03 · What You Need to Know
Understanding Missing Care, Missing Records, and Healthcare Observation
What Does Missing Care Actually Mean?
Missing care refers to a situation in which a healthcare service that was expected, indicated, scheduled, or otherwise relevant to a defined standard was not delivered.
However, the meaning of expected care must be specified. A missed scheduled appointment, an overdue recommended screening, and the absence of a clinically indicated treatment are not necessarily equivalent events.
A patient may appropriately decline a service, a clinician may determine that an intervention is contraindicated, or a treatment plan may change. These circumstances should not automatically be classified as healthcare failure.
For research purposes, define the care expectation using a relevant clinical guideline, documented care plan, eligibility rule, or explicitly stated study criterion.
For example, a study of follow-up attendance might define a missed appointment as a scheduled encounter that the patient did not attend. A study of recommended preventive screening requires additional information about eligibility, timing, and whether screening occurred elsewhere.
The operational definition should match the question being asked rather than rely on a general label such as "missing care."
What Does Missing Data Mean in Healthcare Records?
Missing data occur when information required to determine a variable or event is not available in the accessible dataset.
In healthcare research, this may involve a blank laboratory field, an unavailable consultation note, a missing claims record, or an encounter that occurred outside the contributing institution.
Sometimes the database contains an encounter but lacks a specific measurement. In other cases, the entire encounter is absent.
These situations require different assessments because conventional missing-value methods generally operate on observed units with incomplete fields, while entirely unobserved encounters may not be represented as missing rows at all.
The broader methodological issue is incomplete documentation in routine records, but the present distinction concerns whether an apparent absence of healthcare represents actual nonreceipt or incomplete observation.
Distinguish Four Possible Explanations for an Absent Care Record
| Situation |
What happened? |
What the database may show |
| Care was not delivered |
The expected service did not occur |
No qualifying encounter or procedure |
| Care occurred outside the system |
The patient received care from another provider |
No record in the accessible institution |
| Care occurred but was not documented adequately |
The service was delivered, but the relevant record was omitted or incomplete |
No qualifying structured entry |
| Care was not applicable or no longer indicated |
The patient did not require the service under the applicable circumstances |
No qualifying encounter or procedure |
The same observed database pattern can arise from substantively different circumstances. Researchers should not treat these explanations as interchangeable.
Why Healthcare System Boundaries Matter
Routine health records capture activity within defined organizational, contractual, or technical boundaries.
A hospital electronic health record may contain detailed information about care delivered at that hospital but little information about consultations elsewhere. An insurance claims database may capture reimbursed services during covered periods but omit services paid through other arrangements.
Even an integrated health system may not observe every service a patient receives.
Before interpreting absent encounters, establish which providers contribute records, which service types are included, and whether the patient was observable throughout the relevant period.
Coverage may vary across individuals, institutions, and time. A patient who changes insurers or receives care outside a participating network may appear to have stopped receiving care when only the data source changed.
Where care is distributed across systems, linking records from multiple sources may improve ascertainment, provided the required identifiers, permissions, and linkage-quality safeguards are available.
Establish Whether the Patient Was Observable
An observation period is the interval during which the available data source could reasonably capture the events being studied.
For claims data, enrollment or insurance eligibility information may help establish periods of observable covered care. For institutional electronic records, evidence of registration or encounters may indicate contact with the system, although it does not necessarily establish continuous or complete capture of all care.
Researchers should distinguish three dates: when the patient entered the observable population, when the service was expected, and when the available observation ended.
Suppose a patient was due for a follow-up appointment in December, but the dataset contains records only through October. The absence of a December encounter cannot establish nonattendance.
Likewise, an encounter-free period does not necessarily mean that the patient remained within the data source's effective observation coverage.
Where observability is uncertain, researchers should avoid automatically classifying all event-free intervals as confirmed periods without care.
Distinguish Scheduled Care, Delivered Care, and Recorded Care
Scheduling information, clinical encounter records, and billing transactions may represent different stages of healthcare delivery.
Scheduled care
A service or encounter was planned, but its completion has not necessarily been established.
Delivered care
The service was actually provided, according to an appropriate source of evidence.
Recorded care
A qualifying entry exists in the accessible database, although its precise meaning depends on the record type.
A scheduled appointment does not prove attendance. A completed encounter entry may provide stronger evidence of delivery, but researchers should verify the meaning of local completion statuses.
A billing claim may indicate that a service was submitted for reimbursement, while a medication order may establish prescribing rather than administration or consumption.
Defensible research requires matching the operational measure to the actual event of interest.
How Missing Care and Missing Data Can Produce Different Forms of Bias
When genuine nonreceipt of care is the outcome, identifying it accurately is essential to the study's substantive conclusions.
If care delivered elsewhere is incorrectly classified as missing care, the study may overestimate unmet need or apparent service discontinuity.
If patients who receive little care are less likely to remain visible in the database, excluding them may underestimate barriers to access.
Documentation practices can also differ across demographic groups, providers, or institutions. This may create differential misclassification if the probability of recording care is associated with the exposure or outcome being studied.
For example, a study comparing telehealth users with nonusers could be biased if telehealth encounters are systematically stored in a separate system for some institutions.
The direction of bias depends on the data-generating process. Missing care records do not necessarily inflate or attenuate every estimated association in the same way.
How to Investigate Whether Care Actually Occurred
Begin by examining the source system's recording rules and the types of encounters it captures.
Where feasible, compare information across available sources, such as encounter logs, procedure records, pharmacy dispensing data, referral systems, claims, discharge summaries, and relevant clinical notes.
For example, a procedure absent from a structured procedure table might be documented in an operative report or an external claim.
Consider whether the information sources are independent. Two database fields generated from the same original transaction do not necessarily provide independent confirmation.
Targeted record review can help estimate the frequency of undocumented care, although validation samples should be selected carefully to avoid representing only patients with unusually complete records.
When external data are unavailable, distinguish what is established from what remains uncertain rather than inferring nonreceipt solely from an empty field.
Use Negative Evidence Only When the Recording Process Supports It
An absent record can sometimes provide meaningful negative evidence.
For example, if a system is demonstrably complete for a defined type of reimbursed service during continuous coverage, the absence of a qualifying claim may support the conclusion that no covered service was billed through that system.
That is still narrower than concluding that no care was received anywhere.
The strength of negative evidence depends on the data source's scope, capture completeness, coding validity, and observation period.
This reasoning is closely related to why the absence of a recorded diagnosis does not necessarily establish disease absence.
Decide How to Classify Uncertain Care Status
Depending on the study design, researchers may distinguish confirmed service receipt, confirmed nonreceipt, and uncertain status.
An explicit uncertain category can be useful during data preparation because it prevents unobserved care from being silently coded as zero.
However, retaining uncertainty as a category is not automatically a valid inferential solution. Its treatment in analysis depends on why care status is uncertain and how uncertainty relates to the variables being studied.
Possible approaches include restricting analyses to periods with adequate observation, using supplementary sources, applying appropriate missing-data or measurement-error methods, and conducting sensitivity analyses.
These approaches require defensible assumptions. Statistical imputation cannot automatically reconstruct an entire set of healthcare encounters that never entered the source database.
Report the Scope of Care Ascertainment Precisely
Use language that reflects what the data actually establish.
If the study uses hospital records alone, "no recorded follow-up encounter at the participating hospital" may be more accurate than "the patient received no follow-up care."
If claims are restricted to reimbursed services, report that scope rather than implying comprehensive observation of all healthcare activity.
Describe the expected-care definition, observation period, data sources, exclusions, and procedures used to investigate missing records.
The RECORD statement provides reporting guidance for observational studies using routinely collected health data, including the characteristics of data sources and relevant linkage procedures.
