Methodological Challenges When Studying Distance to Care as an Exposure in Health Research

Strategies to Assess the Effect of Continuous Positive Airway Pressure on Long-Term Clinically Important Outcomes among Patients with Symptomatic Obstructive Sleep Apnea: An Official American Thoracic Society Workshop Report

Continuous positive airway pressure (CPAP) is the first-line treatment for obstructive sleep apnea (OSA). Although CPAP improves symptoms (e.g., daytime sleepiness), there is a lack of high-quality evidence that CPAP prevents many long-term outcomes, including cognitive impairment, myocardial infarction, and stroke. Observational studies suggest that patients with symptoms may be particularly likely to experience these preventive benefits with CPAP, but ethical and practical concerns limited the participation of such patients in prior long-term randomized trials. As a result, there is uncertainty about the full benefits of CPAP, and resolving this uncertainty is a key priority for the field. This workshop assembled clinicians, researchers, ethicists, and patients to identify strategies to understand the causal effects of CPAP on long-term clinically important outcomes among patients with symptomatic OSA. Quasi-experimental designs can provide valuable information and are less time and resource intensive than trials. Under specific conditions and assumptions, quasi-experimental studies may be able to provide causal estimates of CPAP's effectiveness from generalizable observational cohorts. However, randomized trials represent the most reliable approach to understanding the causal effects of CPAP among patients with symptoms. Randomized trials of CPAP can ethically include patients with symptomatic OSA, as long as there is outcome-specific equipoise, adequate informed consent, and a plan to maximize safety while minimizing harm (e.g., monitoring for pathologic sleepiness). Furthermore, multiple strategies exist to ensure the generalizability and practicality of future randomized trials of CPAP. These strategies include reducing the burden of trial procedures, improving patient-centeredness, and engaging historically excluded and underserved populations.

The Relation Between Travel Distance and Overall Survival for HPV-Associated Cancers in a High-Burden State

Purpose

To assess the association between travel distance to an academic health system and overall survival for patients with human papillomavirus (HPV)-associated cancers.

Methods

Using hospital-based cancer registry data from 2005-2019, we calculated unidirectional travel distance from each patient's geocoded address to our academic health center through network analysis. We categorized distance as short (<25 miles), intermediate (25-74.9 miles), or long (≥75 miles). The primary outcome was time from the date of initial diagnosis to the date of death or last contact. We used multivariable Cox proportional hazards regression to evaluate the association between travel distance and overall survival. We also estimated the adjusted observed 5-year survival rate.

Results

Patients with HPV-associated cancers traveling distances that were intermediate (hazard ratio [HR], 1.23; 95% CI, 1.06-1.43) and long (HR, 1.15; 95% CI, 1.01-1.32) had a higher hazard of death than the short-distance group. The adjusted 5-year observed survival rates for HPV-associated cancers were lowest in the intermediate-distance group (60.4%) compared with the long-(62.6%) and short-distance (66.2%) groups.

Conclusions

Our findings indicate that travel distance to an academic health center was associated with overall survival for patients with HPV-associated cancers, reflecting the importance of considering travel burden in improving patient outcomes.

Distance to Multidisciplinary Team Clinic in Gaborone, Botswana, and Stage at Cervical Cancer Presentation for Women Living With and Without HIV

PURPOSE

Cervical cancer is the leading cause of cancer death for women in Botswana. Barriers in access to cancer care can lead to later stages at diagnosis and increased mortality. This study evaluated access, defined as travel time from a patient's residential village to a multidisciplinary team clinic in Gaborone, with stage of cervical cancer at presentation. In addition, because of the high HIV prevalence in Botswana, we explored the association between travel time and HIV status.

METHODS

Eligible patients with cervical cancer presenting to the multidisciplinary team between 2015 and 2020 were included. Data were abstracted from questionnaires and hospital records. Google Maps was used to calculate travel time. Multinomial regression was used to examine travel time and cancer stage, and multivariable logistic regression was used to investigate travel time and HIV status.

RESULTS

We identified 959 patients with cervical cancer of which 70.1% were women living with HIV. The median travel time was approximately 2 hours. Using a reference group of stage I disease and a travel time of < 1 hour, the odds of presenting with stage II increased for patients traveling 3-5 hours (adjusted odds ratio [OR], 2.00; 95% CI, 1.14 to 3.52) and > 5 hours (OR, 2.19; 95% CI, 1.15 to 4.19). There were no significant associations for stage III. For stage IV disease, the odds were increased for patients traveling 3-5 hours (OR, 2.93; 95% CI, 1.26 to 6.79) and > 5 hours (adjusted OR, 4.05; 95% CI, 1.62 to 10.10). In addition, the odds of patients presenting living with HIV increased with increasing travel time (trend test = 0.004).

CONCLUSION

This study identified two potential factors, travel time and HIV status, that influence access to comprehensive cervical cancer care in Botswana.

A spatio-temporal Bayesian model to estimate risk and evaluate factors related to drug-involved emergency department visits in the greater Baltimore metropolitan area

The ongoing opioid overdose epidemic in the United States presents a major public health challenge. Opioid-involved morbidity, especially nonfatal emergency department (ED) visits, are a key opportunity to prevent mortality and measure the extent of the problem in the local substance use landscape. Data on the rate of ED visits is normally distributed by federal agencies. However, state- and substate-level rates of ED visit demonstrate significant geographic variation. This study uses an ongoing sample of ED visits from four hospitals in the University of Maryland Medical System from January 2016 to December 2019 to provide locally sensitive information on ED visit rates and risk for drug-related health outcomes. Using exploratory spatial data analysis and spatio-temporal Bayesian models, this study analyzes both the frequency and risk of heroin-, methadone-, and cocaine-involved ED visits across the greater Baltimore Maryland area at the Zip Code Tabulation Area-level (ZCTA). The Global Moran's I for total heroin-, methadone-, and cocaine-involved ED visits in 2019 was 0.44, 0.56, and 0.53, demonstrating strong positive spatial autocorrelation. Spatio-temporal Bayesian models indicated that ZCTA with a higher score in a deprivation index, with a higher share of Centers for Medicare Services claims, and adjacent to a sampled UMMS hospital had an increased risk of ED visits, with variation in the magnitude of this increased risk depending on the drug-demographic strata. Modeled disease risk surfaces - including posterior median risk and posterior exceedance probabilities - showed distinctly different risk surfaces between the substances of interest, probabilistically identifying ZCTA with a lower or higher risk of ED visits. The modeling approach used a sample of ED visits from a larger health system to estimate recent, locally sensitive drug-related morbidity across a large metropolitan area.

A systematic review of quantitative bias analysis applied to epidemiological research

BACKGROUND

Quantitative bias analysis (QBA) measures study errors in terms of direction, magnitude and uncertainty. This systematic review aimed to describe how QBA has been applied in epidemiological research in 2006-19.

METHODS

We searched PubMed for English peer-reviewed studies applying QBA to real-data applications. We also included studies citing selected sources or which were identified in a previous QBA review in pharmacoepidemiology. For each study, we extracted the rationale, methodology, bias-adjusted results and interpretation and assessed factors associated with reproducibility.

RESULTS

Of the 238 studies, the majority were embedded within papers whose main inferences were drawn from conventional approaches as secondary (sensitivity) analyses to quantity-specific biases (52%) or to assess the extent of bias required to shift the point estimate to the null (25%); 10% were standalone papers. The most common approach was probabilistic (57%). Misclassification was modelled in 57%, uncontrolled confounder(s) in 40% and selection bias in 17%. Most did not consider multiple biases or correlations between errors. When specified, bias parameters came from the literature (48%) more often than internal validation studies (29%). The majority (60%) of analyses resulted in >10% change from the conventional point estimate; however, most investigators (63%) did not alter their original interpretation. Degree of reproducibility related to inclusion of code, formulas, sensitivity analyses and supplementary materials, as well as the QBA rationale.

CONCLUSIONS

QBA applications were rare though increased over time. Future investigators should reference good practices and include details to promote transparency and to serve as a reference for other researchers.

Changes in Behavior with Increasing Pregnancy Attempt Time: A Prospective Cohort Study

BACKGROUND

The extent to which couples change their behaviors with increasing pregnancy attempt time is not well documented.

METHODS

We examined change in selected behaviors over pregnancy attempt time in a North American preconception cohort study. Eligible females were ages 21-45 years and not using fertility treatment. Participants completed baseline and bimonthly follow-up questionnaires for up to 12 months or until pregnancy.

RESULTS

Among 3,339 females attempting pregnancy for 0-1 cycles at enrollment, 250 contributed 12 months of follow-up without conceiving. Comparing behaviors at 12 months versus baseline, weighted for loss-to-follow-up, we observed small-to-moderate reductions in mean caffeine intake (-19.5 mg/day, CI = -32.7, -6.37), alcohol intake (-0.85 drinks/week, CI = -1.28, -0.43), marijuana use (-3.89 percentage points, CI = -7.33, 0.46), and vigorous exercise (-0.68 hours/week, CI = -1.05, -0.31), and a large increase in activities to improve conception chances (e.g., ovulation testing) (21.7 percentage points, CI = 14.8, 28.6). There was little change in mean cigarette smoking (-0.27 percentage points, CI = -1.58, 1.04), perceived stress scale score (-0.04 units, CI = -0.77, 0.69), or other factors (e.g., sugar-sweetened soda intake, moderate exercise, intercourse frequency, and multivitamin use), but some heterogeneity within subgroups (e.g., 31% increased and 32% decreased their perceived stress scores by ≥2 units; 14% reduced their smoking but none increased their smoking by ≥5 cigarettes/day).

CONCLUSIONS

Although many behaviors changed with increasing pregnancy attempt time, mean changes tended to be modest for most variables. The largest differences were observed for the use of caffeine, alcohol, and marijuana, and methods to improve conception chances.

Too poor or too far? Partitioning the variability of hospital-based childbirth by poverty and travel time in Kenya, Malawi, Nigeria and Tanzania

BACKGROUND

In sub-Saharan Africa, women are most likely to receive skilled and adequate childbirth care in hospital settings, yet the use of hospital for childbirth is low and inequitable. The poorest and those living furthest away from a hospital are most affected. But the relative contribution of poverty and travel time is convoluted, since hospitals are often located in wealthier urban places and are scarcer in poorer remote area. This study aims to partition the variability in hospital-based childbirth by poverty and travel time in four sub-Saharan African countries.

METHODS

We used data from the most recent Demographic and Health Survey in Kenya, Malawi, Nigeria and Tanzania. For each country, geographic coordinates of survey clusters, the master list of hospital locations and a high-resolution map of land surface friction were used to estimate travel time from each DHS cluster to the nearest hospital with a shortest-path algorithm. We quantified and compared the predicted probabilities of hospital-based childbirth resulting from one standard deviation (SD) change around the mean for different model predictors.

RESULTS

The mean travel time to the nearest hospital, in minutes, was 27 (Kenya), 31 (Malawi), 25 (Nigeria) and 62 (Tanzania). In Kenya, a change of 1SD in wealth led to a 33.2 percentage points change in the probability of hospital birth, whereas a 1SD change in travel time led to a change of 16.6 percentage points. The marginal effect of 1SD change in wealth was weaker than that of travel time in Malawi (13.1 vs. 34.0 percentage points) and Tanzania (20.4 vs. 33.7 percentage points). In Nigeria, the two were similar (22.3 vs. 24.8 percentage points) but their additive effect was twice stronger (44.6 percentage points) than the separate effects. Random effects from survey clusters also explained substantial variability in hospital-based childbirth in all countries, indicating other unobserved local factors at play.

CONCLUSIONS

Both poverty and long travel time are important determinants of hospital birth, although they vary in the extent to which they influence whether women give birth in a hospital within and across countries. This suggests that different strategies are needed to effectively enable poor women and women living in remote areas to gain access to skilled and adequate care for childbirth.

You Can't Drive a Car With Only Three Wheels

Authors aiming to estimate causal effects from observational data frequently discuss 3 fundamental identifiability assumptions for causal inference: exchangeability, consistency, and positivity. However, too often, studies fail to acknowledge the importance of measurement bias in causal inference. In the presence of measurement bias, the aforementioned identifiability conditions are not sufficient to estimate a causal effect. The most fundamental requirement for estimating a causal effect is knowing who is truly exposed and unexposed. In this issue of the Journal, Caniglia et al. (Am J Epidemiol. 2019;000(00):000-000) present a thorough discussion of methodological challenges when estimating causal effects in the context of research on distance to obstetrical care. Their article highlights empirical strategies for examining nonexchangeability due to unmeasured confounding and selection bias and potential violations of the consistency assumption. In addition to the important considerations outlined by Caniglia et al., authors interested in estimating causal effects from observational data should also consider implementing quantitative strategies to examine the impact of misclassification. The objective of this commentary is to emphasize that you can't drive a car with only three wheels, and you also cannot estimate a causal effect in the presence of exposure misclassification bias.