PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 11763542

Influence function based variance estimation and missing data issues in case-cohort studies.

S D Mark·H Katki
View original source ↗Export metadata ↓

Abstract

Recognizing that the efficiency in relative risk estimation for the Cox proportional hazards model is largely constrained by the total number of cases, Prentice (1986) proposed the case-cohort design in which covariates are measured on all cases and on a random sample of the cohort. Subsequent to Prentice, other methods of estimation and sampling have been proposed for these designs. We formalize an approach to variance estimation suggested by Barlow (1994), and derive a robust variance estimator based on the influence function. We consider the applicability of the variance estimator to all the proposed case-cohort estimators, and derive the influence function when known sampling probabilities in the estimators are replaced by observed sampling fractions. We discuss the modifications required when cases are missing covariate information. The missingness may occur by chance, and be completely at random; or may occur as part of the sampling design, and depend upon other observed covariates. We provide an adaptation of S-plus code that allows estimating influence function variances in the presence of such missing covariates. Using examples from our current case-cohort studies on esophageal and gastric cancer, we illustrate how our results our useful in solving design and analytic issues that arise in practice.

Explore related subjects

Cohort StudiesConfounding Factors, EpidemiologicEsophageal NeoplasmsHealth Care RationingHumansProportional Hazards ModelsResearch DesignStomach NeoplasmsUnited StatesJournal Article

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

S D Mark, H Katki. 2001. Influence function based variance estimation and missing data issues in case-cohort studies.. https://doi.org/10.1023/a%3A1012533130596

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2002

Laparoscopic versus open donor nephrectomy.

BACKGROUND: As compared with open donor nephrectomy (OpenDN), laparoscopic donor nephrectomy (LapDN) offers donors more rapid recovery and recipients equivalent graft function, but LapDN costs remain greater. This study compared LapDN and OpenDN with cost-utility analysis. METHODS: Utilities were assessed with time trade-off, probabilities derived from systematic review of the literature and the costs derived from 27 OpenDN and 34 LapDN patients treated contemporaneously. A societal perspective was taken. Lost employment costs were included. An incremental cost-effectiveness ratio (ICER) was calculated with best- and worst-case scenarios for confidence intervals. Sensitivity analyses assessed robustness. RESULTS: LapDN costs are lower (11,170.71 dollars vs 12,631.91 dollars), whereas quality of life (QOL) is superior (0.7247 vs 0.6585 quality-adjusted life years [QALY], rendering LapDN a dominant strategy. The model was robust to all variables, and LapDN remained dominant from a payer perspective. In a worst-case scenario, the ICER for LapDN was at most 2,231.61 dollars per QALY. CONCLUSIONS: LapDN offers improved QOL at lower costs, despite the fact that this analysis included patients treated during the learning curve of LapDN at our institution. By potentially increasing organ donor rates, LapDN may be further cost saving by decreasing the number of patients receiving dialysis.

Cohort Studies↗
PubMed2002

Health and economic outcomes of vancomycin-resistant enterococci.

BACKGROUND: The health and economic impact of vancomycin-resistant enterococci has not been quantified. METHODS: A retrospective matched cohort study was conducted comparing the outcomes of patients with vancomycin-resistant enterococci (cases) with those of control subjects matched for length of hospital stay until inclusion in the cohort, hospital location, and calendar date. The propensity to be a vancomycin-resistant enterococci case was modeled based on patient characteristics, and included in multivariable models to adjust for confounding. Analyses included the following: (1) conditional logistic regression for mortality, surgery, intensive care unit admission, and discharge to long-term care; (2) linear regression for the logarithm of cost; and (3) accelerated failure time model for length of stay. RESULTS: A total of 233 cases were compared with 647 controls. Groups were similar in age (mean, 62 years), sex (female, 47%), and length of stay before inclusion in the cohort (mean, 8.1 days), but differed in primary diagnosis and comorbidities, past infection or colonization with methicillin sodium-resistant Staphylococcus aureus or Clostridium difficile, and treatment with cephalosporins or metronidazole. These variables were included in the propensity score, which had good to excellent prediction. Outcomes for cases vs controls and adjusted risks (relative risks [RRs]) were as follows: (1) case fatality rate, 17% vs 6% (RR, 2.13; P =.04); (2) length of stay after inclusion in the cohort, 15.1 vs 8.5 days (RR, 1.73; P<.001); (3) hospital costs, $52 449 vs $31 915 (RR, 1.40; P<.001); (4) surgery after inclusion in the cohort, 18% vs 10% (RR, 2.74; P =.001); (5) intensive care unit admission after inclusion in the cohort, 25% vs 14% (RR, 3.47; P<.001); and (6) transfer to an institution, 51% vs 35% (RR, 2.01; P =.001). CONCLUSION: Compared with a matched hospital population, a population with vancomycin-resistant enterococci was associated with severe adverse outcomes: increased mortality, morbidity, and costs.

Cohort Studies↗
PubMed2002

Risk of renal stone events in primary hyperparathyroidism before and after parathyroid surgery: controlled retrospective follow up study.

AIM: To study the risk of renal stone episodes and risk factors for renal stones in primary hyperparathyroidism before and after surgery. DESIGN: Register based, controlled retrospective follow up study. SETTING: Tertiary hospitals in Denmark. PARTICIPANTS: 674 consecutive patients with surgically verified primary hyperparathyroidism. Each patient was compared with three age- and sex-matched controls randomly drawn from the background population. Hospital admissions for renal stone disease were compared between patients and controls. Risk factors for renal stones among patients were assessed. MAIN OUTCOME MEASURES: Number of renal stone episodes; comparison of hospital admissions for renal stones in patients and controls; assessment of risk factors for renal stones. RESULTS: Relative risk of a stone episode was 40 (95% confidence interval 31 to 53) before surgery and 16 (12 to 23) after surgery. Risk was increased 10 years before surgery, and became normal more than 10 years after surgery. Stone-free survival 20 years after surgery was 90.4% in patients and 98.7% in controls (risk difference 8.3%, 4.8% to 11.7%). Patients with preoperative stones had 27 times the risk of postoperative stone incidents than controls. Before surgery, males had more stone episodes than females and younger patients had more stone episodes than older patients. Neither parathyroid pathology, weight of removed tissue, plasma calcium levels, nor skeletal pathology (fractures) influenced the risk of renal stones. After surgery, younger age, preoperative stones and ureteral strictures were significant risk factors for stones. CONCLUSIONS: The risk of renal stones is increased in primary hyperparathyroidism and decreases after surgery. The risk profile is normal 10 years after surgery. Preoperative stone events increase the risk of postoperative stones. Stone formers and non-stone formers had the same risk of skeletal complications.

Cohort Studies↗