Web-accessible critical limits and critical values for urgent clinician notification

Definitions

A critical limit is defined as a low or high quantitative threshold of a life-threatening diagnostic test result. A critical value is defined as a qualitative result (e.g., a positive COVID-19 rapid antigen test) warranting urgent notification. Both demand rapid response, ideally within minutes [8, 9] and potentially life-saving treatment, isolation of the patient, or other timely medical intervention.

Open-source searches

PubMed articles and internet postings were identified using browser keywords (e.g., critical limits, critical values, critical-risk results, significant-risk results, alert values, hypercritical results, and panic values). This yielded 270 articles, editorials, letters to the editor, blogs, and peer-reviewed papers dealing with critical limits, critical values, urgent notifications, relevant artificial intelligence, urgent communication of critical results, and related topics.

Only those articles directly related to raw data, notification lists, national surveys, interpretation, and associated topics (please see MLO [10, 11]) were included in the final 70 selected. Raw data and policy retrievals from the Internet were parsed into university hospital and independent laboratory findings. Multiple search hits zeroed in on the same hospital documents thereby providing assurance that none were missed. This research was conducted April 2023 through April 2024.

Scope

We focused on hospital Web listings in the United States. All possible university and children’s hospital critical notification lists were sought along with an equal number of independent laboratory postings to balance statistical analyses. The 28 states with postings were well distributed across America. Lists were analyzed for changes in the number of measurands, laboratory categories (i.e., disciplines), quantitative thresholds, and qualitative findings in relation to those observed in 1990–93 national surveys [1], [2], [3], [4].

Databases

Summaries of critical limits and critical values for adults and children from past national hospital surveys have appeared in the MLO Clinical Laboratory Reference (CLR) annually from 1992 through 2023 [12] and for adults through 2024 [13]. Raw data from the United States national surveys [1], [2], [3], [4] archived by Knowledge Optimization (Davis, CA) were used to compare 2023 Web listings. That is, we compared changes in raw data from 1990 to 1993 to Web-listed raw data in 2023. Measurands not found or documented in the 1990s surveys are marked by ellipses (…).

Changes over time

Changes in quantitative critical limits were determined by comparing the means and medians of quantitative critical limits in 1990–93 national surveys [1], [2], [3], [4] to those listed on the Web in 2023. In addition to clinical chemistry, hematology, and other laboratory disciplines, notifications studied in detail comprised: (a) cardiac biomarkers – for their current importance in emergency and crisis care, (b) ionized calcium – to assess the impact of potential modifications of electrochemical methods since 1987 [3, 14], and (c) highly infectious diseases, such as COVID-19 – to update actionable tests used during the pandemic.

Statistics

The normality of 2023 data was determined by applying the Shapiro-Wilk test (https://www.statskingdom.com/shapiro-wilk-test-calculator.html) and confirmed by inspecting histograms. Very few 2023 quantitative critical limits were normally distributed. Therefore, we used primarily the Kruskal–Wallis non-parametric test to determine if differences over three decades were statistically significant (https://www.socscistatistics.com/tests/kruskal/default.aspx).

Differences were considered significant when p<0.05, or if p<0.01, then highly significant. Cell entries in the Tables are boldface when significantly different. Please note that the Kruskal–Wallis test does not directly compare medians, but instead the raw data distributions and therefore can generate a significant p-value even if two medians are equal.

In the case where both 1990s and 2023 data were normally distributed, the significant difference was determined using Student’s t-test for the means with unequal variances (https://www.omnicalculator.com/statistics/t-test) and identified with a superscript “ND” (for normally distributed) next to the footnoted p-value.

The order of measurands in the tables was determined from frequencies of 2023 Web survey listings. Listing frequencies were determined by dividing by 26 (2023 Web survey), 92 (1990 adult national survey [1]), or 100 (1993 national survey of ionized calcium critical limits [3]).

Units for critical tests

The percentages of university hospital notification lists using conventional units were: 64 % for chemistry, 100 % for blood gas, 98 % for hematology and coagulation, 89 % for newborn, and 78 % overall for quantitative tests. We report tabulated results in conventional units to allow facile access to statistical summaries without the need for healthcare providers to use conversion factors in the United States or in other countries still using conventional units.

Ethics

The UC Davis IRB deemed this research exempt (ID 2078118-1). Data for the Web survey were extracted from public domain open sources accessible on the Internet. No hospitals or organizations were contacted. There were no interviews. Results are reported anonymously.

Crisis intervention

Critical notifications should reflect life-threatening diagnostic test results, focus on treatable conditions and isolation of highly contagious patients, and avoid false positives. The responding clinical team may need to act immediately. Tests identifying serious abnormalities (e.g., elevated cardiac troponin) for which treatment is indicated or signaling need for isolation (e.g., positive COVID-19 test) take high priority. Published evidence presented next suggests that hospitals should re-evaluate critical limit thresholds that have changed since the 1990s.

Associations with mortality

Of the quantitative critical limits found in 2023 to have statistically significant differences since the 1990s, low and high white cell count; low glucose (adults and newborns), total calcium, and platelet count; high magnesium, partial thromboplastin time, and potassium (newborns); and high hematocrit (newborns) were among the top 15 tests associated with patient death within 24 h of reporting test results according to Yang et al. [19].

Arterial pO₂ (adults and newborns) and pCO₂ (adults) also had statistically significant differences but were not included in the Yang et al. [19] assessment. The new listing of INR reflected standardization since the 1990s. Therefore, to maintain relevance to life-threatening conditions, we recommend that quantitative notification thresholds be reviewed now and updated annually.

Listing frequencies

Critical limits (including blood gases) that had significant changes in notification thresholds over the past three decades comprised 60 % of the top 20 tests ranked by individual institutions, laboratorians, or clinicians, and professional bodies based on the categorization of frequencies of observation documented in a 2016 review [20]. In all three categories (institutions, clinicians, and professional bodies), ionized calcium occupied the 21st rank in frequency of listings [20]. The non-critical span (high minus low critical limits) for ionized calcium agreed well with that originally documented in a 1993 national survey [3] despite technological advances, such as smaller and faster ion-specific Ca⁺⁺ biosensors and processing for direct measurements used by whole-blood analyzers [21].

The magnesium median high critical limit increased. High and low magnesium thresholds should be re-evaluated in support of clinical management [22]. Ionized magnesium was not listed. Critical notifications for cardiac biomarker tests, which were listed by about half of the university hospitals (but not consistently), should meet the urgent needs of rule-in/rule-out protocols for the diagnosis of acute myocardial infarction [23].

Increase in listings

Three decades ago, 95 %  of respondents at university hospitals listed only 16 chemistry, 4 blood gas, and 7 hematology quantitative critical limits with a total of 42 tests for adult patients [1], plus a relatively limited set of qualitative critical values. Web retrievals show that notification lists have expanded. The myriad qualitative notifications partially explain why the current totals of tests listed are high.

University hospitals in the top quartile listed 89–116 tests. Typically, these were large tertiary care academic medical centers with complex patients and several laboratory disciplines serving a broad spectrum of clinical specialties. We recommend that these hospitals assess whether critical results are diligently and timely notified to responsible clinicians, and importantly, if they trigger therapeutic actions that improve patient outcomes.

Infectious diseases and biothreats

Now endemic, COVID-19 variants are precipitating fall-winter surges [24]. Surprisingly, 88.5 % of university hospitals did not identify the detection of this pandemic threat as warranting urgent notification. Those that did (11.5 %) likely benefit from enhanced awareness, accelerated infection control, and mitigated contagion [25, 26]. In contrast to COVID-19, listings of bioterrorism agents (see Table 6A) and pathogens (see Table 6B) demanding rapid detection were encountered frequently. The latter aligned best with National Institutes of Health priorities.

Point-of-care testing

With one exception, urgent notification lists specific for POCT could not be retrieved, an unusual finding following worldwide expansion of POCT for the detection of COVID-19. Not explicitly listing point-of-care tests may increase liability if therapeutic delays occur. Also, actionable bedside critical results may not be communicated consistently to authorized decision makers capable of rapidly assessing and treating patients [27]. A survey abstract titled “POC Programs” stated that “glucometer” critical limits were listed by 95.2 % of 63 respondents; the abstract did not enumerate other measurands [28].

Children’s hospitals

We found only two children’s hospital notification lists on the Web. In 2023–2024 the MLO CLR [13] stopped publishing critical limits for children. Investigation of newborn and pediatric critical limits is necessary going forward because of innovative diagnostics (e.g., transcutaneous CO₂ and bilirubin monitoring), advances in the management of premature births, and updated life-saving therapeutic protocols. While not traditionally the purview of laboratorians, continuous monitoring has evolved to the point where rapid changes could trigger fast decisions and therefore merit notifications.

Trends and cautions

Investigators have addressed critical notifications in anatomic and surgical pathology, cytology, cytogenetics, molecular genetics, neonatology, pediatrics, radiology, surgery, virology, and other specialties. Some authors question the value of repeating critical results to confirm them [29], [30], [31]. Others analyzed false positive critical value results [32]. With COVID-19 rapid antigen and molecular critical value notifications [33], both false positives and false negatives should be of concern, the former for triggering unnecessary isolation and possibly placing a patient in a high-risk ward, and the latter, for spreading disease by those unaware of infection [24], [25], [26].

Global awareness

Searches revealed relevant publications originating from Australia, Brazil, Canada, China, Croatia, Egypt, Ethiopia, Germany, India, Indonesia, Iran, Italy, Japan, Kenya, Korea, Kuwait, Nepal, Netherlands, Paraguay, Saudi Arabia, Singapore, Slovakia, South Africa, Spain, Taiwan, Thailand, Turkey, the UK, and other countries [20]. College of American Pathologists’ Q-Probes provided insight into notification polices and tests in the United States in 2002 and 2007 [34, 35]. This year, the US Joint Commission 2024 safety goals continue to specify that United States hospitals must implement critical values policies, select tests for urgent notification, and monitor timeliness [36]. Ultimately, global awareness may facilitate harmonization [37].

Guidelines

The Clinical Laboratory and Standards Institute guideline EP47 [18] equates critical limits with “alert thresholds” and “significant-risk results,” which may confuse and may not prompt immediate therapeutic intervention. Only one notification list, which used a mixed terminology, “Critical Risk Alert Thresholds,” was titled close to that suggested in EP47. A review [20] of notification practices concluded that outcomes studies and professional collaboration, which could be augmented by Web postings and AI, will be most useful for building future policies and test lists. Therefore, we briefly investigated the capabilities of Web-based AI answer engines.

Artificial intelligence

Artificial intelligence answer engines could help select decision thresholds and track the dynamics of changes, as well as synthesize knowledge bases, and if sufficiently granular, possibly enhance identification of highly infectious threats [38], timely responses to new outbreaks, and favorable quarantine outcomes. Open sourcing critical test lists on the Web would enable AI “data fusion,” that is, heuristic integration of multiple data streams [39]. Sharing both critical test lists and notification policies could enhance acute care, AI-facilitated decision making, and consistency [10, 11, 37].

Please see Supplementary Material 2 for a Venn diagram mapping integration of AI and critical notification practices and Supplementary Material 3 for sample dialogues regarding critical results management and associations with mortality generated by Perplexity (https://www.perplexity.ai), a highly touted and heavily funded San Francisco start-up and answer engine [40]. Artificial intelligence currently appears weak at discovering original Web raw data but could help support laboratory professionals, including facilitating clinician communications and responding to questions [41]. Web postings of decisions thresholds would enable higher level AI responses to questions (Figure S2) and possibly also accelerate a shift from high volumes of urgent notifications to assessments that moderate their frequency and enhance their value [42].

Limitations

The sample of twenty-six university hospital notification lists represents all that were accessible online. Nonetheless, Web results revealed potential changes in mortality-associated critical limits over the past three decades, adequate reason for caution and re-evaluation. Some analyses of quantitative critical limits were affected by non-normal frequency histograms, but the medians did not change.

We found listings of new infectious threats, such as COVID-19, but did not contact hospital staff to determine why some threats were not listed more frequently. Conversely, we found numerous listings of qualitative critical values but cannot fully explain their urgency.

In 1984 in the United States, it was recommended to embed SI units in the digital outputs of instruments, including bedside monitoring devices, and display both SI and conventional units together when reporting results to initiate an educational process of conversion from conventional to SI units [43]. In a 1985 report, the Council on Scientific Affairs of the American Medical Association endorsed changing units to SI in America [44].

In 1986, the Editor of the Journal of the American Medical Association (JAMA) wrote an article promoting the conversion to SI units [45]. The article included tabulated conversion factors for diagnostic tests. Following publication of the first US national survey of critical limits and critical values, which was published in JAMA [1], frequencies of the use of conventional and SI units were extracted from the survey data for a summary report the same year [46] in follow-up to the national initiative. We observe from the current Web survey that the majority of hospitals in the United States still are using conventional units.

Outcomes

We recommend periodic review of quantitative thresholds and qualitative notifications. Rapid response to extremely abnormal test results helps prevent adverse adult and pediatric patient outcomes [19, 20, 47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68]. Secure messaging can improve efficiency by activating teams responding to hypercritical test results [69]. Automated messaging makes pharmacokinetic sense when alerting dynamic changes in drug levels, identifying the presence of toxic elements, quickly triaging patients with critical results, and improving efficiency.

Changes over time

We recommend that hospitals evaluate changes over the past three decades for clinical impact. Figure 1 summarizes changes in need of review, which could have resulted from assay modifications, calibration shifts, elimination of outliers, new therapies altering decision thresholds, or other factors, such as bias in the limited sample size. When assessing the impact of changes over time, both low and high critical thresholds should be evaluated because the non-critical span bracketing urgent treatment decisions may have changed.

Figure 1:

Recommended strategies for improving critical notification practices. This flowchart strategizes process steps that will enhance standards of care (green) for critical notification practices. The first steps are re-evaluation of quantitative critical limits that have changed since 1990–93 (red) combined with review of new and neglected qualitative notifications (yellow). Collaborative sharing (blue) combined with AI-facilitated discovery will produce high impact in the 21st Century. AI, artificial intelligence; COVID-19, Coronavirus disease 2019; POCT, point-of-care testing.

Global threats

We recommend adding new notifications, such as positive COVID-19 test results, whether obtained in the laboratory or community, to urgent notification lists to speed mitigation of contagion and life-saving intervention. Sharing will advance transparency. Posting notification lists on the World Wide Web [70] and facilitating AI should help hospitals to identify new tests and critical values, adjust for changes in quantitative critical limits, and assess the status of bioterrorism agents and pathogens classified as high priority for rapid response.