Use of Text Messaging for Postpartum Depression Screening and Information Provision

Postpartum depression is the most common postpartum mental health issue, with a prevalence of 13% to 15% in high-income countries (1). Postpartum depression leads to maternal disability, impaired mother-child attachment (2), and infant social and cognitive problems (3). Screening is recognized as an effective means of identifying women at risk (4, 5); only 18% to 25% of postpartum depression cases are detected without screening (6).

Although recommended (7, 8), screening is not widely implemented in North America (9). Barriers include lack of physician familiarity with screening tools or diagnostic criteria (10, 11), inability to incorporate screening into routine care (10), and lack of physician/nurse time (11). Many mothers do not seek help due to lack of time, transportation, and child care (12, 13); difficulty identifying psychiatric symptoms (14); and fear that child protective services will be called (15). Many new moms find it challenging to decipher between typical postpartum adjustment and problematic psychiatric symptoms (16). Early, formal, and efficient screening and clear, direct education about postpartum depression are essential. Well-validated screening tools and information packages that are brief, confidential, and easy to administer and disseminate are needed.

Mobile telephones have been used for mental health assessment and education and are free of identified barriers to screening owing to familiarity, ubiquity, affordability, and confidentiality. Text messaging has been used in several different populations, most frequently substance use disorders, schizophrenia, and affective disorders, and predominately in the service of information provision, support, and self-monitoring (17). In the perinatal population, text messaging has been used for information provision and an adjunct to treatment. Broom and colleagues (18) used text messaging as an adjunct to cognitive-behavioral therapy for treating postpartum depression among low-income women from racial and ethnic minority groups; participants perceived this to positively influence their motivation to change and reduce their depressive symptoms. Bhat and colleagues (19) found that text messaging to send appointment reminders and mental health information to women with perinatal depression was acceptable and useful. Jareethum and colleagues (20) discovered that informational text messages reduced anxiety levels and increased confidence among healthy antenatal women. Other research demonstrates the ability of texting technology to improve antenatal care attendance (21, 22) and stimulate healthy behavior (18, 23–25). To date, there are no reports on the use of text messaging for mental health screening in the perinatal population.

The primary objective for this study was to evaluate the feasibility of using text messages to enhance postpartum depression screening and education of women in the immediate postpartum, as a first step toward improved access to care and reducing the significant morbidity associated with such illnesses.

Methods

Participants

A total of 1,608 women ≥18 years of age in their second or third trimester attending the obstetrics and gynecology clinic housed at Mount Sinai Hospital in Toronto were approached by a study recruiter between July 2015 and January 2017. Mobile telephones were available for those who did not own one; everyone approached for participation owned a mobile telephone that was suitable for texting. The study recruiter obtained informed consent and conducted the baseline session. Ethical approval for the study was granted by the Mount Sinai Hospital Research Ethics Board.

Study Process

Baseline.

The study recruiter administered a sociodemographic questionnaire that collected information on demographics, current pregnancy details, history of postpartum mental illness, e-mail address, and information on mobile telephone ownership, telephone number, and data provider. The Edinburgh Postnatal Depression Scale (26), the gold standard in screening for depressive symptoms in perinatal women, was administered. Women who screened positive and were not in active psychiatric care were offered an appointment with a perinatal psychiatrist and included in the study; women in active psychiatric care were excluded. Participants were sent a $10 CAD gift card and a letter thanking them for completing the baseline assessment.

Screening.

This study used a three-phase screening process. To identify all potential cases of postpartum depression, our first phase consisted of sending a screen by text message to all participants. The screen consisted of both questions on the Patient Health Questionnaire (PHQ)–2, which are also the first two questions on the PHQ-9; the PHQ-2 has been validated to screen for symptoms of postpartum depression (27). This same screen was repeated biweekly until 12 weeks postpartum. In the second phase, we identified all those who screened positive. A positive screen alert was communicated to the project coordinator. In phase 3, to determine the ability of the texted screen to detect symptoms of postpartum depression, the project coordinator contacted all women who screened positive to complete the Edinburgh Postnatal Depression Scale via telephone. These women were matched as closely as possible on age, parity, and postpartum week to a subset of women who were also assessed via the Edinburgh Postnatal Depression Scale after screening negative in phase 2. Sensitivity and specificity of the texted screen were determined by the concordance between the rate of positive and negative screens rendered via text message and the rate of positive and negative screens rendered by the Edinburgh Postnatal Depression Scale.

All women who screened positive on the texted screen or Edinburgh Postnatal Depression Scale for potential postpartum depression were offered an appointment with a perinatal psychiatrist. A chart review was conducted for all women who accepted these appointments; presenting and discharge diagnoses were noted.

Informational texts.

All participants were sent three text messages per week designed to bolster awareness and self-care behaviors (e.g., “Recent mom? Feeling depressed? Want more information? Here is a Web site that has information about postpartum depression: http://www1.toronto.ca/postpartum depression”). These texts were developed based on our clinical experience to provide accessible information for women looking to immediately access information or help and to provide information about commonly addressed postpartum mental health issues and concerns that women can address on their own to improve their mental health.

Follow-up survey.

Feasibility of the texting service was assessed via a secure online satisfaction survey sent through e-mail, anonymously completed. This investigator-developed survey had two parts. The first assessed whether participants liked being screened, if they would recommend a texted screening for all postpartum women, and if they would have preferred a different approach and why. The second asked whether they found the informational texts helpful, if they used any of the information provided in the texts, and if they would recommend that all women be sent such texts in the postpartum stage.

Texting Application

A Samsung S3 android mobile telephone continuously running a custom application for the study duration was used to automatically send and receive text messages. At the same time, each day the application sent screening and information texts based on scheduling information stored in a register file. The application continuously listened for incoming texts and logged the responses to a file on the device. The project coordinator checked the log on the device each day to determine with whom to follow up by phone.

Measures

The PHQ-2 has been validated for use with new moms (28) and is an effective screening tool for postpartum depression (29). The questions ask about the two fundamental symptoms of depression, diminished mood and pleasure. The first asks about feeling “down, depressed, or hopeless” in the past two weeks. The second question asks about “little interest or little pleasure in doing things” in the past two weeks. Responses are scored with a Likert-type scale of 0, not at all, to 3, nearly every day. A score of ≥2 on one or both questions represents a positive screen for depressive symptomatology. These two questions mimic the criteria for a diagnosis of major depressive disorder as found in the DSM-5 (30).

The Edinburgh Postnatal Depression Scale is a well-validated 10-item screening tool developed for screening symptoms of postpartum depression (26). It has excellent sensitivity and specificity for postpartum depression (31–33) and is superior to many measures of depression (34). Response options are recorded on a Likert scale of 0 to 3. Validation studies have used thresholds ranging from 9 to 13 points. Some research in pregnant populations recommends a more conservative cutoff (35), given the possible intersection between symptoms of pregnancy and depression. For this study, a cutoff of ≥15 was used at the baseline screen (i.e., during pregnancy); a cutoff of ≥12 was used for the postpartum screens (all other screens). Although the Edinburgh Postnatal Depression Scale is not a definitive diagnostic tool, for the purposes of testing the response characteristics of the texted screen, we used the Edinburgh Postnatal Depression Scale as the gold standard. The specificity and sensitivity reported in this article reflect the Edinburgh Postnatal Depression Scale ≥12 criterion, not a final diagnosis of postpartum depression.

Data Analysis

Summary statistics were used to describe the study population responses and rates. Bivariate associations between descriptive variables and a positive texted screen (score of ≥2 for either question) were assessed using independent t tests and the chi-square statistic with statistical significance set at p<0.05 (two-tailed). Receiver operating characteristic (ROC) curve analysis was performed for the PHQ-2 by using the Edinburgh Postnatal Depression Scale as the gold standard. We report the standard classification statistics, including sensitivity, specificity, Cohen’s kappa, positive predictive values, and negative predictive values. ROC parameters were computed with weights to correct for undersampling of the screen-negative participants.

Sensitivity and specificity range from 0 to 1, with higher values indicating better accuracy. Although there are no commonly recommended thresholds for sensitivity and specificity, a minimum sensitivity and specificity of .70 was considered acceptable to reflect the priority of screening accuracy. For sensitivity, specificity, positive predictive value, and negative predictive value, 95% confidence intervals based on the Wilson interval (36) were calculated.

Cohen’s kappa measured diagnostic agreement between the texted screen and the Edinburgh Postnatal Depression Scale beyond that expected by chance. Kappa values were interpreted following guidelines proposed by Landis and Koch (37): 0.01–0.20, slight; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, substantial; and 0.81–1.00, almost perfect. All analyses were conducted with SPSS version 23.0.

Results

Of the 1,608 women approached about participation, 1,492 (93%) were eligible. Of these, 1,000 consented to participate in the study (67% consent rate). Of these, 63 withdrew or became ineligible over the course of the study (see figure in the supplement), for a final sample size of 937. Table 1 shows the sociodemographic characteristics of the total sample and the subsamples of participants with a positive or negative screen. The pattern seen in Table 1 is consistent with research on risk factors for postpartum depression, which are known to include psychiatric history and low income (38).

Results were analyzed with the texting screen criterion of ≥2 on either question. The kappa value indicated that there was fair agreement (κ=0.37) between screening and the Edinburgh Postnatal Depression Scale. Sensitivity was 0.49 (95% CI=0.38–0.61), and specificity was 0.93 (95% CI=0.91–0.94). The positive predictive value was 0.38 (95% CI=0.31–0.46); the negative predictive value was 0.95 (95% CI=0.94–0.96). Exploration of other PHQ-2 cutoff values showed that making a slight change in the cutoff criterion increased sensitivity greatly. The revised criterion used the total score instead of the original either-or criterion, thereby classifying two responses in the mild range (1+1=2) as a positive screen for depression. This revised criterion provided the best balance between sensitivity and specificity, 0.90 (95% CI=0.81–0.96) and 0.82 (95% CI=0.79–0.85), respectively, and the highest kappa (κ=0.45, moderate). The positive predictive value was 0.32 (95% CI=0.29–0.36), whereas the negative was 0.99 (95% CI=0.98–0.99).

Referrals to the study perinatal psychiatrist were offered to all 126 participants who screened positive to the text messaging screen, regardless of their Edinburgh Postnatal Depression Scale score. Of these, 31 (25%) declined the referral. Ten (8%) were in active psychiatric care elsewhere, and 21 (22%) could not be contacted. Of the 31 who declined the referral, 81% (N=25) were later classified as false positive and likely did not feel a need to be seen.

Of the 64 (51%) women referred, 33 (52%) did not show up for the appointment; 18 of these (55%) were classified as false positive. Thirty-one (48%) participants attended at least one appointment with the study psychiatrist. The chart review revealed that 10 (32%) of these received a diagnosis of postpartum depression, 11 (36%) received a diagnosis of a postpartum psychiatric disorder other than depression, whereas three (10%) exhibited significant psychiatric symptoms. Of the seven women who did not receive a diagnosis or experience symptoms, three (43%) were classified as false positive.

Of the 937 participants who completed the study, 930 (99%) responded to a minimum of one of the six texted screens, whereas 632 (67%) responded to all six. Responsivity was at its greatest at the first texted screen in postpartum week 2 (N=929, 99%), declined by approximately 2% with each texted screen, and fell to 92% (N=858) by the final screen. The majority of positive screens were found at the first screening at 2 weeks postpartum (N=62, 49%), with the smallest percentage found at postpartum week 8 (N=15, 12%).

Of the 937 participants who completed the study, 589 (63%) responded to the online satisfaction survey. The majority (N=459, 78%) indicated that they would recommend that all women in the postpartum period be screened for postpartum depression via text messaging. Eighty-seven percent (N=511) preferred screening via text message in contrast to other approaches (e.g., in person, telephone). A majority of participants either liked (N=253, 43%) or felt neutral (N=302, 51%) about being asked the screening questions.

Most respondents to the survey stated that they read the information texts (N=554, 94%). Half of these found the information in the texts helpful (N=305, 55%), whereas 32% (N=177) of these used the information in the texts. The majority would recommend that all women in the postpartum stage be sent information texts about postpartum depression (N=504, 91%).

Discussion

This study of 937 postpartum women tested the feasibility of using text messaging to screen for postpartum depression. Using the Edinburgh Postnatal Depression Scale as the reference standard, we found good sensitivity for the screen when using the more liberal cutoff score (>1 total score). Although the level of specificity (82%) was not ideal, high sensitivity is preferable over high specificity in a multistage screening procedure whose aim is not diagnosis but rather identification of women who need further follow-up. Although we did not employ a systematic protocol for assessing diagnostic accuracy, our chart review revealed that 68% of those who attended at least one appointment with the study perinatal psychiatrist received a diagnosis of a postpartum psychiatric disorder, whereas another 10% suffered from psychiatric symptoms. These results are suggestive of a diagnostically accurate screening procedure.

Our self-report questionnaire revealed a high degree of satisfaction with the text messaging screen, with the majority endorsing depression screening via text message for all women in the postpartum period. Responsivity to the texted screen was also high, with rates consistently over 90% for each screen and almost 100% in postpartum week 2. These numbers stand in stark contrast to traditional efforts at universal postpartum depression screening in clinical settings (28, 39–41).

We believe the reasons for the high rate of responsivity to the screen were its brevity and the fact that in-person attendance was not a requirement for completion. The postpartum stage is a very difficult time for a mother, frequently characterized by sleep deprivation and high stress levels. Completing lengthy assessment tools in-person at this time can be challenging, particularly among mothers who are experiencing symptoms of depression. Screening women via text circumvents the need for a new mother to present in person for an appointment, with all that entails (e.g., transportation and child care), and allows for instant access to a segment of the postpartum population unlikely to attend postpartum appointments. Given the relationship between maternal depressive symptoms and noncompliance with well-child visits (42), reaching this population is critical. A texted screen also allows for repeated screening over time, facilitating detection of depressive symptoms throughout the postpartum period (10).

Twenty-two percent of those who screened positive could not be contacted, whereas 52% did not attend their appointment with the study perinatal psychiatrist. This is in keeping with nonattendance rates found by others for psychiatrically distressed women in the perinatal period (e.g., 62%−83%) (43–45) and possibly reflects women’s preference for psychiatric treatment provided in the home or obstetric clinic, rather than in specialty psychiatric units (46, 47).

A common barrier to screening for postpartum depression is the lack of knowledge among women about how to distinguish between normal symptoms of postpartum adjustment and problematic psychiatric symptoms (15). Calls have been made to institute educational programs alongside screening protocols for postpartum depression to provide women with this information (10). Our attempt to pair screening with relevant and useful information appears feasible and useful, with over half of the sample finding the information in the texts helpful, and a third of these using the information.

Limitations include the time between a screened positive result on the texted screen and completion of the Edinburgh Postnatal Depression Scale (average 3 days), due to difficulties in immediately contacting some participants to conduct the telephone screen. This could have potentially led to actual symptom change in some cases. The reported validity and reliability statistics could be conservative estimates. Secondly, our sample was largely Caucasian, married, and university educated. Our results may not be generalizable to all women in the postpartum period, particularly disadvantaged women, who may experience higher rates of postpartum depression (48). Third, this study did not include a systematic diagnostic assessment of all those who screened positive. We recognize that screening by itself is insufficient to improve clinical outcomes, and that it must be coupled with appropriate and accessible evaluation, monitoring, and follow-up (47). However, the focus of this study was on the feasibility of a multistaged screening process to improve access to mental health treatment for postpartum depression.

Conclusions

This study found that using text messages to screen for postpartum depression and provide education on postpartum mental health in the immediate postpartum period is practical and feasible. To our knowledge, no studies have applied the use of text messaging in screening for mental illness and specifically for postpartum depression. The strength of this screening procedure is that it can be easily applied to a wide range of perinatal patients, who are often difficult to reach during this crucial period (42). This cost-effective, innovative technology has excellent potential to be used in mental health care for universal screening purposes and as a first step in the pathway to treatment.