facilities have led to the development of the facilities management (FM) discipline. This encourages building owners and users to increase their expectations and requirements of facilities. Facilities managers thus are expected to attain lower operational costs by efficient construction, management, and maintenance of facilities, without compromising their performance. Today’s unexpected rate of changes, and the uncertainties that it brings, challenge those who are charged with the management of property and facilities to acquire a better understanding of the body of knowledge and the procedures related to the built asset. Consequently, integration of different specialisms related to FM (maintenance, performance, risk management, energy and operations, management, and development) presents a rationale for the development of a quantitative model, which may contribute both to the planning of FM activities and to the improved effectiveness of FM units. At present, the discipline of FM lacks quantitative models that support decision making at the strategic and tactic levels, and therefore this development is deemed necessary. The progress in information technology (IT) provides greater data-processing capacities, which are applicable at all stages of the built asset’s life cycle – from design to operations and maintenance. Furthermore, new computational methodologies, which include artificial intelligence (AI) methods (e.g. artificial neural networks (ANN) and case-based reasoning (CBR)), create opportunities for the development of novel attitudes towards a more integrated approach to built asset management. These advances necessitate the development of a quantitative model that will investigate buildings according to core FM parameters and variables. For instance, the effect of average occupancy, which is defined as the number of patient beds per 1,000m2 floor area; and built density, which is defined as the ratio between the total built area (floors and parking areas) and the total ground area must be quantified in such a model. Variables such as preparedness, which is defined as the capacity of the utilities (water and fuel) and of energy reservoirs (air-conditioning, electricity, and
129
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
fire detection); and required performance level necessitate quantitative analysis. Such a model should integrate these different FM domains and ultimately produce performance indicators for consideration and implementation in the making of strategic decisions in FM. The proposed FM model is unique in its capability to handle multivariables with multiple effects among them, a limited number of cases, missing and incomplete data and significant differences between facilities. This paper also describes the rationale and architecture of the model.
AI techniques, which are being introduced increasingly into various practical applications (Clark and Metha, 1997). The current development of computer applications in the area of FM is, however, still slow; primarily due to the fact that it requires comprehensive, structured asbuilt databases (Yu et al., 1997). AI is often developed in terms of various methodologies, such as constraint-based programming, fuzzy logic, genetic algorithms, logic programming, ANN, and CBR, all of which were developed over the past three or four decades (Watson, 1999). In the following, we will examine the different AI techniques, and consider AI’s potential applicability in resolving FM problems: . Constraint-based analysis and programming was developed mainly in order to resolve scheduling problems. Such problems contain constraints that must be satisfied (e.g. production, technological coherence, resource, and capacity constraints), and other preferences such as minimisation of machine time, whose satisfaction is desirable (Hopegood, 1993; Fahle et al., 2002). Thus, this methodology is not suitable for use in this research, since it does not deal with constraints and preference problems. . Fuzzy logic can be considered as a methodology that computes an output vector from an input vector by applying linguistic rules or statements (Costa et al., 1996). The fuzzy rules are read as IF-THEN conditions. Thus, fuzzy logic can successfully deal with multi-variable, non-linear, and time-varying processes (Stylios and Groumpos, 1999; Ligas and Ali, 1996). The architecture of the problem in this research that includes performance and risk evaluation cannot be characterised using linguistic conditional statements; consequently this methodology is deemed unsuitable. . Genetic algorithms are a search and optimisation technique based on genetics and inspired by natural evolution (Lingras, 2001; Leite et al., 2002). Genetic algorithms belong to the category of probabilistic algorithms, and their problem-solving process requires several feasible solutions (Goldberg, 1989). Since the problem in this study is not an optimisation problem (except certain parts of it, e.g. the maintenance policy), and since the scope of data in FM is limited, this technique is inappropriate for successful implementation here. . Logic programming combines logic and procedures (Cercone and McCalla, 1987). Garcia and Chien (1991) described the following four elements that make up logic
Scientific background Healthcare FM In general, the FM discipline is perceived today as a multi-disciplinary theme that aims to integrate between people and the physical workplace of an organisation (BIFM, 2003; EuroFM, 2003; IFMA, 2003). The knowledge base on the subject of healthcare FM, in particular, has grown, especially due to the substantial academic research and practical efforts that have been witnessed during the past decade. The literature abounds with researchers who are known to have dealt with different aspects of healthcare FM, including: . the development of the profession and the role of facilities manager in the healthcare organisation (Gelnay, 2002; Rees, 1997, 1998; Gallagher, 1998); . the development of key performance indicators (KPIs) for healthcare facilities (Fredriksen, 2002; Pullen et al., 2000); and . risk management of such facilities (Okoroh et al., 2002; Holt et al., 2000). Shohet and Lavy (2004) described six core domains of healthcare facilities management, as follows: (1) maintenance management; (2) performance management; (3) risk management; (4) supply services management; (5) development, which includes strategic planning; and (6) IT, the implementation of which may assist in resolving the complexity that is inherent in the subject of healthcare FM.
IT Gallant (1994) outlined the main objectives of AI as the creation of applications that perform as well as humans on tasks involving learning, vision, language, and robotic motion. During recent years, a great deal of interest has been expressed in
130
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
.
.
programming: an alphabet of symbols, a set of inference rules, a set of axioms, and a definition of functions within the logic. The FM problem consists of qualitative and quantitative data, and as a result, no axioms can be defined. Thus, logic programming may be suitable for solving part of the problem, e.g. risk management, but is not suitable for resolving the FM problem presented in this research. The development of ANN was started 60 years ago by McCulloch and Pitts (1943), and was motivated by a desire to try to both understand the brain and imitate some of its strengths. The ANN technique is inspired by biology, and similar to the human brain, it consists of a network of interconnected processing elements (also called neurons, units, cells, and nodes). Its memory uses a weighting system that connects neurons together in several layers (Fausett, 1994; Flood and Kartam, 1994; Garrett et al., 1997; Chao and Skibniewski, 1998; Edwards et al., 2000). The application of ANN has several drawbacks, namely, its ability to deal only with numerical figures, and its requirement of a large body of data sets (Yau and Yang, 1998a). The FM problem is characterized by a limited number of cases, and by missing and incomplete data; therefore the use of ANN is unsuitable. Yu et al. (2000) perceived that “future FM software must be more integrated so that facilities can be managed in a more comprehensive manner during their life cycle”. FM usually requires an analysis of a considerable amount of data, which makes it very difficult to obtain and produce appropriate and suitable knowledge (Christian and Chan, 1993). As can be seen from the characteristics of the above techniques, they are inadequate or only partially suitable for implementation in an integrated FM model. Therefore, we will next examine a potentially suitable AI technique, namely, CBR.
CBR A different problem-solving paradigm approach is known as CBR. CBR was originally motivated by the desire to understand how people remember information, and it was found that people generally solve problems by recalling how similar problems were solved (Watson, 1999). CBR is capable of utilising the specific knowledge obtained in previously experienced concrete cases; thus, it is a technique that solves a new problem by retrieving
previous cases from a “bank” of classified cases, using sets of indicators and rules (Kim and Han, 2001). Aha (1998) described five main steps for achieving a solution to a new problem: (1) retrieve a set of stored cases decided to be similar to the description of the new problem; (2) reuse one or more solutions from these cases; (3) revise: adjust these solutions in order to solve the current problem; (4) review: assess the results attained from applying the proposed solutions to the current problem; and (5) retain: evaluate if the new problem is appropriate to be added to the library of cases as a new case. The literature cites many researchers that have been known to use the CBR approach in solving complex problems, with the majority of applications being in the field of medicine (Ozturk and Aamodt, 1998). In construction engineering, CBR has been used in various applications, including: a steel bridge members plant where it was used to examine and recommend corrective actions for manufacture errors (Roddis and Bocox, 1997); the selection of retaining walls (Yau and Yang, 1998b); the assessment of its application in a house-renovation funding system (Brandon and Ribeiro, 1998); solving of scheduling problems (Dzeng and Tommelein, 1997; Burke et al., 2000); in a study on the bidding behavior of contractors in the procurement of different projects (Chua et al., 2001); and so on. Arditi and Tokdemir (1999) compared the two approaches, CBR and ANN, in order to predict the effectiveness of the techniques in solving construction litigation problems. They found the CBR approach to be more successful, especially due to the kind of information it provides (which includes different definitions of features). In CBR, each case is represented by a number of fields that take on various forms, such as numerical, logical, alphabetical, and strings. Yau and Yang (1998a) concluded that “a CBR application’s input and output are more readable than that of a neural network application”. One important advantage of CBR is its suitability for implementation in cases in which the interactions and relations between variables are not clearly formulated and understood (Cunningham and Bonzano, 1999). As is the case in medicine problems, the problem of healthcare FM is characterised by different kinds of data, including both numerical data (e.g. maintenance budget, physical performance, and energy consumption), and verbal data (e.g. geographical category, type of facility, and maintenance policy). In addition, data for most of the healthcare facilities studied was partially missing or incomplete, making the task at
131
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
hand similar to giving an expert medical diagnosis when some of the information is incomplete or unknown. Furthermore, like in medical problems, the type of solution and the reasoning mechanism involves integration of statistical knowledge of the phenomenon with previous experience and heuristics of other cases. Like the human body, built facilities are system-intensive entities in which malfunctioning of one system propagates to other systems. Since CBR has exhibited high efficiency in the field of medicine, it may prove itself to be a promising approach for diagnosing and treating built facilities as well.
development of this model. The statistical approach was based mainly on previous research and on the statistical analyses (performed in the framework of this research), whereas the heuristic approach made use of the cases obtained in the field survey. The architecture of the proposed model is described below. (4) Computerisation of the model. The conceptual development of the model will be followed by the development of a computer application, which will include input and output interfaces, as well as an evaluator and predictor reasoning phase. (5) Validation. In order to examine and improve the developed model, validation will be carried out using two different methodologies: . several case studies will be performed and their recommendations presented to a group of healthcare FM experts who will evaluate them subjectively; and . the model will be implemented on a healthcare facility and its results measured, analysed, and compared with the model’s predicted values.
Methodology The present research uses the following methodologies: (1) Structured field survey (data gathering). A structured questionnaire survey was distributed among 21 facilities managers of public acute care hospital facilities in Israel. The questionnaire focused on facility parameters and variables such as building parameters, maintenance activities, manpower, energy and operations, and performance. Performance was examined by dividing the buildings into ten main systems, which were in turn subdivided into 52 main building components. Each component was evaluated according to the building performance indicator (BPI) (Shohet, 2003a, b) using a 100-point rating scale, in which 20 represents the lowest score and 100 the highest. The main objective of this survey was to characterise the core parameters and variables that affect FM performance. (2) Statistical analysis. Although the facility managers of 14 of the 21 healthcare facilities that constituted a representative sample of the population of public acute care hospitals in Israel answered the questionnaire (67 per cent), the data available were still incomplete. The data collected in the field survey were analysed using different statistical approaches, which led to the development of a preliminary set of indicators for the evaluation of maintenance, performance, and management of a hospital facility. (3) Conceptual development of the model. The conceptual development of the Integrated Facilities Management Model (IFMM) was conducted based on the statistical analyses of the field survey findings, and the principles put forward in previous stages of the research (Shohet et al., 2003a, b). Statistical and heuristic approaches were used in the
This paper focuses on the first three phases. Implementation and validation of the model will be presented in a separate paper.
Field survey Main findings The field survey carried out in this research comprised 14 public hospital facilities in Israel (12 acute care and two psychiatric hospitals). Its main findings are summarised below. Profile of healthcare facilities Table I presents a profile of the sample population, and summarises the mean and standard deviation (SD) of different parameters examined at the 12 acute care public hospitals. The main findings were as follows. The mean built density was 0.64, with an SD of 0.61. In only two out of the 12 facilities in the sample population, was the built density greater than 1, a value that indicates the presence of tall buildings in relatively dense facilities. This ratio can be used to develop an indicator for the allocation of resources according to the facility’s built density. The high SD in the average number of patient beds indicates a variety in the size of facilities examined: small, medium, and large. However, the average occupancy (patient beds per 1,000m2 floor area) in the public acute care hospital facilities was 8.91 (with a SD of 1.90 beds), which indicates a unified policy in the
132
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
Table I Profile of sample population of Israeli acute care hospitals Parameter
Mean
SD
Number of facilities Floor area (103m2) Parking area (103m2) Total ground area (103m2) Built density (floor parking area per ground area) Number of patient beds Occupancy (patient beds per 1,000m2 floor area) In-house maintenance employees Patient beds per in-house maintenance employee
12 83.4 25.6 241.7
68.1 28.0 235.1
0.64 669
0.61 341
8.91 47.6
1.90 21.8
14.6
3.3
design of hospitals with regard to their occupancy level. The number of in-house maintenance employees does not include gardening, cleaning, kitchen, and security personnel. On average, there were 14.6 patient beds per in-house maintenance employee, or in terms of floor area, 0.63 employees per 1,000m2 floor area (or one in-house maintenance employee per 1,587m2 floor area). The average number of employees per healthcare facility was strongly affected by its policy pertaining to the employment of in-house as opposed to the use of outsourced labour. On average, the main in-house maintenance crews were in the fields of electricity, air-conditioning, water and plumbing, metal workshop, building, and carpentry workshop, which are the core areas of maintenance in healthcare facilities. On the other hand, no in-house crews dealt with elevators, fire protection, and waterproofing. The main reasons for this are the high availability of outsourced workers, the competitiveness of contractors, and their low cost. These issues will be dealt with thoroughly in the development of the proposed FM model. The total number of employees (physicians, nurses, etc.) in the surveyed healthcare facilities was calculated. For public acute care hospitals, the regression analysis between the number of employees and the number of patient beds showed a linear correlation (R2 ¼ 0:86, at the 0.001 significance level). The findings showed that facilities with 350 patient beds employed an average of 930 employees. Each additional patient bed required an average supplement of 3.25 jobs. This regression is valid in the range of 300 to 1,300 patient beds in public acute care hospitals. It was found that these hospitals employed, on average, 2.86 employees per patient bed, with an SD of 0.74.
Performance The performance of 54 buildings, selected as a sample out of the building population of 11 public acute care hospital facilities surveyed, was examined. The total floor area of this sample was calculated to be 482,710m2 (constituting 48 per cent of the total floor area). On average, the area of each building was 8,940m2, and the average actual service life was 23.5 years. Most of the buildings examined had an area of less than 10,000m2 (40 out of 54 buildings, i.e. 74 per cent), and the actual service life of 37 of the buildings (69 per cent) was less than 30 years. Figure 1 presents the percentages of different areas of the buildings according to their designation: hospital wards, offices and administration, utility rooms (laundry, kitchenette, and warehouses), energy facilities, laboratories, clinics, and miscellaneous areas (e.g. lobby, cafeteria). It can be seen from Figure 1 that 42.7 per cent of the surveyed area was designated for hospital wards, 32.6 per cent for utilities, offices and energy, and 18.1 per cent was designated for clinics and laboratories. These figures provide valuable measures for the allocation of resources and priority setting of maintenance, operations, and development activities. The surveyed buildings were divided into components, and each component was evaluated according to the BPI (Shohet, 2003a, b), using a 100-point rating scale. The surveyor, together with the facilities manager for each of the facilities, Figure 1 Average distribution, by designation, of floor area in Israeli public acute care hospital facilities
133
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
examined the performance of each component of the selected buildings. The mean scores for each building system in public acute care hospital facilities are presented in Table II, along with their SDs. Table II also presents the lowest and highest scores for each building system. Mean scores were calculated for each building system, and weighted according to the area of the building in relation to the total facility area. The composition of each system in the total life-cycle costs of the building determined its weight in the overall performance indicator (Shohet et al., 2003). The values obtained reflect the performance level of the facility and its systems, according to the following categories: . A performance level with over 80 points indicates that the system is in good or better than good condition. . A performance level of between 70 and 80 points indicates that the state of the system is such that some of the components are in marginal condition (i.e. some preventive maintenance measures must be taken). . A performance level of between 60 and 70 points indicates a state of deterioration (i.e. preventive and break-down maintenance activities must be carried out). . A performance level of less than 60 points means that the system is in a run-down condition.
points, while the lowest was 67.8 points. The total performance score was 76.6 points, with an SD of 5.9 points. This means that the performance level of systems in this survey was satisfactory but was not yet at a good level (80).
Table II shows that, on average, the system with the highest performance was the medical gases system (87.5 points out of 100), while the system with the lowest performance was the sanitary system (69.9 points). Only two building systems were found to have an average performance level that was higher than 80 points (indicative of a system in good condition), namely, medical gases and communications and low-voltage (82.2). On the other hand, only one system had an average performance level lower than 70 points (a state of deterioration), namely, the sanitary system. The highest performance for a single facility was 85.5 Table II Performance scores of public acute care hospitals Building system Structure Exterior envelope Interior finishes Electricity Sanitary system HVAC Fire protection Elevators Communication Medical gases Total
Sample size
Mean
SD
Lowest
Highest
11 11 11 11 11 11 11 11 11 11 11
79.9 74.4 76.2 77.8 69.9 77.0 76.6 79.6 82.2 87.5 76.6
6.1 10.1 8.8 6.9 9.9 7.1 14.2 5.7 10.1 10.5 5.9
70.6 59.0 63.1 65.5 57.1 64.5 44.2 70.4 70.1 71.4 67.8
91.3 94.1 90.0 88.2 89.2 87.6 93.7 85.9 96.0 100.0 85.5
Energy Energy is one of the most important aspects of healthcare FM, especially due to its implications on efficiency and cost-effectiveness. Various infrastructure parameters, such as the preparedness and the energy consumption, were examined in this field survey as well. Table III summarises the main findings for public acute care hospital facilities. The results are characterised by high SDs in the water and fuel reservoirs and fire protection systems, which emphasises the diversity between the facilities. There are many possible reasons for this diversity, such as different geographical locations, policies, sets of priorities, availability of financial resources, etc. This situation reinforces even further the need to establish a methodology that will compare between the facilities and monitor these parameters. Annual water consumption was calculated for the public acute care hospital facilities. The regression analysis between the water consumption and the total floor area of each facility showed a highly linear correlation (R2 ¼ 0:97, at the 0.001 significance level). The findings show that facilities with an average floor area of 35,000m2 consume 90,000m3 of water per annum, on average. Each additional area of 1,000m2 requires an average supplement of 3,223m3 of water per year. This regression is valid for floor areas between 35,000 and 280,000m2. The water consumption was also analysed in relation to other parameters of the healthcare facility (e.g. number of patient beds, average occupancy, gardening area, etc.). The regression analyses for those cases, however, exhibited low statistical correlations. Similar analyses were conducted in relation to power consumption. In this case, two multi-variable regressions were found that can be used to predict power consumption. The first uses a parabolic function of both the floor and the parking areas of the hospital Table III Average capacity of systems in public acute care hospitals
Parameter
Units
Air-conditioning Electricity (emergency generators) Water reservoir Fuel reservoir Fire protection
Ton KVA m3 Ton m2
134
Capacity per patient bed n SD
Capacity per 1,000m2 floor area n SD
3.5 6.8 2.2 0.21
29.8 58.2 19.7 1.88 675
1.3 1.9 1.1 0.09 –
6.3 16.3 10.4 0.95 201
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
(R2 ¼ 0:99, at the 0.001 significance level), whereas the second uses a parabolic function of the occupancy and a linear function of the total number of employees (R2 ¼ 0:97, at the 0.001 significance level). Use of these two functions may enable prediction of the annual power consumption of a hospital facility. Data obtained from the facilities managers on the subject of the operation of acute care facilities and which are required in order to reach conclusions on the topic, was found to be insufficient. This topic will therefore be investigated later on during the course of this research.
Summary Analysis of the field data establishes a profile of the facility parameters, the performance and the energy consumption of public acute care hospitals in Israel. The findings show that the sample population of hospital facilities was nonhomogeneous in many parameters such as performance, outsourcing of FM services, and so on. These findings illustrate the lack of structured outlines for healthcare FM. The regression analyses of power and water consumption data as against the healthcare facility parameters shows high correlations between the power and water consumption and the total floor area of a healthcare facility and the number of its employees. These statistical models will be implemented in the knowledge base predictor of the energy and operations module of the IFMM.
The healthcare facilities maintenance model Rationale The model proposed in this research (IFMM) provides insight into the assessment of parameters that affect maintenance and operations in healthcare facilities. The proposed model, presented in Figure 2, is divided into three main phases: input interface; reasoning evaluator and predictor phase; and output interface. In general, the input interface receives data from the user on a particular facility, using five main modules (as described below). The input data is then analysed by the reasoning evaluator and predictor, which stores knowledge on both past cases (using CBR) and on developed models (knowledge-based predictor model). Finally, the output interface provides the user with analyses of the facility’s indicators, using the same five modules used by the input interface. The interfaces are forward related, i.e. information received by the input interface is analysed by the reasoning evaluator and predictor, resulting in
indicators and recommendations that are subsequently processed by the output interface. The input interface As presented in Figure 2, the input interface is composed of five main modules: (1) maintenance management; (2) performance and risk management; (3) energy and operations; (4) management; and (5) development. The data are input into these modules by the user. The modules are designed to contain information related to the facility in question, which is later used by the second interface in the evaluation and prediction of the maintenance and operations parameters. Figure 3 presents a conceptual description of three of the five modules, namely, maintenance management, performance and risk management, and energy and operations (see the Appendix for a list of abbreviations). Whereas the two remaining modules, management and development, are certainly part of the model, such modules have already been developed in previous studies (Shen and Spedding, 1998; Shen and Lo, 1999; Shohet and Perelstein, 2003) and hence will not be dealt with here. The three modules of the input interface consist of the following data: (1) Maintenance management. This module includes the profile of the facility and its parameters, in the following categories: . geographic category of the facility (GEC) – coastal, in-land, desert, or mountain; . type of facility (TOF) – university hospital, medium/large, or small/peripheral hospital; . facility area (FAR) – total ground area, floor, parking and gardening areas (m2); . environmental conditions (ENV) – marine or in-land environment; . density of personnel and visitors (DEN) (people/m2/year); . occupancy of patients, and yearly average occupancy (OCP) (patient beds/m2); . actual annual maintenance expenditure per total floor area (AME) ($/m2); . actual maintenance resources (AMR) – inhouse, outsourcing, and materials – for each of the maintenance professions; . availability of labour for each of the maintenance professions (AOL); . facility designation by categories (FAD) – hospital wards, offices and administration, utility areas, energy, laboratory, clinics, and other, for each part of the building; . actual and required service life for each part of the building (ASL and RSL) (years); and
135
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
Figure 2 Outline of the proposed IFMM
.
cost of failures for each of the following building systems (CFn) – structure, exterior envelope, interior finishes, sanitary system, electricity, HVAC, fire protection, elevators, communication, and medical gases, for each part of building ($).
(2) Performance and risk management. This module includes the performance and risk parameters for each of the building’s systems, in the following categories: . required performance (RPn) (0 4 100 BPI scale) – for each building system; . physical performance (PPn) (0 4 100 BPI scale); . required risk level (RRn); and . actual risk level (ARn). (3) Energy and operations management. This module includes the energy and operations data, in the following categories: . energy consumption (ECO) – water, electricity, fuel, and medical gas; . preparedness (PRE) – air-conditioning, water, electricity, fire detection, UPS, and exterior envelope area; and . actual annual operations expenditure (AOE) – security, cleaning, and gardening ($/m2). Input from these modules is passed on to the second stage of the model for assessment, evaluation, and prediction.
The output interface The output interface is schematically similar to the input interface, and is composed of the same five main facilities management modules. This phase of the model produces a set of variables, according to which a given facility is analysed. Figure 4 presents the structure of the output interface. The main outcomes of the five modules are as follows: (1) Maintenance management. Two outcomes are produced: predicted annual maintenance expenditure (PrAME) ($/m2), and recommended maintenance policy for each building system (MPn). (2) Performance and risk management. This module computes the actual facility performance level (BPI), in addition to the following two outcomes: predicted risk level (PrRLn), and predicted facility performance (PrFPn). These three products are obtained for each of the ten building systems. (3) Energy and operations management. Two outcomes are produced: predicted annual energy consumption (PrAEC), and predicted annual operations expenditure (PrAOE). The above-mentioned outcomes are interrelated; for instance, the predicted annual maintenance expenditure is adjusted to the predicted risk level and to the predicted performance, and all of the variables related to the maintenance must be adapted to the required risk and performance level. Such implementation will enable the execution of
136
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
Figure 3 The IFMM – input interface
This part of the model is executed by implementing two methodologies: Each of the indicators predicted at this phase uses either CBR or a knowledge-based predictor model to assess its effect and impact on the output interface (the subsequent interface). The reasoning evaluator and predictor helps estimate different indicators used in the third interface – the output interface. The maintenance management module of the reasoning evaluator and predictor classifies facilities according to past cases, and uses indicators to predict maintenance expenditure and maintenance resources. Figure 5 presents the flowchart of the maintenance management module. It begins with the input interface, in which the facility is classified; continues to the reasoning evaluator and predictor phase, in which the maintenance indicators are calculated, according to a database of performance indicators and a database of previous cases; and concludes with the output interface, described above. Similar procedures are performed in order to classify the building and to predict the performance and risk level in the performance and risk management module, as well as to predict the energy consumption and operations expenditure in the energy and operations management module.
Conclusions Decision making in FM involves multiple parameters and variables analysis. Despite the fact that great research and development efforts have been invested in the topic of healthcare FM (Fredriksen, 2002; Gallagher, 1998; Gelnay, 2002; Holt et al., 2000; Okoroh et al., 2002; Pullen et al., 2000; Rees, 1997, 1998), the currently
Figure 4 The IFMM – output interface
Figure 5 The maintenance management model – flowchart
sensitivity and cost-benefit analyses of alternative requirements.
The reasoning evaluator and predictor The reasoning evaluator and predictor analyses the data that have been input into the input interface.
137
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
available models for FM decision making are still limited, particularly on the strategic level. This may be attributed to the fact that the integration between the different parameters of the facility has not yet been researched thoroughly. This is particularly true concerning the effects of the facility parameters on maintenance and operation. The conceptual frame put forward in this paper proposes an integrated model for healthcare FM. The proposed model is divided into five modules, namely, maintenance management, performance and risk management, energy and operations, management, and development. The field survey carried out in the preliminary phase of the research showed that the facilities’ parameters are non-homogeneous. Nevertheless, the analysis succeeded in producing a profile of an acute care hospital facility, using key parameters and indicators. The statistical analysis developed models for the prediction of power and water consumption as against the facility’s floor and parking area, and number of employees. The output of the conceptual model comprises seven vectors and variables in three modules (Figure 4), as follows: (1) The maintenance management module produces the following outcomes: . predicted annual maintenance expenditure (PrAME) – a value pertaining to the level of expenditure justified by the parameters and the performance of the facility; and . maintenance policy (MPn) – a vector comprising ten variables concerning the preferred maintenance policy for each of the ten building systems. (2) The performance and risk management module (PRMM) produces three outcomes: . BPIs – a vector comprising ten values (Pn), indicating the actual performance of a facility; . predicted risk level (PrRLn) – a vector comprising ten indicators regarding the level of risk for each building system; and . predicted facility’s performance (PrFPn) – the predicted performance of the facility as derived from the implementation of the resources indicated in the preceding outcomes. (3) The energy and operations module produces two outcomes: . predicted annual energy consumption (PrAEC) – based on the predicted energy consumption model presented above; and . predicted annual operations expenditure (PrAOE).
were found to be adequate for this purpose. The proposed model identifies and defines the core body of parameters required in order to characterise healthcare facilities. The model suggests a set of five groups of outcomes that will allow for performance-based management of healthcare facilities. The outcomes of the model provide guidelines for the methodological design and management of healthcare facilities from a life-cycle perspective.
The reasoning phase of the model is based on the application of CBR alongside the statistical models and KPIs, which, in the review of methodologies,
References Aha, D.W. (1998), “The omnipresence of case-based reasoning in science and application”, Knowledge-based Systems, Vol. 11 No. 5/6, pp. 261-73. Arditi, D. and Tokdemir, O.B. (1999), “Comparison of case-based reasoning and artificial neural networks”, Journal of Computing in Civil Engineering, Vol. 13 No. 3, pp. 162-9. Brandon, P.S. and Ribeiro, F.L. (1998), “A knowledge-based system for assessing applications for house renovation grants”, Construction Management and Economics, Vol. 16 No. 1, pp. 57-69. British Institute of Facilities Management (BIFM) (2003), available at: www.bifm.org.uk Burke, E.K., MacCarthy, B., Petrovic, S. and Qu, R. (2000), “Structured cases in case-based reasoning – re-using and adapting cases for time-tabling problems”, Knowledgebased Systems, Vol. 13 No. 2/3, pp. 159-65. Cercone, N. and McCalla, G. (1987), The Knowledge Frontier, Springer-Verlag, New York, NY. Chao, L.C. and Skibniewski, M.J. (1998), “Neural networks for evaluating construction technologies”, in Flood, I. and Kartam, N. (Eds), Artificial Neural Networks for Civil Engineers: Advanced Features and Applications, ASCE, New York, NY, pp. 1-34. Christian, J. and Chan, Y.C. (1993), “Use of an expert system in facilities management”, in Proceedings of the 8th International Conference on Applications of Artificial Intelligence in Engineering, Toulouse, Vol. 2, pp. 193-201. Chua, D.K.H., Li, D.Z. and Chan, W.T. (2001), “Case-based reasoning approach in bid decision making”, Journal of Construction Engineering and Management, Vol. 127 No. 1, pp. 35-45. Clark, G. and Metha, P. (1997), “Artificial intelligence and networking in integrated building management systems”, Automation in Construction, Vol. 6 No. 5/6, pp. 481-98. Costa, A., De Gloria, A. and Olivieri, M. (1996), “Hardware design of asynchronous fuzzy controllers”, IEEE Transactions on Fuzzy Systems, Vol. 4 No. 3, pp. 328-38. Cunningham, P. and Bonzano, A. (1999), “Knowledge engineering issues in developing a case-based reasoning application”, Knowledge-based Systems, Vol. 12 No. 7, pp. 371-9. Dzeng, R.J. and Tommelein, I.D. (1997), “Boiler erection scheduling using product models and case-based reasoning”, ASCE, Journal of Construction Engineering and Management, Vol. 123 No. 3, pp. 338-47. Edwards, D.J., Holt, G.D. and Harris, F.C. (2000), “A comparative analysis between the multilayer perceptron ‘neural network’ and multiple regression analysis for predicting construction plant maintenance costs”, Journal of Quality in Maintenance Engineering, Vol. 6 No. 1, pp. 45-60.
138
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
Euro Facility Management Network (EuroFM) (2003), available at: www.eurofm.org Fahle, T., Junker, U., Karisch, S.E., Kohl, N., Sellmann, M. and Vaaben, B. (2002), “Constraint programming based column generation for crew assignment”, Journal of Heuristics, Vol. 8 No. 1, pp. 59-81. Fausett, L. (1994), Fundamentals of Neural Networks: Architecture, Algorithms, and Applications, Prentice-Hall, Englewood Cliffs, NJ. Flood, I. and Kartam, N. (1994), “Neural networks in civil engineering. I: principles and understanding”, Journal of Computing in Civil Engineering, Vol. 8 No. 2, pp. 131-48. Fredriksen, O. (2002), “Benchmarking facility management of hospitals using the EFFOMETER method”, in Proceedings of the 17th Congress of the International Federation of Hospital Engineering, Bergen, pp. 171-80. Gallagher, M. (1998) Harlow, P. (Ed.), Evolution of Facilities Management in the Health Care Sector, Construction Papers, No. 86, The Chartered Institute of Building, Ascot, pp. 1-8. Gallant, S. (1994), Neural Network Learning and Expert Systems, MIT Press, Boston, MA. Garcia, O.N. and Chien, Y.T. (1991), Knowledge-based Systems: Fundamentals and Tools, IEEE Computer Society Press, Los Alamitos, CA. Garrett, J.H., Gunaratnam, D.J. and Ivezic, N. (1997), “Introduction”, in Kartam, N., Flood, I. and Garrett, J.H. (Eds), Artificial Neural Networks for Civil Engineers: Fundamentals and Applications, ASCE, New York, NY, pp. 1-18. Gelnay, B. (2002), “Facility management and the design of Victoria Public Hospitals”, in Proceedings of the CIB Working Commission 70 – Facilities Management and Maintenance Global Symposium 2002, Glasgow, pp. 525-45. Goldberg, D.E. (1989), Genetic Algorithms in Search Optimization, and Machine Learning, Addison-Wesley, Reading, MA. Holt, B., Edkins, A. and Millan, G. (2000), “Developing a generic risk database for FM”, in Nutt, B. and McLennan, P. (Eds), Facility Management – Risks and Opportunities, Blackwell Science, Oxford, pp. 201-11. Hopegood, A.A. (1993), Knowledge-based Systems for Engineers and Scientists, CRC Press, Boca Raton, FL. International Facility Management Association (IFMA) (2003), available at: www.ifma.org Kim, K. and Han, I. (2001), “Maintaining case-based reasoning systems using a genetic algorithms approach”, Expert Systems with Applications, Vol. 21, pp. 139-45. Leite, P.T., Carneiro, A.A.F.M. and de-Carvalho, A.C.P.L.F. (2002), “Energetic operation planning using genetic algorithms”, IEEE Transactions on Power Systems, Vol. 17 No. 1, pp. 173-9. Ligas, D. and Ali, A. (1996), “Neural net – fuzzy logic rules mapping for dynamic of fuzzy sets boundaries”, Computers and Industrial Engineering, Vol. 31 No. 1/2, pp. 429-33. Lingras, P. (2001), “Statistical and genetic algorithms classification of highways”, Journal of Transportation Engineering, Vol. 127 No. 3, pp. 237-43. McCulloch, W.S. and Pitts, W. (1943), “A logical calculus of the ideas immanent in nervous activity”, Bulletin of Mathematical Biophysics, Vol. 5, pp. 115-33. Okoroh, M.I., Gombera, P.P. and Ilozor, B.D. (2002), “Managing FM (support services): business risks in the healthcare sector”, Facilities, Vol. 20 No. 1/2, pp. 41-51.
Ozturk, P. and Aamodt, A. (1998), “A context model for knowledge-intensive case-based reasoning”, International Journal of Human-Computer Studies, Vol. 48 No. 3, pp. 331-55. Pullen, S., Atkinson, D. and Tucker, S. (2000), “Improvements in benchmarking the asset management of medical facilities”, in Proceedings of the International Symposium on Facilities Management and Maintenance, Brisbane, pp. 265-71. Rees, D. (1997), “The current state of facilities management in the UK National Health Service: an overview of management structures”, Facilities, Vol. 15 No. 3/4, pp. 62-5. Rees, D. (1998), “Management structures of facilities management in the National Health Service in England: a review of trends 1995-1997”, Facilities, Vol. 16 No. 9/10, pp. 254-61. Roddis, W.M.K. and Bocox, J. (1997), “Case-based approach for steel bridge fabrication errors”, Journal of Computing in Civil Engineering, Vol. 11 No. 2, pp. 84-91. Shen, Q. and Lo, K.K. (1999), Priority Setting in Maintenance Management, Hong Kong Polytechnic University, Kowloon. Shen, Q. and Spedding, A. (1998), “Priority setting in planned maintenance – practical issues in using the multi-attribute approach”, Building Research and Information, Vol. 26 No. 3, pp. 169-80. Shohet, I.M. (2003a), “Key performance indicators for maintenance of health-care facilities”, Facilities, Vol. 21 No. 1/2, pp. 5-12. Shohet, I.M. (2003b), “Building evaluation methodology for setting maintenance priorities in hospital buildings”, Construction Management and Economics. Shohet, I.M. and Lavy, S. (2004), “Healthcare facilities management: state of the art review”, Facilities (accepted for publication). Shohet, I.M. and Perelstein, E. (2003), “Decision support model for the allocation of resources in rehabilitation projects”, ASCE, Journal of Construction Engineering and Management. Shohet, I.M., Lavy, S. and Bar-on, D. (2003), “Integrated maintenance monitoring of hospital buildings”, Construction Management and Economics, Vol. 21 No. 2, pp. 219-28. Stylios, C.D. and Groumpos, P.P. (1999), “A soft computing approach for modelling the supervisor of manufacturing systems”, Journal of Intelligent and Robotic Systems, Vol. 26 No. 3, pp. 389-403. Watson, I. (1999), “Case-based reasoning is a methodology not a technology”, Knowledge-based Systems, Vol. 12 No. 5/ 6, pp. 303-8. Yau, N.J. and Yang, J.B. (1998a), “Case-base reasoning in construction management”, Computer-aided Civil and Infrastructure Engineering, Vol. 13 No. 2, pp. 143-50. Yau, N.J. and Yang, J.B. (1998b), “Applying case-based reasoning technique to retaining wall selection”, Automation in Construction, Vol. 7 No. 4, pp. 271-83. Yu, K., Froese, T. and Vinet, B. (1997), “Facilities management core models”, paper presented at the Annual Conference of the Canadian Society for Civil Engineering, Sherbrooke, May. Yu, K., Froese, T. and Grobler, F. (2000), “A development framework for data models for computer-integrated facilities management”, Automation in Construction, Vol. 9 No. 2, pp. 145-67.
139
Development of an integrated healthcare facilities management model
Facilities
Igal M. Shohet and Sarel Lavy
Volume 22 · Number 5/6 · 2004 · 129-140
Appendix. List of abbreviations AME
– Maintenance policy for system n (n ¼ 1 4 10). OCP – Occupancy. – Physical performance for system n PPn (n ¼ 1 4 10). PrAEC – Predicted annual energy consumption. PrAME – Predicted annual maintenance expenditure. PrAOE – Predicted annual operations expenditure. PRE – Preparedness. – Predicted facility performance for PrFPn system n (n ¼ 1 4 10). PRMM – Performance and risk management module. PrRLn – Predicted risk level for system n (n ¼ 1 4 10). – Required performance for system n RPn (n ¼ 1 4 10). – Required risk level for system n RRn (n ¼ 1 4 10). RSL – Required service life. TOF – Type of facility.
MPn
– Actual annual maintenance expenditure. AMR – Actual maintenance resources. AOE – Actual annual operations expenditure. AOL – Availability of labour. – Actual risk level for system n ARn (n ¼ 1 4 10). ASL – Actual service life. BPI – Building performance indicator. – Cost of failures for system n CFn (n ¼ 1 4 10). DEN – Density. ECO – Energy consumption. ENV – Environment. EOMM – Energy and operations management module. FAD – Facility designation. FAR – Facility total floor area. GEC – Geographical category. MMMO – Maintenance management module.
140
