Abstracts

Introduction

Non-monetary factors such as the distance travelled to health facilities are nowadays widely acknowledged as important determinants of health care demand (Lovett et al., 2002LOVETT, A. et al. Car travel time and accessibility by bus to general practitioner services: a study using patient registers and GIS. Social Science and Medicine, Boston, v. 55, n. 1, p. 97-111, 2002.; Puig-Junoy et al., 1998PUIG-JUNOY, J.; SAEZ, M.; MARTINEZ-GARCIA, E. Why do patients prefer hospital emergency visits?: a nested multinomial logit analysis for patient-initiated contacts. Health Care Management Science, Richmond, v. 1, n. 1, p. 39-52, 1998.; Dor et al., 1987DOR, A.; GERTLER, P.; VAN DER GAAG, J. Non-price rationing and the choice of medical care providers in rural Cote d’Ivoire. Journal of Health Economics, York, v. 6, n. 4, p. 291-304, 1987.).

Ever since the seminal work of Acton (1975)ACTON, J. P. Nonmonetary factors in the demand for medical services: some empirical evidence. Journal of Political Economy, Chicago, v. 83, n. 3, p. 595-614, 1975., who found that these non-monetary factors were as powerful as price itself – distance, elasticity of demand for ambulatory care was in the order of -0.96 to -1.4 – that other economic factors beyond price are considered in the budget constraints patients face when seeking for health care.

Following Acton (1975)ACTON, J. P. Nonmonetary factors in the demand for medical services: some empirical evidence. Journal of Political Economy, Chicago, v. 83, n. 3, p. 595-614, 1975., several studies have confirmed the importance of distance to health care facilities in modulating demand for care, regardless of the nature of the care and of the development of the country or its Health System (Haynes et al., 1999HAYNES, R. et al. Effects of distances to hospital and GP surgery on hospital in patient episodes, controlling for needs and provision. Social and Science Medicine, Boston, v. 49, n. 3, p. 425-433, 1999.; Jordan et al., 2004JORDAN, H. et al. Distance, rurality and the need for care: access to health services in South West England. International Journal of Health Geographics, Inverness, v. 3, n. 1, p. 21, 2004.; Friedman et al., 2013FRIEDMAN, J. M. et al. Distance to hospital and utilization of surgical services in Haiti: do children, delivering mothers, and patients with emergent surgical conditions experience greater geographical barriers to surgical care? The International Journal of Health Planning and Management, Staffordshire, v. 28, n. 3, p. 248-256, 2013.; Alegana et al., 2012ALEGANA, V. et al. Spatial modelling of health care utilization for treatment of fever in Namibia. International Journal of Health Geographics, Inverness, v. 11, n. 1, p. 6, 2012.; Malqvist et al., 2010MALQVIST, M. et al. Distance decay in delivery care utilization associated with neonatal mortality: a case referent study in Northern Vietnam. BMC Public Health, London, v. 10, n. 1, p. 762, 2010.). This phenomenon is known as “the distance decay effect”, as demand for health care decreases as distance to health care institutions increases.

In the last few decades, the continuous overcrowding of Emergency Departments (ED) has raised interest on the determinants of the demand for this service and their importance on shifting care towards primary care (Ludwick et al., 2009LUDWICK, A. et al. Distances to emergency department and to primary care provider’s office affect emergency department use in children. Academic Emergency Medicine, Des Plaines, v. 16, n. 5, p. 411-417, 2009.; Lowe et al., 2005LOWE, R. A. et al. Association between primary care practice characteristics and emergency department use in a Medicaid managed care organization. Medical Care, Washington, DC, v. 43, n. 8, p. 792-800, 2005., ²⁰⁰⁹LOWE, R. A. et al. Community characteristics affecting emergency department use by Medicaid enrollees. Medical Care, Washington, DC, v. 47, n. 1, p. 15-22, 2009.).

In Sweden, Magnusson (1980)MAGNUSSON, G. The role of proximity in the use of hospital emergency departments. Sociology of Health & Illness, Cardiff, v. 2, n. 2, p. 202-214, 1980. also identified that ED utilization was significantly associated with the distance patients had to travel to the ED, as did Rudge et al. (2013)RUDGE, G. M. et al. The combined influence of distance and neighborhood deprivation on emergency department attendance in a large English population: a retrospective database study. PloS one, San Francisco, v. 8, n. 7, 2013. in the UK, Sanz-Barbero et al. (2012)SANZ-BARBERO, B.; OTERO GARCIA, L.; BLASCO HERNANDEZ, T. The effect of distance on the use of emergency hospital services in a Spanish region with high population dispersion: a multi-level analysis. Medical Care, Massachusetts, v. 50, n. 1, p. 27-34, 2012. in Spain and Lee et al. (2007)LEE, J. E. et al. Utilization of the emergency room: impact of geographic distance. Geospatial Health, Naples, v. 1, n. 2, p. 243-253, 2007. and Henneman et al. (2011)HENNEMAN, P. L. et al. Geography and travel distance impact emergency department visits. Journal of Emergency Medicine, Milwaukee, v. 40, n. 3, p. 333-339, 2011. in the USA.

In Portugal, to our knowledge, only Santana (1996)SANTANA, P. Utilização dos cuidados hospitalares: uma abordagem da geografia da saúde. In: VAZ, A.; RAMOS, F.; PEREIRA, J. As reformas dos sistemas de saúde. Lisboa: APES, 1996. p. 182-208. studied the effect of distance in the accessibility to health services, namely ED. Santana (1996)SANTANA, P. Utilização dos cuidados hospitalares: uma abordagem da geografia da saúde. In: VAZ, A.; RAMOS, F.; PEREIRA, J. As reformas dos sistemas de saúde. Lisboa: APES, 1996. p. 182-208. also found that distance had an important impact on the utilization of a large urban ED, with distance-elasticity ranging from -2 to -4. Furthermore, the author showed that distance was more important in determining ED demand for patients living at nearby municipalities (<50Km) than further away from the hospital.

We build upon on this previous study in several ways: firstly, we use the Manchester Triage to distinguish between different visits’ severities, under the assumption that low-severity visits are more distance-elastic than high-severity visits; secondly, we use data from 3 EDs, covering all levels of ED care in Portugal, again under the assumption that small rural Basic ED visits are more price elastic than large urban Polyvalent ED visits; finally, we include several other determinants of health care demand, both on the supply and on the demand side, which may be responsible for some of the variations in ED utilization.

The aim of this study is twofold: first, we revisit the “distance decay effect” on the demand for ED in Portugal, re-estimating ED’s distance-elasticity of demand considering different severities on admission; second, we provide further insight over other determinants of ED utilization in Portugal.

The rest of the paper is structured as follows: in the next section we provide a brief review of the Portuguese Health System, namely the organization of ED care and its relationship with primary care; in section 3, we describe the methodology we used, in section 4 we present our main findings and in the final section we discuss the implications of those findings.

Context

Portuguese have the right to health protection through a National Health Service (NHS) created in 1979, which should be general, universal and tend to be free of charge, according to patients’ social and economic status. The non-gratuity of the System means that, when accessing NHS health services (e.g. GP care, hospital ambulatory care and ED care), some patients ought to pay a direct payment (known in Portugal as moderating fees - “taxas moderadoras”), which is maximum for central Polyvalent EDs (20€ per visit – about 5% of the national minimum wage per month). Indirect costs, like transportation to health facilities, are sometimes covered: ambulances transport patients to the health facility in emergent situations and back to their home in special circumstances (e.g. patients with chronic conditions, psychiatric patients, oncologic patients) although the criteria for being eligible for free transport has recently become tighter (for instance, for ED care patients with a non-urgent triage classification are automatically ineligible for free transport).

There are two doors Portuguese may use for entering the NHS: the family doctor, which acts as a gatekeeper in the System; the ED, which have a 24/7 open door policy, i.e., patients may reach to the service without a referral, regardless of their clinical status. The focus of our work is on the latter.

The Portuguese Emergency Network comprises three levels of care – Polyvalent EDs, Medical-Surgical EDs and Basic EDs -, which differ in terms of the complexity of the cases they receive, the resources they have (e.g. human resources, lab and imaging exams), the price the Government pays the hospitals for each ES visit and the level of co-payment required to each patient. Polyvalent EDs are those that receive the most complex patients. Basic EDs usually receive only patients with simple cases. Medical-surgical EDs are at an intermediate level, receiving cases with some complexity, but referring the most complex ones to polyvalent EDs.

Methodology

Data

In this paper, we used ED visits’ electronic data from three Portuguese NHS hospitals, one for each of the three types of ED found in Portuguese NHS hospitals (basic ED, medical-surgical ED and polyvalent ED) for a period of 12 months (January-June 2011 and January-June 2012, to avoid seasonality bias).

For each ED, episodes were grouped according to clinical severity on admission (measured by the Manchester Triage classification) into low-severity (“blue” and “green” visits), intermediate-severity (“yellow” visits) and high-severity visits (“orange” and “red” visits).

ED visits were then linked to freguesias (the smallest administrative area in Portugal) according to the patients’ registered residential address. Our unit of measure is therefore the number of ED visits, for each severity category (low, intermediate and high-severity), in each of the 3 EDs (basic, medical-surgical and polyvalent EDs).

Excluding criteria comprised ED visits from patients with an address outside the Northern Region, because arguably these patients did not travel from their registered addresses to the ED, pediatric ED visits (patients younger than 18 y.o.) and ED visits for which no Manchester Triage classification was assigned.

Model

Was assumed that discrete random variable Y (number of visits to ED) is Poisson-distributed with rate parameter, and is the exposure parameter - the population of the correspondent freguesia. Y is defined by the following density distribution function

Where E[Y], the expected value of Y, equals the variance – V[Y].

The equi-dispersion property of Poisson model (equality of the mean with variance) is frequently violated in real-life data, which made us use the negative binomial regression for estimating our model’s parameters.

The final model that is estimated is:

Whereis the number of ED visits for freguesia represents the number of freguesias of our sample size, is an offset variable that accounts for the resident population of freguesia andis the distance from patients’ municipality centre to the ED (in Kilometres). Vector X includes two key variables we used to stratify our ED visits: a categorical variable “ED Level” (coded as two dummy variables representing the 3 levels of care in Portuguese EDs) and a visit severity variable (coded as two dummy variables representing the 3 levels of ED visits’ severity we defined) and a set of variables related to population’s health care needs such as, demographic information (sex and age at admission), health status, level of income and education and sanitary conditions; health care demand (e.g. psychiatric and ophthalmologic visiting rates) and information on health provision determinants in each freguesia. The full list of health care needs and provision variables we used is presented in Table 2.

All coefficients have been put on an exponential scale, thus interpretation of the parameters obtained from the Negative Binomial Regression models are in terms of incident ratio (exp[]). The expression tell us the change in the incidence of ED demand for each unit increase in the independent variable. Variables with IRR higher (lower) than 1 have a positive (negative) influence on ED utilization.

STATA Ver.12© was used to estimate the multivariate model. ArcGIS (v.10.0, Environmental Systems Research Institute, CA) was used for distance computation.

Results

Throughout the period of analysis, we identified a set of 229528 visits in the selected EDs corresponding to people with residence in 1 490 freguesias from the North of Portugal.

Our model’s coefficients are presented in table 2.We take the natural logarithm of distance and the coefficients from NB regressions may be directly read as elasticity. (Sheu et al., 2004SHEU, M. L. et al. The effect of a major cigarette price change on smoking behavior in California: a zero-inflated negative binomial model. Health Economics, New York, v. 13, n. 8, p. 781-791, 2004.)

Overall, we found a distance-demand elasticity of about -1,65; i.e., an increase of 10% in distance to ED leads to a 16,5% decrease in ED utilization.

Table 2 also shows the significance of other parameter estimates that may influence ED demand; specifically, we found an effect of gender (males have overall lower ED utilization, IRR=0,915) and age (elder have overall higher ED utilization, IRR=1,018) and socio-economic factors – unemployment (IRR=1,022) and homelessness (IRR=1,017) positively affect ED utilization, whereas higher income areas and areas with overcrowding of homes (IRR= 0,987) have lower ED demand (IRR=0,997).

More importantly, variables related with supply of care in each area had higher influence in determining ED demand: availability of doctors was positively associated with ED rates (IRR=1,093) whereas PHC had an inverse influence on ED utilization, with areas having an USF (modern primary health centers with performance-based financing) with the lowest ED visiting rates (IRR=0,801; IRR=0,582).

Our gross estimation for the distance-elasticity of demand may be affected by both the severity of the ED visits and the ED level. In table 3, we present different distance-elasticity, considering each ED level and ED visits’ severity category. The results can be read as follows: in the polyvalent ED, for low severity visits, a 10% increase in distance reduces utilization for approximately 13%. In brief, table 3 shows that the sensitiveness of ED utilization to the distance patients have to travel varies quite considerably if we take into account the different visits’ severity: high severity visits have the lowest distance elasticity, whereas intermediate severity visits seem to be the most sensitive to distance. Furthermore irrespectively of the ED visits’ severity, estimates for distance-elasticity are lower in both the Polyvalent and the Elementary EDs ,i.e., the Medical-Surgical ED’ utilization is more sensitive to distance.Figure 1 illustrates for the Polyvalent ED this different effect of distance, according to ED demand severity.

The analysis of figure 1 also shows that the effect of distance becomes less pronounced for higher distances to the ED. Table 4 shows that considering only freguesias located near the ED produces a clear gradient of distance-elasticity across visits’ severity, with low-severity demand having the highest distance-elasticity and high-severity demand the lowest.

Discussion

Our findings provide support over the hypothesis that distance has a negative influence on Emergency Department utilization in Portugal. A previous national study (Santana, 1996SANTANA, P. Utilização dos cuidados hospitalares: uma abordagem da geografia da saúde. In: VAZ, A.; RAMOS, F.; PEREIRA, J. As reformas dos sistemas de saúde. Lisboa: APES, 1996. p. 182-208.) that looked at the influence of distance on ED utilization has found an estimate for the distance-elasticity as high as -4. We found estimates that range from -1 to -2 (an increase in 10% in distance decreases ED utilization by 10 to 20%), which are closer to others reported in international literature. Taking into account other determinants that may influence ED utilization in each area was crucial for making distance-elasticity more accurate.

Specifically, one major influence we show is that in areas with higher availability of other health services, particularly Primary Health Care, ED utilization is lower, which has also been found in other studies (Mathison et al., 2013MATHISON, D. J. et al. Primary care spatial density and non urgent emergency department utilization: a new methodology for evaluating access to care. Academic Pediatrics, Rochester, v. 13, n. 3, p. 278-285, 2013., Ludwick et al., 2009LUDWICK, A. et al. Distances to emergency department and to primary care provider’s office affect emergency department use in children. Academic Emergency Medicine, Des Plaines, v. 16, n. 5, p. 411-417, 2009., Lowe et al., 2005LOWE, R. A. et al. Association between primary care practice characteristics and emergency department use in a Medicaid managed care organization. Medical Care, Washington, DC, v. 43, n. 8, p. 792-800, 2005.). Interestingly, we also show that the existence of an USF (a new modern typology of Health Centres in Portugal with performance-based financing) in that area intensifies this inverse relationship between PHC and ED utilization, which may suggest that patients are under closer watch in these new units (i.e. have higher accessibility to acute non-urgent care) or that they have baseline lower health care needs. A panel study that looks at the influence of different PHC units on the ED utilization could be helpful in addressing these hypotheses. In addition to this effect of health services’ availability, we also found that areas with lower socio-economic conditions (higher unemployment and homelessness and lower income) have higher ED utilization, as do populations with worse health status. One other strength in our study is that we disentangle the distance effects on ED utilization by the severity of ED demand and by ED level. Similarly to Raknes et al. (2013)RAKNES, G.; HANSEN, E.; HUNSKAAR, S. Distance and utilization of out-of-hours services in a Norwegian urban/rural district: an ecological study. BMC Health Services Research, London, v. 13, n. 1, p. 222, 2013. we also found a gradient of distance-elasticity according to the severity of ED visits, with distance-elasticity increasing as visits’ severity decrease. However, for distances higher than 50Km, distance-elasticity became considerably lower and the gradient across demand severity became less clear. We hypothesize that this may be due to the existence of other alternatives at these higher distances, both in the public and private sector, which we do not take into account in our study. Furthermore, we found that the Polyvalent and the Elementary EDs had the lowest distance-elasticity, regardless of the severity of demand. The finding that Polyvalent EDs may have more “loyal” populations (less sensitive to distance) is not new (Santana, 1996SANTANA, P. Utilização dos cuidados hospitalares: uma abordagem da geografia da saúde. In: VAZ, A.; RAMOS, F.; PEREIRA, J. As reformas dos sistemas de saúde. Lisboa: APES, 1996. p. 182-208.), and we argue that it may be due to a possible easier access to some specialties in the ambulatory care that are only available in Central Hospitals, following the ED consultation. As for the Elementary EDs, we assume that lower distance-elasticity may be related with the specific population these less-resourceful EDs attend, which are usually nearby elder populations.

Conclusion

Distance patients have to travel to the Emergency Department has a sizeable effect in determining demand across all levels of Emergency care in Portugal, after taking into account other factors that also influence ED demand. Policies that wish to reduce geographic inequity in the accessibility to ED should be mindful of these findings and consider both direct and distance costs in the total costs patients incur when seeking care within the Portuguese NHS.

Acknowledgements

We are grateful to Manuel Gomes Mota for his help with the GIS analysis and to the Administrations of the CHSJ and CHMA for providing the data, particularly to José Pedro Almeida and to Antonio Parra for their time and their help on providing technical support.

References

Publication Dates

History