- Email: [email protected]
Preferences to participate in treatment decision making: The adult model
Preferences to Participate in Treatment Decision Making: The Adult Model Lesley F. Degner,
ONCERN ABOUT the degree of control that people are currently able to assume in treatment decision making has intensified as a result of beginning empirical evidence that individuals who exercise such control have better health care outcomes.1-3 One of the greatest challenges facing investigators is how to measure the degree of control that consumers actually want in health care. This article describes the development of the Control Preferences Scale (CPS), a general measure of a unidimensional construct, consumer preferences about participating in treatment decision making. This measure has recently been adapted for use with parents of children with cancer.
C
The Construct The control preferences construct (Fig 1) emerged from a descriptive theory of how treatment decisions are actually made in health care.4 A central factor influencing decision making was who actually controlled the design of treatment. Four patterns of control were identified and described: provider-controlled, patient-controlled, familycontrolled, and jointly controlled decision making. This description provided the initial parameters for development of the control preferences construct. Description of the patterns of control led to the question: “Do patients have preferences about the degree of control they actually want to exercise in treatment decision making?” Participant observation in oncology clinics provided evidence for the existence of such preferences. Some patients refused to
From the Faculty of Nursing, University of Manitoba, Winnipeg, Manitoba. Address reprint requests to Lesley E Degner, PhD, RN; St Boniface Genera/ Hospital Research Centre, 357 Tache Ave, Winnipeg, Manitoba RZHZA6. o 1998 by Association of Pediatric Oncology Nurses. 1043-4542/98/l 503-0102$3.00/O
PhD, RN
become involved in selecting their own treatment, even when urged to do so by their physician. Others indicated a need to discuss the available options with their physician. Rarely, there were patients who had obviously prepared themselves beforehand, making it clear that they would be deciding what treatment they received. These observations led to the formulation of the control preferences construct, defined as “the degree of control an individual wants to assume when decisions are made about medical treatment.” Patients were hypothesized to have differing preferences about keeping, sharing, or giving away control over treatment decision making to their physician. That is, any particular patient would have an individual preference that would fall at a particular point along that psychological continuum.
The Measurement Model The hypothesis that patients have preferences about their level of involvement in treatment decision making guided selection of the measurement model, unfolding theory.5 Unfolding theory is based on the premise that an individual’s preference corresponds to an ideal point on a continuum. This ideal point can be derived by presenting successive paired comparisons of stimuli that fall along the continuum. In the case of the CPS, the stimuli are five cards that describe different roles in treatment decision making (Fig 2). The ideal point is represented by the order in which the subject places the cards, from the most to the least preferred.
Presentation Procedures in a Research Context Administration of the CPS involves subjects sorting a series of cards through successive paired comparisons. The result of these comparisons consists of an ordered permuta-
Journal ofpediatric Oncology Nursing, Vol 15, No 3, Suppl 1 (July), 1998: pp 3-9
3
4
Lesley E Degner
procedure and is committed to minimizing error. An example would be when the investigator has a single research assistant for whom extensive training can be offered, and when a direct test of the hypothesis is essential. It is also restricted to clinical situations in which subject fatigue is not an issue. This sorting procedure involves successive comparisons of all possible combinations of subsets of two of the five cards. The subject would make 10 paired comparisons [n(n-1)/2]. The se pairs can be identified beforehand and arranged to obtain the maximal distance between the maximal number of items.6 These pairs would be AB, BC, CD, DE, AC, BD, CE, AD, BE, and AE. The method ensures that the research subjects determine their rank ordering of preferences only after they have considered every possible combination of two cards. This approach minimizes measurement error, as evidenced by the results of a survey of 436 cancer patients in which 63% fell directly on the hypothesized psychological dimension when this method of administration was used.7
Hypothesized Psychosocial Dimension Ideal Point
PAY
1
Keep Control
PHYSIcyI
Share Control
Give Away Control
FIGURE 1. The control preferencesconstruct.
tion of the letters that represent the five cards. Three different research procedures have been developed and used to date. In each of these, the subject is asked to consider one particular decision in expressing their preferences, such as their initial surgical treatment for breast cancer. Data obtained through administration of the CPS are only meaningful insofar as one particular decision of reference has been clearly identified.
2. Random Order Presentation of the Cards by Hand.
1. Comparing Every Possible Subset of Two Cards by Hand.
The second procedure is useful when multiple data collectors with varying levels of education and ability might be employed, and when a direct test of the hypothesis is
The first procedure is for use where the investigator has good control over the testing
CoIlaborativeRole
Active Role
B.
Ipxefertomakethe final decision about my treatment after saciously considering imy doctor’s opinion.
I
fl
n
I
PassiveRole
I
Partictpating
less important. For example, the dimensionality of the construct in the population being tested might have been established in previous research. The five cards are placed in random order by shuffling them. The first two cards are placed in front of the subject who is asked to select the preferred card. The preferred card is placed on top of the nonpreferred card. Then the next card is removed from the deck and placed beside the new stack of two cards. The subject is asked to compare the new card with the most preferred card. If the subject still prefers the previous card over the new one, the previous card is flipped over and the new card is compared to the next one in the new stack. If the subject prefers the new card, it is placed between the two cards in the new stack; if the previous second card is preferred, the new one is placed last in the new stack. This process continues until the patient’s entire preference order is unfolded. The participant will make between five and 10 paired comparisons, depending on which card is drawn first, the order of the pack of cards, and the patient’s ideal point. Although it is easier to administer in the field, this approach leads to more measurement error. For example, in a survey of the public, 53% of subjects fell directly on the hypothesized dimension (compared to the 63% of patients described above).7
3. Fixed Order Presentation of the Cards by Hand. The five cards are placed in the following fixed order: BDCEA. The same procedure as described in the above random order presentation is used, beginning with the first two cards: B and D. Once the preferred card is selected, the next card (C) is compared to the preferred card and so on as described above. This procedure has been very useful in clinical populations because patients can locate their ideal point in general terms (active, collaborative, passive) in their paired comparisons. A training video has been developed to teach research staff this procedure.
m Decision
Making
5
Presentation Procedures in a Clinical Context The CPS is also useful as a clinical assessment tool. Two different approaches have been developed for application of the CPS in practice.
1. The “Pick 0ne”Approach Patients were interviewed privately before their first visit to the oncologist as part of a nursing intervention.8 They were asked to consider the five cards and select the one that fell closest to their preferred role in the treatment decisions to be discussed with their oncologist that day. Patients were able to read through the cards, which were laid out in the order ranging from most to least control, and select the one that best represented their preferred role in decision making. Then they were asked to talk about the extent to which the statement on the card actually represented their preference. This is important in the clinical context because an individual’s ideal point might fall between two items. Many patients were surprised that they were offered this choice, thinking that they were just going to have to accept their physician’s recommendation about treatment. However, seeing the range of alternatives allowed them to think about participation in decision making in a different way. The CPS has proven to be useful for eliciting input from the patients and helping clinicians gain insight into patient preferences.
2. Fixed Order Presentation of the Cards by Computer. The fixed order presentation approach has been adapted for use on a touch screen computer. After an introductory screen that explains the task, two cards at a time appear side by side on the screen in the fixed order described in the Research Applications section above. The subject just touches the screen over their preferred card and the next two cards automatically appear on the screen. At the end of the task, the program prints a diagram that shows the patient their pre-
6
Lesley E Degner
ferred level of involvement in decision making on the keep-share-give away control dimension (Fig 1) with an arrow indicates the line at their ideal point. This program is currently being evaluated in a clinical trial to determine whether women with breast cancer who are coached with this program and given the printout before their medical visit achieve a greater degree of involvement in treatment decision making than women who are not coached with the computer program.
Preferred Versus Actual Role in Decision Making Sutherland et al9 used the original statements of Strull, Lo, and CharleslO in studying 60 radiotherapy patients at the Princess Margaret Hospital in Toronto. Patients were asked to select their preferred role in decision making using the “pick one” procedure described above, but were also asked to indicate which was the role they actually assumed in making treatment decisions. This approach was used in a recent study of women with breast cancer.‘l Once the patient had completed the card sort and the result had been recorded, the five cards were laid out on a surface in the order the patient had provided. They were then asked to reconsider the cards, and indicate the one that best represented the role they had played with respect to their original surgical treatment. Patients were easily able to distinguish between their preferred and actual roles in treatment decision making.
Data Management and Analysis Comparison of respondents’ most preferred role with the role they believe they actually assumed in decision making provides an important index of how consumers believe different settings are accommodating their preferences. While chi-square analysis gives an indication of whether there has been significant incongruence between preferred and actual roles, displaying respondents’ preferred and actual roles allows more precise examination of where the incongruence has occurred, in what direction, and
how severe it was. Incongruence could occur in the direction of people being asked to make choices they prefer not to make, as well as in the direction of not getting the degree of involvement they want. The further the individual’s actual role is from their preferred role, the more serious the incongruence. This analysis approach was best shown in a recent survey of 1 ,012 women with breast cancer.” Only 42% of all women believed they had achieved their preferred role in decision making for their initial surgical treatment. Analysis of incongruences between preferred and actual roles revealed that only 21% of women who wanted the most active roles in decision making achieved them. However, there was a small group of women (14.9%) who believed they had been pushed to assume more decisional control than they wanted. This type of analysis is very useful in describing patients’ perspectives on their experiences around decisional control in our health care system. The potential to reduce discrepancies between preferred and actual roles in decision making also provides an opportunity for evaluation of specific nursing interventions by measuring a specific outcome.r*
Samples Studied The CPS has been used to determine distributions of preferences in a variety of samples (Table 1). A sample of 436 newly diagnosed cancer patients was compared with a sample of 482 members of the public who were asked to choose the role they would like to play in treatment decision making if they were diagnosed with cancer.7 While the majority of cancer patients wanted physicians to make treatment decisions on their behalf (59%), most members of the public (64%) thought they would want to make their own treatment decisions if they got cancer. Other projects have involved testing the scale in smaller samples of newly diagnosed prostate cancer patientsI and newly diagnosed breast cancer patients.14,15 A larger study of breast cancer patients at a variety of points in their
Partmpating
disease trajectory showed that 22% preferred an active role, 44% a collaborative role, and 34% a passive role in decision making.l’ Another study was undertaken in Britain with 150 newly diagnosed breast cancer patients and 200 women with benign breast problems.r6 The CPS has also been adapted to study preferences for decisional control in kidney dialysisI in vitro fertilization, l8 and childbirth.lg More recently, the CPS has been modified for use with parents of children with cancer,*O and met the scaling criterion for reliability (see below) in a sample of 58 parents.
The construct validity of the CPS was established through definition of the conin the context
of a grounded
theory,
and to date the scale has proven to be a useful tool in epidemiological surveys in TABLE
Makmg
7
which the goal is to establish estimates of the prevalence of different preferences and to determine the best predictors of these pre-
ferences.
However,
the usefulness
of the
measure in longitudinal as opposed to crosssectional prediction remains to be established. Current work is being conducted to determine whether women with breast cancer who desire more involvement in decision making at time of diagnosis and/or who achieve more involvement have less anxiety, depression, and post-decisional regret 3 years after diagnosis as suggested by Fallowfield’s long-term follow-up of breast cancer patients
in England.’ The issue of whether preferences change over time is also being ad-
Evidence for Validity and Reliability
struct
m Decision
dressed in this study. Davison and Degner12 have already shown that a nursing intervention to empower men with prostate cancer did allow men to assume more active roles in decision making as measured by the CPS
1.
SamplesStudied RolePreferences Reference
Sample
Degner & Sloan (1992)7 Newly diagnosedcancer patients (n = 436) Membersof the public (n = 482) Degneret al (1997)” Canadianwomenwith breastcancer at any point in diseasetrajectory (n = 1,012) Bilodeau& Degner Canadianwomennewly (1996)14 diagnosedwith breast cancer (n = 74) Beaver et al ( 1996)16 Britishwomen newly diagnosedwith breast cancer (n = 150) Hack et al ( 1994)15 Canadianwomennewly diagnosedwith breast cancer (n = 35) Davison& Degner Canadianmennewly diagnosedwith pros(1997)12 tate cancer Davisonet al ( 1995)13
Canadianmenwith prostatecancer (n = 57)
Age
Active Collaborative Passive (%) (%) @)
X = 59
12
29
59
X = 42
64
27
9
X = 58.25 (12.7)
22
44
34
X = 57.5 (11.6)
20
37
43
X = 54.8 (10.7)
20
28
52
Rangeof 32-83 years
23
57
20
X = 69.8 (5.0) in control group,
13
50
37
X = 66.0 (8.3) in experimental group x= 71
37
37
27
19
23
58
8
Lesley E Degner
TABLE 2. Reliability of the Control Preferences Scale in Cancer Patients Criterion 50% plus one subjects fall on hypothesized dimension ABCDE
Study
Sample
Degner & Sloan Newly diagnosed (1992)’ (mixed) Beaver et al ( 1996) I4 Newly diagnosed Degner et al ( 1997) l1 (breast cancer) Any point in disease trajectory (breast cancer)
and also resulted in lower anxiety levels for these men. In each of the major studies undertaken with the Control Preferences Scale in cancer populations, unfolding analysis has shown that the scale has met Coombs’ criterion of 50% plus 1 patients falling on the hypothesized dimension (Table 2). Similar results were obtained in members of the general public (56%). Given that there are 60 possible scales that could emerge from n = 5 stimuli and only one has met the scaling criterion in study after study, the CPS has shown reliability in cancer populations. However, similar testing would have to be undertaken should the scale be used to estimate preferences in different populations or about health care decisions other than those concerning medical treatment. As Table 2 shows, method of presentation does influence the degree of measurement error, with random order being less satisfactory than the other two methods in a research context. In Beaver et al’sI study of 200 women with benign breast disease in which random order presentation was used, only 49% of subjects fell on the ABCDE dimension, thereby failing to meet this reliability criterion.
Sample Percent on Size Dimension 436 150 1,012
66 58 63
Method of Presentation Every subset of two Random order Fixed order
Conclusion The CPS offers a new approach to eliciting consumer preferences about participating in health care decision making. The scale has been used by several hundred people, ranging from members of the public through highly stressed populations, and has proven to be easily and quickly administered. Even people with relatively low levels of education can successfully complete the paired comparisons. Scaling procedures have shown that people in a variety of populations do have systematic preferences about keeping, sharing, or giving away control over health care decisions. The CPS should prove to be a useful measure for both clinicians and researchers as they attempt to foster consumer involvement in a manner that leads to improved health care outcomes and better utilization of scarce resources.
Author’s Note: More detailed information about the Control Preferences Scale can be found in Degner, Sloan, & Venkatesh ( 1997) The control preferencesscale. Canadian Journal of Nursing Research, 29: 21-43.
References 1. Fallowfield LJ, Hall A, Maguire GP, et al: Psychosocial outcomes of different treatment policies in women with early stage breast cancer outside a clinical trial. Br Med J 301:575-580,199O 2. Greenfield S, Kaplan S, Ware JE: Expanding patient involvement in care: Effects on patient outcomes. Ann Internal Med 102:520-528, 1985 3. Morris J, Royle GT: Offering patients a choice of surgery for early breast cancer: A reduction in anxiety
and depression in patients and their husbands. Social Sci Med 26:583-585,1988 4. Degner LF, Beaton JI: Life-Death Decisions in Health Care. New York, NY, Hemisphere Publishing Corp, 1987 5. Coombs CH: A Theory of Data. Ann Arbor, Ml, Mathesis Press, 1976 6. Ross RT: Optimal orders in the method of paired comparisons, in G.M. Maranell GM (ed): Scaling: A
Participating
Sourcebook for Behavioural Scientists Chicago, IL, Aldine, 1974, pp 106-109 7. Degner LF, Sloan JA: Decision making during serious illness: What role do patients really want to play? J Clin Epidemiol45:941-950, 1992 8. Neufeld KR, Degner LF, Dick JAM: A nursing intervention strategy to foster patient involvement in treatment decisions. Oncol Nurs Forum 20:631-635, 1993 9. Sutherland HJ, Llewellyn-Thomas HA, Lockwood GA, et al: Cancer patients: Their desire for information and participation in treatment decisions. J R Sot Med 82:260-263, 1989 10. Strull WM, Lo B, Charles B: Do patients want to participate in medical decisions? JAMA 252:29902994,1984 11. Degner LF, Kristjanson L, Bowman D, et al: Decisional preferences and information needs in women with breast cancer. JAMA 277: 1485- 1492, 1997 12. Davison BJ, Degner LF: Empowerment of men newly diagnosed with prostate cancer. Cancer Nursing 20:187-196, 1997 13. Davison J, Degner LF, Morgan TB: Prostate cancer: Information needs and treatment decision making. Oncol Nurs Forum 22:1401-1408, 1995 14. Bilodeau BA, Degner LF: Profiles of information
m Decision
Making
9
needs in women with breast cancer: A pilot study. Oncol Nurs Forum 23:691-696, 1996 15. Hack TF, Degner LF, Dyck D: Preferences for decisional control and illness information among women with breast cancer: A quantitative and qualitative analysis. Sot Sci Med 39:279-289, 1994 16. Beaver K, Luker KA, Owens RG, et al: Treatment decision making in women newly diagnosed with breast cancer. Cancer Nurs 19:8-19, 1996 17. Kaprowy JA: A descriptive study to investigate end stage renal disease patients’ desire for information and preferences about roles in treatment decision making. Unpublished master’s thesis, University of Manitoba, Winnipeg, Manitoba, Canada, 1991 18. Thompson LA: Preference for participation in treatment decision making and informational needs of couples undergoing fertility investigation. Unpublished master’s thesis, University of Manitoba, Winnipeg, Manitoba, Canada, 1990 19. Gupton A: Trial of labour versus repeated cesarean section: Influences on women’s decision making. Unpublished doctoral dissertation, University of Manitoba, Winnipeg, Manitoba, Canada, 1994 20. Pyke-Grimm KA, Degner L, Small A, et al: Preferences for participation in treatment decision making and information needs of parents of children with cancer: A pilot study. J Pediatr Oncol Nurs 15: 108, 1998
ONCERN ABOUT the degree of control that people are currently able to assume in treatment decision making has intensified as a result of beginning empirical evidence that individuals who exercise such control have better health care outcomes.1-3 One of the greatest challenges facing investigators is how to measure the degree of control that consumers actually want in health care. This article describes the development of the Control Preferences Scale (CPS), a general measure of a unidimensional construct, consumer preferences about participating in treatment decision making. This measure has recently been adapted for use with parents of children with cancer.
C
The Construct The control preferences construct (Fig 1) emerged from a descriptive theory of how treatment decisions are actually made in health care.4 A central factor influencing decision making was who actually controlled the design of treatment. Four patterns of control were identified and described: provider-controlled, patient-controlled, familycontrolled, and jointly controlled decision making. This description provided the initial parameters for development of the control preferences construct. Description of the patterns of control led to the question: “Do patients have preferences about the degree of control they actually want to exercise in treatment decision making?” Participant observation in oncology clinics provided evidence for the existence of such preferences. Some patients refused to
From the Faculty of Nursing, University of Manitoba, Winnipeg, Manitoba. Address reprint requests to Lesley E Degner, PhD, RN; St Boniface Genera/ Hospital Research Centre, 357 Tache Ave, Winnipeg, Manitoba RZHZA6. o 1998 by Association of Pediatric Oncology Nurses. 1043-4542/98/l 503-0102$3.00/O
PhD, RN
become involved in selecting their own treatment, even when urged to do so by their physician. Others indicated a need to discuss the available options with their physician. Rarely, there were patients who had obviously prepared themselves beforehand, making it clear that they would be deciding what treatment they received. These observations led to the formulation of the control preferences construct, defined as “the degree of control an individual wants to assume when decisions are made about medical treatment.” Patients were hypothesized to have differing preferences about keeping, sharing, or giving away control over treatment decision making to their physician. That is, any particular patient would have an individual preference that would fall at a particular point along that psychological continuum.
The Measurement Model The hypothesis that patients have preferences about their level of involvement in treatment decision making guided selection of the measurement model, unfolding theory.5 Unfolding theory is based on the premise that an individual’s preference corresponds to an ideal point on a continuum. This ideal point can be derived by presenting successive paired comparisons of stimuli that fall along the continuum. In the case of the CPS, the stimuli are five cards that describe different roles in treatment decision making (Fig 2). The ideal point is represented by the order in which the subject places the cards, from the most to the least preferred.
Presentation Procedures in a Research Context Administration of the CPS involves subjects sorting a series of cards through successive paired comparisons. The result of these comparisons consists of an ordered permuta-
Journal ofpediatric Oncology Nursing, Vol 15, No 3, Suppl 1 (July), 1998: pp 3-9
3
4
Lesley E Degner
procedure and is committed to minimizing error. An example would be when the investigator has a single research assistant for whom extensive training can be offered, and when a direct test of the hypothesis is essential. It is also restricted to clinical situations in which subject fatigue is not an issue. This sorting procedure involves successive comparisons of all possible combinations of subsets of two of the five cards. The subject would make 10 paired comparisons [n(n-1)/2]. The se pairs can be identified beforehand and arranged to obtain the maximal distance between the maximal number of items.6 These pairs would be AB, BC, CD, DE, AC, BD, CE, AD, BE, and AE. The method ensures that the research subjects determine their rank ordering of preferences only after they have considered every possible combination of two cards. This approach minimizes measurement error, as evidenced by the results of a survey of 436 cancer patients in which 63% fell directly on the hypothesized psychological dimension when this method of administration was used.7
Hypothesized Psychosocial Dimension Ideal Point
PAY
1
Keep Control
PHYSIcyI
Share Control
Give Away Control
FIGURE 1. The control preferencesconstruct.
tion of the letters that represent the five cards. Three different research procedures have been developed and used to date. In each of these, the subject is asked to consider one particular decision in expressing their preferences, such as their initial surgical treatment for breast cancer. Data obtained through administration of the CPS are only meaningful insofar as one particular decision of reference has been clearly identified.
2. Random Order Presentation of the Cards by Hand.
1. Comparing Every Possible Subset of Two Cards by Hand.
The second procedure is useful when multiple data collectors with varying levels of education and ability might be employed, and when a direct test of the hypothesis is
The first procedure is for use where the investigator has good control over the testing
CoIlaborativeRole
Active Role
B.
Ipxefertomakethe final decision about my treatment after saciously considering imy doctor’s opinion.
I
fl
n
I
PassiveRole
I
Partictpating
less important. For example, the dimensionality of the construct in the population being tested might have been established in previous research. The five cards are placed in random order by shuffling them. The first two cards are placed in front of the subject who is asked to select the preferred card. The preferred card is placed on top of the nonpreferred card. Then the next card is removed from the deck and placed beside the new stack of two cards. The subject is asked to compare the new card with the most preferred card. If the subject still prefers the previous card over the new one, the previous card is flipped over and the new card is compared to the next one in the new stack. If the subject prefers the new card, it is placed between the two cards in the new stack; if the previous second card is preferred, the new one is placed last in the new stack. This process continues until the patient’s entire preference order is unfolded. The participant will make between five and 10 paired comparisons, depending on which card is drawn first, the order of the pack of cards, and the patient’s ideal point. Although it is easier to administer in the field, this approach leads to more measurement error. For example, in a survey of the public, 53% of subjects fell directly on the hypothesized dimension (compared to the 63% of patients described above).7
3. Fixed Order Presentation of the Cards by Hand. The five cards are placed in the following fixed order: BDCEA. The same procedure as described in the above random order presentation is used, beginning with the first two cards: B and D. Once the preferred card is selected, the next card (C) is compared to the preferred card and so on as described above. This procedure has been very useful in clinical populations because patients can locate their ideal point in general terms (active, collaborative, passive) in their paired comparisons. A training video has been developed to teach research staff this procedure.
m Decision
Making
5
Presentation Procedures in a Clinical Context The CPS is also useful as a clinical assessment tool. Two different approaches have been developed for application of the CPS in practice.
1. The “Pick 0ne”Approach Patients were interviewed privately before their first visit to the oncologist as part of a nursing intervention.8 They were asked to consider the five cards and select the one that fell closest to their preferred role in the treatment decisions to be discussed with their oncologist that day. Patients were able to read through the cards, which were laid out in the order ranging from most to least control, and select the one that best represented their preferred role in decision making. Then they were asked to talk about the extent to which the statement on the card actually represented their preference. This is important in the clinical context because an individual’s ideal point might fall between two items. Many patients were surprised that they were offered this choice, thinking that they were just going to have to accept their physician’s recommendation about treatment. However, seeing the range of alternatives allowed them to think about participation in decision making in a different way. The CPS has proven to be useful for eliciting input from the patients and helping clinicians gain insight into patient preferences.
2. Fixed Order Presentation of the Cards by Computer. The fixed order presentation approach has been adapted for use on a touch screen computer. After an introductory screen that explains the task, two cards at a time appear side by side on the screen in the fixed order described in the Research Applications section above. The subject just touches the screen over their preferred card and the next two cards automatically appear on the screen. At the end of the task, the program prints a diagram that shows the patient their pre-
6
Lesley E Degner
ferred level of involvement in decision making on the keep-share-give away control dimension (Fig 1) with an arrow indicates the line at their ideal point. This program is currently being evaluated in a clinical trial to determine whether women with breast cancer who are coached with this program and given the printout before their medical visit achieve a greater degree of involvement in treatment decision making than women who are not coached with the computer program.
Preferred Versus Actual Role in Decision Making Sutherland et al9 used the original statements of Strull, Lo, and CharleslO in studying 60 radiotherapy patients at the Princess Margaret Hospital in Toronto. Patients were asked to select their preferred role in decision making using the “pick one” procedure described above, but were also asked to indicate which was the role they actually assumed in making treatment decisions. This approach was used in a recent study of women with breast cancer.‘l Once the patient had completed the card sort and the result had been recorded, the five cards were laid out on a surface in the order the patient had provided. They were then asked to reconsider the cards, and indicate the one that best represented the role they had played with respect to their original surgical treatment. Patients were easily able to distinguish between their preferred and actual roles in treatment decision making.
Data Management and Analysis Comparison of respondents’ most preferred role with the role they believe they actually assumed in decision making provides an important index of how consumers believe different settings are accommodating their preferences. While chi-square analysis gives an indication of whether there has been significant incongruence between preferred and actual roles, displaying respondents’ preferred and actual roles allows more precise examination of where the incongruence has occurred, in what direction, and
how severe it was. Incongruence could occur in the direction of people being asked to make choices they prefer not to make, as well as in the direction of not getting the degree of involvement they want. The further the individual’s actual role is from their preferred role, the more serious the incongruence. This analysis approach was best shown in a recent survey of 1 ,012 women with breast cancer.” Only 42% of all women believed they had achieved their preferred role in decision making for their initial surgical treatment. Analysis of incongruences between preferred and actual roles revealed that only 21% of women who wanted the most active roles in decision making achieved them. However, there was a small group of women (14.9%) who believed they had been pushed to assume more decisional control than they wanted. This type of analysis is very useful in describing patients’ perspectives on their experiences around decisional control in our health care system. The potential to reduce discrepancies between preferred and actual roles in decision making also provides an opportunity for evaluation of specific nursing interventions by measuring a specific outcome.r*
Samples Studied The CPS has been used to determine distributions of preferences in a variety of samples (Table 1). A sample of 436 newly diagnosed cancer patients was compared with a sample of 482 members of the public who were asked to choose the role they would like to play in treatment decision making if they were diagnosed with cancer.7 While the majority of cancer patients wanted physicians to make treatment decisions on their behalf (59%), most members of the public (64%) thought they would want to make their own treatment decisions if they got cancer. Other projects have involved testing the scale in smaller samples of newly diagnosed prostate cancer patientsI and newly diagnosed breast cancer patients.14,15 A larger study of breast cancer patients at a variety of points in their
Partmpating
disease trajectory showed that 22% preferred an active role, 44% a collaborative role, and 34% a passive role in decision making.l’ Another study was undertaken in Britain with 150 newly diagnosed breast cancer patients and 200 women with benign breast problems.r6 The CPS has also been adapted to study preferences for decisional control in kidney dialysisI in vitro fertilization, l8 and childbirth.lg More recently, the CPS has been modified for use with parents of children with cancer,*O and met the scaling criterion for reliability (see below) in a sample of 58 parents.
The construct validity of the CPS was established through definition of the conin the context
of a grounded
theory,
and to date the scale has proven to be a useful tool in epidemiological surveys in TABLE
Makmg
7
which the goal is to establish estimates of the prevalence of different preferences and to determine the best predictors of these pre-
ferences.
However,
the usefulness
of the
measure in longitudinal as opposed to crosssectional prediction remains to be established. Current work is being conducted to determine whether women with breast cancer who desire more involvement in decision making at time of diagnosis and/or who achieve more involvement have less anxiety, depression, and post-decisional regret 3 years after diagnosis as suggested by Fallowfield’s long-term follow-up of breast cancer patients
in England.’ The issue of whether preferences change over time is also being ad-
Evidence for Validity and Reliability
struct
m Decision
dressed in this study. Davison and Degner12 have already shown that a nursing intervention to empower men with prostate cancer did allow men to assume more active roles in decision making as measured by the CPS
1.
SamplesStudied RolePreferences Reference
Sample
Degner & Sloan (1992)7 Newly diagnosedcancer patients (n = 436) Membersof the public (n = 482) Degneret al (1997)” Canadianwomenwith breastcancer at any point in diseasetrajectory (n = 1,012) Bilodeau& Degner Canadianwomennewly (1996)14 diagnosedwith breast cancer (n = 74) Beaver et al ( 1996)16 Britishwomen newly diagnosedwith breast cancer (n = 150) Hack et al ( 1994)15 Canadianwomennewly diagnosedwith breast cancer (n = 35) Davison& Degner Canadianmennewly diagnosedwith pros(1997)12 tate cancer Davisonet al ( 1995)13
Canadianmenwith prostatecancer (n = 57)
Age
Active Collaborative Passive (%) (%) @)
X = 59
12
29
59
X = 42
64
27
9
X = 58.25 (12.7)
22
44
34
X = 57.5 (11.6)
20
37
43
X = 54.8 (10.7)
20
28
52
Rangeof 32-83 years
23
57
20
X = 69.8 (5.0) in control group,
13
50
37
X = 66.0 (8.3) in experimental group x= 71
37
37
27
19
23
58
8
Lesley E Degner
TABLE 2. Reliability of the Control Preferences Scale in Cancer Patients Criterion 50% plus one subjects fall on hypothesized dimension ABCDE
Study
Sample
Degner & Sloan Newly diagnosed (1992)’ (mixed) Beaver et al ( 1996) I4 Newly diagnosed Degner et al ( 1997) l1 (breast cancer) Any point in disease trajectory (breast cancer)
and also resulted in lower anxiety levels for these men. In each of the major studies undertaken with the Control Preferences Scale in cancer populations, unfolding analysis has shown that the scale has met Coombs’ criterion of 50% plus 1 patients falling on the hypothesized dimension (Table 2). Similar results were obtained in members of the general public (56%). Given that there are 60 possible scales that could emerge from n = 5 stimuli and only one has met the scaling criterion in study after study, the CPS has shown reliability in cancer populations. However, similar testing would have to be undertaken should the scale be used to estimate preferences in different populations or about health care decisions other than those concerning medical treatment. As Table 2 shows, method of presentation does influence the degree of measurement error, with random order being less satisfactory than the other two methods in a research context. In Beaver et al’sI study of 200 women with benign breast disease in which random order presentation was used, only 49% of subjects fell on the ABCDE dimension, thereby failing to meet this reliability criterion.
Sample Percent on Size Dimension 436 150 1,012
66 58 63
Method of Presentation Every subset of two Random order Fixed order
Conclusion The CPS offers a new approach to eliciting consumer preferences about participating in health care decision making. The scale has been used by several hundred people, ranging from members of the public through highly stressed populations, and has proven to be easily and quickly administered. Even people with relatively low levels of education can successfully complete the paired comparisons. Scaling procedures have shown that people in a variety of populations do have systematic preferences about keeping, sharing, or giving away control over health care decisions. The CPS should prove to be a useful measure for both clinicians and researchers as they attempt to foster consumer involvement in a manner that leads to improved health care outcomes and better utilization of scarce resources.
Author’s Note: More detailed information about the Control Preferences Scale can be found in Degner, Sloan, & Venkatesh ( 1997) The control preferencesscale. Canadian Journal of Nursing Research, 29: 21-43.
References 1. Fallowfield LJ, Hall A, Maguire GP, et al: Psychosocial outcomes of different treatment policies in women with early stage breast cancer outside a clinical trial. Br Med J 301:575-580,199O 2. Greenfield S, Kaplan S, Ware JE: Expanding patient involvement in care: Effects on patient outcomes. Ann Internal Med 102:520-528, 1985 3. Morris J, Royle GT: Offering patients a choice of surgery for early breast cancer: A reduction in anxiety
and depression in patients and their husbands. Social Sci Med 26:583-585,1988 4. Degner LF, Beaton JI: Life-Death Decisions in Health Care. New York, NY, Hemisphere Publishing Corp, 1987 5. Coombs CH: A Theory of Data. Ann Arbor, Ml, Mathesis Press, 1976 6. Ross RT: Optimal orders in the method of paired comparisons, in G.M. Maranell GM (ed): Scaling: A
Participating
Sourcebook for Behavioural Scientists Chicago, IL, Aldine, 1974, pp 106-109 7. Degner LF, Sloan JA: Decision making during serious illness: What role do patients really want to play? J Clin Epidemiol45:941-950, 1992 8. Neufeld KR, Degner LF, Dick JAM: A nursing intervention strategy to foster patient involvement in treatment decisions. Oncol Nurs Forum 20:631-635, 1993 9. Sutherland HJ, Llewellyn-Thomas HA, Lockwood GA, et al: Cancer patients: Their desire for information and participation in treatment decisions. J R Sot Med 82:260-263, 1989 10. Strull WM, Lo B, Charles B: Do patients want to participate in medical decisions? JAMA 252:29902994,1984 11. Degner LF, Kristjanson L, Bowman D, et al: Decisional preferences and information needs in women with breast cancer. JAMA 277: 1485- 1492, 1997 12. Davison BJ, Degner LF: Empowerment of men newly diagnosed with prostate cancer. Cancer Nursing 20:187-196, 1997 13. Davison J, Degner LF, Morgan TB: Prostate cancer: Information needs and treatment decision making. Oncol Nurs Forum 22:1401-1408, 1995 14. Bilodeau BA, Degner LF: Profiles of information
m Decision
Making
9
needs in women with breast cancer: A pilot study. Oncol Nurs Forum 23:691-696, 1996 15. Hack TF, Degner LF, Dyck D: Preferences for decisional control and illness information among women with breast cancer: A quantitative and qualitative analysis. Sot Sci Med 39:279-289, 1994 16. Beaver K, Luker KA, Owens RG, et al: Treatment decision making in women newly diagnosed with breast cancer. Cancer Nurs 19:8-19, 1996 17. Kaprowy JA: A descriptive study to investigate end stage renal disease patients’ desire for information and preferences about roles in treatment decision making. Unpublished master’s thesis, University of Manitoba, Winnipeg, Manitoba, Canada, 1991 18. Thompson LA: Preference for participation in treatment decision making and informational needs of couples undergoing fertility investigation. Unpublished master’s thesis, University of Manitoba, Winnipeg, Manitoba, Canada, 1990 19. Gupton A: Trial of labour versus repeated cesarean section: Influences on women’s decision making. Unpublished doctoral dissertation, University of Manitoba, Winnipeg, Manitoba, Canada, 1994 20. Pyke-Grimm KA, Degner L, Small A, et al: Preferences for participation in treatment decision making and information needs of parents of children with cancer: A pilot study. J Pediatr Oncol Nurs 15: 108, 1998