It will be seen that out of the 625 patients, 256 are classified as falling into group 2, with a diagnosis of schizophrenia, leaving 369 patients with a diagnosis other than schizophrenia. Table 2 shows the breakdown by date of birth and date of current admission for three sets, the total population of 625 cases, and two subgroups consisting of the 256 patients with schizophrenia, and the 369 patients with other diagnosis.

The aim of the survey was to examine the dates of birth and dates of admission for the whole population, and also to study the subgroup of schizophrenics, comparing them with the remainder of the patients. Table 2 indicates that there are two main differences between the schizophrenic subgroup and the subgroup with other diagnoses. In the first place, the schizophrenic patients are younger than the rest, probably because the latter subgroup contains 213 patients with dementia. Secondly, the schizophrenic group tends to show earlier admission dates than the rest, with 102 out of 256 cases being admitted before 1970, as opposed to 44 out of 369 cases in the group with other diagnoses.

In the past, most research into seasonal effects has been conducted using monthly totals of date of birth or date of admission. While this may be understandable for reasons of convenience, in that data is frequently summarised in this form in official records, there is certainly no logical reason why this approach should be used, and using grouped data in this way may conceal important effects that might be revealed by an analysis of the actual day of birth rather than the month of birth. In addition, there is a smaller problem that the months are not all equal in length. In this study, a different approach was taken, and as it will probably be unfamiliar to many readers, it will be described in some detail. Instead of using the division of the year into twelve months, the time of year can be measured by the position of the planet Earth in its orbit round the Sun. The position of the Sun as seen from the Earth on any given day can be taken from an Astronomical Ephemeris, or calculated using a computer. This position is measured from the vernal equinox, (which usually falls on about March 21st) and is usually given in degrees. At the summer solstice, (on about June 21st), the position of the Sun is 90 degrees, at the autumnal equinox, (September 23rd) the position is 180 degrees, and at the winter solstice, (December 21st), the position is 270 degrees.

In this study, the dates of birth and dates of admission were entered into a computer, and the position of the Sun on the date of birth and the position of the Sun on the date of admission calculated for each patient using a modification of a standard ephemeris computer program (Black, 1983).

These positions were then analysed for the group of 625 patients as a whole, and the two subgroups of 256 schizophrenics and 369 non-schizophrenics. Two methods were used to analyse the data. In order to attempt comparisons with earlier work the year was divided into twelve sectors each consisting of 30 degrees in length. Sector 1 contained those patients born between the time of the vernal equinox and the time when the Sun had advanced 30 degrees; sector 2 contained those born with a Sun position of between 30 degrees and 60 degrees and so on up to sector 12 containing those patients born with Sun positions between 330 degrees and 360 degrees. These twelve sectors are perhaps more familiar to the reader as the twelve signs of the zodiac, starting with sector 1 as Aries and ending with sector 12 as Pisces. The same procedure was carried out for the dates of admission. The distribution for dates of birth in the twelve sectors was then analysed using the chi-squared test with 11 degrees of freedom, as in earlier work on seasonality. The same was done for the distribution of dates of admission in the twelve sectors. In both cases the analysis was performed for the total group, the schizophrenic subgroup and the subgroup of patients with other diagnoses.

The date of admission for each patient was then compared with the date of birth for that same patient, using the formula RA = DA - DB, where DA is the Sun position on the day of admission in degrees, DB is the Sun position on the date of birth in degrees, and RA is the Relative Angle between DB and DA, starting from DB. If RA was negative, 360 degrees were added to RA to return a result between 0 and 360 degrees; for example if RA = -75 degrees for a patient, this is equivalent to +285 degrees. This leads to a spread of values for RA for the patients lying between 0 degrees and 360 degrees, which can again be divided into twelve thirty degree sectors containing totals which can be treated by chi-squared analysis using 11 degrees of freedom.

Due to the fact that the Earth orbits the Sun in an elliptical orbit, rather than in a circular one, the sectors are not of exactly equal length in time ranging from 29.5 days (sector 10) to 31.4 days (sector 4). However the sectors are less uneven than would be the case if calendar months were used instead, and it will be shown that the error introduced in assuming that the sectors are of equal size is small.

The second method of analysing the data was to perform a harmonic analysis to determine the presence or absence of cyclical trends in more detail than can be provided by a simple chi-squared test. This analysis allows cyclical trends of shorter period than one year to be analysed, breaking down the distributions into component sine waves for which the amplitude and phase for each component can be determined.

Since the time of day when the patients were born was not available, the position of the Sun was calculated for midday Greenwich Mean Time on the date of birth. The same applies for the Sun position on the date of admission. Even if some of the patients were born overseas in a different time zone, the total error in calculated Sun positions is most unlikely to exceed one part in a thousand.

There was no statistically significant deviation from chance in the distribution of birthdates among the total of all 625 patients, the schizophrenic subgroup or the non-schizophrenic subgroup. The same was true for the distribution of dates of admission, once an artefact due to the nature of the cross-sectional survey producing artificially high admission rates shortly before the census date had been eliminated.

Table 3 shows the sector totals for Sun position at the date of admission relative to the Sun position at date of birth for the whole population of 625 patients, for the schizophrenic group (256 cases) and for the non-schizophrenic group. Thus totals for sector 1 show those patients admitted during the first month following their birthday, totals for sector 2 shows those patients those patients admitted between one month and two months after their birthday and so on, sector 12 totals showing those patients admitted up to one month before their birthday. The distribution is highly significant for the schizophrenic group (p < 0.0005) and not significant for the non-schizophrenic group or for the group as a whole.

The data was then subjected to a full harmonic analysis, using the methods of Edwards (1961) and Addey (1976). The first twenty harmonics were analysed for each of the three patient groups for Sun position at birth, Sun position at admission and Sun position at date of admission relative to Sun position at date of birth. Table 4 shows those harmonics in each group which achieve statistical significance at the p < 0.01 level using the chi-square test with two degrees of freedom, (Edwards, 1961).

The harmonic number refers to the period of a cyclical trend being assessed. Thus harmonic number one is a measure of a cyclical trend of period one year in this study, harmonic number two is a measure of a cyclical trend with a period of six months, and in general harmonic number N examines the extent of a cyclical trend with length 1 / N of a year. The extent of the cyclical trend of a given harmonic is measured by the amplitude expressed as the percentage rise and fall above the mean. Thus a harmonic with a mean value of 100 and an amplitude of 25% will show a maximum value of 125 and a minimum value of 75. The phase of a harmonic determines where the peak of the wave falls in relation to the starting point. For example, if the first harmonic were to show an amplitude of 20% and a phase of 75 degrees, this would show that there is a cyclical trend with a period of one year, with maximum value 20% above the mean value situated in the centre of sector 3. Phase in this study specifically relates to the maximum of the harmonic, though in some branches of mathematics phase is used to describe the position of the ascending node of a harmonic wave rather than the maximum.

From Table 4 it can be seen that there were no harmonics significant at the p < 0.01 level for either the dates of birth or the dates of admission either in the total population or in the two subgroups.

In addition, Table 4 illustrates that when the position of the Sun at the date of admission relative to the position of the Sun at the date of birth is subjected to harmonic analysis, there are no significant harmonics for the total patient group, and that for the schizophrenic group and the remainder there are two harmonics in each group significant at the p < 0.01 level.

However the most striking result is that the first harmonic for the schizophrenic group, significant at the p < 0.0005 level, has a very large amplitude of 35%. The phase is 23.3 degrees, indicating a peak in sector 1, and a corresponding trough in sector 7. This amplitude means that the maximum value of this first harmonic is 2.08 times the minimum value. In simple terms this means that a patient in the schizophrenic group is more than twice as likely to have been admitted in the month following his or her birthday than in the seventh month following his birthday, and more generally that in this study the schizophrenic patients show a strong tendency to be admitted in the half of the year centred on the month after the birthday. This can be seen in a crude way by examining Table 3 once more and observing that a total of 90 schizophrenic patients fall into sectors 5 to 10 inclusive, with 166 patients in the remaining sectors, (p < 0.00001).

The above results are interesting in that they do not appear to support earlier findings concerning date of birth and schizophrenia, and also in that they suggest highly significant deviation from chance in the relationship between date of birth and date of admission in the schizophrenic group, whether this be assessed by analysis of sector totals, (Table 3), or by harmonic analysis, (Table 4), with a strong cyclical tendency with a period of one year.

One of the chief difficulties in this study is that the diagnoses were collected from the index files and H.M.R. 1 forms for the patients, and it may well be argued that there are likely to be inaccuracies in recording the information, and that different doctors may have used differing diagnostic criteria over the years. This is probably true, and it is difficult to estimate the number of patients in the schizophrenic group who might have been diagnosed more appropriately in the non-schizophrenic group, and vice versa. An attempt was made to assess this problem, at least qualitatively, using harmonic analysis of the schizophrenic group divided up by sex. The first twelve harmonics were calculated for the 127 males and the 129 females to see if they appeared to come from the same population to judge from comparison of any significant harmonics as to Amplitude and particularly as to Phase in the male group and the female group.

Only one significant harmonic was found: the lst Harmonic for males, (Amplitude 50%, Phase 21.44, Chi-squared 15.91, p < 0.0005). When compared with the first harmonic for females, (Amplitude = 23%, phase = 29.7, chi-squared 3.5, not significant), it can be seen that they compare very well for phase, the phase difference being only 8.3 degrees, suggesting that they come from the same population. Thus if there are errors in diagnosis, they are likely to be fairly small, unless by chance they happen to affect males and females in approximately the same way. Even if the errors are large, the schizophrenic group seems remarkably consistent in showing deviation from chance.

The error introduced by assuming a circular orbit for the earth is small. On its own it would give rise to a lst harmonic wave of approximate Amplitude 3% and Phase 102 degrees. Not only is this amplitude smaller than the significant amplitudes we have discussed by about a factor of ten, but it is also about 79 degrees out of phase with the lst harmonic wave for the 256 schizophrenics.

It is most unlikely that the differences between the schizophrenic group and the other patients in terms of dates of birth and dates of admission mentioned earlier, (the schizophrenic group being younger and having spent more time in hospital than the others), contribute to the difference in the relationship between date of admission and date of birth in the two groups. This is because this difference is not concerned with age or length of admission per se; it is concerned with a cyclical trend of period one year. The population of 625 patients, together with the two subgroups of schizophrenics and the remainder all span several decades in both their dates of admission and their dates of birth, and there seems to be no obvious reason why the significant result for the schizophrenics should be due to the span of admission dates and birth dates covering an insufficient time span. Given that there is no obvious reason why the relationship between date of admission and date of birth in the schizophrenic group should be spurious for any of the reasons discussed above, one is left with the question of what this relationship means, and whether it is relevant to the epidemiology of mental illness in any way which could have practical applications.

If the relationship between date of admission and date of birth in the schizophrenic group demonstrated in this study is genuine, there are important implications for research into the field of seasonality and mental illness. The results suggest that research should not only include investigation into birth dates and admission dates it should also be concerned with the relationship between the two. Very little work has been done in this field.

In view of the peak in admission rate being situated shortly after the birth date, it suggests that the viral hypothesis in the aetiology in schizophrenia, re-evaluated by Crow, (1984), is open to question. Some factor associated with the birthday would appear to be connected with time of admission in schizophrenia. The birthday could be regarded as being a major life event associated with stress, though why this should be so for schizophrenics more than for patients with other diagnoses is not at all clear. One possibility is that the schizophrenics in the study tended to be admitted soon after the onset of their illnesses, whereas the controls, with a large incidence of dementia, were admitted following a period of illness that might have had an onset many months earlier.

Another possibility is that birthdays may have some special significance for schizophrenics. Perhaps because they signify maturation or ageing, or because the giving of gifts, (or lack of gifts) highlights interpersonal problems within the family, they may trigger an illness. It could also be the case that the focus of attention on a schizoid person, at the time of the birthday, might be the last straw in bringing forward an illness, which would have occurred later in the year.

It seems clear that further research is required in this field, and with the help of the computer this is now rendered considerably less time consuming. Techniques such as harmonic analysis are ideally suited to the computer, and would be very tedious and error prone if carried out by hand.

This is a preliminary study, and it is suggested that there is room for improvement in several areas. Replication of this study using a different hospital population would be an important first step. It would be an improvement to obtain rather more reliable sources than index cards and H.M.R. 1 forms when collecting details of diagnosis for each patient, though it has been indicated above that some effort has been made to show that errors in diagnosis may not be important in this study. It would also be interesting to perform a study into any relationship between date of birth and date of first admission, or even date of birth and original onset of illness in a further attempt to elucidate any seasonal factors. A larger sample would allow other less common conditions to be investigated; it is interesting in this respect that the 82 patients grouped together under a label of affective disorder show a strong second harmonic, Amplitude 51%, Phase 58 degrees, Chi-squared 10.76, p <0.01 for the relationship between date of admission and date of birth, indicating a strong tendency to be admitted either one month or seven months after the birthday, with a period of six months. Unfortunately 82 patients is not sufficiently large a sample for the methods of harmonic analysis used in this study, a minimum of 120 cases being required in any group under study, (Addey, 1976). A larger sample would generate a larger number of cases in each diagnostic category.

In conclusion, this study demonstrates that there is a hitherto undetected association between the dates of admission and dates of birth of schizophrenics in this group of hospital inpatients, and suggests that further work is needed in this field, in an effort to elucidate this relationship further, and to use similar methods to investigate possible further seasonal effects among other diagnostic groups.

ADDEY, J.M. (1976) Harmonics in Astrology. Romford: L. N. Fowler & Co. Ltd.

BARRY, H. & BARRY, H., Jr. (1961) Season of Birth. Archives of General Psychiatry, 5, 100-108

BLACK, A. K. (1983) Sharp Astrology. (Computer Program), London: Sharpsoft Ltd.

CROW, T. J. (1984) A Re-Evaluation of the Viral Hypothesis: Is Psychosis the Result of Retroviral Integration at a Site

Close to the Cerebral Dominance Gene? British Journal of Psychiatry 145, 243-253

DALEN, P. (1968) Month of Birth and Schizophrenia. Acta Psychiatrica Scandinavica, 203, 55-60

DEAN, G. & BISHOP, C. (1976) Harmonic (Fourier) Analysis. In Recent Advances in Natal Astrology, 152-154. Subiaco, Western Australia: Analogic

EDWARDS J. H. (1961) The Recognition and Estimation of Cyclic Trends. Annals of Human Genetics, 25, 83-87

HARE, E.H. & PRICE, J. S. (1968) Mental Disorder and Season of Birth: Comparison of Psychosis with Neurosis. British Journal of Psychiatry, 115, 533-540

KEVAN, S.M. (1979) Season of Life - Season of Death. Social Science and Medicine, 13 D, 227-232

RIHMER, Z. (1980) Season of Birth and Season of Hospital Admission in Bipolar Depressed Female Patients. Psychiatry Research, 3, 247-251