Effects of aggregation of drug and diagnostic codes on the performance of the high-dimensional propensity score algorithm: an empirical example

BMC Medical Research Methodology

Til Stürmer¹  J Bradley Layton¹  Kathleen J Beach²  Victor J Schoenbach¹  M Alan Brookhart¹  Charles Poole¹  Hoa V Le² 

[1] Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, USA;GlaxoSmithKline, Research Triangle Park, Durham, North Carolina, USA

关键词: Rare outcome;    Small sample;    Propensity score;    Infrequent exposure;    Confounding by indication;    Clinical classification software;    Anatomical therapeutic chemical classification;    Aggregation;   

Others  :  866570
DOI  :  10.1186/1471-2288-13-142

 received in 2013-08-05, accepted in 2013-11-14,  发布年份 2013

【 摘 要 】

Background

The High-Dimensional Propensity Score (hd-PS) algorithm can select and adjust for baseline confounders of treatment-outcome associations in pharmacoepidemiologic studies that use healthcare claims data. How hd-PS performance is affected by aggregating medications or medical diagnoses has not been assessed.

Methods

We evaluated the effects of aggregating medications or diagnoses on hd-PS performance in an empirical example using resampled cohorts with small sample size, rare outcome incidence, or low exposure prevalence. In a cohort study comparing the risk of upper gastrointestinal complications in celecoxib or traditional NSAIDs (diclofenac, ibuprofen) initiators with rheumatoid arthritis and osteoarthritis, we (1) aggregated medications and International Classification of Diseases-9 (ICD-9) diagnoses into hierarchies of the Anatomical Therapeutic Chemical classification (ATC) and the Clinical Classification Software (CCS), respectively, and (2) sampled the full cohort using techniques validated by simulations to create 9,600 samples to compare 16 aggregation scenarios across 50% and 20% samples with varying outcome incidence and exposure prevalence. We applied hd-PS to estimate relative risks (RR) using 5 dimensions, predefined confounders, ≤ 500 hd-PS covariates, and propensity score deciles. For each scenario, we calculated: (1) the geometric mean RR; (2) the difference between the scenario mean ln(RR) and the ln(RR) from published randomized controlled trials (RCT); and (3) the proportional difference in the degree of estimated confounding between that scenario and the base scenario (no aggregation).

Results

Compared with the base scenario, aggregations of medications into ATC level 4 alone or in combination with aggregation of diagnoses into CCS level 1 improved the hd-PS confounding adjustment in most scenarios, reducing residual confounding compared with the RCT findings by up to 19%.

Conclusions

Aggregation of codes using hierarchical coding systems may improve the performance of the hd-PS to control for confounders. The balance of advantages and disadvantages of aggregation is likely to vary across research settings.

【 授权许可】

   
2013 Le et al.; licensee BioMed Central Ltd.

【 图 表 】

【 参考文献 】

• [1]Davis RL, Kolczak M, Lewis E, Nordin J, Goodman M, Shay DK, Platt R, Black S, Shinefield H, Chen RT: Active surveillance of vaccine safety: a system to detect early signs of adverse events. Epidemiology 2005, 16(3):336-341.
• [2]Brown JS, Kulldorff M, Chan KA, Davis RL, Graham D, Pettus PT, Andrade SE, Raebel MA, Herrinton L, Roblin D, Boudreau D, Smith D, Gurwitz JH, Gunter MJ, Platt R: Early detection of adverse drug events within population-based health networks: application of sequential testing methods. Pharmacoepidemiol Drug Saf 2007, 16(12):1275-1284.
• [3]Lieu TA, Kulldorff M, Davis RL, Lewis EM, Weintraub E, Yih K, Yin R, Brown JS, Platt R: Vaccine Safety Datalink Rapid Cycle Analysis Team Real-Time vaccine safety surveillance for the early detection of adverse events. Med Care 2007, 45(10):S89-S95.
• [4]Rubin DB: Estimating causal effects from large data sets using the propensity score. Ann Intern Med 1997, 127:757-763.
• [5]Stürmer T, Schneeweiss S, Brookhart MA, Rothman KJ, Avorn J, Glynn RJ: Analytic strategies to adjust confounding using exposure propensity scores and disease risk scores: nonsteroidal antiinflammatory drugs and short-term mortality in the elderly. Am J Epidemiol 2005, 161(9):891-898.
• [6]Stürmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S: A review of applications of propensity score methods showed increased use but infrequently different estimates compared with other methods. J Clin Epidemiol 2006, 59:437-447.
• [7]Glynn RJ, Schneeweiss S, Stürmer T: Indications for propensity scores and review of their use in pharmacoepidemiology. Basic Clin Pharmacol Toxicol 2006, 98:253-259.
• [8]Perkins SM, Tu W, Underhill MG, Zhou XH, Murray MD: The use of propensity scores in pharmacoepidemiologic research. Phamacoepidemiol Drug Saf 2000, 9:93-101.
• [9]Robins JM, Mark SD, Newey WK: Estimating exposure effects by modeling the expectation of exposure conditional on confounders. Biometrics 1992, 48:479-495.
• [10]Rubin DB: On principles for modeling propensity score in medical research. Phamacoepidemiol Drug Saf 2005, 14:227-238.
• [11]Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer T: Variable selection for propensity score models. Am J Epidemiol 2006, 163(12):1149-1156.
• [12]Greenland S: Quantifying biases in causal models: classical confounding vs. collider-stratification bias. Epidemiology 2003, 14(3):300-306.
• [13]Brookhart MA, Stürmer T, Glynn RJ, Rassen J, Schneeweiss S: Confounding control in healthcare database research: challenges and potential approaches. Med Care 2010, 48(6):S114-S120.
• [14]Seeger JD, Kurth T, Walker AM: Use of propensity score technique to account for exposure-related covariates: an example and lesson. Med Care 2007, 45:S143-S148.
• [15]Johannes CB, Koro CE, Quinn SG, Cutone JA, Seeger JD: The risk of coronary heart disease in type 2 diabetic patients exposed to thiazolidinediones compared to metformin and sulfonylurea therapy. Pharmacoepidemiol Drug Saf 2007, 16:504-512.
• [16]Schneeweiss S, Rassen JA, Glynn RJ, Avorn J, Mogun H, Brookhart MA: High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology 2009, 20(4):512-522.
• [17]Le VH, Beach JK, Powell G, Pattishall E, Ryan P, Mera RM: Performance of a semi-automated approach for risk estimation using a common data model for longitudinal healthcare databases. Stat Methods Med Res 2013, 22(1):97-112.
• [18]Toh S, Rodríguez AGL, Hernán AM: Confounding adjustment via a semi‒automated high‒dimensional propensity score algorithm: an application to electronic medical records. Pharmacoepidemiology and drug safety 2011, 20:849-857.
• [19]Bross ID: Spurious effects from an extraneous variable. J Chronic Dis 1966, 19(6):637-647.
• [20]Brookhart MA, Wang PS, Solomon DH, Schneeweiss S: Evaluating short-term drug effects using a physician-specific prescribing preference as an instrumental variable. Epidemiology 2006, 17(3):268-275.
• [21]Silverstein FE, Faich G, Goldstein JL, Simon LS, Pincus T, Whelton A, Makuch R, Eisen G, Agrawal NM, Stenson WF, Burr AM, Zhao WW, Kent JD, Lefkowith JB, Verburg KM, Geis GS: Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: A randomized controlled trial Celecoxib Long-term Arthritis Safety Study. JAMA 2000, 284(10):1247-1255.
• [22]Singh G, Fort JG, Goldstein JL, Levy RA, Hanrahan PS, Bello AE, Andrade-Ortega L, Wallemark C, Agrawal NM, Eisen GM, Stenson WF: Triadafilopoulos G; SUCCESS-I Investigators Celecoxib versus naproxen and diclofenac in osteoarthritis patients: SUCCESS-I Study. Am J Med 2006, 119(3):255-266.
• [23]Deeks JJ, Smith LA, Bradley MD: Efficacy, tolerability, and upper gastrointestinal safety of celecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic review of randomised controlled trials. BMJ 2002, 325(7365):619.
• [24]Moore RA, Derry S, Makinson GT, McQuay HJ: Tolerability and adverse events in clinical trials of celecoxib in osteoarthritis and rheumatoid arthritis: systematic review and meta-analysis of information from company clinical trial reports. Arthritis Res Ther 2005, 7(3):R644-R665. BioMed Central Full Text
• [25]Goldstein JL, Silverstein FE, Agrawal NM, Hubbard RC, Kaiser J, Maurath CJ, Verburg KM, Geis GS: Reduced risk of upper gastrointestinal ulcer complications with celecoxib, a novel COX-2 inhibitor. Am J Gastroenterol 2000, 95(7):1681-1690.
• [26]Goldstein JL: Significant upper gastrointestinal events associated with conventional NSAID versus celecoxib. J Rheumatol Suppl 2000, 60:25-28.
• [27]Ray WA: Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003, 158(9):915-920.
• [28]MarketScan® Commercial Claims and Encounters of Thomson Reuters Healthcare 2011. [http://thomsonreuters.com/products_services/healthcare/healthcare_products/a-z/marketscan_research_analytics/ webcite]
• [29]Raiford DS, Perez Gutthann S, Garcia Rodriguez LA: Positive predictive value of ICD-9 codes in the identification of cases of complicated peptic ulcer disease in the Saskatchewan hospital automated database. Epidemiology 1996, 7:101-104.
• [30]The Anatomical Therapeutic Chemical (ATC) classification 2012. [http://www.whocc.no/ webcite]
• [31]Hoskins H Jr, Hildebrand P, Lum F: The American Academy of Ophthalmology Adopts SNOMED CT as Its Official Clinical Terminology. Ophthalmology 2008, 115(2):225-226.
• [32]Clinical Classifications Software (CCS) for ICD-9-CM 2012. [http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp webcite]
• [33]García Rodríguez LA, Jick H: Risk of upper gastrointestinal bleeding and perforation associated with individual non‒steroidal anti‒inflammatory drugs. Lancet 1994, 343:769-772.
• [34]Gutthann SP, García Rodríguez LA, Raiford DS: Individual nonsteroidal antiinflammatory drugs and other risk factors for upper gastrointestinal bleeding and perforation. Epidemiology 1997, 8:18-24.
• [35]Hernández‒Díaz S, García Rodríguez LA: Association between nonsteroidal anti‒inflammatory drugs and upper gastrointestinal tract bleeding/perforation: an overview of epidemiologic studies published in the 1990s. Arch Intern Med 2000, 160:2093-2099.
• [36]Greenland S, Schwartzbaum JA, Finkle WD: Problems due to small samples and sparse data in conditional logistic regression analysis. Am J Epidemiol 2000, 151(5):531-539.
• [37]Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR: A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996, 49(12):1373-1379.
• [38]Rassen JA, Glynn JR, Brookhart MA, Schneeweiss S: Covariate selection in high-dimensional propensity score analyses of treatment effects in small samples. Am J Epidemiol 2011, 173(12):1404-1413.