Principled Approaches to Missing Data in Epidemiologic Studies

Perkins, Neil J.; Cole, Stephen R.; Harel, Ofer; Tchetgen, Eric J. Tchetgen; Sun, Baoluo; Mitchell, Emily M.; Schisterman, Enrique F.

doi:10.1093/aje/kwx348

Cited by 193 publications

(158 citation statements)

References 28 publications

Supporting

Mentioning

157

Contrasting

Order By: Relevance

“…Because drop‐out or death can lead to informative missing data with respect to outcomes, we used baseline covariates for inverse probability weighting to reduce bias and boost efficiency . Results without weighting were similar to weighted results.…”

Section: Methodsmentioning

confidence: 85%

Symptom burden among older breast cancer survivors: The Thinking and Living With Cancer (TLC) study

Mandelblatt

Zhai

Ahn

et al. 2019

Cancer

View full text Add to dashboard Cite

Background Little is known about longitudinal symptom burden, its consequences for well‐being, and whether lifestyle moderates the burden in older survivors. Methods The authors report on 36‐month data from survivors aged ≥60 years with newly diagnosed, nonmetastatic breast cancer and noncancer controls recruited from August 2010 through June 2016. Symptom burden was measured as the sum of self‐reported symptoms/diseases as follows: pain (yes or no), fatigue (on the Functional Assessment of Cancer Therapy [FACT]‐Fatigue scale), cognitive (on the FACT‐Cognitive scale), sleep problems (yes or no), depression (on the Center for Epidemiologic Studies Depression scale), anxiety (on the State‐Trait Anxiety Inventory), and cardiac problems and neuropathy (yes or no). Well‐being was measured using the FACT‐General scale, with scores from 0 to 100. Lifestyle included smoking, alcohol use, body mass index, physical activity, and leisure activities. Mixed models assessed relations between treatment group (chemotherapy with or without hormone therapy, hormone therapy only, and controls) and symptom burden, lifestyle, and covariates. Separate models tested the effects of fluctuations in symptom burden and lifestyle on function. Results All groups reported high baseline symptoms, and levels remained high over time; differences between survivors and controls were most notable for cognitive and sleep problems, anxiety, and neuropathy. The adjusted burden score was highest among chemotherapy‐exposed survivors, followed by hormone therapy‐exposed survivors versus controls (P < .001). The burden score was related to physical, emotional, and functional well‐being (eg, survivors with lower vs higher burden scores had 12.4‐point higher physical well‐being scores). The composite lifestyle score was not related to symptom burden or well‐being, but physical activity was significantly associated with each outcome (P < .005). Conclusions Cancer and its treatments are associated with a higher level of actionable symptoms and greater loss of well‐being over time in older breast cancer survivors than in comparable noncancer populations, suggesting the need for surveillance and opportunities for intervention.

show abstract

Section: Methodsmentioning

confidence: 85%

Symptom burden among older breast cancer survivors: The Thinking and Living With Cancer (TLC) study

Mandelblatt

Zhai

Ahn

et al. 2019

Cancer

View full text Add to dashboard Cite

show abstract

“…We conducted two sensitivity analyses to evaluate the multiple imputed data . First, to evaluate for baseline differences in patients lost and not lost to follow‐up at 6 months, we compared patient and clinical characteristics (age, male sex, abnormal initial EMS GCS score, Charlson Comorbidity Index, trauma center transport, and presence of traumatic intracranial hemorrhage) of patients with and without GOS‐E scores at 6 months.…”

Section: Methodsmentioning

confidence: 99%

The Association of Trauma Center Transport and Long‐term Functional Outcomes in Head‐injured Older Adults Transported by Emergency Medical Services

Nishijima

Gaona

Faul

et al. 2020

Academic Emergency Medicine

View full text Add to dashboard Cite

Objective It is unclear whether trauma center care is associated with improved outcomes in older adults with traumatic brain injury (TBI) compared to management at nontrauma centers. Our primary objectives were to describe the long‐term outcomes of older adults with TBI and to evaluate the association of trauma center transport with long‐term functional outcome. Methods This was a prospective, observational study at five emergency medical services (EMS) agencies and 11 hospitals representing all 9‐1‐1 transfers within a county. Older adults (≥55 years) with TBI (defined as closed head injury associated with loss of consciousness and/or amnesia, abnormal Glasgow Coma Scale [GCS] score, or traumatic intracranial hemorrhage) and transported by EMS from August 2015 to September 2016 were eligible. EMS providers completed standardized data forms and emergency department (ED) and hospital data were abstracted. Functional outcomes were measured using the Extended Glasgow Outcome Scale (GOS‐E) at 3‐ and 6‐month intervals by telephone follow‐up. Reasons for disabilities were coded as due to head injury, due to illness or injury to other part of body, or due to a mixture of both. To evaluate the association of trauma center transport and functional outcomes, we conducted multivariate ordinal logistic regression analyses on multiple imputed data for 1) all patients with TBI and 2) patients with traumatic intracranial hemorrhage. Results We enrolled 350 patients with TBI; the median (Q1, Q3) age was 70 (61, 84) years, 187 (53%) were male, and 91 patients (26%) had traumatic intracranial hemorrhage on initial ED cranial computed tomography (CT) imaging. A total of 257 patients (73%) were transported by EMS to a Level I or II trauma center. Sixty‐nine patients (20%) did not complete follow‐up at 3 or 6 months. Of the patients with follow‐up, 119 of 260 patients (46%) had moderate disability or worse at 6 months, including 55 of 260 patients (21%) who were dead at 6‐month follow‐up. Death or severe disabilities were more commonly attributed to non‐TBI causes while moderate disabilities or better were more commonly due to TBI. On adjusted analysis, an abnormal GCS score, higher Charlson Comorbidity Index scores, and the presence of traumatic intracranial hemorrhage on initial ED cranial imaging were associated with worse GOS‐E scores at 6 months. Trauma center transport was not associated with GOS‐E scores at 6 months for TBI patients and in patients with traumatic intracranial hemorrhage on initial ED CT imaging. Conclusions In older adults with TBI, moderate disability or worse is common 6 months after injury. Over one in five of older adults with TBI died by 6 months, usually due to nonhead causes. Patients with TBI or traumatic intracranial hemorrhage did not have improved functional outcomes with initial triage to a trauma center.

show abstract

“…However, it is often incomplete because of imperfect data collection systems, workflows not designed to capture mortality data, and patients lost to follow up . The purpose of this study was to examine the potential impact of missing death data in an EHR‐derived oncology data source, which is of critical importance to establishing a research‐grade EHR‐derived database and should provide guidance with respect to an acceptable level of completeness …”

Section: Discussionmentioning

confidence: 99%

An evaluation of the impact of missing deaths on overall survival analyses of advanced non–small cell lung cancer patients conducted in an electronic health records database

Carrigan¹,

Whipple²,

Taylor³

et al. 2019

Pharmacoepidemiology and Drug

View full text Add to dashboard Cite

Purpose The aim of this study was to assess the impact of missing death data on survival analyses conducted in an oncology EHR‐derived database. Methods The study was conducted using the Flatiron Health oncology database and the National Death Index (NDI) as a gold standard. Three analytic frameworks were evaluated in advanced non‐small cell lung cancer (aNSCLC) patients: median overall survival [mOS]), relative risk estimates conducted within the EHR‐derived database, and “external control arm” analyses comparing an experimental group augmented with mortality data from the gold standard to a control group from the EHR‐derived database only. The hazard ratios (HRs) obtained within the EHR‐derived database (91% sensitivity) and the external control arm analyses, were compared with results when both groups were augmented with mortality data from the gold standard. The above analyses were repeated using simulated lower mortality sensitivities to understand the impact of more extreme levels of missing deaths. Results Bias in mOS ranged from modest (0.6–0.9 mos.) in the EHR‐derived cohort with (91% sensitivity) to substantial when lower sensitivities were generated through simulation (3.3–9.7 mos.). Overall, small differences were observed in the HRs for the EHR‐derived cohort across comparative analyses when compared with HRs obtained using the gold standard data source. When only one treatment arm was subject to estimation bias, the bias was slightly more pronounced, but increased substantially when lower sensitivities were simulated. Conclusions The impact on survival analysis is minimal with high mortality sensitivity with only modest impact associated within external control arm applications.

show abstract

Principled Approaches to Missing Data in Epidemiologic Studies

Cited by 193 publications

References 28 publications

Symptom burden among older breast cancer survivors: The Thinking and Living With Cancer (TLC) study

Symptom burden among older breast cancer survivors: The Thinking and Living With Cancer (TLC) study

The Association of Trauma Center Transport and Long‐term Functional Outcomes in Head‐injured Older Adults Transported by Emergency Medical Services

An evaluation of the impact of missing deaths on overall survival analyses of advanced non–small cell lung cancer patients conducted in an electronic health records database

Contact Info

Product

Resources

About