Mobile apps are suggested for supporting home monitoring and reducing emergency department (ED) visits and hospitalizations for children with medical complexity (CMC). None have been implemented. We sought to assess the MyChildCMC app (1) feasibility for CMC home monitoring, (2) ability to detect early deteriorations before ED and hospital admissions, and (3) preliminary impact.
Parents of CMC (aged 1–21 years) admitted to a children’s hospital were randomly assigned to MyChildCMC or usual care. MyChildCMC subjects recorded their child’s vital signs and symptoms daily for 3 months postdischarge and received real-time feedback. Feasibility measures included parent’s enrollment, retention, and engagement. The preliminary impact was determined by using quality of life, parent satisfaction with care, and subsequent ED and hospital admissions and hospital days.
A total of 62 parents and CMC were invited to participate: 50 enrolled (80.6% enrollment rate) and were randomly assigned to MyChildCMC (n = 24) or usual care (n = 26). Retention at 1 and 3 months was 80% and 74%, and engagement was 68.3% and 62.6%. Run-chart shifts in vital signs were common findings preceding admissions. The satisfaction score was 26.9 in the MyChildCMC group and 24.1 in the control group (P = .035). No quality of life or subsequent admission differences occurred between groups. The 3-month hospital days (pre-post enrollment) decreased from 9.25 to 4.54 days (rate ratio = 0.49; 95% confidence interval = 0.39–0.62; P < .001) in the MyChildCMC group and increased from 1.08 to 2.46 days (rate ratio = 2.29; 95% confidence interval = 1.47–3.56; P < .001) in the control group.
MyChildCMC was feasible and appears effective, with the potential to detect early deteriorations in health for timely interventions that might avoid ED and hospitalizations. A larger and definitive study of MyChildCMC’s impact and sustainability is needed.
Children with medical complexity (CMC) have high acute health care use with recurrent emergency department (ED) visits and frequent and/or prolonged hospitalizations.1–8 Although composing <1% of children, CMC account for 30% of pediatric health care costs and 86% of hospital charges in the United States.9–11
CMC have significant daily care needs at home and require continuous parent (caregiver) attention and care coordination.12 These children often have frequent acute health deteriorations and issues with supportive technologies that lead to recurrent ED and/or hospital admissions. For example, in a sample of CMC with gastrostomy tubes, 73% had device-related complications that led to hospitalizations, most of which could have been identified and managed at home.13 Novel approaches are needed to reduce hospitalizations in CMC.
Many programs to support CMC ongoing care outside the hospital (eg, medical home, home nursing care, specialty clinics, etc) exist.14–31 In observational studies, researchers suggest these programs can reduce hospitalizations, lengths of stay, and costs.32–37 However, they are costly, underfunded, nonaccessible to many CMC, and lack home-monitoring systems for early recognition of child’s health deteriorations to avoid admissions.
Mobile health (mHealth) apps have the potential to support home monitoring and reduce the risk of acute deteriorations of CMC’s health.38,39 Although mHealth apps can improve outcomes of patients with chronic diseases,40–46 studies have largely been targeted at single, noncomplex conditions (eg, asthma, hypertension, etc).47–52 No home-monitoring app has been tested for feasibility or impact on CMC outcomes.
In a previous study, our team described preferences and key functionalities that parents found important for a CMC home-monitoring app.53 The results were used to design MyChildCMC and test it for usability with parents. The current pilot study’s objectives were to (1) assess the feasibility of MyChildCMC as a home-monitoring app for CMC, (2) explore the app’s ability to identify early signs of health deteriorations before ED and/or hospital admissions, and (3) determine the preliminary impact of the app on child-health outcomes. We hypothesized that MyChildCMC use would be feasible and achieve >60% enrollment, retention, and engagement rates and54 would help detect early signs of health deterioration and, that, compared with controls, MyChildCMC users would have better outcomes for the child (fewer ED and/or hospital use and hospital days and higher quality of life [QoL]) and parent (higher satisfaction with child’s care).
Methods
Setting
The study was conducted at a 289-bed freestanding, tertiary-care children’s hospital in the Intermountain West in the United States. CMC represent 30% of admissions and 53% of hospital charges. Annually, the hospital has 3500 ED visits, 1880 admissions, 4000 hospital days, and $257 million in hospital charges for CMC.
MyChildCMC Description
MyChildCMC app was designed for parent use to support CMC home monitoring and generate early warnings for acute health deteriorations that often lead to ED and/or hospital admissions.53 MyChildCMC has multiple features (Supplemental Fig 4) including the following:
1. User interface: This allows parents to record daily the following parameters (Supplemental Table 3), determined during focus groups with CMC parents53 : temperature, heart rate (HR), oxygen saturation (O2Sat), respiratory rate, pain, seizure status, fluid intake and feeding, mental status, and level of parent worry about the child’s health.
a. For temperature, HR, and respiratory rate, measured values are recorded.
b. For pain, the numeric rating scale is used to score and record pain level.55 The numeric rating scale is a 1-item parent proxy scale with scores categorized as mild (0–3), moderate (4–7), and severe (8–10) pain.
c. Feeding and fluid intake (1 item) is recorded as 25%, 50%, 75%, or 100% of usual food and fluid intake.
d. Mental status is determined by using the behavior (1 item) question of the pediatric early warning scores, scored as (1) playing and/or appropriate, (2) sleepy, (3) irritable and/or agitated, and (4) lethargic and/or unresponsive.
e. Seizures are recorded as (1) fewer than usual, (2) usual, (3) more than usual, and (4) way more than usual.
f. Caregiver worry, a nonspecific sense of looming health deterioration (even in the absence of abnormal vital signs and/or symptoms), is recorded as (1) not worried, (2) a little worried, (3) worried, and (4) very worried. Excess parental worry about a child’s health increases the likelihood of the child receiving care.56
2. Decision rules: 2 rules are used to identify significant changes in vital signs and symptoms: (1) specific thresholds and (2) run-chart rules to identify deviation from individualized baselines,57,58 which are established with a minimum of 14 days of data, whereas the child is medically stable as recorded by parents (parents indicate if the child’s health status is “typical” or “normal” for each day’s data entry). Subsequent data are compared with the baseline until a significant change is detected, and, then, a new baseline is established with new 14 normal days after patient’s signs and symptoms return to a stable state, and so on.
a. For thresholds, significant changes are defined as O2Sat <90% (Salt Lake City altitude), a pain score >7, temperature <96.8°F or >101.3°F, <75% hydration or ≤25% nutrition, and scores 3 or 4 for mental status and worry.
b. Run-chart rules are used for temperature, HR, respiratory rate, pain, and O2Sat. Significant changes are defined as having the following:
• An outlier (data point >2 SD from patient’s baseline).
• Shift (≥6 consecutive points below or above the patient’s baseline mean but within 2 SD).
• Trend (≥5 increasing or decreasing consecutive data points).
3. Real-time feedback system: includes a simplified summary report page and individual graphs.
a. Summary report: provides a quick view of “today’s” vital signs and symptoms, categorized in 3 zones: green (no outlier, trend, or shift), yellow (trend or shift with data within 2 SD), or red (outlier, shift, or trend with data >2 SD). Vital sign and symptom data in the green zone means “good,” the yellow zone means “could be better,” and the red zone means “not good.”
b. Individual graphs: provide a longitudinal view of changes over time for each vital sign and symptom.
4. Real-time reminder: automated e-mail and/or text reminders sent daily to support user engagement with daily use.
5. Other features: The app works across popular smartphone platforms. Data are stored behind the university firewall within a Health Insurance Portability and Accountability Act compliant database (HeidiSQL).
Study Design, Population, and Eligibility Criteria
We conducted a randomized controlled trial of parents (or primary caregivers) whose CMC were admitted to the hospital. The inclusion criteria were (1) CMC determined by the care team using an operational definition, including having multisystem chronic conditions with (a) ≥3 organ systems affected, (b) ≥3 health care providers involved in their care, (c) ≥2 previous year admissions, and/or d) technology dependence; (2) aged 1 to 21 years; (3) parent owns a smartphone or tablet with Internet access; and (4) English speakers. The exclusion criteria were (1) critically ill, (2) intensive care admission, (3) receiving palliative care, (4) not discharged from the hospital, (5) non-English speakers, and (6) families without a smartphone or tablet or Internet access. The institutional review board approved the study. The trial is registered with ClinicalTrials.gov (identifier: NCT04470193).
Study Procedures
Each morning, research coordinators contacted inpatient-team attending physicians to identify CMC admissions and review the medical records to confirm eligibility. Participants were approached during hospitalization, enrolled, and randomly assigned to the MyChildCMC app or usual care, by using a computer-generated allocation sequence uploaded into REDCap for use by research coordinators. Participants randomly assigned to MyChildCMC received help to download the app and training on use. Research coordinators reviewed vital sign measurements with parents (control and intervention) and contacted their nurse for training if they had difficulty and provided parents in both groups with needed equipment (pulse oximeter and stethoscope). Participants were enrolled between June 10, 2019, and December 19, 2019, and asked to monitor their child’s vital signs and symptoms daily for 3 months after hospital discharge, by using MyChildCMC (intervention group) or paper (control group). There was no other interaction or interference with the participant’s usual care.
Data Collection, Outcome Measures, and Statistical Analysis
QoL was collected at enrollment (baseline) and at 1- and 3-months postenrollment, by using a survey adapted from Ellzey et al,59 (Supplemental Fig 8), assessing physical health, mental health, sleep, pain, activities, and general QoL. Follow-up surveys included QoL, collected at 1- and 3-months postenrollment, and parent satisfaction with overall child’s care, collected at 3 months, by using Research Electronic Data Capture. Parent satisfaction (Supplemental Fig 9) was collected by using a survey adapted from the Client Satisfaction Questionnaire60 and assessed (1) parent confidence in care provision, (2) parent consistency to manage child’s health, (3) medical professional support, (4) positive social support, (5) care structure (monitoring and responding to symptoms) at home, and (6) parental stress. During satisfaction survey, we asked 4 additional questions (only to the intervention group) to assess parental perception of the app’s impact on their child’s health. Both groups received a gift card of $40 for completing each baseline and follow-up QoL surveys.
We used 3 metrics to assess MyChildCMC feasibility: (1) enrollment rate (percent of eligible parents who enrolled after being invited to participate), (2) retention rate (percent of participants who completed the QoL survey at 1 and 3 months), and (3) engagement rate (percent of daily vital signs and symptoms recorded in MyChildCMC out of all opportunities over the study period). The percentage of participants who used MyChildCMC each day during the study is depicted in Supplemental Fig 5.
MyChildCMC’s ability to detect health deteriorations was determined among users by plotting vital sign and symptom data on run charts and exploring any significant change in the 3 months postenrollment and its relationship with subsequent ED or hospital admission.
Preliminary impact was determined by comparing outcomes between the MyChildCMC and control groups. The primary outcome was the child’s QoL. Secondary outcomes included parent satisfaction with the overall child’s care and subsequent child’s ED or hospital admissions and hospital days (3 months pre- and postenrollment), collected after the study was completed, by using the electronic health record, which captures >95% of pediatric ED and hospital admissions in our state.
We used a generalized estimating equation model for repeated assessments, with a common unstructured residual covariance matrix to account for correlation in repeated measurements in the same patient, to compare QoL total score at baseline, 1 month, and 3 months between the groups. Using intention-to-treat analysis, we compared numbers of ED or hospital admissions and hospital days (3-month pre- and 3-month postenrollment; enrollment hospitalization excluded) between the groups using generalized linear models with logarithmic link and Poisson regression, estimating rate ratios (RRs) and 95% confidence intervals (CIs). A generalized linear model was also used to compare parent satisfaction. Covariates included age, sex, and race and ethnicity.
To achieve our planned sample size of 50 subjects, we randomly assigned 68 participants to provide 50 evaluable subjects, allowing up to 16% attrition, to estimate the proportions of patients satisfying enrollment, retention, engagement, and dropout to within margins of error (half-widths of 90% CIs) of ±0.11, 0.11, 0.16, and 0.13, if the true proportions for enrollment, retention, engagement, and dropout, were 0.60, 0.60, 0.60, and 0.20, respectively. The 50 evaluable subjects were used to estimate the SD of QoL, with a margin of error of ∼20%. The sample size also allowed us to provide provisional estimates of the effect size for the QoL outcome to within a margin of error of ±7.1 points, assuming a true SD of 15 points.
Results
Feasibility and Demographic Characteristics
A total of 68 CMC parents were invited to participate: 12 declined enrollments, and 6 did not meet eligibility criteria (5 were non-English speaking parents, and 1 was on palliative care). Of the 50 out of 62 eligible parents who agreed (80.6%) to participate, 24 were randomly assigned to MyChildCMC, and 26 were randomly assigned to usual care. CMC were 42% female, 76% non-Hispanic white, 22% Hispanic, and 2% other race and/or ethnicity, with an overall mean age of 8.5 (SD: 5.7) years, with no significant differences between groups (Table 1). A total of 6 of the 50 parents participated in a previous usability study of MyChildCMC; 3 were randomly assigned to MyChildCMC, and 3 were randomly assigned to usual care.
Demographic Characteristics
| Patient (child) characteristics | Control (n = 26) | Intervention (n = 24) | P |
|---|---|---|---|
| Sex, n (%) | |||
| Male | 14 (53.9) | 15 (62.5) | .536 |
| Female | 12 (46.2) | 9 (37.5) | — |
| Age, mean (95% CI) | 9.65 (7.40–11.91) | 7.29 (4.90–9.69) | .144 |
| Patient race and/or ethnicity, n (%) | |||
| White | 18 (69.2) | 20 (83.3) | .477 |
| Hispanic | 8 (30.8) | 3 (12.5) | — |
| Other | 0 (0) | 1 (4.2) | — |
| Most common clinical conditions among CMC, n (%) | |||
| Technology dependent (n = 24) | 11 (42.3) | 13 (54.2) | .402 |
| Epilepsy and/or seizures (n = 19) | 8 (30.8) | 11 (45.6) | .273 |
| Respiratory failure (n = 17) | 7 (26.9) | 10 (41.6) | .272 |
| Fever (n = 17) | 7 (26.9) | 10 (41.6) | .272 |
| Infection of known cause (n = 15) | 9 (34.6) | 6 (25.0) | .459 |
| Cerebral palsy (n = 14) | 6 (23.1) | 8 (33.5) | .420 |
| GI or feeding issues (n = 10) | 4 (15.4) | 6 (25.0) | .396 |
| Parent or caregiver characteristics, n (%) | |||
| Phone use type | |||
| iPhone | 0 (0) | 12 (50) | — |
| Android | 0 (0) | 12 (50) | — |
| Unknown | 26 (100) | 0 (0) | — |
| Parent relation to patient, n (%) | |||
| Mother | 23 (88.5) | 20 (83.3) | .723 |
| Father | 2 (7.7) | 2 (8.3) | — |
| Unknown | 1 (3.9) | 2 (8.3) | — |
| Parent race and/or ethnicity, n (%) | |||
| White | 20 (76.9) | 19 (79.2) | .716 |
| Hispanic | 5 (19.2) | 2 (8.3) | — |
| Other | 1 (3.9) | 3 (12.5) | — |
| Parent marital status, n (%) | |||
| Married | 19 (73.1) | 15 (62.5) | .291 |
| Divorced | 2 (7.7) | 4 (16.7) | — |
| Widowed | 2 (7.7) | 0 (0) | — |
| Never married | 2 (7.7) | 3 (12.5) | — |
| Unmarried couple | 1 (3.9) | 0 (0) | — |
| Unknown | 0 (0) | 2 (8.3) | — |
| Parent home status, n (%) | |||
| Not a single parent | 20 (76.9) | 15 (62.5) | .109 |
| Single parent | 6 (23.1) | 7 (29.2) | — |
| Unknown | 0 (0) | 2 (8.3) | — |
| Parent education, n (%) | |||
| College incomplete | 13 (50) | 9 (37.5) | .056 |
| College completed | 13 (50) | 12 (50) | — |
| Unknown | 0 (0) | 3 (12.5) | — |
| Parent family income, n (%) | |||
| <25 000 | 6 (24) | 4 (16.7) | .421 |
| 25 000–39 000 | 5 (20) | 4 (16.7) | — |
| 40 000–49 000 | 2 (8) | 1 (4.2) | — |
| 50 000–74 000 | 3 (12) | 5 (20.8) | — |
| 75 000–99 000 | 1 (4) | 4 (16.7) | — |
| 100 000–124 000 | 0 (0) | 2 (8.3) | — |
| 125 000–149 000 | 3 (12) | 1 (4.2) | — |
| >150 000 | 5 (20) | 1 (4.2) | — |
| Unknown | 1 (0) | 2 (8.3) | — |
| Parent employment status, n (%) | |||
| Unemployed | 8 (30.8) | 6 (25) | .066 |
| Employed | 18 (69.2) | 15 (62.5) | — |
| Unknown | 0 (0) | 3 (12.5) | — |
| Patient (child) characteristics | Control (n = 26) | Intervention (n = 24) | P |
|---|---|---|---|
| Sex, n (%) | |||
| Male | 14 (53.9) | 15 (62.5) | .536 |
| Female | 12 (46.2) | 9 (37.5) | — |
| Age, mean (95% CI) | 9.65 (7.40–11.91) | 7.29 (4.90–9.69) | .144 |
| Patient race and/or ethnicity, n (%) | |||
| White | 18 (69.2) | 20 (83.3) | .477 |
| Hispanic | 8 (30.8) | 3 (12.5) | — |
| Other | 0 (0) | 1 (4.2) | — |
| Most common clinical conditions among CMC, n (%) | |||
| Technology dependent (n = 24) | 11 (42.3) | 13 (54.2) | .402 |
| Epilepsy and/or seizures (n = 19) | 8 (30.8) | 11 (45.6) | .273 |
| Respiratory failure (n = 17) | 7 (26.9) | 10 (41.6) | .272 |
| Fever (n = 17) | 7 (26.9) | 10 (41.6) | .272 |
| Infection of known cause (n = 15) | 9 (34.6) | 6 (25.0) | .459 |
| Cerebral palsy (n = 14) | 6 (23.1) | 8 (33.5) | .420 |
| GI or feeding issues (n = 10) | 4 (15.4) | 6 (25.0) | .396 |
| Parent or caregiver characteristics, n (%) | |||
| Phone use type | |||
| iPhone | 0 (0) | 12 (50) | — |
| Android | 0 (0) | 12 (50) | — |
| Unknown | 26 (100) | 0 (0) | — |
| Parent relation to patient, n (%) | |||
| Mother | 23 (88.5) | 20 (83.3) | .723 |
| Father | 2 (7.7) | 2 (8.3) | — |
| Unknown | 1 (3.9) | 2 (8.3) | — |
| Parent race and/or ethnicity, n (%) | |||
| White | 20 (76.9) | 19 (79.2) | .716 |
| Hispanic | 5 (19.2) | 2 (8.3) | — |
| Other | 1 (3.9) | 3 (12.5) | — |
| Parent marital status, n (%) | |||
| Married | 19 (73.1) | 15 (62.5) | .291 |
| Divorced | 2 (7.7) | 4 (16.7) | — |
| Widowed | 2 (7.7) | 0 (0) | — |
| Never married | 2 (7.7) | 3 (12.5) | — |
| Unmarried couple | 1 (3.9) | 0 (0) | — |
| Unknown | 0 (0) | 2 (8.3) | — |
| Parent home status, n (%) | |||
| Not a single parent | 20 (76.9) | 15 (62.5) | .109 |
| Single parent | 6 (23.1) | 7 (29.2) | — |
| Unknown | 0 (0) | 2 (8.3) | — |
| Parent education, n (%) | |||
| College incomplete | 13 (50) | 9 (37.5) | .056 |
| College completed | 13 (50) | 12 (50) | — |
| Unknown | 0 (0) | 3 (12.5) | — |
| Parent family income, n (%) | |||
| <25 000 | 6 (24) | 4 (16.7) | .421 |
| 25 000–39 000 | 5 (20) | 4 (16.7) | — |
| 40 000–49 000 | 2 (8) | 1 (4.2) | — |
| 50 000–74 000 | 3 (12) | 5 (20.8) | — |
| 75 000–99 000 | 1 (4) | 4 (16.7) | — |
| 100 000–124 000 | 0 (0) | 2 (8.3) | — |
| 125 000–149 000 | 3 (12) | 1 (4.2) | — |
| >150 000 | 5 (20) | 1 (4.2) | — |
| Unknown | 1 (0) | 2 (8.3) | — |
| Parent employment status, n (%) | |||
| Unemployed | 8 (30.8) | 6 (25) | .066 |
| Employed | 18 (69.2) | 15 (62.5) | — |
| Unknown | 0 (0) | 3 (12.5) | — |
The table reveals that there were no statistical differences in demographic characteristics between the study groups. GI, gastrointestinal; —, not applicable.
Among the 50 participating parents, the retention rate was 80% (40 of 50) at 1 month and 76% (38 of 50) at 3 months. Among the 24 families randomly assigned to MyChildCMC, 5 families used it only once at the time of enrollment. Of families who used the app after discharge, the engagement (daily vital sign and symptom recordings) was 68.3% at 1 month and 62.6% at 3 months, and the average length of time from the start to discontinuing use of the app was 78.4 days (median: 89 days), with a range of 13 to 91 days. There were no technical issues with the app or safety concerns during the study.
Ability to Identify Health Deteriorations
We created run charts using participant data with >60% engagement in recording signs and symptoms and whose CMC had a hospitalization during follow-up. In 13 of 15 (87%) patients who had subsequent admissions, run charts revealed shifts (warning signs) and reduced variability in signs and symptoms as common changes occurring 1 to 2 weeks before admissions (Figs 1–3, Supplemental Figs 6 and 7). Two patients had shifts without admissions.
Average HR of patient A: positive shift and reduced variability. We depict the health-rate run-chart plot for patient A, revealing a positive shift (multiple points above the mean) and reduced variability, in the week leading to the admission. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules in which data reveal significant changes. Control limits were (red dotted lines) determined with early data points when the parent indicated that the child’s health was “typical,” “usual,” or “normal state.”
Average HR of patient A: positive shift and reduced variability. We depict the health-rate run-chart plot for patient A, revealing a positive shift (multiple points above the mean) and reduced variability, in the week leading to the admission. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules in which data reveal significant changes. Control limits were (red dotted lines) determined with early data points when the parent indicated that the child’s health was “typical,” “usual,” or “normal state.”
Average temperature changes of patient B: negative shift and reduced variability. The figure reveals a negative shift (multiple points below the mean) and reduced variability in the temperature of patient B. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules, in which data reveal significant changes.
Average temperature changes of patient B: negative shift and reduced variability. The figure reveals a negative shift (multiple points below the mean) and reduced variability in the temperature of patient B. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules, in which data reveal significant changes.
Average HR of patient B: positive shift and reduced variability. The figure reveals patient B with a positive shift (multiple points above the mean) in HR and reduced variability, before the admission. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules, in which data reveal significant changes.
Average HR of patient B: positive shift and reduced variability. The figure reveals patient B with a positive shift (multiple points above the mean) in HR and reduced variability, before the admission. The shaded rectangle highlights the admission period, including the admission date and discharge date. The red points are related to run-chart rules, in which data reveal significant changes.
Preliminary Impact of the MyChildCMC App
The mean QoL score was 59.8 (standardized to 100) points in the MyChildCMC group and trended downward from 62.7 at baseline to 59.8 at 1 month and 55.2 at 3 months. Among controls, the mean QoL score was 52.5 and trended downward from 54.4 at baseline to 51.8 at 1 month and 50.8 at 3 months. QoL scores were not statistically significantly different (P > .05) at any assessment times between the groups. Two deaths (not related to the app use) and 1 hospice placement occurred in MyChildCMC group, and none occurred in controls.
The mean parent satisfaction with overall child’s care was high at 26.93 (SD = 2.22) points in the MyChildCMC group, compared with 24.14 (SD = 4.21) points in the control group (RR = 1.11; 95% CI = 1.01–1.22; P = .035), and subscale scores are provided in Supplemental Table 4, revealing the MyChildCMC group scoring significantly higher for medical and social support. Of the 15 participants randomly assigned to MyChildCMC and who responded to the 4 questions about the impact of the app on child’s health, 1 responded “did not know,” and 1 “strongly disagreed,” but 12 (80%) “agreed or strongly agreed” that the app made a positive contribution to their child’s health.
The mean number of ED and hospital admissions was 1.13 (median: 1 [interquartile range (IQR): 2]) before enrollment and 1.08 (median: 1 [IQR: 2]) after enrollment in the MyChildCMC group (RR = 0.96; 95% CI = 0.56–1.65; P = .891) and 0.85 (median: 1 [IQR: 1]) before and 0.88 (median: 0.5 (IQR: 1]) after enrollment in controls (RR = 1.05; 95% CI = 0.58–1.88; P = .882); there was no difference between the groups. The mean number of hospital days for the MyChildCMC group was 9.25 (median: 0 [IQR: 7.25]) days before and decreased to 4.54 (median: 0 [IQR: 5.75]) days (RR = 0.49; 95% CI = 0.39–0.62; P < .001) after enrollment but increased from 1.08 (median: 0 [IQR: 2]) before to 2.46 (median: 0 [IQR: 4.25]) days (RR = 2.29; 95% CI = 1.47–3.56; P < .001) after enrollment in the control group (Table 2).
ED and Hospital Admissions and Hospital Days (3 Months Pre- and 3 Months Postenrollment)
| No. Patients | No. ED and/or Hospital Admissions | Mean | SD | RR | 95% CI | P | |
|---|---|---|---|---|---|---|---|
| ED and/or hospital admissions | |||||||
| Intervention | |||||||
| Pre | 24 | 27 | 1.13 | 1.26 | — | — | — |
| Post | 24 | 26 | 1.08 | 1.25 | 0.96 | 0.56–1.65 | .891 |
| Controls | |||||||
| Pre | 26 | 22 | 0.85 | 1.05 | — | — | — |
| Post | 26 | 23 | 0.88 | 1.11 | 1.05 | 0.58–1.88 | .882 |
| Hospital days | |||||||
| Intervention | |||||||
| Pre | 24 | 222a | 9.25 | 18.30 | — | — | — |
| Post | 24 | 109a | 4.54 | 6.95 | 0.49 | 0.39–0.62 | <.001 |
| Controls | |||||||
| Pre | 26 | 28a | 1.08 | 1.88 | — | — | — |
| Post | 26 | 64a | 2.46 | 3.84 | 2.29 | 1.47–3.56 | <.001 |
| No. Patients | No. ED and/or Hospital Admissions | Mean | SD | RR | 95% CI | P | |
|---|---|---|---|---|---|---|---|
| ED and/or hospital admissions | |||||||
| Intervention | |||||||
| Pre | 24 | 27 | 1.13 | 1.26 | — | — | — |
| Post | 24 | 26 | 1.08 | 1.25 | 0.96 | 0.56–1.65 | .891 |
| Controls | |||||||
| Pre | 26 | 22 | 0.85 | 1.05 | — | — | — |
| Post | 26 | 23 | 0.88 | 1.11 | 1.05 | 0.58–1.88 | .882 |
| Hospital days | |||||||
| Intervention | |||||||
| Pre | 24 | 222a | 9.25 | 18.30 | — | — | — |
| Post | 24 | 109a | 4.54 | 6.95 | 0.49 | 0.39–0.62 | <.001 |
| Controls | |||||||
| Pre | 26 | 28a | 1.08 | 1.88 | — | — | — |
| Post | 26 | 64a | 2.46 | 3.84 | 2.29 | 1.47–3.56 | <.001 |
The table reveals that there was no change in the cumulative number of ED and hospital admissions that occurred pre-enrollment compared with the cumulative number of ED and hospital admissions occurring postenrollment, for both the MyChildCMC and control groups. However, there was a significant reduction in the cumulative number of hospital days for hospital admissions occurring postenrollment for the MyChildCMC group, whereas there was a significant increase in the cumulative number of hospital days in admissions occurring postenrollment for the control group. —, not applicable.
No. hospital days
Discussion
We found MyChildCMC use by parents feasible for CMC home monitoring, with high enrollment, retention, and engagement. Although QoL was higher in the MyChildCMC group, there was no statistical differences with controls. Yet, we found significantly greater parent satisfaction in child’s care and reduced hospital days among MyChildCMC users. MyChildCMC was able to detect signs (shifts) of acute health deteriorations 1 to 2 weeks before ED or hospital admissions in most (13 of 15) cases, meaning the app may be useful in alerting parents and their HCPs to promote timely home care to reduce ED and hospital use. Only 2 patients had shifts without admissions, possibly because of interventions at home.
We found no differences in ED- and hospital-admission frequency between groups. This may be due to our pilot study being underpowered for this outcome. The apparent reduction in hospital days in MyChildCMC users could be explained by real-time feedback empowering parents to be more vigilant at critical times, seek care, or intervene at home at earlier stages of deteriorations when the child’s condition is less severe, thus leading to shorter hospital stays when these children have unavoidable hospitalizations. However, we did not expect an increase in hospital days in controls and do not know the reasons for the increase.
Despite the availability of mobile phones and the relative low-cost opportunity to support the care of patients with chronic conditions,38–46,61 no study has ever been targeted at CMC with home monitoring apps.62–67 Available apps for home monitoring of children and adults are focused on noncomplex, single chronic conditions (eg, asthma, diabetes, etc).47–52 The variable, multisystem conditions of CMC make it difficult to develop home monitoring apps. Only Cheng et al68 reported a home monitoring app for CMC, but this work was focused only on a subset of children with enteral feeding tubes and limited to usability testing, with no evaluation of impact. MyChildCMC’s uniqueness is that it introduces a practical approach to support home monitoring by focusing on crosscutting symptoms that often precede ED and hospital admissions in CMC rather than using multiple apps to monitor each CMC’s chronic conditions individually. With our study, we are the first to provide evidence that an app can help monitor CMC’s health at home, which may lead to improved outcomes, supporting the need for more studies of home monitoring apps for CMC.
Currently, there is no benchmark for feasibility outcomes, and no study has been conducted on CMC to determine enrollment, retention, and engagement targets of a daily symptom monitoring app. Studies of mHealth apps deemed feasible have various feasibility goals. For instance, a study designed to monitor daily depressive symptoms achieved a 43% user-engagement rate after 1 month.69 Similarly, the retention rate was 53% at 1 month in a study monitoring daily psychiatric symptoms.70 Also, an app for monitoring mental health symptoms in young people was declared feasible, with a recruitment of 52% and engagement rate of 45% at 3 months.71 Although the enrollment rate was high (86%) in a diabetes self-management program, the retention rate at 3 months was only 50%, but the study was declared feasible.72 In a study of a pain monitoring app in sickle cell disease, researchers reported a daily engagement of 86% in the first week but only 58% in the fourth week.73 Although our app is different from typical home monitoring apps (which often are targeted at a single disease or symptom) and there is no study for comparison, we selected our feasibility target a priori, with a goal of achieving >60% enrollment, retention, and engagement rates, on the basis of our previous experience with an app for children with asthma74 and the best practice guidance defined for the diet and exercise app in uterine cancer survivors.54
Run-chart rules are widely used in quality improvement, although their application in monitoring vital signs and symptoms is not common.57,58 Because CMC often have atypical baselines for vital signs,75,76 run charts allowed us to (1) define individual patient’s baselines over 14 or more “normal” days as judged by their parents (studies report that ≥14 data points are needed to have adequate power to detect changes)77,78 and (2) identify significant changes from patient’s individualized baselines. We noted reduced variability in HRs as a signal of health status deterioration, which was also observed preceding preterm neonatal sepsis.79 Our plan included adding an alerting system74,80,81 for significant changes for parents and a population health dashboard for alerting HCPs that will allow real-time access to patient data to promote timely care.
Our study has several limitations, and results need to be interpreted with caution. This was a pilot study not powered for secondary outcomes but will inform a larger and definitive study of MyChildCMC’s impact and sustainability. In our study, we did not take into account planned admissions (eg, for sleep studies, surgeries, etc) in the ED and hospital admissions data, which may have biased our results. We grouped ED visits and hospital admissions as an outcome to capture early signs of child’s health deterioration that lead to any ED visit or hospital admission. Also, qualitative assessment was not included in our study but can be an important aspect of feasibility studies.
Although participants were randomly assigned, baseline hospital days were substantially higher in the MyChildCMC group. This may be due to our small sample size and skewing from a few patients, with more hospital days. However, the use of generalized linear models with logarithmic link and Poisson regression accounted for skewness of the data. Also, baselines in the number of ED and hospital admissions were similar in the groups, minimizing the risk of allocation bias. We also noted the random occurrence of 2 deaths and 1 hospice assignment in the MyChildCMC group, suggesting a sicker population in this group. This could be interpreted in 2 ways: our small sample size led to imbalance with regard to illness severity, suggesting caution in interpreting our results, or (2), despite sicker CMC being in the MyChildCMC group, we saw a significant drop in hospital days in these children, whereas hospital days increased in controls, suggesting the app’s effect size may be larger. In addition, we used intention-to-treat analysis to compare ED and hospital admissions and hospital days and did not exclude these 3 patients in our analysis. Even if we had excluded them, our results would reveal the same patterns of changes.
Although established baselines were stable, run charts reveal that vital signs post (subsequent not index) discharges were unstable for several days in some children. This could be due to either early discharge or a “signal” of an upcoming admission. Unfortunately, we did not collect data after 3 months. Our study was conducted in a single hospital with a less diverse patient population; therefore, generalizability of our results to other locations with diverse patients may be limited. Lastly, although parents reported it took ∼5 minutes per day to record vital signs and symptoms, we believe that automating vital sign data collection with a sensor device will minimize data burden and may increase long-term engagement.82
Conclusions
We found MyChildCMC a feasible app for supporting home monitoring of CMC and may be effective in preventing recurrent ED and hospital admissions and/or reducing hospital days. In future studies, researchers will test its impact on a larger sample and will include a real-time alerting system for home parents and professional care teams to allow timely care in the home and facilitate medical decision-making that can impact ED and hospital visits.
Deidentified individual participant data (including data dictionaries), in addition to study protocols, the statistical analysis plan, and the informed consent form, will be made available to other researchers on request. Proposals should be submitted to [email protected].
Dr Nkoy conceptualized and designed the study, drafted the initial manuscript, and oversaw the analysis; Drs Stone participated in the concept and design of the study, data analysis, interpretation of data, and manuscript revision; Drs Hofmann, Fassl, and Murphy and Mrs Zhu and Mrs Mahtta participated in the study concept and design, interpretation of data, and revision of the manuscript; and all authors have approved the final manuscript as submitted.
FUNDING: Supported by a grant from the Health Resources and Services Administration. The sponsors did not participate in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the article. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of the Health Resources and Services Administration.
Comments