Adherence to the neonatal resuscitation algorithm for preterm infants in a tertiary hospital in Spain

Neonatal resuscitation (NR) is the most frequently performed resuscitation in hospitals [1‐3]. Infants that are more immature are more likely to require support. Approximately 85% of very preterm neonates need intervention during transition after birth and their viability and prognosis greatly depend on the care they receive in the delivery room (DR) [4‐6]. Most preterm infants initiate breathing after birth, but they often have a weak, insufficient respiratory drive. Guidelines recommend tactile stimulation (warming, drying and rubbing the back or soles of the feet) to stimulate breathing. Guidelines exist to standardize and optimize resuscitation. However, there is evidence that the sequence and quality of interventions during NR often deviate from guidelines [3, 7‐11]. Video recording has been widely used for educational and clinical quality assessment purposes, with good acceptance by caregivers [12, 13]. It is inexpensive, it does not interfere with resuscitation, and it offers data to assess performance accurately. Video reviewing reinforces teamwork and permits identification and amendment of errors that otherwise could be neglected. Combining the recording of physiological parameters (ECG, pulse oximetry (PO), capnography and respiratory function monitoring) with video images helps audit performance [12‐15]. There is a lack of information about adherence to NR guidelines in Spain. Consequently, we sought to evaluate adherence to NR guidelines in very preterm neonates at our hospital. Our main hypothesis was that resuscitation often deviates from the algorithm. A secondary hypothesis was that staff neonatologists perform better than pediatricians on-call because they work only with neonates and have more experience on average, whereas pediatricians are younger and work with children up to 18 years.

We conducted this observational study at Hospital Clínic de Barcelona, a tertiary referral center in Spain where approximately 150 babies < 32 weeks’ gestational age (GA) are born every year. Our Hospital Ethics Committee approved this project. We recorded and analyzed these infants’ resuscitations after obtaining written parental informed consent. We aimed to analyze as many NRs as possible. However, given the difficulties in obtaining parental consent in such moments of stress, we aimed to analyze a representative sample of at least one-third of all NRs performed. Thus, we decided to record 50 resuscitations. We had planned to obtain the data in 1 year, although it took us longer (16 months), as fewer candidates were born during the study period than expected. This study was the basis for the doctoral thesis of the main author (see link in http://​www.​ub.​edu/​medicina/​doctorat/​lectura.​htm May 27, 2011). However, the data were never published. The authors believe that the results and conclusions may be perfectly applicable today.

Inclusion criteria: All babies < 32 weeks GA were candidates for inclusion in this study. When the pediatrician was required in the delivery room, parents were approached for consent to record the NR. After obtaining written consent, the resuscitator began recording the NR, and that case was included in the study.

All infants were resuscitated under a radiant heater equipped with a neonatal automatic ventilator (Babylog 2, Dräger Medical, Drägerwerk AG & Co. KGaA, Lübeck, Germany) that included an oxygen blender and could provide Continuous Positive Airway Pressure (CPAP) and Positive Pressure Ventilation (PPV) and with a pulse-oximeter (Nellcor™ NPB-295, Minneapolis, MN, USA). A Sony Handycam DCR-SR 32 E (Sony, Tokyo, Japan) digital video camera attached to the upper left side of the radiant warmer recorded the newborn, the hands of the resuscitators and the PO screen. The clinical team turned the recording on before the baby was born.

We designed an evaluation sheet to score 12 domains in each resuscitation (Table 1) according to the algorithm of the Spanish Society of Neonatology, adapted from the ILCOR 2005 guidelines (see Fig. 1). We assigned a numerical score to every resuscitation, following Carbine’s previously described method [16]: we awarded 2 points for every correct decision and proper procedure, 1 point for delayed interventions or inadequate technique, and 0 points for indicated procedures that were omitted or for inappropriate procedures (for details of how we scored each domain, see Table 1). The total score per resuscitation (“resuscitation score”) ranged from 4 to 22 points. A higher score indicated that more intense resuscitation was required. We obtained the percentage of adherence to the algorithm by dividing the score achieved (X) by the maximum possible score per patient that is, X of a potential of (4–22) points, as a percentage. We registered admission temperature and Apgar scores at 1, 5 and 10 minutes (min) as assigned by the caregiver and after video recording review.

We compared two groups of resuscitators: staff neonatologists (group N) and pediatricians on-call (group P). Neonatologists on-call performed a few of the resuscitations after-hours.

Between April 2008 and August 2009, 162 infants < 32 weeks GA were born in our center. We analyzed 50 resuscitations (30.6%), a representative sample of the population. Groups N (staff neonatologists) and P (pediatricians on-call) were homogeneous. Tables 2 and 3 show the characteristics of our population and subgroups.

Global adherence to the algorithm was 80.9 ± 14.2%, with no differences between groups N and P (81.5 ± 12.7% in group N versus 80.7 ± 15.0% in group P, P = 0.93, Mann-Whitney U), and was independent of the number of interventions required. Eight resuscitations (16%) were technically correct; 15/50 (30%) failed in one domain; 12/50 (24%) in two; 5/50 (10%) in three; 6/50 (12%) in four; 2/50 (4%) in five; and 2/50 (4%) in seven. The mean (SD) resuscitation score was 13.5 (3.9) points/resuscitation (range: 6–20). Table 4 analyzes the adherence to the algorithm by domains.

Table 5 shows results from measures to prevent heat loss and its relation to admission temperature. We found no differences between the group that received correct measures to prevent hypothermia and the group that did not. Intubation differentiated intensive (16–20 total possible points) from mild resuscitation (6–16 points). Infants who did not need intubation (n = 36) had a mean global adherence to the algorithm of 83%. Deviations from the algorithm in this group did not correlate with the intensity of resuscitation (R² = 0.0013).

All our patients were resuscitated in our dedicated room for resuscitation, which provides a setting similar to that in the NICU. Vento [4, 6] suggested that incorporating an intensive care environment into the DR could enhance survival and reduce the morbidity of extremely low birth weight (BW) infants. Among our study population, we found a high percentage of hypothermia (73.5%), which led us to make some changes in our resuscitation room to reduce hypothermia: we increased the temperature to 26 °C by keeping the doors locked and installed a heater next to the resuscitation cot. We use heated, humidified gases for ventilation.

Several authors have proven that performance often deviates from guidelines. Our study is the first report on adherence to the neonatal resuscitation algorithm for very preterm infants in a tertiary care center in Spain.

Carbine was the first to use video recording to evaluate NR [16]. We based our study on his publication and adopted his scoring system. Carbine found deviations in 54% of NRs. We evaluated more aspects and may have detected more errors (84%). We believe that our resuscitation score was higher: 22% of Carbine’s patients only required stimulation (whereas 64% of ours needed stimulation); 80% required stimulation and oxygen, and only 7% needed PPV (vs our 80% respiratory support and 32% intubation). Carbine reported errors in the mask ventilation rate. We considered the use of a self-inflating bag an error, as we used an automatic ventilator. Consequently, we did not find this error. Only 28.6% of Carbine’s cases involving PPV had no deviations, which is worse than our 72.7% rate of proper ventilation. Among Carbine’s infants, 58.3% were intubated on the first attempt (vs our 33.3%), and only 33.3% (vs our 87.5%) were intubated within the established time limit. Like Carbine, we observed that perfect resuscitations were more likely for less intense interventions. None of our patients who required intubation received a perfect resuscitation.

Similar to Dekker [17], we observed that stimulation was often indicated but not performed, and when it was applied, it was most often indicated. Dekker’s infants who received no stimulation required intubation more often (18 vs 7%); in our case, 62.5% of intubated infants had not been stimulated and 25% of stimulated infants did not require intubation. However, 18.7% of infants (6/32) who were stimulated were also intubated afterward.

By using video recording at a Nepalese tertiary hospital, Lindbäck [10] identified deviation from guidelines in over 50% of resuscitations. Most errors concerned the use of bag-and-mask ventilation (which we did not evaluate, as ventilating with a bag and mask was an error in our study), suction and excessive use of oxygen. Their results seem more favorable than ours. However, Lindbäck did not focus on preterm infants.

Gelbart [8] reported that the demanding technical skills scored higher than the more basic steps of resuscitation because technique is taught, whereas clinical assessment, communication skills and teamwork need practice. He found that invasive ventilation and surfactant administration were best performed, with median scores of 100%, whereas the performance of preparation and initial steps (69%) and assessment and communication of heart rate (75%) was worse. In our patients, most errors took place during the initial steps as well, whereas administration of CPAP, PPV and surfactant were performed better. Surfactant was administered in 83.3% of our cases when indicated. Technique was always correct, although two patients who required it did not receive it; the PO was not functioning, and the pediatrician preferred to administer it in the NICU with proper monitoring.

Schilleman [9] used video recording to evaluate compliance with NR guidelines in a population similar to ours, although our patients were in better conditions according to the Apgar scores. Schilleman found that deviations mainly occurred within the first 30 s because caregivers needed more time to perform the initial steps and mainly involved the way ventilation was given. As such, Schilleman suggested that 1 min be allowed for the initial evaluation, which is what the current ILCOR guidelines allow.

No intubations were perfectly performed. We analyzed 40 intubation attempts. Success occurred on the first attempt in 33.3% of cases and on the second attempt in 38.9%; more than three attempts (5 and 6) were required in only 11.1% of cases. In all intubation cases, at least one attempt took longer than recommended. Only 12.5% of intubations took place within the allotted 30 s. Other authors have reported similar deficiencies. Lane [18] reported a mean duration for successful attempts of 27.3 s; 30% infants were intubated on the first attempt, 30% on the second, 20% on the third, and 20% required more than three. Success was higher for 30 s, and no infants decompensated between 20 and 30 s; 20% of successful attempts took longer than 40 s. Finer and Rich’s [3, 15] overall success rate for intubation was 33% within the allotted 20 s and 56% within 30 s. They reported an average of at least three attempts to successfully intubate infants < 1000 g. Our intubation success rate was higher than that reported by Finer and Rich. We needed, on average, 2.2 attempts to intubate infants < 1000 g (range 1–5), but unfortunately, it took longer (median (SD) 58.1 s (23.4), range 17–128 s). O’Donnell [19] analyzed intubation attempts in 31 infants (mean GA 28 weeks and BW 1227 g). Intubation attempts were often unsuccessful and successful attempts often took more than 30 s (17% were successful within 20 s, 20% between 20 and 29 s, and 25% > 30 s). Konstantelos [7] needed a median of 2 attempts, and 47 (25–60) s for intubation, and only 11% were successful within the allotted time. Wozniak [20] analyzed intubation attempts in preterm infants (795 g median BW, 25 weeks’ GA) and reported a mean duration of 35 s and 2 attempts. Like Konstantelos [21], we believe that the lack of medication for intubation and surfactant administration are the reasons for the longer time needed for intubation. Because health care providers often underestimate the passage of time during NR, it is difficult to realize when the allotted time has passed. The American Academy of Pediatrics NRP used to allow 20 s for intubation, but since several studies reported that it often took longer [3, 16, 18, 19, 21] and that infants did not decompensate between 20 and 30 s, the current limit is 30 s [20].

As Fig. 2 shows, resuscitators obtained, on average, 10 points per patient regardless of the intensity of resuscitation, which means that caregivers did not score more points in more complex resuscitations. There are two reasons for the constant red line: a) some initial, common mistakes in heat loss prevention, suctioning and a postductal PO probe placement prevented most mild NRs (mean resuscitation score of 13 points) from scoring higher, and b) the points corresponding to almost all complex domains (intubation, chest compressions or epinephrine administration) were lost. The consequence is the flat line in the relationship between mistakes and the obtained score. However, the more mistakes that occur in a NR, the higher the resuscitation score (blue line, R² = 0.895), meaning that complex domains (the area above the red line) are lost in terms of obtained points. Not surprisingly, the relationship between the resuscitation score and the obtained score (Fig. 3) followed a strong logarithmic pattern (R² = 0.705). This finding means that although more points are possible, the resuscitator team would gain little benefit from those complex domains. More emphasis must be placed on the initial steps, which are common to most NRs, and especially on training for complex skills such as intubation.

This study has some limitations. Its purpose was to evaluate adherence to NR guidelines. For this reason, all mistakes counted equally, although it is obvious that not all the deviations are equally serious: some are mild and nontranscendental (for example, duration of oral suctioning) whereas others are potentially harmful (like timing and route of epinephrine administration). The same mistake could have consequences or not, depending on the patient. For example, placing a postductal PO probe in a patient who does not receive oxygen or ventilatory support has no impact on the maneuvers performed but may lead to hyperoxia in an intubated neonate. Only 16% of our resuscitations perfectly followed the algorithm, but the number of mistakes per resuscitation was low, and global adherence to the algorithm was 80.9%. We acknowledge that the value of global adherence in itself has little meaning without the proper analysis of the main and more critical errors. Another limitation of our study is the lack of feedback of our findings to the resuscitation team. We designed the study to assess adherence to the algorithm and find the most common errors. In the pilot study, we did not consider an active intervention with the resuscitation team. Sharing our findings with them would probably improve performance. In most cases, the resuscitation team consisted of two neonatologists or two pediatricians on-call. All of them are trained in neonatal resuscitation, although their expertise varies from more than 30 years to only a few months after completing a residency. However, 38% of our patients were twins, which may worsen performance, as in some cases, there was only one caregiver per patient [22]. All our medical staff was aware of this study when it started, and we periodically reminded them about it. All the neonatologists and pediatricians who work at our hospital participated in this study. The medical staff turned the video camera on, automatically consenting to be recorded, when they were called to the DR. Recording usually began minutes before the neonate was born but sometimes began when the newborn arrived at the resuscitation room.

Similar to many other previous studies, our study demonstrated that deviations from the algorithm exist. Many of the errors have to do with timing: some maneuvers take longer than allotted, and personnel are not aware of this [23].

We compared performance of staff neonatologists with pediatricians on-call. We thought that the neonatologists would perform better since they work with only newborns and are subspecialized in neonatology, whereas after-hour on-call pediatricians who cover this shift often do not work with only newborns. We observed no differences between these two groups, which means that we have a good team of pediatricians who perform as well as neonatologists. This is a positive aspect to consider. Although global adherence to the algorithm was high, mistakes were common despite our staff’s training.

In line with other authors [8, 22‐25], we found a discrepancy in Apgar scores, particularly when the Apgar was not 9/10/10, and the staff attending the delivery commonly overestimated the score. It is easy to score 9/10/10 if no resuscitation is necessary, whereas it is difficult to remember the patient’s situation at 5 and 10 min when resuscitation is required. Video recording scores tended to be lower than scores given by neonatologists (47.1% at 1 min, 73.3% at 5, 88.9% at 10). Gelbart [8] also found overestimation of Apgar scores by a median value of 2 points at 1 and 5 min. As other authors suggest, we believe that memories of a stressful past event can be inaccurate, and Apgar scores are usually calculated afterward [24].

Finally, we are aware that our study sample was small. We aimed to analyze 50 resuscitations due to the difficulty in obtaining written consent before resuscitation started. Nonetheless, we managed to record one-third of our potential cases, which is a significant sample. The ILCOR guidelines have changed twice since we conducted this study, and some actions that we considered mistakes would now be correct, for example, supporting transition rather than keeping timing strict or not suctioning routinely. While it is true that our data are old, and a few aspects are outdated, our aim was to assess our performance in terms of adherence to the algorithm, that is, if our physicians performed according to the written rules, not the appropriateness of the algorithm. Our results would probably be similar today. Even though only one person reviewed the recordings, the camera was in a good position, and the scoring system was clear, so this bias is likely minimal. There was no feedback given to the resuscitation team during the study period, but our findings could serve as both a starting point for further studies and a teaching tool. As far as we know, there are no similar studies published in Spain to date.

Resuscitation of very preterm newborns often deviates from guidelines. Perfectly performed resuscitations are infrequent, although global adherence to the algorithm is high. Resuscitations led by pediatricians on-call and neonatologists are performed equally correct. Intubation training may improve complex resuscitations the most.