- A retrospective cohort study examines medical records to assess the relationship between a specific medication and adverse cardiac events. The study finds that patients who received the medication had more cardiac events, but further analysis reveals that this medication is preferentially prescribed to patients with existing cardiovascular risk factors. What is the most appropriate conclusion about this finding?
Answer: The association is likely due to confounding by indication and requires adjustment
When you encounter a study showing an unexpected association between a treatment and poor outcomes, you need to consider whether the relationship is causal or due to systematic differences in who receives the treatment.
This scenario describes confounding by indication - a specific type of confounding where the very reason a treatment is prescribed (the indication) is also associated with the outcome. Here, physicians logically prescribed the cardiac medication to patients who already had cardiovascular risk factors, creating a non-random distribution of baseline risk between treated and untreated groups. The medication appears harmful not because it causes cardiac events, but because it was given to patients who were already more likely to experience these events.
Answer B correctly identifies this as confounding by indication that requires statistical adjustment (like multivariable regression or propensity score matching) to account for baseline differences between groups.
Answer A jumps to a causal conclusion without considering confounding - a classic error when interpreting observational studies. The association doesn't prove causation.
Answer C incorrectly labels this as selection bias. While there is preferential selection of high-risk patients for treatment, this creates confounding, not selection bias. Selection bias occurs when the study sample isn't representative of the target population.
Answer D misidentifies the issue as information bias (differential measurement error). There's no indication that cardiac events were documented differently between groups - the problem is baseline risk differences, not measurement differences.
Study tip: In observational studies, always ask "Could baseline differences between groups explain this association?" before concluding causation. Confounding by indication is especially common in pharmacoepidemiologic studies.
- An epidemiologist designs a case-control study to investigate risk factors for a rare genetic disease. Cases are identified through a national registry of affected individuals, while controls are selected from the general population through random digit dialing. The response rate is 85% for cases but only 45% for controls. What is the primary methodological concern with this study design?
Answer: Selection bias due to differential participation rates between cases and controls
When evaluating study design issues, always consider whether the methods for selecting and recruiting participants could introduce systematic differences between comparison groups.
In this case-control study, the dramatically different response rates (85% vs 45%) create a serious selection bias problem. When participation rates differ substantially between cases and controls, you're no longer comparing representative samples of each group. The 55% of controls who didn't participate may systematically differ from those who did participate in ways that relate to the disease risk factors being studied. For example, non-participating controls might be more likely to have certain risk factors, creating a false association or masking a true one.
Option A is incorrect because both groups can use the same data collection methods despite different recruitment strategies. The bias stems from who participates, not how data is collected from participants. Option B misses the point - while genetic matching might be important for this disease, the immediate threat comes from differential participation, not unmatched genetic factors. Option D describes recall bias, which could occur in any case-control study, but the question specifically highlights the participation rate disparity as the design flaw.
Remember this pattern: whenever you see markedly different response rates between study groups (especially differences >20-30%), selection bias should be your primary concern. The groups are no longer comparable because different types of people chose to participate in each group, potentially confounding any associations you observe.
- In a case-control study of lung cancer and occupational exposures, researchers interview participants about their work history over the past 30 years. Cases are interviewed within one month of diagnosis, while controls are interviewed at routine health maintenance visits. Interviewers know whether participants are cases or controls. Which bias is of greatest concern in this scenario?
Answer: Information bias because interviewers may probe more thoroughly when interviewing cases
When evaluating bias in case-control studies, you need to identify which systematic error is most likely given the study design and data collection methods. The key concern here is how knowledge of participant status affects data quality.
Information bias occurs when there are systematic differences in how information is collected between cases and controls. In this scenario, interviewers know whether participants are cases or controls (the study is not blinded), creating a high risk that they will unconsciously probe more thoroughly when interviewing lung cancer patients about occupational exposures. This differential questioning intensity can lead to more detailed exposure histories from cases compared to controls, artificially inflating the association between occupational exposures and lung cancer.
Looking at the incorrect options: (A) Recall bias is a legitimate concern since cases might be more motivated to remember potential causes of their illness, but this is secondary to the interviewer bias problem. The timing difference (within one month vs. routine visits) doesn't create the primary bias issue here. (C) Selection bias relates to how participants are chosen for the study, but recruiting cases and controls from different settings is standard practice in case-control studies and doesn't inherently bias results. (D) Confounding bias refers to the influence of unmeasured variables like smoking, but this is an analysis issue, not a data collection bias.
Study tip: In bias questions, focus on the data collection process first. When interviewers aren't blinded to case/control status, information bias should be your top concern, especially when the exposure requires detailed recall or subjective assessment.
- A cohort study investigating air pollution and asthma in children measured pollution exposure using monitors placed at participants' schools. However, children from lower socioeconomic families were more likely to attend schools in high-pollution areas and also had higher rates of respiratory infections, poor nutrition, and crowded housing conditions. The study found a strong association between school air pollution levels and asthma incidence (RR = 2.1, p < 0.001). What is the MOST appropriate interpretation of these findings?
Answer: The association is likely due to confounding by socioeconomic factors that are related to both pollution exposure and asthma risk
When evaluating observational studies, you need to carefully distinguish between different types of bias and validity threats. This scenario presents a classic confounding situation where a third variable influences both the exposure and outcome.
The correct answer is C because socioeconomic status acts as a confounder here. Confounding occurs when a variable is associated with both the exposure (air pollution) and the outcome (asthma), creating a spurious association. Lower socioeconomic status leads to attending schools in high-pollution areas AND independently increases asthma risk through respiratory infections, poor nutrition, and crowded housing. This makes it impossible to determine how much of the observed association (RR = 2.1) is truly due to air pollution versus these socioeconomic factors.
Answer A incorrectly identifies selection bias. Selection bias occurs when study participants are systematically different from the target population, but here the issue is that the groups being compared differ in ways that affect the outcome.
Answer B is wrong because it assumes causality can be inferred despite the clear presence of confounding variables. The strong statistical association doesn't establish causation when confounders aren't controlled.
Answer D misidentifies measurement bias (also called information bias). While monitors at schools may not perfectly represent individual exposure, the primary threat here is confounding, not measurement error.
Remember: In observational studies, always look for confounding variables that could create spurious associations. When you see disparities in both exposure and outcome risk factors between groups, confounding is likely the main concern, not other types of bias.
- A case-control study examines the relationship between pesticide exposure and Parkinson's disease. Cases are recruited from movement disorder specialty clinics, while controls are recruited from general neurology clinics treating patients with headaches and other non-movement disorders. Both groups are asked about lifetime pesticide exposure. What is the most significant threat to the validity of this study design?
Answer: Selection bias due to cases being more severely affected patients seeking specialized care
When evaluating case-control studies, you need to carefully examine how cases and controls are selected, as this fundamentally determines the study's validity. The key question is whether the control group truly represents the population from which the cases arose.
In this study, the selection process creates a critical bias. Cases are recruited from specialized movement disorder clinics, which means they represent patients with more severe disease who have been referred for specialized care. Meanwhile, controls come from general neurology clinics treating completely different conditions. This creates selection bias because the cases don't represent all Parkinson's patients in the population - they're a subset of more severe cases who sought specialized treatment. The control group may also differ systematically in healthcare-seeking behavior, socioeconomic status, and other factors that could be related to pesticide exposure.
Option A describes differential recall bias, which is a concern but secondary to the fundamental selection problem. Option C mentions confounding by demographics, which can be controlled through matching or statistical adjustment. Option D addresses recall bias from long latency periods, which affects both groups equally and is less problematic than differential recall.
The core issue is that this sampling strategy violates a fundamental principle of case-control studies: controls should represent the same source population that gave rise to the cases. When controls are selected from an entirely different clinical population, you can't draw valid conclusions about exposure-disease relationships.
Study tip: In case-control questions, always evaluate whether the control group could realistically have become cases if they had developed the disease.
- A randomized trial comparing two blood pressure medications found that patients randomized to Drug A had their blood pressure measured by research nurses using automated devices, while patients randomized to Drug B had measurements taken by their regular physicians using manual sphygmomanometers during routine visits. The study found Drug A was significantly more effective. Which bias is the MOST serious threat to the validity of this comparison?
Answer: Detection bias due to different measurement protocols potentially favoring one treatment group over the other
When evaluating study validity, you need to identify which type of bias poses the greatest threat given the specific study design flaws described. This question tests your ability to distinguish between different types of systematic error in clinical trials.
The most serious problem here is detection bias (D). The two groups are being measured completely differently: Drug A patients have blood pressure taken by trained research nurses using standardized automated devices, while Drug B patients are measured by various physicians using manual methods during routine care. Automated devices typically give more consistent, often lower readings than manual sphygmomanometers, and research nurses follow strict protocols while regular physicians may be more variable in their technique. This measurement differential could easily create the appearance that Drug A is more effective when the difference is actually due to the measurement method, not the drug itself.
Selection bias (C) is unlikely because randomization should have eliminated systematic baseline differences between groups. Performance bias (B) isn't the primary concern since the question doesn't indicate that knowledge of treatment assignment is affecting patient or provider behavior regarding the actual treatments. Attrition bias (A) could be problematic, but there's no information suggesting differential dropout rates, and even if present, it's less fundamental than the measurement protocol problem.
Study tip: In clinical trial questions, always check if the measurement or assessment methods are the same between groups. Different measurement protocols can create false treatment effects and represent one of the most serious threats to internal validity.
- In a longitudinal study of cognitive decline, researchers used telephone interviews to assess cognitive function. Participants with hearing impairments scored systematically lower on verbal cognitive tests, while those with better hearing performed better. The hearing status was not related to the exposure of interest (educational attainment). After 5 years, participants with hearing impairments were more likely to be classified as having cognitive decline. What is the PRIMARY concern for study validity?
Answer: Differential measurement bias creating false associations between hearing status and cognitive outcomes
When evaluating study validity in biostatistics, you need to identify which type of bias poses the greatest threat to drawing accurate conclusions. This question tests your ability to distinguish between different sources of bias in longitudinal research.
The scenario describes a classic case of differential measurement bias. The telephone-based cognitive assessment systematically measures cognitive function differently based on participants' hearing ability. Those with hearing impairments appear to have worse cognitive function not because their cognition is actually declining faster, but because the measurement tool (verbal tests via telephone) is inherently biased against them. This creates a spurious association between hearing status and apparent cognitive decline that doesn't reflect true cognitive differences. Answer D correctly identifies this as the primary validity concern.
Let's examine why the other options don't fit: A is incorrect because the question explicitly states that hearing status wasn't related to educational attainment (the exposure), ruling out confounding. B describes non-differential measurement bias, but the bias here clearly differs by hearing status, making it differential, not non-differential. C suggests selection bias, but the issue isn't who participates in the study—it's how their cognitive function is measured once they're enrolled.
The key insight is recognizing that measurement bias becomes "differential" when the measurement error systematically varies across groups. When you encounter longitudinal studies with measurement concerns, always ask: "Does the measurement tool work equally well for all participant groups?" If not, differential measurement bias should be your primary concern for study validity.
- In a cohort study examining dietary fat intake and cardiovascular disease, researchers discover that participants with higher education levels tend to both consume less saturated fat and have better access to preventive healthcare. If education level is not controlled for in the analysis, what impact will this have on the observed association between dietary fat and cardiovascular disease?
Answer: It will result in confounding that may distort the true association being studied
When you encounter a scenario where a third variable is associated with both the exposure and outcome in an epidemiological study, you're likely dealing with confounding. This is a fundamental threat to validity that can make associations appear stronger, weaker, or even reverse their direction.
In this scenario, education level meets the classic definition of a confounder: it's associated with the exposure (higher education → lower saturated fat intake) and independently associated with the outcome (higher education → better healthcare access → lower cardiovascular disease risk). When education isn't controlled for, some of the apparent protective effect attributed to lower saturated fat intake may actually be due to better healthcare access among the educated participants. This distorts the true relationship you're trying to measure, making option C correct.
Let's examine why the other options don't fit: Option A describes information bias, which occurs when data collection methods systematically differ between groups - but the scenario doesn't suggest dietary assessment accuracy varies by education. Option B refers to selection bias, which happens during participant recruitment, not during analysis of already-enrolled participants. Option D identifies recall bias, where memory differences affect data quality, but this isn't the primary concern when education is simply not controlled for analytically.
Remember this pattern: whenever you see a third variable that influences both your exposure and outcome independently, think confounding first. The key phrase to watch for is when something "is not controlled for" or "is not adjusted for" in the analysis - this signals potential confounding rather than other bias types.
- An occupational health study finds that factory workers exposed to chemical X have higher rates of liver disease than unexposed workers. However, exposed workers also have mandatory annual liver function tests while unexposed workers only receive testing when symptomatic. Additionally, workers in the exposed group have higher rates of alcohol consumption. What statement best describes the bias and confounding issues in this study?
Answer: Detection bias may overestimate the association, while alcohol consumption may also overestimate it
When evaluating epidemiological studies, you need to identify both bias (systematic errors in data collection) and confounding (when a third variable affects both exposure and outcome). This question tests your ability to determine the direction of these effects.
Detection bias occurs here because exposed workers receive routine annual liver function tests while unexposed workers only get tested when symptomatic. This differential surveillance means liver disease will be detected earlier and more frequently in the exposed group, even if the true disease rates were identical. This systematic over-detection in the exposed group will overestimate the association between chemical X and liver disease.
Alcohol consumption acts as a confounder because it's associated with both the exposure (higher rates in exposed workers) and the outcome (alcohol causes liver disease). Since exposed workers drink more alcohol, and alcohol independently increases liver disease risk, this confounding will also overestimate the apparent effect of chemical X on liver disease.
Answer A incorrectly suggests alcohol consumption underestimates the association. Since exposed workers have higher alcohol consumption, confounding by alcohol would strengthen, not weaken, the observed association. Answer C wrongly claims detection bias underestimates the association—differential surveillance always inflates apparent disease rates in the more intensively monitored group. Answer D ignores the clear detection bias from differential testing protocols.
When analyzing epidemiological studies, always identify the direction of bias and confounding effects. Ask yourself: "Does this factor make the exposure group appear to have more or less disease than they actually do?"
- A cohort study found that coffee consumption was associated with reduced risk of liver cancer (RR = 0.6, 95% CI: 0.4-0.9). However, coffee drinkers in the study were also more likely to be non-smokers, exercise regularly, and have higher socioeconomic status. After stratifying by smoking status, the association remained significant in non-smokers (RR = 0.7) but disappeared in smokers (RR = 1.1). What does this pattern suggest?
Answer: Smoking is an effect modifier, creating different coffee-cancer relationships in smokers versus non-smokers
The different effect estimates in smokers (RR=1.1) versus non-smokers (RR=0.7) indicates effect modification (interaction) rather than simple confounding. If smoking were just a confounder, we would expect similar effect estimates in both strata after stratification. Choice A incorrectly interprets this as confounding control. Choice C mentions residual confounding but doesn't explain the stratum-specific differences. Choice D introduces selection bias without evidence.
- A cohort study follows healthcare workers to examine the relationship between shift work and cardiovascular disease. During the study period, night shift workers are more likely to leave their jobs and drop out of the study compared to day shift workers. Among those who remain, night shift workers show higher cardiovascular disease rates. How should researchers interpret these findings?
Answer: Selection bias from differential dropout may underestimate the true association
When you encounter questions about differential dropout in cohort studies, think about selection bias and how it affects the remaining study population. This is a classic epidemiological concept where participants systematically leave the study in ways that distort your results.
In this scenario, night shift workers are dropping out at higher rates than day shift workers. This creates selection bias because the night shift workers who remain in the study may not be representative of all night shift workers. Specifically, the workers who stay are likely the "healthier worker survivors" - those robust enough to continue both night shift work and study participation despite health challenges.
Since the sickest night shift workers are leaving the study, you're left comparing relatively healthy night shift workers to a more representative sample of day shift workers. This makes the night shift group appear artificially healthier than they truly are, causing you to underestimate the real association between night shift work and cardiovascular disease.
Answer B correctly identifies this underestimation. Answer A is wrong because differential dropout creates bias, not unbiased evidence. Answer C incorrectly suggests overestimation - but we're losing the sickest night workers, which would reduce the apparent association. Answer D is wrong because this is selection bias from dropout, not information bias from different follow-up methods.
Remember: when participants with higher disease risk systematically leave your exposed group, you'll underestimate the true association. Think "healthy worker survivor effect" whenever you see differential dropout patterns.
- A researcher studying the effectiveness of a new diabetes medication conducts a retrospective analysis of electronic health records. Patients receiving the new medication show better glycemic control than those on standard therapy. However, the new medication is expensive and typically prescribed only to patients with good insurance coverage, who also tend to have better access to diabetes education and specialist care. What analytical strategy would best address the primary validity concern?
Answer: Adjust for insurance status and access to care variables in multivariable analysis
When you encounter a retrospective study showing treatment differences, immediately consider confounding—systematic differences between groups that could explain the observed effect rather than the treatment itself. Here, the expensive medication creates a classic confounding scenario where insurance status determines both treatment assignment and access to other beneficial care.
The best approach is multivariable analysis that adjusts for these confounding variables (option B). By including insurance status, access to diabetes education, and specialist care as covariates in your statistical model, you can estimate the medication's effect while controlling for these systematic differences. This allows you to isolate the true treatment effect from the influence of confounders.
Option A (excluding poorly insured patients) eliminates valuable data and doesn't solve the problem—even among well-insured patients, there may be residual confounding. Option C (increasing sample size) won't help because the systematic bias remains regardless of sample size; you'll just have more precise estimates of a biased effect. Option D (restricting to same providers) addresses only one potential confounder while ignoring insurance status and patient access factors.
The key insight is that retrospective studies often suffer from confounding by indication—patients receive different treatments for systematic reasons related to their prognosis or characteristics. Simply observing differences doesn't prove causation.
Study tip: In biostatistics questions about observational studies, when you see systematic differences between treatment groups, think "confounding" and look for the answer that adjusts for or controls these differences, not one that ignores or avoids them.
- Researchers conducted a case-control study of pancreatic cancer by reviewing medical records. Cases were identified through a state cancer registry, while controls were selected from the same hospitals where cases were treated. The study found an association between diabetes and pancreatic cancer (OR = 2.8). However, diabetes documentation was more complete in the records of cancer patients than in records of controls seeking routine care. What is the MOST likely impact on the study results?
Answer: Information bias due to differential recording of diabetes history will likely overestimate the true association
When evaluating bias in case-control studies, you need to carefully consider how differences in data collection between cases and controls might affect your results. Information bias occurs when there are systematic differences in how exposure information is gathered or recorded between the two groups.
In this scenario, diabetes documentation is more complete in cancer patients' records than in controls' records. This differential misclassification creates a serious problem: cases (cancer patients) are more likely to have their diabetes status accurately recorded, while controls are more likely to have diabetes missed or undocumented. This means diabetes exposure appears artificially higher in cases and artificially lower in controls, inflating the odds ratio beyond the true association.
Let's examine why the other options don't fit. Option A incorrectly identifies recall bias, but this is a medical records study with no patient interviews involved. Option B mentions selection bias from hospital controls, but hospital controls typically bias results toward the null (making associations appear weaker), not away from it, and the question specifically describes information bias. Option D suggests confounding by healthcare utilization, but the issue here isn't that sicker patients use more healthcare—it's that their medical records are more thoroughly documented when they have cancer.
Study tip: In case-control studies using medical records, always ask yourself whether exposure documentation might differ systematically between cases and controls. More thorough documentation in cases will typically overestimate associations, while more thorough documentation in controls will underestimate them.
- A case-control study of skin cancer and sunscreen use interviews patients at a dermatology clinic (cases) and patients at an orthopedic clinic (controls). Both groups are asked about their sunscreen use over the past 10 years. Dermatology patients report lower sunscreen use than orthopedic patients. Which combination of biases is MOST likely affecting this study?
Answer: Selection bias, recall bias, and information bias
When evaluating case-control studies, you need to systematically assess potential sources of bias that could invalidate the findings. This study compares skin cancer patients (cases) from a dermatology clinic with controls from an orthopedic clinic, asking both groups to recall sunscreen use over the past 10 years.
Two major biases are clearly operating here. Selection bias occurs because the control group isn't representative of the population that produced the cases. Orthopedic patients may systematically differ from the general population in ways related to sun exposure (perhaps more athletic, outdoorsy individuals). Recall bias is also present because skin cancer patients are likely to remember and report their sunscreen habits differently than healthy controls—they may underestimate past sunscreen use or overestimate sun exposure because they're searching for explanations for their diagnosis.
Looking at the wrong answers: Choice A mentions "information bias," which is too vague—recall bias is the specific type of information bias present. Choice B excludes selection bias, which is clearly problematic given the non-representative control group. Choice D references "confounding bias," but confounding refers to unmeasured variables that affect the outcome, not the biases inherent in the study design itself.
The question asks for the "combination of biases," and since none of the provided options (A-D) correctly identifies both selection bias and recall bias together, the answer must be E (not shown but implied to be the correct pairing).
Study tip: In case-control studies, always evaluate whether controls represent the source population of cases and whether disease status could affect recall of exposures.
- A case-control study investigating the association between cell phone use and brain tumors recruits cases from a specialized neurosurgery center and controls from the general population through random digit dialing. Cases are interviewed in person at the hospital, while controls are interviewed by telephone. Which type of bias is MOST likely to affect the validity of this study's results?
Answer: Information bias due to different interview methods for cases and controls
When evaluating potential biases in epidemiological studies, you need to examine how the study design might systematically distort the relationship between exposure and outcome. This question tests your ability to identify the most problematic methodological flaw.
The key issue here is that cases and controls are being interviewed using completely different methods - cases in person at the hospital versus controls by telephone. This creates a systematic difference in how exposure information is collected between the two groups, which is the hallmark of information bias. In-person interviews typically yield more detailed responses and allow for clarification, while telephone interviews may be more superficial. This differential data collection quality could artificially create or mask associations between cell phone use and brain tumors.
Option A describes selection bias, but while participation rates might differ, this isn't the most direct threat given the clear methodological differences in data collection. Option C identifies recall bias, which could certainly occur since cases might be more motivated to remember past exposures, but this is secondary to the fundamental problem of different interview methods. Option D mentions confounding by age, but there's no indication in the scenario that age distributions differ systematically between groups recruited this way.
The correct answer is B because information bias due to differential data collection methods represents the most immediate and systematic threat to validity.
Study tip: In bias identification questions, always look first for systematic differences in how data is collected between comparison groups - this usually trumps other potential biases in terms of impact on study validity.
- A cross-sectional study examining the relationship between vitamin D levels and depression used an automated laboratory assay that systematically underestimated vitamin D concentrations by 15% for all samples. Additionally, participants with depression were more likely to be recruited from psychiatric clinics, while non-depressed participants were recruited from routine health screenings. Which statement BEST describes the bias implications?
Answer: The systematic measurement error is non-differential and won't bias the association, but the recruitment strategy creates differential selection bias
The 15% underestimation affects all samples equally (non-differential misclassification), which typically biases toward the null but doesn't create false associations. However, recruiting depressed participants from psychiatric clinics versus controls from health screenings creates selection bias that could create or exaggerate associations. Choice A incorrectly states measurement error biases toward null while selection biases away. Choice C is wrong about both being differential. Choice D incorrectly suggests selection bias only affects generalizability.
- In a case-control study of birth defects and maternal medication use during pregnancy, researchers find that mothers of affected children provide more detailed medication histories than control mothers, including over-the-counter drugs and supplements they might have forgotten initially. This differential reporting pattern represents which type of bias?
Answer: Recall bias because cases have enhanced memory for relevant exposures
When analyzing bias in epidemiological studies, you need to distinguish between different types based on when and how the bias occurs in the study process. This question tests your ability to identify recall bias, a specific subtype of information bias.
The scenario describes mothers of children with birth defects providing more comprehensive medication histories than control mothers. This enhanced reporting occurs because cases have a stronger motivation to remember and report exposures that might explain their child's condition. This is classic recall bias - systematic differences in how accurately study participants remember and report past exposures based on their disease status.
Looking at the incorrect options: Choice A describes selection bias, but the issue isn't about who participates in the study - it's about how participants report information once enrolled. Choice B identifies information bias correctly as the broad category, but recall bias (choice C) is the more specific and precise term for this particular pattern. Choice D suggests measurement bias from different assessment tools, but the scenario indicates the same interview process is used for both groups - the difference lies in how participants respond, not in the measurement instruments themselves.
Remember that recall bias is especially problematic in case-control studies because cases often have heightened awareness of potential risk factors. When you see scenarios where cases provide more detailed or different quality information about past exposures compared to controls, think recall bias first. This helps distinguish it from other information biases that might involve interviewer differences or measurement errors.
- A prospective study examining exercise and depression recruits participants through fitness centers and mental health clinics. After two years, researchers find that participants recruited from fitness centers have lower depression rates. However, 40% of fitness center participants dropped out compared to 15% from mental health clinics. What is the primary concern for interpreting these results?
Answer: Selection bias due to differential loss to follow-up between groups
When evaluating prospective studies, you need to carefully assess whether the groups being compared remain representative throughout the study period. The key issue here is what happens to your study population over time.
The dramatic difference in dropout rates (40% from fitness centers vs. 15% from mental health clinics) creates selection bias due to differential loss to follow-up. This is problematic because the people who drop out may systematically differ from those who stay, and this difference varies by group. For example, fitness center participants who develop depression might be more likely to drop out, leaving behind a healthier subset. Meanwhile, mental health clinic participants might stay in the study regardless of their depression status. This means you're no longer comparing equivalent groups by the end of the study.
Option A addresses confounding, but while baseline differences likely exist, the question emphasizes the dropout problem as the primary interpretive concern. Option C suggests information bias from different measurement methods, but there's no indication that depression was measured differently between groups. Option D mentions recall bias about exercise habits, but this is a prospective study where exercise would be measured going forward, not recalled from the past.
The substantial differential dropout has likely created two very different final study populations than what you started with, making any comparison between groups potentially meaningless.
Study tip: In prospective studies, always check the dropout rates between groups. High differential loss to follow-up (typically >20% difference) should immediately raise red flags about selection bias affecting your results.
- In a randomized controlled trial of a new antihypertensive medication, participants assigned to the treatment group experienced more frequent clinic visits for safety monitoring than those in the control group. At the end of the study, the treatment group had significantly more documented cardiovascular events. What is the MOST likely explanation for this finding?
Answer: Detection bias due to increased surveillance in the treatment group leading to more frequent identification of events
The increased clinic visits in the treatment group create detection bias (surveillance bias) - more opportunities to detect and document cardiovascular events compared to the control group. This differential surveillance can create apparent differences in event rates. Choice B (performance bias) relates to behavioral changes, not detection differences. Choice C (selection bias) is prevented by proper randomization. Choice D (attrition bias) isn't mentioned in the scenario.
- A case-control study of lung cancer included cases diagnosed at a university hospital and controls selected from patients visiting the same hospital for routine procedures. The study found a strong association between residential radon exposure and lung cancer (OR = 3.2). However, the university hospital primarily serves a rural population with older homes, while urban residents with newer homes typically use other healthcare facilities. How does this design choice MOST likely affect the study results?
Answer: Creates selection bias that may overestimate the radon-lung cancer association due to restricted population representation
Using one hospital that serves a specific population (rural, older homes) creates selection bias by restricting the study to a non-representative population where radon exposure may be higher. This limits generalizability and may inflate the association. Choice B describes confounding but this is more fundamentally a selection issue. Choice C mentions recall bias but doesn't address the core problem of population restriction. Choice D mentions misclassification without clear evidence.
