Abstract
This quality improvement study assessed how responses changed after including a nonresponse option for a firearm access question as part of mental health screening.
Introduction
Firearms are the most common means of suicide, and suicide is the most common way people die by firearm in the US.1 Health care encounters prior to firearm suicide are common, particularly primary care visits, where evidence-based practices, like safe storage counseling and safety planning, can help mitigate risk.2,3 Firearm access screening may help identify patients who could benefit from these resources; however, underreporting is common—in one study, approximately one-half of patients who received and answered a question about firearm access and reported no access died by firearm suicide within the next 12 months.4 Moreover, qualitative research among firearm owners highlights notable opportunities to improve firearm access screening, including offering patients the choice to answer and improving transparency about the question purpose.5 Therefore, this study evaluated how these changes were associated with self-reported firearm access. We used a quasi-experimental interrupted time series study design to test the hypothesis that these changes would reduce proportions of patients reporting no access.
Methods
This quality improvement study was approved by the Kaiser Permanente institutional review board with a waiver for informed consent because use of protected health information involved only minimal risk. The study followed SQUIRE reporting guideline for quality improvement studies.
Firearm owners, suicide prevention experts, clinicians, and primary care staff codesigned changes, including the addition of a “prefer not to answer” nonresponse option to a screening question about firearm access, to mental health monitoring questionnaires for all adult patients with a mental health diagnosis (eMethods in Supplement 1). They also recommended the following descriptive tagline: “We ask because firearms are a leading cause of injury and death. Your care team may offer firearm storage resources for prevention.” The question changes went live December 11, 2023. Electronic health record (EHR) data were used to describe demographic characteristics of the primary care population who received the questionnaires and their self-reported mental health symptoms. A quasi-experimental interrupted time series design assessed the association of question changes with self-reported firearm access. Generalized estimating equations were used to test for changes in proportions of patients responding no or yes, analyzed separately (at the questionnaire level), after the question changes.6 All models adjusted for period (before vs after change), time (in weeks), their interaction, questionnaire type, and urbanicity vs rurality using national county-level data to account for regional differences in firearm ownership.7 Empirical sandwich variance estimates were used for inference to account for potential correlation misspecification due to repeated measures.8 For prefer not to answer, we calculated proportions of responses after the change period once it was added as a response option. Statistical significance 2-tailed and determined as 95% CIs that did not cross 1. Analyses were conducted using R version 4.4.0 and RStudio version 1.3.1056 (both from R Project for Statistical Computing) from May 13, 2025, to July 30, 2026.
Results
During the study period from May 10, 2023, to July 11, 2024, firearm access screening was completed on 47 396 questionnaires across 37 614 adult primary care patients (13 846 [37%] aged 40-64 years; 25 699 [69%] female), including 25 806 preimplementation (before December 10, 2023) and 21 590 postimplementation (December 11, 2023, onward), with similar patient characteristics and reported mental health symptoms across periods (Table 1). Postimplementation, a total of 1413 patients reported prefer not to answer across 1525 (7%) questionnaires (Table 2). Comparing preimplementation to postimplementation, there were reductions in the odds of reporting no (odds ratio [OR], 0.81; 95% CI, 0.69-0.97) and yes (OR, 0.82; 95% CI, 0.69-0.97) to the firearm access question. A crossover cohort sensitivity analysis (10 922 questionnaires across 4067 patients) yielded similar patterns as our primary analyses but lacked precision.
Table 1. Primary Care Patient Characteristics of Questionnaire Responders During the Study Periods.
| Characteristic | Patients, No. (%) | ||
|---|---|---|---|
| Overall (N = 37 614) | Prechange (n = 22 745) | Postchange (n = 18 936) | |
| Age group, y | |||
| 18-39 | 13 726 (37) | 7903 (35) | 6952 (37) |
| 40-64 | 13 846 (37) | 8516 (37) | 6780 (36) |
| 65-84 | 9310 (25) | 5880 (26) | 4837 (26) |
| ≥85 | 732 (2) | 446 (2) | 367 (2) |
| Sex | |||
| Female | 25 699 (69) | 15 436 (68) | 13 177 (70) |
| Male | 11 915 (32) | 7309 (32) | 5759 (30) |
| Race and ethnicitya | |||
| American Indian or Alaska Native | 592 (2) | 354 (2) | 312 (2) |
| Asian | 3761 (10) | 2281 (10) | 1784 (9) |
| Black | 2157 (6) | 1247 (6) | 1115 (6) |
| Hispanic | 2566 (7) | 1535 (7) | 1313 (7) |
| Multiracial | 1427 (4) | 835 (4) | 732 (4) |
| Native Hawaiian or Pacific Islander | 486 (1) | 288 (1) | 236 (1) |
| White | 27 956 (74) | 17 044 (75) | 14 147 (75) |
| Unknown | 2111 (6) | 1180 (5) | 1081 (6) |
| Otherb | 1201 (3) | 736 (3) | 582 (3) |
| Urbanicity or ruralityc | |||
| Rural | 1107 (3) | 677 (3) | 569 (3) |
| Suburban | 11 002 (29) | 6522 (29) | 5887 (31) |
| Urban | 25 505 (68) | 15 546 (68) | 12 480 (66) |
| Questionnaire-reported depression, anxiety, and alcohol, mean (SD)d | |||
| Depression (PHQ-9 Score) | 5.9 (5) | 5.6 (5) | 6.2 (6) |
| Anxiety (GAD-2 Score) | 1.6 (2) | 1.5 (2) | 1.6 (2) |
| Alcohol use (AUDIT-C Score) | 1.7 (2) | 1.7 (2) | 1.6 (2) |
Abbreviations: PHQ-9, Patient Health Questionnaire 9-item; GAD-2, Generalized Anxiety Disorder 2-item; AUDIT-C, Alcohol Use Disorders Identification Test Consumption.
Not mutually exclusive, obtained from health system administrative records based on patient self-report.
Patient self-reported race and ethnicity as other.
Based on the National Center for Health Statistics county urban-rural categorization (eMethods in Supplement 1).
Obtained via self-report on the same questionnaires as the firearm access screen (eMethods in Supplement 1).
Table 2. Questionnaire Self-Reported Firearm Access Before and After Implementation of Changes to a Standardized Firearm Access Screening Question.
| Questionnaire response | Preimplementation (n = 25 806) | Postimplementation (n = 21 590) | Change in postimplementation vs preimplementation level, OR (95% CI) | ||
|---|---|---|---|---|---|
| No. (%) | Slope, OR (95% CI)a | No. (%) | Slope, OR (95% CI)a | ||
| No | 20 690 (80) | 1.00 (0.99-1.02) | 16 605 (77)b | 1.00 (1.00-1.01) | 0.81 (0.69-0.97) |
| Yes | 5116 (20) | 0.99 (0.98-1.01) | 3460 (16)b | 1.00 (1.00-1.01) | 0.82 (0.69-0.97) |
| Prefer not to answer | NA | NA | 1525 (7)b | NA | NA |
Abbreviations: NA, not applicable; OR, odds ratio.
Comparing odds in any given week to a time period 4 weeks in the past.
Percentages calculated using all 21 590 questionnaires in the postimplementation period. Restricting to the 20 065 questionnaires with a no or yes response, the no and yes percentages were 83% and 17%, respectively.
Discussion
In this quality improvement study, adding a nonresponse option and a brief description of the question purpose was associated with significantly decreased odds of answering both no and yes to a question about firearm access. The additional response option was not associated with a selective reduction in the proportion of patients answering no. This finding underscores an important consideration for health systems that currently or plan to screen for firearm access. Specifically, findings underscore how firearm safe storage counseling and resources should be offered, regardless of how patients respond to screening questions, including to patients who choose not to answer. Limiting safe storage counseling to only those reporting firearm access will miss people at risk of firearm injuries.
It is methodologically impossible to determine whether these results were driven by the new nonresponse option or descriptive tagline because they were added simultaneously. Some patients may have perceived the tagline as stigmatizing or avoided answering yes because they did not want to receive safe firearm storage resources.
The study findings also highlight potential opportunities for clinicians and health care systems to use the patient-reported preference not to respond to potentially help identify patients at risk of firearm suicide and offer risk mitigation resources. For example, future research could evaluate how independent effects of patient-reported firearm access (including prefer not to answer) impact accuracy of suicide risk prediction algorithms that use EHR data to identify patients at highest risk of self-harm.9 Future human-centered design research could help clinicians engage patients in firearm risk mitigation when patients actively indicate their preference not to report their firearm access status. Screening is the backbone of preventive medicine and should be designed to support trusting relationships between clinicians and patients,10 a vital component of high-quality health care.
eMethods.
Data Sharing Statement
References
- 1.Centers for Disease Control & Prevention . Web-based Injury Statistics Query and Reporting System (WISQARS) NVDRS violent deaths report. Updated May 2, 2026. Accessed May 2, 2026. https://wisqars.cdc.gov/nvdrs/
- 2.Angerhofer JE, Cruz M, Shaw J, et al. Health care use preceding suicide by firearm compared with suicide by other means—Alaska, Colorado, and Washington, 2020-2022. MMWR Morb Mortal Wkly Rep. 2025;74(21):365-371. doi: 10.15585/mmwr.mm7421a2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Angerhofer Richards J, Cruz M, Stewart C, et al. Effectiveness of integrating suicide care in primary care: secondary analysis of a stepped-wedge, cluster randomized implementation trial. Ann Intern Med. 2024;177(11):1471-1481. doi: 10.7326/M24-0024 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Richards JE, Boggs JM, Rowhani-Rahbar A, et al. Patient-reported firearm access prior to suicide death. JAMA Netw Open. 2022;5(1):e2142204. doi: 10.1001/jamanetworkopen.2021.42204 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Richards JE, Kuo ES, Whiteside U, et al. Patient and clinician perspectives of a standardized question about firearm access to support suicide prevention: a qualitative study. JAMA Health Forum. 2022;3(11):e224252. doi: 10.1001/jamahealthforum.2022.4252 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. John Wiley & Sons; 2012. [Google Scholar]
- 7.Goldstick JE, Carter PM, Cunningham RM. Current Epidemiological Trends in Firearm Mortality in the United States. JAMA Psychiatry. 2021;78(3):241-242. doi: 10.1001/jamapsychiatry.2020.2986 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42(1):121-130. doi: 10.2307/2531248 [DOI] [PubMed] [Google Scholar]
- 9.Simon GE, Johnson E, Lawrence JM, et al. Predicting suicide attempts and suicide deaths following outpatient visits using electronic health records. Am J Psychiatry. 2018;175(10):951-960. doi: 10.1176/appi.ajp.2018.17101167 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bell NR, Thériault G, Singh H, Grad R. Measuring what really matters: Screening in primary care. Can Fam Physician. 2019;65(11):790-795. [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods.
Data Sharing Statement
