* Candidate for the Lee B. Lusted Student Prize Competition
Purpose: Control of pandemic influenza by social distancing measures such as school closures is a major theme of pandemic planning. However, the extent to which these measures actually impact the progression of a pandemic has never been established. Two waves of pandemic influenza (pH1N1) were observed in North America in 2009. A unique opportunity to identify and quantify the causes of the large reduction in incidence between the two waves is presented by the Canadian province of Alberta, which conducted widespread, unrestricted virological testing continuously until the middle of the second wave in October.
Method: We examined the complete database of 35,510 specimens submitted for virological testing to the Alberta Provincial Laboratory for Public Health from 19 April 2009 to 2 January 2010, which included both waves of the pH1N1 pandemic. We compared the pH1N1 age-structured incidence pattern with weather time series and the pattern of closing and opening of schools. We used mathematical models to investigate the mechanisms that could have given rise to the observed incidence patterns.
Result: The data strongly indicate that the ending and restarting of school terms had a major impact in attenuating the first wave and sparking the second. Mathematical models suggest that school closure reduced transmission among schoolchildren by more than 50% (95% CI 52–100%), and that this was a key factor in interrupting transmission.
Conclusion: Unrestricted virological testing during an influenza pandemic has made it possible to provide unequivocal evidence that closing schools can have dramatic effects on transmission of pandemic influenza. School closure appears to be a potentially effective strategy for slowing the spread of pandemic influenza in countries with social contact networks similar to those in Canada.
Purpose: A frequently used approach for assessing the health and economic impact of vaccination programs is the static model where externalities resulting from interactions between members of a community are ignored. The objective of this study was to characterize the difference between the estimates of the incremental cost-effectiveness ratios (ICER) of vaccination programs under the static approach and the dynamic approach (where externalities are incorporated) for diseases where transmission occurs.
Method: We used a general SIRS (susceptible-infected-removed-susceptible) model featuring a vaccine with several properties to illustrate the differences between the two approaches using both analytic and numerical techniques. Two special cases of this model were studied analytically focusing mainly on the behavior of the models in the steady state. We then used several sets of realistic parameters values to numerically simulate the general model and trace the transient dynamics leading to the steady state. Finally, we conducted a probabilistic sensitivity analysis (PSA).
Result: All clinical and economic measures of vaccine effects differed across the static and dynamic models. Also, the steady-state ICER of the static model is always (barring some unrealistic scenarios) higher than that of the dynamic model. The difference between the ICERs widens with increases in vaccine cost, waning immunity, quality of life, and reproduction number (Figure 1). The ratio of indirect to direct effects of a vaccine that does not fully protect against infection is always higher than that of a vaccine that confers full protection. This ratio ranges from a small to a large value, implying that the indirect effects can be even greater than the direct effects. We found that the steady-state analysis fails to capture important dynamics that may have profound effects on the results. For example, the ICER under the dynamic model may rise and then fall as the analytic time horizon expands. In all the PSA scenarios, the ICER of the dynamic model is lower than the ICER of the static model and the number of influential parameters is always higher under the dynamic model.
Conclusion: Use of a static model to evaluate the impact of vaccines can result in inaccurate predictions, and may lead to wrong conclusions. We recommend use of a dynamic model when the vaccine is likely to have important effects on transmission.
Purpose: Stroke is a major cause of mortality and severe disability and accounts for considerable amounts of healthcare resources. There is evidence that stroke units with or without early supported discharge (ESD) are effective in treating patients when compared with conventional care. However, their comparative cost-utility is unknown. The aim of this study was to assess the cost-utility of care in ordinary stroke units compared with the care provided in stroke units with ESD and general medical wards.
Method: Analyses were done in NorCaD, a Markov-model based on Norwegian incidence data and treatment costs. The model was run on 70-year-old men with average risk for further cardiovascular diseases. The patients were followed until death or 100 years of age. Effectiveness of the strategies was based on meta-analyses of published randomized controlled trials identified by a systematic search. Quality of life data were extracted from published literature. The model calculated quality-adjusted life year (QALY) gained with different strategies and life time costs related to stroke. We also analysed males at 50 years of age and females at both 50 and 70 years of age. In addition, we performed probabilistic sensitivity analyses to get an impression on uncertainty surrounding our analyses.
Result: Care in ordinary stroke units provided 0.33 additional QALYs and reduced lifetime costs for the health care system with NOK 338,000 (USD 56,000) compared with care in general medical wards. Hence, stroke units are dominant relative to general medical wards. Stroke units with ESD resulted in a QALY gain of 0.17 and reduced lifetime costs (NOK 127,000; USD 21,000) compared with ordinary stroke units, and hence ESD is a dominant strategy. Because ESD dominates ordinary stroke units, and ordinary stroke units dominate general medical wards, ESD also dominates general medical wards. Probabilistic sensitivity analyses documented little uncertainty regarding that stroke units with ESD are the most cost-effective strategy compared to ordinary stroke units and general medical wards. The results showed also little sensitivity with gender and age variation.
Conclusion: Ordinary stroke unit care is cost-effective relative to conventional care (general medical ward). Moreover, stroke units with ESD are cost-effective compared to ordinary stroke units.
Purpose: The optimal therapeutic approach for low-risk clinically-localized prostate cancer (CaP) is unknown: over 50% of screen-detected men are overtreated and treatment is associated with significant side effects (SE). This analysis examines the cost-effectiveness of radical prostatectomy (RP), radiation therapy (IMRT), brachytherapy (BT), proton beam therapy (PBT) and active surveillance (AS) in these men.
Method: A state transition model was constructed and analyzed using Monte Carlo simulation. Men received treatment or AS and incurred SE for 1-2 y and costs until death of CaP/other cause. Men on AS could elect therapy or be treated at progression (both with IMRT). The base case used 65 yo men and included therapy and patient time costs. Transition probabilities and utilities were developed from literature review. Sensitivity analysis on key parameters was performed. Main outcomes were costs (2008US$) and quality-adjusted life-years (QALYs), both discounted at 3%/y, and incremental cost-effectiveness ratios (ICERs).
Result: AS was most effective, providing 8.58 QALYs at a cost of $30422. Compared to RP, AS provided an additional 9.1 mo of QALE at an added cost of $2074 (ICER $2729/QALY). Among initial therapies, BT was most effective and least expensive, providing an additional 3.5 mo of QALE at a cost savings of $2743 vs. RP. IMRT and PBT were more expensive than BT, RP, or AS.
Conclusion: In this model, AS is associated with higher QALE than initial therapy and carries a minimal additional cost relative to RP or BT. AS should be strongly considered in these patients.
Purpose: Current anticoagulation guidelines suggest that the length of anticoagulation (AC) for unprovoked venous thromboembolism (VTE) should be determined by an individual risk assessment, balancing bleeding risk due to AC with the risk of VTE recurrence. Among individuals who are heterozygous for the factor V Leiden (FVL) mutation, however, not only is VTE risk greater, but the risk for bleeding is lower, suggesting standard recommendations may not apply and longer term AC be considered.
Methods: We constructed a Markov model to compare lifetime anticoagulation vs. shorter durations in 20-year-old FVL patients with an unprovoked VTE. Risks of major, minor, and fatal bleeding with and without AC, VTE morbidity and mortality, and quality of life utilities were obtained from the literature. We used sensitivity analyses to determine model parameter values favoring lifelong AC in FVL patients. Outcomes are in quality adjusted life years (QALYs), discounted at 3%/yr.
Results: In general population groups (where VTE relative risk and odds ratio for AC-related bleeding are 1.0), the short-term AC strategy has 0.17 QALYs more than lifetime AC. In FVL patients, lifetime AC was favored if their VTE relative risk was > 1.1 or if their bleeding odds ratio was <0.85. A 2-way sensitivity analysis shows the effects of varying those parameters (Figure). Results were relatively insensitive to variation of other parameter values.
Conclusion: Lifelong anticoagulation may result in greater benefit than risk in individuals with FVL and previous idiopathic VTE; however, further research better documenting FVL-specific bleeding risk and recurrent VTE risk is needed.
Purpose: Conceptually simple, Monte Carlo calibration (i.e., random search) is frequently used in the development of disease history models for economic evaluation. We evaluate whether MC calibration determines at least approximately correct values of unknown inputs to a hypothetical model.
Methods: Hypothetical model: a simplified history model of pressure ulcers (i.e., bed sores) in individuals receiving home care. The Markov model includes 3 health states (i.e., ulcer stage 0-1, 2, 3-4) with four transition parameters: weekly incidence of developing a stage-2 ulcer (°0 ), healing rates of stage-2 (q1) and stage 3-4 ulcers (q3), and progression rate from stage 2 à 3-4 (q2). “True” values of °0 and q-q3 were a-priori estimated. Base case analysis: Given the incidence°0, the model was calibrated to observed stage-specific prevalence data to determine the calibration parameters q1-q3. Prevalence was generated from the model using Kolmogorov's forward equations: for observed prevalence, we used true values of °0 and q1- q3; and for projected prevalence, the true value of°0 and randomly generated values for q1-q3. Sensitivity analysis: MC calibration was evaluated with respect to: i) uncertain incidence°0, ii) multiple calibration targets (i.e., prevalence observed at multiple time points), iii) target misalignment (i.e., different timing between observed and projected prevalence), iv) goodness-of-fit assessment (i.e., Pearson's, likelihood-ratio fit-statistics), v) acceptance criterion for good-fit parameter sets, vi) prior ranges of q1-q3, vii) sampling methods (i.e., random or Latin-hypercube sampling), and viii) sample size (e.g., 1000 to 100,000 random parameter sets). Outcome measures: i) number of good-fit parameter sets from the MC calibration, ii) number of unbiased good-fit parameter sets (i.e., calibrated q1-q3 were within 95% confidence intervals of their true values), and iii) relative errors of individual good-fit parameters.
Results: The MC calibration yielded an ensemble of the good-fit parameter sets, representing post-calibration uncertainty. MC calibration performed well with accurate input data, multiple calibration targets, and perfect alignment. Otherwise, the number of biased good-fit parameter sets increased. MC calibration was robust with respect to variation in methods for goodness-of-fit assessment, acceptance criterion, varied ranges of calibration parameters, sampling methods, and sample size of the random parameter sets.
Conclusions: Our results provide evidence in support of recently proposed components for the standardized calibration reporting checklist, and suggest areas for further methodological development of model calibration.