ORAL ABSTRACTS: MEASURING ECONOMIC AND HEALTH OUTCOMES
Feng Xie, PhD
Department of Clinical Epidemiology and Biostatistics
Methods: To determine BUSS-P validation and reliability and to ensure generalizability, multi-centre field testing was conducted at an academic (University Health Network, Toronto, ON) and a community (Trillium Health Partners, Mississauga, ON) hospital. Purposive sampling was used to accrue 112 BCa patients with varying disease severity and treatment history. Patients completed the BUSS-P and five other HRQOL and utility instruments (FACT-Bl, BCI, EQ-5D, SF-36, TTO). Whole and subscale Spearman’s rank correlations (rs), as well as comparisons of BUSS-P scores across known-groups were used to assess construct validity. Reliability was assessed at two time-points, four weeks apart.
Results: The BUSS-P was found to have high whole-scale correlations with the FACT-Bl (rs=0.82, p<0.0001), the EQ-5D (rs=0.65, p<0.0001), and the SF-36v2 (PCS rs=0.63, p<0.0001; MCS rs=0.65, p<0.0001). Likewise, high subscale correlations were observed between the BUSS-P and the EQ-5D (emotional wellbeing: rs=0.69, p<0.0001), the FACT-Bl (physical wellbeing: rs=-0.70, p<0.0001), and the BCI (urinary issues: rs=-0.62, p<0.0001). Median BUSS-P scores (M) were found to be significantly different (Kruskal-Wallis p=0.0008) across patients with differing disease severity: non-muscle invasive BC (M=85.0), cystectomy (M=80.0), and metastatic BC patients (M=67.50). Similarly, significant differences (p<0.0001) were noted between the median BUSS-P scores of patients with low comorbidity scores (Charlson Index <4, M=85.0) versus high comorbidity scores (Charlson Index ≥4, M=69.17); a finding that persists when cancer is excluded as a comorbid condition (Charlson Index <2 vs. ≥2, p=0.036). Lastly, excellent agreement was observed between BUSS-P scores at test and retest (ICC=0.79).
Conclusions: These results indicate that the BUSS-P is a valid and reliable instrument to measure HRQOL among all BCa patients (localized to metastatic disease). Future work collecting patient- and community member-generated utility weights to convert the questionnaire into a disease-specific utility instrument is under way.
Our objective was to examine the performance of EQ-5D-5L (index score and the anxiety/ depression dimension), and the VR-12 version 2 (feeling downhearted/depressed item, mental health (MH) domain, and the Mental Composite Summary (MCS) score) in identifying individuals with depressive symptoms.
Data were from an on-going cohort study of adults with type 2 diabetes in Alberta, Canada. The EQ-5D-5L index score and MCS were categorized into quintiles, and the MH domain into quartiles. Both EQ-5D-5L anxiety/depression dimension and VR-12 feeling downhearted/depressed item have five levels. Depressive symptoms were measured using the patient health questionnaire 8 items (PHQ), and were categorized into two severity levels: 1) any depressive symptoms (PHQ³10) vs. absent depressive symptoms (PHQ<10); 2) moderate-severe depressive symptoms (PHQ³15) vs. absent moderate-severe depressive symptoms (PHQ< 15). We calculated sensitivity (Sn), specificity (Sp), and area under receiver operator curve (AUROC) for each of the measures for the two levels of depressive symptoms.
For any level of depressive symptoms: the levels that optimized performance of the measures were quintile 4 for EQ-5D-5L index score (Sn =83.4%; Sp =69.1%, AUROC=0.76), level 2 for the anxiety/depression dimension (Sn =91.7%; Sp =63.8%; AUROC=0.78), level 3 for the feeling downhearted/depressed item (Sn =72%; Sp =80.9%; AUROC=0.76), quartile 3 for the MH domain (Sn =85.4%; Sp =70.8%; AUROC=0.78), and quintile 4 for the MCS (Sn =90%; Sp =71.7%; AUROC=0.81). Overall AUROC were highest for MCS (0.90) and EQ-5D anxiety/depression (0.87).
For moderate-severe depressive symptoms: the levels that optimized performance of the measures were quintile 4 for EQ-5D-5L index score (Sn =92.9%; Sp =64.3%; AUROC=0.79), level 3 for the anxiety/depression dimension (Sn =77.3%; Sp =88.6%; AUROC=0.83), level 3 for the feeling downhearted/depressed item (Sn =83.8%; Sp =76%; AUROC=0.80), quartile 3 for the MH domain (Sn =95.5%; Sp =65.4%; AUROC=0.80), and quintile 5 for the MCS (Sn =87.8%; Sp =86.3%; AUROC=0.87). Overall AUROC were highest for MCS (0.90) and EQ-5D anxiety/depression (0.90).
The EQ-5D anxiety/depression
dimension performed similarly to the VR-12 MCS and slightly better than the VR-12
MH domain in identifying symptoms of depression in this sample of adults with
type 2 diabetes.
This study aimed to evaluate the cost-effectiveness of haemodialysis (HD) and two forms of peritoneal dialysis (continuous ambulatory peritoneal dialysis [CAPD] and automated peritoneal dialysis [APD]) for patients with end-stage renal disease (ESRD) in Singapore.
A Markov model was developed for patients who started dialysis with HD, CAPD or APD in a time horizon of 10 years. Event data (death, hospitalization, and transplantation) was taken from a hospital database and the national renal registry; health utility data came from published studies of Singaporean dialysis patients, and costs data was obtained from a local hospital and dialysis services providers. Outcome measures were 10-year costs (in 2015 Singaporean dollars [SG$]), quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratios (ICERs).
The base-case was a hypothetical cohort of 60-year-old non-diabetic ESRD patients who started dialysis with one of the three modalities and had no contradictions to any modality. We performed base-case cost-effectiveness analysis, one-way sensitivity analysis, and probabilistic sensitivity analysis with Monte Carlo simulation. A high-risk group of 60-year-old diabetic ESRD patients was also analyzed.
The base-case analysis showed that the QALYs were 3.27 with CAPD, 3.48 with APD and 4.69 with HD and the total costs were SG$169,872 for CAPD, 201,509 for APD and 306,827 for HD. The analysis of high risk group showed that the QALYs were 2.50 with CAPD, 2.54 with APD and 3.69 with HD. The total costs were SG$144,972 for CAPD, 169,282 for APD and 271,446 for HD.
For both base-case and high-risk groups, CAPD and HD had extended dominance over APD. The ICER of HD versus CAPD was SG$96,447 per QALY for base-case and 106,281 for high-risk group, respectively.
One-way sensitivity analyses showed that the ICER of HD versus CAPD was most sensitive to the utility for HD for base-case and high-risk groups. Probabilistic sensitivity analysis demonstrated that the probability of CAPD being the optimal choice was 36.2% for the base-case and 44.9% for the high-risk group at a willingness-to-pay threshold of SG$60,000 (US$43,000) per QALY.
CAPD may be a cost-effective therapy compared with HD and APD in ESRD patients in Singapore. These findings are potentially useful to all stakeholders of the dialysis services in Singapore.
Purpose: Preoperative breast MRI performed on women with a recent breast cancer diagnosis has increased in recent years, despite a lack of standard recommendations or definitive evidence of its value. Our objective was to quantify the long-term outcomes and cost-effectiveness of preoperative breast MRI use among women diagnosed with early stage (I-II) invasive unilateral breast cancer.
Methods: We developed a Markov state-transition model to compare the cost and quality-adjusted life years (QALYs) among women with vs. without receipt of preoperative breast MRI. We modeled five initial treatment patterns including: breast-conserving surgery (BCS), BCS with radiation therapy (RT), unilateral mastectomy, unilateral mastectomy with RT, and bilateral mastectomy. After initial treatment, women transitioned annually among three health states associated with their initial treatment: disease-free post initial treatment, disease-free post locoregional second cancer event, and post-metastatic cancer event. Initial treatment probabilities were adjusted estimates from logistic regression models from two data sources: Breast Cancer Surveillance Consortium linked with Medicare (BCSC-Medicare); Surveillance, Epidemiology, and End Results linked with Medicare (SEER-Medicare). Various ages of diagnosis (46, 56, 66) and time horizons (10, 20, 30 years), were used to model numerous scenarios. All second cancer event probabilities were obtained from published randomized control trials. All costs were adjusted to 2015 USD. Both costs and QALYs were discounted at an annual rate of 3%.
Results: Regardless of age at diagnosis, length of time post-diagnosis modeled, and initial treatment data source used, receiving an MRI is the preferred strategy due to increased QALYs compared to no receipt of MRI (Figure). All strategies evaluated incurred increased costs, although all were found to have an incremental cost effectiveness ratio (ICER) below the accepted threshold ($100,000/QALY). Both age of diagnosis and time horizon impacted the results.
Conclusions: Preoperative MRI use among women diagnosed with early stage breast cancer had improved health outcomes at a slightly increased cost. Further clinical investigation to evaluate the impact of breast MRI on health outcomes is warranted.
Methods: We used 2007-14 claims from Truven Health Analytics MarketScan Databases to obtain average monthly per-patient expenditures for orally administered anticancer drugs approved by the FDA between 2000 and 2013. We exploited exogenous variation in the age of diagnosis for different cancer types - and therefore the proportion of individuals diagnosed with different cancers that became eligible for Medicare’s prescription drug coverage – to isolate the impact of Part D on the average monthly per-patient expenditures for oral chemotherapy. We defined the proportion of each cancer type that was Medicare-eligible as the proportion of patients who were age 65 or older at the time of initial cancer diagnosis, which we obtained for 15 unique cancer types from the Surveillance, Epidemiology and End Results program database. We used fixed-effects regression models with robust standard errors to evaluate changes over time in monthly costs of oral chemotherapy within each cancer site. All costs were inflated to 2014 USD.
Results: Across all cancer types, the average monthly cost of oral anticancer drugs increased 8.7% per year over inflation between 2007 and 2014. Monthly costs increased significantly faster over time for cancer sites with a larger exposure to the Medicare market. The average monthly cost of using oral chemotherapy rose 1.9% (95% CI: 0.3, 3.5) faster per year for every 1% increase in the share of the cancer site eligible for Medicare. These cancer-specific trends were driven primarily by greater shifts in utilization towards newly FDA-approved and more expensive oral anticancer therapies rather than larger increases in the costs of existing drugs.
Conclusions: We observed substantial increases in the monthly costs of oral anticancer drugs that may be partly attributable to the reimbursement incentives produced by Medicare Part D. The impact of Medicare Part D’s mandate to include all anticancer drugs on plans’ formularies on pharmaceutical innovation in the market for oral chemotherapy warrants further exploration.
To compare Thais' estimated and ideal distributions of governmental spending on healthcare services for the rich and poor.
The survey was conducted by face-to-face interviews in Thailand from February to April 2015. A nationally representative, probability-based, random sample of Thais (N = 3,500) was asked to estimate the distribution of governmental spending on healthcare services for Thais in each of the five income quintiles. Respondents then reported their ideal distributions: how they thought the government should allocate healthcare spending to rich and poor Thais.
We compared estimated and ideal distributions to each other, and to the actual distributions of healthcare spending for the rich and poor. We focused our analyses on government healthcare spending on the richest (top 20%) and poorest (bottom 20%) groups.
Results were analyzed in aggregate and stratified by participants' gender, age, education, and income. Statistical significance was determined at p < 0.05.
Three thousand and five hundred participants completed the interviews. The response rate was 72.4% (3,500 out of 4,833). Mean age was 41.1 years and half of participants (50.8%) were female.
Respondents underestimated how much of the healthcare budget (i.e., proportion of total healthcare budget) was spent on the richest group (21.9% versus 54.8%) and preferred that a smaller proportion be used for the richest group than their estimate (12.0% versus 21.9%), p < 0.001 for both comparisons. Respondents overestimated how much of the healthcare budget was spent on the poorest group (21.1% versus 5.7%) and preferred that a larger proportion be used for the poorest group compared to their estimate (33.1% versus 21.1%), p < 0.001 for both comparisons. See Figure 1 for full results.
Importantly, respondents preferred that the government spend a significantly higher proportion of healthcare budget on the poorest group compared to the richest group (33.1% versus 12.0%), p < 0.001 – the exact opposite pattern of current spending on the poorest (5.7%) and richest (54.8%).
We found similar results in all subgroup analyses regardless of respondents' gender, age, education and income.
Thais underestimate the disparities in governmental healthcare spending on the rich and poor. They prefer that the government allocate a higher proportion of healthcare budget to the poor and that the proportion for the poor is larger than the proportion for the rich.
Bottom of Form