The Medical Home and Hospital Readmissions

This prospective cohort study enrolled randomly selected pediatric patients and their caregivers during an acute hospitalization lasting >24 hours from October 1, 2012, to January 1, 2014, at a 100-bed tertiary children’s hospital. Participants were enrolled from any medical or surgical service on the inpatient ward or PICU. Inpatients who were hospitalized for <24 hours, transferred to other facilities, died, were age >18 years, or were in newborn nursery or NICU were excluded. These exclusions were chosen because such hospitalizations were assumed to have distinct primary care relationships and discharge transition processes that were less relevant to the study hypothesis. Patients previously enrolled in the study were excluded from reenrollment; therefore, children were in the data set only once.

The enrollment hospitalization was defined as the index hospitalization. The primary outcome, unplanned 30-day readmission, was defined as an unplanned hospitalization to any hospital occurring within 30 days of the index discharge, identified through either parent report or administrative sources. We excluded readmissions for chemotherapy, radiation, dialysis, rehabilitation, or labor and delivery. Planned readmissions were further identified and excluded based on 2 factors: (1) parents reporting readmissions as planned and (2) chart review of index discharge summaries noting that rehospitalization was scheduled over the subsequent 30 days. When discrepancies existed between parent report and discharge summaries, parental responses were chosen as correct for 2 reasons. First, children could have an unplanned readmission despite having a separate scheduled readmission; and second, the discharge summary does not reflect decisions made to schedule a hospitalization subsequent to the index discharge.

The secondary outcome, 7-day ED visits, was defined as any ED visit within 7 days of the index discharge that did not result in an admission. This duration was chosen based on an assumption that ED visits within 7 days of discharge would more likely relate to deficiencies in discharge processes or follow-up care. Readmissions and ED visits within the study site’s health system were identified from administrative data, whereas utilization outside the health system was identified through parent report.

The predictor of interest was primary care experience before index hospitalization. Patients were categorized as having or not having a medical home, based on the Maternal and Child Health Bureau’s (MCHB’s) definition of a medical home using questions from the 2005/2006 National Survey of Children with Special Healthcare Needs.¹⁸ To better understand relationships between the medical home concept and outcomes, we examined each of the 5 medical home components: personal doctor or nurse, usual source for sick and well care, family-centered care, no problem with referrals, and coordinated care. To be categorized as a medical home, each of these components needs to be experienced when relevant. A detailed description of this measure is available elsewhere.¹⁸ We also sought to identify individual relationships between medical home components and utilization.

To adjust for potential confounders, we included covariates commonly used in previous studies: age, gender, race/ethnicity, primary language, and payer category.^3,5 Medical complexity was accounted for through presence of complex chronic conditions,^19,20 technology assistance,²¹ seeing ≥3 subspecialists,^21,22 length of index stay, and number of hospitalizations and ED visits in the year before enrollment. Additional variables hypothesized to associate with both readmissions and medical home prevalence were included: distance between home and the children’s hospital (calculated from zip code), primary care provider (PCP) scope (general pediatrician or subspecialist), and parent confidence to avoid a readmission or ED visit after discharge.

Parent confidence was assessed by using questions developed for this study. Caregivers were asked to rate “How confident are you that [child’s name] will stay out of the hospital for the next 30 days?”, with instructions to refer to unplanned hospital visits only. Responses were reported on a 4-point Likert scale (1 = very confident, 2 = somewhat confident, 3 = somewhat unconfident, 4 = not very confident, or “don’t know/don’t remember/not applicable”). For primary analyses, responses were dichotomized into very confident (ie, “1”) or not very confident (ie, “2–4”) and we conducted a secondary analysis by using the entire 4-point Likert scale. A similar question was asked for ED visits.

Research staff (AS) assigned computer-generated random numbers to patients admitted the previous calendar day. Families were approached sequentially, following random order, until either the daily enrollment target was reached or there were no additional patients to approach. The research team aimed to enroll 3 new participants daily, from the average of 10 admissions,⁷ to complete enrollment within the study time frame.

After obtaining informed consent, trained research assistants collected parent-reported data using structured questionnaires in English or Spanish, according to parent preference. Questions related to the medical home were asked in person during the index admission in either the patient’s room or a close private location. Parent confidence and parent-reported ED visit or readmission outcomes were collected through 2 separate phone calls within 7 days (confidence) and after 30 days (outcomes) after the index discharge. As in previous work,²³ parent confidence questions were asked after discharge to avoid biasing clinical care and revisit risk. Administrative and chart review data were gathered and merged with parent-reported data at study conclusion.

Thirty-day readmission rates were estimated from 2008 to 2010 data.⁷ Assuming 44.8% medical home prevalence among California’s CSHCN from the 2005/2006 National Survey of Children with Special Healthcare Needs,²⁴ which applied the same medical home definition, our study had 83% power to detect a difference in readmissions rates of 8 percentage points between those with and without a medical home with 700 children enrolled.

Descriptive statistics characterized relevant covariates between those with and without a medical home. Bivariate followed by multivariate logistic regression models were created with each outcome and covariates having bivariate P values ≤.2 or believed to be conceptually essential. Because complex chronic conditions included medical technology codes,²⁰ we did not include technology assistance as a separate variable in our multivariate models. We verified the absence of collinearity between included variables before constructing the final models. All analyses were completed by using Stata (SE version 13.1, College Station, TX).

This study was approved by the institutional review boards at the University of California–Los Angeles and the University of Wisconsin.

From October 2012 to January 2014, we approached 816 potential participants and enrolled 701 (86%), 99% of whom had administrative data available and 97% of whom completed all follow-up questionnaires (Fig 1). The 30-day unplanned readmission rate was 12.4% and the 7-day ED visit rate was 5.6%. Variation in readmission rates based on data sources and definition strategies are illustrated in Fig 2. Discharge summaries described readmission plans for 11% of the readmissions identified by administrative data. Among readmissions reported by parents as planned, index discharge summaries indicated a plan for readmission only 20% of the time. There was 93.9% agreement on all-cause parent-reported readmissions to our institution and administratively identified readmissions to our institution. Supplemental Tables 3 and 4 summarize the most common all-patient refined diagnostic related groups (APR-DRGs) and complex chronic conditions (CCCs) for index hospitalizations and unplanned readmissions. The most common index APR-DRG was for seizures (24.3%), and that for readmission was postoperative, posttraumatic, or other device infections (7.5%).

Most participants (65.3%) did not have a medical home using the study definition. Despite almost all patients without a medical home reporting having a personal provider (96.7%), only 65% had a usual source for sick and well care. Nearly 98% of the sample did have an appropriate usual source for well care. The majority of patients without a usual source for sick and well care were categorized as such because they reported using the ED as their usual place for sick care. The medical home component reported least often was coordinated care (33.9%).

Participant characteristics, stratified by presence or absence of a MCHB-defined medical home, are summarized in Table 1. Patients with a medical home more often had commercial insurance, received primary care from a generalist, were less medically complex, and were more confident to avoid a readmission or an ED visit at the time of discharge.

In unadjusted models (Table 2), readmissions were significantly associated with age, race/ethnicity, PCP scope (generalist versus subspecialist), medical complexity, and confidence to avoid readmissions or ED visits. Distance to the children’s hospital was not associated with readmissions. The highest readmission odds were among those with ≥2 hospitalizations in the past 12 months (odds ratio [OR] 3.40, 95% confidence interval [CI] 2.04–5.68 versus those with no hospitalizations) and those who were not very confident to avoid readmission (OR 3.14, 95% CI 1.92–5.14 versus those who were very confident). In unadjusted models, ED visits were significantly associated with age, selected medical complexity variables (technology assistance, hospitalizations in the past 12 months, and length of stay), and confidence to avoid readmissions or ED visits.

Readmissions and ED visits did not have statistically significant bivariate associations with the MCHB medical home composite measure, so attention was turned to the medical home components. Readmissions were significantly associated with a key component: having a usual source for well and sick care (OR 0.39, 95% CI 0.25–0.63 versus those with no usual source for well and sick care).

In multivariate models (Fig 3), lower readmission odds remained for those with a usual source for sick and well care compared with those without (adjusted OR 0.54, 95% CI 0.30–0.96). Higher readmission odds were also observed among those with a subspecialist PCP, longer length of index stay, ≥2 hospitalizations in past 12 months, and lack of confidence to avoid a readmission. In the final model, ED visits were associated only with age and lack of confidence to avoid an ED visit.

Although the medical home concept has become a standard of care for pediatrics,²⁵ uncertainty remains as to its impact on certain outcomes such as hospital utilization.^26,–28 To our knowledge, this is the first study outside of the asthma population to explore relationships between early postdischarge hospital utilization and medical home experience.

The definitions used in this study (which combined parent report and administrative data, included readmissions or ED visits to any hospital, and identified unplanned readmissions through multiple modalities) may have provided a higher degree of validity than definitions often used in other studies. These differences in methodology likely explain our observed readmission rate being higher than other published rates (∼6%).⁴ Our findings demonstrate the wide range in readmission rates possible depending on how data are collected and whether definitions include readmissions to other hospitals.

When using administrative data and readmissions only to our institution, our institutional readmission rate was as low as 7%. However, based on our review of index hospital discharge summaries, >10% of those may have actually been planned. On the other hand, the readmission rate could be nearly 15% when including parent-reported readmissions to any institution. This finding surprised us, given our underlying assumption that highly subspecialized populations seek acute care at single institutions. Less densely populated regions with fewer hospitals than Los Angeles might have a different experience with this phenomenon and observe fewer readmissions to different hospitals.

Because most hospitals do not have access to payer data, and because visits to other hospitals appear to be an important dimension of postdischarge utilization, our findings are relevant for health systems and accountable care organizations. Parent report may be one of the few practical ways to obtain data on readmissions to other hospitals, and it could be an effective way to distinguish planned from unplanned utilization. Our findings suggest agreement between parent-reported and administratively identified readmissions; however, discrepancies did occur, and future work should determine the most reliable way to collect data on such outcomes.

One particular component of the MCHB medical home definition, having a usual source for well and sick care, was associated with fewer unplanned 30-day readmissions. Our findings are consistent with a recent study looking at medical home quality measured by the Parent’s Perception of Primary Care tool and asthma readmissions within 1 year.²⁹ In that study, children with the poorest access, measured by availability of routine care, sick care, evening/weekend advice, and ease of travel to PCP, had the highest readmission risk. The investigators similarly found no associations with other dimensions of medical home quality as measured by the tool: continuity, contextual knowledge, communication, comprehensiveness, or coordination of care.

As noted above, nearly one-quarter of our sample did not meet the definition for having a usual source for well and sick care because they reported using the ED as their usual place for sick care. A pediatric study across multiple institutions in Michigan comparably found that despite >90% reporting having a PCP, nearly 24% reported using the ED as their usual source for sick care.³⁰ This prompts the question as to what extent using the ED for sick care is an explanatory variable behind variations in readmission rates. Conceivably, a patient recently discharged from the hospital who presents to the ED with an acute problem may be more likely to be readmitted compared with a similar patient presenting to their PCP’s office.

Recent work has suggested that non-ED outpatient encounters occurring 4 to 29 days after discharge are associated with fewer readmissions among children with medical complexity.¹² There are limitations to important contextual details from these data, although they could suggest that outpatient clinic follow-up outside of the ED, whether urgent or nonurgent, may be associated with less postdischarge utilization. Continuity with pediatric providers has been associated with fewer ED visits, hospitalizations,³¹ and potentially preventable hospitalizations,³² but this question remains largely unexplored for readmissions.

One key unknown is whether primary care factors (eg, night and weekend hours) or patient/family factors (eg, severity of acute illnesses) make certain families more likely to use the ED as their usual source for sick care (and simultaneously increase the likelihood of being admitted). Convenience, appointment wait time, and satisfaction with the PCP are some variables proposed to drive ED use for less urgent problems among patients with regular sources of primary care.^33,–35 Enhanced primary care access (particularly evening hours) has been associated with less ED use among children more generally.³⁶ Conversely, perception of an acute problem’s severity may be more related to ED use than having a regular source for primary care.³⁷ 

Future work could determine where patients seek care for acute problems shortly after discharge (primary care practice or ED) and whether patients with similar levels of acute illness severity have different risks for being readmitted based on the setting in which they first present. With these scenarios in mind, it was interesting that in adjusted regressions, our 7-day ED visit rates were no different between those with and without a usual source for care. It is possible, given that the ED outcome was a secondary outcome, that we were underpowered to find a difference.

There are several possible explanations for why our initial hypothesis, that children with a medical home before hospitalization would have fewer readmissions or ED visits after discharge, was not confirmed. First, aside from access to care, many of the medical home concepts may simply not be related to postdischarge utilization. Second, the measure we used to characterize medical home experience might not have fully captured the concept. Although the MCHB medical home definition was designed to be used as a dichotomous measure, we did consider creating a medical home scale to assign higher scores to participants with more of the medical home components. We observed fewer readmissions and ED visits among children experiencing more medical home components; however, we did not pursue this strategy further because the components were not well correlated in our sample and we had concerns about the scale’s reliability. Third, many other approaches have been designed to measure “medical home-ness,” each having its own strengths and weaknesses.³⁸ Different results might have been obtained if different medical home definitions or measures of primary care quality had been used.

Mounting research highlights the impact of factors outside the hospital on readmission rates, such as sociodemographics and outpatient experiences.^39,40 Although we did not find a relationship between distance to hospital care and postdischarge utilization, we did find associations with PCP scope of practice (generalist versus subspecialist) and parent confidence. Based on our study design, one cannot conclude that there was a difference in quality of services between primary and subspecialty PCPs, since report of a subspecialty PCP was also highly associated with each medical complexity indicator. Differences in complexity might explain these differences in readmission rates.

We were surprised to note the strength and independence of the association of parent confidence on both readmissions and ED visits. Furthermore, our question, when analyzed using the 4-point Likert scale, demonstrated a dose–response relationship with readmission and ED odds (each category of lower confidence was associated with a higher odds of either outcome), providing support that the degree of parent confidence may predict postdischarge utilization. Confidence likely represents an aspect of the more broad discharge readiness concept,⁴¹ which appears to predict less difficulty with parental coping, which in turn predicts less use of postdischarge health services.

The observation that parents may be able to identify whether their child is at high risk is reminiscent of the finding by Berry et al²³ that parents who strongly agreed that their child was healthy enough to leave the hospital had significantly lower likelihood of readmission. Similarly, parents who strongly agreed that they understood how to manage their child’s health over the weeks after going home and felt comfortable contacting their child’s PCP for help had lower bivariate readmission odds.

These findings are important because parents may be well positioned to identify high-risk patients, and such information would be very feasible for inpatient providers to collect. Parents lacking confidence at discharge may even be amenable to interventions designed to mitigate such risk of readmission or early ED visits.

Finally, substantially less research to date has focused on early postdischarge ED visits than readmissions. Readmissions and ED visits are at times combined into a single revisit measure,^42,–44 the appropriateness of which is not well documented. Inconsistent or even disparate impacts on ED visits and readmissions from interventions designed to reduce utilization after hospital discharge suggest they may be less related outcomes.¹¹ 

In this tertiary single-center study with a high prevalence of complex chronic conditions, generalizability is limited. The observational design precludes causality determinations. Parent-report data are subject to response biases, and certain measures could not be verified, such as utilization reported outside of the study’s health system. It is possible that the study selectively enrolled families with characteristics that allowed them to be available at the bedside for consent and simultaneously influenced primary care experience and postdischarge utilization. Analyses did not include care coordination activities during hospitalization or postdischarge variables that might have also influenced utilization, such as outpatient follow-up visits. The study may have been underpowered for the ED visit outcome, and further work with larger samples across multiple sites would be beneficial to confirm findings. Parent confidence questions were developed for this study, and their psychometric properties are unknown. Ensuring that no ED visits occurred before asking the ED visit confidence questions was not possible; however, these data were collected within 72 hours of discharge for >80% of patients with a 7-day ED visit. Future work should ensure that these questions are asked as close to discharge as possible and before any ED visits.

Any attempt to measure the medical home concept has limitations. Our measure was chosen because it has face validity, allows comparison with state and national data, and was feasible to use with families admitted to the hospital. Although the original questions were designed to be asked by telephone, in-person interviews were not difficult. We do not know how assessment of primary care experience might be influenced by an acute hospitalization, and responses might be different when asked at a time that the child is clinically stable. We also note that our medical home prevalence was lower than rates published from California; however, we were not surprised given that the study inherently selected for sicker and more medically complex children dependent on tertiary care as opposed to a population-based sample.

Institutional readmission rates may vary widely based on definitions and data sources. Using the ED as a usual source for sick care is associated with unplanned 30-day readmissions. Recognizing barriers and facilitators that families and medical home providers face trying to provide a usual source of sick care is an important next step to redesigning systems to prevent postdischarge utilization and better implement the medical home ideal. Other medical home components do not appear to be associated with ED or hospital utilization shortly after discharge. Parent confidence at the time of discharge is strongly associated with ED visits and readmissions, and its assessment may be used to identify high-risk patients or target specific readmission reduction efforts. Interventions informed by these findings could focus on crisis or access planning as well as supporting parent self-efficacy.

APR-DRG

all-patient refined diagnostic related group

CI

confidence interval

ED

emergency department

MCHB

Maternal and Child Health Bureau

OR

odds ratio

PCP

primary care provider