Evidence level: C
用体质标签先把人筛一遍、再用 ASSR 精选,省下的每一分钟测试时间都要用敏感度和漏检率来付账——筛查医学有现成账本,体质标签这一侧的账还没建。
正文
选钵流程目前面临一个真实的成本问题:ASSR 校准要求对每个人做逐频点的跟随响应测量,还要跨天重测确认信度,时间开销不小;而体质标签(量表加医师分型)几乎零成本。于是出现一个自然的设计:先用体质标签把人群粗分,只让「预判会响应」的亚组进入 ASSR 精选,其余人直接按标签选钵——这就是「两阶段方案」。这个结构在临床筛查里不是新发明,新生儿听力筛查就是标准的两阶段:耳声发射初筛,未通过者再进自动听性脑干反应复筛。
关键在于:两阶段设计的效果是可以算账的,不需要争论。2025 年一项系统综述加 meta 分析专门汇总了两阶段新生儿听力筛查的质量指标(PMID 40308920);微模拟建模被用来比较不同筛查流程的成本效果(PMID 33306547);2026 年儿童代谢相关脂肪性肝病筛查用「腰高比初筛 → 影像确诊」的两阶段结构做了正式的成本效用分析(PMID 42194435);更早还有把「全量齐测」与「序贯检测」头对头做成本效果比较的诊断策略研究(PMID 16117745)。这些工作共同的底层指标是筛查医学三件套:敏感度、特异度、阳性预测价值(PMID 9709269)。
折损有三个来源,每一个都能量化。第一是漏检:两阶段听力筛查的多中心实测发现,确实存在一部分永久听力损失婴儿在两阶段都被判「通过」而漏掉(PMID 16140706)——预筛门槛画在哪里,直接决定漏检率。第二是预筛偏倚:通过初筛的样本不再代表全体人群,用这批人测出的「ASSR 精选效度」会被系统性高估,外推回全量人群就失真;临床试验领域已经把预筛数据本身当作招募建模的正式输入来管理(PMID 37131229),说明预筛环节的结构性影响被明确承认。第三是返工:门槛放太松,进 ASSR 的人几乎没少,初筛白做;放太紧,漏检激增。而确认端 ASSR 自身也有精度边界——阈值估计精度受刺激参数与受试状态等多个变量影响,已有 meta 分析定量汇总(PMID 17978952),相关综述与儿童阈值估计研究给出了适用范围(PMID 22436114、30197427)。
裁决设计因此很明确:同批受试者全量跑 ASSR 校准并完成体质分型,然后离线模拟不同预筛门槛,输出三条曲线——每人节省的 ASSR 测试分钟数、漏检率、精选后响应预测 AUC 的变化——找帕累托折点。这个裁决有一个前置依赖:体质标签对 ASSR 响应的判别力(AUC)本身还未测出,它是 #758、#760、#764 系列裁决的下游问题。体质侧已有的支撑是 Ayurgenomics 链:Prakriti 分型人群在倾斜应激下呈现体质特异的 HRV 响应(PMID 36106418),且 L1 边表中「dosha 分型—HRV 响应」是一条 A 级证据边(PMID 29269119)。
冲突面必须摆出来:听力筛查两阶段成功的前提,是初筛工具的敏感度已知且足够高;体质标签对 ASSR 响应的敏感度从未被测过,阳性先例不能直接照搬。
门槛扫参的判读规则也要预写死,否则三条曲线出来仍可能各说各话。扫参范围建议从「全部放行」到「只放行标签最强的四分之一」连续取七个档位;每档同时报告三个量并绑定成组,不许只报省时这一项。判定规则预写三条:若漏检率在任何档位超过预设容忍线(例如响应排在人群前 30% 者被挡掉超过 5%),该档位直接出局,无论省时多少;若省时曲线与 AUC 折损曲线在同一个档位出现交叉,取交叉点为帕累托折点;若所有档位 AUC 折损都超过 0.05,则裁决两阶段不可行,回到全量 ASSR。另一条容易被忽略的成本是标签采集本身的时间——医师分型虽便宜但占用人力,若改用纯量表,还需先过标签信度校正这一关,否则初筛噪声会顺着漏检率与偏倚两个通道同时放大。更根本的一条:筛查医学要求目标终点本身明确且可测(PMID 9709269),而选钵响应的终点目前仍在锁定过程中——64 通道脑电研究未发现颂钵对 HRV 各参数的离散效应(PMID 35630011),阴性面提示终点通道未定。地基比听力筛查薄得多,这正是本文把证据等级压在 C 的原因。
证据与局限
已验证: 两阶段筛查结构的成本效果分析范式成熟(两阶段听力筛查、腰高比两阶段肝病筛查、序贯检测比较);ASSR 阈值估计的精度边界有 meta 分析;Prakriti 分型与 HRV 体质特异响应有人体数据。
部分验证: 体质分型在 L1 对应网络中登记为量表层面的个体差异调节变量,与 HRV 响应的关联有 A 级边,但标签对 ASSR 跟随响应的判别力未测。
待验证: 两阶段选钵方案的三条曲线(时间节省、漏检率、AUC 折损)与帕累托折点;预筛后样本的效度高估幅度;终点通道(主观放松、HRV、ASSR 签名)本身的选择。
相关阅读
L1 网络状态:L1 对应网络中无本主题裁决边;相邻登记为「| 体质分型 | 量表 | 个体差异的调节变量 |」(EDGE_TABLE 无两阶段筛查/成本相关边)。该主题尚未进入 L1 对应网络。
证据等级: C
参考文献: PMID 40308920, PMID 33306547, PMID 42194435, PMID 16117745, PMID 9709269, PMID 16140706, PMID 37131229, PMID 17978952, PMID 22436114, PMID 30197427, PMID 36106418, PMID 29269119, PMID 35630011
证据等级:A=有PubMed直接支撑,B=科学类比/间接证据,C=传统文化共识/初步探索
免责说明:本文属传统知识与科学文献的整理、对照与研究设计讨论,仅供理解与参考,不构成医疗建议,不能替代专业诊断与治疗。文中涉及的效应量、频率区间与暴露结论多为研究报告层面的推断,个体情况差异很大;如有身体不适或健康疑虑,请就医并由专业医师判断。
Constitutional Tags as Low-Cost Initial Screening: Cost-Validity Trade-offs in the Two-Stage Selection Bowl Process
Evidence level: C
Using constitutional tags to pre-screen individuals and then refining the selection with ASSR saves every minute of testing time; however, each saved minute must be paid for in terms of reduced sensitivity and increased false-negative rates. While screening medicine has established accounting books, the ledger on the constitutional tag side remains to be built.
Body
The current process for selecting bowls faces a genuine cost constraint: ASSR calibration requires individual, frequency-point-by-frequency-point follow-up response measurements, along with cross-day re-testing to confirm reliability, incurring significant time overhead; conversely, constitution labels (scale plus physician typology) are nearly zero-cost. This creates a natural design: first use constitution labels to broadly stratify the population, allowing only the subgroup predicted to respond into ASSR refinement, while others are assigned directly by label—this constitutes the “two-stage scheme.” This structure is not novel in clinical screening; newborn hearing screening follows exactly this two-stage protocol: otoacoustic emission initial screening, with those failing proceeding to automatic auditory brainstem response re-screening.
The critical point is that the efficacy of a two-stage design can be quantified without debate. A 2025 systematic review and meta-analysis specifically aggregated quality metrics for two-stage newborn hearing screening (PMID 40308920); microsimulation modeling was employed to compare cost-effectiveness across different screening workflows (PMID 33306547); a 2026 cost-utility analysis of pediatric metabolic-associated fatty liver disease screening utilized the “waist-height ratio initial screen → imaging confirmation” two-stage structure (PMID 42194435); earlier still, diagnostic strategy research performed head-to-head cost-effectiveness comparisons between “all-at-once testing” and “sequential testing” (PMID 16117745). The shared underlying metrics across these works are the triad of screening medicine: sensitivity, specificity, and positive predictive value (PMID 9709269).
Three sources of loss must be accounted for, each quantifiable. First is missed detection: multi-center empirical testing of two-stage hearing screening confirmed that a subset of infants with permanent hearing loss were incorrectly classified as “pass” in both stages (PMID 16140706)—where the pre-screen threshold is set directly determines the miss rate. Second is pre-screen bias: samples passing the initial screen no longer represent the full population; thus, “ASSR refinement validity” measured on this subset becomes systematically overestimated, leading to distortion when extrapolated back to the full cohort; clinical trial methodology has already formalized pre-screen data itself as an input for recruitment modeling (PMID 37131229), acknowledging the structural impact of the pre-screen stage. Third is rework: setting the threshold too loosely allows nearly everyone into ASSR, rendering the initial screen pointless; setting it too tightly causes missed detections to surge. Furthermore, the confirmation-stage ASSR itself has precision boundaries—threshold estimation accuracy is influenced by multiple variables including stimulus parameters and subject state, as quantitatively summarized in meta-analyses (PMID 17978952), with related reviews and pediatric threshold estimation studies defining applicable ranges (PMID 22436114, 30197427).
Therefore, the decision design is clear: run all subjects through ASSR calibration and constitution typology, then offline simulate different pre-screen thresholds, outputting three curves—minutes of ASSR testing saved per person, miss rate, and change in post-refinement response prediction AUC—to identify the Pareto knee. This decision has one prerequisite dependency: the discriminative power of constitution labels for ASSR response (AUC) itself has not yet been measured; it remains a downstream issue of decisions #758, #760, and #764. Existing support from the constitution side comes from the Ayurgenomics chain: Prakriti-typed populations exhibit constitution-specific HRV responses under tilted stress (PMID 36106418), and the L1 side table establishes “dosha typology—HRV response” as an A-level evidence edge (PMID 29269119).
The conflict must be laid bare: the success of two-stage hearing screening presupposes that the initial screen tool has known and sufficiently high sensitivity; however, the sensitivity of constitution labels for ASSR response has never been measured, so positive precedents cannot be directly transplanted.
The rules for interpreting threshold sweeps must be pre-written, otherwise the three curves may speak different languages. The sweep range should span continuously from “all-pass” to “only the top quarter by label strength,” with seven discrete levels; each level must report all three quantities bound together, not just time saved. Three judgment rules are pre-specified: if the miss rate exceeds the preset tolerance line at any level (e.g., blocking more than 5% of those ranked in the top 30% of responders), that level is immediately discarded regardless of time saved; if the time-saving curve and AUC loss curve intersect at the same level, the intersection point defines the Pareto knee; if AUC losses exceed 0.05 at all levels, the two-stage design is deemed infeasible and full ASSR is resumed. Another often-overlooked cost is the time of label acquisition itself—while physician typology is cheap, it consumes human resources; if pure scales are used instead, label reliability correction must first be passed, otherwise initial screen noise will amplify simultaneously through both the miss rate and bias channels. More fundamentally, screening medicine requires the target endpoint itself to be clearly defined and measurable (PMID 9709269), yet the endpoint for bowl selection response remains under lock—64-channel EEG studies found no discrete effects of singing bowls on any HRV parameters (PMID 35630011), indicating the endpoint channel is undetermined. The foundation is far thinner than that of hearing screening, which is precisely why this paper assigns evidence level C.
Evidence and Limitations
Verified: The cost-effectiveness analysis paradigm for two-stage screening structures is mature (two-stage hearing screening, waist-to-height ratio two-stage liver disease screening, sequential testing comparisons); meta-analyses establish precision boundaries for ASSR threshold estimation; human data supports Prakriti typology and HRV somatic-specific responses.
Partially Verified: Somatotype classification is registered in the L1 corresponding network as an individual-difference moderator at the scale level, with an A-level edge regarding its association with HRV responses; however, discriminative power of labels for ASSR-following responses remains unmeasured.
Pending Verification: The three curves of the two-stage bowl-selection scheme (time savings, false-negative rate, AUC loss) and Pareto breakpoints; the degree of overestimation for post-screening sample efficacy; and the selection of endpoint channels (subjective relaxation, HRV, ASSR signatures) themselves.
Related Reading
- Three Causes: Dual Model Separation of “Constitutional Tags vs. Imbalanced States”
- ASSR Signature × Constitutional Tags “Shared Awakening Dimension” Mediation Model
- Singing Bowl “Response Predictor” Continuous vs. Discrete Gradient
- Five Research Pools
L1 Network Status: L1 corresponds to the network lacking decision edges for this topic; adjacent registration is marked as “| Constitutional Subtyping | Scale | Moderator of Individual Differences |” (the EDGE_TABLE contains no two-stage screening/cost-related edges). This topic has not yet entered the L1 network.
Evidence Level: C
References: PMID 40308920, PMID 33306547, PMID 42194435, PMID 16117745, PMID 9709269, PMID 16140706, PMID 37131229, PMID 17978952, PMID 22436114, PMID 30197427, PMID 36106418, PMID 29269119, PMID 35630011
Evidence level: A = direct PubMed support; B = scientific analogy/indirect evidence; C = traditional consensus/preliminary exploration.
Disclaimer: This article compiles traditional knowledge and scientific literature for comparison and research-design discussion only. It is not medical advice and cannot replace professional diagnosis or treatment. Effect sizes, frequency ranges and exposure conclusions cited here are largely inferences at the research-report level, and individual circumstances vary widely. If you have any discomfort or health concerns, please consult a physician.
Leave a Reply