Article Text
PDF
Abstract
Objective To describe the associations between patient-to-nurse staffing ratios and rates of mortality, process of care events and vital sign documentation.
Design Secondary analysis of data from the evaluating processes of care and outcomes of children in hospital (EPOCH) cluster-randomised trial.
Setting 22 hospitals caring for children in Canada, Europe and New Zealand.
Participants Eligible hospitalised patients were aged>37 weeks and <18 years.
Primary and secondary outcome measures The primary outcome was all-cause hospital mortality. Secondary outcomes included five events reflecting the process of care, collected for all EPOCH patients; the frequency of documentation for each of eight vital signs on a random sample of patients; four measures describing nursing perceptions of care.
Results A total of 217 714 patient admissions accounting for 849 798 patient days over the course of the study were analysed. The overall mortality rate was 1.65/1000 patient discharges. The median (IQR) number of patients cared for by an individual nurse was 3.0 (2.8–3.6). Univariate Bayesian models estimating the rate ratio (RR) for the patient-to-nurse ratio and the probability that the RR was less than one found that a higher patient-to-nurse ratio was associated with fewer clinical deterioration events (RR=0.88, 95% credible interval (CrI) 0.77–1.03; P (RR<1)=95%) and late intensive care unit admissions (RR=0.76, 95% CrI 0.53–1.06; P (RR<1)=95%). In adjusted models, a higher patient-to-nurse ratio was associated with lower hospital mortality (OR=0.77, 95% CrI=0.57–1.00; P (OR<1)=98%). Nurses from hospitals with a higher patient-to-nurse ratio had lower ratings for their ability to influence care and reduced documentation of most individual vital signs and of the complete set of vital signs.
Conclusions The data from this study challenge the assumption that lower patient-to-nurse ratios will improve the safety of paediatric care in contexts where ratios are low. The mechanism of these effects warrants further evaluation including factors, such as nursing skill mix, experience, education, work environment and physician staffing ratios.
Trial registration number EPOCH clinical trial registered on clinical trial.gov NCT01260831; post-results.
- paediatrics
- mortality
- nurses
- hospitals
- human resource management
- health & safety
Data availability statement
No data are available. Data from the EPOCH trial are not available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
This study is based on a multicentre international prospective data dataset of >210 000 patient admissions in 22 hospital sites in 7 high-income countries.
The measure of patient-to-nurse staffing was from a randomly selected sample of beds in the inpatient ward areas where studied patients received care.
Staffing levels and clinical outcomes represent an average over three 6-month periods.
Individual patient-level risk adjustment and a wider range of variables to control for case mix and general hospital resources would have increased the precision of our results.
Unmeasured factors may be confounding the association between patient-to-nurse ratio and the study outcomes.
Introduction
Nurses are the primary human resource for clinical observation and the delivery of patient care in paediatric hospitals. They are essential components of rapid response systems with roles that include monitoring clinical condition, communicating to others in the primary care team and escalating the intensity of care as determined by the child’s severity of illness.1 Low patient-to-nurse ratios are a well-accepted strategy to mitigate already identified patient risk in intensive care units (ICUs).2–5 In observational studies, lower patient-to-nurse ratios have been associated with lower mortality in adult patients6 7 and with lower hospital readmission at 15–30 days in children.8 Standards of paediatric nursing care on hospital wards require comprehensive and continuous patient assessments, vital signs measurement and documentation9 10 to provide information on early signs of clinical deterioration.11 High patient-to-nurse ratios increase workload for each nurse, which may reduce the frequency of vital signs monitoring, and compromise the effectiveness of interventions and communication.12–15
While there is evidence on the association of nurse staffing levels and patient mortality in the adult inpatient setting, this evidence is limited in paediatric settings.16 17 A study using nurse staffing administrative data in California from 1996 to 2001 examining 3.65 million discharges from 286 general and children’s hospitals reported no association of patient-to-nurse staffing and mortality in children.18 Another study using the Healthcare Cost and Utilization Project Kids’ Inpatient Database in California showed no association of nurse ratios and registered nurses (RNs) Full Time Equivalents (FTEs) with lower risk-adjusted mortality, risk-adjusted complications and risk-adjusted resource utilisation for paediatric cardiac surgical services.19 Moreover, a study performed in Korea including over 600 000 children admitted to 46 tertiary care hospitals showed an increased risk of failure to rescue associated with lower patient-to-nurse staffing ratios (grade 1 (beds/nurse<2), OR 1.39; 95% CI 1.15 to 1.70; compared with grade 3 (2.5≤beds/nurse<3). In the same study, in addition, an association with a better composite outcome of cardiac arrest, shock or respiratory failure was found with lower patient-to-nurse ratios (grade 1 compared with grade 3 OR 0.48; 95% CI 0.40 to 0.58).20 In the neonatal ICU setting, there is some evidence of an association between patient-to-nurse staffing levels or increased proportions of nurses with neonatal certifications and reduced neonatal mortality.21 In paediatric ICU (PICU), an association with lower mortality and lower odds of complications was found with increased years of nursing experience and nurses with bachelor or higher degrees.22–25 Current evidence on safe patient-to-nurse ratios for the paediatric setting is limited and much needed to support future nursing workforce planning.17 Therefore, the objective of this study was to determine the association between patient-to-nurse staffing and rates of clinical deterioration events and processes and perceptions of care on in-hospital paediatric wards.
Methods
We performed a secondary analysis of data from 217 714 patient admissions in acute care paediatric wards the 22 hospitals included in the ‘evaluating processes of care and outcomes of children in hospital’ (EPOCH) cluster-randomised trial.26 Outcomes were available for all patients without loss to follow-up. The hospitals were located in Belgium (n=1), Canada (n=11), the UK (n=5), Ireland (n=2), Italy (n=1), New Zealand (n=1) and the Netherlands (n=1) and had a total of 2085 eligible inpatient unit beds.26 Three hospitals had >200 beds, 10 (45%) had a 24/7 rapid response team and 20 (91%) were University affiliated. 11 hospitals were randomised to implement the Bedside Paediatric Early Warning System (BedsidePEWS).
Main exposure
The main exposure of interest was patient-to-nurse staffing on inpatient wards of participating hospitals. Within each of three 26-week study periods at each hospital, we recorded the total number of patients cared for by the primary nurse of five randomly selected patients, who were eligible on an inpatient ward for more than 24 hours. For each week in each hospital, the study coordinating centre provided a randomly generated list of 20 bed spaces generated from a list of the bed spaces in the paediatric wards (not ICU) where eligible patients may receive care. Coordinators sought eligible patients in the indicated bed spaces, beginning at the top of the list and progressing sequentially until five eligible patients were enrolled.
The mean value of the (5 patients) × (26 weeks) = 130 measurements collected in each 26-week period was used to represent typical staffing levels on the inpatient wards of each hospital for that period.26
Primary and secondary outcomes
The primary outcome was hospital mortality. This was also the primary outcome of the EPOCH trial. Secondary outcomes were: (1) five events reflecting process of care, also collected on all EPOCH patients: clinical deterioration events, late ICU admissions, resuscitation team calls, stat calls and PICU consultations; (2) the frequency of documentation for each of eight vital signs on the random sample of patients; and (c) four measures describing nursing perceptions of care.
Clinical deterioration events
Among secondary outcomes reflecting process of care, clinical deterioration events were defined as deaths on the ward or urgent admission to a PICU, which itself was defined as an admission with departure from the event location in less than 6 hours from the time the PICU admission was initiated.26 The timeliness of urgent PICU admissions was classified using the Children Resuscitation Intensity Scale, with scores that range from 1 (no major interventions) to 7 (death before or within the first hour after ICU admission); a score>2 was classified as a late ICU admission.26
Vital signs documentation
Vital signs were documented on the randomly selected patient records. We abstracted the number of documented measurements over a 24-hour period for each patient for each of eight clinical observations: heart rate, respiratory rate, systolic blood pressure, oxygen therapy, oxygen saturation, capillary refill time, respiratory effort and temperature. We defined vital sign recording as complete when the first seven of these (the clinical indicators in the BedsidePEWS) were documented. We also calculated the BedsidePEWS Score (if at least five of the seven measurements were available) from the last available set of vital signs.
Nurses perceptions of quality of patient care
Nurses’ perceptions of quality of patient care were recorded once in each study period using a Documentation and Interaction Survey, developed and used extensively in the parent study. This survey had the aim of exploring the perception of the documentation system and quality of care. Questions were judged to have high face validity by the study group, although not formally validated. Responses to the communication quality question ‘how do you rate communication about patients on your team?’ were on a 9-point scale from 1, (extremely poor) to 9 (excellent). Responses to the timely care question “please indicate your agreement/disagreement with the statement: ‘patients have received the care that they needed when they needed it’” were on a 5-point scale from 1 (strongly disagree) to 5 (strongly agree). Responses to the apprehension question ‘when calling a physician after hours to review the patient or their management plan, how apprehensive did you feel?’ were on a 9-point scale from 1 (extremely apprehensive) to 9 (not at all apprehensive). Responses to the influence on care question ‘how confident did you feel that you could influence the plan of care?’ were on a 9-point scale from 1 (extremely confident) to 9 (not at all confident), which we reversed, so that a 9 became ‘extremely confident’.
Adjustment variables
In recognition of the modest number of site-by-period observations in the study data set, we carefully considered options for adjustment variables. After consideration of seven options describing the hospital resources—(1) the proportion of full-time (>90% full-time equivalent) RNs, (2) number of paediatric beds, (3) presence of a transplant programme, (4) overnight in-house consultant paediatrician, (5) overnight in-house senior trainee (fellow), (6) 24/7 medical emergency team and (7) mean severity of illness—four adjustment variables were used. Two reflected patient risk for deterioration—the presence of a transplant programme and the average patient severity of illness (measured as the mean BedsidePEWS Score at the site in the period). Two reflected non-nurse staffing resources—the presence of in-house overnight senior trainee (fellow) and having a medical emergency team. We also adjusted by a fifth variable indicating whether the period was assigned to the EPOCH intervention. Online supplemental table 1A summarises the sources and roles of all variables included in our analyses.
Supplemental material
Statistical analysis
Clinical outcomes were aggregated over each 26-week period and represented as rates. Mortality was summarised as deaths per 1000 patient discharges. Five process of care outcomes (clinical deterioration events, late PICU admissions, immediate calls to a physician to attend at the bedside of a patient, resuscitation team calls and PICU consultations) were summarised as events per 1000 patient days. To assess the dependence of hospital mortality on the patient-to-nurse ratio, we entered aggregate death and admission data for each site in each period into a random effects logistic regression, with site-specific random effects and site-and-time specific predictors (EPOCH intervention, patient-to-nurse ratio, presence of transplant team, the presence of in-house overnight senior trainee presence of a medical emergency team and the mean BedsidePEWS).
Similar analyses were used for process of care events, except that random effects negative binomial models were used, with event counts as the outcome, and the logarithm of site period totals of patient ward days as an offset. Analyses of documentation of vital signs also used a random effects negative binomial models, with aggregated counts of documentation as the outcome and logarithm of the total number of assessments on the randomly selected patients as the offset.
A random effects proportional odds regression model was used to assess the dependence of nurse-reported perceptions of quality of care on the patient-to-nurse ratio, with site-specific random effects and the same site-and-time-specific predictors as above. All models were fitted using Bayesian models in the brms package in R and measures of association are presented with a 95% credible interval (CrI) and the Bayesian posterior probability of a reduction in the outcome with a higher patient-to-nurse ratio. After inspection of the data, a post hoc sensitivity analysis was performed in which we excluded a hospital that was an outlier (site 21) with respect to the patient-to-nurse ratio.
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, reporting, or dissemination plans of this research.
Results
Data from 22 hospitals participating in the EPOCH trial26 included 217 174 patients and 849 798 patient days. Random sampling selected 8282 patients for whom we assessed vital sign documentation and patient-to-nurse ratios. The median (IQR) number of patients cared for by an individual nurse was 3.0 (2.8–3.6). Three hospitals (14%) had a mean patient-to-nurse ratio greater than 4, with one outlier having a patient/nurse ratios of 6.5–7.0 over the three study periods (figure 1).
Average patients per nurse and total number of ward days by hospital. The total number of ward days in a period at a site against the average patient-to-nurse ratio in that period at that site, with periods at the same site plotted in the same colour. The boxplots in the margin show the 25th and 75th percentiles and medians across site periods.
Hospital mortality occurred in 360 patients with an overall rate of 1.65/1000 patient discharges (1.57 in period 1), and other events occurred at rates 0.53 to 5.9/1000 patient days in period 1 (table 1).
Model-based estimates for associations between average patients per nurse and clinical outcomes
In univariate analyses, we found a higher number of patients per nurse was associated with lower odds of hospital mortality and lower rates of clinical deterioration events, late ICU admissions and resuscitation team calls. In multivariable models, point estimates for these four associations were similar to the estimates from univariate analyses; the probability of a reduction in mortality with a higher patient-to-nurse ratio increased, but other probabilities of associations decreased (table 1).
Figure 2 plots the association of each of these study outcomes with nurse staffing by 26-week study period, with the estimated univariate association overlaid. In the sensitivity analysis excluding the hospital with a high patient-to-nurse ratio, the associations with hospital mortality and late ICU admissions were stronger, but the evidence for all other associations with clinical events was weaker (right-hand columns of table 1 and online supplemental figure 1).
Supplemental material
The association of the mortality and process of care study outcomes and nurse staffing. (A–F) The rate of the outcome in a period at a site against the average patient-to-nurse ratio in that period at that site, with the areas of the circles being proportional to the number of patient discharges or number of ward days. The fitted values from the univariate random effects models are shown by the solid line (mean) and shaded area (95% credible interval). ICU, intensive care unit; PICU, paediatric intensive care unit.
In multivariable models, nurse perceptions tended to be less favourable in hospitals where nurses were caring for more patients: point estimates of ORs were all <1, meaning that the odds of more favourable perceptions decreased as the patient-to-nurse ratio increased. However, it was only for influence on care that there was a high probability (>95%) that the OR was <1 in either the univariate or multivariable model (table 2).
Model-based estimates of the OR for a better score on the DIS outcomes with increasing values of average patients per nurse
Findings were largely unchanged in the sensitivity analysis. Online supplemental figure 2 shows the nurse perceptions for surveys from site periods where the patient-to-nurse ratio was in the bottom quartile, middle two quartiles and top quartile.
Supplemental material
Documentation of several clinical observations was also related to nurse staffing. Adjusted analyses found, as the patient-to-nurse ratio increased, greater than 95% probability of reduced documentation for each vital sign, with two exceptions, temperature and respiratory effort. Documentation of the complete set of vital signs was also reduced with increased patient-to-nurse ratio (table 3).
Model-based estimates of the association between average patients per nurse and vital signs documentation
With the exclusion of the outlying hospital, there were reductions in documentation with an increased patient-to-nurse ratio, but the probabilities of reductions were lower (online supplemental table A2).
Discussion
We evaluated the associations of nurse staffing with mortality, clinical process measures reflecting failure to rescue, documentation of vital signs and nurse perceptions of care in a secondary analysis of prospective clinical trial data from 217 174 patient discharges in 22 hospital sites in 7 high-income countries.26 There are three main findings and related implications.
First, we found that higher patient-to-nurse ratio was associated with lower mortality and fewer resuscitation team calls, clinical deterioration events and late ICU admission. The robustness of our findings is suggested by the consistency of observed effects across (1) single variable analyses, (2) multivariable analyses that include adjustment for severity of illness and some organisational factors, (3) a sensitivity analysis where a potential outlier hospital was excluded and (4) by concordance of the direction of effects for different related process of care outcomes. Our results contrast with observational studies in adult hospital settings7 27–31 and in the neonatal ICUs where more patients per nurse have been associated with worse outcomes—including mortality.21 32–35
Differences between paediatric wards settings and adult inpatient wards, paediatric and neonatal ICU settings are numerous. Our review of studies found two studies evaluating the nurse staffing in paediatric inpatient wards: one in California (3.65 million admissions in 286 hospitals) found no association of staffing with mortality18 and another from Korea (608 017 admissions in 46 hospitals) found more nurses per patient was associated with increased rates of failure to rescue.20 An additional study evaluating paediatric cardiac surgical services in California showed no association of nurse ratios and RN FTEs with lower risk-adjusted mortality.19 All used coarse methods to calculate nurse staffing, used administrative data and may not have been able to discern patient location (ICU or ward) at the time of events from the administrative data that was used to identify selected events.
Common practice is for sicker patients—those of higher acuity—to be assigned to nurses who are caring for a smaller number of other patients. This practice may lead to an underestimation of the true effect of nursing staffing in unadjusted analyses.36 Another interpretation of the study results is that they illustrate that (at hospital level) the allocation of nurses to the patients in paediatric wards is matched to the patient’s risk of mortality. Thus, nurse staffing is a consequence of expected patient risk of mortality—rather than being its determinant. More studies are needed to understand what other factors related to the work environment, interprofessional collaboration, staffing of other professions, communication, parental advocacy and escalation processes may contribute to failure to rescue and may help identify effective solutions.37–40
Second, adjusted analyses found that documentation of heart rate, respiratory rate, systolic blood pressure, oxygen saturation, capillary refill time, oxygen therapy and complete sets of vital signs was less frequent with higher patient-to-nurse ratio. While the association or higher level of nurse staffing and the completeness and timeliness of vital signs monitoring is reported in adult settings, the effect of increased nurse staffing on this process is small.15 We note that in unadjusted analyses the observed effect of staffing was less than found in adjusted analyses—suggesting that overall documentation was similar and that once acuity and patient complexity were accounted for then documentation was reduced more if the nurse was looking after a greater number of other patients.
Third, nurses working in hospitals with higher patient-to-nurse ratios reported greater apprehension when calling a physician after hours, and perceived worse communication quality, timeliness of care and reduced ability to influence care; the finding was strongest for influence of care. This result is consistent with published observational studies suggesting worse patient outcomes and quality of care in hospitals with less staffing in adult inpatient wards,6 28 29 31 41 neonatal ICU,21 32 33 42 and PICUs,22–25 and of studies linking less favourable perceptions of quality and safety with lower nurse staffing.43–45 Our findings of less favourable nurse perceptions associated with higher patient-to-nurse ratio is consistent with our finding of reduced documentation. It also contrasts with our objective data of lower mortality and fewer clinical events with less intensive nurse staffing. We hypothesise that increased workload in hospitals with higher patient-to-nurse ratio may be influencing perceptions of care quality. It is likely that for most respondents, their frame of reference is dominated by the hospital in which they work; that is, the local culture may lead to different expectations of quality and safety that are separate from the objectively observed trial data. We hypothesise that greater expectations of frontline staff may be mitigating any adverse consequences of higher patient-to-nurse ratio and contributing to lowered rates of adverse outcomes.
Limitations
There are several limitations of this study. First, our measure of patient-to-nurse staffing was from a randomly selected sample of beds in the inpatient ward areas where studied patients received care. Ideally, the nurse staffing would have been recorded for all patients on all days of the 78 weeks of the study. Other approaches, including aggregating nurse staffing each patient day, or sampling a larger sample, may have increased the precision of our description of staffing. This limitation is shared with other studies of inpatient paediatric care.18 20 Our finding of relative stability of estimates from the 130 patients sampled from each 26-week period in the study suggests the approach used was not subject to major random variation. Second, unmeasured factors may be confounding the association between patient-to-nurse ratio and the study outcomes. Possible factors include physician staffing ratios46 that may be related to a ‘teaching hospital’ effect; the availability of licensed practice nurses27 or vocational nurses to moderate the nursing workload; and other patient confounding variables such as case mix, age and comorbidities. The parent study was performed in relatively well-staffed tertiary care hospitals, where the mean patient-to-nurse ratio was relatively low and may have concealed reduced risk from a higher baseline risk in the most acute and complex patients. Possibly, in well-staffed settings, other solutions need to be found to further reduce patient mortality and failure to rescue.
Third, unaccounted differences in nurse education and skill mix might have confounded our results. Increased rates of RNs and nursing support staff,31 41 nurses with bachelor degrees,7 47 48 nurses with higher levels of education49 or improved working environments43 have been found to be essential in reducing safety failures, in-hospital failure to rescue and deaths in the adult patient population.29
Strengths
The multicentre international prospective data collection including 217 174 patient admissions in 22 hospital sites in 7 high-income countries is a strength of this study. This improves the generalisability of the results of the findings.
Conclusions
Our 22-hospital evaluation found that mortality, clinical deterioration and resuscitation team activation were about 10%–20% less frequent with each additional patient per nurse. Our findings were consistent across multiple analyses and in six related measures, and contrast with prior observational data in the adult population showing increased nurse staffing was associated with reduced events. In contrast, these results from paediatric inpatient units suggest that reductions to current patient-to-nurse ratios may not improve the quality and safety of care. Our findings are hypothesis generating and emphasise the value of considering other factors, including skill mix, education requirements, experience and physician: patient ratios before implementing well-intentioned decisions to increase nurse staffing to improve patient safety and quality of care.
Data availability statement
No data are available. Data from the EPOCH trial are not available.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the SickKids Research Ethics Board, Toronto, Canada (Approved REB # 1000062622). The need for written informed consent for patient or clinician participation was waived in all jurisdictions for the EPOCH clustered randomised trial. This 21-centre cluster-randomised clinical trial was coordinated by the Center for Safety Research at the Hospital for Sick Children, in Toronto, Ontario, Canada, and was overseen by the study executive steering committee and the Canadian Critical Care Trials Group.
Acknowledgments
We would like to thank Catia Genna and Kiara Ros Thekkan for their helpful support in editing this paper.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
X @ors_gaw, @dryden_palmer
OG and CSP contributed equally.
Contributors OG and CSP led the study design. GT and LS performed the statistical analyses. OG and CSP wrote the first draft of the manuscript. MLCdA, MR, CC, ET, ID'O, ARJ and KD-P contributed to the interpretation of the findings and the final draft of the paper. MR and ET are the guarantors.
Funding Work in part funded by an unrestricted grant from the Gluskin Sheff and Associates Pediatric Critical Care Research Endowment. The funding agency had no role in the design, conduct, interpretation nor decision to publish this work.
Competing interests CSP is a named inventor of the Bedside Paediatric Early Warning System and has shares in a decision support company in part owned by SickKids that was established to commercialise the Bedside Paediatric Early Warning System. All other authors declare no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.