Socio-economic deprivation, which measures the disadvantage of an individual or group relative to the local community or wider society (Reference TownsendTownsend, 1987), is an indicator of socio-economic position. Three studies have shown an association between area deprivation and length of psychiatric admission (Reference HirschHirsch, 1988; Reference Thornicroft, Bisoffi and De SalviaThornicroft et al, 1993; Reference Glover, Robin and EmamiGlover et al, 1998), but did not control for potentially important confounders. In this study we used the level of deprivation of area of residence as an indicator of individual socio-economic position (Reference Salmond and CramptonSalmond & Crampton, 2001).

METHOD

Counties Manukau Mental Health Services cover the mostly urban South Auckland district which has 378 000 residents. South Auckland has a high proportion of Maori (18%) and Pacific Islanders (17%) and is deprived relative to most of New Zealand. The study site was the 45-bedded psychiatric in-patient unit, which is managed by three consultants. Community care is provided by five teams, with no day hospital. The cohort comprised consecutive admissions from within the district from 1 November 1999 to 31 July 2000. We excluded patients from outside the area, patients readmitted during the study period and homeless people who had no address to code for area deprivation.

Deprivation was measured using the NZDep96 deprivation index, which was created from 1996 census data (Reference Salmond, Crampton and SuttonSalmond et al, 1998) available for all small areas in New Zealand. A small area is defined as one meshblock (the smallest geographical area for statistical purposes (median population about 90 persons)) or two mesh-blocks. The NZDep96 index is a weighted combination of the proportions, in a small area, of nine variables, such as being on a means-tested benefit or lacking a specified resource (e.g. qualifications or a household telephone). The index is split into a quintile scale where 1 represents the least deprived 20% of small areas and 5 the most deprived 20%.

An independent firm assigned a geographical small area code to each patient's address at the time of admission, which enabled the correct area deprivation score to be derived. We defined the most deprived as those living in one of the areas ranked as the most deprived 20% (in accordance with the New Zealand definition of ‘poor populations’), the least deprived as those living in areas ranked among the least deprived 60% and the moderately deprived as those living in areas ranked among the intermediate 20%.

The length of hospital stay was the number of days for the index admission in the study period. Potential confounding variables at individual patient level were obtained from case notes and from interviews with the patient's primary nurse, using structured questionnaires, such as the 10-item form of the Health of the Nation Outcome Scales (Reference Amin, Singh and CroudaceAmin et al, 1999), the Global Assessment of Functioning (American Psychiatric Association, 1994) and the Reasons for Admission schedule (Reference Flannigan, Glover and WingFlannigan et al, 1994). We used the DSM–IV principal diagnosis and any comorbid diagnosis stated in the discharge summary.

We analysed the data using STATA version 6 using the log of the length of stay and the geometric mean, because of the log-normal distribution of length of stay. We used generalised linear modelling for the effect of deprivation, in three categories, using a multiplicative model.

RESULTS

Of the 379 index admissions in the study period, 7 patients were homeless and 50 were from outside the catchment area. For 291 of the remaining 322 patients (90%), enough information was available to enable coding at small area level. There were 166 males (57%) and the mean age of the sample was 36 years. Ninety-nine patients (34%) identified themselves as Maori, 116 as European (40%), 32 as Pacific Islander (11%) and 44 as Asian or other (15%). Three-quarters of the sample were single, widowed, divorced or separated and 43% lived in areas defined as ‘most deprived’, compared with 20% of the national population. One hundred and forty-three patients (49%) had a principal diagnosis of schizophrenia, 108 (37%) of a mood disorder and 41 (14%) another principal diagnosis, with 140 (48%) having a comorbid diagnosis. The geometric mean length of hospital stay was 16.6 days. One hundred and ninety-eight patients (68%) had been admitted involuntarily. The mean number of previous admissions to the psychiatric in-patient unit was 1.6 and the mean length of illness was 101 months.

For those from most deprived areas, the length of hospital stay was 21 days, compared with 12 days for those from the least deprived areas. After full adjustment for confounding variables (Table 1), this was 22 days for those from the most deprived areas, compared with 15 days for those from the least deprived areas. Those from moderately deprived areas also had a longer length of stay than those from the least deprived areas.

Table 1 Multiple regression showing the effect of deprivation on the length of hospital stay for index admissions (n=291)

Deprivation	Comparison with least deprived group	Length of stay, days1	Likelihood ratio test for effect of deprivation, (d.f.=2) χ2	P
Least deprived (quintiles 1-3)		11.8
Moderately deprived (quintile 4)	0.52	19.8	13.48	0.001
Most deprived (quintile 5)	0.55	20.5
Adjusted for demographic factors; diagnosis; severity; non-adherence; physical disability; and service factors2
Least deprived (quintiles 1-3)		15.3
Moderately deprived (quintile 4)	0.33	21.3	6.09	0.048
Most deprived (quintile 5)	0.36	21.9

Principal diagnosis was the main contributor to variance (13%), followed by psychiatric symptom severity/function/chronicity (8%), small area deprivation (6%) and the identity of the consultant psychiatrist (3%).

Individual measures of socio-economic position (individual unemployment, occupational class, housing tenure, being on a benefit) each added only 1–2% to the explanatory power of a model for length of hospital stay containing demographic, clinical and service factors.

DISCUSSION

Lower socio-economic position, as measured by deprivation of small area of residence, was independently associated with increased length of hospital stay. Although principal diagnosis explained more of the variance, the association between deprivation and length of hospital stay remained after accounting for demographic and clinical factors and differences between clinicians. This is consistent with ecological studies (Reference HirschHirsch, 1988; Reference Glover, Robin and EmamiGlover et al, 1998) and with a study which stratified according to diagnosis (Reference Thornicroft, Bisoffi and De SalviaThornicroft et al, 1993). Our findings may be at variance with a study that found no association (Reference Weinberg, Greaves and CreedWeinberg et al, 1998) because we used a measure of deprivation (the NZDep96) that is less prone to measurement error, being applied at a spatial level of 90 persons (Reference Salmond, Crampton and SuttonSalmond et al, 1998; Reference Salmond and CramptonSalmond & Crampton, 2001).

Selection bias is an unlikely explanation, as healthcare is geographically sectorised and little private care is available. Furthermore, the association remained after diagnosis, severity and length of illness had been controlled for. We controlled for most potentially important confounders other than social support. We are not able to say whether the effect of deprivation is at the individual, household or area level. Our data suggest that place may be at least as important as person and that moderate deprivation also has an effect.

Conditions in deprived neighbourhoods (few employment opportunities, restrictive work environments, social fragmentation and poor services) might have an adverse effect on those with mental disorders and their carers (Reference Macintyre, Ellaway and CumminsMacintyre et al, 2002; Reference Allardyce, Gilmour and AtkinsonAllardyce et al, 2005). Several study participants would have been left alone all day if discharged, either because they lived alone or their families worked long and unsociable hours. This, combined with poor opportunities for local employment and poor public transport, contributed to a long length of hospital stay while awaiting daytime placement. Other patients had comorbid physical illness which was aggravated by poor housing. Individual, household and neighbourhood social circumstances should be taken into account in discharge planning and in optimising access to community care. Research is needed to develop hypotheses about group-level factors that may explain the onset and outcome of mental disorders (Reference O'CampoO'Campo, 2003).