Long-Term Impact of Hospitalization for Kawasaki Disease on Health-Related Quality of Life Free

Kawasaki disease (KD) is an acute vasculitis of childhood with predilection for the coronary arteries.^1,2  When untreated, coronary artery inflammation can result in aneurysm formation with subsequent thrombosis, stenosis, or both.^3–5  A subset of patients with coronary aneurysms exhibit lifelong cardiac disability, and some may even experience myocardial infarction.⁶  The etiology of KD remains elusive, and the diagnosis is based on symptoms and laboratory values that overlap with other childhood disease processes, which sometimes lead to challenges in diagnosis and delays in treatment initiation.^1,2,7  Moreover, the addition of potentially serious cardiovascular sequelae with vague long-term prognosis contributes to prolonged uncertainty and psychological distress in patients and families.^8,9 

The World Health Organization and the Centers for Disease Control and Prevention define health-related quality of life (HRQoL) as a principal patient-reported outcome that quantifies disability by accounting for multidimensional domains of physical, social, and psychological functioning, and the use of this measure in evaluating treatment success has been expanding.^10–12  Patient-reported outcomes, including HRQoL measures, provide important information about clinical and behavioral interactions in a wide variety of medical conditions.^12,13  The Pediatric Quality of Life Inventory (PedsQL) 4.0 Generic Core Scales and the PedsQL Infant Scales are standardized population HRQoL instruments with recognized reliability and validity in assessing physical and psychological functioning in both the outpatient and inpatient settings.^14–20  The PedsQL physical domain assesses problems with running, pain, and energy level, whereas the psychosocial domain evaluates trouble with sleeping, feeling angry or scared, and whether a child can get along with other children or can do things that other children of the same age can do (see Supplemental Fig 3 for a sample PedsQL Inventory).

We previously evaluated the acute impact of KD on psychophysical well-being and assessed, using the PedsQL 4.0 Generic Core and Infant Scales,²¹  the deterioration of HRQoL experienced by children hospitalized with KD compared with other childhood diseases. We hypothesized that parents perceive that their children with KD exhibit physical and psychosocial deterioration similar to that experienced by those with other newly established acute and chronic diseases. We then compared the KD HRQoL results with those obtained from patients with newly diagnosed cancer and pneumonia. Our results revealed that patients with KD experienced a significantly greater HRQoL decline from baseline to admission than the other 2 groups with a diagnosis of cancer or pneumonia. KD diagnostic subtype (complete or incomplete) and coronary artery dilatation were not associated with HRQoL outcomes. However, intravenous immunoglobulin (IVIG) nonresponders showed greater HRQoL decline than responders. These results indicated that children with KD experience acute and significant HRQoL impairment exceeding that of children with newly diagnosed cancer and uncomplicated pneumonia.

The principal objective of the present study was to determine if the impact of KD on HRQoL persists or even recurs over a longer term after the initial diagnosis and hospitalization. Acute KD is considered a self-limited disease that generally requires a single hospitalization for treatment. Accordingly, we included patients admitted for acute pneumonia, also usually requiring a single hospitalization, as a comparison group. In this way, we could control for the impact of hospital admission and treatment on HRQoL scores. Because cancer is considered a more chronic disease that requires long-term treatment and multiple admissions, we did not include a cancer cohort for comparison in this long-term study. We sought to determine whether specific KD characteristics, including KD subtype, responsiveness to IVIG, and coronary artery dilatation, influence HRQoL decline and its persistence at long-term follow-up.

We performed a single-center, prospective cohort study that used the framework of 2 existing institutional databases, the KD Research Program and the Outcomes Assessment Program (OAP), at our 400-bed freestanding children’s hospital in the western United States. Key data from the original KD-pneumonia–matched cohort²¹  can be found in Supplemental Tables 3–7. KD data were extracted from the OAP and KD Research Program databases to create a merged KD-HRQoL dataset. The OAP was also queried for patients with a diagnosis of community-acquired pneumonia to construct the pneumonia comparison cohort. Within 72 hours of hospital admission, caregivers were instructed to complete a baseline PedsQL form to reflect HRQoL at 1 month before admission and another form to reflect HRQoL at the time of admission. A short-term follow-up PedsQL form was completed 2 to 12 weeks after hospital discharge as part of the 3 total surveys performed by the OAP. We then performed follow-up long-term PedsQL surveys at least 1 year after hospital discharge. Each parent proxy response is reverse scored in a scale format and linearly converted to a scale of 0 to 100. Those individual items are combined and divided by the number of items answered to derive the total physical and psychosocial scores, respectively. Higher total scores indicate better HRQoL. Improvement scores were derived by calculating the difference between the total, physical, and psychosocial scores on admission and scores at follow-up. Return to baseline was defined as a follow-up score no >4.5 points below the baseline value on the basis of previously published data defining the minimal clinically important decrease in HRQoL.^17,19 

This analysis contains the subset of patients for whom long-term follow-up was available. We used 2 approaches to minimize the impact of differences in cohort demographic and other characteristics. First, results for the entire (unmatched) cohort adjusted for significant differences were calculated. Second, propensity score (PS)–matched cohorts were constructed from the unmatched cohorts. The PS model included patient age, sex, race, parent age, parent education level, insurance type, and baseline physical and psychosocial HRQoL scores. Once a PS for each patient was obtained from the model fitted values, patients with KD were matched, using a 1:1 greedy matching algorithm, to those with pneumonia.

Baseline characteristics were compared using the t test or Fisher’s exact test. Changes from baseline HRQoL to subsequent time points within the KD cohort were analyzed using analysis of covariance models adjusted for imbalanced covariates and baseline scores for the unmatched cohorts. Similar models adjusted only for baseline scores in the matched cohort. These analysis of covariance models were used to calculate least squares means (LSMs) and 95% confidence intervals (CIs). Odds ratios (ORs) for the chance of return to baseline scores were calculated using logistic regression models, with identical covariate adjustments for the unmatched and PS-matched cohorts.

We performed exploratory analyses comparing KD subgroups, recognizing that we would have limited statistical power. Patients with KD were divided into subgroups according to (1) whether they had complete or incomplete KD, (2) responsiveness to IVIG treatment, (3) coronary artery status, and (4) whether they required continued cardiology follow-up and treatment past 6 weeks from the time of diagnosis. The analyses within the KD subgroup are unadjusted and because of the size of the subgroups, should be interpreted with caution. Mean differences in HRQoL scores were evaluated using the t test for independent samples at baseline and the paired t test for follow-up measures. Two-sided P < .05 was considered statistically significant.

A total of 266 patients (118 with KD and 148 with pneumonia) were identified and contacted. The caregivers of 125 patients (57 with KD and 68 with pneumonia) either were not reachable by phone after 3 attempts or did not consent to participate. Long-term follow-up HRQoL data were collected on 141 patients (61 with KD and 80 with pneumonia) whose caregivers completed the parent proxy PedsQL Generic Core Scales for children and adolescents (2–18 years), providing a response rate of 51.7% for the KD and 54.1% for the pneumonia cohorts. Clinical and sociodemographic characteristics for the total and PS-matched KD-pneumonia cohorts are summarized in Table 1 and were well balanced across groups. Baseline total scores for both cohorts approximated the published mean values for healthy children, demonstrating the validity of our results (KD, 88.8 ± 12.3 and pneumonia 87.8 ± 11.9 vs the published value of 90¹⁸ ) (Table 1).

HRQoL scores for the full cohort are shown in Fig 1 and Supplemental Table 3. Based on models adjusted for parent age and baseline score, the PedsQL total score at hospital admission was lower for KD than for pneumonia (mean, 40.5 vs 49.5; P = .03) (Fig 1A). The mean within-patient reduction from baseline to admission was significantly different (−45.9 decrease for KD vs −37.2 decrease for pneumonia; P = .04).

The PedsQL physical subscale was lower at admission for the patients with KD than for those with pneumonia (mean, 29.4 vs 39.7; P = .04). Long-term follow-up scores for the physical subscale were significantly higher for patients with KD than for those with pneumonia (mean, 93.3 vs 84.7; P = .04) (Fig 1B). None of the mean within-patient changes in the PedsQL physical subscale from baseline differed between the KD and pneumonia cohorts. There were no significant differences for the mean scores or within-patient changes for any time point for the psychosocial subscale (Fig 1C).

HRQoL total scores for the PS-matched cohort are shown in Fig 2 and Supplemental Table 3. There were no significant differences in PedsQL total scores at any time point (Fig 2A). The mean within-patient change from baseline to admission was significantly different (−46.6 [95% CI, −53.2 to −39.9] decrease for KD vs −36.5 [95% CI, −43.2 to −29.8] decrease for pneumonia; P = .04).

There were no differences in the PedsQL physical subscale total scores (Fig 2B) or within-patient changes between KD and pneumonia in the PS-matched cohort. For the PedsQL psychosocial subscale, the score at admission was lower for KD than for pneumonia (50.3 vs 61.0; P = .04) (Fig 2C). Analogously, the mean within-patient changes showed greater reductions from baseline to admission in the KD cohort (−36.2 decrease for KD vs −25.3 for pneumonia; P = .03).

ORs adjusted for imbalances and baseline scores in the unmatched cohort and adjusted for baseline scores in the PS-matched cohort are presented in Table 2. At short-term follow-up, 65% to 78% of patients had returned to their baseline scores. For the PS-matched cohort at long-term follow-up, the rates of return to baseline were higher at 76% to 89%. In the unmatched cohort, a higher proportion of patients in the KD group returned to baseline PedsQL physical score at short-term follow-up (87.3% vs 75.0%; P = .04). However, this difference was attenuated in the PS-matched cohort.

Overall, patients with KD experienced a 54% decline in total PedsQL score from baseline to admission (LSM, 87.8 at baseline and 40.5 at admission) (Supplemental Table 3). Subscale PedsQL score analysis demonstrated a greater impact on the physical than on the psychosocial domain (67% vs 40%).

There were 43 patients with complete and 18 with incomplete KD. No differences were found in total or component scores at admission or short-term follow-up between the 2 groups (Supplemental Table 4).

Fifty-two of the patients (85%) responded to the first dose of IVIG, whereas 9 patients required 2 doses (nonresponders). At admission, the nonresponders had lower total PedsQL scores than the responders (mean reduction, −64.5 vs −46.3; P = .02) (Supplemental Table 5). Subscale PedsQL score analysis demonstrated a greater impact on the physical domain (mean reduction, −78.7 vs −57.8; P = .04) compared with the psychosocial domain (P not significant). At short- and long-term follow-up, there was no difference in total or subgroup scores between the 2 groups.

The prevalence of any type of coronary artery abnormality was 47.5%. Thirty-two (52.5%) of the 61 patients had normal coronary arteries on their initial and subsequent echocardiograms. Thirteen (21.3%) of the 61 patients had minimal coronary artery dilatation (z score <2.5), and 16 (26.2%) had coronary artery dilation and/or aneurysm (z score ≥2.5) (Supplemental Table 6). Of the 16 patients with z score ≥2.5, 7 (43.8%) continued to have abnormal coronary arteries at the time of their 6-week follow-up, whereas 9 (56.2%) had complete normalization. All 13 patients with a mild degree of coronary artery dilation (z score < 2.5) demonstrated normal coronary arteries at 6-week follow-up. Coronary artery dilatation of any degree did not reveal any association with lower PedsQL scores on admission or short-term follow-up. Score change from baseline to long-term follow-up was slightly higher among the 16 patients with coronary artery dilation than among those without (9.38 vs 1.50; P = .05).

There was no significant difference in the baseline demographics of patients who continued to have coronary changes (z score of ≥2.5) at the time of their 6-week follow-up and required continued cardiology care (Supplemental Table 7). In addition, no difference was found in hospital length of stay, baseline total scores, KD type, or number of IVIG doses. There was no difference in total scores at short- or long-term follow-up between patients who had persistent coronary changes at 6 weeks and those who did not.

All patients with KD remained on a 6-week antiplatelet regimen and presented for outpatient clinic surveillance per standard KD management protocol. Patients with giant aneurysms required additional treatment with low-molecular-weight heparin injections and more frequent outpatient follow-up. There was no difference in HRQoL scores between patients who had persistent coronary artery changes and required longer-term follow-up and anticoagulation versus patients who discontinued aspirin at 6 weeks.

We performed the first prospective study evaluating longitudinal HRQoL scores in a hospitalized KD cohort over a relatively long term. Our previous investigation evaluated the impact of acute admission and early convalescence on HRQoL. We have extended the study, expanded the KD and pneumonia cohorts, and added PS matching to account for potential differences in demographic factors between the cohorts. For the most part, the findings conform to results from our initial study with regard to impact of hospitalization on short-term HRQoL. More importantly, our current data reveal that patients with KD are resilient and recover HRQoL over time. Most patients return to baseline function and maintain that status for at least 1 year. The results can serve to ameliorate parental concerns about the long-term impact of KD, particularly in patients who do not exhibit coronary artery abnormalities.

In our previous study, we used the PedsQL instrument to assess function of children with KD in real time to avoid potential errors introduced in previous retrospective studies that used long-term recall. We used a greedy matching algorithm, which used variables of age (within 1 year), sex, and race to construct disease cohorts with similar characteristics. Our current dataset revealed that our unmatched cohorts showed significant differences in some baseline HRQoL scores as well as parental age between the KD and pneumonia cohorts. To account for potential bias created from these discrepancies and affecting HRQoL, we matched patients by using a PS, which included multiple variables. Although, PS matching removes some confounding variables, loss of unmatchable subjects from the dataset reduces statistical power. Accordingly, we do have some differences in output when comparing unmatched versus PS-matched cohorts. However, the data suggest that some factors, such as parental age, may have affected differences between the KD and pneumonia cohorts.

In the KD cohorts, unmatched and PS matched, ∼75% of patients achieved their baseline HRQoL total and component status in the short term, and nearly 90% achieved it over the longer term. Although recovery was similar with that of the pneumonia cohort, we sought to explore KD-specific elements that could delay or inhibit HRQoL recovery. Presentation for KD is highly variable. Families with children presenting with incomplete KD criteria often encounter issues related to lack of diagnostic clarity. Qualitative research has revealed that ambiguity and delay on the part of the physician in establishing the correct diagnosis as key issues for heightened parental anxiety in KD.⁹  Our study, however, reveals that incomplete KD does not lead to a greater deterioration in HRQoL, and we did not observe any differences overall in stress burden in patients with complete versus incomplete KD.⁹ 

To date, the factors that determine KD response to IVIG remain undefined, and retreatment of IVIG-refractory patients could increase family anxiety and affect the child’s function.²²  Our data reveal that IVIG nonresponders have significantly lower HRQoL scores statistically at admission than patients who responded to a single dose of IVIG. When evaluating HRQoL scores at short- and long-term follow-up, no differences were found in scores between the IVIG responders and nonresponders, revealing that even though IVIG nonresponders have a lower HRQoL at admission, they have a rapid recovery in the short term that persists at long-term follow-up.

Coronary artery dilatation did not reveal any association with lower HRQoL at admission or short- and long-term follow-up. Similarly, the patients who continued to have persistent coronary changes with z scores ≥2.5 at 6-week follow-up and required longer-term evaluation by a cardiologist did not experience lower HRQoL scores at any stage. The rate of coronary changes in our population was comparable with that reported in previous publications.^2,4,9,23,24  Important cardiovascular sequelae in the form of severe dilatation or coronary aneurysms developed in only 16 (21%) of 61 patients with KD, and of these patients, only 7 had persistent changes at 6-week follow-up. Therefore, our study may have been underpowered to determine the effect of coronary artery involvement on HRQoL scores.

In short-term follow-up, 25% of the population with KD did not meet their premorbid levels of age-appropriate function at 4 to 12 weeks after presumed recovery. However, at long-term follow-up, 89% of the PS-matched cohort had reached their baseline HRQoL. This finding reveals that although patients with KD experience a substantial decrease in HRQoL at hospital admission, over time, their recovery is no worse than that of patients with a common pediatric disease like pneumonia and that the rebound in HRQoL is durable between short- and long-term follow-up. Most previous studies evaluating anxiety or psychosocial dysfunction in KD had limitations inherent in retrospective analyses requiring long-term recall.^11,24–29 

At least 1 recent study centered in Ontario used an administrative database to suggest that patients with KD exhibit long-term anxiety disorders beginning at 1 year after the acute episode.³⁰  However, we report prospectively collected quantitative and longitudinal data from individual parent surveys. This discrepancy highlights problems and potential inaccuracies when using administrative databases, which often depend on diagnostic coding. Through our study, we can provide some reassurance for parents that acute KD does not generally lead to long-term functional deficits. The study was not adequately powered to determine if patients with persistent coronary artery abnormalities exhibit functional deficits.

We speculate that higher parental education level has a protective effect on HRQoL, although no studies in the literature have evaluated this relationship. On the other hand, a number of studies have revealed a positive correlation between higher socioeconomic status and high HRQoL.^31–33  Interestingly, 86.5% of patients with KD and 81.1% with pneumonia had commercial insurance, indicating a likely higher socioeconomic status, whereas 67.6% of patients with KD and 59.5% with pneumonia had parents with at least a 4-year college degree. Given this demographic distribution, it is likely that acute illness and hospitalization might have had a lesser impact on the HRQoL of our patient cohorts and that these results might not be generalizable to other patient populations. The results reinforce the strength of the immediate impact on patients and families during the acute KD episode. We believe that much of this effect on function relates to multiple interventions, tests, and continued uncertainty surrounding the KD diagnosis, which relies primarily on clinical algorithms. A definitive diagnostic test will likely reduce this impact, particularly with patients diagnosed with American Heart Association–defined incomplete criteria. Parents have described KD as “an emotional rollercoaster dominated by their child’s acute suffering” at short- and long-term follow-up.⁹  Care providers should be alert to acute decompensation and stress syndromes exhibited by families during and immediately after diagnosis and treatment of acute KD. Reassurance that this study reveals limited, if any, long-term effects on psychosocial or physical functioning will be helpful.

Limitations include those inherent to single-center studies, as the results might not be generally applicable. More specifically, the majority of our patients had commercial insurance, indicating a likely higher socioeconomic status, and a large proportion of parents had higher education attainment than noted in the most recent US census report, which may reflect a bias toward those willing to complete multiple surveys. However, no significant difference was found between the KD and pneumonia cohorts for these factors. Although PS matching reduced differences between the pneumonia and KD cohorts for measured covariates, there is always a potential risk that differences in unmeasured covariates are responsible for cohort differences. We used all available demographic data to construct similar groups, but in the absence of a uniformly collected medical history, there is a risk of confounding from unmeasured covariates. A proportion of patients with KD diagnosed during the study period were not reachable by phone, and we therefore have no information on the disease severity of these patients. Similarly, given the cohort size and small number of patients with coronary changes and IVIG nonresponse, this study was underpowered to fully investigate the effect of persistent coronary artery changes or IVIG nonresponse on HRQoL. Additionally, because of small numbers, we could not adequately perform PS matching in these subgroups. Thus, the KD subgroup analyses should be considered exploratory. Finally, even though PedsQL surveys were collected prospectively for hospital admission and short- and long-term follow-ups, the baseline survey collected at the time of hospital admission was retrospective and might be subject to recall bias.

To our knowledge, we are the first to evaluate with a prospective study the long-term effects of KD on HRQoL. Even though patients with KD experience a significant decline in HRQoL at the time of hospital admission, we have shown that these patients achieve rapid and persistent recovery and that almost all patients achieve their baseline HRQoL scores at long-term follow-up. Future research should focus on identifying higher-risk patients who warrant more intense monitoring after hospitalization to reach their premorbid physically and developmentally appropriate functional levels.