- In a survey study of mental health symptoms among healthcare workers during a pandemic, researchers achieve a 35% response rate. Non-response analysis shows that younger workers and those in emergency departments were less likely to respond. When documenting limitations, which statement BEST characterizes the primary concern?
Answer: The direction of non-response bias cannot be determined without additional information about non-responders' characteristics
When evaluating non-response bias in survey research, you need to understand that the direction and magnitude of bias depends on how non-responders differ from responders on the outcome of interest. This is a classic limitation that researchers must acknowledge honestly.
The correct answer is D because determining bias direction requires knowing whether non-responders have better or worse mental health than responders. While we know younger workers and emergency department staff responded less, we cannot assume their mental health status. Younger workers might have better mental health (leading to underestimation) or worse mental health due to different stressors (leading to overestimation). Similarly, emergency department workers could be more resilient or more traumatized than other healthcare workers.
Answer A assumes non-responders have better mental health, but this is speculation without evidence. Answer B makes the opposite assumption - that non-responders have worse mental health - which is equally unsupported. Answer C incorrectly identifies this as selection bias rather than non-response bias. Selection bias occurs when the study population itself isn't representative, but here the issue is differential response rates within an appropriate target population.
The 35% response rate is concerning, but the key limitation isn't the low rate itself - it's that we cannot determine whether the 65% who didn't respond differ systematically on mental health outcomes.
Study tip: In biostatistics questions about survey limitations, remember that bias direction can only be determined when you have information about how non-responders differ on the outcome variable, not just on demographic characteristics.
- Researchers conduct a multi-site clinical trial with standardized protocols, but discover that one site consistently reports higher adverse event rates than others, despite treating similar patient populations. When documenting this finding as a study limitation, which explanation should be prioritized?
Answer: Reporting bias may occur due to variations in site personnel training or institutional culture
When analyzing study limitations in multi-site trials, you need to distinguish between different types of bias based on the specific circumstances described. This question tests your ability to identify the most likely source of systematic error when one site shows consistently different results despite standardized protocols.
The key insight here is that while protocols are standardized, the implementation and interpretation of those protocols can vary significantly between sites. When one site consistently reports higher adverse event rates with similar patient populations, this suggests differences in how site personnel are trained to recognize, document, or report these events. Some sites may have more aggressive reporting cultures, better-trained staff, or institutional policies that encourage comprehensive documentation. This represents reporting bias - systematic differences in how information is collected and communicated rather than true differences in event occurrence.
Choice A is incorrect because site-specific confounding would require actual unmeasured differences in patient characteristics, but the question states populations are similar. Choice B incorrectly identifies detection bias - while monitoring practices could vary, the issue described is more about reporting patterns than detection methods. Choice D suggests measurement bias from definitional differences, but standardized protocols should minimize this issue.
The critical distinction is between what's happening (similar adverse events) and how it's being captured and reported (differently across sites). When you see consistent site-level differences despite standardized protocols, think about human factors in implementation rather than protocol design flaws. Focus on how training, culture, and reporting practices can create systematic differences even with identical written procedures.
- Researchers conduct a randomized trial comparing an online intervention to usual care for depression. The primary outcome is measured using a validated depression scale administered by research staff who are aware of participants' treatment assignments. When documenting potential bias, which concern should receive priority?
Answer: Observer bias may affect outcome measurement due to lack of blinding in assessment
When evaluating potential bias in clinical trials, you need to identify which type of bias poses the greatest threat given the specific study design elements described.
Observer bias occurs when researchers' knowledge of treatment assignments influences how they measure or interpret outcomes. In this scenario, research staff who know participants' treatment assignments are conducting depression assessments using a validated scale. Even with a standardized instrument, the staff's awareness of whether someone received the intervention or usual care can unconsciously influence how they administer the assessment, interpret borderline responses, or record scores. This creates a systematic error that could favor one treatment group over another.
Looking at the other options: Performance bias (B) refers to differences in care received by participants due to their knowledge of treatment assignment, but the question specifically highlights that research staff are unblinded during assessment. Detection bias (C) involves systematic differences in how outcomes are ascertained between groups, but there's no mention of different follow-up procedures—the issue is assessor awareness. Measurement bias (D) addresses the inherent subjectivity of depression scales, but validated instruments are designed to minimize this concern, and subjectivity alone doesn't create systematic between-group differences.
The correct answer is A because observer bias directly stems from the specific design flaw mentioned: unblinded research staff conducting outcome assessments.
Study tip: In bias identification questions, match the specific study design feature mentioned to the corresponding bias type. Unblinded assessors = observer bias risk, regardless of other potential biases that might theoretically exist.
- A longitudinal study follows 1,200 participants for 10 years to examine predictors of cognitive decline. The analysis uses linear mixed-effects models with random intercepts and slopes. During follow-up, 23% of participants miss at least one visit, with higher missingness among older participants and those with lower baseline cognitive scores. Which assumption regarding the missing data mechanism should be most carefully documented?
Answer: The missing at random assumption may be violated if missingness is related to unobserved cognitive decline not captured by baseline measures
When analyzing longitudinal studies with missing data, you need to understand the three missing data mechanisms and their implications for your statistical models. The key is identifying which assumption is most critical to document given the study's specific missingness pattern.
Answer D correctly identifies the core concern. While missingness relates to observed baseline characteristics (age, baseline cognitive scores), the real threat to validity occurs if missingness is also related to unobserved factors—specifically, the rate of cognitive decline itself. If participants with faster cognitive decline are more likely to drop out, and this decline isn't fully captured by baseline measures, then the Missing at Random (MAR) assumption is violated. This would bias the mixed-effects model estimates because the missing data mechanism depends on unobserved outcomes.
Answer A mischaracterizes the situation. Since missingness relates to observed baseline characteristics, this actually supports MAR rather than violating Missing Completely at Random (MCAR). MAR is the standard assumption for mixed-effects models, not a problem requiring multiple imputation.
Answer B incorrectly suggests that intermittent missingness compromises analysis validity. Modern mixed-effects models handle irregular follow-up patterns well under MAR, and complete case analysis would actually reduce power and potentially introduce bias.
Answer C overstates the problem. Differential missingness across observed subgroups doesn't automatically indicate non-ignorable missing data—it could still be MAR if properly modeled.
Study tip: In longitudinal studies, focus on whether missingness might be related to the unobserved trajectory of your outcome variable. This is the most common and serious threat to valid inference in repeated measures analyses.
- An epidemiologist conducts a cross-sectional study using an online survey to investigate the prevalence of depression symptoms in healthcare workers during a pandemic. The survey link is distributed through professional social media groups and yields 2,847 responses from a target population of approximately 50,000 healthcare workers in the region. Which sampling-related limitation requires the most prominent documentation?
Answer: The voluntary response sampling method introduces selection bias, as participants may systematically differ from non-respondents in depression symptom severity
Choice A correctly identifies that voluntary response sampling is the primary limitation - people who choose to respond to a depression survey may be systematically different (more likely to have symptoms or strong opinions) than non-respondents, biasing prevalence estimates. Choice B misinterprets response rate as a power issue rather than bias concern. Choice C addresses coverage bias but voluntary response bias is more fundamental. Choice D incorrectly suggests the sampling method determines the statistical analysis approach.
- A systematic review examines interventions for childhood obesity, but most included studies are conducted in high-income countries with well-resourced healthcare systems. Only 2 of 25 studies are from low- or middle-income countries. When documenting limitations, which assumption about generalizability should be prioritized?
Answer: The assumption that intervention effectiveness is similar across different economic and healthcare contexts
When evaluating systematic reviews with limited geographic diversity, you need to assess which assumptions about generalizability are most problematic given the evidence base.
This systematic review suffers from severe geographic bias—23 of 25 studies come from high-income countries with well-resourced healthcare systems. The most dangerous assumption here is that interventions proven effective in wealthy, well-resourced settings will work equally well in different economic and healthcare contexts. Answer A correctly identifies this as the priority limitation because intervention effectiveness often depends heavily on available resources, healthcare infrastructure, cultural factors, and implementation capacity—all of which vary dramatically between high-income and low-/middle-income countries.
Answer B is less concerning because childhood obesity does share many biological and physiological mechanisms across populations, making cause-and-consequence assumptions more defensible than intervention effectiveness assumptions.
Answer C misframes the issue—the review isn't claiming to represent global disease burden, but rather global intervention effectiveness, making this a secondary concern.
Answer D, while related to feasibility concerns, focuses on implementation rather than effectiveness. The more fundamental assumption being made is that the interventions will actually work across contexts, not just whether they're implementable.
Study tip: In systematic review questions, always prioritize limitations that threaten the core conclusions. When geographic diversity is limited, question assumptions about intervention effectiveness across different resource settings—this is usually more problematic than assumptions about disease biology or burden.
- A researcher analyzes administrative claims data to compare 30-day readmission rates between two hospitals, adjusting for patient age, sex, and primary diagnosis. Hospital A shows significantly lower readmission rates (OR = 0.78, 95% CI: 0.65-0.94). When documenting assumptions about causal interpretation, which limitation should receive primary emphasis?
Answer: The administrative data source lacks clinical severity measures, creating potential confounding by unmeasured patient acuity differences between hospitals
Choice A correctly identifies the key limitation for causal interpretation - administrative data typically lacks clinical severity measures, so sicker patients may systematically go to certain hospitals, confounding the comparison. Choice B addresses selection but doesn't specify the key unmeasured confounder issue. Choice C discusses post-discharge factors but these would affect both hospitals similarly. Choice D incorrectly suggests propensity scores solve the unmeasured confounder problem.
- A cross-sectional study uses electronic health records to examine the prevalence of diabetes complications among patients at a large healthcare system. The analysis includes only patients who had at least one clinical visit in the past year. Which limitation should be prioritized in documentation?
Answer: Prevalence estimates may not represent the true population prevalence due to healthcare-seeking behavior
When analyzing study limitations in cross-sectional research using healthcare data, you need to distinguish between different types of bias and identify which poses the most fundamental threat to your study's validity and generalizability.
The correct answer is D because this limitation strikes at the heart of what makes prevalence estimates meaningful. When you restrict your sample to patients who had recent clinical visits, you're fundamentally altering the population you're studying. Patients who seek healthcare regularly may have different baseline health characteristics, socioeconomic status, or disease severity compared to the general diabetic population. This creates a systematic difference between your study sample and the target population, making your prevalence estimates potentially misleading for public health planning or clinical decision-making.
Option A describes a real phenomenon but isn't the primary concern here since the study specifically examines patients already in the healthcare system. Option B represents information bias, which is certainly present but is a data quality issue that can potentially be addressed through standardization or validation procedures. Option C identifies selection bias correctly but focuses too narrowly on patients receiving care elsewhere, missing the broader issue of healthcare-seeking behavior patterns.
The key distinction is that while A, B, and C represent manageable limitations or secondary concerns, D identifies a fundamental flaw that undermines the external validity of the entire study. On biostatistics exams, prioritize limitations that threaten the study's ability to answer its primary research question or generalize findings to the intended population.
- A longitudinal study examines the relationship between social media use and mental health among adolescents, with assessments every 6 months for 2 years. Researchers note that social media platforms and usage patterns change rapidly during the study period. When documenting assumptions, which temporal issue should receive priority?
Answer: The assumption that the exposure measurement captures relevant aspects of social media use across changing platforms
When analyzing longitudinal studies, you need to consider how changes over time might threaten the validity of your measurements and conclusions. This question tests your understanding of measurement validity in rapidly changing environments.
The correct answer is C because exposure measurement validity is the most fundamental concern here. If your instruments can't accurately capture what social media use actually means as platforms evolve (from Facebook to Instagram to TikTok, for example), then your entire study becomes meaningless. You could be measuring yesterday's social media landscape while drawing conclusions about today's reality. This threatens the core validity of your exposure variable.
Option A is incorrect because assuming constant relationships over time is actually unrealistic and not a priority assumption to document. Longitudinal studies often aim to detect changing relationships, so this assumption would be counterproductive.
Option B is wrong because expecting stable usage patterns contradicts what we know about adolescent behavior and technology adoption. This assumption would be both unrealistic and unnecessary for a valid study.
Option D, while important, is less critical than C because mental health assessment tools typically have established validity across adolescent age ranges. The tools themselves don't become invalid just because participants age two years.
Remember: In longitudinal studies involving rapidly changing exposures (technology, medications, behaviors), always prioritize whether your measurement tools can keep pace with the evolving reality you're trying to study. Invalid exposure measurement undermines everything downstream.
- A systematic review examines antibiotic prophylaxis effectiveness in surgical procedures. The included studies span 15 years and show considerable variation in outcome definitions: some report surgical site infections at 30 days, others at 90 days, and some include only deep infections while others include superficial infections. When documenting the impact on evidence synthesis, which methodological limitation should be prioritized?
Answer: The heterogeneous outcome definitions create measurement bias that may invalidate pooled effect estimates due to non-comparable endpoints
When evaluating methodological limitations in systematic reviews, you need to distinguish between issues that affect individual studies versus those that compromise the entire meta-analysis. The core question here is: what most fundamentally threatens the validity of combining these studies?
Option D correctly identifies the primary concern. When studies define outcomes differently—mixing 30-day versus 90-day follow-ups and superficial versus deep infections—you're essentially comparing apples to oranges. These aren't measuring the same clinical endpoint, so pooling their results produces a meaningless estimate. The measurement bias occurs because each study is capturing a different aspect of post-surgical infection, making any combined effect estimate scientifically invalid.
Option A incorrectly focuses on temporal changes. While surgical practices may evolve over 15 years, this historical bias affects individual study validity but doesn't prevent meaningful synthesis if outcomes are consistently defined. Option B mischaracterizes the problem as classification bias requiring subgroup analysis. However, you can't simply stratify your way out of fundamentally incomparable outcomes—the definitions are too heterogeneous for meaningful grouping. Option C identifies differential ascertainment bias, which is a real concern, but frames it as systematically underestimating shorter follow-up studies rather than recognizing that different follow-up periods create entirely different outcomes.
Remember this principle for systematic review questions: heterogeneous outcome definitions are a synthesis-killer. Before worrying about study quality or bias within individual studies, first ensure you're actually measuring the same thing across studies. Non-comparable endpoints make meta-analysis meaningless, regardless of how sophisticated your statistical methods are.
- A randomized controlled trial tests a mobile app intervention for diabetes management. Participants must own smartphones and be comfortable using technology. The control group receives standard care without app access. When documenting generalizability limitations, which assumption should receive primary focus?
Answer: The assumption that the study population's characteristics allow generalization to all diabetes patients
When evaluating generalizability in clinical trials, you need to assess how well the study population represents the broader target population you want to apply results to. This question tests your understanding of external validity and the assumptions underlying generalization of research findings.
The primary concern here is whether findings from smartphone-owning, tech-comfortable diabetes patients can be generalized to all diabetes patients. Answer C correctly identifies this core generalizability assumption - that the study population's characteristics (smartphone ownership, tech comfort, and likely associated demographics like age, income, and education) allow meaningful generalization to the entire diabetes population. This is the fundamental external validity question.
Answer A focuses too narrowly on just health outcomes among smartphone owners, missing the broader population representation issue. The concern isn't whether smartphone owners have different health outcomes, but whether they represent the full spectrum of diabetes patients across all relevant characteristics.
Answer B addresses effect modification rather than generalizability. While technology comfort might moderate effectiveness, this doesn't directly relate to whether findings can be generalized beyond the study population.
Answer D concerns implementation rather than generalizability. Variation in standard care affects internal validity and practical application, but doesn't address whether the study population represents the broader diabetes community.
Remember: generalizability questions always center on population representativeness. When you see studies with restrictive inclusion criteria (like requiring specific technology or skills), immediately ask whether those restrictions create a study population that differs meaningfully from the target population for intervention implementation.
- A study examining workplace stress and blood pressure recruits participants through employee wellness programs at large corporations. Participation is voluntary, and employees are assured that individual results will not be shared with employers. Which selection-related limitation should be prioritized in documentation?
Answer: Self-selection bias may occur if participation correlates with both stress levels and health awareness
When evaluating selection bias in research studies, you need to identify which bias creates the most problematic confounding relationship between the exposure (workplace stress), outcome (blood pressure), and the selection mechanism itself.
Self-selection bias occurs here because participation in voluntary wellness screenings likely correlates with both variables of interest. Employees who choose to participate may have higher health awareness, leading them to both recognize their stress levels more acutely AND be more likely to have their blood pressure monitored or managed. This creates a confounding pathway where the act of self-selecting into the study is associated with both the exposure and outcome, potentially distorting the true relationship between workplace stress and blood pressure.
Option A incorrectly assumes healthy worker bias applies, but this study recruits from existing employees, not comparing workers to non-workers. The healthy worker effect typically relates to employment status, not voluntary program participation. Option B describes volunteer bias too narrowly, focusing only on health concerns rather than the broader issue of how health-conscious individuals might differ systematically in both stress perception and health monitoring behaviors. Option D misidentifies the selection mechanism - the bias stems from individual voluntary participation, not from which companies offer wellness programs.
For biostatistics questions about selection bias, always ask: "What characteristic drives people into the study, and how might that same characteristic relate to both the exposure and outcome?" The most concerning biases are those that create three-way relationships between selection, exposure, and outcome variables.
- A case-control study investigates the relationship between prenatal vitamin use and birth defects. Cases are identified from a birth defects registry, while controls are selected from birth certificates in the same geographic region. Maternal vitamin use is assessed through telephone interviews conducted 2-3 years after delivery. Which bias should be prioritized when documenting limitations?
Answer: Information bias may result from mothers' knowledge of their child's health status influencing responses
When evaluating bias in case-control studies, you need to consider which type of bias is most likely to systematically distort your results based on the study design and data collection methods.
Information bias occurs when there are systematic differences in how information is collected between cases and controls. In this study, mothers of children with birth defects (cases) are being interviewed about their prenatal vitamin use while already knowing their child has a birth defect. This knowledge creates a powerful motivation to recall their behavior differently than mothers of healthy children (controls). Case mothers may be more likely to underreport vitamin use (feeling guilty) or overreport it (trying to show they were responsible), while control mothers have no such emotional investment in their responses. This differential reporting based on outcome knowledge makes option B the primary concern.
Looking at the other options: A) describes recall bias, but this is actually a subset of information bias and doesn't capture the key issue of differential recall between groups. C) selection bias from registry completeness could affect study validity, but it's less directly related to the specific data collection method described. D) misclassification bias from the time interval would affect both groups equally, not create the systematic difference between cases and controls that's most problematic here.
Study tip: In case-control studies where participants know their outcome status during data collection, always prioritize information bias as your primary concern. The knowledge of being a "case" fundamentally changes how people report their exposures.
- A longitudinal study examines cognitive decline in elderly adults over 5 years. Participants undergo cognitive testing annually, but 40% drop out before study completion, with higher dropout rates among those with lower baseline cognitive scores. Which assumption about missing data should be documented as MOST problematic?
Answer: The assumption that participants who drop out have similar cognitive trajectories to those who remain
When analyzing longitudinal studies with dropout, you need to evaluate assumptions about missing data mechanisms and their impact on study validity. The key concern is whether missing data introduces systematic bias that affects your conclusions.
Option B identifies the most problematic assumption because it directly contradicts the study's observed pattern. The scenario tells us that participants with lower baseline cognitive scores are more likely to drop out, which strongly suggests these individuals may be experiencing faster cognitive decline—exactly what the study aims to measure. Assuming dropouts have similar trajectories to continuing participants ignores this systematic difference and would lead to underestimating the true rate of cognitive decline in the population.
Option A describes Missing Completely at Random (MCAR), which would actually be the least problematic assumption, though unrealistic given the described dropout pattern. Option C addresses prediction accuracy rather than missing data bias—while baseline scores may not perfectly predict trajectories, this doesn't create the systematic bias that missing data does. Option D raises a methodological concern about measurement frequency, but this affects data quality for all participants equally, not specifically the missing data problem.
The fundamental issue is that when people experiencing the outcome of interest (cognitive decline) are more likely to drop out, any analysis assuming similar trajectories will systematically underestimate the true effect size.
Study tip: In longitudinal studies, always ask whether the reason for dropout is related to the outcome being measured. When it is, assumptions about similar trajectories become highly suspect and create the most serious validity threats.
- A researcher conducts a cross-sectional study to examine the relationship between smartphone use and sleep quality among college students. Participants are recruited through social media advertisements and complete an online survey about their daily screen time and self-reported sleep duration. Which assumption is MOST critical to document when interpreting the study's generalizability?
Answer: The assumption that the sample represents the broader population of college students
When evaluating the generalizability of any study, you need to assess how well the findings can be applied beyond the specific sample studied. The most fundamental threat to generalizability is sampling bias—when your sample doesn't adequately represent the target population.
In this cross-sectional study, the researchers used social media recruitment for an online survey. This sampling method creates significant selection bias because it only captures college students who: use social media, see the advertisements, have internet access, and are motivated to complete surveys. This could systematically exclude students with different demographic characteristics, technology use patterns, or attitudes toward research participation, making the sample unrepresentative of all college students.
Option A addresses measurement validity, which affects the accuracy of findings within the sample but doesn't directly impact whether findings generalize to other populations. Option C involves causal inference, but cross-sectional studies cannot establish causation regardless—they only show associations at a single time point. Option D concerns internal consistency reliability of the measurement instrument, which affects data quality but not generalizability to broader populations.
The sampling method is the primary threat here because even if the data is perfectly accurate (A), the instrument highly reliable (D), and researchers avoid causal claims (C), biased sampling would still prevent you from confidently applying these findings to college students who weren't represented in the sample.
Study tip: When evaluating study limitations, always consider sampling methods first for generalizability questions. Representative sampling is the foundation that makes external validity possible.
- A randomized trial comparing two surgical techniques uses surgeon self-assessment to measure technical difficulty and procedure time. Surgeons rate their own procedures immediately after completion. When documenting potential measurement bias, which assumption should receive primary attention?
Answer: The assumption that surgeons can objectively assess their own technical performance
When evaluating measurement bias in clinical research, you need to identify which assumption poses the greatest threat to data validity. Self-assessment creates inherent conflicts between objectivity and personal investment in outcomes.
Option A correctly identifies the primary concern: surgeons objectively assessing their own performance. This assumption is fundamentally problematic because it violates basic principles of unbiased measurement. Surgeons have direct personal and professional investment in their performance, making truly objective self-evaluation nearly impossible. They may unconsciously (or consciously) rate procedures as less difficult to maintain confidence or reputation, creating systematic bias that affects the entire study's validity.
Option B addresses inter-rater reliability between surgeons, which is certainly a concern but secondary to the fundamental self-assessment bias. Even if surgeons were consistent with each other, they could all be systematically biased in the same direction.
Option C focuses on timing of assessment. While immediate post-procedure rating might miss some aspects of true difficulty, this represents measurement error rather than systematic bias. The timing issue doesn't inherently favor one surgical technique over another.
Option D considers experience bias, where familiarity with techniques affects ratings. This could introduce bias, but it's technique-specific rather than affecting the overall measurement approach. Proper randomization and stratification can address this concern.
Study tip: In biostatistics questions about measurement bias, always prioritize assumptions that create systematic directional bias over those that create random measurement error. Self-assessment scenarios should immediately trigger concern about objectivity conflicts.
- A cohort study follows 1,000 adults for 10 years to examine the relationship between coffee consumption and cardiovascular disease. At baseline, coffee consumption is assessed using a single 24-hour dietary recall. Which limitation should be prioritized when documenting potential sources of bias?
Answer: Misclassification bias may occur because a single day may not represent typical coffee consumption patterns
When evaluating bias in cohort studies, you need to consider how the exposure measurement method affects the validity of your findings throughout the entire follow-up period. The key issue here is whether a single baseline measurement adequately captures participants' true exposure status over the 10-year study duration.
Misclassification bias occurs when participants are incorrectly categorized regarding their exposure status. In this study, using a single 24-hour dietary recall to classify coffee consumption creates a high risk of misclassification because people's daily coffee intake varies considerably. Someone might drink no coffee on the assessment day but consume 3 cups daily most other days, leading to their misclassification as a "non-coffee drinker." This measurement error affects the fundamental exposure categorization that drives the entire analysis, making option C the priority concern.
Option A incorrectly identifies recall bias, but 24-hour dietary recalls typically have good accuracy for recent consumption since participants are recalling yesterday's intake, not distant memories. Option B describes selection bias, but the question doesn't indicate systematic differences in who participates based on coffee consumption. Option D mentions information bias from underreporting, but coffee consumption isn't typically considered a socially undesirable behavior that people would systematically underreport.
For cohort study questions, always evaluate whether the exposure measurement adequately captures participants' true exposure status over the follow-up period. Single-point measurements of variable behaviors are particularly prone to misclassification bias, which can severely weaken study validity by diluting true associations.
- A retrospective cohort study uses insurance claims data to examine the relationship between medication adherence and hospitalization rates. Medication adherence is calculated from prescription fill dates, assuming patients consume medications at prescribed intervals. Which assumption requires careful documentation?
Answer: The assumption that prescription fills accurately reflect actual medication consumption patterns
When you encounter questions about retrospective studies using administrative data, focus on which assumptions most directly affect the validity of your exposure measurement and outcome assessment.
In this study, researchers are using prescription fill dates to infer actual medication consumption, which creates a significant measurement validity issue. Option A identifies the most critical assumption because there's often a substantial gap between filling a prescription and actually taking the medication as prescribed. Patients may fill prescriptions but not take them, take them irregularly, or discontinue them without refilling. This assumption directly affects the accuracy of the primary exposure variable and requires careful documentation because it's largely unverifiable from claims data alone.
Option B, while important, is less problematic because insurance claims typically provide comprehensive hospitalization records for covered individuals, and missing hospitalizations would likely be random rather than systematic. Option C incorrectly assumes that adherence consistency is a major concern - in fact, most adherence studies expect and account for variation over time. Option D misunderstands the scope of this particular study, which would typically focus on specific medication classes rather than making broad generalizations across all medications.
The key distinction is between assumptions that affect measurement validity (how accurately you're measuring what you think you're measuring) versus assumptions about completeness or consistency. In biostatistics, always prioritize identifying threats to measurement validity, especially when using proxy measures like prescription fills to infer actual behavior.
- In a randomized controlled trial of a new antihypertensive medication, 15% of participants in the treatment group and 8% in the control group discontinued the study before completion. The final analysis included only participants who completed the full study protocol. Which assumption underlying the statistical analysis is most directly violated, and what bias does this introduce?
Answer: Missing at random assumption is violated, introducing attrition bias through differential dropout related to treatment assignment
Choice C correctly identifies that differential dropout between groups violates the missing at random (MAR) assumption, as missingness is related to treatment assignment. This creates attrition bias in the complete-case analysis. Choice A confuses independence with missing data assumptions. Choice B mischaracterizes the issue as random sampling rather than missing data. Choice D incorrectly focuses on variance assumptions rather than the missing data mechanism.
- A researcher conducts a retrospective cohort study examining the relationship between statin use and cardiovascular events using electronic health records from a single hospital system. The analysis shows a significant protective effect of statins (HR = 0.65, 95% CI: 0.52-0.81). When documenting study limitations, which statement most accurately reflects the primary concern regarding generalizability?
Answer: The single-center recruitment may limit external validity to populations with different demographic and clinical characteristics
Choice B correctly identifies the key generalizability limitation - single-center studies may not represent broader populations due to institutional differences in patient demographics, practice patterns, and case mix. Choice A describes internal validity concerns rather than generalizability. Choice C addresses information bias but doesn't specifically relate to generalizability. Choice D incorrectly characterizes cohort studies, which establish temporal sequence.
