-
Spanish Translation of StartSmartTM using the Beaton Process to Ensure Tech Equity
Objective. Perinatal outcomes in the United States are poor among Hispanics, the country's fastest-growing population. Hispanics experience higher rates of obesity, diabetes, gestational diabetes, preterm birth, and pregnancy-related hypertension, as well as 20% of infant deaths. StartSmarttm, an mHealth app, screens for risk and protective factors. This project translated and culturally adapted StartSmarttm for Spanish-speaking pregnant women. Methods. Using the Beaton process for translation and cultural adaptation, we completed a six-stage process. Stage I: Translation into Spanish, Stage II: Synthesis of the translations, Stage III: Back-translation, Stage IV: Expert committee review, Stage V: Pretesting with Spanish-speaking pregnant women, Stage VI: Appraisal of process. Results. StartSmarttm was acceptable and feasible for use during prenatal visits. Participants reported ease of use and understanding the material. No issues regarding literacy were identified, and participants provided helpful feedback for refining StartSmarttm for Spanish-speaking women. Conclusions. The Beaton process ensures culturally appropriate translation.
Decision support technology, individualized patient education, prenatal screening for risk and protective factors, tech equity, Technology Acceptance Model, Beaton Process for Translation and Cultural Adaptation
Many conditions and lifestyle factors during pregnancy affect a woman's health as well as the health of her child—some effects lasting long after birth.1 Despite [End Page 85] advances in perinatal health care, infant morbidity and mortality have seen little improvement in the United States (U.S.),2 and rates of severe maternal morbidity have continued to rise.3 Hispanics are the fastest-growing population of childbearing women in the U.S. In 2018, nearly a quarter of live births in the U.S. were to Hispanic mothers,4 and the U.S. Census Bureau projects a 74% increase in births among Hispanic women.1 Preterm births in Hispanic women account for over one in five (22%) of the total preterm births in the U.S.5 Hispanic women have higher rates of obesity, diabetes, gestational diabetes, and pregnancy-related hypertension compared with White women, resulting in poor birth outcomes.6 Hispanic pregnant women also receive fewer mental health services than any other racial/ethnic group.7 Neural tube defects, birth defects that can be prevented by prenatal folic acid, are also significantly higher in Hispanics.6 Given these facts, it is essential to identify risk and protective factors during pregnancy in Hispanic women in order to improve perinatal outcomes and prevent health problems for the mother and child. Prevention and early intervention depends on effective screening, education, and referral.8,9 The American College of Obstetricians and Gynecologists (ACOG) and the American College of Nurse Midwives (ACNM) have published guidelines for screening and counseling women on these risk and protective factors.10,11
Evidence points to a link between poor perinatal outcomes and pregnancy health risks such as mental health conditions, substance use, and chronic diseases, including obesity and diabetes, during pregnancy.12 Protective factors that influence pregnancy outcomes include immunizations, nutrition, folic acid, physical activity, and sleep, yet pregnant women are not consistently counseled on these.13 Discussions of prenatal guidelines are often constrained by clinical work-flow and time limits; language can be an additional barrier for Hispanic women.14
In an effort to address these barriers, self-administered, computerized instruments may be used in order to elicit more accurate responses than face-to-face interviews in both the general population and specifically among pregnant women.15,16 In one study screening pregnant women on sensitive information and risky behaviors including alcohol, drug, and tobacco use, as well as intimate partner violence, respondents reported a higher prevalence of these behaviors when using technology than was reported in studies using other data collection methods.17
Barriers to the application of perinatal guidelines also exist for providers. Lack of time, awareness, resources, availability of specialist referral, and discomfort with discussing sensitive topics can be barriers to guideline adherence.18–24 Similarly, providers have expressed value in technologies that support clinical implementation of perinatal guidelines, and there is significant evidence for decision-support technology to improve patient outcomes in ambulatory care settings.25–28,29 The aim of our work has been to extend the reach of technology to a Spanish-speaking population.
Our team has developed StartSmartTM,30,31 a tablet-based, English-version technology to facilitate comprehensive screening with self-administered questionnaires and to automate decision support to improve brief motivational interviewing counseling and referral for treatment when indicated. The technology was developed using the Davis' Technology Acceptance Model (Figure 1), which is an iterative development process [End Page 86]
Iterative development using the Davis' Technology Acceptance Model.
with end-users (patients and providers) involved at each step.32 During the initial development, this process was limited to English-speaking participants.
The system uses validated screening instruments to gather information from pregnant women and treatment algorithms based upon the guidelines to create individualized counseling and education information (see Table 1). The potential for algorithm bias is minimized through the use of instruments that have been validated for Spanish-speaking populations including those for anxiety, depression, sleep, physical activity.33–35 The rigorous Beaton process was used for translation and cultural adaptation of measures as described below. The Spanish versions were compared with the translated versions when they were available. StartSmart™ consists of two web-based applications: a touch screen interview and a webpage for entry of measurements and generation of patient and provider summaries. Both applications are accessed through standard web browsers on internet-connected computers. The patient completes the screening prior to the encounter. The measurement-entry application is accessed by staff where the height, weight, blood pressure, laboratory findings, and immunizations are entered into the system. A risk summary and individualized recommendations are given to the patient to increase the understanding of risks and provide the patient-specific recommendations. The provider handout includes prompts for ordering laboratory assessments, brief interventions with motivational interviewing, and referrals as indicated.
Due to the escalating birth rates and the prevalence of poor birth outcomes in Hispanics, it is imperative that providers screen all Hispanic women including Spanish-speaking-only women for related risk and protective factors and provide appropriate [End Page 87]
SCREENING INSTRUMENTS AND TREATMENT ALGORITHM
[End Page 90]
counseling to improve birth outcomes. StartSmartTM has the potential to fill this need, but the technology was not available in Spanish.
The translation and cultural adaptation of a self-administered health questionnaire such as StartSmartTM requires a refined method to reach equivalence between the original source and the target version of the questionnaire.36 In order for measures to be used across cultures, the items must not only be translated well linguistically, but also must be adapted culturally to maintain the content validity of the instrument at a conceptual level across cultures.37–39
Attention to this level of detail allows increased confidence that the impact of the condition and treatment is described in a similar manner.36 "Cross cultural adaptation" is a process that looks at both language (translation) and cultural adaptation issues in the process of preparing a questionnaire and education materials for use with another population.36 Thus, the purpose of this study was the translation and cultural adaptation of StartSmartTM, using the Beaton process to enhance usability and acceptability for Hispanic pregnant women and their providers.
Methods
Beaton process for translation and cultural adaptation
The Beaton guidelines for translation and cultural adaptation of self-report measures (Figure 2) were developed based upon a review of medical, sociological, and psychological literature and was used to guide this study.36 The Colorado Multiple Institution Review Board approved the study and re-reviewed the adapted interview process using Zoom (video conferencing software; www.zoom.us) when the COVID-19 pandemic halted all clinical research at our institution.
Stage I
Translation included forward translation of the screening instrument and patient education materials from bilingual translators (T1 & T2) who were native Spanish speakers. Both women were born in Mexico and now live in the U.S. One translator was aware of the concepts being examined and one was not aware of the content. Each translator provided a written report that highlighted challenges and their rationale for their choices.
Stage II
Synthesis of the two translations occurred with the two translators and the primary investigator (PI) working from the original questionnaires and Spanish versions when available (some of the screening instruments used in the app are available in Spanish), to create a synthesis by consensus. The PI reviewed the two separate translations and created a document with all discrepancies for discussion. A Zoom meeting was conducted between the two translators and the PI to discuss the best translation given the intent of the question or educational material and the possible audience.
Stage III
Back-translation occurred working from the synthesized translation created in Stage II. Two bilingual, English first-language translators back-translated the documents into English and provided a written report (BT1 & BT2) of their process. These translators were blind to the original documents and project.
Stage IV
Expert committee review was conducted by the research team, the translators, and language specialists with expertise in nuances of Spanish language to consolidate all versions and reach consensus on the version to pretest. The PI reviewed [End Page 91]
Beaton process for translation and cultural adaptation.
the back-translation for consistency with the original instruments when available in Spanish and educational materials and discussed with the team to reach consensus on the best translation.
Stage V
Pretesting of the translated and culturally-adapted application was conducted guided by the Technology Acceptance Model32 with attention to usability and acceptability from patient and provider perspectives. The Spanish version of StartSmartTMwas pilot-tested in a federally qualified health center (FQHC) in the Southwestern U.S. Due to the restrictions imposed by the COVID-19 pandemic, a bilingual pre-doctoral fellow mentored by the PI who was a provider at the clinic was allowed to recruit, obtain informed consent, facilitate the patients' completion of the screening, and connect the patients with a bilingual interviewer via Zoom. Spoken consent was obtained to video-record the Zoom interviews. Participants received a $25 gift card for participation.
The women completed prenatal screening on a tablet-based touch screen for risk and protective factors, including: weight status and related conditions (underweight, overweight/obese, gestational weight gain, gestational diabetes mellitus), substance use (alcohol, drugs, tobacco), emotional conditions (anxiety, depression, domestic violence), and protective factors (immunizations, prenatal vitamins, physical activity, sleep). The clinic staff then entered measurements, laboratory findings, and immunizations, and generated an individualized summary of risk and protective factors for the provider and an education packet for the woman based upon her responses to the screening questions.
A bilingual research team member interviewed the patients regarding their experience with completion of the app screening, their perspectives on the education materials, the providers' counseling using the education materials, and their intent to make [End Page 92]
PARTICIPANT INTERVIEW QUESTIONS
changes based upon the recommendations. Participants were asked the 12 questions outlined in Table 2.
Recruitment continued until saturation was achieved. Previous data suggest the majority of usability problems will be detected with three to four participants and that seven participants will identify the majority of problems.40 Acceptability and cultural appropriateness were addressed in interviews with the 17 participants.
Additionally, a focus group with providers was conducted via Zoom at the end of the study to get their perspectives on: 1) the feasibility of using the technology in their clinic and with their population; 2) the usability of the technology, with special emphasis on the ease of use and usefulness for conducting comprehensive screening per guidelines and counseling women on the areas of need; 3) the effects on clinic flow; and 4) suggestions for improving the screening and/or education materials for use with their population. Spoken consent was obtained to record the focus group. Lunch was provided for participants.
Data from interviews and focus groups were transcribed verbatim. Transcripts were analyzed independently using directed content analysis. Discrepancies in interpretation were discussed and consensus reached. The themes were agreed upon as well as the plans for adaptations to the mHealth app.
Stage VI
Appraisal by the committee that developed StartSmartTM took place to [End Page 93]
DEMOGRAPHICS
determine whether the Beaton process was followed to translate and culturally adapt the system and to suggest refinements to the technology.
Population studied
Demographic characteristics
We recruited a diverse population of Spanish-speaking women at a federally qualified health center (FQHC) (see Demographics in Table 3). Seventeen women consented to complete the tablet-based screening and be interviewed. Due to problems with Internet connectivity, data from screening were not captured for two participants. Thus, interview data from 17 women and screening data for 15 women are included in this report. Women aged 20.5 to 40.5 were asked to complete the tablet-based screening, were counseled by their provider using the individualized patient educational materials that were generated, and then were interviewed by a bilingual research team member. Most women (15 of 17) identified as Hispanic. When responding to the second question about race, many of the participants did not understand that even if they responded Hispanic to ethnicity they were also to indicate race. Thus, responses to race included: seven responded other, five Caucasian, and four were missing data.
Screening data suggested this was the first pregnancy for the majority of women (n=7), but we were able to recruit a diverse sample, with three participants experiencing their second pregnancy, one in her third pregnancy, and four in their fourth pregnancy. Fourteen participants identified their current pregnancy as a singleton and one unknown.
Measures
Anxiety
The Generalized Anxiety Disorder 7-item instrument (GAD-7) was used to assess anxiety.41 It consists of seven questions concerning the preceding two weeks; they ask how often the individual has been bothered by the following problems: feeling nervous, anxious, or on edge; not being able to stop or control worrying; worrying too much about different things; trouble relaxing; being so restless that it's hard to sit still; become easily annoyed or irritable; feeling afraid as if something awful might happen. Responses are rated on a 0–3 Likert scale where 0 = not at all; 1 = several days; 2 = over half the days; 3 = nearly every day. Scores are summed and [End Page 94] interpretation ranked as 0–4 minimal anxiety, 5–9 is mild anxiety, 10–14 is moderate anxiety, and 15–21 is severe anxiety. The GAD-7 has a sensitivity of 89% and specificity of 82%.41 A study using the GAD-7 in pregnant women reported good reliability with Cronbach's alpha of 0.89.34
Depression
The Patient Health Questionnaire (PHQ) is used to screen for depression.42 The PHQ-2 (first two questions on the instrument) are used as a preliminary screen: little interest or pleasure in doing things, and feeling down, depressed, or hopeless. Responses are on a 0–3 Likert scale with 0 = not at all, 1 = several days, 2 = more than half the days, and 3 = nearly every day. If the patient score is greater than 3, the remaining PHQ-9 questions are added. The PHQ-2 has a sensitivity of 87% and a specificity of 86% for any depressive disorder.43 Cronbach's alpha for PHQ-2 was 0.767 and PHQ-9 0.851.44
Intimate partner violence (IPV)
The Abuse Assessment Scale was used to assess for IPV.45 It is a five-item questionnaire that has yes/no responses asking about history of abuse, abuse in the last year, abuse during pregnancy, sexual abuse, and fear of abuser. A positive response to any question is followed by a safety assessment. Sensitivity is reported as 93–94% and specificity ranges from 55–99%.
Substance use
The National Institute on Drug Abuse (NIDA) quick screen was used to assess for alcohol, prescription or illegal drug, tobacco, and marijuana use in pregnancy.46 There is one question for each substance type, "Since you have been pregnant, how often have you used. …" The quick screen has a sensitivity of 79.7% and a specificity of 82.8%.
Obesity
A body mass index (BMI) was generated based upon the height and weight entered by the clinic personnel using the standard formula (weight [kg]/height [m]2).47 A BMI less than 18.5 kg/m2 is classified as underweight, 18.5–24.9 kg/m2 is normal weight, 25–29.9 kg/m2 is overweight, and greater than or equal to 30 kg/m2 is obese. Guidelines suggest early screening for gestational diabetes for Hispanic women with a BMI greater than 25 kg/m2 and for all women at 28 weeks gestation.48
Physical activity
The Godin-Shepherd Physical Activity Questionnaire was used to assess physical activity.49 It consists of three questions: During a typical seven-day period (one week), how many times on average do you do the following kinds of exercise for more than 15 minutes during your free time? a) strenuous exercise (heart beats rapidly—like running, swimming laps, biking [10 miles per hour or more], aerobic exercise); b) moderate exercise (not exhausting—like walking briskly, biking [less than 10 miles per hour]); and c) mild exercise (not much effort—like slow walking, light housework). The activity score is calculated using the following formula (9 X Strenuous activity) + (5 X Moderate activity) + (3 X Mild activity) = Weekly Activity Score. Norms are ≥ 24 units is active, 14–23 units are moderately active, and less than 14 units is sedentary of insufficient activity. Sensitivity is reported at 62% and specificity ranges from 68–97%.
Sleep
The Insomnia Severity Index was used to assess sleep.50 It consists of three questions concerning a) difficulty falling asleep; b) difficulty staying asleep; and c) problems waking up too early. Participants select a response on the Likert 0–4 scale for each: 0 = None, 1 = Mild, 2 = Moderate, 3 = Severe, 4 = Very Severe. Responses are summed and interpreted using the following cut points: 0–7 = No clinically significant [End Page 95] insomnia; 8–14 = subthreshold insomnia; 15–21 = moderate clinical insomnia; and 22–28 = severe clinical insomnia. Sensitivity is reported at 86% and specificity is 87%. Cronbach's alpha is excellent at 0.90.51
Results
Patient risk and protective factor clusters
Participants scored 0–6 on the GAD-7, with a mean of 1.5. The majority of participants (n=13) were classified as having minimal anxiety, with two participants having mild anxiety. Depression scores ranged from 0–2 with a mean of 0.5. None of the participants in our study had a score greater than 3, so the additional questions were not needed. The majority of participants reported no abuse (n=13), and one participant reported a past history of abuse but none in past year during pregnancy. One participant reported using marijuana and tobacco during pregnancy. No other substance use (alcohol, prescription drug, illegal drug) was reported. No reportable clinical outcomes were disclosed.
Over half of our participants were overweight or obese, and only one was underweight (see Table 4). Five participants had glucose screening, three had early screening for a BMI greater than 25 kg/m2, two participants over 28 weeks were screened, and four were missing the date of their last menstrual period (LMP) so weeks gestation could not be calculated. Our results indicated that seven (47%) of the women were active, two (13%) had moderate activity, and six (46%) were sedentary. The sedentary women reported zero activity in all categories. Of the sedentary women, three (50%) were obese, one (16%) was overweight, and two (33%) were normal weight. The majority of women (n=10 [66%]) indicated no insomnia and five (33%) met the criteria for subthreshold insomnia.
Patient interviews
Seventeen women were interviewed regarding the completion of screening with the StartSmartTM app, the educational materials generated for them, the provider's counseling using the educational materials, and their intent to make changes based upon the recommendations. Participant interviews took place after the prenatal visit and were relatively short, ranging from seven to 12 minutes.
The major themes from the interviews are summarized below for usability, acceptability, intent to change, and challenges and suggestions for refinement.
Usability and acceptability
Participants reported that StartSmartTM was very easy to use, the Spanish language used was very easy to understand, the handouts were acceptable, and graphics were clear. One participant suggested, "The design was beautiful to me, as they have babies, but a baby with a mother would be beautiful, because I did not see that." Several women reported that the physical activity questions were difficult to answer since they were divided across separate screens for strenuous, moderate, and mild activity. One participant stated, "It was good for me, I think they give you the screening and then you have confidence and open yourself up if you have a problem… to be able to tell if something is happening to you."
Intent to change
Participants also reported appreciation for the individualized educational materials and conversation with the provider (e.g., "I did not know that physical activity was important for me and my baby;" and "Women go through things [like violence and depression] and it is not talked about, it is delicate and you need [End Page 96]
RISK AND PROTECTIVE FACTORS
information"). All of the participants indicated they would follow the recommendations provided. One participant stated she would follow the recommendations, "Because it helps one maintain a healthy pregnancy, which is what one wants, that the delivery goes well for me."
Challenges and suggestions for refinement
The major challenges mentioned by several participants was that there was a glitch in the app causing the program to revert to English when the participant hit the back button. The other issue raised by multiple participants was understanding the physical activity questions spread across multiple screens. Programmers are working to resolve these issues. Another participant reported [End Page 97]
Screenshot of more diverse babies.
grammatical issues with translation (masculine word was used instead of the feminine version), which are now being addressed by the translators.
Provider interviews
Three providers were interviewed via Zoom. Overall, feedback from providers was positive. They reported StartSmartTM was useful and allowed them to complete a comprehensive screening on all patients. The providers suggested that some Spanish-speaking clients do not read Spanish very well, and they noted that some participants struggled with completing the questions in 45 minutes. They suggested adding the following question, "Would you like to have someone read the questions to you?" Providers also wanted to have the ability to edit the estimated due date based upon ultrasound findings. The app currently asked women the date of their last menstrual period (LMP) and then estimated a due date based upon it. Providers felt that ultrasound provided a more accurate gestational age when compared with a patient remembering her LMP and requested the ability to edit the due date. Providers also wanted more details when a woman screens positive, specifically, which questions she responded positively to on the validated instruments. They also requested an additional question on the mental health screening to include: Have you ever had care for a mental health problem in the past?
Refinement
Based upon the feedback from the participants (both patients and providers), the following refinements were made to the mHealth prenatal screening app including: 1) language revisions as agreed upon through the process, 2) graphics changed to represent more Hispanic babies (see Figure 3), 3) physical activity questions reformatted on one page to improve understanding (see Figure 4), 4) glitches in the program addressed so that app did not change to English language when participants hit the back button, 5) due date can be edited by the provider based upon the ultrasound findings, 6) the provider summary is being reformatted to include more details specific to the patient's response to individual items when she falls into the medium or high-risk categories, and 7) additional question regarding past treatment for mental health will be added. [End Page 98]
Screenshot of refined physical activity questions.
Discussion
Birth outcomes in the U.S. consistently rank lower than other developed countries, and significant health disparities exist between racial and ethnic minority groups and Whites in the U.S.6,52–55 Prenatal health care counseling is associated with positive health outcomes.56,57 Women with greater understanding of risky behaviors have been shown to be more likely to change those behaviors than women whose beliefs are not as strong.58 Provider advice is a catalyst for behavior change.58 Patient-provider communication, however, can be challenging for patients who do not speak English.59
Research also suggests that technology decision support, self-management, and behavior change programs have increased healthy behaviors23–26 and may improve patient outcomes.27,28,56 StartSmartTM is screening technology that provides individualized patient education and provider support for counseling on risk and protective factors.31 To maximize accessibility, the technology had to be translated and culturally adapted for use by Spanish-speaking pregnant women.
This article describes the Spanish translation and cultural adaptation of StartSmartTM, an mHealth technology designed to provide a comprehensive assessment of risk and protective factors in pregnant women, individualized patient educational materials, a provider summary of risk and protective factors, and prompts for counseling on risk and protective factors using motivational interviewing. The technology was developed using the Technology Acceptance Model that includes an iterative process using end user (patient and provider) feedback at each step in the process.32 The translation was guided by the Beaton process, which includes a rigorous translation and back-translation procedure and the cross-cultural adaptation that includes attention to both languages in the translation and cultural adaptation issues in preparing the instruments for use in a particular population.36 After translation and back-translation, the translated instrument was pretested with a diverse population of Spanish-speaking women including immigrants from Mexico, Central America, and South America. The profile of risk and protective factors in this small, clinical sample revealed a diverse sample with a wide [End Page 99] age range, range in the number of pregnancies, and range of risk and protective factors, allowing us to get feedback on a variety of presentations of the educational materials that are specific to the participants' risk and protective factors.
A key tenet of the Beaton process is that translation alone does not automatically provide culturally nuanced language concerning health, and this was carefully monitored throughout the process. In the final testing, we continued to concern ourselves with how well the material performed through questions about the acceptability of the questions and educational materials.36 Participants were also asked if there was a better way to present the material for people like them. The pretesting was guided by the Technology Acceptance Model,32 with feedback from both patients and providers. Refinements based upon the feedback are currently underway.
There are multiple strengths and weaknesses to our study. The process of translation was guided by the rigorous Beaton process. We engaged a total of four bilingual translators—two native Spanish-speakers to complete the forward-translation, and two native English-speakers to complete the back-translation. Final agreement on the Spanish translation was reached by consensus of the translators and research team.
We completed pretesting in a small federally qualified health center (FQHC) with a diverse population of Spanish-speaking women, but the pretest results only represent feedback from one clinic. There were challenges due to the COVID-19 pandemic that limited the researchers' access to the clinic; therefore, we engaged a pre-doctoral fellow who was a clinician working at the clinic to perform the consent procedures with participants. We also adapted the data collection plan to conduct the interviews via Zoom.
Weaknesses of the study included a few technical issues and a small provider sample. A few issues with Internet connectivity affected two of the participants' experience with the mHealth screening. Only three providers in this FQHC see pregnant women, so feedback from providers was limited to this small sample. The study was also limited by the small sample of Spanish-speaking participants who completed the screening with a limited number of risk factors reported other than weight. Thus, our ability to draw conclusions about the cultural appropriateness of the assessment of the stigmatizing behaviors is limited.
Future research is planned to expand our understanding of the utility of StartSmartTM in screening and counseling Spanish-speaking pregnant women. A comprehensive evaluation of algorithm bias was not conducted and should be examined in the future. When refinements based upon this provider and patient feedback are completed, large-scale testing must be completed to establish the efficacy of the revised application on birth outcomes for mothers and babies. Feedback from a larger population of providers is needed in future studies to optimize its use with diverse populations. Additional studies are planned to examine the prevalence of risk and protective factors in Spanish-speaking populations.
BONNIE GANCE-CLEVELAND, SUSAN YATES, ANA T. WILLIAMS, CLAUDIA AMURA, MIA ROBERTS, JESSICA ANDERSON, and PRISCILLA NODINE are affiliated with the College of Nursing, University of Colorado Anschutz Medical Campus, Aurora, CO. JENN LEIFERMAN is affiliated with the Colorado School of Public Health, University of Colorado Anschutz Medical Campus, Aurora, CO. JEN HYER is affiliated with the Denver Health and Hospitals Authority. Please address all correspondence to: Bonnie Gance- Cleveland, College of Nursing, University of Colorado Anschutz Medical Campus, 13120 E. 19th Ave, Mail Stop C288-18, Aurora, CO 80045; Phone: 303-724-7443; Fax: 303-724-6185; Email: Bonnie.Gance- Cleveland@CUAnschutz .edu.
