2027 Missing Data and Loss to Follow-Up: Questions to Ask When Reading Nursing Studies
Audited and Verified:
Our Editorial and Data Verification Process
Research.com uses government datasets, official education statistics and primary institutional information to support its career and degree guides. Our review process checks factual accuracy, occupational definitions, source dates and the requirements associated with the credentials discussed.
How This Article Was Fact-Checked
Our editorial team verifies salary estimates, employment projections, education requirements and program information against original data providers and official organizations. Salary figures are checked against the relevant occupational definitions and reporting periods. Information about licensure, accreditation and required in-person training is reviewed separately.
Fact-checked and verified on:
A nursing study may report a promising improvement in patient outcomes, yet its conclusion depends partly on whose outcomes were actually measured. If participants leave the study or records are incomplete, the reported result may not represent everyone who started. For nursing students, clinicians, and graduate researchers reading evidence for coursework or practice decisions, the useful question is not simply whether data are missing. It is whether the amount, pattern, and handling of missing data could change the conclusion. This guide shows you where to check and how to judge what a study can support.
Key Things You Should Know
- Compare the number enrolled with the number analyzed for each outcome, group, and follow-up time; one overall dropout percentage can hide an important imbalance.
- There is no universally acceptable loss-to-follow-up cutoff. Even a small amount can bias results if missingness is related to patients' outcomes or differs between groups.
- Check why outcomes are missing, what assumptions the analysis makes, and whether sensitivity analyses show the conclusion survives plausible alternative assumptions.
What Are Missing Data and Loss to Follow-Up in Nursing Studies?
Missing data are information a study intended to collect but did not obtain. A patient may have a baseline pain score but no discharge score; a nurse may complete an intervention survey but skip a burnout question. Loss to follow-up is narrower: researchers cannot obtain a participant's planned later assessment. Someone can remain in a study while leaving one question unanswered, so the terms are not interchangeable.
Consider a hypothetical nurse-led discharge study that enrolls 100 patients in each group and measures medication understanding at discharge and 30 days later. If all 200 complete the discharge measure but only 170 complete the 30-day measure, the later outcome has 30 missing assessments. To evaluate the result, you still need to know which patients are missing, from which group, and why.
Also distinguish a missing measurement from an outcome that did not occur. "No readmission documented in this hospital's record" is not necessarily the same as "no readmission," especially if patients can receive care elsewhere.
Prospective students may have experience with missing data, especially when comparing entry barriers at nursing schools. A lack of published graduate outcomes or student support can prevent you from drawing the necessary conclusions to arrive at a decision. In nursing studies, this missing data can lead to faulty generalizations or wrong assumptions.
Why Do Missing Data and Participant Dropout Matter in Nursing Research?
Missing outcomes reduce the information available to estimate an effect and can make confidence intervals wider. The more consequential problem is selection bias: participants with observed outcomes may differ systematically from those without them. For example, patients struggling with a new self-management plan may be less likely to return a follow-up survey. A result based only on respondents could then overstate the plan's benefit. This is especially crucial to consider in nontraditional practice formats in RN-to-BSN programs.
The clinical setting helps determine what missingness might mean. Patients discharged to another facility, home-health clients who cannot be reached, and nurses who leave a high-turnover unit may each be missing for reasons related to the outcome under study. A statistically precise result does not resolve that concern if the analysis relies on an implausible account of the missing observations.
This distinction matters when deciding whether evidence is strong enough for a class paper, an evidence-based practice proposal, or a change in local care. A study can be useful while still warranting a narrower conclusion than its abstract suggests.

How Can Loss to Follow-Up Bias Nursing Study Results?
Bias arises when the people who remain under observation are not representative of those who began the study in ways that matter to its result. Random assignment protects the initial comparison between trial groups; it does not guarantee that the groups remain comparable after outcome data are lost.
Here is a hypothetical example, not a finding from a published nursing study. Both groups start with 100 patients, and the outcome is an unfavorable event:
| Group | Patients enrolled | Patients with measured outcomes | Events among measured patients |
| Nurse-led intervention | 100 | 70 | 14 of 70 (20%) |
| Usual care | 100 | 90 | 9 of 90 (10%) |
The observed data suggest more events in the intervention group. But outcomes are unknown for 30 intervention patients and 10 usual-care patients. If none of the 30 missing intervention patients had an event while all 10 missing usual-care patients did, the full-group event counts would be 14 versus 19. That deliberately extreme scenario does not establish what happened; it demonstrates that the missing outcomes could change the direction of the comparison.
When reading a study, ask whether a more clinically plausible range of outcomes among missing patients could materially change the estimate - not merely whether researchers reported a statistically significant result among those they measured. Developing that appraisal habit early—even while reviewing entry-level nursing degree comparisons as a prospective clinician—helps you evaluate clinical evidence critically from the start of your training.
First identify what is missing. Baseline characteristics, exposure or intervention records, repeated measurements, and final outcomes have different implications. A missing baseline risk factor can weaken adjustment for confounding; a missing primary outcome can directly change the main result. Missingness may also occur at a single visit, in one survey item, or across every later visit after a participant withdraws.
Researchers often describe the likely mechanism using three terms. These are assumptions about why values are missing, not labels that a dataset can prove on its own:
- Missing completely at random: the chance of missingness is unrelated to observed or unobserved values relevant to the analysis.
- Missing at random: after accounting for observed information - such as baseline health or prior survey responses - missingness does not additionally depend on the unobserved value itself.
- Missing not at random: missingness still depends on the unobserved value, even after accounting for the observed information.
For a nursing example, patients with higher recorded baseline symptom scores might be harder to reach; accounting for those scores could make a missing-at-random analysis more credible. If patients are less likely to respond specifically because their unrecorded current symptoms worsened, that assumption may fail. The reason is often uncertain, which is why sensitivity analyses matter—a practical distinction worth learning across all nursing pathways, including for prospective students exploring faster routes into practical nursing education.
No single percentage makes loss to follow-up acceptable or unacceptable. Its importance depends on the size of the expected effect, the number of missing outcomes, whether missingness differs by group, and whether the reasons for missingness are associated with the outcome. Losing a few participants can matter in a small study or one with a rare event; a larger proportion may be less threatening if outcome information is recovered through a dependable alternative source and the remaining uncertainty is addressed.
Calculate the proportion separately for each group and time point: number without the required outcome ÷ number expected to provide it. Keep the original eligible or assigned group in view rather than using only the people who completed follow-up as the denominator. For a 12-month outcome, do not count someone as lost merely because the article has not yet reached that person's scheduled 12-month assessment; look for the study's stated follow-up window.
Be wary of applying a familiar percentage threshold as a pass-fail rule. Much like applicants comparing GPA requirements in accelerated BSN programs must look past a single numeric cutoff to gauge overall readiness, a more useful test for attrition is whether plausible outcomes among those missing could erase, substantially shrink, or reverse the reported effect.

Start with the participant-flow diagram or results section, then reconcile its counts with the tables. Trial reports may follow CONSORT reporting guidance, while observational studies may use STROBE guidance; neither label guarantees that every missing-data question has been answered. Supplementary materials and a study protocol can add details that the abstract omits.
Check these locations in sequence so you can trace who contributed to the reported result:
- Methods: Find the planned follow-up schedule, primary outcome, rules for including participants, and proposed missing-data analysis.
- Participant flow: Record how many were enrolled or assigned, received the intervention, completed each follow-up, and entered each analysis.
- Tables and figures: Compare denominators for each outcome and time point; they may differ even within the same group.
- Results, supplement, and limitations: Look for dropout reasons, comparisons between respondents and nonrespondents, and sensitivity analyses.
If the abstract says "200 patients were studied" but the primary outcome table includes 142, use the table's actual denominator when describing that finding. If the counts cannot be reconciled, note the reporting gap rather than filling it with an assumption. Checking full documentation carefully—similar to verifying exact admissions criteria when planning nursing applications without TEAS or HESI—ensures you rely on verified figures rather than incomplete summaries.
Common explanations include withdrawal of consent, illness or death, transfer of care, inability to contact a participant, scheduling or transportation barriers, and incomplete questionnaires. In nurse-focused studies, staff turnover, shift changes, and competing work demands may interrupt follow-up. These categories are more informative when reported by study group and, when relevant, by assessment time.
A reason does not automatically remove the risk of bias. "Unable to contact" may conceal a change in housing, health, or care access related to the outcome. Death may make a quality-of-life measure impossible to obtain, but excluding those patients without explanation can distort interpretation. Stopping an intervention is also different from withdrawing permission to measure outcomes: researchers may still be able to obtain a later outcome under the study's approved procedures.
Look for vague categories such as "other" or "protocol deviation" that contain a large share of the missing participants. In studies involving working clinicians—such as nurses juggling clinical shifts while comparing fast LPN-to-RN nursing programs—schedule-driven attrition can easily be masked by generic labels. If a report gives no reasons, you cannot judge whether dropout was likely unrelated to the outcome; that uncertainty belongs in your appraisal.
Researchers can analyze only observed cases, use models that draw on available repeated measurements, apply weighting, or estimate missing values through methods such as multiple imputation. Each approach has conditions under which it is useful; none reconstructs an outcome that was never measured with certainty.
Complete-case analysis is easy to understand but can discard substantial information and produce bias when people with complete records differ from others. Multiple imputation creates several plausible completed datasets using observed information and combines their estimates. Its credibility depends on the variables included and assumptions about missingness. Repeated-measures models and inverse-probability weighting likewise require readers to assess whether their assumptions fit the clinical setting.
Check whether researchers performed a sensitivity analysis that asks how results change under alternative, clinically plausible assumptions - particularly when missing outcomes may depend on unobserved health status. Carrying a patient's last observed score forward is not inherently conservative; symptoms can improve or worsen after that visit. Likewise, saying a trial used an "intention-to-treat" analysis does not explain how it dealt with missing outcomes among assigned participants. Building familiarity with these foundational biostatistics concepts—often highlighted during direct-entry MSN prerequisite comparisons—helps readers recognize when analytical assumptions need closer scrutiny.
Finally, distinguish an outcome that is missing from one that cannot occur in the same way after a competing event, such as death before a later symptom survey. That situation may require a different outcome definition or analysis, not simply an imputed survey score.
Judge the reported effect and its uncertainty together with the missing-data problem. A credible interpretation connects the study design, outcome, participant flow, reasons for missingness, statistical assumptions, and sensitivity analyses. Strong reporting makes these checks possible; it does not by itself prove that the result is unbiased.
Confidence is stronger when follow-up is clearly reported for each group, outcome data are obtained from appropriate sources, reasons for missingness are documented, and the main finding remains similar under reasonable alternative analyses. Confidence falls when attrition is unexplained or uneven, the primary result relies on a small subset of those enrolled, or modest assumptions about missing patients could change the conclusion.
For an evidence-based practice decision, also ask whether the measured outcome reflects something meaningful in your patient population or work setting. A statistically robust survey response may still be less relevant than an adequately measured patient outcome. Conversely, a study with some missing data may still contribute useful evidence if you describe its limitations and weigh it alongside other studies rather than treating it as decisive on its own—an appraisal skill emphasized across advanced nursing curricula, including lower-cost online BSN-to-MSN degree routes.
What Questions Should You Ask Before Citing a Nursing Study?
Before using a study to support a nursing assignment, literature review, or practice proposal, answer the following questions from the full text. They help you decide both whether to cite the finding and how strongly to word your claim:
- What was the primary outcome, when was it measured, and how many participants were expected to provide it?
- How many provided usable data for that outcome in each group, and do the counts agree across the flow diagram, text, and tables?
- What reasons were given for missing outcomes, and could those reasons be related to the outcome?
- Did the analysis account for relevant baseline differences or earlier measurements, and what assumptions did its missing-data method require?
- Would plausible outcomes among missing participants materially change the result, and did the authors test that possibility?
- Does your proposed sentence reflect the analyzed participants and the uncertainty, rather than implying that every enrolled participant had a measured benefit?
If key answers are unavailable, cite cautiously: for example, say an intervention "was associated with improvement among participants with follow-up data" rather than that it "improved outcomes for all enrolled patients." For graduate research—whether you are already completing coursework or mapping out ADN-to-MSN program planning—record the missing-data limitation in your evidence table so it remains visible when you compare studies. If the finding is central to a consequential decision, seek clarification from the authors or additional evidence instead of treating an unreported denominator as a minor detail.
Other Things You Should Know About Nursing
No. It may mean the event was not documented in that system. Check the study's outcome definition and whether it captured care outside the organization.
They can provide some outcomes without a new patient visit, but only for events the records reliably capture. Coverage, linkage, and documentation gaps still need evaluation.
Not automatically. Reviewers can assess risk of bias and test how excluding weaker studies affects conclusions, but pooling results does not recover participants' unmeasured outcomes.
Publication does not grant access to restricted data. Use an authorized public dataset or obtain the required permissions and institutional guidance before accessing patient-level information.
- Reporting attrition in randomised controlled trials - PMC - NIH https://pmc.ncbi.nlm.nih.gov/articles/PMC1444839/
- Missing data: Issues, concepts, methods | sciencedirect.com https://www.sciencedirect.com/science/article/pii/S1073874624000082
- Missing Data and Multiple Imputation | Columbia University Mailman School of Public Health https://www.publichealth.columbia.edu/research/population-health-methods/missing-data-and-multiple-imputation
- Item 21c: Missing data | Consort Spirit https://www.consort-spirit.org/item21c-missingdata
- Chapter 1: Introduction | Cochrane https://www.cochrane.org/authors/handbooks-and-manuals/handbook/archive/v5.2.0
- Read "The Prevention and Treatment of Missing Data in Clinical Trials" at NAP.edu https://www.nationalacademies.org/read/12955/chapter/6