Implementing and evaluating a Paediatric Early Warning System (PEWS) at a National Referral Hospital in Botswana

Abstract

Background

Mortality among hospitalised children in low-resource settings remains much higher than in high-resource environments. Paediatric Early Warning Systems (PEWSs) have been shown to improve vital signs collection, strengthen interprofessional communication, lower healthcare costs and reduce paediatric hospital mortality in multiple low- and middle-income countries. Providers at Botswana’s national referral center, Princess Marina Hospital (PMH), face significant challenges in identifying children at risk for clinical deterioration.

Methods

We used PEWS previously validated in resource-limited settings to create the PMH PEWS. We piloted the PMH PEWS from December 2022 to March 2023. We assessed (1) effectiveness of PEWS at reducing unplanned escalations of care by comparing pre-implementation and post implementation rates of clinical deterioration events (unplanned intensive care unit (ICU) transfer; use of inotropic medications, mechanical ventilation or mannitol; cardiopulmonary resuscitation; non-palliative mortality), (2) acceptability and feasibility of PEWS adoption using the acceptability and feasibility of implementation measures and (3) barriers and facilitators to implementation through stakeholder interviews structured around Consolidated Framework for Implementation Research domains.

Results

The relative frequency of clinical deterioration events changed post-PEWS implementation (p=0.01) such that initiation of mechanical ventilation (12.3% vs 23.2%) and inotropes (18.5% vs 35.7%) decreased while ICU transfers increased (27.7% vs 8.9%). Both doctors and nurses found PEWS to be acceptable and feasible. Staff universally reported that PEWS improved patient care, increased clinician provider accountability for deteriorating patients and strengthened interprofessional communication. Nurses reported greater engagement with PEWS adaptation than doctors. Physical resource limitations and inter- and intra-professional hierarchies were widely endorsed barriers to implementation.

Conclusions

PEWS has the potential to improve the care of hospitalised children in Botswana by strengthening interprofessional communication and increasing clinician accountability for deteriorating patients.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Paediatric early warning systems (PEWSs) are important tools for identifying hospitalised children at risk of clinical deterioration in low-resource settings.

WHAT THIS STUDY ADDS

• We report on the implementation of PEWS in a novel setting, Botswana’s national referral and teaching hospital. We rigorously applied implementation research methods in evaluating barriers and facilitators to, as well as the acceptability and feasibility of, PEWS implementation in our context.

HOW THIS STUDY MAY IMPACT PRACTICE OR POLICY

• This successful pilot serves as a template for the expansion of PEWS to other inpatient facilities in Botswana and other, similar environments. We comment on health system practice changes necessary to support successful PEWS implementation.

Background

Mortality rates for hospitalised children in low- and middle-income countries (LMICs) are significantly higher than in high-resource environments.^{1 2} Identification of hospitalised children at risk for clinical deterioration is a major challenge for health systems in LMICs. Providers face high patient-to-staff ratios, insufficient continuous monitoring technologies and very limited access to intensive care resources.³

Paediatric Early Warning Systems (PEWSs) are a critical tool for reducing intensive care unit (ICU) admissions and hospital mortality in LMICs. PEWSs use vital signs and nursing assessment to score children based on their risk for deterioration; an associated response algorithm then guides escalation of care based on a child’s score.⁴ Multiple versions of PEWSs have been implemented in hospitals in Latin America and sub-Saharan Africa, where their use is associated with improved vital signs collection practices, nursing empowerment^{5 6} and interprofessional communication,^{4 7} as well as lower rates of ICU admission⁸ and hospital mortality.^{9 10}

Here, we present findings from a pilot PEWS initiative at a national referral hospital in the southern African nation of Botswana. PEWS was adopted at Princess Marina Hospital (PMH) with support of Project PASHA (PEWS Adaptation to Support Hospitals in the Alliance), a collaborative to scale up PEWS use globally, after a quality improvement project demonstrated that nurses working in the hospital’s paediatric units only correctly identified vital signs that were abnormal for age in 22% of cases, and only documented notifying a doctor in 9%.¹¹ PMH is the first hospital in Botswana to use PEWS.

Our convergent, parallel mixed methods implementation study aimed to: (1) evaluate the impact of implementing a PEWS at PMH on the rate of clinical deterioration events (CDEs); (2) assess the feasibility and acceptability of PEWS among healthcare providers and (3) identify facilitators and barriers to successful PEWS implementation in a novel environment.

Methods

Setting

PMH is Botswana’s national referral and teaching hospital, located in the capital Gaborone. The Paediatric Medical Ward (PMW) has a 40-bed capacity. One bay is designated for ‘high dependency’ patients and has two electronic monitors. Nurse-to-patient ratios are 1:6 during the day and 1:12 at night; doctor-to-patient ratios are 1:10 during the day and 1:40 at night. Vital signs are collected every 8 hours on stable patients and every 6 hours on ‘high dependence’ patients. Overnight, one paediatric resident is present in each unit and one paediatric consultant is on call for each ward from home. All medical records and patient registers are hand-written.

There is no paediatric intensive care unit (PICU) at PMH. Patients >5 kg who require ICU care are admitted to the 8-bed adult ICU when space is available and transferred to a private hospital when space is unavailable. Critically ill infants <5 kg admitted from home are admitted to the 6-bed PMH neonatal intensive care unit (NICU) until it reaches capacity and then to a private hospital. The cost of private ICU admission under these circumstances is covered by the government for citizen children. Due to limited capacity, children with an irreversible, pervasive neurodevelopmental disorder or a terminal illness were not transferred to any ICU or NICU at the time of the study. The cost of private ICU care is not covered for non-citizen children. Cardiopulmonary resuscitation (CPR) is typically not performed on children deemed ICU-ineligible. Ward policy is to counsel families of children who are either deemed ICU-ineligible and for whom the cost of private ICU care is not covered if the child begins to deteriorate.

PMH was invited to participate in a PEWS mentorship programme designed and organised by St. Jude Children’s Research Hospital called PASHA¹² adapted from an effective PEWS implementation strategy from hospitals in Latin America.^{12 13} The PEWS implementation team at PMH received training and feedback on their PEWS materials from PASHA before and during implementation.

Measurement of CDEs

The innovation effectiveness¹⁴ of PEWS in the PMW at PMH was measured by comparing CDE frequencies and rates before and after PEWS implementation. Using definitions provided by PASHA,¹⁵ a CDE was classified as any one of the following occurrences in the ward: (1) unplanned transfer to an ICU; (2) use of inotropic medications; (3) use of non-invasive or invasive mechanical ventilation; (4) CPR; (5) use of mannitol (added by our implementation team to the PASHA definitions) or (6) non-palliative mortality. We classified any death that was unanticipated prior to admission as non-palliative, including those among children who were deemed ICU ineligible and non-citizen children whose families could not afford private ICU care. If multiple eligible events occurred during a single resuscitation, the first documented event determined the CDE type. No event that occurred in the emergency room or in an ICU was counted as a CDE. A CDE was declared over when the patient: (1) remained in the ward without requiring a CDE-eligible medication or device for ≥24 hours; (2) died within 24 hours of the start of a CDE; (3) was discharged from an ICU within 24 hours of the start of a CDE; or (4) died in an ICU following a CDE.

To identify CDEs, study staff attended departmental handover every weekday morning, at which all admissions and mortalities from the previous night are presented. Study staff then contacted providers on call the previous night to ask about events involving previously admitted patients and to clarify event details. Study staff pulled all relevant patient charts in the ward to confirm event details and entered each CDE directly into a REDCap database using a tablet. Study staff followed patients who experienced CDEs through hospital discharge. For patients transferred to a private facility, an update was obtained from the consultant on service at the time of transfer when available.

Patient volume was measured in number of PMW admissions per month, obtained from the ward register at the end of each month. CDE rate was calculated as number of events per 100 admissions. Event mortality was defined as any one of the following: (1) death on the ward <24 hours after CDE end; (2) death in an ICU after transfer during a CDE; (3) death at home <24 hours after CDE end or (4) death in a non-ICU outside facility <24 hours after CDE end. All analyses were conducted in Stata BE V.17.

Pre-implementation activities

The PASHA team mentored PEWS implementation at PMH based on previously described methodology.¹² Figure 1 depicts the project timeline and key interventions undertaken by the implementation leadership team.

Figure 1. Project timeline and key interventions. PASHA, PEWS Adaptation to Support Hospitals in the Alliance; PEWS, Paediatric Early Warning System; PMH, Princess Marina Hospital; PMW, Paediatric Medical Ward.

Adaptation of an existing PEWS

The PMH PEWS was adapted from the validated Escala de Valoración de Alerta Temprana (EVAT),¹⁶ an English translation of which was provided by the PASHA team.¹² A few small revisions were made based on resource availability (eg, exchanging FiO2 for L/min). No changes to the published EVAT scoring rubric were required. The escalation algorithm was adapted based on resource availability by author MRH, PMH’s paediatric intensivist and PMH’s paediatric subspecialists, and then presented to the University of Botswana Department of Pediatrics for feedback. The final scoring rubric and algorithm can be found in online supplemental appendix 1.

Adaptation of nursing documentation

Author MRH developed three prototypes for nursing documentation that incorporated PEWS scoring based on models provided by PASHA¹² and others.⁵ Nursing leaders identified one prototype as being the simplest to use and the least resource-intensive, while fulfilling existing documentation requirements. After minor revisions, we introduced a draft of this form to the ward prior to formal PEWS training, where bedside nurses used it for a week and then provided feedback on their experience via a brief, anonymous survey. We iteratively refined the form three times and the response algorithm once based on nursing feedback during core implementation. The final versions can be found in online supplemental appendix 2.

Pre-implementation training and pilot

We held nurse training on 4 days in November 2022. Nursing leadership helped the implementation team devise the training schedule to ensure that every nurse was trained, either by the implementation team or on-the-job by a peer. Nurse training sessions were led by authors STM and WBT, both University of Botswana School of Nursing lecturers, in a conference room adjacent to PMW. Physicians were trained on-the-job by author MRH who received training on PEWS from the PASHA programme. PMW piloted the PMH PEWS from December 2022 to March 2023 (12 weeks). PEWS was calculated with every vitals check. At the start of the pilot, study staff reviewed all patient charts twice weekly on randomly selected days to calculate error frequencies and identify providers who required additional support with PEWS. Nurses with complete and accurate PEWS documentation were acknowledged at morning huddles and received a small performance incentive. Successful implementation was defined as <15% errors in completion, accuracy and documented algorithm adherence across all nursing staff, per PASHA recommendations^{12 13}; this target was achieved in February 2023 and sustained through May 2023.

Post implementation activities

Survey of acceptability and feasibility

Authors MRH and CZW-H designed a survey based on Weiner et al’s Acceptability of Implementation (AIM) and Feasibility of Implementation (FIM) instruments, which can be found in online supplemental appendix 3.¹⁷ Acceptability is users’ (in this case PMW staff’s) perception that an innovation (ie, PEWS) is agreeable or satisfactory, while feasibility is the users’ perception of the extent to which the innovation can be successfully used in the study context (PMW).¹⁷ The survey was reviewed by the remaining authors for clarity and by a non-author American implementation researcher. All nurses, residents, medical officers and paediatric consultants working in PMW during implementation were invited to take the survey, administered via REDCap on a tablet. Survey participants provided written consent prior to the survey.

Semi-structured interviews and qualitative analysis

Author MRH developed an interview guide structured around the Consolidated Framework for Implementation Research (CFIR) 2.0, a widely used model that explores a variety of factors that influence implementation success across multiple domains including inner setting (eg, available resources), outer setting (eg, policies and laws), individuals (eg, implementation facilitators), processes (eg, assessing needs), and the characteristics of the ‘innovation’ (eg, relative advantage, adaptability).^{18 19} CFIR 2.0 was selected for this pilot study as no prior data on PEWS or implementation determinants were available for our context. We also explored the use of the Exploration Preparation Implementation Sustainment (EPIS) framework²⁰ to inform future work on PEWS in Botswana (see online supplemental appendix 5). Three clinicians who were not affiliated with the study and either currently practise or previously practised in Botswana reviewed and provided feedback on the guide. All nurses, paediatric residents and medical officers (medical school graduates who have not completed a residency) who had been working on the ward for ≥6 months prior to PEWS introduction were invited to participate in a recorded interview with author MRH between May and August 2023. All interviews were conducted in person and in English, the official language of medical practice in Botswana. Interviews lasted 15 min on average (range 10–35). Participants provided verbal consent prior to their interview.

Interviews were transcribed by author KG and transcripts were reviewed by author MRH. Authors MRH, STM, SG and WBT used an integrated approach to coding and analysis of the interviews.²¹ We developed and applied deductively derived codes based on the CFIR 2.0 model as well as inductively derived codes that emerged from a close reading of the interviews. The study team met regularly to refine codes and establish a stable codebook with areas of disagreement resolved through group discussion until coders reached >80% agreement. After coding, each code was summarised and examined for patterns. Interview participants were recruited until we reached saturation of themes.

Final coded themes were mapped back to relevant CFIR constructs to ensure alignment with the model. Figure 2. depicts the sub-set of CFIR 2.0 constructs we selected for inclusion. We then compared frequency of themes between genders and across clinician roles (eg, nurse, resident physician). Chi-squared and Fisher’s exact tests assessed differences at the alpha=0.05 level using Stata.

Figure 2. CFIR 2.0 domains and constructs: constructs included in the final qualitative analysis are highlighted in red.^{1 2} CFIR, Consolidated Framework for Implementation Research.

Patient and public involvement statement

Our findings were formally disseminated by author KG to the Botswana-UPenn Partnership Community Advisory Board, composed of members of the public, at the conclusion of the study.

Results

Clinical outcomes

151 CDEs during 2129 admissions were captured between July 2022 and July 2023, 56 before PEWS implementation and 65 post implementation, resulting in 4% power to detect a 10% difference in CDE frequency at α=0.050.²² figure 3 depicts the CDE rate, event-related mortality and adherence by staff to the PEWS algorithm over the duration of the study. Data on CDEs captured during the 3-month pilot period were excluded from final analysis but are available in online supplemental appendix 4. Table 1 contains CDE characteristics before and after PEWS implementation. The event rate per 100 admissions did not change following PEWS implementation (6.8 vs 7.4, p=0.6). The relative frequency of unplanned ICU transfer CDEs increased following implementation (8.9% vs 27.7%, p=0.013), even though the absolute frequency of ICU transfers did not change (39.3% vs 43.1%, p=0.34) and a higher percentage of patients who experienced CDEs were deemed ineligible for ICU transfer after implementation (35.7% vs 55.6%, p=0.002). The relative frequencies of inotrope initiation (18.5% vs 35.7%) and mechanical ventilation initiation (12.3% vs 23.3%) also decreased (p=0.013). There was no change in CDE-attributed mortality following PEWS implementation (40.7% vs 50.0%, p=0.33, with 18% power to calculate a 10% difference in mortality at α=0.05.²² There were no changes in the characteristics of patients experiencing CDEs post implementation (table 2).

Figure 3. Clinical deterioration event frequency, event-related mortality and adherence to PEWS algorithm. PEWS, Paediatric Early Warning System.

Table 1. Characteristics of clinical deterioration events (CDEs) before, during and after PMH-PEWS implementation. Reported as No. (%) except where otherwise indicated.

Event characteristics, No. (%)			All CDEs (n=151) 2129 admissions	Pre-PEWS CDEs (n=56) 827 admissions	Post implementation CDEs (n=65) 873 admissions	P value, pre-implementation PEWS vs post implementation
Event rate per 100 admission, n=151			7.1	6.8	7.4	0.60
	Initiation of invasive or non-invasive ventilation	Event rate per 100 admissions, n=151	1.4	1.6	1	0.23
		No. (%), n=151	30 (19.9)	13 (23.2)	8 (12.3)	0.013
	Initiation of vasoactive medications	Event rate per 100 admissions, n=151	1.7	2.4	1	0.12
		No. (%), n=151	37 (24.5)	20 (35.7)	12 (18.5)	(see above)
	CPR	Event rate per 100 admissions	2.1	2.1	3	0.53
		No. (%), n=151	45 (29.8)	17 (30.4)	22 (33.9)	(see above)
	Initiation of mannitol	Event rate per 100 admissions	0.3	0.12	0.50	0.24
		No. (%), n=151	7 (4.6)	1 (1.8)	4 (6.2)	(see above)
	Unanticipated ICU transfer	Event rate per 100 admissions	1.4	0.6	2.1	0.01
		No. (%), n=151	30 (19.9)	5 (8.9)	18 (27.7)	(see above)
	Non-palliative mortality	Event rate per 100 admissions	0.1	0	0.1	0.51
		No. (%), n=151	2 (1.3)	0 (0)	1 (1.5)	(see above)
Nursing shift	Morning		47 (31.1)	13 (23.2)	22 (33.9)	0.43
	Afternoon		51 (33.8)	22 (39.3)	23 (35.4)
	Night		53 (35.1)	21 (37.5)	20 (30.8)
Patient was transferred to an ICU			61 (40.4)	22 (39.3)	28 (43.1)	0.34
Patient was deemed ineligible for ICU transfer (n=72)			32 (44.4)	10 (35.7)* (n=28)	15 (55.6)* (n=27)	0.002
Median CDE length in hours (IQR) (n=143)			33.9 (1.7–119.3)	60.3 (7.3–173.3) (n=54)	32.3 (0.8–119.3)* (n=59)	0.10
Median maximum PEWS score in 24 hours preceding the CDE (IQR) (n=60)			5.2 (2.3–10.3)	–	6.5 (2.5–10.4)	–
Patient died following the CDE (n=142)			69 (48.6)	22 (40.7)* (n=54)	29 (50)* (n=58)	0.33

^*Number of CDEs for which these data were available differs from the total n used for all other calculations and is thus provided. Some data are missing for patients transferred to a private facility, as addressed in the Discussion section.

CPR, cardiopulmonary resuscitation; ICU, intensive care unit; PEWS, Pediatric Early Warning System; PMH, Princess Marina Hospital.

Table 2. Characteristics of patients who experienced clinical deterioration events (CDEs) before, during and after PMH-PEWS implementation.

Patient characteristics, No. (%)		All CDEs (n=151)	Pre-PEWS CDEs (n=56)	Post implementation CDEs (n=65)	P value, Pre-implementation PEWS vs post implementation
Median age in years (IQR)		0.5 (0.1–2)	0.3 (0.1–1.0)	0.5 (0.2–1.8)	0.16
Female sex		82 (54.3)	30 (53.6)	33 (50.8)	0.86
Nationality	Motswana	130 (86.1)	51 (91.1)	58 (89.2)	0.65
	Zimbabwean	20 (13.3)	5 (8.9)	6 (9.2)
	Undocumented	1 (0.7)	0	1 (1.5)
HIV status	Unexposed	101 (66.9)	38 (67.9)	43 (66.2)	0.51
	Exposed, PCR-negative	18 (11.9)	6 (10.7)	9 (13.8)
	Exposed, status unknown	8 (5.3)	5 (8.9)	2 (3.1)
	Positive	2 (1.3)	0	1 (1.5)
	Undocumented	22 (14.6)	7 (12.5)	10 (15.4)
History of prematurity (n=120)		26 (21.7)	12 (22.6)	8 (17.0)	0.48
Malnourished on admission (n=146)		37 (25.3)	14 (25.9)	15 (23.8)	0.79
Preexisting conditions	Congenital heart disease	25 (16.6)	9 (16.1)	12 (18.5)	0.80
	Epilepsy	13 (8.6)	1 (1.8)	2 (3.1)	0.65
	Cerebral palsy	4 (2.6)	0	2 (3.1)	0.19
	Neurosurgical condition	3 (2.0)	1 (1.8)	2 (3.1)	0.65
	Chronic renal disease	4 (2.6)	0	3 (4.6)	0.10
	Chronic hepatic disease	2 (1.3)	1 (1.8)	1 (1.5)	0.92
	Diabetes	1 (0.7)	0	1 (1.5)	0.35
	Cancer	9 (6.0)	3 (5.4)	2 (3.1)	0.55
	Non-oncologic haematologic disease	2 (1.3)	1 (1.8)	0	0.28
	Congenital syndrome	3 (2.0)	1 (1.8)	2 (3.1)	0.65
	Other	8 (5.3)	2 (3.6)	4 (6.2)	0.51
Admission indication	Acute respiratory illness	57 (37.8)	17 (30.4)	28 (43.1)	0.75
	Dehydration/shock	40 (26.5)	18 (32.1)	16 (24.6)
	Cardiac problem	15 (9.9)	7 (12.5)	6 (9.2)
	Seizure	7 (4.6)	3 (5.4)	3 (4.6)
	Sub-specialist review	5 (3.3)	2 (3.6)	2 (3.1)
	Other	27 (17.9)	9 (16.1)	10 (15.4)
Median ICU LOS in days (IQR) (n=143)		7 (3–10)	8 (5–10.5)	7 (3–12)	0.83
Median LOS in days (IQR) (n=125)		9 (2–14)	9.5 (2–18.5)	10 (3–14)	0.71

ICU, intensive care unit; LOS, Length of Service; PEWS, Pediatric Early Warning System; PMH, Princess Marina Hospital.

Acceptability and feasibility

We received survey responses from 42 of 55 (76.4%) eligible staff. Of those who did not respond, the majority were consultants and sub-specialists. The survey completion rate was 88.1% (37/42); all of those who did not complete the survey were nurses. Both nurses and doctors of all professional levels working in PMW found PEWS both acceptable and feasible (tables35).

Table 3. Participant characteristics (n=43).

Characteristics		n (%)
Respondent type	Nurse	26 (61.9)
	Doctor	16 (38.1)
Nursing role	Diploma holder	17 (65.4)
	Degree holder	7 (26.9)
	Nurse manager	1 (3.9)
	Matron	1 (3.9)
Doctor role	Medical officer-general paediatrics	3 (18.8)
	Medical officer-haematology/oncology	2 (12.5)
	Resident	10 (62.5)
	Paediatrician	1 (6.3)
Median year of experience in the Paediatric Medical Ward (IQR)		2.5 (1–5)
Years of experience	<5	28 (66.7)
	5–9	9 (21.4)
	≥10	5 (11.9)

Table 5. Feasibility of implementation measure (FIM) presented as median score (IQR).

FIM statement	All respondents, n=37	Nurses, n=21	Doctors, n=16	P value*	<5 years experience in PMW, n=25	≥10 years experience in PMW, n=3†	P value*
PEWS seems realistic	5 (4–5)	5 (4–5)	5 (4–5)	0.39	4 (4–5)	5 (5–5)	0.11
PEWS seems doable	5 (5–4)	5 (4–5)	5 (4–5)	0.94	4 (4–5)	5 (5–5)	0.10
PEWS seems challenging	2 (2–3)	2 (2–4)	2 (2–3)	0.76	2 (2–4)	1 (1–2)	0.03
PEWS seems easy to use	4 (4–5)	4 (4–5)	5 (4–5)	0.41	4 (4–5)	5 (5–5)	0.07

1=Complete disagree; 2=Disagree; 3=Neither agree nor disagree; 4=Agree; 5=Completely agree.

^*Groups compared using Wilcoxon ranked sums.

^†Nurses and one doctor who completed the survey had worked in the unit for ≥10 years.

PEWS, Pediatric Early Warning System; PMW, Paediatric Medical Ward.

Table 4. Acceptability of implementation measure (AIM), presented as median score (IQR) (n=37).

AIM statement	All respondents, n=37	Nurses, n=21	Doctors, n=16	P value*	<5 years experience in PMW, n=25	≥10 years experience in PMW, n=3†	P value*
PEWS meets my approval	5 (4–5)	5 (4–5)	5 (4–5)	0.33	5 (4–5)	5 (5–5)	0.19
I have no objection to PEWS	5 (4–5)	4 (4–5)	5 (4–5)	0.53	4 (4–5)	5 (5–5)	0.08
PEWS is pretty good	4 (4–5)	5 (4–5)	4 (4–5)	0.85	4 (4–5)	5 (5–5)	0.03
I like PEWS	4 (4–5)	5 (4–5)	4 (4–5)	0.52	4 (4–5)	5 (4–5)	0.21
I welcome the use of PEWS	5 (4–5)	5 (4–5)	5 (4–5)	0.88	5 (4–5)	5 (5–5)	0.15

1=Complete disagree; 2=Disagree; 3=Neither agree nor disagree; 4=Agree; 5=Completely agree.

^*Groups compared using Wilcoxon ranked sums.

^†Nurses and one doctor who completed the survey had worked in the unit for ≥10 years.

PEWS, Pediatric Early Warning System; PMW, Paediatric Medical Ward.

Thematic analysis

We reached saturation after 20 interviews. Table 6 displays key participant demographics.

Table 6. Characteristics of interviewed providers (n=20).

Characteristic		n (%)
Gender	Man	4 (20)
	Woman	16 (80)
Provider type	Nurse	9 (45)
	Doctor	11 (55)
Provider role	Bedside nurse	7 (35)
	Nursing leader	2 (10)
	Paediatrics resident	8 (40)
	Medical officer	3 (15)
Median years in clinical practice (IQR)	5 (4–8)
Median years in practice at PMH (IQR)	4 (3–5)

PMH, Princess Marina Hospital.

We identified 34 themes aligned with 15 CFIR constructs. Online supplemental table 3b displays the mapping of themes to CFIR constructs and illustrative quotations. Online supplemental appendix 5 depicts an exploratory mapping of these same themes to the EPIS framework.²⁰ Below we organize key findings into implementation facilitators and barriers.

Facilitators of PEWS implementation

Improved patient care and detection

PEWS enhanced identification of children at risk for deterioration earlier, leading to better patient care outcomes. Participants universally expressed that they believe PEWS improves patient care. Many participants specified that PEWS enabled the more rapid identification of patients at risk of, or already experiencing, clinical deterioration who they believe would not have been identified as quickly prior to PEWS introduction.

Accessibility of resources

The use of posters and lanyards increased the usability of PEWS by making vital sign references and scoring algorithms readily available. Many participants alluded to their frequent use of the PEWS vital sign references and scoring algorithms made available to them in the form of ID badge cards and large-format posters affixed to provider worktables, and described these materials as important teaching tools.

Increased accountability

The introduction of PEWS heightened the sense of self-accountability and team accountability among healthcare providers, prompting timelier and coordinated responses to patient deterioration. Many providers described feeling greater pressure to react to abnormal vital signs because PEWS obligates nurses to report elevated scores to doctors and document that notification process. Doctors also reported feeling increased accountability for deteriorating patients because nurses notify them of elevated PEWS scores.

Provider motivation and engagement

Providers expressed a strong motivation to continue using PEWS, indicating sustained engagement and perceived value of the tool. Participants universally described themselves as ready to continue using PEWS.

Opportunities for feedback and adaptation

Nurses had frequent opportunities to offer feedback on PEWS, and their input was visibly incorporated into the tool’s iterative improvements. Nurses described being asked regularly by PMW nursing leaders for feedback on PEWS both during morning huddles and via their unit’s WhatsApp group. Multiple nurses expressed that they could see their input reflected in serial modifications to PEWS.

Barriers to PEWS implementation

Initial complexity and resistance

The complexity of PEWS initially created resistance among nurses, although this was eventually overcome with practice and persistent use. Multiple nurses described PEWS as being challenging to adopt, but by the time of the interview, these same nurses now felt that PEWS was easy to use. When asked what they believe helped them overcome the challenge of PEWS, these nurses identified time and persistent use as the key factors.

Hierarchical communication challenges

Persistent hierarchical dynamics between nurses and doctors, as well as between junior and senior doctors, could hinder effective communication and implementation of PEWS recommendations. Nurses reported that, especially during down-staffed night and weekend shifts, doctors were not always receptive to their reporting of elevated PEWS scores. Doctors reported that high PEWS scores did not often facilitate more regular communication with their supervising consultant. Some doctors attributed this phenomenon to feeling uncomfortable questioning the judgement of a more senior doctor who does not believe the patient requires an escalation in care, even when PEWS scores are elevated.

Dependence on external ICU resources

The necessity of transferring patients to private ICU facilities due to limited internal ICU resources often complicated the escalation process recommended by PEWS. PEWS recommends ICU consultation at scores of 5 and above but multiple participants identified the lack of in-house ICU facilities as a barrier to responding appropriately to elevated PEWS scores.

Staffing shortages

Short staffing, particularly during nights and weekends, disrupted the consistent use of PEWS and delayed appropriate patient care. Participants identified nights and weekends, when fewer staff are on shift, and the frequent re-allocation of staff throughout the health system, as disruptive to implementation.

Inadequate equipment

Broken or insufficient technical equipment, such as monitors and respiratory support devices, impeded accurate PEWS scoring and escalation efforts. In particular, participants named ageing or broken vital signs collection equipment, electronic patient monitors and absent or broken respiratory support equipment as barriers to accurately scoring PEWS and escalating care for children with elevated PEWS scores.

Lack of doctor feedback opportunities

Doctors denied being given opportunities to provide feedback on PEWS. When asked about specific venues at which the implementation leadership team had solicited feedback from doctors, participants reported that they had not been aware of these opportunities and had not been in attendance.

Low doctor engagement with the PEWS algorithm

Most doctors interviewed reported non-adherence to the PEWS escalation algorithm when reviewing a patient with an elevated PEWS score. Some felt that their own clinical judgement was sufficient. Others did not remember to refer to the algorithm when reviewing patients.

Stationery and printing issues

Participants highlighted frequent stock outs of PEWS forms due to printer malfunctions, stock outs of ink or toner, and the absence of a unit clerk assigned to keep clinical stationery in stock as barriers to using PEWS.

Differences in interview participant responses by participant characteristics

Doctors were more likely than nurses to report intra-professional barriers to communication around PEWS (19.2% vs 12.3%, p=0.01). Men were more likely than women to describe themselves as having felt engaged with PEWS instrument adaptation (29.8% vs 14.2, p=0.002).

Discussion

This convergent parallel mixed methods implementation study at Botswana’s national referral hospital found PEWS to be both acceptable and feasible to providers. Qualitative analysis of provider interviews identified major facilitators of PEWS implementation to be the perceived positive impact on patient care, increased accountability for deteriorating patients and the active inclusion of nursing in PEWS adaptation for PMW. Providers identified key implementation barriers as insufficient staff, interrupted access to PEWS nursing forms, persistent intra-professional and interprofessional hierarchies, and the lack of PICU facilities at PMH. Providers universally endorsed a belief that PEWS improved their care of children. While we were underpowered to detect a change in the rate of CDEs post implementation, the introduction of PEWS may have shifted the care of deteriorating children from the ward to better equipped, better staffed ICU settings.

Providers in Botswana reported some of the same barriers elicited from providers in four countries in Latin America following PEWS implementation,²³ including inadequate medical equipment, limited ICU capacity, understaffing and a hierarchical professional culture. As in Latin America, initial resistance related to the perceived complexity of PEWS was ultimately overcome and providers described a high level of motivation to use PEWS. In contrast to providers in Latin America, providers in our study did not endorse resistance from hospital leadership as a barrier. Nurses in Botswana experienced some of the same improvements in interprofessional communication reported by nurses in Rwanda following PEWS adoptions.⁵

We observed an increase in the relative frequency of ICU transfers after PEWS implementation, while the relative frequency of initiation of mechanical ventilation and inotropes in the ward decreased. This observation was corroborated by interviewed providers, who observed that sick children were being identified and transferred to an ICU earlier in their illness. However, no change in the rate of either ICU transfers or resuscitations was observed, and we did not observe the decrease in either unplanned ICU transfers post-PEWS implementation reported by Agulnik et al.⁸ This discrepancy is most likely due to insufficient power. To our knowledge, the background frequency of systematically categorised clinical deteriorations among hospitalised children was unavailable prior to this study. Botswana has a small population²⁴ and future studies on this topic may require multiple years of data collection at a single site or data from multiple facilities. The CDE rate metric available to us was also imprecise, as patient volume is only reported by PMH in terms of admissions, not hospital days or individual bed occupancy, and is only certified at the end of each month. Another possible contributor is that PEWS adoption was associated with more patients being labelled as ICU-ineligible—likely due to PEWS prompting more frequent or in-depth conversations within the medical team about ICU eligibility—which may have impacted the number of ICU transfers.

While documented adherence to PEWS was acceptable (see figure 3), interviewed doctors also reported low engagement with the PEWS algorithm, which specifies increased monitoring of children with abnormal vital signs who are not yet critical. This may have led to persistent delays in altering the management of stable children with worsening vital signs, and thus the persistence of CDEs and unplanned ICU transfers. The Department of Paediatrics was consulted on the structure and content of the PEWS algorithm, but, as front-line providers, residents received additional training that consultants did not, and nurses and residents were trained separately. While PEWS alone cannot address staff shortages, team-based training might help to further break down communication barriers between providers at different levels and encourage consultant accountability for deteriorating children who are not yet critically ill.^{6 7}

In addition to the challenges addressed above, our study was limited by the pre-post evaluation design and the collection of data during a single, 12-month window, which prevented us from seasonally normalising our observed CDE and mortality frequencies. The true background CDE rate and event mortality rate may vary from what we have reported here due to seasonal epidemiology. Our ability to measure a difference in event mortality was also limited by the small number of CDEs and mortalities captured during the study window, as discussed above. Final disposition and survival data were not available for all children transferred to private facilities for ICU care, as there is no formal care coordination process between the public and private health sectors. This may have led to inaccuracy in our observed event mortality. A multi-group interrupted time sequence analysis would have been a more rigorous way to measure the impact of PEWS. As the PMW was the only paediatric unit in the singular public teaching hospital in Botswana at the time, there was no practical control group against which to compare PEWS effectiveness in the PMW.

The men we interviewed were more likely than the women to describe themselves as engaged with the PEWS implementation process. There were fewer men in the study than women (4 vs 16), which reflects the general gender distribution among paediatric providers. Of the four men we interviewed, two have mentorship responsibilities in the ward. This may have led to them feeling more engaged than their women colleagues, or simply more comfortable sharing their feedback throughout implementation and during the interview.

Our study findings can be used to inform a portfolio of implementation strategies to support increased PEWS uptake. Figure 4 illustrates the implementation strategies recommended by the Expert Recommendations for Implementation Change (ERIC)²⁵ which map to the CFIR 2.0 constructs identified by this pilot study and could be employed by other facilities in Botswana planning to adopt PEWS. Clinical teams could consider including a structured reference to PEWS data during morning departmental handover or as a structured check-in during nightly intervals. Developing capacity for ongoing support could be developed through remote coaching and technical assistance^{13 26} to replace the role of the implementation team over time. Finally, macrosocial policy work is needed to support ICU resourcing and transfer processes, staffing workforce needs and ongoing equipment needs.

Figure 4. Matching of CFIR constructs identified by this study with implementation strategies recommended by the Expert Recommendations for Implementing Change (ERIC).¹ CFIR, Consolidated Framework for Implementation Research; PEWS, Paediatric Early Warning System; PMH, Princess Marina Hospital.

Conclusions

In conclusion, providers at Botswana’s national referral hospital found PEWS to be both acceptable and feasible. Providers unanimously reported that they believe it improves their care of hospitalised children. While PEWS alone cannot overcome health system-wide barriers to the early identification of deteriorating children, such as staffing shortages or the lack of an on-site PICU facility, our findings indicate that PEWS does have the potential to improve inpatient care by strengthening interprofessional communication and increasing provider accountability for deteriorating patients.

Data availability statement

Data are available upon reasonable request.