Using discrete event simulation to compare the performance of family health unit and primary health care centre organizational models in Portugal

General information

Primary health care services of good quality are widely recognized as critical for the improvement of health care systems [1–3], in particular when health care systems are being challenged by aging populations, an increased prevalence of chronic diseases, complexities of team-based contemporary practice and limited funding. Policy makers have been experimenting with different models of primary care delivery in order to enhance comprehensiveness, integration and accessibility [4]. In Portugal a major reform of primary health care was started in 2006 in order to address those challenges.

The Portuguese National Health Service (NHS) is defined as universal, nearly free at the point of use and funded by general taxation [5]. Although there are overlapping health systems (health insurance schemes from employer organizations and private health insurance plans) and a high use of private services in some specialties, most primary care services are provided by the NHS [6]. Primary care represents the first level of contact with the health system: general practitioners/family doctors (GPs) act both as care providers and gatekeepers to secondary care. Patients must register with a GP from a health centre, preferably in their area of residence, so they can be entitled to home visits if the need arises.

In 2005, primary care services were provided in 351 Primary Health Care Centers (PHCCs) and most of these (254) provided 12 or 24-hour health services for acute illness and some emergency care [7]. In the same year, on average, each PHCC had 30388 registered patients, employed 20 GPs and 20 nurses, and provided 75000 consultations, with an average of 1497 registered patients per GP. An additional amount of non-registered patients (on average 3376, approximately 10% of the total number of patients) sought care in PHCCs. An average of 15000 urgent or emergency consultations per PHCC was performed in that same year [7].

Until 2005, PHCCs had little autonomy and were depending on Regional Health Administrations for relevant management decisions [8]. At the time several problems in the primary care system were acknowledged. These problems included (i) small and inadequate numbers of GPs in some regions, together with the retirement of a growing number of GPs; (ii) high number of people not registered with a GP (10.6% according to [7]); (iii) high dissatisfaction of patients and doctors with primary care provision; and (iv) unnecessary overuse of hospital emergency services related to difficult access to primary care services [9]. In 1998 an experimental organizational model was launched by the Health Ministry, and 20 groups of GPs organized themselves in small autonomous functional units inside the existing health centres. The payment system included a capitation fraction. The evolution of these groups was closely monitored by the Ministry of Health, and relevant efficiency gains were identified [9, 10]. This observation led to the development, in 2006, of a new organizational model for primary health care units, which were designated family health units (FHUs) [11]. These new units loosely resemble family health networks in Canada [12] and British practices. In PHCCs, health care centres operate under a rigid chain of command and control, and professionals are employed as civil servants through a complicated bureaucratic process. FHUs, on the other hand, are self-organized multi-professional teams formed by GPs, nurses, managers and other professionals. These teams have the autonomy to define their own working processes and to negotiate goals to be met with local health authorities [13]. A new payment system was set for these units.

Some differences between the PHCC and FHU organizational models should be highlighted for a better understanding of the changes involved. In some PHCCs acute cases are treated in separate facilities, staffed by the GPs of the PHCC, with opening hours varying between 3 and 24 hours, depending on the location of the PHCC; in FHUs acute cases are treated by GPs during their normal working hours. In most PHCCs only medical consultations are scheduled; in FHUs regular nursing appointments must also be scheduled. In PHCCs a fixed salary is the norm; in some FHUs (known as model B FHUs) remuneration is compounded by a smaller fixed salary fraction plus a series of supplements: (i) capitation (up to a defined ceiling); (ii) a complement for the provision of specific services under contract beyond the basic job description; (iii) a premium for achieving negotiated goals; and (iv) fee-for-service payments for house calls [14]. Many FHUs adopted model B as their remuneration system. In this paper, we focus on the FHU with a model B remuneration system.

By the end of 2010 299 FHUs were in place and in 2011 78 new ones are opening [15]. FHUs are expected to (i) increase motivation and satisfaction of both patients and health care professionals, (ii) outperform PHCCs in terms of quality, access and efficiency [8, 11], and (iii) implicitly confirm the importance of team building and collaboration in the delivery of primary care services [16]. The FHU organizational model brought major changes in appointment scheduling, acute care delivery and staff payment system [17, 18].

FHUs show a positive impact in preliminary evaluations of the reform process. According to Campos [9], these positive signs started to become visible at the end of 2007, with decreased demand of out-of-hours appointments, better doctor-patient relationship and higher degrees of satisfaction and motivation both from patients and professionals. In two complementary studies, Gouveia et al. [19, 10] analysed the cost differences between PHCCs and FHUs, using econometric analysis. Results from those two studies show that, in spite of a higher level of GPs' remuneration in FHUs, global costs were lower in FHUs due to comparatively lower costs in key components of health care, such as on the spending of diagnostic tests, drugs and other procedures. However, and as far as we know, no studies compared the impact of FHUs on accessibility and efficiency. Also, the effect of converting PHCCs into FHUs has not been quantified. Most studies evaluating health care reforms in other countries did not quantify the impact of new organizational models and, when they did, they mostly analysed the impact on a narrow range of indicators.

Simulation modelling in health care

Health care simulation models typically attempt to provide support for better operational decision making and planning [20]. Several health care administrators used discrete event simulation (DES) models as effective tools for allocating resources in the improvement of patient flow, while reducing health care delivery costs and increasing patient satisfaction [21]. The choice of a DES model to model PHCCs and FHUs is justified by its capacity for reproducing a systems' behaviour through the modelling of the relation between the inputs of a primary care delivery system (including patients, doctors, nurses, patient scheduling and patient routing) and various outputs measures (e.g. waiting times, patient throughput, staff utilization) [22]. Systems simulated with DES models consist of discrete entities which occupy discrete states that will change over time. Adapting examples from Pidd [23], a primary care centre will include individual patients who are entities and whose states may include 'being admitted', 'waiting for a consultation', 'being in a consultation with doctor', and so on. An entity is thus an object whose behaviour within the model will be explicitly tracked as the simulation proceeds. Similarly, doctors and nurses who treat the patients may themselves be regarded as entities that change state. Hence, a DES model aims to capture the important features of the system in terms of entities and states. DES models have a time dimension as entities change state through time; have activities (e.g. consultations) that may require the co-operation of more than a single class of entity (e.g. between doctors and patients); and contain processes in the form of a chronological sequence of activities through which an entity must or may pass (e.g. patients pathways). Each class of entity (e.g. patients of types A and B) will have one or more processes associated with it (e.g. will have associated different pathway(s)), and when an entity that is a member of a specific class (e.g. a single patient A) appears in the simulation, each process becomes a route through which the entity will pass (e.g. single patient A will pass through its associated pathway(s)). Thus, building a DES model needs a set of logical statements, expressed in a computable form, describing how the entities change state. Another feature of DES models, being stochastic simulation models, is that they allow the accounting for uncertainty in the demand and delivery of health care (e.g. in the time a patient spends in a consultation). A detailed explanation of DES models can be found elsewhere [23, 24].

Despite the potential of DES models to analyse quality and efficiency improvements in health care systems [25], we did not find any studies using DES to evaluate organizational reforms in the primary care sector. Most simulation and DES studies were developed at the micro level, mainly focusing on the problems of scheduling and capacity planning (available reviews of DES models can be consulted in [21, 20, 25]). These studies seldom modelled whole health care units [25].

Few studies compared the performance of PHCCs and FHUs and, to the best of our knowledge, no study assessed the impact of converting PHCCs into FHUs. In this work, we used DES to model PHCC and FHU organizational models and to analyse the impact of expanding current primary care reforms in Portugal. We developed DES models to compare the impacts of adopting the PHCC or FHU organizational models on accessibility, productivity and costs and assessed the gains that could potentially be achieved with FHU adoption. This paper brings attention to the usefulness of DES models in the evaluation of organizational models and to the potential impacts of expanding the present Portuguese primary care reform.

Conceptual models

We started by representing two conceptual models that included the processes and flows displayed in Figure 1. Flows are shown by five different sets of arrows that represent: (i) patients entering the health care unit, (ii) internal flows of patients inside the unit, (iii) patients re-entering the system for follow-up consultations, (iv) exit of patients from the unit and (v) the model's specific remuneration type.

Case study and computational implementation

The conceptual models for the PHCCs and FHUs were implemented using the Simul8 DES software [26]. The Simul8 software is a visual interactive modelling system for DES [23] that enables the creation of a computer model that replicates a process containing discrete entities and events occurring at discrete times. Figure 2 provides a screenshot of the Simul8 software showing, for a PHCC, patients' pathways, waiting rooms, doctors' cabinets, entry and exit points. The Simul8 software includes the Visual Logic programming language that was used to program the multiple features of PHCCs and FHUs. These features included units' opening times, doctors' working hours, and the routes to be followed by patients requiring different types of care.

The models were applied to a real case study including a convenience sample of nineteen primary health care units (thirteen PHCCs and six FHUs) from the Greater Lisbon area. This area comprises nine municipalities, and our work included units from three of these municipalities - Lisboa, Oeiras and Cascais. This is an urban and densely populated area, with a population growth rate above the national average [9], which places increased pressure on health care provision.

The PIs described in Table 2 are process indicators related to the provision of primary care services that, using the taxonomy of Donabedian [34], capture two dimensions of quality in health care delivery: the ease with which persons can obtain care, i.e. accessibility; and the ability to lower the cost of care without diminishing attainable improvements in health, i.e. efficiency. Efficiency was further decomposed into productivity indicators - on the production per unit of input (for example, consultations per GP) -, and on the costs incurred to produce that amount (for example, total costs for delivering those consultations). One should note, however, that given the structure of the DES model, indicators of accessibility and productivity in Table 2 are related - for example, if each GP provides a higher level of consultations, not only his/her productivity increases, but also accessibility as measured by waiting times for consultations is expected to decrease.

The key parameters of the model (Table 3) were calibrated with real data from the year 2007 and included information regarding resources, production and operational costs for each unit. Given the variability and uncertainty associated with some parameters of the model, as shown in Table 3 several parameters are associated with a probability distribution. As a result, the outputs/PIs of the model are not point estimates but probability distributions. While some PIs produced by the DES model could be compared with real data, the model also generated new information for which real data was not available (for example, on waiting times). Because of difficulties in obtaining costs from FHUs, following the assumptions used by Gouveia et al. [10], we used cost data from the experimental units previously described, and that were in operation in 2005, to simulate FHU costs.

Modelling the conversion of a PHCC into a FHU

Converting thirteen PHCCs into FHUs demanded for the outline of a set of assumptions regarding the legislation that defined FHUs features and the guidelines for FHUs creation (resources, composition, size, operating mode and financing rules) [36, 37]. The following procedure was adopted: first, depending on its dimension, each PHCC was divided into smaller subunits with six to nine physicians, six to nine nurses and, on average, five administrative staff; each of these smaller subunits corresponded to an independent FHU. Second, in each of these subunits the permanent acute care unit was discontinued and GPs working in that unit were assigned exclusively for medical consultations in the FHU. In this way, as previously explained, acute cases started to be treated in the FHU regular working hours. Third, nursing care started to be scheduled, instead of patients showing up (randomly) for unscheduled nursing consultations. Fourth, remuneration of the health care professionals was changed from the salary without incentives to the model B remuneration system previously described. Finally, the demand from each PHCC was split between the corresponding FHUs, as a proportion of the GP resources allocated to each FHU.