The impact of implementing a preprinted order form for inpatient management of otherwise healthy children admitted to a tertiary care centre with a diagnosis of bronchiolitis

Abstract

Objective

To examine the effect on length of stay (LOS) of a preprinted order (PPO) set for children admitted to hospital with a diagnosis of bronchiolitis, as well as on quality improvement measures, medical errors, and resource utilization.

Patients and methods

A retrospective chart review was performed of children admitted to the inpatient units at the Children’s Hospital of Eastern Ontario (CHEO) with a discharge diagnosis of bronchiolitis. Primary and secondary outcomes were compared between the pre-PPO (December 1, 2014 to June 30, 2015) and post-PPO (December 1, 2015 to June 30, 2016) periods. The primary outcome was LOS in days. Secondary outcomes included the proportion of quality improvement measures reached, the number of medical errors, and resource utilization.

Results

A total of 245 patients were included; 122 patients from the pre-PPO period (December 1, 2014 to June 30, 2015) and 123 patients from the post-PPO period (December 1, 2015 to June 30, 2016). Mean LOS was 3.1 days (2.7 to 3.5 days) and 2.8 days (2.4 to 3.2 days) in the pre- and post-PPO periods, respectively (multivariate analysis, P-value = 0.13). There were significant differences between the pre- and post-PPO periods in a number of quality improvement measures, although not in the number of medical errors. Significant reductions in oxygen, corticosteroid, antibiotic, and bronchodilator use were noted post-PPO implementation.

Conclusion

Although our bronchiolitis PPO did not significantly shorten LOS, it led to important improvements in quality improvement measures and in resource utilization efficiency.

Bronchiolitis affects more than one-third of children in the first 2 years of life. It is the most common reason for hospital admission in infancy (1,2) and causes significant financial burden on the health care system (3). Despite supportive care being bronchiolitis’ hallmark of treatment, investigations and treatments are still commonly prescribed, and significant variation in management practices remains (4,5).

Studies have shown that preprinted order sets (PPOs) decrease practice variability and improve compliance with established evidence-based practices, ultimately leading to more cost-effective care (6,7). PPOs also improve quality of care for a variety of conditions by reducing prescribing errors and length of hospital stay (LOS) (8–10). In an attempt to decrease LOS, standardize, and improve bronchiolitis care, our team developed a PPO (Supplementary File) based on current clinical guidelines and existing literature (1,2).

The goal of the present study was to evaluate if our PPO affected LOS, medical errors, quality improvement (QI) measures, and resource utilization. We anticipated that children hospitalized with bronchiolitis and treated as per our PPO would have a significantly shorter LOS than children treated without it. We also hoped to improve quality of care by standardizing monitoring and management, increase patient safety by decreasing potential for medical errors, and overall, improve children’s health.

METHODS

Population

We conducted a retrospective chart review of children admitted to the inpatient medical units of the Children’s Hospital of Eastern Ontario (CHEO) with a discharge diagnosis of bronchiolitis. CHEO, a tertiary paediatric care centre located in Ottawa, Ontario, Canada, admits approximately 400 children with bronchiolitis yearly. We excluded patients with congenital heart disease requiring medical therapy, chronic lung diseases such as cystic fibrosis and bronchopulmonary dysplasia, syndromes and/or genetic conditions, and neurological impairment, to decrease the potential impact of comorbidities on LOS. The review period comprised of a pre-PPO implementation period (December 1, 2014 to June 30, 2015) and of a post-PPO implementation period (December 1, 2015 to June 30, 2016). Of note, a nursing directive addressing oxygen saturation thresholds to trigger oxygen supplementation was implemented in May, 2016. The study was approved by the Institutional Ethics Board.

Data collection

The Department of Health Records at CHEO produced a computer-generated, randomly selected list of patients admitted during the pre- and post-PPO study periods. The charts of patients not meeting inclusion criteria were removed. Clinical data was manually extracted from patient charts and entered into an electronic data collection platform (REDCap). Collected data included baseline characteristics (age, gender, prematurity, concurrent diagnosis of bacterial pneumonia, previous bronchiolitis admissions in the last 12 months), QI measures (use of PPO, ordering of feeds, preferential use of frequent and superficial nasal suction over deep suction, oxygen monitoring and discontinuation, use of 90% as trigger for oxygen supplementation, ordering of daily weights, monitoring of intake and output, and of electrolytes if on intravenous fluids, documentation of respiratory assessment and family education), medical errors, as well as resource utilization data (oxygen supplementation, use of steroids, antibiotics, nebulizers (normal saline, epinephrine, hypertonic saline, salbutamol), and salbutamol metered dose inhalers, and prescription medication on discharge). Decisions on which QI and resource utilization data to include were both based on evidence (1,2) and expert opinion.

Sample size

Based on CHEO’s bronchiolitis LOS data from February 1 2014 to January 31 2015, a sample size calculation was performed. LOS was found to be skewed; therefore, a log transformation was used. The mean log-transformed LOS was 1.19 days (standard deviation [SD]=0.62), which, when back-transformed, corresponded to a median of approximately 3.29 days. A 20% reduction in LOS, felt to be clinically significant, would correspond to a reduction in the mean log-transformed LOS of log(0.8)=−0.22. The sample size calculation was thus performed on the log scale, based on detecting a difference of −0.22 assuming an SD of 0.62. Fixing the probability of type-I error at 5%, equal samples of 121 patients from both the pre- and the post-PPO implementation periods would provide 80% power to detect a 20% reduction in median LOS.

Study outcomes

The primary outcome was LOS, measured in days, for both the pretest and post-test periods. Secondary outcomes included the proportion of QI measures reached, the proportion of medical errors, and resource utilization in both study periods.

Data analysis

Baseline characteristics and secondary outcomes were summarized descriptively with median and interquartile range for continuous variables, and with frequencies and percentages for categorical variables. They were compared between pre- and post-test periods using the Wilcoxon rank sum for continuous variables and Fisher’s exact test for discrete variables. The primary outcome, LOS, was compared between pre- and post-test periods using univariate and multivariate linear regressions. Because LOS is positively skewed, regressions were performed using log-transformed LOS. Multivariate linear regression was used to investigate possible subgroup effects of age and history of prematurity, that is, interactions between period and age and between period and history of prematurity, along with the main effects of age and history of prematurity. Because interactions were not detected, the interaction terms were dropped from the multivariate model, and the model was refitted. Two-sided P-values less than 0.05 were considered statistically significant.

RESULTS

Demographics

A total of 245 patient charts were included in the final analysis; 122 pre-PPO and 123 post-PPO. There were no significant differences between groups. Table 1 provides characteristics of the study patients.

Table 1.

Baseline characteristics

Variable	Pre-PPO NPre/NT (%)	Post-PPO NPost/NT (%)	P-value
Overall	122/245 (49.8)	123/245 (50.2)
Sex (male)	76/122 (62.3)	76/123 (61.8)	1.00
Prematurity	25/122 (20.5)	19/123 (15.4)	0.32
Concurrent diagnosis of bacterial pneumonia	36/122 (29.5)	27/123 (22.0)	0.19
Any previous bronchiolitis admissions in the last 12 months	16/122 (13.1)	8/123 (6.5)	0.09
Variable	Pre-PPO Median (IQR)	Post-PPO Median (IQR)	P-value
Age (months)	10.8 (6.9, 16.5)	8.3 (5.6, 14.8)	0.05

IQR interquartile range; NT Total number of patients; NPre Total number of patients in the pre-PPO phase; NPost Total number of patients in the post-PPO phase.

Primary outcome

Multivariate analysis estimated mean LOS to be 3.1 (2.7, 3.5) days for the pre-PPO period and 2.8 (2.4, 3.2) days for the post-PPO period. No statistically significant interactions were detected between period and age (P=0.21) or between period and history of prematurity (P=0.23). There was no statistically significant difference in LOS between the two periods, using univariate or multivariate analysis. Neither age nor history of prematurity was statistically significantly associated with LOS (Figure 1) (Supplementary File).

Figure 1.

Length of stay compared between pre- and post-PPO time periods. Each circle represents an individual observation. The shaded boxes extend from the 25th percentile to the 75th percentile. The heavy horizontal line indicates the median. PPO Preprinted order.

Secondary outcomes

Quality improvement measures

There were significant differences between the pre- and post-PPO periods in multiple QI measures (Table 2).

Table 2.

Secondary outcomes: quality improvement measures

Variable	Pre-PPO N = 122 (%)	Post-PPO N = 123 (%)	P-value
There is a copy of the PPO filled out by a physician in the patient’s chart	0 (0.0)	88 (71.5)	0.09
Patient to be NPO when respiratory rate exceeds 60 breaths/min	37 (30.3)	91 (74.0)	<0.001
NG feeds or IV therapy to be initiated when patient NPO	6 (4.9)	63/ (51.2)	<0.001
Superficial nasal suction to be performed prior to feeds	15 (12.3)	71/ (57.7)	<0.001
Superficial nasal suction to be performed prior to inhalation therapy	1 (0.8)	68/ (55.3)	<0.001
Was a cut-off of 90% ordered for administration of supplemental oxygen	15 (12.3)	95 (77.2)	<0.001
Daily weights ordered	18 (14.8)	80 (65.0)	<0.001
Daily weights recorded	13 (10.7)	19 (15.4)	0.34
Strict ins and outs ordered	110 (90.2)	119 (96.7)	0.04
Respiratory assessment was documented pre- and post-each nebulization	58 (47.5)	41 (33.3)	0.03
Weight, height and head circumference were documented on admission to unit	97 (79.5)	102 (82.9)	0.52
Family education was documented	48 (39.3)	54 (43.9)	0.52
Electrolytes obtained on a daily basis for patients on full maintenance IV fluids (N=126)a	15/66 (22.7)	23/60 (38.3)	0.08
Discontinue continuous monitoring when patient no longer required oxygen (N=139)a	10/77 (13.0)	47/62 (75.8)	<0.001

Bolded values are significant at p<0.05. IV Intravenous; NG Nasogastric; NPO Nil per os; PPO Preprinted order.

^aOf note, denominators for the last two variables are different from the first 13 variables; only 126 of all reviewed patients were receiving full IV fluid maintenance, and only 139 patients were documented to be on continuous oxygen monitoring.

Medical errors

There were 13 charts with documented medical errors in the pre-PPO period as compared with 7 in the post-PPO period (P=0.17). These were considered to be nonserious medical errors. Eleven pre-PPO period errors related to erroneous dosing of medication, one was a nursing transcription error, and one was an inappropriate intravenous fluid choice. Two post-PPO period errors related to erroneous dosing of medications, one was a medication not ordered, three were transcription or missed orders, and one was inappropriate PPO use for a premature infant.

Resource utilization

There was a significant decrease in some resource utilization measures post-PPO implementation (Table 3).

Table 3.

Secondary Outcomes: Resource Utilization

Variable	Pre-PPO N=122 (%)	Post-PPO N=123 (%)	P-value
O2 supplementation	73 (59.8)	55 (44.7)	0.02
Steroid use	28 (23.0)	12 (9.8)	0.01
Antibiotics use	75 (61.5)	51 (41.5)	0.002
Normal Saline nebulizers	57 (46.7)	52 (42.3)	0.52
Epinephrine nebulizers	71 (58.2)	65 (52.8)	0.44
Hypertonic saline nebulizers	28 (23.0)	40 (32.5)	0.12
Salbutamol nebulizers	38 (31.1)	18 (14.6)	0.002
Salbutamol metered dose inhaler	28 (23.0)	16 (13.0)	0.05
Prescription medication on discharge	79 (64.8)	60 (48.8)	0.01
Chest x-rays ordered	31 (25.4)	27 (22.0)	0.55

Bolded values are significant at p<0.05. PPO Preprinted order.

DISCUSSION

Despite a trend toward decreased LOS, our study did not show a statistically significant reduction in LOS for children hospitalized with bronchiolitis when treated with our PPO, as compared to before implementation. This is surprising given the literature showing that PPOs reduce LOS in many medical conditions (8–13). However, there are multiple possibilities as to why this happened. Our baseline LOS was already low at a mean of 3.1 days prior to PPO implementation, therefore, the 20% change in LOS our study was powered for may have been too ambitious. Interestingly, although we did not achieve statistical significance, a 10% decrease in LOS was noted post-PPO implementation. Our baseline LOS is also important to acknowledge when looking at the use of hypertonic saline pre- and post-PPO, which was not significantly different. A 2017 Cochrane review (14) reported that hypertonic saline reduces LOS in infants with bronchiolitis if LOS is greater than 72 hours at baseline, while Silver et al. (15) showed that hypertonic saline has no effect if LOS is shorter than 72 hours. This may partially explain why, despite a trend toward increased hypertonic saline use post-PPO implementation, our LOS did not decrease significantly. Additionally, there was no delay between PPO implementation on the inpatient units and study data collection. As PPOs represent shifts in practice habits, there likely was a transitional period delaying the PPO being used to its full capacity, which could have impacted LOS. As there are no safe established discharge criteria in the literature, our PPO did not incorporate rigorous discharge criteria. This may have resulted in clinical practice variation between physicians and subsequent impact on LOS. Lastly, attempts to wean oxygen were not always documented, which may have resulted in prolonged oxygen use and increased LOS. Improved oxygen weaning practices are known to improve LOS and therefore must be a point of emphasis moving forward (16). Surprisingly, we did not show an association between prematurity and LOS, nor between age and LOS. It may be that infants most affected by complications of prematurity were excluded based on exclusion criteria. As for age, our study included only a small number of younger children, and studies have shown that LOS is most affected for children under the age of 3 months (17,18).

Many QI measures showed significant improvement during the study period. For instance, in the post-PPO period, superficial frequent suctioning practices, despite equivocal evidence (2), were encouraged over deep suctioning and long intervals between suctions, which are known to increase LOS (19). Additionally, continuous oxygen monitoring was discontinued more frequently after PPO implementation. While we did not calculate the individual effect of our oxygen weaning practices on LOS, protocols involving oxygen use and monitoring guidelines have been shown to reduce time on oxygen, time on continuous monitoring, and decrease LOS (20). These are important changes as pulse oximetry and oxygen therapy prolong LOS (21). It is important to note that a nursing directive to target and oxygen saturation of >90% while awake and >88% while asleep was introduced in May, 2016. This hospital wide practice change was further consolidated by our PPO. This change is consistent with current recommendations (1) and allows for decreased oxygen utilization (20) while maintaining a practice that is equally safe and clinically effective as using higher saturation targets (22).

The reduction in medical errors in our study did not reach statistical significance. Standardization of medication prescribing through the use of PPOs has been shown to decrease medical errors (8,9,23,24). The number of medical errors in both of our study periods was small, thus limiting our ability to draw strong conclusions

Finally, our PPO resulted in a decrease in resource utilization, consistent with current literature (13,25–27). There was a significant decrease in bronchodilator use in the post-PPO period. In 2014, a systematic review demonstrated that bronchodilator use in bronchiolitis had no effect on LOS nor on disease course (28). However, bronchodilators carry substantial costs. Although we did not perform a cost analysis, it is likely that a reduction in salbutamol doses would decrease hospitalization costs, as reported in prior studies investigating the impact of clinical pathways on bronchodilator use (29). Bronchodilators such as salbutamol also have the potential for toxicity (i.e., agitation, hypokalemia, tachycardia), therefore reducing their use can potentially improve patient safety (30). In addition, we observed a decrease in systemic steroid utilization post-PPO implementation. A 2013 Cochrane review evaluating the utility and safety of systemic steroids in bronchiolitis reported no effect on LOS (31). In addition, another study published six years prior reported no effect of dexamethasone on LOS, adverse events, or respiratory status in bronchiolitis treatment, also supporting decreasing steroid usage (32).

We did not observe a significant change in the number of patients receiving a chest radiograph (CXR) between the two study periods, despite clear guidelines that CXRs should not be routinely ordered in the management of bronchiolitis (1). This could be because most CXRs are ordered in the emergency department, prior to the inpatient PPO taking effect. Studies show that CXRs do not change acute management in bronchiolitis (33), and may in fact result in a greater proportion of patients treated with antibiotics (34,35). The data surrounding the effect of PPOs on antibiotic use in bronchiolitis is somewhat inconsistent, with reports of no reduction (13,25), and reports of decreased use (12). However, in our study, the number of children receiving antibiotics in the post-PPO-period was significantly reduced, which also translated in a significant decrease in patients given medications on discharge. This decrease may stem from a combination of increased awareness of prescribing practices with use of our PPO, as well as implementation of an institutional antimicrobial stewardship program (36). Continued antibiotic stewardship in these patients is crucial as antibiotics have been shown to have no effect on LOS in bronchiolitis (32) and their excessive use may increase bacterial resistance.

There are several limitations to our study. First, its quasi-experimental independent pretest post-test design includes an intrinsic difficulty in directly inferring causality (37,38). However, collecting data over two back to back bronchiolitis periods minimizes the possibility that major practice changes in bronchiolitis treatment, unrelated to the use of the PPO, occurred during the study period. Retrospective chart reviews are also limited by the information that is charted, such as missing data and variability in the quality or detail of documentation between practitioners (39). As suggested in the literature, we developed a data collection form, adopted an electronic database system, and employed a trained research team of data abstractors, statisticians, and health care professionals to mitigate this limitation (40). There is also variation in the description of the clinical definition of bronchiolitis, which can impact selection of patients to be managed on the PPO and theoretically could impact external validity (4). However, our study represents real-life situations in which the PPO is applied—for children whose diagnosis is made clinically by the admitting team. PPO adherence was documented in 72% of patients in the post-PPO period. This may have resulted in children not being treated in accordance to current guidelines and subsequently impacting LOS and other outcomes as well. Since the end of the study period, all medical documentation and ordering, including PPOs, have been converted to a computerized system, which has been shown to improve adherence (12). Finally, as discussed above, our limited sample size as well as our ambitious expected change in LOS may have undermined our ability to show a statistically significant reduction in LOS.

CONCLUSION

Bronchiolitis is an important and prevalent paediatric condition with high hospitalization rates that result in burdensome costs to health care, morbidity to children, and impact on families. This widespread impact makes it imperative that we continue to refine our management protocol in an effort to improve these outcomes. Although our study did not demonstrate a statistically significant decrease in LOS, it adds to a growing body of bronchiolitis literature by suggesting that PPOs can have a positive impact on QI measures, as well as significantly decrease the use of certain hospital resources. This has the potential for important implications on both hospital economics and patient care. We hope that by addressing the potential gaps in our PPO, such as outlining discharge criteria and achieving consistent provider implementation, we will shorten LOS, improve adherence to clinical guidelines, while continuing to provide cost-effective and safe patient care.