To contents page

To Psychiatric Activism Page

Chapter 1

Readmission: An overview of empirical research

Myself when young did eagerly frequent

Doctor and Saint, and heard great argument

About it and about: but evermore

Came out by the same Door as in I went.

- Edward Fitzgerald 1809-1883

Psychiatry does not have a distinguished research tradition in South Africa, qualifying for little more than a cursory mention in both Brink's (1988) and Malan's (1988) histories of medical research. Although it could be argued that these histories are biased against psychiatry as one of the less glamorous medical specialties, empirical evidence on publication impact and frequency confirms South African psychiatry's relatively low research profile. While other branches of South African medical research are highly regarded internationally - local work in 'General and Internal Medicine' ranks seventh in the world according to bibliometric and citation indices (Pouris, 1989) - psychiatry is listed among those scientific disciplines "in which South Africa does not make any international contributions" (p. 625). Thus when the Medical Research Council (MRC) established its Clinical Psychiatry Research Unit in 1980 there was a considerable backlog of research and a need to identify the most urgent priorities. Gillis (1987), who was appointed as first head of the unit, summarised the unit's research agenda as follows: "The areas which were selected as being of greatest concern were mental illness in the aged, the high readmission rate to South African psychiatric hospitals (approaching 50% overall and rising), and substance abuse, particularly marijuana, as a concomitant or cause of mental illness" (p. 797). The importance of the readmission question was later endorsed in an official statement by the Society of Psychiatrists of South Africa (Ben-Arie & Nash, 1986).

It is disturbing that in the face of issues such as the lack of recreation areas and poor sanitation in many South African psychiatric facilities (highlighted in a survey by Visser, Haasbroek and Bodemer, 1989), the mental health effects of poverty and apartheid, the torture of detainees by security police, the generally appalling conditions in psychiatric institutions for blacks and the extreme racial disparities in outpatient facilities (all of which were highlighted in a paper by Dommisse, 1987, published in the same year as Gillis' policy statement), the MRC and the Society of Psychiatrists should have opted for a relatively safe topic such as readmission.

Nevertheless, from the point of view of emulating a certain international research tradition, they chose well (as will be shown) in fixing on readmission as a priority area, and within five years of its founding the Clinical Psychiatry Research Unit was able to report (Gillis, Sandler, Jakoet & Dickman, 1985) on what it construed as an alarming rise in readmissions to South African psychiatric hospitals, with 45% of 1984 admissions nationwide being readmissions. In one study of patients in a mental hospital in Cape Town (Gillis, Sandler, Jakoet & Elk, 1986) it was found that 26.5% of white, 41% of coloured and 42% of black patients were readmitted within a year.

What causes South African psychiatric patients to be readmitted so frequently? The answers thus far suggested by Gillis and his colleagues (Gillis et al., 1985; Gillis et al., 1986; Gillis, 1987, 1988; Sandler & Jakoet, 1985) invoke both general historical factors (which account for the overall rise in readmissions) and specific causal variables (which account for individual differences in readmission frequency).

The historical factors are explained as follows by Gillis (1988): "Major changes in the pattern of admissions, readmissions and community and outpatient attendances have occurred in the last 10-15 years. Inpatient admissions have decreased to a remarkable extent and outpatient and community services have increased about 400%" ( p. 303). The result of this shift is that there are now upwards of half a million community and outpatient attendances per year as compared to around 30 000 inpatient admissions and, given this change of emphasis to community (as opposed to custodial) care, cases of relapse are now less likely to occur in the hospital setting and are therefore more likely to require (temporary) rehospitalisation.

The process of deinstitutionalisation has been uneven in South Africa, with many facilities (particularly for Black patients) remaining essentially custodial in nature (Dommisse, 1987; Freeman, 1991). George (1988) reports that, despite continued population growth, there was a decrease in the number of patients resident in institutions financed by the state from 25 881 in 1975 to 19 576 in 1985. In the then exclusively white psychiatric hospital where the empirical study reported on in this dissertation was conducted, the number of hospital beds were reduced from 200 (Moross, 1969) to around 100 over a 25 year period. The Oranje Hospital in Bloemfontein, on the other hand, which caters mainly for Black patients, achieved a similar reduction (from 235 to 100 beds) over a three year period from 1985 to 1987 (Fourie & Gagiano, 1987)(1). Taken overall, psychiatric inpatient numbers in South Africa appear to have started declining sharply from the second half of the 60s. Between 1964 and 1976, inpatient admissions declined by 27%, while outpatient visits went up by 834%; altogether 8800 psychiatric inpatient beds were done away with between 1961 and 1989 (Visser, Haasbroek & Bodemer, 1989). However, by the late 1980s South Africa was still lagging in terms of deinstitutionalisation with around one bed in a psychiatric institution for every 1000 population, compared to the WHO recommendation for Europe of 1 per 2000 population (Freeman, 1991).

In terms of specific causal variables, Gillis and his colleagues (1986) were able in one study to isolate as significant predictors (for white patients): Living alone, mixed substance abuse and having no supportive relationship, and (for coloured patients) being male and being diagnosed as schizophrenic. The only variable which was predictive across racial groups was having more than one previous admission in the past five years. Other factors which have been blamed for readmission in local research include 'job-related difficulties', lack of social contact, dependency on others, the effects of stigma, non-compliance with medication (Strong, 1987; Gillis, Trollip, Jakoet & Holden, 1987; Gillis, Koch & Joyi, 1989), and high levels of 'expressed emotion' (Ben-Arie, 1988) or a lack of understanding in patients' families (Moross, 1969).

Although the MRC team suggested various steps which may be taken to limit readmission (more careful surveillance of outpatient compliance with medication, bolstering social support systems and so on), none of these interventions have been shown to have a major impact on readmission (Ben-Arie, Koch, Welman, & Teggin, 1990). However, as Gillis et al. (1986) are quick to point out, the failure of any intervention to prevent readmission should not necessarily be interpreted as an index of the inefficacy of psychiatric treatment, but may simply be due to the chronic and recurrent nature of mental illness itself.

The main themes characteristic of South African academic literature on readmission outlined above - contextualising the problem within an historical trend towards outpatient care for the mentally ill, a (usually unsuccessful) search for specific causal variables and preventative interventions, and an attempt to normalise readmission as an inevitable part of mental illness - closely resemble similar trends in the international psychiatric literature, which is reviewed below.


Although the term 'deinstitutionalisation' only came into use in the mid-70s (Morrisey, 1982), the ideal of phasing out mental hospitals started in the early 50s and was already well under way during the 60s. As in South Africa, the readmission problem is internationally ascribed to this process (Dorwart & Hoover, 1994). Lehman, Possidente and Hawker (1986) report that "since deinstitutionalization began more than two decades ago, the inpatient census of public mental hospitals has been reduced by more than 70 percent." (p. 901) According to Levine and Perkins (1987) there was a drop in the mental hospital population in the United States in this period from 600 000 to 200 000, with a concomitant growth in community mental health centres from nil to around 750. Expressed in terms of 'market share' (counted as care episodes), state mental hospitals in the United States have seen their share drop from nearly 50% in 1955 to only 9% in 1975, with resident patients accounting for 75% of episodes in 1955, but only 29% in 1980 (Morrisey, 1982).

The process of deinstitutionalisation, although a global phenomenon, varied considerably in precipitating factors and chronology.

Krauss and Slavinsky (1982) attribute deinstitutionalisation to the discovery of more effective psychiatric medications in the 1950s: "The widespread use of chlorpromazine and related neuroleptics was the single greatest contributor to a reduction in the number of patients residing in mental hospitals, to the shortened length of stay, and to a shift to community-based care." (p. 74). Chien (1981) also endorses this view, pointing out that by the early 80s more than 10 000 scientific papers had been published on psychotropic medication, and that the pharmaceutics industry which had arisen around psychiatry was already worth billions of dollars per year.

Mechanic (1986) disputes this view, arguing that economic, rather than scientific, factors were mainly responsible for deinstitutionalisation in the United States:

Superficial historical overviews link the reduction of public mental health populations to the introduction of the phenothiazines in the middle 1950s, but in reality individual states varied significantly in the timing and rates of deinstitutionalization. Large scale relocation of chronic mental patients began only ten years later with the introduction of Medicaid and improvements in the social welfare system (p. 892).

Mechanic (1986) argues that expectations that outpatient care would be either cheaper or a less direct tax burden were the main driving force behind deinstitutionalisation, with American states shifting costs to federal programs by transferring patients from mental hospitals to alternate facilities such as nursing homes.

A third precipitating factor, and one which is often invoked when discussing the undesirable consequences of deinstitutionalisation, is the ideological critique of institutional psychiatry developed by the anti-psychiatry movement (reviewed in Dain, 1989).

Finally, explicit political initiatives, such as John F. Kennedy's 1963 address to Congress in which he called for a 'bold new approach' to mental illness and which resulted in the Mental Retardation facilities and Community Mental Health Centers Act of 1963 (Krauss & Slavinsky, 1982), also contributed to deinstitutionalisation. In Italy, which had perhaps the most radical approach to deinstitutionalisation (Basaglia, 1981; Crepet, 1990), Law 180 of 1978 brought in sweeping changes aimed at the dismantling of mental hospitals, to be replaced by mental wards in general hospitals and community care. According to Law 180 there were to be no first admissions after May of 1978 and no new admissions after December 1981. Surprisingly, this has been largely complied with (although patients have sometimes been disguised as 'guests') and the mental hospital census in Italy is still dropping (Tansella & Williams, 1987).

It appears that internationally the readmission crisis, which followed on deinstitutionalisation, set in almost immediately. Already in 1968 Crumpton could complain:

Most admissions to psychiatric hospitals are discharged within six weeks. However, our problems now are not so much new admissions, but of readmissions (p. 15).

In 1964 for every new schizophrenic patient admitted to a British mental hospital, two more were readmitted, and by 1968 patients who had been readmitted ten or more times were by no means rare (Crumpton, 1968). This state of affairs has been termed the 'revolving door' of psychiatric admissions and readmissions.

Of course not all patients released from mental hospitals become part of the 'revolving door' pattern. Lerner, Popper and Zilber (1989) followed a 10% nationwide random sample (n=832) of all psychiatric patients hospitalised in Israel in 1980 over a four year period and found three main hospitalisation patterns: 50% were admitted for a single, short hospitalisation; 22% accumulated a year or more of hospitalisation over several rehospitalisations during the period; and 8% were admitted for one long hospitalisation of at least a year. The mean number of hospitalisations during the four and a half years of follow-up was 2.62. Some studies have reported a readmission rate as high as 64% (Schnur, Friedman, Dorman, Redford & Kesselman, 1986).

Psychiatry is not the only branch of medicine to have had to contend with high readmission rates. According to Soeken, Prescott, Herron and Creasia (1991), reported one-year readmission rates range from 3% for surgical patients to 66% for patients with chronic congestive heart failure. As in other fields of medicine, however, it appears that there is a group of patients who are particularly prone to rehospitalisation, and the challenge becomes to empirically identify the distinguishing features of these revolving door patients, also known as 'new chronics' (Casper, Romo & Fasnacht, 1991) and heavy or frequent users (Casper & Pastva, 1990), in order to predict and possibly prevent future readmission.

Predicting and preventing readmission

A typical example of the kind of research conducted with this purpose in mind is a study by Zeff, Armstrong, Crandell and Folen (1990). They set out, in a sample of 246 psychiatric patients in a military psychiatric facility (mainly first admissions and mainly young males), to isolate those factors predictive of rehospitalisation within 90 days. Not having any particular theoretical perspective, and since a host of factors had been implicated in previous readmission research, they used no fewer than twenty independent variables (including age, sex, branch of service, rank, marital status, number of children, diagnosis, number of days hospitalised, past psychiatric admissions and medication). Of these, four proved to be statistically significant in predicting rehospitalisation; however, together these variables accounted for only about 10% of the variance in readmission outcomes.

While these sorts of results are disappointing, it is through the accumulation of findings from such studies, so it is hoped, that a solid scientific understanding of the causes and cures for readmission will arise. Empirically-minded researchers will readily acknowledge that there may be technical difficulties in rendering findings from different studies comparable, for instance due to wide variation in sample composition and the ways in which readmission is measured, but the specific facts yielded by each study are believed to add to the growing body of knowledge on readmission. The facts thus far accumulated are surveyed below.

Demographic factors

Anecdotally, revolving door patients tend to be young males (Bender, 1986), and demographic variables are therefore often seen as potential risk factors for readmission. Hughes, Joyce and Staley (1987) compared 110 readmitted psychiatric patients (of whom 19 were multiple readmissions within 1 year) to 100 non-readmissions and found that there were no demographic differences between the two groups. In a much smaller sample (n=38) of depressed and non-depressed patients Hirschfeld, Klerman, Andreasen, Clayton and Keller (1986) similarly found that none of a range of demographic variables predicted readmission over a two-year period. However, in some studies particular demographic variables have been found to be statistically significant in predicting readmission.

Gender, although not a significant predictor in Zeff, Armstrong, Crandell and Folen's (1990) study (which used a largely male sample), may have some predictive value. Both Colenda and Hamer (1989) and Zilber, Popper and Lerner (1990) found that male gender was predictive of low probability of readmission for first-time patients, although not for non first-time patients, with first-time nonwhite females being at greatest risk in Colenda and Hamer's study. It is not clear how these findings relate to the stereotype of 'new chronic' patients as consisting mostly of young males.

There is some evidence that age may play a role (e.g. Kastrup, 1987; Hadley, McGurrin, Pulice & Holohean, 1990) with patients who over-utilise services tending to be relatively young. However, in other studies (e.g. Zeff, Armstrong, Crandell and Folen, 1990) age was non-significant. As with gender, age may be differentially related to readmission for first-time and non-first-time patients. Zilber, Popper and Lerner (1990) found that for patients hospitalised for the first time in their life, age was negatively correlated with probability of readmission (but not cumulative hospitalisation), but for patients who were not first-timers, age was positively correlated with cumulative length of stay.

Similarly, Zilber, Popper and Lerner (1990) found that marital status was conditionally related to readmission in that being single was predictive of long cumulative stay for non-first-time patients, but not for first time patients. Zeff, Armstrong, Crandell and Folen (1990) found marital status to be non-significant as a predictor of rehospitalisation.

Race may play a role in that Colenda and Hamer (1989) found that for first-time patients nonwhite females are at greatest risk. However for non-first-time patients race was not significant. Zilber, Popper and Lerner (1990) also found 'ethnic origin' to be unrelated to readmission for non first-time patients.

Number of children has been investigated in a single study (Zeff, Armstrong, Crandell and Folen, 1990) and found to be non-significant.


A considerable amount of work has been done relating psychiatric diagnosis to probability of readmission and to the related variable of length of hospital stay (reviewed in Parks and Josef, 1997). Several studies have suggested that being diagnosed as suffering from a major mental illness (Colenda & Hamer, 1989; Hadley, McGurrin, Pulice & Holohean, 1990; Kastrup, 1987) or the severity of the diagnosis (Zeff, Armstrong, Crandell & Folen, 1990) may be predictive of readmission.

However, some authors have found no differences in diagnostic category (e.g. Hughes, Joyce & Staley, 1987), while others (e.g., Bender, 1986; Paris, 1988; Zeff, Armstrong, Crandell and Folen, 1990) have suggested that a diagnosis of personality disorder is most predictive of readmission. Somewhat by contrast Zilber, Popper and Lerner (1990) found that (for first-time admissions) personality disorder patients have significantly shorter hospital stays than schizophrenics. Studies by both Grossman, Harrow and Goldberg (1991) and Mojtabai, Nicholson and Neesmith (1997) found that patients with a diagnosis of schizophrenia had much higher relapse rates, particular when compared to patients diagnosed with depression. Mojtabai, Nicholson and Neesmith (1997) also found differential readmission rates for patients diagnosed with personality disorder, with those diagnosed as schizoaffective most at risk and those diagnosed as suffering from adjustment disorder least at risk.

A major problem with this research is the notorious variability of psychiatric diagnosis. Commenting on their chart review study of 99 readmitted psychiatric patients in San Francisco, Surber, Winkler, Monteleone, Havassy, Goldfinger and Hopkin (1987) write: "Even though the patients were almost always diagnosed as having a major mental illness, 40 percent were difficult to diagnose definitively from the records because they received multiple diagnoses or no consistent diagnosis over time" (p. 1113). In Kastrup's (1987) study which tracked all Danish psychiatric patients over a period of 10 years, only 43.5% kept their original diagnosis. Fennig, Craig and Tanenberg-Karant (1994) have shown that clinical diagnoses of psychiatric disorders frequently differ sharply from psychiatric disorders based on structured interviews.

Length of hospitalisation

The major characteristic of psychiatric hospitalisation in the post-deinstitutionalisation era is of course that it no longer consists of lengthy and continuous incarceration, but rather of one or more much shorter stays. Perhaps, so the argument goes, patients are discharged before they are quite ready for life outside the hospital, or before the hospital treatment has had an opportunity to exercise its beneficial effect.

In their review of studies relating readmission to length of hospital stay, Caton and Gralnick (1987) differentiate between uncontrolled and controlled studies. Uncontrolled studies have mostly found no differences in outcome between long and short stays, except that short stays appear to be indicated for nonchronic, nonpsychotic patients. There have been six major controlled studies; five found no differences, while one found that short stays lead to more re-hospitalisation.

De Francisco, Anderson, Pantano and Kline (1980), who in their study did find shorter stays to be related to more readmissions, explain the effect in terms of greater family involvement in the patient's treatment (since there is more time to involve the family), more family involvement in discharge planning, and better discharge planning generally for longer-stay patients. Axelrod and Wetzler (1989) explain the possible beneficial effects of longer hospital stays as follows: "It appears that as patients remain in the hospital, they become better stabilized, develop greater insight into the need for aftercare, and become more willing to comply with recommendations for aftercare" (p. 400-401). Appleby, Desai and Luchins (1993) and Mojtabai, Nicholson and Neesmith (1997) suggest similar reasons for their findings that shorter stays are related to increased risk of readmission.

Taken to its extreme, the contention that longer hospital stays lead to fewer readmissions can of course not be other than true, in that those who are permanently incarcerated are at zero risk of readmission.

Quality of aftercare

As early as 1968 Crumpton blamed the high readmission rate on uncoordinated aftercare, and this continues to be a prominent theme in the literature on readmission. Mechanic (1986) argues that greater co-operation between various agencies and better case management would help substantially in reducing the readmission rate. Empirically, almost any kind of aftercare seems to be effective in reducing relapse (Axelrod & Wetzler, 1989). In one study Bond, Witheridge, Wasner, Dincin, McRae, Mayes and Ward (1989) found that although there were no dramatic differences in the effectiveness of different forms of aftercare, hospitalisation could be 'deflected' in 68% of cases. It is now generally accepted that at least some of the blame for the rise in readmission rates can be attributed to inadequate aftercare (Hadley, Turk, Vasko & McGurrin, 1997). This situation is similar to that obtaining for criminal recidivism where interventions aimed at altering prisoners' life circumstances outside the prison have met with some success (Bedell, Challis, Cilliers, Cole, Corry, Nieuwoudt, Phayane & Zachariades, 1998).

As with length of stay, there is an inherent contradiction in using quality of aftercare as predictor of readmission. The nursing care residences into which many 'deinstitutionalised' patients were 'transinstitutionalised' (Morrisey, 1982) in the 60s and 70s, were no doubt very effective in preventing readmission, but were as confining and dehumanising as the mental hospitals from which they were supposed to protect patients. This point is further elaborated in Chapter 3.


Although patients in mental hospitals are notoriously unreliable at taking the medication prescribed for them (as many as 20% of inpatient schizophrenics avoid taking their medication), outpatients are even more unreliable, with up to 45% defaulting (Crumpton, 1968). Axelrod and Wetzler (1989) found that greater compliance was related to continuity of treatment (particularly having the first outpatient appointment soon after discharge), more incidents of hospitalisation, longer hospital stays, less denial of the need for treatment, and greater perceived need for medication.

Unfortunately, findings on the relapse-preventing properties of psychotropic medication are equivocal. With regard to antidepressant medication, Baldessarini (1989) contends that:

Evidence for true prophylaxis against subsequent recurrences of major depression is still meagre. Although a few controlled, prospective follow-up studies of 1 to 3 years' duration indicate the TCAs and perhaps lithium may have a moderate preventive (or recurrence-aborting) effect in the treatment of recurrent depression, little is known about the optimal choice of drug or of dose or about safety or efficacy after several years of such treatment. Moreover, the average degree of benefit of TCA and lithium treatment over a placebo is not impressive after the first 6 to 12 months (p. 124).

Neuroleptic medication is not entirely unproblematic as a form of rehospitalisation prevention either. Commenting on the results of a study in which a group of schizophrenic patients had their dosage cut by half, Faraone, Cirelli, Curran and Brown (1988) write:

It is striking that 45% of our original 29 patients remained stable for three years on 50 percent of their previous neuroleptic dose considering that they had been treated for a mean of 23 years, that none were being treated with megadoses before the dose reduction, and that their current doses are in most instances below the usual therapeutic range. Clearly, a substantial subgroup of schizophrenic patients can be maintained on neuroleptic doses far below the current standard (p. 1208).

Previous admissions

In a much-cited paper, Rosenblatt and Mayer (1974) concluded that the number of previous admissions was the only variable consistently predictive of rehospitalisation in the studies they reviewed, and this appears still to be the case for more recent studies (e.g. Bond, Witheridge, Wasner, Dincin, McRae, Mayes & Ward, 1989; Buel & Anthony, 1973; Casper, Romo & Fasnacht, 1991; Casper & Pastva, 1990; Colenda & Hamer, 1989; Hughes, Joyce & Staley, 1987; and Zilber, Popper & Lerner, 1990).

However, a single study (Zeff, Armstrong, Crandell & Folen, 1990) found number of previous admissions to be statistically unrelated to readmission frequency.

The Family

A patient's family inevitably plays a large role in the course of her illness, and is often the base to which she returns after hospitalisation. Even where a patient lives away from family, they may continue to exert a strong influence. In a study of 73 chronic patients living away from home, Wilk (1988) found that only 26% had not seen their families in the past two weeks.

Writing relatively soon after the initial wave of deinstitutionalisation, and at a time when antipsychiatric ideas still held some sway, Simmons and Freeman (1967) claimed that the clearest indicator of a return to hospital is a lack of family tolerance for deviance. More recently the literature has tended to emphasise not so much family members' acceptance of the patient's 'eccentric' behaviour, but rather their understanding of and cooperation with the official psychiatric constructions and management of the problem.

Grunebaum and Friedman (1988) identify four key areas for mental health professionals wishing to build 'collaborative relationships' with families of the mentally ill:

1. Give the family a chance to be heard.

2. Impart information on the rules and structures of the mental hospital and the nature of the patient's illness, e.g. its prognosis and treatment.

3. Help families deal with feelings arising from hospitalisation, e.g. guilt, fear, anger, and depression.

4 Identify the family's coping pattern, e.g. denial, hypercontrol/intrusiveness, lack of distance/separateness.

5. Assist with the ethical dilemma of the family's versus the patient's needs.

In a similar vein Jacob, Frank, Kupfer, Cornes and Carpenter (1987) describe a day-long workshop for unipolar depressed patients and their families, the objectives of which were, inter alia, "to validate unequivocally that major depression is a legitimate medical illness over which the patient cannot exercise voluntary control" (p. 969) and to inform patients and family "of the high likelihood (50 percent) of having a recurrence of depression within one's lifetime and of the exceptionally high recurrence rates (up to 80 percent) among patients who already had three episodes" (p. 970).

There is some empirical evidence that such efforts to recruit families into psychiatric ideologies and treatment plans help to reduce the probability of rehospitalisation (Hughes, Joyce & Staley, 1987; De Francisco, Anderson, Pantano, & Kline, 1980).

Expressed emotion

Antipsychiatrists such as Laing and Cooper have long maintained that madness - when it is not "some sort of lost truth" (Cooper, 1967, p. viii) - must be the result of the duplicitous and disempowering ways in which society and the family communicate with those who become mad. This idea survives, in a reified and insipid form, in research on Expressed Emotion (EE) in families. Invented in the early 70s by the group of British psychiatrists around Vaughn and Leff (1976a, b), the degree of Expressed Emotion in families, and in particular the number of critical comments aimed at the schizophrenic family member, has repeatedly been found to be predictive of relapse and readmission (reviewed in Leff, Berkowitz, Shavit, Strachan, Glass & Vaughn, 1989). What makes the EE concept particularly attractive for mainstream psychiatrists is that it may be useful in identifying that group of patients who are readmitted despite adequate medication and compliance (Herz, 1984; Miklowitz, Goldstein & Nuechterlein, 1988; Schnur, Friedman, Dorman, Redford & Kesselman, 1986).

However, the EE literature has also been strongly criticised. In their review Kanter, Lamb and Loeper (1987) point out that EE is only related to florid positive symptoms in schizophrenics (i.e. the kinds of symptoms which are likely to lead to rehospitalisation), not the more lasting negative symptoms such apathy, passivity, and withdrawal. The direction of causality in EE is also unclear - is it patients' impossible behaviour which causes families to become hyper-critical or are patients driven mad by hyper-critical families? Finally, EE may actually be harmful because it tends to place the blame on families who may be innocent.

EE has also been criticised for the unnecessary complexity of the construct and of the instrument (the Camberwell Family Interview) used to measure it. Hooley and Teasdale (1989) found that the single best predictor of relapse was not Camberwell Family Interview scores, but a patient's response to the simple question "How critical is your spouse of you?"

Integrating empirical findings

There are at least two mainstream psychiatric responses to the kinds of piecemeal findings relating to gender, age, marital status, race, number of children, diagnosis, length of hospital stay, medication, previous admissions, quality of aftercare, family involvement and expressed emotion presented above.

The first is to pretend that, although as yet inconclusive, the findings will eventually - and probably sooner rather than later - add up to something more substantial and coherent. Thus despite the rather dismal results of their review of studies relating to length of hospital stay and readmission (discussed above), Caton and Gralnick (1987) remained confident that "an empirically based policy on length of stay is within reach" (p. 862). Similarly, when all the variables (age, marital status, diagnosis, previous hospitalisations, etcetera) in Zilber, Popper and Lerner's (1990) study accounted for only 14.7% of the variance in the mean duration of hospitalisation (in non-first-time patients) they did not conclude that the enterprise is hopeless, but suggested that "adding variables related to personality characteristics and availability of services would probably explain part of the residual variance" (p. 148). Such visions of an imminent (but somehow always deferred) resolution to empirical problems is typical of psychiatric research (and most probably of positivist research generally) and is reviewed at greater length in Chapter 4.

Steps may even be taken to hasten the hoped-for breakthrough, for instance by advocating the use of larger sample sizes (Mojtabia, Nicholson & Neesmith, 1997) or quantitative meta-analyses of previous research. However, Kastrup's (1987) work with a very large sample (the entire population of Danish psychiatric patients) suggests that increasing the sample size is in itself not likely to be a solution. Due to the large sample size virtually all variables (diagnosis, sex, age group, proximity to hospital, size of municipality, referring agency, discharge destination, marital status) were statistically significantly related to readmission. However, the amount of variance accounted for remained negligible.

Although no quantitative meta-analyses of the psychiatric readmission literature appear to have been done, Soeken, Prescott, Herron and Creasia's (1991) meta-analysis of 44 non-psychiatric readmission studies is suggestive of what sorts of results may be expected. They found that (among a host of other potential variables) medical readmissions were statistically significantly related only to diagnosis, age, length of initial hospital stay, and prior admissions. All relations were trivial in absolute terms. A meta-analysis of twelve intervention studies showed a non-significant overall treatment effect. Once again, despite these discouraging results, Soeken et al. (1991) are upbeat in their conclusions: "If risk factors for readmission can be identified, then providers can focus on developing interventions aimed at reducing unnecessary and preventable readmissions" (p. 264).

From readmission to chronicity

A less charitable view would be that as findings accumulate, readmission research will become ever more fragmented (readmission is statistically related to factor x, but only for young black male first-time patients diagnosed as schizophrenic) and ever more prone to circular restatements of the problem: The longer patients stay in hospital, the smaller the probability that they will be readmitted; the more a patient has been readmitted, the more likely she is to be readmitted; the more community care comes to resemble hospital care, the more effective it will be in deflecting readmission.

Although piecemeal empirical readmission research as such is not usually criticised in this fashion, there is an admission among psychiatric researchers that the problem itself needs to be retheorised. Not only is a theoretical model necessary if readmission research is to get anywhere (Hughes, Joyce & Staley, 1987), but it is argued that readmission should be accepted as an inevitable consequence of the nature of mental illness. Where readmission has in the past often been taken as sign of failure and used to compare different institutions (Kastrup, 1987; Chambers & Clarke, 1990), it should now be realised that chronicity is part and parcel of mental illness (Mechanic, 1986). Mental health professionals should not blame themselves for failing to prevent readmission, but rather ask if deinstitutionalisation was a good idea in the first place (Gralnick, 1985).

This movement, from considering readmission as an unfortunate and essentially preventable side-effect of deinstitutionalisation to being an indicator of an inherent psychiatric chronicity, and thus of the need for large scale reinstitutionalisation, is described in greater detail in the next two chapters and contextualised within longer-term historical oscillations in psychiatric orthodoxy.

1. According to Gagiano (1990) the reduction was from 415 to 100 inpatients between 1985 and 1989, with a concomitant increase from 20 000 to 50 000 in patients being cared for in the community.

To contents page

To Psychiatric Activism Page