Abstract
In marked contrast to the burgeoning studies of first presentation psychosis in adolescents and young adults, there are few reports of first presentation psychosis in elderly persons [1,2]. To date, studies of psychosis in the elderly have recruited subjects irrespective of whether they were presenting in their first episode of illness or had had multiple previous episodes. Features found to be commonly associated with the development of psychosis in late-life include systematised delusional ideas, often of the ‘partition’ type [3], and hallucinations in several modalities. Thought disorder and negative symptoms, in contrast, are uncommon in this cohort of patients [4]. Other associations include a predominance of females [5], sensory impairment [6], family history of schizophrenia [7,8] or affective disorder [9]. A number of recent studies have found greater abnormalities on neuroimaging in subjects with lateonset psychosis than in controls. These changes include increased ventricular-brain ratio [10], cortical atrophy [11], silent cerebral infarction [12] and increased deep white matter hyperintensities [13]. Interpretation of these findings remains inconclusive as there is still debate about their specificity for lateonset psychosis, and the extent to which neuroimaging changes can be regarded as ‘normative’ for this age group [14,15]. Subtle neuropsychological deficits have also been identified in, at least, a subgroup of late-onset psychosis subjects [16]. Comparison between studies, however, has been limited by different lower age cut-offs (45 years or 60-65 years [17]) for determining whether psychosis can be defined as late-onset. Increased cognitive deficits have been noted in the studies using older samples, suggesting that age per se may be a compounding factor in this finding [18].
Few studies of newly diagnosed late-onset psychosis subjects have used a prospective methodology. Most commonly, cases have been identified from psychiatric registers or hospital records, and then followed up by clinical interview if the patient was alive and able to give informed consent [5,19]. Using the Camberwell Psychiatric Case Register, Holden [19] identified 37 cases with late-onset psychosis and followed these up for 10 years or until death. Although all cases appeared ‘functional’ initially, significant heterogeneity was found over the longer term. Thirty-five percent of the sample progressed to dementia and were differentiated by having greater sensory impairment, less family history of psychiatric illness, poorer response to treatment and greater mortality over the 10-year period of the study. The remaining 65% had a greater frequency of first-rank symptoms and their course of illness was more typical of early onset schizophrenia. The Camberwell Psychiatric Case Register was also utilised by Castle and Murray [5] to identify 470 cases of psychosis with first psychiatric service contact between 1965 and 1984. They found significant gender differences, with only 4% of male patients first manifesting their illness after 65 years of age in contrast to 18% of female patients presenting after this age. Jorgensen and MunkJorgensen [20] studied hospital records of 106 first admission patients aged 60 years and over with a diagnosis of schizophrenia or paranoid disorder. In their study, the female to male ratio was 2:1. Seventy-seven percent of the sample had persecutory delusions, 45% had hallucinations, and only 1% had evidence of thought disorder. Flint et al. [12] reviewed all cases over 60 years of age presenting to the Clarke Institute of Psychiatry between 1972 and 1987 and identified 33 cases (total number of cases not provided) with delusions and/or hallucinations in the absence of affective disorder or dementia. They found that those patients without perceptual disturbance, that is, consistent with a diagnosis of late-onset paranoia (36%), were more likely to have unsuspected cerebral infarction detected by computerised tomography (CT) brain imaging. They postulated that cerebrovascular disease may play a significant role in the development of the less ‘schizophrenic’ types of psychosis that develop in old age.
Probably the most comprehensive analysis of issues pertaining to late-onset psychosis has been provided by a prospective study of 101 subjects who fulfilled Roth's [21] criteria for ‘late paraphrenia’, a broad diagnostic construct which includes schizophrenia-like and delusional states that arise in later life. In a number of published reports [3,9,11,16,17] this Maudsley research group investigated issues concerning the nosology, phenomenology, demography, family history and neuroimaging changes associated with late-onset psychosis. These authors also explored the unique delusional phenomena (i.e. ‘partition delusions’) that have been described as characteristic of these elderly patients [3,4]. Using cluster analytic techniques [16], they identified two subgroups on the basis of clinical, neuroimaging, and neuropsychological variables. Compared with the cognitive performance of elderly controls, one group were only impaired in specific areas of executive functioning, while the other group showed widespread cognitive impairment. In addition, the higher functioning group exhibited more florid psychotic symptomatology, but fewer neurological signs and negative symptoms. The authors postulated that different pathogenic mechanisms may underly these two ‘types’ of late-onset psychosis, but that further empirical testing is required. Methodologically, this study was limited by recruiting a sample comprised of elderly psychotic patients presenting for the first time with psychosis, together with patients with previous treatment histories.
The rationale for recruiting late-onset psychosis subjects in their first presentation of illness derives substantially from the growing literature in this area relating to early onset psychosis [1,2,22,23]. As the relationship between early and late-onset psychosis is still being vigorously debated [4,5,8,17], the applicability of this research approach for the latter warrants consideration. McGlashan [1], McGorry et al. [2] and others [23,24] have found, through their studies of premorbid vulnerability factors and early phenomenological changes in first presentation samples, that early detection increases the prospects of better treatment response to positive symptoms, minimises the emergence of deficit syndromes and produces better psychosocial outcomes. This ‘seachange’ in the approach to the investigation and treatment of early onset psychosis has implications for the study and management of similar syndromes emerging in later life. Although deficit or negative symptoms are not prominent in late-onset psychosis, there is increasing research evidence that a subgroup who present initially with a ‘functional’ psychosis in old age will deteriorate cognitively, while others will continue to function independently if their psychotic symptoms are controlled [16]. Early identification of these apparent subgroups would assist greatly in developing appropriate management strategies for these patients. As has been the experience in younger psychotic cohorts, the study of first presentation samples would also facilitate clarification of early clinical changes without the contaminating effects of prior hospitalisation [1], and the iatrogenic effects of neuroleptic medication that commonly occur in elderly patients [25].
With our ageing population, the incidence of late-onset psychosis will inevitably increase and recent reviews [26,27] have argued for further empirical research that focuses on the interaction of aetiological factors across several domains. Henderson and Kay [26] propose that this investigative strategy should include personality traits, genetic vulnerability, social environmental factors and degenerative changes in brain function. Recruiting subjects during the early phase of their illness [1,23] and using a prospective methodology, preferably with longer term follow-up, would be the most reliable approach for obtaining this data. However, this is not an easy undertaking as the base rate for presentation is relatively low. Henderson and Kay [26] found that hospital records and psychiatric registers indicated an annual treated incidence rate for ‘late paraphrenia’ of approximately 15–20 per 100 000. Population-based studies employing DSM-III-R [28] criteria showed mean annual incidence rates of three per 100 000 for late-onset schizophrenia, and 16 per 100 000 for late-onset delusional disorder. Despite the inevitable time commitment that is required to recruit samples of sufficient size, the first presentation psychosis studies undertaken in younger populations have clearly demonstrated the worth of this approach. As a consequence, there is now a much greater understanding of the antecedent factors, course of illness, and successful intervention strategies with regard to psychosis developing in early adult life [29].
This paper is the first report of an ongoing study of subjects with late-onset psychosis. Although a ‘convenience’ sample was recruited, this methodological approach has been used by other studies in this area [30,31]. In the present report, sociodemographic, psychotic symptom and personality profiles of the sample are described. In a subsequent report, these late-onset subjects will be compared with young first-episode psychosis cases from the Early Psychosis Prevention and Intervention Centre data base [2,32] who fulfil the same DSM-III-R [28] diagnostic criteria. Additionally, heterogeneity within the late-onset psychosis sample will be explored by cluster analytic techniques.
Method
The study recruited subjects from four of six psychogeriatric services across metropolitan Melbourne between November 1994 and July 1997. Both community and inpatient services were requested to refer elderly patients who fulfilled the following criteria: (i) the onset of delusions and/or hallucinations after the age of 60 years; (ii) no prior hospitalisation or treatment for psychosis; (iii) an absence of significant cognitive impairment; (iv) an absence of a primary affective disorder; (v) no history or clinical signs of stroke, degenerative neurological conditions, medical illnesses compromising cerebral function, or significant substance abuse; (vi) a sufficient command of English to be able to undertake the assessment procedure; (vii) willingness to give informed written consent to participate in the study.
Sixty-three patients were initially referred and assessed by the investigator (AH). All subjects were seen within the first few weeks of being referred for psychiatric assessment. Seventeen subjects did not subsequently fulfil inclusion criteria after a screening assessment was undertaken. This comprised an explanation of the subject's involvement in the study, the Standardised Mini-Mental State Examination (SMMSE) [33], the Beck Depression Inventory (BDI) [34], the Bech-Raphaelson Mania Scale (BRMS) [35], the Brief Psychiatric Rating Scale (BPRS) [36], and a physical examination. Ten subjects refused to give written consent after initially agreeing verbally to participate in the study; four were considered by the investigator to have significant cognitive impairment (SMMSE less than 24); three had clinical signs of cerebrovascular disease on physical assessment.
Of the 46 subjects who fulfilled inclusion criteria and gave written consent to participate in the study, the mean scores for the screening instruments were as follows: SMMSE 27.9 (SD = 1.6), BPRS 14.3 (SD = 6.0), BDI 6.3 (SD = 4.0), BRMS 1.0 (SD = 0).
subjects who fulfilled the inclusion criteria for the study were administered the following instruments.
(1) Royal Park Multidiagnostic Instrument for Psychosis (RPMIP) [37]. This semistructured diagnostic instrument was developed for the assessment of an acute psychotic episode. Utilisation of this instrument enabled subsequent comparison with a sample of young first-onset psychosis subjects from the data base at the Prevention Early Psychosis and Intervention Centre (EPPIC) in Melbourne [2].
(2) Component symptoms of partition delusions [3]. Partition delusions have been defined as the belief that somebody or something is operating in the ceiling above, the floor below, or beyond the wall separating an adjacent room and is interfering with the patient's life and circumstances [38]. This particular type of delusional phenomenon has been noted in several studies of late-onset psychosis [3,7,38]. The only developed scale for measuring these symptoms [3] was administered to the subjects in the study.
(3) The Cambridge Mental Disorders of the Examination Elderly, Cognitive Examination (CAMCOG) [39]. The C AMDEX is a widely used instrument for the diagnosis of dementia [39], including a cognitive section (CAMCOG) of 60 items with a maximum score of 107.
(4) Revised N EO Personality Inventory (NEO PI-R) [40]. This 240-item personality questionnaire was developed through rational and factor analytic methods to measure the dimensions (Neuroticism, Extraversion, Openness to Experience, Agreeableness, Conscientiousness) of the Five-Factor Model [FFM] of personality, a widely accepted model of personality based on the continuity of traits from adaptive to maladaptive [41].
(5) Head CT scan findings were graded as following: (i) normal, if no abnormalities were present except mild changes consistent with current age; (ii) mild changes, if one or two areas of abnormality, not identified clinically, were present; (iii) moderate to severe changes, if multiple areas of abnormality were present which may or may not have been identified clinically.
(6) Hearing impairment was assessed on a threepoint scale, from normal to severe hearing impairment, as undertaken by Howard et al. (1994) [30] in their study of ‘late paraphrenic’ subjects.
(7) Response to treatment was defined by the same three-point scheme used by Howard et al. (1994) [30].
Subjects were assessed at their place of residence or while hospitalised for treatment of their first psychotic episode. The assessment process for the 46 subjects who fulfilled the inclusion criteria was spread over three or more sessions to minimise fatigue as a complicating factor. Except for the personality inventory, all of the assessment components were undertaken within 2–4 weeks of the subjects consenting to participate in the study. As well as the above measures, sociodemographic information was obtained at the first session. Only subjects who had a complete or partial response to treatment, scoring less than 4 on the psychotic subscale of the BPRS (suspiciousness, unusual thought content, hallucinatory behaviour, conceptual disorganisation) [36], were administered the NEO PI-R [40]. For 38 subjects, this occurred within 4 months of the initial assessment. Three subjects were administered the NEO PI-R at 6, 8 and 14 months, respectively, because of initial poor compliance with treatment or slowness to respond to medication. All of the assessment instruments were administered by a single investigator (AH).
Results
Sociodemographic details of the 46 subjects recruited to the study are shown in Table 1. The mean age of the 10 subjects who refused to participate in the study was 78.9 (SD = 6.1) years. This refusal group comprised three male and seven female subjects.
Sociodemographic variables for the late-onset psychosis sample
Clinical features of the sample are shown in Table 2. The duration of symptoms before presentation varied significantly across the sample. The majority of subjects reported the presence of psychotic symptoms for less than 3 years, but a smaller group described psychotic symptoms which had been present for several years. For only 11 of the 46 subjects was corroborative information about the duration of their symptoms available from relatives.
Clinical variables in the late-onset psychosis sample
Approximately 50% of subjects were not admitted to hospital for treatment of their psychotic symptoms. Self-neglect, barricading themselves in their homes and harassment of neighbours were common precipitants for admission to hospital. Of the subjects in the study who were admitted to hospital for initial treatment, all subsequently returned to their previous accommodation, either home or a hostel. Of the 10 subjects who refused to participate in the study, relatively more were treated in the community (n = 8) than admitted to hospital (n = 2).
Two-thirds of subjects fulfilled DSM-III-R [28] diagnostic criteria for schizophrenia. Schizophreniform disorder was diagnosed in less than 10% cases, reflecting the frequent duration of psychotic symptoms for months or years before presentation.
Hearing impairment was not found to be significantly increased in this sample; 22.9% subjects had varying degrees of deafness which was similar (21.1%) to that reported by the Australian Bureau of Statistics for the general elderly population [45]. Both schizophrenia-spectrum and affective disorders were more common in the first-degree relatives of study subjects. Head C T scans were only obtained on approximately 60% subjects. Refusals came from patients assessed and treated in their home setting and reflected their fear of venturing beyond their familiar environment. In over 90% of the sample, psychotic symptoms significantly attenuated or resolved with antipsychotic medication. Administered medication consisted of conventional neuroleptics (eight patients had depot medication) or the atypical neuroleptic, Risperidone. Only three subjects refused to accept any treatment.
The most prominent psychotic symptoms assessed with the RPMIP [37] were systematised persecutory delusions (93.5%), delusions of reference (91.3%) and misinterpretation or misidentification (91.3%). Sexual (28.3%) and somatic (26.7%) delusions were less common. Auditory hallucinations were described by 78.3% subjects, visual hallucinations by 39.1% subjects, and olfactory hallucinations by 30.4% subjects. Loosening of associations was evident in only one subject (2.2%).
Over 90% of subjects in this sample had ‘partition’ delusions as defined by Howard et al. (1992) [3]. These were grouped into the same categories and compared with the findings of Howard et al. (1992). Most commonly, subjects believed that they were being watched or overheard through the walls of their dwelling, or that there were human intruders, commonly involving theft.
The CAMCOG [39] total and subscale scores are shown in Table 4. The total mean score of 83.8 (median 84) indicated the presence of some cognitive deficits, but falling short of the cut-off value of 79//80 found by Roth et al. (1986) [39] to discriminate between demented and normal subjects. Short-term memory and language tasks were the most poorly performed.
Frequencies of ‘partition delusions’ compared with the sample of Howard et al. (1992) [3]
The Cambridge Mental Disorders of the Examination Elderly, Cognitive Examination total and subscale scores for the late-onset psychotic sample
The NEO PI-R [40] was only administered to the 41 subjects who scored less than four (mean = 2.7; SD = 1.6) on the psychotic subscale of the BPRS [36] following treatment, as the validity of the instrument has not been established when significant psychotic symptoms are present [40]. Of the other five subjects, three were not compliant with antipsychotic medication and remained floridly psychotic, and two subjects became physically unwell before the NEO PI-R could be completed.
Mean (SD) scores for the five personality factors compared with adult norms for the Revised NEO Personality Inventory (NEO PI-R) [40]
As openness to experience has been postulated to have conceptual links with the DSM Cluster 1 group of personality disorders (paranoid, schizoid and schizotypal) [46], a series of six one-sample t-tests were conducted on the six facets of this personality factor for this sample. The findings were as follows: fantasy [t(40) = −6.34, p = 0.000]; aesthetics [t(40) = 0.67, p = 0.506]; feelings [t(40) = −2.21, p = 0.033]; actions [t(40) = −10.07, p = 0.000]; ideas [(40) = −9.32, p = 0.000]; values [t(40) = −9.31, p = 0.000]. With the Bonferroni adjustment for multiple tests, four of the six comparisons for the facets exceed the adjusted significant level of 0.008.
Discussion
This paper describes sociodemographic and clinical findings from a cohort of first-presentation elderly psychotic subjects. Similar to previous studies in this area [4,19,47], social isolation was found to be a prominent feature of subjects even before the onset of their illness. Also, a predominance of females, low marriage rates and few children [4,5,19,30,47] were previously noted findings replicated in this study. Family history suggested a genetic loading for both schizophrenia and affective disorder, although corroborative information was limited. Of the few studies addressing this issue, the most recent controlled family study of lateonset non-affective psychosis [9] found no genetic association with early onset schizophrenia, but an increased history of depression in the relatives of cases than controls. Unlike previous studies [6,19,30], this report did not find an increased rate of hearing loss in patients who develop psychosis in old age. Although deafness exacerbates social isolation and predisposes an individual to misinterpret their environment, its potential role as a specific risk factor for developing psychosis in old age must be considered in the context of its high prevalence in old age [45].
Systematised persecutory delusions involving intrusion into the person's home, multimodal perceptual disturbance and the absence of thought disorder, as found in this study, were phenomenological features which particularly impressed Roth [21] when he proposed the term ‘late paraphrenia’ for psychosis emerging in late life. Subjects in this study fulfilled the criteria for this diagnostic construct and, further, had a high frequency of ‘partition’ delusions as described by previous authors [4,30]. Howard et al. [3] postulated that this relatively unique type of delusion may reflect a profound psychological boundary disturbance in which the elderly person perceives a breakdown in the barriers between their abode and a hostile outside environment. Premorbidly, these elderly individuals are usually described as very selfreliant in terms of regular employment and coping style, but with limited capacity for interpersonal relationships [21,27]. Roth concluded that his late-onset psychotic patients were premorbidly schizoid or paranoid on the basis of anecdotal information about their isolated and ‘eccentric’ lifestyles [21]. Other studies [4,48], utilising standardised instruments of disordered personality functioning, have found paranoid features, but caution that this finding has to be interpreted in the light of some subjects being actively psychotic at the time of assessment [49]. In order to avoid, as much as possible, the contaminating effects of state-type phenomena on long-standing trait functioning, subjects in this study did not undergo personality assessment until their psychotic symptoms had significantly settled or resolved. The NEO PI-R [40], which measures personality traits on a continuum from adaptive to maladaptive, was used in preference to an instrument which only focused on abnormal personality functioning and generated personality disorder diagnoses [43,44]. The Five Factor Model (FFM), from which this instrument is derived, has the most compelling empirical support for a dimensional model of normal personality functioning [40,41]. There is now a growing literature that suggests this model should provide a basis for discriminating among conceptions of personality disorders, which were constructed with a paucity of empirical validation [43,44]. Assessed with the NEO PI-R [38], the personality style of the subjects in this study was not typical of DSM Cluster A dysfunction [44] as previously proposed [4,19,21]. Significant abnormalities were found, however, when the data were compared with validated (albeit North American) adult norms [40]. This was most evident with regard to openness to experience, the factor that encompasses attentiveness to inner feelings, creativity, intellectual curiosity, and tolerance for change. Individuals who score low on this factor tend to be conventional in behaviour, conservative in outlook and restricted in their emotional responses [42]. Of note, the subjects in this study scored particularly low on the facets of openness that related to cognitive flexibility, behavioural repertoire and creative pursuits. This might suggest that they would have difficulty in finding alternative explanations for their delusional beliefs since their low scores on openness suggested that this was a general characteristic of the sample.
In terms of the existing literature, this sample is most readily compared with studies that have focused on psychosis developing after the age of 60 years [19,21,30], rather than the 45 years age cutoff employed in most North American studies [8,31]. Because of the low incidence of late-onset psychosis [5], prospective studies have, to date, recruited subjects at different stages of illness and with varying treatment histories [4,21,30,31]. These case-finding issues have limited the comparison between studies, even though consistent themes are emerging. In particular, cerebral organic factors are now postulated to have a more significant role in the development of late-onset psychosis than previously considered [10,13,16,31]. Neuroimaging changes [10,13] and neuropsychological deficits [16,31] in late-onset psychosis samples are being increasingly identified and characterised. These findings have also been compared with those found in early onset schizophrenia and dementia [4,13,50]. To date, no definite conclusions can be drawn; however, Almeida et al. [16] suggest that the degree of cognitive impairment may divide late-onset psychosis into two subgroups; a less cognitively impaired group with clinical features similar to early onset schizophrenia and a group that has more in common with the early stages of dementia. Cognitive changes were evident in the majority of subjects in this study, even though a diagnosis of dementia could not be made at the time of assessment. Memory and language functions were the most impaired and this will be examined further in another report. Similarly, the presence of neuroimaging abnormalities has been proposed as a possible discriminating factor for subdividing within late-onset psychosis [11,12], although findings in this regard are also inconclusive. The small number of subjects who consented to neuroimaging in this study did not allow any contribution to be made to the continuing debate in this area.
The sample size for this study was limited by the requirement that only newly diagnosed subjects be recruited. Subjects who were initially hospitalised for treatment, as well as those who accepted community-based treatment, were included. Although this meant that a spectrum of presentations was sampled, possible biases were then introduced into the assessment procedure. In particular, neuroimaging was more difficult to obtain on the subjects who were not admitted to hospital. Also, relatively more refusals to participate in the study came from subjects treated in the community, as compared to those who were admitted to hospital. Other prospective studies in the area [30,31] have recruited larger samples by including subjects at different stages of illness. This methodological approach inevitably introduces contaminating influences from the effects of previous treatment and contact with psychiatric services [1,22,24]. A control sample of ‘normal elderly’ was not recruited for this study as this would not have yielded a useful comparison on most variables. In a subsequent report, a comparison of these late-onset psychosis subjects with a sample of early onset subjects will be described.
A further methodological concern in this study involves the time to assessment of the subjects' personality. This varied according to the time taken for subjects' psychotic symptoms to attenuate sufficiently that ‘state’ effects could be minimised. As the study was ‘naturalistic’, treatment regimes were not standardised and subjects were managed by their respective clinicians. All subjects received low-dose neuroleptic medication and community follow-up by a case manager; however, the length of time for psychotic symptoms to totally or partially resolve was influenced by medication-responsiveness, emergence of side effects and compliance issues. Administration of the NEO PI-R [40] at varying lengths of time after the initial assessment reduced the standardisation of the assessment process, but facilitated the evaluation of the subjects' personality without the presence of florid psychotic symptoms, particularly persecutory delusional ideation. Five subjects were not assessed with the NEO PI-R, thus introducing a bias in that their personality profiles could not be included in the reported findings. The presence of early cognitive deficits was another factor which may have influenced subjects' responses to the NEO PI-R questions; nevertheless, subjects in this study did not fulfil diagnostic criteria for dementia. Of note, the findings reported in this paper do not attempt to establish the relationships between personality variables and the onset of psychosis in late life. For example, it is not clear whether low scores on openness to experience are associated with cognitive deficits or social isolation, and to what extent any of these variables make independent contributions to predicting types of psychotic symptoms. To address this issue, correlational analyses and hierarchical regression analyses will be conducted and then cluster analyses will be used to determine if subgroups within the sample can be identified. Findings obtained will be reported in subsequent papers, as will a comparative analysis with a young first-onset psychosis sample.
In conclusion, this descriptive study found sociodemographic and phenomenological similarities to other studies of late-onset psychosis. To what extent the differences found in this study, particularly regarding hearing loss and premorbid personality style, can be attributed to the first presentation nature of this sample is not clear at this stage. The findings suggest that phase of illness may be a variable that needs to be taken into greater consideration when studying late-onset psychotic subjects. Research into early onset psychotic disorders has demonstrated the value of differentiating between the first presentation and more chronic stages of illness and this approach may also have value in advancing our understanding of the risk factors and aetiological mechanisms associated with late-life psychotic syndromes.
Footnotes
Acknowledgement
My thanks to Henry Jackson for his comments in the preparation of this article.