Confounding in observational studies based on large health care databases: problems and potential solutions – a primer for the clinician

Figures & data

Table 1 Different types of confounders and potential solutions on how to control for these in observational studies based on health care databases

Type of confounders	Examples	Strategy
Measured	Age and sex	Restriction Matching Stratification Standardization Regression analysis Propensity scores
Unmeasured but measurable	Smoking Body mass index Disease severity	External adjustment Proxy measures Imputation
Unmeasurable	Frailty	Self-controlled design Instrumental variable Mendelian randomization Active comparator Regression discontinuity design Sensitivity analyses

Table 2 Examples of instrumental variables recently used in published studies

Association examined	Instrumental variable	Reference
NSAID treatment for persistent ductus arteriosus and mortality and moderate/severe bronchopulmonary dysplasia	Institutional variation in NSAID treatment frequency	Slaughter et al^Citation53
Nonparental childcare attendance and childhood obesity	The number of relatives who live close to the family	Isong et al^Citation54
Use of NIV in patients with pneumonia and 30-day mortality	Differential distance, ie, the difference between 1) the distance from a patient’s residence to the nearest high NIV use hospital and 2) the distance from a patient’s residence to the nearest hospital of any type	Valley et al^Citation55
Psychosocial assessment of patients in the emergency department and risk of repeat self-harm	Time of day of hospital presentation	Carroll et al^Citation56
Second-generation versus third-generation oral contraceptives and risk of venous thromboembolism	Proportion of prescriptions for third-generation oral contraceptives by the general practitioner in the year preceding the current prescription	Boef et al^Citation57
Readmission destination and risk of mortality after major surgery	Regional index hospital readmission rates	Brooke et al^Citation58

Abbreviations: NIV, noninvasive ventilation; NSAID, non-steroidal anti-inflammatory drug.

Figure 1 First CD4 count and mortality hazard rate in an HIV-positive population.

Notes: Predicted hazards are displayed as solid lines. Dashed line shows extrapolated prediction if all patients were treatment eligible at first CD4 count. Dots are hazards predicted for CD4 count bins of width 10 cells. Copyright © 2014 by Lippincott Williams & Wilkins. Figure originally published by Bor et al. Regression discontinuity designs in epidemiology: causal inference without randomized trials. Epidemiology 2014;25:729–737.⁴⁶

Figure 2 Causal diagram showing an ideal negative control exposure B for use in evaluating studies of the causal relationship between exposure A and outcome Y.

Notes: B should ideally have the same incoming arrows as A. U is the set of uncontrolled confounders. L is assumed measured and controlled for. Modified with permission from Lipsitch et al. Negative controls: a tool for detecting confounding and bias in observational studies.in Epidemiology 2010;21(3):383–388. https://www.ncbi.nlm.nih.gov/pubmed/20335814.⁴⁹

SlaughterJLReaganPBNewmanTBKlebanoffMAComparative effectiveness of nonsteroidal anti-inflammatory drug treatment vs no treatment for patent ductus arteriosus in preterm infantsJAMA Pediatr Epub201713

 Web of Science ®Google Scholar

IsongIARichmondTKawachiIAvendanoMChildcare attendance and obesity riskPediatrics20161385e2016153927940780

 PubMed Web of Science ®Google Scholar

ValleyTSWalkeyAJLindenauerPKWienerRSCookeCRAssociation between noninvasive ventilation and mortality among older patients with pneumoniaCrit Care Med Epub2016105

 Google Scholar

CarrollRMetcalfeCSteegSPsychosocial assessment of self-harm patients and risk of repeat presentation: an instrumental variable analysis using time of hospital presentationPLoS One2016112e014971326918579

 PubMed Web of Science ®Google Scholar

BoefAGSouvereinPCVandenbrouckeJPInstrumental variable analysis as a complementary analysis in studies of adverse effects: venous thromboembolism and second-generation versus third-generation oral contraceptivesPharmacoepidemiol Drug Saf201625331732426749045

 PubMed Web of Science ®Google Scholar

BrookeBSGoodneyPPKraissLWGottliebDJSamoreMHFinlaysonSRReadmission destination and risk of mortality after major surgery: an observational cohort studyLancet2015386999688489526093917

 PubMed Web of Science ®Google Scholar

BorJMoscoeEMutevedziPNewellMLBarnighausenTRegression discontinuity designs in epidemiology: causal inference without randomized trialsEpidemiology201425572973725061922

 PubMed Web of Science ®Google Scholar

LipsitchMTchetgenTECohenTNegative controls: a tool for detecting confounding and bias in observational studiesEpidemiology201021338338820335814

 PubMed Web of Science ®Google Scholar