Unique health care utilization patterns in a homeless population in Ghent

Homelessness is worldwide an important societal issue as it can be considered as unacceptable and unfair. In the USA, each year over three million people experience homelessness of which 1,3 million children [1]. In England 99.5000 households were officially recognised as newly homeless in 2007 [2]. In the Flemish region each year 12.000 people get support in residential care for homeless persons [3].

Compared to a decade ago, the homeless population today is younger (between 30 and 50 years) with a growing number of women and children. This trend continues both in the USA and in Europe [4–7]. The way people become homeless can be attributed to a number of factors such as a disruptive family environment characterized by extreme poverty, marital discord, addiction, financial problems (no job, no money) and mental health problems [7–10].

The National Health Survey is a rich source of information when studying health and health care use. This cross-sectional survey is used to collect data from a representative sample of the non-institutionalised population residing in Belgium. For the 2004 survey, 12.945 Belgians were interviewed using both a face-to-face survey and a self-registering survey. Sampling has been based on a combination of stratification, multistage sampling and clustering. Stratification was done at the regional level and the level of the provinces. The number of selected individuals per province was determined by the proportions of the Belgians living in that province. In order to keep the fieldwork manageable, the number of interviews in each municipality should at least be 50. Per stratum, households were selected by randomly selecting so called 'reference persons' from the State Register. In every selected household, maximum four persons were questioned: the 'reference person', his/her partner and one or two other household members. Two hundred and thirty five inhabitants of Ghent were questioned. Finally, the data were anonymized by a trusted third party (the National Institute for Statistics) and made available for external researchers. For this study a variable was add to the dataset, indicating whether the participant is an inhabitant of Ghent.

However, homeless people are not included in the study. Therefore additional data from homeless people in Ghent was collected using a face-to-face survey, according to an adapted form of the National Health Survey protocol and using a reduced version of the NSH questionnaires.

The final database, used for this paper, contains data from 235 inhabitants with a fixed address in Ghent and data from 122 homeless people staying in a shelter in Ghent.

Concerning health care utilisation, bivariate analysis demonstrate a significant association between being homeless and the likelihood to consult a GP (p < 0,001), secondary care (p = 0,005) and emergency care (p < 0,001), with higher likelihoods for all three forms of care in the homeless people.

The model focusing on likelihood to consult secondary care explained 8,7% of the variance and demonstrated after adjusting for age and gender, being homeless increases the likelihood to consult secondary care. The odds ratio for the homeless is 3,271 representing the homeless persons' increase in likelihood to consult secondary care (95% CI = 1,800 - 5,943).

The comparable model for likelihood to consult an ED explained 27,7% of the variance and demonstrates that being homeless also increases the likelihood to consult an ED, after adjusting for age and sex (see table 3). Homeless patients have a 13,351 times higher likelihood to consult an ED compared to the general population (OR: 13,351 95% CI = 5,114 - 34,857).

Regression models additionally adjusting for perceived health status follow the same trend. The explained variance increases to 21,5% for the model focussing on likelihood to consult a GP, to 11,9% for the model focussing on likelihood to consult secondary care and to 33,3% for the model focussing on likelihood to consult an ED. In these models, the OR for a homeless person to be likely to consult a GP is 2.592, to consult secondary care 2,683 and to consult an ED 10,752.

Homeless people have a disproportionately higher burden of health problems than the general population. Existing studies (mainly from the UK and the US) describe higher utilisation rates for hospital-based care and emergency care, and lower rates for primary care by homeless people compared to the general population. Some studies indicate that homeless people do not access primary care at all and use emergency care as the only form of health care [6, 16, 20]. Homeless people are importantly hindered and/or steered in their health care use by barriers directly related to the organisation of care [6, 7, 21, 24].

In this study we describe the likelihood to consult a GP, secondary care and emergency care of homeless people living in an area with a primary health care system in which homeless people with health problems are actively guided to low-threshold but universal accessible (i.e. not selective) community health centres.

Hereto the Belgian National Health Survey Database was extended with data of 122 homeless persons residing in one of the night shelters or temporarily accommodations for homeless people in Ghent.

The homeless respondents in this study are significantly younger, lower educated, more likely to smoke, have more often a non-Belgian nationality and report to be in worse health than the general population. These findings are concordant with other studies reporting that homeless people have worse health, suffer more from multimorbidity, have higher smoking prevalence (up to 80%), and are mostly less educated, decreasing the chance to find a stable job [4, 6, 14, 18, 31, 32].

Literature shows a growing number of women in the homeless population [33]. This could not be confirmed in this study in which over 80% of the respondents are male. This could partly be due to a selection bias; women (often with children) who become homeless usually stay in crisis centres for women and children which were not included in this study. However, possibly there are also less homeless women in Belgium than in other countries because of the relatively good safety net for women with children; they have priority in getting social housing or -in case of the unavailability of social housing-emergency housing (temporarily housing in a premises owned or rented by the council).

Multivariate analyses show that homeless persons in Ghent have a higher likelihood to consult a GP (i.e. have consulted the GP at least once in the 2 months prior to the interview) than the general population after adjusting for sex, age and perceived health status. This finding indicates that in Ghent the homeless people find the way to primary health care and use it.

This contrasts with studies from other countries. Several US based studies show lower utilisation rates of ambulatory care and up to 5 times higher utilisation rates of acute hospital-based care and emergency care in the homeless population compared to the US average [6, 16, 20]. Studies from the United Kingdom describe the same trend: homeless people contact secondary services more frequently than the housed population. Some studies describe that a considerable proportion of the homeless population does not find the way to primary health care at all: a survey conducted by Crisis (the national charity for solitary homeless people in the UK) in 2002 reported that homeless people were three times more likely not to have had contact with a GP in the last year and were nearly five times more likely than the general population to attend the emergency departments [34].

A 2002 survey conducted in the UK also looked at the likelihood to consult a GP. They found that homeless persons were 3 times more likely not to contact their GP in the last year than the general population[34].

This higher likelihood in Ghent to consult a GP in the homeless population compared to the general population could be explained by the fact that primary health care for homeless people in Ghent is organised in a way that contributes to the reduction of some of the barriers demonstrated in others types of primary health care organisation such as financial barriers (insurance status), registration, accessibility, stigmatisation,... [7, 11, 14, 21, 23, 24].

Tracing homeless people with health problems and active guidance towards primary care centres by social workers, might contribute to an easier access to the GP. The Community Health Centres are situated close the shelters and social restaurants, so transportation is often not an issue. There are also agreements on the payment procedure so the homeless people do not need to worry about money, reducing the financial barrier. Additionally, Community Health Centres are known not to stigmatise or refuse patients because of being homeless or having no money. This hypothesis is supported by the finding that the homeless people who visit a GP, do this considerably more often at a group practice (including the health centres) (56% versus 5,6% in the general population).

Also, the universal approach of the primary care centres (providing care to everyone residing in the geographical area) leads to less stigmatization and labelling than in a selective setting focussing solely on the homeless population. In this study, the higher likelihood to consult a GP does not go hand in hand with a lower likelihood to consult the emergency department or secondary care. This indicates that homeless people, although they have a higher likelihood to make use of the primary health care system, have not totally abandoned the way to emergency care. Probably this can partly be explained by the particularity of the health problems homeless people experience (e.g. postponing care until it's really urgent or care during the evening or night). Homeless people are also more often the victim of drug related accidents and road accidents and suffer from health problems that need more emergency care than the general population (e.g. hypothermia)[35]. Additionally, this study has a cross-sectional design which makes that we do not know whether the homeless people would use more frequently the emergency services if there was the universal primary care system would have been organised differently. Moreover, since this study is the first to study the health care use of homeless people in Ghent, it is not possible to describe an increase or decrease in the emergency care use of the homeless population over time.

This study has some limitations. First of all there is no reliable information available on the response rate of the homeless people nor on the respondents excluded from the study. The study was presented to all homeless persons present at different moments in one of the shelters in Ghent during the five months inclusion period and everyone was invited when fulfilling the inclusion criteria. However, there was no registration of any information concerning the people who did not want to participate. This was mainly due to anonymity reasons thrown up by the managers of the shelters. Also the exact total number of homeless people visiting shelters and homeless centres is not available. Most of the crisis centres have no registration of their visitors.

A second and important limitation of this study is that only likelihood to consult is studied. The National Health Survey contains self reported frequencies of health care use, so analogously these frequencies were also included in the survey for the homeless people. However, most of the respondents found it very hard to remember the exact number of their visits to a care provider. For a more nuanced view on health care use, information on the number of consultations is necessary. The concept of likelihood to consult is in any case usable to evaluate whether people know and find their way to a certain care provider but not to evaluate the intensity of their health care use. Although this study cannot prove the impact of a better accessible primary health care system on emergency care utilisation rates or on secondary care utilisation rates, a recent study commissioned by the Belgian government gives an important indicator of the possible impact. In this report the authors compared the traditional, fee-for-service based primary health care services in Belgium with capitation based primary health care services such as the community health centres. Their findings complement the findings of our study as they state that the community health centres indeed succeed in reaching the most vulnerable of society. Additionally this study provides evidence that people who found their way to these centres tend to contact less frequently the emergency department or secondary care[36].

The merits of this study lie in the fact that a relatively large sample of homeless people was questioned (an estimated third of the homeless population in Ghent visiting a homeless centre in the research period), following the research protocol and instruments of the National Health Survey as close as possible so that a comparison with National Health Survey data was made possible. However, prudence is recommended in the generalisation of the results. Data were collected from homeless people seeking help or assistance, mainly for food or a place to sleep, in the homeless centres or night shelters in Ghent. Without doubt their help seeking behaviour towards health care is different from that of homeless people not seeking help in the homeless centres. Also no information is available on the response rate and the characteristics of the people who refused to participate in the study, which makes it very difficult to evaluate the selection bias.

Further research addressing the health care use of the homeless people living in Belgium is needed to provide policymakers with the evidence needed to take action in order to make the health care system accessible and suitable for every person disregarding their different socio-economic backgrounds. The performance of universal primary health care services, as demonstrated in this study, may stimulate the debate between a universal or a selective approach.