Health Information Technology use among healthcare providers treating children and adolescents with obesity: A systematic review

Abstract

Purpose of review:

A systematic review was performed to synthesize the latest evidence of health information technology (HIT) used by healthcare professionals to address overweight/obesity among children and adolescents.

Recent findings:

This literature has not been summarized since 2012 despite the rapidly changing nature of HIT, the millions of federal healthcare dollars invested in the last decade, and the continued burden of childhood obesity and associated metabolic diseases.

Summary:

Three databases were searched from January, 2012 to August, 2020. Experimental and quasi-experimental studies, including pilot and feasibility studies, using HIT to address obesity and behavior (physical activity and nutrition) change among children and adolescents aged 2–18 years were included. Studies were required to report the impact on clinical/behavioral outcomes or care processes (e.g., BMI screening). Twenty-five studies met inclusion criteria. In sum, HIT tools improved care processes (e.g., screening rates), yet it remains unclear whether improvements in care processes translate to changes in clinical/behavioral outcomes (e.g., BMI). The majority of studies included implementation outcomes; yet outcomes were mainly operationalized as satisfaction and lacked diversity. Studies commonly used a patient-centered approach, yet rarely employed precision-medicine to individualize care based on the patient’s data. Few studies specifically targeted disadvantaged populations or social determinants of health and/or used behavioral theory. High quality, rigorous efficacy and effectiveness studies are needed to address the knowledge gaps in this emerging field of science.

Introduction

The World Health Organization declared obesity prevention an international public health priority. However, for over the last three decades, obesity prevalence among children and adolescents has increased.[1] Worldwide, over 20% of children and adolescents have obesity, with even higher rates among disadvantaged populations.[2] Once established, obesity is difficult to reverse,[3] and adolescents with obesity are at greater risk for developing metabolic risk factors (e.g., dyslipidemia, hypertension, glucose intolerance) than their normal weight counterparts.[4, 5] Children and adolescents who are overweight or have obesity are also at greater risk of health problems as adults including cardiovascular disease, cancer, insulin resistance, osteoarthritis, and disability.[1] Pediatric primary prevention and treatment of metabolic risk factors, commonly known as metabolic syndrome (MetS), are key. While wide variation exists in defining and measuring MetS among children and adolescents,[6–8] the prevention and treatment of obesity is recommended as a first-line approach to attenuate risk factors associated with MetS (e.g., dyslipidemia and hypertension).[5] Interventions targeting physical activity and nutrition decrease children’s weight and also improve risk factors associated with MetS.[6, 9, 10]

Healthcare professionals, particularly primary care providers, reach a large number of adolescents and have the potential to address and treat obesity during clinical visits.[11] Yet less than half of children who see a pediatric provider receive body mass index (BMI) screening and/or counseling for physical activity and nutrition as recommended by the American Academy of Pediatrics.[12–15] Within the clinical care setting, healthcare professionals lack evidence-based training and tools (e.g., health information technologies) that are effective and sustainable. Providers need rigorously tested, efficacious and sustainable tools to help them manage the burden of childhood obesity, particularly among disadvantaged populations facing greater barriers to health and wellness.[16, 17]

Health Information Technology (HIT) has rapidly emerged to improve the efficiency, cost-effectiveness, quality, and safety of medical care.[18–20] Common forms of HIT include electronic health records (EHRs), computerized decision support systems, applications, websites, telemedicine, text message and telephone support. HIT tools can be combined with informatics to support clinical decision-making, care delivery processes and/or population health strategies.

These tools have the potential to incorporate precision medicine [21] techniques (i.e., advanced analytics, machine learning) to risk stratify individuals into distinct groups. Once stratified, HIT can help provide individually targeted interventions based on factors such as social determinants of health (SDOH) or behavioral indicators. In essence, HIT and precision medicine can build a sociobehavioral phenotype and align specific evidence-based interventions to sociobehavioral disparities to generate equitable change across populations.[22] Moreover, HIT tools can systematically generate evidence and new knowledge about these approaches through routine practice.[23] Such evidence is necessary to build the scientific foundations to support and improve HIT tools. For example, our understanding is limited around issues of clinical end-user adoption, utilization, interface design and usability, and how these concerns impact workflow, productivity, and the sustainability of HIT.[23–25]

Several systematic reviews have examined the use of digital health (including HIT and mobile health [mHealth]) for behavior change and mainly identified technologies designed for school-based or community settings.[26–28] One review by Smith and colleagues (2013) focused on the use of HIT in clinical settings for the screening and treatment of childhood obesity.[29] HIT studies published from 2006–2012 (N=13) improved screening rates and access to treatment yet did not demonstrate substantial impact on behaviors and weight outcomes.[29] No review has examined this body of literature since 2012 despite the rapidly changing nature of HIT and the millions of federal healthcare dollars invested in the last decade.[30] The purpose of this review was to develop an understanding of the current state of the science. We sought to synthesize the evidence of HIT used by healthcare professionals to prevent or treat overweight/obesity among children and adolescents. Additionally, this review sought to examine whether HIT tools are: 1) providing the necessary implementation evidence (e.g., adoption, utilization, feasibility) needed to support optimal integration and sustainability of HIT in clinical settings; 2) being used to individualize care (e.g., precision medicine); 3) targeting disadvantaged populations and SDOH; and 4) using behavioral theory in their design.

Methods

This systematic review followed PRISMA reporting guidelines as shown in the checklist (Appendix 1).

Data Sources and Search Terms

We used a combination of standardized index terms and plain language terms to identify English language, experimental studies using HIT to prevent and treat obesity in children (Appendix 2: full search strategy). We searched EBSCO CINAHL, EBSCO Medline, and Scopus from 2012 to 2020. We exported 9,251 search results to EndNote. After deduplication, the remaining 6,826 citations were exported into Covidence systematic review software.[31] Covidence automatically located and removed an additional 204 duplicate citations, leaving 6,622 unique citations to screen (Figure 1).

Figure 1: PRISMA 2009 Flow Diagram.

Note: The shaded vertical boxes on the left represent each review stage. The arrows between the center boxes depict steps to the subsequent stage. Arrows to the right connecting to boxes on the right represent articles omitted at each screening stage.

Selection Criteria

Experimental and quasi-experimental studies (e.g., randomized controlled trials, non-randomized trials, pre-post studies), including pilot and feasibility studies, targeting obesity and behavior change among children and adolescents aged 2–18 years were considered. Studies were included if at least one component of the intervention included HIT used by a healthcare professional (e.g., physician, nurse, social worker, dietician). HIT was defined broadly as: electronic health records (EHRs), computerized decision support tools (CDSTs), telehealth (telemedicine, video conferencing, telephone support), smartphone applications, websites, text messaging, and emails. Studies were required to report measures of clinical effectiveness (changes in weight-related measures, physical activity, food intake, or treatment adherence) and/or care processes (e.g., BMI screening, comorbidity testing, counseling on behavior change).

Screening Procedures

Our team consisted of four coders. We used a dual independent process in which two coders reviewed title and abstracts using Covidence [31] and voted anonymously on inclusion or exclusion. Coders reviewed discrepancies on a weekly basis; if coding pairs could not resolve discrepancies, the first author made a final decision. A similar approach was used at the full text review stage, with the addition of an explanatory exclusion code assigned to all articles not included in the study. We applied a hierarchical exclusion code such that the first applicable pre-determined exclusion rule was applied in descending order (e.g. if an article could be excluded on exclusion code 4 and code 6, we applied code 4; Appendix 3).

Data Extraction and Synthesis

Two authors developed and piloted a data extraction procedures manual (Appendix 4), and the full team provided feedback on procedures and variables for extraction. Upon finalizing the manual, we conducted dual non-independent data extraction in which a primary coder extracted all relevant data and a second coder checked extraction for completeness and accuracy. For studies with multiple articles reporting results from the same trial, we extracted the oldest articles first. We then reviewed the more recent article(s) to determine if any new information was available to add to the extraction. Each study was represented by a single row in the extraction database (and subsequent results tables). Coders met on a weekly basis to resolve discrepancies. Authors resolved disagreements in consultation with the first author.

Extraction elements included information about: 1) study characteristics (e.g., location, target population, inclusion of parents, study design, intervention duration and follow-up); 2) outcomes (primary and secondary measures, and reported change); and 3) intervention characteristics (e.g., intervention description, HIT type, intended user, patient-centered care, and/or precision medicine, a focus on disadvantaged populations/SDOH, and use of behavioral theory). Outcomes and change in outcomes were extracted for all clinical/behavioral outcomes (e.g., weight, fruit and vegetable intake) and care processes (e.g., BMI screening). Additionally, we extracted implementation outcomes, previously explicated by Proctor and colleagues [32] These included: acceptability, adoption, appropriateness, cost, feasibility, fidelity, penetration, and sustainability. Such measures were of interest as they are critical to providing evidence to combat concerns regarding the uptake and routine use of HIT during its emergence, particularly for obesity management.

Precision medicine was defined as the use of an innovative approach that takes into account individual differences in patients’ genes, environments, and lifestyles.[21] Studies were coded as using a precision medicine if the approach took action based on an individuals’ “below the skin” networks of metabolism, epigenetics, and genetics and/or “above the skin” factors such as psychosocial barriers and families’ economic or social circumstances.[33] Only approaches that used data to systematically tailor/individualize care (e.g., tailored lifestyle change program based on current behaviors) were considered to use precision medicine. Patient-centered care was defined as the patient and the healthcare provider interacting dynamically, collaborating and deliberating about the best plan of action based on a broad range of personal information.[34]

Quality Assessment

Raters conducted dual independent assessment of study quality using criteria applied in previous studies of HIT interventions.[29, 35–37] Disagreements were resolved in the same manner as the procedures for data extraction consensus described above. The rating scale consisted of criteria assessing the following 5 aspects of methodological rigor: 1) method of allocation to study group; 2) unit of allocation; 3) baseline differences between groups that could potentially be linked to study outcomes; 4) outcome objectivity; and 5) rates of attrition. Each criterion was scored two, one, or zero points; see Table 2 for full scoring information. The scores on each criterion were summed to determine an overall quality score for each study; the maximum quality score was 10 points.

Table 2:

Study Quality (N=25)

Study	Allocation method	Unit of allocation	Baseline differences between groups	Outcome objectivity	Attrition	Total
Pilot Studies (N=7)				Pilot study mean (SD): 4.7 (2.1)
Altamirano-Diaz, 2017 [52]	0	0	1	2	2	5
Fleischman, 2016 [38]	2	0	2	2	1	7
Moschonis, 2019 [39]	2	0	2	2	1	7
Pbert, 2016 [63]	0	0	2	2	2	6
Slusser, 2016 [54]	0	0	0	0	2	2
Tripicchio, 2017 [64]	0	0	1	2	0	3
Wright, 2013 [40]	2	0	0	0	1	3
Full Trials (N=10)				Full trial mean (SD): 5.8 (2.2)
Bohlin, 2017 [85]	2	0	2	2	0	6
Currie, 2018 [55]	2	0	2	2	0	6
Lipana, 2013 [51]	0	0	1	0	0	1
Mameli, 2018 [57]	2	0	2	1	0	5
Nguyen, 2013 [56]	2	0	2	2	0	6
Patrick, 2013 [62]	2	0	2	2	0	6
Rhodes, 2017 [53]	2	2	2	1	2	9
Sherwood, 2019 [58]	2	0	2	2	1	7
Taveras, 2017b [50]	0	0	2	2	0	4
Connect for Health (Bala 2019, Taveras 2015 & 2017) [59–61]	2	0	2	2	2	8
Quality Improvement Studies (N=8)				QI study mean (SD): 4.6 (2.3)
Andrade, 2020 [45]	0	2	0	1	2	5
Bode, 2013 [46]	0	2	1	2	0	5
Keehbauch, 2012 [47]	1	1	1	2	2	7
Myers, 2020 [65]	0	0	0	2	0	2
Wray, 2018 [86]	0	0	0	1	0	1
HEALTH COP (Shaikh 2014 & 2015) [48,49]	0	1	2	1	0	4
HealthSmart Kids (Gance-Cleveland 2016 & 2017) [42,43]	2	2	0	1	0	5
STAR (Ayash 2013, Taveras 2014) [41,44]	2	2	2	2	0	8

Note: Quality scoring criteria: a. Method of allocation to study group (2= random assignment, 1= quasi-random assignment, 0= concurrent or no controls); b. Unit of allocation (2= cluster, 1=health professional, 0= patient or not applicable); c. Baseline differences between groups that may be linked to study outcomes (2= no baseline differences or differences adjusted with appropriate statistical adjustment, 1= baseline difference reported and no statistical adjustment, 0=baseline characteristics not reported); d. Outcome objectivity (2= objective outcomes, or subjective outcomes with blinding, 1= validated subjective outcomes [e.g. validated survey], 0=subjective outcomes poorly defined); e. Attrition (2= less than 10%, 1= 10–20%, 0= greater than 20%, not reported, or not applicable)

Results

Abstract/title screening was performed on 6,622 articles, yielding 124 full text articles examined for data extraction. Ninety-four articles were excluded at the full text review stage; the most common reasons for exclusion were ineligible study designs (N=30) and ineligible intended users (N=28). A total of 30 articles were included for data extraction (see Appendix 1). There were 25 unique studies; 21 studies were represented by a single article, while 4 studies had two or more published articles that were extracted together. We report on the results at the study level (N=25).

Description of Studies

Nearly three-quarters of studies (N=18, 76%) were conducted in the United States; three studies were conducted in Canada, three in European countries, and one in Australia. Most studies took place in an outpatient pediatric or primary care clinic (N = 19, 76%), with several studies occurring in more than one clinical setting (e.g. children’s hospital, family clinics). Of the 25 included articles, 7 (28%) were pilot/feasibility studies, 8 (32%) were quality improvement (QI) studies that aimed to improve care processes, and 10 (40%) were full trials (i.e. studies powered to detect statistically significant differences between pre- and post-intervention measurement and/or between intervention and control groups). Thirteen studies (52%) were RCTs including 3 pilot/feasibility RCTs [38–40] (including 1 cross-over RCT [38]) and 2 quality-improvement RCTs.[41–44] Five studies used some form of pre and post design, 4 of which reviewed EHR data.[40, 45–47] The remaining studies were quasi-experimental,[48–51] randomized chart review,[47] single-arm quantitative longitudinal, [52] and single-arm mixed methods feasibility [45] studies. Most studies (N=20, 80%) included an intervention period 12 months or less in duration, and only 7 studies included any long-term follow-up measurement (i.e. measurement at an additional time point beyond the assessment immediately post-intervention). Twelve studies (48%) had sample sizes over 100 participants, of which 6 were full trials with all others being QI studies. A similar number (N=9, 36%) were smaller studies with 50 or fewer participants. Nine (36%) family interventions targeted parents and children together, while the majority of studies (N=13, 52%) did not target parents or only targeted parents in a supplementary role; only three (12%) studies included parents as the primary target. See Table 1 for full report of study characteristics.

Table 1:

Study characteristics (N=25)

Study	Setting	Intervention Target Pop.	Inclusion of parents	Study Design	Intervention Duration	Follow-up measurement
Pilot Studies (n=7)
Altamirano-Diaz, 2017 [52]	health center (cardiology)	youth previously diagnosed with congenital heart disease (34)	did not include parents	single-arm quantitative longitudinal	12 months	no
Fleischman, 2016 [38]	pediatric clinic, children’s hospital	children 10–17 years BMI 95th+ percentile (38)	targeted parent and child	crossover RCT	6 months	12 months
Moschonis, 2019 [39]	children’s hospitals (2)	children 6–12 years (65)	included parents, but not as primary targets	RCT	3 months	no
Pbert, 2016 [63]	pediatric primary care clinic	children 8–12 years (37)	parents were primary target	nonrandomized intervention study with contemporaneous control	3 months	no
Slusser, 2016 [54]	children’s hospital, school clinic, family clinic	children, average 11 years (28)	included parents but not as primary targets	pre-post	one clinic visit	no
Tripicchio, 2017 [64]	pediatric clinics	children 2–18 years (64)	targeted parent and child	non-randomized, pre-post comparing 3 intervention groups	3 months	no
Wright, 2013 [40]	pediatric primary care clinic	children 9–12 years (50)	targeted parent and child	RCT	3 months	No
Full Trials (n=10)
Bohlin, 2017 [85]	pediatric clinic	children 5–14 years (37)	targeted parent and child	RCT	18 months	10 months
Currie, 2018 [55]	pediatric obesity clinic	adolescents 12–18 years (64)	did not include parents	RCT	7 weeks	no
Lipana, 2013 [51]	weight management clinic	Children mean age 11 years (243)	did not include parents	retrospective quasi-experimental comparative effectiveness trial	NA	no
Mameli, 2018 [57]	hospital obesity clinic	children 10–17 years BMI 95th+ percentile (43)	included parents but not as primary targets	RCT	3 months	no
Nguyen, 2013 [56]	community health center, children’s hospital	adolescents 13–16 years (151)	targeted parent and child	RCT	12 months	24 months
Patrick, 2013 [62]	pediatric clinics (18)	adolescents 12–16 years (101)	targeted parent and child	RCT	6 months	12 months
Rhodes, 2017 [53]	pediatric/family practice clinics (8)	children 5–10 years (22)	parents were primary target	RCT	5 weeks	no
Sherwood, 2019 [58]	pediatric primary care clinics (20)	children 5–10 years (421)	parents were primary target	RCT	12 months	24 months
Taveras, 2017b [50]	federally qualified health centers	children 2–12 years (1,479)	included parents but not as primary targets	quasi-experimental with comparator group	24 months	no
Connect for Health (Bala 2019, Taveras 2015 & 2017) [59–61]	pediatric primary care clinics	children 2–12 years (721)	targeted parent and child	RCT	12 months	no
Quality Improvement Studies (n=8)
Andrade, 2020 [45]	primary care clinics (5)	children 18 months −5 years (280)	included parents but not as primary targets	single-arm mixed methods feasibility	12 months	NA
Bode, 2013 [46]	U.S. military adolescent medicine clinic	adolescents 12–19 years (661)	did not include parents	pre-post	1 month	no
Keehbauch, 2012 [47]	family medicine clinics (2)	children 2–19 years (2340); physicians (63)	did not include parents	randomized chart review	NR	no
Myers, 2020 [65]	academic children’s hospital (2)	children 3–17 years (NR); physicians (15)	did not include parents	pre-post	12 months	12 months
Wray, 2018 [86]	pediatric primary care clinics	children 4–7 years (882)	did not include parents	pre-post	6 months	no
HEALTH COP (Shaikh 2014 & 2015) [48, 49]	rural health clinics	children 2–11 years (288)	included parents but not as primary targets	pre-post quasi-experimental	9 months	no
HealthSmart Kids (Gance-Cleveland 2016 & 2017) [42, 43]	school-based health centers (24)	2016: children 5–12 years charts (2153); providers (33) 2017: parents of children 5–12 (570); providers (31)	targeted parent and child	prospective mixed methods, cluster randomized quality improvement study (2016); Comparative effectiveness RCT (2017)	NR (2016 & 2017)	6 months (2016 & 2017)
STAR (Ayash 2013, Taveras 2014) [41, 44]	pediatric clinics (14)	children 6–12 years (549)	targeted parent and child	cluster RCT (2014)	12 months	no

Note: Numerical values in parentheses represent sample sizes; NR= not reported; NA= not applicable

Methodological Quality Assessment

Quality assessment results are reported in Table 2. Included studies were of moderate quality (mean = 5.12, SD = 2.18). Studies varied considerably in quality, ranging from 1–9 out of a possible 10 points. Full trials tended to be of higher quality (mean score = 5.8, SD = 2.2), while QI studies (mean score = 4.6, SD = 2.3) and pilot studies (mean score = 4.7, SD = 2.1) had nearly equal mean quality ratings, on average scoring slightly below the scale mid-point. Only one study earned a non-zero score for each of the five quality assessment criterion.[53] Unit of allocation was the least frequently met criteria (i.e. allocation occurred at the patient-level which may introduce provider-level biases). Overall, studies scored well on outcome objectivity (i.e. used objective measures and/or blinded subjective measures) and attrition (i.e. attrition rates at or below 10%).

Study Outcomes and Measurement

Sixteen studies (64%) included clinical and/or behavioral primary outcomes (e.g., change in BMI, change in physical activity), while the eight QI studies examined care processes (e.g. change in EHR documentation) as primary outcomes. One full-trial included both clinical and care processes as primary outcomes.[51] Sixteen studies (64%) included an implementation-focused outcome as either a primary or secondary outcome, though only one focused on implementation as an exclusive primary outcome,[54] which measured patient satisfaction (satisfaction was classified as acceptability per the Proctor et al. framework).[32] Nearly all studies focused on changes related to BMI as primary outcomes. Fourteen studies (56%) measured changes in BMI, while seven studies (28%) examined BMI-related documentation or correct diagnoses in the EHR as their primary outcome. Three studies (12%) [52, 53, 55] reported behavioral outcomes as one of their primary targets, with an additional ten studies including them as secondary outcomes. MetS risk factors were assessed as secondary outcomes in four studies.[38, 52, 54, 56] Two QI studies assessed MetS related practices: ordering of lipid panels[47] and laboratory assessment adherence.[42]

All ten full trials examined a primary behavioral or clinical outcome. Eight examined change in a BMI indicator (e.g. BMI z-score, BMI standard deviation), while one examined weight,[57] one targeted dietary behavior change,[53] and one included both BMI and physical activity.[55] Only one study found a statistically significant change in BMI, and only for participants in one of the two study sites.[50] Four trials saw decreases in BMI that approached but did not reach statistical significance, or both intervention and control groups decreased but the differences between the groups were not statistically significant.[55, 56, 58–61] The trial examining dietary behavior change as a primary outcome saw a non-statistically significant change in glycemic load, but no change in fat intake.[53] Similarly, the trial reporting physical activity change demonstrated a non-statistically significant change in number of steps.[55] Three [51, 57, 62] of the six studies that included behaviors as secondary outcomes reported significant results. One study included MetS risk factors as secondary outcomes with significant impact on systolic blood pressure, but no impact on the eight other risk factors examined.[56]

Six of seven pilot trials also examined primary behavioral or clinical outcomes (one [54] had clinical indicators as secondary outcomes). Although we would not expect pilot studies to be powered to detect statistically meaningful changes in primary outcomes, all detected a positive signal that may be further tested for significance in larger trials. Two studies found a statistically significant change in BMI, though with small sample sizes (N=34–37).[52, 63] An additional four pilot studies [38–40, 64] found changes in BMI that were not statistically significant or in which both comparison and control group BMIs changed, but the difference between the groups was not statistically significant. Of the five studies that examined behavioral outcomes, one study reported at least one statistically significant behavioral change.[39] Across three studies, there was no statistically significant impact on MetS risk factors included as secondary outcomes.[38, 52, 54]

All QI studies examined changes in care processes as primary outcomes. Seven of these studies were concerned with BMI or BMI-related (e.g. growth curve) documentation in the EHR. Of these, six saw a statistically significant change. Two studies also found statistically significant differences in secondary outcomes related to BMI diagnosis, including dietary counseling, ordering lipid labs, and referral to weight management services.[47, 65]

HIT interventions and novel approaches

Table 4 provides a brief description of the study intervention, including the type of HIT tested. Five studies (20%) tested a bundle of two or more HITs together or compared multiple HIT approaches; the majority of studies (N=20, 80%) tested only one kind of HIT. Telehealth, which included virtual care visits, telephone support, and video conferencing, was the most frequently utilized HIT (N=15, 60%). Eleven studies (44%) utilized a CDST (e.g. EHR best practice alert, automated calculation of BMI percentile). All other HITs, including smartphone applications, websites, email, and text messaging were used in seven instances across three studies, and were always used in conjunction with another HIT. The majority of studies (N=14, 56%) utilized a patient-centered approach, such as tailored counseling and targeted goal setting. Seven studies (28%) integrated precision medicine approaches (e.g. individualized risk scores based on clinical and non-clinical factors, tailored recommendations based on risk data) and four studies (16%) used patient-centered and precision medicine approaches together. Six studies (24%) explicitly indicated using HIT with a disadvantaged population (e.g. low income, uninsured) or to intervene on SDOH (e.g. providing community resources).[39, 40, 42, 45, 57, 62] Only four studies (16%) reported using a behavioral theory (e.g. social cognitive theory, theoretical model of behavior change) to inform their HIT intervention.[40, 58, 62, 63]

Table 4:

Description of HIT interventions

Author	Intervention	HIT Type	Intended HIT user	Comparator	Patient-centered	Precision Medicine
Pilot Studies
Altamirano-Diaz, 2017 [52]	tailored telephone counseling focused on education, behaviors, and family engagement	telehealth	patients, dietitians, fitness specialists	NA	yes; tailored counseling	no
Fleischman, 2016 [38]	virtual visits with obesity specialists	telehealth	dietitians, psychologists	UC or no-contact waitlist control	yes; tailored counseling	no
Moschonis, 2019 [39]	computerized CDST to provide personalized lifestyle recommendations and weekly meal plans	CDST	pediatric healthcare professionals	attention control without technology component	no	yes; tailored meal plans and lifestyle recommendations using decision tree algorithms
Pbert, 2016 [63]	telephone counseling sessions (FITLINE) to help child make dietary and activity changes; physician referral to and communication from FITLINE	telehealth	dieticians	UC or no-contact waitlist control	yes; tailored counseling with motivational interviewing, targeted lifestyle goal setting.	no
Slusser, 2016 [54]	telemedicine clinic visit for obesity care	telehealth	pediatricians, dieticians, psychologists, surgeons	NA	no	no
Tripicchio, 2017 [64]	family-based behavioral program plus virtual health coaching sessions; a patient fitness smartphone application (FITNET)	telehealth, smartphone application	trained health coaches	a. active control without technology component b. receiving intervention with active control (app)	Yes; tailored health coaching using motivational interviewing.	no
Wright, 2013 [40]	telephone counseling delivered by automated IVR software; EHR behavioral counseling tool	telehealth, CDST	primary care providers	UC or no-contact waitlist control	yes; tailored counseling	yes; counseling and clinical decision support tailored based on data from IVR phone calls; IVR technology tailors communication based on patient response
Full Trials
Bohlin, 2017 [85]	family-based telephone counseling	telehealth	nurses	UC or no-contact waitlist control	yes; tailored and flexible frequency counseling	no
Currie, 2018 [55]	weekly telephone counseling; adolescents provided with pedometer	telehealth	clinical research assistants	UC or no-contact waitlist control	yes; tailored counseling with motivational interviewing, goal setting and action planning	no
Lipana, 2013 [51]	telehealth delivered weight management therapy	telehealth	providers	UC or no-contact waitlist control	no	no
Mameli, 2018 [57]	web platform populated with patient energy expenditure data from a wristband and energy intake data from smartphone app used to develop personalized lifestyle programs; SMS text messages to patients	website, text messaging	pediatric endocrinologists, sports medicine physicians, dieticians	active control without technology component	no	yes; tailored lifestyle change program based on objectively measured physical activity data
Nguyen, 2013 [56]	telephone coaching, short-message-service and/or email communication added as follow-up to a 12-month weight-management program	telehealth, text messaging, email	dieticians, community health workers	active control without technology component	no	no
Patrick, 2013 [62]	interactive website for adolescents and parents; adolescents received pedometer and scale; text-messaging with health counselor	website, text messaging	health counselors	a. UC or no-contact waitlist control b. receiving intervention with active control (web only) c. receiving intervention with active control (web + group sessions)	yes; tailored counseling with goal setting and focus on health challenges.	yes; counseling and text messages tailored based on data from patient-facing interactive website.
Rhodes, 2017 [53]	weekly telephone dietary counseling sessions	telehealth	dieticians	receiving intervention with active control	yes; tailored counseling	no
Sherwood, 2019 [58]	pediatric primary care provider counseling (in-person); parent-targeted telephone coaching for obesity	telehealth	trained health coaches	attention control with technology component	yes; tailored counseling using motivational interviewing for behavior, self-assessment, behavior change prioritization, and goal setting.	no
Taveras, 2017b [50]	advanced provider training, CDST for obesity management; multidisciplinary weight management program within the FQHCs; integration of CHWs; FQHC environmental changes	CDST	primary care providers	UC or no-contact waitlist control	yes; tailored counseling using motivational interviewing and goal-setting techniques	no
Connect for Health (Bala 2019, Taveras 2015 & 2017) [59–61]	CDST to alert and assist in treating children with overweight or obesity; telephone and video health coaching conducted with families	CDST, telehealth	physicians, health coaches	active control without technology component	yes; tailored health coaching using motivational interviewing and shared decision making.	yes; text message counseling tailored to patient self-assessment feedback.
Quality Improvement Studies
Andrade, 2020 [45]	EHR-integrated nutrition risk screening tool; flagged eligible patients in EHR; parent resource materials and community referrals	CDST	nurses, dietitians	NA	no	yes; generates individualized risk score based on clinical and non-clinical factors
Bode, 2013 [46]	BMI percentage and display of growth curve in the EHR	CDST	primary care team	NA	no	no
Keehbauch, 2012 [47]	automated BMI percentage calculation in EHR; provider education; pocket algorithm decision aids for screening and management	CDST	physicians	active control without provider education	yes; tailored counseling	no
Myers, 2020 [65]	online provider training; CDST (EHR prompts, computerized physician order entry sets and template documentation) and other processes (e.g., scripts) to improve the identification and management of obesity	CDST	physicians	NA	no	no
Wray, 2018 [86]	EHR reminder alert tool for height and weight documentation	CDST	primary care providers	NA	no	no
HEALTH COP (Shaikh 2014 & 2015) [48,49]	videoconferencing learning sessions; clinician decision support; clinical information systems training; community resource database; delivery system design (including standardized documentation forms); telehealth visits	telehealth	providers	NA	no	no
HealthSmart Kids (Gance-Cleveland 2016 & 2017) [42,43]	HIT system w/clinical decision support which generates tailored patient education materials	CDST	providers	active control without technology component	yes; tailored counseling using motivation interviewing and goal setting.	yes; CDST provides tailored behavior change recommendations based on patient’s clinical information (e.g., obesity, risk factors)
STAR (Ayash 2013, Taveras 2014) [41,44]	computerized, point-of-care alert at well-child visit for a child with a BMI 95th+ percentile; information regarding child’s BMI to parents prior to visit	CDST	pediatricians	a. UC or no-contact waitlist control b. receiving intervention with active control	no	no

Note: EHR= electronic health record; CDST = clinical decision-support tool; Telehealth= virtual care visits, telephone support and videoconferencing; UC = usual care; IVR= Interactive Voice Response

Conclusions

The pediatric obesity epidemic continues to challenge healthcare systems to identify novel solutions to prevent and manage obesity to reduce the burden of metabolic diseases and associated health care costs.[5, 66, 67] HIT has rapidly emerged with potential to improve healthcare quality and inform new approaches (e.g., precision medicine) to healthcare delivery for childhood obesity that will promote population health and reduce health disparities.[23] This review synthesizes the evidence of HIT interventions used by healthcare professionals to address pediatric overweight/obesity to update the literature from a review published in 2014.[29] We further expand the literature by documenting: 1) aspects of implementing HIT that may inform solutions to adoption and sustainability, and 2) the use of approaches (e.g., precision medicine) that may further the impact of HITs by tailoring evidence-based approaches to individuals, populations, and environments.

Our findings are in line with the previous review;[29] HIT tools improve care processes (e.g., screening rates), yet these improvements rarely translate to changes in clinical outcomes (e.g., BMI). The effect of HIT on behavioral outcomes was not the primary focus of included studies and findings varied. While clinical outcomes were primarily measured objectively, behaviors tended to be measured using self- or parent-reported surveys rather than objective methods (e.g., accelerometry) that reduce data bias.[68] Studies had short intervention durations and follow-up periods that may not be sufficient to produce or detect change in behavioral or clinical outcomes. Furthermore, HIT tools may be insufficient without a robust behavior change curriculum or intervention and/or the healthcare provider may not be the critical end-user of these HITs. HIT components directed to the child and/or parent or in conjunction with the provider may be necessary for behavior change. Our ability to draw conclusions, even from full trials, was limited by small sample sizes and study quality. It is important to note that HIT, particularly for pediatric obesity management, is an emergent and relatively young discipline, and thus an area in which small-scale studies may be expected. Compared to studies (n=13) published from 2006–2012,[29] HIT interventions for pediatric obesity management have increased in number and quality (3.7 vs. 5.2 mean score) indicating some level of growth in this area of research.

Beyond efficacy and effectiveness, this review determined if and how studies are assessing implementation outcomes that generate evidence for the adoption and sustainability of HIT tools for obesity management in clinical settings. The majority of studies included implementation measures, yet the types of implementation measures were limited. Most studies included patient and/or provider acceptability (i.e. satisfaction); other important implementation outcomes (e.g., appropriateness, cost, penetration, and sustainability) were not measured which may result in critical gaps in evidence and scientific advancement. Without information on these types of outcomes, it is difficult to ascertain whether the limited evidence for the effectiveness of the HITs is due to the tools themselves, or the extent to which they were optimally and sustainably integrated into clinical care. The use of implementation science theory and principles is necessary to generate a richer understanding of the complex “real-world” issues that influence the translation of HIT from research to practice to widespread adoption and maintenance.[18, 69]

To our knowledge, this is the first review to explore the use of patient-centered care[34] and precision medicine[21, 33] in HIT interventions directed at healthcare providers for preventing or treating pediatric obesity. HIT improves the feasibility of individualizing care and these approaches improve healthcare quality for diverse populations while reducing costs.[70–72] Precision medicine gives clinicians the tools to better understand the complex biological and environmental mechanisms underlying a patient’s health to better align prevention and treatment for the individual. Additionally, patient-centered care can account for the complexity and multiplicity of culturally-based beliefs, traditions, weight norms, and the larger social contexts in which individuals live through deliberate patient-provider collaboration. HIT interventions were commonly delivered using a patient-centered approach (e.g., tailored telephone counseling with individual goal-setting); however, studies rarely identified their approach as patient-centered and often lacked details regarding how providers collaborated with patients to identify the best plan based on a broad range of personal factors. When cited, methods for patient-provider interaction included motivational interviewing, goal setting and prioritization, and shared decision making and are in alignment with other patient-centered research.[72–74] Conversely, few studies used a precision-medicine approach that systematically used patient data to individualize treatment strategies. These few studies tailored approaches based on behavioral/clinical data (e.g., physical activity, obesity-status) and patient preferences; no studies were tailored based on genomic, environmental or other individual data. Inclusion of such data may maximize the potential of precision medicine.[75] HIT interventions in this review rarely targeted disadvantaged populations or SDOH thus limiting their ability to benefit populations particularly vulnerable to pediatric obesity and MetS. Continuing to develop and use HIT approaches designed for the average patient that fail to reflect the social diversity of real-world settings may exacerbate disparities. Failing to include minoritized and socially disadvantaged populations during this resource-intense era of HIT development and research continues a troubling trend in healthcare.[76]

HIT interventions lacked an underlying behavioral theory which may further limit impact on patient behaviors that are essential for the prevention and management of obesity.[77] Lack of the use of theory in the design and implementation of HIT is not new [78–82] and continues to limit the effectiveness of the interventions and scientific advancement.[83] Future research would likely benefit from the use of behavioral theory to underpin not only the behavior of lifestyle improvement by the patient, but also the behavior of HIT use by healthcare providers and patients. All behaviors are important to consider when designing and evaluating HIT tools to support clinical care and improve patient outcomes.

Strengths of this review include the comprehensive search strategy, systematic methods and inclusion of recent evidence. Although several studies included MetS outcomes, our search strategy focused on weight-related and behavioral outcomes and may not have identified studies in which MetS was a primary focus. A range of study designs were included as the field is emerging and randomized trials may not be feasible or appropriate to understand HIT impact (e.g., in a single practice-based setting or with widespread HIT adoption).[84] However, this limits the ability to determine causality and make comparisons across study types (e.g., RCTs, quasi-experimental designs, QI studies). Meta-analyses were not performed given the heterogeneity of included study designs and outcomes measured. As noted previously, the majority of included studies were small and/or of poor methodological quality indicating a continuing need for more rigorous research. The vast majority of trials occurred in the United States; however, our review did not include non-English publications, and databases searched may be United States-centric. As a result, our conclusions may have limited applicability across cultural contexts and health systems.

HIT is able to improve care processes for childhood obesity; yet the impact of HIT on children’s health behaviors and clinical outcomes including weight and MetS remains unclear. Pediatric obesity HIT interventions are beginning to incorporate novel approaches to provide individualized care and address health disparities. Additional research is needed to understand the effectiveness and implementation potential of these approaches. Understanding the benefits and limitations of HIT will enable healthcare professionals to design, implement, and sustain HIT interventions that fit the needs of their patients and maximize the impact on pediatric obesity across diverse populations.

Table 3:

Outcomes and Measurement

Study	Outcome types	Primary Outcomes	Measures primary outcomes	Change in primary outcomes	Secondary outcomes	Measures secondary outcomes	Change in secondary outcomes
Pilot Studies (N=7)
Altamirano-Diaz, 2017 [52]	behavioral/clinical	a. BMI z-score b. waist circumference z-score c. waist-to-height ratio z-score d. lean body mass e. body fat f. cardiorespiratory exercise capacity	a. height, weight b. clinic measurement c. clinic measurement d. dual x-ray absorptiometry e. dual x-ray absorptiometry f. VO2 uptake, CO2 production, heart rate	a. yes b. yes c. yes d. yes e. no f. no	biochemical markers	HDL, LDL, TG, fasting plasma glucose, hemoglobin A1C, insulin levels, HOMA-IR	no
Fleischman, 2016 [38]	behavioral/clinical; implementation	a. BMI z-score b. BMI percentile	a. height, weight b. height, weight	a. yes, but not stat sig b. yes, but not stat sig	a. waist circumference b. triceps skinfold c. blood pressure d. glycemic load e. physical activity f. patient satisfaction	a. clinic measurement b. calipers c. auscultation d. interview e. interview f. survey	a. no b. no c. no d. yes e. no f. yes, but not stat sig
Moschonis, 2019 [39]	behavioral/clinical	BMI z-score	height, weight	yes, but not stat sig	a. 17 dietary intake outcomes b. food intake c. waist circumference d. physical activity	a. 24-hour food recall b. food frequency questionnaire c. clinic measurement d. survey	a. 2-yes; 4-yes, but not stat sig; 11-no b. yes, but not stat sig c. yes, but not stat sig d. no
Pbert, 2016 [63]	behavioral/clinical; implementation	BMI	height, weight	yes	a. 9 physical activity behaviors b. 11 dietary behaviors c. provider acceptability	a. survey b. survey c. group debriefing session	a. 1- yes, but not stat sig; 8- no b. 5- yes, but not stat sig; 6-no c. NA (post only)
Slusser, 2016 [54]	behavioral/clinical; implementation	patient satisfaction	survey	NA (post only)	a. provider satisfaction b. BMI c. blood pressure	a. survey b. height, weight c. EHR extraction	a. NA (post only) b. yes, but not stat sig c. yes, but not stat sig
Tripicchio, 2017 [64]	behavioral/clinical; implementation	BMI z-score	height, weight	yes, but not stat sig	a. retention b. feasibility/acceptability of HIT c. technology engagement	a. attendance b. survey c. app data	a. NA (post only) b. NA (post only) c. NA (post only)
Wright, 2013 [40]	behavioral/clinical; implementation	a. BMI z-score b. BMI percentile c. BMI	a. height, weight b. height, weight c. height, weight	a. yes, but not stat sig b. yes, but not stat sig c. no	a. kilocalorie consumption b. fat consumption c. fruit & vegetable consumption d. hours/week of television time e. acceptability and feasibility of intervention	a. survey b. survey c. survey d. survey e. survey	a. no b. no c. no d. no e. NA (post only)
Full Trials (N=10)
Bohlin, 2017 [85]	behavioral/clinical; implementation	degree of obesity	BMI standard deviation score	no	participant perceived acceptability, feasibility	survey	NA (post only)
Currie, 2018 [55]	behavioral/clinical; implementation	a. BMI b. physical activity	a. height, weight b. pedometry	a. yes, but not stat sig b. yes, but not stat sig	intervention delivery fidelity	number of telephone call attempts and duration	NA (post only)
Lipana, 2013 [51]	behavioral/clinical; care process	a. BMI b. change in diagnosis c. change in treatment	a. height, weight b. NR c. NR	a. NR b. yes c. yes	a. diet improvement b. physical activity c. screen time d. weight status	a. survey b. survey c. survey d. physician report	a. yes b. yes c. yes d. yes
Mameli, 2018 [57]	behavioral/clinical; implementation	weight change	weight	No, not stat sig	a. BMI b. energy and nutrient intake c. level of commitment to intervention d. burden of intervention among children e. satisfaction with intervention f. burden of intervention with parents g. awareness of importance of lifestyle change h. lifestyle habit change	a. BMI standard deviation score b. food diary c. survey d. survey e. survey f. survey g. survey h. survey	a. no b. yes c. no d. no e. no f. no g. no h. no
Nguyen, 2013 [56]	behavioral/clinical; implementation	a. BMI z-score b. waist to height ratio	a. height, weight b. waist circumference, height	a. yes, but not stat sig b. yes, but not stat sig	a. systolic BP b. diastolic BP c. metabolic outcomes (7) d. Psychosocial well-being e. dietary intake f. physical activity & sedentary behaviors g. patient & parent program satisfaction	a. clinic measurement b. clinic measurement c. blood test d. survey e. survey f. survey g. survey	a. yes b. no c. no d. no e. no f. no g. NA (post only)
Patrick, 2013 [62]	behavioral/clinical	BMI z-score	height, weight	no	a. fruit & vegetable intake b. physical activity c. sedentary behavior d. health-related QOL e. depressive symptoms f. self-esteem g. skills to execute behavior change strategies	a. food frequency questionnaire b. 7-day physical activity recall c. survey d. survey e. survey f. survey g. survey	a. no b. no c. yes d. no e. no f. no g. yes
Rhodes, 2017 [53]	behavioral/clinical; implementation	a. dietary GL b. dietary fat	a. dietary recall b. dietary recall	a. yes, but not stat sig b. no	a. total energy intake b. parent satisfaction with intervention c. child engagement	a. dietary recall b. survey c. survey	a. yes, but not stat sig b. NA (post only) c. NA (post only)
Sherwood, 2019 [58]	behavioral/clinical; implementation	a. BMI percentile b. BMI z-score	a. height, weight b. height, weight	a. yes, but not stat sig b. yes, but not stat sig	a. minutes of physical activity b. total energy intake c. fruit & vegetable intake d. sugar sweetened beverage intake e. screen-time f. treatment fidelity g. parent satisfaction	a. accelerometry b. 24-hour diet recall c. 24-hour diet recall d. 24-hour diet recall e. survey f. survey g. survey	a. yes, but not stat sig b. yes, but not stat sig c. no d. no e. yes, but not stat sig f. NA (post only) g. NA (post only)
Taveras, 2017b [50]	behavioral/clinical	BMI z-score	height, weight	yes (one site); yes, but not stat sig (one site)	NR	NR	NR
Connect for Health (Bala 2019, Taveras 2015 & 2017) [59–61]	behavioral/clinical; implementation	a. BMI b. health-related quality of life	a. BMI z-score b. survey	a. yes, not stat sig b. yes, not stat sig	a. weight status b. health-related quality of life c. parent resource empowerment d. parent satisfaction and acceptance of intervention e. feasibility	a. CDC cut points for BMI b. survey c. survey d. survey e. % of participants that completed intervention components	a. yes b. yes c. yes, but not stat sig d. NA (post only) e. NA (post only)
Quality Improvement Studies (N=8)
Andrade, 2020 [45]	care process	screening rates	EHR data	NA (post only)	NR	NR	NR
Bode, 2013 [46]	care process	inclusion of BMI percentage & BMI growth curve	EHR chart review	yes	NR	NR	NR
Keehbauch, 2012 [47]	care process	BMI documentation in EHR	EHR chart review	yes	a. dietary counseling b. ordering lipid labs c. scheduling follow up visit	a. chart review b. chart review c. chart review	a. yes b. yes c. no
Myers, 2020 [65]	care process; implementation	BMI documentation in EHR	EHR chart review, trends over time	yes	a. obesity diagnosis b. referral to weight management services c. clinician confidence d. clinician satisfaction with the intervention e. receipt of obesity intervention f. discharge referral	a. EHR, trends over time b. survey c. survey d. survey e. electronic health record data f. electronic health record data	a. NR b. yes b. yes, but not stat sig c. NA (post only) d. NA (post only) e. NA (post only) f. NA (post only)
Wray, 2018 [86]	care process; implementation	BMI documentation in EHR	EHR extraction	yes	a. clinician satisfaction b. physicians’ practices related to weight management, healthy eating and physical activity	a. survey b. survey	a. NA (post only) b. NA (post only)
HEALTH COP (Shaikh 2014 & 2015) [48,49]	care process	a. BMI documentation in EHR b. PA and nutrition counseling	a. EHR BMI documentation score b. survey	a. yes b. yes	change in PA and nutrition behavior	interview	yes, but not stat sig
HealthSmart Kids (Gance-Cleveland 2016 & 2017) [42,43]	care process; implementation	a. provider satisfaction (2017) b. provider proficiency in counseling and assessment practices (2017) c. parent perceptions of provider adherence and care quality (2017) d. BMI documentation in EHR (2016) e. blood pressure documentation in EHR (2016)	a. survey b. survey c. survey d. chart audit e. chart audit	a. NA (post only) b. NA (post only) c. NA (post only) d. yes, but not stat sig e. yes, but not stat sig	a. provider attitudes b. overweight diagnosis c. obesity diagnosis d. laboratory assessment adherence	a. survey b. chart audit c. chart audit d. chart audit	a. NA (post only) b. yes c. yes d. no
STAR (Ayash 2013, Taveras 2014)	care process	diagnostic code for BMI percentile	HEDIS ICD codes in EHR	yes	NR	NR	NR

Note: CG = comparison/control group; NR = not reported; NA = not applicable; yes, but not stat sig = yes, but not statistically significant (significance determined pre-post for single arm studies; significance determined between intervention and control groups for comparison studies); Studies with multiple outcomes are reported using alphabet character ordering that corresponds with subsequent measurement and changes.