To estimate the benefit of PSA-based screening for prostate cancer from the patient and societal perspectives.
A partially observable Markov decision process (POMDP) model was used to optimize PSA screening decisions. We considered age-specific prostate cancer incidence rates and the mortality rates from prostate cancer and competing causes. Our model trades off the potential benefit of early detection with the cost of screening and loss of patient quality of life due to screening and treatment. PSA testing and biopsy decisions are made based on the patient’s probability of having prostate cancer. Probabilities are inferred based on the patient’s complete PSA history using Bayesian updating.
The results of all PSA tests and biopsies done in Olmsted County, Minnesota from 1993 to 2005 (11,872 men and 50,589 PSA test results).
Perspective and Outcome Measures
Patients’ perspective: maximize expected quality-adjusted life years (QALYs); societal perspective: maximize the expected monetary value based on societal willingness to pay for QALYs and the cost of PSA testing, prostate biopsies and treatment.
From the patient perspective the optimal policy recommends stopping PSA testing and biopsy at age 76. From the societal perspective the stopping age is 71. The expected incremental benefit of optimal screening over the traditional guideline of annual PSA screening with threshold 4.0 ng/mL for biopsy is estimated to be 0.165 QALYs per person from the patient perspective, and 0.161 QALYs per person from the societal perspective. PSA screening based on traditional guidelines is found to be worse than no screening at all.
PSA testing done with traditional guidelines underperforms and therefore underestimates the potential benefit of screening. Optimal screening guidelines differ significantly depending on the perspective of the decision maker.
Online tools such as Adjuvant! provide tailored estimates of the possible outcomes of adjuvant therapy options available to breast cancer patients. The graphical format typically displays four outcomes simultaneously: survival, mortality due to cancer, other cause mortality, and incremental survival due to adjuvant treatment.
To test whether simpler formats that only present baseline and incremental survival would improve comprehension of the relevant risk statistics and/or affect treatment intentions.
Randomized experimental manipulation of risk graphics shown included in Internet-administered survey vignettes about adjuvant therapy decisions for breast cancer patients with ER+ tumors.
Demographically diverse, stratified random samples of women ages 40–74 recruited from an Internet research panel.
Participants were randomized to view either pictographs (icon arrays) that displayed all four possible outcomes or pictographs that showed only survival outcomes.
Comprehension of key statistics, task completion times, graph evaluation ratings, and perceived interest in adjuvant chemotherapy.
In the primary study (N=832), participants who viewed survival-only pictographs had better accuracy when reporting the total chance of survival with both chemotherapy and hormonal therapy (63% vs. 50%, p<0.001; higher graph evaluation ratings (M=7.98 vs. 7.67, p=0.04), and less interest in adding chemotherapy to hormonal therapy (43% vs. 50%, p=0.04; adjusted OR=0.68, p=0.008). A replication study (N=714) confirmed that participants who viewed survival-only graphs had higher graph evaluation ratings (M=8.06 vs. 7.72, p=0.04) and reduced interest in chemotherapy (OR=0.67, p=0.03).
Studies used general public samples; actual patients may process risk information differently.
Taking a “less is more” approach by omitting redundant mortality outcome statistics can be an effective method of risk communication and may be preferable when using visual formats such as pictographs.
decision aids; risk; patient education as topic; audiovisual aids
Women with localized breast cancer face difficult decisions about adjuvant therapy. Several decision aids are available to help women choose between treatment options. Decision aids are known to affect treatment choices and may therefore affect patient survival. The authors aimed to model the effects of the Adjuvant! decision aid on expected survival in women with early stage breast cancer.
Patients and Methods
Data were obtained from a randomized trial of Adjuvant! (n =395). To calculate the effects of the decision aid on survival, the authors used the Adjuvant! survival predictions as a surrogate endpoint. Data from each arm were entered separately into statistical models to estimate change in survival associated with receiving the Adjuvant! decision aid.
Most women (~85%) chose a treatment option that maximized predicted survival. The effects of the decision aid on outcome could not be modeled because a small number of women (n =12, 3%) chose treatment options associated with a large (5%–14%) loss in survival. These women—most typically estrogen receptor positive but refusing hormonal therapy—were equally divided between Adjuvant! and control groups and were not distinguished by medical or demographic factors.
Expected benefit from treatment is a key variable in understanding patient behavior. A small number of women refuse adjuvant treatment associated with large increases in predicted survival, even when they are explicitly informed about the degree of benefit they would forgo. Investigation of the effects of decision aids on cancer survival is unlikely to be fruitful due to power considerations.
Adjuvant!; breast cancer; decision aids; women’s health; oncology; outcomes research
Health State Preferences, Utilities, and Valuation; Health status indicators; Spine diseases; Quality of life; Economic evaluation; SPORT; Scale Validation; Cost-utility Analysis; Cost-effectiveness Analysis
Microsimulation models are important decision support tools for screening. However, their complexity creates a barrier, making it difficult to understand models and, as a result, limiting realization of their full potential. Therefore, it is important to develop documentation that clarifies assumptions. We demonstrate this problem and explore a solution for the natural history, using three independently developed colorectal cancer screening models.
We begin by projecting the cost-effectiveness of colonoscopy screening for the three microsimulation models. Next, we provide a conventional presentation of each of them, including information that would usually be published with a decision analysis. Finally, for the three models, we provide the simulated reduction in clinical cancer incidence following a one-time complete removal of adenomas and preclinical cancers. We denote this measure as maximum clinical incidence reduction (MCLIR).
There are considerable between-model differences in projected effectiveness. Conventional documentation describes model structure and associated parameter values. Given only this information, it is very difficult to compare models, largely because differences in structure make parameter values incomparable. In contrast, the MCLIR clearly shows the differences in assumptions on the key issue of the natural history: the dwell time of progressive preclinical disease, explaining between-model differences in projected effectiveness.
The simulated “maximum clinical incidence reduction” adds to the insight in dwell time, the critical characteristic of the natural history of disease, and how it differs between models. Inclusion of the MCLIR as a standard description would clarify the implications of assumptions for models applied to screening questions.
In the United States, African Americans are more likely to experience lower quality patient/provider communication and less shared decision making (SDM) than whites, which may be an important contributor to racial health disparities. Patient factors have not been fully explored as a potential contributor to communication disparities.
The authors analyzed cross-sectional data from a survey of 974 patients with diabetes seen at 34 community health centers (HC) in 17 midwestern and west-central states. They used ordinal and logistic regression models to investigate racial differences in patients’ preferences for SDM and in patients’ behaviors that may facilitate SDM (initiating discussions about diabetes care).
The response rate was 67%. In bivariate and multivariate analyses, race was not associated with patient preference for a shared role in the 3 measured SDM domains: agenda setting (odds ratio [OR]: 1.13 [0.86, 1.49]), information sharing (OR: 1.26 [0.97, 1.64]), or decision making (OR: 1.16 [0.85, 1.59]). African Americans were more likely to report initiating discussions with their physicians about 4 of 6 areas of diabetes care—blood pressure measurement (66% v. 52%, P < 0.001), foot examination (54% v. 47%, P = 0.04), eye examination (57% v. 46%, P = 0.002), and microalbumin testing (38% v. 29%, P = 0.01)—but not HbA1c testing (39% v. 43%, P = 0.31) or cholesterol testing (53% v. 51%, P = 0.52). In multivariate analysis, African Americans were still more likely to report initiating conversations about diabetes care (OR: 1.78 [1.10, 2.89]).
The authors found that African Americans in this study preferred shared decision making as much as whites and were more likely to report initiating more discussions with their doctors about their diabetes care. This research suggests that, among diabetes patients receiving care at community health centers, patient preference or patient behaviors may be an unlikely cause of racial differences in shared decision making.
randomized trial methodology; risk factor evaluation; population-based studies; scale development/validation
The decision to participate in a research intervention or to undergo medical treatment should be both informed and voluntary.
The aim of the present study was to develop an instrument to measure the perceived voluntariness of parents making decisions for their seriously ill children.
A total of 219 parents completed questionnaires within 10 days of making such a decision at a large, urban tertiary care hospital for children. Parents were presented with an experimental form of the Decision Making Control Instrument (DMCI), a measure of the perception of voluntariness. Data obtained from the 28-item form were analyzed using a combination of both exploratory and confirmatory factor analytic techniques.
The 28 items were reduced to nine items representing three oblique dimensions of Self-Control, Absence of Control, and Others’ Control. The hypothesis that the three-factor covariance structure of our model was consistent with that of the data was supported. Internal consistency for the scale as a whole was high (0.83); internal consistency for the subscales ranged from 0.68 to 0.87. DMCI scores were associated with measures of affect, trust, and decision self-efficacy, supporting the construct validity of the new instrument.
The DMCI is an important new tool that can be used to inform our understanding of the voluntariness of treatment and research decisions in medical settings.
voluntariness; decision making control; informed consent; ethics
Background. When planning to use a validated prediction model in new patients, adequate performance is not guaranteed. For example, changes in clinical practice over time or a different case mix than the original validation population may result in inaccurate risk predictions. Objective. To demonstrate how clinical information can direct updating a prediction model and development of a strategy for handling missing predictor values in clinical practice. Methods. A previously derived and validated prediction model for postoperative nausea and vomiting was updated using a data set of 1847 patients. The update consisted of 1) changing the definition of an existing predictor, 2) reestimating the regression coefficient of a predictor, and 3) adding a new predictor to the model. The updated model was then validated in a new series of 3822 patients. Furthermore, several imputation models were considered to handle real-time missing values, so that possible missing predictor values could be anticipated during actual model use. Results. Differences in clinical practice between our local population and the original derivation population guided the update strategy of the prediction model. The predictive accuracy of the updated model was better (c statistic, 0.68; calibration slope, 1.0) than the original model (c statistic, 0.62; calibration slope, 0.57). Inclusion of logistical variables in the imputation models, besides observed patient characteristics, contributed to a strategy to deal with missing predictor values at the time of risk calculation. Conclusions. Extensive knowledge of local, clinical processes provides crucial information to guide the process of adapting a prediction model to new clinical practices.
clinical prediction rules; methodology; decision rules; provider decision making; statistical methods
As the complexity of microsimulation models increases, however, concerns about model transparency are heightened.
We conducted model “experiments” to explore the impact of variations in “deep” model parameters using three colorectal cancer (CRC) models. All natural history models were calibrated to match observed data on adenoma prevalence and cancer incidence, but varied in their underlying specification of the adenoma-carcinoma process. We projected CRC incidence among individuals with an underlying adenoma or preclinical cancer vs. those without any underlying condition and examined the impact of removing adenomas. We calculated the percentage of simulated CRC cases arising from adenomas that developed within 10 or 20 years prior to cancer diagnosis, and estimated dwell time – defined as the time from the development of an adenoma to symptom-detected cancer in the absence of screening among individuals with a CRC diagnosis.
The 20-year CRC incidence among 55-year-old individuals with an adenoma or preclinical cancer was 7 to 75 times greater than in the condition-free group. The removal of all adenomas among the subgroup with an underlying adenoma or cancer resulted in a reduction of 30% to 89% in cumulative incidence. Among CRCs diagnosed at age 65, the proportion arising from adenomas formed within 10 years ranged between 4% and 67%. The mean dwell time varied from 10.6 years to 25.8 years.
Models that all match observed data on adenoma prevalence and cancer incidence can produce quite different dwell times and very different answers with respect to the effectiveness of interventions. When conducting applied analyses to inform policy, using multiple models provides a sensitivity analysis on key (unobserved) “deep” model parameters and can provide guidance about specific areas in need of additional research and validation.
Microsimulation models (MSMs) for health outcomes simulate individual event histories associated with key components of a disease process; these simulated life histories can be aggregated to estimate population-level effects of treatment on disease outcomes and the comparative effectiveness of treatments. Although MSMs are used to address a wide range of research questions, methodological improvements in MSM approaches have been slowed by the lack of communication among modelers. In addition, there are few resources to guide individuals who may wish to use MSM projections to inform decisions.
This article presents an overview of microsimulation modeling, focusing on the development and application of MSMs for health policy questions. We discuss MSM goals, overall components of MSMs, methods for selecting MSM parameters to reproduce observed or expected results (calibration), methods for MSM checking (validation), and issues related to reporting and interpreting MSM findings (sensitivity analyses, reporting of variability, and model transparency).
MSMs are increasingly being used to provide information to guide health policy decisions. This increased use brings with it the need both for better understanding of MSMs by policy researchers, and continued improvement in methods for developing and applying MSMs.
Discrete Event Simulation; Decision Analytic Models
Uncertainty is a pervasive and important problem that has attracted increasing attention in health care, given the growing emphasis on evidence-based medicine, shared decision making, and patient-centered care. However, our understanding of this problem is limited, due in part to the absence of a unified, coherent concept of uncertainty. There are multiple meanings and varieties of uncertainty in health care, which are not often distinguished or acknowledged although each may have unique effects or warrant different courses of action. The literature on uncertainty in health care is thus fragmented, and existing insights have been incompletely translated to clinical practice. In this paper we attempt to address this problem by synthesizing diverse theoretical and empirical literature from the fields of communication, decision science, engineering, health services research, and psychology, and developing a new integrative conceptual taxonomy of uncertainty. We propose a three-dimensional taxonomy that characterizes uncertainty in health care according to its fundamental sources, issues, and locus. We show how this new taxonomy facilitates an organized approach to the problem of uncertainty in health care by clarifying its nature and prognosis, and suggesting appropriate strategies for its analysis and management.
Antiviral coverage is defined by the proportion of the population that takes antiviral prophylaxis or treatment. High coverage of an antiviral drug has epidemiological and evolutionary repercussions. Antivirals select for drug resistance within the population, and individuals may experience adverse effects. To determine optimal antiviral coverage in the context of an influenza outbreak, we compared 2 perspectives: 1) the individual level (the Nash perspective), and 2) the population level (utilitarian perspective).
We developed an epidemiological game-theoretic model of an influenza pandemic. The data sources were published literature and a national survey. The target population was the US population. The time horizon was 6 months. The perspective was individuals and the population overall. The interventions were antiviral prophylaxis and treatment. The outcome measures were the optimal coverage of antivirals in an influenza pandemic.
At current antiviral pricing, the optimal Nash strategy is 0% coverage for prophylaxis and 30% coverage for treatment, whereas the optimal utilitarian strategy is 19% coverage for prophylaxis and 100% coverage for treatment. Subsidizing prophylaxis by $440 and treatment by $85 would bring the Nash and utilitarian strategies into alignment. For both prophylaxis and treatment, the optimal antiviral coverage decreases as pricing of antivirals increases. Our study does not incorporate the possibility of an effective vaccine and lacks probabilistic sensitivity analysis. Our survey also does not completely represent the US population. Because our model assumes a homogeneous population and homogeneous antiviral pricing, it does not incorporate heterogeneity of preference.
The optimal antiviral coverage from the population perspective and individual perspectives differs widely for both prophylaxis and treatment strategies. Optimal population and individual strategies for prophylaxis and treatment might be aligned through subsidization.
mathematical models; economic evaluation; decision analysis
Use of instrumental variables is gaining popularity as a method of controlling for confounding by indication in observational studies of treatments.
To illustrate how unmeasured instrument-level treatment substitution can distort effect size estimates using as an example an instrumental variable analysis of phototherapy for neonatal jaundice.
Retrospective cohort study.
Northern California Kaiser Permanente Hospitals.
The authors studied 20,731 newborns ≥2000 g and ≥35 weeks' gestation born 1995–2004 with a “qualifying” total serum bilirubin (TSB) level within 3 mg/dL of the 2004 American Academy of Pediatrics (AAP) phototherapy threshold who did not have a positive direct antiglobulin test.
The intervention was inpatient phototherapy within 8 hours of the qualifying TSB. The outcome was a TSB level exceeding the AAP exchange transfusion threshold <48 hours from the qualifying TSB. The instrumental variable was a measure of the frequency of phototherapy use at the newborn's birth hospital. The unmeasured substituted treatment was supplementation with infant formula, assessed by chart review in a sample from the same cohort.
In total, 128 infants (0.62%) exceeded the exchange transfusion threshold. Logistic and propensity analyses yielded crude odds ratios of ∼0.5 for phototherapy efficacy, decreasing to ∼0.2 with control for confounding by indication. Instrumental variable analyses suggested much greater phototherapy efficacy (e.g., odds ratios of 0.02–0.05). However, chart reviews revealed greater use of infant formula (which also lowers bilirubin levels) in hospitals that used more phototherapy (r = 0.56; P = 0.02), an association not present at the individual level (r = 0.13).
Instrumental variable analyses may provide biased estimates of treatment efficacy if there are cointer-ventions or confounders associated with treatment at the level of the instrument, even when these associations may not exist in individuals.
randomized trial methodology; risk factor evaluation; population-based studies; scale development
Effectively controlling the HIV epidemic will require efficient use of limited resources. Despite ambitious global goals for HIV prevention and treatment scale up, few comprehensive practical tools exist to inform such decisions.
We briefly summarize modeling approaches for resource allocation for epidemic control, and discuss the practical limitations of these models. We describe typical challenges of HIV resource allocation in practice and some of the tools used by decision makers. We identify the characteristics needed in a model that can effectively support planners in decision making about HIV prevention and treatment scale up.
An effective model to support HIV scale-up decisions will be flexible, with capability for parameter customization and incorporation of uncertainty. Such a model needs certain key technical features: it must capture epidemic effects; account for how intervention effectiveness depends on the target population and the level of scale up; capture benefit and cost differentials for packages of interventions versus single interventions, including both treatment and prevention interventions; incorporate key constraints on potential funding allocations; identify optimal or near-optimal solutions; and estimate the impact of HIV interventions on the health care system and the resulting resource needs. Additionally, an effective model needs a user-friendly design and structure, ease of calibration and validation, and accessibility to decision makers in all settings.
Resource allocation theory can make a significant contribution to decision making about HIV prevention and treatment scale up. What remains now is to develop models that can bridge the gap between theory and practice.
Inadequate adherence to highly active antiretroviral therapy (HAART) may lead to poor health outcomes and the development of HIV strains that are resistant to HAART. We developed a model to evaluate the cost effectiveness of counseling interventions to improve adherence to HAART among men who have sex with men (MSM).
We developed a dynamic compartmental model that incorporates HIV treatment, adherence to treatment, and infection transmission and progression. All data estimates were obtained from secondary sources. We evaluated a counseling intervention given prior to initiation of HAART and before all changes in drug regimens, combined with phone-in support while on HAART. We considered a moderate-prevalence and a high-prevalence population of MSM.
If the impact of HIV transmission is ignored, the counseling intervention has a cost-effectiveness ratio of $25,500 per QALY gained. When HIV transmission is included, the cost-effectiveness ratio is much lower: $7,400 and $8,700 per QALY gained in the moderate- and high-prevalence populations, respectively. When the intervention is twice as costly per counseling session and half as effective as we estimated (in terms of the number of individuals who become highly adherent, and who remain highly adherent), then the intervention costs $17,100 and $19,600 per QALY gained in the two populations, respectively.
Counseling to improve adherence to HAART increased length of life, modestly reduced HIV transmission, and cost substantially less than $50,000 per QALY gained over a wide range of assumptions, but did not reduce the proportion of drug-resistant strains. Such counseling provides only modest benefit as a tool for HIV prevention, but can provide significant benefit for individual patients at an affordable cost.
Cost Effectiveness; Adherence; HIV; Counseling; Computer Simulation
Colon cancer screening recommendations for patients aged 75 years and older should account for variation in older adults’ health states, life expectancies, and potential to benefit from screening.
To assess if resident physicians incorporate health state and life expectancy information when making recommendations about colon cancer screening for adults aged 75 years and older.
Resident physicians at a university internal medicine program completed a survey in which they made life expectancy estimates and screening recommendations for hypothetical 75- and 85-year-old women patients with good, fair, or poor health states. Outcomes of interest included accuracy of residents’ life expectancy estimates (compared with life table data), effect of health state and life expectancy on screening recommendations, and whether providing life table information affected the initial screening recommendation for the 85-year-old hypothetical patients.
Residents’ life expectancy estimates demonstrated moderate agreement with life table estimates. Their recommendations for colon cancer screening for the 75-year-old patient vignettes varied appropriately by health state and by their estimates of life expectancy. Receiving information about life expectancy from life tables affected residents’ recommendations for one of the three 85-year-old hypothetical patients, the woman in good health. Many resident physicians reported uncertainty about the potential to benefit from screening for each patient scenario.
Resident physicians appropriately used life expectancy and health state to make colon cancer screening recommendations for older adults. Residents reported substantial uncertainty with regard to the potential benefit of screening.
decision making; aged; aged 80 and older; mass screening; colorectal cancer
Patients should understand the risks and benefits of cancer screening in order to make informed screening decisions.
Evaluate the extent of informed decision making in patient-provider discussions for colorectal (CRC), breast (BrCa), and prostate (PCa) cancer screening.
National random-digit dial telephone survey.
English-speaking U.S. adults aged 50 and older who had discussed cancer screening with a health care provider within the previous two years.
Cancer screening survey modules that asked about sociodemographic characteristics, cancer knowledge, the importance of various sources of information, and self-reported cancer-screening decision-making processes.
Overall, 1,082 participants completed one or more of the three cancer modules. Although participants generally considered themselves well informed about screening tests, half or more could not correctly answer even one open-ended knowledge question for any given module. Participants consistently overestimated risks for being diagnosed with and dying from each cancer and overestimated the positive predictive values of PSA tests and mammography. Providers were the most highly rated information source, usually initiated screening discussions (64–84%), and often recommended screening (73–90%). However, participants reported providers elicited their screening preferences in only 31% (CRC women) to 57% (PCa) of discussions. While over 90% of the discussions addressed the pros of screening, only 19% (BrCa) to 30% (PCa) addressed the cons of screening.
Recall bias is possible because screening process reports were not independently validated.
Cancer screening decisions reported by patients who discussed screening with their health care providers consistently failed to meet criteria for being informed. Given the high ratings for provider information and frequent recommendations for screening, providers have important opportunities to ensure that informed decision-making occurs for cancer screening decisions.
prostatic neoplasms; breast neoplasms; colorectal neoplasms; early detection of cancer; decision making
To determine radiologists’ reactions to uncertainty when interpreting mammography and the extent to which radiologist uncertainty explains variability in interpretive performance.
The authors used a mailed survey to assess demographic and clinical characteristics of radiologists and reactions to uncertainty associated with practice. Responses were linked to radiologists’ actual interpretive performance data obtained from 3 regionally located mammography registries.
More than 180 radiologists were eligible to participate, and 139 consented for a response rate of 76.8%. Radiologist gender, more years interpreting, and higher volume were associated with lower uncertainty scores. Positive predictive value, recall rates, and specificity were more affected by reactions to uncertainty than sensitivity or negative predictive value; however, none of these relationships was statistically significant.
Certain practice factors, such as gender and years of interpretive experience, affect uncertainty scores. Radiologists’ reactions to uncertainty do not appear to affect interpretive performance.
medical decision making; physician uncertainty; medical malpractice; mammography interpretation
Some aspects of the natural history of metachronous colorectal cancer (MCRC), such as the rate of progression from adenomatous polyp to MCRC, are unknown. The objective of this study is to estimate a set of parameters revealing some of these unknown characteristics of MCRC.
The authors developed a computer simulation model that mimics the progression of MCRC for a 5-year period following the treatment of primary colorectal cancer (CRC). They obtained the inputs of the simulation model using longitudinal data for 284 CRC patients from the Mayo Clinic, Rochester.
Five-year MCRC incidence and all-cause mortality were 7.4% and 12.7% in the patient cohort, respectively. Statistical analysis showed that 5-year MCRC incidence was associated with gender (P = 0.05), whereas both all-cause and CRC-related mortalities were associated with age (P < 0.001 and P = 0.01). Estimated annual probabilities of progression from adenomatous polyp to MCRC and from MCRC to metastatic MCRC were 0.14 and 0.28, respectively. Annual probabilities of mortality after MCRC and metastatic MCRC treatments were estimated to be 0.06 and 0.26, respectively. The estimated annual probability of mortality due to undetected MCRC was 0.16.
The results imply that MCRC, especially in women, may be more common than suggested by previous studies. In addition, statistics derived from the clinical data and results of the simulation model indicate that gender and age affect the progression of MCRC.
colorectal cancer; discrete event simulation; simulation methods; operations research
To explore public attitudes toward the incorporation of cost-effectiveness analysis into clinical decisions.
The authors presented 781 jurors with a survey describing 1 of 6 clinical encounters in which a physician has to choose between cancer screening tests. They provided cost-effectiveness data for all tests, and in each scenario, the most effective test was more expensive. They instructed respondents to imagine that he or she was the physician in the scenario and asked them to choose which test to recommend and then explain their choice in an open-ended manner. The authors then qualitatively analyzed the responses by identifying themes and developed a coding scheme. Two authors separately coded the statements with high overall agreement (kappa = 0.76). Categories were not mutually exclusive.
Overall, 410 respondents (55%) chose the most expensive option, and 332 respondents (45%) choose a less expensive option. Explanatory comments were given by 82% respondents. Respondents who chose the most expensive test focused on the increased benefit (without directly acknowledging the additional cost) (39%), a general belief that life is more important than money (22%), the significance of cancer risk for the patient in the scenario (20%), the belief that the benefit of the test was worth the additional cost (8%), and personal anecdotes/preferences (6%). Of the respondents who chose the less expensive test, 40% indicated that they did not believe that the patient in the scenario was at significant risk for cancer, 13% indicated that they thought the less expensive test was adequate or not meaningfully different from the more expensive test, 12% thought the cost of the test was not worth the additional benefit, 9% indicated that the test was too expensive (without mention of additional benefit), and 7% responded that resources were limited.
Public response to cost-quality tradeoffs is mixed. Although some respondents justified their decision based on the cost-effectiveness information provided, many focused instead on specific features of the scenario or on general beliefs about whether cost should be incorporated into clinical decisions.
decision making; public opinion; cost-benefit analysis; health care rationing
To understand mammographers’ perception of individual women’s breast cancer risk.
Materials and Methods
Radiologists interpreting screening mammography examinations completed a mailed survey consisting of questions pertaining to demographic and clinical practice characteristics, as well as 2 vignettes describing different risk profiles of women. Respondents were asked to estimate the probability of a breast cancer diagnosis in the next 5 years for each vignette. Vignette responses were plotted against mean recall rates in actual clinical practice.
The survey was returned by 77% of eligible radiologists. Ninety-three percent of radiologists overestimated risk in the vignette involving a 70-year-old woman; 96% overestimated risk in the vignette involving a 41-year-old woman. Radiologists who more accurately estimated breast cancer risk were younger, worked full-time, were affiliated with an academic medical center, had fellowship training, had fewer than 10 years experience interpreting mammograms, and worked more than 40% of the time in breast imaging. However, only age was statistically significant. No association was found between radiologists’ risk estimate and their recall rate.
U.S. radiologists have a heightened perception of breast cancer risk.
perception; risk; pretest probability
Statistical evaluation of medical imaging tests used for diagnostic and prognostic purposes often employ receiver operating characteristic (ROC) curves. Two methods for ROC analysis are popular. The ordinal regression method is the standard approach used when evaluating tests with ordinal values. The direct ROC modeling method is a more recently developed approach that has been motivated by applications to tests with continuous values, such as biomarkers.
In this paper, we compare the methods in terms of model formulations, interpretations of estimated parameters, the ranges of scientific questions that can be addressed with them, their computational algorithms and the efficiencies with which they use data.
We show that a strong relationship exists between the methods by demonstrating that they fit the same models when only a single test is evaluated. The ordinal regression models are typically alternative parameterizations of the direct ROC models and vice-versa. The direct method has two major advantages over the ordinal regression method: (i) estimated parameters relate directly to ROC curves. This facilitates interpretations of covariate effects on ROC performance; and (ii) comparisons between tests can be done directly in this framework. Comparisons can be made while accommodating covariate effects and comparisons can be made even between tests that have values on different scales, such as between a continuous biomarker test and an ordinal valued imaging test. The ordinal regression method provides slightly more precise parameter estimates from data in our simulated data models.
While the ordinal regression method is slightly more efficient, the direct ROC modeling method has important advantages in regards to interpretation and it offers a framework to address a broader range of scientific questions including the facility to compare tests.
comparisons; covariates; diagnostic test; markers; ordinal regression; percentile values
Bivariate random effect models are currently one of the main methods recommended to synthesize diagnostic test accuracy studies. However, only the logit-transformation on sensitivity and specificity has been previously considered in the literature. In this paper, we consider a bivariate generalized linear mixed model to jointly model the sensitivities and specificities, and discuss the estimation of the summary receiver operating characteristic curve (ROC) and the area under the ROC curve (AUC). As the special cases of this model, we discuss the commonly used logit, probit and complementary log-log transformations. To evaluate the impact of misspecification of the link functions on the estimation, we present two case studies and a set of simulation studies. Our study suggests that point estimation of the median sensitivity and specificity, and AUC is relatively robust to the misspecification of the link functions. However, the misspecification of link functions has a noticeable impact on the standard error estimation and the 95% confidence interval coverage, which emphasizes the importance of choosing an appropriate link function to make statistical inference.
meta-analysis; bivariate random effect models; sensitivity; specificity; receiver operating characteristic curve; area under the ROC curve
To examine the effects of communicating uncertainty regarding individualized colorectal cancer risk estimates, and to identify factors that influence these effects.
Two web-based experiments were conducted, in which adults aged 40 years and older were provided with hypothetical individualized colorectal cancer risk estimates differing in the extent and representation of expressed uncertainty. The uncertainty consisted of imprecision (otherwise known as “ambiguity”) of the risk estimates, and was communicated using different representations of confidence intervals. Experiment 1 (n=240) tested the effects of ambiguity (confidence interval vs. point estimate) and representational format (textual vs. visual) on cancer risk perceptions and worry. Potential effect modifiers including personality type (optimism), numeracy, and the information’s perceived credibility were examined, along with the influence of communicating uncertainty on responses to comparative risk information. Experiment 2 (n=135) tested enhanced representations of ambiguity that incorporated supplemental textual and visual depictions.
Communicating uncertainty led to heightened cancer-related worry in participants, exemplifying the phenomenon of “ambiguity aversion.” This effect was moderated by representational format and dispositional optimism; textual (vs. visual) format and low (vs. high) optimism were associated with greater ambiguity aversion. However, when enhanced representations were used to communicate uncertainty, textual and visual formats showed similar effects. Both the communication of uncertainty and use of the visual format diminished the influence of comparative risk information on risk perceptions.
The communication of uncertainty regarding cancer risk estimates has complex effects, which include heightening cancer-related worry—consistent with ambiguity aversion—and diminishing the influence of comparative risk information on risk perceptions. These responses are influenced by representational format and personality type, and the influence of format appears to be modifiable and content-dependent.
When using state-transition Markov models to simulate risk of recurrent events over time, incorporating dependence on higher numbers of prior episodes can increase model complexity, yet failing to capture this event history may bias model outcomes. This analysis assessed the tradeoffs between model bias and complexity when evaluating risks of recurrent events in Markov models.
We developed a generic episode/relapse Markov cohort model, defining bias as the percentage change in events prevented with two hypothetical interventions (prevention and treatment) when incorporating 0–9 prior episodes in relapse risk, versus a model with 10 such episodes. We evaluated magnitude and sign of bias as a function of event and recovery risks, disease-specific mortality, and risk function.
Bias was positive in the base case for a prevention strategy, indicating that failing to fully incorporate dependence on event history overestimated the prevention’s predicted impact. For treatment, the bias was negative, indicating an underestimated benefit. Bias approached zero as number of tracked prior episodes increased, and average bias over 10 tracked episodes was greater with the exponential than linear functions of relapse risk and with treatment than prevention strategies. With linear and exponential risk functions, absolute bias reached 33% and 78%, respectively, in prevention, and 52% and 85% in treatment.
Failing to incorporate dependence on prior event history in subsequent relapse risk in Markov models can greatly impact model outcomes, overestimating the impact of prevention and treatment strategies by up to 85%, and underestimating impact in some treatment models by up to 20%. When at least four prior episodes are incorporated, bias does not exceed 26% in prevention or 11% in treatment.