Table 5.

Summary of the limitations and methodological issues of the reviewed studies.

DHIa type, reference, and type of methodological issue			Description
Audiovisual consultation [24]
	Study design	Low generalizability; selection bias—telephone and telemedicine consultations not randomly assigned; children in DHI group were younger than those in the control group
	DHI related	Low telemedicine use and likely overestimated operation cost because of small cohort
	Clinical data	No patient follow-up data to monitor potential postdischarge health problems
Telephone consultation [25]
	Study design	Low generalizability (a single specialist referral center with high socioeconomic status); measuring differences in clinical outcomes but not in cost outcomes; double counting of parental time costs
	Clinical data	Some information collected in self-reported questionnaires was subject to recall bias
Audiovisual and telephone consultation [26]
	Clinical data	Parent self-reported measures leading to incidental misreporting because of memory errors (long trial period); no data collected on diagnostic remission; missing demographic data for a large percentage of the sample
	Cost data	Costs associated with accessing mental health services not included
	DHI related	Inconsistency between treatment arms as DHI was delivered one-to-one and usual care was delivered in group format
	Study design	No blinding to random allocation; DHI self-selection bias; generalizability limited (narrow age range)
SMS text messaging [27]
	Clinical and cost data	Lack of primary data (patient recruitment challenges); incomplete data records for approximately 50% of participating women upon exit interviews
	DHI related	Most of the fixed costs of DHI did not vary with changing program scale
SMS text messaging [28]
	DHI -related	Limited empirical data and evidence on large-scale mHealthb programs for pregnancy; thus, numerous assumptions about population and service coverage inputs
	Economic evaluation	Model does not incorporate complexities between preventive and curative care
SMS text messaging [29]
	Cost data	Cost data collected from a single district and did not include costs incurred by pregnant or postpartum women to seek care or to the health system to collect data
	Study design	Methodological weaknesses in study design and data collection methods (sampling and survey tool)
SMS text messaging [30]
	Study design	Number of participants divided into 2 groups was not equal or adequately balanced
	DHI related	Study not able to verify whether SMS text message reminders were received and further read by clients in the DHI group
SMS text messaging [31]
	Study design	Quasi-experimental design—not enough statistical power and adjustment for confounding factors
	Cost data	Cost adjustment for standardized estimations to 1 million population may not incorporate changes with scaling up; household and service provision costs associated with DHI not included
	DHI related	Intervention was a reminder, not provision of care, implying that the health outcome could have been influenced by other determinants of access and quality of care in the system
SMS text messaging [32]
	Study design	Short follow-up; new “test and treat” malaria case management policy not tested under trial conditions; patients aged ≥5 years not included
Mobile phone app [33]
	Cost data	Did not assess health care input cost or time spent by health workers in training and supportive supervision by medical officers and health care professionals; total cost of implementing DHI assumed to be the same in per-protocol and intention-to-treat analyses, which may be an overestimation in the per-protocol analysis
Mobile phone app [34]
	Cost and clinical data	Not able to track provision of services for all 10 antenatal care interventions
	Clinical data	No follow-up data to monitor compliance with hypertension management; no appropriate data to track subsequent second tetanus toxoid vaccination
	Economic evaluation	Assumed full compliance of women given iron folate and malaria prophylaxis
Mobile phone app [35]
	DHI related	Not clear whether the effect of several simultaneous interventions was additive, multiplicative, or otherwise
	Cost data	Intervention scale-up costs estimated in ideal conditions without bottlenecks in implementation
	Study design	Intervention was not randomly assigned, leading to possible confounding
Web portal [36]
	Clinical data	Available measures of quality of life may lack validity for children and adolescents with anxiety disorders
	Study design	Active control condition rated as being less credible at week 3; results may not be generalizable to the entire patient population (most patients were self-referred and from educated families); participants with missing data were more severely ill at baseline and in the comparator vs DHI group; short follow-up
Web portal [37]
	Cost data	Uncertainties on ongoing vs sunk costs; ongoing costs of DHI may have been overestimated because of small sample size
	Study design	No conclusions on noninferiority can be drawn (this was an intention-to-treat rather than per-protocol analysis)
Web portal [38]
	Study design	Crossover from usual treatment to iCBTc after 10-week follow-up; high educational level of parents may reduce external validity; inclusion criterion of basic reading and writing skills excluded newly arrived or marginalized immigrants; not possible to blind patients and therapists to treatment assignment
Web portal [39]
	Study design	Moderate sample size; measurements at 2 time points (before and after the intervention); short follow-up
Web portal [40]
	Study design	Most participants were self-referred, with potential confounding effects of higher motivation to work compared with typical patients with SADd
	Economic evaluation	Comparator less credible than iCBT
Web portal [41]
	Cost data	Tax-funded universal health system in Sweden may affect interpretation of the results; other health care resources and societal costs were assessed retrospectively with a parent-reported measure
	Study design	Stepped care may result in delayed treatment response and, thus, is not the preferred choice for policy makers
Web portal [42]
	Study design	Study narrowly focused on health benefits linked to prevention of incidence of depression only
	Economic evaluation	Model assumed that preventive interventions for depression led to a reduction in depression incidence based on the outcomes of meta-analyzed RCTe studies with short time frames; excluded evidence from RCT studies assessing depression symptom changes
	Cost and clinical data	Data limitations (old intervention pathways, effectiveness data with high risk of bias, and lack of cost data)
Web portal [43]
	Economic evaluation	Short time horizon because of lack of evidence on longer-term effects of DHI; no active comparator
	Clinical data	No multi-attribute utility instrument for dimensions affected by IBSf in adolescents
Web portal [44]
	Cost data	Cost data estimated retroactively; prospective monitoring of costs could yield more precise estimates
	Study design	Short follow-up data collection period (2 weeks)
Web-based symptom monitoring [45]
	Study design	Frequency of outpatient visits may differ in clinical practice compared with RCT conditions
	DHI related	DHI partly combined with usual care in the intervention arm

^aDHI: digital health intervention.

^bmHealth: mobile health.

^ciCBT: internet-delivered cognitive behavioral therapy.

^dSAD: social anxiety disorder.

^eRCT: randomized controlled trial.

^fIBS: irritable bowel syndrome.