Article: Mosquera RA, Avritscher EBC, Pedroza C, Bell CS, Samuels CL, Harris TS, Eapen JC, Yadav A, Poe M, Parlar-Chun RL, Berry J, Tyson JE. Hospital Consultation From Outpatient Clinicians for Medically Complex Children: A Randomized Clinical Trial. JAMA Pediatr. 2021;175(1):e205026.
Fellow: Kristin Kalita, MD; University of Colorado
Summary: Though the percentage of children with medical complexity (CMC) is low among all children, CMC account for an increasing percentage of children admitted to US hospitals. Admission often leads to discontinuity of care for CMC, which is likely a factor in the higher hospital charges and higher readmission rates among CMC. This single-center, randomized quality improvement trial assessed whether a hospital consultation (HC) service from outpatient physicians from the institution’s comprehensive care (CC) clinic for high-risk CMC reduced hospital days compared with usual hospital care (UHC). In the study, 342 CMC that received outpatient care in the CC clinic were randomized to either HC (n = 167) or UHC (n = 175). HC included discussions between the CC physicians and emergency providers about need for admission, recommendations from the CC physician to the inpatient teams caring for admitted CMC, assistance with discharge preparation and coordination of discharge follow up by the CC physician, and conversations between the CC physician and the parents of admitted CMC when possible. The primary outcome was total hospital days per child-year. Hospitalizations, mean hospital length of stay, pediatric intensive care unit (PICU) admissions, PICU days, serious illnesses, emergency department visits within 30 days after discharge, 30-day readmissions, parental rating of hospital clinicians, and health system costs were all secondary outcomes.
Bayesian analyses were conducted. Bayesian analyses are conceptualized around the probability of benefit (or harm) of an intervention rather than around the idea of a result being significant versus non-significant, as in frequentist (more traditional) statistics. Bayesian risk ratios are also calculated along with credible intervals. Credible intervals are similar to confidence intervals but are not meant to be interpreted strictly by whether or not they include 1.
Using a conservative prior probability that assumed there would be no effect of the intervention, the group found a 91% probability that HC reduced hospital days compared with UHC (Bayesian rate ratio [RR], 0.61; 95% credible interval [CrI], 0.30-1.26). They also found a 98% probability that HC reduced hospitalizations (RR, 0.68; 95% CrI, 0.48-0.97), an 89% probability of reduced PICU days (RR, 0.59; 95% CrI 0.26-1.38), and a 94% reduction in mean total health system costs (cost ratio, 0.67; 95% CrI, 0.41-1.10). In a secondary analysis using a less conservative prior probability based on expert opinion, the probability that HC reduced hospital days compared with UHC was 96%, and the probability of reduced total health system costs was 97%.
Strengths: This study had a large sample size that included all of the patients within the CC clinic for high-risk CMC who met enrollment criteria, and the randomization process was clearly outlined. The group performed an intention-to-treat analysis, and there were similar numbers of patients in each group (HC or UCH) for whom the intervention was discontinued due to moving out of the area, transitioning out of the clinic, changing insurance, or death. Finally, the fact that the group performed a primary analysis using a conservative prior probability and a secondary analysis of the total hospital days and total health system costs using a more optimistic prior probability determined by expert opinion allows for a broader acceptance of the intervention’s effect given the high probability of benefit seen in both analyses.
Limitations: Firstly, this was an intervention performed by a single institution that had a robust outpatient program for CMC prior to the initiation of the HC service and that has a unique population, which may limit its generalizability to other institutions. Additionally, the lack of a discussion about the costs of the HC service itself and the time commitment of the providers inhibits a full understanding of the resources needed to maintain such a program. Finally, the use of Bayesian statistics, though appropriate for this study, may limit the buy-in of hospital administrators who may be more accustomed to interpreting frequentist statistics and may therefore be more hesitant to act upon findings that utilize more novel analyses.
Major Takeaway: The involvement of an HC service provided by outpatient physicians of high-risk CMC resulted in a high probability of reduced hospital days and PICU days, fewer hospitalizations, and a reduction in mean total health system costs when compared to UHC.
How this article should impact our practice: Hospitals should consider the addition of an inpatient consultation service from outpatient physicians of high-risk CMC as a way to improve outcomes for these children and reduce overall medical costs. Hospitalists could have a role in advocating for such services. Even in hospitals without the means to provide such consultation services, hospitalists could use the results from this study as a reminder of the importance of involving primary care physicians in the inpatient care of CMC.
Fellow: Kristin Kalita, MD; University of Colorado
Summary: Though the percentage of children with medical complexity (CMC) is low among all children, CMC account for an increasing percentage of children admitted to US hospitals. Admission often leads to discontinuity of care for CMC, which is likely a factor in the higher hospital charges and higher readmission rates among CMC. This single-center, randomized quality improvement trial assessed whether a hospital consultation (HC) service from outpatient physicians from the institution’s comprehensive care (CC) clinic for high-risk CMC reduced hospital days compared with usual hospital care (UHC). In the study, 342 CMC that received outpatient care in the CC clinic were randomized to either HC (n = 167) or UHC (n = 175). HC included discussions between the CC physicians and emergency providers about need for admission, recommendations from the CC physician to the inpatient teams caring for admitted CMC, assistance with discharge preparation and coordination of discharge follow up by the CC physician, and conversations between the CC physician and the parents of admitted CMC when possible. The primary outcome was total hospital days per child-year. Hospitalizations, mean hospital length of stay, pediatric intensive care unit (PICU) admissions, PICU days, serious illnesses, emergency department visits within 30 days after discharge, 30-day readmissions, parental rating of hospital clinicians, and health system costs were all secondary outcomes.
Bayesian analyses were conducted. Bayesian analyses are conceptualized around the probability of benefit (or harm) of an intervention rather than around the idea of a result being significant versus non-significant, as in frequentist (more traditional) statistics. Bayesian risk ratios are also calculated along with credible intervals. Credible intervals are similar to confidence intervals but are not meant to be interpreted strictly by whether or not they include 1.
Using a conservative prior probability that assumed there would be no effect of the intervention, the group found a 91% probability that HC reduced hospital days compared with UHC (Bayesian rate ratio [RR], 0.61; 95% credible interval [CrI], 0.30-1.26). They also found a 98% probability that HC reduced hospitalizations (RR, 0.68; 95% CrI, 0.48-0.97), an 89% probability of reduced PICU days (RR, 0.59; 95% CrI 0.26-1.38), and a 94% reduction in mean total health system costs (cost ratio, 0.67; 95% CrI, 0.41-1.10). In a secondary analysis using a less conservative prior probability based on expert opinion, the probability that HC reduced hospital days compared with UHC was 96%, and the probability of reduced total health system costs was 97%.
Strengths: This study had a large sample size that included all of the patients within the CC clinic for high-risk CMC who met enrollment criteria, and the randomization process was clearly outlined. The group performed an intention-to-treat analysis, and there were similar numbers of patients in each group (HC or UCH) for whom the intervention was discontinued due to moving out of the area, transitioning out of the clinic, changing insurance, or death. Finally, the fact that the group performed a primary analysis using a conservative prior probability and a secondary analysis of the total hospital days and total health system costs using a more optimistic prior probability determined by expert opinion allows for a broader acceptance of the intervention’s effect given the high probability of benefit seen in both analyses.
Limitations: Firstly, this was an intervention performed by a single institution that had a robust outpatient program for CMC prior to the initiation of the HC service and that has a unique population, which may limit its generalizability to other institutions. Additionally, the lack of a discussion about the costs of the HC service itself and the time commitment of the providers inhibits a full understanding of the resources needed to maintain such a program. Finally, the use of Bayesian statistics, though appropriate for this study, may limit the buy-in of hospital administrators who may be more accustomed to interpreting frequentist statistics and may therefore be more hesitant to act upon findings that utilize more novel analyses.
Major Takeaway: The involvement of an HC service provided by outpatient physicians of high-risk CMC resulted in a high probability of reduced hospital days and PICU days, fewer hospitalizations, and a reduction in mean total health system costs when compared to UHC.
How this article should impact our practice: Hospitals should consider the addition of an inpatient consultation service from outpatient physicians of high-risk CMC as a way to improve outcomes for these children and reduce overall medical costs. Hospitalists could have a role in advocating for such services. Even in hospitals without the means to provide such consultation services, hospitalists could use the results from this study as a reminder of the importance of involving primary care physicians in the inpatient care of CMC.