Dr Staines' Evidence
88 Dr Anthony Staines is the College Lecturer in Epidemiology at University College Dublin. He is a qualified medical doctor with degrees from both Trinity College and London. He obtained his Phd in Epidemiology from the University of Leeds. He has a background in paediatrics as well as a scholarly history in epidemiology. The latter field he defined as the study of factors which influence the health of people, particularly in population groups. It is the study of the factors that lead to particular causes of death and the acquisition of particular diseases.
89 Dr Staines had been provided with two reports from Dr Shavelle. These had been compiled in December 2000. In his first report of 7 December 2000 Dr Shavelle, a Researcher with the Department of Statistics at University of California, had expressed various estimates of the plaintiff's life expectancy. The information on which he had based his estimates was taken from a document known as a "Patient Evaluation Questionnaire" (PEQ) which had been completed on 21 November 2000 in Australia by Dr Jayne Antony who had, on the first defendant's behalf, seen and examined the plaintiff. Dr Shavelle also had the benefit of other materials sent to him by the solicitors for the first defendant. There were varying estimates provided in the first of these two early reports because Dr Shavelle, who is not a medical doctor, considered that some of the facts he had been given about the plaintiff were not clear. The most favourable of his estimates was a further 33.3 years.
90 (For completeness, and in order to understand the ensuing sequence, it is necessary to say that Dr Shavelle provided a further estimate on 26 February 2001 in which he computed the plaintiff's life expectancy to be 34 additional years. This was following additional information provided to him by Dr Bowers, the Rehabilitation Physician retained on behalf of the first defendant. On 26 March 2001 he provided the defendant's solicitors with a new report in which he said that the life expectancy in his report of 26 February 2001 (an additional 34 years) was incorrect. There was a coding error in his computation and, in addition, the model he had used was arguably inappropriate. Because of these errors, he revised his estimate and computed the plaintiff's life expectancy, on a revised basis, at 40 additional years. He also noted that, based on Dr Bowers evaluation, he had assumed that the plaintiff had no functional use of her hand. He was asked to provide a revised estimate which assumed that the plaintiff "uses raking motion or grasp with hand". On this assumption he gave an alternative computation of life expectancy at 42.9 additional years.
91 On 28 March 2001, he provided a final report which, inter alia, gave details of the manner in which the coding error had been made in the earlier report and why it was that the model chosen in that report was considered inappropriate. Although it appears that the two March reports were in the possession of the first defendant's solicitors when Dr Staines gave his evidence on 28 and 29 March 2001, it appears, presumably for reasons of forensic advantage, they were not provided to the plaintiff's solicitors until after Dr Staines had completed his evidence and returned to Ireland. For this reason, Dr Staines did not have the opportunity in examination in-chief to comment on the errors which had been made by Dr Shavelle, nor was he asked directly about such matters in cross-examination.
92 Dr Staines had been asked to comment on Dr Shavelle's December reports. He was asked, essentially, whether they were a reasonable or reliable method of estimating the plaintiff's remaining life expectancy. He was also asked to consider whether Dr Shavelle's statistical approach to estimating future life expectancy was methodologically sound from "an epidemiological point of view". Dr Staines was retained and instructed on 4 January 2001 and provided a succinct but thorough report on 24 February 2001.
93 First, Dr Staines explained the methods used by Dr Shavelle. He described these as "sophisticated and appropriate statistical methods" in the analysis of the survival of people with cerebral palsy in California. These methods estimate a quantity known as "the hazard rate". This can most simply be thought of as the risk of dying immediately after having lived to a given age. He gave as an example a twenty year old woman. What is estimated for her, as an example, can be thought of as the risk of her dying before the age of twenty one. This estimate is based on the actual survival of twenty year old women in the population under study. More precisely, what is estimated is the instantaneous hazard rate - that is the risk of dying immediately upon reaching the age of twenty. For practical purposes, the hazard is usually calculated over a period of time, typically a single year, for the hazard of dying between age twenty and twenty one.
94 Dr Staines explained that, given a set of estimates of hazard, it is straight forward to calculate a life expectancy. At this point, he made an important distinction. He said that it was very important to be clear as to what an estimated life expectancy is. It is the expected duration of life from a specified age, for a hypothetical person who experiences the estimated hazard rates in question. Life expectancies are usually calculated for entire populations not for individuals. He said that it was important to emphasise that no real person experiences this set of risks of death.
95 Dr Staines developed this point in his oral evidence. First, he said that individuals do not have life expectancies. Life expectancies are properties of groups of individuals only. It only makes sense to talk about the life expectancy of a large number of people. When one comes to examine the likely remaining years for an individual, it is not a straight forward process and it cannot be simply done by taking the number shown as life expectancy for the group and applying it to the individual. The second point made by Dr Staines in this context is that epidemiologists normally would not be concerned to estimate the remaining years of an individual within a group possessing particular characteristics. Epidemiologists study populations the whole time. As a rule, they do not have a clinical background, although in his case he did have a clinical background in paediatrics. He expected that a treating clinician would be able to estimate the remaining years of an individual patient. A clinician would take the life expectancy of a large group of people with similar characteristics as background. He would then apply that knowledge, and the other knowledge generally inherent in the clinician's experience, to the individual characteristics of the person and, based on all those matters, anticipate the likely number of years left for that person to live.
96 The third point he made in connection with this introduction was that "calculated life expectancies are not exactly what a non statistician might expect. They are not estimates of how long an individual can expect to live, rather they are convenient summaries of a set of hazards". He went on to explain that this does not invalidate their use. For example, calculating life expectancies in this way permits the profitable operation of Life Assurance Companies. While life expectancies are tools for summarising sets of hazards rates for populations, the application of the method to an individual, he cautioned, was "fraught with difficulty" (Report page 10 section 5.1).
97 (I interrupt this summary of Dr Staines' discussion to state that, I shall from this point endeavour to refrain, unless the context requires it, from using the expression "the plaintiff's life expectancy" but rather speak of "the plaintiff's remaining years". It must be said that when examining the various medical and statistical reports, the two expressions are often used without the necessary and relevant distinction. This applies also to the entirety of the transcript where the subject is discussed).
98 The second matter discussed at length by Dr Staines in his report is an examination of the caution which needs to be applied to the database itself and to the method of prescribing a life table for Ms Simpson, as Dr Shavelle had done. These matters were discussed to make points regarding the possibility that statistical errors might arise; and to examine the precision of the estimates. First, he pointed to the quality of the Californian Mental Retardation Database itself. Dr Staines made the point that cerebral palsy is not a diagnosis. It has a range of different origins. This fact, in itself, may result in a source of statistical difference. Secondly, he pointed to the methods used to gain the information which is contained within the database. The data comes from forms completed as part of the routine activities by people from many different professional backgrounds. More importantly, Dr Staines pointed to his own experience with large scale epidemiological studies. This experience suggested that, without suitable training and the most scrupulous supervision of the participants, the quality of data recorded may well be variable. Some will be of good quality and some will be of poor quality. Thirdly, he pointed to the PEQ which Dr Antony had given to Dr Shavelle. Dr Shavelle had calculated an estimated life table for the plaintiff using this PEQ. It was quite apparent to Dr Staines that the information which Dr Antony had obtained was, in a number of important respects, inaccurate in so far as it purported to give details of the plaintiff's characteristics. There were clear discrepancies. These included an estimate of the plaintiff's mental condition, and the report that she suffered epilepsy (although she has had no seizures for years). Also of importance was the statement that she "had no functional use of her hand" when in fact she can use the thumb of her left hand to point to her communication board; she has increased function in her left hand and can control the wheelchair with her left hand and has some voluntary control of her left upper limb. There were other matters as well including the PEQ information that she "does not crawl, creep or scoot" whereas the plaintiff was able to move around in her bed; she could roll over in the bed and on the ground, and was able to pull herself forward by use of her limbs when lying prone on the floor.
99 The point that Dr Staines made was that any discrepancy between the plaintiff's actual condition and the data provided to Dr Shavelle through the PEQ was likely to lead to errors in the estimation that he gave. The statistical result achieved would be rendered less precise. The greater the errors, so too would be the extent of the imprecision. The potential magnitude of such errors was illustrated by the various estimates given by Dr Shavelle in his first report which varied between 15.5 and 33.3 for further years of life. No doubt, had Dr Staines been given the two March reports of Dr Shavelle, he might well have made an equally telling point that the coding error and inappropriate modelling contained in the earlier report of Dr Shavelle also demonstrated tellingly the potential for error arising out of inherent imprecisions in the process.
100 A related potential imprecision was the document which enabled the collation of material in the Californian Database. This was known as the California Client Development Evaluation Report (CDER). Although Dr Staines had not seen a copy of this document at the time, the point he was making was that, for one reason or another, inaccuracies or ambiguities could arise in the completion of this document which might further lead to less precision in the final estimate.
101 (I should interrupt to say that Dr Antony herself gave evidence in the proceedings. She explained how it was that the discrepancies had occurred when the PEQ form was completed. She confirmed that she would perhaps have answered some of the questions in the form differently if she had seen the videos of the plaintiff's physical abilities (Exhibits "EEE" and "PPP") which were shown to her in court. Of more importance, she agreed that there were a number of problems with the PEQ classification system itself. She gave examples of this: For example, she was not given any definitions which aided her completion of the form and she said she had to make "her own interpretations" (T 2385 lines 35-53). She agreed that there were some confusing aspects of the form which "confounded" the examiner when it came to filling in detail (T 2388 lines 15-18)).
102 Dr Staines was also concerned that there may be room for confusion in relation to the completion and interpretation of forms occurring in different countries. He said that doctors and health care professionals in different countries use the same terms with systematically different meanings. This is a further source of potential error.
103 Another matter which required considerable caution was this: The plaintiff was, by any standards, an unusual person. She had severe motor difficulties but essentially normal intellectual ability. The implication of Dr Shavelle's comments in his report as to the small number of people in the database who were comparable to the plaintiff meant that the estimate made by Dr Shavelle would be "less precise" than his estimates for people with more common patterns of cerebral palsy. Dr Staines was at pains to say that this was no criticism of Dr Shavelle's professional competence but, "an inevitable limitation of epidemiological methods when applied to sparse or limited data".
104 The next matter discussed related to the precision of the estimates themselves. This introduced first, the difficult concept of confidence intervals. Differences in confidence intervals, however, might affect the precision of the statistical estimates. Dr Staines turned then to another complicated issue. This related to the fact that there had been a need to include two different components in the life table prepared for the plaintiff. First, Dr Shavelle had calculated life expectancy from age 20 to age 35 directly from the Californian database. Secondly, he calculated life expectancy from age 35 onwards with reference to the Australian Life Table for the years 1996-1998. In other words, he made an assumption that the plaintiff had an elevated risk of death at age 35 equal to the ratio of mortality amongst 35 year old Australian females and that among 35 years females with cerebral palsy in the Californian Database. The assumption was then made that the extra risk of death experienced by the plaintiff, as compared to other Australian women, would decline smoothly towards a zero excess risk by the time the plaintiff reached the age of 100. From a statistical point of view, this assumption of a linear decline in the log of the relative risk of death with age, was reasonable. However, as no other studies had documented survival among people with cerebral palsy born much before 1960, the assumption was both untested and untestable.
105 There was a further area which raised issues of precision particularly in the light of a recent unpublished Western Australian study which suggested that the population of people with cerebral palsy born before the 1960's in Australia was quite different from populations born in the later decades of the century.
106 Dr Staines, having made these various cautionary points, then turned to the estimates themselves and to the central question he had been asked. First, he made a point of great significance. This was that there was a manifest difference between the approaches of Dr Wise and Dr Buckley on the one hand and Dr Shavelle on the other. It was in this context that he made the point already mentioned that the sophisticated statistical methods used by Dr Shavelle, although perfectly correct for selecting hazard rates for populations, were "fraught with difficulty" when applying the methods to an individual. Dr Wise and Dr Buckley were, in reality, answering a different question, namely - is there any specific reason to suppose that the plaintiff will die early, and if so, how early? Dr Shavelle was posing and answering quite a different question, namely, one which was concerned with a population of people like the plaintiff, as obtained from the Californian database.
107 In this context, Dr Staines returned to the point he had made in relation to the distinction which he thought necessary in relation to the term "life expectancy". Dr Staines took as an example one of the Shavelle estimates, namely the figure of an additional 33.3 years. This was stated to be "Ms Simpson's life expectancy" by Dr Shavelle . This was statistically correct, Dr Staines said, but "potentially very misleading". The obvious interpretation of the estimate was the plaintiff could expect to be dead by the age of 55. This was not correct, and such an inference represented an unwarranted shift from the very precise and technical meaning of expectancy, in probability theory, to the every day meaning of expectation.
108 Dr Staines then added back into the equation the various cautions he had expressed. With the addition of these matters to the equation he thought that "great caution" should be exercised in accepting the estimate so far as it related to the individual.
109 Finally, Dr Staines came to his conclusion. He did so against the background that certain clinical and related features that he had been informed of regarding the plaintiff suggested that her "life expectancy" (in the technical sense) might be "quite good" relevant to other people with cerebral palsy. For a number of reasons, he said that the survival experience of the Californian population in Dr Shavelle's study was not likely to fairly represent the plaintiff's future expectation of life. An important aspect of this opinion was the rarity of people like the plaintiff in the Californian database. Dr Staines expressed his conclusion in these terms:-
"For the various reasons discussed above it does not seem to me that any very credible estimate of Ms Simpson's real life expectancy can be made based on the Californian data, and the PEQ provided to Dr Shavelle. This does not reflect errors in Dr Shavelle's methods, which seem to me to be correct. Rather it reflects a failure to acknowledge the many sources of error in the data, in the clinical information available, and the statistical imprecision inherent in all estimates. The fact that Ms Simpson is a rather unusual person only makes matters worse.
It is not disputed that most people with cerebral palsy will have a reduced life expectancy. The evidence for this is overwhelming. The issue is the degree of credibility which can be placed in Dr Shavelle's estimate that Ms Simpson can expect to live a further 33.3 years as opposed to the 60.8 which other Australian women of her age can expect.
While it seems plausible that Ms Simpson will live fewer years than her non disabled counterparts rather than more, because of the multiple sources of uncertainty, it is not clear to me that any useful estimate of Ms Simpson's own life expectancy can be made in statistical terms. In this circumstance an estimate made on the basis of an individual assessment, by a person with suitable clinical experience would be useful."
110 It is necessary to note that Mr Brereton SC was very careful in cross-examination of Dr Staines to eliminate, to a degree, some of the cautionary concerns Dr Staines had about the accuracy and utility of the Californian database. Some of his concerns were clearly alleviated, for example, the possibility of there being inconsistency between the PEQ and CDR forms. However, despite this careful cross-examination, an overall impression was left that there were various areas in which imprecision might well remain. In this regard, Dr Staines made a further useful point. Statisticians are used to dealing with imprecision, they are able to make allowances for it. However, imprecision can often turn into unreliability and this is so, particularly where small sampling is involved (T 963 lines 26-37; line 47 to T964 line 10).
111 Secondly, Mr Brereton SC secured Dr Staines agreement to a proposition which he had conceded in the text of his report, namely that the plaintiff's life expectancy would not be the same as that of a non disabled counterpart (T 982 lines 10-15); and finally, Senior Counsel secured the witness' agreement to the proposition that the statistical material and the Shavelle type exercise provided "a starting point" for the adjustment to be made. This was a more accurate approach than that of Drs Buckley and Wise who had taken as their starting point the Australian Life Tables. This agreement was, of course, subject to the qualification as to whether the results of the Shavelle exercise were correct (T 969 lines 30-45).
112 There was one final matter dealt with during cross-examination which requires comment. Between pages T 976 and T 982, there was considerable dialogue between Mr Brereton SC and Dr Staines in relation to the proper methodology involved for clinicians who had taken the statistical material as a starting point for assessing the position of an individual and determining remaining life span in that context. The entirety of these pages is instructive for the methodology I regard as appropriate to the present case. There was a particular passage at T 976 line 45 which, to my mind, succinctly emphasised a real danger of imprecision in applying an estimate of life expectancy based on statistical material to the issue of a particular plaintiff's remaining years. The passage occurs against the background of a discussion that centred on the fact that only 22 people involved in the vast Californian database were similar to the plaintiff. Counsel had asked whether the fact that there were 22 like people would be likely to give a far better basis for a starting point than any experience that would be involved in a single clinician's experience. Dr Staines replied that because the sample was so small the 22 might be "atypical". He said that the possibility or even "modest errors" in the CDER could substantially affect precisely the make up of the 22 persons. He said at line 45: -
"I cannot speak to the quality of the California database, I know nothing about it. I have been unable to find, from the papers and documentation that I have been able to secure, what the procedures are for ensuing quality in the database.
I have very considerable experience working with other large routine health databases and I would be very concerned about using other large routine health databases to make inferences about rare events or about small groups of people."
113 Dr Staines said that, in his view, this was "potentially a major issue here" (T 976 line 35).