Abstract
Keywords
Evidence has been accumulating since the 1930s that needs for mental health care are higher among those exposed to relatively greater socio-economic deprivation [1–4]. Psychiatric service utilization correlates with individual-level socio-economic indicators, such as low social class and living alone [5, 6] and with composite indices of deprivation derived from census data [7, 8]. A key implication of this is the need to take area deprivation into account when allocating resources in mental health.
In New Zealand, over 90% of secondary mental health care is provided free of charge via government funding. The national mental health strategy gives benchmarks for a range of service components, such as beds [9]. However it does not discuss the gradient required in those benchmarks in relation to deprivation. Counties Manukau District Health Board Mental Health Services operates with a smaller number of acute adult beds (11.6 per 100 000 total population, benchmark 12.8), of extended care beds (3 per 100 000, benchmark 11.2), of 24 h nurse-staffed residential places (31 per 100 000, benchmark 37.3), and of community mental health service staff (22 per 100 000, benchmark 42) than have been recommended (Sue Hallwright, Northern District Support Agency, pers. comm. 2002) [9–11]. This area also has one of the greatest concentrations in New Zealand of people living in areas ranked as most deprived [12]. The first aim of this study was to describe the relationship between deprivation and measures of utilisation. The annual period prevalence of admission (i.e. the number of persons admitted per 100 000 per year) is considered to be less susceptible to bed availability than the total admission rate and thus to be more robust [8]. Total bed days have also been recommended as a proxy for the cost of care [7]. The second aim of the study was to compare expected bed numbers with available beds.
Methods
Sample
The sample consisted of all persons admitted to Tiaho Mai, the hospital psychiatric unit (HPU) at Middlemore Hospital, Counties Manukau, between 1 June, 1998 and 31 December, 2000 (= 2.58 years). Admissions were recorded routinely by the ward clerk and entered onto the computerised Patient Information Management System (PIMS).
Persons admitted from outside the catchment area were excluded. These were identified by address at admission. In addition, the homeless were excluded, as they had no address to geocode.
Accuracy of routine dataset
A validity check of the PIMS system was carried out by daily ward visiting to collect admission and discharge details on 255 consecutive admissions. Two hundred and forty-five out of 255 (96%) details matched identically on the recording of name, National Health Index unique identifier (NHI), date of admission and date of discharge. Errors in the PIMS were mostly due to inaccuracy in recording of the NHI number, resulting in additional persons apparently admitted. A database check on names and NHI numbers was undertaken, and false records were excluded from the analysis presented here. Discharge date errors were mostly due to the hospital practice of ‘discharging’ persons who were temporarily transferred for general hospital care. Fifty-five out of 482 repeat admissions proved to be ‘false’ as these persons had been temporarily discharged within the hospital, for example for a tooth extraction, and then readmitted to the psychiatric unit within 1–10 days. These 55 were re-coded as having one continuous psychiatric admission in the analysis presented here.
Measures
An admission is defined as a person admitted onto the ward by the duty psychiatrist and entered onto the PIMS. First admissions and repeat admissions were separated using the unique patient identification number.
Discharge was recorded on the PIMS on the day that the in-patient consultant discharged his/her responsibility for the patient.
Occupied bed days were calculated by subtracting the date of admission from the date of discharge. Leave days were included as admission days. Bed days contributed by all those admitted in the study period were included.
The estimated population at risk in the catchment area for 1999 was 247 488 persons aged 15–64 years [13]. Deprivation in an area was measured by the NZDep96 index of deprivation for small areas [14, 15]. NZDep96 was created from 1996 census data and is available for all 20 166 small areas, each having approximately 100 persons living in them. The index is a weighted combination of the proportions of nine variables in a small area. Each variable indicates the proportion of inhabitants in the small area lacking a specified advantage, such as a household telephone. Table 1 shows all variables included in the index. These were standardized for age group and gender, and two were adjusted for household size and composition. Principle components analysis of the variables (standardized small-area proportions) was then used to create the index, which is the score from the first principle component. The overall distribution of the index has been split into deciles to produce a 10-point scale of deprivation where 1 is the least deprived 10% of small areas, and 10 is the most deprived 10%. Weighted average deprivation scores may be calculated for aggregated collections of meshblocks, for example census areas units, as in the present paper.
Description of variables used to de velop NZDep96 (small ar ea index of socio-economic deprivation)
The dataset was geocoded, i.e. a geographical area code was assigned to the street address at admission for each person. The area codes related to census area units (CAU) levels, which have an average population size of 3286 persons in the Counties Manukau District Health Board area. Following geocoding, a deprivation score using NZDep96 was assigned to each patient record.
Principal diagnosis was coded after patient discharge. A trained coding clerk routinely combines the DSM-IV Axis diagnosis from the discharge summary with information from the clinical notes. This is entered into an ‘N coder programme’ (Mental Health Diagnosis Conversion Table ICD-10-AM/DSM-IV) which produces an ICD-10 diagnosis. This resulted in approximately 100 different codes, which were reduced to 10 broad diagnostic categories.
Outcomes
Annual period prevalence of admission = number of individual persons admitted per 100 000 population at risk per year.
Rate of all admissions per 100 000 population at risk per year (i.e. new plus subsequent admissions in the study period).
Rate of total occupied bed days per 100 000 population at risk per year.
Statistical analysis
Data analysis was carried out using STATA Version 6 [16]. The accumulated person years in the different NZDep96 quintiles during the period studied were estimated by using the age and gender-specific population in the Counties Manukau District Health Board area in 1999. The annual period prevalence of admission and the rates of occupied bed days were calculated for the different quintiles of deprivation. The direct method of standardization was used to take account of the different age structures in the different quintiles of deprivation. The New Zealand population in the 1996 census was used as the standard population. Standardized rates are presented as per 100 000 person years.
Following the method developed by Glover [8], the admission prevalence was expressed as a ratio, i.e. the standardized period prevalence of admission for each quintile of deprivation was divided by the standardized average rate of admission across the quintiles. The rate of total occupied bed days was also expressed as a ratio using the same method.
Results
One thousand and twenty-two people were admitted between 1 June, 1998 and 31 December, 2000, of whom 872 were from within the catchment area and comprise the study sample of ‘new admissions’. Two hundred and thirty-eight out of 872 were admitted two or more times during the 2.58-year study period, giving 1299 total admissions.
Characteristics of the 872 people admitted are shown in Table 2. The various characteristics are shown for the first (or only) admission in the study period. The principal diagnoses were as follows: 38% (332/872) schizophrenia and schizoaffective disorder; 24% (210) bipolar disorders; 12% (108) depressive disorders; 8% (73) paranoia/unspecified psychosis; and 8% (69) adjustment disorders.
Characteristics of per sons admitted
The crude annual period prevalence of admission was 137 per 100 000 per year. The crude total admission rate was 203 per 100 000 per year. The crude rate of total occupied bed days was 5714 occupied beds days per 100 000 per year.
As seen in Table 3, the annual period prevalence of admissions, standardized for age and gender, increases with relatively greater deprivation. The standardized prevalence of admission ratio was 1.41 for deciles 9 and 10 (most deprived) and 0.50 for deciles 1 and 2 (least deprived), giving a risk ratio of 2.8. People living in most deprived areas have 2.8 times the use of in-patient beds compared with those living in least deprived areas. The 95% confidence intervals do not overlap comparing deciles 1–4 with deciles 7–10, indicating a significant difference in admission prevalence between these levels of deprivation. The ratio of total admissions (n = 1299) was also calculated, standardized for age and gender. It was 1.67 for deciles 9 and 10 (most deprived) and 0.41 for deciles 1 and 2 (least deprived) giving a risk ratio for total admissions of 4.1.
Characteristics of g eographical areas at dif ferent levels of deprivation in the catc hment zone of Counties Manukau District Health Board 1998–2000, and the ratio of standardized admission prevalence
A ratio of occupied bed days was calculated for the different quintiles, derived from the standardized rate of occupied bed days. This is shown in Table 4. The ratio was 1.46 for deciles 9–10 and 0.42 for deciles 1–2, giving a risk ratio of 3.8. People living in most deprived areas use 3.8 times the number of bed days compared to people living in least deprived areas. The confidence intervals indicate a significant difference between deciles 1–4, deciles 5–6 and deciles 7–10.
Total occupied bed days and deprivation in the catc hment zone of Counties Manukau District Health Board, 1998–2000, and the ratio of standardized rates of occupied bed days
As seen in Table 5 we applied the weights to the different population numbers to calculate estimated bed needs, based on the New Zealand national benchmark of 13 acute psychiatric beds per 100 000 [9]. This method gives a figure of 56 acute beds for Counties Manukau; currently there are 45 beds available. Table 5 also illustrates that those living in Counties Manukau in areas of ‘average’ deprivation (deciles 4–6) have a need for beds close to the national benchmark. A higher bed ratio is required for those living in deciles 7–10, and lower bed ratio for those living in deciles 1–4.
Estimated acute mental health beds for catc hment area of Counties Manukau District Health Board
Discussion
We found a strong association between area deprivation and psychiatric bed utilization. Persons living in most deprived areas of Counties Manukau have approximately three times the admission prevalence of those living in least deprived areas. Total occupied bed days were four times as great for those living in most versus least deprived areas. A similar gradient in admission rates for mental health services has been demonstrated in other countries [7, 8, 17]. This is also consistent with preliminary data for New Zealand populations [18, 19]. Deprivation, as measured by NZDep96, appears a useful predictor of psychiatric bed use. This is not surprising as it includes several indicators that have been shown to be strongly associated with need for mental health care including low income, unemployment, lack of access to a car and single parent households [7, 8, 17].
We also have to consider other possible explanations for the apparent relationship between deprivation and psychiatric bed use. The wide confidence intervals show that there is some imprecision around the estimates of bed utilization. However, the non-overlapping confidence intervals, between deciles 1–4 and deciles 7–10, indicate chance is an unlikely explanation for the findings. Selection bias may be an explanation if people in less deprived areas were going outside the catchment area for psychiatric care. This is unlikely as: (i) health care is sectorized so people have little choice to be admitted elsewhere; and (ii) very little private in-patient care is available. However, in the same way that 15% of total admissions were excluded from the analyses as they were admitted from ‘out of area’, a similar percentage of those from the Counties Manukau District Health Board catchment area needing admission are admitted elsewhere when the local beds are full. While there is no reason to expect these to be selected systematically on the basis of deprivation, it is conceivable that in the unusual circumstance of two consumers needing admission at the same time, that the receiving hospital may select the one with fewer complicating social problems. Confounding is another possible explanation for the association between bed use and deprivation. We were able to take account of age and gender structure of the population, but not, for example, of urban residence, clinician, or ethnicity. Urban rates of schizophrenia are known to be around twice those of rural areas [20]. Clinician or provider preferences are known to have major effects on the way in-patient beds are used [21, 22], although little data exists on the relationship between clinical practice and deprivation of consumers. For this study, the decision to admit was made by the crisis team and/or by the on-call registrars, both working across the whole catchment area. It is therefore unlikely that particular practitioners working in particular settings of deprivation can explain the findings. Maori are known to have higher rates of psychiatric admission than Pakeha (European New Zealanders) for psychoses, and to be over-represented compared to Pakeha in areas of social deprivation [18, 23]. It would be important to know if the gradient is the same after stratifying for ethnicity on a larger sample. Another mediating mechanism between deprivation and bed use may be social isolation. For example, single marital status is associated with severe mental disorder [20], and with having a lower income relative to those with a partner [24].
Much of the explanation of the relationship between socio-economic deprivation and psychiatric bed use will lie in the higher prevalence of mental disorders, such as schizophrenia and depression, in deprived areas [25]. For schizophrenia, Eaton et al. [20], concluded that the difference in rates between lowest and highest social classes is 3: 1. Are such disorders common in poor areas because unwell people select to live there, being unable to afford to live elsewhere, or because conditions in these areas cause the disorders? Both selection and causation may be implicated in explaining the link between low socio-economic status (SES) and severe mental illness (SMI). Selection appears to be an important explanation, especially in schizophrenia, whereby the disability impinges on occupational, income and housing opportunities, and those with the disorder may move progressively downward, most likely because of the initial onset in adolescence when social and occupational skills are learned [26]. Causative factors may work at individual and/or neighbourhood level. At an individual level, causative factors associated with deprivation may work via a range of intermediaries such as being unable to afford to buy psychiatric medication or attend followup; high rates of adverse life events, of substance abuse and of coexisting physical illness [27]; exposure to infections and poor obstetric care in utero [25]; and adverse coping styles and health-damaging behaviours patterned in early life [28]. Contextual conditions in more deprived neighbourhoods, such as poor quality housing and few employment opportunities [29, 30], or negative attitudes to mental illness and lack of community support networks, may also act as risk factors for SMI and for secondary service use.
Access to secondary care
At first glance it appears that access to secondary mental health care is good for people living in the most deprived areas. However, these data are not adjusted for need, which, as discussed above, is likely to be very high in the most deprived areas [30]. Also, higher admission rates may, at least in part, reflect poor access to primary care, poor quality of primary care [31], or poor access to outpatient mental health care [32]. Access to effective, local primary health care is determined by supply factors such as inequitable general practitioner distribution (MPC, 1997) or quality and cultural sensitivity of primary care services, and demand factors such as belief systems about illness [33] and uptake of preventive services [34]. Follow-up care, referral to secondary care, or psychotherapy, may be even less likely for those from ethnic minority groups [32, 35]. Maori first admissions and readmissions have risen over the past 15 years [36], with a greater percentage of involuntary admissions than Pakeha, and a greater percentage of admissions being due to schizophrenia [36, 37]. Durie, reviewing the data on Maori pathways to care, considers several indications that Maori have inadequate access to primary health care (PHC) [38]. These include relatively high referral rates from prisons and other law enforcement agencies, and relatively low referral rates from medical practitioners, day-patient facilities and domiciliary nursing services. Several factors may mitigate against use of PHC including cultural barriers, language barriers [39], health belief systems and the New Zealand system of charging for PHC [40]. Staff may perceive greater barriers in working with those living in extreme deprivation. Such barriers may be influenced by a lack of social capital, whereby those from most deprived areas have experienced low levels of interpersonal trust, reciprocity and mutual aid [41].
Interventions to reduce health inequalities
These will be diverse, ranging from policies to minimize the effect of illness on SES, to reducing inequalities in SES and to increasing the supply of health care in lower-SES groups [27, 42]. In reviewing mental health care, Goldberg [43] has recommended five key areas, including: (i) adjustment of the resource allocation formula to take account of the much greater mental health needs of deprived inner-city areas; (ii) improved facilities for working women; (iii) improved preschool teaching for disadvantaged groups; (iv) interventions to reduce substance abuse; and (v) a public information campaign targeted at ethnic minority populations and deprived elderly. Equitable resource allocation and initiatives to improve mental health in ethnic minority groups have been highlighted in parts of Australia [44, 45]. This may allow development of a spectrum of services, including supported accommodation, assertive outreach teams and in-patient beds [46, 47]. It may also allow care to be provided for a range of disorders outside of psychotic illnesses, as is more the case in less deprived areas [31, 43]. In terms of bed numbers, our data suggest that a higher bed-to-population ratio than 13 per 100 000 is needed for more deprived areas. Counties Manukau has one of the highest concentrations in New Zealand of people living in areas ranked as the most deprived [13, 48]. Its bed ratio per head of population, and its ratio of supportive accommodation places and of community mental health nurses, should be among the highest in New Zealand, yet currently they are amongst the lowest (Sue Hallwright, Northern District Support Agency, pers. comm. 2002). This relative under-supply of mental health resources in areas of high deprivation has been described in other settings [49]. Goldberg argues that the likely benefits of more beds for deprived areas will include less ward overcrowding, less waiting for basic items of service, including community mental health care; less premature discharge of dangerous consumers into the community, and less staff burnout [43].
In New Zealand, further research is needed to see if a similar gradient in mental health service utilization, including use of community mental health services, is found across the country. Any interventions to reduce inequalities should be carefully evaluated [50].
Footnotes
Acknowledgements
We thank Counties Manukau District Health Board Mental Health services, especially Sue Wyeth, Rob Kydd and Murray Patton, for their support. We also thank the Oakley Mental Health Research Foundation and Auckland Medical Research Foundation, Sharon Pearce, Sue Hallwright and Clare Salmond.