Development and psychometric properties of the Sleep Parenting Scale for Infants (SPS-I)

Abstract

Although infants’ sleep behaviors are shaped by their interactions with parents at bedtime, few tools exist to capture parents’ sleep parenting practices. This study developed a Sleep Parenting Scale for Infants (SPS-I) and aimed to (1) explore and validate its factorial structure, (2) examine its measurement invariance across mothers and fathers, and (3) investigate its reliability and concurrent and convergent validity. SPS-I was developed via a combination of items modified from existing scales and the development of novel items. Participants included 188 mothers and 152 mother-father dyads resulting in 340 mothers and 152 fathers; about half were non-Hispanic white. Mothers and fathers completed a 14-item SPS-I for their 12-month old infant. Exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were used to explore and validate SPS-I’s underlying structure. Multi-group CFA was used to examine measurement invariance across mothers and fathers. Reliability was examined using Cronbach’s alpha. Concurrent validity was assessed using linear regressions examining associations between SPS-I factors and parent-reported infants nighttime sleep duration. Convergent validity was assessed using paired-sample t-tests to test whether the SPS-I subscale scores were similar between mothers and fathers in the same household. EFA and CFA confirmed a 3-factor, 12-item model: sleep routines, sleep autonomy and screen media in the sleep environment. SPS-I was invariant across mothers and fathers and was reliable. Concurrent and convergent validity were established. SPS-I has good psychometric properties, supporting its use for characterizing sleep routines, sleep autonomy and screen media in the sleep environment by mothers and fathers.

INTRODUCTION

Sleep problems at night such as short sleep time, fragmented sleep and frequent/prolonged awakenings are common among infants and young children.^1,2 On average, infants and children sleep less than the recommended sleep duration by at least 37 minutes.³ If not properly managed, sleep problems can negatively impact infants’ physical, cognitive and socio-emotional developments. In terms of physical health, recent studies have found that shortened infant sleep duration is associated with increased risk of developing overweight or obesity during preschool and later years.^4–6 Other sleep dimensions linked with increased risk of obesity include delayed bedtimes,^7,8 poor sleep quality and sleep efficiency.^9–11 Although the pathways linking sleep problems and childhood overweight and obesity are not fully understood, potential mediators include unhealthy dietary behaviors, physical inactivity and disrupted appetite hormone regulations.¹² Besides their negative impacts on physical health, short sleep duration and later bedtime during infancy are also linked to a higher likelihood of developing socio-emotional problems or internalizing behaviors during childhood.^13–15

Sleep parenting, which refers to the interactive behaviors between parents and infants at bedtime, plays an important role in shaping infants’ sleep behaviors.¹⁶ To cultivate healthy sleep patterns among infants, the Self-determination Theory (SDT) suggests that the optimal sleep parenting style should convey autonomy support, adequate structure and positive parental involvement in infants’ sleep environments.¹⁷ Autonomy support encompasses actions that nurture the infant’s capacity to self-regulate when parents are not around.¹⁸ In the context of sleep parenting, it refers to parents’ efforts to allow the infant to exercise autonomy in regulating their sleep, particularly going to sleep. The provision of structure refers to parents creating a structured environment that facilitates children’s competence in pursuing the desired behaviors.^17,19 For example, a structured sleep environment for infants can be achieved through establishing regular bedtime activities/routines, sleep duration and sleep locations. Parental involvement encompasses parents’ affection, interest, dedication and attention to the child to foster desired behaviors.^17,20 Through continuously engaging in the infant’s bedtime activities and routines, parents are seen to provide support to foster their infant’s confidence and self-direction in pursuing expected bedtime behaviors.

As infants’ sleep behaviors are highly dependent on their interaction with parents in the sleep environment, it is imperative to identify and better understand how specific sleep parenting practices influence infants’ sleep. However, while there are a small number of measures assessing parents’ sleep parenting (e.g., Brief Infant Sleep Questionnaire,²¹ Infant Sleep Questionnaire,²² Parental Interactive Bedtime Behaviour Scale,²³ Sleep and Settle Questionnaire²⁴ and Baby Care Questionnaire²⁵), several weaknesses limit their utility. First, some of these tools were developed for infants at risk of sleep disorders or other psychological problems.^21,22 Their suitability for use among healthy populations is unclear. Second, no scales to our knowledge have been tested for use with fathers.^21–25 Given fathers’ increasing involvement in caregiving,^26,27 the lack of tools that are valid for use with fathers limits our understanding of sleep parenting in two parent households and households without mothers (e.g., male-male and single father households). Finally, none of the existing scales capture parenting in reference to children’s access to and use of screen media in the sleep environment. Research shows that televisions and mobile media devices in the sleep environment disrupt sleep in children, adolescents and adults.²⁸ There is also evidence that parents give children access to mobile media devices at increasingly younger ages. A recent study of parents with children aged 6 months to 4 years reported that 14% of children younger than 1 year and 26% of 1-year-old infants had their own tablet or smart phone; what is more, 28% of parents gave their children access to such devices at bedtime.²⁹ A second study found that approximately 51% of infants 6 to 11 months used a touchscreen device on a daily basis and that higher daily use was linked with reduced sleep and delayed sleep onset.^30,31 Such patterns have prompted the American Academy of Pediatrics to recommend that media use should be avoided in children under 18 months, with the exception of video chat.³² In short, research highlights the need to carefully measure parents’ approaches to allowing screen media in infants’ sleep environment because it runs counter to positive parental involvement outlined in SDT and has implications for children’s sleep.

To address these limitations and advance knowledge on mothers’ and fathers’ sleep parenting practices, we developed a novel multidimensional measure of sleep parenting, the Sleep Parenting Scale for Infants (SPS-I), and rigorously tested its psychometric properties in mothers and fathers with 12-month old infants. More specifically, we examined the scale’s factorial structure, measurement invariance across mothers and fathers, reliability and concurrent and convergent validity.

METHODS

Study sample

The present study was part of a longitudinal observational study examining associations between infant sleep patterns and growth from birth to 24 months: the Rise and SHINE (Sleep Health in INfancy and Early childhood) cohort. Details of the original cohort have been published elsewhere.³³ Briefly, 433 mother-infant pairs were recruited between May 2016 and June 2018 after delivery at Massachusetts General Hospital in Boston, Massachusetts. Beginning in October 2016, fathers (n=224) were recruited along with mothers and infants from the delivery ward. Eligibility criteria for mothers included fluency in either English or Spanish, at least 18 years of age, biological and birthing mother of the infant, singleton birth, and no significant health conditions. Eligible fathers were biological fathers of the infants and living in the same household as the mothers and infants. Data were collected when infants were one, six, 12 and 24 months. Although variables of interest of the present study, including infant sleep patterns and sleep parenting, were measured at different data collection timepoints, 12 months is the age at which sleep parenting was measured using the full scale among both mothers and fathers. Therefore, our analytical sample included mothers (n=340) and fathers (n=152) who completed the SPS-I at 12 months post-birth between 2017 and 2019: 152 mother-father dyads living in the same household and 188 independent mothers. Data were analyzed in 2020. Participants’ demographics are outlined in Table 1. The Partners Health Care Institutional Review Board reviewed and approved all the study activities (protocol # 2015P002292). All participants included in the study provided their written informed consent.

Table 1.

Participants’ demographics

Characteristics	Total sample (n=492)	Mothers random-sample A for EFA (n=170)	Mothers random-sample B for CFA (n=170)	Fathers sample for CFA (n=152)
Age (years; mean, SD)a	34.60 (5.00)	33.76 (5.11)	34.04 (4.64)	36.18 (4.94)
Race/ethnicity (n, %)b
White	253 (51.8)	75 (44.6)	84 (49.7)	94 (62.3)
Black/African American	32 (6.6)	13 (7.4)	10 (5.9)	9 (6.0)
Asian	73 (15.0)	24 (14.3)	25 (14.8)	24 (15.9)
Hispanic/Latino	130 (26.6)	56 (33.3)	50 (29.6)	24 (15.9)
Education (n, %)b
Less than 12th grade	31 (6.3)	16 (9.4)	9 (5.3)	6 (4.0)
High school or GED	40 (8.2)	15 (8.8)	11 (6.5)	14 (9.3)
Some college or an associate’s degree	52 (10.6)	21 (12.4)	22 (12.9)	9 (6.0)
Bachelor’s degree	124 (25.3)	43 (25.3)	39 (22.9)	42 (28.0)
Graduate degree or higher	243 (49.6)	75 (44.1)	89 (52.4)	79 (53.7)
Annual household income in USD (n, %)b
Less than 80,000	156 (33.2)	67 (41.1)	59 (35.3)	30 (21.4)
80,000 or more	314 (66.8)	96 (58.9)	108 (64.7)	110 (78.6)
Employment status (n, %)b
Employed full-time	327 (66.7)	94 (56.0)	97 (57.1)	136 (89.5)
Employed part-time	84 (17.1)	37 (22.0)	39 (22.9)	8 (5.3)
Not employed	74 (15.1)	36 (21.4)	32 (18.8)	6 (4.0)
Student	5 (1.0)	1 (0.6)	2 (1.2)	2 (1.3)
Parent self-report nighttime sleep duration (hours; mean, SD)a	6.78 (1.03)	6.87 (1.01)	6.76 (1.05)	6.72 (1.02)
Mother-report infant nighttime sleep duration (hours; mean, SD)a	-	10.15 (1.19)	10.23 (1.25)	-

If Ns in each category do not add to the total N, the difference is due to missing data.

CFA, confirmatory factor analysis; EFA, exploratory factor analysis; GED, General Educational Development; SD, standard deviation; USD, United States Dollars.

^aNo evidence for difference between mothers random-sample A and mothers random-sample B based on t-test (p>0.05).

^bNo evidence for difference between mothers random-sample A and mothers random-sample B based on chi-square test (p>0.05).

Development of the Sleep Parenting Scale for Infants (SPS-I)

Development of the SPS-I items was guided by the SDT. SDT has been widely used in parenting research, including research on food parenting³⁴ and physical activity parenting,³⁵ because of its unique focus on the dynamic interactions of autonomy support, provision of structure and parental involvement, which is seldom considered in other parenting frameworks related to obesity-related behaviors.¹⁷

To begin the item development process, we reviewed existing scales measuring parenting in the context of child sleep including the Brief Infant Sleep Questionnaire,²¹ Infant Sleep Questionnaire,²² Parental Interactive Bedtime Behaviour Scale,²³ Sleep and Settle Questionnaire,²⁴ Baby Care Questionnaire,²⁵ Adolescent Sleep Hygiene Scale³⁶ and the Bedtime Routine Questionnaire.³⁷ Some of these existing scales have not been tested among infants but they served as the starting point for creating the SPS-I. Items related to infant sleep and SDT parenting dimensions were initially selected with wording modified appropriate for infants as needed. Additional items were developed to further reflect the overarching constructs of autonomy support, structure and involvement³⁸ and to ensure there were at least four items per construct.

The initial scale included 14 items, including six items from existing scales and eight newly developed items. Table 2 lists all the items and their sources. Briefly, among the six existing items, four measured sleep routines and two measured sleep autonomy. To this, we added one new item measuring sleep routines (i.e., “My child gets out of bed at about the same time each morning”), three new items measuring sleep autonomy (i.e., “I put my child to bed after s/he is already asleep”, “My child falls asleep on his/her own after I have left the room” and “I feed my child each time s/he wakes at night”) and four new items measuring screen media in the sleep environment (i.e., “My child uses mobile or screen-based devices 1 to 2 hours before they go to bed”, “My child has access to a mobile or screen-based device (such as an iPad or smartphone) while in bed”, “My child falls asleep while using a mobile device” and “A television is usually playing in the room when my child goes to sleep at night”). The screen media in the sleep environment subscale captures parents’ approaches to allowing both passive exposure to (e.g., television playing in the background) and active use of (e.g., direct interaction with mobile devices) screen media in the infants’ sleep environment. Items used a four-point response scale: 1=“Strongly Disagree”, 2=“Disagree”, 3=“Agree” and 4=“Strongly Agree”. SPS-I items align with the key concepts of the SDT. Sleep routines provide structure. Sleep autonomy, or providing the opportunity for the infant self soothe and fall asleep unassisted, reflects autonomy support. Allowing televisions and mobile media devices in infants’ sleep environment represents less parental involvement.

Table 2.

Sleep Parenting Scale for Infants factor structure and standardized item loadings from exploratory and confirmatory factor analyses

Item #	Item	Item source	Mean score out of 4 points (SD)			Exploratory Factor Analysisa					Confirmatory Factor Analysis
			Mothers random-sample A	Mothers random-sample B	Fathers	Factor 1	Factor 2	Factor 3	Communalities	Decision	Sleep routines	Sleep autonomy	Screen media in the sleep environment
1	My child has a bedtime routine	Adolescent Sleep Hygiene Scale (ASHS)	3.38 (0.75)	3.48 (0.71)	3.67 (0.59)	−0.02	−0.28	0.59	0.57	Kept	0.52
2	My child goes to bed at about the same time each night	Bedtime Routine Questionnaire	3.37 (0.66)	3.55 (0.62)	3.57 (0.63)	−0.02	−0.07	0.78	0.66	Kept	0.53
3	My child gets out of bed at about the same time each morning	Author developed	3.33 (0.59)	3.39 (0.64)	3.45 (0.60)	0.06	0.06	0.70	0.44	Kept	0.40
4	My child sleeps in the same room or location each night	Bedtime Routine Questionnaire	3.65 (0.52)	3.75 (0.46)	3.72 (0.54)	−0.04	−0.08	0.49	0.30	Kept	0.32
5b	I put my child to bed after s/he is already asleep	Author developed	2.76 (1.02)	2.83 (1.04)	2.77 (1.07)	0.53	0.21	0.16	0.39	Kept		0.60
6b	My child sleeps at least some part of the night in my bed	Parental Interactive Bedtime Behavior Scale	2.79 (1.19)	2.98 (1.11)	3.07 (1.12)	0.76	−0.22	−0.19	0.55	Kept	.	0.88	.
7b	I rock my child to sleep	Parental Interactive Bedtime Behavior Scale	2.81 (0.96)	2.91 (1.01)	2.78 (1.05)	0.40	0.25	0.20	0.27	Removed: low factor loading		.
8	My child falls asleep on his/her own after I have left the room	Author developed	2.75 (1.08)	2.81 (1.10)	2.80 (1.10	0.68	−0.10	−0.01	0.41	Kept		0.68
9b	I feed my child each time s/he wakes at night	Author developed	2.74 (1.12)	2.78 (1.07)	3.17 (0.91)	0.39	−0.02	−0.22	0.25	Removed: low factor loading	.	.	.
10b	I lie with my child until s/he falls asleep	Parental Interactive Bedtime Behavior Scale	2.76 (1.13)	2.92 (1.11)	3.16 (1.00)	0.85	0.00	−0.01	0.71	Kept		0.89
11	My child uses mobile or screen-based devices 1 to 2 hours before they go to bed	Author developed	1.59 (0.82)	1.51 (0.79)	1.34 (0.68)	−0.04	0.65	−0.05	0.43	Kept			0.47
12	My child has access to a mobile or screen-based device (such as an iPad or smartphone) while in bed	Author developed	1.36 (0.58)	1.31 (0.58)	1.20 (0.56)	0.00	0.69	−0.09	0.53	Kept			0.48
13	My child falls asleep while using a mobile device	Author developed	1.32 (0.57)	1.25 (0.50)	1.12 (0.40)	−0.14	0.83	−0.08	0.63	Kept			0.41
14	A television is usually playing in the room when my child goes to sleep at night	Author developed	1.53 (0.83)	1.44 (0.73)	1.18 (0.54)	0.09	0.55	−0.03	0.38	Kept			0.43

SD, standard deviation.

Significant coefficients from the pattern matrix from exploratory factor analysis are those ≥0.50 and appear in boldface.

^aThree factor promax-rotated solution was used for standardized regression coefficients for exploratory factor analysis.

^bReverse-coded items.

To assess infants’ nighttime sleep duration, mothers responded to the following question: “How much total time does your baby spend sleeping during the night (between 7 in the evening and 8 in the morning)?” As infants’ temperament could influence their sleep behaviors,³⁹ mothers were also asked to report their infants’ irritability using a 4-item subscale from the Baby Pediatric Symptom Checklist.⁴⁰ Responses were on a three-point scale: 0 = “Not at All” to 2 = “Very Much”. An irritability score was calculated by summing the responses. Both mothers and fathers reported their age, race, ethnicity, educational attainment, annual household income, employment status and nighttime sleep duration.

Statistical Analyses

We randomly split mothers (n=340) into two groups, in which exploratory factor analysis (EFA) was performed using mothers random-sample A (n=170) and confirmatory factor analysis (CFA) was performed using mothers random-sample B (n=170). Splitting the mothers sample into two random groups permits replication and cross validation without requiring data from a new independent sample.^41–43 As the sample size of fathers was insufficient (i.e., less than 300) to split them into two groups for factor analyses, the fathers sample was retained as a complete sample for CFA. Descriptive statistics (means/standard deviations, counts/percentages) were used to characterize demographics of the full sample and each subsample. Characteristics between mothers random-samples A and B were compared using t-tests for continuous variables and chi-square tests for binary and categorical variables.

The psychometric properties of the SPS-I were tested in four phases as outlined below. All the analyses were performed using R version 3.6.3 and SAS version 9.4 software (SAS institute Inc., Cary, NC, USA). Type I error rate was set at 0.05.

Phase 1: Factor exploration

We first used EFA to determine scale dimensionality and guide item reduction with data from mothers random-sample A. Three participants did not have complete data and were excluded (analytical n=167). Prior to EFA, the bivariate correlation matrix of all items was calculated, and no bivariate correlation was above 0.80. With the 14 items, we had a ratio of 11.9 observations per variable for EFA (i.e., 167 mothers/14 items).

As our item responses were ordinal and data were non-normally distributed, we used a principal axis factor estimator to extract factors.⁴⁴ Based on an examination of the scree plot and eigenvalues, along with theoretical considerations, we extracted three factors. An oblique (promax) rotation was used to account for the possible correlations among the extracted factors.⁴⁴ Items were removed at this phase if they had a factor loading <0.50 or they cross-loaded across factors (i.e., had moderate to high factors loadings for multiple factors).

Phase 2: Factor confirmation

Next, we used CFA to validate the factor structure established from the EFA using 322 participants with mothers random-sample B (n=170) and the fathers sample (n=152). We used a maximum likelihood estimator with robust (Huber-White) standard errors to ensure robustness to the variance structure.⁴⁴ Missing data were handled by using full-information maximum likelihood estimation (1 mother and 11 fathers had missing data).⁴⁵

To confirm the three-factor structure, we specified a three-factor CFA using the items retained from EFA. We allowed correlations among the three factors and fixed each factors’ variance to 1.0 to identify the model. Adequate model fit was indicated by ≥0.90 for robust comparative fit index (CFI) and <0.08 for the robust root-mean-square error of approximation (RMSEA) and the robust standardized root-mean-square residual (SRMR).^46–49

Phase 3: Factorial invariance

Multi-group CFA was used to compare the factor structure (i.e., test for measurement invariance) across mothers and fathers from the same sample used in Phase 2. Measurement invariance testing begins with testing configural invariance followed by metric, scalar and residual invariance. Details of measurement invariance testing have described by others.⁵⁰

Briefly, configural invariance testing examined whether the factor structure of the SPS-I measurement model is the same for mothers and fathers. Metric invariance testing examined whether specific items in the SPS-I contribute to their respective latent factor similarly across mothers and fathers. Scalar invariance testing examined whether mothers and fathers would have similar item scores if they have similar latent factor scores. Residual invariance testing examined whether measurement error of individual items is similar across mothers and fathers. If configural, metric, and scalar invariances are satisfied, case factors can be interpreted in the same manner and be compared across mothers and fathers.

Invariance at each step is determined by comparing a model with constraints to a model with fewer constraints (the specific constraints are dependent on the step of testing) and comparing the fit. We used the cut-offs suggested by Chen (2007) to assess model fit for measurement invariance because our study aligns well with the conditions outlined by Chen when using those cut-offs: sample size above 300; similar sample size between fathers (n=152) and mothers (n=170); and mixed loadings across the two groups.⁵¹ The criteria for violating metric invariance are a change in CFI ≥0.010, paired with either a change in RMSEA ≥0.015 or a change in SRMR ≥0.030. The criteria for violating scalar and residual invariances are a change in CFI ≥0.010, paired with either a change in RMSEA ≥0.015 or a change in SRMR ≥0.010.

Phase 4: Reliability and validity

The SPS-I’s reliability and its concurrent and convergent validity were tested using the full sample. To assess reliability, Cronbach’s alpha for each factor was computed with values ≥0.70 indicating acceptable reliability.⁵² Thereafter, we assessed SPS-I’s concurrent validity, which is a form of criterion validity and is a rapid way to test the validity of a scale by measuring the extent to which SPS-I subscale scores correlate with infants’ nighttime sleep duration assessed at the same time.⁵³ To accomplish this, linear regression models were used to examine the associations between infants’ nighttime sleep duration (dependent variable) and sleep routines, sleep autonomy and screen media in the sleep environment measured using the SPS-I (independent variables). Each SPS-I subscale was analyzed in a separate model. Subscale scores were calculated by taking the mean of the subscale items. Higher subscale scores indicate greater levels of sleep routines, sleep autonomy and screen media in the sleep environment. As infants’ sleep was reported by the mothers, mothers and fathers were analyzed separately to avoid the dependence between reporters from the same household. For each model, mothers’/fathers’ race/ethnicity, education, and their self-reported nighttime sleep duration were added as covariates. Each model also controlled for infants’ irritability score, accounting for the potential links between infants’ temperament and their sleep.³⁹ Finally, to assess SPS-I’s convergent validity, which is a form of construct validity measuring how closely fathers’ and mothers’ reported measures are related,⁵⁴ paired-sample t-tests were utilized to test whether the SPS-I subscale scores were similar between mothers and fathers in the same household.

RESULTS

Participants were mostly white (51.8%, n=253) or Hispanic/Latino (26.6%, n=130) with a mean age of 34.60 (SD = 5.00). The majority of participants had completed a bachelor’s degree or beyond (74.9%, n=367). Close to 70% had an annual income ≥80,000USD (n=314) and were employed full-time (n=327). As expected, there were no demographic differences between mothers in random-samples A and B.

Phase 1: Factor exploration

A scree plot and eigenvalues suggested a three-factor structure with all items loading on the expected factors. The factor loading of the initial extraction is presented in Table 2 with the three-factor structure explaining 51.0% of the variance. Items 7 and 9 were removed due to their low factor loadings (<0.50). Although item 4 had a factor loading of 0.49 for sleep routines, we decided to keep this item as sleep location is linked to young children’s sleep quality.⁵⁵

Phase 2: Factor confirmation

The three-factor CFA using 12 items demonstrated adequate model fit (robust χ²[51]=100.7; robust CFI=0.959; robust RMSEA=0.062, 90% CI=0.044–0.080; robust SRMR=0.053). Factor loadings ranged from 0.32 to 0.89 (Table 2).

Phase 3: Factorial invariance

Results of the measurement invariance testing are summarized in Table 3. Our configural model (M1) exhibited good model fit, indicating the factor structure is the same across mothers and fathers. When data were fitted separately, adequate model fit was also observed for the mothers group (robust χ²[51]=77.9; robust CFI=0.962; robust RMSEA=0.061, 90% CI=0.031–0.086; robust SRMR=0.060) and fathers group (χ²[51]=85.7; robust CFI=0.946; robust RMSEA=0.071, 90% CI=0.043–0.096; robust SRMR=0.066).

Table 3.

Tests of factorial validity and measurement invariance for Sleep Parenting Scale for Infants across mothers and fathers

Model	Robust χ2 (df)	Robust CFI	Robust RMSEA (90% CI)	Robust SRMR	Model compared	Δ Robust χ2 (df)	Δ Robust CFI	Δ Robust RMSEA	Δ Robust SRMR	Decision
M1: Sleep parenting model (baseline)	163.4 (102)	0.954	0.066 (0.046, 0.084)	0.063	-	-	-	-	-	Accept
M2: Metric invariance	174.6 (111)	0.952	0.064 (0.045, 0.082)	0.069	M1	11.2 (9)	0.002	0.002	0.006	Accept
M3: Scalar invariance	200.7 (120)	0.940	0.069 (0.052, 0.086)	0.072	M2	26.1 (9)	0.012	0.005	0.003	Accept
M4: Residual invariance	221.1 (132)	0.922	0.075 (0.057, 0.092)	0.081	M3	20.4 (12)	0.018	0.006	0.009	Accept

N= 322; fathers group, n= 152; mothers sub-sample B group, n= 170.

CFI, Comparative Fit Index; CI, confidence interval; df, degree of freedom; M, model; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual.

Both the metric invariance model (M2) and the scalar invariance model (M3) exhibited adequate model fit. The changes of goodness-of-fit indices support both metric and scalar invariance (M2 vs M1; M3 vs M2), suggesting measurement invariance across mothers and fathers. Residual invariance was also established (M4 vs M3).

Phase 4: Reliability and validity

Table 4 summarizes the mean and standard deviation for each factor for mothers and fathers, along with its Cronbach’s alpha. There was adequate reliability across all subscales among mother and fathers (α=0.79 to 0.85). The SPS-I demonstrated concurrent validity. We found that higher levels of sleep routines and autonomy were positively associated with infants’ nighttime sleep duration, and higher levels of screen media in the sleep environment were negatively associated with infants’ nighttime sleep duration; these patterns were present for mothers (routines b=0.42, 95% CI=0.16 – 0.68, p=0.002; autonomy b=0.32, 95% CI=0.16 – 0.47, p<0.001; screen media in the sleep environment b=−0.45, 95% CI=−0.72 – −0.19, p=0.001) and fathers (routines b=0.41, 95% CI=0.03 – 0.79, p=0.03; autonomy b=0.36, 95% CI=0.12 – 0.60, p=0.004; media in the sleep environment b=−0.95, 95% CI=−1.40 – −0.51, p<0.001). Finally, based on the paired-sample t-test analyses, the scale demonstrated convergent validity; that is, parent reports of sleep routines (t[150]=−0.59, p=0.557), autonomy (t[143]=0.18, p=0.856), and screen media in the sleep environment (t[146]=1.31, p=0.191) did not differ for mothers and fathers. Final version of the SPS-I and its scoring system are provided as supplementary materials.

Table 4.

Subscales mean and reliability for the Sleep Parenting Scale for Infants

	Mothersa		Fathers
Subscale	Mean out of 4 points (SD)	Cronbach alpha	Mean out of 4 points (SD)	Cronbach alpha
Sleep routines	3.49 (0.49)	0.80	3.60 (0.49)	0.84
Sleep autonomy	2.83 (0.87)	0.80	2.96 (0.84)	0.79
Screen media in the sleep environment	1.42 (0.54)	0.81	1.21 (0.44)	0.85

SD, standard deviation.

^aCombined mothers random-samples A and B.

DISCUSSION

In this study, we tested the factor structure of SPS-I for 12-month old infants, examined its measurement invariance across mothers and fathers, and assessed its reliability and validity. Factor analyses revealed acceptable fit for three factors including sleep routines, sleep autonomy and screen media in the sleep environment. Measurement invariance was established across mothers and fathers indicating that the interpretation of responses on the scale is equivalent for mothers and fathers. All three factors exhibited adequate reliability, and both concurrent and convergent validity were established.

To our knowledge, the SPS-I is the first to capture parents’ approaches to allowing screen media in infants’ sleep environment. This is a timely contribution given that parents are increasingly allowing their infants to own and use mobile screen devices^29,31 and theses devices can be easily carried into infants’ sleeping space. While scores for this subscale appear to be very low in this sample, it should be kept in mind that this subscale needs to be reverse coded before it can be compared to the other two subscales; this manipulation results in a mean closer to 2.6 for mothers and 2.8 for fathers, with higher scores indicating parents allowing less screen media in their infants’ sleep environment (i.e., positive sleep parenting methods), which is not dissimilar to the other two subscales. It is worth noting that, across all sleep parenting subscales, associations between sleep parenting and infants’ nighttime sleep duration were the strongest for screen media in the sleep environment. At a minimum, this finding illustrates that there is value in measuring screen media in the sleep environment even among infants. Our sample included few low income parents; the use of screen media in the sleep environment and links with infant sleep may be even stronger in more socioeconomically diverse samples.²⁹

Our study found that the latent structure of the SPS-I is invariant across mothers and fathers. In other words, the SPS-I factors can be interpreted in the same way and compared across mothers and fathers.⁵⁰ The SPS-I offers the opportunity to capture sleep parenting practices from parents regardless of the parent’s gender. This is a valuable contribution as fathers have been more involved in caregiving,^26,27 and the SPS-I will be useful to measure sleep parenting in households without mothers such as male-male and single father households.

The SPS-I also exhibited both concurrent and convergent validity. As expected, higher reported use of sleep routines and methods promoting sleep autonomy were linked with higher sleep duration in infants, whereas, greater degree of allowing screen media in infants’ sleep environment was linked with lower sleep duration in infants. These findings are consistent with prior investigations among young children.^23,56–58 SPS-I’s concurrent validity suggests that SPS-I may be a useful tool for practitioners to assess infants who may be at risk for short nighttime sleep durations. In addition, SPS-I’s convergent validity suggests that SPS-I could be used by either the father or mother in the same household to assess sleep parenting for their infants.

The development of the SPS-I provides opportunities for future research, particularly with regards to understanding associations between various sleep parenting practices and infants’ sleep. For example, future studies can examine how subscales of the SPS-I are linked to other sleep dimensions (e.g., sleep timing, sleep quality and sleep efficiency) and infant weight status. Future longitudinal studies should also consider the use of the SPS-I to better understand the bidirectional relationships among sleep parenting, infant sleep, weight-related behaviors and weight status. Given the growing evidence linking sleep to weight gain,¹² the knowledge generated from these investigations will be important for informing the development of relevant childhood health promotion and obesity prevention strategies. Practitioners who work with infants with sleep problems may also use the SPS-I to better understand the contextual factors around infant sleep problems of their clients.

There are some limitations that need to be acknowledged. First, parents in our study were predominately white and with high socioeconomic status. Considering parenting practices are different across different ethnocultural and socioeconomic groups,^59,60 and childhood obesity disproportionally affects certain populations,⁶¹ further testing of the SPS-I is needed among more diverse samples. Second, infants’ nighttime sleep duration was reported by mothers, which may be subject to recall and social desirability biases. Testing of the SPS-I’s concurrent validity with objective sleep measurements is needed. Third, we only tested SPS-I among infants who were 12-months old. Further studies are needed to examine its performance among parents with younger infants. Fourth, the SPS-I might not have captured all the sleep parenting practices that are relevant to infants’ sleep. Further refinement and modifications are encouraged.

CONCLUSIONS

The SPS-I has good psychometric properties, supporting its use for characterizing sleep routines, sleep autonomy and screen media in the sleep environment across mothers and fathers. Further understanding of these sleep parenting practices will help better inform future interventions.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the senior author, KKD, upon reasonable request.