Medication administration record (MAR): what it contains and how to read one
For nurses, legal nurse consultants, attorneys and claims reviewers who have to read a MAR after the fact. You get the parts of the form, the federal timing and error rules, a worked example with 6 entries that need a question, a blank MAR template, a review log and a records request.
A medication administration record (MAR) is the part of a patient's chart where nurses record every dose as it is given, held, refused or missed: the drug, dose, route, scheduled time, actual time and the initials of the person who gave it. Hospitals and nursing homes keep it under federal rules, and in a claim or lawsuit it is the main evidence of what the patient actually received.
Orders say what a clinician wanted the patient to get. The pharmacy profile says what was dispensed. The medication administration record is the only document that claims to say what reached the patient, dose by dose, with a time and a set of initials. When a case turns on a missed antibiotic, an insulin dose or an opioid, the MAR is where the argument starts.
The MAR in 8 numbers
What a medication administration record is
A medication administration record, almost always shortened to MAR, is a grid. Each row is 1 active medication order. Each column is a date, and inside each day sit the scheduled times. When a nurse gives a dose, the nurse records it in the cell for that time: an initial and the actual time on paper, or a bar-code scan and an electronic signature on an eMAR. When a dose is not given, the cell should carry a code that says why.
The MAR sits between 2 other documents. Upstream is the order, written by a physician, nurse practitioner or other prescriber and verified by a pharmacist. Downstream is the patient's response, charted in vital signs, pain scores, glucose logs and nursing notes. The MAR is the hinge. It is where the plan becomes an act, and it is usually the first place anyone looks when a patient was harmed by a drug or by the absence of one.
Every setting that gives medications keeps some version of it. Hospitals, nursing homes, assisted living, home health, hospice, psychiatric units, jails and schools all use MARs, and many state forms for group homes and assisted living are simply blank MAR grids. The layout changes from place to place. The questions a reviewer asks do not.
MAR, eMAR and BCMA
- MAR
- Medication administration record. Paper or electronic, 1 row per order, 1 cell per scheduled dose.
- eMAR
- The electronic MAR inside the EHR. Orders flow in from the prescriber and pharmacy, so nobody transcribes them by hand.
- BCMA
- Bar-code medication administration. The nurse scans the patient's wristband and the drug's bar code; the eMAR checks the match before the dose is charted.
- Scheduled dose
- A dose due at a set time (09:00, 21:00). The cell exists before the dose is given.
- PRN dose
- "As needed." Given for a stated reason, such as pain 4 to 6, so the entry needs the reason and the result.
- Held, refused, omitted
- 3 different things. Held is a clinical decision, refused is the patient's decision, omitted means the dose did not happen and nobody recorded a reason.
- ADC
- Automated dispensing cabinet. The locked cabinet on the unit that logs every removal by user and time. Its report is a separate record from the MAR.
- Waste
- The part of a controlled-substance dose that is discarded, usually recorded with a second person as witness.
The parts of a MAR, field by field
A printed MAR looks crowded because it carries 7 kinds of information on 1 page. The mock sheet below marks each one. The layout is invented; every real facility's form differs, but the same 7 parts are on all of them.
- 1Header and allergiesThe allergy line is checked against every order. A drug in the same class as a listed allergy is a question, even when the dose was given without harm.
- 2The order as transcribedDrug, strength, dose, route, frequency and any hold parameter. Compare it word for word with the signed order. A hold parameter on the MAR that the order never had, or the reverse, is a transcription discrepancy.
- 3Start, stop and verificationThe dates the order was active. Doses charted before the start or after the stop are unauthorized; doses due inside the window with nothing charted are omissions.
- 4Scheduled timesSet by facility policy, not by the prescriber: "every 12 hours" becomes 09:00 and 21:00. The timing window in chapter 4 is measured from these times.
- 5The administration cellInitials, actual time and, where the order has a parameter, the value that was checked (heart rate, blood pressure, glucose). A code replaces the initials when the dose was not given.
- 6PRN entriesA PRN dose needs a reason that fits the order and a follow-up result. Pain 7 on an order for pain 4 to 6 and a blank effect field are 2 separate questions.
- 7The signature legendEvery set of initials on the sheet should resolve to a named person with a credential. Without the legend you cannot tell who gave what.
Codes for doses that were not given
Every facility writes its own legend, usually printed at the foot of the paper MAR or in the eMAR's help text. The table shows the categories you will meet; the exact letters and numbers change from one form to the next, so always get the legend for the form in front of you.
| Category | What it should mean | What to look for next |
|---|---|---|
| Held | A nurse decided not to give the dose, usually under a hold parameter or a prescriber's instruction | The value that triggered the hold, and whether policy required telling the prescriber |
| Refused | The patient declined | A note with the reason, education given and notification if the drug is one the policy lists |
| NPO | Nothing by mouth for a test, procedure or swallow concern | The NPO order and whether a non-oral route was ordered instead |
| Off unit | The patient was away (radiology, dialysis, surgery) | The other department's times and whether the dose was given on return |
| Not available | Pharmacy had not supplied the drug | The dispensing record and how long the delay lasted |
| Blank | Nothing. A blank cell is not a code | Any other record that covers the time: notes, scan logs, cabinet reports |
Abbreviations add a second layer. The Joint Commission's Do Not Use list bans U for units, IU, Q.D., Q.O.D., trailing zeros ("1.0 mg") and missing leading zeros (".5 mg"), and MS, MSO4 and MgSO4 for morphine and magnesium sulfate. You will still see them on older paper MARs. Our guide to medical abbreviations for legal professionals covers how to read them without guessing.
The federal rules behind the MAR
No federal regulation prescribes a MAR form. The rules set what the record must contain and how drugs must be given, and CMS surveyors read the MAR to check both. For a hospital that takes Medicare, the starting point is the medical records Condition of Participation:
"All practitioners' orders, nursing notes, reports of treatment, medication records, radiology, and laboratory reports, and vital signs and other information necessary to monitor the patient's condition."
The same section requires every entry to be "legible, complete, dated, timed, and authenticated" by the person responsible (482.24(c)(1)). On a MAR that means each dose has a date, a real time and a signature or initials that can be tied to a person. CMS surveyor guidance adds the point that matters most in litigation: "Documentation is expected to occur after actual administration of the medication to the patient." A dose charted at 08:00 for a patient who was in radiology until 08:50 is a documentation problem, whatever happened to the drug.
| Rule | What it requires | How it shows up on the MAR |
|---|---|---|
| 42 CFR 482.23(c)(1) | Drugs prepared and given under federal and state law, the practitioner's orders and accepted standards of practice | Every entry is measured against the order and the facility's policy |
| 42 CFR 482.23(c)(5) | A procedure for reporting transfusion reactions, adverse drug reactions and errors in administering drugs | An error on the MAR should have a matching report and, often, a note that the prescriber was told |
| 42 CFR 482.24(c)(1) and (c)(4)(vi) | Medication records in every chart; entries legible, complete, dated, timed and authenticated | Blank cells, missing times and unresolvable initials are all documentation gaps |
| 42 CFR 482.25(b)(5) | Automatic stop orders for drugs not specifically limited by the prescriber | Doses charted after an automatic stop date are unauthorized unless the order was renewed |
| 42 CFR 483.45(c) | Nursing homes: a licensed pharmacist reviews each resident's drug regimen at least once a month and reports irregularities | The monthly review report should mention MAR problems a reviewer finds later |
| 42 CFR 483.45(d) to (f) | No unnecessary drugs; PRN psychotropics limited to 14 days; error rate under 5%; no significant medication errors | PRN psychotropic entries past day 14 need a documented extension; see chapter 5 for errors |
The 5 rights and the 5 stages
CMS expects hospital policy to require 5 checks before every dose, the "5 rights": right patient, right medication, right dose, right route, right time. The same guidance reminds surveyors that the 5 rights cover only administration. Errors start earlier, and the MAR sits in the middle of a 5-stage process:
A clean MAR can hide a prescribing error, and a messy MAR can sit on top of care that was fine. That is why a reviewer reads the order, the MAR and the monitoring together.
Timing windows, PRN doses and held doses
There is no single federal deadline for every dose. The CMS surveyor guidance for 482.23(c) sets 3 windows. Each hospital decides which drugs fall in which window, so the policy is the first document to request.
A dose 90 minutes late is inside the window for a daily vitamin and outside it for a time-critical antibiotic. The drug's category decides whether a late entry is a deviation.
CMS lists examples of drugs a hospital may treat as time-critical: antibiotics, anticoagulants, insulin, anticonvulsants, immunosuppressive agents, non-IV pain medication, drugs ordered within a set time of the order, drugs that must be spaced from other drugs, and anything ordered more often than every 4 hours. The list is illustrative. A drug can be time-critical for 1 patient and not for another, and the policy must say how nurses know which is which.
Missed and late doses
The same guidance requires a policy on doses that fall outside their window: a patient off the unit for a test, a refusal, a patient who cannot swallow, a drug that was not available. The policy sets when a nurse may reschedule on their own judgment and when the prescriber must be told first. A late dose with no reason code and no note is the entry that draws the most questions, because the policy almost always required one of the two.
PRN entries
An as-needed dose has 3 parts, and the MAR often records only 1. The reason (pain 5, agitation, nausea) has to match the order's indication and range. The time has to respect the minimum interval ("every 4 hours"). The effect, a reassessment after the dose, is usually required by policy within a set time and is the field most often left blank. In nursing homes, PRN psychotropic orders run out after 14 days unless the prescriber documents a rationale and a duration, and PRN antipsychotics cannot be renewed without a new evaluation (42 CFR 483.45(e)(4) and (5)).
Held doses and hold parameters
Orders for blood pressure drugs, heart rate drugs, insulin and anticoagulants often carry a parameter: hold for heart rate under 60, hold for glucose under 70. A correct entry shows the value that was checked, and the vitals flowsheet should show the same value at about the same time. A dose charted as given when the flowsheet shows a value below the hold line is one of the most common findings in a MAR review.
Medication errors and how they show up on a MAR
The clearest federal definition of a medication error comes from the nursing home survey guidance for tags F759 and F760 in Appendix PP. An error is a preparation or administration of a drug that departs from 1 of 3 things: the prescriber's order; the manufacturer's specifications (not recommendations); or accepted professional standards. A significant error is one that causes the resident discomfort or puts their health and safety at risk.
Surveyors mostly find errors by watching a medication pass, then comparing what they saw with the MAR and the orders. The guidance also lets them find errors in the record itself, including "discrepancies in the MAR that lead to identification of a medication error", with interviews and other documents to back it up.
The error rate calculation
The nursing home rule at 42 CFR 483.45(f) requires an error rate under 5% and no significant errors at all. The surveyor's formula is simple:
| Step | Example |
|---|---|
| Doses observed being given | 40 |
| Doses ordered but not given during the observation | 2 |
| Opportunities for error (40 + 2) | 42 |
| Errors observed, significant and not | 2 |
| Error rate: 2 divided by 42, times 100 | 4.8%, under the 5% line |
2 rules decide close cases. A rate below 5% cannot be rounded up, so 4.6% stays under the line. And any significant error is cited under F760 whatever the overall rate, so 1 missed dose of a heart drug can be a citation on its own.
What makes an error significant
The guidance gives 3 tests: the resident's condition, the drug category, and how often the error happened. Drugs with a narrow therapeutic index, where the effective dose sits close to the toxic dose, raise the stakes; the guidance names phenytoin, carbamazepine, warfarin, digoxin, theophylline and lithium. Its examples show the range. An omitted daily dose of metoprolol succinate is listed as significant; a missed multivitamin is not. Losartan due at 08:00 and given at 09:30 is listed as not significant; 2 oxycodone tablets due 20 minutes before a painful treatment and given after it are significant.
| Error type | What it looks like on the MAR | What proves it |
|---|---|---|
| Omission | A due cell left blank, or a code with no reason | The active order, the cell, and no other record of the dose |
| Unauthorized drug | A dose charted with no active order, or after a stop date | The order history and the pharmacy profile |
| Wrong dose | Charted amount differs from the order or the sliding scale | The order, the scale, the glucose or lab value at the time |
| Wrong route or form | Route or product on the MAR differs from the order | The order and the dispensing record |
| Wrong time | Actual time outside the policy window | The policy's drug categories and the charted time |
| Parameter breach | Dose given when a hold value was reached | The vitals or glucose value charted near the dose |
Worked example: 1 day of a hospital MAR
| Due | Order | MAR says | Other source | Page |
|---|---|---|---|---|
| 09:00 | Enoxaparin 40 mg subcut daily | Given 08:55, RN-B | None needed | 88 |
| 09:00 | Metoprolol 25 mg PO, hold HR under 60 | Given 09:10, RN-B | HR 56 at 09:02 | 88, 60 |
| 09:00 | Ceftriaxone 1 g IV daily (time-critical) | Given 10:14, RN-B | Note: IV infiltrated, restarted 10:05 | 88, 111 |
| 11:30 | Insulin lispro per scale, before meals | 4 units, 11:34 | Glucose 286 at 11:20; scale says 6 units for 251 to 300 | 89, 72 |
| PRN | Oxycodone 5 mg PO every 4 hrs, pain 4 to 6 | 13:20, pain 7, effect blank | Next pain score 17:40 | 90, 61 |
| 16:30 | Insulin lispro per scale, before meals | (blank) | Note: NPO from 15:00 for swallow study | 89, 112 |
| 18:05 | Oxycodone (cabinet removal) | No entry | Removed 18:05 by RN-B; no waste recorded | 140 |
| 21:00 | Metoprolol 25 mg PO | R (refused), LPN-C | Note: dizzy, prescriber told 21:15 | 88, 113 |
8 entries, 6 questions. 1 has an explanation in the notes, 2 need a policy answer, and 3 need a witness.
What each flag means
- Metoprolol given at a heart rate of 56. The order says hold under 60. The MAR carries no heart rate, and the flowsheet shows 56 at 09:02. A recheck that was never charted is possible. On paper this is a parameter breach.
- Ceftriaxone 74 minutes late. Outside a 30-minute window if the policy lists IV antibiotics as time-critical, which this one does. The note explains the delay: the IV failed and was restarted. The remaining question is whether the policy required the prescriber to be told.
- 4 units instead of 6. The scale on the order and the glucose value do not match the dose charted. The next glucose value tells you whether it mattered.
- PRN opioid with no reassessment. The reason was outside the order's range (pain 7 against an order for 4 to 6) and no reassessment was charted within 60 minutes.
- Blank 16:30 insulin cell. The patient was NPO, so holding the dose may have been right. The MAR should say so with a code; a blank is not a code.
- Cabinet removal with no MAR entry. Oxycodone left the cabinet at 18:05 under RN-B's login, and nothing on the MAR or in the waste record accounts for it. This is the entry most likely to draw a follow-up.
The same day on 1 timeline
- 09:02Heart rate 56
Below the hold parameter on the metoprolol order.
Vitals p. 60 - 09:10Metoprolol charted as given
No heart rate on the MAR entry.
MAR p. 88 - 10:05IV restarted after infiltration
Explains the antibiotic delay.
Note p. 111 - 10:14Ceftriaxone given
74 minutes after the scheduled time.
MAR p. 88 - 11:20 to 11:34Glucose 286, then 4 units
The scale calls for 6.
Glucose log p. 72; MAR p. 89 - 13:12 to 13:20Oxycodone removed, then given
Removal and dose match.
Cabinet report p. 140; MAR p. 90 - 13:20 to 17:40No pain reassessment
Policy called for one by 14:20.
Flowsheet p. 61 - 15:00NPO for swallow study
Explains a held meal-time insulin dose, if a code had been charted.
Note p. 112 - 18:05Second oxycodone removal
No MAR entry, no waste record.
Cabinet report p. 140 - 21:00Metoprolol refused
Reason charted, prescriber told 21:15.
MAR p. 88; note p. 113
Read alone, the MAR has 1 blank. Merged with 4 other sources, it has 6 entries that need a question and 1 that needs a witness.
What a MAR can show
When the entries are complete
- That a dose was charted as given, by whom, and at what time.
- That a dose was charted as held or refused, with a reason code.
- That a due dose has no entry at all.
- Dose, route and product as charted, to compare with the order.
What it cannot show alone
Where you need another record
- That the charted time is when the dose was given (scan logs and cabinet reports help).
- Whether the dose was clinically right (the order and the monitoring decide).
- Why a blank is blank (notes, testimony).
- Whether a controlled substance removed from the cabinet reached the patient.
How to review a MAR, step by step
The method below works on paper and electronic MARs, in hospitals and in nursing homes. It is slow by hand on a long stay, which is why the order of steps matters: build the list of doses that should exist before you read the doses that do.
- Collect the full set. Every MAR page for the date range, the signed orders, the pharmacy profile, the vitals flowsheet, the glucose log, the nursing notes, and for controlled substances the cabinet removal and waste reports.
- Build the dose skeleton from the orders. For each order, list every dose due between its start and stop, using the facility's standard times.
- Match every MAR entry to a due dose. Mark each as given, held with a code, refused, or blank. Mark any entry with no matching order.
- Check timing against the policy. Only after you know which drugs the policy lists as time-critical.
- Check parameters and scales. Heart rate and blood pressure against hold lines, glucose against the sliding scale.
- Check PRN entries. Reason inside the ordered range, interval respected, effect charted.
- Reconcile controlled substances. Every cabinet removal should match a MAR entry, a waste record or a return.
- Log each question with a page. 1 line per due dose, not per blank, so the log can be checked against the file.
A blank or late entry: the decision path
1. MAR review log
1 line per due dose. Keep the page cite in every line so anyone can check it.
MAR REVIEW LOG
Patient: [ID] Facility: [NAME] Dates reviewed: [FROM] to [TO]
Policy used for timing: [POLICY NO., REVISION DATE]
Date | Due | Order (drug, dose, route, frequency) | Order page
| MAR entry (given / code / blank) | Actual time | Initials
| Parameter or scale value at the time | Source page
| Other record covering the dose | Source page
| Question | Category (omission, timing, dose, parameter,
PRN, controlled substance, unauthorized)
Example:
03/03 | 09:00 | Metoprolol 25 mg PO q12h, hold HR under 60 | p. 44
| Given | 09:10 | RN-B
| HR 56 at 09:02 | p. 60
| None | n/a
| Given below hold parameter; recheck not charted | parameter2. Blank MAR template
A plain grid for 1 week. Adapt the times and codes to your facility's policy and your state's form rules.
MEDICATION ADMINISTRATION RECORD
Name: ____________ Date of birth: ________ Record no.: ________
Allergies: ______________________ Weight: ______ Month: ______
Medication, dose, route, frequency: ___________________________
Hold parameters: ____________ Start: ______ Stop: ______
Prescriber: ____________ Verified by pharmacy: ______
Time | Day 1 | Day 2 | Day 3 | Day 4 | Day 5 | Day 6 | Day 7
_____|_______|_______|_______|_______|_______|_______|______
_____|_______|_______|_______|_______|_______|_______|______
_____|_______|_______|_______|_______|_______|_______|______
PRN DOSES
Date | Time | Medication and dose | Reason (score) | Initials
| Effect | Time rechecked | Initials
CODES (edit to match policy)
H = held (write the value) R = refused N = NPO
O = off unit (write where) NA = not available from pharmacy
SIGNATURES
Initials | Full name | Credential | Signature3. Request list for medication records
A plain list of what to ask for. Adapt it to your court's rules or your facility's release process. Not legal advice.
MEDICATION RECORDS FOR [PATIENT], [DATES] 1. Every medication administration record page, paper or electronic, including PRN and one-time dose sections. 2. The code legend and signature legend for each MAR form. 3. All medication orders, with order history showing entry, verification, modification and discontinuation times. 4. The pharmacy medication profile. 5. Bar-code scan records for each dose, including overrides and the reason entered for each override. 6. Automated dispensing cabinet removal, return and waste reports for the patient. 7. Vital signs, glucose and pain score records. 8. Policies in force on the dates of care on medication administration timing, time-critical medications, missed and late doses, PRN reassessment and controlled-substance waste. 9. Any medication error or adverse drug event report for the patient, to the extent not privileged under [STATE STATUTE].
Before you rely on a MAR printout
0 of 6 checked.
When a MAR looks edited after the fact, the method changes. Our guides on altered medical records and charting by exception cover late entries and blank cells in more depth.
eMAR, bar codes and AI review
Most hospital MARs are now electronic, and many are tied to bar-code scanning. The best-known evidence that this reduces errors comes from a study at an academic medical center that compared units before and after bar-code eMAR. Poon and colleagues observed thousands of administrations directly and found a clear drop in errors other than timing errors.
776 errors fell to 495, a 41.4% relative reduction. Fewer errors is still not zero, and scan overrides leave their own trail.
For a reviewer, the eMAR changes what exists. Paper MARs give you initials and a handwritten time. An eMAR has a second layer behind the printout: scan times, override reasons, the order version on screen, and who charted when. The printout rarely shows it, so ask for it by name, as the request list in chapter 7 does.
Where AI helps with a MAR
A long stay produces hundreds of MAR pages, and the review in chapter 7 is mostly matching: this dose against that order, this time against that vital sign. That is the part AI medical record review does well. A grounded system can pull every dose into a table, place each one on a timeline with the vitals, glucose values and notes, and give every line a page-level citation back to the source, which is what turns a MAR into an AI medical chronology you can check.
It has limits you should plan for. Scanned paper MARs need OCR, and initials in tiny grid cells are hard for any handwriting recognition model. A large language model (LLM) asked to summarize a MAR without citations may produce a fluent answer with a dose that is not on the page; hallucination is the reason every line has to point to its source. And no tool knows the facility's timing policy or legend unless you have them. The human in the loop decides what a flag means.
Medrecords AI works this way on the files you upload. It builds a cited medical chronology with every entry linked to its page, routes each page through OCR matched to its content, including handwritten entries, answers questions like "every metoprolol dose with the heart rate nearest to it" through cited Q&A, and flags records that should exist but were not produced. Flags are signals, not verdicts. It does not read eMAR scan logs or cabinet exports as data, it does not decide whether an entry is an error, and it does not retrieve records from providers. It runs under SOC 2 and HIPAA with a signed business associate agreement (see security and HIPAA). Self-Service bills 10 cents a deduplicated page, down to 5 cents at volume, duplicates free; Enterprise On-Prem is an annual license.
If you are comparing tools, our guide to medical record review covers what to ask any medical chronology software vendor. See how MAR review fits medication error cases, nursing home neglect and legal nurse consultant work.
See every dose on 1 timeline, cited to its page.
Book a demo on a file with a medication question, then run your first case free on us. Every line comes back cited to its source page. You review, you revise, you sign.
Scheduling only. No records move from a public page.
Frequently asked questions
- What is a medication administration record?
- The part of the chart where nurses record each dose as given, held, refused or missed, with the drug, dose, route, scheduled time, actual time and initials. Hospitals and nursing homes keep it as part of the medical record.
- What is the difference between a MAR and an eMAR?
- An eMAR is the electronic version inside the EHR. Orders flow into it from the prescriber and pharmacy, and it is often paired with bar-code scanning. It also stores scan times and overrides that a paper MAR cannot.
- What should be on a MAR?
- Patient identifiers and allergies; each order with drug, dose, route, frequency and any hold parameter; start and stop dates; scheduled times; an entry for every dose with time and initials or a code; PRN reasons and effects; and a signature legend.
- What are the 5 rights of medication administration?
- Right patient, right medication, right dose, right route and right time. CMS expects hospital policy to require all 5 to be confirmed before each dose. Some nursing texts list more, up to 9.
- How late can a scheduled medication be given?
- Under CMS hospital guidance, within 30 minutes either side for drugs the hospital lists as time-critical, 1 hour either side for drugs given more than daily up to every 4 hours, and 2 hours either side for daily, weekly or monthly doses. Facility policy decides the categories.
- What does a blank cell on a MAR mean?
- That nothing was charted. It may be an omitted dose, a dose given but not charted, or a held dose missing its code. Other records (notes, scan logs, cabinet reports) usually settle which.
- How long do hospitals keep medication records?
- At least 5 years under 42 CFR 482.24(b)(1), as part of the medical record. Many states require longer.
- Can a patient get a copy of their MAR?
- Yes. The MAR is part of the medical record, so it falls in the designated record set covered by the HIPAA right of access. The provider has 30 days to act, with 1 extension of up to 30 more days (45 CFR 164.524).
- Can AI read a scanned or handwritten MAR?
- It can extract most typed entries with OCR and many handwritten ones, but initials and times in small grid cells are where errors cluster. Use a tool that cites every value to its page and check the flagged cells by hand.
- Can software find medication errors in a MAR?
- Software can flag mismatches: a dose charted against a hold value, a dose outside the order's range, a due dose with no entry. Whether a mismatch is an error depends on the order, the policy and the clinical facts, which a person decides.
- Is it HIPAA compliant to upload a MAR to an AI tool?
- It can be, with a vendor that signs a business associate agreement, holds a SOC 2 report and commits in writing not to train on your data. A consumer chatbot without a BAA is a different matter.
Sources and method
Regulations were read in full text from primary sources in September 2026, and quoted text is verbatim. CMS surveyor guidance was read from the published transmittal and appendix. Study figures come from the published paper. The MAR sheet, patient, policy, times and pages in chapters 2, 6 and 7 are hypothetical. Product facts come from this site's product pages. Nothing here is legal or medical advice.
- 42 CFR 482.24(b)(1), (c)(1), (c)(4)(vi) and (c)(4)(viii): retention, entry standards, medication records, completion within 30 days.
- 42 CFR 482.23(c)(1) and (c)(5): drug administration and reporting of administration errors.
- 42 CFR 482.25(b)(5): automatic stop orders.
- 42 CFR 483.45(c), (d), (e) and (f): nursing home drug regimen review, unnecessary drugs, PRN psychotropic limits, error rate.
- CMS State Operations Manual Appendix A, Transmittal 151 (November 20, 2015): interpretive guidance for 482.23(c): the 5 rights, the 5 stages, timing windows, time-critical examples, missed doses, documentation after administration.
- CMS State Operations Manual Appendix PP, F759 and F760 (Rev. 173, 2017): medication error definitions, error rate formula, significance tests and examples.
- 45 CFR 164.524: HIPAA right of access, 30 days plus 1 extension.
- The Joint Commission, Do Not Use list of abbreviations.
- Poon et al., "Effect of bar-code technology on the safety of medication administration", New England Journal of Medicine 362(18), 2010: non-timing errors 11.5% without and 6.8% with bar-code eMAR, a 41.4% relative reduction.