Death certification in a psychiatric hospital

Aims and method A retrospective review of death certificates issued at a large psychiatric hospital in North Cheshire during the 1980s and 1990s. Results Dementia, which was the recorded clinical diagnosis in 78% of all deceased, was reported in 31% of death certificates, while other psychiatric disorders (22% of all deceased) appeared in only 2% of certificates. Autopsy appears to have very little or no value in improving the quality of death certificates in psychiatry. Clinical implications The onus is on the clinicians to produce adequate death certificates. Recording chronic conditions present at death, such as dementia and other psychiatric disorders, in addition to those directly causing or contributing to death would improve the epidemiological value of death certificates.

older adults waiting for long-term placement should be explored further, as it is capable of being audited, though it is recognised that the present financial arrangements for nursing home and residential home care militate against rapid transfer despite an apparent excess of supply in these facilities in some areas.
Finally, perhaps not surprisingly, consultants were overwhelmingly in favour of retaining decision-making power over admission to and discharge from NHS beds. HEALTH SERVICE COMMISSIONER (1996)  Death certification in a psychiatric hospital

Emad Salib
Aims and method A retrospectivereview of death certificates issued at a large psychiatric hospital in North Cheshire during the 1980sand 1990s. Results Dementia, which was the recorded clinical diagnosis in 78%of all deceased, was reported in 31%of death certificates, while other psychiatric disorders (22% of all deceased) appeared in only 2% of certificates. Autopsy appears to have very little or no value in improving the quality of death certificates in psychiatry. Clinical implications The onus Is on the cliniciansto produce adequate death certificates. Recording chronic conditions present at death, such as dementia and other psychiatric disorders, in addition to those directly causing or contributing to death would Improve the epidemiological value of death certificates.
Death certificates, despite their inadequacies, remain one of the most commonly used sources of data for the estimation of the prevalence of most diseases. However, morbidity and mortality statistics that arise from death certificates are seriously flawed, mainly because of the omission of important materials (O'Sullivan, 1996). Cer tain diseases are grossly underestimated on the death certificates such as dementia (Burns et al, 1990: Macera et al, 1992. Possible sources of error in the use of statistics based on death certificates may arise from inadequate infor mation provided on the certificate, errors in the pre-mortem diagnosis or non-documentation of the confirmed diagnosis on the death certificate (Macera et al, 1992). Deciding what should or should not be included on death certificates may be difficult and become a matter of opinion and not fact, which it purports to be. What needs to be recorded is not just the 'diseases or conditions directly leading to death', but 'with what diseases or conditions' did death occur (Ashworth, 1991). These conditions may not be directly related to  1980-1986 and 1990-1996 Year Total death, n death, but could be of importance to public health.
The main aim of this review is to examine cause of death of psychiatric in-patients, as reported in death certificates over two decades during which Winwick Hospital has retracted considerably. The effect of hospital retraction on patients' mortality, completeness of death certi ficates, the frequency and value of performing post-mortem examination in increasing the accuracy of death certificates are examined. The review also attempts to evaluate the extent to which mortality data underestimate preva lence of psychiatric conditions, particularly dementia.

The study
Winwick Hospital is a 100-year-old psychiatric hospital receiving all psychiatric patients from the two districts in North Cheshire with a total population of 350000. The hospital had a large population of ageing psychiatric long-stay pa tients. Beds have been reduced from over 2000 to less than 300 during the past 20 years and the hospital will be replaced by much smaller units later this year.
This retrospective review examined all death certificates issued at Winwick Hospital, Warrington in two periods of seven years each: 1980-1986 and 1990-1996. Data included all patients, long-stay or newly admitted, who died in hospital during the review period. The collected informa tion included age, gender, duration of hospital stay, whether post-mortem was carried out, year and month of death, ICD-9 diagnosis (World Health Organization, 1978) and recorded cause of death.
Cause of death was divided into: (a) direct cause of death, which is the same as Part la in the present death certificate form: (b) underlying cause of death leading to death (Parts Ib and c in death certificate form): and (c) associated condi tions that existed at death and may have contributed to it. although not part of the sequence leading directly to death which corres ponds with part II of death certificates. Where autopsy was performed, the cause of death identified by the pathologist was noted. For analysis, the diagnostic categories were recoded into major groups: dementia and all other psychiatric diagnoses added together.
In both study periods, 5% of all autopsies (1.5% of total hospital deaths) were requested by clinicians, whereas 95% of post-mortems were requested by the coroner's office to establish causes of sudden or unexpected deaths when no medical officer was in a position to sign the death certificate. Three coroner's inquests involving deaths due to unnatural causes were held, one in the 1980s and two during the 1990s. One verdict of suicide and two open verdicts were recorded. Death rate appears to have declined by an average of 1% with hospital retraction, noticeably so since 1994 which may have resulted from the reduction in numbers of elderly beds.
Completeness of death certificates is shown in Table 2. All sections of death certificates were completed in only 16% of cases, while in 50% of all death certificates, whether based on autopsy or not, neither the underlying cause of death nor the associated conditions were stated. Dementia, the recorded clinical diagnosis in 78% of all deceased, was reported in 31% of death certifi cates and psychiatric diagnoses other than dementia (22% of all deceased) in only 2% of death certificates (Table 4).
Cause of death (la) in the two study periods is shown in Table 3.
Associated conditions at death (part II) were not reported in 661 (49%) and 271 (59%) of all death certificates in the 1980s and 1990s respectively (Table 2). Table 4 shows almost equal proportions of dementia (31 and 29%) and senility (12 and 7%), and only 2% stated the presence of a psychiatric condition at time of death in the two periods. Autopsy-based certifi cates did not report senility but recorded dementia in only 1 and 5% in the 1980s and 1990s, respectively.
Post-mortem examinations - Tables 1 and 3 show the difference in findings of autopsy during the two periods. Apart from some reduction in the rate of autopsy in the 1990s, there was no significant difference between post-mortem find ings and how they differ from the clinically-based death certificates. Autopsy-based death certifi cates reported fewer cases with bronchopneumonia as direct cause of death compared with those completed by medical staff without autopsy (P<0.001). More findings were also reported with myocardial infarction and other cardiovascular causes (P<0.05) in the autopsybased certificates. Hypothermia was recorded on three occasions, presumably on patients who Table 3. Direct cause of death as reported in death certificates with and without autopsy (Part la of death certificates) 1980-1986 1990-1996 Direct cause of death  went off wandering in cold weather. One incident of suicide (by asphyxia, using a plastic bag) was recorded in the seven years during the 1980s and two incidents that attracted an open verdict: drowning in the hospital bath and by hanging, had been recorded in autopsy during a similar period in the 1990s. The underlying cause of death was twice as likely to be reported in death certificates based on autopsy as 75% of autopsybased certificates had the underlying cause of death completed, higher than the 35% in the clinically-based certification (P<0.01) for both periods. Table 4 shows that a higher proportion of autopsy-based certificates did not report any associated conditions in part II of the certificates (72%), compared with 53% of the clinically-based certificates (P<0.05).

Comment
The review shows that little has changed in doctors' perceptions of death certificates over two decades in a large psychiatric hospital. In this study bronchopneumonia was reported as direct cause of death in over 55% of all deaths of psychiatric long-stay patients, which is not dissimilar to around 70% reported by Burns et al (1990), Whitehead & Hunt (1982) and Kay (1962). Bronchopneumonia is also the most common mode of death found in elderly medically ill geriatric patients (Burns et al. 1990;Macera et al, 1992). Kay et ai (1962) showed that subjects with senile dementia died from non-specific causes such as bronchopneumonia, whereas functionally-ill patients had similar proportions of specific and non-specific deaths compared with the general population. However, in this review bronchopneumonia was significantly less frequently recorded if post-mortem examination was carried out. Cardiovascular conditions were, on the other hand, more frequently reported in autopsy-based certificates.
Ninety-five per cent of all post-mortems were not selective hospital autopsies requested by clinicians. The main reason for the coroner's requesting autopsy was to establish cause of death which, in all but three, had been due to natural causes. So autopsy-based certification in this study involved a random sample, probably representative of all hospital deaths. Therefore, positive or negative findings of the effect of autopsy on the accuracy of death certificates may apply to all deaths reported from the same population. The randomness of the vast majority of autopsy cases may explain why the percen tages have remained so constant over the review periods. Corsellis (1962), in a neuropathological study, found that only 20% of dementia patients died of bronchopneumonia at autopsy, not dissimilar to the 14% found in this review more than 35 years later. The underlying cause of death was twice as likely to have been reported in autopsy-based death certification. The rate of hospital autopsy in this study of only 5% of all post-mortems during the study periods (1.5% of total hospital deaths) is considerably lower than other specialities (Harris et al. 1993) and half that of the national average (Ashworth, 1991) and the previously reported autopsy rate in psychiatry of 10% (Newens et al, 1993). It may be that autopsies were not considered of great value in long-stay psychiatric patients, the majority of whom had a well-documented medical history.
The contribution of dementia, as a chronic syndrome associated with varying frequency and severity and with many comorbid disorders, to death varies and may be difficult to assess in the individual case. In this review, more than 48% of cases with dementia, present at death, was not recorded as having contributed directly (Ib and c) or indirectly (II) to the death. Dementia was recorded as the immediate cause of death in 2%, similar to that previously reported by Newens (1993). Martyn & Pippard (1988) pointed out that the use of the underlying cause of death as reported in mortality statistics would be an economical way of investigating regional differences in the incidence of dementia. In this study, dementia was recorded in 31% of death certificates compared with 57% reported by Martyn & Pippard (1988) and similarly by Newens et al (1993). The review findings clearly show that underlying cause of death is an unreliable source of information concerning the presence of de mentia and other psychiatric conditions at death.

Discussion
Inadequate reporting of cause of death leads to serious inaccuracies. Clinicians should be more aware of the importance of recording dementia on the death certificates. But could more hospital autopsies add anything to death certifi cates or improve their accuracy?
Autopsy is likely to have little value in ensuring higher rates of recording conditions such as dementia, particularly Alzheimer's disease, but may be of value in providing more information regarding underlying cause of death in the elderly. The main value of autopsy in psychiatry is to verify the diagnosis of Alzheimer's disease and other dementias, especially as almost half of clinically diagnosed Alzheimer's disease are not confirmed neuropathologically (Molsa et al, 1985;Homer et al, 1988). But neuropathological examination of the brains may be outside the remit of hospital pathologists carrying out a post-mortem. Even by conforming with Royal Colleges recommended autopsy rates of 35% (Kamal et al, 1997), 65% of cases will still depend on the way the doctor completes the death certificates and information that he decides to include, especially chronic conditions such as dementia. Pathologists cannot diagnose the presence of dementia from their autopsy findings alone and recording the diagnosis of dementia rests almost entirely with the clinician. The review has shown that very little has changed in doctors' perceptions of death certification during the 1980s and 1990s. Findings of studies carried out 36 years ago in Sweden (Kay, 1962) and England (Corsellis. 1962) have remained virtually unchanged. The onus is on clinicians to produce adequate death certificates that may provide valid statistical data to be used in estimating prevalence, trends, risk factors and other aspects for future research. MARTYN