Purpose: Past studies have found that even at relatively low levels of intake, long-term use of alcohol can adversely influence the prognosis for patients chronically infected with Hepatitis C Virus (HCV). Continued alcohol use with HCV can lead to cirrhosis and/or hepatic cancer, suggesting abstinence is best. Research has demonstrated differences among physicians in what they tell patients about whether various levels of alcohol consumption are safe. Gastroenterology (GI) specialists almost universally advise HCV patients to never drink alcohol. Despite a strong message from a GI specialist, some patients may process health information differently based upon socio-demographic, cultural, or personality characteristics. The purpose of this study is to examine the frequency and characteristics associated with HCV patients who have seen a GI specialist and who report their physician told them to never drink.

Methods: HCV patients (N=349) who saw a GI specialist at an urban Midwest teaching hospital were interviewed over the telephone. Patients answered questions about their alcohol use, healthcare behaviors, psychological traits, and coping styles. All patients were asked if they had been told by a physician to never drink. Standard clinical practice at this hospital is for all GI specialists to advise HCV patients to never drink alcohol. We assume that all patients in the sample were told by a physician to never drink. The dependent variable for the logistic regression was Did a physician tell you to never drink alcohol? (yes/no).

Results: Sixty-eight percent (95% CI, 63-73) of the patients reported a physician told them to never drink. Males (OR=.46, p=.01) and patients with lower incomes (OR=.19, p=.01) were more likely to report being told to never drink. Likewise, individuals scoring higher on a neuroticism scale (OR=1.11, p=.04) and those who take a planful approach to coping (OR=1.14, p=.02) were more likely to report being told to never drink. Future analyses will examine the conditional effects of gender and race.

Conclusions: Socio-demographic and psychological differences are associated with patients' reports about abstaining from alcohol. Physicians need to be sensitive to these differences to equip their patients with the knowledge needed to make the best decision about their alcohol use. To achieve this goal, we will need to examine various ways to present the same message to different types of patients.

Purpose: The purpose of this study was to assess associations between cognitive deficits and ability of patients with dementia (PWD) to state their health care preferences. Method: In this cross-sectional study, visual analog scale, time trade-off, and standard gamble health state valuation (HSV) techniques were used. Sixty PWD were asked to provide HSV for 2 practice states (dependence on glasses; blindness), their current health, and 3 hypothetical disease states (mild memory problems; mild to moderate dementia; severe dementia). The outcome variable, ability to perform HSV, was scored dichotomously. Patients received scores of 1 if they completed and provided logically consistent responses to both practice states and at least one additional HSV using any 1 of the 3 techniques. Those not able to perform at this level received scores of 0. Patients completed a neuropsychological test battery to assess 4 cognitive domains: executive function, language, attention and concentration, and verbal memory. Logistic regression was used to examine associations between cognitive domain scores and HSV performance. Results: The sample was 55% male and had a mean age of 74.9 (s.d.=9.4). Most participants (87%) had at least a high school education. Participants had mild to moderate dementia (Mini Mental State Exam mean = 22.8, s.d. =5.5). Cognitive subscale items were standardized before computing the domain scores. Means and standard deviations were as follows: executive function (-0.04, 1.38), language (0.12, 2.77), attention and concentration (0.03, 2.21) and verbal memory (0.13, 2.47). Higher language and verbal memory scores increased the odds of completing the HSV (OR=4.24, p=0.002; OR =1.79, p=0.04, respectively). There was a trend for higher scores on attention and concentration to be associated with a decreased odds of completing the HSV tasks (OR=0.44, p=0.059). The association between executive function scores and completion of the HSV tasks was not significant. The model c statistic was 0.90 (p<0.001). Conclusion: Dementia patients with relatively preserved language and verbal memory abilities are better able to complete the complex tasks involved in HSV. Clinically, these patients may still be able to state their health care preferences. The findings regarding attention and concentration are puzzling. Further study of the relationship between these cognitive functions and ability to perform HSV are needed.

Purpose: Patients often consult informal sources, including the internet, to aid them with treatment decisions about their disease. In light of the growing use of complimentary and alternative medicine practices (CAM) such as the use of herbs, supplements, acupuncture, etc, among patients with hepatitis C (HCV), we explored the topics surrounding patients' decisions to use CAM therapies for this disease as discussed on the internet.

Methods: Two sources of cross-sectional textual data were collected from the internet and analyzed for this study: (1) 307 electronic illness narratives posted on the topic of HCV from eight English language websites; and (2) 264 threaded discussions that included 534 discussants from four internet sites on the topic of HCV. The data were analyzed by first importing the text files into NVIVO (qualitative data management software) and coding any reference to CAM. Next, using the procedure of open-coding, the researchers independently reviewed the CAM text and generated a list of descriptive codes (topics). The researchers then met to compare lists and agree on a final version of the codes and their sub-categories.

Results: Nine descriptive codes, related to information sharing and decision making about the use of CAM for HCV, emerged from analysis of the qualitative data: 1) Reasons for Using CAM (liver health, dissatisfaction with traditional treatment, treatment of symptoms and side effects); 2) Research and Education on CAM (internet, books, seeking advice from others); 3) Types of CAM Used (herbs, supplements, behavioral, CAM provider based therapies); 4) Types of CAM Providers Seen (homeopaths, naturopaths, herbalists, acupuncturists, etc); 5) Therapeutic Regimens of CAM (combinations, dosages, frequency); 6) Perceived Results of CAM (positive, negative); 7) Cost Issues of CAM (price, insurance); 8) Re-evaluation of CAM Use (continue, discontinue); and 9) Sharing Experiences/Giving Advice about CAM Use.

Conclusions: The internet is a venue where a large amount of information about CAM is shared and a powerful tool for aiding patients in the medical decision making process. In order to assist patients with HCV in choosing therapies that are efficacious and safe, healthcare providers need to be aware of the wide range of topics that impact on the decision to use CAM therapies, and be able to discuss them with their patients.

Purpose. Currently, patients waiting for a cadaveric liver have little information regarding their position on the waiting list. This study aims to evaluate whether knowledge of specific location on the waiting list changes the optimal accept/reject decision for a given donor organ and shows the value of making the waiting list information transparent. Methods. We model the problem of accepting or rejecting a liver offer as a Markov decision process (MDP). The state of the system is composed of the triplets (h,k,l), where ‘h' represents the health status of the patient, ‘k' is the rank of the patient in the waiting list for the current liver offer, and ‘l' is the quality of the liver as determined by the characteristics of the donor such as age, gender, race, etc. The transition probabilities of the system are estimated using the natural history of liver disease as well as a previously validated national liver allocation simulation model. Optimal accept/reject decisions are found using the policy iteration algorithm. Results. The optimal accept/reject decisions take a ‘control-limit' form as shown in the figure. The horizontal axis shows the health of the patient as measured by MELD score. (Higher MELD scores represent sicker patients). The possible ranks of the patient are displayed on the vertical axis, where a zero rank is interpreted as the patient being off of the waiting list at the current stage. Each line represents a different organ quality where organ quality decreases from 1 to 14. For each organ quality, the area is split into two regions by the threshold line, where the optimal action is to ‘reject' the offer below the line (inclusive) and to ‘accept' the offer above the line. For the first 5 organ qualities (i.e., the top five quality organs) the optimal action is always to ‘accept' regardless of the health/rank combination (i.e., the lines associated for these organs coincide with the horizontal axis). If the organ quality further decreases, the rank threshold increases. In general, the reject region becomes larger as the quality of the organ decreases. Conclusion. The probabilistic model used shows that the optimal decision changes significantly if the waiting list information is known to the decision maker (patient/surgeon).

Purpose: Physicians caring for patients with Chronic Hepatitis C Virus (HCV) need to ascertain how much their patients are drinking. Patients with HCV may improve their prognosis by reducing or abstaining from alcohol use, even if they are not candidates for HCV drug treatments. A two phase investigation was employed to determine whether patient reports of quitting drinking are the same as total abstinence.

Methods: To be eligible for the study, patients had to be HCV positive, non-dependent drinkers. In Phase I, 42 subjects were recruited using a sampling grid to ensure a balance of race (White, African American, and Hispanic) and gender. Eligible patients who consented in Phase I were interviewed using a semi-structured format. Qualitative software was used to organize and analyze interview transcripts. In Phase II, 398 patients completed a telephone survey. Alcohol use was assessed using the Alcohol Use Disorders Identification Test (AUDIT) and the question “Have you quit drinking?” Both samples were drawn from the population of HCV patients at an urban teaching hospital.

Results: In Phase I, 20 patients said that they quit drinking or were a non-drinker. 12 of these 20 described in their interviews that they had continued to consume some alcohol. In Phase II, of those who reported quitting drinking 5% (95% CI, 2-7) had an AUDIT score greater than zero. A logistic regression of 64 covariates was used to compute the propensity score for the 398 Phase II subjects. The propensity score was used to create ten subclasses of patients matched by their propensity to have quit drinking. Mean scores on the AUDIT were compared within each of the ten subclasses. Patients who reported they quit drinking (N=272) had a weighted mean AUDIT score of .11 (95% CI, .05-.17) while patients who did not quit drinking (N=126) had a weighted mean AUDIT score of 2.97 (95% CI, 2.69-3.25).

Conclusions: Nearly all patients who reported having quit drinking abstained from alcohol completely in the past year. However for some HCV patients, quitting may imply a reduced pattern of drinking behavior rather than total abstinence. Doctors should discuss alcohol use with their patients using unambiguous questions that address the quantity and frequency of alcohol consumption, as well as the date of the last drink.

Purpose: Primary care practitioners have had difficulty implementing published guidelines for the diagnosis and management of asthma. Computerized decision support systems are tools that have been proven to improve compliance with published guidelines. This study's objective is to assess the accuracy of a computerized asthma decision support system in recognizing severe asthma using expert asthma clinicians as the reference standard.

Methods: This study is an accuracy assessment of 69 consecutive patients seen in an urban asthma subspecialty clinic between August 1 and November 30, 2004. Patients' asthma was classified as “severe” or “mild” both by the decision support system and expert asthma clinicians who were blinded to the assessment of the decision support system. Accuracy was quantified using percent accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and positive and negative likelihood ratios using the expert asthma clinicians as the reference standard.

Results: The decision support system had a 91% accuracy of recognizing severe asthma defined by expert asthma clinicians. The sensitivity of the decision support system for severe asthma was 96%, the specificity was 73%, the positive predictive value was 93%, the negative predictive value was 85%, the positive likelihood ratio was 3.6 (95% CI: 1.6, 8.4), and the negative likelihood ratio was 0.05 (95% CI: 0.012, 0.21).

Conclusions: The asthma decision support system was able to discern “mild” from “severe” asthma in a similar fashion to expert asthma clinicians. The effect of the decision support system on patient outcomes should be assessed.

Purpose: Although the potential for mammography screening to reduce breast cancer mortality risk is generally accepted for older women, some authors argue the benefits for younger women are less certain, because disease incidence and screening test accuracy are lower in younger women. Alternatively, the American Cancer Society Guidelines for Breast Cancer Screening state annual screening is more important for younger women since the disease is more aggressive in younger women. The primary objective of this research is to resolve these opposing viewpoints on selecting mammogram screening intervals for both life phases.

Method(s): We formulate a partially observed Markov decision process (POMDP) model exploring the relationship between mammography screening frequency and lifetime mortality risk. This approach incorporates the unobservable disease progression, the possibility of false results, and the possibility of unsuccessful treatment upon detection. We use this model to evaluate different screening policies and to construct tradeoff curves that plot “policy effort” versus mortality risk.

Results: We conduct a numerical study for a 25 year-old patient without cancer. By our analysis, we determine which policies are dominant, draw conclusions regarding the interaction between problem parameters (disease incidence, disease aggression, comorbidities, screening efficacy, screening start age, switch age, and stop age) and answer open questions concerning the value of dynamic (two-phase), versus routine (single-phase), screening policies.

Conclusions: A patient can achieve an intermediate breast cancer mortality risk using a dynamic policy rather than a “routine” policy. Our analysis provides the following preliminary insights: (i) annual screening beginning at 40 is efficient (ii) most switch ages for efficient two-phase policies are in the 50s; (iii) the most common screening start ages are 30 and 35, these policies also require the highest policy effort; (iv) two-phase policies prescribing more frequent screening early in life and policies prescribing more frequent screening later in life are both on the frontier; and (v) screening policies that stop screening early tend to result in higher mortality risk.

Purpose: To estimate the cost-effectiveness and mortality reduction associated with implementation of a rotavirus vaccination program in India.

Methods: We compared three strategies: no vaccination, 1- and 2-dose vaccination with Rotarix - a new live attenuated rotavirus vaccine. We assumed the doses would be administered through an existing immunization program with known coverage rates. Published cumulative incidences of infection were used to generate interval-specific probabilities of infection. These probabilities were revised according to vaccination status using efficacies from a clinical trial. Since no trials of Rotarix in Asia have been completed, efficacies were estimated using India-specific strain distributions and known strain-specific vaccine efficacies. Conditional probabilities for hospitalization and mortality were derived from population figures. Baseline vaccine cost used was $7/dose, the wholesale price recently paid by Brazil. We varied this parameter from $1 to $50. A recent trial evaluating treatment of diarrhea in Indian children provided most treatment cost data. Since the quality-of-life impact was over such a short period of time in relation to life expectancy of an infant, we used life-years saved (LYS) as our effectiveness measure. The conservative cost-effectiveness criterion used was $3,300 US (1 x India's estimated 2005 per capita GDP) per LYS.

Results: Baseline analysis yielded an incremental cost-effectiveness ratio (ICER) of $170.56 per LYS for moving from a strategy of no vaccination to a strategy of one dose of Rotarix, and an ICER of $350.77 per LYS for moving from one to two doses. This corresponded to estimated annual mortality reduction of 25,555 (27.0%) for one dose and 37,369 (39.5%) for two doses. Estimated India-specific vaccine efficacies used were 50.6% against any infection and 59.7% against severe infection (versus 72% and 85%, respectively, from a trial in Finland). 2-dose vaccination was preferred in one-way sensitivity analyses of all variables. Threshold vaccine costs per dose for 1- and 2-dose strategies were $133.13 and $65.25. For a 2-dose strategy, vaccination remained cost-effective at efficacies as low as 6.25% against any infection and 7.35% against severe infection.

Conclusions: Until there is a Rotavirus vaccine made from strains isolated in India, Rotarix appears to be a cost-effective and lifesaving intervention. This is despite the fact that differences in endemic strains would likely make Rotarix less effective in India than trial data might suggest.

Purpose: In cluster-randomized trials (CRTs), identifying clusters that are balanced on key features before assigning them to study groups is critical for fair and powerful comparisons. The key measure of balance in CRTs is the intra-cluster correlation coefficient (ICC), with better values being closer to zero, and larger values associated with significantly reduced power relative to a randomized controlled trial of the same size (where treatment allocation occurs at the patient level.) This report demonstrates the use of EMR data in executing a well-balanced design for a CRT of clinical decision support (CDS).

Methods: The Diabetes Improvement Group – Intervention Trial (DIG-IT) compares the impact of alternative approaches to CDS on the care and outcomes of 15,000 diabetic patients cared for by 220 primary care physicians (PCPs) in 24 diverse clinical practices of 2 health care systems. DIG-IT interventions apply to practices, but outcomes (diabetes-related parameters, rates of appropriate tests, treatments, immunizations, and resource use) occur at the patient level. We considered all feasible groupings of System A's 10 practice sites into two clusters and System B's 14 sites into three clusters. Before assigning clusters to study groups, we assembled practice-level data from each System's EMR to examine these groupings along an array of baseline clinical, demographic, and utilization characteristics, including trends in diabetes-related parameters. Blinded to site identifiers, DIG-IT investigators reached consensus about the groupings that most effectively balanced these pre-specified characteristics. Interventions were then randomly allocated to clusters in those groupings.

Results: System A's 10 practices were split into two clusters of 5 practices (pre-intervention, 2085 and 2281 patients, respectively); System B's 14 practices were split into three groups of 4, 6, and 4 practices (2069, 3115, and 3185 patients). ICCs for most pre-specified characteristics within each System were below 0.005, including A1c trend and mean, last systolic blood pressure, current smoking, "no show" rates, and percent with ED visits in the past year. The achieved balance was superior to that of potential groupings by demographics alone, by geography, or unstratified practice-level randomization, and it also was excellent for characteristics not explicitly accounted for in the design.

Conclusions: Rich data available through EMRs provide opportunities for designing state-of-the-art community-based CRTs, including health care delivery trials of clinical decision support.