A comparative analysis of APGAR score and the gold standard in the diagnosis of birth asphyxia at a tertiary health facility in Kenya

Background Birth asphyxia is a consistent key contributor to neonatal morbidity and mortality, notably in sub-Saharan Africa. The APGAR score, though a globally used diagnostic tool for birth asphyxia, remains largely understudied especially in resource-poor settings. Objective This study determined how effectively the APGAR score is used to diagnose birth asphyxia in comparison to the gold standard (umbilical cord blood pH <7 with neurologic involvement) at Moi Teaching and Referral Hospital (MTRH), and identified healthcare provider factors that affect ineffective use of the score. Methods Using a quantitative cross-sectional hospital-based design, term babies born in MTRH who weighed ≥2500g were randomly and systematically sampled; and healthcare providers who assign APGAR scores were enrolled via a census. Umbilical cord blood was drawn at birth and at 5minutes for pH analysis. APGAR scores assigned by healthcare providers were recorded. Effective use of the APGAR score was determined by sensitivity, specificity, positive and negative predictive values. At a significance level of 0.05, multiple logistic regression analysis identified the independent provider-associated factors affecting ineffective use of the APGAR score. Results We enrolled 102 babies, and 50 (49%) were females. Among the 64 healthcare providers recruited, 40 (63%) were female and the median age was 34.5years [IQR: 31.0, 37.0]. Assigned APGAR scores had a sensitivity of 71% and specificity of 89%, with positive and negative predictive values of 62% and 92% respectively. Healthcare provider factors associated with ineffective APGAR score use included: instrumental delivery (OR: 8.83 [95% CI: 0.79, 199]), lack of access to APGAR scoring charts (OR: 56.0 [95% CI: 12.9, 322.3]), and neonatal resuscitation (OR: 23.83 [95% CI: 6.72, 101.99]). Conclusion Assigned APGAR scores had low sensitivity and positive predictive values. Healthcare provider factors independently associated with ineffective APGAR scoring include; instrumental delivery, lack of access to APGAR scoring charts, and neonatal resuscitation.

Background : Birth asphyxia is a consistent key contributor to neonatal morbidity and mortality, notably in sub-Saharan Africa. The APGAR score is a globally used diagnostic tool for birth asphyxia; but there is paucity of data comparing this score with the gold standard (umbilical cord blood pH <7 with neurologic involvement), as well as limited studies assessing how effectively healthcare providers apply the score. Objective: This study determined how effectively the APGAR score is used in Moi Teaching and Referral Hospital (MTRH) compared to the gold standard in diagnosis of birth asphyxia and identified the healthcare provider factors that affect the effective use of the score. Methods : Using a quantitative cross-sectional hospital-based design, term babies born in MTRH who weighed ≥2500g were randomly and systematically sampled; and healthcare providers who assign APGAR scores were enrolled via a census. Umbilical cord blood was drawn at birth and at 5minutes for pH analysis. APGAR scores assigned by healthcare providers were recorded. Effective use of the APGAR score was determined by sensitivity, specificity, positive and negative predictive values. At a significance level of 0.05, multiple logistic regression analysis identified independent healthcare provider factors associated with effective use of the APGAR score. Results: We enrolled 102 babies, and 50 (49%) were females. Among the 64 healthcare providers recruited, 40 (63%) were female and the median age was 34.5years [IQR: 31.0, 37.0]. Assigned APGAR scores had a sensitivity of 71% and specificity of 89%. Babies who were resuscitated, those born via instrumental delivery, and those scored by healthcare providers without access to APGAR scoring charts were significantly more likely to get incorrect APGAR scores.
Conclusion: Assigned APGAR scores had low sensitivity and positive predictive values. Healthcare provider factors independently associated with effective APGAR scoring include instrumental delivery, lack of access to APGAR scoring charts, and neonatal resuscitation. Background: Birth asphyxia is a consistent key contributor to neonatal morbidity and 27 mortality, notably in sub-Saharan Africa. The APGAR score is a globally used diagnostic 28 tool for birth asphyxia; but there is paucity of data comparing this score with the gold 29 standard (umbilical cord blood pH <7 with neurologic involvement), as well as limited 30 studies assessing how effectively healthcare providers apply the score.    Healthcare provider factors independently associated with effective APGAR scoring 51 include instrumental delivery, lack of access to APGAR scoring charts, and neonatal 52 resuscitation.

54
Keywords: APGAR Score, umbilical cord pH, Birth Asphyxia, Western Kenya. Perinatal mortality is still a significant public health issue, with its top three aetiologies identified 61 as; prematurity (28%), neonatal infections (26%), and birth asphyxia (23%) [1,2]. Birth asphyxia 62 is a neonatal condition characterized by failure to initiate or sustain breathing at birth, and it has 63 been a consistent major aetiology of neonatal mortality in the last decade [3]. Globally, neonatal 64 mortality accounts for about half of under-five mortality [4]. From 2003-2014, childhood mortality 65 trends in Kenya showed that, among neonatal, infant, and under-5 mortalities, neonatal mortality 66 showed the slowest decrease [5]. Therefore, to decrease the mortality rate of children below five 67 years of age, it is imperative to decrease neonatal mortality [6]. To further curb neonatal mortality, 68 there is a profound need to explore the aetiologies of neonatal mortality such as birth asphyxia, 69 especially in resource-poor settings [7]. Sub-Saharan countries bear the greatest burden of birth 70 asphyxia, and these asphyxia-related deaths are for the most part preventable deaths [8] . 71 The APGAR score, though widely used for birth asphyxia diagnosis can either overestimate or 72 underestimate asphyxia due to its subjective nature [9]. The combined APGAR score has shown 73 some superiority to the conventional APGAR score in birth asphyxia diagnosis [10,11], and over 74 the years, severely asphyxiated babies have been found to have normal APGAR scores [12,13]. 75 However, in low-and middle-income countries, the conventional APGAR score continues to be 76 solely used for birth asphyxia diagnosis. An APGAR scores of less than <7 depict asphyxia; to 77 ascribe this asphyxia to an intrapartum hypoxic-ischemic event, the American College of 78 Obstetricians and Gynaecologists and American Academy of Paediatrics require an add-on 79 umbilical cord pH value of <7, evidence of neurologic involvement (e.g., seizures, altered tone, 80 coma) and evidence of multiorgan involvement [14]. On the other hand, umbilical cord blood pH 81 has been shown to have better diagnostic value in birth asphyxia [15,16], and when combined with 82 the APGAR score, higher sensitivity in birth asphyxia diagnosis has been observed [17,18]. 83 Birth asphyxia is a major contributor cause of neonatal morbidity and mortality. Globally, it 84 contributes to a quarter of all neonatal deaths. In Kenya, according to the KDHS 2014 report, birth 85 asphyxia is the leading cause of neonatal deaths at 31.6% [5]. It has steadily been identified as one 86 of the major aetiologies of neonatal mortalities in the last decade [3], this creates the need to assess 87 the performance of APGAR score as a diagnostic tool for birth asphyxia in comparison to the gold 88 standard. Failure to optimally diagnose birth asphyxia directly interferes with the timeline of 89 managing neonates, further increasing the risk of neonatal mortality. 90 Published studies report varying proportion of APGAR score sensitivities and specificities of 62.8-91 99% and 41-81% respectively [11,19,20]. Consistency in APGAR scoring among healthcare providers has been shown to vary greatly, with the heart rate component of the score having the 93 least inconsistency. Healthcare provider factors influencing APGAR scoring include among others: 94 years of experience of the healthcare provider, cadre of the healthcare provider, number of staff per 95 delivery, nature of resuscitation done by healthcare provider, and the type of delivery [21][22][23]. 96 Exploring these factors in this study is pivotal in fostering improved usage of the APGAR score by 97 healthcare providers in identifying neonates with asphyxia. 98 Improvements in antenatal care, delivery and neonatal resuscitation services have been observed in 99 many countries with developing economies. This has led to the establishment of national 100 intervention programs such as "Help Babies Breathe" in Kenya [24,25]. Yet, birth asphyxia 101 consistently remains a major cause of neonatal mortality, creating the need for effective use of 102 diagnostic tools such as the APGAR score. If the diagnosis of birth asphyxia is not effectively done, 103 it cannot be promptly and adequately managed. Consequently, its short-and long-term morbidity, 104 and mortality persistently occur as reflected by the above-mentioned data. This study therefore 105 aimed to determine how effectively the APGAR score is applied in the labour ward and theatre of 106 a tertiary hospital in Kenya. Furthermore, it assessed healthcare provider factors associated with 107 this effective use of APGAR score. This knowledge will aid in addressing and halting asphyxia-108 related morbidity and mortality through evidence-based data. 109

110
This was an analytical cross-sectional hospital-based quantitative study carried out in the babies are admitted to NBU monthly, creating an optimal study setting for birth asphyxia.

117
The researchers randomly and systematically sampled 102 term neonates (37-42weeks 118 gestation) who had a birthweight of at least 2500 grams. To assess healthcare provider 119 factors, 64 healthcare providers who assign APGAR scores to neonates after delivery were Fisher's exact test if any expected cell count was <5) was used to assess significant 135 differences in categorical variables across asphyxia and no asphyxia groups. The Wilcoxon 136 rank sum test was used to test association between predictors and outcomes for continuous 137 variables. Effective APGAR score use was computed as sensitivity, specificity, positive 138 and negative predictive values. Intraclass correlations did not identify the need for multi-139 level regression models in this study; thus, one-level logistic regression was used at a 95% 140 confidence level to determine independent factors associated with effective APGAR score 141 use, with the dependent variable being incorrect APGAR scores.

144
This study enrolled 102 neonates, of whom, majority were males 52 (51% Among all the 102 neonates enrolled, 21 (21%) had birth asphyxia defined as cord blood  with the presence of clinically apparent seizures and/or an altered tone (gold standard).   light what healthcare provider factors are associated with the effective use of the score in our 218

setting. 219
Effective use of the APGAR score 220 We estimated a sensitivity of 71% and a specificity of 89%. Our specificity was more comparable 221 to that of a study prospective cohort study conducted among 464 neonates admitted to a tertiary 222 hospital in Iran at 81% [11]. On the other hand, we had a lower sensitivity of 71% compared to the 223 81% in Iran [11]. Our results therefore show that, the APGAR scores assigned in MTRH need to 224 be improved upon for proper identification of neonates with asphyxia. 225 We postulate that, our low sensitivity can be explained by a form of ascertainment bias, which is 226 also in line with findings from a review on the APGAR score by the score's founder, Dr. Virginia 227 Apgar. She recommended in this review that, the person carrying out the delivery shouldn't be the 228 one assigning the APGAR score as they could have a vested interest in the neonatal outcome, and 229 also have divided attention in view of the need for continual maternal management [13,26]. The 230 practice for most deliveries in MTRH except caesarean deliveries is that the person carrying out 231 the delivery also assigns the APGAR score to the neonate. This could have therefore resulted in 232 incorrectly high APGAR scores, and true asphyxia being reported (per APGAR score) as no asphyxia, translating to low sensitivity. On the flip side, the scores observed in our study were 234 better at determining babies who did not have birth asphyxia. 235 According to a multicentre study that assessed umbilical cord blood parameters in the diagnosis of 236 birth asphyxia, a sensitivity of 99% and a specificity of 41% were obtained [20]. These findings 237 were dissimilar to ours. The difference could be explained by the fact that, in this study, APGAR 238 scores in the first minute were used; while, in our study, we used APGAR scores in the fifth minute 239 as our operational definition for birth asphyxia was adopted from the combined consensus of 240 ACOG/AAP. they had a much lower positive predictive value when compared with ours. This difference could 245 be explained by two things. Firstly, they applied the APGAR score in their study as a predictive 246 tool for neonatal mortality, while we used it for diagnosis of asphyxia in our study. Secondly, they 247 included extremely low birth weight babies in their study, and we did not. 248

Healthcare provider factors affecting effective use of APGAR score 250
In the current study, mode of delivery, need for resuscitation and lack of access to an APGAR 251 scoring chart were significantly associated with the effective use of the APGAR score. 252 Mode of delivery was significantly associated with incorrect APGAR scores, whereby, babies born 253 via instrumental deliveries were 8 times more likely to have incorrect scores when compared with 254 neonates born via spontaneous vaginal delivery. Additionally, neonates born via caesarean 255 deliveries were slightly less likely to be assigned incorrect scores with an odds ratio of 0.13. Our 256 findings can be attributed to the concept of normalcy by which, deliveries requiring intervention 257 are already viewed as a deviation from the norm -spontaneous vaginal deliveries and have a 258 beforehand expectation of poor neonatal outcome. Therefore, it is possible that this could have 259 negatively influenced how effectively the APGAR score was used in deliveries requiring 260 intervention in our study. This was however more applicable to instrumental deliveries than 261 caesarean births in our study and we think it was so because, we included babies born via elective 262 caesarean sections with healthy outcomes in which case, the assigning of APGAR scores is 263 forthright. Our findings agreed to a previous study which reported that neonates born via normal 264 vaginal deliveries were more likely to be better scored than their counterparts born through other 265 methods of delivery [22]. 266 Babies who underwent resuscitation in our study were 23.83 times more likely to be assigned 267 incorrect APGAR scores. We presuppose that, this finding could be attributed to the conventional 268 APGAR score does not take into account neonatal resuscitation. Additionally, given that the person 269 resuscitating is also the one who scores the neonate, there may have been an element of 270 ascertainment bias whereby, some scores could have been assigned incorrectly to evade 271 responsibility of negative neonatal outcomes. Our findings mirror those reported in Cameroon 272 where a significant variation of APGAR scores assigned to resuscitated versus non-resuscitated 273 babies was noted [27]; and those from the Netherlands, where the authors [28] noted a great 274 variability of the respiratory component of the APGAR score among healthcare providers in the 275 context of resuscitation, especially if a neonate is intubated. Furthermore, in Richmond-Virginia 276 [21], subjective APGAR scoring by healthcare providers in the context of neonatal resuscitation 277 was noted, making the authors recommend the use of other scores such as the Expanded APGAR 278 score or Neonatal Resuscitation and Adaptive score to replace the conventional APGAR score in 279 resuscitated babies. We also highlight this need for alternative scoring methods for resuscitated 280

babies. 281
Healthcare practitioners without access to an APGAR scoring chart were significantly more likely 282 to incorrectly classify asphyxia compared to healthcare practitioners with access to an APGAR 283 scoring chart. We postulate that, having access to an APGAR scoring chart could have resulted in 284 better utilization of the score in terms of more objective application of the score parameters, as 285 opposed to attempting to apply the score off by heart without the score chart. Our findings were 286 identical to those reported in Nigeria [29], where a similar low-resource setting like the current 287 study was used. Additionally, in Malawi, the authors [30] recommended a review of the guidelines 288 of the use of the APGAR score among healthcare providers. 289

290
This study reports low sensitivity and positive predictive values of assigned APGAR scores 291 at 71% and 62% respectively among newborns attending a tertiary hospital in Western 292 Kenya. Newborns who were resuscitated, born by instrumental deliveries, and assigned 293 APGAR scores by healthcare providers without access to the APGAR scoring chart were 294 significantly more likely to get incorrect APGAR scores. With the foregoing, there is need 295 to adopt the gold-standard on birth asphyxia diagnosis into the hospital-based neonatal 296 management guidelines at the tertiary hospital and other resource constrained healthcare 297 settings. In addition, improved APGAR scoring is attainable by augmenting availability of 298 APGAR scoring charts, continuous medical education programs on the score, and on 299 neonatal resuscitation. Additional studies should be conducted to explore the applicability 300 of other neonatal assessment scores for resuscitated neonates.