Caring Contacts, a schedule of brief, nondemanding messages, carries a weak recommendation in the 2024 VA/DoD suicide guideline, written for VA and DoD care: periodic caring communications, in addition to usual care, for 12 months after hospitalization related to suicide risk, to reduce suicide attempts. On November 2, CARE, a randomized trial in American Indian and Alaska Native communities, reported no detectable benefit from a culturally adapted version over enhanced usual care.
The result is easy to overread in either direction. It does not show that Caring Contacts fails. It says something specific about what messages added in one setting, and reporting problems limit how firmly even that can be stated.
What was compared
CARE recruited adults in five communities in Alaska, the Northern Plains and the Southern Plains from 2018 to 2022; the paper analyzes 709 randomized participants. Entry required a Suicidal Ideation Questionnaire-Junior (SIQ-Jr) score of 32 or higher or a suicide attempt in the past year; 85.2% reported a lifetime attempt. Most were recruited outside formal health care settings.
Everyone received enhanced usual care (EUC): an individualized card summarizing behavioral health, crisis, traditional healing and community resources, reviewed with study staff, who facilitated warm handoffs when appropriate and remained available to help with access. Staff did not provide psychotherapy or ongoing clinical management. Everyone also completed a baseline assessment, which 72% of later survey respondents remembered as very positive. The intervention arm was also assigned 25 messages over 12 months, written by local staff, with individualized replies when participants wrote back.
So CARE did not test messages against nothing. It tested whether messages add to a personal start that included a tailored resource plan and an offer of help. The trials behind the guideline made different comparisons. For one, in patients admitted to an Iranian hospital after self-poisoning, the guideline reports that the trial's authors described usual care as poor and inadequately coordinated.
What the results show
No between-group difference was statistically significant:
Suicidal ideation at 12 months: adjusted coefficient 0.51 SIQ-Jr points (standard error 1.31), slightly higher, meaning worse, with Caring Contacts. By our approximate calculation, the 95% interval runs from about 2.1 points lower to 3.1 points higher with Caring Contacts, on a 0 to 90 scale.
Any suicide attempt during follow-up: 34 of 240 (14.2%) with Caring Contacts versus 39 of 238 (16.4%) with EUC; adjusted odds ratio 0.90 (95% CI, 0.55 to 1.50).
Suicide-related hospitalization: 13 of 208 (6.3%) versus 14 of 187 (7.5%); adjusted odds ratio 0.78 (95% CI, 0.34 to 1.77).
Nonsignificance is not equivalence. The attempt and hospitalization intervals are compatible with meaningful benefit, no effect, or harm. The ideation interval is narrow in scale points. We did not establish a validated adult threshold for clinically important change, so the interval does not show that any effect is trivial.
Among those assessed, both arms improved substantially: mean SIQ-Jr fell from the low 40s to about 23, and past-year attempts fell from 48.0% to 14.2% with Caring Contacts and from 43.4% to 16.4% with EUC. That is not evidence that EUC or the assessment worked. Without an untreated arm, the change cannot be separated from regression to the mean in people enrolled near a peak of risk, natural course, or effects of participation, which the commentary calls a placebo effect. The design cannot distinguish these.
Why the null is harder to read than it looks
Follow-up reporting is the largest problem. The text reports that 454 of 709 participants (64.0%) provided 12-month data, below the 70% retention the revised power calculation assumed; noncompleters were more likely to be male, never married, less educated and lower income. But the flow diagram shows 386 participants completing 12-month follow-up, 68 more labeled as completing 18-month follow-up, and 454 analyzed. Outcome denominators differ again: 478 for attempts, about 390 for the ideation thresholds. The paper does not explain which assessments entered which analyses, and we cannot resolve it. Multiple imputation, briefly described, assumed data missing at random, which cannot be verified. Deaths, including suicide deaths, occurred in both arms, too few for any comparison.
The trial was also smaller than planned. The original grant proposal called for 1,200 participants. A revised power analysis and smaller sample went to the monitoring board in January 2020, before the pandemic; the published target was 730, powered for relatively large effects, such as a risk ratio of 0.6 for attempts. The authors say the pandemic then slowed recruitment and contributed to attrition, and medical record review was discontinued, leaving hospitalizations to self-report. Assessors were blinded; participants were not.
Some outcome changes were documented in advance of this report. The published protocol records that a fourth primary outcome, still meeting the active-suicidality entry threshold, was added after the trial began, with the monitoring board, and that rare event counts could be analyzed as any versus none, as the paper did. Planned connectedness measures, a mediation analysis and 18-month results are not reported here, which does not show they were withheld.
Measurement and delivery raise further questions. The commentary argues that the SIQ-Jr has been validated in Native communities only in children and younger adolescents, and that the adapted attempt measure is untested. We did not establish whether adult validation exists. Uncertain measurement could affect eligibility, sensitivity to change and transferability; its effect on the comparison between arms is unknown. The paper describes real-time monitoring of messages and replies but reports no aggregate delivery or reply counts. In a supplementary reception survey, more than half of responding intervention participants were classified as having gotten fewer than half the messages; how that relates to actual delivery is not stated.
Failed or negative?
Michael Allen's commentary argues there are reasons to consider this "a failed rather than a negative trial" and that "it seems premature to conclude that Caring Contacts does not work." That is a fair warning against overreading. But the failed-trial label should not become a way to discount an unwelcome result. CARE was community-partnered, randomized after baseline assessment, and specified its intervention in detail. Whether assignment to a message program helps under real community conditions is itself a meaningful service question.
The most defensible reading is narrow. In these communities, against this comparator, assignment to messages produced no detectable incremental effect, behavioral outcomes remain imprecise, and the follow-up reporting leaves unclear which data support each estimate. The result does not describe American Indian and Alaska Native communities as a whole, and the data do not show that culture explains it.
The earlier evidence warrants modesty, not dismissal. According to the guideline, its 43% pooled reduction in attempts (compatible range 20% to 60%) comes from two trials. In a military trial that randomized 658 service members, both groups received standard care; the most common services were individual therapy and medication. Its baseline assessment was designed to be supportive. Both primary outcomes were null, but two secondary outcomes favored messages: any suicidal ideation during follow-up (79.6% versus 87.7%) and any suicide attempt (9.0% versus 14.9%). That trial's secondary benefits occurred alongside substantial background treatment. Cross-trial differences do not establish why CARE was null; populations, outcomes, delivery and era also differed.
What to do with it
In VA and DoD care, CARE does not overturn the guideline's weak suggestion. Elsewhere, the guideline is relevant evidence rather than a directive, and CARE is a reason not to assume messages add much when people already have a tailored resource plan and an offer of help.
Treat messages as the guideline frames them: an addition to usual care, not a replacement. A message schedule is not a follow-up plan.
CARE does not establish the value of what both arms shared, such as the resource card, staff availability or the baseline contact, or how messages compare with them.
Programs that send messages should weigh the considerations the guideline names: nondemanding, culturally adapted content, delivery barriers, and replying to time-sensitive responses. Recording delivery and receipt would clarify implementation; evaluating incremental benefit locally requires a credible comparison.
Sources
Comtois KA, Hebert LE, Bogic M, et al. Caring Contacts and suicidality among American Indian and Alaska Native communities: a randomized clinical trial. JAMA Netw Open. 2026;9(11):e2639023. doi:10.1001/jamanetworkopen.2026.39023. Published November 2, 2026. Full publisher text reviewed online November 4, 2026, including Methods, Tables 1 to 3, the Figure 1 flow diagram and Limitations. Funded by the National Institute of Mental Health; authors reported no conflicts.
Comtois KA, et al. Supplement 1, Trial Protocol and Statistical Analysis Plan, and Supplement 2, eTables 1 and 2 and eAppendix, to the article above. Full PDFs reviewed November 4, 2026; Supplement 1, a 124-page bundle including the original grant application, a February 2020 data and safety monitoring board memo and the published protocol, was reviewed in targeted sections.
Allen MH. A negative or a failed trial of Caring Contacts: understanding a disappointing result. JAMA Netw Open. 2026;9(11):e2638969. doi:10.1001/jamanetworkopen.2026.38969. Invited commentary; full publisher text reviewed online November 4, 2026. Interpretation, not new data.
Hebert LE, Fruhbauerova M, Evanson A, et al. Caring Texts, a strength-based, suicide prevention trial in 5 Native communities: research design and methods. Contemp Clin Trials. 2022;123:106966. doi:10.1016/j.cct.2022.106966. Author manuscript, full text reviewed November 4, 2026.
US Department of Veterans Affairs and Department of Defense. VA/DoD Clinical Practice Guideline for Assessment and Management of Patients at Risk for Suicide. Version April 2024. Full guideline reviewed November 4, 2026; recommendation 16 and discussion, pages 65 to 66. The pooled estimate and the description of the Iranian trial's usual care are cited as the guideline reports them; the underlying systematic review and Iranian trial were not independently reviewed.
Comtois KA, Kerbrat AH, DeCou CR, et al. Effect of augmenting standard care for military personnel with brief caring text messages for suicide prevention: a randomized clinical trial. JAMA Psychiatry. 2019;76(5):474-483. doi:10.1001/jamapsychiatry.2018.4530. Full text reviewed online November 4, 2026; supplements not reviewed.
Source note: This is commentary on interpreting one randomized trial of Caring Contacts against enhanced usual care. It does not establish that Caring Contacts is ineffective, that enhanced usual care or study participation caused the improvement seen in both arms, or that results transfer to other communities or care settings.
Educational Disclaimer: The Psychiatric Record provides general educational information for psychiatric and mental-health professionals. Content does not constitute medical, legal, regulatory, compliance, billing, or other professional advice; does not establish a standard of care; and is not a substitute for independent professional judgment. Appropriate assessment and treatment depend on the individual, setting, and current authoritative guidance. Patients should coordinate treatment decisions with their treating clinicians.