BACKGROUND AND OBJECTIVES:

Acute hematogenous osteomyelitis (AHO) causes significant morbidity in children. Quality improvement (QI) methods have been used to successfully improve care and decrease costs through standardization for numerous conditions, including pediatric AHO. We embarked on a QI initiative to standardize our approach to the inpatient management of AHO, with a global aim of reducing inpatient costs.

METHODS:

We used existing literature and local consensus to develop a care algorithm for the inpatient management of AHO. We used the Model for Improvement as the framework for the project, which included process mapping, failure mode analysis, and key driver identification. We engaged with institutional providers to achieve at least 80% consensus regarding specific key drivers and tested various interventions to support uptake of the care algorithm.

RESULTS:

Fifty-seven patients were included. There were 31 patients in the preintervention cohort and 26 in the postintervention cohort, of whom 19 were managed per the algorithm. Mean inpatient charges decreased from $45 718 in the preintervention cohort to$32 895 in the postintervention cohort; length of stay did not change. Adherence to recommended empirical antimicrobial agents trended upward.

CONCLUSIONS:

A simple and low-cost QI project was used to safely decrease the cost of inpatient care for pediatric AHO at a tertiary care children’s hospital. A robust local consensus process proved to be a key component in the uptake of standardization.

Acute hematogenous osteomyelitis (AHO) causes significant morbidity in children1,3 and often requires prolonged and costly treatment.1,3,6 A recent large study from Spain revealed that children with osteomyelitis were hospitalized for an average of 13.5 days, that 20% required surgery, and that 2.3% developed long-term sequelae, such as leg-length discrepancy.3 Early transition to oral therapy has been shown to be an effective and less costly approach compared with prolonged parenteral therapy.5,13

Despite efforts at our institution to reduce costs through early transition to oral therapy,14 our inpatient costs remained high. In preliminary discussions among specialties involved in caring for children with AHO at our institution, substantial variation in practices from 1 patient to another was identified. We hypothesized that standardizing our practice in all areas would lead to an overall cost reduction and improved care.

We embarked on a quality improvement (QI) initiative to standardize the approach to the inpatient management of AHO, with a global aim of reducing inpatient costs. In initial work, we attributed more than half of inpatient charges to length of stay (LOS). Therefore, we set out to develop and implement a local care algorithm on the basis of existing evidence15 and local consensus, focusing on factors affecting LOS. Our specific aims at the outset of the project were (1) to achieve 80% use of algorithm-recommended antimicrobial agents and (2) to decrease the average inpatient LOS of patients with AHO from 6.3 to 5 days within 12 months.

Cincinnati Children’s Hospital Medical Center is a free-standing tertiary care children’s hospital. It is the primary provider of pediatric acute care in our 8-county region and has >600 beds (440 as acute care medical beds). Before development of the care algorithm, most patients with AHO were admitted to the inpatient setting under the Division of Hospital Medicine (HM), with consultation routinely provided by the Division of Orthopedics. The Division of Infectious Diseases (ID) provided inpatient and outpatient consultative services for the majority of patients with AHO. Inpatient ID consultation often occurred late in the admission when transition to outpatient was being contemplated. Occasional patients were admitted under orthopedics, with HM, ID, or both consulting. We intentionally included opinion leaders from each division on the core team. We also consulted extensively with additional groups who commonly participate in caring for children with AHO, such as radiology, emergency medicine, and the pediatric residency program.

We included patients aged 3 months to 21 years with a clinical history, examination, and imaging (plain radiograph, MRI, or technetium bone scan) consistent with AHO.1 We excluded patients with any nonhematogenous source, including a previous fracture, surgery (including hardware), penetrating trauma, or a pressure ulcer at or near the site of the infection. We also excluded patients with chronic osteomyelitis (ie, antecedent symptoms lasting >2 weeks) and those with underlying chronic illnesses or those who were immunocompromised. Patients with multifocal infections (including septic thrombophlebitis and septic pulmonary emboli) or a protracted bacteremia (defined as positive blood culture results on ≥3 consecutive days) were excluded from all analyses except the postintervention analyses of algorithm antimicrobial adherence and time to effective therapy, in which they were included for operational reasons.

Using a manual chart review, we collected a preintervention (baseline) data set by retrospectively identifying cases of AHO among all patients who were seen for hospital follow-up in the ID clinic during 2015. Inpatients with AHO are routinely referred to the ID specialty clinic at our institution. For postintervention case identification, we used real-time notification by an enterprise intelligence resource (VigiLanz; VigiLanz Corp, Minneapolis, MN). This software allowed for us to create a rule that automatically sent an e-mail to a team member every time a radiology report was generated that included the word “osteomyelitis,” including in the indication for testing. We also used active surveillance16 by conferring regularly with frontline providers and reviewing indications for all inpatient ID consults and outpatient referrals for the management of a suspected bone infection. Any patients identified by either method underwent a detailed chart review to see if they met the algorithm’s inclusion criteria.

The Model for Improvement provided the framework for the project, which included process mapping, failure mode analysis, and key driver identification.17,19 We reviewed the existing literature and similar algorithms used by other children’s hospitals. We analyzed our preintervention data set to evaluate our local demographics (Table 1), clinical characteristics, current processes (ie, antimicrobial selection and surgical interventions), and outcomes (ie, LOS and costs). Information thus obtained constituted the foundation for a subsequent consensus building process. The team then developed a care algorithm that contained recommendations (Supplemental Information) with at least an 80% consensus agreement among local providers across all included divisions and that was used to target key drivers (Supplemental Fig 5) of our process and outcome measures. The algorithm provided recommendations on the initial workup (including criteria for expedited MRI), the timing and selection of empirical antimicrobial agents, surgical indications on initial and subsequent assessments, early transition to enteral antimicrobial agents for most patients, specific criteria for timing of transition to enteral antimicrobial agents, and specific discharge criteria.

TABLE 1

Characteristics of Patients With Pediatric AHO, Cincinnati Children’s Hospital Medical Center, January 2015–May 2017

Patient CharacteristicsAll Patients (n = 57)Preintervention (n = 31)Postintervention (n = 26)
Age, y, mean, median (range) 7, 6.4 (0.7–17.8) 6.9, 6.4 (1.2–14.6) 7.3, 6.6 (0.7–17.8)
Age groups, y, n (%)
<1 1 (2) 1 (4)
1–5 25 (44) 14 (45) 11 (42)
6–12 26 (46) 15 (48) 11 (42)
13–18 5 (9) 2 (6) 3 (12)
Male sex, n (%) 38 (67) 21 (68) 17 (65)
Treatment, n (%)
Completed algorithma — — 19 (73)
Surgery 25 (44) 14 (45) 11 (42)
Anatomic site, n (%)
Tibia and fibula 26 (45) 18 (58) 8 (31)
Pelvis 9 (16) 2 (6) 7 (27)
Other 9 (16) 3 (10) 6 (23)
Femur 8 (14) 7 (23) 1 (4)
Arm 5 (9) 1 (3) 4 (15)
Microbiology, n (%)
Culture results negative 18 (32) 10 (32) 8 (31)
S aureus 35 (61) 21 (68) 14 (54)
Kingella kingae 2 (4) 2 (8)
Streptococcus pyogenes 1 (2) 1 (4)
Bartonella henselaeb 1 (2) 1 (4)
Patient CharacteristicsAll Patients (n = 57)Preintervention (n = 31)Postintervention (n = 26)
Age, y, mean, median (range) 7, 6.4 (0.7–17.8) 6.9, 6.4 (1.2–14.6) 7.3, 6.6 (0.7–17.8)
Age groups, y, n (%)
<1 1 (2) 1 (4)
1–5 25 (44) 14 (45) 11 (42)
6–12 26 (46) 15 (48) 11 (42)
13–18 5 (9) 2 (6) 3 (12)
Male sex, n (%) 38 (67) 21 (68) 17 (65)
Treatment, n (%)
Completed algorithma — — 19 (73)
Surgery 25 (44) 14 (45) 11 (42)
Anatomic site, n (%)
Tibia and fibula 26 (45) 18 (58) 8 (31)
Pelvis 9 (16) 2 (6) 7 (27)
Other 9 (16) 3 (10) 6 (23)
Femur 8 (14) 7 (23) 1 (4)
Arm 5 (9) 1 (3) 4 (15)
Microbiology, n (%)
Culture results negative 18 (32) 10 (32) 8 (31)
S aureus 35 (61) 21 (68) 14 (54)
Kingella kingae 2 (4) 2 (8)
Streptococcus pyogenes 1 (2) 1 (4)
Bartonella henselaeb 1 (2) 1 (4)

No patients were enrolled during the January 2016–April 2016 washout period. —, not applicable.

a

Patients in the postintervention cohort who met initial inclusion criteria and never developed exclusionary findings (patients who developed exclusionary findings during treatment were taken off the care algorithm).

b

Diagnosed in 1 patient with vertebral osteomyelitis on the basis of history and serologies. Bartonella serologies were not routinely performed.

On the basis of identified key drivers, we tested multiple interventions using plan-do-study-act cycles to support reliable uptake of the care algorithm’s recommendations in clinical practice. We presented the algorithm in person or through e-mail to each of the divisions to obtain consensus and trained the residents in the rationale, content, and application of the algorithm in clinical care. After these interventions, which occurred over 4 months between January 2016 and May 2016, we began using the algorithm clinically. Throughout the project, a team leader reviewed each case shortly after discharge. The entire algorithm team reviewed selected cases, such as positive or negative outliers or those with unexpected outcomes, and discussed them with inpatient providers. This discussion afforded an opportunity to increase awareness of the new algorithm and solicit suggestions from frontline providers for additional improvements. Subsequently, we tested 2 additional interventions. First, we developed an internal radiology protocol to expedite MRI for patients managed per the algorithm. Second, we provided progress reports, reinforced education, and performed reviews of selected negative outliers with prolonged LOSs with the ID and HM faculty and fellows.

Inpatient charges per episode was our primary outcome measure for cost. We defined inpatient charges as the sum of all charges (provider and facility) generated by our billing software for the initial admission. All charges are presented in US dollars. As a related process measure, we managed inpatient LOS using the standard definition of inpatient time from midnight to midnight. We used algorithm antimicrobial adherence as a surrogate process measure for uptake of the algorithm’s bundle of recommendations. We defined this as receipt of an algorithm-recommended antimicrobial during the first 24 hours after antimicrobial agents were begun and no more than 1 dose of a nonrecommended antimicrobial. As a balancing measure, we managed time to effective antimicrobial therapy (defined as the time from admission to the initiation of a course of an antimicrobial that was effective against the causative organism). We defined a course as receipt of ≥2 consecutive doses of the same antimicrobial. Using standard breakpoints, we defined effective antimicrobial agents as agents to which the patient’s recovered organism tested susceptible.20 We excluded patients in whom no causative organism was identified from this metric.

We used run charts (algorithm antimicrobial adherence) and X and moving range (MR) process control charts (time to effective therapy, LOS, and inpatient charges) to track our results quarterly. We analyzed the data using statistical process control methods.19,21,23 We defined special cause variation as ≥8 consecutive points above or below the mean, any single point outside the control limits, or 2 of 3 consecutive points in the outer one-third of the difference between the centerline and a control limit. To be accepted, such a finding had to be present with a plausible explanation for a system change occurring at that time. Points that fell outside the control limits on the MR companion charts were considered to be operating outside the existing system entirely. Such points were “ghosted” (ie, excluded from the calculation of centerline means, control limits, and special cause trends if a plausible explanation for why they fell outside the usual system could be determined).19

Baseline values were calculated by using 12 months of preintervention data (January 2015–December 2015). We included a 4-month washout period (January 2016–April 2016) to account for possible influences of algorithm development on clinical care. Postintervention data are reported for 13 months (May 2016–May 2017). For the analysis of average inpatient charges, charge subcategories were grouped by using business intelligence software (Business Objects; SAP SE, Walldorf, Germany) on the basis of revenue center codes assigned by EPIC Resolute (Epic Systems Corp, Verona, WI). The change from pre- to postintervention by category was calculated by subtracting the mean preintervention charges per category from the mean postintervention charges per category (a negative number is consistent with cost-savings postintervention).

This QI project was considered exempt from formal review by the Cincinnati Children’s Hospital Medical Center Institutional Review Board.

In total, 57 patients were included; 31 patients were included in the preintervention cohort, and 26 patients were included in the postintervention cohort, of whom 19 were managed per the care algorithm for the duration of their admission. All 7 patients who were taken off the care algorithm during their initial admission were determined to have multifocal disease or a protracted bacteremia. Patient characteristics in each cohort are described in Table 1. Overall, the median patient age was 6.4 years, and 67% of patients were boys. Lower extremity long bones constituted the most common site of infection followed by the pelvis and upper extremity long bones. An organism was recovered in 68% of cases, with Staphylococcus aureus constituting the majority of isolates. Among S aureus isolates, 71.4% were methicillin susceptible, and 65.7% were clindamycin susceptible. Of patients, 44% underwent at least 1 surgical procedure during the admission.

We observed good algorithm adherence, with the percentage of guideline antimicrobial adherence increasing from 0% in the preintervention data to 60% by the project’s end (Fig 1). Overall, mean inpatient charges decreased from $45 718 in the preintervention cohort to$32 895 in the postintervention cohort (Fig 2, Supplemental Fig 6), and this finding met special cause rules. A subcategory analysis of the absolute change in average inpatient charges is shown in Fig 3. Small increases in expenditures for diagnostic imaging and professional billing occurred. All other charge categories, most notably room and board, decreased, resulting in overall savings. LOS appeared to decrease without meeting special cause rules (Fig 4, Supplemental Fig 7). Our baseline mean time to effective therapy was 15 hours and did not reveal any special cause variation (data not shown).

FIGURE 1

Guideline antimicrobial adherence: run chart of the percentage of patients receiving guideline-adherent antimicrobial agents in ascending chronological order. Each point represents a group of 5 consecutive patients. Interventions and their timing are depicted by the yellow boxes. Note that no centerline shift is present because the data did not meet special cause rules.

FIGURE 1

Guideline antimicrobial adherence: run chart of the percentage of patients receiving guideline-adherent antimicrobial agents in ascending chronological order. Each point represents a group of 5 consecutive patients. Interventions and their timing are depicted by the yellow boxes. Note that no centerline shift is present because the data did not meet special cause rules.

Close modal
FIGURE 2

AHO inpatient charges X chart (January 2015–May 2017): statistical process control X chart of mean inpatient charges per patient in ascending chronological order. Each point represents a single patient. Charges are presented in US dollars. Interventions and their timing are depicted by the yellow boxes. A shift in the centerline and the control limits denotes special cause variation. The MR companion chart is shown in Supplemental Fig 6.

FIGURE 2

AHO inpatient charges X chart (January 2015–May 2017): statistical process control X chart of mean inpatient charges per patient in ascending chronological order. Each point represents a single patient. Charges are presented in US dollars. Interventions and their timing are depicted by the yellow boxes. A shift in the centerline and the control limits denotes special cause variation. The MR companion chart is shown in Supplemental Fig 6.

Close modal
FIGURE 3

Subcategory analysis of inpatient charges: subcategory analysis of the change in mean inpatient charges from preintervention to postintervention. A negative number (green bar) is indicative of cost savings postintervention compared with preintervention. a Examples include billing for ancillary services (eg, occupational, physical, vascular access team), nursing procedures, blood products, etc.

FIGURE 3

Subcategory analysis of inpatient charges: subcategory analysis of the change in mean inpatient charges from preintervention to postintervention. A negative number (green bar) is indicative of cost savings postintervention compared with preintervention. a Examples include billing for ancillary services (eg, occupational, physical, vascular access team), nursing procedures, blood products, etc.

Close modal
FIGURE 4

AHO LOS (MN to MN) X chart (January 2015–May 2017): statistical process control X chart of LOS per patient in ascending chronological order. Each point represents a single patient. LOS is presented as an integer of the number of calendar days the patient was present in the hospital. Interventions and their timing are depicted by the yellow boxes. There is no shift in the centerline and the control limits because special cause variation did not occur. The MR companion chart is shown in Supplemental Fig 7. MN, midnight.

FIGURE 4

AHO LOS (MN to MN) X chart (January 2015–May 2017): statistical process control X chart of LOS per patient in ascending chronological order. Each point represents a single patient. LOS is presented as an integer of the number of calendar days the patient was present in the hospital. Interventions and their timing are depicted by the yellow boxes. There is no shift in the centerline and the control limits because special cause variation did not occur. The MR companion chart is shown in Supplemental Fig 7. MN, midnight.

Close modal

Of note, 3 patients (2 in the preintervention cohort and 1 in the postintervention cohort) were ghosted, as described previously, on the LOS chart (Fig 4, Supplemental Fig 7). Similarly, 2 of the same 3 patients were ghosted on the inpatient charges chart (Fig 2, Supplemental Fig 6). An investigation of these instances revealed that in each case, the treatment team had elected to provide long-course intravenous therapy in the inpatient setting.

With our project, we demonstrate a substantial reduction in inpatient charges through standardization using QI methodologies and a robust local consensus process. We successfully accomplished our global aim of reducing the cost of care although we were unable to detect a significant reduction in LOS. The local consensus process itself appeared to be our most powerful intervention.

AHO is a relatively infrequent condition even at a high-volume institution. This results in substantial time lags in assessing both the desired and unintended impacts of interventions. Reviewing individual cases throughout and learning from each patient (ie, “n-of-one testing”) mitigated this problem in our plan-do-study-act cycles. In addition, our results reveal the power of statistical process control methods to quantitatively measure the effects of interventions even in small cohorts. This has important implications for measuring the effects of QI efforts used to target rare conditions that cause significant morbidity, mortality, or expense. Beyond the inherent benefits of standardization, QI reveals promise as an alternative approach for advancing the standard of care because it enables us to discover the most cost-effective combinations of existing diagnostic and treatment modalities.

Randomized controlled trials used to study the treatment of pediatric AHO have proven difficult because of its infrequency and rare poor outcomes.11 Current clinical care is primarily based on clinical experience, expert opinion, and observations from small cohorts. As a result, there have been few consensus guidelines available to guide the management of pediatric AHO, and the approach to management has varied considerably.5,6 Even when national consensus guidelines have been formulated, uptake of the guidelines remains inconsistent.3,24 Using local consensus building, we overcame the standardization challenges associated with the involvement of multiple specialties and the lack of high-quality evidence or national guidelines. This could be an effective alternative approach to national consensus guidelines for conditions for which available scientific evidence is limited, and care is largely based on observational data and expert opinion.

We also noted another instance in which QI methods added specific context information to our project that was not available from current literature. During the consensus process, many providers expressed concerns that several of our recommendations would result in delays in the initiation of effective antimicrobial therapy. Our decision to monitor time to effective antimicrobial therapy as a balancing measure was reassuring to providers. Ultimately, this problem did not materialize.

Our time-series approach establishes a close temporal association between our interventions and the changes observed. Our project took place in a carefully defined population at a single institution. Previous studies have revealed an effect on LOS for osteomyelitis through standardization15 as well as an effect on decreasing both costs25,26 and events that could lead to increased costs, such as catheter-associated bloodstream infections and surgical site infections.27,28 These factors can be used to make a compelling argument for a causal association between our care algorithm implementation and the reductions in charges that we observed.

Specifically, special cause variation for inpatient charges appeared immediately after the initial rollout of the algorithm, suggesting that the consensus process, the initial education efforts, and the ongoing feedback mechanism represented the primary effective interventions. We initiated subsequent interventions after negative outlier reviews that suggested effect decay29 had begun to occur before special cause variation was identified. We speculate that these interventions may have been important in preventing decay because we did not observe any subsequent special cause in either direction. The impact of individual key decisions on inpatient charges and LOS were likely small, with our findings resulting from their impact in aggregate. On the basis of an analysis of the charge subcategories, we speculate that most of our charge savings were related to an effect on LOS that was below the threshold of detection for our chosen LOS measure and sample size. Increased LOS is associated with increased hospital charges in many circumstances.30,31 With our findings, we suggest that LOS may be a lagging measure of cost, and inpatient charges may, in practice, be a more sensitive measure of incremental LOS-related costs.

Our small sample size precludes a statistically meaningful analysis of the individual charge subcategories. In addition to the indirect effects of a reduced LOS (ie, patients discharged from the hospital cannot receive inpatient pharmaceuticals or laboratory testing), it is worth considering that our standardization of laboratory testing and antimicrobial agents may also have contributed to charge reductions in those categories through reduced intensity or redundancy in testing and treatment. Similarly, our algorithm recommended increased imaging and subspecialty consultation, and both of those subcategories revealed modest increases in charges. Surgical management was not substantially addressed in our algorithm, and our surgical rates were similar pre- and postintervention. Therefore, although the charge reduction in this category appears large, we conclude that it most likely reflects common cause variation.

We do note some limitations to the generalizability of our findings. First, our project took place at a single institution in a carefully selected population. Nonetheless, our cohort demographics are consistent with those of previously published cohorts,1,2,6,12,32 and our recommendation for early transition to oral therapy is widely supported by evidence from other studies.5,6,8,11 Second, although consensus building and the Model for Improvement have been used successfully in a diverse array of circumstances, context is believed to affect organization and practice change, and our project took place in an institution with a long-standing culture of QI.14,33,34 Third, the magnitude of expected change in inpatient charges could vary substantially across institutions.

With respect to internal validity, our approach to case identification and data extraction poses a risk of both selection and differential detection bias. Retrospectively, we relied on clinical documentation, which can be incomplete. Prospectively, we were unable to blind our assessors, who may have been aware of information not included in the chart. We attempted to mitigate this concern using objective inclusion criteria, but some remained partially subjective. Our primary outcome of inpatient charges is objective and was extracted via billing software and so should not be subject to detection bias. Additionally, bias is unlikely to entirely explain our findings given the magnitude of change observed. Discovering additional episodes of special cause in our time series will clarify this issue, and we plan ongoing surveillance.

Our use of statistical process control methods and n-of-one testing shortened the lag time in measuring the effects of interventions on infrequent outcomes in a high-cost, low-volume condition. Furthermore, a robust local consensus process overcame the standardization challenges posed by a lack of high-quality evidence or national consensus standards and the need for cooperation among diverse groups of specialists. Our project reveals the feasibility and effectiveness of combining a robust local consensus, statistical process control measurement, and n-of-one testing to safely decrease the cost of care for pediatric AHO. Cost savings and improved clinical care are not mutually exclusive aims when change is undertaken in a systematic fashion with measurement in place.

• AHO

acute hematogenous osteomyelitis

•
• HM

hospital medicine

•
• ID

infectious diseases

•
• LOS

length of stay

•
• MR

moving range

•
• QI

quality improvement

Dr Robinette conceptualized and designed the quality improvement project and acted as the project team leader, designed the data collection instruments, collected, analyzed, and interpreted the data, drafted the initial manuscript, and reviewed and revised the manuscript; Dr Brower designed the data collection instruments and reviewed and revised the manuscript; Dr Schaffzin assisted with the initial drafting of the manuscript and reviewed it critically for important intellectual content; Drs Whitlock, Shah, and Connelly conceptualized and designed the quality improvement project and revised and reviewed the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

We thank Judy Spenlau for help with collecting and compiling financial data and Andrew Beck, MD, for help with preparing the article.

1
Dartnell
J
,
Ramachandran
M
,
Katchburian
M
.
Haematogenous acute and subacute paediatric osteomyelitis: a systematic review of the literature.
J Bone Joint Surg Br
.
2012
;
94
(
5
):
584
595
[PubMed]
2
Saavedra-Lozano
J
,
Mejías
A
,
N
, et al
.
Changing trends in acute osteomyelitis in children: impact of methicillin-resistant Staphylococcus aureus infections.
J Pediatr Orthop
.
2008
;
28
(
5
):
569
575
[PubMed]
3
Calvo
C
,
Núñez
E
,
Camacho
M
, et al;
Collaborative Group
.
Epidemiology and management of acute, uncomplicated septic arthritis and osteomyelitis: Spanish multicenter study.
Pediatr Infect Dis J
.
2016
;
35
(
12
):
1288
1293
[PubMed]
4
Maraqa
NF
,
Gomez
MM
,
Rathore
MH
.
Outpatient parenteral antimicrobial therapy in osteoarticular infections in children.
J Pediatr Orthop
.
2002
;
22
(
4
):
506
510
[PubMed]
5
Zaoutis
T
,
Localio
AR
,
Leckerman
K
,
S
,
Bertoch
D
,
Keren
R
.
Prolonged intravenous therapy versus early transition to oral antimicrobial therapy for acute osteomyelitis in children.
Pediatrics
.
2009
;
123
(
2
):
636
642
[PubMed]
6
Keren
R
,
Shah
SS
,
Srivastava
R
, et al;
Pediatric Research in Inpatient Settings Network
.
Comparative effectiveness of intravenous vs oral antibiotics for postdischarge treatment of acute osteomyelitis in children.
JAMA Pediatr
.
2015
;
169
(
2
):
120
128
[PubMed]
7
Bachur
R
,
Pagon
Z
.
Success of short-course parenteral antibiotic therapy for acute osteomyelitis of childhood.
Clin Pediatr (Phila)
.
2007
;
46
(
1
):
30
35
[PubMed]
8
Jaberi
FM
,
Shahcheraghi
GH
,
M
.
Short-term intravenous antibiotic treatment of acute hematogenous bone and joint infection in children: a prospective randomized trial.
J Pediatr Orthop
.
2002
;
22
(
3
):
317
320
[PubMed]
9
Jagodzinski
NA
,
Kanwar
R
,
Graham
K
,
Bache
CE
.
Prospective evaluation of a shortened regimen of treatment for acute osteomyelitis and septic arthritis in children.
J Pediatr Orthop
.
2009
;
29
(
5
):
518
525
[PubMed]
10
Le Saux
N
,
Howard
A
,
Barrowman
NJ
,
Gaboury
I
,
Sampson
M
,
Moher
D
.
Shorter courses of parenteral antibiotic therapy do not appear to influence response rates for children with acute hematogenous osteomyelitis: a systematic review.
BMC Infect Dis
.
2002
;
2
:
16
[PubMed]
11
Peltola
H
,
Pääkkönen
M
,
Kallio
P
,
Kallio
MJ
;
Osteomyelitis-Septic Arthritis Study Group
.
Short- versus long-term antimicrobial treatment for acute hematogenous osteomyelitis of childhood: prospective, randomized trial on 131 culture-positive cases.
Pediatr Infect Dis J
.
2010
;
29
(
12
):
1123
1128
[PubMed]
12
Peltola
H
,
Unkila-Kallio
L
,
Kallio
MJ
;
the Finnish Study Group
.
Simplified treatment of acute staphylococcal osteomyelitis of childhood.
Pediatrics
.
1997
;
99
(
6
):
846
850
[PubMed]
13
Vinod
MB
,
Matussek
J
,
Curtis
N
,
Graham
HK
,
Carapetis
JR
.
Duration of antibiotics in children with osteomyelitis and septic arthritis.
J Paediatr Child Health
.
2002
;
38
(
4
):
363
367
[PubMed]
14
PW
,
Brinkman
WB
,
Simmons
JM
, et al
.
Oral antibiotics at discharge for children with acute osteomyelitis: a rapid cycle improvement project.
BMJ Qual Saf
.
2014
;
23
(
6
):
499
507
[PubMed]
15
Spruiell
MD
,
Searns
JB
,
Heare
TC
, et al
.
Clinical care guideline for improving pediatric acute musculoskeletal infection outcomes.
J Pediatric Infect Dis Soc
.
2017
;
6
(
3
):
e86
e93
[PubMed]
16
Nsubuga
P
,
White
ME
,
Thacker
SB
, et al
. In:
Jamison
DT
,
Breman
JG
,
Measham
AR
, eds.
Disease Control Priorities in Developing Countries
, 2nd ed.
Washington, DC
:
The International Bank for Reconstruction and Development/The World Bank
;
2006
17
Langley
GJ
,
Moen
RD
,
Nolan
KM
,
Nolan
TW
,
Norman
CL
,
Provost
LP
.
The Improvement Guide: A Practical Approach to Enhancing Organizational Performance
, 2nd ed.
San Francisco, CA
:
Jossey-Bass
;
2009
18
Lloyd
R
.
On Demand: An Introduction to the Model for Improvement
.
Boston, MA
:
Institute for Healthcare Improvement
;
2016
19
Provost
LP
,
Murray
SK
.
The Health Care Data Guide: Learning From Data for Improvement
.
San Francisco, CA
:
Jossey-Bass
;
2011
20
Clinical and Laboratory Standards Institute
.
Performance Standards for Antimicrobial Susceptibility Testing
. 27th ed.
Wayne, PA
:
Clinical and Laboratory Standards Institute
;
2017
21
Carey
RG
.
How do you know that your care is improving? Part II: using control charts to learn from your data.
J Ambul Care Manage
.
2002
;
25
(
2
):
78
88
[PubMed]
22
Carey
RG
.
How do you know that your care is improving? Part I: basic concepts in statistical thinking.
J Ambul Care Manage
.
2002
;
25
(
1
):
80
87
23
Benneyan
JC
.
Use and interpretation of statistical quality control charts.
Int J Qual Health Care
.
1998
;
10
(
1
):
69
73
[PubMed]
24
Saavedra-Lozano
J
,
Calvo
C
,
Huguet Carol
R
, et al
.
SEIP-SERPE-SEOP consensus document on the treatment of uncomplicated acute osteomyelitis and septic arthritis [in Spanish].
An Pediatr (Barc)
.
2015
;
82
(
4
):
273.e1
273.e10
25
Lee
VS
,
Kawamoto
K
,
Hess
R
, et al
.
Implementation of a value-driven outcomes program to identify high variability in clinical costs and outcomes and association with reduced cost and improved quality.
JAMA
.
2016
;
316
(
10
):
1061
1072
[PubMed]
26
James
BC
,
Savitz
LA
.
How Intermountain trimmed health care costs through robust quality improvement efforts.
Health Aff (Millwood)
.
2011
;
30
(
6
):
1185
1191
[PubMed]
27
Schaffzin
JK
,
Harte
L
,
Marquette
S
, et al
.
Surgical site infection reduction by the solutions for patient safety hospital engagement network.
Pediatrics
.
2015
;
136
(
5
). Available at: www.pediatrics.org/cgi/content/full/136/5/e1353
[PubMed]
28
Nuckols
TK
,
Keeler
E
,
Morton
SC
, et al
.
Economic evaluation of quality improvement interventions for bloodstream infections related to central catheters: a systematic review.
JAMA Intern Med
.
2016
;
176
(
12
):
1843
1854
[PubMed]
29
Luria
JW
,
Muething
SE
,
Schoettker
PJ
,
Kotagal
UR
.
Reliability science and patient safety.
Pediatr Clin North Am
.
2006
;
53
(
6
):
1121
1133
[PubMed]
30
Kato
MG
,
Erkul
E
,
Nguyen
SA
, et al
.
Extracapsular dissection vs superficial parotidectomy of benign parotid lesions: surgical outcomes and cost-effectiveness analysis.
.
2017
;
143
(
11
):
1092
1097
[PubMed]
31
Korbel
L
,
Easterling
RS
,
Punja
N
,
Spencer
JD
.
The burden of common infections in children and adolescents with diabetes mellitus: a Pediatric Health Information System study.
Pediatr Diabetes
.
2018
;
19
(
3
):
512
519
[PubMed]
32
McNeil
JC
,
Forbes
AR
,
Vallejo
JG
, et al
.
Role of operative or interventional radiology-guided cultures for osteomyelitis.
Pediatrics
.
2016
;
137
(
5
):
e20154616
[PubMed]
33
Kaplan
HC
,
PW
,
Dritz
MC
, et al
.
The influence of context on quality improvement success in health care: a systematic review of the literature.
Milbank Q
.
2010
;
88
(
4
):
500
559
[PubMed]
34
Tomoaia-Cotisel
A
,
Scammon
DL
,
Waitzman
NJ
, et al
.
Context matters: the experience of 14 research teams in systematically reporting contextual factors important for practice change.
Ann Fam Med
.
2013
;
11
(
suppl 1
):
S115
S123
[PubMed]

## Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.