Suicide ranks as a leading cause of death in America, with recent estimates up 2.6% from 2021 to 2022 following a 5% increase the year before.[1] However, for each documented suicide death, there are an estimated 3 hospitalizations for self-harm and another 265 people who seriously considered suicide.[2] Many patients seeking behavioral health care will be in this latter group, supporting the need for clinicians to identify and treat suicidal ideation to improve outcomes for these patients. While assessment of suicidal ideation, and ultimately suicide risk, is routine in behavioral health settings, there remains room for refinement and increased understanding.
Suicidal ideation, defined as thinking about, considering, or planning suicide, is an important predictor of risk for suicide and often a key component of treatment and discharge planning. The role of suicidal ideation in the overall wellbeing and safety of behavioral health patients prompted a deeper look at suicidal ideation in the MHO dataset. Within MHO data, there are two distinct methods to identify suicidal ideation:
1. Diagnosis: Episodes of care with a primary or secondary diagnosis code of R45.851 are indicative of a psychiatrist’s diagnosis
2. Self-Report Item: Did the patient endorse suicidal ideation items on self-report measures at admission? Endorsement of the PHQ-9 [3] Item “Thoughts that you would be better off dead or of hurting yourself in some way” or BASIS-32 [4] Item “Suicidal feelings or behavior”.
Notably, while the diagnostic process employed by each psychiatrist remains unknown, the self-report items are structured to facilitate positive responses. Research in primary care settings reveals questions about suicidal ideation framed for positive responses, rather than “no” responses, can elicit more information[5]. Furthermore, research has shown that single suicidal ideation items from other depression questionnaires have demonstrated correlation with more comprehensive suicide assessments [6]. Therefore, we aimed to examine and compare both methods of identifying suicidal ideation in our dataset.
MHO’s database contained 373,995 episodes of care for adult (18+) inpatients and PHP/IOP patients during 2022 and 2023, where both diagnostic and patient self-report suicidal ideation item data were available. A primary or secondary diagnosis of suicidal ideation (R45.851) was present for 59.7% of care episodes and 61.5% endorsed any suicidal thinking at admission on suicidal ideation items while 22.9% endorsed the highest severity on these items.
Suicidal Ideation Rates
Suicidal ideation rates by demographic groups were compared for the different methods of identification (See Figure 1). While there do appear to be differences in suicidal ideation rates across demographic groups, the two methods of identifying suicidal ideation are generally agreeable in those rates with notable differences in the PHP/IOP setting, for patients who have a primary substance use problem, and in patients aged 65+. In both instances, patients are much more likely to self-report suicidal ideation than to have a diagnosis.
Agreement in patient-self report and diagnosis
Of real interest is determining the degree of agreement between the different methods. We found modest agreement (к=.36, p<.001) between patient self-report and diagnosis; Patient self-report of suicidal ideation and the psychiatrist’s diagnostic impression matched in 70% of episodes of care (See Figure 2). It’s worth pointing out that a 30% mismatch rate is approximately equal to 1 in 3 episodes of care. Amongst all mismatched episodes of care, mismatches were evenly split between who did and did not endorse suicidal ideation (psychiatrist did but patient did not, or psychiatrist did not but patient did).
When the same analysis is performed with a more strict definition of a positive patient self-report of suicidal ideation, that is those patients who select the “worst” possible response to the suicidal ideation items, the rate of patients who self-report but do not have the accompanying diagnosis falls from 16.1% to only 4.6%.
This fair to modest agreement is found for most patient demographic groups as well (к range from .28 to .39), with the exception of level of care. For inpatients compared to PHP/IOP patients, there was better agreement between patient self-report items and psychiatrist diagnosis (к=.34, p<.001 and к=.19, p<.001 respectively).
Implications
While generally patients and psychiatrists agree on suicidal ideation[7], there are some important reasons why both pieces of information are important. Both the psychiatrist’s assessment of the patient and the patient’s assessment of their own symptom severity play a shared role in overall patient care.
First, the use of standardized survey items allows consistent information gathering[8]. Data from these items can be used by psychiatrists for assessment and treatment planning. Furthermore, when patients use self-report items to report suicidal thoughts, this can identify areas where additional standardized in-depth suicide risk assessment can be used for refining risk.
Second, patient diagnoses become part of the medical record that transmits health care information from provider to provider. In contrast, a patient survey item may not be recorded for other providers. Making sure each patient’s set of diagnoses provides as complete information as possible enhances continuity of care and informs care as patients move through treatment.
What do discrepancies in reporting indicate? When patients do not endorse suicidal ideation items, but a diagnosis of suicidal ideation is recorded, it may be that the patient felt more comfortable disclosing this information during the clinical interview or as the therapeutic relationship progressed. The reverse scenario is more challenging to interpret. Though a patient may indicate suicidal ideation through self-reported admission questionnaire items, the physician’s assessment may not lead to a formal diagnosis of suicidal ideation once they gain more details about severity and intention. And since assessment is ongoing during care, the patient may end up later reevaluating their initial assessment or demonstrating fluctuations in suicidal ideation [9]. It is also possible the diagnostic record did not accurately document the patient’s suicidal ideation. If the latter is the case, it can impede the continuity of care.
Regardless of how we identify suicidal ideation in our dataset, there is evidence that more than half of adult patients seen in our psychiatric units are experiencing suicidal thoughts The agreement between these methods of identifying suicidal ideation suggests responses to individual survey items can play a critical role in gathering information about patients. This information can then be used by the psychiatrist to support assessment and treatment success for our patients who are experiencing suicidal ideation.