The National Program for Quality Indicators In Community Healthcare: Methodological Issues Orly Manor 7th meeting of the Eastern Mediterranean Region of the International Biometric Society (EMR-IBS) Tel-Aviv April 22-25, 2013 2 “Efforts to improve quality require efforts to measure it” (Casalino, 2000:NEJM) Healthcare in Israel 3 National Health Insurance law (1995) Universal healthcare Standard basket of medical services Four health plans (kupot cholim) “Justice, equity and solidarity…medical services will be offered based on medical considerations, with reasonable quality.” Healthcare in Israel 4 Health Tax - progressive , paid to the National Insurance Institute (funds are distributed to the health plans according to a capitation formulae) + modest copayment directly to health plan for specific services Open enrollment, no option to reject applicant, annual option to switch plans Managed competition between plans (uniform benefits) Competition is based on quality and nature of services Israel National Program for Quality Indicators in Community Healthcare (QICH) 5 Supervisory bodies were established “to follow and assess the influence of the NHI law on health services in Israel, their quality, effectiveness and cost" QICH started as a research project (Porath & Rabinovitz, 2002) and later adopted as a national program. Full cooperation and support by all four health plans Mission 6 To provide the public and policy-makers information on the quality of community healthcare provided in Israel. This information covers various health categories and is intended to promote and improve the standard of healthcare in Israel. Main Product 7 Annual report presenting national results of quality indicators in community healthcare Enables the evaluation of developments and changes in healthcare over time the early identification of risk factors in the Israeli population and in sub-populations the comparison of healthcare quality in Israel with other countries 8 "Not everything that counts can be counted, and not everything that can be counted counts." Einstein Quality Indicators- Methods 9 Measures of clinical performance (structure, process, outcome) Based on electronic health records from the four health plans All indicators are rates Some indicators are conditional on others Covariates: age, sex, SEP (proxy) Quality indicators (2013) 10 Health promotion BMI Smoking cessation Cancer Children and adolescents Breast cancer screening Anemia screening (infants) Colon cancer screening Weight and height documentation Elderly Influenza vaccination (seasonal) Pneumoc. vaccination Asthma Secondary prevention Appropriate use of control medication Influenza vaccination (seasonal) Cardiovascular health Diabetes Primary prevention Secondary prevention Cholesterol assessment Blood pressure documentation Cholesterol documentation Blood pressure documentation BMI Blood glucose documentation Eye exam Kidney care Influenza vaccination (seasonal) Pneumococcal vaccination Secondary prevention LDL modifier use Effectiveness of care Cholesterol control Effectiveness of care Cholesterol control Blood glucose control Blood pressure control Example: Diabetes 11 Blood glucose levels of individuals with diabetes are directly related to the development of complications: cardiovascular disease, blindness, kidney failure Monitoring blood glucose by periodic hemoglobin A1c testing and achieving adequate glycemic control Example: Diabetes 12 Prevalence measure: Rate of individuals with diabetes mellitus from the entire population (overall and by age and gender) Process measure: Rate of individuals with diabetes with documented levels of hemoglobin A1c (HbA1c) Outcome (intermediate) measure: Rate of individuals with controlled levels of HbA1c from patients with diabetes with documented levels of hemoglobin A1c (HbA1c) Methodological Issues 13 1. Criteria 2. Population coverage 3. Data quality-measurement error 4. Data sources 5. Consistency of measures 6. Reporting Criteria 14 Importance & relevance Evidence Ability to quantify Availability & accessibility of electronic data Population Coverage 15 Population-based, near-complete coverage (not sample) Transfers between health plans Births/deaths Other populations: e.g., soldiers Data Quality- misclassification error 16 Estimating the prevalence of a medical condition in the absence of neither a gold standard nor an additional classification. We wish to estimate - prevalence, sensitivity and specificity , yet df=1. We can use a Bayesian approach- simultaneous inferences of the prevalence, sensitivity and specificity and positive and negative predictive value (Joseph et al 1995) Selecting priors- experts’ opinion, understanding sources of data and errors (Greenland 2009) Data Quality- Consistency of Measures 17 Uniform definitions across health plans Membership Data collection period Numerator and denominator Data Sources 18 1. Medical records (e.g., documentation of BMI) 2. Nurse’s records (e.g., documentation of vaccination) 3. Pharmacy claim records (e.g., medication purchase) 4. Laboratory results (e.g., HbA1c levels) 5. Hospital procedure codes (e.g., CABG) 6. Other (e.g., mammography) 19 Data Quality – Sources of Data and Sources of Error 1. Medical records (e.g., documentation of BMI) Automated manual data of input 2. Nurse’s records (e.g., vs documentation vaccination) (variation between and within health plans) Pop-up options vs typing in vaccine name, 3. Pharmacy claim records (e.g., medication purchase) historical data ATC vsbetween YARPA/LARGO (variation health plans) 4. Laboratory results (e.g., HbA1c levels) (variation between health plans) Standardized for calibration 5. Hospital procedure codes values (e.g., CABG) (variation between laboratories) MOH codes used for billing 6. Other (e.g., mammography) (too broad) Self reported, billing-based Data Quality – Checks and Audit 20 1. Internal (health plans) Data checks (BI) Feedback loops/criterion validity 2. Quasi-external (directorate) Between and within health plan data checks (outliers) Comparison with existing national data 3. External (independent auditor) Process audit of infrastructure changes Process audit of indicator implementation Reporting 21 Transparency + court ruling of the public reporting of indicators by health plan Case mix: substantial differences between health plans by SEP Limited available data on SEP Friedberg et al. (2011). Rand Corporation: Methodological Issues in Public Reporting Reporting - Adjusting for Case Mix 22 Israel is divided into statistical areas. The Israel Central Bureau of Statistics calculates SEP scores for each statistical area using recent census information Currently-using GIS each person’s address (unidentified data) is linked to his statistical area and the respective SEP score Reporting -Benchmarking 23 Setting benchmarks: setting, testing, comparing?? Diabetes care 24 Diabetes care 25 Diabetes care includes routine monitoring and proper control of: • Blood glucose levels (93%) • Cholesterol levels (90%) • Blood pressure measurement (92%) • BMI assessment (86%) • Eye examination (65%) • Influenza (55%) and pneumococcal vaccination (77%) Diabetes care 26 International comparisons – Diabetes care (2010) 27 *US data from the National Committee for Quality Assurance, HEDIS data set for 2010 Diabetes care – Israel and England (QOF) (2009) Thank you 29 Directorate: Orly Manor, Arie Ben-Yehuda, Amir Shmueli, Ora Paltiel, Ronit Calderon and Dena Jaffe Directorate staff: Wiessam Abu Ahmad, Galit Shefer. Maccabi Clalit Yair Birenboim Chaim Bitterman Einat Elran Orit Ya’akobson Nesya Gordon Arnon Cohen Rachel Marom Margalit Goldsprecht Guy Levy Tamara Koren External auditor: Aliza Lukach Israel National Institute for Health Policy Research Advisory boards Meuhedet Leumit Liora Valinsky Daniel Vardi Yossi Zini Eran Matz Alon Yaffe Doron Dushnitzky Nirit Peretz Importance and Relevance 30 Evidence – Moving Target 31 HbA1c control <7% Newer cases <7% 18-74 years <7% <8% 75-84 years <8% Older cases (10+ years) “Intensive therapy was stopped after a mean of 3.5 years due to increased mortality” The Action to Control Cardiovascular Risk in Diabetes Study Group N Engl J Med 2008; 358:2545-2559 . Ability to Quantify – Defining Diabetes 32 Definition 2002-2010 • Purchase of medications for diabetes Definition 2010-current • Purchase of medications for diabetes • Laboratory values (high glucose or high HbA1c) Uncontrolled HbA1C >9% (males) Reporting SES 0-4 yrs 5-17 yrs 18-24 yrs 25-34 yrs 35-44 yrs 45-54 yrs 55-64 yrs 65-74 yrs 75-84 yrs 85+ yrs 33 Low SES 0-4 yrs 5-17 yrs 18-24 yrs 25-34 yrs 35-44 yrs 45-54 yrs 55-64 yrs 65-74 yrs 75-84 yrs 85+ yrs High SES 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% Quasi-external Audit 34 Comparison old v. new 2009 Three-year trend 2008-2010 Relative difference New Report Measure name Old 2009 New 2009 2010 20082009 20092010 Diff Rel Diff 2008 2009 Influenza vaccination 0.605 0.605 0.00 0.00% 0.60 0.60 0.58 0.92% -4.60% Pneumococcal vacciation 0.500 0.796 0.81 0.79 0.78 -1.62% -1.75% HbA1c documentation 0.931 0.931 0.00 0.00% 0.92 0.93 0.93 0.93% -0.06% Controlled HbA1c (0-74 yrs) 0.477 0.455 -0.02 -4.61% 0.44 0.46 0.44 3.01% -2.50% Uncontrolled HbA1c 0.096 0.096 0.00 0.00% 0.10 0.10 0.09 -6.62% -8.55% Absolute value >3% Absolute value 1-3%