SEARCH

SEARCH BY CITATION

Keywords:

  • decision-making;
  • decision support techniques;
  • hormone replacement therapy;
  • menopause;
  • women education

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

Objectives  To understand the factors associated with a post-menopausal woman deciding to take hormone replacement therapy (HRT) after reviewing a decision aid (DA) and having a counselling visit with her physician as well as the factors associated with the act of taking HRT 2 months after the counselling interview.

Design  A secondary analysis of data collected for a randomized controlled trial evaluating two DAs.

Main outcome results  Although 28% of women were uncertain regarding their decision after the counselling interview, only 2.4% of women, at the assessment at 2 months, had not made a decision. The most significant factor associated with the decision to take HRT, after the physician visit, was the physician preference (OR: 62, 95% CI: 13.3, 289.7). Physician preference (OR: 78, 95% CI: 6.2, 975) remained the most significant factor for taking HRT 2 months after the counselling interview followed by low uncertainty about the decision (OR: 0.4, 95% CI: 0.2, 0.7).

Conclusion  Physician preference was the factor that was most associated with the woman's decision following counselling and 2 months later. Qualitative evaluation of the interview process involving the patient and physician would determine whether the patient and physician are reaching a shared decision or is the physician preference influencing the patient.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

The decision by a woman to take hormone replacement therapy (HRT) can be undertaken with considerable uncertainty. Physicians and health-care practitioners remain polarized on who should receive HRT due to the conflicting evidence regarding the benefits and risks of taking HRT.1,2 Practice guidelines from the American College of Physicians, the Society of Obstetricians and Gynaecologists of Canada, and the Osteoporosis Society of Canada recommend that women should participate in the decision to take HRT and physicians should take into account a women's personal probabilities of outcomes and personal values.3–5

O'Connor6 has developed a Decision Support Framework to guide counselling appropriate for such difficult decisions. The goal of the Decision Support Framework is to enable the patient and the practitioner to make a decision that is informed, consistent with personal values and acted upon. Decision support is provided to address the modifiable determinants of the decision or decision-making process that are suboptimal. For example, patients and practitioners may base their decisions on inadequate knowledge, unrealistic expectations, unclear values, and biased perceptions of others' opinions. Decision support tools provided to patients and practitioners may improve decision-making by providing information about options, outcomes, and probabilities, clarifying the patients' values, providing balanced perceptions of others and structuring the counselling to support shared decision-making.

The framework has been empirically validated by O'Connor et al.6–9 in that decision aids (DA) based on the framework do reduce uncertainty, improve knowledge and expectations, and clarify values. However, these improvements do not always result in changes in decisions, particularly for HRT. It would therefore be illuminating to examine which factors predict choices during and following the process of providing decision support.

This study evaluated the determinants of the decision to take HRT and the decision to follow through with the treatment decision after the counselling interview and 2 months later, respectively.

Data source

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

We conducted a secondary analysis of data collected during a randomized controlled trial that evaluated the effectiveness of two decision support interventions (DSI) for the use of HRT. The DSIs were a brief pamphlet of the American College of Physicians describing the options and outcomes in general terms and a detailed decision aid (audio-tape, booklet and worksheet) that included outcome probabilities tailored to clinical risk, values clarification, and guidance in the steps of decision-making. The data collected for the original study was obtained from post-menopausal women recruited from a random sample of practices of community-based family physicians in the Ottawa area. Physicians eligible for recruitment fulfilled the following criteria: they were >5 years since graduation from medical school and <60 years of age. Women eligible for recruitment into the study were 45–69 years of age, post-menopausal for at least 1 year, able to read English, had never used HRT, had no history of osteoporosis-associated fractures, and had no absolute contraindications to HRT.

The family physicians were randomized to one of the DSIs. Therefore, all women recruited by their family physician received the same DSI.

Variables

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

Data and variables, evaluated as factors influencing the decision to take HRT in the analysis, were guided by the Ottawa Decision Support Framework (see Table 1).6 Baseline data included the patient and physician demographic and clinical determinants of the decision, as well as the length of the relationship between the physician and the patient, and the patients' preferred role in decision-making. Two quality of life tools were used at baseline: the Short Form 1210 (SF-12) and the Menopause Quality of Life Scale11 (MENQoL). Each woman was assessed for her opinion regarding whether or not she would take HRT before the study was started. This variable was not considered as one of the factors for inclusion in the model as the woman's knowledge, values and beliefs were gathered explicitly according to the decision support framework before the interview and would still be operating post-consultation.

Table 1.  Independent variables
Time framePatient variablesPatient–physician variablesPhysician variables
BaselineAge, employed, education,  hysterectomy, smoker, years  to last menstrual period,  medical history assessing  risk for coronary heart  disease, osteoporosis,  and breast cancer, quality  of life scales i.e. MENQoL  and SF-12Length of relationship,  patient's desired role  in decision-makingAge, gender, CCFP certification,  year of graduation from medical  school after 1978, satisfaction  with autonomy in practice, usual  prescribing pattern of HRT
After counselling  interviewUncertainty, knowledge,  realistic expectations, valuesDecision satisfaction, patient's  actual role in decision-makingPhysician preference  for HRT
Composite baseline/ post-counselling variable Achievement of desired  role in decision-making 

Following review of the DSI (either the decision aid or pamphlet) and within 24 h of a counselling interview with their physician their ‘perception of the decision’ was measured. Based on the Decision Support Framework6 knowledge of HRT, expectations of benefits and risks and whether they were realistic, uncertainty measured using the decision conflict scale12 and personal values regarding the benefits and risks of HRT were assessed. The decision satisfaction inventory13 was used to measure satisfaction with the decision and the decision-making process. The patients' actual role in the decision-making process was assessed and a composite variable ‘achieved role’ was constructed based on a match between the patients' desired and actual role in decision-making. Afterwards, the preference of the physician was assessed by asking ‘if the decision to start HRT for this patient was up to them alone would they prefer the patient to start HRT?’ The preference of the physician was included for consideration in the model as we wanted to capture all the variables that were operating pre-consultation and the impact of the decision aid and the counselling interview on the effect of the decision of the woman to take HRT. The DA serves as a tool for the interview and the patient thinking on whether or not to take HRT.

Analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

Data from the first 126 patients and their associated physicians were used to develop two statistical models predicting the decision immediately after the counselling interview and whether the woman was taking HRT 2 months after the counselling interview. The decisions were classified as ‘yes’ or ‘no’, and any woman who was uncertain was excluded at the time point where the model was developed was from that particular model.

The statistical analysis was performed using SAS© (Ver. 7; SAS Institute, Inc., Cary, NC, USA). Univariate analysis was performed using the student's t-test for continuous variables and chi-square (χ2) for categorical variables. Generalized estimating equations (GEE) were used in the univariate analysis to adjust for physician, as the physician was the unit of randomization.14–16 Any variables achieving a significance of P < 0.10, adjusted or unadjusted for the physician, were considered for inclusion in the model. The physician was accounted for in the multivariate model development stage using GEE with a forward stepwise method for addition and removal of variables with a significance level of 0.05 using a hierarchical model building approach.17

The intervention was a design variable and therefore it was included in all multivariate models. Possible colinearity was identified based on clinical experience and the variable added first was retained in the model. First-order interactions were assessed for variables where clinical experience suggested a possible interaction. First-order interactions between the DSI and each variable included in the models were also assessed.18

The variables were reviewed for contribution to the predictive ability of each model using the Hosmer and Lemeshow goodness-of-fit statistic (HL).19 Models were tested for overfitting using the generally accepted rule that there must be at least 10 outcome events for each covariate and the heuristic shrinkage estimator of van Houwelingen and le Cessie (a shrinkage coefficient <0.85 suggests overfitting).17 The reproducibility of the final model was evaluated with the next 46 patients enrolled in the study using the HL test to assess the goodness-of-fit of the model.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

One hundred and twenty-six patients and 31 physicians were used to develop the statistical models. Sixteen physicians and their 74 patients were randomized to the decision aid group and 15 physicians and their 52 patients were randomized to the pamphlet group.

Table 2 summarizes the variables measured in the current study. The 126 women in the development model had an average age of 55.6 years, were well educated, and employed outside the home. There were slightly fewer female than male physicians and the Canadian College of Family Physicians (CCFP) certified the majority of the physicians.

Table 2.  Descriptive statistics for the variables assessed for the development and validation sets
Variable categoryVariable nameDevelopment set (n = 126)Validation set (n = 46)
  • DA-decision aid, SD-standard deviation, IQR-interquartile range, LMP-last menstrual period.

  • a

     Menopause Quality of Life Scale (MENQoL) range (0 symptoms least bothersome to 6 most bothersome).

  • b

     Short Form (SF) 12 Physical and Mental scales are scored using a norm-based method to have a mean of 50 and a SD of 10 in the general U.S. population with a higher number worse.

  • c

     Knowledge and Realistic Expectations are scored out of 100.

  • d

     Value Scale (range 0–10) higher number more valued.

  • e

     Decision Conflict Scale (DCS) (range 1–5) higher number more decisional conflict.

  • f

     Satisfaction with decision (range 0–15), satisfaction with process (range 0–50) higher number more satisfied.

DesignRandomized to DA74 (58.7%)17 (37.0%)
Baseline opinion‘I would consider taking HRT.’16 (12.7%)4 (8.7%)
Baseline patient characteristicsEmployed outside home72 (57.1%)37 (80.4%)
At least some post-secondary education86 (68.3%)34 (73.9%)
Current smoker14 (11.1%)2 (4.3%)
Hysterectomy35 (27.8%)5 (10.9%)
Coronary heart disease6 (4.8%)0.0 (0.0%)
Hypertension31 (24.6%)5 (10.9%)
Diabetes5 (4.0%)4 (8.7%)
High cholesterol32 (25.4%)7 (15.2%)
Fibroids23 (18.3%)6 (13.0%)
Low bone density22 (17.5%)5 (10.9%)
Family history of heart disease41 (32.5%)8 (17.4%)
Family history of breast cancer20 (15.9%)4 (8.7%)
Mean (SD), age (years)55.6 (6.3)54.0 (4.2)
Median (IQR) L MP (years)7.6 (2.2–13.6)4.5 (2.0–9.2)
MENQoL Vasomotor subscalea[Mean (SD)]2.8 (1.9)3.2 (1.8)
MENQoL Psychosocial subscale [Mean (SD)]3.0 (1.6)3.4 (1.8)
MENQoL Physical subscale [Mean (SD)]3.3 (1.3)3.4 (1.4)
MENQoL Sexual subscale [Mean (SD)]2.5 (1.9)2.8 (2.0)
SF 12 Physical subscaleb[Mean (SD)]26.3 (5.8)26.1 (5.8)
SF 12 Mental subscale [Mean (SD)]51.9 (7.0)52.7 (6.4)
Baseline physician characteristicsSatisfaction with autonomy100 (79.4%)40 (87.0%)
Year of graduation after 197864 (50.8%)27 (58.7%)
No. patients/week >12051 (40.5%)16 (34.7%)
Female gender59 (46.8%)24 (52.2%)
Canadian College of Family Physicians90 (71.4%)26 (56.5%)
After visit patient characteristicsKnowledgec[Mean (SD)]76.7 (17.5)79.1 (17.3)
Realistic expectations [Mean (SD)]32.2 (22.5)30.5 (21.7)
Value of heart diseased[Mean (SD)]7.5 (2.8)7.1 (2.6)
Value of osteoporosis [Mean (SD)]7.1 (2.9)7.4 (2.6)
Value of side-effects [Mean (SD)]7.9 (2.4)8.0 (2.1)
Value of breast cancer [Mean (SD)]8.2 (2.5)8.3 (2.2)
Decision Conflict Scalee[Mean (SD)]2.1 (0.6)2.5 (0.5)
DCS Uncertainty subscale [Mean (SD)]2.6 (1.0)3.1 (1.0)
DCS Knowledge subscale [Mean (SD)]1.9 (0.6)2.2 (0.6)
DCS Supported subscale [Mean (SD)]2.0 (0.6)2.4 (0.5)
DCS Values subscale [Mean (SD)]2.0 (0.6)2.3 (0.6)
DCS Satisfaction subscale [Mean (SD)]2.0 (0.7)2.5 (0.6)
Patient–physician characteristicsRelationship >5 years68 (54.0%)28 (65.1%)
Patient achieved decision role67 (53.2%)20 (46.5%)
Satisfaction with decisionf[Mean (SD)]11.3 (2.0)11.2 (1.6)
Satisfaction with processf[Mean (SD)]46.8 (7.5)45.9 (7.2)
After visit physician characteristicPhysician preference for HRT65 (51.6%)25 (54.3%)

The descriptive statistics for the women analysed in the validation set are also presented in Table 2.

The variables that were significantly related to the decision to take HRT in the univariate analysis are presented in Table 3. For the decision after the counselling interview a number of vari- ables became significant when an adjustment was made for the physician (Table 2): whether or not the physician held CCFP certification, and if the length of the patient–physician relationship was >5 years. These variables were variables that were measured at the physician level.

Table 3.  Characteristics associated with the decision to take hormone replacement therapy (HRT). Analysis Results for decision measured immediately and 2 months after the counselling visit (significant variables only)
Variable namePost-visit HRT usePost-visit (2 months) HRT use
Yes (34)No (57),Yes (29),No (94)
  • Legend: Same as Table 2.

  • *

     P < 0.1 only for analysis unadjusted for correlation.

  • **

     P < 0.1 only for analysis adjusted for correlation.

I would consider HRT (not in model)14 (41.2%) 0 (0.0%)12 (41.3%) 3 (3.2%)
Baseline patient characteristics
 Hysterectomy16 (47.1%)11 (19.3%)15 (51.7%)19 (20.2%)
 Fibroids10 (29.4%) 8 (14.0%) 
 FHx of heart disease15 (44.1%)14 (24.6%)17 (58.6%)24 (25.5%)
 FHx of breast cancer 2 (5.9%)16 (28.1%) 1 (3.4%)19 (20.2%)
 Age (years) [mean (SD)]53.7 (5.2)56.4 (6.1)*52.9 (4.5)56.4 (6.4)
 MENQoL Vasomotor subscale [Mean (SD)] 3.6 (2.2) 2.6 (1.7) 3.5 (2.3) 2.6 (1.7)
 MENQoL Psychosocial subscale [Mean (SD)] 3.7 (1.4) 2.8 (1.5) 4.0 (1.4) 2.7 (1.5)
 MENQoL Physical subscale [Mean (SD)] 3.8 (1.1) 3.0 (1.2) 4.0 (1.2) 3.1 (1.3)
 MENQoL Sexual subscale [Mean (SD)] 3.0 (2.1) 2.2 (1.7) 3.4 (2.3) 2.2 (1.7)
 SF 12 Physical subscale [Mean (SD)]28.7 (5.2)24.6 (5.4)28.7 (5.6)25.6 (5.7)
Baseline physician characteristics
 Female gender11 (32.4%)29 (50.9%)**10 (35.0%)46 (49.4%)**
 CCFP24 (70.6%)42 (73.7%)** 
After visit patient characteristics
 Value of heart disease [Mean (SD)] 8.9 (2.0) 6.6 (3.2)* 9.0 (2.2) 7.2 (3.0)
 Value of osteoporosis [Mean (SD)] 7.6 (2.7) 6.5 (2.9) 
 Value of side-effects [Mean (SD)]  6.9 (3.0) 8.0 (2.5)**
 Value of breast cancer [Mean (SD)] 7.4 (3.0) 8.9 (1.9) 7.0 (3.0) 8.6 (2.2)
 DCS Uncertainty subscale [Mean (SD)]  2.3 (0.8) 2.6 (1.0)**
 DCS Values subscale [Mean (SD)]  1.9 (0.5) 2.1 (0.6)
 DCS Satisfaction subscale [Mean (SD)] 1.8 (0.5) 1.8 (0.6)** 1.8 (0.5) 2.1 (0.7)
Patient–physician characteristics
 Relationship >5 years18 (52.9%)32 (56.4%)** 
 Physician preference for HRT30 (88.2%)15 (26.3%)27 (93.1%)36 (38.7%)
 Satisfaction with decision [Mean (SD)]11.3 (2.2)12.3 (1.6) 
 Satisfaction with process [Mean (SD)]45.2 (8.2)48.7 (7.2) 

The final models developed for the two decisions regarding the use of HRT are outlined in Table 4. After the counselling interview, the relative odds of taking HRT for a woman with physician preference was 62 (95% CI: 13.3, 290) than without physician preference. For each unit increase in the physical subscale of the MENQoL (more physical symptoms), the odds increased by a factor of 2.3 (95% CI: 1.3, 4.0) that patients would take HRT. For each unit increase in the vasomotor subscale of the MENQoL (more vasomotor symptoms), the odds increased by a factor of 1.4 (95% CI: 1.1, 1.7) that patients would take HRT.

Table 4.  Models for decisions of hormone replacement therapy
 Post-visit OR (95%) CIPost-visit (2 months) OR (95%) CI
  1. OR = odds ratio.

  2. CI = confidence interval.

Variables
 Intervention0.3 (0.1, 1.0)0.6 (0.1, 2.3)
 Physician preference62.0 (13.3, 290)78.0 (6.2, 975)
 MENQoL Physical subscale2.3 (1.3, 4.0) 
 MENQoL Vasomotor subscale1.4 (1.1, 1.7) 
 MENQoL Psychological subscale 2.0 (1.4, 3.0)
 DCS Uncertainty subscale 0.4 (0.2, 0.7)
 FHx of heart disease 4.0 (1.4, 11.8)
Model Statistics
 Hosmer and Lemeshow8.04 (P = 0.43)6.53 (P = 0.59)
 Shrinkage0.880.86
 External validation12.04 (P = 0.15)12.45 (P = 0.13)

These data were validated using the data set of 55 patients, nine patients were excluded as the data was missing for physician preference leaving 46 patients. Of these 46 patients, eight were uncertain, 15 decided to take HRT after the counselling interview and 23 did not. The HL was 8.40 (P = 0.20) indicating that this model was a good fit.

The relative odds of HRT 2 months after the counselling interview for women with physician preference was 78.0 (95% CI: 6.2, 975) than without physician preference. For each one unit increase in the uncertainty subscale of the decision conflict scale (measured on a scale of 5), a woman was less likely to take HRT with an odds ratio of 0.4 (95% CI: 0.2, 0.7). The odds of taking HRT were increased by a factor of 4.0 (95% CI: 1.4, 11.8) if the woman had a positive family history of heart disease. For each unit increase in the psychological subscale of the MENQoL (more physical symptoms), the odds increased by a factor of 2.0 (95% CI: 1.4, 3.0). The data were validated using the data set of 46 patients. After excluding five unsure women, 41 remained, of which 14 were taking HRT 2 months after the counselling interview. The HL was 12.452 (P = 0.13) indicating that this model was a good fit.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

This is the first study undertaken to identify the factors associated with the decision to take HRT in the context of the Decision Support Framework. The Decision Support Framework is a construct that is useful in clinical situations with value-laden decisions. Identification of the preference of the physician as a strong factor associated with the decision to take HRT is an important finding. Physician preference was assessed after the patient had reviewed a DA and the patient and physician had a counselling interview. The physician was asked if the decision to take HRT was up to them alone would they recommend HRT for a particular patient or not.

Several hypotheses may account for why physician preference was so influential. First, the physician may be confirming the patient's preference for HRT. Indeed, many of the women who started the visit with a baseline opinion that they would consider HRT ended up in favour of HRT. Secondly, the physician and the women may be reaching a decision together in a shared decision-making process. Lastly, the physician may be dominating the decision-making process and pushing the women towards the choice of HRT, although the physician's usual prescribing pattern of HRT was not significantly associated with the decision to take HRT.

Our study is the first study to explore the patient–physician relationship and the decision-making process between the women and her physician for the decision to take HRT and demonstrate the complexities of the interaction between the patient and the physician. Previous work has shown that the advice of the physician is an important factor in identifying women who are or were taking HRT or would consider taking HRT.20–24 Our study is the first to assess the influence of the physician prospectively and at the time when the actual decision is being made.

Future studies should explore the patient–physician relationship to understand better the amount of negotiation and patient-centred discussion, and its effect on the decision.

Women who were taking HRT 2 months after the counselling interview had less uncertainty in their choice, i.e. they had a lower decisional conflict on the uncertainty subscale of the decision conflict scale than those who did not choose HRT. This factor did not contribute to the model for the decision to take HRT immediately after the counselling interview. Immediately after the counselling interview, 35 women were excluded from model development, as they had not made a decision to take HRT. Two months after the counselling interview all but three women had made a decision. Therefore, the higher uncertainty in the group that declined HRT use 2 months after the interview could be a default decision. The decision conflict scale and the uncertainty subscale could be evaluated in future studies to examine if further information, counselling, or more support is required for these women in the decision-making process. We need to better understand how the physician can aid in the decision-making process for women who are still uncertain in their decision after the physician visit.

These results may not be generalizable for a number of reasons. The models developed were exploratory or hypothetical in nature due to the small sample size. Our patient population was predominantly of Caucasian race, married, and well educated.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References

The original study was supported by the Arthritis Society Research Grant no. 864 and the Medical Research Council Grant no. GR-13304.

AM O'Connor is a career scientist supported by Ontario Ministry of Health. ID Graham is a scholar funded by the Medical Research Council of Canada.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Data source
  6. Variables
  7. Analysis
  8. Results
  9. Discussion
  10. Acknowledgements
  11. References
  • 1
    Rabin DS, Cipparrone N, Linn ES, Moen M. Why menopausal women do not want to take hormone replacement therapy. Menopause, 1999; 6: 6167.
  • 2
    The North American Menopause Society. A decision tree for the use of estrogen replacement therapy or hormone replacement therapy in postmenopausal women: consensus opinion of The North American Menopause Society. Menopause, 2000; 7: 7686.
  • 3
    Anonymous. Guidelines for counseling postmenopausal women about preventive hormone therapy. American College of Physicians. Annals of Internal Medicine, 1992; 117: 10381041.
  • 4
    Society of Obstetricians and Gynecologists of Canada. Proceedings of the Canadian Menopause Conference. JSOGC, 1997; 16: 140.
  • 5
    Josse RG. Prevention and management of osteoporosis: consensus statements from the Scientific Advisory Board of the Osteoporosis Society of Canada. 3. Effects of ovarian hormone therapy on skeletal and extraskeletal tissues in women. Canadian Medical Association Journal, 1996; 155: 929934.
  • 6
    O'Connor AM, Tugwell P, Wells GA et al. A decision aid for women considering hormone therapy after menopause: decision support framework and evaluation. Patient Education and Counseling, 1998; 33: 267279.
  • 7
    O'Connor AM, Fiset V, DeGrasse C et al. Decision aids for patients considering options affecting cancer outcomes: evidence of efficacy and policy implications. Monographs of the National Cancer Institute, 1999; 25: 6780.
  • 8
    O'Connor AM, Tugwell P, Wells GA et al. Randomized trial of a portable, self-administered decision aid for postmenopausal women considering long-term preventive hormone therapy. Medical Decision Making, 1998; 18: 295303.
  • 9
    O'Connor AM, Tugwell P, Wells G. Testing a portable, self-administered decision aid for post-menopausal women considering long-term hormone replacement therapy to prevent osteoporosis and heart disease. Medical Decision Making, 1994; 14: 438.
  • 10
    Ware JE, Kosinski M, Keller SD. SF-12: How to Score the SF-12 Physical and Mental Summary Scales, 2nd edn. 1995. Boston, MA: The Health Institute, New England Medical Center.
  • 11
    Hilditch JR, Lewis J, Peter A et al. A menopause-specific quality of life questionnaire: development and psychometric properties. Maturitas, 1996; 24: 161175.
  • 12
    O'Connor AM. Validation of a decisional conflict scale. Medical Decision Making, 1995; 15: 2530.
  • 13
    Holmes-Rovner M, Kroll J, Schmitt N et al. Patient satisfaction with health care decisions: the satisfaction with decision scale. Medical Decision Making, 1996; 16: 5864.
  • 14
    Donner A, Klar N. Statistical considerations in the design and analysis of community intervention trials. Journal of Clinical Epidemiology, 1996; 49: 435439.
  • 15
    Donner A, Klar N. Methods for comparing event rates in intervention studies when the unit of allocation is a cluster. American Journal of Epidemiology, 1994; 140: 279289.
  • 16
    Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics, 1986; 42: 121130.
  • 17
    Harrell FEJ, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Statistics in Medicine, 1996; 15: 361387.
  • 18
    Gunst RF, Mason RL. Multicollinearity effects. In: OwenDB (ed.) Regression Analysis and its Application: a Data-Oriented Approach. New York, NY: Marcel Dekker Inc., 1980: 290314.
  • 19
    Hosmer DW, Lemeshow S. Assessing Fit. Applied Logistic Regression. New York: John Wiley & Sons, 1989: 135175.
  • 20
    Kadri AZ. Hormone replacement therapy – a survey of perimenopausal women in a community setting. British Journal of General Practice, 1991; 41: 109112.
  • 21
    Ferguson KJ, Hoegh C, Johnson S. Estrogen replacement therapy. A survey of women's knowledge and attitudes. Archives of Internal Medicine, 1989; 149: 133136.
  • 22
    Leveille SG, Lacroix AZ, Newton KM, Keenan NL. Older women and hormone replacement therapy: factors influencing late life initiation. Journal of the American Geriatrics Society, 1997; 45: 14961500.
  • 23
    Walsh JME, Brown JS, Rubin S, Kagawa M, Grady D. Postmenopausal hormone therapy: factors influencing women's decision making. Menopause, 1997; 4: 3945.
  • 24
    Blumberg G, Kaplan B, Rabinerson D, Goldman GA, Kitai E, Neri A. Women's attitudes towards menopause and hormone replacement therapy. International Journal of Gynecology and Obstetrics, 1996; 54: 271277.