Validation of the Chinese SAD PERSONS Scale to predict repeated self-harm in emergency attendees in Taiwan

Suicide can cause significant years of life lost and bring impacts to service providers and the society as a whole [1]. The challenge of high suicide numbers but relatively few research in Asian countries have called for more attention in prevention strategies in Asia [2]. In this context, predicting suicidal behaviour for appropriate interventions has become an emerging issue for the clinicians [3, 4]. Given the limited evidence on suicide risk assessment in predicting risk levels [5‐7], more research are needed to investigate the validity of risk screening tools.

Past and repeated self-harm or suicide attempts were long-term risks to completed suicide [8‐10]. Recognizing repetition risks of suicide by non-mental health professionals was an important variable that could improve patient outcome [11]. Such risk identification task can be performed through the nurses and motivated general health workers [12, 13]. A simple scale and adequate risk assessment skills for first-line physicians and nurses would provide basic objective indicators of risk level [14] and raise awareness of the clinicians to offer professional support and attention to facilitate help-seeking behaviour among people with self-harm [15]. Though there was limited evidence on the accuracy of screening tools [7], risk screening under guidance of an effective and efficient tool appeared to be a critical and promising method for high-risk identification, follow-ups and further interventions [16, 17].

Few established scales of suicide risk assessment were developed and studied [18], with some assessing indirect risks such as suicide ideation [19], depression or other general health indicaters [20]. Though depression scales such as the Patient Health Questionnaire provided acceptable diagnostic properties [21], its predictive capability may not be equal to that of direct risk assessment given reliable responses acquired. Supporting evidence for suicide scale validation was still limited [22]; even more limited in the Chinese populations. Moreover, few studies compared the differences among scales that directly assess suicide-related risks such as suicide intention (e.g. the Pierce Suicide Intention Scale or Beck Suicide Intention Scale), hopelessness (e.g. the Beck Hopelessness Scale) or a relatively comprehensive suicide risk assessment (e.g. the SAD PERSONS Scale) that was originally developed to assess the need to be hospitalized among emergency attendees with self-harm. The aim of the study was to demonstrate the psychometric properties and an adequate cut-off point of the Chinese version of the SAD PERSONS Scale (CSPS) to predict 6-month self-harm repetition and its applicability in general medical settings.

The patient participants were consisted of 147 people with self-harm from the Emergency Room at baseline and the 6^th month and collected 284 patients without self-harm from the Family Medicine Department in the same hospital. The response rate at 6-month follow-up was 74.8%. Table 1 showed sample characteristics. In this female predominating (72.1%) sample mean-aged 35.6 (standard deviation 13.9) years, the 6-month repetition rate was 21.8% (regardless of self-harm types).

The study provided evidence in identifying an optimal cut-off value of the CSPS to predict self-harm repetition with an acceptable concurrent and predictive validity. The qualitative findings also highlighted its applicability in general medical settings by non-mental health professionals. The scale was relatively brief and valid as well as adequate for suicide risk assessment in the clinical sample. Given its low internal consistency derived from different components of risk factors, the ten-item scale appeared to predict a higher risk for self-harm repetition. At cut-off of 5 point and above, the CSPS could reasonably classify a relatively high-risk group for future self-harm.

The findings of the CSPS validation were limited to comparability due to rarity of such studies [7, 34]. In a Canadian study, the CSPS was not found to accurately predict future suicide attempts owing to a low sensitivity (40%) and low positive predictive value (PPV) (7.4%) [22]. In applying the 2 point cutoff approach to predict repetition as used in the above study, our results similarly indicated that the CSPS had good classification performance (sensitivity 56.5%, PPV 87.4%) and supported the predictive value of the scale. But we failed to proof its predictive performance by logistic regression analysis suggested by the above study in using separate CSPS items or total score alone to predict 6-month repetition, as we found only the item of Depression could predict future self-harm risk rather than many other items (data shown on request). Although using the CSPS alone might result in considerable false-positives, personnel workload, and extra resources allocated in staff training, it was evident that the scale was comparatively an appropriate option for suicide risk prediction than other scales tested in our study. Besides, the relatively low predictive ability of the PSIS shown in Figure 1 worth attention; it was possibly related to item contents which rated the severity of suicidality from one single self-harm episode that mostly reflects cross-sectional severity rather than a person’s consequent suicide risk level. Whereas in comparing to other scales with the CSPS, it also performed significantly better than depression, hopelessness, or suicide ideation measurements in self-harm prediction. Therefore it would be beneficial to build up staff awareness and train the first-line clinicians to use the CSPS for suicide risk assessment. This implication was similar to that of the others which emphasizing the importance of nesting risk screening in routine care by general nurses in general medical settings [35, 36]. But according to nurses’ opinions, specific training may be needed on how to establish therapeutic relationship and get reliable responses for Item 4, 8, 10 in the CSPS.

The study was to the best of our knowledge the first that applied a relatively comprehensive method to validate the CSPS in a Chinese population and provided evidence-based data for its validation. We believe that the concerns for low PPV of the CSPS might be significantly improved if risk assessment is combined with another short scale for suicide risk evaluation such as depression. The strategy of augmenting risk prediction ability by using two brief and valid tools concurrently to screen for suicide risk level was also suggested by other study [37]. For example, one may include the CSPS with another brief depression scale such as the BSRS [19, 38], the PHQ-2 [39] or other tool specifically for a certain age group such as the elderly (e.g. the Geriatric Depression Scale) to increase precision in predicting the risk level. The CSPS might provide a wider clinical usage in the general medical context for suicide risk identification rather than focusing on people with psychiatric illness, given that the latter group of people had different set of risk factors compared to the general population [34].

The study was among the few that discussed validation of the SAD PERSONS scale and presented both concurrent and predictive validity for relatively objective conclusions. The strengths included that it explored feasibility and applicability of clinical use in general hospitals and involved service providers’ (i.e. the nurses’) opinions which complement the study results. Our findings were salient in its mixed data sources with rigorous quality control performed across two major general hospitals. However, the results should be interpreted under several limitations. Firstly, it could be a concern that we tested the scale validity in the emergency department at one hospital but evaluated its applicability based on the nurses’ opinions from another hospital. Nevertheless, in the medical context of Taiwan, people are likely to visit doctors between general hospitals due to physician’s reputation and affordable medical costs. The risk for subsequent self-harm in the patients from the two hospitals and the source of patients that nurses care for in these two hospitals are regarded as similar due to the approximate characteristics (e.g. mean age and gender ratio) of the patient source to that of a case register study in Taipei [40]. Thus recruitment from the two medical settings would not raise serious problem in interpretation of results. Secondly, the outcome of scale prediction was repeated self-harm. The results derived from such measurement may not be comparable to that of completed suicide. But repetition of self-harm was recognized as an important risk factor more prevalent than suicide ideation and has been adopted in other studies [22, 37]. Therefore we regarded this to be an acceptable outcome measurement. Thirdly, we failed to recruit more patients and/or follow up the participants for longer than 6 months due to resource shortage, thus generalization was limited to emergency attendees with self-harm and the exclusion of those with severe physical complications after the index self-harm acts, cognitive dysfunctions, or those who refused to participate at baseline assessment. Given the fact that we failed to collect 6-month data in the control group which may affect the richness of the results and limit the analysis of this group, we focused on establishing the psychometric properties using the case group in this study and had the follow-up period longer than that of other study, e.g. three months [37]. To have a better understanding of patient source of the study groups and the limitation of the hospital controls in this study, refer to our previous publication [41]. Finally, there were factors not considered to be tested in the study but might regarded as repetition risk predictors, such as impulsivity, drug abuse, and life event. However, some of those factors were unforeseen (e.g. life event) and not detrimental to predict 6-month repetition of self-harm [37], so the influence of not including those in the study would be small. Future studies could validate using the CSPS in predicting completed suicide in longitudinal studies or generalizing the results to longer term risk prediction for higher risk of repeated self-harm in a wider clinical populations.

We concluded that the CSPS is a valid and brief instrument suitable for self-harm repetition prediction. It was a considerable instrument for suicide risk screening based on its brevity and feasibility. However, cautions should be paid to level of therapeutic relationships during assessment, staff workload and adequate training for wider clinical applications.

Mr. Wang is a 47 year-old man who admitted to the Metabolic Ward, day five after his debilitating condition with poor sugar control. He was mostly alone at bedside without a caregiver. His primary nurse found several parallel scars on his left arm during physical check-up. He just briefly explained that he got those scars from his work. According to electronic medical records, he had two prior admissions to the emergency room due to cutting wounds at left wrist with unknown causes. He had no other major diseases or service attendance records. The night shift nurse observed that he had poor sleep and almost kept awake for the night since admission. He disclosed his bad mood and poor concentration and also showed limited activity level or interests in watching television or reading. His appetite became poor since last few weeks, which caused his body weight to lose from 68 kg to 59 kg. He could take care of himself in basic activities of daily living, but he was lack of motivation to move or talk and appeared slow-motioned than ever before. He felt he was not himself like before, and these changes have seriously affected his life quality and work. While the primary nurse who concerned about his situation asked what had happened to him, he replied, “I have been feeling tightness on my chest for weeks, and difficult breathing too. I’m wondering why I have to stay alive and face all these problems? My family hates me because I drink, but that’s because I have so much tension from work that no one could imagine … I have suffered enough from my work. Even I worked hard, I still got sacked and then my wife and kid left me. Now I’m alone. I feel so sad but no one understands me, I just feel like ending my life…”