DISMEVAL A STUDY ON DISEASE MANAGMENT EVALUATION IN EUROPE BERT VRIJHOEF PH.D., PROFESSOR OF CARE FOR THE CHRONICALLY ILL INTEGRATED CHRONIC DISEASE MANAGEMENT FORUM, DEPARTMENT OF HEALTH, 29TH AUGUST 2011, MELBOURNE A few facts The Netherlands • • • • • • 41,526 sq km Amsterdam (capital) 16,6 million inhabitants Infant mortality rate: 4.4/1000 Life expectancy: 78.3 yrs for men and 82.3 yrs for women GDP: 843 billion US dollars www.state.gov Australia • • • • • • 7,7 million sq km Canberra (capital) 22,6 million inhabitants Infant mortality rate: 4.7/1000 Life expectancy: 79.3 yrs for men and 83.9 yrs for women GDP: 1.3 trillion US dollars www.state.gov A few facts The Netherlands • Hockey Champions Trophy 2011 Australia • Tour de France 2011 A few facts The Netherlands • Planking Australia • Planking Contents Disease Management Disease Management Evaluation DISMEVAL, the project DISMEVAL, the Netherlands Take home messages ‘Disease management’ presents a spectrum of chronic care improvement programmes ranging from no till substantial attempt to redesign practice Measuring disease management programme performance at the population level in a scientifically sound fashion while being practicable for routine operations remains a challenge DISMEVAL illustrates that other designs (than RCT) constitute valuable designs for the evaluation of complex interventions and provide in-depth insights Future evaluation of ‘disease management’ should focus on the identification of subgroups and target these subgroups DISEASE MANAGEMENT • ‘Disease management’ is a global buzzword in health care and it is often used by different people to mean different things • Structured ‘disease management ‘has been proposed as a means to improve quality and reduce the costs of health care and to improve health outcomes for the chronically ill • ‘Disease management’ presents a wide spectrum of chronic care improvement programmes DISEASE MANAGEMENT • Coleman et al. [2008] distinguish programmes by the degree to which they work to redesign the patients’ medical care DISEASE MANAGEMENT Deficiencies in current health care systems: • rushed practitioners not following evidence based guidelines • lack of care coordination • lack of active follow-up to ensure the best outcomes • patients inadequately trained to manage their illness DISEASE MANAGEMENT Why system change? • to put emphasize on system instead of only on physician • to categorize characteristics of successful interventions in an useful way • to appreciate commonalities across chronic conditions Contents Disease Management Disease Management Evaluation DISMEVAL, the project DISMEVAL, the Netherlands DISEASE MANAGEMENT EVALUATION • What we know about the impact of interventions to manage chronic disease(s) tends to be based, mainly, on small studies that frequently focus on high risk patients, and are often undertaken in academic settings (Bodenheimer, Wagner and Grumbach, 2002). • The effects of large, population-based programmes and initiatives are less well understood (Mattke, Seid and Ma, 2007). • Four examples: Steuten et al. (2006); Mattke et al. (2007); Coleman et al. (2008, 2009); Peikes et al. (2009). DISEASE MANAGEMENT EVALUATION • Steuten et al. (2006) found that a link between aims of disease management and evaluated indicators does not exist in a substantial part of published studies on disease management of diabetes and asthma/COPD, especially when efficiency of care is concerned. DISEASE MANAGEMENT EVALUATION • Mattke et al. (2007) report that the evidence on the role of disease management in reducing utilization of health service is inconclusive. Support for disease management “is more an article of faith than a reasoned conclusion grounded on well researched facts”. DISEASE MANAGEMENT EVALUATION • Coleman et al. (2008) found that the degree to which disease management redesigns practice predicts more positive outcomes. The impact of the Chronic Care Model on health care costs and revenues remains uncertain and probably varies by condition (2009). DISEASE MANAGEMENT EVALUATION • Peikes et al. (2009) conclude that care coordination, as practiced by the programs participating in the demonstration from 2002 to 2006, holds little promise of reducing total Medicare expenditures for beneficiaries with chronic illnesses. DISEASE MANAGEMENT EVALUATION • The overall results of disease management evaluation studies are hard to interpret. • This is in part because of the variation in interventions, their constituting components, and the lack of standardization of initiatives (e.g. design, scope, scale, operational detail and providers involved vary widely). • Another challenge is the lack of widely accepted evaluation methods to measure and report programme performance at the population level in a scientifically sound fashion and that are also practicable for routine operations. • Contents Disease Management Disease Management Evaluation DISMEVAL, the project DISMEVAL, the Netherlands DISMEVAL • Aim: to support the evaluation of disease management by identifying and validating evaluation methods and performance measures for disease management programmes or equivalent approaches and to make recommendations to policymakers, programme officials and researchers on best practices that are both scientifically sound and operationally feasible. • DISMEVAL brings together a multi-disciplinary team of 10 partners in 7 EU countries. It is funded under the European Commission’s 7th Framework Programme, theme Health and runs from 1 January 2009 to 31 December 2011. DISMEVAL Evaluating chronic disease management in Europe -- A review of evaluation methods and performance measures in use by Conklin A, Nolte E, Vrijhoef HJM (submitted) 1/3 • Review on approaches to chronic disease management in 12 European countries and data on methods and metrics to evaluate these approaches • Austria, Denmark, England, Estonia, France, Germany, Hungary, Italy, Latvia, Lithuania, Netherlands, Switzerland • key informants per country were identified and asked to collect data on: (i) chronic disease management approach and (ii) evaluation categories as defined in template • data synthesis: narrative approach DISMEVAL Evaluating chronic disease management in Europe -- A review of evaluation methods and performance measures in use by Conklin A, Nolte E, Vrijhoef HJM (submitted) 2/3 • most chronic disease management approaches underwent some form of evaluation, except in Latvia and Lithuania • nature and scope of evaluations varied considerably • mix of controlled and non-controlled studies measuring predominantly clinical process measures, patient behaviour and satisfaction, cost and utilisation • effects were usually observed over 1-3 years on a population of patients with a single, commonly prevalent, chronic disease DISMEVAL Evaluating chronic disease management in Europe -- A review of evaluation methods and performance measures in use by Conklin A, Nolte E, Vrijhoef HJM (submitted) 3/3 • majority of approaches were targeting diabetes and/or cardiovascular disease(s) • most commonly aimed to assess the performance of a given approach with rarely specified information about targets against effects might be compared • about half of evaluations used non-experimental designs without means for comparison • large differences in indicators of effect, length of observation • evaluations are carried out by different actors DISMEVAL Disease management evaluation – A comprehensive review of current state of the art by Conklin A, Nolte E (RAND 2010) 1/3 • aim: to help advance the methodological basis for chronic disease management evaluation by providing a comprehensive inventory of current evaluation methods and performance measures, and by highlighting the potential challenges to evaluating complex interventions such as disease management • two-stage literature review: (i) methodology papers, (ii) evaluation of disease management • 111 papers (89 journal articles, 6 book chapters, 10 working papers, 6 other) DISMEVAL Disease management evaluation – A comprehensive review of current state of the art by Conklin A, Nolte E (RAND 2010) 2/3 • challenges identified are conceptual, methodological, and analytical in nature • conceptual: clarification of characteristics of disease management intervention and the selection of evaluation measures • methodological and analytical: the establishment of the counterfactual is a key challenge -- in the context of multi-component, multi-actor disease management initiatives, RCTs are frequently not applicable because randomisation is not possible (or desirable) for reasons such as cost, ethical considerations, generalisability, and practical difficulties of ensuring accurate experimental design. DISMEVAL Disease management evaluation – A comprehensive review of current state of the art by Conklin A, Nolte E (RAND 2010) 3/3 • as alternative strategies become less of a controlled experiment, there are more threats to the validity of findings from possible sources of bias and confounding which can undermine the counterfactual and reduce the utility of the evaluation • other analytical challenges: selection of suitable control group, determining statistical power, case mix, etc. • a clear framework of the mechanisms of action and expected effects that draws on an understanding of the characteristics of disease management, those of the intervention and target populations, an adequate length of observation to measure effects and the logical link between performance measures and the intervention’s aims, elements and underlying theory driving the anticipated behaviour change Contents Disease Management Disease Management Evaluation DISMEVAL, the project DISMEVAL, the Netherlands DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • most experience in disease management approach in NL is related to care for people with diabetes • disease management is organized by ‘care groups’: provider networks based in primary care • insurers pay a single fee to care groups, to cover an integrated bundle of care for a specific chronic disease over a fixed period of time (usually one year) • minimal required services are stipulated in national care standards DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • aim: to advance current methods of disease management evaluation by testing and validating potentially valuable research designs on data from existing programmes set in the Netherlands • assumption: advanced methods are necessary to go beyond a ‘grand mean effect of disease management’ and gain insight into what (combination of) components of the approach work(s) and for whom. • use was made of data from 9 care groups, already collected by Dutch National Institute for Public Health and the Environment, and an other 9 care groups retrospectively retrieved for this study DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • experimental comparisons are not feasible, as the DM approach to T2DM care has been dispersed across all regions of the Netherlands • two research designs were applied, i.e. meta-analysis and metaregression, which are assumed to be particularly suitable for evaluating a heterogeneous care strategy that differs across settings in terms of the specific interventions being offered, targeted patients, and level of implementation • included outcomes are restricted to clinical endpoints, as data on more patient-centered measures, such as quality of life or satisfaction with care, as well as on the costs of care could not be retrieved retrospectively. DISMEVAL: The Netherlands >8000 2000-8000 <2000 ●●● Frontrunner groups ●●● Non-frontrunner groups • 18 care groups • data of over 105,000 diabetes patients • baseline 8-12 months and follow-up 12 months DISMEVAL: The Netherlands DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • meta-analysis of individual patient data (IPD), i.e. the raw individual level data for each study, which is considered by some as the ‘gold standard’ of systematic review (Thompson & Higgins, 2005; Stewart, Tierney & Clarke, 2011). We applied meta-analysis methods to determine the impact of the programmatic approach on IPD concerning changes in eight clinical outcomes between baseline and follow-up. • meta-regression can examine multiple individual and group level factors together, the results of which may facilitate stratified medicine, and adjust for baseline factors Care group 1 Frontrunner status1 Size of patient population2 Price of care bundle (€) 3 Collaboration with specialist care4 351,00 Limited 367,48 Limited 1 No 2 No 3 No 18,531 345,72 Limited 4 No 7913 455,23 Extensive 5 No 1308 358,00 Limited 6 No 9392 380,04 Extensive 7 No 5778 390,00 Extensive 8 No 7192 458,20 Limited 9 No 13,754 298,92 Limited 10 Yes 348 - - 11 Yes 4948 - - 12 Yes 1539 - - 13 Yes 9756 - - 14 Yes 1072 - - 15 Yes 7636 - - 16 Yes 1893 - - 17 Yes 1614 - - 18 Yes 2174 - - 1480 8728 Frontrunners are subsidized by the government as part of the experimental pilot into bundled payments for integrated diabetes care; 2 Size of patient population during groupspecific research period of either 18 or 24 months (see figure 2); 3 Price per care bundle per year on 01/01/2010; 4 Limited collaboration is defined as ‘primary care providers having an agreement with specialists that enables them to consult the latter for advice’, whereas extensive collaboration implies ‘more intensive cooperation, for instance in the form of patients being seen by specialists in the primary care setting’. Care group Frequency (median) HbA1c Range (%)1 Cholesterol LDL HDL Triglycerides SBP DBP BMI 1 1 1 1 1 1 4 4 3 69.7 2 3 1 1 1 1 4 4 2 55.7 3* - - - - - - - - 86.7 4 3 1 1 1 1 3 3 2 44.4 5 2 1 1 1 1 4 4 3 65.4 6* - - - - - - - - 84.6 7* - - - - - - - - 84.5 8 2 2 2 2 2 4 4 3 74.7 9 4 1 1 1 1 3 3 1 44.4 10 2 1 1 1 1 3 3 3 57.2 11 2 1 1 1 1 3 3 3 74.0 12 4 1 1 1 1 4 4 4 76.8 13 2 1 1 - 1 4 4 4 73.0 14 2 1 1 1 1 3 3 3 64.1 15* - - - - - - - - 65.6 16 1 1 1 1 1 2 2 1 73.9 17 1 1 1 1 1 3 3 2 55.1 18 4 1 1 1 1 4 4 2 81.6 Range was operationalized as the percentage of patients having registrations of all eight included clinical outcomes during the follow-up year. *Care groups 3, 6, 7, and 15 registered solely the endpoints measured during the annual check-up of patients. Hence, information on frequency of measurements was unavailable for these groups. 1 Characteristics Patients for whom characteristic is known Estimate (total =105,056) Baseline age Baseline diabetes duration Sex % (N) Mean ± SD 99.9 (105,014) 65.7 ± 11.9 71.9 (75,498) 4.8 ± 5.6 % (N) % (N) 99.3 (104,369) Male 49.3 (51,421) Female Medication 50.7 (52,948) 65.0 (68,298) No 36.0 (24,606) Yes Type of medication 64.0 (43,692) 41.6 (43,692) Oral 80.5 (35,163) Insulin 7.9 (3460) Both Comorbidity1 11.6 (5069) 94.5 (99,278) None 84.2 (75,357) One or more Smoking status 15.8 (14,165) 74.6 (78,384) No or Ex-smoker 81.6 (63,943) Current smoker 18.4 (14,441) % (N) Mean ± SD Baseline HbA1c (mmol/mol) [target <53] 71.5 (75,127) 50.2 ± 9.8 Baseline total cholesterol (mmol/l) 58.4 (61,376) 4.5 ± 1.0 Baseline LDL (mmol/l) [target < 2.5] 55.9 (58,697) 2.6 ± 0.9 Baseline HDL (mmol/l) 51.8 (54,456) 1.2 ± 0.4 Baseline triglycerides (mmol/l) 58.1 (61,078) 1.8 ± 0.9 Baseline SBP (mmHg) [target < 140] 69.9 (73,437) 140.4 ± 18.0 Baseline diastolic blood pressure (mmHg) 69.6 (73,115) 78.60 ± 9.6 Baseline BMI (kg/m2) [target < 25] 60.3 (63,341) 29.7 ± 5.2 1Included were four major comorbidity associated with diabetes mellitus: angina pectoris, myocardial infarction (MI), stroke, transient ischemic attack (TIA). DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • Main results from meta-analysis: Endpoint Care groups (N) Patients (N) Mean difference [95%CI] Heterogeneity (I2, P) HbA1c (%) 18 75,127 0.17 [-0.60, 0.93] 98%* Total cholesterol (mmol/l) 18 61,376 -0.10 [-0.14, -0.06]* 90%* LDL (mmol/l) 18 58,697 -0.09 [-0.13, -0.05]* 93%* HDL (mmol/l) 17 54,456 0.02 [0.00, 0.03]* 92%* Triglycerides (mmol/l) 18 61,078 -0.05 [-0.07, -0.03]* 75%* SBP (mmHg) 18 73,437 -0.95 [-1.25, -0.64]* 57%* DBP (mmHg) 18 73,115 -0.80 [-0.93, -0.67]* 34% BMI (kg/m2) 18 63,341 -0.04 [-0.10, 0.02] 0% * Statistically significant (p<0.05) DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • subgroup analysis were performed on patient, care group and process characteristics. Most notably: • DM is more effective for patients who have poor baseline clinical values (HbA1c, LDL, SBP) than for those with better baseline health • impact of DM for BP is greater in patient in higher age categories • care group characteristics did not reduce heterogeneity • patients whose cholesterol was measured more frequently show better results DISMEVAL: The Netherlands • Main results from meta-regression: Groups (N) HbA1c (mmol/mol) Care group characteristics Change in variance Intercept Frontrunner subsidy Care group size Process characteristics Change in variance Intercept Measurement frequency Measurement range Length of follow-up Patient characteristics Change in variance Intercept Age Disease duration Baseline HbA1c (mmol/mol) Comorbidity Smoking status Significant covariates Change in variance Intercept Measurement range Length of follow-up Age Disease duration Baseline HbA1c (mmol/mol) Smoking status * Statistically significant (p<0.05) 18 18 14 13 13 Patients (N) 75127 75127 48235 42659 42659 Coefficient τ2 0.1810 -0.5230 0.9698 0.000038 -0.4845 -0.02744 -0.09095* 0.1141* 20.8566* -0.03388* 0.1471* -0.3869* 0.1954* 0.2458* 21.1761* -0.1441* 0.06357* -0.03288* 0.1468* -0.3873* 0.2477* σ2 ICC 2.8001* 2.6095* -6.8% 62.6311* 62.6311* 0% 3.5274 26.0% 62.5406 -0.1% 2.4510* -12.5% 47.9356* -23.5% 2.5665 -8.3% 47.8876 -23.5% 4.28% DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • meta-regression revealed that most variance exists within groups • no evidence was found for relationship between care group characteristic on any clinical outcome • process characteristics: as the duration of care increases, the positive effects of DM diminish; higher measurement frequency is associated with better outcomes • patient characteristics: impact of DM is greater as patients’ baseline values are poorer and this is more determinative than other characteristics DISMEVAL: The Netherlands Going beyond the ‘grand mean’ – Advancing approaches to disease management evaluation in the Netherlands by Elissen A, Duimel-Peeters IGP, Spreeuwenberg C, Vrijhoef HJM • from a simple pre-post comparison one would conclude that DM does not achieve the intended goals • from mete-regression and meta-analysis it was found that the variance within care groups is much higher than between care groups • the poorer patients’ baseline values of a particular endpoint are, the more beneficial frequent measurement of that outcome is • in other words: intensive DM programs targeting patients at high risk of diabetes complications have great potential for costeffectiveness Take home messages ‘Disease management’ presents a spectrum of chronic care improvement programmes ranging from no till substantial attempt to redesign practice Measuring disease management programme performance at the population level in a scientifically sound fashion while being practicable for routine operations remains a challenge DISMEVAL illustrates that other designs (than RCT) constitute valuable designs for the evaluation of complex interventions and provide in-depth insights Future evaluation of ‘disease management’ should focus on the identification of subgroups and target these subgroups THANK YOU H.J.M.VRIJHOEF@UVT.NL INTEGRATED CHRONIC DISEASE MANAGEMENT FORUM, DEPARTMENT OF HEALTH, 29TH AUGUST, MELBOURNE