The DSM is based on extensive observations of patients, with suggestions on categories going back over 100 years. The originators commented that the models were not entirely adequate and needed further modifications¹. Current models, too, have been called “a first approximation” needing additional features to achieve better utility and validity². Specific issues identified as needing improvement include reliability, validity, completeness and utility^{3, 4}.

While standard clinical practice does employ DSM diagnoses in making treatment decisions, it often emphasizes additional information from patient assessment. That is, physicians often use a broad problem solving rather than a diagnosis specific approach⁵.

Explicitly targeting utility and completeness, we asked a sample of clinicians, by an online RedCap survey, how they use DSM diagnoses in the context of other clinical information in assessing and treating psychotic disorders (i.e, schizophrenia spectrum and bipolar and major depressive disorder with psychotic features). Psychiatrists surveyed were at 27 academic centers in the US, as they are the greatest users of DSM and are most engaged in ongoing consideration of how to choose and use DSM criteria. Answers were anonymous and physicians did not receive any compensation for completing the survey. The study was approved and classified as exempt by the Partners Healthcare institutional review board.

Respondents ranked the importance in their practice of nine clinical assessment considerations (DSM-5 diagnosis, specific presenting signs and symptoms, severity of signs and symptoms, history of signs and symptoms, comorbidities, treatment history, social assessment, family history, and medication history), rated for each of four clinical decision and intervention domains (prognosis, recommended level of care, recommended medications, and recommended psychosocial therapies), using a five-choice Likert-type scale ranging from not important (assigned a value of 1) to extremely important (assigned a value of 5).

Of 566 psychiatrists who were invited to participate in the survey, 129 (22.8%) responded. They represented both sexes, and many ages, regions, sites and types of practice. Results indicated that all of the nine assessment considerations were considered at least moderately important for at least one clinical purpose. Primary hypothesis testing found highly significant evidence of a greater mean rating for current signs and symptoms than other clinical assessment considerations (X²=667, p<0.001). Using a secondary intersection-union approach, we found strong evidence that psychiatrists rate current signs and symptoms as more important than every other assessment consideration included in the survey (mean importance rating=4.46, t=5.86, p<0.001). DSM-5 diagnosis had the lowest observed mean importance rating (mean=2.77).

Post-hoc t-tests found evidence that the mean for DSM-5 diagnosis was significantly lower than the mean for every other ­assessment consideration (mean>3.58, t_121-123<–9.65, p<0.001) except family history (mean=2.84, t₁₂₃=–0.77, p=0.44). Post-hoc tests using linear regression found no association of the difference in mean importance ratings between current signs and symptoms and DSM-5 diagnosis with age (t₁₂₂=–0.43, p=0.67); sex (t₁₂₀=1.04, p=0.30); US region (X₍₄₎ ²=1.21, p=0.88); site (categorized as hospital only, hospital and other, private practice only, and clinic only, X₍₃₎ ²=2.37, p=0.50); and number of patients seen (X₍₄₎ ²=0.97, p=0.91).

We did not sample all possible elements that clinicians use in assessments, but had an open question where psychiatrists could note factors not surveyed. Factors suggested included: previous diagnoses, age, cognitive function, risk or history of suicide or violence, forensic history, legal status, cultural background, social networks, work history, family involvement, insight, acceptance of illness and treatment, preferences among treatments, rapport between doctor and patient, and financial resources.

While our survey was being completed, a worldwide screen of expert opinion from mental health clinicians, assessing the value of ICD-11, which is similar to DSM in its categorical approach and content, was published⁶. This global survey addressed all the categories in the ICD and DSM, exploring the relative use of ICD/DSM for administrative purposes, managing treatment, communicating with other treaters, and teaching. Our survey targeted only US psychiatrists, focused on psychotic disorders, and obtained relative rankings of the use of DSM diagnoses versus other clinical findings in choosing and guiding treatment. Thus, the two studies were partially overlapping. Consistent with our project, the authors of the global survey concluded that the ICD and DSM categories are most useful for administrative and billing purposes and for communicating with other clinicians. They are least used and substantially less useful for choosing individual treatments or advising on prognosis.

Our results suggest that, among patients with psychotic disorders, the DSM-5 diagnosis is less important than identifying other individual features of illness, especially type and severity of symptoms, but also comorbidities and some aspects of personal history. Relevant factors noted by other investigators include suicidality, recreational drug use, obstetric complications, early or recent adverse events, social cognition and neurocognition⁵. The use of these factors allows more flexibility in description than categories alone. Course can be included as well.

Notable for interpreting the responses, we only contacted clinicians at well-known academic centers. The majority (70.5%) of respondents had hospital-based practices, but this might be expected for those who treat many patients with psychotic disorders. The results represent opinions of clinicians who teach and perform research, in addition to their clinical practices. Most psychiatrists did not respond. Nonetheless, the response rate (22.8%) was typical of online surveys⁷. Possibly, those who did respond were interested in the subject and might have thought about the matters raised. We are not suggesting that responders were representative of US psychiatrists, but it might be noted that the suggestions, made a century ago, on which ICD and DSM are based, were also from clinical observations, largely from clinicians in select sites. They were not made or since have been confirmed on the basis of other validators¹.

Lastly, an argument has been made that changes in DSM and ICD should strive to improve utility and accuracy⁸. Accuracy in choosing treatments and predicting outcome might be enhanced by incorporating factors that clinicians cite as most important into formal diagnostic systems. That these factors are already in use for making clinical decisions shows that they are practical and suggests that they may be valid. An improved system might incorporate both categorical entities and additional features, such as those provided by recognizing individual symptoms and severity of those symptoms, in new models⁹. Such models can be tried and tested, then implemented if they show advantages compared to existing systems.

DSM 基于对患者的广泛观察，其分类建议可追溯到 100 多年前。创始人评论说，这些模型并不完全足够，需要进一步修改¹ 。当前的模型也被称为“第一近似”，需要额外的特征来实现更好的实用性和有效性² 。确定需要改进的具体问题包括可靠性、有效性、完整性和实用性^{3, 4} 。

虽然标准临床实践确实采用 DSM 诊断来制定治疗决策，但它通常强调来自患者评估的附加信息。也就是说，医生经常使用广泛的问题解决方法，而不是特定的诊断方法⁵ 。

明确以实用性和完整性为目标，我们通过在线 RedCap 调查询问了一些临床医生，他们如何在其他临床信息的背景下使用 DSM 诊断来评估和治疗精神障碍（即精神分裂症谱系和伴有精神病的双相情感障碍和重度抑郁症）特征）。接受调查的精神科医生来自美国 27 个学术中心，因为他们是 DSM 的最大用户，并且最积极地持续考虑如何选择和使用 DSM 标准。答案是匿名的，医生没有因为完成调查而获得任何报酬。该研究得到了 Partners Healthcare 机构审查委员会的批准并列为豁免。

受访者对九项临床评估考虑因素（DSM-5诊断、具体表现的体征和症状、体征和症状的严重程度、体征和症状的病史、合并症、治疗史、社会评估、家族史和用药史）的重要性进行了排名），对四个临床决策和干预领域（预后、推荐的护理水平、推荐的药物和推荐的心理社会治疗）中的每一个进行评分，使用五项选择的李克特量表，范围从不重要（指定值为 1）到极其重要（分配值为 5）。

在受邀参加调查的 566 名精神科医生中，有 129 名（22.8%）做出了回应。他们代表了不同的性别、不同的年龄、地区、地点和实践类型。结果表明，所有九个评估考虑因素都被认为对于至少一种临床目的至少具有中等重要性。主要假设检验发现了非常显着的证据，表明当前体征和症状的平均评分高于其他临床评估考虑因素（X ² =667，p<0.001）。使用二次交叉联合方法，我们发现强有力的证据表明，精神科医生认为当前的体征和症状比调查中包含的所有其他评估考虑因素更重要（平均重要性评级=4.46，t=5.86，p<0.001）。 DSM-5 诊断的观察到的平均重要性评级最低（平均值 = 2.77）。

事后 t 检验发现证据表明 DSM-5 诊断的平均值显着低于其他所有评估考虑因素的平均值（平均值>3.58，t _121-123 <–9.65，p<0 id=111>123 =– 0.77，p=0.44）。使用线性回归的事后测试发现，当前体征和症状与 DSM-5 诊断之间的平均重要性评级差异与年龄没有关联（t ₁₂₂ =–0.43，p=0.67）；性别（t ₁₂₀ =1.04，p=0.30）；美国地区（X ₍₄₎ ² =1.21，p=0.88）；地点（分类为仅医院、医院和其他、仅私人诊所和仅诊所，X ₍₃₎ ² =2.37，p=0.50）；以及就诊的患者数量（X ₍₄₎ ² =0.97，p=0.91）。

我们没有对临床医生在评估中使用的所有可能元素进行抽样，但有一个悬而未决的问题，精神科医生可以注意到未调查的因素。建议的因素包括：既往诊断、年龄、认知功能、自杀或暴力的风险或历史、法医病史、法律地位、文化背景、社交网络、工作经历、家庭参与、洞察力、对疾病和治疗的接受程度、治疗偏好、医患关系和经济资源。

在我们的调查即将完成的同时，发布了一份全球范围的心理健康临床医生专家意见筛选报告，评估了 ICD-11 的价值，其分类方法和内容与 DSM 类似⁶ 。这项全球调查涉及 ICD 和 DSM 中的所有类别，探讨了 ICD/DSM 在行政目的、管理治疗、与其他治疗者沟通和教学方面的相对使用。我们的调查仅针对美国精神科医生，重点关注精神障碍，并获得了 DSM 诊断与其他临床发现在选择和指导治疗方面的使用相对排名。因此，这两项研究有部分重叠。与我们的项目一致，全球调查的作者得出结论，ICD 和 DSM 类别对于管理和计费目的以及与其他临床医生的沟通最有用。它们使用最少，对于选择个体治疗或建议预后的作用也大大降低。

我们的结果表明，在精神障碍患者中，DSM-5 诊断不如确定疾病的其他个体特征重要，尤其是症状的类型和严重程度，以及合并症和个人病史的某些方面。其他研究人员指出的相关因素包括自杀、消遣性药物使用、产科并发症、早期或近期不良事件、社会认知和神经认知⁵ 。与单独的类别相比，使用这些因素可以使描述更加灵活。课程也可以包括在内。

值得注意的是，我们只联系了知名学术中心的临床医生，以解释这些答复。大多数（70.5%）受访者都有医院执业经历，但这对于那些治疗许多精神障碍患者的人来说可能是意料之中的。结果除了临床实践之外，还代表了教学和研究的临床医生的意见。大多数精神科医生没有做出回应。尽管如此，响应率 (22.8%) 仍是在线调查的典型⁷ 。那些做出回应的人可能对这个主题感兴趣，并且可能考虑过所提出的问题。我们并不是说回应者代表了美国精神病学家，但值得注意的是，一个世纪前提出的 ICD 和 DSM 所依据的建议也来自临床观察，主要来自选定地点的临床医生。它们不是由其他验证者¹做出的，也不是已经被确认的。

最后，有人认为 DSM 和 ICD 的改变应努力提高实用性和准确性⁸ 。通过将临床医生认为最重要的因素纳入正式的诊断系统，可以提高选择治疗方法和预测结果的准确性。这些因素已经用于做出临床决策，这表明它们是实用的，并且表明它们可能是有效的。改进的系统可能会在新模型中纳入分类实体和附加特征，例如通过识别个体症状和这些症状的严重程度提供的特征⁹ 。如果这些模型与现有系统相比具有优势，则可以尝试和测试，然后实施。