Article Text
Abstract
Objective To assess differences in rates of postpartum hospitalisations among homeless women compared with non-homeless women.
Design Cross-sectional secondary analysis of readmissions and emergency department (ED) utilisation among postpartum women using hierarchical regression models adjusted for age, race/ethnicity, insurance type during delivery, delivery length of stay, maternal comorbidity index score, other pregnancy complications, neonatal complications, caesarean delivery, year fixed effect and a birth hospital random effect.
Setting New York statewide inpatient and emergency department databases (2009–2014).
Participants 82 820 and 1 026 965 postpartum homeless and non-homeless women, respectively.
Main outcome measures Postpartum readmissions (primary outcome) and postpartum ED visits (secondary outcome) within 6 weeks after discharge date from delivery hospitalisation.
Results Homeless women had lower rates of both postpartum readmissions (risk-adjusted rates: 1.4% vs 1.6%; adjusted OR (aOR) 0.87, 95% CI 0.75 to 1.00, p=0.048) and ED visits than non-homeless women (risk-adjusted rates: 8.1% vs 9.5%; aOR 0.83, 95% CI 0.77 to 0.90, p<0.001). A sensitivity analysis stratifying the non-homeless population by income quartile revealed significantly lower hospitalisation rates of homeless women compared with housed women in the lowest income quartile. These results were surprising due to the trend of postpartum hospitalisation rates increasing as income levels decreased.
Conclusions Two factors likely led to lower rates of hospital readmissions among homeless women. First, barriers including lack of transportation, payment or childcare could have impeded access to postpartum inpatient and emergency care. Second, given New York State’s extensive safety net, discharge planning such as respite and sober living housing may have provided access to outpatient care and quality of life, preventing adverse health events. Additional research using outpatient data and patient perspectives is needed to recognise how the factors affect postpartum health among homeless women. These findings could aid in lowering readmissions of the housed postpartum population.
- obstetrics and gynaecology
- health services research
- womens health
Data availability statement
No data are available. HCUP requires researchers to sign Data Use Agreements (DUAs) before data are released to them. This DUA prohibits sharing or re-release of individual-level data. However, aggregate statistics and the statistical analysis plan are available.
Statistics from Altmetric.com
Introduction
Pregnancy and delivery account for approximately 4 million annual US hospital admissions,1 with 1%–2% resulting in a postpartum readmission.2–6 Hospitalisations reflect care-seeking behaviour for more severe, costly conditions, serving as a global indicator of overall health and well-being.7–12 Postpartum complications are a significant public health issue,13 with social determinants of health contributing significantly to readmissions and inequity in quality of perinatal care.14 Women experiencing homelessness face arduous challenges, with higher rates of chronic and acute illnesses15 and risky behaviours that can result in adverse birth outcomes and delivery complications, impacting rates of postpartum readmissions.16–19 Rates may reflect lower quality of care in some cases, as well as social exclusion, limited access to healthcare services and transportation,15 poor daily health management and hesitation in accessing timely services.20–23 Current literature evaluating the risk of postpartum readmission among homeless women is limited.
Patient education before hospital discharge8 and postpartum check-ups help limit readmissions.9 The American College of Obstetricians and Gynecologists recommends a check-up within 3 weeks of delivery, and a comprehensive check-up by week 12.9 Most readmissions stem from medical conditions treatable in an outpatient setting if diagnosed in a timely manner.24 Thus, evaluation of inequities in social determinants of health in perinatal care is critical to the improvement of patient and child health. For example, postpartum readmissions are lower among women with private insurance than uninsured women.25 Most homeless women in the USA have public insurance or are uninsured.26
Pregnancy can also increase a woman’s risk of homelessness,16 27 28 with pregnancy among homeless women reportedly twice the rate (10%) of US reproductive-age women.29–31 Homeless women have a higher rate of comorbidities, including pre-eclampsia.32 Children born to homeless women risk lower birth weight and higher utilisation of neonatal intensive care units than children born to non-homeless women.18 33–36
In this study, we evaluated differences in rates of postpartum readmissions and emergency department (ED) visits within 6 weeks of delivery hospitalisation discharge among postpartum homeless women compared with non-homeless women. We further explored rates of postpartum readmissions and ED visits stratified by four income levels in a sensitivity analysis. ED visits could serve as a proxy for barriers to preventive care, resulting in readmissions. Low-income and homeless individuals use emergency services at a higher rate than their higher-income counterparts due to poor health management, limited access and delayed care utilisation.20–23 Heightened awareness can help healthcare stakeholders develop and direct more effective interventions designed specifically for perinatal homeless women.
According to Department of Housing and Urban Development homeless population counts, 223 578 women experienced homelessness in 2020.37 New York (NY) was selected for this analysis because of suitable coding quality of the homeless indicator and because it is home to the second largest female homeless population in the USA (18.7%), second only to California (23.9%).37 Beginning in 2012, the NY State Department of Health’s Office of Insurance Programs enrolled individuals experiencing homelessness into Medicaid Managed Care.38 By streamlining Medicaid service access, some barriers were likely removed, but data regarding incentivisation to seek services was not available.
Methods
Database
We used 2009–2014 discharge data from the NY State Inpatient Databases (SID) and State Emergency Department Databases (SEDD), compiled by the Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality (AHRQ).39 40 The SID and SEDD capture all inpatient discharge and ED discharge data, respectively, within NY.41 Therefore, we were able to capture all eligible cases and track the same individuals, even if they used different healthcare facilities throughout the year. While 2016 data were the most recent available at the time of analysis, we chose to use years 2009–2014 as information regarding housing status was discontinued beginning in 2016,42 and 2015 data contained an unusually low number of patients who delivered and were coded as homeless.
Identification of patients
We identified delivery hospitalisations based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes, using previously described methodology.43 Patients with an ‘H’ listed under the zip code variable were categorised as ‘homeless’.42 Patients who had a numeric zip code listed for this variable were categorised as ‘non-homeless’. We restricted our sample to patients who were coded as homeless and non-homeless patients with a numeric NY zip code.
We defined a postpartum readmission or ED visit within a 6-week time frame after the discharge date of the delivery hospitalisation, which corresponds roughly with the uterine involution period.25 44 45 Since personal identifiers in HCUP datasets are not maintained across calendar years, we excluded women with delivery hospitalisation discharges in November or December. We excluded records without a unique identifier, and cases of deaths during delivery hospitalisations which would not have readmission records. Finally, we excluded women missing any of the following delivery hospitalisation data: (1) race/ethnicity, (2) insurance type, (3) age, (4) length of stay (delivery-LOS) (5) discharge month or (6) survival status. Delivery-LOS was necessary to number the days between delivery hospitalisation discharge and the next readmission or ED visit because HCUP provides the number of days between admissions. We included only the first delivery hospitalisation for women who had multiple delivery hospitalisations, yielding a final analytical cohort of 1 109 785 women (online supplemental etable 1)
Supplemental material
Measurements
The primary exposure variable was homelessness. The primary outcome of interest was a binary variable indicating postpartum readmission within 6 weeks after the discharge date of the delivery hospitalisation. The secondary outcome of interest examined women who had a postpartum ED visit.
Adjustment variables were based on previous studies46–49 and included (1) age in quintiles (<23 years, 23–27 years, 28–30 years, 31–34 years and >34 years); (2) race/ethnicity (non-Hispanic white, Black or African American, Hispanic or other); (3) insurance type at delivery (public, private, self-pay or other); (4) delivery-LOS as a categorical variable (≤1 day, 2 days, 3 days, 4 days and ≥5 days); (5) maternal comorbidities, used to create a validated maternal comorbidity index score based on prenatal conditions50 (maternal comorbidity index conditions listed in online supplemental etable 2); (6) other pregnancy complications25 44; (7) neonatal complications (identified within maternal records using ICD-9-CM codes); (8) a binary variable to indicate caesarean delivery; (9) year; and (10) birth hospital ID (detailed adjustment variable information is provided in online supplemental etable 3). Delivery-LOS was included in the models to flag either potentially greater disease severity or more time under care to identify complications emerging during the delivery hospitalisation or comorbidities, which can influence need for follow-up care after discharge. Previous studies51–53 reported a generally longer LOS among the homeless population. In addition, an association between longer postoperative LOS and decreased risk for postpartum hypertension-related readmissions has been reported among women who delivered via caesarean section.54 Therefore, impact of the delivery-LOS requires a necessary adjustment variable. We investigated the correlation between the maternal comorbidity index, other pregnancy complications and method of delivery prior to the analyses. In addition, although maternal age is part of the maternal comorbidity index score, we included it as a separate predictor in all models.
Analytic approach
Descriptive statistics were tabulated and compared between homeless and non-homeless women using χ2 tests for categorical variables and t-tests for continuous variables.
Multivariable logistic regression models were used to assess the primary and secondary outcomes (ie, association between homelessness and postpartum readmissions, and association between homelessness and postpartum ED visits). All regression models were adjusted by age, race/ethnicity, insurance type at delivery, delivery-LOS, maternal comorbidity index score, other pregnancy complications, neonatal complications, and a caesarean delivery indicator with a year fixed effect and birth hospital random effect. A birth hospital random effect was included because some women delivered at the same hospitals, which would lead to residual correlations and affect SEs for estimates and p values. We used marginal standardisation55 to calculate risk-adjusted rates for each outcome. After accounting for adjustment variable associations, we examined whether homelessness was associated with differences in postpartum readmissions. To do this, we performed likelihood ratio tests (LRTs) to compare models that included only adjustment variables with models that included homelessness as an adjustment variable, in addition to the 10 adjustment variables listed previously.
Sensitivity analyses
Six sensitivity analyses were conducted. First, a delivery-LOS was included as a continuous variable in the model. Because it was positively skewed, we log-transformed delivery-LOS and included it in the model. Second, the majority of women in our sample had a delivery-LOS lasting between 2 and 4 days. Vaginal deliveries generally require a 2-day stay.56 According to a recent study by Federspiel et al,56 most US women stayed between 2 and 3 days following uncomplicated caesarean sections (87.5%), with only 1.2% staying more than 4 days. Thus, we only included women with a delivery-LOS between 2 and 4 days. Wen et al 54 reported that 90.6% of readmissions among women who underwent a caesarean delivery occurred within 10 days of discharge. We reassessed this interval for the current data, showing that most postpartum readmissions and ED visits occurred within 10 days after discharge (online supplemental efigures 1–4). This second sensitivity analysis excluded women who had a long delivery-LOS (ie, >4 days), who may have been kept in the hospital for conditions that would likely cause a readmission or ED visit. Third, to assess postpartum readmissions and ED visits for a longer period of observation, we included only women discharged from their delivery hospitalisation in January, and evaluated primary and secondary outcomes during the next 11 months, rather than 6 weeks. Previous studies have reported that postpartum mortality due to suicidal behaviour occurred later in the postpartum period.57–59 Therefore, it was important to test whether our findings were sensitive to our choice of a 6-week observation period following the discharge from delivery. Fourth, we included a birth hospital fixed effect, rather than a random effect, in the model because hospital characteristics may be causing the effect rather than homelessness. A hospital fixed effect controls all time-invariant factors, such as teaching status, profit status, bed size and so on. Fifth, HCUP’s patient identifiers do not cross calendar years, which leaves opportunities to potentially include the same women multiple times if they had multiple deliveries between 2009 and 2014. Therefore, we conducted a sensitivity analysis using only 2014 data to minimise chances of using data from the same mother multiple times. Last, we categorised non-homeless women into four groups based on residential zip code-level median household income and included the income quartile variable in the model. We also estimated risk-adjusted rates of each outcome for each of the five categories, that is, homeless women and the four income quartiles of non-homeless women, with the first quartile representing the lowest income group.
Supplemental material
Supplemental material
Supplemental material
Supplemental material
A two-sided p value ≤0.05 was considered statistically significant. All analyses used SAS V.9.4. (SAS Institute, Cary, NC, USA).
Patient and public involvement
No patients were involved in setting the research question or outcome measures, or in developing plans for the design or implementation of the study. No patients were asked to advise on interpretation or writing of the results.
Results
The analytical cohort included 82 820 homeless and 1 026 965 non-homeless women discharged alive after delivering babies between January and October from 2009 to 2014. The mean ages were 27.7 and 29.3 years among homeless and non-homeless women, respectively. Compared with non-homeless women, homeless women were less likely to be non-Hispanic white (3.6% vs 49.0%), more likely to be Black (32.4% vs 14.2%), Hispanic (28.0% vs 16.4%) or Other-race-identifying (36.0% vs 20.4%), have public insurance (74.8% vs 44.2%) or self-pay which likely encompasses uninsured (23.8% vs 1.6%) and have higher rates of comorbidities. For example, pre-eclampsia occurred more often among homeless women than non-homeless (4.7% vs 2.7% for mild pre-eclampsia and 2.3% vs 1.5% for severe pre-eclampsia or eclampsia) (all p<0.01) (table 1).
We found a strong correlation between preterm labour and early onset of delivery. Therefore, we created a composite binary variable, indicating ‘1’ if preterm labour or early onset of delivery was recorded, or ‘0’ if otherwise (online supplemental etable 4). After adjusting by the variables listed previously, we found that homeless women in NY’s SID and SEDD databases had lower risks of readmissions than non-homeless women within 6 weeks after delivery (risk-adjusted rates: 1.4%, 95% CI 1.2% to 1.6%, vs 1.6%, 95% CI 1.6% to 1.7%; adjusted OR (aOR) 0.87, 95% CI 0.75 to 1.00, p=0.048) (tables 2 and 3). The same trend (ie, a significantly lower risk) was observed among homeless compared with non-homeless women when we examined postpartum ED visits (8.1%, 95% CI 7.4% to 8.8%, vs 9.5%, 95% CI 9.0% to 10.1%; aOR 0.83, 95% CI 0.77 to 0.90, p<0.001) (tables 2 and 3). Results from the LRT showed that homelessness was meaningfully associated with differences in postpartum readmissions and ED visits (online supplemental etable 5).
The results of all sensitivity analyses mirrored the direction of our main analyses, although differences were not always statistically significant (table 4). This is likely explained by decreased power due to the smaller samples included in the sensitivity analyses. When the non-homeless population was subcategorised into four groups based on residential zip-code-level median income, postpartum readmission and ED visit rates increased as the income level decreased. However, the homeless population had the lowest rates of readmission and ED visits, with statistically significant differences observed between the homeless population and housed women in the lowest income quartile for readmissions and between the homeless population and housed women in the lowest three income quartiles for ED visits (tables 2 and 4; online supplemental efigures 5 and 6).
Supplemental material
Supplemental material
Discussion
Our study’s findings of significantly lower rates of readmissions and ED visits within 6 weeks after discharge from delivery hospitalisation among postpartum homeless women compared with non-homeless women are unique. Literature reported higher rates of general hospitalisations and ED visits for other health conditions among the homeless population.53 60–63 Categorisation of our non-homeless sample into quartiles based on zip code-level median household income demonstrated that postpartum readmissions and ED rates increased as income level decreased, yet rates among the homeless population did not follow this trend. This trend is likely indicative of a combination of contributing factors.
Homeless women face numerous barriers in their attempt to access and use healthcare services. NY expanded Medicaid benefits to low-income adults in 2000, but began transitioning homeless individuals and families onto Medicaid Managed Care plans starting in 2012.38 However, 23.8% of our cohort lacked access, having no medical insurance at delivery. Obstacles to using postpartum healthcare for homeless mothers, beyond securing food or housing, include distance to care coupled with lack of transportation, irregular insurance coverage, distrust due to poor treatment by providers, lack of childcare, low health literacy, and fear of being reported to Child and Family Services for homelessness or substance abuse.64 65 As a Medicaid recipient, homeless women also report a stigma associated with their insurance status,66 67 which can result in difficulties finding providers who accept Medicaid plans, and sometimes waiting upwards of 4 hours at a time to be seen.66 67
Several studies have suggested reasons as to why individuals of a low socioeconomic status (SES) are generally expected to have higher rates of postpartum hospitalisations, compared with individuals of a higher SES.25 64 68 69 DiBari et al 68 reported that lower-income women enrolled in Medicaid were more likely to skip postpartum care visits,68 yet these women may be more at risk for health conditions that can result in postpartum readmissions. Clapp et al 25 found that patients who experienced a postpartum readmission were more likely to be in the lowest income quartiles. Our own study resulted in similar findings to Clapp et al 25 when we only evaluated the non-homeless population.
Physical and emotional barriers to scheduled postpartum care have also been cited among mothers in the general population. Bennett et al 69 highlighted obstacles to the customary 6-week postpartum check-up, such as the mother’s negative delivery experience and her sense of stress adjusting to her new role, infant health problems and difficulty adjusting to an infant’s unpredictable schedule, apprehension surrounding additional healthcare related to the delivery and a lack of appropriate services such as childcare. These barriers typically lead to poor health management at the outpatient setting, with limited access to postpartum follow-up care. It is also recognised that non-white patients often have unpleasant and even intimidating experiences with healthcare providers.70 71 Greater adverse experiences during delivery may occur among homeless populations marginalised because of their race or ethnicity and subsequently deter women from accessing postpartum follow-up care.70 71 Without care, a greater volume of readmissions or ED visits among women of a low SES are expected, compared with women of a higher SES. Conversely, our results for homeless women, who are generally among the lowest SES, displayed an even lower trend of readmissions and ED visits.
Given this unexpected trend among postpartum homeless women, a second factor may be involved. Lower odds of readmissions and ED visits may be attributable to certain quality of care standards practised by practitioners who accept Medicaid or agencies that advocate for the homeless, specifically in the state of NY. For instance, some organisations who treat homeless patients may have engaged in patient-centred comprehensive care.72 In this case, care providers may have linked women with health and social services at discharge.72 In fact, NY state has legislated waivers to its Medicaid managed care models for lower-income populations to provide respite and convalescent care, including women with more problematic deliveries and comorbidities.73 As a result, even though homeless women may have been unable to stay in the hospital beyond the period medically needed, respite and convalescent care may have been available for more problematic deliveries.73 In addition, substance abuse services, including sober living homes, are often arranged for women with addictions who must attain or maintain sobriety to maintain custody of their infant.74 These treatment and living facilities can improve homeless mothers’ quality of life, and prevent adverse health events that lead to a postpartum readmission or ED visit. Moreover, this coordinated care effort may facilitate outpatient visits when needed. Khran et al 75 reported that women experiencing homelessness who gave birth and subsequently needed to adapt to the needs of their child experienced high rates of poor mental health and/or substance use. In a systematic review, the authors found that housing programmes implemented throughout the perinatal period were effective in improving the overall health of women.75 The quality of care arising from NY’s patchwork of comprehensive care programmes may have reduced the need for postpartum hospitalisation among homeless women.72 Moreover, comprehensive care engages medical providers to improve patient outcomes by planning beyond the delivery hospitalisation, facilitating coverage for later treatments in an outpatient setting.72 If these connections prevented adverse health events, discharge planning and care coordination among this group could be applied to the low-income population, who had the highest postpartum readmission rate among non-homeless women. Additional qualitative research is needed to understand the quality of services homeless women in NY, and more broadly the USA, have access to and use.
This study contained several limitations. First, our study focused only on NY databases. Therefore, results may not be generalisable to all states. Moreover, healthcare utilisation was only measured through readmissions and ED visits, as outpatient data were not available. In addition, other than public information regarding NY’s social service and medical care partnerships for homeless patients, utilisation data for respite or sober living housing were unavailable. The NY data did not reflect the findings of Clapp et al,5 who found minor readmission rates, which suggests postpartum readmission rates may be insignificant to the quality of perinatal care. However, the study by Clapp et al 5 was not targeted to homeless women, examined 30-day readmissions after delivery rather than 42-day readmissions and surveyed 21 states, many of which have hospitals with much lower facility volumes than NY state’s average. We found a larger readmission rate than that reported by Clapp et al, 5 that is, median of 1.73% versus 1.06%. In addition, our secondary outcome of postpartum ED visit rates occurred more often than readmissions, but still mirrored the main analyses. Hospitalisations reflect care-seeking for more severe and costly health conditions and are often used as a global indicator of patient health and well-being for many different conditions.7–12 Furthermore, ED services often serve as a primary source of care for low-income and homeless populations due to factors such as lack of health insurance or delays in seeking treatment until the condition cannot go unaddressed.20–23 Therefore, we believe that readmissions and ED visits serve as a quality indicator, and the parallel of trends for readmissions and ED visits indicates robust findings.
Second, the postpartum period was defined from the delivery hospitalisation discharge and not the delivery date, which could not be established for vaginal deliveries. We chose the discharge date, when a woman is considered able to continue rehabilitation on her own, for this 6-week window, as often done in previous literature.44
Third, administrative datasets did not provide additional clinical detail beyond ICD-9-CM codes. We could not eliminate the risk of unmeasured confounders, such as number of previous births or parity, or limited demographic information in administrative databases. Length of homelessness was unavailable for our analyses, and we could not confirm definition(s) of homelessness used by EDs and hospitals. We designated women as homeless who were coded as homeless at the time of delivery. It is possible the homeless designation was incorrectly recorded. Women could have listed the address of extended family or friends they were staying with as their own if they did not consider themselves homeless, or wished to avoid the stigma. These women would thus be misclassified as non-homeless. However, it is unlikely that non-homeless individuals were coded as homeless. Therefore, the bias created from this misclassification is towards the null. A study from Massachusetts (MA) used linked emergency shelter usage data and Medicaid records and identified 2384 deliveries per year. Assuming that the same proportion of deliveries to women experiencing homelessness that occurred in NY also occurred in MA, we would expect approximately 2300 deliveries among homeless women per year in MA,16 which supports the quality of homeless coding within the NY SID and SEDD during study years.
Fourth, we could not include eligible readmissions or ED visits that occurred outside NY or cases who died without being hospitalised or visiting an ED. We also could not determine whether some hospitals readmitted women pre-emptively during the postpartum period to prevent harm or deterioration that could befall them due to housing insecurity as they recovered physically.
Fifth, although we only included the first deliveries within the calendar year to avoid including the same individuals multiple times in the analyses, the same individuals could have been included more than once, as personal identifiers do not cross calendar years. Results of a sensitivity analysis designed to address this limitation by using only 2014 data showed similar differences between postpartum homeless and non-homeless women as the main analyses.
Last, we focused on healthcare utilisation within 6 weeks after the delivery discharge. Although the postpartum period is defined as 1 year following delivery, personal identifiers did not cross calendar years. Eleven months was the longest we could evaluate healthcare utilisation, which is after a discharge in January. The sensitivity analysis evaluating postpartum readmissions and ED visits 11 months after delivery discharge demonstrated similar results to our main analyses. In addition, the peak of readmissions occurred within the first 10 days postpartum, as in other studies.13 24 Therefore, we believe our study focused on the most critical postpartum time period.
Our study revealed lower odds of postpartum readmissions or ED visits during the postpartum period among homeless women compared with non-homeless women in NY. Additional research is necessary to explore whether barriers prevent homeless women from using services when needed or if more comprehensive postpartum care services in NY state have improved outcomes for postpartum homeless mothers. The next step will be to conduct interviews among postpartum women experiencing homelessness in NY, as well as to combine evidence from outpatient data with ED and hospital records. If interviews identify barriers in quality of care, or inequities within the homeless postpartum population, further research and collaboration with stakeholders in NY is needed to improve health outcomes. If output data and interviews show that additional services resulted in lower readmission and ED visit rates among homeless women compared with non-homeless women in our study, NY state’s discharge planning and care coordination specific to the homeless population could become a model for comprehensive care among the non-homeless population, as well as for other states.
Data availability statement
No data are available. HCUP requires researchers to sign Data Use Agreements (DUAs) before data are released to them. This DUA prohibits sharing or re-release of individual-level data. However, aggregate statistics and the statistical analysis plan are available.
Ethics statements
Patient consent for publication
Ethics approval
The Institutional Review Board at The Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center approved the study under the ‘exempt’ category.
Acknowledgments
This study used 2009–2014 NY State Inpatient Databases (SID) and State Emergency Department Databases (SEDD), Healthcare Cost and Utilization Project (HCUP), compiled by the Agency for Healthcare Research and Quality (AHRQ). The authors would like to acknowledge all HCUP partners (https://www.hcupus.ahrq.gov/partners.jsp).
References
Footnotes
Twitter @SakaiBizmark
Contributors RS-B conceptualised and designed the study, developed the models, contributed to data interpretation, drafted the initial manuscript and revised the manuscript. HK conceptualised and designed the study, developed the models, contributed to data interpretation and the discussion section, and reviewed and revised the manuscript. DE and SN performed data analyses, contributed to data interpretation, created the figures, and reviewed and revised the manuscript. LEMB conducted literature searches, created the tables, contributed to the introduction and discussion sections, and reviewed and revised the manuscript. LAM and EHM conducted literature searches and reviewed and revised the manuscript. MGR supervised the study overall, confirmed interpretation of the results, and critically reviewed and revised the manuscript. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.
Funding RS-B had financial support from National Institutes of Health (NIH) Research Scientist Development Award (NHLBI K01HL141697) for the submitted work.
Disclaimer The contents of this work are solely the responsibility of the authors and do not represent the official views of the National Heart, Lung, and Blood Institute.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.