Sign Up
Login
Try for Free
Login
The Geriatric Depression Scale (GDS) is a public-domain, yes-or-no self-report depression screen for older adults, with a 30-item original (Yesavage and colleagues, 1982) and the 15-item GDS-15 short form (Sheikh and Yesavage, 1986). Primary care, geriatric, and behavioral health clinicians use it to decide who needs a full mood assessment; it screens, never diagnoses. This page covers how to document and interpret GDS-15 results, with a fictional sample note.
Patient self-report on paper, or read aloud by any clinician or trained staff member with the patient's own answers recorded; no license, training requirement, or publisher qualification level applies; the treating clinician interprets and documents
Primary care and geriatric clinicians, nurses and nurse practitioners, psychologists and behavioral health clinicians, assisted-living and home-health teams, quality-measure abstractors, payers and auditors reviewing wellness-visit and screening claims
3 to 6 chart lines (form and mode, cognitive status, score with the threshold applied, interpretation, separate risk note, follow-up plan) · patient completion about 5 to 7 minutes
Brief self-report screen (15 yes-or-no items, one point per answer in the depressive direction, total 0 to 15, past-week frame; the 30-item original and the 5- and 4-item derivatives are separate forms with their own ranges)
Medicare annual wellness visits and annual depression screening, geriatric and primary care intakes, assisted-living and community reviews, comprehensive geriatric assessment in hospital and oncology, and repeat measurement during treatment; never as a suicide screen
Yesavage and colleagues 1982 (GDS-30) and Sheikh and Yesavage 1986 (GDS-15); the original scale is public domain per the Stanford author site; an example tool in CMS Quality ID 134 (2026) and in CCSMH guidance (2021), mandated by no law; described here for documentation, no items reproduced
The Geriatric Depression Scale is a self-report screen for depressive symptoms in older adults, developed by Jerome Yesavage, T. L. Brink, and colleagues at Stanford and the Palo Alto VA. The 30-item original appeared in the Journal of Psychiatric Research (volume 17, pages 37 to 49; PubMed dates it 1982 and the Stanford site 1983, so both years circulate), after an early report by Brink and colleagues in Clinical Gerontologist in 1982. The 15-item short form, the GDS-15, comes from a 1986 chapter by Javaid Sheikh and Jerome Yesavage in Clinical Gerontology: A Guide to Assessment and Intervention (Haworth Press), and it is the form most practices mean when they write GDS. Every item is answered yes or no about the past week and scores one point when answered in the depressive direction; some items count when answered yes and others when answered no, so the total is not a count of yes answers. The developers started from a 100-item pool and kept the items that best separated depressed from non-depressed elders, which is why complaints about sleep, appetite, and energy do not appear on it and why it is less confounded by medical illness than the PHQ-9. Two shorter derivatives are common: the five-item GDS-5 of Hoyl and colleagues (Journal of the American Geriatrics Society, 1999) and the four-item set of D'Ath and colleagues (Family Practice, 1994), and a 2021 review counted five different four-item and seven different five-item versions in the literature, so a short form is identified by its author as well as its length. Stanford also hosts a 15-item informant version (Brown and Schinka, 2005) and more than 40 translations, with the note that the authors cannot vouch for the accuracy of the translations.
Three facts carry the documentation. First, the threshold belongs to the form: the author site says that a GDS-15 total higher than 5 should prompt a careful interview, its scoring page reads totals above 10 as almost always depression, and the validation studies mostly evaluate a cut point of 5 or more, so the note has to name the form, the denominator, and the threshold it applied. A bare "GDS 7" is positive on 15 items, sits inside the original normal range on 30, and is impossible on a four- or five-item form. Second, a positive screen is not a diagnosis, and the informal severity labels that circulate for the GDS-15 are conventions rather than diagnostic strata: an individual participant data meta-analysis in 2024 found that a GDS-15 threshold of 5 or more identified 34.2 percent of older adults as depressed against 14.8 percent by structured interview. Third, the scale has no suicide item, so a positive result does not discharge the duty to ask about suicide risk, and a low score says nothing about it; the PHQ-9 carries a self-harm item and its own follow-up rules, and the suicide risk assessment page covers the direct assessment. A fourth fact governs who can be screened at all: the GDS is self-report, so every entry states the patient's cognitive status and ability to answer, using the GDS through mild and moderate impairment where the patient can follow yes-or-no questions and moving to an observer-rated instrument such as the Cornell Scale for Depression in Dementia when self-report is no longer reliable.
In the United States the GDS-15 lives in the Medicare annual wellness visit and the annual depression screening benefit, where CMS asks for a standardized tool recognized by national professional medical organizations and names none, and in the 2026 Quality ID 134 specification, which lists the GDS and the Cornell Scale among acceptable adult tools. Family physicians, internists, geriatricians, and nurse practitioners chart it at wellness visits and intakes, assisted-living and home-visit teams use it for community residents, and hospital and oncology teams place it inside a comprehensive geriatric assessment because it avoids the somatic items that fatigue, anorexia, and treatment effects would inflate on a PHQ-9. Nursing homes and home health agencies are the exception: the MDS 3.0 resident mood interview and the OASIS mood section run on the PHQ family, so there the GDS is a clinical adjunct rather than the required record. In Canada the CCSMH guideline (2021) recommends the GDS or the PHQ-9 for older adults without significant cognitive impairment, and in Australia the 75-plus health assessment must cover psychological function including cognition and mood without naming a tool. The neighbors are fixed: a Mini-Cog, MoCA, MMSE, or SLUMS result is the cognitive statement every GDS entry references; the Cornell Scale takes over when self-report is unreliable; the suicide risk assessment or C-SSRS carries the risk question the GDS cannot ask; serial results follow the outcome measure note; and a positive screen that leads to treatment continues in the medication management note or a psychotherapy progress note.
No law, payer, or author prescribes a GDS note format. What exists is an author site with scoring guidance, a quality measure that requires the tool's name and a dated follow-up plan, and regulatory assessment systems in nursing homes and home health that use a different mood instrument. What survives review is an entry that names the form and its denominator, states how the scale was administered and whose answers it holds, records the patient's cognitive status, gives the total with the threshold applied, labels the result a screen, documents a separate suicide-risk question, and closes with a dated follow-up plan. Each element below carries the pitfall that most often undermines it.
Form, denominator, and time frame. Write GDS-15, GDS-30, or the short derivative with its author (Hoyl five-item, D'Ath four-item), the denominator (7 of 15), the language or translation, and the past-week frame as worded. The forms have different item sets, ranges, and threshold conventions, and short forms of the same length are not one instrument, so the form is the first thing a later reader, an abstractor, or a payer needs. Pitfall: "GDS 7" with no form: positive on 15 items, inside the original normal range on 30, impossible on 4 or 5.
Mode and answer source. Record whether the patient self-completed the form on paper or a tablet or a clinician read the items aloud and recorded the patient's own answers, whether an interpreter or a validated translation was used, and whether any workaround for vision, hearing, or aphasia was needed. A relative's answers are not the patient's GDS: if collateral report is the only option, use the informant version (GDSI-15) or an instrument built for informant input and label the result as informant-rated. Pitfall: A daughter's answers charted as the patient's GDS-15, or a translated form charted as plain GDS-15 with no language.
Cognitive status and capacity to answer. Name the cognitive screen and its result (Mini-Cog 4 of 5, MoCA 22 of 30), the absence or presence of delirium, any established diagnosis with its stage, and whether the patient followed and answered yes-or-no questions. The evidence supports the GDS-15 through mild and moderate impairment and not in severe impairment, and completion falls sharply below an MMSE of about 10; no authority sets a MoCA or MMSE score at which validity stops, so document the observed capacity rather than a cutoff. When self-report is unreliable, say so and switch to the Cornell Scale. Pitfall: A GDS-15 in advanced dementia with no cognitive statement, or "valid" declared from a diagnosis label alone.
Score, key, and completeness. Give the total over the denominator, confirm the key was applied to both answer directions, and say who computed the score. If items were missed, apply the author site's rule (prorate the completed items to the full denominator and round up) and chart the prorated figure as prorated; the same site suggests, as the author's own estimate, that more than five missing items invalidate the scale. Keep the item-level answers on the flowsheet rather than in the narrative. Pitfall: A count of yes answers charted as the score, or a prorated total with no note that items were missing.
Threshold applied and the screening conclusion. State the convention and its source, then apply it: higher than 5 for a careful interview (author site), 5 or more as the practice protocol or the cut point most validation studies used, above 10 as almost always depression (author site scoring page). Write the result as a screen: "positive screen at the stated threshold; depression not established." The four-band severity labels that circulate for the GDS-15 have no primary-source authority and do not map onto DSM-5 severity, so they never enter the diagnosis line. Pitfall: "GDS-15 9, moderate depression" charted as a diagnosis, or "positive" with no threshold named.
Suicide risk assessed separately. The GDS contains no suicide or self-harm item. Ask directly, record the question asked and the answer, name the tool if one was used (C-SSRS screener), and file the formulation and any safety plan in the risk section of the record. A low total documents nothing about ideation and a high total nothing about intent; the Hartford Institute guidance is that any positive short-form screen prompts a fuller assessment and an evaluation for suicidality. Pitfall: "GDS-15 3, no safety concerns" with no direct question asked.
Follow-up plan, claim, and record. Document the plan a positive screen triggers (referral for evaluation, treatment, or another intervention for depression) on the date of the encounter or within two calendar days, and put the tool's name in the record; Quality ID 134 requires the name, not the score, and does not accept a repeat screen or a suicide risk assessment as the plan. Record the claim (G0444 with a subsequent wellness visit; it is bundled into the IPPE and the initial visit), the setting's other required mood instrument if any, and the repeat plan. Pitfall: A positive screen with no plan, or a repeat screen or suicide assessment logged as the plan.
GDS DOCUMENTATION BLOCK Date: [ ] Setting: [ ] Clinician: [ ] Encounter: [wellness visit / intake / assisted living / hospital or oncology / monitoring] Form: [GDS-15 / GDS-30 / GDS-5 (author) / GDS-4 (author)] Language: [ ] Mode: [self-completed / read aloud, patient's own answers / interpreter / informant version, labeled as such] Time frame: [past week, as worded] Cognitive status: [screen and result; delirium absent or present; diagnosis and stage; followed yes-or-no questions: yes / no] Score: [ ] of [15] Key applied to both answer directions: [ ] Completeness: [all items answered / missing items; prorated, rounded up] Threshold applied: [convention and source] Result: [positive / negative] Prior: [date, same form and mode] [ ] of [15] Change: [ ] Interpretation: [a screen, not a diagnosis; congruence with interview; grief, medical, medication, substance contributors; function] Suicide risk (no GDS item): [direct question asked; tool; findings; formulation and safety plan filed at ...] Follow-up plan: [referral / treatment / other intervention; dated on the encounter date or within two calendar days] Record: [tool name in the record; code claimed (bare number); required setting instrument (MDS or OASIS mood interview) completed separately] Clinician signature / credentials: Date:
Free to use and share, no signup. The PDF includes a one-page cheat sheet with element-by-element pitfalls and a pre-sign checklist; the DOCX is the blank documentation block, ready to adapt. Neither reproduces the scale items.
Scenario: a US internal medicine practice's Medicare subsequent annual wellness visit, where a 79-year-old widow's paper GDS-15 comes back positive, the physician documents the cognitive screen, the threshold, a direct suicide-risk question, the interview, and a same-day follow-up plan, and records the claim. All details are fictional.
Patient: E.R., 79 · Setting: Internal medicine, Medicare subsequent annual wellness visit · Clinician: M. Okafor, MD · Note date: 09/15/2026
Screen: GDS-15 (15-item short form, English), self-completed on paper in the waiting room on 09/15/2026, past-week frame as worded, all 15 items answered, scored by the medical assistant with the key applied to both answer directions and confirmed by me. Total 7 of 15. Positive at the practice threshold of 5 or more, and above the author site's interview trigger of a total higher than 5. Item-level answers are on the screening flowsheet. Prior: 2 of 15 on the same form at last year's wellness visit.
Cognitive status: Mini-Cog completed today as the wellness visit's cognitive assessment: 4 of 5 (recall 2 of 3, clock normal), negative under the practice protocol; no delirium, alert and oriented, followed the yes-or-no format without difficulty. No cognitive diagnosis. Self-report judged reliable, so the GDS-15 result is interpreted at face value.
Interpretation: Positive depression screen; the GDS-15 is a screen and does not establish a diagnosis. In the interview E.R. describes low mood most days, tearfulness, and loss of interest in her garden club since her husband's death five months ago, with sleep and appetite preserved and no psychomotor change. No prior depressive episode, no alcohol use, medication list unchanged for two years. Working impression: depressive symptoms in the context of bereavement, major depressive disorder not established today; DSM-5 criteria and duration will be assessed at the follow-up visit. Result is congruent with the interview and with the change from last year's score.
Risk: The GDS-15 contains no suicide item. Asked directly: denies thoughts of death, of being better off dead, or of harming herself, now or since her husband died; C-SSRS screener negative; no prior attempts; daughter lives nearby and visits daily. Formulation and the standing offer of the practice's crisis contact recorded in the risk section of the record.
Follow-up plan and claim: Plan discussed with E.R. and documented today: referral placed to the practice's collaborative care behavioral health clinician for a diagnostic assessment and grief-focused counseling, information given on the hospital's bereavement group, TSH and B12 added to today's labs, nurse phone check on 09/22/2026, office visit in three weeks to assess criteria and discuss treatment options if symptoms persist. Screening reviewed and addressed on the encounter date; screening time 8 minutes. Annual depression screening claimed under G0444 with the subsequent wellness visit under G0439, supported by the named tool, result, and plan above; Quality ID 134 met (tool named, positive result, follow-up plan documented on the encounter date). E.R. was told the screen is not a diagnosis and agreed with the plan.
This sample is fictional and for educational purposes. It does not describe a real patient or record; the scores, dates, and details are invented to show documentation structure and are not clinical guidance.
Writing these after every session? BastionGPT drafts complete notes from bullets, dictation, or a transcript.
Generate a note from bulletsUnited States: the instrument is a convention; the rules around it are real. No federal or state LAW names the GDS or prescribes a screening note format. The Medicare annual wellness visit (42 CFR 410.15; PAYER POLICY) makes review of potential risk factors for depression a required element of the initial visit and tells clinicians to select from standardized screening tools recognized by national professional medical organizations, naming none; detection of cognitive impairment is a separate required element, which is why a GDS entry and a cognitive screen belong in the same note. Medicare's annual depression screening benefit (NCD 210.9, effective October 14, 2011; PAYER POLICY) covers one screen per year, up to 15 minutes, in a primary care setting with staff-assisted depression care supports, and states that CMS does not identify specific screening tools; it is billed under G0444, which is bundled into the IPPE and the initial wellness visit and separately payable with subsequent visits. CMS Quality ID 134 (2026 MIPS specification; PAYER POLICY with quality-measure teeth) requires an age-appropriate standardized tool, lists the Geriatric Depression Scale and the Cornell Scale for Depression in Dementia among the adult examples alongside the PHQ-9, BDI, CES-D, and others, requires the name of the tool in the medical record but not the score, and requires that a positive screen carry a follow-up plan (referral for evaluation, pharmacological intervention, or another intervention for depression) discussed during the encounter and documented on the date of the encounter or within two calendar days; a further screening or a suicide risk assessment does not count as the plan, positive results suggesting high self-harm risk call for more urgent action, and cognitive, functional, or motivational limitations that may affect accuracy are an allowed medical reason for not screening. Under LAW (the federal resident and patient assessment systems), the MDS 3.0 Resident Assessment Instrument (manual v1.20.1, effective October 1, 2025) runs Section D on the PHQ-2 to 9 resident interview and the PHQ-9 staff-observation version, validated for MDS 3.0 by Saliba and colleagues in 2012 (weighted kappa 0.69 against a gold-standard interview; 86 percent of residents could complete it), and the home health OASIS-E2 data set (effective April 1, 2026) carries its own mood section, so in nursing homes and home health a GDS is a clinical adjunct that never replaces the required mood items. HEDIS depression screening measures likewise accept only the instruments in their published value sets, so a practice reporting them should check the current specification before treating a GDS result as measure credit. The result is a HIPAA record; nothing about it invokes 42 CFR Part 2.
Canada, Australia, and the evidence on cognition. In Canada no federal or provincial LAW or payer names the GDS. The Canadian Coalition for Seniors' Mental Health guideline (2021 update; CONVENTION, professional-society guidance) recommends the GDS or the PHQ-9 for older persons without significant cognitive impairment in general medical or geriatric settings, and recommends an observer-rated instrument such as the Cornell Scale for Depression in Dementia instead of the GDS for patients with moderate to severe cognitive impairment, both graded B and unchanged since 2006. British Columbia's cognitive-impairment guideline appendix lists the short GDS among tools for minor cognitive loss or early-stage dementia and the Cornell Scale for advanced impairment, and tells clinicians to choose by the person's remaining cognitive abilities and to follow any positive screen with assessment. Provincial long-term care runs on the interRAI RAI-MDS 2.0 in most provinces, whose Depression Rating Scale is an observer-rated score on a 0 to 14 range, not a GDS, so a "depression score" in a Canadian facility record needs its instrument named. In Australia, the older person's health assessment (MBS items 701, 703, 705, and 707; PAYER POLICY) is claimable once every 12 months for people 75 and over who are not inpatients or residential care residents and must include an assessment of psychological function including cognition and mood, with no instrument named, so a GDS-15 fits and is not required. In residential aged care the Aged Care Funding Instrument carried a modified Cornell Scale until the AN-ACC funding model replaced it in October 2022 with no depression measure, according to the Australian Institute of Health and Welfare, and Victoria's standardised care process for depression (October 2023; CONVENTION) recommends the Cornell Scale on admission and when symptoms emerge, paired with the Psychogeriatric Assessment Scales cognitive screen, and states that it is not diagnostic. The evidence behind the cognitive-status rule is now specific. In 24,674 US participants stratified by Clinical Dementia Rating, Bennett and Wang (2024) found negligible bias in GDS-15 scores at questionable, mild, and moderate impairment and significant bias at severe impairment, and concluded the scale can be used with mild or moderate impairment but not severe; a 2026 analysis of 14,077 people in the same national dataset found differential item functioning in 13 of 15 items, with cognitive status the strongest influence, which is the reason to record cognition rather than declare the scale valid; among 834 people aged 85 and over, Conradsson and colleagues (2013) found the GDS-15 completed by 1.5 percent of those with an MMSE of 0 to 4, 41.7 percent at 5 to 9, 65.5 percent at 10 to 14, and more than 90 percent at 20 and above; and in nursing homes the self-rated and the informant GDS both failed to detect depression adequately in residents with dementia (areas under the curve 0.57 to 0.70; Li and colleagues, 2015). Meta-analyses of medical settings found accuracy not significantly influenced by the presence of dementia (Mitchell and colleagues, 2010), so the boundary is functional, whether this patient can understand and reliably answer, and the chart should say what was observed. A translated GDS is charted with its language and version, because the author site hosts more than 40 translations it cannot vouch for and the cut point can shift with the version.
The Geriatric Depression Scale is in the public domain; the authors are Jerome A. Yesavage, T. L. Brink, and colleagues for the 30-item scale (Journal of Psychiatric Research, 1982 to 1983) and Javaid I. Sheikh and Jerome A. Yesavage for the 15-item short form (Clinical Gerontology: A Guide to Assessment and Intervention, Haworth Press, 1986), and the Stanford author site states that the original scale is in the public domain because it was partly the result of federal support, hosts the forms, the informant version, the translations, and free scoring apps, and asks users to look for its versions because other versions circulate. That statement covers the original scale; the five-item form of Hoyl and colleagues (Journal of the American Geriatrics Society, 1999) and the four-item set of D'Ath and colleagues (Family Practice, 1994) were published in journals, and at least one instrument-licensing database lists rights conditions for the short forms, so a commercial product, app, or EHR vendor should confirm the rights of the exact version it embeds rather than assume every derivative shares the original's status. BastionGPT is not affiliated with, or endorsed by, any of these publishers. This page reproduces no test items, stimuli, norms, or scoring materials.
The numbers behind these errors are specific. An individual participant data meta-analysis of 14 studies and 3,602 older adults found that a GDS-15 threshold of 5 or more classified 34.2 percent as depressed against 14.8 percent by structured interview, with 8 or more the closest match (Parsons and colleagues, 2024); at a cut point of 5 the pooled sensitivity is 0.89 and specificity 0.77 across 32 studies, with the authors warning of selective reporting of cut points (Pocklington and colleagues, 2016), and 86 and 79 percent across 53 studies and 17,018 participants (Krishnamoorthy and colleagues, 2020). A 2021 review found five four-item and seven five-item versions in use with different accuracy at the same numerical threshold and rated the evidence low certainty. In 24,674 adults the GDS-15 showed negligible bias through moderate cognitive impairment and significant bias in severe impairment (Bennett and Wang, 2024), and in nursing homes neither the self-rated nor the informant version detected depression adequately in residents with dementia (Li and colleagues, 2015). No published audit counts how often the form, the mode, or the cognitive statement is missing from GDS chart entries; the Quality ID 134 and Medicare rules described under compliance considerations are what turn those omissions into a measure or claim problem. The BastionGPT Clinical Advisory Board sees the same errors most often in GDS-15 documentation reviews:
BastionGPT is specifically trained, tuned, and clinically tested on behavioral health progress notes and screening documentation.
See how clinicians use it day to day on the AI therapy notes page.
Many BastionGPT users report saving more than 90 minutes per day on documentation.
HIPAA-compliant with a signed BAA on every plan. Your data is never used to train models. BastionGPT drafts, you review and sign.
Each of the 15 items is answered yes or no about the past week and scores one point when answered in the depressive direction; some items count when answered yes and others when answered no, so the total (0 to 15) is not a count of yes answers and the key has to be applied. There is no single mandated cut point. The Stanford author site says patients with totals higher than 5 should be interviewed carefully, most validation studies evaluate a threshold of 5 or more (pooled sensitivity 0.89 and specificity 0.77 in a 2016 meta-analysis), and the author site's scoring page, echoed in secondary guidance, reads totals above 10 as almost always depression. The informal bands (normal, mild, moderate, severe) that circulate for the GDS-15 come from toolkits rather than a validation paper and do not map onto DSM-5 severity. Chart the form, the total over 15, the threshold you applied with its source, and the result as a screen; if items were missed, use the author site's proration rule (scale the completed items to 15 and round up) and say so.
They are separate forms with different item sets, ranges, and threshold conventions, not one scale at different lengths. The GDS-30 is the 1982 original, scored 0 to 30, with the original validation treating totals up to about 10 as the normal range. The GDS-15 (Sheikh and Yesavage, 1986) is the form in general use, scored 0 to 15, with the interview trigger above 5 described above. The GDS-5 of Hoyl and colleagues (1999) selected the five GDS-15 items most associated with a clinical diagnosis in frail outpatients and is conventionally read as positive at 2 or more, with the development study's strong accuracy figures coming from a small veteran outpatient sample. The four-item set of D'Ath and colleagues (1994) has no stable threshold, because moving the boundary by one point trades most of its sensitivity for specificity. A 2021 systematic review counted five four-item and seven five-item versions in the literature, so a short form is charted with its author as well as its length. The practical test is the bare score: a 7 is positive on 15 items, normal on 30, and impossible on 4 or 5, which is why the form is the first thing in the entry.
Through mild and moderate impairment, when the patient can understand and answer yes-or-no questions, and with the cognitive status written into the entry; not in severe impairment. In 24,674 US participants, Bennett and Wang (2024) found negligible bias in GDS-15 scores at questionable, mild, and moderate impairment (CDR 0.5 to 2) and significant bias at severe impairment (CDR 3), and a 2026 analysis of 14,077 people found that 13 of the 15 items behaved differently by cognitive status, which is why the chart records what was observed instead of declaring the scale valid. Completion collapses as cognition falls: among people aged 85 and over, 1.5 percent with an MMSE of 0 to 4 and 41.7 percent with 5 to 9 could complete it, against more than 90 percent at 20 and above. The CCSMH guideline (2021) recommends the GDS or PHQ-9 for older adults without significant cognitive impairment and an observer-rated instrument such as the Cornell Scale for Depression in Dementia instead of the GDS in moderate to severe impairment, and British Columbia's guideline places the short GDS with minor cognitive loss or early dementia and the Cornell with advanced impairment. No authority sets a MoCA or MMSE score at which validity stops, so document the screen and result, delirium status, the diagnosis and stage, and whether the patient followed the format, and when self-report is unreliable say so and use the Cornell.
Not as the patient's GDS. Reading the items aloud and recording the patient's own answers is an accepted administration mode (the author site even describes a point-board for aphasia), and that is charted as read aloud. A relative answering in the patient's place changes the measurement source, which is why Brown and Schinka (2005) built a separate 15-item informant version, the GDSI-15, hosted on the Stanford site. Its limits matter: in nursing homes the informant version performed worse than self-report in residents without dementia (area under the curve 0.67 against 0.75 to 0.79) and neither version detected depression adequately in residents with dementia (Li and colleagues, 2015), and proxy and self-ratings often disagree. So: use self-report while the patient can answer; if collateral report is the only option, use the informant version or an instrument designed around informant input such as the Cornell Scale, chart it as informant-rated with the informant's relationship, and keep the collateral history in the interview rather than inside the patient's score.
Either satisfies the wellness visit. CMS asks for a standardized tool recognized by national professional medical organizations and names none, the annual screening benefit (G0444, once a year, in primary care with staff-assisted depression care supports) leaves the tool to the clinician, and the 2026 Quality ID 134 specification lists the Geriatric Depression Scale and the Cornell Scale among acceptable adult tools beside the PHQ-9, so a GDS-15 earns measure credit when the tool's name is in the record and a positive screen has a documented follow-up plan. Choose the GDS-15 when somatic illness would make sleep, appetite, and energy questions hard to read; choose the PHQ-9 when you want DSM-mapped symptom domains, a built-in self-harm item, and a validated severity scale for monitoring. The exceptions are regulatory: nursing homes complete the MDS 3.0 Section D mood interview (PHQ-2 to 9, or the staff-observation version) and home health agencies complete the OASIS-E2 mood section, so in those settings a GDS is an adjunct and does not replace the required items. On repetition, G0444 is limited to once per 12 months as a benefit rule, not a psychometric one; repeat the GDS-15 when it will monitor treatment or after a clinical change, on the same form and mode, following the outcome measure note conventions.
A separate, direct assessment, every time the screen is positive and whenever the interview or history raises concern. Unlike the PHQ-9, the GDS carries no item about death or self-harm, so a positive total does not discharge the duty to ask and a low total does not rule anything out; the Hartford Institute's guidance is that any positive short-form screen should prompt an in-depth assessment and an evaluation for suicidality. Document the question asked in plain terms, the answer, the tool if one was used (a C-SSRS screener, for example), protective factors, and where the formulation and any safety plan are filed, using the structure on the suicide risk assessment page. One quality-measure detail: Quality ID 134 encourages a suicide risk assessment when appropriate but does not count it as the follow-up plan, so the plan still needs a referral or intervention for depression documented in its own right.
Both facts, then the interview that reconciles them. Write the result as a screen ("GDS-15 7 of 15, positive at the stated threshold; patient does not identify herself as depressed"), record what you observed (affect, engagement, tearfulness, psychomotor change), and document the interview: duration, loss of interest, function, grief, medical contributors, medications, alcohol, and sleep, with a working impression that leaves the diagnosis open until criteria are assessed. Older adults often endorse the mood and outlook items while rejecting the word depression, and the GDS-15 threshold of 5 or more classified 34.2 percent of older adults as depressed against 14.8 percent by structured interview in a 2024 meta-analysis, so a positive screen with a denial is common and neither fact wins on its own. Then document the same-day plan the measure requires (a return visit for a fuller assessment, a referral, or an intervention), the separate suicide-risk question, and any labs; the follow-up conversation belongs in a narrative progress note or the psychiatric diagnostic evaluation if a diagnosis is made.
At the topic level the items ask about mood, satisfaction with life, interest and activity, outlook and hopelessness, energy and concentration as the person experiences them, and self-worth, over the past week; by design there are no items about sleep, appetite, or physical symptoms and none about suicide. This page does not print or paraphrase the items; the Stanford author site is the source of record and hosts the forms, the scoring guidance, the informant version, more than 40 translations it cannot vouch for, and free scoring apps. The site states that the original scale is in the public domain because it was partly the result of federal support, so printing it, building it into a health system's own record for administration, and citing the authors are all ordinary uses. Two cautions: that statement covers the original scale, and the short derivatives (Hoyl's five-item form, D'Ath's four-item set) were published in journals, with at least one licensing database listing rights conditions for the short forms, so a commercial app, an EHR vendor, or a public website should confirm the rights of the exact version it embeds; and a translation posted on the site is not a validated version, so chart the language and, where one exists, the validation study.
Yes. Give it the facts (form and language, mode, the cognitive screen and result, the total and denominator, any missing items, the threshold your practice applies, the prior score, the interview findings, the suicide-risk question and answer, and the plan) and it drafts the full entry: the screen result with the threshold named, the cognitive statement, a screening-level interpretation, the separate risk line, the dated follow-up plan, and the claim, ready for your review. It can also check a finished note for a form or denominator missing, proxy answers charted as the patient's score, no cognitive statement, a yes-count or hidden proration, a band charted as a diagnosis, a risk conclusion drawn from the total, or a positive screen with no dated plan. BastionGPT is HIPAA-compliant with a signed BAA on every plan, and your data is never used to train models.
The instrument facts and compliance claims on this page trace to these sources, last verified September 2026:
Educational content, not legal or billing advice. Sample notes are fictional. Follow your organization's policies and your board, payer, and jurisdiction requirements.