Purpose: When using health care costs, it is common practice to apply costing data from one time point in one country to another time point in another country. This requires converting across currencies, health care systems, and time. The conventional recommendation is to first convert to the desired currency using purchasing power parity, then adjust for inflation using the local context to determine the rate of adjustment [1]. However, this recommendation was based on untested assumptions that may not consistently hold. This study aims to demonstrate the implications of using different methods for converting health care costs between countries and across time. [1] Drummond et al. Issues in the cross-national assessment of health technology. Int J Technol Assess Health Care. 1992;8(4):671-82.

Methods: Using a preliminary convenience sample of nine common drugs, we extracted costing data for 2006 and 2009 from the drug formularies for the Ontario Drug Benefit Program and the United Kingdom National Health Service. We examined differences in accuracy (defined as percent error between calculated and actual cost) for two different possible conversion routes: 1) convert currency, then inflate or 2) inflate, then convert currency, crossed with two different currency exchange mechanisms: a) purchasing power parity or b) exchange on currency markets. This yields four different possible conversion methods: 1a (recommended method as per [1]), 1b, 2a and 2b.

Results: Even in this very small sample, there were significant differences in accuracy for the four different conversion methods, whether calculating Ontario costs from NHS data (F(1,8)=14.16, p=.006) or NHS costs from Ontario data (F(1,8)=75.94, p<.001). Across drugs and methods, Ontario costs were underestimated by up to 47% and overestimated by up to 249%. UK costs were never underestimated and were overestimated by as much as 578%. Best accuracy for Ontario came from methods 2b (2 drugs) and 1b (7 drugs). Best accuracy for calculating UK costs was achieved with method 2b for all drugs. The recommended method (1a) yielded results that differed from the most accurate method for a given drug by up to 73%.

Conclusions: Differences in methods for cost conversion lead to vastly different results. Within this sample, the currently recommended method never yielded the most accurate results.

Purpose: Micro-costing is often conducted to determine incremental costs of an intervention for cost-effectiveness analysis, but may not be consistent with budgetary costs used for implementation. We describe these differences using a case study of implementation of rapid HIV testing and counseling in a substance abuse treatment program following a clinical trial.

Method: During the clinical trial, we used micro-costing methods to determine the cost of HIV testing in substance abuse treatment programs to conduct a cost-effectiveness analysis. Time and materials were from study records (including start and stop times for time conducting on-site testing and counseling) and site interviews; labor costs assume full capacity and were valued at local labor rates; and overhead was calculated from site financial records and applied as a percentage of labor costs. Costs include counselor and other labor, rapid HIV test and materials, supervision, quality control, and overhead. After the trial, one site implemented on-site rapid HIV testing with risk-reduction counseling in its detoxification program for 30 weeks. We compared projected costs in 2009 US dollars of implementation at this site based on micro-costing to budgetary costs reported by the site.

Result: The site administered 184 rapid HIV tests during the implementation period. Projected total costs for this period using micro-costing were $13,900 versus $20,300 budgetary costs. Labor costs based on micro-costing were $5,500 (245 hours) versus $16,700 (784 hours) budgeted for staff assigned to implementation. Overhead based on micro-costing was $5,500 versus $3,500 budgeted. Costs of tests and counseling supplies were estimated at $2,200 using micro-costing, whereas in the implementation the tests and supplies were provided from public health sources without cost to the site. Quality assurance costs using micro-costing were $700 but these costs were not separately budgeted in the implementation.

Conclusion: Cost estimates developed for cost-effectiveness analysis using micro-costing should not be indiscriminately applied when planning for implementation. Micro-costing may underestimate some costs (e.g. by assuming full capacity labor utilization) and overestimate others (e.g. by not considering donated materials and services). Micro-costing, however, may also identify cost categories not fully covered by implementation budgets (e.g. overhead and quality assurance).

Purpose: This study explores how health technology assessment (HTA) and research-funding agencies might effectively and efficiently use value-of-information (VOI) analysis to inform priorities for systematic reviews.

Methods: We reviewed 1) priority setting processes used by 13 international HTA and research-funding agencies, and 2) methods applied in 75 VOI studies from the literature. Following this, we developed an algorithm for deciding about the most effective and efficient approach to analyzing the value of systematic reviews in specific contexts.

Results: Our review revealed that the use of VOI and modeling is rarely applied in prioritizing systematic reviews. We identified conditions under which four alternative VOI approaches may be used for this purpose. The construction of “maximal” models of a broad disease process – often including multiple interventions to screen, diagnose and treat patients - can be worthwhile for prioritizing reviews when topics cluster in particular domains, such as diabetes, heart disease, and prostate cancer. VOI analyses commonly involve full modeling of a disease and its treatment but such exercises are generally too complex and too costly for prioritizing systematic reviews. Modeling can be minimized when existing comparative effectiveness studies provide appropriate data on comprehensive measures of health outcomes. Another approach is “conceptual VOI”, which uses information about the multiplicative elements of VOI, such as the burden of illness, uncertainty in treatment benefits, and the expected clinical use or implementation of research evidence, to provide informative bounds on the value of systematic reviews. Our algorithm describes a multi-stage process for deciding about the analysis of VOI in reviewing evidence. This process begins with clustering review topics and decisions about the use of maximal models, followed by conceptual VOI and then minimal modeling approaches. Although full models may aid in the planning and design of future research and HTA, we find limited conditions for the effective and efficient use of this traditional approach in prioritizing systematic reviews. 

Conclusion: An algorithmic approach that includes maximal modeling, full modeling, minimal modeling and conceptual VOI analysis may be useful in informing priorities for systematic reviews. In future work, we will illustrate the application of the algorithm for prioritizing review topics nominated to the Agency for Healthcare Research and Quality (AHRQ) Evidence-based Practice Centers.

Purpose: Highly publicized sudden cardiac deaths (SCD) in asymptomatic children and young adults have stimulated public interest in pre-athletics and school-based screening for asymptomatic cardiac disorders to avert these tragedies. However, the performance and trade-offs of the electrocardiogram (ECG) as a screening tool for the most common of these cardiac conditions is less understood.

Method: We systematically reviewed published literature on hypertrophic cardiomyopathy (HCM), long QT syndrome (LQTS), and Wolff-Parkinson-White syndrome (WPW), the three most common disorders associated with SCD and detectable by ECG. Using this information, we estimated (1) phenotypic prevalence, (2) sensitivity and specificity of ECG in detecting these disorders, (3) and predictive values using the illustrative point where sensitivity and specificity were equally weighted and the illustrative point where specificity was maximized.

Result: We identified and screened 6,954 abstracts, yielding 396 articles, and extracted data from 30. Summary prevalence estimates per 100,000 asymptomatic children were low at 45 (95% CI: 10, 79) for HCM; 7 (95% CI: 0, 14) for LQTS; and 136 (95% CI: 55, 218) for WPW. The areas under the receiver operating characteristic (ROC) curves for ECG were 0.91 for detecting HCM and 0.92 for LQTS.  When sensitivity and specificity were weighted equally, the positive predictive value (PPV) of detecting either HCM or LQTS using ECG was less than 1%, there were many false positives per case detected (399 for HCM and 2,323 for LQTS), and the false negative rate was 15% for HCM and LQTS. However, when specificity was maximized, the PPV increased to 2% for HCM and 1% for LQTS, the false positives per case detected declined (57 for HCM and 135 for LQTS), as did the false negative rate (<1% for HCM and LQTS). Regardless of sensitivity and specificity cut-point, the negative predictive value (NPV) was near 100% and the false reassurance rate was low (<45 per 100,000 screened) for HCM and LQTS. 

Conclusion: Because HCM, LQTS, and WPW have very low prevalence rates, population screening with ECG would yield substantial false positives. Guidelines regarding ECG screening will need to balance trade-offs between identification and treatment of affected individuals against the additional costs and risks associated with post-screening cardiac evaluations to rule out these disorders as well as potential overdiagnosis and overtreatment of asymptomatic individuals.

Purpose: To obtain population-based estimates of direct healthcare costs for prostate cancer (PC) from diagnosis to death for health states in a state-transition model.

Method: PC patients, diagnosed in 1993, 1994, 1997, 1998, 2001, and 2002, and residing in three regions of Ontario, Canada, were selected from the Ontario Cancer Registry. We retrieved pathology reports to identify patient name, referring physician, and tumour information. With consent from referring physicians, we contacted patients and family of dead patients for consent to review charts. We visited physicians’ clinics and hospitals and reviewed charts to obtain data describing PC diagnosis, treatment, and outcome. We developed clinical criteria to allocate each patient’s observation time to 11 PC-specific Markov health states, including active surveillance, treatments, follow-up, recurrence, metastases, and death. We linked these data to health care administrative databases to calculate healthcare resource use and costs per health state, using previously developed costing methods. Mixed model multivariable regression determined predictors of costs. To assess model validity, we compared predicted costs estimated from the model with actual costs using the root mean square error and mean average error.

Result: The final sample numbered 829 patients (mean age = 67 years). Over 50% had T2 to T4 disease, and 5% were metastatic at diagnosis. The most costly primary treatment was radical prostatectomy ($4,702 per 100 days). The least costly health state was post-prostatectomy ($731 per 100 days). Costs before death and for hormone-refractory metastatic disease were high at $11,008 and $6,324 per 100 days, respectively. Costs increased with age (p<0.001), comorbidity (p<0.001), and advanced PC at diagnosis (p<0.05). Radical prostatectomy, metastatic disease, and final (pre-death) health states were significantly more costly than active surveillance (p<0.05), while post-prostatectomy and post-radiation therapy states cost significantly less (p<0.0001). The validity of the model was assessed; the root mean square error was $4,206 and mean average error was $1,873, relatively small compared with observed mean and median costs per 100 days, $4,344 and $2,338, respectively.

Conclusion:   Combining chart reviews and administrative data is feasible to estimate mean adjusted costs and the effects of covariates on costs for state-transition models. However, this approach is very costly and time consuming. Administrative data alone may be sufficient for applications that do not require a high level of clinical detail.

Purpose: Recent research has highlighted the importance of subgroup analysis to facilitate the use of comparative effectiveness research in shared decision making (Basu, MDM, 2009). Obtaining sufficient data for such analyses may require the use of large observational data sets, particularly where adverse events/failures are rare or occur over extended time periods. Potential pitfalls can still arise when examining small differences across subgroups. We illustrate these issues, in the context of a high-profile example, prosthesis selection for primary total hip replacement (THR). Here decision-makers require cost-effectiveness results for pre-defined age and gender groups.

Method: A Markov model of THR was populated with data from three large databases to compare the cost-effectiveness of cemented, uncemented and hybrid prostheses. Patient reported outcomes on THR are now routinely collected in England providing Generic (EQ5D) and condition specific QoL data before and after THR (n = 10,000). Data on prosthesis survival was taken from the National Joint Register (NJR) of England and Wales (n=217,000) and THR admissions data for English National Health Service Hospitals (HES) (n=457,000). Ordinary least squares regression analysis was used to report QoL following THR with each prosthesis type, for different patient subgroups, adjusting for baseline differences. Alternative model specifications were considered using measures of model fit such as AIC. Combination of data from HES and NJR allowed a semi-parametric consideration of prosthesis survival up to ten years with parametric extrapolation beyond ten years by patient subgroup.

Result: Across the age range considered (60 to 80), cemented prostheses were cheaper and offer superior survival, but hybrid prostheses provide larger gains in QoL. The regression results suggest that the relative gains in QoL for hybrid prostheses may be greater for younger patients. After inclusion of subgroup interactions cemented prostheses dominate hybrid and uncemented prostheses for eighty year olds. Hybrid prostheses are the most cost-effective alternative for sixty and seventy year olds (at λ=£20,000 per QALY the incremental net benefit for females age 70 are: uncemented, £181,000; cemented, £183,000; hybrid £184,000).

Conclusion: Large observational databases can allow crucial parameters in CEA models such as QoL and survival gains to be estimated both overall and for subgroups of high policy interest. This can help both providers and patients make more informed choices about competing alternatives.