1. School of Public Health, The University of Sydney, Sydney, NSW, Australia 2006,

Transcription

1 Community preferences for the allocation of donor organs for transplantation Kirsten Howard 1, Stephen Jan 2, John M. Rose 3, Michelle Irving 1,4, Germaine Wong 1,4, Allison Tong 1,4, Jonathan C. Craig 1,4, Steven Chadban 5,6, Richard D. Allen 6, Alan Cass 2,7 1. School of Public Health, The University of Sydney, Sydney, NSW, Australia 2006, 2. The George Institute for Global Health, Missenden Rd, Camperdown, Sydney, NSW, Australia Institute for Transport and Logistics Studies, University of Sydney, NSW, Australia, Centre for Kidney Research, the Children s Hospital Westmead, Sydney, NSW, Australia Central Clinical School, Bosch Institute, The University of Sydney, Sydney, NSW, Australia Department of Renal Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney, NSW, Australia Menzies School of Health Research, PO Box 41096, Casuarina, NT, Australia 0811 Word Count: tables (total 7563) 1

2 ABSTRACT Background Whilst solid organ transplantation is the treatment of choice for those with organ failure, the demand for organs for transplant far exceeds supply. Organs are generally allocated to patients using a computer algorithm that is primarily based on time on waiting list, medical urgency, human leukocyte (HLA) matching, sensitization, and paediatric status. However, there is ongoing debate about the relative weighting that should be given to these, and other possible criteria. Although organs for transplant can be viewed as a community held resource, little is known about community preferences for how they should be allocated. This study uses discrete choice methods to assess community preferences for organ allocation. Methods Respondents recruited from the community via a web-based panel answered 30 scenarios presenting two hypothetical patients, and were asked to indicate which patient should receive a transplant if there was only one organ available. Patients were described in terms of factors such as age, sex, number of previous transplants, whether they or their family are registered organ donors; whether they have dependents or caring responsibilities, likely adherence, time on waiting list, likely survival with and without a transplant, likely quality of life (QOL) with and without a transplant, other comorbidities, and other lifestyle factors such as smoking status. All analyses were conducted in NLOGIT 4.0; a mixed logit model was used for analysis. Results 2051 respondents completed the survey. Respondents were representative of the age distribution of the Australian population (18+) with a mean age of 44 (range 18-83); 51% were female. All attributes significantly influenced choice of recipient, with the exception of sex. Compared to patients aged 40, younger patients were given higher preference, while older patients were less preferred. Family members being registered as donors, having dependents or caring responsibilities, and spending a longer time on waiting lists all increased priority; patients with possible non-adherence to treatment, smoking, drinking, obesity and some clinical comorbidities were less preferred. Patients with lower pre-transplant life expectancy and lower pre-transplant QOL were given a higher preference (need); as were patients who had a higher post-transplant life expectancy (LE) and a higher posttransplant QOL (utility). Pre transplant life expectancy was valued more highly than post-transplant life expectancy; one fewer year of pre-transplant life expectancy took an increase of 1.49 years in post-transplant life expectancy to compensate, but post-transplant QOL was valued more highly that pre-transplant QOL. When LE and QOL were combined into a single multidimensional measure of quality adjusted life years (QALYs) there was a higher preference for giving organs to patients with lower pre-transplant QALYs, compared to higher posttransplant QALYs with one fewer pre-transplant QALY requiring an extra 2.12 post-transplant QALYs to compensate. Conclusions Many factors considered in current organ allocation algorithms are consistent with community preferences, such as preference given to children and a longer time on the waiting list. Other factors were also considered important, including donor status of the patient or family; number of previous transplants, and whether the recipient had dependents or caring responsibilities. Lower pre-transplant life expectancy and QALYs (which can be thought of as need ) were more important than higher post-transplant life expectancy or QALYs. This suggests that whilst there is a current move toward allocation algorithms favouring high life expectancy after transplant (utility), emphasis also ought to be given to need as measured by pre-transplant life expectancy and quality adjusted life expectancy. 2

3 INTRODUCTION All across the world, the demand for organs far outstrips supply (insert reference), making them a scarce community resource, the supply of which is reliant on public willingness to donate. Patients waiting for kidney transplants make up by far the greatest proportion of patients on waiting lists around the world 1. The current median waiting time for a deceased donor kidney in Australia is between four and six years, but can exceed 10 years, depending on the clinical status of the patient 1. In the US, this waiting time is between two to three years depending on blood type 2. With waiting lists as they are, the challenge clearly rests in, not only increasing availability of donor organs, but also in deciding how these scarce organs should be allocated. It is currently unclear how community preferences for the allocation of donor organs are actually incorporated into existing allocation algorithms 3, or the extent to which existing algorithms adequately account for community preferences. In most countries, allocation algorithms consider factors such as time on the waiting list, medical urgency, human leukocyte antigen (HLA) matching, antibody sensitization, blood group, and paediatric status 4;5. There is ongoing debate about the relative weighting that should be attached to these criteria 4;6;7, and some scepticism exists amongst both patients 8 and the general community about the fairness of allocation. Recent changes to allocation protocols such that they now explicitly favour expected graft and post-transplantation life-expectancy at the expense of other criteria have also heightened controversy. In the US, the Kidney Committee of the Organ Procurement and Transplantation Network have proposed an allocation system that would preferentially allocate the top quintile of donor kidneys with the best expected graft survival to the top quintile of transplant candidates with the longest predicted post-transplant survival. For the remaining donors and recipients there would be broad age-matching within a 30 year age range. 6. This approach seeks to maximize graft survival, but at the expense of older and sicker patients. Similarly in Australia, recent changes limit the eligibility for inclusion on the transplant waiting list, a necessary step prior to even the possibility of receiving a donor organ. The Transplantation Society of Australia and New Zealand (TSANZ) organ allocation protocols explicitly only allow waitlisting of candidates if they have a sufficient estimated post-transplant life expectancy 5. For example kidney transplant recipients are only considered eligible for inclusion on the transplant waiting list if they have an 80% chance of 5 year survival post-transplant, and liver transplant recipients are expected to have at least a 50% probability of 5 year survival respectively 5. These requirements are likely to lead to significant disparities in the allocation of organs, with diabetics, older recipients (over 65 years) and Indigenous people losing opportunities to benefit 1;9 simply because they are not even able to be included on the waitlist, and are therefore never eligible for a donor organ. This approach seeks to maximize graft and recipient survival, however, by limiting the transplant opportunities for older, sicker patients (those that potentially stand to gain the most), we may not be maximizing the incremental health gains possible from transplantation. As in allocation of almost all healthcare resources, there are two competing imperatives: the desire to maximize value and the desire to consider the distributional consequences of allocation decisions and be fair. In the health economics literature, this is referred to as the equity-efficiency trade-off 10 ; in the clinical transplantation literature it is often characterized as the trade-off between clinical utility/maximizing benefit and need 6;11. Clinical utility is about getting maximal benefit from a donor organ, and often focusses on either graft, or post-transplant recipient survival. It could also refer to maximizing post-transplant quality of life, however this aspect of clinical benefit is referred to less often. Need can refer to a variety of factors: time on the waiting list, low pre-transplant survival or quality of life, or in some circumstances, older age 12, or more comorbidities, because those patients are likely to have reduced survival and quality of life. Where the balance should sit between maximizing benefit and need/ equity in organ allocation is currently unclear, little is known about community preferences for how organs should be allocated, or about the relative weighting of clinical utility factors vs need/equity factors. This study uses discrete choice methods to assess community preferences for organ allocation. 3

4 METHODS This study formed part of a broader study, the PArADOx Study, examining community preferences for the donation and allocation of organs 13. Study Population and recruitment This study examines the preferences of the Australian community, using a discrete choice experiment (DCE) for the allocation of donor organs to transplant recipients. The DCE was conducted as a web-based survey using an existing online research panel administered by an external organisation (Survey Sampling International, SSI). From this panel, SSI alerted men and women aged over 18 that the survey was available. Participant information sheets and consent were provided online before the survey commenced. Participants accrued points by completing the survey. The Discrete Choice Experiment Community preferences were assessed using a DCE 14-16, a quantitative technique that assumes a health care intervention can be described by its characteristics (attributes). Attribute levels are varied systematically in a series of questions and respondents choose the option that they prefer for each question. In the current study, attributes described the characteristics of a potential transplant recipient, rather than a programme or intervention. DCEs can determine which attributes are driving preferences (e.g. for or against a particular recipient) and the trade-offs between attributes that people are willing to accept (e.g. the amount of health gain willing to be foregone to prioritise a patient who has been waiting longer). Attributes and attribute levels Attributes used to describe potential transplant recipients were based upon systematic reviews of the literature concerning community 3 and healthcare provider preferences on allocation 17, focus groups conducted in patients 8 and the community 12 and one on one interviews with clinicians involved in transplantation and management of patients on the transplant waiting list 11. Table 1 describes the attributes and levels used in the final DCE. From these sources, both survival and quality of life were important factors that people identified; these were often balanced against the idea of need. Survival and quality of life (QOL) were discussed both in the context of maximising the value of a transplant (often referred to as utility, or capacity to benefit), and in terms of need, where lower pre-transplant survival and quality of life was often viewed as having a greater need. We therefore specified two survival / life expectancy variables, pre -and post-transplant, to capture respectively, need and utility. Quality of life was also considered separately as pre- and post-transplant, and was described on a scale of 0-10, where 0 was being dead, and 1 was full health. Clinical comorbidities were based upon common comorbid conditions in patients waiting for transplant. 4

5 Table 1 Attributes and levels used to describe potential organ recipients Attribute Level description corresponding to each Attribute description name code Age Recipient age 5 years old; 15 years old; 25 years old; 40 years old; 55 years old; 70 years old sex Recipient sex MALE; FEMALE prev Whether they have had previous transplants no previous transplants; one previous transplant; 2 or more previous transplants donor Whether they, or their family members are registered organ donors Not registered as donors; Yes, registered as donors depend Whether they have dependents or caring No dependents or caring responsibilities; Yes, they responsibilities have dependents or caring responsibilities adher How well they follow their doctor's advice always takes tablets according to doctors advice; occasionally misses a tablet (eg 1 per month or less); sometimes misses a tablet (eg 1 per week) wait Time on the waiting list 6 months; 1 year; 2 years; 5 years; 10 years LE_pre How long they are likely to survive if they DO NOT receive a transplant 1 year or less; 2 years; 5 years; 10 years; 15 years LE_post How long they might survive if they DO receive a transplant 2 years; 4 years; 8 years; 12 years; 25 years Their quality of life if they DO NOT receive a QOL_pre transplant (on a scale of 0= being dead, 10= full health for someone their age) 1/10; 3/10; 5/10 Their quality of life if they DO receive a QOL_post transplant (on a scale of 0= being dead, 10= full health for someone their age) 3/10; 6/10; 9/10 COMORB None; diabetes; previous cancer; diabetes and Whether they have other health conditions that previous cancer; hepatitis or other viral illness (being might also affect their life expectancy treated) SMOK Whether they smoke does not smoke; ex-smoker; current smoker DRINK Whether they drink alcohol does not drink alcohol; drinks alcohol occasionally; drinks alcohol on 5 or more nights per week OBES Whether they are obese not obese; obese Study design and questionnaire A two-step pilot study was conducted using an initial efficient design; 10 adults completed the DCE in a face to face interview to ensure that attributes were understandable, and a further 108 respondents aged over 18 completed an online pilot questionnaire. The pilot consisted of 24 questions presenting two alternatives where participants were asked to choose the one patient they would give the donor organ two. An opt-out or status quo alternative was not included because it was clinically implausible; if a donor organ becomes available, the scarcity of available organs means it is always allocated to a potential recipient. The pilot study included two attributes describing recipient comorbidities, this was simplified to one attribute and the number of levels was reduced. A mixed logit model was estimated to inform the final study design. A d-efficient design is reached by testing alternative attribute level combinations, with designs with lower d-errors representing more statistically efficient designs. The aim of these designs is to produce data that will maximise the t-ratios for the design parameters 18;19. The final DCE design was a d-efficient design of 150 choice tasks, blocked into 5 blocks of 30 questions, created using NGENE design software ( (d-error of 0.004; s-estimate of 404). Respondents completed 30 choice questions presenting two hypothetical patients, and were asked to indicate which patient should receive a transplant if there was only one organ available (Table 2). Additional sociodemographic information, including education, income, employment, marital status, organ donor status, prior experience with donation for themselves and family/friends, and their views on organ donation was also collected. Constraints were placed on the design such that recipients who were children (aged 5 or 15) could not smoke or drink, could not have dependents or caring responsibilities, and based on clinical plausibility, could not have a comorbidity of hepatitis or other viral illness (being treated) ; children aged 5 could not have been on a waiting list for longer than 5 years; and post-transplant survival and quality of life was specified as equal or greater than pre-transplant survival and quality of life (based on the existing clinical literature about the effectiveness of transplant as a treatment for organ failure). The study was approved by the University of Sydney Human Research Ethics Committee. 5

6 Table 2: example scenario Analysis We used a mixed logit (ML) model with a panel specification. Categorical variables were effects coded. For recipient age, 40 years old was specified as the base level, as this is the mean age of patients on a transplant waiting list in Australia; linear variables were specified as normal distributions. The initial model was estimated to consider life expectancy and quality of life separately; another model considered life expectancy and quality of life combined into a single, multidimensional health state, the quality adjusted life year (QALY). All respondent demographic variables were effects coded. A mixed logit model allows consideration of the full distribution of a parameter estimate, and estimates random parameters. Random parameter implies that each individual has an associated parameter estimate on the specified distribution. Whilst the exact location of each individual s preferences on the distribution may not be known, estimates of individual-specific preferences can be accommodated by deriving the individual s conditional distribution, based within sample on their choices (i.e. prior knowledge) 15 Additional discussion is available elsewhere 14;15;20. Results are reported as beta coefficients and also as the odds (and 95% CI) of being allocated an organ. We estimated the following value functions; explanatory variables are the attributes describing the potential recipient. U = V+ε = β 0 +β 1 Age5 +β 2 Age15+β 3 Age25+β 4 Age55+β 5 Age70+ β 6 Sex+β 7 Prev + β 8 Donor+ β 9 Depend+ β 10 Adher + β 11 Wait +β 12 LE_Pre+β 13 LE_post+β 14 QOL_pre+β 15 QOL_post+ β 16 Diab+β 17 PrevCan + β 18 DiabCan+ β 19 Viral + β 20 exsmok +β 21 CurrSmok +β 22 Drink1+β 23 Drink2+β 24 Obes+ε U = V+ε = β 0 +β 1 Age5 +β 2 Age15+β 3 Age25+β 4 Age55+β 5 Age70+ β 6 Sex+β 7 Prev + β 8 Donor+ β 9 Depend+ β 10 Adher + β 11 Wait +β 12 QALY_Pre+β 13 QALY_post + β 14 Diab+β 15 PrevCan + β 16 DiabCan+ β 17 Viral + β 18 exsmok +β 19 CurrSmok +β 20 Drink1+β 21 Drink2+β 22 Obes+ε We checked the direction of estimated coefficients to verify whether they were consistent with a priori expectations; and examined goodness of fit using pseudo R 2 and AIC. All analyses used NLOGIT Version (Econometric Software, Castle Hill, NSW, Australia, RESULTS Respondent characteristics Two thousand and fifty one (2051) respondents aged completed 30 choice questions, giving a total of choice observations. Respondents had a mean age of 44, and 51% were female. Respondent characteristics are presented in Table 3. Education and income are approximately equivalent to the Australian population as a whole; and the proportion of respondents with private health insurance was also reasonable comparable (53% in our sample, 45% in Respondent views on organ donation are also congruent with previous reports 22, with around 75% of respondents indicating there were in favour, or strongly in favour of organ donation. 6

7 Approximately 45% of respondents indicated that they had registered a donation intention (43.8% as donors, 1.3% as non-donors); however, only around 40% of them have done so through the Australian Organ Donor Registry (through Medicare) (361 of 2051 (17.6% of total respondents). Most indicated that the registration was done via their driver s license, however the legal standing and validity of this registration varies by state jurisdiction. The Donor Registry reports that legally valid consent registrations from around 11% of the ABS estimated population and intent registrations from approximately 24% of the ABS estimated population. 23 7

8 Table 3 Respondent characteristics (Australian community sample aged over 18) N (total = 2051) % Mean age (SD) 44.5 (15.6) Range: Sex (female) % Education Did not complete high school % Completed high school % TAFE/technical/trade Certificate or Diploma % Completed university (Bachelor or higher) % Marital status Married/living together % Current partner % Divorced/widowed/single % Employment Full time % Part time % Home duties % Retiree / pensioner % Not working/ full time student % Annual Household Income < $35, % $35,000 - $65, % $65,001 $95, % >$95, % Did not answer % Language spoken at home English % Have private health insurance % Pre-survey views on organ donation Against / very much against % Neither for nor against % In favour / very much in favour % Current OD views influenced by religious beliefs 1 Not at all % % 3 somewhat influenced % % 5 very influenced % Registered donation intention No % Yes, registered as donor % Yes, registered intention to NOT donate % Means of registration Driver s license % (of registered) Organ donor register through Medicare % (of registered) Ever discussed donation decision with family Yes % (of registered) No % (of registered Ever received transplant % Family member or friend received transplant % Currently on transplant waiting list % Family member or friend currently on waiting list % Family member or friend ever been on waiting list % Family or friend ever been a donor % 8

9 Preferences Most attributes significantly influenced choice of recipient, with the exception of sex, and the presence of diabetes (Model 1 and 2), or previously treated cancer as a comorbidity (Model 2), and occasionally drinking alcohol (Model 2); attributes also behaved in the a priori expected directions. As would be expected, recipient age significantly influenced the likelihood of being allocated an organ. The base age group was 40 years as this is the approximate mean age of patients on a transplant waiting list in Australia. Respondents had a significantly higher preference for allocating organs to 5, 15 and 25 year old recipients compared to 40 year old recipients. The odds of preferring younger recipients compared to 40 year old recipients ranged from to for Model 1 and to (Model 2). Respondents also had a significantly lower preference for allocating organs to 55 or 70 year old recipients, compared to patients who were 40. Whether the recipient themselves, or a family member was a registered donor, increased the odds of being allocated the donor organ (OR= (Model 1); (Model 2)), as did the potential recipient having caring responsibilities or dependents (OR=1.421 (Model 1); (Model 2)). As was expected from our previous qualitative work, a longer time on the transplant waiting list was also associated with a higher chance of being allocated a donor organ. Each year spent on the waiting list increased the odds of being allocated an organ by (Model 1); (Model 2) Increasing numbers of previous transplants led to a lower preference for allocating the donor organ (OR (Model 1); (Model 2)) for every previous transplant. The likelihood of receiving a transplant decreased slightly as non-adherence increased, for example, moving from always taking tablets according to doctors advice to occasionally missing a tablet (1 per month or less), or moving from occasionally missing a tablet to sometimes missing a tablet (eg 1 per week), decreased the odds of being allocated an organ by between 3% and 6% (OR = 0.942, (Model 1); (Model 2)). The influence of clinical comorbidities on the likelihood of being allocated an organ was also assessed. Comorbidities included diabetes, previous cancer, diabetes plus previous cancer and currently treated hepatitis or other viral illness. Compared to patients with no comorbidities, patients with diabetes alone were not significantly more or less likely to be allocated an organ in either model, and in Model 2, the likelihood of being allocated an organ was also not influenced by a past history of previously treated cancer. Having diabetes and previously treated cancer (OR= ), and having treated hepatitis or other viral illness ( ) were all associated with significantly lower preferences for allocation. Lifestyle factors such as smoking, drinking and obesity also reduced the likelihood of being allocated an organ. Compared to never having smoked, the recipient being an ex-smoker had a lower likelihood of being allocated an organ (OR= ), and being a current smoker reduced the likelihood even further (OR= ), to the lowest likelihood of any attribute. Drinking alcohol also influenced respondent preferences: In modle 1, drinking occasionally reduced the odds to 0.911; however in model 2, having an occasional drink did not significantly influence the likelihood of transplant (OR=0.992). Drinking more than 5 nights per week reduced the odds to Obesity was also associated with a lower preference (OR= ). In Model 1, patients with lower pre-transplant life expectancy (OR = per 1 year less life expectancy) lower pre-transplant QOL (OR=1.059 per 1/10 unit lower QOL) and lower pre-transplant QALY (OR=1.16 per 1 QALY less pre-transplant) were given a higher preference (need). Patients who had a higher post-transplant life expectancy (OR = per extra year), a higher post-transplant QOL (OR=1.119 per 1/10 unit higher QOL) (capacity to benefit / utility) and higher post-transplant QALYs (OR=1.07 per additional QALY) were also given a higher preference. Pre-and post-transplant life expectancy and QOL were not valued equivalently. Pre transplant life expectancy was valued more highly than post-transplant life expectancy; one year less of pre-transplant life expectancy was equivalent to an increase of 1.49 years in post-transplant life expectancy. On the other hand, post-transplant QOL was valued more highly that pre-transplant QOL; a 1 unit lower pre-transplant QOL score required only an increase of 0.51 in post-transplant QOL score for equivalence. When life expectancy and quality of life were combined to a QALY measure (Model 2), pre-transplant QALYs were valued more highly than post-transplant QALYs. One less QALY pre-transplant took an increase of 2.12 QALYs post-transplant to compensate (i.e. one additional post-treatment QALY took only 0.45 fewer pretransplant QALYs to be equivalent). 9

10 Table 4 Community preferences for allocation (mixed logit) Model 1 (life expectancy and QOL separate Model 2 (life expectancy and QOL combined to QALYs) OR of OR of 95% 95% Recipient Characteristic β SE SD of β receiving P value β SE SD of β receiving P value CI CI organ organ Constant > >0.05 Age 5 years old ( ) < ( ) < years old * ( ) < * ( ) < years old ( ) < ( ) < years old (base level) years old ( ) < ( ) < years old ( ) < ( ) < Previous Transplants (for each previous transplant) ( ) < ( ) < Adherence (for each level of reduced adherence) * ( ) < * ( ) < Female (vs male) * (NS) ( ) > * ( ) >0.05 Family registered as donors (vs not registered) ( ) < ( ) < Recipient caring responsibilities or dependents (vs not) ( ) < ( ) < Time on waiting list (for each extra year on waiting list) ( ) < ( ) < Life expectancy without transplant (for every 1 year less of pre-transplant survival) ( ) < Life expectancy with transplant (for every extra year of post-transplant survival) ( ) < Expected QOL without transplant (per 1 unit less of pre-transplant QOL) ( ) < Expected QOL with transplant (per 1 unit increase in post-transplant QOL) ( ) < Expected QALYs without transplant (for every1 less QALY of pre-transplant survival) ( ) < Expected QALYs with transplant (for every extra QALY of post-transplant survival) ( ) < Comorbidities (vs none) diabetes (NS) ( ) > ( ) >0.05 previous cancer ( ) ( ) >0.05 diabetes and previous cancer * ( ) < * ( ) < hepatitis or other viral illness (being treated) * ( ) < * ( ) < Smoking status (vs non-smoker) ex-smoker * ( ) < * ( ) < current smoker ( ) < ( ) < Drinking status (vs non-drinker) drinks alcohol occasionally * ( ) * ( ) >0.05 drinks alcohol on 5 or more nights per week ( ) < ( ) < Obese (vs Not obese) ( ) < ( ) < Model Fit LL(0) LL(β) R Adjusted R AIC (normalised) Number of parameters Number of respondents (observations) 2051 (61530) 2051 (61530) *SD of random parameter non-significant at p=0.05, all others significant at p=

11 DISCUSSION Receiving a donor organ is two stage process; patients are listed on a transplant waiting list, and then when a donor organ becomes available, it is allocated to patients with the highest priority from that waiting list. Organs are currently allocated to potential recipients based upon algorithms which consider factors such as human leukocyte antigen (HLA) matching, antibody sensitization, time on the waiting list, medical need/ urgency, and paediatric status 4;5. Recently the US has proposed an allocation system where the best organs go to the best recipients, ie those defined as having the longest predicted post-transplant survival 6. In Australia, the current TSANZ allocation protocols preclude even the waitlisting of patients with less than, for example 80% chance of 5 year survival posttransplant for kidney transplants and a 50% chance of 5 year survival for liver transplants. 5 Whilst these approaches maximize the clinical utility of a donor organ, the potential gain in post-transplant survival comes at the expense of equity and fairness for patients waiting for organs. Qualitative research suggests that the community considers both clinical utility and need to be important in allocating donor organs 3;12, but does not explicitly quantify the influence of various factors on the likelihood of being allocated a donor organ. Our study examines the extent to which the community is willing to trade between various patient attributes (including those characterizing clinical utility and need) in deciding about organ allocation. We found that many of the criteria used in allocation algorithms were broadly consistent with community preferences, such as higher preference for younger recipients and longer time on the waiting list. Need or medical urgency can be characterised by lower pre-transplant life expectancy, lower QOL or lower pre-transplant QALYs (quality adjusted life years, a measure that combines both survival and quality of life). Our respondents gave a higher preference to patients with these characteristics, which is consistent with the principle of medical urgency or need outlined in many allocation protocols 5. Clinical utility was also considered, and patients with higher posttransplant life expectancy, higher QOL and higher post-transplant QALYs were also given higher preference. Preand post-transplant life expectancy, QOL and QALYs were not valued equivalently. Pre-transplant life expectancy was valued more highly than post-transplant life expectancy; and post-transplant QOL was valued more highly that pre-transplant QOL. When combined into a single measure, QALYs, there was a significantly higher preference, for allocation to patients with lower pre-transplant expected QALYs with one pre-transplant QALY requiring more than two post-transplant QALYs for equivalence. Other factors were also considered important in determining the allocation of donor organs: the likelihood of allocation increased if they, or a family member, were a registered donor, or if they had caring responsibilities or dependents, whereas a higher number of previous transplants and lower adherence to medical advice both decreased the likelihood of allocation. Israel recently adopted a system designed to encourage more organ donations by giving a higher priority for immediate family members of people who are registered donors 24, a system that also exists in Singapore, based on the idea of reciprocity - if you are willing to accept an organ, then you should also be willing to donate. Reciprocity was also identified as an important factor in our own qualitative work of factors influencing the decision to become a donor 25. It is interesting to note the community respondents in our survey have rewarded this principal of reciprocity with an increased likelihood of being allocated an organ for those who are registered, or whose family are registered as donors. The presence of some clinical comorbidities and lifestyle factors such as drinking and smoking decreased the likelihood of being allocated an organ, however, it is important to note that although the likelihood of being allocated an organ decreased to a greater or lesser degree, the public did not believe that these factors should completely preclude being allocated an organ. Few previous studies have quantified community preferences for the allocation of organs using a stated preference method. Ratcliffe et al 26;27, used a DCE to examine preferences for allocation of donor livers in a sample of 303 respondents recruited from academic and non-academic staff of a UK university. Broadly consistent with our results, respondents had a higher preference for patients with a higher post-transplant survival, and a longer waiting time; while older patients and those where alcoholic liver disease was the reason for a liver transplant were given a lower priority. Interestingly, unlike our study, patients who were receiving a re-transplant were given a higher preference. Cass et al 28 conducted a survey of 183 Australian nephrologists to assess decision making about suitability of various patients for kidney transplantation. Although not strictly a DCE, respondents were shown a series of hypothetical patients described in terms of age, sex, ethnicity, weight, compliance, diabetes, smoking status and cardiovascular disease, and asked to indicate whether they would recommend transplantation. Sex and ethnicity were not significant independent predictors; however, consistent with our study, increasing age was associated with a lower likelihood of recommendation for transplant compared to patients aged 30-49, year old patients (OR 0.65) and year 11

12 old patients (OR 0.15) were significantly less likely to be recommended for transplant. Patients who were obese (OR 0.50), who smoked (OR 0.36), who were non-compliant (OR 0.17), or who had cardiovascular disease (OR ranged from ) were all less likely to be recommended for transplant compared to patients who did not have these characteristics. Although the community also viewed some of these factors as influencing the likelihood of a transplant, the penalty applied by the community was much less severe than that applied by nephrologists. Patients who had diabetes (OR=0.67) were also less likely to be considered for transplant, whereas in our study, diabetes did not significantly influence the likelihood of allocation. Nephrologists are likely to have interpreted the clinical consequences of diabetes differently to the general population; they may have been imputing the downstream consequences (in terms of reduced survival) associated with diabetes in kidney disease patients, even though survival was not a specified attribute. This analysis specifically focuses on the individual recipient characteristics, as would be specified in a potential allocation policy algorithm 5. Although we have identified significant heterogeneity in many of the parameters, this current analysis does not consider interactions between attributes, or between respondent characteristics and attributes. Our rationale for this is that in a policy sense, respondent characteristics would not be included in any potential allocation algorithm that might be informed by this work. Additional analyses are planned that will examine interactions, for example interactions between recipient age and waiting time, or comorbidities, where the value attached these factors may vary with recipient age. Similarly, it is possible that the value attached to clinical utility attributes (post-transplant life expectancy and QOL) may be influenced by need, as defined by either longer waiting time, or poorer current health (pre-transplant survival and QOL). It is also possible that some attributes could be modelled differently, for example pre and post- transplant survival attributes could be subtracted to estimate the gain in life expectancy, the value of which might also be dependent upon the starting point (ie pre-transplant life expectancy), or recipient age, or waiting time. A similar approach could be used for QOL. In addition, although they would not be used to inform any allocation algorithm, we will also examine interactions between sociodemographic characteristics of respondents and attributes, to better understand preference heterogeneity. Our respondent sample may also not be completely representative of the Australian adult population. They are broadly similar on characteristics such as education and income, and approximately similar proportions also report being in favour of organ donation as has been previously reported 22. Nevertheless, the proportion of respondents reporting some previous experience with donation or waitlisting (either themselves personally, or family/friends) seem somewhat higher than might be expected given the number of donors in Australia each year, and the number of people currently on waiting lists, although there are no data available from the general population about these factors to support or refute these concerns. Implications for allocation policy and conclusions Some factors currently considered in organ allocation algorithms are broadly consistent with community preferences, such as preference given to children and a longer time on the waiting list. Other factors not currently considered, were also deemed important, including donor status of the patient or family; number of previous transplants, and whether the recipient had dependents or caring responsibilities, as well as lifestyle factors and clinical comorbidities. While age of recipient significantly influenced allocation, with a higher preference for younger recipients, this was not limited to just children under 18, as specified in current allocation protocols. Younger adults (eg 25 years) were also viewed more favourably, suggesting that the current points system where eg for kidney transplant patients under 18 get bonus points perhaps should encompass a broader age range, or perhaps a sliding scale of additional points based on age, and relative preference could be developed instead. It is also important to note that although there was a lower preference for older recipients, this should not be interpreted as meaning that older recipients should not get any organs. Indeed, the fact that there is also a strong preference for people who have been on the waiting list for longer, and for those with lower pre-transplant life expectancy and quality of life, and that older recipients are more likely be represented in these groups, suggests that there are likely to be trade-offs and interactions between age and these other factors. Post-transplant QOL was valued more highly that pre-transplant QOL, suggesting that if allocation algorithms continue to focus on maximising post-transplant health outcomes, quality of life improvements, rather than survival, are important to consider. Lower pre-transplant life expectancy and quality adjusted life expectancy (QALYs) (need) were more important than higher post-transplant life expectancy/ QALYs (clinical utility), on a year per year 12

13 basis. This suggests that whilst there is a current move toward allocation algorithms favouring high life expectancy (utility) after transplant, emphasis also ought to be given to need as measured by pre-transplant life expectancy. Most importantly when QOL and life expectancy were combined into a single multidimensional measure of health outcome (QALYs) to evaluate their joint effect on allocation, the importance of need far outweighed the importance of utility, with one less pre-transplant QALY requiring more than 2 post-transplant QALYs for equivalence. Considering only patients with high post-transplant survival 5 as potential recipients, and precluding those with lower life expectancy or quality adjusted life expectancy from even being eligible for the transplant waiting list appears to be misaligned with community preferences. 13

14 Acknowledgements: The PArADOx (Preferences for the Allocation And Donation of Organs) Study is funded by an Australian Research Council Discovery Project Grant DP The funding organisations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review or approval of the manuscript. Reference List (1) Australia and New Zealand Organ Donation Registry (ANZOD). ANZOD 2012 report. Available from: URL: (2) UNOS. United Network for Organ Sharing (UNOS) (USA) Available from: URL: (3) Tong A, Howard K, Jan S, Cass A, Rose J, Chadban S et al. Community preferences for the allocation of solid organs for transplantation: a systematic review. Transplantation 2010; 89(7): (4) Leichtman AB, McCullough KP, Wolfe RA. Improving the allocation system for deceased-donor kidneys. New England Journal of Medicine 2011; 364(14): (5) The Transplantation Society of Australia and New Zealand. Organ Transplantation from Deceased Donors: Consensus statement on eligibility criteria and allocation protocols Version 1.2. Available from: URL: (6) Hippen BE, Thistlethwaite JR, Jr., Ross LF. Risk, prognosis, and unintended consequences in kidney allocation. New England Journal of Medicine 2011; 364(14): (7) Axelrod DA, Pomfret EA. Race and sex disparities in liver transplantation: progress toward achieving equal access? JAMA 2008; 300(20): (8) Tong A, Jan S, Wong G, Craig JC, Irving M, Chadban S et al. Patient preferences for the allocation of deceased donor kidneys for transplantation: a mixed methods study. BMC Nephrology 2012; 13:18. (9) Yeates KE, Cass A, Sequist TD, McDonald SP, Jardine MJ, Trpeski L et al. Indigenous people in Australia, Canada, New Zealand and the United States are less likely to receive renal transplantation. Kidney International 2009; 76(6): (10) Wagstaff A. QALYs and the equity-efficiency trade off. Journal of Health Economics 1991; 10: (11) Tong A, Howard K, Wong G, Cass A, Jan S, Irving M et al. Nephrologists' perspectives on waitlisting and allocation of deceased donor kidneys for transplant. American Journal of Kidney Diseases 2011; 58(5): (12) Irving MJ, Tong A, Wong G, Chadban S, Allen R, Craig JC et al. Community preferences for the allocation of organs for transplantation. Nephrol Dial Transplant 2013; Accepted March 18, (13) Howard K, Jan S, Rose J, Chadban S, Allen RD, Irving M et al. Community Preferences for the Allocation & Donation of Organs--the PAraDOx Study. BMC Public Health 2011; 11:386. (14) Hensher DA, Rose JM, Greene WH. Applied Choice Analysis. A Primer. 1st ed. Cambridge: Cambridge University Press; (15) Hensher DA, Greene WH. The mixed logit model: The state of practice. Transportation 2003; 30:

15 (16) Bridges JF, Hauber AB, Marshall DA, Lloyd A, Prosser LA, Regier DA et al. Conjoint Analysis Applications in Health a Checklist: A Report of the ISPOR Good Research Practices for Conjoint Analysis Task Force. Value in Health 2011; 14(4):5. (17) Tong A, Jan S, Wong G, Craig JC, Irving M, Chadban S et al. Rationing scarce organs for transplantation: healthcare provider perspectives on wait-listing and organ allocation. Clinical Transplantation 2012; DOI: /ctr (18) Bliemer, M. C. J., Rose, J. M., and Hess, S. Approximation of Bayesian efficiency in experimental choice designs. <[11] Source> [<[29] Medium>] <[05] Last Update> [cited <[28] Access Date>]; <[12] Volume>(<[15] Edition>) (19) Rose, J. M. and Bliemer, M. C. J. Sample optimality in the design of stated choice experiments. <[11] Source> [<[29] Medium>] <[05] Last Update> [cited <[28] Access Date>]; <[12] Volume>(<[15] Edition>) (20) Lancsar E, Louviere J. Conducting discrete choice experiments to inform healthcare decision making: a user's guide. Pharmacoeconomics 2008; 26(8): (21) Commonwealth Department of Health and Ageing. Private Hospital Insurance Available from: URL: (22) Mathew T, Faull R, Snelling P. The shortage of kidneys for transplantation in Australia. Medical Journal of Australia 2005; 182(5): (23) Medicare Australia. Australian Organ Donor Register report, June Available from: URL: (24) Chandler JA, Burkell JA, Shemie SD. Priority in organ allocation to previously registered donors: public perceptions of the fairness and effectiveness of priority systems. Progress in Transplantation 2012; 22(4): (25) Irving MJ, Tong A, Jan S, Cass A, Chadban S, Allen RD et al. Community attitudes to deceased organ donation: a focus group study. Transplantation 2012; 93(10): (26) Ratcliffe J. Public preferences for the allocation of donor liver grafts for transplantation. Health Economics 2000; 9(2): (27) Ratcliffe J, Buxton M, Young T, Longworth L. Determining priority for liver transplantation: a comparison of cost per QALY and discrete choice experiment-generated public preferences. Applied Health Economics & Health Policy 2005; 4(4): (28) Cass A, Cunningham J, Anderson K, Snelling P, Colman S, Devitt J et al. Decision making about suitability for kidney transplantation: Results of a national survey of Australian nephrologists. Nephrology 2007; 12: