* Candidate for the Lee B. Lusted Student Prize Competition
Purpose: To determine if particular values clarification exercises included in a decision aid had discernible impact on patients with early-stage prostate cancer.
Method: A randomized controlled trial comparing two versions of a computerized decision aid was conducted in four centres. Patients were centrally randomized (stratified by location) to either a decision aid that included information structured to assist decision making (Inf) or a decision aid containing the same information plus two types of values clarification exercises (ValEx): (a) a ranking exercise to help the patient identify attributes affecting his decision, and (b) a bar exercise to help him determine the attributes’ relative weighting. Patients went through the 1-1.5 hr Decision Aid (DA) between diagnosis and decision making. Telephone follow-up interviews were conducted just after patients made their decisions with their physician (Followup1), 3 months later (Followup2) and a mailing >1 year later (Followup3). Outcome measures included Decisional Conflict Scale (scale 13-65), Satisfaction with Preparation for Decision Making Scale (scale 8-40), and Regret (scale 5-25).
Result: A total of 156 patients participated, 75 Inf and 81 ValEx subjects. The groups not differ significantly in their distributions of age or of education but did in their marital-status distributions: fewer married patients were in Inf (72%) than in ValEx (86%), χ2=5.2, p=.02. The groups did not differ significantly on any outcome evaluated at the time of the Decision Aid. In both groups, decisional conflict decreased continuously from before [means 34.7(Inf) vs 34.1 (ValEx)] to immediately after using the DA [means 26.6 (Inf) vs 25.8 (ValEx), F=104.0, p=.00]. Between-group differences emerged with time. The ValEx group reported better Satisfaction with Preparation at Followup1 [mean 28.9 (Inf) vs 31.5 (ValEx), t=2.36, p=.02] and again at Followup3 [mean 28.8 (Inf) vs 31.7 (ValEx), t=2, p=.046]. Regret did not differ between the groups at Followup2 [means 7.7 (Inf) vs 7.2 (ValEx), t<1] but did at Followup3 [means 8.5 (Inf) vs 7.2 (ValEx), t=2, p=.047].
Conclusion: These data suggest that including the two values clarification exercises leads to better Satisfaction with Preparation for Decision Making and to less regret but that it is necessary to evaluate the outcomes only after adequate time has passed for the decision processes to have impact.
Purpose: We sought to determine if physicians’ personal moral intuitions were associated with their judgments about using cost-effectiveness data in clinical decisions as well as their judgments about cost-containment strategies.
Method: We surveyed 2000 practicing U.S. physicians regarding moral and social issues including their moral intuitions related to five constructs – harm, fairness, ingroup, authority, and purity. These constructs have been used in social psychology using a 6 point ordinal scale in a 30-item instrument with 5 subscales for each construct. We hypothesized that harm and fairness subscale ratings would be associated with favorable perceptions of using cost-effectiveness data and cost-containment strategies. We asked physicians to rate their degree of moral objection (none, moderate, or strong) to using cost-effectiveness data in clinical decisions. We then asked, on a 4-point scale, to what degree they agreed with limiting reimbursements for expensive drugs and procedures in order to expand coverage to uninsured patients (cost-containment). We used logistic regression to examine associations between harm and fairness subscale scores and judgments about cost-effectiveness and cost-containment (using dichotomized versions of these items).
Result: 1032 of 1895 physicians (54%) responded. In unadjusted analyses, harm ratings were significantly associated with moral objection to cost-effectiveness. For every 1-unit increase in harm subscale scores (0-5), there was a 20% greater chance of objecting to cost-effectiveness analysis in clinical practice (OR= 1.2 [1.0-1.4]). After adjusting for age, gender, region, and specialty, that association was no longer significant. Fairness scores were not associated with judgments about using cost-effectiveness data in clinical practice. Both harm and fairness were significantly associated with judgments about cost-containment. For every 1-unit increase in mean harm score, there was a 20% increased chance of agreeing with cost-containment (OR=1.2 [1.0-1.4]). Similarly, every 1-unit increase in mean fairness scores was associated with a 70% greater chance of agreeing with cost-containment. These associations were unchanged after adjusting for age, gender, region, and specialty. There was no association between ingroup, authority and purity and cost-effectiveness or cost-containment judgments.
Conclusion: Differences of opinion among U.S. physicians related to cost-effectiveness data and cost-containment measures may arise from differences in the relative importance they place on key moral intuitions. Efforts to enlist the support of physicians in cost-containment may need to account for this diversity of moral intuitions.
Purpose: When asked for their preference between death and colostomy, most people say that they prefer colostomy. However, when given the choice of two hypothetical treatments that differ only in that one has four percent chance of colostomy while the other has four percent additional chance of death, approximately 25% of people who say that they prefer colostomy actually opt for the additional chance of death. This study examined whether probability-sensitive preference weighting may help to explain why people make these types of treatment choices that are inconsistent with their stated preferences.
Method: 1656 participants in a demographically diverse online survey were randomly assigned to indicate their preference by answering either, “If you had to choose, would you rather die, or would you rather have a colostomy?” or, “If you had to choose, would you rather have a 4% chance of dying, or would you rather have a 4% chance of having a colostomy?” They were then asked to imagine that they had been diagnosed with colon cancer and were faced with a choice between two treatments, one with an uncomplicated cure rate of 80% and a 20% death rate, and another with an uncomplicated cure rate of 80%, a 16% death rate, and a 4% rate of colostomy.
Result: Consistent with our prior research, most people whose preferences were elicited with the first question stated that they preferred colostomy (80% of participants) to death (20%), but many then made a choice inconsistent with that preference (59% chose the treatment with higher chance of colostomy; 41% chose the treatment with higher chance of death). Compared to the first group, participants whose preferences were elicited with the 4% question preferred death (31%) over colostomy (69%) more often (Chi-squared = 24.31, p<.001) and their treatment choices were more concordant with their stated preferences (64% chose the treatment with higher chance of colostomy; 36% chose the treatment with higher chance of death, Chi-squared for concordance = 36.92, p<.001).
Conclusion: Our experiment suggests that probability-sensitive preference weighting may help explain why people’s medical treatment choices are sometimes at odds with their stated preferences. These findings also suggest that preference elicitation methods may not necessarily assume independence of probability levels and preference weights.
Purpose: There has been increasing demand that measures of uncertainty be presented alongside point estimates when reporting cost-effectiveness analyses. However, a number of methodological studies have argued that optimal strategies can be identified using expected value alone, and that the probability distribution of the cost-effectiveness metric is irrelevant to this choice. Under this reasoning, the distributional information is only useful for determining the expected value of future research, decisions about which are seen as independent from the strategy choice. I examine the conditions under which it is appropriate to choose optimal strategies based solely on the expected value of the cost-effectiveness metric, and issues that arise when these conditions do not exist.
Method: In reality, decision-makers face an ongoing sequence of decisions, with new information becoming available periodically, and each occasion allowing us to update prior beliefs and revise earlier decisions. A simple mathematical model is developed to demonstrate the consequences of making decisions using expected value alone versus using the joint distribution of incremental costs and outcomes. In this model the decision-maker chooses between two strategies at t0 based on probabilistic information about costs and outcomes. Consequences of the chosen strategy accrue until t1, when, with probability p, new information on costs and outcomes becomes available. The decision-maker can then revise their strategy choice, and consequences continue to accrue until t2. The decision-maker’s goal is to maximize total expected utility.
Result: I find an important condition for choosing strategies based solely on expected value is that the strategy set available at future decisions—and pay-offs—are independent of the current decision. If independence holds, each decision becomes a one-shot game, and decisions based on expected value will maximize overall outcomes. However, independence does not hold in two common situations – where changing strategy incurs transition costs, and where decision-makers are unwilling to consider cost-effective strategies that produce worse health outcomes. In these scenarios, I find that decisions made using the joint probability distribution of incremental costs and outcomes will achieve (weakly) greater total expected utility compared to using expected value alone, and that decisions about optimal strategy cannot be made independently from decisions about optimal research investment.
Conclusion: In two commonly encountered situations, making decisions based on expected value alone can lead to suboptimal outcomes.
Purpose: Seasonal influenza is a major public health concern and the first line of defense is the flu shot. Antigenic drifts and high rate of influenza transmission require annual updates in the flu shot composition. We propose a mathematical model to optimize the strain selection decisions for the annual flu shot. We analyze the trade-offs involved with various different policy issues.
Methods: We take the view of the Vaccines and Related Biological Products Advisory Committee of FDA; and optimize strain selections based on a production plan that is designed by the manufacturers exogenously. To select the strains for the flu shot, the committee meets for the first time at the end of February. In this initial meeting, recommendations are made for prevalent strains that have sufficient production yields. On the other hand, if the information is insufficient to select a strain for a category, the final decision of that category is deferred to the next meeting, which is held after four weeks. We propose a multi-stage stochastic mixed-integer program to determine the best flu shot composition, and the optimal time to select it. We consider all three flu strain categories (A/H1N1, A/H3N2, and B). We calibrate the cross-protective immunity among the candidate strains using a shape space model in which only the vaccine strain that has the smallest antigenic distance triggers immune response. The strain selection decisions are made to maximize the expected benefit of immune response minus the expected shortage cost under various different scenarios.
Results: Selecting the strain of each category independent from the others results in a loss of up to 14% of the optimal benefit. The cost of considering only the most prevalent strains might be as high as the 45% of the optimal benefit. Our model allows incorporating more than three strains in the flu shot; hence it can be used to assess the benefits of a tetravalent flu shot. We find that incorporating a fourth strain into the flu shot would potentially prevent over a million flu cases.
Conclusions: Integrating the composition and timing decisions is crucial to design the best flu shot. The uncertainties associated with the flu shot preparation campaign should analytically be incorporated into the strain selection decisions.
Purpose: Effective control of the global HIV epidemic requires judicious allocation of scarce resources between testing, treatment and prevention. Existing models used to inform the decision making process are too simplistic, while theoretical models are complex and impractical. Collaborating with UNAIDS, we designed a customizable, user-friendly model for decision making on HIV resource allocation that accounts for the effects of different levels of investment in interventions, effects of overlapping interventions, and epidemic dynamics.
Method: We created a spreadsheet-based model with a simple user interface. Default values characterizing the epidemic, key risk behaviors and available interventions are suggested for all parameters based on the literature. The user can customize these parameters. Epidemic dynamics are estimated by independent modules accounting for mother-to-child transmission (MTCT) and key risk groups: injection drug users (IDU), commercial sex workers (CSW) and men who have sex with men (MSM). Each module is designed as a dynamic compartmental model. Interventions considered include voluntary counseling and testing (VCT), HIV treatment (HAART), blood safety programs, prevention of MTCT, condom promotion, outreach for CSW and MSM, harm reduction (substitution therapy, needle exchange), and education. Production functions model program effect as a function of investment, and the effects of overlapping interventions are considered. The user inputs budget allocations that determine the scale of the interventions. The model estimates HIV prevalence, infections averted, QALYs gained and cost per QALY gained over a 20-year time horizon. The model also determines the optimal allocation of investment between interventions.
Result: We populated the model with data from Ukraine, Russia, Tanzania and Zambia. We calibrated our results against current practice in these countries and compared them against the optimized budget allocation. We showed that in Ukraine and Russia significant additional benefits can be obtained by increasing the budget allocation for harm reduction and treatment for IDUs. For heterosexually driven epidemics where concurrent partnerships occur, it may be optimal to increase investment in interventions that reduce the number of partners.
Conclusion: Our model provides a much needed bridge between research and practice, and can improve the HIV resource allocation process in settings around the world.