Timothy P. Copeland

Epidemiologist specializing in causal inference, registry-based epidemiology, and statistical software development.

Department of Clinical Neuroscience, Karolinska Institutet

Publications

Stata Packages

1,294

Citations

h-index

i10-index

Featured Publications

View all →

Opioid use patterns and persistent use after hemodialysis access surgery

Copeland, T.P., et al.

Journal of Vascular Surgery, 2025

Journal PubMed Google Scholar

OBJECTIVE We evaluated the risk of persistent opioid use related to opioid pain management for hemodialysis access creation and identify patterns of opioid dosage and use.METHODS This retrospective cohort study included Medicare-enrolled opioid-naïve patients in the US Renal Data System who initiated hemodialysis for end-stage kidney disease between April 2015 and June 2019. Persistent opioid use was defined as an opioid prescription 90 to 180 days after surgery. The Cox proportional hazards model used censored patients at the time of hospitalization, subsequent hemodialysis access procedures, death, and end of Medicare enrollment. A mixed-effect logistic regression modeled opioid prescription, and a mixed-effect linear regression modeled opioid quantity in 5-mg hydrocodone equivalents.RESULTS A total of 28,404 patients (median age, 72 years; 25th, 75th percentile, 66, 79 years; 15,322 men [53.9%]) underwent index access creation, of whom 14,265 patients filled an opioid prescription at access creation (50.2%), with 3289 opioid prescriptions (23.1%) for ≥42 tablets of 5-mg hydrocodone equivalents (ie, ≥2 weeks prescription). Patients receiving ≥42 tablets of 5-mg hydrocodone equivalents (≥2 weeks) had 1.28 times the risk of persistent opioid use (95% confidence interval, 1.08-1.52) relative to those who did not fill an opioid prescription. In contrast, receipt of 20 or fewer tablets of 5-mg hydrocodone equivalents was not associated with an increased risk of persistent opioid use (hazard ratio 1.00; 95% confidence interval, 0.85-1.18).CONCLUSIONS Given the risk of persistent opioid use stemming from hemodialysis access procedures, the development of formal recommendations for conservative hemodialysis access pain management may be worth considering.

Dialysis facility staffing ratios and kidney transplant access among adolescents and young adults

Bicki, A.C., et al.

JAMA, 2024

Journal PubMed Google Scholar

Importance Patient to staff ratios vary across US dialysis facilities and have been associated with patient outcomes in older adultsObjective To determine whether patient to nurse or patient to social worker staff ratios are associated with access to kidney transplant for adolescents and young adults.Design, Setting, and Participants Retrospective cohort study including patients aged 12 to 30 years who started dialysis between 2005 and 2019 at 8490 US facilities according to the US Renal Data System, the national end-stage kidney disease registry.Exposures Time-updated quartile of patient to nurse and patient to social worker ratios at dialysis facilities.Main Outcomes and Measures Fine-Gray models were used to relate the exposure to the incidence of waitlisting and kidney transplant, accounting for the competing risk of death. Subgroup analysis by age at dialysis initiation (&lt;22 vs ≥22 years) was performed. Follow-up was censored in January 2020Results A total of 54 141 participants were included (median age, 25 years [IQR, 21-28]; 54.4% male; 4.3% of Asian race, 35.3% of non-Hispanic Black race). The median patient to staff ratios were 14.4 patients per nurse (IQR, 10.3-18.9) and 91.0 patients per social worker (IQR, 65.2-115.0). During a median follow-up of 2.6 years, 39.9% of patients (n = 21 598) received a transplant. In adjusted analysis, the highest (vs lowest) quartile of patient to nurse ratios was associated with 14% lower incidence of transplant (subhazard ratio [SHR], 0.86 [95% CI, 0.82-0.91]). The highest (vs lowest) quartile of patient to social worker ratios was associated with lower incidence of waitlisting (SHR, 0.95 [95% CI, 0.91-0.99]) and transplant (SHR, 0.85 [95% CI, 0.81-0.89]). For both staff ratios, there was an interaction with age at dialysis initiation, such that the association was more pronounced in patients starting dialysis at younger than 22 years (SHR, 0.71 [95% CI, 0.65-0.78] for the highest vs lowest quartile for nursing; SHR, 0.74 [95% CI, 0.68-0.80] for social work) compared with those 22 years and older (SHR, 1.00 [95% CI, 0.94-1.06] for nursing; SHR, 0.96 [95% CI, 0.91-1.02] for social work) for the outcome of transplantConclusions and Relevance Adolescents and young adults receiving care at dialysis facilities with higher patient to staff ratios had reduced access to waitlisting and transplant, particularly if they were younger than 22 years of age at dialysis initiation.

Surgeon Factors Have a Larger Effect on Vascular Access Type and Outcomes than Patient Factors

Copeland, T.P., Lawrence, P., & Woo, K.

Journal of Surgical Research, 2021

Journal PubMed Google Scholar

BACKGROUND AND OBJECTIVES Though patient factors are frequently linked to hemodialysis vascular access selection and outcomes, variability by surgeon and surgeon specialty may play a role as well. The objective of this study is to examine the extent to which individual surgeons influence selection of vascular access type, removal of tunneled hemodialysis catheter (THC), and repeat vascular access. DESIGN, SETTING, PARTICIPANTS, &MEASUREMENTS A national claims database was used to identify patients initiating hemodialysis via a THC between 2011 and 2017. Likelihood of repeat AVF/AVG was analyzed using mixed-effects logistic regression. Time from initial arteriovenous fistula (AVF)/graft (AVG) to THC removal and time to repeat AVF/AVG were analyzed using Weibull proportional hazard models. Individual surgeon identifier served as the random effect in all models.RESULTS 6,908 AVF/AVG met the inclusion criteria: 5366 (78%) AVF and 1,542 (22%) AVG. Surgeon specialty only had a significant influence on access type, with vascular surgeons having 26% greater odds of performing AVG compared to general surgeons (P = 0.006). Relative to the other independent variables, individual surgeon identifier had the greatest magnitude of effect on access type (median odds ratio, 2.36; 95% CI, 2.09-2.72). Individual surgeon identifier had the second greatest magnitude of effect likelihood of THC removal (median hazard ratio, 1.66; 95% CI, 1.58-1.77) and second access (median hazard ratio, 1.83; 95% CI, 1.66-2.05), in both cases second only to the effect of AVG, which was associated with greater likelihood of THC removal (hazard ratio 1.91; 95% CI, 1.77-2.07) and lower likelihood of second access (hazard ratio 0.44; 95% CI, 0.38-0.52).CONCLUSION Individual surgeons are associated with greater variation in vascular access type and likelihood of repeat access than surgeon specialty and measurable patient demographics/co-morbidities. Future research should focus on identifying which surgeon factors are associated with improved outcomes.

A deep learning model to predict a diagnosis of Alzheimer disease by using 18F-FDG PET of the brain

Ding, Y., et al.

Radiology, 2019

Journal PubMed Google Scholar

Purpose To develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease (AD), mild cognitive impairment, or neither at fluorine 18 (18F) fluorodeoxyglucose (FDG) PET of the brain and compare its performance to that of radiologic readersMaterials and Methods Prospective 18F-FDG PET brain images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (2109 imaging studies from 2005 to 2017, 1002 patients) and retrospective independent test set (40 imaging studies from 2006 to 2016, 40 patients) were collected. Final clinical diagnosis at follow-up was recorded. Convolutional neural network of InceptionV3 architecture was trained on 90% of ADNI data set and tested on the remaining 10%, as well as the independent test set, with performance compared to radiologic readers. Model was analyzed with sensitivity, specificity, receiver operating characteristic (ROC), saliency map, and t-distributed stochastic neighbor embeddingResults The algorithm achieved area under the ROC curve of 0.98 (95% confidence interval: 0.94, 1.00) when evaluated on predicting the final clinical diagnosis of AD in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity; P < .05). Saliency map demonstrated attention to known areas of interest but with focus on the entire brainConclusion By using fluorine 18 fluorodeoxyglucose PET of the brain, a deep learning algorithm developed for early prediction of Alzheimer disease achieved 82% specificity at 100% sensitivity, an average of 75.8 months prior to the final diagnosis. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Larvie in this issue.

Latest Publications

View all →

Association between estimated time with low glomerular filtration rate and access to transplant among youth with advanced CKD

Ku, E., Copeland, T.P., et al.

Pediatric Nephrology

Journal PubMed Google Scholar

Background Slower chronic kidney disease (CKD) progression allows more time for transplant preparation. Whether differences in CKD progression by race/ethnicity associate with preemptive or living donor transplantation in youth has not been well studiedMethods We examined the association between time spent with low eGFR (between 10–30 mL/min/1.73 m 2 ) and odds of preemptive or living donor transplantation among youth with CKD. eGFR was estimated using the bedside Schwartz (if < 18 years) and CKD-EPI 2021 equations (if ≥ 18 years) and the CKiD U25 equation in sensitivity analysis. Time spent with low eGFR was compared by race/ethnicityResults Among 333 youth with CKD (median age 11 years [IQR 7,14]), median time spent with low eGFR was 28.8 months. 77% were preemptively waitlisted, and 45% received preemptive transplantation (56% of White and 24% of Black youth). Black (vs. White) youth had shorter time with low eGFR (–6.5 months; 95% CI –11.5, –1.4). Time with low eGFR did not differ across other groups. Findings were similar using the U25 equation. Every additional year spent with low eGFR was associated with higher odds of preemptive (OR 1.45; 95% CI 1.24–1.70) and living donor transplantation (OR 1.42; 95% CI 1.21–1.67), but not preemptive waitlisting (OR 0.96; 95% CI 0.83–1.11) in unadjusted analysesConclusions Longer time spent with low eGFR is associated with greater odds of preemptive and living donor transplantation. Earlier transplant referral for all children, especially Black youth, may help improve access to preemptive and living donor transplantation. Graphical abstract

Oocyte maturation triggering in high responders in IVF treatment: a systematic review and network meta-analysis

Beebeejaun, Y., Copeland, T.P., et al.

Frontiers in Endocrinology

Journal PubMed Google Scholar

Objective The aim of this study was to compare and rank the efficacy and safety of four final oocyte maturation trigger strategies—human chorionic gonadotropin (hCG), gonadotropin-releasing hormone agonist (GnRHa), dual, and double trigger—in predicted high responders undergoing in vitro fertilization (IVF) with GnRH antagonist protocols, using a network meta-analysis (NMA) approachMethods A systematic search of MEDLINE, EMBASE, CENTRAL, clinical trial registries, and the Cochrane Database of Systematic Reviews was conducted through December 2024. Eligible studies were randomized controlled trials (RCTs) including high responders, defined by elevated antral follicle count, anti-Müllerian hormone, or estradiol levels. Studies using GnRHa triggers followed by fresh embryo transfer were included only if intensive luteal phase support was provided. Oocyte donation cycles, quasi-randomized designs, and trials lacking outcome data were excluded. Data extraction and risk of bias assessment were independently conducted by two reviewers. Study integrity was evaluated using the TRACT checklist. NMA was performed in STATA (v16), and treatment ranking was based on Surface Under the Cumulative Ranking curve (SUCRA)Results Seven high-quality RCTs comprising 632 women were included. There were no significant differences in the number of oocytes retrieved between GnRHa and hCG triggers (mean difference [MD] 1.08, 95% CI –1.06 to 3.22), dual and hCG (MD 0.61, 95% CI –1.53 to 2.74), or GnRHa and dual (MD 1.08, 95% CI –1.06 to 3.22). Similarly, there were no significant differences in mature oocyte yield, clinical pregnancy rate (CPR), or miscarriage rate across comparisons. However, GnRHa trigger significantly reduced the risk of moderate to severe ovarian hyperstimulation syndrome (OHSS) compared with hCG (RR 0.23, 95% CI 0.07–0.82). There were no significant differences in OHSS risk between dual and hCG (RR 0.28, 95% CI 0.05–1.64) or between GnRHa and dual (RR 0.28, 95% CI 0.05–1.64)Conclusion GnRHa, hCG, and dual triggers demonstrate similar efficacy in terms of oocyte yield, maturity, and clinical pregnancy rates in predicted high responders. The GnRHa trigger, however, offers a superior safety profile by significantly lowering the risk of OHSS. Larger multicenter RCTs are required to evaluate live birth outcomes and the potential role of the double trigger in this population. Systematic review registration https://www.crd.york.ac.uk/prospero/ , identifier CRD42022351423.

Understanding the relationship between severe donor Diabetes Insipidus, KDPI, and recipient kidney function

Calthorpe, L. et al.

Transplantation Direct

Journal PubMed Google Scholar

Background Severe diabetes insipidus (DI) causes dynamic changes in creatinine before organ donation. It is unknown whether terminal or peak creatinine is the optimal measure of graft quality among donors with severe DIMethods Adult donors after brain death were identified from the United Network for Organ Sharing database, years 2015–2022. Kidney Donor Profile Index (KDPI) was computed using both terminal creatinine (KDPI-Terminal) and peak creatinine (KDPI-Peak). Model fit for the prediction of graft failure and reduced glomerular filtration rate (GFR) was compared. Logistic regression was used to determine the association between KDPI group reclassification and graft functionResults Of 28 718 kidney transplants, 29% of donors met criteria for severe DI. No differences in estimated GFR (eGFR) or all-cause graft failure were observed on the basis of the presence of DI. Among donors with severe DI, KDPI-Peak improved prediction of reduced eGFR at 1 y (eGFR <60 mL/min/1.73 m 2 , eGFR <30 mL/min/1.73 m 2 ). Additionally, patients who received organs from donors with severe DI whose KDPI group would be reclassified on the basis of the use of peak rather than terminal creatinine had 2-fold increased odds of reduced eGFR at 1 y (eGFR <30 mL/min/1.73 m 2 ; odds ratio, 1.92; 95% confidence interval, 1.37-2.69)Conclusions This national analysis confirms that while the presence of severe DI alone is not associated with graft outcomes, the use of peak creatinine significantly improves risk stratification of donor graft quality. Precise risk stratification is critical for efforts to improve appropriate kidney allocation and increase confidence in organ function after transplant.

Association between time spent with low post-transplant glomerular filtration rate and access to relisting and re-transplantation

Ku, E., Copeland, T.P., et al.

Kidney Medicine

Journal PubMed Google Scholar

RATIONALE & OBJECTIVE For transplant recipients with low estimated glomerular filtration rate (eGFR), early preparation is needed to optimize the transition back to dialysis or retransplantation. Our objective was to determine if the amount of time spent with low eGFR was associated with a higher likelihood of relisting or retransplantation within 1 year of graft failure.STUDY DESIGN Retrospective cohort study.SETTING & PARTICIPANTS We included adults who developed graft failure between 2010 and 2021 according to the US Renal Data System.EXPOSURE We modeled the time spent with eGFR between 20 and 5 mL/min/1.73 m2 using available serum creatinine values reported annually (henceforth referred to as time with low eGFR).OUTCOME Relisting or retransplantation within 1 year of graft failure.ANALYTICAL APPROACH We used Fine-Gray models accounting for the competing risk of death to determine the association between time spent with low eGFR (as a categorical variable) and outcomes of interest.RESULTS We included 14,680 patients; median age at graft failure was 53 years (IQR, 42-61). Within 1 year of graft failure, ∼18% (n = 2,690) of patients were relisted and 13% (n = 1,854) were retransplanted. In adjusted models, spending 2-4 years (SHR, 1.43; 95% confidence interval [CI], 1.16-1.75) or >4 years (SHR, 1.28; 95% CI, 1.04-1.57) with low eGFR was associated with a higher likelihood of relisting (vs spending <2 years with low eGFR). Similarly, spending 2-4 years (SHR, 1.37; 95% CI, 1.09-1.72) or >4 years (SHR, 1.73; 95% CI, 1.37-2.19) with low eGFR was also associated with a higher likelihood of retransplantation within 1 year of graft failure.LIMITATIONS Residual confounding; observational nature of study.CONCLUSIONS Timely preparation of patients with low eGFR for retransplantation could increase patients' likelihood of early relisting and retransplantation. For transplant recipients with low kidney function, the vast majority of patients will return to dialysis after graft failure, and many will not receive a second transplant. Our objective in this study was to determine if a longer time spent with low kidney function before the return to dialysis was associated with a higher likelihood of relisting or retransplantation in transplant recipients using national registry data. We found that spending more than 2 years with low eGFR was associated with a much higher likelihood of relisting and retransplantation than spending less than 2 years with low eGFR. Early preparation for retransplantation is needed to improve transplant patients’ likelihood of receiving a second transplant.

Letter concerning Comparative Safety Profiles of Ocrelizumab and Rituximab in Multiple Sclerosis Treatment

Copeland, T.P.

Annals of Neurology

Journal PubMed Google Scholar

Research Areas

Causal Inference Registry-Based Research Neurology Clinical Epidemiology Fertility Medicine Vascular Surgery Radiology Oncology Statistical Software Target Trial Emulation Pharmacoepidemiology