Registry-Based Research CSI Seminar 25/05/2016 Magnus Ekstrm MD, - PowerPoint PPT Presentation

Registry-Based Research CSI Seminar 25/05/2016 Magnus Ekström MD, PhD Department of Respiratory Medicine and Allergology Clinical Sciences Lund University, Lund, Sweden pmekstrom@gmail.com

Karlskrona

Research

Focus of the Talk • Examples of registry-based studies • Data sources in Sweden • The process to obtain data

Linkage Drugs Health care Swedevox • Home oxygen • Ventilator Mortality • CPAP Diagnoses

Ekström MP, Franklin KA, Ström KE. Chest 2010;137 Material: Patients starting LTOT for COPD 1987 – 2004 in Swedevox. Causes of Death Register. 2005 1987 Lund University / Faculty of Medicine / Department of Respiratory Medicine & Allergology

Excess mortality for the main causes of death Women Men Diagnosis entity Obs / Exp SMR (95% CI) SMR (95% CI) Obs/Exp 292.2 (279.2 – 305.8) COPD 1860 / 6.4 155.4 (148.3-162.8) 1762 / 11.3 5.2 (4.5 – 6.0) 3.8 (3.4 – 4.3) Ischemic heart disease 191 / 37.0 280 / 73.6 15.8 (13.2 – 19.0) 8.7 (7.3 – 10.3) Lung cancer 116 / 7.3 123 / 14.2 9.2 (7.3 – 11.6) 4.1 (3.1 – 5.4) Heart failure 70 / 7.6 50 / 12.2 1.7 (1.3 – 2.3) 1.0 (0.7 – 1.3) Stroke 45 / 26.8 34 / 35.7 5.9 (3.5 – 9.8) 3.1 (2.1 – 4.7) Aortic aneurysm 15 / 2.6 22 / 7.1 4.7 (2.8 – 8.2) 6.7 (4.3 – 10.3) Venous thromboembolism 13 / 2.7 20 / 3.0 68.9 (42.9 – 110.8) 23.0 (11.0 – 48.2) Tuberculosis 17 / 0.3 7 / 0.3 1.4 (0.7 – 2.7) 1.1 (0.6 – 1.8) Pneumonia 8 / 5.9 13 / 12.3 2.4 (1.4 – 4.1) 1.2 (0.6 – 2.4) Colon cancer 13 / 5.4 8 / 6.8 Total SMR Women 12.0 (95% CI, 11.6 - 12.5) Men 7.4 (95% CI, 7.1 - 7.6) Lund University / Faculty of Medicine / Department of Respiratory Medicine & Allergology

Cardiovascular drugs and survival time Time-dependent analysis of drug exposure E 1 = 0 E 2 = 0.5 E 3 = 1 E 4 = 0 E 5 = 0 E 6 = 0.4 E 7 = 1 End of follow-up Start of follow-up Adjusted for age, sex, hypoxemia, PaCO 2 air, body mass index, WHO performance status och comorbidity (anemia, diabetes mellitus, renal failure and cardiovascular disease). Lund University / Faculty of Medicine / Department of Respiratory Medicine & Allergology

Ekström M et al. AJRCCM 2013; 187

Safety of Benzodiazepines and Opioids COPD, N = 2,449 Hospitalization Death from all causes 3 months Dose of Adjusted for : age, sex, FEV 1 , blood medication gases, BMI, WHO performance status, morbidities, prev hospitalizations, and oral glucocorticoids.

Swedish Registry of Palliative Care Thorax 2016; 71

Combined Two Quality Registries SRPC=Swedish Registry of Palliative Care

Ahmadi et al. Thorax 2016; 71

Ahmadi et al. Thorax 2016; 71

Ahmadi et al. Thorax 2016; 71

Data Sources in Sweden • Health care quality registries (QR) • Governmental ”public” registries • Study databases

Quality Registries (QR) Emilsson. J Intern Med 2015; 277 • Single data source • Combine (QR, public reg, study) • Recruitment base (observ or intervent)

No. Quality Registries Per Discipline Emilsson. J Intern Med 2015; 277

Emilsson. J Intern Med 2015; 277

Swedish Cancer Registry https://www.socialstyrelsen.se/register/halsodataregister/cancerregistret/inenglish • 1958 - (4% missing diagnoses) • All malignant disease nationwide • Reporting: laboratory + clinician • PID, age, sex, place of residence • Diagnosis type and date • Tumour site, hist type, stage (2004 -) • + more disease-specific data in sub-registries … (INCA registries)

Swedish Registry of Palliative Care http://palliativ.se/ • 2011 – • About 60% of deaths in Sweden • > 85% of deaths in people with cancer • Data on last 7 days of life • Reported by staff (nurses) within 7 days after death • Unit, resources, preference for place of death • Presence of symptoms, level of relief, validated assessment • ’As needed ’ medications Emilsson. J Intern Med 2015; 277

End of Life Questionnaire ….

How are QRs Governed? • Funded by the government (liberal ----> ?) • Responsible authority (Country Council) • QR does not own but nurture the data • Social security number • Public Access to Information and Secrecy Act • … but formal process not devoid of bureaucracy

How QRs are run Government (Sw Assoc of Local Authorities and Regions); oversight/strategy Responsible Steering commitee Registry County Council Data management unit Care units

How are Public Registers Governed? • National Board of Health and Welfare (Socialstyrelsen) • Mortality : ≈ 100% coverage; • Hosp and diagn: > 90% for somatic; • Drugs: 100% of outpatient prescriptions • Standardized process to request and obtain data • Data provided de-identified • Study ID instead of social security number • Can save key to social security number > 3 yrs

The Process to Obtain Data 1. Quality Registry as Lone Data Source Regional Ethics Identify authority/region Project plan Data Registry County Council

The Process 2. Aggregated public data (rates of hospitalization, diagnoses, mortality …) Project plan Data No ethics approval needed

The Process 3. Combination of Registries / Study Data Saves ID key x 1 No ID Regional Ethics Identify authority/region ID Project plan Data Registry, x 1..n study County Council

Potential Advantages Of Registry-Based Studies • Large sample sizes and high precision • Capture rare conditions or events • Routinely collected, feasability, cost-effective • Prospective data (exposures before outcomes) • Hard endpoints (death) • High rate of follow-up (depending on outcome) • Representativeness, generalizability

Potential Limitations Of Registry-Based Studies • Bias (generally unaffected by increasing sample size) • Data quality • ”Real -world studies”?

Registry-Based RCT (R-RCT) Outcome Intervention Assessment Registry Randomization • Registry • • Research unit Consent Comparison • Eligibility James et al. Nat Rev Cardiol 2015; 12

The first TASTE of the R-RCT

Included 85% of eligible patients 61% of all with STEMI in Sweden Fröbert et al. NEJM 2013; 369

Mortality at 30 days Complete follow-up Fröbert et al. NEJM 2013; 369

Suitable in Advanced Disease? • Facilitate recruitment • Follow-up • Cost-effective • Challenges • Suitable outcomes (PROMs) • Assessment of adverse events Fröbert et al. NEJM 2013; 369

R-RCT of Long-Term Oxygen Registry data Death LTOT 24 h/d Hospitalizations Diagnoses Randomization Prescribed drugs • Eligibility Questionnaire LTOT 15 h/d PROMs • Consent

Last Words • Rich network of registry data in Sweden • Possibilities for collaborations • Needs ideas, time, some bucks • Development of R-RCTs

Thanks a lot! pmekstrom@gmail.com