The Apgar score in clinical research: for what, how and by whom it is used

Introduction

Seventy years after Dr. Virginia Apgar published her eponymous scoring system for describing the condition of the newborn infant in the first minute after birth [1], it is used in the vast majority of deliveries around the world. Yet over the decades, concerns about its reliability [2], [3], [4], prognostic value [5], [6], [7], how to account for interventions [8] and prematurity [9], as well advances in the care of the infant immediately after delivery [10] have raised questions about its continued value [11], [12], [13]. This has led to studies of modified Apgar scoring or even entirely new scores [3, 14, 15].

Before undertaking any effort to update a newborn assessment score, there is a need to define the purpose of such a score. Dr. Apgar’s original purpose for her score was to “predict survival, to compare … methods of resuscitation and … to compare perinatal experience in different hospitals.” [1]. She later added that it helped ensure closer observation of the infant, and endorsed its value for neonatal research and in predicting neurodevelopmental deficits after testing in the late 1950s to early 1960s showed an association between low scores and both poor neurodevelopment and death [16].

The purpose of this study was to examine how the Apgar score is used, based on an examination of published peer-reviewed clinical research papers indexed in PubMed. We also looked at how this may have changed over time by comparing results of our reviews between 1989–90 and 2018–19.

Methods

The PubMed database of published medical articles was searched using the PubMed search engine in February 2020 and again in December 2020. All titles listed using the MeSH Terms “apgar score” OR “apgar” AND “score” AND “humans” AND “English language” published between 1/1/1989–12/31/1990 or 1/1/2018–12/31/2019 were reviewed by one of the investigators. The first epoch was chosen because electronic access to articles was more common by this epoch, reducing the risk of selection bias. Those that described 3 or fewer cases, were primarily reviews, editorials or letters, or that were about the Surgical Apgar Score [17] or the Family Apgar [18] were eliminated (see Figure 1). The remaining articles were reviewed for (a) the medical or professional specialty of the first author; (b) the location of the work if specified, or of the first author (classified as Africa, Asia, Europe, Oceania, North America, or South America); (c) the primary purpose for which the Apgar score was included using a priori definitions (Table 1); (d) how the scores were recorded for use (which we termed usage) such as completely (0–10), divided into categories (e.g., 0–3, 4–6, 7–10), with a threshold (e.g., <7), both the one and 5 min scores or just a single value.

Figure 1:

Consort diagram of search retrieval, reasons for elimination and final medical literature papers used in the 1989–90 and 2018–19 epochs.

Differences in the geographic distribution, primary specialty, purpose, and score use between the two epochs were compared by Chi-square or Fisher’s Exact test. Significance was defined as p<0.05.

Results

Figure 1 details the initial number of papers identified and how many were eliminated from the final analysis for the specified two epochs. There was an absolute increase of 61.1% in the number of publications in the more recent 2018–19 epoch compared with the earlier 1989–90 sample. However, when accounting for the fourfold increase in the overall number of publications indexed in PubMed from 1990 to 2019 [19], the rate of Apgar score – related publications did not change (Figure 2). There was a significant shift in the geographic distributions, with an increase in papers published from Africa, Asia and Europe, accompanied by a 70% drop in the proportion originating from North America (Table 2). Regarding the distribution of authors’ profession (Table 2), while statistically different between epochs, those professions focused on conditions prior to delivery (obstetrics and anesthesia) constituted 60.1% in 1989–90 vs. 58.9% in 2018–19, and similarly, pediatricians/neonatologists were 21.5 vs. 19.6% between the earlier and later epochs.

Figure 2:

Articles indexed in PubMed listed by search using “Apgar score” per 100,000 articles by year of publication. Source – http://esperr.github.io/pubmed-by-year.

Table 3 shows that between the two time periods, there were significant changes in the purpose for what the Apgar score was reported. Of note, there was a 10% drop in the proportion of studies where the score was used to describe the neonate, from 24.4 to 14.0%, while the papers using the score as an outcome measure of conditions and treatments in the mother went from 1/5 to over 1/3 (20.3–34.9%). In both the earlier and later epochs, there was a wide variety of ways the score was defined as important (Table 3), totaling over 50 permutations of time and either individual total score, or a threshold (e.g.<7) or by categories (e.g. 0–3, 4–6, 7–10). The most common usage in 1989–90 was the actual scores at 1 and 5 min (31.8%) while in 2018–19 it was a 5-min score <7 (38.1%). The 1 min score was used alone in 11.9 and 15.2% in 1989–90 and 2018–29 respectively, the 5 min score in 29.2 and 46.7% and both the one- and 5 min scores in 53.0 and 31.2% (p<0.0001).

We examined whether the profession (Table 4) or score purpose (Table 5) influenced the way a paper defined what a significant score was, i.e. usage. Comparing score usage definitions between those involved with mothers (OB and anesthesia) vs. those unfocused on the child (Neonatal and pediatric), there was a significant difference in 2018–19 but not 1989–90. By purpose, the distribution of usage did differ in the earlier time (p=0.019) and even more so in the more recent epoch (p<0.00002).

Discussion

There are several unique observations within this study. The first is that between 1989–90 and 2018–19, there was far stronger rise in the number of papers that used the Apgar score from Africa/Asia/South America (155 more publications, a 267% increase) than from Europe/Oceania/North America (35 more publications, a 13.8% increase). This could be due to one or more factors, reflecting growth in obstetric or newborn research in the former regions or wider use or acceptance of the Apgar score. Alternatively, it could reflect an overall increase in access to publishing in established English-language medical journals. Another observation is that in both epochs, the majority of first or primary authors were in professions that focus on the mother rather than the baby. While most of the specialties within the ‘other’ category would be considered child-centered, when these are combined with the neonatal/pediatric authors, no more than one third of the authors in each epoch could be identified as primarily pediatric. This would seem to indicate that as a research tool, the Apgar score is viewed as a way to measure the effect on the fetus or baby of something occurring or done to the mother. It is a reminder that Dr. Apgar was an obstetric anesthesiologist when she developed her score and in her original paper, the actual scoring was performed by anesthesia residents [1]. It is also important to note that one of the major aims of her score was prognostic [1], that she validated her score against in-hospital mortality in her second report [20], and that the sharp rise in the number of Apgar score publications between 1969 and 1974 followed the publication of the reports from the Perinatal Collaborative Project results on the relationship between Apgar score and mortality [21] and morbidly [22]. It is not surprising therefore, that despite concerns about the predictive ability of the score dating back decades [23, 24], 18.5% of papers published in 2018–19 used scores as risk factors for later outcomes.

The fact that there is such a variety of ways investigators define the impact or significance of the Apgar score likely illustrates the lack of validation for the score in any specific use. As noted, there were more than 50 different usages. In 1989, the actual value between 0 and 10 was used in 48.6% of the published papers, but only two thirds of these included both the 1- and 5 min. The usages that defined a threshold below which a score was considered significant used thresholds as high as 9 and as low as 2. The most common usage did change between 1989–90 and 2018–19, from the 1- and 5-min actual scores to the score just at 5 min below a defined threshold. The most common threshold selected was <7, which was listed in 142 papers or 28.3% of the 501 published in 2018–19. Yet despite the relative popularity of this specific usage, its scientific rigor is not well established. In its 1986 statement warning about the use and abuse of the Apgar score [23], the America Academy of Pediatrics stated that “a 5-min apgar [sic] score of 7–10 is considered “normal”” and referenced the work of Nelson and Ellenberg from 1981 [5]. This paper used the data from the National Perinatal Collaborative and showed that among those > 2,500 g birthweight, 3.3% had a score of <7, and in those there was a 10.1% incidence of death or cerebral palsy, compared with a 1.2% chance in those with a score of 7–10. Population-based studies confirmed that the incidence of later neurological problems or death was more than 14-fold higher with lower scores at 5 min (although it was also true that 91.5% of the affected children has a 5-min score ≥7) [25]. However, more recent studies such as from Cnattingius and colleagues from Sweden demonstrate that, while the relative risks for CP or seizures are far higher in term with a 5-min score of 0–3, even those with a score of 9 have a significantly increased risk compared to those with a score of 10 [26]. There also appears to be a trend toward more specialization in including the Apgar in research reports. In 2018–19 there were significant differences in the usage of the score based on profession and by study purpose but either was not present or was not as significant in 1989–90.

This is, as far as we know, the first comprehensive examination of the role of the Apgar in clinical research but there are limitations to our analysis that must be noted. We used the first author and their listed department affiliation to define the profession, but this may not always be an accurate way to assess this variable. Multicenter studies involving sites from different continents posed issues as well as when, for example, data from hospitals in Africa were analyzed and the study written by researchers in Europe. In a few studies, the main purpose of the study was not the purpose to which the Apgar data and analysis contributed, and while Apgar purpose was defined a priori in our study, it was not always clear cut. Each of these concerns affected only a small number of the over 800 papers included in our analysis but could add up to a sum that could influence the analysis.

It is not surprising how, after almost seventy years, the use of a universal numeric assessment tool of newborns has expanded from Dr. Apgar’s original intentions. This is analogous to how new applications for drugs are often found that were not included in the initial approval process. The variety of ways that the significance of a score is defined, however, does seem to reflect the lack of rigorous validation studies that would provide far more confidence in the score as an outcome variable or a risk factor. This may be especially concerning if these clinical research studies reflect how clinicians use the Apgar score in practice.

The results from this study should inform any effort to update newborn assessment scoring. It points to the need to include multiple constituencies in the planning, including obstetricians, anesthesiologists, epidemiologists, pediatricians, midwives, and others. Any scoring system should include directions or guidelines on how to interpret the scores, rather than have a wide variety proliferate over time, as seen here. Finally, the continued and expanding use of the Apgar score indicates the that a newborn assessment system is still useful and needed.