* Candidate for the Lee B. Lusted Student Prize Competition
Purpose: We assessed how much, if anything, people would pay for laboratory testing to predict their future disease status, even if the test had no immediate consequences for treatment.
Method: We administered a questionnaire via an internet-based survey to a random sample of 2,223 adult U.S. respondents. Respondents were asked whether they would pay for a new blood test that would determine whether they would one day develop a particular disease. Each respondent answered questions about two different hypothetical scenarios, each of which specified: 1 of 4 randomly selected diseases (Alzheimer’s, arthritis, breast cancer, or prostate cancer); an ex ante risk of developing the disease (randomly designated 10% or 25%); and test accuracy (randomly designated perfect or “not perfectly accurate”). Willingness to pay (WTP) was elicited with a double-bounded, dichotomous-choice approach, which presented respondents with a binary bidding game, with respondents randomized to one of several starting bids.
Result: Most respondents were inclined to take the test across all diseases (70-88%, depending on the scenario). Inclination for testing was highest for prostate cancer (85-88% of respondents, depending on disease risk and test accuracy) and lowest for Alzheimer’s disease (71-74%). WTP was lower for Alzheimer’s and higher for prostate cancer compared with arthritis, and rose somewhat with the stated disease prevalence and for the perfect vs. imperfect test. Median WTP varied from $109 for the imperfect arthritis test to $263 for the perfect prostate cancer test. Regarding what they would do with a positive test, respondents most frequently stated that they would obtain a second medical opinion; seek medical care from a medical specialist; sign an advance directive; spend more time with family; and get their finances in order.
Conclusion: Most people preferred diagnostic tests even in the absence of direct treatment consequences -- and were willing to pay reasonably large amounts for the opportunity. People valued diagnostic information for a host of health and non-health related reasons. The results held across multiple diseases and across different magnitudes of disease prevalence, test accuracy, and adverse events. This “value of knowing” seems in part a desire for reassurance, and a desire for information even in light of possible bad news, and suggests that conventional cost-effectiveness analyses may underestimate the value of diagnostics.
Purpose: To identify trends in the use of direct and indirect utility elicitation techniques in published cost-effectiveness analyses.
Method: We analyzed data extracted from cost-effectiveness analyses (CEAs) included in the Tufts Medical Center Cost-effectiveness Analysis Registry, a database with detailed information on CEAs published in the peer-reviewed medical and economic literatures. Using studies from 1991-2008, we analyzed the proportion of utility weights elicited using direct vs. indirect methods, type of direct or indirect elicitation method, source of weights, age of the population affected by the disease, and disease category. Trends over time were assessed by comparing the 1991-93 and 2006-08 periods.
Result: For CEAs published from 1991 to 2008, 42% of utility weights were elicited using direct elicitation methods, 35% using indirect methods, and methods were unknown for 23%. For adult utility weights, the rating scale was the most commonly used direct method (25% for ‘91-‘93 and ‘06-‘08). For children, author/clinician judgment was most commonly reported as the “direct” method in earlier years (91%) compared with the standard gamble later (31%). For the elderly, the time trade-off was the most commonly reported direct method for both periods (33% and 18%, with 50% unreported in both periods). The most commonly used indirect method in the later period was the EQ-5D for adults (28%) and elderly (23%), and the HUI for children’s states (25%). Source of the utility weight also varied over time: for adult weights, the most common source changed from clinicians to patients for children, from authors to proxies; and for elderly, from proxies to patients. Specific characteristics of utility weights were missing for 6-60% of utility weights depending on the year, with fewer missing in later years. Few CEAs specific to children or elderly target populations were reported in earlier years.
Conclusion: Published guidelines recommend the use of generic instruments and trends over time show increasing adherence to these recommendations. In recent years, fewer studies have used authors or clinicians to estimate utility weights and more have used indirect methods. Nevertheless, the substantial proportion of CEAs using direct elicitation methods suggests there may be a continued role for direct elicitation. Future research should explore conditions under which a direct elicitation approach may be warranted.
Purpose: Non-steroidal anti-inflammatory drugs (NSAIDs) used to manage osteoarthritis (OA) can differ in their benefit and risk profiles. Shared-decision making requires that informed OA patients choose among these options, yet little is known about patient preferences for NSAIDs. We estimated the preferences for NSAIDs among a sample of OA patients from the UK and tested for differences in defined patient subgroups.
Methods: We applied a conjoint analysis across relevant benefits (ambulatory pain, resting pain, stiffness, and difficulty doing daily activities) and risks (bleeding ulcer, stroke, heart attack, and hypertension) of NSAIDS in a sample of OA patients aged >44 y recruited from a chronic disease panel. Respondents randomly received one of four blocks of 10 forced-choice, paired-comparison conjoint choice tasks drawn from an efficiency experimental design. Preferences were estimated via a mixed-effects logit and differences in standardized preference weights - ranging from least preferred (0) to most preferred (10) - were assessed by subgroups defined by age and current medication use.
Results: Of the 294 patients who met eligibiity criteria, 65% were female, 62% married. 56% were diagnosed with OA >4y and had an average age of 59y. 76% were on prescription OA medications, 49% had hypertension and 36% were using PPIs. Patients ranked ambulatory pain (6.32; 95% CI 5.0- 7.6) and difficulty doing daily activities (6.32; 95% CI 5.0- 7.6) as the most important benefit followed by resting pain (2.80, 95% CI 1.8-3.8), and stiffness (2.65; 95% CI .9-4.4). Incremental changes (3%) in the risk of heart attack or stroke were assessed as the most important risk (10.00; 95% CI 8.2-11.8; and 8.90; 95% CI 7.3-10.5, respectively). A 2.5% incremental change in one-year ulcer risk (3.61; 95% CI2.6-4.6) and the risk of hypertension (3.02; 95% CI 2.8-3.2) were valued less. Overall we identified no significant differences in preferences across subgroups, but patients with hypertension weighed CV risks higher and patients on PPIs placed less weight on the risk of ulcer.
Conclusions: Patients do have well defined preferences across NSAID-related benefits and risks. These preferences can be estimated and used to examine acceptable tradeoffs between benefits and risks. The observed differences in defined subgroups serve to validate our results, but need to be more rigorously examined in a larger sample of OA patients.
Purpose: To map the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) onto the EQ-5D in patients with knee osteoarthritis (OA).
Method: A consecutive sample of patients (n=258) completed the WOMAC and the EQ-5D. Regression models with the ordinary least squares (OLS) or the censored least absolute deviations (CLAD) as the estimator were used to establish the mapping function. Different presentations of the WOMAC scores were the explanatory variables. Goodness-of-fit criteria included mean absolute error (MAE) and root mean squared error (RMSE). An iterative random sampling procedure was used to account for variability in these goodness-of-fit diagnostics. Prediction precision was evaluated at both individual and group levels. At individual level, the prediction error was computed using the difference between observed and predicted EQ-5D scores for each of the 258 patients. At the group level, the prediction error was estimated by applying a non-parametric bootstrapping with replacement method. Specifically, various group sizes of patients (n=50, 100, 200, and 400) were randomly sampled. A patient was randomly chosen from the original dataset and his/her predicted EQ-5D score and prediction error were recorded. This process was repeated until the sample size of each group was reached. For each group, mean predicted EQ-5D scores and mean prediction error were calculated which formed one bootstrapping replicate. By repeating the above-mentioned process 5000 times, we generated a distribution for group mean predicted EQ-5D scores and corresponding group mean prediction errors for each of the groups. The 2.5th and 97.5th percentiles of the distribution were therefore used to estimate 95% CI for the prediction error.
Result: The model using OLS estimator and WOMAC domain scores as the explanatory variables was chosen as the preferred model: EQ-5D = 0.83414 - 0.00166×WOMAC pain score – 0.00092×WOMAC stiffness score – 0.00330 × WOMAC function score The prediction error at individual level exceeded the maximal tolerance value (i.e. the minimally important difference of the EQ-5D) in about 16% of the patients. At the group level, the width of the 95% CI of prediction errors varied from 0.0176 at a sample size of 400 to 0.0359 at a sample size of 100.
Conclusion: EQ-5D scores can be predicted using WOMAC domain scores with an acceptable precision at both individual and group levels in patients with knee OA.
Purpose: Mapping is a widely used method to convert scores from condition-specific instruments to utility values. There is limited knowledge of the extent to which use of different generic preference-based measures for mapping gives rise to differences in estimates of cost-effectiveness. We compared estimates of cost-effectiveness derived from two different generic preference-based measures for a proposed therapy in the management of knee osteoarthritis.
Method: An economic model was used to estimate cost-effectiveness of glucosamine sulphate therapy for knee osteoarthritis, using evidence from a systematic review of randomized controlled trials of clinical effectiveness, together with data from observational studies. The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) was the condition-specific instrument used in clinical effectiveness trials, and scores were converted to utility weights using published mapping algorithms for the Health Utilities Index Mark 3 (HUI3) and the EQ-5D measures. The mapped values were then used to estimate quality of life gains associated with therapy, and to predict cost-effectiveness relative to current care.
Result: Using the HUI3 mapping algorithm, trials showed that therapy was associated with an annual quality of life gain of 0.005. Taking a lifetime horizon, the incremental cost per quality adjusted life year (QALY) gain for adding glucosamine sulphate to current care was approximately £21,000. At a cost per QALY gain threshold of £20,000, the likelihood that glucosamine sulphate was more cost-effective than current care was 0.43, whilst at a threshold of £30,000, the probability rose to 0.73. Using the EQ-5D mapping algorithm however, therapy was associated with an annual quality of life loss of 0.006. In this situation, current care was predicted to be more cost-effective than therapy at standard cost-effectiveness thresholds.
Conclusion: Cost-effectiveness estimates were highly sensitive to the choice of generic-based measure used in mapping. This suggests that future studies should be encouraged to use generic-preference based instruments in the first instance rather than rely on mapping as a means to estimate cost-effectiveness.
Purpose: To test a brief consumer-led intervention consisting of three generic “Consumer Questions” (See http://www.abc.net.au/rn/healthreport/stories/2006/features/questions.htm) designed to
encourage doctors to discuss evidence related to treatment options with their patients. They are: 1. “What are my options?” 2. “What are the possible outcomes of those options?” 3. “How likely is each of the outcomes to occur?” Our study aimed to evaluate the effects on (1) communication about evidence related to treatment options, and (2) patient involvement.
Method: We utilized a randomized design,
set in family medicine practices in Sydney, Australia. Two professional actors were trained as Standardised Patients (SP) portraying two female roles with moderate depression. The roles were similar in clinical presentation but one of the SPs was trained to ask the Consumer Questions (the other did not ask the intervention questions but could ask other questions). The SPs made an unannounced visit to each of 18 participating doctors 5-24 weeks apart (order allocated randomly). Consultations were covertly audio-recorded. Doctors consented to the SP visits but did not know when they would occur, and were kept blind to the study purpose. Audio-recordings were transcribed verbatim, and analysed using (1) the ACEPP (Assessment of Communication about Evidence and Patient Preferences) scale, specifically designed for this study to measure the quality of communication about clinical evidence and (2) the OPTION scale (Elwyn 2003) to measure patient involvement. Two trained coders, kept blind to the study purpose, coded the transcripts.
Result: Scores in intervention consultations were significantly higher on both scales (Table).
|Intervention score (mean)||Control score (mean)||Difference (mean)||P value|
|ACEPP Scale (range 0-40)||21.4||16.6||4.8||<0.001|
|OPTION Scale (range 0-100)||38.32||25.27||13.05||<0.001|
The questions increased communication about evidence and patient involvement. The questions are a simple, inexpensive and sustainable consumer-led intervention which is widely applicable across health decisions, and does not require expensive development and updating processes. Potential benefits are increased evidence-based practice, improved safety and quality of care, as well as better decision quality.