Nursing informatics data standards and devices

To show community leaders where high diabetes rates overlap with food deserts across neighborhoods, the most effective visualization is a:

1. run chart.
   A run chart tracks one measure over time; the question is about where things overlap, not when they change.
2. scatter plot.
   A scatter plot can show that two variables are correlated across neighborhoods, but it drops the map, so leaders cannot see where the clusters are.
3. geographic heat map.
   A geographic (choropleth) heat map shades each neighborhood by its value, so overlapping clusters of diabetes and food deserts are visible at a glance to community leaders.
4. pie chart by neighborhood.
   A pie chart shows parts of a single whole and cannot display two variables or their spatial overlap.

Heart-rate data from patients' wearables contain gaps and implausible spikes caused by loose sensors. Which characteristic of big data does this problem concern?

1. Velocity
   Velocity is the speed at which data arrive and must be processed; fast arrival does not explain implausible values.
2. Variety
   Variety refers to many different data types and formats; the problem here is the quality of one data type.
3. Veracity
   Veracity is the trustworthiness and quality of incoming data; gaps and artifact spikes from loose sensors are a veracity problem.
4. Volume
   Volume is the sheer amount of data; continuous wearable streams are large, but size is not what makes these readings unreliable.

A hospital sends a complete, human-readable discharge summary that is persistent, can be legally authenticated, and contains coded sections. Which standard is best suited?

1. HL7 C-CDA
   The Consolidated CDA is HL7's document standard: a persistent, human-readable, legally authenticable clinical document with coded sections, such as a discharge summary.
2. A DICOM structured report
   DICOM structured reports belong to the imaging world and carry imaging findings, not hospital discharge summaries.
3. An HL7 v2 ORU message
   An ORU message carries observation results as a transient message between systems; it is not a signed, persistent document.
4. A FHIR Observation
   A FHIR Observation is one discrete resource, such as a single vital sign, not a complete authenticated summary.

Registration staff often create new records for returning patients whose names were misspelled. Which control best reduces duplicate records?

1. A multi-field search for an existing record before adding one
   Requiring registrars to search on several fields, such as name, date of birth, and phone, catches misspelled returning patients before a duplicate is created.
2. A monthly report of possible duplicates for HIM review
   A duplicate report detects records after they are created; it cleans up the problem rather than preventing it.
3. Requiring a Social Security number for every patient
   Requiring a Social Security number blocks registration for the many patients who lack or decline to give one, and does nothing about misspelled names.
4. Allowing registrars to merge records whenever names match
   Merging on matching names alone risks combining two different people, an overlay that is more dangerous than a duplicate.

In a data governance program, accountability for defining 'pressure injury present on admission' and monitoring the accuracy of that data element belongs to the:

1. database administrator.
   The database administrator runs the database technically; defining what a clinical data element means is outside that role.
2. IT data custodian.
   A custodian manages storage, security, and backup of data; confusing custodian with steward is the classic governance error.
3. clinical data steward.
   A data steward is the business or clinical owner accountable for an element's definition, quality, and appropriate use, such as a nurse expert defining present-on-admission pressure injury.
4. data governance committee chair.
   The governance committee sets policy and settles disputes, but day-to-day accountability for one element sits with its steward.

An encounters table has one row per visit. To report how many different patients were seen last month, the query should use:

1. COUNT(*) with GROUP BY patient_id
   Grouping by patient returns one row per patient with that patient's visit count, not a single total of different patients.
2. COUNT(*)
   COUNT(*) counts every row, so it returns the number of visits, not the number of patients.
3. COUNT(DISTINCT patient_id)
   COUNT(DISTINCT patient_id) counts each patient once, no matter how many visits that patient had.
4. COUNT(patient_id)
   COUNT(column) skips only nulls; it still counts a returning patient once per visit.

A nurse charts a heart rate of 72 from memory, while the monitor showed 112. The charted value passes every range check. Which data quality characteristic is deficient?

1. Completeness
   Completeness means required data are present; a heart rate was recorded, so nothing is missing.
2. Accuracy
   Accuracy is whether a value matches the true state; 72 charted from memory against a monitor reading of 112 is inaccurate even though it is plausible.
3. Timeliness
   Timeliness is not in question: the value was charted at the expected point; the problem is what was charted.
4. Validity
   Validity means a value fits the allowed format and range; the stem says the 72 passed every range check, so validity is met.

A readmission risk model trained mainly on insured urban patients performs poorly for rural uninsured patients. This problem is best described as:

1. data leakage from future variables.
   Data leakage, a model learning from information not available at prediction time, inflates measured accuracy for every group; it does not explain one under-represented group doing worse.
2. sampling bias in the model's development data.
   When the patients a model learns from do not represent the population it is used on, it performs worse for the under-represented group; that is bias from an unrepresentative sample.
3. model drift after deployment.
   Drift is performance decay as the population or practice changes over time; here the gap is built in from the start.
4. low sensitivity from a high alert threshold.
   A threshold setting would affect all patients alike; it does not explain why one population is served poorly.

A sepsis alert validated in the ICU is deployed on general medical units, where sepsis is much less common. Which measure is expected to fall?

1. Sensitivity of the alert rule
   Sensitivity is a property of the test among patients who have the condition and does not change just because prevalence is lower.
2. Positive predictive value
   Positive predictive value depends on prevalence: with fewer true sepsis cases on general units, a larger share of alerts are false positives.
3. Negative predictive value
   Negative predictive value rises, not falls, when a condition becomes rarer, because a negative result is even more likely to be correct.
4. Specificity
   Specificity is measured among patients without the condition, so in principle it stays stable as prevalence falls.

A nurse manager notices that falls entered this morning do not appear on the warehouse-based quality dashboard. What is the most likely explanation?

1. The falls were not saved in the EHR correctly
   Falls not saved in the EHR is possible for one entry, but it would not explain every fall entered this morning missing in the same way.
2. The warehouse is loaded in a nightly batch
   Data warehouses are usually refreshed by scheduled batch ETL, so events charted today appear only after the next load.
3. The dashboard filters out same-day falls
   A deliberate same-day filter is an unusual design; the pattern fits the normal load schedule of a warehouse.
4. The warehouse deletes data older than a day
   Warehouses keep history for trend analysis; deleting old data would remove past falls, not hide today's.

Two health information exchanges use compatible standards and terminologies but still cannot share data because their consent policies and legal agreements differ. Which interoperability level is lacking?

1. Foundational
   Foundational interoperability is the basic ability to send and receive data, which compatible standards already provide.
2. Organizational
   Organizational interoperability covers governance, policy, consent, legal agreements, and trust between organizations; differing consent rules block exchange at this level.
3. Structural (syntactic)
   Structural interoperability concerns the format and syntax of the message, already met by the compatible standards.
4. Semantic
   Semantic interoperability is about shared meaning through common terminologies, which the stem says the exchanges already have.

A nursing unit is renamed, and the new name must be changed in 300 rows because the unit name is stored in every patient-assignment record. This problem is best prevented by:

1. denormalizing the table for faster reporting.
   Denormalizing adds more redundancy, which makes this kind of update problem worse.
2. archiving assignment records older than a year.
   Archiving moves old rows elsewhere but does not change the design that copies the name into each record.
3. adding an index on the unit name column.
   An index speeds lookups but leaves the name repeated in every row, so the rename still touches all 300 rows.
4. normalizing the unit name into its own table.
   Normalization stores each unit name once and links assignment records to it by a key, so a rename touches one row and update anomalies disappear.

A fall-risk model scores 98% accuracy on its training data but only 70% on new patients from the same hospital. This problem is:

1. overfitting.
   A model that scores far higher on its training data than on new patients from the same setting has learned noise in the training set; that is overfitting.
2. underfitting.
   An underfit model is too simple and performs poorly on training data as well, not just on new patients.
3. model drift.
   Drift is decay over time as the population changes; the drop here appears immediately on new patients from the same hospital.
4. class imbalance.
   Class imbalance, too few falls among non-falls, can inflate accuracy, but it does not explain a large gap between training and new data.

Two hospitals both use HL7 v2.5, yet their laboratory result interface still needs custom field mapping. The most likely reason is that HL7 v2:

1. cannot carry coded laboratory results.
   HL7 v2 routinely carries coded laboratory results in the OBX segment.
2. requires a RESTful API for each exchange.
   RESTful APIs belong to FHIR; HL7 v2 messages typically travel over interface engines and MLLP.
3. sends complete documents rather than messages.
   HL7 v2 is a messaging standard; complete documents are the domain of CDA.
4. permits optional elements and local variation.
   HL7 v2 leaves many fields optional and allows site-specific use of segments and Z-segments, so two sites on the same version still need field mapping.

Many discharge dispositions are entered in the EHR several days after patients leave, so the daily bed availability report is wrong. Which data quality characteristic is most affected?

1. granularity
   Granularity is the level of detail captured, such as hourly versus daily; the disposition detail itself is adequate.
2. accuracy
   Accuracy is not the main failure: the dispositions are presumably correct once entered, and the report is wrong because they arrive late.
3. consistency
   Consistency means the same fact is represented the same way across systems; nothing suggests conflicting values.
4. timeliness
   Timeliness is whether data are recorded and available when needed; dispositions entered days late leave the daily bed report out of date.

During interface testing, a patient is admitted in registration but never appears on the nursing unit census. The interface to check first is the:

1. ADT (admission) feed.
   ADT messages carry admission, transfer, and discharge events from registration to downstream systems such as the unit census.
2. results (ORU) feed.
   ORU messages carry results such as lab values after a patient is already in the system.
3. pharmacy order (ORM) feed.
   ORM messages carry orders; a missing census entry is not an order problem.
4. imaging (DICOM) feed.
   DICOM moves medical images between modalities and PACS and has no role in placing a patient on a census.

Nursing assessment questions in a new flowsheet will be shared with other organizations. To support semantic interoperability, each assessment question should be mapped to:

1. CPT codes.
   CPT codes describe billable procedures and services, not the questions on a flowsheet.
2. RxNorm codes.
   RxNorm normalizes clinical drug names; it has no codes for assessment questions.
3. LOINC codes.
   LOINC codes observations and assessment questions, including nursing assessment items, so receivers know exactly what was asked.
4. ICD-10-CM codes.
   ICD-10-CM is a diagnosis classification used for billing and statistics, not a code set for assessment questions.