4J
HEART DISEASE: DECISION MAKING AND CLINICAL GUIDELINE IMPLEMENTATION
* Finalists for the Lee B. Lusted Student Prize
Purpose: Discount factors are critical in assessing the cost effectiveness of treatments as they reflect the time value of costs and benefits. We considered the trade-off between the benefits and side effects of statins to elicit the discounted present value of each future life year for current lipid management guidelines.
Method: We formulated the statin initiation problem for Type 2 diabetes patients as a Markov decision process where the state of the process is defined as the patient's total cholesterol and high density lipoprotein (HDL) levels, and the statin initiation decision is revisited annually. We considered six guidelines: Adult Treatment Panel (ATP) III, Canadian, British, European Union, Australian and New Zealand. For each guideline, we used inverse optimization to find the discounted present value of each future life year that makes the guideline non-dominated with respect to total expected QALYs prior to a first major cardiovascular event and the risk of a first major cardiovascular event. For patients diagnosed with Type 2 diabetes at age 40, we used Mayo Clinic electronic medical records to model the progression of their total cholesterol, HDL, triglycerides, systolic blood pressure, and HbA1c. We incorporated the adverse effects of statins into the model using a constant disutility factor.
Result: Among all guidelines we considered, the underlying present values of future life years for males were never more than those of females. The present value of a life year for males varied from 20% to 70% of that for females between ages 45 and 65. All guidelines were consistent in discounting future with time-varying annual discount factors. Guidelines which initiate statins at earlier ages, such as ATP III, valued the far future more than those which initiate statins relatively later. For males, all guidelines exhibited a sharp decrease in discounting life years after age 45; for females this threshold was at age 55. While no guideline discounted the value of a life year immediately following the diagnosis of Type 2 diabetes, all guidelines were consistent with a present value of <0.001 years beyond age 80.
Conclusion: Our analyses show that guidelines discount future life years with time-variant annual discount factors and show a substantial difference between the implied present value of life years for males and females.
Method: A Markov model (six-month cycle-length; life-time horizon; NHS perspective), developed by the National Clinical Guideline Centre to inform the National Institute for Health and Care Excellence (NICE) guidelines, on treatment options for adults with primary hypertension (CG127) was used as the basis for this analysis. The model structure captured seven relevant health states for patient cohorts (starting age 55-years) with hypertension prescribed either a: thiazide diuretic (TD); calcium-channel blocker (CCB); beta-blocker (BB); angiotensin-converting enzyme inhibitor/angiotensin-II receptor blocker (ACE/ARB). The relevant comparator was defined as ‘no intervention’. Pay-off times were numerically captured, by assessing when the cumulative benefits associated with treatment exceeded the cumulative harms. Both benefits and harms were measured using the quality-adjusted life-year. The model was re-run (‘iteration’) for: starting age; patient gender; assumed harm-related utility decrement (0 or 0.01 per cycle). Adaptations were also made to the Excel® produced Markov model, to generate visual profiles of cumulative QALY gains which clearly show the benefit/harm trade-off over time.
Result: The pay-off time was calculated for each iteration of the model. In the base-case analysis (no harms; utility decrement = 0) the estimated pay-off time was positive in cycle one (at six-months) for each of the five anti-hypertensives in a 55-year old male patient cohort. In comparison, when there was an assumed ongoing treatment-induced harm (constant 0.01 utility decrement), the estimated pay-off time was: 8.3 years for CCB; 8.6 years for ACE/ARB; 8.7 years for TD; 10.1 years for BB.
Conclusion: This study illustrates the potential for implementing quantitative estimates of the ‘pay-off time’, together with visual cumulative QALY-gain profiles, within current economic models. Both concepts represent potentially useful tools to aid decision-making when prioritising treatment options.
Two-thirds of Medicare beneficiaries have multiple chronic conditions (MCCs), or two or more chronic conditions, but clinical practice guidelines (CPGs) often provide recommendations for single conditions. We developed and evaluated an automated method to determine the extent to which disease-specific CPG recommendations mention comorbid conditions.
Method:
This study focuses on guidelines for the 15 most prevalent Medicare chronic disease diagnoses, excluding cancer given the breadth of the term, and adding obesity due to its high prevalence and clinical significance. We compiled a corpus of text from CPG summaries available in the National Guideline Clearinghouse. CPGs were included if only one of the 15 diseases was mentioned in the title and the target population was the general adult, non-pregnant population. Using disease synonyms obtained from ontologies accessible via BioPortal at the National Center for Biomedical Ontology, we developed a direct text matching algorithm to identify comorbid disease terms in the recommendation sections of included CPGs. We tabulated the proportion of CPGs mentioning comorbid disease terms, and the number of comorbid disease terms mentioned in each CPG. To evaluate the automated approach, manual annotation by two annotators with medical expertise performed sentence-level annotation on five randomly selected CPGs for different diseases to generate a preliminary reference standard.
Result:
We obtained 2,503 guideline summaries, and 243 met inclusion criteria. Guidelines for concordant diseases (diseases that are part of the same pathophysiologic risk profile), such as hypertension, diabetes, hyperlipidemia, and obesity, mentioned one another most frequently. Only hypertension was mentioned across all CPGs, while Alzheimer’s disease and osteoporosis were mentioned the least. Annotators agreed on 95.5% of 561 sentences in the reference standard. Compared to the reference standard, precision (or positive predictive value) of the automated method was 0.79, recall (or sensitivity, or true positive rate) was 0.86, and F-measure (the harmonic mean of precision and recall) was 0.83.
Conclusion:
We developed and evaluated an automated method that identifies comorbid disease terms in CPG recommendations. An annotation guide to improve the reference standard is in development and will guide algorithm improvements. This method may be useful to inform gaps in guideline recommendations regarding comorbid diseases and therefore identify opportunities for guideline improvement. Further investigation is needed to understand the context and variation of comorbid disease mentions in CPGs.
There are numerous health-related quality of life (HRQol) measurements in coronary heart disease (CHD) in the literature. Only those measured with preference-based instruments (e.g. EQ-5D), or mapped from different HRQoL measurements can be applied in cost-utility analyses (CUAs).
The aim of this study is to synthesize instrument-specific preference-based values in CHD while accounting for study-level covariates and correlation between those values using multivariate meta-regression.
Method:
A systematic review was conducted to identify studies reporting preference-based (EQ-5D, SF-6D, 15D, HUI3, QWB) and non-preference-based HRQoL measurements in CHD. Non-preference based measurements were further mapped onto EQ-5D and SF-6D estimates. A multivariate random-effects meta-regression model was applied to synthesize the HRQoL measurements. Study-level covariates examined were: underlying form of CHD (i.e. stable angina, acute coronary syndrome (ACS), general CHD), age, time point of measuring HRQoL, publication year.
Result:
A total of 34 studies providing preference-based values and 47 studies providing SF-36 and SF-12 measurements were detected. Two data sets were built from the collected data. The base-case data set included only published preference-based values. The joint data set comprised both published and mapped EQ-5D and SF-6D estimates.
The synthesized estimates using the base-case data set were: 0.77 (EQ-5D UK ”tariff”), 0.80 (EQ-5D US ”tariff”), 0.79 (EQ-5D European ”tariff”), 0.69 (SF-6D), 0.85 (15D), 0.46 (HUI3) and 0.62 (QWB). The estimates summarized using the joint data set were: 0.74 (EQ-5D UK ”tariff”) and 0.73 (SF-6D). No significant improvement in model fit was found after adjusting for any study-level covariates. Analysis for the EQ-5D UK and US “tariff” values were stratified for stable angina, ACS and general CHD. A large heterogeneity unexplained by meta-regression modeling was observed on all analyses.
Conclusion:
The choice of the most robust value from the abundance of HRQoL measurements is not trivial given its impact on the CUA outcome and subsequent interpretation of CUA results. Notably, it is this sensitivity of CUA outcomes on the HRQoL that urges for an accurate estimate of HRQoL. Given the abundance of HRQoL measurements in CHD and the requirement for applying a single, state-specific value in a CUA, synthesized estimates could be highly applicable in CUAs.
Purpose: The majority of patients with diabetes and hypertension require additional medications over time in order to maintain intensive glucose and blood pressure control. However, it is unclear whether patient expectations are closely aligned with the natural histories of these conditions. We interviewed patients to identify their future expectations for their diabetes and hypertension.
Method: We recruited adults, aged 40-70 years old, living with co-occurring diabetes (duration <10 years) and hypertension from an academic primary care clinic (N=35). Eligibility criteria included taking oral medications for both conditions. Patients taking insulin were excluded. We conducted semi-structured, in-person interviews which asked subjects 3 open-ended questions about their future expectations: 1) how much longer they would need to take diabetes and hypertension medications, 2) if they expected to need additional diabetes and hypertension medications, and 3) if they believed their diabetes and hypertension could be cured.
Result: Subjects had a mean age of 59 years, 66% were female, and 89% were African American. Subjects had been taking diabetes and hypertension medications for a mean of 4 and 11 years, respectively. About half of subjects expected to stop taking their diabetes (57%) and hypertension medications (53%) within 3 years (Table). In contrast, 14% of subjects expected to need diabetes medications and 32% expected to need hypertension medications for the rest of their lives. About two-thirds of subjects did not expect to need additional medications for their diabetes [pills (62%) or insulin (68%)] or hypertension (69%) in the future. An additional quarter of subjects expected to probably not need additional medications for their diabetes [pills (29%) or insulin (23%)] or hypertension (26%). The majority of subjects believed that their diabetes could be cured (60%) or probably cured (20%). Fewer subjects believed that their hypertension could be cured (50%) or probably cured (15%).
Conclusion: Despite the fact that most patients with diabetes and hypertension require medications for life, we found that patients are very optimistic about the future management of their diabetes and hypertension. Most patients with co-occurring diabetes and hypertension expect to be medication-free and cured in the future. Educating patients on the natural history of diabetes and hypertension may be an important step to managing expectations and could be used to motivate patients to enact lifestyle changes.
Method: Encounter data were studied from state Medicaid Managed Care Organizations for 18-64 year old patients with a diagnosis of HF over 5 years ending in 6-30-10. Exploratory models assessed the prevalence of COPD, CVD, stroke, and renal dysfunction, respectively and shown by age, race and gender, as diagnosed in HF patients within 3 months after HF diagnosis.
Result: Of 14,149 HF patients, 36% were 45-54 years old (yo), 35% 55-64 yo and others under 45; 56% females, 60% African-Americans. Most had hypertension (11130). Diabetes (6369), COPD (4297 ), renal dysfunction (4012) and or stroke (3208) were major comorbidites. HF Patients with renal dysfunction were younger (27% under 45), than counterparts with hypertension (25%), diabetes (20%), stroke (20%) and or COPD (16%). Those with COPD were largely Caucasian (46%), while counterparts with hypertension (62%), renal dysfunction (66%), diabetes (60%) and or stroke (60%) were largely African American. Except for renal dysfunction (48% female), most patients with COPD (60%), hypertension (62%), renal dysfunction (66%), diabetes (59%) and or stroke (55%) were females.
Conclusion: The prevalence of COPD, stroke, hypertension, diabetes and renal dysfunction in heart failure is high. Those comorbidities however affect HF patients differently according to race, age and gender, which calls for customized treatments for high risk but relatively younger, more diverse and female HF populations, typically under-represented in clinical trials.